table of contents

Endnotes

1. See, e.g., John Harwood, Despite the McVeigh Case, Curbs on Executions Are Gaining Support, Wall St. J., May 22, 2001, at 1 (lead story):

    [P]aradoxically, the dawn of George W. Bush's presidency is bringing a swing in the pendulum away from executions in America. Though most Americans continue to back capital punishment, support has been dropping in recent years in tandem with declining rates of violent crime. Advances in DNA testing and scandals involving the prosecution of major offenses have underscored the fallibility of evidence in capital cases.

    *****

    Recent allegations of misleading testimony by an Oklahoma police chemist who served as a frequent prosecution witness, as well as the FBI's mishandling of records in the McVeigh case, are only adding to pressure for better safeguards.

    *****

    "The politics of the death penalty are clearly changing . . . because of the blunders of the system," says Oklahoma Gov. Frank Keating. Though he staunchly supports capital punishment, the conservative Republican says he favors establishing a higher standard of proof in capital cases, even if that makes death sentences more difficult to obtain. Just five years ago, such a change was unthinkable. But it reflects a broader reconsideration taking place across the spectrum of . . . issues.

Accord Tom Brune, Nation Examines Death Penalty: Amid Capital Punishment Concerns, Federal, State Executions Declining, Newsday, June 10, 2001, at A5; Laurie Goodstein, Death Penalty Falls From Favor as Some Lose Confidence in its Fairness, N.Y. Times, June 17, 2001; Thomas Healy, Death Penalty Support Drops as Debate Shifts: Foes Turning Focus from Moral Issues to Flaws in the System, Balt. Sun, July 25, 2001; Kenneth Jost, Rethinking the Death Penalty, Cong. Q. Res., Nov. 16, 2001; Eric Lichtblau, Death Penalty Reforms Gather New Momentum: With Dozens of Death Row Inmates Freed, a Cry Rises for Precautions, Such as DNA Tests, L.A. Times, June 26, 2001; Toni Loci, Push to Reform Death Penalty Growing; Mistakes Could Shake Confidence in System, USA Today, Feb. 20, 2001; Emilie Lounsberry, Death Penalty's Fairness Debated Nationwide, Phil. Inq., June 8, 2001; Richard Willing, Support for the Death Penalty Might Be Declining, USA Today, Dec. 14-16, 2001.

Reexamination of the death penalty has continued since the events of September 11, 2001. In addition to Kenneth Jost's November 2001 in-depth study of the topic in the Congressional Quarterly Researcher (Jost, supra), Richard Willing's article in USA Today, cited above, and Taylor Bright's five-part series in the Birmingham Post-Herald, Is Fairness Missing from the State's Use of Capital Punishment? (Dec. 2-7, 2001), see Patrik Jonsson, Governors Soften on Death Penalty, The Christian Science Monitor, Oct. 23, 2001; Indiana's Other Lottery: Death Penalty, Multi-part series published in five Indiana newspapers during the week of October 22, 2001; other recent articles cited infra notes 2 & 84 (continuing death penalty debate in Oklahoma prompted in part by a pattern of faulty forensic tests by state experts), 7 (legislative testimony and debate in Nebraska following issuance of a comprehensive study there calling for limits on use of death penalty in the state), 14 (ongoing study of death penalty in Nevada), 23 & 37 (increasing legislative support for moratorium on executions while system's fairness is studied; Kentucky governor's recent endorsement of ban on juvenile executions; surge in court reversals of Kentucky capital verdicts), 27 (progress of legislative proposal to abolish death penalty in Oregon), 30 (discussing study issued by Virginia's Joint Legislative Audit and Review Commission in December 2001 finding "'significant inconsistencies' in Virginia's use of the death penalty" and that overly strict review procedures "may have let stand the convictions of some death row inmates who did not receive fair trials"; reform proposal following study in Indiana), 34 (discussing late Fall 2001 ruling banning execution of mentally retarded defendants in Tennessee), 35 (Chicago Tribune's December 16-19 series on false confessions in capital and other cases), 43 (discussing strong bipartisan support among gubernatorial candidates in Illinois for maintaining the state's moratorium on executions), 44 (Fall 2001 statements by New Mexico Governor Gary Johnson, a long-time death penalty supporter, that "eliminating the death penalty in the future may prove to be better public policy given the reality of the sentence today," followed by his withdrawal of proposed legislation to shorten the review process in capital cases, which he now opposes because it might permit innocent prisoners to be executed), 69 (speech by Supreme Court Justice Sandra Day O'Connor to Nebraska Bar Association reiterating her view that the death penalty as currently administered risks executing the innocent), 73 (recent successes in California and Nebraska of ongoing campaigns to secure moratorium resolutions from local governments calling on state officials to halt executions while studies occur), 75 (decline in number of executions, continuing through end of 2001), 894 (December 29, 2001 decision of the Kansas Supreme Court limiting the application of the state's death penalty to highly aggravated cases and requiring new sentencing hearings for all of the state's death row inmates), 968 (decision of the Illinois House of Representatives Speaker to forgo attempt to override the governor's veto of bill expanding number of aggravating circumstances in Illinois's capital statute).

See also Dion Nissenbaum, Riordan Outlines Education Plan, Touches on Death Penalty, Jose Mercury News, Jan. 24, 2002 ("Republican Richard Riordan [who is running for governor in California] waded into two touchy issues Wednesday, telling a Silicon Valley crowd that as governor he might try to eliminate the state's elected schools chief and would be open to halting executions if evidence emerged that they were being unfairly imposed."); Michael Saul, Mayor-Elect Mike Wants To Be a Major GOP Player, N.Y. Daily News, Nov. 12, 2001 (Mayor-elect Michael Bloomberg has a message for his fellow Republicans: When it comes to certain hot-button issues . . . you're dead wrong. . . . 'I am . . . against the death penalty, as are most people in New York City,' he said. . . . Bloomberg . . . said he 'absolutely' plans to use his new position to help shape the Republican Party agenda."); Peter Whoriskey, Kaine Edges Out Katzen For State's No. 2 Office, Wash. Post, Nov. 7, 2001 (discussing election of Democrat Timothy M. Kaine as lieutenant governor of Virginia despite his opposition to the death penalty and support for a moratorium on executions, and despite the longstanding political wisdom that staunch support for the penalty is a prerequisite for statewide office in Virginia).

Note: This Report's discussion of recent events is current to November 30, 2001. Occasional references are made to matters that came to our attention after that date, but the demands of the production process have prevented us from comprehensively canvassing events after November 30.

2. See Editorial, Assuring Justice, Columbus Dispatch, July 29, 2001:

    Americans disagree passionately about the death penalty, but those on both sides of the issue should be able to find common ground on this one concept:

    Innocent people ought not be executed.

    This is the premise behind the Innocence Protection Act, a proposed federal law that would require action to help ensure that defendants convicted of capital murder are, in fact, guilty. The bill would make DNA testing more available to condemned inmates and would set minimum standards for defense lawyers representing indigents.

See also Goodstein, supra note 1 ("After a decade in which there appeared to be an unshakable near consensus in favor of the death penalty, Americans say they are now rethinking and debating capital punishment . . . . Interviews in six states this week reflect the poll numbers, which show that while there is still a majority in favor of the death penalty, the size of the majority is shrinking. While many people cited the biblical command to take 'an eye for an eye' . . . others said they had recently changed their minds after concluding that the death penalty was administered unfairly. Some said that what persuaded them was the news that 13 prisoners on death row in Illinois were discovered to be innocent-a revelation that led Gov. George Ryan, a Republican, to declare a statewide moratorium on the death penalty last year. Others said they were troubled by reports that the death penalty may be disproportionately imposed on blacks and Hispanics.").

A major focus of concern is Oklahoma, where faulty forensics work may have compromised a number of death row cases, including those of some inmates who have been executed. As the Dallas Morning News recently reported:

    In a 21-year career, Oklahoma City police chemist Joyce Gilchrist was a prosecutor's dream: She delivered supportive lab analysis and convincing testimony that helped send hundreds to prison --at least 23 people to death row.

    Ms. Gilchrist may turn out to be a prosecutor's worst nightmare: So much of her work was questioned by appeals courts and forensics experts that she was suspended and fired. Investigators are digging through 1,197 of her cases to see whether anyone is behind bars because of false or misleading testimony.

    And now--in a year when Oklahoma leads the nation in carrying out the death penalty, and with suspect convictions being reviewed even beyond the Gilchrist cases--some are pondering the unthinkable: Has Oklahoma executed the innocent?

    *****

    Separately, the Oklahoma City Police Department has asked the task force to review as many as 10 cases handled by a second former police chemist, the late Janice Davis. In one such case, Ms. Davis used hair and fiber analysis to link Dewey George Moore to a 1984 murder. The task force ordered DNA tests to determine whether Mr. Moore has wrongly spent 16 years on death row.

    ******

    Meanwhile, the state task force reviewing Ms. Gilchrist's files has ordered DNA testing on two death row cases: John Michael Hooker, who was convicted in the 1987 stabbing deaths of his girlfriend and her mother; and Michael Edward Hooper, convicted in the 1993 shooting deaths of his former girlfriend and her two young children.

Chemist's Errors Stir Fear: Did Oklahoma Execute Innocent? Questions Force Experts to Dig Through Thousands of Cases, Dallas Morning News, Oct. 22, 2001. See also infra notes 50, 79, 84, 232 (discussing various aspects of the Gilchrist case).

Not everyone is equally concerned about the possibility of executing the innocent. For an argument that executing the innocent is an acceptable collateral effect of the existing death penalty, see Ernst van den Haag, In Defense of the Death Penalty: A Practical and Moral Analysis, in Hugo Adam Bedau, The Death Penalty in America 325 (3d ed. 1982). See also ABC, Innocence and the Death Penalty, Nightline, July 14, 1997 (transcript), quoted in part in Jackie Cooperman, ABCNEWS.com, July 15, 1997, http://more.absnews.go.com/sections/us/death715/index2.html (statement of Rep. Bill McCollum, R-Fla.):

    I don't think there's any question that someday somebody who is innocent will be executed in this country. It's happened before. . . . [O]n the whole, considering the total number of murders there are in this country, the total number of people who are on death row . . ., the interest of the public and the victims in seeing justice done, and what I believe is a deterrent effect for the death penalty with respect to those who commit premeditated murder, I think it is a relatively low risk and one that we have to accept. We try not to let it happen . . . but realize it probably will happen.

3. Brian Bakst, O'Connor Questions Death Penalty, Associated Press, July 2, 2001.

4. See, e.g., William Clairborne, Ill. Governor, Citing Errors, Will Block Executions: Wash. Post, Jan. 31, 2000, at A1; Steve Mills & Ken Armstrong, Gov. George Ryan Plans to Block the Execution of Any Death Row Inmates, Chi. Trib., Jan. 30, 2000.

5. See, e.g., Jo Thomas, New Death Penalty Rules Are Issued in Illinois, N.Y. Times, Jan. 23, 2001, at A17 ("The Supreme Court of Illinois has adopted new rules governing the way death penalty cases are handled. The rules . . . set requirements for training and experience for all defense lawyers and assistant prosecutors handling the cases[,] . . . require regular training for [capital] judges . . ., remind prosecutors that their duty is 'to seek justice, not merely to convict,' . . . require prosecutors not only to give defense lawyers any evidence that may tend to exonerate their client-like a statement that someone else committed the crime, or a scientific test result that is not incriminating-but also to identify clearly which information may be mitigating . . .[, and] specify that prosecutors must hand over any relevant information relating to DNA evidence, including reports explaining any discrepancies in the testing, observed defects or laboratory errors, the reasons for these errors and the effect of these mistakes." (discussing Ill. R. Prof. Conduct 3.8; Sup. Ct. R. 43, 411, 412, 416, 417, 701, 714)).

6. See Ray Long & Maurice Possley, House Acts to Restrict Jailhouse Informants; Bills Also Target Prosecutor Abuses, Chi. Trib., Apr. 24, 2001 ("The Illinois House overwhelmingly approved a sweeping legislative package," sponsored by former prosecutor and Republican state representative Jim Durkin and supported by the Illinois Association of Police Chiefs, which is "designed to crack down on wrongful convictions and misconduct by prosecutors. . . . The legislation would put strict limits on the testimony of jailhouse informants, whose goal of shortening their sentences by providing testimony for prosecutors often has prompted them to twist the truth at the expense of justice . . . [and] also would place a new burden on prosecutors who fail to turn over [exculpatory] evidence to the defense, . . . gives a person facing charges that carry a potential death sentence . . . the chance to challenge the veracity of witnesses by taking depositions before trial, . . . [and] would require pretrial reliability hearings before prosecutors would be allowed to put jailhouse informants on the witness stand.").

7. See, e.g., See Ken Hambleton, Researcher: Limit Call for Executions, Lincoln J. Star, Oct. 18, 2001 (discussed infra p. 396 & n.886); Robynn Tysyer, Execution Suspension Approved: Senators Hand Johanns Life-and-Death-Decision, Omaha World-Herald, May 20, 1999, at 1; Robynn Tysyer, Death Penalty Study OK'd, Omaha World-Herald, May 28, 1999, at 1.

8. See The Justice Project, 2001 State Legislation on Death Penalty Reform at a Glimpse, July 19, 2001.

9. See Howard Libit, Data Hint at Race Bias: Researchers Advise Md. To Halt Executions Until Studies End, Baltimore Sun, Feb. 23, 2001.

10. See, e.g., Daniel LeDuc, Md. House Backs 2-Year Moratorium on Death Penalty, Wash. Post, Mar. 25, 2001, at C1; Daniel LeDuc & Lori Montgomery, Md. Bill to Suspend Executions Dies, Wash. Post, Apr. 10, 2001, at A1; Jost, supra note 1 ("In Maryland, supporters believed they had enough votes to win passage of a moratorium bill in the state Senate, but opponents blocked a vote in April with a filibuster as the legislature was about to adjourn.").

11. See, e.g., Editorial, Maryland's Execution Pause, Wash. Post, Apr. 15, 2001, at B6 ("The Maryland Court of Appeals on Thursday accomplished what the state's legislature failed to do a few days earlier-put a temporary halt on executions in the state.").

12. See Ed Vogel, Nevada Assembly Rejects Death Penalty Halt, Las Vegas Review-Journal, May 17, 2000.

13. See Ed Vogel, Bill to Protect Mentally Retarded Passes, Las Vegas Rev.-J., Apr. 25, 2001 (discussing Nevada Senate's passage of bill banning execution of mentally retarded persons by a vote of 28-11).

14. See Eric Neff, Lawmakers Tackle Tough Issue of Death Penalty, Las Vegas Sun, Oct. 30, 2001 ("No issue dogged this year's Legislature the way Nevada's death penalty did. . . . On Monday an interim legislative committee began in the first of six meetings to tackle the task of recommending bills to the 2003 Legislature that will improve what many believe are inherent flaws in the way capital punishment is carried out in Nevada, which has the largest per capita death row population in the nation. . . . 'I think it's one of the most important topics nationally, as well as in the state, of our day,' said Assemblywoman Sheila Leslie, D-Reno, who chairs the interim study committee."); Cy Ryan, Death Penalty Foe to Head Panel: Legislators to Consider Future of Capital Punishment in Nevada, Sept. 7, 2001; Sean Whaley, Interim Study Sessions: Lawmakers Enter Research Season; Eight Topics Funded for In-Depth Study Evaluation in Preparation for 2003 Legislature, Las Vegas Rev.-J, July 23, 2001 (discussing death penalty study ordered by state legislature, which "will look at a variety of issues, including the use of DNA testing, the cost of implementing the death penalty as opposed to life imprisonment and whether people under age 18 or who are mentally retarded should be sentenced to death").

15. See, e.g., Lounsberry, supra note 1.

16. See Christy Hoppe, Governor Enacts Criminal Justice Laws, Dallas Morning News, June 15, 2001.

17. See Raymond Bonner, Ban on Execution of the Retarded Is Vetoed in Texas: Exception to U.S. Trend, N.Y. Times, June 18, 2001, at A1 (lead story).

18. See Christy Hoppe, Texas Senate Approves 'No Parole' Option, Dallas Morning News, May 4, 2001.

19.Unless otherwise noted, these actions are documented in Death Penalty Information Center, Changes in the Death Penalty Around the U.S., http://www.deathpenaltyinfo.org/Changes.html (visited Oct. 27, 2001). See infra note 75 (regarding number of executions in 2000 and 2001). For a useful summary, see Christy Hoppe, A Shift in Scales of Justice: Besieged System Gets an Overhaul, Dallas Morning News, May 29, 2001, at A1:

    The swaggering and cocky Texas justice system sat down for a reflective session and came away with something akin to a death-row conversion.

    By Monday, gone from law-and-order legislators were their defense of sleeping lawyers and executing the mentally retarded. Silenced was the refrain that innocents never get the death penalty.

    Instead, after the national spotlight of a presidential election, lawmakers accepted responsibility for their criminal justice policies and voted overwhelmingly to change them-although many of the same proposals had faced defeat time and again in earlier years.

    "Sometimes a picture can say 1,000 words, and the picture of a sleeping lawyer defending someone who is facing the death penalty was so vivid in the minds of people in this country that it didn't take a rocket scientist to figure out that that is not competent representation," said Rep. Juan Hinojosa, D-McAllen.

    ******

    [M]ost lawmakers hailed the bills as a major victory-perhaps the most far-reaching efforts of the legislative session-ending the status quo and starting toward a fairer system. "You know, this is a law-and-order state, and I think we made the point that you can be tough on crime and still be compassionate," said Sen. Rodney Ellis, D-Houston, who has championed indigent defense reform and prohibiting the execution of the mentally retarded.

See also Steve Mills, Texas Revisits Death Penalty: Legislators Weigh Reform in Nation's Execution Leader, Chi. Trib., March 25, 2001; John Moritz, Texas May Be Yielding to Execution Criticisms, Fort Worth Star-Telegram, Aug. 17, 2001 ("Texas hasn't gone soft when it comes to the death penalty, but its leaders seem to be increasingly sensitive to criticism of the state's use of capital punishment, several observers said a day after a court halted the execution of a man who killed at age 17."); Jim Yardley, Texas Retooling Criminal Justice in Wake of Furor on Death Penalty, N.Y. Times, June 1, 2001, at A1 (lead story) ("Texas, which leads the nation in executions and endured withering criticism of its death penalty system during the presidential campaign last year, is poised to make significant changes in its criminal justice laws and so, supporters of the overhaul say, create a fairer system of capital punishment."); Jim Yardley, Of All Places: Texas Wavering on Death Penalty, N.Y. Times, Aug. 19, 2001.

For an article discussing other legislation given serious consideration by Texas lawmakers, see Paul Duggan, Texas Legislators Review Use of Death Penalty; National Criticism During Presidential Campaign Reverberates in a Spate of Bills, Wash. Post, May 14, 2001, at A3: "Texas, which has executed more convicted murderers in the last two decades than most nations of the Western world, is considering a surprising array of capital punishment reforms that could reduce the number of death sentences imposed here, lawmakers said. After meeting resistance in the Texas House and Senate in past years, several bills aimed at tempering the capital punishment system have advanced this spring, [including]":

  • "a measure that would bar executions of convicted killers who are mentally retarded," which died in a committee last year after then-Governor George W. Bush opposed it;

  • a bill "that would prohibit capital punishment for defendants accused of committing murders before age 18," which cleared a key House committee, and stands in stark contrast to the corresponding bill in the previous session, which would have lowered the minimum age to 16, "the youngest allowed by the U.S. Supreme Court";

  • a bill "designed to improve the quality of court-appointed defense lawyers in capital murder trials" by "set[ting] minimum professional standards for the lawyers and offer[ing] $22 million in indigent-defense grants to counties that limit the role of judges in the attorney-selection process," which contrasts with a "much weaker" bill that passed both houses two years ago but was vetoed by then-Governor Bush after local judges lobbied against its removal of their patronage power over appointments;

  • a "bill that would allow jurors in capital cases to impose sentences of life without parole . . . [which] could make a jury less inclined to vote for execution," which "died in House and Senate committees in 1999 [but t]his year . . . has passed the Senate and won committee approval in the House" and received cautious support from the current governor; and

  • even a bill calling for a referendum on a proposed execution moratorium, which cleared key committees in both houses of the legislature . . . .

For an influential view that the reforms Texas has adopted are not yet enough, see Editorial, Death Row Decisions-Once Again, the System Is in Question, Dallas Morning News, Aug. 15, 2001 ("[T]he glare of publicity [two recent cases in which Texas pursued plans to execute a juvenile offender and defended in court a capital sentence imposed on a defendant whose lawyer slept through much of the trial] they have attracted once again draws attention to the need to improve the justice system in Texas.").

20. See Eric Frazier, New Death Penalty Law Put in Action: Suspect in Murder Won't Face Execution, Charlotte Observer, May 25, 2001, at A1:

    Mecklenburg prosecutors brought an important new capital punishment law into dramatic focus Thursday when they announced they won't seek the death penalty against a man they nearly put on trial for his life days earlier.

    The law, signed by Gov. Mike Easley last week, gives prosecutors more discretion in whether to seek the death penalty. Mecklenburg Assistant District Attorney Marsha Goodenow said her office plans to review its pending murder cases in light of the change.

    ******

    The old law forced prosecutors to seek death when they found one or more of 11 aggravating circumstances in a murder case. . . .

    Critics of the N.C. capital punishment system have pointed to the old law mandating death penalty trials as a reason for the state's higher-than-average death sentencing rates.

21. Stephen Dinan, Time Limit Lifted on DNA Testing for Felons, The Wash. Times, May 3, 2001; see, e.g., Craig Timberg, Time Limit Lifted for DNA Appeals, Wash. Post, May 3, 2001 ("Gov. James S. Gilmore III approved a new right today for death row inmates, adding Virginia to the growing list of states that give the condemned access to DNA testing long after their convictions.").

22. See Bill Blakemore, Support for Death Penalty Drops, ABC World News Tonight, May 2, 2001.

23. See Mark R. Chellgren, Momentum May Be Growing Toward Moratorium on Capital Punishment, AP State & Local Wire, Nov. 21, 2001 (discussing legislative support for 2-year moratorium on executions in Kentucky while state "studies whether capital punishment is fairly administered.").

24. See supra p. 1.

25. See supra pp. 1-2.

26. See supra p. 2.

27. See Changes in the Death Penalty Around the U.S., supra note 19; Jost, supra note 1 ("Opponents of the death penalty . . . came very close to abolishing capital punishment in two states [in 2001]-New Hampshire and New Mexico."). On Oregon, see Mark L. Thompson, Repeal of Death Penalty Sought, Salem Statesman J., Nov. 9, 2001 ("Capital punishment opponents will begin their campaign to repeal Oregon's death penalty next week").

28. See Editorial, Waning Penalty, Boston Globe, March 14, 2001 ("The 32-vote margin in the Massachusetts House Monday against reinstating the death penalty-compared to the nine-vote difference in 1999- . . . signals a sea change in death-penalty politics . . . ."); Emmet Meara, Maine Death Penalty Bill Denied, Bangor Daily News, May 17, 2001. See also Jost, supra note 1 ("Death penalty bills failed to advance [in 2001] in two . . . other . . . states, Michigan and Minnesota.").

29. See Death Penalty Report Released, Associated Press, Aug. 4, 2001 ("A state panel reviewing Arizona's use of the death penalty released an interim report that recommends additional resources for defense attorneys [i.e., "a statewide public defender's office to represent defendants in death penalty cases"], commutation of sentences against the mentally ill, and a ban on executing [offenders for crimes committed as] minors.").

30. See, e.g., Associated Press State & Local Wire, Panel Recommends Change in State's Death Penalty Law, Dec. 14, 2001 (recommendation by Indiana's Criminal Law Study Commission, which reviewed the state's death penalty at Governor Frank O'Bannon's request, that "Indiana's death penalty law should make it clear that only people who intended to kill someone should be executed"); Associated Press, Revised Death Penalty Wins Final Legislative Approval, June 7, 2001 (Connecticut "state lawmakers approved a bill that would . . . require a comprehensive study of how the state imposes the death penalty."); Suzan Bibisi, Death Penalty Bill Becomes Law, Waterbury Republican-American, July 7, 2001 (noting that Governor Rowland of Connecticut signed legislation establishing a comprehensive study of Connecticut's death penalty system); Raymond Bonner, Justice Dept. Set to Study Death Penalty in More Depth, N.Y. Times, June 14, 2001 ("The Department said today that it would undertake a comprehensive study of the federal death penalty to determine whether the system is racially or ethnically biased."); Changes in the Death Penalty Around the U.S., supra note 19 (discussing studies in Arizona, Illinois, Indiana, Maryland, Nebraska, North Carolina and Virginia); Frank Green, Study Critical of Death Penalty: Statute Audit Panel Cites 'Significant Inconsistencies,' Richmond Times-Dispatch, Dec. 11, 2001 ("A state study found 'significant inconsistencies' in Virginia's use of the death penalty and concluded that strict adherence to rules by federal and state appeals courts may have let stand the convictions of some death row inmates who did not receive fair trials. The study released yesterday by the Joint Legislative Audit and Review Commission also suggested the Virginia Supreme Court was overly strict in its use of state law in determining whether a death sentence was excessive or disproportionate. In all 119 death cases that have come before the state justices since the death penalty resumed in 1977, none was found excessive."); Robert Schwaneberg, AG Studies Disparities in Capital Punishment, Newark Star-Ledger, Aug. 21, 2001 ("[New Jersey] Attorney General John Farmer Jr. has launched a review of standards that prosecutors use for seeking the death penalty, following a landmark study [issued by the New Jersey state supreme court] that found wide differences in how aggressively capital punishment is sought in the 21 counties. Farmer said he will head a panel of four former attorneys general and a former director of criminal justice to study why some counties are far more likely than others to seek the death penalty and whether the state should impose a more uniform approach to the way it is applied."); supra notes 7, 12-14 & accompanying text (discussing Nebraska and Nevada studies).

31. Robynn Tysver, Inmate-DNA Bill Gets 2nd-Round Approval, Omaha World-Herald, May 10, 2001.

32. See, e.g., Akron Beacon J., Nov. 12, 2000 (Ohio); Diane Plumberg, Gov Signs Tuition Hike, DNA Testing Funds Into Law, The Oklahoman, http://www.oklahoman.com/cgi-bin/show_article?ID=689458&-pic=none&TP=getarticle (Oklahoma); Changes in the Death Penalty Around the U.S., supra note 19. See also Uniform Statute for Obtaining Post-Conviction DNA Testing, Unanimously Approved at the Ninth Meeting of the National Commission on the Future of DNA Evidence, Apr. 9-10, 2000.

On some limitations of these statutes (particularly Florida's) that keep them from providing full protection to capital prisoners, see, e.g., Jagger, Death Row Case Demands DNA Testing, Miami Herald, Aug. 28, 2001.

33. See, e.g., In re Adoption of a new subdivision 29.17, entitled, "Post-conviction Motion for Forensic Testing not Available at Trial," of Supreme Court Rule 29 (Mo. S. Ct. Feb. 20, 2001); Francis X. Clines, Inmate Access to DNA Tests Gains Ground Across Nation, N.Y. Times, Dec. 19, 2000 (discussing administrative order of Ohio Attorney General Betty D. Montgomery, requiring DNA testing of death row inmates who show that "a favorable [DNA test] result for the inmate [would] 'by itself' . . . produce exoneration"); DNA Tests Okd for Texas Cons, N.Y. Daily News, Apr. 6, 2001; Legislators Reevaluate Death Penalty with New Initiatives; The Aims: Proper Representation for Those on Death Row and New Trials If DNA Evidence Shows a Need, Phil. Inq., July 10, 2001; Changes in the Death Penalty Around the U.S., supra note 19; supra pp. 2 & n.15 (discussing recent DNA legislation in Texas and Virginia).

34. See, e.g., Associated Press, Revised Death Penalty Wins Final Legislative Approval, June 7, 2001 ("After years of making changes to strengthen the death penalty, [Connecticut] lawmakers approved a bill that would exempt mentally retarded people from execution, remove one category of crime from qualifying for capital punishment and require a comprehensive study of how the state imposes the death penalty."); Bibisi, supra note 30 (noting that Connecticut Governor John Rowland signed legislation banning execution of retarded individuals); Bill Bell, Jr., Legislation Sent to Holden Would Ban Death Penalty for Mentally Retarded, St. Louis Post-Dispatch, May 11, 2001 ("Capital punishment in Missouri for the mentally retarded would become a thing of the past under legislation on its way to the governor. House members gave final approval to the bill Friday, voting 107-19; the Senate approved the same measure Thursday night. The sponsor is Sen. David Klarich, R-Ballwin. Gov. Bob Holden indicated Friday he probably would sign the bill. 'I've always said that someone who meets the definition of mentally retarded, I would have serious concerns about ever putting an individual like that to death,' he said."); Amber McDowell, Court Denounces Execution of Retarded, Associated Press, Dec. 4, 2001 ("Executing the mentally retarded is cruel and unusual punishment prohibited by both the federal and state constitutions, the Tennessee Supreme Court ruled Tuesday. Such executions violate 'the evolving standards of decency that mark the progress of a maturing society both nationally and in the state,' Justice Riley Anderson wrote in the majority opinion."); Holden Signs Ban on Death Penalty for Mentally Retarded, St. Louis Post-Dispatch, July 2, 2001; Lounsberry, supra note 1 (discussing Arizona's adoption of ban on executing mentally retarded persons in April 2001); Jeb Bush Signs Bill Barring Executing the Retarded, N.Y. Times, June 13, 2001("Joining a rising number of states that prohibit the execution of individuals who are mentally retarded, Gov. Jeb Bush extended the ban to Florida today under a bill he signed into law."). See also Raymond Bonner, Argument Escalates on Executing Retarded, N.Y. Times, July 23, 2001; Raymond Bonner, Drawing a Line on Death, N.Y. Times, June 24, 2001 ("Capital punishment is a divisive topic in this county, and recently a heated debate has arisen about whether convicted murderers who are mentally retarded should be executed. President Bush jumped in the fray when he said, "We should never execute the mentally retarded."); Duggan, supra note 19 (as of 1989, "only two states barred capital punishment for mentally retarded defendants"; in the ensuing 11 years (as of 2000), 11 more adopted a ban).

On North Carolina, see, e.g., Associated Press, Easley Signs Law Banning Execution of Mentally Retarded, Herald Sun, Aug. 4, 2001 (despite reservations, "Gov. Mike Easley . . . signed legislation that bans executions of the mentally retarded," noting that support for the bill by "the state district attorney's association and Attorney General Roy Cooper were 'compelling factors' in his decision"); Dana Damico, Death-Penalty Reforms Remain on Legislative Agenda: Backers Push Proposals Ranging from Outright Ban to Additional DNA Tests, Winston-Salem J., Apr. 30, 2001 (legislation to ban the execution of retarded defendants passed the North Carolina Senate by a vote of 31-18); Stan Swofford, Decision Could Nullify Appeal, Greensboro News & Record, July 26, 2001 (noting that North Carolina ban on executing retarded individuals may spare "as many as 20 percent of the state's 220 Death Row inmates"; "[n]ationwide, an estimated 300 mentally retarded people are on death row awaiting execution.").

35. On Albert Burrell's case, see Tom Guarisco, Former Death Row Inmate Wants to Repair Cars Again, Baton Rouge Advocate, January 3, 2001:

    The nearly fatal turn Albert Ronnie Burrell's life took 14 years ago finally ended Tuesday when he walked away from Angola's death row a free man.

    Burrell was living with his mother in rural north Louisiana in 1987 when a jury convicted and sentenced him to die for the double murder of an elderly Union Parish couple.

    Last week, the state Attorney General's Office dismissed all charges against Burrell and his supposed accomplice, Michael Ray Graham Jr. of Roanoke, Va. The Attorney General's Office said the case against the men was weak, and DNA testing of blood found at the scene did not match their blood.

    *****

    Burrell's release came late Tuesday afternoon after a court hearing in Union Parish. Burrell was once as close as 17 days away from execution. He is mentally retarded[, cannot read or write,] and did not understand how and why he ended up on death row, but he fully expected to die, according to his attorneys and family.

On the Porter and Washington cases, see infra pp. 27-34. In regard to the proclivity of retarded persons, children, suspects under coercion from police officers and others to confess to crimes they did not commit, particularly in capital and other murder cases, see The Chicago Tribune's recent four-part series, Ken Armstrong, Steve Mills & Maurice Possley, Cops and Confessions, Chi. Trib., Dec. 16-19, 2001; Editorial, New Doubts About Confessions, Chi. Trib., Dec. 19, 2001.

36. In McCarver v. North Carolina, 121 S. Ct. 1401 (2001) (mem.), the Supreme Court agreed to reconsider its decision in Penry v. Lynaugh, 492 U.S. 302 (1989), that the Eighth Amendment Cruel and Unusual Punishment Clause does not bar execution of mentally retarded individuals. See Lounsberry, supra note 1 ("Twelve years after approving such executions, the [U.S. Supreme Court] said it would . . . determine if imposing the death penalty on the mentally retarded violates the Eighth Amendment ban on cruel and unusual punishment"). After North Carolina adopted legislation retroactively barring executions of the mentally retarded prisoners, see supra note 36 and accompanying text, the Court dismissed the McCarver case and immediately agreed to review the same claim made by a Virginia death row inmate, Daryl Atkins. See Gina Holland, Court Reviews Executing Mentally Ill, Associated Press, Sept. 25, 2001 ("The Supreme Court made clear Tuesday that it will soon decide if it is constitutional to execute mentally retarded killers, substituting a moot North Carolina inmate's case with one from Virginia.").

37. See Changes in the Death Penalty Around the U.S., supra note 19; supra pp. 1-2 & nn.14, 19 (Nevada, Texas). On Kentucky, see John Cheves, Judges Are Overturning Death Sentences, Lexington Herald-Leader, Oct. 31, 2001 (Kentucky "death penalty opponents have found a new ally: Gov. Paul Patton announced this month that he supports a proposed bill to abolish the death penalty for those who committed their crimes as 16- and 17-year-olds.).

38. Harwood (Wall St. Journal), supra note 1:

    Other [states], including Arkansas and North Carolina, have indirectly curbed [the death penalty's] application by beefing up standards or taxpayer funds for the representation of indigent defendants. . . .

    Just last week, the Texas House voted to create the state's first standards for court-appointed lawyers. The Texas Senate had already passed similar legislation. The Supreme Court this fall is scheduled to revisit whether to bar the execution of mentally retarded inmates.

    The pendulum swing is occurring even in Oklahoma City, where Mr. McVeigh bombed the Alfred P. Murrah Federal Building six years ago, killing 168 people. There is early evidence that Oklahoma convicts are receiving fewer death sentences in the wake of the state's decision to improve legal counsel for poor defendants and expand access to DNA testing. . . .

    [Under the Oklahoma reforms, a] state board . . . provides lawyers for poor defendants.... In the past, if a lawyer assigned to represent an indigent defendant "had vital signs, he was determined to be competent," says [Board chairman Jim] Bednar. "In theory I'm not opposed to the death penalty. But it's the practice we need to look at. The system is flawed."

    [Bednar] began to overhaul the indigent-defense agency by winning funding increases to hire better-quality lawyers. The agency is now sending the message that attorneys for poor inmates "are really going to show up and do our job," Mr. Bednar says.

    Because of stiffer opposition, prosecutors are becoming "more hesitant to seek the death penalty," he adds. In fiscal year 1998 . . ., prosecutors in the area served by his Norman office, which covers roughly the western half the state, sought death sentences in 36 cases. They obtained the punishment in four cases. Last year, prosecutors sought 26 death sentences and obtained only one.

See also Changes in the Death Penalty Around the U.S., supra note 19; The Justice Project, supra note 8. For evidence of a similar decrease in the use of the death penalty as a result of better qualified and compensated counsel in Indiana, see Norman Lefstein, Reform of Defense Representation in Capital Cases: The Indiana Experience and Its Implications for the Nation, 29 Ind. L. Rev. 495 (1996) (discussed infra note 937).

On Virginia, see Associated Press, Court to Help Set Standards-Death-Penalty Lawyers Affected, Jan. 4, 2002 ("Beginning Jan. 1, the state Supreme Court becomes involved in setting standards for lawyers qualified to represent people facing the ultimate legal trouble: the death penalty. Lawyers appointed to represent indigent Virginia clients receive the lowest compensation in the nation. The change in the law represents a step by the state to compile a better list of attorneys with the skills and training necessary to defend people accused of capital crimes.").

39. See supra p. 2 & n.16.

40. See Changes in the Death Penalty Around the U.S., supra note 19.

41. See, e.g., Hon. Bill Delahunt, Protecting the Innocent, Harvard Crimson, May 14, 2001; Harwood, supra note 1:

    In Congress, legislation that would create financial incentives for states to expand access to DNA testing and set standards for legal representation of defendants in capital cases is gathering support in both parties. In the Senate, its 19 co-sponsors include four Republicans and last year's Democratic vice presidential candidate, Joseph Lieberman, who declined to back the bill a year earlier. Its . . . co-sponsors in the House include several members of the GOP's conservative wing.

    GOP Rep. Mark Souder of Indiana, one of the co-sponsors, says, "I support the death penalty, [but] I'm a little uncomfortable. We want to be more sure."

See also Jost, supra note 1 (as of November 2001, bill had 21 co-sponsors in Senate); Brooke A. Masters, Executions Decrease For the 2nd Year: Va., Texas Show Sharp Drops Amid a National Trend, Wash. Post, Sept. 6, 2001 ("The federal Innocence Protection Act, which would provide DNA testing and set minimum standards for court-appointed defense lawyers, also continues to make progress. The House version has 210 sponsors, close to a majority. In the closely divided Senate, several moderate Republicans have recently come out for the bill."). On Senator Warner's views, see infra note 77.

42. See Ryan Says He Won't Run for Re-Election, Chi. Trib. Aug. 9, 2001 ("Above all, [Ryan] pointed with pride to his death-penalty moratorium. Ryan vaulted into the national spotlight in January 2000 by halting executions indefinitely after a string of men were released from death row either because evidence showed they were innocent or because they received unfair trials. Since Illinois reinstated capital punishment in 1977, 12 death row inmates have been executed while the sentences of 13 others have been overturned. A Ryan-appointed commission is studying the state's death penalty system. 'During the past three years, I've talked to people all over the world about the shocking condition of our capital punishment system,' Ryan said. 'I knew then as I do now that I had no choice but to call a halt to what was obviously a deeply flawed process.'"); 7 of 8 Candidates for Governor Would Prolong Execution Freeze, St Louis Post Dispatch, Oct. 29, 2001 ("7 of the 8 candidates for Illinois governor [including the state's conservative Republican attorney general and one other leading Republican candidate] say they would keep the state's freeze on executions until it is proved that innocent people aren't being sentenced to death. It's a political oddity in these 'law-and-order' times-one that dramatizes how jittery Illinoisans are about a justice system that has had to release more death row inmates than it has executed."); supra p. 1 & nn.4-6.

43. See Gerald F. Seib, Bush's Race Issue: What's the Role of Death Penalty, Wall St. J., Feb. 28, 2001 (quoting Governor Rowland, a death penalty supporter, stating that Republicans and especially the Bush Administration need to take seriously African American citizens' doubts about the fairness of the death penalty).

44. See Facing Death, Santa Fe New Mexican, Oct. 28, 2001 ("Gov. Gary Johnson-for years a staunch advocate of the death penalty-now says that eliminating capital punishment might be good public policy. . . . [I]n a letter being sent this week to the hundreds of people from around the world who have written to him about [an impending] execution, Johnson, who campaigned in 1994 and 1998 as a supporter of capital punishment, said, 'My mind is not closed on the subject.' 'I am of the opinion that swift and sure punishment deters crime,' Johnson wrote. 'Currently, I do not believe that New Mexico's death penalty serves as an effective preventative measure because it is neither swift nor sure. The time period currently allowed for appeals under the process is too long and yet I have come to believe that innocent people might be put to death if these safeguards are not in place.' . . . . '[E]liminating the death penalty in the future may prove to be better public policy given the reality of the sentence today.'"). See also Gilbert Gallegos, Johnson Yanks Support from Time Cap on Sentence Appeals, Albuq. Trib., Nov. 6, 2001 (discussing Governor Johnson's decision to withdraw legislation he had previously proposed that would have placed a two-year cap on the length of capital appeals; Johnson now believes that "limiting death-row appeals would probably lead to innocent people being executed," a view he bases in part on "a case in the 1970s in which four members of a motorcycle gang were wrongly accused and convicted for murdering a University of New Mexico student. The four men were later released from death row after another man admitted to the murder. 'That case proved that had I passed my law, someone like that might have been sentenced to death,' Johnson said. 'That scared me to death. I've had several things happen to me causing me to reexamine my position.'" (discussed infra p. 419 & n.966)). See also Barry Massey, As the 2002 Legislature Convenes, Gov. Gary Johnson Says He Will Consider Repeal of the Death Penalty, Santa Fe New Mexican Jan. 16, 2002.

45. See John Harwood, Bush May Be Hurt by Handling of Death-Penalty Issue, Wall St. J., Mar. 21, 2000 (noting that "independent Gov. Jesse Ventura of Minnesota ha[s] abandoned his former support for capital punishment"); David Shaffer, Though Most in State Back Death Penalty, Support is Decreasing, Minneapolis-St. Paul Star Trib., Mar. 20, 2000 ("Gov. Jesse Ventura, a one-time death penalty advocate, also has changed his mind. In February, he said on 'Meet the Press' that he no longer supports it because the risk of putting an innocent person to death bothers his conscience. The government has no right to take someone's life, he said.").

46. See Robert Reno, Support for Death Penalty Goes Wobbly, Des Moines Reg., June 12, 2000 ("The most recent defector from . . . capital-punishment . . . is Oliver North," who recently declared, "'I think capital punishment's day is done in this country. I don't think it's fairly applied."). See also Murray Campbell, Capital Punishment: Bush Faces a Shift in Public's Mood, Toronto Globe & Mail, Feb. 21, 2001 ("The [death penalty] has ceased to split Democrats and Republicans; conservatives such as Pat Robertson, Oliver North and George Will have criticized the death penalty. Even John DiIulio, director of Mr. Bush's new White House Office of Faith-Based and Community Initiatives, is reported to have abandoned his support for capital punishment."); infra note 57 (John DiIulio's views).

47. See George Will, Innocent on Death Row, Wash. Post Apr. 6, 2000 at A23.

48. See Bruce Fein, Death Penalty Ignominy, Wash. Times, Mar. 20, 2001 ("For a select category of barbaric crimes, the death penalty is justified . . . [b]ut it is disgraceful for the government in cases to deny an indigent accused at least mediocre defense counsel to bolster what may be the chief safeguard against executing the innocent . . . .").

49. See Paul Craig Roberts, Wash. Times, June 19, 2001 ("An ever-growing number of books, innocence projects and overturned convictions speak to the unreliability of conviction. A surprising number of death row inmates have been discovered to be innocent of the capital offense for which they were convicted. A criminal justice system that convicts innocents on the serious charge of murder is certain to convict innocents on less serious charges as well.").

50.Dreher's views are particularly relevant to the impact of legal error on faith in the capital system:

    FBI Director Louis Freeh did what Pope John Paul II has not been able to do: turn this law-and-order Catholic conservative against the death penalty.

    Timothy McVeigh's guilt is unquestionable, as is the gravity of his crime. If ever there were a case where it was important for the government to play by the rules, it was this one.

    And still, they fumbled.

    To be sure, the 3,000 pages of documents the FBI failed to turn over to McVeigh's lawyers will not exonerate him. McVeigh did it. He admits he did it. He deserves to die.

    That's not the same as saying the state should put killers to death. After the McVeigh debacle, who can trust our government to administer capital punishment?

    We know all about McVeigh's saga. But what of the anonymous cases where the defendant's guilt is less obvious, and law enforcement feels less of an obligation to do things by the book?

    Consider the shocking scandal unfolding in Oklahoma City, of all places, involving the work of police chemist Joyce Gilchrest. Over 14 years, the state won hundreds of felony convictions based in part on her expert testimony. But she has long been criticized for sloppy work.

    Last week, a judge released a convicted rapist who had been behind bars for 15 years after an independent lab analysis contradicted her findings.

    At least that poor sod has his life. Eleven men have been executed for convictions won partly on the basis of Gilchrest's testimony. Twelve others sit on death row. The state is investigating.

    ******

    So why is McVeigh a tipping point? I can't say for sure, but there's something about the psychology of this case that magnifies the FBI's relatively insignificant error. If they screw up even with McVeigh . . .

    That sets a hairline crack in the foundation of our justice system, one that collapses the rock-solid faith in it one must have if one is to support state-sanctioned killing.

    In the end, McVeigh will probably be executed, and justice will have been done. On that day, I won't feel any compassion for the guy, but I will feel terrible for those unknown innocents yet to be executed so we could be free to whack Tim McVeigh.

    Think of men like Anthony Faison and Charles Shepard, who were freed from prison yesterday after someone else confessed to a 1987 murder that sent them up the river-and a key state

    witness admitted to lying on the stand.

    Had they been convicted in a death-penalty state, which New York was not at the time, they might be dead today.

    ******

    We conservatives cannot afford to let our justified outrage at unrepentant killers like McVeigh make us morally indifferent to the deadly and irrevocable peril in which society places the truly guiltless on trial for their lives.

    At some point in this death-penalty debate, the sanctity of innocent life demands that men and women of conservative conscience have to say: Enough.

Rob Dreher, Feds Shake Faith in Death Penalty, N.Y. Post, May 15, 2001.

51. See Seib, supra note 43 (urging President Bush to rely on his "death-penalty credentials" to "open the door to a national debate on the justice system" and especially the "death penalty [which] hits blacks disproportionately").

52. See Death Penalty Information Center, Conservative Leaders Join Moratorium Group in Urging President Bush to Suspend Federal Executions, http://www.deathpenaltyinfo.org/whatsnew.html (visited June 27, 2001) ("Citizens for a Moratorium On Federal Executions(CMFE), a Washington-based group of prominent citizens concerned about the death penalty, sent a letter to President Bush this week calling for a halt to federal executions until lingering questions about its fairness can be resolved. The letter was signed by both well-known death penalty opponents and conservative leaders such as Emmett Tyrell Jr., editor in chief of The American Spectator and John Whitehead, founder of the Rutherford Institute."); R. Emmett Tyrrell, Jr., Capital Brutality, Am. Spectator, Feb. 1, 2001, at 13.

53. See Milt Bearden, Death Penalty Would Hinder Anti-Terrorism, Wall St. J., June 4, 2001. See also Wayne Woodlief, Gutsy Reps Voted Bravely-Death Penalty Nays Stood Up for True Justice, Boston Herald, Dec. 23, 2001 (conservative columnist praising five Massachusetts congressmen for voting against a "death penalty add-on" to an extradition bill, which "makes it harder to bring terrorist suspects arrested abroad to justice in the United States," given the refusal of Canada, European nations, South Africa and other countries "to extradite suspects from the al-Qaeda network if they will face execution" in the U.S.).

54. James Orenstein, In this Death Penalty Case, the Choices Were Too Few, N.Y. Times, June 17, 2001 (concluding that the recent execution of federal prisoner Juan Raul Garza "forces us to confront troubling questions about not only the general fairness of a capital punishment system that has a disproportionate impact on African-Americans and Hispanics, but also the fairness of depriving Mr. Garza of a basic protection that every federal inmate on death row has received," i.e., an instruction informing the jury that it had the alternative of sentencing the defendant to life without parole").

55. See Brooke A. Masters, Standards for U.S. Executions Proposed: Nonpartisan Committee Seeks Broad Consensus on Reforming Death Penalty, Wash. Post, June 27, 2001, at A4.

56. See Harwood (Wall St. Journal), supra note 1:

    Increasing opposition to capital punishment among religious leaders helped fuel the shift in opinion. Catholic bishops have called for the abolition of capital punishment as part of the "ethic of life" that leads to their opposition to abortion. In early 1999, then-Missouri Gov. Mel Carnahan commuted the death sentence of one inmate after receiving a personal plea from the Pope. Last year, televangelist Pat Robertson, a former Republican presidential candidate, called for a moratorium on capital punishment, after earlier unsuccessfully lobbying Mr. Bush to spare the life of convicted Texas murderer Karla Faye Tucker.

57. See John J. DiIulio, Abolish the Death Penalty, Officially, Wall St. J., Dec. 15, 1997, at A23.

58. See Ken Camp, Texas CLC Seeks Stay of Executions, The Baptist Standard, Apr. 23, 2001. The Baptist Life Commission is the agency that studies ethical issues for the Texas Baptist Convention, which is affiliated with the Southern Baptist Convention, a denomination with a strong position in favor of capital punishment. See National Catholic Reporter, June 30, 2000 (quoting recent resolutions by Southern Baptists supporting capital punishment).

59. See State Methodist Organization Calls for Moratorium on Death Penalty, The Associated Press June 9, 2001.

60. See John W. Whitehead, Passing a Moratorium on Federal Executions Would be a Giant Step for a Compassionate Conservatism, Letter to President George W. Bush, Apr. 3, 2001.

61. See Caryle Murphy, 'An Eye for an Eye' Challenges Faith, Wash. Post, May 13, 2001 ("Most Christian evangelicals, believing that the Bible condones the death penalty, continue to back it. But some are rethinking it since Texas's 1998 execution of Karla Faye Tucker, who became a born-again Christian in prison. 'One should not underestimate the significance of [her] execution to the psyche of American evangelicals,' said Richard Cizik, spokesman for the National Association of Evangelicals. 'It's left a long-standing impression. It certainly did with me. It was wrong.' Her death also led the influential evangelical magazine Christianity Today to declare that the death penalty 'has outlived its usefulness.'). See also Peter Steinfels, Belief, N.Y. Times, May 12, 2001:

    Only a few years ago, few lost causes seemed quite as irretrievably lost as opposing the death penalty. Not only did 8 out of 10 Americans favor capital punishment in 1994, but that proportion had been growing for over a quarter of a century.

    Such overwhelming support may still exist for the execution of Timothy J. McVeigh, but not for the death penalty generally. This shift in public opinion is remarkable . . . in itself. But so are two other things. First, a major factor behind the shift is religion. Second, virtually no one has thought to complain about that fact.

62. See Alan Johnson, Justice Has Change of Heart on Ohio Death-Penalty Law, Columbus Dispatch, Apr. 26, 2001:

    Campaigning for attorney general in 1990, Paul E. Pfeifer tore into Democrats for not zealously enforcing the death-penalty law that he'd helped write nine years earlier as a state senator.

    *****

    Now, as an Ohio Supreme Court justice responsible for weighing life-and-death cases, the 58-year-old Pfeifer sees the issue far differently.

    *****

    He doesn't oppose the death penalty, but he clearly is wrestling with the ramifications of the law he co-wrote and steered as Senate Judiciary Committee chairman.

    *****

    [Recently h]e has taken a close look at how capital-punishment cases are reviewed by the high court, a process he says is handcuffed by a lack of information and time.

    Further, he has called for Gov. Bob Taft to form a blue-ribbon panel to evaluate all 201 Death Row cases to see how many could be commuted to life in prison without parole.

    "We know we're not going to execute all these people on Death Row," Pfeifer said. "This state is not going to start executing people at the clip of 10 a year, which is what you would have to do to stay current at the pace we're putting them in."

63. See Rethinking Justice: a Cleburne Judge Who Helped Write the State's Laws Says He Now Has Concerns about the Death Penalty, Fort Worth Star-Telegram, July 24, 2001:

    30 years after helping craft Texas' capital punishment laws, Senior State District Judge C.C. "Kit" Cooke is questioning the fairness of the death penalty.

    During a statewide legal seminar last week in Corpus Christi, Cooke recalled the 11 death penalty cases he has presided over during his 23-year judicial career. He talked about how those trials have altered his feelings regarding Texas' ultimate punishment.

    "I think the mood is changing in this country and people are realizing there are deficiencies in the system," said Cooke, 54, of Cleburne . . . "We always think we've got the right person, but the system is not infallible. . . . ."

    Although Cooke still supports the death penalty, he has concerns about possible deficiencies in the system such as inadequate legal representation, access to DNA testing and the racial disparity of those executed. "I was looking at it as a young politician, with about 90% of my district supporting the death penalty," said Cooke, referring to his service in Johnson County as a state representative when he was 24. "Now, from a judge's perspective and taking care of people's rights, I think it has a lot of flaws . . . ."

64. See Lise Olsen, The Death Penalty: Uncertain Justice, Seattle Post-Intelligencer, Aug. 6, 2001 (noting that death penalty "[c]ritics have a respected ally in former [Washington] Supreme Court Chief Justice Richard Guy, who spearheaded numerous reforms before retiring this year. Guy concedes they didn't go far enough. He's convinced the state's capital punishment system is broken-appeals take too long and cost too much. 'The legislature has the obligation to look at whether the cost is worth what we pay for it in terms of money, time and facts that have torn people apart,' he said."); Lise Olsen, State's Chief Justice Calls for Reform in Death Penalty Defense System, Seattle Post-Intelligencer, Aug. 18, 2001 ("Concerned about a capital punishment system marred by bad lawyering, the chief justice of the Washington Supreme Court is recommending that the state assume more responsibility for funding death penalty defense teams and screening lawyers assigned to the high-stakes cases. Chief Justice Gerry Alexander said the state should pick up at least part of the tab for capital defense statewide to take pressure off rural counties and help ensure local judges appoint qualified attorneys and not cut corners on legal fees.").

65. For example:

  • Legislative reform efforts in Illinois are being led by a former prosecutor and Republican state representative, Jim Durkin, and are supported by the Illinois Association of Police Chiefs. See supra note 6.

  • "Ohio's capital-punishment law . . . is under fire from an unlikely source-conservative, faith-based Republicans . . . [who are] lining up with liberal Democrats . . . to support a study of the death-penalty law." See Alan Johnson, Ohio Study of Execution Law Sought, Columbus Dispatch, June 18, 2001.

  • Conservative Nevada Senator Mark James (R-Las Vegas), "who has worked nearly a decade to toughen Nevada's criminal laws," is urging the state Assembly to adopt a moratorium on executions while Nevada's capital punishment system is studied. "[N]ever has the death penalty been the subject of a study in this state that I know of," said James. "And the system is broken." James also expressed concern about who receives the death penalty in Nevada. "The Assembly passed a bill unanimously which prohibits racial profiling in traffic stops and yet 40 percent of those on death row are African Americans and nobody says a thing about it." He added, "There is not a single person on death row that had a fully funded private defense. If you're rich, you're not going to get capital punishment-period." Senator Says Justification for Death Penalty Study Obvious, Associated Press (AP) Newswires, April 21, 2001.

  • Virginia Delegate Frank Hargrove (R-Hanover County), recently introduced a bill in the Virginia General Assembly to abolish the death penalty. Hargrove, who once proposed bringing back hanging as a method of execution in Virginia, said that he was now troubled by the possibility that mistakes could lead to the execution of an innocent person. Another conservative Virginia Republican, Jeannemarie Devolites (R-Fairfax County), is sponsoring a moratorium bill which would suspend executions until after the Joint Legislative Audit and Review Commission completes its review of the state's death penalty system. "I think there's a lot of concern, not just from legislators, but from the public as a whole, that we could be executing innocent people," said Devolites. A third Virginia legislator, Delegate Robert Marshall (R-Prince William), has proposed legislation that would require that capital juries be told that Virginia nearly executed an innocent man, Earl Washington, who was eventually exonerated by DNA evidence. Death Penalty Information Center, Changes in the Death Penalty Around the U.S.: Virginia, http://www.deathpenaltyinfo.org/newvoices.html#VA (visited June 27, 2001) (quoting Associated Press, Jan. 20, 2001).

  • "In eight years as St. Louis [Missouri's] circuit attorney, Dee Joyce-Hayes stood by her assistants as they argued for the death penalty . . . . Despite that history and 12 years as a trial prosecutor herself, Joyce-Hayes, just seven months out of office, denounced the death penalty . . . .'I began over a period of time to have experiences that certainly caused me to question at least the efficacy of the death penalty, if not the moral implications of it,' said Joyce-Hayes." Elizabethe Holland, Joyce-Hayes Speaks Against Death Penalty, St. Louis Post-Dispatch July 30, 2001.

  • The strongly pro-death penalty Orlando Sentinel recently published an editorial calling for a study and moratorium:

      Capital punishment historically has played a small, but significant, role in Florida's criminal-justice system, a role this newspaper has supported for years. Fairness and the integrity of that system, however, demand a new look. Florida leads the nation in the number of people-21-who have been removed from death row in the past two decades because the defendants were later found innocent or because of serious flaws in the way cases were handled. Those flaws included misconduct by police and prosecutors and inept defense lawyers. Illinois is a distant second with 13 people removed from its troubled death row. In fact, during the past 25 years, Florida has executed 51 people. Is there any chance that an innocent person was among them? Considering the number of people removed from death row, that question should disturb any reasonable Floridian. . . . Gov. Jeb Bush, . . . should convene a special bipartisan commission to scrutinize capital punishment, and he should delay any further executions until that investigation is completed.

    Editorial, Revisit the Death Penalty: Gov. Bush Should Create a Panel on the Death Penalty and Delay Executions, Orlando Sentinel, Aug. 19, 2001.

  • "Last month, the name of [Texas's] death row, the Terrell Unit, was changed after Charles Terrell, a former chairman of the Texas Board of Criminal Justice, asked that his name be removed. Mr. Terrell made the request in part because he questions the fairness of the capital punishment system in his home state." Yardley, Texas Wavering on Death Penalty, supra note 19.

See also Peter Beinart, Mercy Seat, New Republic, June 11, 2001 ("Washington is several years behind public opinion and the states on the issue [of the death penalty]. And . . . it is often state and local Republicans who have taken the lead [in reform efforts]. Last year, the Republican governor of Illinois announced a moratorium on executions. Nebraska's GOP-controlled state legislature passed one as well. In overwhelmingly Republican New Hampshire, the state legislature passed legislation outlawing the death penalty altogether. . . . [And] the Texas state legislature, including the Republican-led State Senate, [have] now pass[ed] a series of reforms . . . .").

66. See Cheyenne Hopkins, Keating Proposes Death Penalty Standard, The Oklahoman, June 23, 2001 (quoting National Press Club speech citing decline in faith in the capital system and calling for a "higher threshold" of certainty about guilt before convicting defendants subject to the death penalty that would "requir[e] a juror to go deeper in his own deliberations than the 'reasonable doubt' standard requires"); John Podhoretz, Why DNA Will Save the Death Penalty, NY Post, June 19, 2001 (advocating changes in "[s]tate and federal sentencing guidelines [that] make it possible to impose the death penalty only in cases where the physical evidence makes it absolutely certain that the accused is indeed the killer. When the evidence is circumstantial, the death penalty will not be sought."); supra note 1 (further discussion of Governor Keating's proposal). See also Death Penalty Information Center, Changes in the Death Penalty Around the U.S.: Ohio, http://www.deathpenaltyinfo.org/ Changes.html#OH (visited June 27, 2001) ("On February 15, 2001, Rep. Jones proposed a bill (HB 101) to prohibit the death penalty unless the defendant is convicted beyond any doubt of aggravated murder and an aggravating circumstance and the aggravating circumstances outweigh beyond any doubt the mitigating factors").

67 Joseph Hoffmann, Violence and the Truth, 76 Ind. L.J. 939, 940-41 (2001); see id. at 941 ("public support" for the death penalty is "dropping sharply"). See also conservative commentator John Podhoretz's view that "arguments in opposition to the death penalty are receiving a receptive hearing in a way they haven't since the early 1970s. For more than two decades, Americans approved the use of the death penalty by a three-to-one margin. Lately, that has shrunk to two-to-one, still overwhelming, . . . but a significant drop in support nonetheless." Podhoretz, supra note 66; accord Jost, supra note 1 ("'There's no doubt that public support for executions has declined.'" (quoting Dudley Sharp, resource director of a leading pro-death penalty group, Justice for All)).

68 See, e.g., Lounsberry, supra note 1.

69 "Noting that Minnesota does not have the death penalty, O'Connor noted, 'You must breathe a big sigh of relief every day.'" Bakst, supra note 3; see Charles Lane, Justice O'Connor Questions Death Penalty, Wash. Post, July 3, 2001 ("Justice Sandra Day O'Connor, whose vote often decides close cases at the Supreme Court, has added her voice to the growing chorus of skepticism about the administration of capital punishment in the United States."); Alan Berlow, A Supreme Court Shocker: Sandra Day O'Connor's Criticisms of the Death Penalty Couldn't Have Come from a More Unlikely Source, Salon, July 3, 2001, http://www.salon.com-/news/feature/2001/07/04/oconnor/index2.html; Healy, supra note 1 ("More than just a surprise, however, O'Connor's statement was the latest and perhaps clearest sign of how sharply the debate over the death penalty has shifted in recent years."). Justice O'Connor elaborated on these warnings at a Nebraska Bar Association meeting several months later. See John Fulwider, O'Connor Lectures Lawyers, StatePaper.com, Nebraska, Oct. 18, 2001 (in Justice O'Connor's speech to the Nebraska Bar Association, she said that "innocent people may well continue to receive the death penalty if lawyers don't . . . start doing more pro bono work for indigent defendants. . . . 'More often than we want to recognize, some innocent defendants have been convicted and sentenced to death,' O'Connor told members of the Bar Association . . . . 'That will continue to happen unless qualified lawyers take up their cases,' she said.")

    For a similar view from a front-line federal judge, see Judge Michael Posner, Life, Death and Uncertainty, Boston Globe, July 8, 2001:

    To the [federal district] judge in charge, the murder trial of Kristen Gilbert offered an unsettling lesson-and inescapable conclusion-about the ultimate cost of the death penalty. . . .

    *****

    Presiding over this, the 1st death penalty case in Massachusetts in several decades, was the most complicated and stressful thing I've ever done (aside, perhaps, from raising teenagers).

    The experience left me with one unavoidable conclusion: that a legal regime relying on the death penalty will inevitably execute innocent people . . . . Any honest proponent of capital punishment must face this fact.

See also the like-minded concerns of Ohio Supreme Court Justice Paul E. Pfeifer and Texas Senior State District Judge C.C. "Kit" Cooke, both of whom-like Justice O'Connor when she was majority leader of the Arizona state legislature-played important roles in adopting those states' modern death penalty laws.

70 See Blakemore, supra note 22 (also discussed in Richard Morin & Claudia Deane, McVeigh's Execution Approved, While Principle Splits Public, Wash. Post, May 3, 2001, at A9) (even when questioned on the eve of the McVeigh execution, only 63% of Americans-the lowest proportion measured in years-expressed support for the death penalty on a recent ABC-Washington Post poll); Jonsson, supra note 1 ("While a majority of Americans still support the death penalty, that number has slipped from 77 percent five years ago to 63 percent, according to polls); Jost, supra note 1 ("Support for the death penalty has dropped since reaching its peak in 1994. During roughly the same period, the percentage of people against capital punishment doubled.").

Other recent polls have measured support for the death penalty between 59% to 67%-in all cases the lowest each respective poll has measured in years. See Peter D. Hart Research Associates, Inc., Poll Finds Support for Death Penalty Alternatives and for System Reforms, Mar. 27, 2001 ("A national poll recently conducted by Peter D. Hart Research Associates found only 60% favored the death penalty for persons convicted of murder"); Jeffrey M. Jones, Two-Thirds of Americans Support the Death Penalty, Gallup News Service, Mar. 2, 2001 ("Since [1994], support [for the death penalty] has declined, dropping to the current level of 67%. Even news of the Oklahoma City bombing and the federal government's seeking (and obtaining) the death penalty for McVeigh in 1995 did not reverse the downward trend."); The Declining Support for Execution, N.Y. Times, May 10, 2001 (discussing various polls including one by the Pew Research Center pegging support for the death penalty at 66%); Sparing Innocents, Boston Globe, Apr. 3, 2001 (reporting Harris Poll showing national support for death penalty at 64%); Richard Willing, Even for Death Penalty Foes, McVeigh Is the Exception, USA Today, May 4, 2001, at 1A (support for death penalty drops to 59% when respondents are given choice between supporting the death penalty, opposing it except for mass murder Timothy McVeigh (22%), and opposing it in all cases (16%)). See also Lichtblau, supra note 1 ("Recent polls show that, while a majority of Americans still favor the death penalty, the numbers are shrinking. California saw a particularly sharp drop, with support declining from 78% in 1990 to 58% last year, according to a Los Angeles Times Poll."). See generally Samuel R. Gross & Phoebe C. Ellsworth, Second Thoughts: Americans' Views on the Death Penalty at the Turn of the Century, in Capital Punishment at Century's End: New Insights, Old Doubts (Stephen P. Garvey ed. 2001).

71 See Duggan, supra note 19 ("Of the 38 states with capital punishment laws, 35 offer juries the life-without-parole choice . . . ."); Editorial, McVeigh Errors Raise Doubts About Other Capital Cases, USA Today, May 16, 2001 (reporting that Wyoming recently adopted life without parole as an alternative to the death penalty, making it, the 36th death penalty state (out of 38) to do so); The Justice Project, supra note 8 (reporting that, during its 2001 legislative session, Oregon adopted life without parole as the alternative to a death sentence, making it the 37th death state (out of 38) to do so). On Texas, see supra p. 2 & n.18.

72 See, e.g., Morin & Deane, supra note 70 (in a May 2001 poll, 46% favored the death penalty over life without parole, while 45% favored life without parole-up from 38% measured two years ago); Peter D. Hart Research Associates, supra note 70 (in national poll in mid-March 2001, 38% favor death penalty versus 48% favoring alternative of life without parole and restitution to victim); Jones, supra note 70 (in national poll in late February 2001, 54% favor death penalty versus 42% favoring life without parole). See also Illinois Poll Finds Drop in Death Penalty Support, Peoria J. Star, Jan. 30, 2001 ("More Illinoisans would rather see a murderer in prison for life without the possibility of parole [47% favored this option] than sentenced to death [33% favored this option], according to a survey just released . . . .").

73 See, e.g., Morin & Deane, supra note 70 (in recent ABC-Washington Post poll, "fifty-one percent of those interviewed favored halting all executions until a commission is established to determine whether the death penalty is being administered fairly while 43 percent opposed a halt. . . . More than four in 10 [death penalty supporters] supported a moratorium . . . [, and] the proportion [of all respondents] who favored a halt in executions rose to 57 percent when respondents were reminded that the governor of Illinois recently stopped all executions in his state while a commission reviews how the death penalty has been applied."); Peter D. Hart Research Associates, supra note 70 ("72% [of Americans polled in national survey] favored suspension of the death penalty until questions about its fairness can be studied, up from 64% in August 2000"); Jeffrey M. Jones, Americans Closely Divided on Death Penalty Moratorium, Gallup News Service, Apr. 11, 2001 ("The public's support for a moratorium ranges between 53% and 42% depending on exactly how the concept is presented to them."); Governor Asked to Halt Executions, Las Vegas Sun, Oct. 31, 2001 ("A recent Field Poll found that as many as 73 percent of Californians support a moratorium."). See also Healy, supra note 1:

    In the early to mid-1990s, . . . polls showed that 75 percent to 80 percent of Americans supported the death penalty. It was the highest level of support since polling began in the 1930s, and few experts predicted that it would drop any time soon.

    But that is precisely what has happened over the past two or three years [with] polls show[ing] that the level of support has fallen to between 60 percent and 65 percent, the lowest point since the mid-1970s.

    *****

    Though public support for capital punishment has not fallen below 60 percent in nearly three decades, researchers says a further drop in support is not unthinkable. Of the 60 percent to 65 percent of Americans who say they still favor the death penalty, roughly half express some doubts, which means they could eventually switch sides altogether.

For polling information indicating the source of public concern about the death penalty, see Morin & Dean, supra note 70 (reporting that in recent ABC-Washington Post poll, "68 percent said the death penalty is unfair because 'sometimes an innocent person is executed,'" while "63 percent agreed that capital punishment is unfair because 'it's applied differently from county to county and state to state"; for the first time since the question has been asked, starting in 1985, a majority of Americans believe that the death penalty is not an effective deterrent, a view shared by 40% of death penalty supporters).

In response to a nationwide death penalty moratorium campaign, nearly 60 cities and counties around the nation have adopted resolutions calling on state governors and legislatures to halt executions while the issue is studied. See, e.g., Council on Record Favoring Death Penalty Moratorium, Lincoln J. Star, Oct. 30, 2001 ("The Lincoln [Nebraska] City Council tilted 4 to 3 Monday in favor of a death penalty moratorium . . . follow[ing] 3 hours of thoughtful debate."); Martin Dyckman, Death Penalty Moratorium Has Victory, St. Petersburg Times, Jan. 13, 2002 (discussing vote by Tallahassee City Commission calling for a moratorium on executions in Florida); Maria Alicia Gaura, Santa Clara County Wants Execution Suspended; Davis Asked to Review Process, S.F. Chon., Oct. 31, 2001 (discussing Santa Clara County Board of Supervisors's adoption of a resolution urging the governor to impose a death penalty moratorium in California; six other California municipalities, including Oakland and San Francisco, have previously passed such resolutions). At the forefront of this movement are 14 city counsels in North Carolina-those of Asheville, Carrboro, Cary, Chapel Hill, Charlotte, Cofield, Davidson, Durham, Fayetteville, Greensboro, Hillsborough, Orange, Thomasville, and Winston-Salem. Other cities that have adopted such resolutions are Atlanta, Baltimore, Buffalo, Detroit, Harrisburg, Hartford, New Haven, Philadelphia, Pittsburgh, Rochester, San Francisco, Tucson and Wilmington (Delaware).

74 See, e.g., Bureau of Justice Statistics Bulletin, Capital Punishment 1999 (NCJ 184795) and Capital Punishment 2000 (NCJ 190598) tbl.5 & App. 1 (Dec. 2000, 2001) (only 214 death sentences were imposed in 2000, compared to 272 in 1999 and 300 in 1998; the number of death sentences imposed in 2000 was the lowest since 1980; the drop from 1999 to 2000 was the largest in any year since 1976-1977); Bill Sloat, Fewer Killers Going to Ohio Death Row, Cleveland Plain Dealer, Aug. 21, 2001 ("Only three convicts were sentenced to death row [in Ohio] last year, an all-time low. And so far this year, there have been five. Criminal justice statistics show a steep decline in the overall percentage of aggravated-murder convictions carrying the death sentence starting in 1999. Juries began turning to a punishment relatively new in Ohio: Life without parole. . . . Last year, 27 percent of Ohio's aggravated murder convictions carried life without parole, the highest number ever. Four percent were for death, the lowest percentage since records have been kept. . . . It's a dramatic shift from 1998, when sentences for death and life without parole were nearly even. 'Usually we've been up in the double digits with 15, sometimes 17 a year,' Reginald A. Wilkinson, director of the state prison system, said of the inmates entering death row. 'Now the numbers are way down.'. . . [C]riminologists, corrections officials, academics and prosecutors say there could be other reasons for the decline. Some speculate that it may be related to falling crime rates. . . . Other experts suggest the number dropped because juries and judges might be reluctant to sentence people to death because they now know executions actually take place in Ohio. . . . Others say it could be that there are concerns there might be innocent people on death row."); Willing, Support for the Death Penalty Might Be Declining, supra note 1 ("U.S. juries last year sentenced fewer convicts to death than in any year since 1980, according to new Justice Department research that, along with a drop in executions, reflects what some analysts say appears to be waning enthusiasm for capital punishment. . . . The numbers suggest that questions about whether the death penalty is being applied fairly might be resonating across the USA."); supra note 30 (discussing downturn in death sentences in Oklahoma); infra note 233 (reporting decline in death sentences being imposed by California jurors). See also Harwood, supra note 1 ("The number of people annually sentenced to death in the U.S. has fallen in three of the last four years for which statistics are available, to 272, in 1999, since peaking at 319 in 1994 and 1995."); Death Penalty Information Center, Death Row Declines, http://www.deathpenaltyinfo.org/whatsnew.html (noting that "the number of prisoners on death row as of April 1, 2001 is 3,711, down from earlier this year. This decrease reverses a long trend since the death penalty was reinstated [30 years ago] of consistently increasing death row size." (emphasis added)).

75 See Death Penalty Information Center, The Death Penalty in 2001: Executions Decline; Death Row Numbers Also Drop (Dec. 2001) (noting that Texas and Virginia, which have accounted for around half of all executions for about a decade, accounted for less than 30% of the 2001 executions); Jim Yardley, Number of Executions Falls For Second Straight Year, N.Y. Times, Dec. 14, 2001 (66 executions in 2001, "down from 85 in 2000 and 98 in 1999 . . . is the first time since executions resumed in 1977 that the number of executions has fallen in consecutive years . . . . Equally notably, Texas, the perennial leader in executions saw a decline to 17 executions in 2001, compared with a record 40 last year, and trailed Oklahoma, which led the nation by executing 18 convicted murderers. The drop in executions concludes a year in which capital punishment was debated in statehouses across the country as polls showed growing public concern that an innocent person could be put to death."). See also Brune, supra note 1 ("[T]he fast pace of state executions appears to have slowed as the nation takes stock of how capital punishment works."); Brooke A. Masters, Executions Decrease For the 2nd Year: Va., Texas Show Sharp Drops Amid a National Trend, Wash. Post, Sept. 6, 2001 ("[F]ive of the 10 states that have executed the most people-Louisiana, South Carolina, Alabama, Arizona and Georgia-have not executed anyone in 2001 [as of September 6]. In Alabama, all four scheduled executions were stopped by the state or federal courts. Maryland has not carried out an execution since 1998.").

76 See Patrik Jonsson, Governors Soften on Death Penalty: Numbers of Commuted Sentences, While Sill Small, Rise as Fairness Questions Grow, The Christian Science Monitor, Oct. 23, 2001:

    [North Carolina Governor Mike] Easley's uncertainty about the fairness of the conviction [of Robert Bacon, Jr., leading to Easley's recent commutation of Bacon's death sentence] is in fact part and parcel of an uptick in the number of death- row commutations granted by U.S. governors and parole boards.

    Indeed, the trend may be emblematic of a stirring debate from the Carolinas to California- not about the morality of the death penalty, but how fairly and accurately it's applied.

    As it is, the number of commuted sentences has gone from an average of one per year since the Supreme Court reinstated the death penalty in 1976 to an average of three per year since 1999. (The one-a-year average in the 1980s is high, skewed by anti-death penalty governors like New Mexico's Tony Anaya, who once commuted five people at once.) By contrast, the recent decisions come from a broad swath of social conservatives such as Oklahoma Gov. George Keating and pro-death penalty Democrats like Easley.

See also Estes Thompson, Easley Commutes Sentence of Man who Denies Killing Girlfriend, Associated Press, Jan. 10, 2002 ("A condemned man who steadfastly denied beating his girlfriend to death in 1990 escaped execution Thursday as [North Carolina] Gov. Mike Easley commuted his death sentence to life in prison. . . . No one saw the killing and no blood or fingerprint evidence connected the attack to Alston, who had been convicted about six weeks earlier of assaulting Perry. Alston, a brick mason, contended his innocence would be proved by DNA tests on evidence that has disappeared.").

77 See Death Penalty Information Center, Innocence and the Death Penalty, http://www.death-penaltyinfo.org/innoc.html (visited Jan. 7, 2002) ("Since 1973, 99 people in 22 states have been released from death row with evidence of their innocence."). See also John Aloysius Farrell, DNA Scrutiny Tests Judicial System, Boston Globe, June 26, 2001 ("public opinion polls suggest that the 'innocence' argument has done more to undermine support for capital punishment than any other . . . ."); Masters, Executions Decrease, supra note 75 (explaining the decline in support for the death penalty to its "lowest [point] in two decades" and a "sharp" decline in executions recently to the fact that "[t]wenty-one people have been released from death row in the past three years after DNA tests or other new evidence cast doubt on their convictions," and quoting statement of Republican Senator John Warner of Virginia that "'[t]he number of cases of inmates being taken off death row says to the public that this system has faults and we've got to take greater steps to ensure guilt beyond a reasonable doubt'").

78 See supra note 77; NAACP Legal Defense and Educational Fund, Inc., Death Row U.S.A. Fall 2001; infra p. 24 & nn.127, 128.

79 Recent events in Oklahoma have aggravated concerns that innocent people have been executed:

    When Jeffrey Pierce was convicted of rape in 1986, he lost his freedom and his family. He and his wife decided to divorce and she left Oklahoma to raise their twin infant sons as if he did not exist. To survive in prison, he learned to do two things: mind his own business and lift weights.

    But today, after maintaining his innocence throughout the 15 years he spent behind bars, Mr. Pierce, 39, was freed because DNA testing refuted the crucial testimony against him from an Oklahoma City police chemist long accused of shoddy work and now the focus of one of the most wide-ranging investigations into a police laboratory.

    *****

    Mr. Pierce had been a landscaper who happened to be working near the scene of the rape. The DNA test results that cleared Mr. Pierce, as well as a separate review of his case by the Federal Bureau of Investigation, set in motion the larger inquiry into the chemist, Joyce Gilchrist. Last week, the federal Justice Department began an investigation while Gov. Frank Keating of Oklahoma ordered a review into every felony conviction linked to Ms. Gilchrist to make certain that no one else has been wrongly convicted.

    Among those hundreds of cases are 11 in which the defendant was executed and 12 in which the defendant is on death row. Mr. Keating has expressed confidence that no innocent person has been executed.

    The investigations into Ms. Gilchrist, who analyzed forensic evidence like blood, hair, semen and fibers from 1980 until she was promoted in 1994 to a supervisory position, come as other police laboratory scientists are under scrutiny in Illinois, West Virginia and Florida. . . .

Jim Yardely, Forensic Expert Under Scrutiny as DNA Test Frees 'Rapist', N.Y. Times, May 8, 2001. See supra notes 2, 50; infra notes 84, 232.

80 See, e.g., Farrell, supra note 77 (noting that "the families of three executed men in Texas and Virginia have gone to court to try to use DNA to prove their relatives were not guilty" after officials refused to make the relevant evidence available for testing); Frank Green, DNA Tests Not Likely After an Execution: Virginia Opposing Third Request of its Kind, Richmond Times-Dispatch, Mar. 26, 2001 (discussing refusal of Virginia to release biological material for testing that would reveal whether three inmates executed for separate offenses in that state were accurately convicted and executed for rape murders) ; Brooke A. Masters, New DNA Testing Urged in Case of Executed Man: Post, Others Ask Va. Court to Release Evidence, Wash. Post, Mar. 28, 2001, at B1 (discussing thus far unsuccessful efforts of press, charities and members of the family of executed individuals to secure access to DNA-testable biological samples in cases of individuals executed by Virginia, despite consistent claims of innocence); Mr. Washington's Release, Wash. Post, Feb. 12, 2001 (discussing efforts by "Virginia officials . . . to block posthumous DNA testing that could resolve . . . remaining questions" about guilt or innocence of inmates executed by that state).

So far as we are aware, posthumous DNA testing in cases of executed individuals has occurred only twice in this country:

    Frank Lee Smith spent 14 years awaiting execution in Florida. Then, in December [2000], a DNA test cleared him of the 1985 murder of an 8-year-old girl. Unfortunately for Smith, the state opposed DNA testing until six months after he died of cancer on death row at age 52.

    *****

    Apparently the only time a court anywhere in the country has ordered DNA testing in a case where an execution has already been carried out was last year in Houston County, Ga. That testimony was requested by the [Boston] Globe, which argued that there was an "overriding right to know the result of modern DNA testing on evidence maintained in a death penalty case." Authorities . . . did not oppose the Georgia request. Testing results have thus far been inconclusive.

Green, supra. On the Frank Lee Smith case, see infra pp. 34-35.

81 See Green, supra note 80 ("The Catholic Diocese of Richmond sought the DNA in the O'Dell and Barnabei cases [both O'Dell and Barnabei consistently proclaimed their innocence before and after trial and in their last words before being executed] using a state law that permits evidence to be donated to a charity once it is no longer needed. The requests were turned down and, in the O'Dell case, the evidence was destroyed by Virginia Beach Circuit Court order [on request of the local prosecutor]."); Associated Press State & Local Wire, DNA Evidence in Coleman Case to Stay in California, Aug. 25, 2001 (discussing ruling of Virginia circuit judge that the Boston Globe, Washington Post, Richmond Times-Dispatch, Virginian-Pilot and Princeton, N.J.-based Centurion Ministry would not be permitted to test DNA of Roger Keith Coleman who was executed in 1992 for a rape murder he consistently contended he did not commit; noting that the "Virginia attorney general's office [has] opposed the new testing," and that a "Virginia court has never allowed DNA testing on evidence in a case where the convicted person has been executed [despite a string of] post-execution requests [including] in the cases of Joseph O'Dell III, who was executed in 1997, and Derek R. Barnabei, who was executed last year"). Cf. id. (quoting the clerk of the Norfolk Circuit Court, who has custody of the files in the Barnabei case: "'As long as I am the clerk, we will not destroy the evidence without a court order."). See also Brooke A. Masters, Va. Evidence Destroyed Despite Warnings to Clerk, Wash. Post, Oct. 18, 2001 ("An Arlington courthouse clerk threw away all the evidence from a 1999 death penalty case in violation of Virginia law, despite warnings from two colleagues that the material contained DNA and that the inmate's appeals were pending, according to court documents filed yesterday. . . . At least 50 exhibits, including the murder weapon, were thrown out May 23.").

States-including Louisiana and Virginia in 2001-and localities are even adopting laws and policies permitting the systematic destruction of evidence that could confirm or disconfirm the accuracy of executions. See id. ("A new law passed in the recent [Virginia] General Assembly says evidence in a capital case need only be kept by the state's Division of Forensic Science until the execution has been carried out."); Barbara Bradley, DNA Testing in Crime Cases Causing Distrust in the Criminal Justice System, NPR Morning Edition, Aug. 29, 2000, transcript available at 2000 WL 21481402 ("DNA has become one of the most powerful tools to prove the guilt or innocence of a criminal defendant. And in recent years, dozens of convicted prisoners have been released after the original biological evidence was tested again. But now prisoners wanting to take advantage of this new science are hitting a roadblock. Police and courts across the country are destroying the biological evidence that could determine whether a person has been wrongly convicted."; also discussing decision of officials in Houston, Texas to destroy biological evidence contained in court files of scores of rape and rape-murder cases).

82 See Paul F. Enzinna, Afraid of a Shadow of a Doubt (Op-ed), Wash. Post, May 7, 2000, at B8 (quoting a Virginia prosecutor's successful argument in court that Virginia should be allowed to destroy DNA samples that could have proven that Joseph O'Dell did not commit the rape murder for which he was executed in 1997).

83 Cf. Editorial, Review the Errors, Ariz. Daily Star, June 29, 2001 (editorial endorsing proposal of Arizona Supreme Court Chief Justice Thomas Zlaket to establish a special commission to determine the circumstances under which wrongful convictions occur in Arizona); Carol Sowers, Arizona Lawyers Seek 'Innocence' Panel, Ariz. Repub., Nov. 12, 2001 ("Two Valley Lawyers want Arizona to be the first state with an Innocence Commission, a public panel that would investigate when the wrong people are sent to prison.").

In every case in which an innocent person is shown to have been wrongfully convicted of murder in Canada (which does not have the death penalty), an ad hoc commission is appointed under the direction of a former Supreme Court Justice or other respected jurist or lawyer, to inquire into what went wrong and to propose reforms. See James Lockyer, Guilt Revisited: A Comparative Perspective on Canada, the United Kingdom and the United States, talk delivered at DNA and Human Rights: An International Conference, University of California, Berkeley, CA. Apr. 27, 2001. For the report of the first such commission-the Kaufman Commission, which examined the wrongful conviction of Guy Paul Morin, discovering serious abuses in the use of jail house informants and proposing ameliorative legislation that a number of Canadian provinces have adopted-see www.attorneygeneral.jus.gov.on.ca:80-/morin/morin.htm. The United Kingdom has established on ongoing Criminal Cases Review Commission with authority to examine the accuracy of convictions that have become final in the relevant nations and to propose that the Court of Appeal overturn ones shown to have reached a false conclusion. The Commission thus far has referred to the Court of Appeal about 70 wrongful conviction recommendations, over 25 in homicide cases; the court has accepted nearly all of them, including three in cases of individuals wrongfully executed in the 1950s and 1960s before the UK abolished the death penalty. See James Lockyer, Associates in the Defense of the Wrongfully Convicted, Canada, http://globetrotter.berkeley.edu/humanrights/DNA/.

84. Arnold Hamilton, Chemist's Errors Stir Fear: Did Oklahoma Execute Innocent?, Dallas Morning News, Oct. 22, 2001. The Oklahoma Attorney General is currently conducting a study of hundreds of convictions based on testimony by a compromised laboratory employee, Judith Gilchrist. See supra notes 1, 50, 70; infra note 232. Despite strong indications that Gilchrist fabricated evidence in at least one case that led to an execution, the state has given the lowest priority to cases in where the defendant was executed after being convicted in part based on Gilchrist's testimony, and has refused in the mean time to permit others to examine the forensic evidence. See, e.g., Hamilton, supra ("In Oklahoma, the 12 pending death row cases that Ms. Gilchrist worked on are being reviewed first by the state task force, according to OSBI spokeswoman Kym Koch. Those with sentences of life without parole will be reviewed next, followed by those with lesser sentences. The 11 Gilchrist cases in which inmates have already been executed will be considered last, if at all."); Lois Romano, Police Chemist's Missteps Cause Okla. Scandal, Wash. Post, Nov. 26, 2001 ("the most sensitive aspect of the [Gilchrist] investigation is whether the state of Oklahoma, relying on Gilchrist, may have executed an innocent man-or someone who would have been sentenced to life without her critical testimony. Twenty-three capital cases have been identified in which Gilchrist provided testimony. Of those, 11 convicted murderers have already been put to death, and 12 sit on death row. . . . As the Gilchrist investigation lumbers past its seventh month, the state multi-agency task force has agreed to test only three of the 12 current death row cases connected to Gilchrist-and none of the 11 in which the perpetrators have already been executed. . . . The Oklahoma Indigent Defense System, though, wants to conduct forensic testing on several additional Gilchrist death row cases, and seven of the 11 individuals already executed-but has met resistance from state Attorney General Drew Edmondson."). For further background, see Henry Weinstein, Evidence Questioned in Execution: Police Memos Contradict Chemist's Forensic Testimony at a 1982 Oklahoma Murder Trial, L.A. Times, Aug. 30, 2001:

    The investigation of a controversial Oklahoma City police chemist has produced evidence raising questions about whether she testified falsely in the 1982 rape and murder trial of a man who was executed last year, protesting his innocence to the end.

    On Wednesday, Oklahoma Atty. Gen. Drew Edmondson released two memos written last month by four Oklahoma City Police Department forensic scientists saying tests they had conducted of material found at the murder scene contradicted testimony given by Joyce Gilchrist [that semen matching Johnson's was found at the crime scene]. . . .

    Edmondson, who released the memos after they were described in [newspaper] stories . . ., said it was possible that "we might reexamine the evidence down the road." Now, however, Edmondson said his office was focused on reviewing numerous cases in which Gilchrist testified and the defendant received a long prison term or a death sentence not yet carried out. . . .

    [A lawyers working on Johnson's behalf] . . . filed a lawsuit seeking access to all records in the case. The city has opposed this request and has acknowledged that it had lost some of the records. . . .

    [DNA technology only became available after Johnson's trial, at which point he] attempted to secure DNA testing of the biological evidence. The state attorney general's office resisted, saying that under federal laws governing death penalty appeals, Johnson was ineligible for testing. Federal judges agreed, so no testing was done.

85. This is not to say that officials never hide actions that threaten or take innocent life. But when they do, their suppression of the facts is typically thought to be improper. This is not so in the death penalty context, which makes it unique.

86. Many private enterprises no longer distinguish between quality control, safety and other kinds of inspections for faults, recognizing that high rates of any kind of error signal a high risk of all kinds of error and thus are likely to indicate that the production process itself is inefficient. See, e.g., Frederick H. Abernathy, et al., A Stitch in Time: Lean Retailing and the Transformation of Manufacturing-Lessons from the Apparel and Textile Industries (1999).

87. Of the 4546 death sentences imposed during the 1973-1995 study period that were finally reviewed on direct appeal, state supreme courts found serious error and reversed 1852 (41%) of them. The raw numbers reported here, but not the percentages, are slightly different from those reported in A Broken System, Part I. Since that Report was published, we have continued to check the numbers, including based on new information received from readers of the first report, to assure that our figures are as accurate as possible. As here, this has led to modest adjustments, none of which affect our conclusions. See infras note 90, 174.

88. See infra pp. 17-18 (explaining why the 10% figure substantially understates how frequently state post-conviction courts discovered reversible capital error).

89. See supra pp. 6-7.

90. See supra note 87 (explaining why some of the numbers in the National Report Card are slightly different from those in the National Report Card in A Broken System, Part I). Although the raw numbers changed slightly, the percentages are the same, with three exceptions: The overall reversal rate for only direct appeal and state post-conviction, excluding errors discovered on federal habeas dropped from 47% to 46%; the reversal rate for only direct appeal and federal habeas reversals, but not those occurring on state post-conviction, rose from 64% to 65%; and the percentage of cases awaiting direct review rose from 21% to 22%.

91. See infra pp. 17-18 (explaining why the reversal rate at the state post-conviction stage is probably substantially higher than 10%).

92. At the state post-conviction stage, the phase of the trial affected by the reversal is known in 336 cases. Of those reversals, 126 (37.5%) overturned phase-one verdicts, and 210 (62.5%) overturned phase-two verdicts.

93. At the federal habeas stage, the phase of the trial affected by the reversal is known in 239 cases. Of those reversals, 102 (43%) overturned phase-one verdicts, and 139 (58%) overturned phase-two verdicts. (These figures sum to 101% because there were a small number of cases in which relief was granted on claims undermining both phase-1 and phase-2 verdicts.)

Latzer and Cauthen reviewed a sample of undifferentiated state direct appeal and state post-conviction reversals of capital verdicts occurring between 1990 and 1999, concluding that 39% affected the first-phase determination, while 61% affected the second-phase determination. See Barry Latzer & James N.G. Cauthen, Capital Appeals Revisited, Judicature, Sept.-Oct. 2000, at 64, 66. As we develop elsewhere, Latzer & Cauthen's sampling technique was skewed against finding phase-one error, see James S. Liebman, Jeffrey Fagan & Valerie West, Death Matters: A Reply to Latzer and Cauthen, Judicature, Sept.-Oct. 2000, at 76 (finding that 83% of cases Latzer and Cauthen's sampling technique failed to count were phase-one reversals), and they failed to distinguish between direct appeal and state post-conviction reversals. Together, our more complete state post-conviction data, see supra note 92 and accompanying text, and Latzer and Cauthen's estimates after accounting for the effect of their sampling biases, suggest that well more than 40% of direct appeal reversals during the 1990-1999 period were phase-one reversals. (Together, the fact that their bottom-line number for an undifferentiated set of reversals from both the direct appeal and state post-conviction stages is based on a systematic undercount of first-phase reversals as a result of biased search criteria, and the fact that the number they come up with is close to the accurate number for the state post-conviction phase taken by itself, suggest that an accurate count of all direct appeal reversals taken by themselves must be well above that number.)

94. See, e.g., Lowenfield v. Phelps, 484 U.S. 231 (1988) (states may impose death sentences only for murders that are capitally aggravated, but may make that determination at either the first or second phase of capital trials).

95. See Jeffrey Fagan, James S. Liebman & Valerie West, Death Is the Whole Ball Game, Judicature, Nov.-Dec. 2000, at 144-45; Death Matters, supra note 93, at 72-77, 99.

96. At the state post-conviction stage, the part of the trial tainted by reversal is known in 336 cases. Of those reversals, at least 166 (49%) undermined the decision whether the defendant committed aggravated murder, because they tainted (1) the finding of guilt of some level of murder, as determined at the first-phase of trial, and/or (2) the finding that the murder was capitally aggravated, as determined at the first phase of trial in a minority of states and at the second phase of trial in most states. At the federal habeas corpus stage, the basis for reversal is known in 239 cases. Of those reversals, at least 122 (51%) undermined the decision whether the defendant committed aggravated murder, for one or both of the same two reasons.

These figures substantially undercount reversals of decisions determining the level of offense committed. The data we collected reveal the number of death verdicts reversed based on prosecutorial suppression of evidence and incompetent lawyering at the second and third phases of trial but do not reveal the precise determinations made at those stages that were compromised by incompetent defense counsel or prosecutorial suppression of evidence. To assure a conservative estimate of the proportion of guilt-related reversals, we accordingly count all second-phase reversals on these two grounds as "sentence-related" error, even though some of these reversible errors in fact tainted the finding of aggravated murder and thus should be counted as guilt-related error.

Collecting information on the basis for reversal, and the aspect of the capital trial that it compromised, is costly and time-consuming. The limited resources available for our research permitted us to collect that information for the second stage of review (state post-conviction) and the third stage of review (federal habeas) but not for the first stage of review (direct appeal), where the number of reversals is larger. For reasons discussed at pp. 45-47 below, it is it unlikely that the proportion of reversals affecting the determination of guilt of aggravated murder is greatly different from the nearly identical rates for the two later review stages.

97. These points are developed at greater length in Death Matters, supra note 93; Death Is the Whole Ball Game, supra note 95.

98. Woodson v. North Carolina, 428 U.S. 280, 305 (1976).

99. See supra note 86 and accompanying text.

100. See infra note 680 (noting statistically significant tendency of federal judges to use "per curiam" opinions, which often are unpublished decisions, more often when reversing than when affirming death verdicts). For evidence of efforts on the part of state reviewing courts to avoid controversy generated by reversals of capital verdicts, see infra pp. 194, 218-19, 236, 257.

101. See James S. Liebman, Jeffrey Fagan & Valerie West, A Broken System (Part I): Error Rates in Capital Cases, 1973-1995, at 19-20, 26-27 & nn.39, 132; App. C, pp. C-1 to C-2 (2000) [hereinafter, A Broken System, Part I].

102. Mr. Shiffman graciously provided his data to the lead author of this Report. His report on his study appeared in the Nashville Tennessean on July 23, 2001.

103. Nationally, the proportion of death verdicts passing inspection on direct review was 59%: 100% - 41% reversed on direct appeal = 59%. If the 300% Tennessee adjustment holds nationally, the proportion passing both direct appeal and state post-conviction inspection is 41%: 59% passing direct review inspection - 30% of those 59% that were reversed on state post-conviction = 59% - 18% = 41%. And in that event, the proportion passing all three inspection stages is 25%: 41% passing inspection at the first two stages - 40% of those 41% that were reversed on federal habeas = 41% - (.40 x 41%) = 41% - 16% = 25%. The overall error rate, in that case, is 75%.

To see how large a difference it makes to account for the many verdicts that got stuck in the review process and were not finally reviewed at the state post-conviction stage, consider a state that imposed 170 death verdicts during the study period, 41% (or 70) of which were reversed on direct appeal, leaving 100 verdicts available for state post-conviction review. If there were 10 state post-conviction reversals in the state during the period, we conservatively estimated the error rate as 10% (10/100). But if at least 20% of those 100 cases were not in fact finally reviewed at the state post-conviction stage during the study period (as is almost certainly true of all states), the actual rate of reversible error at the state post-conviction stage was at least 30% higher than our estimate. (10/75 = 13%. 13%-10% = 3%. 3/10 = 30%.) If, instead, 40% of the available cases were unreviewed (as was certainly true in a number of states), then A Broken System underestimated the error rate by 70%. (10/60 = 17%. 17%-10% = 7%. 7/10 = 70%.)

Our best estimate is that around 45% of all verdicts available for state post-conviction review during the study period were not finally reviewed during the period. We know that a total of 2694 verdicts cleared direct review and thus were available for state post-conviction and federal habeas review during the study period. We also know that 257 of those verdicts were reversed on state post-conviction review, leaving 2437 verdicts (2694-257) to be accounted for. Of those, only 598 (25%) were finally reviewed on federal habeas. The remaining 1839 verdicts (2437-598), comprising fully 68% of the verdicts that were available for some type of post-direct appeal review during the study period (1839/2694), got stuck in the system either at the state post-conviction phase or the federal habeas phase. Given the steadily and substantially shrinking pools of cases clearing each of the three stages, it is reasonable to think that most of the unaccounted for verdicts were stuck in state post-conviction process, while a much smaller proportion of the original verdicts were stuck in federal habeas proceedings. It thus is reasonable (if probably conservative) to allocate two-thirds of the unaccounted for verdicts (.67 x 1839 = 1232) to the state post-conviction phase, and one-third (607) to the habeas stage. On that assumption:

  • The proportion of verdicts available for review on state post-conviction that were not reviewed is 46% (1232/2694).

  • The number of verdicts finally reviewed on state post-conviction is 1462 (2694-1232).

  • The state post-conviction reversal rate is 18% (257/1462)-which is 80% higher than we estimated in A Broken System.

  • And the overall reversal rate, combining all three review stages, is 71%, not the 68% we conservatively report. (100% - 41% reversed on direct appeal = 59% available for state post-conviction review. 59% - 18%(59%) reversed on state post-conviction = 48% available for federal habeas review. 48% - 40%(48%) reversed on federal habeas = 29% available for execution, and 71% reversed.)

See also infra note 315.

104. 408 U.S. 238 (1972).

105. See Roberts v. Louisiana, 428 U.S. 325 (1976); Roberts v. Louisiana, 431 U.S. 633 (1977).

106. See Woodson v. North Carolina, 428 U.S. 280 (1976).

107. See Lockett v. Ohio, 438 U.S. 586 (1976).

108. Those states are Alabama, California, Indiana, Kentucky, Mississippi, Montana, New Mexico, South Carolina, Tennessee, Virginia, Washington and Wyoming.

109. See State v. Voclain, 369 So.2d 190 (La. 1978); State v. Adams, 367 So.2d 8 (La. 1978); State v. McGraw, 366 So.2d 1278 (La. 1978); State v. Jones, 363 So.2d 714 (La. 1978); State v. Willie, 360 So.2d 813 (La. 1978); State v. Yates, 357 So.2d 541 (La. 1978); State v. Forrest, 356 So.2d 945 (La. 1978); State v. Hamilton, 356 So.2d 422 (La. 1978); State v. Brooks, 351 So.2d 1197 (La. 1977); State v. Smith, 351 So.2d 1191 (La. 1977); State v. Buggage, 351 So.2d 97 (La. 1977); State v. Sheppard, 350 So.2d 615 (La. 1977); State v. Preston, 349 So.2d 1252 (La. 1977); State v. Roberts, 350 So.2d 130 (La. 1977); State v. Gleason, 346 So.2d 206 (La. 1977); State v. Williams, 346 So.2d 204 (La. 1977); State v. Williams, 343 So.2d 1026 (La. 1977); State v. Perkins, 343 So.2d 730 (La. 1977); State v. MacHeard, 343 So.2d 730 (La. 1977); State v. Tyler, 342 So.2d 574 (La. 1977).

110. The prisoners whose death verdicts were overturned on successive petitions during the study period, but whom we counted as having had their death verdicts affirmed are: Levis Aldridge (Florida: 925 F.2d 1320), Stephen Booker (Florida: 922 F.2d 633), Frank Smith (Florida: 61 F.3d 815), Carl Songer (Florida: 769 F.2d 1488), and Willie Watson (Louisiana: unpublished).

We also counted at least eight death verdicts as having been affirmed at all three stages, though the defendants were later found to be innocent and released. Among those were some whose exoneration occurred as a result of a post-study-period successive federal habeas petition. See infra note 128 (listing the eight cases); infra pp. 25-27 (discussing the exoneration of Lloyd Schlup, whose death verdict is counted in our study as having been affirmed at all three review stages, but who later was exonerated in a successive federal habeas proceeding).

111. See supra p. 8 n.* for a discussion of the slightly different rounding procedures used in this example for ease of exposition, and in the National Report Card at p. 9 above.

112. See Valerie West, James S. Liebman & Jeffrey Fagan, Look Whose Extrapolating?: A Reply to Hoffmann, 76 Ind. L.J. 951 (2001).

113. See infra pp. 91-93 & Figure 10.

114. Between 1973 and 1995, federal courts finally reviewed about 600 death verdicts. As of 1995, however, hundreds of other verdicts imposed in the same years were still undergoing federal habeas review, and were not counted in our reversal rates. Figure 10, p. 93 below, shows (1) that flawed verdicts are over-represented in the part of each sentence-year cohort of cases that we did not count because they still were awaiting review as of the cut-off date, and (2) that federal habeas cases without reversible error are over-represented in the verdicts we counted. It thus is likely that the reversal rates we report for each sentence-year cohort of cases are lower than they eventually turn out to be, once all verdicts in the cohort are finally reviewed.

115. Although for the reasons discussed at pp. 17-18 above, we substantially underestimated state post-conviction reversal rates, we have no reason to believe that our cut-off date, and our decision not to extrapolate outcomes of decisions delayed beyond that cut-off date, either exacerbated or moderated this undercount.

116. See Hoffmann, supra note 67, at 957.

117. See, e.g., id.

118. See id. at 958.

119. See infra p. 36.

120. See, e.g., Williams v. Taylor, 529 U.S. 362 (2000).

121. See, e.g., Mitchell v. Kemp,762 F.2d 886 (11th Cir. 1985) (denying habeas relief to capital prisoner whose lawyer made only a minimal pretrial inquiry into whether the death penalty was appropriate in his case, on the ground that the incomplete investigation did not prejudice the defendant (discussed in Mitchell v. Kemp, 483 U.S. 1026 (1987) (Marshall, J., dissenting from denial of certiorari)); Glen v. Tate 71 F.3d 1204 (6th Cir. 1995) (despite conclusion that defense counsel were incompetent-they "never took the time to develop" evidence about the mental impairments of the 19-year-old brain damaged and "mentally deficient" defendant who acted under the influence of his older brother; "never spoke to any of [the defendant's] numerous brothers and sisters," "never examined his school records," and "never examined his medical records (including an emergency room record prepared after he collapsed in court one day) or records of mental health counseling they knew he had received"-court upholds capital conviction citing a lack of prejudice); Agan v. Singletary, 12 F.3d 1012 (11th Cir. 1994) (although defense counsel provided incompetent representation-pleading defendant guilty to a capital offense without any agreement on sentence, spending only seven hours total investigating the case, never contacting the attorney who had previously handled the case, never investigating the defendant's mental disabilities which were well documented, and never examining state investigation file indicating the investigator's belief that someone else committed the killing and containing correspondence from the victim documenting death threats from other suspects-court upholds the capital conviction, citing a lack of prejudice); Brimmer v. State, 1998 WL 612888 (Tenn. Cr. App. Sept. 15, 1998) (denying state post-conviction relief from capital conviction imposed at trial where the defendant was represented by a lawyer who admittedly was intoxicated and abusing drugs throughout the trial period on the ground that lawyer's incompetence did not prejudice the defendant at the guilt-innocence stage of trial); infra note 160 (discussing several similar cases in Texas).

122. See, e.g., Strickler v. Green, 524 U.S. 978 (1998) (despite the fact that police and prosecutors failed to disclose (1) that the central identification witness against the defendant had given conflicting statements to police, including that she could not identify the defendant, and (2) that the witness had steadily embroidered her story during a series of police interviews, conforming it to the police officers' theory of the case, Court upholds capital verdict, citing a lack of prejudice); Townes v. Murray 68 F.3d 840 (4th Cir. 1995) (citing waiver, court refuses to consider claim that the trial court improperly failed to determine if the mentally impaired defendant was competent to represent himself at the sentencing phase of his capital trial after his trial attorney withdrew, because his appellate lawyer's brief on the issue in the Virginia Supreme Court- which clearly challenged the trial court's failure to determine if the defendant could competently represent himself at the trial for his life-failed to specify the particular time during the trial when the trial court's inquiry should have been made); Smith v. Zant, 887 F.2d 1407 (11th Cir. 1989) (although the confession the police took from the mentally retarded defendant violated the Constitution because the defendant did not understand that he had a right to consult a lawyer before talking to the police, and although the confession was the prosecution's central evidence of capital murder and was inconsistent with defense evidence that the defendant acted in response to a physical provocation (in which case the offense was manslaughter, not murder, much less capital murder), the court upholds the capital conviction, ruling that the admission of the unconstitutional confession was harmless); Reams v. State, 909 S.W.2d 324 (Ark. 1995) (reviewing 18-year-old African American defendant's capital sentence for a murder committed by a co-defendant during a robbery; although the verdict was imposed by an all-white jury after the prosecutor exercised peremptory challenges to strike all the potential black jurors, and although faulty instructions may have prevented the jury from considering defendant's youth and other extenuating circumstances as bases for a sentence less than death, the court upholds the verdict, noting that "even in death penalty cases, a defendant must have raised the allegations of error at the trial court level by having made a specific, timely objection" and ruling that because of counsel's failure to object at trial, errors could not be corrected on appeal); Cannaday v. State 455 So.2d 713 (Miss. 1984) (although jailer's questioning of 16-year-old retarded defendant outside the presence of her attorney clearly violated the Constitution, and although the statement was extremely inflammatory as reported out of context to mean something different from what the young woman evidently meant (according to a third-party witness), court upholds capital conviction, citing lack of prejudice); Commonwealth v. LaCava, 666 A.2d 221 (Pa. 1995) (although the prosecutor repeatedly made irrelevant and highly prejudicial and inflammatory statements to the jury at both the guilt-innocence and sentencing phase of trial-including that because the defendant was known to sell drugs, he should be assumed to have sold them to children (there was no evidence that the defendant sold drugs to children, and the offense charged had nothing to do with drugs or drug dealing)-the court upholds the capital conviction, finding the errors harmless and ruling that, although the defendant's trial was flawed, it was fair enough).

123. See, e.g., Ken Armstrong & Maurice Possley, Break Rules, Be Promoted, Chi. Trib., Jan. 14, 1999, at N1 ("Between 1993 and 1997, there were 167 published opinions in which the Illinois Appellate Court or Illinois Supreme Court found that prosecutors committed some form of misconduct that could be considered harmless. In 122 of the cases-or nearly three out of four times-the reviewing court affirmed the conviction, holding that the misconduct was 'harmless'"; also quoting retired Illinois appellate judge Dom Rizzi, criticizing courts' and chief prosecutors' tendency to ignore "troublesome pattern[s]" of misconduct); Ken Armstrong & Steve Mills, Justice Derailed, Chi. Trib., Nov. 14, 1999, at N1; Spencer Hunt, Clouded Cases: Prosecutors' Conduct Risks Reversals, Cincinnati Enquirer, Sept. 10, 2000, at A1 (reporting, based on study of court records, that, although the Ohio Supreme Court "repeatedly has criticized [Cincinnati] prosecutors for making improper courtroom statements to win 14 death penalty cases over the past 12 years," and "has written at least four lengthy opinions since 1988 telling prosecutors to stop the misconduct, the record shows justices are more than willing to forgive these mistakes and uphold death sentences" on grounds of a lack of prejudice-as the court has done in 13 or 14 cited cases, with the 14th still under review; also quoting the Chief Justice of the Ohio Supreme Court, stating: "We better stop complaining about it if we're not going to do something about it."); Prosecutorial Restraint: Death Penalty Allows No Margin for Error, Columbus Dispatch, July 15, 2000, at 6A (citing a number of recent capital cases in which the Ohio Supreme Court had "express[ed] frustration" and "'mounting alarm'" over the increasing incidence of misconduct by prosecutors" and trial judges; in nearly all cases, however, the error was found harmless and the death verdict was upheld).

124. We also know (1) that over four-fifths of the death verdicts approved on federal habeas raised claims of ineffective assistance of counsel or prosecutorial suppression of evidence, and (2) that one of the two main reasons these claims are denied is a lack of "prejudice"-i.e., that the court would not reverse due to malfeasance or nonfeasance by defense counsel or the prosecutor because the defendant could not prove that the misconduct probably changed the outcome of the trial. Regrettably, however, collecting information on whether claims were denied on "no prejudice" grounds was beyond our capacity, given finite time and resources, and we thus cannot say how many verdicts were approved on "no prejudice" grounds, in addition to those approved on "harmlessness" grounds.

125. See, e.g., Gray v. Netherland, 518 U.S. 152 (1996); Sawyer v. Whitley, 501 U.S. 333 (1992); Coleman v. Thompson, 501 U.S. 722 (1991); Smith v. Murray, 477 U.S. 527 (1986); Dugger v. Adams, 489 U.S. 214 (1989); Steve Mills & Ken Armstrong, Lawyers Err, and Appeals Door Shuts, Chi. Trib., July 10, 2001, at N1 ("After Jeffrey Rissley landed on Illinois' Death Row, he seemed to have solid grounds for an appeal: His trial lawyer, a probate and real estate attorney, had advised him to plead guilty without bargaining for any leniency-a move practically unheard of in capital cases. But the attorney advising Rissley about his appeal misinterpreted a new state law, leading Rissley to file the appeal six days after a crucial deadline. That tardy filing prompted the Illinois Supreme Court to rule that Rissley . . . had forfeited the right to raise his trial attorney's incompetence-virtually his only issue-in further appeals."); infra note 216 (discussing Giarratano case).

126.

Number and Percent of Cases Where Default or Harmless Error
 
Reason for Denial Number %
 
Waiver 113 32
Harmless Error 63 18
 
One or Both of Above 156* 44
 
Total Denials of Relief 358

  • The above two figures do not sum to 156 because procedural default and harmless error were sometimes found in the same case. Source: HCDB.
  • 127. On the 95th exoneration, see Man on Death Row for Five Years Found Innocent, Associated Press, May 26, 2001:

      A north Alabama man who spent five years on prison's death row has been found innocent at retrial. Gary Wayne Drinkard, 455, of Falkville was released Friday after a Morgan County jury found him innocent of capital murder charges in the robbery-slaying of Decatur businessman Dalton Pace.

      Drinkard's first conviction had been overturned by the Alabama Supreme Court because prosecutors were allowed to talk about Drinkard's prior criminal history. . . .

      On the 96th exoneration, see Pat Leisner, Jurors Acquit Man in Slaying Retrial, AP Online, June 6, 2001:

      A 30-year-old man who had faced the death penalty was acquitted [in Florida] of two counts of first-degree murder Wednesday in a retrial.

      Joaquin Martinez, a Spanish citizen living in the United States, fought back

      tears at the jury's verdict. His mother, Sara Martinez, wept openly.

      *****

      Martinez spent nearly three years on death row until the Florida Supreme Court overturned his conviction and death sentence and ordered a new trial.

      There was no physical evidence linking Martinez to the slayings in a case that attracted the attention of the Pope and the King of Spain . . . .

    On the 97th exoneration, see Associated Press, Sheets Released from Death Row, Lincoln (Nebraska) J. Star, June 13, 2001 ("Former death-row inmate Jeremy Sheets . . . was released from prison Tuesday, three years after he was ordered to die for the kidnaping, rape and killing of an Omaha girl. . . . His release came after the Nebraska Supreme Court threw out a taped statement used to convict Sheets . . . . Without the taped statement, prosecutors said, they did not have enough other evidence to proceed with another trial.").

    On the 98th exoneration, see Raymond Bonner, Death Row Inmate is Freed After DNA Test Clears Him, N.Y. Times, Aug. 24, 2001:

      Charles Fain has been on death row for almost 18 years for the rape and murder of a 9-year-old girl [Daralyn Johnson] who was snatched off the street in Nampa [Idaho], a small town west of here.

      But this afternoon, Mr. Fain, 11 days shy of his 53rd birthday, walked out of the maximum security prison here into the blazing sun, a free man. Two hours earlier, a state judge ordered the charges against him dismissed on the basis of DNA tests indicating that hairs found on the girl's body, which had been used to convict Mr. Fain, were not his.

      ******

      "Justice requires the action we have taken today," David L. Young, the Canyon County prosecutor, said today at a news conference . . . . "It also requires that we do everything we can to solve this case."

      Mr. Young added, "The killer has not yet been apprehended."

      Today the Johnson family seemed to accept Mr. Fain's release.

      "We would like to say we are in complete support of the judicial system and all those involved in the reinvestigation of this case," the family said in a statement. "We are confident that we will have closure and that all those involved will be brought to justice."

    See also Henry Weinstein, DNA Tests Give Idahoan Freedom After 17 Years on Death Row, L.A. Times, Aug. 24, 2001(quoting Fain's lawyer saying, "This case demonstrates that those who presuppose that only the guilty get the death penalty are wrong.").

    On the 99th exoneration, see Phil Long & Amy Driscoll, Inmate on Death Row Goes Free After 17 Years, Miami Herald, Jan. 4, 2002:

      A chance discovery two years ago of an old legal transcript in a lawyer's files led to freedom Thursday for Juan Melendez-17 years after he was sent to Florida's Death Row for a murder another man claimed to have committed.

      Polk County prosecutors effectively set Melendez free when they announced Thursday that they do not have sufficient evidence to re-try him for the 1983 slaying of a Central Florida beauty school owner.

      *****

      Melendez, 50, was sentenced to death in 1984 for the murder of Delbert "Mr. Del" Baker in his Auburndale beauty salon. Melendez lost several rounds of appeals and his death sentence was upheld. He was nearing the end of his appeals when his former defense lawyer, Roger Alcott, discovered a key transcript as he moved old boxes of files following his appointment as a Polk County circuit judge in early 2000.

      The transcript details a conversation taped about a month before Melendez's trial. On the tape, Vernon James, a now-deceased witness in the case, admitted being involved in the murder and said that Melendez was not at the scene. . . .

      Melendez's appellate lawyers . . . [found] that James, once a suspect in the murder, had told up to 20 other people-including a former law enforcement officer-of his involvement in the murder. Some said he had confessed to the killing.

      Armed with new evidence, the lawyers returned to state court to appeal Melendez's conviction. . . . Circuit Judge Barbara Fleischer . . . found that the trial prosecutor, Hardy Pickard, had failed to disclosed potentially damaging information to the defense, including serious inconsistencies in statements by John Berrien, one of two major state witnesses. [Berrien subsequently recanted his testimony.]

      Additionally, the judge said Pickard misled the jury about testimony from the other main state witness, David Luna Falcon, by telling jurors that Falcon had "nothing to gain" from testifying. But Falcon, who testified at trial that Melendez had confessed, had struck a deal with prosecutors to reduce his own prison time in exchange for testimony.

      ******

      The new evidence, along with Pickard's withholding of information, "combine to undermine the confidence in the outcome of the defendant's original trial," [Judge Fleischer] concluded.

    128. Among the factually or legally innocent inmates whose capital verdicts were affirmed at all three stages, who thus were approved for execution and in whose cases A Broken System assumes there were no errors at all, are: Herbert Bassette (VA), Paris Carriger (AZ), Ronald Monroe (LA), Donald Paradis (ID), Anthony Porter (IL), Lloyd Schlup (MO), Frank Lee Smith (FL), Joseph Spaziano (FL), and Earl Washington (VA). Among the factually or legally innocent inmates whose capital verdicts were affirmed at two stages, and in whose cases A Broken System assumes there were no errors at both those stages, are: Blazak (AZ), Bowen (OK), Brown (FL), Clemmons (MO), Creamer (GA), Drake (GA), Fain (ID), Guerra (TX), Harris (WA), Jacobs (FL), Jent (FL), Knapp (AZ), Kyles (LA), Macias (TX), Melendez (FL), Miller (FL), Wallace (GA), and Williamson (OK). Among the factually or legally innocent inmates whose capital verdicts were affirmed at one stage, and in whose cases A Broken System assumes there were no guilt-innocence errors at that stage, are: Adams (TX), Bloodsworth (MD), Burrell (LA), Burrows (IL), Brandley (TX), Cobb (IL), Cook (TX), Cruz (AZ), Cruz (IL), Deeb (TX), Dexter (MO), Farmer (CA), Gauger (IL), Graham (LA), Hernandez (IL), Jimenez (NV), Jimerson (IL), Jones (CA), Jones (IL), Lucas (TX), McManus (TX), Miller (OK), Miranda (NV), Mitchell (TX), Munson (OK), Nelson (GA), McMillian (AL), Nieves (PA), Peek (FL), Robison (AZ), Smith (IL), Smith (LA), Smith (IN), Dennis Williams (IL), and Darby Williams (IL).

    129. State v. Schlup, 724 S.W.2d 236, 241-42 (Mo. 1987).

    130. Schlup v. State, 758 S.W.2d 715, 717 (Mo. 1988).

    131. Schlup v. Armontrout, 941 F.2d 631, 639 (8th Cir. 1991).

    132. Schlup used a procedure allowing a small class of prisoners with newly discovered evidence to file a "successive" petition raising claims of error that courts at the three regularly available review stages had previously denied. (Because successive petitions are not available as of right, our study did not count reversals occurring through this procedure. See supra pp. 19.) In 1996, Congress forbade all such successive petitions. See 28 U.S.C. § 2244(b).

    It is in the nature of valid innocence claims like Schlup's that they can be proved only through successive litigation in which testimony at each prior stage is revealed to be false by newly discovered evidence that became relevant for the first time when the false testimony was given. Since 1996, therefore, it has been in the nature of valid innocence claims like Schlup's that existing court review mechanisms cannot be relied upon to reveal innocence.

    133. Brooke A. Masters, Missteps On Road To Injustice; In Va., Innocent Man Was Nearly Executed, Wash. Post, Dec. 1, 2000, at A1.

    134. Brooke A. Masters, DNA Clears Inmate in 1982 Slaying, Wash. Post, Oct. 3, 2000, at A1.

    135. Washington v. Commonwealth, 323 S.E.2d 577, 585-86 (Va. 1984).

    136. See Masters, Missteps on Road to Injustice, supra note 133.

    137. Washington v. Murray, 952 F.2d 1472, 1477-78 (4th Cir. 1991).

    138. Washington v. Murray, 4 F.3d 1285, 1290 (4th Cir. 1993).

    139. State v. Porter, 489 N.E.2d 1329, 1335-36 (Ill. 1986).

    140. State v. Porter, 647 N.E.2d 972, 975-76 (Ill. 1995).

    141. Id. at 974-75.

    142. Id.

    143. United States ex rel. Porter v. Warden, 1996 WL 167340 (N.D. Ill. Apr 4, 1996) (quoting Strickland v. Washington, 466 U.S. 668, 691, 694 (1984)).

    144. Porter v. Gramley, 112 F.3d 1308, 1313 (7th Cir. 1997).

    145. Tom Ragan, Years After Death Row Travesty, Killer Gets Due, Chi. Trib. Sept. 8, 1999, at M1.

    146. See Steve Mills, Simon Also Suspected in Milwaukee Slaying, Chi. Trib., Mar. 10, 1999, at 1.

    147. Steve Mills, Porter Case Had Wrongs at Each Turn, Chi. Trib., Feb. 12, 1999, at 1.

    148. Shannon O'Boyle, Paula McMahon & Ardy Friedberg, Death Row Prisoner Dies; Now DNA Test Clears Him, South Florida Sun-Sentinel (Ft. Lauderdale Fla.), Dec. 15, 2000, at 1A.

    149. Smith v. State, 515 So.2d 182, 183-84 (Fla. 1987).

    150. See, e.g., Dave Maass, Some Counties Simply Can't Afford To Pursue Capital Offense Cases, Arizona Capitol Times, May 25, 2001, at A1:

      A May 4 decision by the Arizona Supreme Court to send a death row inmate back for re-sentencing has left some county officials wondering whether they can afford to pursue the death penalty in capital offense cases. . . .The court held in its opinion, written by Chief Justice Thomas Zlaket, that "so long as the law permits capital sentencing, Arizona's justice system must provide the adequate resources to enable indigents to defend themselves in a reasonable way."The problem is many smaller counties don't have the resources to pay for a pre-sentencing investigation. It's a choice between the death penalty and fiscal death.

    See also Russell Gold, Regional Report: Counties Struggle With High Cost of Prosecuting: Death-Penalty Cases; Result Is Often Higher Taxes, Less Spending on Services; 'Like Lightning Striking', Wall St. J., Jan. 9, 2002 ("Just prosecuting a capital crime can cost an average of $200,000 to $300,000, according to a conservative estimate by the Texas Office of Court Administration. Add indigent-defense lawyers, an almost-automatic appeal and a trial transcript, and death-penalty cases can easily cost many times that amount.... The cost, county officials say, can be an unexpected and severe budgetary shock-much like a natural disaster, but without any federal relief to ease the strain. To pay up, counties must raise taxes, cut services, or both.").

    151. See also Rhonda Cook, System Prolongs Death Row Wait, Atlanta Const., Oct. 27, 2001 ("The average time a Georgia death row inmate waits for his or her sentence to be carried out is 11 years, but 64 of the 128 currently awaiting execution have been under death sentences even longer. Waiting the longest-28 years-has been Carl Isaacs, and his execution date remains undetermined.").

    152. The discussion in this section is partially reprinted, and expanded, in James S. Liebman, An "Effective Death Penalty?": AEDPA and Error Detection in Capital Cases, forthcoming in the Brooklyn Law Review, 2002.

    153. See, e.g., sources collected in James S. Liebman, The Overproduction of Death, 100 Colum. L. Rev. 2030, 2112-14 n.197 (2000).

    154. See id.

    155. See, e.g., Candidates for Lieutenant Governor Hold Opposing Views on the Death Penalty, Roanoke Times, Oct. 30, 2001 ("Support for the death penalty has long been considered almost a prerequisite for becoming a major party candidate for statewide office in Virginia. 'If you ran for statewide office, it was assumed you had to be for the death penalty,' said Robert Holsworth, director of the Center for Public Policy at Virginia Commonwealth University. 'Being opposed to it was viewed almost as a disqualifier for office.'").

    156. See supra pp. 5 & n.77, 24 & n.127.

    157. For a more extensive discussion of this issue, with numerous examples and citation of a variety of sources, see 100 Colum. L. Rev., supra note 153, at 2111-14 & nn.194-99. See also infra notes 161, 209 (discussing voters' removal of judges from California Supreme Court because of their votes to reverse capital verdicts).

    158. See A Broken System, Part I, supra note 101, at 9.

    159. See infra pp. 333-34.

    160. On the tendency of the Illinois appellate system to allow serious error in capital cases to go unremedied on harmless error and waiver grounds, see the investigative reports cited supra notes 123, 125. On the tendency of Ohio appellate judges to deem disturbing capital error to be "harmless," see the articles cited supra note 123. On the same tendency in Texas capital cases, see, e.g., Janet E. Elliott, Prisoner Loses Attorney Ruling; Court Says 'Competent Counsel' Right Does Not Apply in Appeals, Houston Chron., Jan. 2, 2002 ("Death row inmates don't have a right to capable counsel in their appeals,, the state Court of Criminal Appeals ruled Wednesday" in a 6-3 decision; as a result, appellate lawyer's incompetent failure to challenge errors on these appeals will bar all subsequent consideration of the error); James Kimberly, Guilty . . . Or Merely Proven Guilty?: Once on Death Row, it Might Not Matter, Houston Chron., Feb. 6, 2001, at 1:

      Colella . . . is by no means the only death row defendant to have his claims of innocence ignored by the [Texas] Court of Criminal Appeals. As former Presiding Judge Michael McCormick pointed out, the court is not particularly concerned with claims of innocence.

      "Actual innocence claims have to go through the clemency process. That's what it is there for," said McCormick . . . .

      The chairman of the Texas Board of Pardons and Paroles, however, says he leaves innocence questions to the courts.

      *****

      The Court of Criminal Appeals has never been particularly kind to defendants. . . . But never has the court been as reluctant to reverse a conviction or a death sentence as it is today.

      *****

      Three times the Court of Criminal Appeals has upheld convictions in cases in which there were allegations the defense attorney slept during portions of the trial. The court also once ruled against a capital defendant who presented evidence that her attorney was a close friend of the key prosecution witness. Another time, the court ruled against a death row inmate whose attorney was accused of drinking and snorting cocaine outside the courtroom during the capital murder trial.

      The Texas Defender Service . . . reviewed 103 of the 187 habeas corpus appeals filed with the Court of Criminal Appeals over five years.

      The service said almost half the petitions-typically a defendant's final appeal in state court-were less than 30 pages, indicating the court-appointed attorney either did not know how to properly write a habeas writ or made little attempt to investigate the case.

      Eighteen of the habeas petitions for death penalty defendants were less than 15 pages long. Two were only three pages.

      The Court of Criminal Appeals never raised questions about the competency of the attorney in any of the cases. Instead, the court just ruled against the defendants, Marcus said.

    See also Ken Armstrong & Steve Mills, Gatekeeper Court Keeps Gates Shut: Texas Judges Prove Reluctant to Nullify Cases, Chi. Trib., June 12, 2000, at N1:

      By the time the Court of Criminal Appeals of Texas ruled on Cesar Fierro's request for a new trial in 1996, it was clear his confession had been obtained under disturbing circumstances.

      Nearly everyone involved with the case had come to agree that Fierro confessed to the murder of an El Paso taxi driver only after police across the border in Juarez, Mexico, raided his parents' house, held them captive and threatened to torture his stepfather with electrical current from a generator attached to his genitals.

      The prosecutor who put Fierro on Death Row now believes that. So does the local district court judge who reviewed how the confession was obtained. He recommended that Fierro get a new trial.

      Even the Court of Criminal Appeals, Texas' highest criminal court, agreed that Fierro's rights were violated and police committed perjury.

      But the court . . . stopped there. It refused to grant Fierro a new trial.

      In Texas, the Court of Criminal Appeals is the state's court of last resort, a gatekeeper that is supposed to remedy injustice, correct fundamental errors that occur at trial, and ensure that convicted defendants receive a fair hearing on appeal.

      But a Tribune investigation found that it has not always done that. To handle Death Row appeals, the court has appointed attorneys with previous disciplinary records or little experience. In its rulings, the court has frequently proved tolerant of flawed convictions and reluctant to acknowledge holes in the prosecution's case.

      In Fierro's case, the majority on the nine-member court ruled that the violation of his rights was "harmless." The court kept Fierro on track for execution, and he has now been on Death Row for two decades.

      *****

      The Court of Criminal Appeals also has refused relief to Death Row inmates represented by an attorney who slept at trial. It refused relief to a defendant who, a psychologist testified, was more likely to commit future acts of violence because he is Hispanic-testimony that even the state attorney general's office found objectionable. The U.S. Supreme Court last week ordered a new sentencing hearing for the defendant, and the Texas attorney general has said six other inmates may receive new sentencing hearings on the same grounds.

      The court has even refused relief to a convicted rapist, Roy Criner, even though DNA testing conducted after trial showed the semen found in the victim wasn't his. . . .

      The Tribune's investigation of the death penalty in Texas found deep-seated problems that call into question the system's integrity.

    161. Rather than picking and choosing among judges based on our subjective judgments about how trustworthy they are when it comes to identifying serious error, we counted all reversals by all judges during the study period. We did this for three reasons. First, to resort to our own conclusions that Judge A's rulings count, but Judge B's don't-notwithstanding that both were elected or appointed to review death verdicts and make the final rulings A Broken System counted-would be inconsistent with acceptable social scientific practice and susceptible to bias and inaccuracy. Second, we counted only final reversals, meaning ones subject to scrutiny and validation by at least two and often three or four levels of courts, with the ruling at each level besides the lowest one being made by panels of at least three and often as many as seven or nine (or even more) judges. When added to the screening of judges that takes place through elections and appointments, the multi-tiered review process at each stage and the multi-member review panels at all but the first levels of review provide strong assurances that findings of reversible error in capital verdicts are not the result of a single judge's idiosyncracies but the mutually verified conclusions of multiple judges, the vast majority of whom in the aggregate are, by political and professional disposition, reluctant to reverse capital verdicts. Third, we used this method because it is conservative: It moderates reversal rates. Given the political and professional realities already discussed, it is probably the case that for every Rose Bird, the California Supreme Court Chief Justice who some claim was too quick to overturn death verdicts-but then was voted out of office for that reason, see infra note 209-there are several judges like Malcolm Lucas, who replaced Rose Bird and, without fear of political discipline, voted to affirm virtually every capital verdict he reviewed. If we chose to eliminate decisions in which any particular "outlying" judge concurred (assuming there were objective criteria for identifying such judges), the result would be the elimination of many more affirmances joined by Chief Justice Lucas and similar judges (see infra note 209) than reversals joined by Chief Justice Bird and similar jurists-with the effect that reversal rates would be even higher than A Broken System finds.

    162. See case studies cited supra notes 121-23, 125, 160.

    163. See supra p. 24 & n.128.

    164. About 15% of the death row exonerations (releases from death row due to findings of factual or legal innocence) occurred as a result of direct appeal findings that there was not enough evidence to allow a reasonable person to conclude that the defendant was guilty. See Innocence and the Death Penalty, supra note 77. Exonerations following state post-conviction and federal habeas review almost never occurred for this most glaring of reasons, and instead were generally prompted by the kind of violations that are the most hidden from view: prosecutorial suppression of evidence of innocence and other exculpatory evidence, and an incompetent lawyer's failure to find similar evidence. Our federal habeas data reveal no reversals because the evidence of guilt was insufficient to convince a reasonable jury, and only two reversals based on a conclusion that there was insufficient evidence to justify a death sentence.

    165. See, e.g., the debates in both Houses of Congress that preceded the adoption of the Antiterrorism and Effective Death Penalty Act of 1996. Statements by many supporters of the Act juxtaposed favorable views of the properly discerning capacities of state direct appeal judges with claims that federal habeas judges tend to have excessively low standards for reversible error. See, e.g., 142 Cong. Rec. S3362 (Apr. 16, 1996) (Sen. Hatch) ("I have to say, one of the biggest problems [is] loony judges in the federal courts who will basically grant a habeas corpus petition for any reason at all."). Pennsylvania Senator Arlen Specter, for example, extended similar criticisms to, and proposed similar limits on, the state post-conviction stage. For an overview, see Larry W. Yackle, A Primer on the New Habeas Corpus Statute, 44 Buff. L. Rev. 381 (1996). These views explain the various rules adopted by the U.S. Supreme Court during the late 1980s and early 1990s and by Congress in 1996 that curbed federal habeas review, all of which proceeded on the assumption that direct appeal courts catch the most serious errors at the first stage of review, and that life-tenured federal habeas judges at the last stage of review are more likely than elected state judges to reverse based on non-serious error. Similar rules adopted in Texas and elsewhere imposed similar constraints on state post-conviction review based on like comparisons of the appropriately tough-minded decisions of state direct appeal judges and laxer review thought to occur at the state post-conviction stage. See supra p. 27 & n.132; infra pp. 63-64 & nn.202-06 (discussing recently expanded constraints on state post-conviction and federal habeas review).

    166. See infra pp. 40-41 (further discussion of the seriousness of error found at the direct appeal stage).

    167. Of the 351 state post-conviction reversals, the reason for reversal in known in 299 cases. Of those 299:

    • 116 (39%) were reversed because of egregiously incompetent defense lawyers;

    • 57 (19%) were reversed because of police or prosecutorial suppression of evidence of innocence or that a death sentence was not appropriate (47 cases) or because of other prosecutorial misconduct (10 cases);

    • 58 (19%) were reversed due to misleading instructions to jurors on how they were required to decide the case; and

    • 13 (4%) were due to judge or jury bias.

    • 239 (80%) were reversed due to one of these four serious flaws.

    See Appendix C for all the relevant information and summary tables. The first four figures above do not sum to the last figure (239) because five verdicts were reversed due to both incompetent lawyering and prosecutorial suppression of evidence. For the reasons discussed supra notes 87, 90, these figures vary slightly from those reported in A Broken System, Part I.

    168. See the Wallace case discussed in Appendix D, p. D-6.

    169. Reasons for federal habeas reversals are known for 220 cases, as set out below. Because there often were multiple bases for reversal, we counted each basis as its proportion of the total number of bases for relief. Where there was one basis for reversal, that claim is counted as 1; where there were two bases for reversal, each basis is counted as .5; each is counted as.33 (one-third)) where there are three bases for relief; and so forth.

    Reasons for Reversals on Federal Habeas
    Reason for Relief No. of Verdicts Overturned for this Reason % of Reversals Where Reason Is Known
    Incompetent defense lawyer 50.17 23%
    Lawyer denied altogether .25 .1%
      (All denials of compentent lawyer
    (50.42) (23%)
    Prosecutorial suppression of evidence 12.33 6%
    Prosecutors' intentional exclusion of black jurors 3.33 1.5%
    Other police, prosecutor misconduct 11.28 5%
      (All police, prosecutor misconduct)
    (26.94) (12%)
    Invalid instructions on burden of proof 38.00 17%
    Invalid instructions on aggravating circumstances 13.80 6%
    Invalid instructions on mitigating circumstances 18.03 8%
    Other invalid capital sentencing instructions 2.20 1%
      (All invalid instructions)
    (72.03) (33%)
    Juror bias caused by third parties during trial 1.50 .7%
    Juror bias due to pretrial influences 5.00 2%
    Illegal exclusion of blacks from jury pool 8.50 4%
      (All biased decision makers)
    (15.00) (7%)
    Exclusion of exculpatory evidence 2.00 .9%
    Defendants denied funds for adequate defense 2.08 .9%
    Defendant was mentally incompetent to be tried 5.33 2%
    Involuntary confession 5.00 2%
    TOTAL 178.80 81%

    Source: HCDB.

    170. A small number of reversals (2% of federal habeas reversals; 0 state post-conviction reversals) occurred because capital defendants asked for, but were not allowed to see, lawyers during in-custody interrogations. As the United States Supreme recently held, such reversals are not technicalities. Rather, as the Court noted, the purpose of requiring defendants to be informed of their right to an attorney and to be provided with a lawyer (if they ask for one) before additional interrogation takes place is to assure the reliability of statements made to police. See Dickerson v. United States, 530 U.S. 428 (2000). For an example of the miscarriages that can result when suspects make unreliable statements without understanding their rights and without having a lawyer present, see the case study of Earl Washington's case, supra pp. 27-30. See also the recent Chicago Tribune study of false confessions, supra note 35.

    171. The reasons for reversal counted here include incompetent defense lawyering, prosecutorial suppression of exculpatory evidence, rulings excluding such evidence and instructions to ignore such evidence. Because our state post-conviction data do not distinguish among improper instructions on these and other bases, we conservatively assumed that only 40% of the misinstructions were on this ground. We also conservatively omitted reversals for other forms of prosecutorial misconduct and improper instructions on burden of proof, even though some of these violations involve prosecutorial mishandling of exculpatory evidence, prosecutorial statements calling on jurors to ignore such evidence and judicial instructions keeping jurors from giving weight to such evidence. Even when counted in this conservative manner, error fitting in this category accounts for 60% of all state post-conviction reversals and 50% of all federal habeas reversals.

    172. See infra pp. 319-24.

    173. See supra pp. 17-18; See A Broken System, Part I, supra note 101, at 19-20, 26-27 & nn.39, 132; App. C, pp. C-1 to C-2.

    174. For full information and summary tables, see infra Appendix C. New information obtained since the publication of A Broken System decreased the proportion of retrial verdicts imposing sentences less than death from 75% to 73%, but increased the proportion of acquittals from 7% to 9%, leaving the overall proportion of outcomes less than death the same, 82%. See supra notes 87, 90.

    175. See Fed. R. Evid. 612 (applicable in about 40 states, as are comparable rules elsewhere).

    176. See, e.g., Fed. R. Evid. 804(b)(1), 803(5) (applicable in about 40 states, as are comparable rules elsewhere); see also Fed. R. Evid. 801(d)(1). In point of fact, the need to use prior testimony instead of current testimony is infrequent because most witnesses in capital cases-law enforcement officers, co-offenders, and surviving victims (usually of robberies of commercial establishments)-were working age or younger when the first trial occurred.

    177. See supra pp. 40-41.

    178. See supra pp. 22-23, 26, 30-32, 41.

    179. See infra pp. 319-24.

    180. See sources and examples cited in 100 Colum. L. Rev., supra note 153, at 2078-84 & nn.142, 145; id. at 2144-45 & n.271.

    181. Insufficient evidence almost always is challenged on direct appeal, and state legal rules almost always bar claims raised on direct appeal from being raised again in state post-conviction proceedings. See also supra note 164.

    182. See supra pp. 40-41.

    183. For additional discussion of this study for the Nashville Tennessean, see supra pp. 17-18 & n.102.

    184. See Shiffman, supra note 102.

    185. For examples of how overall reversal rates are calculated, see supra pp. 8 n.*, 19; supra note 103.

    186. The six states that had death verdicts finally reviewed at only the direct appeal stage are Colorado, Connecticut, New Jersey, New Mexico, Ohio and Oregon.

    187. The two states for which we only have data on outcomes at the direct appeal and federal habeas stages review are Delaware and Washington.

    188. See Figure 1A, p. 50 below, for a list of these 26 states.

    189. Conversely, the more random the pattern of bars is, the less evidence the chart provides of a relationship between reversal rates and the factor being measured. For discussion in addition to that in the text just below about why our subsequent multiple regression analyses are a more comprehensive and reliable method of testing relationships between two variables-such as reversal rates and homicide rates-than are the bivariate comparisons permitted by graphs such as those discussed here in text, see infra pp. 82-83.

    190. See infra pp. 89-91, 99, 140-42 & Table 4, 154-56, 177, 194-95, 258-59, 372-75.

    191. For a discussion of how combined, or overall, reversal rates are calculated, see supra supra pp. 8 n.*, 19; supra note 103.

    192. Judges reviewing death verdicts in later years are different from judges reviewing verdicts in earlier years, and the same judges face political and other pressures in some years (say, election years) that are different from pressures they face in other years. Analyzing differences in reversal rates based on the year of the final decision directly assesses the effect of having different decision makers in different years and of different pressures on the same decision makers.

    This same measure of change may also be indirectly affected by changes in the quality of capital verdicts over time: Death verdicts reviewed later tend to be ones that were imposed later, so analyses of differences in reversal rates for verdicts reviewed in later, versus earlier, years may detect differences in the quality of verdicts imposed in later, versus earlier, years. As a result, analyses of changes in reversal rates across all study years in which death verdicts were reviewed will be partly sensitive to changes over time in how reliably trial actors reached a verdict. Nonetheless, because the effect of sentence-year conditions on review-year outcomes is indirect, sentence-year conditions probably have less of an effect on differences in reversal rates measured by review year than do the conditions described in the first paragraph of this footnote.

    193. Trials held in later years may be funded differently from, or staffed by better trained or less well-trained professionals than, trials held earlier. Analyzing differences in reversal rates based on the year death verdicts were imposed directly assesses the effect of these trial-specific conditions on reversal rates. This same measure of change may also be indirectly (and, so, we think, less substantially) affected by changes in the attitudes and behaviors of reviewing judges: Verdicts imposed in a given year can only be reviewed by judges on the bench in that or (more probably) a later year, not by the different set of judges who were on the bench earlier. Consequently, analyses of changes in reversal rates based on the year death verdicts were imposed will be indirectly sensitive to changes in how carefully judges scrutinized capital verdicts for error in later, as opposed to earlier, periods.

    194. Of the 2349 reversals occurring during the study period, 1852 (79%) occurred at the direct appeal stage, and 240 (10%) occurred at the federal habeas stage. The remaining 257 (11%) occurred at the state post-conviction review stage. For discussion of our regression findings on changes in direct appeal reversal rates over time after controlling for other factors, see infra pp. 194-95, 258-59.

    195. See supra pp. 11-18.

    196. See supra pp. 17-18.

    197. See decisions cited supra notes 105-09 and accompanying text.

    198. See supra p. 43 & n.173.

    199. Information on California, Florida, Illinois, Nevada, and Washington are collected in A Broken System, Part I, supra note 101, at 146-47 n.241.

    On Alabama, see Letter from James S. Liebman to Senator Patrick Leahy, Chair, Senate Judiciary Committee, July 2, 2001 (analyzing capital reversal data supplied to the Judiciary Committee by the Alabama Attorney General, which reveals a 74% reversal rate from 1977 to 2000, compared to the 77% reversal rate we discovered for the 1977-1995 period).

    On Kansas, see infra p. 400 & n.894.

    On Kentucky, see John Cheves, Judges Are Overturning Death Sentences, Citing Mistakes at Trial, Lexington Herald-Leader, Oct. 31, 2001 ("[S]uddenly . . . judges are overturning death sentences, citing mistakes at trial, twice as often as juries recommend them. Six men were taken off Kentucky's Death Row in the past year, while just three more men arrived. . . . . 'I don't ever remember a series of reversals coming down like this before,' said Joe Bouvier, a Fayette County prosecutor who studies capital punishment issues for the Kentucky Commonwealth's Attorneys Association.").

    Information about Louisiana and North Carolina comes from conversations with capital punishment lawyers in those states. But cf. Ames Alexander & Liz Chandler, Errors, Inequities Often Cloud Capital Cases in the Carolinas, Charlotte Observer, Sept. 10, 2000, at 1A (from 1977 to 2000, "more than half of all death sentences imposed [whether or not they were finally reviewed] have been thrown out because of flawed trials").

    On Ohio, see sources cited supra note 123 (noting reluctance of Ohio Supreme Court to reverse capital verdicts despite errors identified by that court).

    On Pennsylvania, see Debbie Garlicki, Death Row Defenders Rroil Legal Waters, Allentown Morning Call, Jan. 13, 2002 ("Over the last six years, judges have . . . ordered new trials in at least 14 [Pennsylvania capital] cases and new sentencing hearings in at least 23 cases involving the defenders."); Debbie Garlicki, Yet Another Death Penalty Is Set Aside, [Allentown] Morning Call, Aug. 28, 2001;.

    On Tennessee, see Shiffman, supra note 102 (study showing state direct appeal reversal rate of about 50% for the 1980-2001 period, compared to the 29% reversal rate we discovered for our study period).

    On Virginia, see, e.g., Alan Cooper, Death Sentence Ruled Subject to New Hearing, Richmond Times-Dispatch, Sept. 28, 2001; Frank Green, Death Penalty Cases Scrutinized More: Hearings Are Being Ordered in Virginia, Richmond Times-Dispatch, Apr. 9, 2001 ("Bumps and potholes are slowing the pace of capital cases on Virginia's once smooth road from death sentence to execution."); Brook A. Masters, Legal Scrutiny Slows Pace of Executions in Virginia, Wash. Post, Dec. 26, 2000 ("The number of executions in Virginia fell by nearly half this year, as state and federal appeals courts intervened more often in the commonwealth's fastest-in-the-nation system of capital punishment."); Death Sentence Reversed in Va., Wash. Post, June 8, 2001 ("Kevin Green . . . is the second death row inmate in seven weeks-and the sixth since 1999-to win a reversal from the Virginia Supreme Court"; case is thought by some "legal analysts" to "reflect the Virginia Supreme Court's growing concern" with the reliability of Virginia capital verdicts); With Plea, Virginia Convict Moves Off Death Row, Wash. Post, Apr. 29, 2000 ("For the first time in a decade, an inmate learned yesterday that he would walk away from Virginia's death row alive without the governor's help.").

    On Washington, see also Lise Olsen, The Death Penalty: Uncertain Justice, Seattle Post-Intelligencer, Aug. 6, 2001 ("30 men have been sent to death row since Washington's death-penalty law was reinstated in 1981-but 15 have had their death sentences overturned. Only three have been executed, and 2 of them 'volunteered.' The leading causes of the reversals are ineffective assistance of counsel and judicial error-both of which figured in 6 rulings. . . . Statewide, 22% of the condemned have gotten off death row due to ineffective assistance of counsel."); Olsen, supra note 64 (discussing reversal of James Leroy Brett's death sentence because of incompetent lawyering, noting that "[i]t was one of three death sentences overturned so far this year by the state Supreme Court-the most in a single year since capital punishment was reinstated in 1981.").

    200. See supra pp. 22-47.

    201. See supra pp. 1-5.

    202. See, e.g., Alan Berlow, The Wrong Man, The Atlantic Monthly, Nov. 1999, at 84 ("In 1995 Congress weighed in on the need for speedier executions when it eliminated the $20 million annual budget for Post-Conviction Defender Organizations which had provided some of the most sophisticated and effective counsel for death-row inmates in twenty death-penalty states); Roscoe C. Howard, Jr., The Defunding of the Post-Conviction Defense Organizations as a Denial of the Right to Counsel, 98 W. Va. L. Rev. 863 (1996). See also, e.g., Editorial, Death Penalty Blunder, Wash. Post, July 12, 2001 ("The Virginia Supreme Court has adopted guidelines that effectively slash the fees of court-appointed attorneys in death penalty cases. The move comes at a time when national attention has been focused on the low quality of counsel in capital cases."); Death Row Takes a Toll on Lawyers; Money, Resources Are in Short Supply, New Orleans Times-Picayune, Aug. 20, 2001 ("After Congress pulled federal funding for post-conviction appeals in 1996, the number of capital cases in Louisiana kept climbing and defense lawyers working pro bono for death row inmates already were scarce. With urging by the Louisiana Supreme Court, which has put specific death penalty appeals on hold until the defendants had lawyers, the state responded in 1999 and enacted a law that gives those on death row the right to representation. But legislators didn't provide the $1.3 million they themselves said was needed for the program.").

    203. At the national level, see the Anti-Terrorism and Effective Death Penalty Act of 1996 ("AEDPA"), Pub. L. 104-132, 110 Stat. 1214 (1996), which made a number of changes in preexisting federal habeas review, all designed to cut back on the amount of review and relief. See Randy Hertz & James S. Liebman, 1Federal Habeas Corpus Practice and Procedure §§ 3.2-3.5 (4d ed. 2001); supra p. 27 & n.132. U.S. Department of Justice data show that since AEDPA was adopted, federal habeas courts at the appellate level have used "procedural terminations" to decline to decide the legality of underlying convictions and sentences about 15-20% more often than had been the case before the Act was adopted-with no appreciable improvement in the amount of court time spent on the cases. See http://www.ojp.usdoj.gov/bjs/abstract/ppfc96.htm (data do not distinguish capital and non-capital cases; most cases are non-capital cases).

    At the state level, see, e.g., Florida Court Issues New Rules to Speed Up Appeals Process for Death Row Inmates, Associated Press Newswires, July 12, 2001; Duggan, supra note 19 ("The most significant death penalty change enacted during [George W.] Bush's six years as governor [of Texas] was a [1995] law that toughened and accelerated the appeals process for condemned inmates. During Bush's tenure, the state carried out 152 executions, an average of one every other week."). For more on these and other state-level changes, see 100 Colum. L. Rev., supra note 153, at 2136-38 & nn.252-53.

    204. For evidence that it has become more difficult to provide lawyers for defendants challenging death verdicts in state post-conviction and federal habeas proceedings, see Elizabeth Amon, The State of Pro Bono 2001, Nat'l L.J., Jan. 7, 2002 ("Lawyers in death penalty work say they've seen a drop in volunteerism. Capital cases are expensive and demanding, and those involved in placing attorneys on them say large firms in particular have shied away from them. 'It's a real crisis,' says Robin Maher, director of the American Bar Association pro bono death penalty representation project. 'In Alabama alone, there are 30 people on death row who don't have lawyers.'").

    205. Consider, e.g., Lloyd Schlup's exoneration, which occurred as a result of a type of federal habeas procedure that Congress withdrew in 1996. If a similar case arose today, there would be no judicial mechanism available to the innocent prisoner to avoid execution. See supra p. 27 & n.132.

    206. The discussion in this section is reprinted and expanded in "Effective Death Penalty?," supra note 152.

    207. See A Broken System, Part I, supra note 101, at 61, Figure 8; 102-07 & Figure 33.

    208. The combined state direct appeal and state post-conviction reversal rate and the federal habeas reversal rate have a .341 correlation (significance = .044, 1-tail; .088, 2-tail).

    209. See Maura Dolan, Execution Issue Clouds Davis' Judge Selection, L.A. Times, Nov. 13, 1999, at A1 (reporting complaints that California Governor Gray Davis, "haunted by the memory of former Chief justice Rose Bird . . . whom voters ousted because she never voted to uphold a death sentence," is demanding, in the opinion of Peter Keane, Dean of Golden Gate University Law School, that his appointees to the state's trial and appellate courts profess support for the death penalty as 'the greatest thing since sliced bread'"). See also Stephen B. Bright & Patrick J. Keenan, Judges and the Politics of Death: Deciding Between the Bill of Rights and the Next Election in Capital Cases, 75 B.U. L. Rev. 759, 760-65 (1995) (describing California voters' removal of three state supreme court justices due to their votes to reverse capital verdicts).

    210. See Paul Elias & Rinat Fried, A Failure to Execute, The Recorder, Dec. 15, 1999, at 1 (quoting federal judges blaming lax review by California state court judges for the time federal courts must devote to federal review of California death verdicts and for high federal court reversal rates; also quoting the chief justice of the California Supreme Court acknowledging that lax state court review in the past, during the tenure of Chief Justice Malcolm Lucas, may have placed a burden on federal courts).

    211. See Green, Study Critical of Death Penalty, supra note 30 (discussing December 2001 study by Virginia's Joint Legislative Audit and Review Commission concluding that "the strict adherence to rules by federal and state appeals courts may have let stand the conviction of some death row inmates who did not receive fair trials" and "suggest[ing that] the Virginia Supreme Court was overly strict in its use of state law in determining whether a death sentence was excessive or disproportionate").

    212. See A Broken System, Part I, supra note 101, at 106-07.

    213. See infra pp. 389-90.

    214. For press treatments of this issue, see, e.g., Linda Greenhouse, Death Penalty Gets Attention of High Court, N.Y. Times, Oct. 30, 1999, at A1 (noting the Supreme Court's "unusual flurry" of grants of review in Virginia/Fourth Circuit capital cases); Brooke A. Masters, 4th Circuit Is Steering Hard to the Right, Wash. Post, July 5, 2000, at B1 (finding in recent Supreme Court decisions a message "that the 4th Circuit had gone too far in death-penalty cases"); Tim McGlone, State's Death-Row Cases Draw High-Court Scrutiny, The Viginian-Pilot, Dec. 9, 1999, at A1 (describing recent Supreme Court decisions as "applying the highest level of scrutiny in modern times to [Virginia's] handling of death penalty cases"). For citation of the relevant cases, see Hertz & Liebman, supra note 203, § 3.2; 100 Colum. L. Rev., supra note 153, at 2060 n.105. The progression of Virginia cases in which the Court has granted certiorari continued recently. See Mickens v. Taylor, 121 S. Ct. 1651(Apr. 21, 2001) (mem.) (granting review of decision adverse to Virginia capital prisoner in Mickens v. Taylor, 240 F.3d 348 (4th Cir. 2001)); supra note 36 (discussing Court's October 2001 decision to review the constitutionality of the death sentence Virginia has imposed on retarded prisoner Daryl Atkins).

    215. See A Broken System, Part I, supra note 101, at App. C-49 (discussing (1) Virginia Governor's execution-eve grant of clemency to Earl Washington, due to doubts about his guilt, which placed him under a life sentence without parole until a subsequent Governor fully exonerated him based upon proof of his innocence, see supra pp. 27-30; and (2) conditional pardon of Joseph Giarratano, permitting him to secure a new trial, after the state and federal courts had declined to provide that same relief, notwithstanding the acknowledged illegality of Giarratano's initial conviction, because of their strict application of the technical waiver rules barring claims a prior lawyer had failed to present in the proper way at the proper time). See generally, Brooke A. Maters, New Evidence Overturns Murder Conviction in Va., Wash. Post, Nov. 14, 2001 ("Since 1989, Virginia has freed five inmates after DNA testing cast doubt on their guilt, but all five were released through gubernatorial clemency rather than the courts.").

    216. See infra pp. 218-19, 236.

    217. See infra pp. 323-24, 333-34.

    218. See supra p. 43.

    219. See supra pp. 22-35.

    220. See supra pp. 8-9, 11.

    221. See supra p. 11.

    222. See supra p. 69.

    223. See supra pp. 24 & n.128, 25-35.

    224. See supra p. 36 & n.150.

    225. See supra pp. 24 & n.128, 25-35.

    225a. Sources and Periods Covered in Figures 8A and *B:

    * Homicde and arrest data are for each of the 3 study states during the portion of the 1973-1995 study period when the state had a valid post-Furman death-sentencing statute. The Uniform Crime Reporting Program defines homicdes as "the willful (non-negligent) killing of one human being by another. The classification of this offense is based solely on police investigation as opposed to the determination of a court, medical examiner, coroner, jury, or other judicial body. Not included in the count for this offense classification are deaths caused by negligence, suicide, or accident; justifiable homicde; and attempts to murder or assults with intent to murder,which are scored as aggravated assults."

    † Homicide arrest data are only available fro 1977 to 1995. Arrests from 1973 to 1976 are calculated by determining the average number of arrests per homicide for each state from 1977 to 1995, then multiplying that number times the number of homicide for each study state during the relevant years from 1973 to 1976. Outliers were removed and replaced with an estimate based on the number of homicides. Missing values for particular states and years are estimed in the same manner.

    ‡ Information on murder convictions following homicide arrests from 1973 to 1985 is not available. the Source Book of Criminal Justice Statistics provides information on the number of murder convictions for all states combined for each even-numbered year between 1986 and the present. For even numbered years from 1986 to 1994, we began with the 50-state murder conviction statistic reported in the Source Book. For odd numbered years between 1987 and 1995, we estimed murder convictions as the average of the prior and following years. We then used that actual or estimated number of murder convictions for all 50 states to derive a corresponding number for each of the 34 study states. To do this we (1) derived a national rate of murder convictions per homice arrest for each year from 1986 to 1995 by dividing the number of murder convictions in each year by the number of homicide arrests for that year, then (2) multiplied the homicde arrests in each state and year by the national rate of convictions per arrest. To estimate murder convictions for the 1973-1985 period, we used the same procedure -- multiplying the national rate of murder convictions per homicide arrest for the post-1985 period that we derived from the Source Boo times the number of homicide arrests in each of the study states in each of the relevant years between 1973 and 1985. The estimates achieved in this manner are extremely rough because they assume that there is little variation among states in the rate of murder convictions per homicde, and that murder convictions prior to 1986 have the same relationship to homice arrests as we true in the period after 1985. We base no conclusions on the number of murder convictions reported in text and present the number only to provide as much context as possible, given the limited datat that are available, for the death-sentencing events in the table about which more is known. the authors of the Source Book define "murder" for this purpose as homicides (using the definitionin note * above) minus manslaughter offenses.

    ˆ Some dispositions occurred in the absence of final review by the courts. those dispositons are indicated on Figure 8A but not on Figure 8B, which only coveres dispositions following final review by the courts. Dispositions in advance of final court review include the execution of individuals who shose to give up their appeals and be executed consensually, as well as clemencies and exonerations that occurred prior to the completion of final court review. All dispositions are for death sentences and, where relevant, for court review occurring between 1973 and 1995.

    226. R = -.71; significance = .001. Figure 9 uses non-consensual executions for the reasons set out infra note 292, and to be consistent with Figures 6, 20, and 21. If, instead, all executions are considered, the "X" pattern of the two lines in the graph is nearly the same, and the negative correlation between those lines remains high and sifnificant: R = -.70; significance = .001.

    227. For examples, see supra notes 134, 146, 148 and accompanying text; 100 Colum. L. Rev., supra note 153, at 2130 n.242. See also Alexander & Chandler, supra note 199 ("Capital punishment in the Carolinas is so tainted with mistakes, inequities and incompetence that the system risks executing innocent people, while sparing some of the state' most vicious killser, a [Charlotte] Observer investigation has found."

    228. for a thoughful descriptionof the devastating effect on victims of reversible error in capital cases, see William H. Brill, Finality? Not for Us, and It's 17 Years Later, Wash., Post, Apr, 29, 2001 (Outlook), at B3:

    When James Strong was sentenced to death in Pennsylvania for the murder of my cousin Janes husband nearly two decades ago, there was s bitter sence of relief for our family. it wasn't that it was over -- noting like that can really be over -- but at least Jane and their 9-year-old son, jan, could begin to reorganize their lives. Jane moved out of the motel near the court, where she had been staying for the trial. She went bad to her job, back to her house; and Jay came home fromthe neighbors' house where he had beenstaying while Jane was at the trail.

    Justice had been served. The man who killed John Strock, would die for his crime. Or so we believed.

    But it didn't work that way. The death sentence -- with its protracted process of appeals -- locked Jane into a 17-year relations shp with her husband's murderer and even with his family. And it's not over yet.

    As someone who has made a career of styding violence, I'm as familiar as anyone with the arguments for and against the death penalty: that is is proper vengeance, and that some people deserve to die; or that it is inhumane and ofter applied unfairly. But I had never given much thought to its effect on members of the victim's family, to what it really means for them. I don't think our family was unususal in believing, in the simplest terms, that the death penalty would make us feel better, that it would provide, as they say, "closure."

    Our family approved of the verdict, particularly Jane. She was anguised and she wated to see her husband's murderer die. She appreciated the support she got: Members of the local police force where she lived outside Philadelphia even offered to go with her to witness the execution.

    But Strong, the convicted man, didn't die. He stayed on death row.

    Oh, we expected that to happen -- at least for a time. But the years went on -- two, five years, then 10,1 5. While Jane waited for some resolution, Jay grew up. She tried not to dwell on Strong or his execution. there wer football games to watch, a college to choose, checks to write.

    For the extended family, thoughts about Strong gradually receded. It has always been awkward to talk about him, anywan....So we all kind of put it aside. Which meant kind of putting the memory of Jane's husband aside.

    I was no different. Even though I had worked as a professor, a security consultant and an expert witness in criminal behavior, and had traveled the country analyzing serious criminal cases, I found it just as difficult to inquire. To do so would bring it all up again. How John has picked up two hitchhickers on his way home from visiting his parents in West Virginia in 1983; how they made him dirve off the highway onto a deserted rad new Wilkes Barre, PA; how they marched him offo into the woods, robbed him and shot him with a sawed-off shotgun.

    *****

    From time to time after the conviction and sentencing, Jane would speak about the delays. Always with confusion. "I don't know what they're doing," she would say. "It's a legal thing, something about him going into a different pool, getting placed on a different schedule."...

    I never pressed her. I could tell she hadted to call the district attorney's office. But I also knew that she was wondering whether the man who killed her husband was ever going to be executed....

    Then, four years ago Jane got a call. It wasn't the call she had expected. "Just a routine hearing," she reported being told. "You don't even need to come up." But Jane felt she had to go. "Someone ought to be there for John," she told me. And so she took time off from her job in marketing and drove back to Wilkes Barre. Same trip, but a different courtroom.

    Jane called me when she got back. "It was horrible," she said. "He was there, only this time he looked clean-cut. and he was in a suit, not al scraggly with long, wild hari like he was before." His family was also there, still saying they were sorry about what had happened, still insisting that their "Jimmy" didn't do it.

    "But what about the hearing?" I asked. "I could hardly follow it," she said. "It seemed like they went over the same old thing. But they told me when it was over: No problem. That it was all routine."

    I heard nothing more until last December, when a large manila envelope arrived in my mail. It looked like a calendar, probably a Christmas present, from Jane. I opened it and discovered it was an opinion from the Pennsylvania Supreme Court. Strong's conviction has been overturned. "This matter is remanded for a new trial," the document read.

    the grounds for that decision were that letters had just come to light suggesting that the DA's office had made a deal at the time of Strong's trial with the attorney for his partner in the crime, James Alexander, and had failed to disclose this to the court or to the jury--or even to the assistant DAs who prosecuted the case. Alexander has been higchiking with Strong when John picked them up, and it was he wo had testified that Strong was the shooter. During Strong's trial, Alexander denied having made any deals. Later, in his own trial, he pleaded guilty to a lesser charge of third-degree murder, he was sentenced to 40 months in prison.

    And so it continues. Within the next few months, the whole process will begina again--with a new trial set for September and pretrial motions are due in June. Only this time it's quite possible that Strong will leave the courtroom a free man. Alexander, whose testimony probably wouldn't have been believable anyway one the jury heard there had been a deal, has since died of a heart attack. ...Who knows what the jury will decide? That it was Alexander and not Strong who shot John?

    Of course, there could have been a reversal in any case. Nobody wants a decision to be made on incomlete evidence, or a prosecutor to fail to disclose relevant facts. Nobody wants an innocent man to be put in prison for life. let alone put to death. What's more, I can't help wondering what the possibility of getting a death penalty verdict couldmean to a prosecutor. Is it a trophy? In Strong's case, could it have encouraged the prosecutor to overrreach, even to the point of cutting a bad deal with a codefendant? Those are all ver important questions. But they're not my main point.

    My point is that the death sentence had lulled Jane and the rest of us into believeing thatshe could put one part of her life behind her, that Strong's death would indeed bring her closure.

    I won't ever know for sure how Jane would have felt if Strong had been executed. Given the questions that have prompted the new trial, that could not have been a good thing for the cause of justice. The delays that have characterized this case are typical, though. And becausr of them, Jane was watched and waited year after unsettling year, only to discovered now that the man whom she believe all these years murdered her husband may go free. Not long ago, she wondered out loud to me what kind of manhe is, whether he's been rehabilited. "You don't suppose he will come after me, do you?" she asked.

    So Jane's uneasy relationship witht he man convicted of murdering her husband contineus. Whatever some people believe about the value of the death penalty, it hasn't solved anything for our family, and has probably made it all the worse.

    229. See supra p. 75, Figure 7A.

    230. See Figure 10, p. 93 below.

    231. The relevant studies of the cost of the death penalty and of curing its mistakes are collect in 100 Colum. L. rev., supra note 153, at 2129-33 & nn.243-46.

    232. See, e.g., State Law Seeks to Provide Strong Defense: Most Capital Cases are Handled Well, But Examples of Inadequate Representation Show Lapses in the System, South Bend Trib., Oct. 22, 2001 ("Prosecutors tend to agree the death penalty's requency of being carried out -- and any deterrent factor that may result--is lessened when a sentence is overturned.").

    233. See, e.g., Paul Craig Roberts, supra note 49 (quoted supra note 49). Much of the commentary in the wake of the discover that even Timothy McVeigh's state-of-the-art federal capital prosecution was marred by serious error on the part of the FBI in failing to turn over thousands of pages of documents to which the defense was entitled has focused on the damage done to the integrity of the process. An editorial in USA Today is illustrative:

    Editorial, McVeigh Errors Raise Doubts About Other Capital Cases; Our View: FBI Blunders Spotlight How Often Death-Penalty Cases Are Flawed, USA Today, May 16, 2001. See also Jordana Horn Marinoff, If FBI Has Erred on McVeigh, What About Other Cases, Milwaukee J. Sentinel, May 20, 2001. For another example, see Editorial, Imperfect Justice, Wash. Post, May 28, 2001; Page A22:

    234. See supra p. 52.

    235. See supra pp. 40-43 & nn.167, 169.

    236. See 100 Colum. L. Rev., supra note 153, at 2101-36, for a detailed demonstration and documentation of this point.

    237. See id. For evidence that the public is catching on to problems with the capital system that have long been known only to the professionals who operate it, see, e.g., Editorial, Seattle Post-Intelligencer, Aug. 12, 2001 ("Last week a special report in the Post-Intelligencer . . . shed light on a horrible truth known mostly, but well, to the professionals who prosecute, defend, adjudicate and review the first-degree aggravated murder cases that are the only ones that can lead to lethal injection in this state. Of the existing system's well-documented flaws, none is greater than the economic disadvantage that some capital defendants face in trying to secure justice in Washington.").

    238. See supra pp. 70-81.

    239. We follow the usual convention of noting relationships where the probability that they exist by chance is 10% or less, while placing primary emphasis on, and drawing policy implications from, relationships where the probability that they exist by chance is 5% or less. For additional discussion of statistical significance, see infra pp. 109-10, 148.

    240. See supra pp. 70-76.

    241. See supra pp. 79-80 & Figure 9.

    242. See supra pp. 48-61.

    243. See Furman v. Georgia, 408 U.S. 238 (1972).

    244. See Woodson v. North Carolina, 428 U.S. 280 (1976); supra p. 18 & n.108.

    245. See supra p. 18.

    246. See infra pp. 89-91, 99, 140-42 & Table 4, 154-56, 177, 194-95, 213, 216-17, 258-59, 372-75.

    247. See infra pp. 102-03, 222-33 (discussing Analyses 14 and 15), 107-08, 272-78 (discussing Analyses 16 and 17), 334 (discussing one group of explanatory factors studied within the context of Analysis 19 that did not include time).

    248. See supra pp. 20-21.

    249. See infra pp. 89-91, 99, 140-42 & Table 4, 154-56, 177, 194-95, 213, 216-17, 258-59, 372-75.

    250. See 100 Colum. L. Rev., supra note 153, at 2129-36.

    251. See infra pp. 140-42 & Table 4.

    252. The one exception is Analysis 10A, which examines state- and county-level factors associated with county reversal rates at the state direct appeal stage.

    253. See supra pp. 17-18; infra Appendix C, pp. C-1 to C-2; A Broken System, Part I, supra note 101, at 19-20, 26-27 & nn.39, 132; App. C, pp. C-1 to C-2.

    254. See infra p. 99 & n. 257.

    255. See supra pp. 88-92.

    256. See supra pp. 14-81.

    257. All of our binomial analyses are over-dispersed and logistic, even where we omit those descriptions for ease of exposition.

    258. See supra pp. 89-92; infra pp. 140-42 & Table 4, 154-56, 177, 194-95, 213, 216-17, 258-59, 372-75.

    259. By adding an offset to transform the dependent variable into a rate, our Poisson analyses may be used to predict rates as well as counts. All of our Poisson analyses are over-dispersed and logarithmic, even where we omit those descriptions for ease of exposition. For discussions of the rationale for using both Poisson as well as binomial regressions, see supra p.97; infra p. 109.

    260. See supra pp. 21, 88-89.

    261. See supra pp. 95-96.

    262. See supra p. 100.

    263. See supra pp. 17-18; infra Appendix C, pp. C-1 to C-2.

    264. See supra pp. 17-18 & n.103.

    265. See supra p. 99.

    266. The numbering of analyses jumps from 6 to 14 because we conducted all our comparisons of the states' 23-year experiences last (Analyses 14-17), after first completing our state analyses (Analyses 1-6) and county-state analyses (Analyses 7-13) in which we disaggregated the analysis by year.

    267. Although just over 1000 counties imposed death sentences during the study period, the year of the death sentence is known for verdicts imposed in only 967 of those counties. In studies in which the role of time is studied, we use this 967 cohort of counties. The additional counties are included in studies in which the year the death verdict was imposed is not a consideration. See infra page 107.

    268. See infra p. 251 (discussing this feature of Analysis 7).

    269. See supra p. 103 & n.267.

    270. See supra pp. 88, 100 (discussing the rationale for treating state and year as random effects).

    271. See supra pp. 102-03.

    272. See also supra pp. 87, 99-100.

    273. See supra pp. 97-98.

    274. See infra pp. 238, Table 7; pp. 283, Table 8; pp. 286, Table 9.

    275. See supra pp. 85-86.

    276. See supra pp. 52, 82-83.

    277. For more on measuring fit, see infra pp. 149-50 & n.311.

    278. Using a random intercept (measuring unexplained variation in reversal rates from one place to another), and a measure of slope (measuring unexplained variation over time), we compare two sets of explanatory factors to see which leaves more unexplained variation. A lower valueCi.e., less unexplained varianceCis better. In this test, the explanations generated by a given set of factors, and by the baseline analysis, are compared to see if the specific factors leave less unexplained variation than the baseline analysis. In the test discussed in the last item in the list in text, the explanations generated by two or more sets of factors are compared to see which one leaves less or the least unexplained variation.

    279. A relationship that is not statistically significantCe.g., one as to which there is an 89% probability that it does not appear by chance but an 11% probability that it doesCconveys some useful information about factors that may be related to capital reversal rates. Nonetheless, we choose to base no findings on, and to draw no conclusions from, relationships about which we are not highly confident.

    280. Our effect-size estimates are coded as "newestimates" in the detailed results of each of our Analyses 1-18 in Appendix G. For a description of how these effect-size estimates are calculated, see infra notes 283, 286.

    281. See generally supra pp. 97-98, 99-100, 110.

    282. Assume an increase in the homicide rate from 5 to 6 homicides per 100,000 residents. Assume, as well, that the "odds" of reversal where the homicide rate is 5 per 100,000 residents are 1 to 2, which is equivalent to a probability of 33%. (Odds of x to y can be changed to a probability using the following formula: x/(x + y). So, if the odds are 1 to 2, the probability is 1/(1+2) = 1/3 = .33.) Given these assumptions, and given a "newestimate"--or effect-size estimate--of 1.4, the regression analysis predicts that the odds of reversal where the homicide rate is 6 per 100,000 residents are (1 x 1.4) to 2, or 1.4 to 2, which is equivalent to a 41% probability. Where the homicide rate is 7 per 100,000 residents, the predicted odds of reversal are (1 x 1.4 x 1.4) to 2, or 1.96 to 2, which is equivalent to a 49% probability.

    283. Factors with values that are not integers (i.e., values such as .43, 1.22, 10.54 as opposed to 1, 4 and 7), and particularly factors that are scaled by reference to some other population (e.g., homicides per 100,000 residents) are often logged to make their effect size easier to interpret. In the example in note 282 above, we assumed an increase in the homicide rate from 5 to 6. But states rarely have an actual homicide rate per 100,000 residents that is an interger (e.g., 2) as opposed to, say, 1.3 or 4.6. To help interpret effect size, therefore, the values are logged to the base 10. (For any value, n, we calculate n = 10x, and use x as the new value. So if n = 100, then x = 2, because 10 = 102..) The effect is to compress the values onto a much narrower scale. All logged factors in our results are indicated by an "l" at the beginning of the variable name in our study results (e.g., "ldor").

    Effect size for logged factor values in binomial logistic analyses is calculated using the following formula. For each doubling of the value of an explanatory factor, the predicted "odds" of reversal increase by a factor of x, with x being the effect-size estimate (i.e., the "newestimate") reported in our study results. So, if homicide rates are logged in an analysis, and if the "newestimate" is 1.4, an increase from 2 to 4 homicides per 100,000 residents increases the predicted odds of reversal by a factor of 1.4. If the odds of reversal where the homicide rate is 2 per 100,000 residents are 1 to 2 (33%), the predicted reversal rate of a state with a logged homicide rate of 4 per 100,000 residents is (1 x 1.4) to 2 = 1.4 to 2 (41%).

    The method we used to calculate the "newestimates" for unlogged variables in binomial logistic analyses is to exponentiate the parameter estimate for the factor of interest (coded "Estimate" in the results in Appendix G). The method we used to calculate the "newestestimate" for logged variables in binomial logistic analyses is to calculate 2 to the power of the parameter estimate for the factor of interest.

    Numeric effect-size estimates for variables that are also components of interaction effects variables (e.g., "bltot," which is a component of our "bltot*fac_clsd" variable), need to be interpreted in conjunction with the effect-size estimates for the interaction variable. In our effect-size graphsCon which we exclusively rely in the text of this Report when discussing effect size for variables that are components of interaction variablesCwe display the joint effect of the two variables, but we have not performed that additional calculation in the numeric effect-size estimates.

    284. The formula used to calculate effect size in the graphs of factors in our binomial analyses is (all values for the factor of interest) x (that factor's parameter estimate (coded "Estimate" in the detailed results in Appendix G)) + (the sum of (each other factor's parameter estimate x that factor's mean value)). The estimates obtained from this formula then were transformed using an inverse logit transformation, which we multiplied times 100 to derive a percentage, which then served as the y-axis value on the graph. Where a factor of interest was also a component of an interaction effect, the interaction effect was taken into consideration in calculating effect size in the graphs.

    285. Assume that the homicide rate increases from 5 to 6 per 100,000 residents, and that the predicted reversal rate where the homicide rate is 5 per 100,000 residents is 33%. In this event, the analysis predicts a reversal rate in states with a homicide rate of 6 per 100,000 residents of (.33 x 1.4), or .46 (46%). Where the homicide rate is 7 per 100, 000 residents, the predicted reversal rate is (.33 x 1.4 x 1.4), or .65 (65%). And so on.

    286. Effect size for logged factor values in Poisson logarithmic analyses (see supra note 283) is calculated using the following formula: For each doubling of the value of an explanatory factor, the predicted rate of reversal increases by a factor of x, with x being the effect-size estimate (i.e., the "newestestimate") reported in our study results. So, if homicide rates are logged in a study, and if the "newestimate" for that factor is 1.4, an increase from 2 to 4 in homicides per 100,000 residents increases the predicted probability of reversal by a factor of 1.4. If the reversal rate associated with a logged homicide rate of 2 is 33%, the predicted reversal rate for a logged homicide rate of 4 is (.33 x 1.4) = .46, or 46%.

    The method we used to calculate the "newestimate" for unlogged variables in Poisson logarithmic analyses is to exponentiate the parameter estimate for the factor of interest (coded "Estimate" in the results in Appendix G). The method we used to calculate the "newestimate" for logged variables in Poisson analyses is to calculate 2 to the power of the parameter estimate for the factor of interest. Although we used the same method to calculate "newestimates" for binomial and Poisson analyses, see supra note 283, the interpretation of the "newestimates" is different.

    Numeric effect-size estimates for variables that are also components of interaction effects variables (e.g., "bltot," which is a component of our "bltot*fac_clsd" variable), need to be interpreted in conjunction with the effect-size estimates for the interaction variable. In our effect-size graphsCon which we exclusively rely in the text of this Report when discussing effect size for variables that are components of interaction variablesCwe display the joint effect of the two variables, but we have not performed that additional calculation in the numeric effect-size estimates.

    287. The formula we use to calculate effect size in the graphs of factors in our Poisson logarithmic analyses is (all values for the factor of interest) x (that factor's parameter estimate (coded "Estimate" in the detailed results in Appendix G)) + (the sum of (each other factor's parameter estimate x that factor's mean value)). The estimates obtained from this formula then were transformed using an exponential transformation, which served as the y-axis value on the graph. Again, where a factor of interest was also a component of an interaction effect, the interaction effect was taken into consideration in calculating effect size in the graphs.

    288. The percent change is calculated using the following formula: Moving from a homicide rate of x1 to a homicide rate of x2 (as indicated on the horizontal axis) leads to a predicted percent change in reversal rates of ((y2 - y1)/y1) x 100 (with the y values being those indicated on the vertical axis). In the example given: ((.4-.2)/.2) x 100 = (.2/.2) x 100 = 100%.

    289. See infra pp. 311-12.

    290. See supra pp. 37-39, 65-68 & Table 5.

    291. Figure 16 is the same as Figure 1A above.

    292. Figure 21 is the same as Figure 6 above. As do Figures 6 and 9, Figures 20 and 21 examine non-consensual executions, meaning ones where the prisoner challenged his or her death verdict in court and, as a result, secured a judicial determination of the quality of the verdict. Excluded are executions occurring consensually, because a prisoner decided he preferred execution to prison and did not appeal his verdict or, therefore, secure a judicial determination whether the verdict was seriously flawed. Because our goal is to gauge relationships between various conditions and the presence or absence of serious, reversible capital error, verdicts carried out in the absence of any judicial determination of whether or not they were seriously flawed are less informative in this context.

    293. See, e.g., Associated Press State & Local Wires, supra note 30 (discussing New Jersey "state Supreme Court report released [in August 2001] that found vast differences in how often prosecutors seek capital punishment" in the state); Stefan C. Friedman, District Attorneys Above the Law, N.Y. Post, Aug. 22, 2001 (discussing decision of some district attorneys in New York state to forgo use of the death penalty); supra note 30 (discussing study by legislative advisory committee finding evidence of disparities in capital-charging in Virginia); infra p. 396 & n.880 (discussing evidence of local disparities in capital-charging found by legislatively commissioned study of Nebraska capital practice).

    294. See supra pp. 84-85.

    295. There are almost no alleged constitutional or other violations that have a statistically significant relationship to the probability of reversal. Claims such as ineffective assistance of counsel that are the most common reasons for reversal (see supra pp. 40-41) are raisedCand failCin most cases and thus are not associated with a higher (or lower) probability of reversal. The one claim that is significantly related to a higher probability of relief addresses an idiosyncratic practice of a single state (Texas) that was eventually held to be unconstitutional by the U.S. Supreme Court. See Estelle v. Smith, 451 U.S. 454 (1981).

    296. See infra pp. 157-59, 160-63, 165-66.

    297. See supra pp. 89-92, 99.

    298. See supra pp. 89-93 & n.10, 99.

    299. See supra pp. 20-21, 91-93 & Figure 10.

    300. See supra pp. 89-92.

    301. Additional notes on data and sources:

    1. DADB: For a direct appeal to be "finally decided" within the study time period, the highest state court with jurisdiction to review capital judgments in the relevant state must have taken one of two actions during the study period: (1) affirmed the capital judgment, or (2) overturned the capital judgment (either the conviction or sentence) on one or more grounds. If one of those two actions occurred prior to or during 1995, and the United States Supreme Court thereafter denied certiorari review, the case is included in the study, because the Supreme Court's action did not affect the finality of the state decision. If the Supreme Court instead granted certiorari in a case but did not decide the case before or during 1995, the case is omitted from the study because the Supreme Court's action withdrew the finality of the decision.

    2. HCDB: All information on federal habeas petitions was collected from judicial decisions handed down during the study period. Starting with a list of names of individuals sentenced to death, habeas data were collected during the summers from 1994 to 1999 using either a Westlaw or a Lexis search engine. For a habeas case to be "finally decided" within that time period, all of the following events must have occurred in the case within the study period: (1) a United States District Court must have denied habeas corpus relief, thereby approving the capital judgment, or granted habeas relief from the capital judgment (either the conviction or sentence) on one or more grounds; (2) if an appeal was timely filed, a United States Court of Appeals must have approved or reversed action (1); and (3) if certiorari review was timely filed, the United States Supreme Court must have either (a) denied review or (b) granted review and affirmed or reversed action (1).

    3. DRCen: Our information on death verdicts is catalogued by the name of the prisoner and state that imposed the death sentence. Some individuals appear on the list more than once because they had multiple capital verdicts during the study period. Multiple capital verdicts for the same individual may result from verdicts imposed at different trials for different offenses; verdicts imposed at different trials for crimes against different victims of the same offense; or new verdicts imposed following court reversal of earlier verdicts. Our unit of analysis is the outcome of appeals of capital verdicts. If a single court judgment was imposed at trial covering more than one victim or offense and was reviewed on appeal as a single judgment, it is counted as a single verdict. If an appellate court disaggregated a judgment imposed at trial into multiple verdicts covering different offenses or victims, affirming one or more and reversing one or more, we followed the same practice.

    4. USCen: Census Bureau data on the racial composition of states are from Census Data Set PE-19 1970-79; State Estimates by Age, Sex, and Race; Estimates of the Population of States by Age, Sex, Race and Hispanic Origin: 1981 to 1989; Estimates of the Population of State by Age, Sex, Race and Hispanic Origin: 1990 to 1998; Summary Tape File 3C(STF3C).

    5. UCRDB: U.S. Department of Justice, FBI, Uniform Crime Reports for the United States: Crime in the United States, 1973-1996, available in spreadsheets at http://www.ojp.usdoj.gov/bjs/datast.htm.

    6. HomVic: Data for 1973-78 are from Vital Statistics of the United States, Mortality Detail Files, 1968-1978 (ICPSR STUDY 7632). Data for 1979-87 are from CDC Wonder, the Centers for Disease Control data extraction engine at http://wonder.cdc.gov. Data for 1988-95 are from Centers for Disease Control and Prevention National Center for Health Statistics, Compressed Mortality File, 1989-98 CD-ROM Series 20, No 2C ASCII Version. Through 1988, the relevant data sources list homicide victims by county of death. After 1988, the relevant data source lists homicide victims by county of residence.

    7. PrisCen: U.S. Department of Justice, Bureau of Justice Statistics, Source Book of Criminal Justice Statistics 1973-1996.

    8. CtCaLd: Inter-University Consortium for Political and Social Research, State Court Statistics 1985-1994 (ICPSR 9266, 1995). Information on the filings and dispositions of state court cases are only available for 1985 to 1995.

    9. CtExpen: Expenditure and Employment Data for the Criminal Justice System [United States]: CJEE Extract Files 1971-1995, available from the Inter-University Consortium for Political and Social Research, ICPSR datasets: 2257, 2940, 6006, 6459, 6579, 6795, 7618, 8382, 8455, 9160, 9161, 9162, 9396, 9554, 9773.

    10. Messner county data: Professor Messner and his colleagues collected data on every county in the United States for the decennial census years 1970, 1980 and 1990. The intervening years were interpolated using SPSS's linear interpolation procedure.

    11. PolPres: See C. Flango & D. Rottman, Appellate Court Procedures (Nat'l Center for State Courts 1998).

    302. See supra pp. 97, 100, 110.

    303. See generally supra pp. 109-17.

    304. See supra p. 140.

    305. Full results of these analyses and all others for Analyses 1 and 2 are available in Appendix G.

    306. p = .0006 in both Analyses 1 and 2. We indicate the significance levels here to assist the reader in comparing this first discussion of results to the more fuller description of results in Appendix G. Below, we do not give specific significance levels in every case and instead describe the significance levels more generally (making the usual assumption that a level of .05 or less is "significant," that a level of less than .0001 is "highly significant," and that a level greater than .05 but less than .1 is worth noting. See supra pp. 85-86 & n.239, 109-10; infra p. 148. In regard to the random intercept, see supra note 278.

    307. p = .01 (Analysis 1) and .009 (Analysis 2). In regard to slope, see supra note 278.

    308. See supra pp. 85-86 & n.239, 109-10.

    309. See supra p. 112; infra note 311.

    310. On the baseline analysis, see supra pp. 113 & n.278, 147.

    311. Our measure of fit is the -2 Res Log Likelihood. In general, one expects an improvement in fit of 1 for each additional factor tested beyond those included in the baseline analysis. Improvements in fit compared to the baseline inquiry of about 2 or more per added factor tend to be statistically significant in analyses with the number of factors that our analyses add to the baseline inquiry. See Critical Values of the Chi-Square Distribution, in Jeffrey M. Wooldridge, Introductory Econometrics: A Modern Approach App. G. (1999). Below we list the number of factors added to the baseline by Analyses 1A-2B and the improvement in fit achieved by each. All are highly significant.

      No. of Factors Added to Baseline Improvement in Fit
    Analysis 1 (binomial):
      Analysis 1A
    10 86
      Analysis 1B
    10 99
    Analysis 2 (Poisson):
      Analysis 2A
    10 151
      Analysis 2B
    10 89

    312. See supra pp. 146-47 & nn.306-07.

    313. Statistical significance is discussed supra pp. 85-86 & n.239, 109-10, 148. As the coefficients decrease, indicating less unexplained variance, significance values (in parentheses in Table 5) get higher, meaning that the probability that the remaining variation appears merely chance is growing. A significance value higher than .05 indicates that one cannot be confident that the variance left unexplained by the set of factors being analyzed is more than a chance occurrence—i.e., that the set of factors has explained most of the variance.

    Correlation matrices are provided in Appendix F. They show that our measure of the effectiveness of law enforcement (arrests, convictions and incarcerations per 100 FBI Index Crimes) is correlated with the passage of time, and that our measure of general court cases is correlated with our measure of population structure. We left these factors in the model because we believe that all of them are conceptually important in order to test our hypotheses, and also because the correlated factors operate in distinct conceptual domains—i.e., they measure conditions whose sources are independent. As a check, we omitted the weaker of the two correlated variables (given significance levels and effect size) from Analysis 1A to see if doing so changed our results. We first omitted only the measure of general court caseloads. Then, we omitted both that factor and time trend. In each case, the impact on fit was minimal, and the other factors remained significant or were between the .05 and .1 level. (When the measure of general court caseloads is omitted from Analysis 1A, p = .06 for white compared to black homicide victimization rates, and also for the interaction between that factor and the African-American proportion of the population. When both general court caseloads and time are omitted, p = .09 for those two factors. All other factors remain significant at the .05 level or better.) Removing the two variables did, however, weaken the explanatory power of the model, leaving substantially more unexplained variance than Analysis 1A, while not substantively changing the interpretation or implications of the analysis or the important variables. Under these circumstances, we conclude that more is learned by including the two variables than by excluding them, for three primary reasons. First, some degree of multi-colinearity is inevitable in measures of real-world processes or phenomena. Second, the correlated variables are conceptually important to test our hypotheses, as is discussed at length below. Finally, the correlated variables are not correlated with the explanatory variable on which we premise our principal conclusion: states' and counties' death-sentencing rates.

    314. See supra pp. 89-91, 99, 140-42.

    315. During the study period, 5826 verdicts were imposed. Of those, 2109 were reversed at the first two review stages and thus were removed from the system before reaching the final review phase. 598 verdicts were finally reviewed at the third, federal habeas stage. That leaves 3119 verdicts (54% of the total) that were imposed but were not finally reviewed during study period. We know that during that period, 1280 verdicts (22% of the total) were not finally reviewed at any of the three review stages. Our best estimate is that about the same number of verdicts—1232—cleared the first stage of review but were stuck at the second stage of review as of the end of the study period, and that about 607 verdicts cleared the second stage of review but did not complete the third stage by the end of the study period. See supra note 103.

    316. See supra pp. 88-90.

    317. See supra pp. 89-90.

    318. This factor is coded "bltot" in the detailed results in Appendix G. At pp. 89-90 above, we explain why we made this variable a count, not a rate tied, e.g., to each state's population, number of homicides or number of death sentences. In brief, we designed the variable to treat big and small states the same, given their susceptibility to the same bottle-necking effect of having a single state high court responsible for two of the three review phases. In any event, our use of an explanatory factor for population size, see infra p. 168, introduces a control for differences in the size of state populations.

    319. This four-part "factor" is coded "fac_csld" in the detailed results in Appendix G.

    320. This factor is coded "rgrtbklg" in the detailed results in Appendix G. Both alternative measures of all cases awaiting review overlap our measure of backlogged capital cases, because capital cases are included in the categories of filed felony cases, and all filed cases. Capital cases, however, make up a tiny proportion of all filed, and even all filed felony, cases.

    321. As we develop below, effect size for this factor is too small to warrant much attention to it.

    322. p = .02 (Analysis 1B and 2A); p = .06 (Analysis 1A).

    323. p = .01.

    324. See supra p. 93, Figure 10.

    325. See supra pp. 89-91, 99, 140-42, 152-53.

    326. See supra pp. 20-21, 91-93 & Figure 10.

    327. This factor is coded "year-n" in the detailed results in Appendix G.

    328. Below, we compare the baseline inquiry and our analyses of specific factors using three measures—t values, effect size and significance. A negative sign indicates an inverse relationship between the passage of time and reversal rates. Weaker relationships between the passage of time and error rates as one moves from the baseline to the specific analysis, are indicated by lower t values, smaller distances between the effect size coefficient and 1, and declining significance (indicated by a higher probability that the relationship appears by chance). Of the three measures, effect size is the most useful for present purposes.

    Strength of the Relationship Between the Passage of Time
    and Capital Error Rates: Analyses 1 and 2
      t Value Effect Size Significance
    Analysis 1 (binomial)
    Baseline Analysis -9.76 .82 <.0001
    Analysis 1A -2.77 .94 .006
    Analysis 1B -2.63 .95 .009
     
    Analysis 2 (Poisson)
    Baseline Analysis -6.36 .89 <.0001
    Analysis 2A -2.47 .96 .014
    Analysis 2B -2.85 .95 .005

    329. See infra pp. 194-95, 258-59.

    330. See, e.g., McCleskey v. Kemp, 481 U.S. 279 (1987) (opinion of Stevens, J., dissenting) (summarizing relevant studies); David C. Baldus et al., Equal Justice and the Death Penalty: A Legal and Empirical Analysis (1990); David C. Baldus, George Woodworth & Charles A. Pulaski, Jr., Reflections on the >Inevitability' of Racial Disrimination in Capital Sentencing and the >Impossibility' of its Prevention, Detection, and Correction, 51 Wash. & Less L. Rev. 359, 386 n.115 (1994) (citing studies); J.C. Beck & R. Shumsky, A Comparison of Retained and Appointed Counsel in Cases of Capital Murder, 21 L & Hum. Behav. 525, 534 (1997); William J. Bowers, The Pervasiveness of Arbitrariness and Discrimination Under Post-Furman Capital Statutes, 74 J. Crim. L. & Criminol. 1067 (1983); William J. Bowers & Glenn L. Pierce, Arbitrariness and Discrimination Under Post-Furman Capital Statutes 26 Crime & Delinq. 563 (1980); Samuel R. Gross & Robert Mauro, Patterns of Death, 37 Stan. L. Rev. 27 (1984); Joseph E. Jacoby & Raymond Paternoster, Sentencing Disparity and Jury Packing: Further Challenges to the Death Penalty, 73 J. Crim. L. & Criminol. 379 (1982); T. J. Keil & G. F. Vito, Race and the Death Penalty in Kentucky Murder Trials: 1976-1991, 20 Am. J. Crim. Justice 17 (1995); Raymond Paternoster & A. Kazyaka, Racial Considerations in Capital Punishment: The Failure of Evenhanded Justice, in Challenging Capital Punishment: Legal and Social Science Approaches 113-48 (K. C. Haas & J. A. Inciardi eds., 1988); Raymond Paternoster, Prosecutorial Discretion and Capital Sentencing in North and South Carolina, in The Death Penalty in America: Current Research (R. M. Bohm ed. 1991); J. R. Sorensen& D. H. Wallace, Capital Punishment in Missouri: Examining the Issue of Racial Disparity, 13 Behav. Sci. & L. 61 (1995); E. Thomson, Research Note: Discrimination and the Death Penalty in Arizona, 22 Crim. Justice Rev. 65 (1977); U.S. Gen. Accounting Office, Death Penalty Sentencing: Research Indicates Patterns of Racial Disparities, Report # GGD-90-57 (Feb. 26, 1990).

    331. This factor is coded "pctblack" in the detailed results in Appendix G.

    332. We also considered whether the proportion of nonwhites in a state's population, and not just the proportion of African-Americans, is associated with high error rates, but the proportion of African-Americans had more explanatory power in terms of significance, fit explained variance and effect size.

    333. In a number of these analyses, we used factors (focused on the race of the victims of the offenses for which death sentences were imposed) for which we were missing data for 23 of the 519 observed "state-years." This required us to omit those state-years, diminishing the number of observations to 496, which automatically lowers the fit value somewhat. The fit for the Analysis 1A models with 23 missing values thus cannot be compared to the fit for analyses with all 519 observations.

    334. One factor focused on the race of people sentenced to die did have a significant effect in some Analysis 1R models, but that relationship tends to confirm the importance of the size of the state's African-American population generally. The factor is the percentage of blacks among those the state sentenced to die minus the percentage of blacks in the state's population (coded "pctpdiffnew" in the detailed results in Appendix G). In some analyses, this factor was negatively related to error rates, meaning states with lower proportions of blacks in their population than on death row tend to have lower reversal rates, while states with higher proportions of blacks in their population than on death row tend to have higher reversal rates. This factor was only occasionally significant, and added nothing to the power of the pre-existing set of factors, so we give it little weight. But it tends to support the finding of Analyses 1 and 2 that it is the proportion of African-Americans in the states' population, not their proportion among the people states sentence to die, that is associated with higher error rates.

    335. We do not calculate these reversal rates by race for the state post-conviction stage, because we do not have sufficient information for that stage to compute them accurately. See supra pp. 17-18; Appendix C, pp. C-1 to C-2.

    336. The only racial disparity of any note suggested by Table 6 is for death verdicts imposed on defendants who are neither African-American nor white, which have a somewhat lower reversal rate, especially at the habeas stage. The numbers are small, however, especially at the habeas stage, where the relationship is not significant when other factors are considered.

    337. See, e.g., United States General Accounting Office Report on Pattern of Racial Disparities 330 (collecting and evaluating studies documenting effects of victim's race on decision to seek the death penalty and concluding that, in "82 percent of the studies, race of victim was found to influence the likelihood of being charged with capital murder or receiving the death penalty, i.e., those who murdered whites were found to be more likely to be sentenced to death than those who murdered blacks. This finding was remarkably consistent across data sets, states, data collection methods, and analytic techniques. The finding held for high, medium, and low quality studies."); David C. Baldus et al., Equal Justice and the Death Penalty, supra note 330, at 370-93; David C. Baldus, et al., Racial Discrimination and the Death Penalty in the Post-Furman Era: an Empirical and Legal Overview, with Recent Findings from Philadelphia, 83 Cornell L. Rev. 1638, 1658-61 & n.69 (1998) (collecting numerous additional sources, many of them conducting studies of particular capital states); John Blume, Theodore Eisenberg and Martin T. Wells, Explaining Death Row's Population and Racial Composition, Cornell Law School (Draft Oct. 19, 2001); Bowers & Pierce, supra note 330, at 594 tb. 2; Deon Brock, Nigel Cohen & Jonathan Sorensen, Arbitrariness in the Imposition of Death Sentences in Texas: An Analysis of Four Counties by Offense Seriousness, Race of Victim and Race of Offender 22 (November 2001) ("Across the state, and within each of the major jurisdictions . . ., the prevalence and consistency of disparities based on the race of the victim indicate a pattern of arbitrary sentencing. These findings are consistent with other studies performed in Texas and elsewhere . . . ."); David Cole, No Equal Justice: Race and Class in the American Criminal Justice System 133-50 (1999); Samuel R. Gross, Robert Mauro, Patterns of Death: An Analysis of Racial Disparities in Capital Sentencing and Homicide Victimization, 37 Stan. L. Rev. 27 (1984); Randall Kennedy, Race, Crime and the Law 328-42, 450-51 nn.50-51 (1997) (extensively canvassing the literature); Death Row USA, Summer 2001 (reporting that 80.7% of executions were white victim cases), available at http://www.death-penaltyinfo.org/DRUSA-ExecUpdate.html; Associated Press State & Local Wires, supra note 30 (discussing New Jersey state supreme court report released in August 2001 that "found that those convicted of killing white victims are far more likely to face the death penalty than those convicted of killing black victims" in New Jersey). For other research, see sources cited supra note 330.

    338. The classic demonstration of this point, extensively documented, is Kennedy, supra note 337, at 29-75, 326-50, 400-10. See also Baldus, et al., supra note 337; Blume, Eisenberg & Wells, supra note 337, at 1-2, 32-33; Bowers & Pierce, supra note 330, at 601; Richard H. Burr, III, Representing the Client on Death Row: The Politics of Advocacy, 59 U.M.K.C. L. Rev. 1 (1990); James Alan Fox & Jack Levin, The Will to Kill 167 (2001); Richard C. Dieter, The Death Penalty in Black and White: Who Lives, Who Dies, Who Decides (Death Penalty Information Center, June 1998) (Figure 7, summarizing studies); Gross & Mauro, supra note 337; Gary Kleck, Racial Discrimination in Criminal Sentencing: A Critical Evaluation of the Evidence with Additional Evidence on the Death Penalty, 46 Am. Soc. Res. Rev. 783 (1981); Raymond Paternoster, Race of Victim and Location of Crime: The Decision to Seek the Death Penalty in South Carolina, 74 J. Crim. L. & Criminol. 754 (1983).

    339. See infra pp. 213, 224-26, 253, 280.

    340. This factor is coded "wbrtst" in the detailed results in Appendix G.

    341. See infra Appendix G (reporting the mean and maximum values on this "wbrtst" factor among states and years in our study, showing that the white homicide victimization rate is usually lower than the black homicide victimization rate).

    It is not uncommon in social scientific and criminological research to discover that conditions that might be expected to have an impact on a phenomenon being studied do so only when measured comparatively, as opposed to absolutely. To give a well known example, although absolute differences in average income among different metropolitan areas and different times periods are not significantly related to differences in crime rates, relative differences in incomeCmeasured by dividing the income of the most well-off cohort of citizens by the income of the least well-offCis a strong predictor of crime rates (which tend to be higher in communities with greater income disparities than in communities where there is less income inequality). See John Kaplan, Robert Weisberg & Guyora Binder, Criminal Law 74 (4th ed. 2000) (summarizing studies). Our measure of comparative homicide rates among white and blacks is in this same tradition.

    342. p = .02 to .05 for Analyses 1A-2B.

    343. See supra pp. 157-59.

    344. These three factors are coded, respectively, "ldfwvrt," "wvldfst" and "facvic2l" in the results of Analysis 1R in Appendix G.

    345. See infra pp. 313-33.

    346. This interaction is coded as pctblack*wbrtst in the detailed results in Appendix G.

    347. The interaction was significant at the .02 level in Analysis 1A; at the .05 level in Analyses 2A and 2B; at the .06 level in Analysis 1B; and at the .03 to .04 level in Analyses 1RA, 1RB and 1RC.

    348. See supra pp. 158-62 & Table 6.

    349. See, e.g., Chisom v. Roemer, 501 U.S. 380, 385-86 (1991) (noting that no African-American has ever been elected to the Louisiana Supreme Court); HLA v. Attorney General of Texas, 501 U.S. 419, 421-22 (1991) (noting that only 5% of the trial judges in Harris County (Houston), Texas were African-American, although its population is 20% African-American); Nicholas O. Alozie, Black Representation on State Judiciaries, 69 Soc. Sci. Q. 979 (1988) (finding that key factor in under-representation of blacks on state judiciaries is their under-representation among members of state bar); Associated Press, Second Black Alabama Supreme Court Justice Sworn In, Columbus (Ga.) Ledger-Enquirer, Nov. 2, 1993, at B2 (noting that there was only one African-American among Alabama's 17 appellate court judges, and only 12 blacks among the state's 255 circuit and district court judges); Stephen B. Bright, Discrimination, Death and Denial: The Tolerance of Racial Discrimination in Infliction of the Death Penalty, 35 Santa Clara L. Rev. 433, 437-38, (1995) ("Members of racial minorities continue to be excluded as judges, jurors, prosecutors, lawyers, and law enforcement officials in the criminal justice system"); Bureau of Justice Statistics, Local Police Departments, 1997, NCJ 17329, at 3 tbl. 5 (1998) (racial makeup of nation's police forces); Law Enforcement Management and Administrative Statistics, 1997: Data for Individual State and Local Agencies with 100 or More Officers, NCJ 171681 (1998) (same) ; Mark Curriden, Racism Mars Justice in U.S. Panel Reports, Atlanta J. & Const., Aug. 11, 1991(observing that only 6 of Georgia's 134 Superior Court judges were African-American, and those six were in three judicial circuits; none of state's elected district attorneys in Georgia are black); Draft Final Report of the Special Committee on Race and Ethnicity to the D.C. Circuit's Task Force on Gender, Race, and Ethnic Bias (Jan. 1995), at 31 n.50 (noting that 3.4% of nation's practicing attorneys are African-American); Dwight L. Greene, Abusive Prosecutors: Gender, Race & Class Discretion and Prosecution, 39 Buff. L. Rev. 737, 780-81, 802 n.102, (1991) ("African-American judges constitute approximately 3.8% of all state court judges, but are disproportionately concentrated in part-time and limited jurisdiction courts [without jurisdiction in capital cases]. The percentage of African-Americans on the bench roughly mirrors the percentage of lawyers in the country who are African-Americans, approximately 3%." (citing data from Graham, Judicial Recruitment and Racial Diversity on State Courts, 74 Judicature, 28, 30, 34 (1990) (discussing Athe problem of the lack of racial diversity on the state bench")); Alex M. Johnson, Jr., The Underrepresentation of Minorities in the Legal Profession, 95 Mich. L. Rev. 1005, 1007-11 (1997) (comprehensive review of statistics); Joint Center for Political Studies, Black Judges in the United States (1986); Lewis A. Kornhauser & Richard L. Revesz, Legal Education and Entry into the Legal Profession: The Role of Race, Gender, and Educational Debt, 70 N.Y.U. L. Rev. 829 (1995); Jonathan P. Nase, Pennsylvania's Appellate Judges, 1969-1994, at 514 & Exhibit 11, 33 Duq. L. Rev. 377 (1995); Sourcebook of Criminal Justice Statistics 67-68 & tbls. 1.77, 1.78 (2000) (racial makeup of federal judiciary); Rorie Sherman, Is Mississippi Turning?, Nat'l. L. J. , Feb. 20, 1989, at 1, 24 (only 2.6 % of all state court judges in the United States are black); Jesse Smith & Robert Johns, Statistical Record of Black America 774 -75 (3d ed. 1995).

    350. See, e.g., Cole, supra note 337, at 102-05; Albert Alschuler & Andrew G. Deiss, A Brief History of the Criminal Jury in the United States, 61 U. Chi. L. Rev. 867, 894-96 (1994); Joan Biskupic, The Push Is on for More Diverse Juries, USA Today, Aug. 28, 2001 ("interviews with more than three dozen court officials, defense lawyers, prosecutors and law professors indicate that minorities are significantly under-represented in jury pools. Scattered lawsuits across the USA include statistics supporting that notion."); Tanya E. Coke, Race Neutrality and the Ideology of Representative Juries, 69 N.Y.U. L. Rev. 327, 345-46 (1994); Hiroshi Fukarai, Race and the Jury 21-24, 64 (1993); Morris B. Hoffman, Peremptory Challenges Should Be Abolished: A Trial Judge's Perspective, 64 U. Chi. L. Rev. 809, 851 & n.192 (1997); Kennedy, supra note 337, at 232-37, 435-36 n. 4 ("Even in the absence of illegal racial discrimination, traditional methods of jury selection often yield substantial >underrepresentation' of blacks . . . mean[ing] the percentage of blacks serving on juries is lower than the percentage of blacks living in the jurisdictions from which jurors are drawn."); Nancy J. King, Racial Jurymandering: Cancer or Cure? A Contemporary Review of Affirmative Action in Jury Selection, 68 N.Y.U.L. Rev. 707, 712-20 (1993) (detailing reasons why "minorities remain underrepresented on juries"); Stephanie Domitrovich, Jury Source Lists and the Community's Need to Achieve Racial Balance on the Jury, 33 Duq. L. Rev. 39, 42 (1994).

    351. See, e.g., Blume, Eisenberg & Wells, supra note 337, at 32-33; Bright, supra note 349, at 454-59 ("This >death qualification' process often results in the removal of more prospective jurors who are members of minority groups than those who are white . . . [because] minority jurors [more often] have reservations about the death penalty . . . ."); James Alan Fox et al., Death Penalty Opinion in the Post-Furman Years, 18 N.Y.U. Rev. L. & Soc. Change 499, 503 (1990-91); Samuel R. Gross, Update: America Public Opinion on the Death Penalty--It's Getting Personal, 83 Cornell L. Rev. 1448, 1451 (1998) ("Race and sex, the two major demographic predictors of death penalty attitudes, continue to be influential on every survey. On the 1996 GSS, for example, blacks were far less likely to favor the death penalty than whites (51% to 75%)."); Phoebe C. Ellsworth & Samuel R. Gross, Hardening of the Attitudes: Americans' Views on the Death Penalty, 50 J. Soc. Issues 19, 21 Summer 1994. In regard to the automatic exclusion of prospective jurors who oppose the death penalty, see, e.g., Lockhart v. McCree, 476 U.S. 162 (1986); Wainwright v. Witt, 469 U.S. 412 (1985).

    352. See sources cited supra note 351.

    353. See Albert W. Alschuler, The Surpreme Court and the Jury: Voir Dire, Peremptory Challenges, and the Review of Jury Verdicts, 56 U. Chi. L. Rev. 153, 210-11 (1989); David C. Baldus, George Woodworth, David Zuckerman, Neil Alan Weiner & Barbara Broffitt, The Use of Peremptory Challenges in Capital Murder Trials: A Legal and Empirical Analysis, 3 U. Pa. J. Const. 3 (2001) ("Our findings indicate that venire member race was a major determinant in the use of peremptories . . . ."); Bright, supra note 349, at 454-59 ("Often the >death qualification' process reduces the number of minority jurors to few enough that those remaining can be eliminated by the prosecutor with peremptory strikes. Even when jurors who express reservations about the death penalty indicate they can put aside their personal views and consider it, the prosecutor may justify his or her strikes with the hesitancy of those jurors to impose the death penalty"; documenting specific instances and systemic policies in particular prosecutor's offices of using peremptory strikes against blacks because of their tendency to vote against the death penalty); Cole, supra note 337, at 115-23; Andrew Hammell, Discrimination and Death in Dallas: A Case Study in Systematic Racial Exclusion, 3 Tex. Forum Civ. Rts. & Civ. Lib. 187 (1998); Michael Higgins, Few Are Chosen, A.B.A. J. Feb. 1999, at 50-51; Steve McGonigle & Ted Timms, Prosecutors Routinely Bar Blacks, Study Finds, Dallas Morning News, March 9, 1986; Barry Siegel, Storm Still Lingers Over Attorney's Training Video, L.A. Times, Apr. 29, 1997.

    The evidence in this and the preceding notes suggests another hypothesis: Efforts to keep black jurors from deciding capital cases may be especially common in states with large black populations, and especially likely to lead to capital reversals. Some capital cases are indeed reversed because of unlawful efforts to keep blacks off of capital juries. But the number of such reversals is too small—less than 5% of all reversals at each of the two review stages where the bases for reversal are known, see supra pp. 40-42 & n.169—to account for our regression finding. In addition, this explanation predicts that error will be more common in black-defendant and white-victim cases, because it is in those cases that prosecutors have the greatest incentive to keep black jurors from sitting. But the evidence is to the contrary: Error is no more, or less, common in black-defendant and white-victim cases. See supra pp. 158-62 & Table 6.

    354. See, e.g., Baldus, Woodworth, Zuckerman, Weiner & Broffitt, supra note 353, at 124-25; William J. Bowers, Benjamin D. Steiner & Marla Sandys, Death Sentencing in Black and White: An Empirical Analysis of the Role of Jurors' Race and Jury Racial Composition, 3 U. Pa. J. Const. L. 171, 259-60 (2001); Theodore Eisenberg, Stephen P. Garvey & Martin T. Wells, Forecasting Life and Death: Juror Race, Religion, and Attitude Toward the Death Penalty, 30 J. Legal Stud. 277, 308-09 (2001) (empirical study of capital trial outcomes concluding that "[d]eath sentences depend not only on the defendant's race, and not only on the victim's race but on the juror's race as well. All else being equal, white jurors are more apt to vote for death that black jurors . . . .").

    355. See supra pp. 158-62.

    356. See, e.g., infra pp. 166-67.

    357. See sources cited supra note 337.

    358. The social scientific literature here is voluminous and consistent in regard to the tendency on the part of white individuals, including actors in the criminal justice and trial system, to, in Professor Kennedy's phrase, "use race as a proxy for an increased risk of criminality." Kennedy, supra note 337, at 137. Extensively canvassing the empirical literature and providing case studies and examples documenting the intensity and inaccuracy of the association between African-Americans and crime, particularly violent crime, are, e.g., Kennedy, supra note 337, at 136-67, 420 n.2; Ericka L. Johnson, "A Menace to Society": The Use of Criminal Profiles and its Effects on Black Males, 38 Harv. L.J. 629 (1995); Cynthia Kwei Yung Lee, Race and Self-Defense, Toward a Normative Conception of Reasonableness, 81 Minn. L. Rev. 367, 402-33 (1996); Jon Hurwitz & Mark Peffley, Public Perceptions of Race and Crime: The Role of Stereotypes, 41 J. of Pol. Sci. 375, 380, 393-94, 399-401 (1997) (extensive review of literature and citation of sources documenting consensus based on Arecent national surveys" "that the image of blacks as a violent underclass has become a central component of contemporary white stereotypes of African-Americans," and that "one of the most popular negative beliefs expressed abou 'most' blacks is that they are 'violent and aggressive'; reporting results of authors' own empirical study finding that much thinking about crime in fact is not rooted in racial stereotypes with one exception: "Only when crimes are violent and when policies are punitive are negative stereotypes substantially more likely to see blacks as guilty of crimes, to envision more crimes in the future, and to favor harsher punishments."). See also, e.g., F.C. Dane & L.S. Wrightsman, Effects of Defendants' and Victims, Characteristics on Jurors' Verdict, in The Psychology of the Courtroom 83-115 (N.L. Kerr & R.M. Bray (eds. 1982); Birt L. Duncan, Differential Social Perception and Attribution of Intergroup Violence: Testing the Lower Limit of Stereotyping of Blacks, 4 Personality & Soc. Psychol. 590, 592-96 (1976); Randall A. Gordon, Jennifer L. Michels & Caroline L. Nelson, Majority Group Perceptions of Criminal Behavior: The Accuracy of Race-Related Crime Stereotypes, 26 J. Applied Soc. Psych. 148-59 (1988) (empirical study in which whites consistently overestimated the number of violent crimes blacks commit, particularly motor vehicle theft, rape and criminal homicide); Mark Peffley & John Hurwitz, Racial Stereotyping in Contemporary White Society: Sources and Political Consequences, in Perception and Prejudice: Race and Politics in the United States (John Hurwitz & Mark Peffley eds. 1998); M Sunnafrank & N.E. Fontes, General and Crime Related Racial Stereotypes and Influence of Juridic Decisions, 17 Cornell J. Soc. Rel. 1 (1983).

    359. The highest proportion of a state's population that was African-American in any state and year in our study is 36%.

    360. Strong empirical support for this proposition is found in existing social scientific research. See Teven E. Barkan & Steven F. Cohn, Racial Prejudice and Support for the Death Penalty by Whites, 31 J. Res. Crime & Delinq. 202-09 (1994) (reviewing study of interviews in connection with general nationwide survey of 1150 white adults indicating that white support for the death penalty is associated with antipathy to blacks and with racial stereotyping and discussing adverse impact these racial attitudes may have on legislative policy making in regard to the death penalty); Hurwitz & Peffley, supra note 358, at 393-94 (finding link between racial stereotypes associating blacks with past and assumed future violent crimes and support for harsher punishments).

    361. See supra p. 125.

    362. See John Blume & Theodore Eisenberg, Judicial Politics, Death Penalty Appeals and Case Selection: An Empirical Study, 72 S. Cal. L. Rev. 465, 469 (1999).

    363. In contrast to our data, which we gathered ourselves from official court reports, the BJS information is provided to the Justice Department by state prison administrators based on their counts of people admitted to death row in state prisons. The BJS information is less complete than ours, as a result of (1) BJS's unconfirmed, word-of-mouth method of collecting data; (2) its failure to count death-sentenced individuals who were not on death row on a reporting date, either because they remained in local jails or entered and left death row between reporting dates; (3) its focus on the number of individuals on death row, not the number of capital verdicts they represent, so that an individual sentenced to die by different juries at different trials for different offenses, or reconvicted and/or resentenced to death at a new trial by a different jury after the first verdict was reversed only counts as one event; (4) inaccuracies in its scheme for dating death sentences, which sometimes reports death sentences as having occurred in a later year than the person reached death row; (5) BJS's failure to distinguish wholesale reversals due to invalid capital statutes and reversals due to erroneous applications of presumptively valid statues (see supra p. 18); and most importantly for present purposes (6) BJS's failure to collect any information reflecting the stage of review (state direct appeal, state post-conviction or federal habeas) at which court reversals occurred.

    364. This factor is coded "dor" in the results collected in Appendix G—following Blume and Eisenberg's "death obtaining rate" terminology.

    365. In any event, this factor was highly significant, even when we controlled for the number of death verdicts imposed and awaiting review, see supra pp. 153-54, and for the size of the population, see infra p. 170.

    366. See supra pp. 165-66.

    367. This factor is coded "pnindx" in the detailed results in Appendix G. FBI index crimes are murder, forcible rape, aggravated assault, robbery burglary, larceny and auto theft.

    368. See e.g., Erin Emery, Mystery Remains in Deputy's Shooting, Denver Post, Dec. 27, 2001 (noting drive to recall Arapahoe County District Attorney Ed Rogers for failing to seek death penalty in fatal shooting of a county sheriff's deputy); 100 Colum. L. Rev., supra note 153 at 2078-82 & nn.138-40 (extensively documenting political pressures that can be, and have been successfully, brought to bear on prosecutors to use the death penalty).

    369. See Angela J. Davis, The American Prosecutor: Independence, Power, and the Threat of Tyranny, 86 Iowa L. Rev. 393, 450-51 (2001) (discussing widespread use of elections to select district attorneys).

    370. Analyses 1A, 2B, and 1RA-C use the first index of political pressure on judges (coded "ppindx" in the detailed results in Appendix G). Analyses 1B and 2A use the second index (coded "ppindx2"). Effect sizes are slightly greater for the second than for the first index, as the "newestimates" in Analyses 1A, 1B, 2A and 2B reveal. From the standpoint of statistical significance, it usually did not matter which index we used in a particular analysis; where one was statistically significant, both usually were. The two indexes did, however, modestly affect the "fit" and amount of unexplained variance. It was on those bases, therefore, that we identified two analyses that included the first political pressure index (Analyses 1A and 2B), and two other analyses with included the second index (Analyses 1B and 2A), as the preferred Analysis 1 and Analysis 2 models.

    371. See, e.g., Steven F. Messner, Luc Anselin, Robert D. Baller, Darnell F. Hawkins, Glenn Deane & Stewart E. Tolnay, The Spatial Patterning of County Homicide Rates: An Application of Exploratory Spatial Data Analysis, 15 J. Quant. Criminol. 423 (1999); Robert D. Baller, Luc Anselin, Steven F. Messner, Glenn Deane, and Darnell F Hawkins (2001); Structural Covariates of U.S. Country Homicide Rates: Incorporating Spatial Effects, Criminology 39 (3): 561-590. We are grateful to Professor Messner and his colleagues for sharing their county homicide and related data with us. Our state version of this factor is coded "psst" in the detailed results in Appendix G.

    372. Analysis 1A (p = .004); Analysis 1B (p = .002); Analysis 2A (p = .008); Analysis 2B (p = .03); Analysis 1RA (p = .002); Analysis 1RB (p = .0003); Analysis 1RC (p = .0002).

    373. See supra pp. 152-55.

    374. Interaction effects are calculated by multiplying the two component factorsCi.e., here, our measure of backlogged capital verdicts, coded "bltot" (described supra pp. 153-54 & n.318) and our four-factor measure of caseloads of all sorts, coded "fac_csld" (described supra pp. 154-55 & n.319). This interaction effect is coded "bltot*fac_clsd" in the detailed results in Appendix G. In Analysis 2B, we used the alternative, one-factor measure of general court congestion (coded rgrtbklg), see supra p. 154 & n.320, and accordingly included that factor in the interaction effect (coded "bltot*rgrtbklg").

    375. See supra pp. 114-17 & nn.282, 285.

    376. See supra pp. 88-98.

    377. See supra pp. 20-21, 88-89 (explaining why true error rates are reversals as a proportion of actually reviewed verdicts).

    378. See supra pp. 162-63 (discussing the interaction between high proportions of African-Americans in the state population and high rates of white compared to black homicide victimization); supra pp. 170-71 (discussing the interaction between high backlogs of capital cases awaiting review and high per capita rates of court filings). Effect size for factors that also are components of interaction effects are a combination of the main effect and the interaction effect.

    379. See supra pp. 89-90, 97, 99, 153-54, 173-74.

    380. As is indicated in Appendix F, the average number of capital cases awaiting appeal for all states and years in Analyses 1 and 2 is between 5 and 6, and the maximum is 49.

    381. These factors are discussed supra p. 154. Except in Figure 23B, the direction of the (very small) effect on reversal rates is upwardCeven though in the regressions themselves this factor was negatively related to reversal rates. The reason for this change in direction is that effect size takes into consideration not only the effect of the factor itself, but also the effect of any interaction variable that includes the factor as one of its components. As is discussed above, these measures of general court caseloads are also a component of an interaction variable in Analyses 1 and 2 that measures the combined effect of capital and noncapital backlogs. See supra pp. 170-71 & n.374. Moreover, that interaction variable was positively related to reversal rates. See id. Evidently, these court-caseload factors' upward influence on reversal rates as components of the interaction effect is stronger than their downward effect on reversal rates, when considered by themselvesC accounting for the slight upward direction of the lines in Figures 23A, 23C and 23D.

    382. See supra p. 93, Figure 10; p. 142, Table 4.

    383. See supra pp. 89-91, 99, 140-42 & Table 4, 154-56.

    384. See supra pp. 20-21, 91-93 & Figure 10, 140-42, 155.

    385. See supra pp. 89-90, 97, 99, 154-56.

    386. See supra pp. 152-53.

    387. See infra pp. 194-95, 258-59.

    388. See supra pp. 91-92.

    389. See infra pp. 194-95, 258-59.

    390. See supra p. 160 (discussing the definition of this factor).

    391. In a small number of states and years, there were no homicides of African-Americans. In those instances, the value in the denominator of our calculation of this variable (white homicide rate ? black homicide rate) is 0. Because it is inappropriate to divide a value by 0, and because there is no measurable racial disparity in homicide victimization rates in those particular states and years, we recorded a value of 0 in those instances. Because effect-size associated with those few instances is difficult to interpret, we omit them in these graphs. Doing so is also conservative, because the reversal rates associated with those values are very low and tend to exaggerate the degree of change in predicted reversal rates across the spectrum of states and years in the study. We chose to omit them to give a more conservative estimate of the size of the impact of this variable.

    392. See supra pp. 166-68.

    393. See supra pp. 20-21, 88-89, 172.

    394. See supra pp. 173-74.

    395. See supra pp. 166-67, 168-69.

    < name = "396">396. See supra p. 168-69 & n.367; infra Appendix E (defining FBI Index crimes).

    397. "When life is at hazard in a trial, it sensationalises the whole thing almost unwittingly; the effect . . . [is] very bad." Felix Frankfurter, Of Law and Men 81 (1956). See, e.g., William J. Bowers, The Pervasiveness of Arbitrariness and Discrimination under Post-Furman Capital Statutes, 74 U. Crim. L. & Criminology 1067, 1076-77 (1983); Stephen B. Bright & Patrick J. Keenan, Judges and the Politics of Death: Deciding Between the Bill of Rights and the Next Election in Capital Cases, 75 B.U. L. Rev. 759 (1995); Samuel R. Gross, Lost Lives: Miscarriages of Justice in Capital Cases, 61 Law & Contemp. Probs.,Autumn 1998, at 125; Norman Lefstein, Reform of Defense Representation in Capital Cases: The Indiana Experience and its Implications for the Nation, 29 Ind. L. Rev. 495, 511-12 (1996); Tina Roseberg, Deadliest D.A., N.Y. Times, July 16, 1995, (Magazine), at 21, 23-24; Richard Willing & Gary Fields, Geography of the Death Penalty, USA Today, Dec. 20, 1999, at 1A (quoting Florida prosecutor Harry Shorestein); Eric Zorn, Daley's Oversight of Prosecutors Didn't Do Justice to the Job, Chi. Trib., Nov. 16, 1999, at N1; Symposium, Politics and the Death Penalty: Can Rational Discourse and Due Process Survive the Perceive Political Pressure, 21 Fordham Urb. L.J. 239 (1994).

    398. See supra pp. 171-73; infra Appendix E.

    399. See supra pp. 171-73.

    400. See supra pp. 20-21, 88-89, 172, 183.

    401. See supra pp. 20-21, 88-89, 172, 183.

    402. See supra pp. 20-21, 91-93 & Figure 10, 140-42, 155, 177.

    403. See supra pp. 17-18 & nn.101, 103; infra Appendix C, pp. C-1 to C-2.

    404. See infra note 406.

    405. The fit inquiry described supra note 311 generates the following results in regard to Analyses 3A, 3B, 4A-4C:

    Analysis no. Number of explanatory factors
    added to baseline analysis
    Improvement in fit Significant at what level
    3A 10 25 <.01
    3B 7 37 <.01
    4A 9 28 <.01
    4B 9 26 <.01
    4C 7 42 <.01

    406. The table below indicates the amount of unexplained state-to-state and year-to-year variance in the Analysis 3 and 4 baseline analyses, and in the analyses of more specific explanatory factors.

    Analyses 3 and 4: Coefficient Indicating Amount of Unexplained Variance
    (on both measures, less is better; significance levels are in parentheses)

      variance from state to state variance from year to year
    Analysis 3
      Baseline Analysis
    .52 (.002) .064 (.02)
      Analysis 3A
    .32 (.01) .003 (NS)
      Analysis 3B
    .37 (.004) .007 (NS)
    Analysis 4
      Baseline Analysis
    .16 (.002) .017 (.03)
      Analysis 4A
    .087 (.01) 0 (NS)
      Analysis 4B
    .089 (.01) 0 (NS)
      Analysis 4C
    .11 (.004) .0003 (NS)

    407. In all five analyses, p < .0001.

    408. As is discussed above, one measure of effect size is the effect-size coefficient for each explanatory condition in each analysis (coded "newestimate" in our detailed results in Appendix G). See supra pp. 114-17 & nn.282, 285; pp. 171-73. In binomial regression studies, each unit increase in a given explanatory condition is expected to change the odds of reversal rates by a factor of x, with x being the amount of the effect-size estimate. In Poisson regression studies, each unit increase in a given explanatory condition is expected to change the reversal rate by a factor of x, with x being the amount of the effect-size estimate. In both types of studies, effect-size estimates less than 1 mean that reversal rates are expected to decline with each unit increase in the explanatory condition. The larger the distance between 1 and the effect-size estimate, the greater is the predicted decline in reversal rates. As the table below reveals, the predicted decrease in reversal rates for each additional death verdict awaiting review is considerably greater (because the distance between 1 and the estimate is greater) in Analyses 1 and 2 than in Analyses 3 and 4.

    Effect Size of Backlogged Death Verdicts
      Effect-size estimate
    Analyses of Three Review Stages
      Binomial Analysis 1A
    .82
      Binomial Analysis 1B
    .81
      Poisson Analysis 2A
    .89
      Poisson Analysis 2B
    .90
    Direct Appeal Stage Analyses
      Binomial Analysis 3A
    .92
      Binomial Analysis 3B
    .92
      Poisson Analysis 4A
    .95
      Poisson Analysis 4B
    .95
      Poisson Analysis 4C
    .96

    The same point--that effect size for capital backlogs is smaller in direct appeal Analyses 3 and 4 than in Analyses 1 and 2 of all three review stages—is made by comparing the effect-size graphs for this factor in Analyses 3 and 4, see Figures 31A and 31B, p. 201 below, to effect-size graphs for this factor in Analyses 1 and 2, see Figures 22A-22D, at p. 175 above.

    409. See supra pp. 89-91, 99, 140-42 & Table 4, 154-56, 177. Analysis 10 below is also a reliable measure of the effect of time trend.

    410. In regard to fit, see supra note 405. In regard to unexplained variance, see supra note 406. The table below compares the baseline analyses in Analyses 3 and 4 to the best analyses of specific factors, revealing not only the switch in the direction of the relationship between the passage of time and reversal rates from negative to positive, but also how much stronger the relationship is in the analyses of specific factors. As is discussed in more detail supra note 328, higher t values (irrespective of the sign), effect sizes further from 1 (irrespective of whether they are lower than 1 or higher than 1), and smaller significance values (i.e., a lower probability that the relationship appears by chance) indicate a stronger relationship.

    Relationship of the Passage of Time and Capital Error Rates
    Analyses 3 and 4
      t Value Effect Size Significance
    Analysis 3 (binomial):
      Baseline Analysis
    -3.67 .95 .0003
      Analysis 3A
    +6.42 1.18 < .0001
      Analysis 3B
    +6.10 1.17 < .0001
    Analysis 4 (Poisson):
      Baseline Analysis
    -3.62 .97 .0003
      Analysis 4A
    +6.37 1.09 <.0001
      Analysis 4B
    +6.39 1.09 <.0001
      Analysis 4C
    +6.33 1.09 <.0001

    411. For Analyses 3B and 4A-4C, .02 ? p ? .05.

    412. For Analyses 3A, p = .07.

    413. Analysis 3A, p = .04; Analysis 3B, p = .067; Analysis 4A, p = .02; Analysis 4B, p = .03; Analysis 4C, p = .057.

    414. p = .12, .07 and .09. Effect size cannot be reliably calculated for this or other interaction effects.

    415. p = .16 (Analysis 3B), .12 (Analysis 4C).

    416. Analysis 3A is the only one in which a second measure of judicial politics appears. This measure (coded "pajid") measures the "ideology" of state judges based on how they are selected and their political party. This factor was rarely significant, and its occasional significance disappeared when other factors (including our own measure of political pressure) were tested. Characteristically, the factor was not significant in Analysis 3A, but it did modestly improve fit and decrease unexplained variance. Its direction was positive, suggesting that death verdicts imposed by more conservative judges are more likely to be reversed—but, as noted, the effect was not significant.

    417. We included this factor in Analysis 3A (p = .02), Analysis 4A (p = .007) and Analysis 4B (p = .008).

    418. This factor is coded "dir_exrt" in the detailed results in Appendix G.

    419. The minimum annual direct expenditure on state courts per resident in the relevant states and years was about $.02 in Arkansas in 1974; the maximum was $5.98 in North Carolina in 1994; and the average was $2.05.

    420. During the years studied, the 34 study states spent from $.02 per resident on their courts (in Arkansas in 1974) to $5.98 (in North Carolina in 1994), with the average expenditure being $1.73.

    421. See supra pp. 17-18 & nn.101, 103; infra Appendix C, pp. C-1 to C-2 (explaining why state post-conviction reversal rates must be studied in this way). Poisson analysis is used because the rates studied in Analysis 5 are reversals as a proportion of all verdicts available for review, which include many cases that got stuck in the review process and were not finally reviewed in the study period. The measure artificially deflates reversal rates, therefore—in contrast to Analyses 3 and 4, which examine reversals as a proportion of finally reviewed verdicts and provide a more realistic picture of error rates. See supra pp. 20-21, 88-89, 172, 183, 194-95. In addition, the reversal rates we estimate for the second, state post-conviction review stage appear are lower on average than the reversals we count at the first, state direct appeal and second, federal habeas, review stage. See supra pp. 9, 17-18 & n.103.

    422. See supra note 103.

    423. See infra note 424.

    424. Fit for the baseline was 1510. For Analyses 5A and 5B, it was 1507 and 1516. Neither Analysis 5A's improvement of 3 points nor Analysis 5B's decline of 6 points is statistically significant. As for unexplained variance, see the table below:

      variance from
    state to state
    variance from
    year to year
    Analysis 5
      Baseline Analysis
    .46 (.01) .140 (.07)
      Analysis 5A
    .38 (.02) .074 (NS)
      Analysis 5B
    .36 (.03) .167 (.06)

    On effect size, see infra p. 214-15.

    425. See supra pp. 198-99.

    426. See supra pp. 89-90, 97, 99, 153-54. This factor is coded "pcbl" in the detailed results in Appendix G. Because we were unable to count state post-conviction affirmances for the reasons discussed supra pp. 17-18; infra Appendix C, pp. C-1 to C-2, our measure of backlogs in Analysis 5 (verdicts available for review minus verdicts reviewed and reversed) is composed not only of verdicts as to which review was delayed but also verdicts that were reviewed and approved because no reversible error was found. To this latter extent, the factor's significant negative relationship to reversal rates is obvious and uninteresting because the factor is essentially the same as the condition being studiedCwhether verdicts were reversed or not reversed.

    427. See supra pp. pp. 89-91, 99, 140-42 & Table 4, 154-56, 177, 194-95.

    428. See supra pp. 59-61 & Figures 3A and 3B.

    429. This factor is coded "wvrtst" in the detailed results in Appendix G.

    430. p = .09.

    431. This analysis is based on the numerical effect-size estimate for this unlogged factor of 1.21. For an explanation of how effect size is calculated using these kinds of estimates, see supra pp. 114-17 & nn.282, 285.

    432. This analysis is based on the numerical effect-size estimate for this unlogged factor of 8.53, and on the fact that values for this factor in particular states and years range from .05 to .54. For an explanation of how effect size is calculated using these kinds of estimates, see supra pp. 114-17 & nn.282, 285.

    433. The numerical effect-size estimate for this logged factor in Analysis 5A is 2.45, and in Analysis 5B is 2.07. For an explanation of how effect size is calculated using these kinds of estimates, see supra pp. 114-17 & nn.282, 285.

    434. This kind of analysis is suited to distributions of rates to be explained that include many at the high as well as the low end of the range from 0 to 100.

    435. The baseline fit was 628Cthe same for Analysis 6A, and 2 points less than (but, statistically speaking, the same as) for Analysis 6B.

    436. Analysis 6: Coefficient Indicating Amount of Unexplained Variance

      variance from state to state variance from year to year
    Analysis 6
      Baseline Analysis
    .49 (.04) .08 (NS)
      Analysis 6A
    0 (NS) 0(NS)
      Analysis 6B
    0(NS) 0(NS)

    On effect size, see infra Figures 40A, 40B, 40C-1, 40C-2 and 40D, pp. 220-21, below.

    437. For discussion of that non-error-related downward effect of the passage of time on reversal rates in Analyses 1, 2 and 5, see supra pp. 88-91, 99, 140-42, 152-53, 154-56, 177, 194-95, 213.

    438. See supra pp. 91-93, 140-42, 155, 177, 193. As of the end of the study period, the average time from imposition to final federal habeas decision was just over 11 years for affirmed death verdicts and 12.5 or more years for reversed death verdicts. Despite this difference, nearly all death verdicts imposed early in the 23-year study period—whether flawed or not—were finally reviewed on federal habeas by the end of that study period (assuming the verdicts were not reversed at an earlier review stage). But as death verdicts became more recent, more of them were likely to have been in federal habeas proceedings towards the end of the study period. Among those verdicts, the shorter period needed to finally review verdicts without discoverable flaws, and the longer period needed to review verdicts with those flaws, meant that a disproportionately large number of unflawed verdicts imposed relatively late in the study period were finally reviewed and counted in our study, while a disproportionately large number of flawed verdicts imposed relatively late in the period were not finally reviewed by the end of the study period or, thus, counted in the study. This systematic undercounting of flawed cases imposed late in the study period would lead the passage of time to be associated with decreasing reversal rates because of discrepancy in the time needed to review flawed and unflawed verdicts and not because fewer flawed verdicts were imposed in later years.

    439. p = .0002 (Analysis 6A); p = .0007 (Analysis 6B).

    440. For the size of this at least partly non-error-related effect, see Figure 40A (Analysis 6A), p. 220 below.

    441. Federal habeas is the only stage at which this factor was important.

    442. See supra pp. 16, 37-39, 65 n.208.

    443. See supra pp. 198-99, 212.

    444. See supra pp. 169-70 & n.370 (discussing the two political pressure indexes).

    445. Effect size can also be compared based on their numeric effect-size estimates (coded "newestimates" in Appendix G). For a discussion of how to interpret these estimates, see supra pp. 114-17 & nn.282, 285. Effect size is compared by comparing the distance between the estimate and 1. The greater that distance, the greater the effect size. For the political pressure indexes in Analyses 1, 3 and 6, that comparison is as follows:

    Effect Size for Political Pressue Indexes:
    Binomial Analyses 1,3 and 6

      Effect-size estimate
    Binomial Analysis 1 (3 review stages combined)  
      Analysis 1A (first political pressure index)
    1.25
      Analysis 1B (second political pressure index)
    1.30
    Binomial Analysis 3 (direct appeal stage)  
      Analysis 3A (second political pressure index)
    1.19
      Analysis 3B (second political pressure index)
    1.15
    Binomial Analysis 6 (federal habeas stage)
      Analysis 6A (first political pressure index)
    1.44
    Analysis 6B (second political pressure index) 1.63

    446. p = .01 (Analysis 6A); p = .0009 (Analysis 6B).

    447. See supra pp. 37-38, 65 & n.209.

    448. See supra pp. 194.

    449. The numbering of analyses jumps from 6 to 14 because we conducted all our comparisons of the states' aggregate 23-year experiences last (Analyses 14-17), after first completing the state Analyses 1-6 and county-state Analyses 7-13 in which state and time were given more conventional treatment.

    450. Analyses 14 and 15: Fit, and Coefficient Indicating Amount of Unexplained Variance
    (on both measures, less is better)

      Fit
    (amound and signficance of improvement)
    Coefficient indicating amount of unexplained state-to-state variance
    (siginificance level in parentheses)
    Analysis 14
      Baseline Analysis
    1952 .40 (.002)
      Analysis 14A
    1806 (-146, HS) .12 (.03)
    Analysis 15
      Baseline Analysis
    1409 .15 (.002)
      Analysis 15A
    1111 (-298, HS) .026 (.04)
      Analysis 15B
    1113 (-296, HS) .027 (.03)

    451. p < .0001.

    452. See supra pp. 89-90, 97, 99, 152-53, 173-74, 212.

    453. See supra pp. 174, 176.

    454. p < .0001 (Analyses 14A, 15A), .002 (Analysis 14B), .02 (Analysis 15B). See infra p. 224-25 (linking the higher significance levels in Analyses 14A and 15A, compared to Analyses 14B and 15B, to the inclusion of homicide rates as a factor in the latter two analyses but not the former two.

    455. See supra note 454.

    456. This factor is coded "hrst" in the detailed results in Appendix G. In Analysis 14B, p = .11; in Analysis 15B, p = .002.

    457. Because there is no homicide rate factor in Analysis 15A, there is no Figure 41E-1.

    458. Because homicide rates vary by year with some consistency across states, Analyses 14 and 15's removal of the assumption, programmed into all earlier analyses, that reversal rates for given years are more like each other than ones for other years, may also account for the importance of homicide rates in these, but not prior, analyses.

    459. R = .56 (p < .0001).

    460. See supra note 454.

    461. See supra pp. 165-66.

    462. p = .08 (Analysis 14A), .095 (Analysis 14B).

    463. See supra pp. 224-25.

    464. As with percent black, this factor and homicide rates are correlated. R = .34 (p < .0001). States with high homicide rates tend to be ones in which the homicide risk to whites approaches or surpasses that to blacks.

    465. p = .10 (Analysis 14A), .11 (Analysis 14B), .02 (Analysis 15A), .06 (Analysis 15B).

    466. See supra p. 167-68.

    467. Cf. supra pp. 198-99, 212, 217-18, 227.

    468. See supra pp. 224-25 & n.458.

    469. This finding does not suggest that time be removed from all of our analyses. As we note above, time is included in many analyses to facilitate simultaneous comparisons of states that began imposing the death penalty under constitutional statutes at different times, starting in 1973. See supra pp. 87-88.

    470. See supra p. 146-91.

    471. Here, we look to see if one or more "best" sets of explanatory factors within each Analysis 1-6, 14 and 15 satisfied the relevant diagnostic criterion.

    472. Unless noted, the baseline analysis left both state-to-state and the year-to-year differences to be explained, and the differences were significant at the .05 level or better. Analyses 14 and 15 only examine state-to-state differences.

    473. Year to year variance was significant at the .07 level.

    474. State-to-state variance was significant, but year-to-year variance was not.

    475. Percent of Significant Factors* in Each Analysis That Were Significant in, and

    Have Same Relationship to Reversal Rates as, Analyses of One or Two Other Review Stages

      One Other
    Review Stage
    Two Other
    Review Stages
    Analysis 1 89% 67%
    Analysis 2 89% 67%
    Analysis 3 78% or 89%+ 56% or 63%+
    Analysis 4 78% or 89%+ 56% or 63%+
    Analysis 5 100% 100%
    Analysis 6 50% 50%
    Analysis 14 89% 67%
    Analysis 15 89% 67%

    * For purpose of this diagnostic test, we treat different measures of capital-cases backlogs applicable to different stages of review as the same factor. We exclude interaction effects from this analysis given the difficulty encountered in interpreting them.

    + For reasons discussed supra notes 88-91, 99, 140-41, 149-55, 175, 190, 201. Analyses 3 and 4 are the only ones that are completely reliable measures of the effect of time on error rates, and the only ones in which the passage of time is not only significantly but positively associated with reversal rates. The fact that no other analysis reveals that same relationship arguably should not be held against Analyses 3 and 4. If that factor is omitted, the two analyses' rate of congruence with analyses of one or more other review stages rises from 78% to 89%, and their rate of congruence with analyses of two or more other review stages rises from 56% to 63%.

    476. See supra pp. 102-03, 222 (distinguishing analyses that divide all states and years into two different gridsCone for all years in the same state, and all states in the same yearCand those that divide states and years only into one grid comparing states in different years).

    The two percentages presented for Analysis 5

    477. The comparable percent for Analysis 6 is 33%. For purpose of this diagnostic test, we treat different measures of capital-cases backlogs applicable to different stages of review as the same factor. We exclude interaction effects from this analysis. The 80% score for Analysis 5 rises to 100% if one treats the "white homicide victimization rate" factor there as sufficiently analogous to the "white-vs.-black homicide victimization rate" factor in Analyses 14 and 15 to qualify as a match.

    478. During the study period, there were 2349 reversals, 1852 (79%) of which occurred at the direct appeal stage.

    479. See supra pp. 88-91, 99, 140-42, 152-53, 154-56, 177, 194-95, 213.

    480. More particularly, the greater the reviewing judges' insulation from local political pressures, the stronger the positive relationship between political pressure on state judges and capital error rates. See supra pp. 198-99, 212, 217-18, 227.

    481. See supra pp. 218-19.

    482. Notes to Table 7

    † All significant factors are positively associated with reversal rates (reversal rates tend to increase when the amount or intensity of the explanatory factor increases), unless the factor is preceded by a negative sign ( in which case the factor is negatively associated with reversal rates (reversal rates tend to decrease when the amount or intensity of the condition increases).

    # Analysis 1R is treated as a variant of Analysis 1, rather than as a separate analysis. Explanatory factors

    -pctbdiffnew, -dswvrt, and wvdsst only appear in Analysis 1R. Because Analyses 14 and 15 explain variation in reversal rates measured as a composite of each state's 23-year experience, without regard to particular years in the study period, see supra pp. 102-03, 222, time is excluded as a possible explanatory value in these analyses. As a result, time could only be tested in six of the eight analyzes examined in this table.

    & The relative homicide rates for whites and blacks just failed to reach significance in Analysis 14 (p = .08); the interaction between that factor and the proportion the population that is black also just failed to reach significance in Analyses 4 and 14 (p = .08); and the indexes of political pressure were just above significance in Analyses 3 and 4 (p =.07, .06).

    ** Explanatory factors that were significant in analyses of all three review stages in which all "state-years" were grouped for analysis both by state and by year (Analyses 1and 2), and in analyses of all three review stages in which "state-years" were grouped only by states (Analyses 14 and 15).

    ^ In Analyses 1 and 2, the relationship of the passage of time to reversal rates is an unreliable indication of the changing quality of death verdicts over time. This is because reversal rates in those analyses are the number of death verdicts imposed in a given year that were reversed, as a proportion of all death verdicts imposed that year, whether or not they were reviewed. Because it takes 2 to 5 years, on average, for final review to occur at even just the first (direct appeal) stage of review, and many more years for later stages of review to occur, there are almost no death verdicts imposed in 1994 and 1995 that secured any review as of the end of the study period in 1995, and only small proportions of death verdicts imposed in 1991, 1992 and 1993 that were reviewed even at the first stage as of 1995. More generally, the later a death verdict was imposed, the less likely it is that the verdict was finally reviewed at any review stage, and especially at all three review stages, by the analysis's cut-off date of 1995. Death verdicts imposed in later years thus are less likely than those imposed in earlier years to have been reviewed as of the end of the study period, imposing a downward trend on death verdicts over time. That decrease, however, is a function of the successively smaller number of death verdicts imposed in later years that were reviewed, and does not indicate that later verdicts were less likely to be flawed. As a result, any decline in reversal rates over time is at least in part, and may be entirely, a function of the number of unfinished appeals, rather than a function of changes in the amount of serious error. See supra pp. 88-91, 99, 140-42 & Table 4, 152-53, 154-56, 177, 194-95, 213, 216-17. Likewise, in Analysis 5, the measured effect of the passage of time is unreliable because reversal rates there are calculated as the number of state post-conviction reversals of death verdicts imposed in a given year as a proportion of the number of verdicts imposed in that year that were available for state post-conviction review, whether or not they were actually reviewed. The result is the same as is described in the previous paragraph: The rate of reversals as a proportion of all verdicts available for review, whether or not they actually were reviewed, is subject to a downward trend that is a consequence of the time needed to complete state post-conviction review and is not an indication of less error over time.

    Analyses 3, 4 and 6 analyze reversals as a proportion of the number of reviewed death verdicts. In these analyses, therefore, low rates of review do not automatically lead to low rates of reversal irrespective of the quality or error-proneness of death verdicts. These analyses thus are more reliable measures of the effect of the passage of time on rates of serious error than Analyses 1, 2 and 5. Analysis 6 has a different source of unreliability, however. That analysis examines the federal habeas stage of review, where reversals take one to two years longer on average than affirmances to be ordered. As a result, the pool of death verdicts imposed during the study period that were not finally reviewed on habeas during that period includes a disproportionately large number of the flawed verdicts imposed in a given year, while the pool of verdicts that were reviewed and whose review results were counted in our study includes a disproportionately large number of the unflawed verdicts imposed in any such year. See supra pp. 20-21, 91-93 & Figure 10, 140-42, 155, 177, 193, 216-17 & n.438.

    ‡ This factor had low effect size.

     

    483. For this purpose, the stages considered were state direct appeal, state post-conviction, federal habeas, and the three combined.

    484. In six of the analyses, the measure of capital backlogs was the number of capital verdicts awaiting review at all three stages; in our state post-conviction analysis (Analysis 5) the measure was the number of capital verdicts awaiting review at the state post-conviction stage.

    485. In Analysis 5, the homicide rate among whites, by itself, was significant. In that analysis, the white homicide victimization rate relative to the black homicide victimization rate fell just above the .05 significance level.

    486. In this one instance, we include a factor where the three-fourths rule is satisfied only by considering results that were significant at above the .05 level. We do so because this factor, which was significant at less than the .05 level in 5 analyses, fell just barely above significance in 2 additional analyses, meaning the factor is not qualitatively different from our very strongest explanations, which were significant in 7 of 8 analyses. In addition, our interpretation of this factor is based on its having only a moderate or intermediate effect at the state direct appeal stage—the stage covered by the two analyses in which the factor fell just above the .05 levelCwhich assures that we are interpreting the factor based on its actual, intermediate outcome in those analyses, rather than ascribing more than they show. See supra pp. 198-99, 212, 217-18, 227, 236; infra p. 336.

    487. In a fourth and a fifth analysis (Analyses 4, 14), this factor fell just above the .05 significance level.

    488. See supra pp. 224-26.

    489. See the effect-size estimates (i.e., "newestimates") in the detailed results in Appendix G, and by an array of effect-size graphs displayed above.

    490. See supra pp. 198-99, 212, 217-18, 227, 236.

    491. See supra pp. 218-19, 236.

    492. Only six of the eight analyses undertook to measure the effect of the passage of time. The other 2, by design, rendered the passage of time irrelevant by averaging each state's reversal rates and explanatory conditions over the entire study period. Those six analyses all used the approach to weighting the states.

    493. See supra pp. 88-91, 99, 140-42 & Table 4, 152-53, 154-56, 177, 194-95, 213, 216-17, 235-36.

    494. See supra p. 235-36 & n.478.

    495. See supra pp. 52-58 & Figures 2A-2D.

    496. See supra pp. 194-96.

    497. Of the 2349 total reversals, 240 were at the federal habeas stage.

    498. See supra pp. 20-21, 91-93 & Figure 10, 140-42, 155, 177, 193, 216-17 & n.438, 439.

    499. See supra pp. 65, 67 & Figure 5.

    500. See, e.g., Supreme Court decisions cited supra notes 122, 125.

    501. See supra pp. 133-34.

    502. See supra p. 133.

    503. To help readers account for this point, we indicate the number of counties in each state, at the top of each state's bar in Figure 42A.

    504. Figures 42A and 42B are based on 1004 capital counties in 34 states. Our regression analyses study 1002 counties. The difference arises because the best available county data on factors such as homicide rates, described supra p. 170 & n.371, aggregates a two small independent cities into adjoining counties, forcing us to do the same in order to make use of the data.

    505. This point is developed further at pp. 268-69, 272-82 below.

    506. 3054 / 967 = 3.16.

    507. Our earlier analyses leave little reason to think that binomial and Poisson analyses generate different outcomes in important respects. Because Analysis 7 is merely exploratory7#151;aiming to maximize the number of factors to consider in later analyses, rather than to exclude all but the strongest effects—we concluded that a single regression technique was sufficient. We used Poisson analysis because its coefficients are easier to interpret.

    508. Some of the county-level data analyzed in our county studies were supplied by Professor Messner and his colleagues, for which we are grateful. See supra p. 170 & n.371; supra note 504.

    509. For most counties, homicide is a rare event that does not occur in most years. In order to account for this phenomenon, we calculated death-sentencing rates somewhat differently for counties than for states, by averaging the number of homicides in the year each death verdict was imposed and in the preceding year. In the event that there were no homicides in the death-sentencing and preceding year, we averaged the number of homicides in the death-sentencing year and the two prior years.

    510. In the detailed results in Appendix G, these factors are coded, respectively, as "year-n," "lcntydor2," "pctblk," "hr2," "wbrt" and "wvrt,""ps," and "fp."

    511. As noted above, see supra pp. 89-90, 97, 99, 153-54, 173-74, 212, 223, backlogs of capital cases awaiting review automatically lower rates of court review. That, in turn, automatically lowers reversal rates where they are calculated as the proportion of imposed death verdicts that are reversed, not the proportion of reviewed death verdicts that are reversed. This factor is coded "bltot."

    512. Detailed results for the county (and all other) analyses are reported in Appendix G.

    513. Analysis 7: Fit, and Coefficient Indicating Amount of Unexplained Variance (on both measures, less is better; significance levels are in parenthesis)

          Fit Coefficient indicating amount of unexplained variance
       County-to-county variance Year-to-year variance
    Analysis 7
      Baseline Analysis
    9373 .17 (.HS) .003 (HS)
      Analysis7A
    9434 .11 (.0007) .001 (.0003)

    514. See supra pp. 88-91, 99, 140-42 & Table 4, 152-53, 154-56, 177, 194-95, 213, 216-17, 235-36.

    515. See supra pp. 89-90, 97, 99, 153-54, 173-74, 212, 223.

    516. See supra pp. 224-26.

    517. See infra pp. 266, 285, 286-87, 336, 363.

    518. See supra pp. 250-51.

    519. See supra pp. 88-91, 99, 140-42, 152-53, 154-56, 177, 194-95, 213, 216-17, 235-36.

    520. Analyses 8-10: Fit, and Coefficient Indicating Amount of Unexplained Variance
    on both measures, less is better; significance levels are in parentheses)

      Fit Coefficient for random intercept indicating amount of unexplained variance
    Analysis 8
      Baseline Analysis
    13045 .53 (HS)
      Analysis 8A
    13257 0 (NS)
    Analysis 9
      Baseline Analysis
    9331 .14 (HS)
      Analysis 9A
    9497 0 (NS)
    Analysis 10
      Baseline Analysis
    10129 .24 (HS)
      Analysis 10A
    10235 0 (NS)

    521. In all three analyses, p < .0001.

    522. See supra p. 194.

    523. p < .0001.

    524. For the difference between the two indexes, see supra pp. 169-70 & n.370.

    525. p < .0001 (Analyses 8A, 10A); p = .0017 (Analysis 9A).

    526. p = .32 (Analysis 8A); p= .0009 (Analysis 9A); p < .0001 (Analysis 10A). On the analogous county-level factor, see infra p. 266 & n.533.

    527. The latter result is anomalous, given the tendency in other analyses of homicide rates to be either non-significant or positively associated with reversal rates. See supra pp. 224-26, 253; infra p. 280.

    528. See supra pp. 88-91, 99, 140-42 & Table 4, 152-53, 154-56, 177, 194-95, 213, 216-17, 235-36, 255-56.

    529. p = .06 (Analysis 8A), .38 (Analysis 9A).

    530. As is noted above, see pp. 252 & n.511, Analysis 7 used a county-level factorCthe number of verdicts the county imposed in each year that had not been fully reviewed by the end of the study periodCas a control for the effect of delay. In county-state Analyses 8-10, that same purpose was served by including the same factor as a state-level predictor, which is more appropriate because the factor measures the effect of activities in state supreme courts and other courts with statewide, as opposed to county-wide, jurisdiction.

    531. See supra pp. 258-59, 264 & Figures 43P, 43Q.

    532. p =.003 (Analysis 8A), .04 (Analysis 9A), .009 (Analysis 10A).

    533. p = .008 (Analysis 8A); p < .0001 (Analysis 9A); p = .07 (Analysis 10A). In regard to the analogous state-level factor in Analysis 8, see supra p. 258 & n.526.

    534. See supra pp. 224-26. These two racial conditions were significantly related to differences in county reversal rates when measured at the state level but not when measured at the county level.

    535. See supra pp. 246-50.

    536. The coefficients for unexplained state-level variance and their significance levels in Analyses 11 and 12 are derived from the baseline inquiries for Analyses 1 and 2. Those coefficients are displayed both supra p. 151, Table 5, and infra note 540. The nonrandom variance in state reversal rates that is left to be explained by the baseline analyses of state, year and time trend is significant at the .0006 level in both Analyses 11 and 12.

    537. See infra note 540.

    538. For evidence that this is the case, see supra pp. 246-50; infra pp. 272-78, 279-82, 286-87.

    539. See supra pp. 246-50.

    540. Analyses 11 and 12: Fit, and Coefficient Indicating Amount of Unexplained Variance
    (on both measures, less is better; significance levels are in parentheses)

          Fit Coefficient indicating amount of unexplained variance
       State-to-state or
    County-to-county variance
    Year-to-year variance
    Analysis 11
      State Level Analysis:
      Baseline Analysis
    1813 .52 (.0006)) .29 (.01)
      Analysis11A
    1714 .15 (.01) .15 (.03)
      County Level Analysis:
      Baseline Analysis
    13452 .072 (.015)) N/A
      Analysis11A
    13745 .064 (.021) N/A
    Analysis 12
      State Level Analysis:
      Baseline Analysis
    1260 .14 (.0006)) .30 (.009)
      Analysis12A
    1109 .03 (.02) .22 (.01)
      County Level Analysis:
      Baseline Analysis
    9711 .007 (.09)) N/A
      Analysis12A
    9790 .004 (NS,.17) N/A

    541. See supra pp. 241-45.

    542. For evidence that this is the case, see supra pp. 246-50; infra pp. 279-82, 286-87.

    543.Effect size is indicated by the "newestimate" coefficients in the detailed results in Appendix G. The interpretation of those estimates is discussed supra pp. 113-15 & nn.282, 285.

    544. It does include a random intercept to measure the amount of unexplained variance.

    545. Analysis 13: Fit, and Coefficient Indicating Amount of Unexplained Variance
    on both measures, less is better; significance levels are in parentheses)

      Fit Coefficient for random intercept indicating
    amount of unexplained
    county-to-county variance
    Analysis 13
      Baseline Analysis
    9353 .12 (HS)
      Analysis 13A
    9449 .03 (.002)
      Analysis 13B
    9575 .01 (.05)

    546. p = .52 (Analysis 13A), p = .25 (Analysis 13B).

    547. See supra pp. 88-91, 99, 140-42 & Table 4, 152-53, 154-56, 177, 194-95, 213, 216-17, 235-36, 255-56.

    548. See supra pp. 258-59. In fact, although non-significant, the relationship between reversal rates and the passage of time in Analyses 13A and 13B is positive, meaning that some upward force on reversal rates over time (after accounting for other factors) has entirely neutralized the downward pull on reversal rates from the effect of unfinished appeals.

    549. p < .0001 (Analyses 13A and 13B).

    550. p < .0001 (Analyses 13A and 13B). Effect size is similar to that for the same explanatory factor in Poisson Analysis 2. See supra pp. 183-84 & Figures 27C, 27D.

    551. p < .0001 (Analyses 13A and 13B). Effect size for this factor is slightly larger in Analysis 13 than in Analysis 2. See supra pp. 179-80 & Figures 25C, 25D.

    552. p < .0001 (Analyses 13A and 13B). Effect size is about 15 to 25% larger for this factor in Analysis 13 than for the same factor in Analysis 2. See supra pp. 181-82 & Figures 26C, 26D.

    553. p < .0001 (Analyses 13A and 13B). Effect size is difficult to interpret for interaction effects.

    554. p < .0001 (Analyses 13A and 13B, both using the second index of political pressure). Effect size is about the same for this factor in Analysis 13 as in Analysis 2A. See supra pp. 187-88 & Figure 29C.

    555. p < .0001 (Analyses 13A and 13B). Effect size is slightly larger than that for the same factor in Analysis 2 . See supra pp. 185-86 & Figures 28C, 28D.

    556. p < .0001 (Analyses 13A and 13B). Effect size for this factor is about the same in Analysis 13 as in Analysis 2. See supra pp. 189-90 & Figures 30C, 30D.

    557. p < .0001 (Analysis 13B). Effect size remains negligible in Analysis 13, as in Analysis 2A. See supra pp. 175-76 & Figure 23C.

    558. p < .0001 (Analysis 13B). Effect size is difficult to interpret for interaction effects.

    559. p = .05 (Analysis 13A), .47 (Analysis 13B).

    560. For state homicide rates (Analysis 13A), p = .86; for county homicide rates, p = .13.

    561. See supra pp. 246-50, 268-69.

    562. Analyses 16-17: Fit, and Coefficient Indicating Amount of Unexplained Variance
    (on both measures, less is better; significance levels are in parentheses)

      Fit Coefficient for random intercept indicating
    amount of unexplained
    county-to-county variance
    Analysis 16
      Baseline Analysis
    3661 .48 (.0003)
      Analysis 16A
    3651 .11 (.08)
    Analysis 17
      Baseline Analysis
    2534 .18 (.0002)
      Analysis 17A
    2529 .04 (.06)

    Analyses 16 and 17 present more potential multi-colinearity questions than other studies, see infra Appendix F, and for that reason would not be a basis for any firm conclusions by themselves. The fact that their results are very similar to those of the other analyses, but that effect size is consistently smaller than in other analyses, suggests that multi-colinearity is deflating the estimates (dampening the clarity of the relationships) in these two studies.

    563. p < .0001 (Analyses 16A, 17A). General court backlogs were tested as an explanatory factor in both analyses but were not significant.

    564. p = .02 (Analysis 16A); .04 (Analysis 17A).

    565. p = .09 (Analysis 16A), .07 (Analysis 17A).

    566. p = .01 (Analysis 16A), .006 (Analysis 17A).

    567. p = .0002 (Analysis 16A); p < .0001 (Analysis 17A). The interaction of this factor and the preceding, "percent black" factor was not significant in this model and was excluded from our best analyses.

    568. p = .0002 (Analysis 16A), .0008 (Analysis 17A). Both analyses use the second political pressure index.

    569. p = .02 (Analysis 16A), .008 (Analysis 17A).

    570. p = .01 (Analysis 16A), .02 (Analysis 17A).

     

    571. See Furman v. Georgia, 408 U.S. 238 (1972).

    572. Florida, Georgia and Texas as Proportion of 34 Study State Totals

      FL GA TX 3 States 34-State Tot. % of Tot.
    No. of counties with greater than or equal to 1 death verdict during study period 53 88 105 246 1002 25%
    No. of death verdicts during study period 889 339 734 1962 5826 34%
    No. of death verdicts reversed during study period 480 199 218 897 2349 38%
    No. of death verdicts finally reviewed during study period 539 233 321 1093 2707 40%

    Florida and Texas imposed more death verdicts than any other state during the study period. California was third (with 533), and Georgia was fourth. Because California got started several years after the other three states, however, and its verdicts experience such extended delays in the review process, it has many fewer court decisions than the other three states.

    573. p = .08 (binomial), .08 (Poisson).

    574. The situation remained the same when we attempted to maximize variance by dropping year as a random effect, removing time trend from the baseline analysis, and switching county from a random effect to a subject variable.

    575. Analysis 18, Florida: Fit, and Coefficient Indicating Amount of Unexplained Variance
    (on both measures, less is better; significance levels are in parentheses)

      Fit Coefficient indicating amount unexplained
    county-to-county variance
    Analysis 18: Florida, Poisson
      County and Year = Random Effects:
      Baseline Analysis
    849 .016 (.08)
      Analysis 18A
    809 .001 (NS)
      Analysis 18B
    822 .001 (NS)

    576. See supra pp. 88-91, 99, 140-42, 152-53, 154-56, 177, 194-95, 213, 216-17, 235-36, 255-56. For this factor, p = .0002 (Analysis 18A), .001 (Analysis 18B).

    577. Effect-size estimates are in the detailed results in Appendix G. See supra pp. 114-17 & nn.282, 285 (discussing the calculation and interpretation of effect-size estimates).

    578. See supra pp. 89-90, 97, 99, 152-53, 173-74, 212, 223. For this factor, p < .0001.

    579. p = .0006 (Analysis 18A), <.0001 (Analysis 18B).

    580. p < .0001 for both analyses.

    581. p =.004 (Analysis 18B). The information in note 575 above reveals a significant drop in fit as a result of including this factorCthe only factor in Analysis 18B that is not in Analysis 18A.

    582. See supra pp. 224-26, 266.

    583. See, e.g., supra pp. 268-69, 272-78, 279-80.

    584. See supra pp. 246-50.

    585. See supra pp. 119-20.

    586. Notes to Table 8

    * Although there are 10 county analyses, only 8CAnalyses 8-13, 16 and 17Cinclude both county and state factors. Analyses 7 and 18 include only county factors. Those studies are surveyed in Table 9, p. 285 below.

    † Analyses 8-10, 13 explain variation among counties treated as entities nested within particular states (meaning the statistical analysis is programmed to assume that there are more similarities among counties within the same state than among counties from different states). Explanatory factors that were significant in these analyses are indicated by "cnty-in-st." Analyses 11 and 12 use the entire set of state explanatory factors from our best state-only Analyses 1 and 2 to identify predicted reversal rates for all counties in each state, then test those predicted rates to see how closely associated they are with counties' actual reversal rates. Explanatory factors that were significant in these analyses are indicated by "predval." Analyses 16 and 17 explain variation among states, where each state is treated in effect as the composite of all of its counties, which in turn are treated as their aggregate values across all study years in which they had at least one death verdict. Explanatory factors that were significant in this study are indicated by "st'comp/cntys." All but one of the analyses collected here explain variation in aggregate reversal rates at all three stages of review; the one exception is Analysis 10, which explains variation in reversal rates at only the direct appeal stage. Explanatory factors that were significant in that single-stage analysis are indicated by "da."

    ^ In these analyses, the linear trend for time is unreliable because the base number (the denominator or offset) for reversals is the raw number of death verdicts imposed in each relevant year, not the number of verdicts imposed in that year that were actually reviewed. Because it takes two to five years, on average, for final review to occur at even just the first (direct appeal) stage of review (and even more years for later stages of review to occur), there will be almost no death verdict imposed in 1994 and 1995 that secured any review as of the end of the study in 1995, and only very small proportions of death verdicts imposed in 1991, 1992 and 1993 will have been reviewed as of 1995. See supra p. 142 & Table 4. For this reason, there will be a low number of reversals, and a low percentage of death verdicts imposed in those later years that were reversedCmaking it appear that the linear trend was toward lower numbers and rates of reversals in the later years and, so, over time. That lower number and rate of reversals for death verdicts imposed in the later years is, however, an artifact of the smaller number of death verdicts imposed in those years that were reviewed, rather than an indication that fewer of the verdicts that were reviewed were reversed. Because Analyses 16 and 17 explain variation in reversal rates that are aggregates of counties' 23-year capital experiences, time is excluded as a possible explanatory value. The passage of time thus could be tested in only six of the eight analyses analyzed in this table.

    + See supra note 527.

    588. See supra pp. 224-26, 266.

    589. Notes to Table 9

    * Analyses 8-13, and 16-18.

    † Analysis 7 studies variation among counties, by year, as entities independent of states. Explanatory factors that were significant in this analysis are indicated by "cnty." Analyses 8-10 and 13 explain variation among counties treated as entities nested within particular states (meaning the statistical analysis is programmed to expect more similarities among counties within the same state than among counties from different states). Explanatory factors that were significant in these analyses are indicated by "cnty-in-st." Analyses 11 and 12 use the entire set of state explanatory factors from our best state-only Analyses 1 and 2 to identify predicted reversal rates for all counties in each state based on the number of death verdicts imposed by each county, then test those predicted rates to see how closely associated they are with the counties' actual reversal rates. Explanatory factors that were significant in these analyses are indicated by "predval." Analyses 16 and 17 explain variation among states, where each state is treated in effect as the composite of all of its counties, which in turn are treated as their aggregate values in all years in which they had at least one death verdict. Explanatory factors that were significant in this analysis are indicated by "st'comp/cntys." Analysis 18 explains variation in capital reversal rates among Florida counties. Explanatory factors that were significant in this analysis are indicated by "fl." All but one of the analyses collected here explain variation in aggregate reversal rates at all three stages of review. The one exception is Analysis 10, which explains variation in reversal rates at only the direct appeal stage. Explanatory factors that were significant in that single-stage analysis are indicated by "da."

    + One factor treated as county-level influence in the two county-only analyses is in fact a state-level variable introduced at the county level for a particular statistical purpose. As is discussed above, in any analysis of all three review stages combined using proportions of imposed as opposed to reviewed death verdicts, the reversal rate is partly a function of the review rate: Lower review rates automatically lead to lower reversal rates. See supra pp. 89-90, 97, 99, 153-54, 173-74, 212, 223. Our analyses use the number of death verdicts awaiting review to winnow out the effect of delay. This explanation in fact operates at the state level, because it is the product of the action and inaction of the state's unified appellate court system. In county-only Analyses 7 and 18, however, it is necessary to introduce a county-level analogue of that explanation to serve as a control for delay. That factor's significance in Analyses 7 and 18 does not reveal a separate county-level factor, and simply reflects the importance of a state-level factor the state analyses already demonstrate. See supra p. 252.

    ^ The measures of the effect of the passage of time are unreliable in these analyses for the same reasons given above in regard to other analyses of the three review stages combined. See supra pp. 89-91, 99, 140-42 & Table 4, 154-56, 177, 194-95, 258-59; supra note 482, note^.

    590. Of course, the states we study are also only ones that impose the death penalty. But their vastly different rates of doing so, and a variety of other death-sentencing-focused criteria illustrated by Figures 11-21 above, establish a high degree of inter-state variability. See supra pp. 120-32. Indeed, as we develop in connection with Figures 42A and 42B above, one important way in which capital states differ is in how few or many of their counties use the death penalty. See supra pp. 246-49.

    591. See supra not 504.

    592. See supra pp. 246-50 & Figures 42A, 42B.

    593. See supra pp. 250.

    594. See supra note 504 (explaining that in our regression analyses, we have 1002 counties, not the 1004 discussed here, because the source of most of the county demographic and other data used in our regression analyses combined two contiguous counties in each of two states for purposes of analysis, forcing us to do the same).

    595. Thus, if a state adopted a valid post-Furman death-sentencing statute in 1973, the number of homicides attributed to a county in that state is the number of homicides committed in the county during the full, 1973-1995 study period. On the other hand, if a state did not adopt a valid death-sentencing statute until 1978, then the number of homicides attributed to that county is the number committed there between 1978 and 1995. In either case, the death-sentencing rate is the number of death verdicts imposed under the modern death-sentencing statute divided by the number of homicides committed in the year that statute was adopted and in every subsequent year through 1995 multiplied by 1000.

    596. See infra pp. 347-49 & Tables 19 and 20.

    597. Overall reversal rates in this section are for the two of the three review stages: direct appeal and federal habeas. Nationally, during the study period, the overall reversal rate at those two stages was 65%. See supra p. 9. For discussion of how overall reversal rates are calculated, see supra pp. 8 & n.*, 19; supra note 103.

    598. Murder is the only offense for which state and federal law permit the death penalty. See Coker v. Georgia, 433 U.S. 584 (1977) and allied decisions (forbidding executions for rape, armed robbery and kidnaping and suggesting that Court would not permit the penalty for non-homicide offenses); cf. Enmund v. Florida, 458 U.S. 782 (1982), as modified by Tison v. Arizona, 481 U.S. 137 (1987) (setting some modest limits on the types of murders for which the death penalty may be imposed). Moreover, although some jurisdictions retain the penalty for aggravated rape of minors and for crimes against the state such as espionage and treason, homicide was the only offense for which the 34 states imposed the penalty during the study period. Because murder is a form of homicide, and because homicide is the only relevant category of offenses on which public agencies consistently collect and report statistics, the number of homicides is the best available indicator of the number of the chances each jurisdiction had to impose the penalty. On why it is useful to compare counties with similar number of homicides, see supra pp. 287-88.

    599. Here and throughout the remainder of this section, we consider only counties with five or more death verdicts during the study period, which are the counties collected in Table 11A in Appendix B.

    600.We also conducted this comparison using the four counties' exact numbers of homicides (from 986 to1361) to establish an upper and lower bound on the number of homicides in comparison counties. The results were nearly identical.

    601. Again, these counties are limited to those with five or more death verdicts during the study period.

    602. See supra pp. 8 & n.*, 19; supra note 103 (discussing how overall reversal rates are calculated).

    603. The link between high county death-sentencing rates and high capital error rates has been noted in the press recently. See, e.g., Lise Olsen, Unstable Marshall Is Fighting for his LifeCToday; Pierce County Shouldn't Have Sent Mentally Ill Man to Death Row, Court Rules, Seattle Post-Intelligencer, Aug. 7, 2001 (noting that "Pierce County [Washington] authorities, who over the years have sent more men to death row than prosecutors in any other county, . . . haven't been able to make many sentences stick. Of the eight Pierce County men condemned to die since 1981, only three remain on death row. Two of them were sent there this year and have not begun their appeals processes. The sentences of five others were reversed--convictions of three of those five also were overturned, and one of the five was set free. It's a capital-punishment machine that cost Pierce County $1.3 million in 1999 alone, according to state records. This year, the county--where the sheriff's office regularly has to forgo investigating many crimes for lack of funds--will spend well over $1 million on capital cases."). See also Lise Olsen, One Killer, Two Standards, Seattle Post-Intelligencer, Aug. 7, 2001 (Pierce County Prosecutor ALadenburg helped send two men to death row for shooting a single victim in a holdupCthe only such cases among the 30 men sentenced to death in the past 20 years, though one recently had his conviction overturned on appeal. . . . Ladenburg is the only prosecutor in the state whose death-penalty cases have been overturned for misconduct by his office--it's happened twice. . . . Prosecutors in Pierce [Coutny] lead the pack in seeking the death penalty, doing it in about 60 percent of all aggravated-murder cases, more than twice as often as King County [Seattle], according to ... a study by the Washington Association of Prosecuting Attorneys.)

    604. See supra pp. 134-35.

    605. The closest the research community has come to this ideal is the study David Baldus and colleagues conducted of the influence of race on capital sentencing in Georgia in the late 1970s and early 1980s. No similar study, however, has reached across state boundaries and decades.

    606. For example, state post-conviction decisions in Nevada, Tennessee (in the first half of the study period) and Texas are frequently unpublished even at the appellate stage. Virtually no trial-level state post-conviction decisions are published.

    607. See 28 U.S.C. '' 2254(b), (c).

    608. See supra pp. 15.

    609. Because the entire case drops out of a multiple regression analysis if any of the many conditions under consideration for that case is not known, we had to exclude from consideration factors as to which there were more than a handful of cases where the presence or absence of a trait was unknown. Traits are unknown when the reviewing judges didn't choose to mention it one way or the other in their opinions. Among the traits that were excluded for this reasons were the defendant's age at the time of trial, the last year of school completed by the defendant and the exact number of prior crimes committed by the defendant. In some cases, we could meaningfully change the question from whether a trait was present in the case, to whether any reviewing decision referred to the trait. In this event, "unknown" became "no." Judges' decisions to mention particular traits of cases, or not, were rarely significant, howeverCas one would expect, given the many, essentially conflicting, reasons why a judge might not mention a trait of a case (e.g., it was not present; the judge didn't know or wasn't sure it was present; it was present but the judge didn't think it was important enough to mention; it was present and played a role in the judge's decision or in the decision of another judge who took part in the case, but the writing judge did not choose or remember to mention it; etc.).

    As an example of the problem of insufficient variance, we knew the gender of the defendant in all cases, but there were so few women (6) that there was not enough variance to analyze.

    610. See supra pp. 85-86 & n.239, 109-10, 148.

    611. Aggravating and mitigating factors are treated somewhat differently by law, which accounts for our somewhat non-parallel treatment of them in this index. Most importantly, while essentially all capital statutes enumerate aggravating factors at least one of which must be present to justify a death sentence, not all statutes enumerate mitigating circumstances. On the other hand, while a number of states limit the aggravating factors jurors may consider to those enumerated in the statute, the federal Constitution requires that the sentencer consider all mitigating factors in the case, whether or not enumerated in the statute. In any event, we also constructed another index (the next one noted in text) which added some non-statutory aggravating factors, thus moderating the non-parallel treatment of aggravating and mitigating factors.

    612. These factors are discussed further at pp. 320-21 below.

    613. Scores on this index were not significantly related to the probability of federal habeas reversal.

    614. See supra pp. pp. 85-86 & n.239, 109-10, 148 (discussing statistical significance).

    615. This test provides effect-size information analogous to that provided in our state and county analyses. See supra pp. pp. 114017 & nn.282, 285; 171-73.

    616. Some factors have only two possibilitiesCe.g., the prisoner's lawyer was or was not from out of state; a state evidentiary hearing was or was not held; a majority of the members of the panel of federal judges who finally decided the case were or were not appointed by Republican Presidents. In that event, the exponentiated B value indicates that the probability of reversal increases (if the value is above 1) or decreases (if the value is below 1) by the amount of the value times the original probability. If the exponentiated B value is 1.94, then the probability of reversal increases by 94% (1 + 1(.94)) if the relevant condition is present. On the other hand, if the exponentiated B value is .65, then the probability of reversal decreases by 35% (1?1(.65)) if the condition is present. Other factors have a rangeCe.g., the number of aggravating circumstances minus the number of mitigating circumstances, and the number of supplemental aggravating circumstances. Here, the exponentiated B value indicates how much the probability of a reversal either increases or decreases with each increase or decrease of 1 in the relevant condition (e.g., in the number of aggravating circumstances net of mitigating circumstances, or in the number of supplemental aggravating circumstances). So, if the exponentiated B value for supplemental aggravating circumstances is .78, then for every supplemental aggravating that is present in the case, the likelihood of reversal decreases by 22% (1?1(.78)).

    617. This test, like the fit test discussed above, measures the distance between the federal habeas outcomes predicted by the set of explanatory factors and the actual outcomes. See supra pp. 149-50. We also examine results of three other diagnostic tests that tend to track the fit test:

    618. The relevant factors are coded as follows in the detailed results in Appendix G: state evidentiary hearing held ("seh2"); defense lawyer at final federal habeas stage is not from sentencing state ("dlos ffd"); federal evidentiary hearing was held ("feh2"); the number of statutory aggravating factors minus the number of mitigating factors ("agg mit"); and index of seven other aggravating factors relating to the offender and victim ("ofvcindx").

    619. See supra pp. 311-12 & n.616.

    620. Fit = 758.5 (5 explanatory factors).

    621. See supra pp. 311 & n.616 (discussing interpretation of exponentiated B's as a measure of effect size).

    622. See, e.g., Williams v. Taylor, 529 U.S. 420 (2000).

    623. See, e.g., Kyles v. Whitley, 514 U.S. 419 (1995).

    624. See, e.g., Amadeo v. Zant, 486 U.S. 214 (1988).

    625. See, e.g., Estelle v. Smith, 451 U.S. 454 (1981).

    626. See 28 U.S.C. '' 2254(b), (c).

    627. See 28 U.S.C. '' 2254(d)(2), 2254(e)(1), and pre-1996 statute, 28 U.S.C. ' 2254(d).

    628. See, e.g., Townsend v. Sain, 372 U.S. 293 (1963). Usually, the relevant hearing must be held after trial. Sometimes, a pre-trial hearing will sufficeCas where a defendant claims the police coerced him into confessing or denied him a requested lawyer during post-arrest interrogation and the trial court holds a hearing on the matter before trial and takes testimony from, e.g., the defendant and the arresting police officers.

    629. See supra pp. 199-200, 209.

    630. See, e.g., supra pp. 170-71, 228, 244.

    631. The descriptions of federal habeas practice and practitioners in this section are based on 1 Federal Habeas Practice & Procedure, supra note 203, '' 2.2, 4, 7.1, 7.2, 12.1-12.5, 13-20, and the many sources cited there.

    632. For two recent detailed profiles of some of of the "approximately 50" lawyers nationally in this category, see Rare Breed: Death Penalty Lawyers Defend Rights of Politically Invisible, Nat'l Catholic Reporter, Oct. 5, 2001; Dan McAllister, Not Dead Yet: Penalty Cases Are Finding Fewer Takers at Big Firms. But Some Lawyers Still Fight Life-and-Death BattlesCand Win, Amer. Lawyer, Dec. 4, 2001.

    633. Federal funding for such lawyers did not become available in all cases until 1988, when Congress provided for the mandatory appointment and funding of federal habeas lawyers in capital cases. See 21 U.S.C. ' 848(q)-848(r). Many of the cases in our study were decided before this law was passed.

    634. See, e.g., 1 Federal Habeas Practice & Procedure, supra note 203, '' 2.2, 13-20 and sources cited there.

    635. In a few states such as Georgia and Texas, members of local private law firms volunteer their services to in-state capital defendants undergoing federal habeas review. Most such federal habeas lawyers, however, are from large law firms in such cities as Boston, New York, Philadelphia, Chicago, Minneapolis, Denver, Los Angeles, San Francisco, Seattle and Washington, D.C., who volunteer to represent capital habeas petitioners in other states.

    636. Because of the high travel costs associated with representing a capital prisoner in another state, almost all of the out-of-state lawyers who volunteer to represent such prisoners are from large, well-funded private law firms. Such firms tend to hire only employees with strong academic records or proven track records in other firms, and they also tend to pay their lawyers well and provide ample support for investigators and experts.

    637. See supra pp. 308, 311 & n.611; infra notes 639, 643.

    638. See supra note 611; infra note 643.

    639. The aggravating circumstances present in the case are formally listed in published decisions somewhat more consistently than mitigating circumstances. For that reason, we used two alternative measures of aggravationCone focused entirely on the number of aggravating circumstances found in the case, the other on that number minus the number of mitigating circumstances. The results using each version were consistently very similar, leading us to choose the latter version because it includes a bit more information.

    640. This circumstances is not a duplicate of the common statutory aggravating circumstance noted above. This factor considers whether the defendant has any criminal record; most statutory aggravating circumstances consider a record of "violent" or "assaultive" crime.

    641. High status victims include law enforcement officers, fire fighters, public officials, and well-known and respected personalities in the community.

    642. See infra p. 324.

    643. The two aggravating circumstance indexes are not highly correlated. R = .038; p = .35. This indicates that the two indexes analyze different sets of information about the seriousness of the offense, the moral blame attached to it and the extent to which the defendant is responsible for the offense and its morally repugnant characteristics.

    As noted above, we constructed a third index of traits of cases, i.e., a count of the types of evidence used to convict defendants, such as fingerprints, eyewitness identification, blood or other bodily fluids, a weapon. We originally hoped this index might serve as a measure of the strength of guilt, because no direct measure of that condition is available. A review of the cases revealed the index's weakness for this purpose, however. First, unlike aggravation and mitigation, which are routinely and systematically listed in judicial decisions in most capital cases, descriptions of the evidence of guilt are inconsistent across cases and incomplete in most. Second, the presence of many types of evidence (even when we know it was present) is not a good indicator of the strength of the evidence of guilt. Under some circumstances, an eye-witness identification of the defendant as the killer is strong evidence of guilt. Under other circumstances, such identifications are highly unreliable. Complicating matters, analyses show that jurors are not always good at distinguishing reliable from unreliable eyewitness identifications. See, e.g., Richard O. Lempert, et al., A Modern Approach to Evidence 243-51 (2000) and sources cited. Habeas decisions' brief summaries of the facts of the case rarely indicate whether an eye-witness identification in the case fits in the "strong" or "weak" category, much less whether the jury accurately believed it was strong or weak. The same is true of many kinds of forensic evidence (e.g, hair, carpet sweepings, footprints), weapons and the like. So, when a case is coded as having an eye-witness identification or one of these other types of evidence, there is reason to expect the factor to be inconclusive as an indicator of strength of guilt, because it could point in opposite directions. (Although some kinds of evidence are more consistently strongCe.g., fingerprints and DNACthey are present in so few cases that there is too little variation to study. See supra note 609. This is especially so of DNA evidence, which was not available for trials until the early 1990s, and thus did not start appearing in cases under appellate review until the mid-1990s, at the very end of our study period. ) Again, evidence of aggravation and mitigation tends to be different in this regard. Which way a circumstance cuts is often stated by law, and when it is not so stated, the issue of relevanceCwhether the circumstance tends to increase moral blameCis often fairly clear on the face of the circumstance.

    The "types of evidence" index consistently had the same relationship to federal habeas outcomes as the level of aggravation: The fewer the types of evidence, the higher the probability of reversal. But unlike aggravation levels, fewer "types of evidence" was not a statistically significant predictor of reversals, because there is more than a 5% probability that the relationship appears by chance. See supra pp. 85-86, 109-10, 148 (on statistical significance). Given the difficulty of getting adequate data on this point, and given the ambiguity of available data, this result does not indicate that the strength of the evidence of guilt is irrelevant to the probability of reversal, but only that we cannot tell with any confidence whether it is relevant.

    The "types of evidence" factor has a stronger (although still not statistically significant) relationship to the probability of reversal when it is considered by itself than when it is considered along with the two indexes of aggravating circumstances. This suggests that the two indexes may partly measure the strength of the evidence of guilt, leaving less of a role for the "types of evidence" factor in this regard. This is likely, because many aggravating and mitigating circumstances reflect fairly directly on the quality of the evidence of guilt. Aggravating factors do this by establishing motives or mental states that increase the probability of guilt. Mitigating factors do it by establishing mental conditions or statuses that decrease the probability that the defendant had the mental state needed to establish first-degree murder or increase the likelihood that someone besides the defendant planned and committed the killing.

    644. See supra pp. 165-66, 168-69, 185, 210, 225-26, 243, 266.

    645. See supra pp. 157-59, 160-63, 165-66, 322-23.

    646. See supra pp. 22-23 & n.121 (citing sources), 125-35.

    647. See, e.g., Herrerra v. Collins, 506 U.S. 390 (1993); supra pp. 22-23.

    648. See supra pp. 25-35.

    649. See supra pp. 43-45.

    650. See supra pp. 314-15.

    651. See supra pp. 40-41.

    652. See supra p. 134 & n.295.

    653. Above, for example, we discovered that capital defendants with lawyers with volunteer, out of state lawyersCones who tend to be more skilled and better fundedCare more likely to have their capital verdicts reversed by federal habeas judges. See supra pp. 315-19. This could mean that skilled lawyers are better at uncovering error than other lawyers. Or it could mean that, although they are no better at uncovering error, they are more willing to volunteer to represent capital prisoners with stronger than with weaker legal claims. By controlling for the strength of claims, we can eliminate this latter hypothesis, because then, it appears that out of state lawyers secure reversals more often than other lawyers, even after controlling for the strength of the claims in the federal habeas petition.

    >654. See, e.g. 28 U.S.C. § 2254(e)(2).

    655. See id.

    656. See Federal Habeas Corpus Practice and Procedure, supra note 203, at ' 20.1.

    657. See id.

    658. R = .024.

    659. For example, the stronger the claim is, the more likely it is that the federal court will excuse the failure to raise the claim at trial or on appeal based either on prior counsel's incompetence, or on the state Court's unreasonableness in not providing a hearing.

    660. See Federal Habeas Practice and Procedure, supra note 203, § 20.4.

    661. A table in Appendix F examines whether the various factors found significant in Analysis 19 are correlated. None of those factors is significantly correlated with the "federal evidentiary hearing" factor.

    662. This factor is coded "sentyr" in the results in the Technical Appendix.

    663. Fit = 722 (6 explanatory factors).

    664. See supra pp. 20-21, 91-93 & Figure 10, 140-42, 155, 177, 193, 216-17.

    665. See supra pp. 91-93 & Figure 10.

    666. See supra pp. 20-21, 91-93 & Figure 10, 140-42, 155, 177, 193, 216-17.

    667. See supra pp. 194-96, 210, 235, 244, 259, 264 (Figure 43Q). For impressionistic evidence of an increase in state post-conviction reversal rates over time, see supra pp. 59-61 & Figures 3A and 3B. Cf. supra p. 213.

    668. See supra pp. 65, 67 (Figure 5), 244-45. If the proportion of flaws remained steady over time, while state direct appeal courts got better at catching them, the resulting increase in direct appeal reversals could explain a parallel decline in federal court reversals. We cannot tell whether this happened.

    669. See supra pp. 194-96, 210, 235, 244, 259, 264 (Figure 43Q).

    670. See supra pp. 65, 67 (Figure 5), 244-45.

    671. See id.

    672. This factor is coded "claimno" in the results in the Technical Appendix.

    673. There was one case in which both sentence year and the number of claims was missing, so Analysis 19C has 33, not 34, fewer cases than Analysis 19B.

    674. Fit = 656 (7 explanatory factors).

    675. The relationships here are complex, because the correlation between state evidentiary hearings and the number of claims, although positive, is very weak. R = .114; p = .007.

    676. We rejected two other interpretations of this factor. First, we considered the possibility that court decisions granting reversal may understate the actual number of claims raised at the final federal stage, because once reversal is granted on one claim, there may be no need to address, and thus to mention or list, other claims. We discounted this interpretation for two reasons. (a) Federal appellate courts (which usually serve as the final federal habeas stage) are often clear about the number of claims raised at that stage, and where they are not clear the cases were coded as "unknown," meaning the case was not counted in this analysis. (b) Most of the cases in which courts were unclear about the number of claims raised were affirmances, not reversals, of the defendant's capital verdict. Second, we considered the possibility that better lawyers may exercise more control over the number claims raised, suggesting a correlation between number of claims and bad lawyering (and between bad lawyering and denials of federal habeas relief). We discounted this hypothesis, after discovering that out-of-state lawyers (who by and large are at the top end of the spectrum of low to high quality lawyers) raised almost exactly the same number of claims as did other lawyersCin other words, the correlation between out-of-state lawyers and number of claims was very close to 0C requiring some other explanation besides the quality of lawyering for variation in the number of claims raised.

    677. See 28 U.S.C. § 2253. During the 1973-1995 study period, the general practice was to permit an appeal of as many claims as the lawyer chose to raise, if the federal courts were convinced that at least one of those claims was "substantial." A 1996 change in the law has led most federal courts to alter this practice, and hear only the particular claims on which appeal was permitted.

    678. Sometimes, capital defendants who win at the trial level will "cross-appeal," arguing that the lower court should have granted relief on more claims that it did. Because this puts lawyers in the awkward position of both defending and attacking the lower court decision, they typically limit cross-appeals to one or two very strong claims.

    679. This factor is coded "repmaj" in the results in Appendix G.

    680. The deciding judges are not named in 36 decisions reversing capital verdicts, but only 18 affirming them. The statistically significant correlation between unsigned decisions and reversals supports a point made earlier: Judges more often decline to publish or publicly associate themselves by name with decisions reversing capital verdicts than with decisions approving verdictsCperhaps because reversals are more controversial than affirmances. See supra p. 16 & n.100.

    681. Most of the cases lacking information on the deciding judges also lacked information on the number of claims raised, so the number of cases to be studied dropped by 24, not the full 54.

    682. Fit = 613 (8 explanatory factors).

    683. See supra pp. 38-39.

    684. See supra pp. 328-30.

    685. This factor was significant at the .05 level in some analyses and at other times dropped to between there and .10.

    686. For reasons given at pp. 328-30 above, Analysis 19 does not reliably reveal any relationship between the age of capital verdicts and their probability of reversal.

    687. See, e.g., supra pp. 165-66, 168-69, 185, 210, 225-26, 243, 266.

    688. See supra pp. 170-71, 228, 244.

    689. See supra pp. 199-200, 209.

    690. See supra pp. 25-35, 323-324.

    691. See supra pp. 323-24.

    692. See supra pp. 315-18.

    693. See supra pp. 16, 218-19, 266, 285. See also supra pp. 194, 257.

    694. See supra pp. 198-99, 212, 217-18, 227, 236.

    695. See supra pp. 216-17.

    696. See supra pp. 91-93 & Figure 10, 140-42, 155, 177, 193, 216-17, 328-30.

    697. See supra pp. 25-35.

    698. See supra pp. 287-306 & Tables 10-16.

    699. Brooke A. Masters, Executions Decrease For the 2nd Year: Va., Texas Show Sharp Drops Amid A National Trend, Wash. Post, Sept. 6, 2001.

    700. A Governor's Role in Death Penalty Cases, Burden of Proof with Greta Van Susteren, CNN, Aug. 21, 2001, www.CNN.Com/Burden.

    701. See supra pp. 166-68, 183-84.

    702. See supra pp. 166-68, 183-84, 197, 214, 224, 239-42, 252, 256-57, 269, 271, 274, 280-81 (significant to highly significant in Analyses 2-5, 7-15, 18; just barely above significance in Analyses 16 and 17).

    703. See supra pp. 166-68, 183-84, 224, 239-42, 252, 256-57, 269, 271, 274, 280-81 (Analyses 1, 2, 7-9, 11-15; 18; just barely above significance in Analyses 16 and 17).

    704. See supra pp. 197, 214 (Analyses 3-5).

    705. See supra pp. 319-23 (Analysis 19).

    706. Regarding state reversal rates, see supra pp.166-68, 183-84, 197, 214, 224, 239-42,(Analyses 1-4, 14, 15). Regarding county reversal rates, see supra pp. 252, 256-57, 269, 271, 274, 280-81 (Analyses 7-13, 18; just above the .05 significance level in Analyses 16 and 17).

    707. See supra pp. 250, 265-66, 272, 281.

    708. See supra pp. pp. 166-68, 183-84, 224, 239-42, 252, 256-57, 269, 271, 274, 280-81 (compare Analyses 7-13, 16-18 to Analyses 1-5, 8-17).

    709. See supra pp.287-306.

    710. See supra pp. 86-98.

    711. See supra pp. 183-84 & Figures 27A-D.

    712. See, e.g., supra pp. 183-84 & Figures 27A-D, 197, 205 (Figures 35A, 35B), 229 (Figure 41C), 260 (Figures 43C, 43D), 271 & n.550.

    713. In one case, discussed below, we examine rankings and predicted reversal rates based on direct appeal Analysis 3A.

    714. Each state's weighted average for each factor is calculated using the following formula: (((number of death sentences in year x1) x (factor value in year x1)) + ((number of death sentences in year x2) x (factor value in year x2)) + . . . ((number of death sentences in year xn) x (factor value in year xn))) ? total number of death verdicts in years x1 through xn.

    715. See supra pp. 70-73 & Figure 6.

    716. See supra pp. 246-49 & Figures 42A, 42B. A majority of counties in 59% (20) of the 34 capital states imposed no death sentences during the study period.

    717. The data are for the counties in which each city is located. Where county and city names differ, the county names are Maricopa (Phoenix), Harris (Houston), Dade (Miami), Cook (Chicago), Clark (Las Vegas), Pinellas (St. Petersburg), Hillsborough (Tampa), Duval (Jacksonville), Jefferson (Birmingham), and Broward (Ft. Lauderdale). Sources for this table are DRCen, DADB, HCDB, Vital Statistics.

    718. Two exceptions to this caveat are Shreveport, Louisiana and Dayton, Ohio, which had fewer than five death verdicts during the study period.

    719. The Florida counties are Leon, Marion, St. Johns and Volusia.

    720. The Florida counties are Bay, Brevard, Escambia, Martin, Okaloosa, Pinellas, Putnam and Indian River.

    721. The additional Arizona counties are Yavapai and Yuma. The five additional Florida counties are Bradford, Citrus, Columbia, Pasco and Taylor.

    722. The Georgia counties are Cook, Douglas, Jones, Meriwether, Seminole and Wayne. The Alabama counties are Blount, Coffee, Colbert, Monroe and Talladega. The Arizona county is Mohave. The Florida counties are Hernando, Santa Rosa, Sumter and Union.

    723. The rates set out here are for the counties in which listed cities are located. Where that name is different from the listed city, the counties are as follows, in the order of locales listed in text: Shreveport (Cado Parish, LA), Dayton (Montgomery County, OH), Newark (Essex County, NJ), Atlanta (Fulton County, GA), Kansas City (Jackson, MO), Nashville (Davidson, TN), Albuquerque (Bernalillo, NM), Las Vegas (Clark County, NV), Reno (Washoe County, NV), suburban Baltimore (Baltimore County, MO), Akron (Summit County, OH), Jefferson City (Cole County, MO).

    724. These counties are listed supra notes 719-22.

    725. For recent articles contrasting relatively high death-sentencing areas like Houston, Philadelphia, suburban Baltimore County, Danville, Virginia, Columbus and Baldwin County, Georgia and Cincinnati with relatively low death-sentencing areas like Dallas, Pittsburgh, Baltimore City, Richmond, Virginia, Atlanta, and Columbus, Ohio, see 100 Colum. L. Rev., supra note 153, at 2068-69 n.114. See also Brooke A. Masters, Death Penalty, Location Are Linked in Va. Study: Execution Sought Most Often in Suburbs, Wash. Post, Dec. 11, 2001 ("Suburban prosecutors are significantly more likely to seek capital murder indictments and ask juries for a death sentence than their counterparts in rural and urban areas, the Joint Legislative Audit and Review Commission concluded after a year-long study. . . . The Virginia study concluded that prosecutors in medium-density jurisdictions, such as Prince William County and Danville City, sought the death penalty in 45 percent of possibly capital cases, compared with 16 percent in urban areas such as Richmond and Norfolk and 34 percent in rural areas."); Lise Olsen, One Killer, Two Standards, Seattle Post-Intelligencer, August 7, 2001:

    726. See supra pp. 287-306 & Tables 10-16; Appendix B.

    727. See supra pp. 250, 265-66, 272, 281.

    728. See supra pp. 319-20 & n.639.

    729. See supra pp. 313, 319-20, 328, 330, 333.

    730. See supra pp. 322-24. See also 157-59, 160-63, 165-66.

    731. See supra pp. 168, 185-86, 197, 226-27, 256-57, 268-69, 271, 274. This result was reached by all studies of state-level factors related to state and county capital error rates at all three review stages combined and at the state direct appeal stage (Analyses 1-4, 8-17). Although we express the finding in the text as one about state reversal rates in states with poor law enforcement records, the finding also applies to county reversal rates in such states.

    732. See supra pp. 185-86, 197, 206, 226-27, 232, 256-57, 262, 271 & n. 274, 277& n. 55, & Figures 28A-D, 36A, 36B, 41G, 43K, 43L, 44F.

    733. See infra pp. 370-72.

    734. See supra pp. 165-66, 168-69, 185, 210, 225-26, 243, 266, 321-22, 335, 349-50.

    735. Both a general explanation for high error rates (heavy use of the death penalty) and a related specific explanation (concerns about the ineffectiveness of the state's response to serious crime, triggering heavier use of the penalty) can be significant at the same time, if (1) there are multiple reasons for heavy death-sentencing, and (2) some reasons are more closely linked to error than others. In that event, an indicator of the intensity of one of the important reasons for heavy use of the penalty leading to error (e.g., evidence that non-capital law-enforcement strategies are ineffective) will only partly explain high error rates, leaving the rest to be explained by indicators of the other important pressures, or by a general measure of all pressures to use the death penalty (e.g., high death-sentencing rates). Below, we explain why the four separate pressures to use the death penalty addressed in this and the next three sections may be particularly conducive to high rates of capital error, and thus why it is not every additional use of the death penalty, but only the penalty=s use in weakly aggravated cases, that increases error rates. See infra pp. 359-60, 367.

    736. See supra p. 343.

    737. See supra pp. 51 (Figure 1B).

    738. See supra pp. 169-70, 187-88, 198, 217-18, 227, 258, 269, 271, 274 (Analyses 1-4, 6, 8-17). In regard to direct appeal Analyses 3 and 4 in which this result fell just barely above the .05 level (p = .06), see supra note 486.

    739. See id.

    740. See supra pp. 119-20, 133.

    741. See supra pp. 169 & n.369.

    742. See supra pp. 187-88, 199, 207, 217-18, 221, 227, 232, 258, 262, 271 & n.554, 274, 277, & Figures 29A-D, 37A, 37B, 40C-1, 40C-2, 41H, 43I, 43J, 44E.

    743. See supra pp. 69-70 & n.370.

    744. Virginia in fact got only the second lowest score that is possible on the index, namely, a score of 2. But that was the lowest score among the 34 study states. None of the 34 study states scored a 1, because none uses gubernatorial or nonpartisan appointment procedures in addition to immunizing judges entirely from regular or at least retention or recall elections.

    745. The formula for calculating this factor is white homicide victims per 100,000 whites ÷black homicide victims per 100,000 blacks. As is discussed supra p. 160 & n.341, the homicide rate among blacks is usually higher than among whites. In most cases, that is, this factor compares states based on how much lower the white homicide rate is than the black homicide rateCor, conversely, how closely the white homicide rate approaches the black homicide rate.

    746. See supra pp. 159-61, 181-82, 196, 213-14, 225-26, 257, 269, 271, 274. This result was reached by analyses studying state and county capital error rates at all three review stages combined and at the state direct appeal stage separately (Analyses 1-4, 8-17). This factor was just above the .05 significance level in our single-stage analysis of the state post-conviction stage, where, in addition, there was a significant relationship between higher reversal rates and higher homicide rates among whites (apart from any comparison to the homicide rate among blacks) (Analysis 5).

    747. See supra pp. 181-82, 196, 204, 213-14, 225-26, 231, 257, 262, 269, 271 & n.552, 274, 276, & Figures 26A-D, 34A, 34B, 41F-1, 41F-2, 43G, 43H, 44D.

    748. Although a few other states have much lower predicted reversal rates, their extremely low black populations lead us to exclude them from the comparison made in text.

    749. See supra pp. 224-26, 253, 280 (Analysis 7, 15 and 18). See also supra p. 213 (significance of homicide victimization rate among whites in Analysis 5).

    750. The one exception was Analysis 5, in which the white homicide rate by itself was a slightly better predictor of error rates at the state post-conviction stage than the white/black homicide rate. See supra pp. 213-14.

    751. See Cole, supra note 337; Kennedy, supra note 337; other sources cited supra notes 337, 338, 358, 360. See also sources cited supra notes 349-51, 353, 354.

    752. See sources cited supra note 337, 338.

    753. See supra note 735 (discussing conditions under which not only high death-sentencing rates themselves, but also particular pressures to increase death-sentencing rates, could both be significant).

    754. See supra pp. 157-59, 160-63, 165-66 & Table 6.

    755. See supra pp. 322-24, 351.

    756. The relevant policies appear to be related to the statewide distribution of the risk of homicide among whites and blacks, not to its local distribution. In no analysis of county-level factorsCnot even Analysis 7, which omitted state-level factors, giving county-level factors the maximum opportunity to explain reversal ratesCwas there any significant relationship between the countywide distribution of the homicide risk between whites and blacks and county reversal rates.

    757. See supra pp. 157-59, 179, 196, 224, 257, 269, 271, 274. This result is reached by all our analyses of state factors associated with state and county reversal rates at all three review stages combined and at the state direct appeal review stage by itself (Analyses 2-4, 8-17).

    758. See supra pp. 217 (Analysis 6).

    759. See supra pp. 179-80, 196, 203, 224, 257, 261, 271 n.551, 274, 276 & Figures 25A-25D, 33A-33D, 41D-1, 41D-2, 43E, 43F, 44C.

    760. See supra pp. 217, 220 & Figure 40B (Analysis 6). Like other effects, these ones appear to operate at the level where policy is made, not at the level of individual cases. Capital verdicts imposed on black defendants are no more likely to be overturned due to serious error than those imposed on white or other defendants. See supra pp. 157-59 & Table 6. (Because nearly all capital defendants are poor, but information on how poor is not kept by officials, differences in error rates linked to the economic status of capital defendants cannot be studied.)

    Reflecting another pattern noted above, the relevant capital policies seem to be related to the proportion of African-Americans in the state, not the county, population. No analysis of county-level factorsCnot even Analysis 7, which omitted state-level factors, giving county-level factors the greatest opportunity to explain reversal ratesCrevealed any significant relationship between proportion of blacks in a county's population and its capital reversal rates.

    761. See supra pp. 163-65.

    762. See supra pp. 157-59 & Table 6, 160-63; supra note 760.

    763. This research is collected supra notes 358, 360.

    764. See supra p. 361.

    765. See supra pp. 5 & n.77, 24-35 & nn.134, 146, 148; supra p. 80 & n.227.

    766. See sources cited supra notes 358, 360.

    767. See, e.g., supra p. 160 & n.341.

    768. See sources cited supra note 358.

    7694. See id.

    770. See supra pp. 224-26, 266, 285, 336.

    771. See id.

    772. This factor was significant to highly significant in Analyses 1, 2, 8-10, 12, and 13 and fell just above the .05 significance level in Analyses 4, 11 ( p = .056), and 14. See supra pp. 162-63, 226, 257, 269, 271.

    773. See sources cited supra notes 337, 338.

    774. states' Rank and Value Based on Interaction of Race of Population and of Homicide Victims

    State

    Interaction of Race of Population and of Homicide Victims

     

    Rank

    Value

    Connecticut

    21

    0.014

    Kentucky

    16

    0.020

    Maryland

    8

    0.046

    Tennessee

    11

    0.036

    Mississippi

    2

    0.094

    Oregon

    29

    0.003

    California

    18

    0.019

    New Jersey

    15

    0.022

    Idaho

    32

    0.001

    Montana

    34

    0.000

    Georgia

    4

    0.066

    Arizona

    27

    0.007

    Alabama

    5

    0.060

    Colorado

    26

    0.010

    Washington

    28

    0.005

    Wyoming

    33

    0.000

    Florida

    13

    0.030

    Oklahoma

    17

    0.019

    Indiana

    25

    0.010

    Arkansas

    12

    0.034

    North Carolina

    6

    0.059

    Nebraska

    30

    0.002

    Nevada

    20

    0.016

    South Carolina

    1

    0.100

    Utah

    31

    0.001

    Louisiana

    3

    0.070

    Illinois

    14

    0.022

    Pennsylvania

    23

    0.013

    Texas

    10

    0.040

    Missouri

    22

    0.013

    Delaware

    7

    0.047

    New Mexico

    24

    0.012

    Ohio

    19

    0.016

    Virginia

    9

    0.044


    Source: Analysis 1A

    775. See supra pp. 157-59, 160-63, 165-66, 322-24, 351, 360-61 & n.756, 362 & n.760.

    776. See supra pp. 169 & n.369, 356.

    777. See supra note 735.

    778. See supra pp. 153, 173, 194, 212, 253, 257, 269, 271, 273, 280 (Analyses 1-5, 7-18).

    779. See supra p. 173-74 & Figures 22A, 22B. See also pp. 201, 229, 260, 276 (Figures 22A-D, 31A, 31B, 41A, 43A, 43B, 44A). In several analyses, high per capita filings of court cases of all types operate similarly. Effect size is too small to warrant additional attention, however. See supra pp. 154, 174, 176, 223, 229, 258, 263 & Figures 23A-D, 41B, 43N.

    780. See supra pp. 90, 173-75 & Figures 22A-D.

    781. See supra p. 342 & n.714.

    782. See supra pp. 89-90, 153-54, 173-74, 212, 223, 351.

    783. See supra pp. 89-91, 99, 140-42 & Table 4, 154-56, 177, 194-95, 213, 216-17, 258-59.

    784. See supra pp. 20-21, 88-89, 172, 183, 194-95, 343 & n.713.

    785. See, e.g., supra p. 21 & n.116.

    786. See supra 91-93 & Figure 10, 140-42, 155, 216-17, 177, 193, 328-30, 336.

    787. See supra pp. 194, 257.

    788. The table below reports state rankings, and the difference between their predicted reversal rates based only on their capital backlogs, holding other factors at their averages. As is discussed in the text, the high reversal rates predicted for states with low capital backlogs (e.g.., Nebraska), and the low predicted reversal rates for states with large backlogs and delays in capital appeals (e.g., California) occurs because of the perverse tendency of delay to depress reversal rates. Because this factor does not accurately reflect the risk of errorCand instead reflects the effect of delay—we report its results here, rather than in Table 18.

    States' Rank, and Comparison to Predicted 34-State Average Error Rate,
    Based on Capital Backlogs (Analysis 1A), Holding Other Factors at the 34-State Average

    State

    Backlog of Capital Appeals (Higher Value = Lower Review Rate = Lower Reversal Rate)

     

    Rank

    Value

    Difference from 34-State Avg. Error Rate

    Connecticut

    4

    0.8

    +40.7%

    Kentucky

    16

    2.8

    +30.7%

    Maryland

    6

    1.1

    +38.9%

    Tennessee

    23

    5.4

    +19.0%

    Mississippi

    14

    2.3

    +33.2%

    Oregon

    12

    1.8

    +35.3%

    California

    34

    26.9

    -10.7%

    New Jersey

    9

    1.4

    +37.4%

    Idaho

    13

    1.9

    +34.8%

    Montana

    2

    0.5

    +41.9%

    Georgia

    21

    4.0

    +25.1%

    Arizona

    25

    6.5

    +14.4%

    Alabama

    27

    7.9

    +9.5%

    Colorado

    8

    1.3

    +37.8%

    Washington

    3

    0.5

    +41.7%

    Wyoming

    5

    0.8

    +40.4%

    Florida

    32

    17.6

    -7.5%

    Oklahoma

    26

    6.9

    +13.1%

    Indiana

    17

    2.9

    +29.9%

    Arkansas

    7

    1.1

    +38.8%

    North Carolina

    29

    9.9

    +3.9%

    Nebraska

    1

    0.4

    +42.1%

    Nevada

    22

    4.4

    +23.1%

    South Carolina

    19

    3.5

    +27.4%

    Utah

    11

    1.6

    +36.4%

    Louisiana

    15

    2.4

    +32.3%

    Illinois

    28

    8.7

    +7.0%

    Pennsylvania

    31

    14.6

    -4.6%

    Texas

    33

    18.5

    -8.1%

    Missouri

    24

    5.7

    +17.8%

    Delaware

    18

    3.2

    +28.7%

    New Mexico

    10

    1.5

    +36.7%

    Ohio

    30

    12.4

    -1.4%

    Virginia

    20

    3.8

    +25.7%


    Source: Analysis 1A.

    789. See supra pp. 170, 227, 269, 272, 273-74 (Analyses 1, 2, 11-17).

    790. See supra pp. 199-200 (Analyses 3, 4).

    791. See supra pp. 199-299, 209 & Figures 39A, 39B (Analyses 3 and 4). See also supra p. 54 & n.194.

    792. See supra p. 41.

    793. See infra pp. 413-18. For citation and discussion of numerous government, bar association, judicial and press reports thoroughly documenting the relationship between low funding levels and incompetent capital lawyering, and the especially high demands that capital cases place on lawyers and legal support services, see 100 Colum. L. Rev., supra note 153, at 2102-10 & nn.175-91.

    794. See supra pp. 314-15.

    795. See supra p. 342 & n.714.

    796. See supra pp. 199-200 & Figures 39A, 39B.

    797. States' Value and Rank, Holding Other Factors Constant, for Interaction of Backlog of Capital Appeals and General Court Caseloads

    State

    Interaction of Backlog of Capital Appeals and General Court Caseloads

     

    Rank

    Value

    Connecticut

    13

    -0.2

    Kentucky

    27

    -1.6

    Maryland

    11

    0.8

    Tennessee

    10

    0.9

    Mississippi

    21

    -0.9

    Oregon

    22

    -0.9

    California

    1

    69.0

    New Jersey

    12

    0.7

    Idaho

    28

    -1.6

    Montana

    15

    -0.3

    Georgia

    9

    1.9

    Arizona

    33

    -3.5

    Alabama

    24

    -1.3

    Colorado

    19

    -0.8

    Washington

    14

    -0.3

    Wyoming

    20

    -0.8

    Florida

    2

    54.5

    Oklahoma

    17

    -0.5

    Indiana

    30

    -1.9

    Arkansas

    18

    -0.6

    North Carolina

    6

    4.4

    Nebraska

    16

    -0.4

    Nevada

    34

    -4.2

    South Carolina

    29

    -1.8

    Utah

    25

    -1.5

    Louisiana

    26

    -1.5

    Illinois

    4

    12.9

    Pennsylvania

    5

    12.3

    Texas

    3

    20.5

    Missouri

    8

    2.0

    Delaware

    32

    -3.3

    New Mexico

    23

    -1.1

    Ohio

    31

    -2.7

    Virginia

    7

    4.3


    Source: Analysis 1A.

    798. See supra pp. 52-61 & Figures 2C, 2D, 3A, 3B; Broken System, Part I, supra note 101, at 38, Figure 3.

    799. As long as some proportion of imposed death verdicts are flawed, and as long as reversal rates are calculated by taking the number of reversals as a proportion of all imposed verdicts (reversals divided by imposed verdicts), any drop in the number of finally reviewed verdicts will drive down reversal rates, because fewer of the flawed verdicts will be available to be reversed while the number of imposed verdicts stays the same. The numerator (reviewed and reversed verdicts) shrinks while the denominator (imposed verdicts) stays the same, causing the rate to drop.

    800. See supra pp. 89-90, 97, 99, 152-53, 154-56.

    801. See supra pp. 89-91, 99, 140-42 & Table 4, 154-56, 177, 194-95, 213, 216-17, 258-59.

    802. See supra pp. 20-21, 91-93 & Figure 10, 140-42, 155, 177, 193, 216-17, 328-30, 336, 353, 369.

    803. In that event, delay affects the denominator as much as the numerator. See supra note 799.

    804. See supra pp. 194-95, 202, 258-59, 263 & Figures 32A, 32B, 43Q. See also supra pp. 20-21, 88-89, 172, 173, 352-53. Effect size for Analysis 3 is graphed in Figure 32A, p. 202. Analysis 4 likewise predicts and 8-fold increase in error rates over the study period (Figure 32B, p. 202). When county as opposed to state reversal rates are analyzed, the predicted increase in reversal rates over the 23-year period is 3-fold (Figure 43Q, p. 263).

    805. See supra note 194.

    806. See supra pp. 89-91.

    807. See supra pp. 89-91, 99, 140-42 & Table 4, 154-56, 177, 194-95, 213, 216-17, 258-59, 372-75.

    808. See supra pp. 216-17, 328-30 (Analyses 6, 19).

    809. See supra pp. 154-56, 213, 216-17, 253, 258-59, 280 (Analyses 1, 2, 5-7, 11, 12, 18).

    810. See supra pp. 177-78 & Figures 24A-D (Analyses 1A, 1B, 2A, and 2B). See also the effect-size estimates in Appendix G for Analyses 5, 7, 11 and 12 (infra pp. G-7, G-9, G-13, G-14) and supra pp. 280 and n.577 for a discussion of the low effect size for this factor in Analysis 18.

    811. See supra pp. 259, 271 & n.548, 264 (Figure 43P) (discussing and displaying the results of Analyses 8, 9 and 13, in which there was no significant relationship between error rates and the passage of time).

    812. See supra pp. 259, 271.

    813. See supra pp. 54, 57, 58 & Figures 2C and 2D.

    814. See supra pp. 52-53 & Figure 2A; A Broken System, Part I, supra note 101, at 38, Figure 3.

    815. See supra pp. 65, 67 (Figure 5), 245 & nn.499, 500..

    816. See supra pp. pp. 52-53 & Figure 2A. See also supra pp. 89-91, 99, 140-42, 154-56, 177, 194-95, 213, 216-17, 258-59, 351-52.

    817. See supra p. 24.

    818. See supra pp. 1-5.

    819. See supra pp. 63-64 & nn.202-03.

    820. See Editorial, State's Record in Death Cases Cause for Study, Tallahassee Dem., Dec. 14, 2001:

    821. See supra pp. 6-7 (explaining why it is much harder in the capital than in these other areas to tell whether egregious harm has —including because officials are permitted to withhold and destroy evidence on the question).

    822. Classic treatments of this problem in the Chicago Tribune are: Ken Armstrong & Steve Mills, Flawed Murder Cases Prompt Calls for Probe, Chi. Trib., Jan. 24, 2000; Ken Armstrong & Maurice Possley, Break Rules, Be Promoted, Chi. Trib., Jan. 14, 1999 (detailing patterns of error by Illinois police officers, prosecutors, and judges that went unnoticed and unremedied by reviewing judges), and Ken Armstrong, "Cowboy Bob" Ropes Wins—But at Considerable Cost , Chi. Trib., Jan. 10, 1999 (same, Oklahoma City prosecutors). For other examples, see Sara Rimer & Raymond Bonner, Texas Lawyer's Death Row Record a Concern, N.Y. Times, June 11, 2000 (same, Texas defense lawyer); Shiffman, supra note 102 (Tennessee courts' failure to review comparative information kept in capital cases, which has sat in files, entirely unused, for 30 years); 100 Colum. L. Rev., supra note 153, at 2089-91 n.151 (various sources discussing pattern of misconduct in capital cases by police at particular Chicago precinct house), 2094-95 n.160 (Armstrong & Possley discussing Chicago prosecutor; Hunt discussing Cincinnati prosecutors; Rosenberg, discussing Philadelphia prosecutors), 2101 n.173 (Armstrong & Possley, discussing New Orleans police and prosecutors), 2104 n.178 (various sources discussing continued appointment of bar-disciplined capital defense lawyers), 2119-29 7 nn.227-33 (esp 231). See also notes 123, 160 (tendency of judges in Illinois, Ohio and Texas to pass over error, as "harmless"); infra note 941 (additional examples of repeated appointment of same poorly prepared defense lawyers in Texas and elsewhere).

    823. These reports are collected and discussed in 100 Colum. L. Rev., supra note 153, at 2119-29.

    824. See id. at 2120-21.

    825. See id. at 2121-27.

    826. See id. at 2078-82 & nn.137-40 (citing numerous examples of susceptibility of prosecutors to political pressures in potentially capital cases); supra pp. 169, 187.

    827. See supra pp. 194, 257, 368-69.

    828. See supra pp. 315-18.

    829. See supra p. 63 & n.202.

    830. See supra pp. 63-64.

    831. See supra pp. 70-80.

    832. See, e.g., supra pp. 25-35 (four cases studies).

    833. See supra pp. 5 n.77, 24.

    834. See supra pp. 25-35.

    835. See supra pp. 37-38.

    836. See supra pp. 319-24.

    837. See supra pp. 169-70, 187-88, 198, 217-18, 227, 258, 269, 271, 274, 354-56.

    838. See supra pp. 198-99, 212, 217-18, 227, 230, 240 n.486, 336.

    839. See supra pp. 218-19, 236. See also infra note 876 (cataloguing studies in which this factor was significant and had considerable—in some cases quite large—effect size).

    840. By "more" and "less" homicides and death verdicts, we mean numerically, not per homicide.

    841. See supra pp. 218-19, 236.

    842.States' Rank, and Comparison to Predicted 34-State Average Error Rate,
    Based on Population Size and Density, Other Factors at the 34-State Average

    State

    Population Size and Density

     

    Rank

    Value

    Difference from 34-State Avg. Error Rate

    Connecticut

    8

    0.92

    +4.5%

    Kentucky

    19

    0.15

    -6.8%

    Maryland

    7

    0.94

    +4.7%

    Tennessee

    14

    0.39

    -3.4%

    Mississippi

    22

    -0.30

    -12.8%

    Oregon

    27

    -0.53

    -15.6%

    California

    1

    1.62

    +14.9%

    New Jersey

    2

    1.61

    +14.8%

    Idaho

    31

    -1.50

    -25.6%

    Montana

    33

    -1.96

    -29.1%

    Georgia

    13

    0.42

    -3.0%

    Arizona

    26

    -0.48

    -15.0%

    Alabama

    20

    0.12

    -7.2%

    Colorado

    24

    -0.39

    -14.0%

    Washington

    17

    0.16

    -6.7%

    Wyoming

    34

    -2.26

    -31.1%

    Florida

    6

    1.15

    +7.9%

    Oklahoma

    21

    -0.25

    -12.1%

    Indiana

    12

    0.60

    -0.4%

    Arkansas

    25

    -0.45

    -14.7%

    North Carolina

    10

    0.64

    +0.2%

    Nebraska

    28

    -1.01

    -20.9%

    Nevada

    32

    -1.59

    -26.3%

    South Carolina

    18

    0.16

    -6.7%

    Utah

    29

    -1.14

    -22.2%

    Louisiana

    15

    0.25

    -5.4%

    Illinois

    5

    1.16

    +8.0%

    Pennsylvania

    3

    1.29

    +10.0%

    Texas

    9

    0.82

    +2.9%

    Missouri

    16

    0.23

    -5.6%

    Delaware

    23

    -0.34

    -13.2%

    New Mexico

    30

    -1.30

    -23.7%

    Ohio

    4

    1.24

    +9.2%

    Virginia

    11

    0.62

    0.0%


    Source: Analysis 1A.

    843. See supra pp. 194, 201, 212, 257.

    844. See supra pp. 194, 257.

    845. The Federalist, No. 83, at 499 and No. 81, at 486 (Clinton Rossiter ed. 1961).

    846. See supra pp. 333-34.

    847. See supra pp. 321-24.

    848. See supra pp. 25-25, 321-24.

    849. See supra pp. 6-7.

    850. See supra pp. 1-2 & n.3, 4-5 & n.69.

    851. See supra p. 11.

    852. See supra pp. 350-66.

    853. See supra pp. 37-69.

    854. See supra pp. 24-25.

    855. See supra pp. 16 & n.100, 194, 198-99, 212, 217-18, 227, 230, 257, 240 n.486, 336, 382-84.

    856. See supra pp. 5 n.70, 24.

    857. See 100 Colum. L. Rev. supra note 153, at 2050-51 n.84 (collecting sources); supra pp. 25-35.

    858. See supra pp. 343, 353-54. This is true even when Table 18 is supplemented by the information in the tables in notes 774, 788, 797 and 842 above.

    859. See supra notes 774, 788, 797, 842.

    860. See supra note 714.

    861. Moreover, as is noted above, Connecticut's and Colorado's highest "risk" ranking is, perversely, a result of how few delays the two states have in processing capital appeals, as is reflected by their low backlogs of pending capital appeals. See supra pp. 382-84 & n.797. As we develop above, delays in reviewing capital verdicts decrease the number of reversals (because there are fewer outcomes of any sort) and depress reversal rates measured as proportions of imposed verdicts (for the same reason). Large backlogs of pending capital appeals also seem to make reviewing courts more willing to tolerate error that would lead to reversal if the courts had fewer cases backed-up awaiting review. See supra pp. 194, 257, 382-84. As a result, states like Connecticut and Colorado where reversal rates are solely a function of error, and are not confounded by delay, are at "risk" of having higher reversal rates than states such as California, Florida and Texas, where the number and rate of reversals are held down by low rates of decided appeals.

    862. See supra note 714.

    863. See supra pp. 194, 257, 382-84; supra note 861.

    864. See supra pp. 303-04 & Table 16.

    865. See Innocence and the Death Penalty, supra note 77.

    866. Georgia has two rankings of 21 out of 34.

    867. See sources cited supra notes 160, 165. See also Editorial, Harmful ErrorCCriminal Appeals Court Eroding Justice for all Texans, Houston Chron., Jan. 21, 2002 ("The Texas Criminal Court of Appeals is the state's highest authority on criminal justice matters. In recent years, a majority of the Court's jurists have demonstrated an affectionate tolerance of incompetent judging and lawyering . . . . In an opinion earlier this month, the court ruled that a defendant facing execution has no right to have a competent lawyer handle his appeal.").

    868. See supra pp. 68-69.

    869. See id.

    870. See supra pp. 341-45, 354-56.

    871. These Review Commission=s findings are discussed in note 30 above.

    872. See infra pp. 405-06, 411-12, 418.

    873. See supra p. 3.

    874. See supra pp. 350-66.

    875. Such doubts evidently tempt policy makers to broaden the use of the death penalty as a quick and dramatic demonstration of a will to combat serious crime. See supra pp. 350-66. Above we explain why this response to the problem of serious crime is likely to make matters worse, given the association between high capital-sentencing rates and high rates of serious capital error. Obviously, erroneously convicting and capitally sentencing the wrong person, while leaving the actual killer at largeCas has repeatedly occurred, see sources cited supra notes 227—is not an effective law enforcement strategy. Moreover, the death penalty is unlikely to provide an effective across-the-board solution to problems with a state's response to serious crime, given that the Constitution bars the death penalty for any crime other than wilful homicide, including crimes that affect the largest number of citizens—rape, robbery, burglary, car jacking, drug dealing, theft, assault and even most homicides (e.g., ones during domestic arguments and spur-of-the-moment fights between acquaintances). Instead, attempting to cure across-the-board deficiencies in a state's law enforcement capacity by extending the death penalty to some additional, marginally aggravated homicides is likely to divert millions of dollars per capitally sentenced inmate away from effective crime control and into additional trials, multi-stage appeals and retrials that in the vast majority of cases will end with non-capital outcomes.

    876. See supra pp. 170, 189-90 & Figures 30A-D, 199, 206 (Figures 36A, 36B), 215, 227, 232 (Figure 41I), 253, 258, 263 (Analysis 43M), 266, 269, 271 & n.556, 274, 277 (Figure 44G) (Analyses 1-5, 7, 9-17).

    877. A cure for the crime problem eluded policy makers for decades beginning in the mid-1960s. Even in the recent period of declining crime rates, there has been widespread puzzlement about why the declines occurred, and insofar as the drop in crime is a result of the recent period of sustained economic prosperity and dropping crime rates, there is reason to fear that the motivating trends have ended. See, e.g., Andy Newman, Giuliani's Last Crime Report Shows Sharp Drop Despite National Upward Trend, N.Y. Times, Jan. 1, 2002 ("Elsewhere in the country [besides New York City], crime seemed to rebound in 2001. As of mid-December, murders were up by more than 60 percent in Boston and Phoenix and several other big cities, including St. Louis, Houston and Atlanta, which all posted double-digit percentage increases in murders."); Willing, supra note 1 ("The [recent] decline in death sentences has followed a steep drop in the nation's murder rate, which fell nearly 21% from 1996 to 2000. . . . Analysts say public support for capital punishment could begin rising again, if violent crime--which has ticked upward in some citiesCcontinues to rise and Americans feel less secure, particularly in light of terrorism threats.").

    878. See supra pp. 354-56.

    879. See supra pp. 370-72.

    880. See supra pp. 195-96, 244.

    881. See supra pp. 195-96, 244, 322-26.

    882. See supra pp. 157-59, 160-63, 165-66, 322-24, 350, 359-60 & n.756, 362 & n.760.

    883. See supra pp. 70-81.

    884. See supra pp. 392-93.

    885. See supra pp. 195-96, 244, 322-26.

    886. See Ken Hambleton, Researcher: Limit Call for Executions, Lincoln J. Star, Oct. 18, 2001. Professor Baldus proposed that the Nebraska Legislature or Nebraska prosecutors limit capital prosecutions to "cases with two or more aggravating factors." This is similar to proposals made by Professor Baldus, and endorsed by four United States Supreme Court Justices in a dissenting opinion in McCleskey v. Kemp, 481 U.S. 279 (1987) (Stevens, J., dissenting) as a cure for pronounced racial disparities—mainly associated with the race of the victims—in who receives the death penalty in Georgia.

    887. Several of the proposals made here track ones made recently by the Committee to Prevent Wrongful Executions. This Committee is "a nonpartisan panel of judges, former prosecutors and victims advocates," some of them "strong death penalty supporter[s]," that recently "spent a year studying capital punishment in the United States" and made "18 proposals . . . intended to make the death penalty more reliable and less open to chance." Brooke A. Masters, Standards for U.S. Executions Proposed; Nonpartisan Committee Seeks Broad Consensus on Reforming Death Penalty, Wash. Post, June 27, 2001, at A4. Proposals made below parallel the Committee's recommendations that states "[s]et mandatory-minimum standards for defense lawyers and pay them adequately, and make it easier for death row inmates to get new trials if they had bad lawyers"; "[f]ully inform juries by giving them the explicit option of life without parole"; "[p]reserve and test DNA after convictions and make it easier for inmates to get new trials based on newly discovered evidence"; and "urge[] prosecutors to open all their files to the defense before trial." Id.

    For further development of some of the proposals made here in the context of a more comprehensive and specific reform proposal, see James S. Liebman, Opting for Real Death Penalty Reform, 63 Ohio St. L.J. ____ (forthcoming 2002).

    888. See supra p. 4 & n.66; supra note 1. See also Margery Koosed, Averting Mistaken Executions by Adopting the Model Penal Code's Exclusion of Death in the Presence of Lingering Doubt, 21 N. Ill. U.L. Rev. 41 (2001).

    889. See supra pp. 25-35.

    890. See supra pp. 25-35. See also 100 Colum. L. Rev., supra note 153, at 2114-19.

    891. See sources cited in Lempert, supra note 643, at 1243-48.

    892. See, e.g., Boyde v. California, 494 U.S. 370 (1990); Blystone v. Pennsylvania, 494 U.S. 299 (1990).

    893. See Walton v. Arizona, 497 U.S. 639 (1990).

    894. Court Finds Death Penalty Is Misused in Kansas, N.Y. Times, Dec. 30, 2001 ("The Kansas Supreme Court has found that a crucial aspect of the way the state's death penalty is handed down is unfair and must be changed," requiring new sentencing hearings for all death row inmates in the state).

    895. On the need for clear instructions, see Theodore Eisenberg & Martin T. Wells, Deadly Confusion: Juror Instructions in Capital Cases, 79 Cornell L. Rev. 1, 11-12 (November 1993) (finding, based on South Carolina data, that "[j]uror comprehension appears to be worse when mitigating factors are considered"); Craig Haney, The Capital Jury Project: Taking Capital Jurors Seriously, 70 Ind. L.J. 1223,1229 (Fall 1995) (same); see also Stephen P. Garvey, et al, Correcting Deadly Confusion: Responding to Jury Inquiries in Capital Cases, 85 Cornell L. Rev. 627, 637 (March 2000) (reporting that 41% percent of jurors interviewed "erroneously believed that the law required them to impose a death sentence if [at least one aggravating circumstance was present, e.g., if] the evidence proved that the defendant's crime was >heinous, vile, or depraved"); Theodore Eisenberg et al., Jury Responsibility in Capital Sentencing: An Empirical Study, 44 Buff. L. Rev. 339, 361 (1996) (finding, based on interviews of 153 jurors who sat in South Carolina capital cases, that "[n]early one-third of the jurors were under the mistaken impression that the law required a death sentence if they found heinousness or dangerousness"); Study Finds Jurors Confused in Capital Trials, St. Louis Post-Dispatch, Mar. 26, 1995, at Dll (reporting study released by Indiana University School of Law finding that Anearly 43 percent [of jurors] thought they had to impose a death sentence if the crime was >heinous, vile or depraved,' 32.6 percent believed that death was the required punishment if the evidence proved that the defendant posed a future danger to society, [and that] 42 percent mistakenly thought that the jury had to be unanimous before it could decide that a particular factor, such as a defendant's mental retardation or lack of a prior criminal record, justified a sentence other than death."); Joseph L. Hoffmann, Where's the Buck? — Juror Misperception of Sentencing Responsibility in Death Penalty Cases, 70 Ind. L.J. 1137 (1995) (Indiana University Law School study findings).

    896. See supra p. 3 & nn.34-36.

    897. Jost, supra note 1 (quoting Kent Scheidegger).

    898. See supra pp. 3 & n.35, 27-34.

    899. See supra pp. 27-31.

    900/a>.. See Thompson v. Oklahoma, 487 U.S. 815 (1988).

    901. See supra pp. 2-3 & nn. 14, 19, 37.

    902. See, e.g., Johnson v. Texas, 509 U.S. 350 (1993); Eddings v. Oklahoma, 455 U.S. 104 (1982).

    903. See, e.g., the Wallace case in Appendix D. Wallace was convicted and sentenced to die during a period when he was so mentally disordered that he could not understand the proceedings against him or assist his lawyer. After his conviction and death sentence were reversed on this ground and after he was given treatment and restored to mental competence, he was acquitted at a retrial.

    904. For examples, see the Buttrum, Christy, Jones, Jurek, Estelle, Smith, Wade, Wallace, Wilkins, Williams and Williamson cases in Appendix D.

    905. James Q. Wilson, What Death Penalty Errors?, N. Y. Times, July 10, 2000, at A19.

    906. William S. Sessions, Primary Goal Is Justice, Not Execution, San Antonio Express News, July 19, 2000; see also Beck v. Alabama, 447 U.S. 625 (1980) (recognizing that the absence of an appropriate sentencing option at the guilt stage of trial creates the danger of an unwarranted death sentence).

    907. See, e.g., Ramdass v. Angelone, 530 U.S. 156 (2000) (reporting that after the jury delivered a death verdict, members of the jury contacted by petitioner's counsel "expressed the opinion that a life sentence would have been imposed had they known [defendant] would not be eligible for parole"); Simmons v. South Carolina, 512 U.S. 154, 159 & 161 (1994) (plurality opinion) (citing public opinion survey finding that more than 75% of those surveyed indicated that the amount of time a convicted murderer actually would have to spend in prison would be an "extremely important" or a "very important" factor in choosing between life and death, and recognizing that jurors might impose death because they underestimate the legally mandated alternative to death in the case before them); William J. Bowers & Benjamin D. Steiner, Death by Default: An Empirical Demonstration of False and Forced Choices in Capital Sentencing, 77 Tex. L. Rev. 605, 645-48 (1999); Anthony Paduano & Clive A. Smith, Deathly Errors: Juror Misperceptions Concerning Parole in the Imposition of the Death Penalty, 18 Colum. Hum. Rts. L. Rev. 211 (1987).

    908. See supra pp. 352-54.

    909. See Ala. Code 13A-5-42 to -47 (1982); Del. Code Ann. Tit. 11 4209(d) (1979); Fla. Stat. Ch. 921.141 (1991); Ind. Code 35-50-2-9(e) (1986).

    910. See supra p. (Table 18).

    911. For a recent critical examination of Alabama's override system, see Taylor Bright, When a Jury's Choice Doesn't Matter: Judicial Overrides Send Many to Chair; Overrides Viewed as Political Leverage; Critics: Popularity Can Outweigh Justice, Birmingham Post-Herald, Dec. 13, 2001.

    912. See Harris v. Alabama 513 U.S. 504, 519-22 (1995) (Stevens, J., dissenting) (stating that a jury verdict "expresses a collective judgment that we may fairly presume to reflect the considered view of the community"); Spaziano v. Florida, 468 U.S. 447, 487-489 (1984) (Stevens, J., concurring in part and dissenting in part) ("[T]he jury provides a better link to community values than does a single judge . . . , they are more representative institutions than is the judiciary; they reflect more accurately the composition and experiences of the community as a whole, and inevitably make decisions based on community values more reliably, than can that segment of the community that is selected for service on the bench."); Witherspoon v. Illinois, 391 U.S. 510, 519 n.15 (1968) ("The jury also is a significant and reliable objective index of contemporary values."); Gregg v. Georgia, 428 U.S. 153, 181 (1976) (opinion of Stewart, Powell, and Stevens, JJ.) (calling the jury "a significant and reliable objective index of contemporary values"). See generally Harris, 513 U.S. at 521 (Stevens, J., dissenting) ("Death sentences imposed by judges, especially against jury recommendations, sever the critical >link between contemporary community values and the penal system.' They result in the execution of defendants whom the community would spare.") (quoting Witherspoon, 391 U.S. at 519, n. 15).

    913. See Bright, supra note 911 (investigative press report documenting influence of politics on judicial overrides in Alabama).

    914. See Dobbert v. Florida, 432 U.S. 282, 296 (1977).

    915. Statistics reveal that override systems generally result in more judicial overrides of death sentences than non-death sentences. The relevant data for each of the four states are follows:

    Alabama. Between 1982 and July 2000, there were 82 cases in which the judge imposed a death sentence over a jury recommendation of life, compared to 6 cases in which the judge rejected a jury verdict of death. Correspondence from Eva Ansley, Equal Justice Initiative of Alabama, Aug. 13, 2000. See also Harris v. Alabama, 513 U.S. at 522 (Stevens, J., dissenting) (discussing statistics collected by the Alabama Prison Project showing that, as of 1995, there were only 5 cases in which an Alabama judge rejected a jury verdict of death, compared to 47 in which an Alabama judge imposed a death sentence over a jury recommendation of life). Since 1995, "nearly all of the [Alabama] overrides [35 of 36] are from life to death--there have only been a handful [actually one] in the other direction." Correspondence from Ruth Friedman, Equal Justice Initiative of Alabama, July 5, 2000. Alabama has 184 people on death row now, "about a quarter of [whom] got there after a judge overturned a jury life sentence." Id.

    Florida. Between 1972 and early 1992, Florida judges imposed death sentences over 134 jury recommendations of life imprisonment, but overrode only about 51 death recommendations. See Michael Radelet and Michael A. Mello, Death-to-Life Override: Saving the Resources of the Florida Supreme Court, 20 Fla. St. U. L. Rev. 195, 196 & 210-11 (1992). During this same time period, 20% of those sentenced to die had originally received jury recommendations of life. Id. at 196. See also Michael A. Mello, The Jurisdiction to do Justice: Florida's Jury Override and the State Constitution, 18 Fla. St. U. L. Rev. 923, 926 (1991) (reporting that between 1972 and 1988, 1of every 5 death sentences in Florida involved an override of a jury's life recommendation). In Boyett v. State, 688 So. 2d 308 (Fla. 1996), the Florida Supreme Court discouraged the practice of overriding life recommendations. Since 1995, there have been seven overrides in Florida, all overrides of life verdicts. Correspondence from Michael Radelet, July 5, 2000.

    Indiana. In Indiana, between 1980 and 2000, judges used overrides to impose 10 death sentences, compared with 9 life sentences. Correspondence from Monica Foster, Defense Attorney, Indianapolis, Indiana, July 18, 2000. See also Harris, 513 U.S. at 522) (Stevens, J., dissenting) (citing data from the Indiana Public Defender Council, reporting that between 1980 and early 1994, judges had used overrides to impose eight death sentences, and four life sentences); Indiana Death Row Statistics, avail. at http://www.clarkpro-secutor.org/html/death/rowstats.ghtm. Note that in Indiana, unlike in other states, all jury recommendations must be unanimous. In other words, both a life and a death determination require the concurrence of all 12 jurors. If the jury is hung even by one vote, the judge sentences as if sentencing had been to the court alone (meaning that all available sentencing options are still on the table). See Ind. Code 35-50-2-9(e) (1986). The relatively even balance of "life" and "death" overrides in Indiana may be a function of the discipline imposed both by the Indiana Supreme Court's insistence on deference to the jury's verdict absent clear error, and by the unanimity requirement, which puts judges contemplating override in the position of disagreeing with all 12 jurors.

    Delaware. Since 1991, there have been seven judicial overrides of jury death verdicts in Delaware, and no overrides of jury life verdicts. Correspondence from Kevin O'Connell, Office of the Public Defender, Wilmington, Delaware, August 11, 2000. Delaware's unique resistance among override states to overrides of life recommendations may be attributable both to the Delaware Supreme Court's insistence on strong deference to jury recommendations of life (as in Florida and Indiana) and to the relatively nonpolitical nature of judicial selection in Delaware, as opposed to the other override states. See Correspondence from Kevin O'Connell, Office of the Public Defender, Wilmington, Delaware, July 5, 2000. See also Fred B. Burnside, Comment: Dying to Get Elected: A Challenge to the Jury Override, 1999 Wis. L. Rev. 1017, 1043 (same) (citing telephone interview with Nan Perillo, Attorney, Delaware Public Defenders Office (Feb. 12, 1999)). One of the drafters of Delaware's override statute is reported to have said: "We would have been substantially more reticent to change to judicial sentencing if judges were elected." Burnside, supra at 1043 (citing telephone interview with Steve Wood, Chief Prosecutor, New Castle County, Delaware (Feb. 16, 1999)).

    916. As the data in note 915 above reveal, the disproportionate use of jury overrides to impose death sentences (1) is extremely pronounced in states that give trial judges essentially unfettered discretion to overturn jury votes for life the case for any or no reason at all, and (2) is stronger in states where judges are subject to direct election. In the first regard, compare Florida before 1995 and Alabama throughout the modern death-sentencing era (when both states gave trial judges relatively free reign to override jury verdicts), to Florida since 1995 and to Indiana and Delaware (which have relatively strict standards governing when a judge may override a jury verdict).

    In the second regard, compare override results in Alabama, Florida and Indiana, where judges face election, and overrides of life sentences are generally far more common than overrides of death sentences, to Delaware, where judges are not elected, and the few overrides that occur are all in the direction of life; no Delaware judge has ever overridden a jury's majority verdict for life. Even more specifically, as our analyses would predict, there "appears to be a sliding scale of jury override use depending not only on [whether] elections [are used to select judges], but the type of election." Burnside, supra note 915, at 1049. Judges in Alabama, which has partisan elections and places no constraints on a judge's discretion to override life verdicts, override almost 10 jury life-sentence recommendations for every vetoed death sentence recommendation; Florida, a state with non-partisan retention elections, has an approximate 3-to-1 ratio in favor of overriding juries' recommendations for life sentences; and Indiana, also with non-partisan retention elections, has a ratio close to 1-to-1 that only slightly favors overrides of juries' recommendations for a life sentence. Id. at 1043.

    The Table below compares the four override states based on the percentage of judge overrides that imposed death sentences in cases in which jurors voted for life (as opposed to overrides imposing life sentences in cases in which the jury imposed death), and on the states' scores on our two indexes of the

    political pressure judicial selection techniques place on judges (with higher scores indicating more political pressure). As our analyses would predict, judges in states where judicial selection methods make them more politically vulnerable have a greater propensity to use overrides to impose additional death verdicts.

    State

    % Judge Overrides Imposing Death

    Score on 1st Political Pressure Index

    Score on 2nd Political Pressure Index

    Alabama, 1982-2000

    93% (82/88)

    8

    7

    Delaware, 1991-2000

    0% (0/7)

    3

    3

    Florida, 1972-2000*

    73% (141/192)

    7

    7

    Indiana, 1980-2000

    53% (10/19)

    8

    5

    In Alabama, there also "is a statistically significant correlation between [the likelihood of] judicial override [in favor of death] and election years in most of the counties where these overrides take place." Symposium, Politics and the Death Penalty: Can Rational Discourse and Due Process Survive the Perceived Political Pressure?, 21 Fordham Urb. L.J. 239, 256 (1994).

    917. The Florida experience documents the risk of error in override cases and the burden excessive overrides place on appellate courts. See Gerald B. Cope, Jr., Discretionary Review of the Decisions of Intermediate Appellate Courts: A Comparison of Florida's System with those of the other States and the Federal System, 45 Fla. L. Rev. 21, 99-100 (1993) (finding that jury recommendations of life constitute the vast majority of override cases, and that the Florida Supreme Court usually reverses death sentences imposed by judges contrary to the life recommendation of a jury, either based on a conclusion that the override was improper or due to other errors in the case; also reporting that in 1991, the reversal rate for Florida capital verdicts imposed following overrides of life verdicts was 91%, and concluding that based on these high reversal rates, eliminating override cases would reduce the death penalty workload of the Florida Supreme Court by 21% and the Court's overall workload by 6 to 8%). See also Gary Caldwell, Capital Crime Decisions: 1992 Survey of Florida Law, 17 Nova L. Rev. 31, 64 n. 261 (1992) (reporting that from 1986 through 1992, the Florida Supreme Court upheld death sentences in only seven cases where judges overrode life verdicts).

    918. See supra pp. 321-21, 336, 341, 385.

    919. See supra pp. 22-23, 323-24 & n.647.

    920. See, e.g., . Tennessee article on forgone review. John Shipman

    921. See supra pp. 394-95.

    922. State Law Seeks to Provide Strong Defense; Most Capital Cases Are Handled Well, But Examples of Inadequate Representation Show Lapses in the System, South Bend Trib., Oct. 22, 2001 (quoting Professor Karlson). According to Karlson, financial savings are not the only reason to spend more time and energy assuring the accuracy of capital verdicts from the outset: "While prosecutors . . . complain about [spending money on capital defense], Karlson said it is imperative to prevent an innocent person from being executed. Such a mistake, he reasoned would lead to fewer executions of defendants deserving to die." Id.

    923. See, e.g., sources catalogued in 100 Colum. L. Rev., supra note 123, at 2078-94. See also Lise Olsen, One Killer, Two Standards, Seattle Post-Intelligencer, August 7, 2001 (noting, in examining potential causes of problems in Washington state capital cases, that "[p]rosecutors may seek the death penalty before getting a full picture" because "Washington has no standardized process to guide prosecutors in reviewing capital cases," leaving them free to seek "death sentences quickly, after brief conversations with defense attorneys," rather than following the practice elsewhere of using "trained teams to review lengthy documents detailing the defendant's upbringing, criminal history and mental health".

    924. See, e.g., sources collected in 100 Colum. L. Rev., supra note 123, at 2078-94.

    925. See, e.g., sources collected in id at 2144-45.

    926. Lise Olsen, 1 Killer, 2 Standards, Sidebar: Solutions, Seattle Post-Intelligencer, Aug. 7, 2001 (listing reforms reportedly proposed by state's Attorney Birkett).

    927. See supra p. 6.

    928. See supra pp. 40-41.

    929. See supra p. 71.

    930. The standard allows police and prosecutors to withhold evidence of innocence or mitigation and evidence undermining the state's case of guilt or aggravation, if in their judgment there is no "reasonable probability" that but for their suppression of the evidence, "the outcome of the trial"—which has not yet taken place—"would be different." See, e.g., Strickler v. Greene, 527 U.S. 263 (1999).

    931. See supra pp. 319-24, 349-50.

    932. See supra note 930.

    933. See supra pp. 405-06; infra p. 418.

    934. See supra pp. 354-56.

    935. See supra pp. 16 & n.100, 194, 198-99, 212, 217-19, 227, 230, 236, 240 & n.486, 257, 382-84.

    936. See Ring v. Arizona, 122 S. Ct. 865 (2002) (granting certiorari to resolve apparent conflict between Jones v. United States, 526 U.S. 227 (1999) and Apprendi v. New Jersey, 530 U.S. 466 (2000), on the one hand, and Walton v. Arizona, 497 U.S. 639 (1990), on the other hand, over whether the Constitution requires a jury finding of the prerequisites for a capital as opposed to a lesser sentence). On the greater insulation of jurors than judges from political pressure, see supra pp. 405-06 & nn.911-16.

    937. On Indiana, see Norman Lefstein, Reform of Defense Representation in Capital Cases: the Indiana Experience and its Implications for the Nation, 29 Ind. L. Rev. 495, 496-504, 506-07, 509-12, 518-26, 533 (1996) (discussing effect of Indiana's early 1990s adoption of legislation making state funds available to local jurisdictions that satisfy a state commission's guidelines for appointment of qualified counsel in capital cases, and the commission's incorporation within its guidelines of a state supreme court rule (1) requiring the appointment of two lawyers in capital cases with recent extensive training in capital defense and with, respectively, at least five and three years criminal litigation experience that includes at least five and three felony jury trials, (2) disqualifying lawyers with excessive workloads, (3) setting minimum hourly rates that are relatively generous, though they remain well below the rates prevailing among retained attorneys, and (4) assuring ">adequate funds for investigative, expert, and other services necessary to prepare and present an adequate defense at every stage of the proceeding, including the sentencing phase'"; noting that prosecutorial requests for death sentences dropped from 23/year in the two years before, to 10/year in the three years after, the adoption of these reforms and reporting agreement among state capital prosecutors and defense counsel interviewed before the data were available that the reforms (1) had improved the quality of capital defense lawyering in the state, especially by increasing the use of expert witnesses at the mitigation phase, (2) had attracted more and better defense lawyers to the work, (3) had probably generated better police and prosecutorial preparation and decreased the likelihood that the resulting (smaller number of) capital judgments would be reversed on appeal, and (4) in the words of prosecutors "ha[d] >definitely put a damper on [their] asking for the death penalty," ">put some economic judgment into the decision-making about whether to seek the death penalty" and made them A>think two or three times' before filing a death penalty request@ not only because of the greater cost of trying cases but also because of the increased "risk [of] losing"; comparing the Indiana experience to that of Ohio, which adopted similar reforms but compensates defense lawyers at only two-thirds the rate in Indiana and provides funds for expert witnesses and mitigation specialists far less frequently than in Indiana and has experienced smaller declines in the death-sentencing rate; concluding that there is "strong[]" reason to believe that the "ability of defense counsel, the cost of the prosecution [which in turn is affected by the quality and resources of defense counsel], and the burden on the prosecutor's staff" affect prosecutorial charging decisions in capital cases). See also Ten Who Have Faced the Death Penalty in Indiana Have Been Found Not Guilty, Evansville Courier & Press, Oct. 23, 2001 ("There have been few cases resulting in acquittals in recent years because prosecutors are exerting more caution in seeking the death penalty and thus filing it less often, both defenders and prosecutors agree.").

    On Oklahoma, see supra note 38.

    938. See, e.g., supra p. 32 & n.143.

    939. See supra pp. 40-42.

    940. See supra pp. 199-200, 370-72.

    941. See, e.g., Paul Barrett, Lawyer=s Fast Work on Death Cases Raises Doubts About System, Wall St. J., Sept. 7, 1994, at A1; Vivian Berger, The Chiropractor as Brain Surgeon: Defense Lawyering in Capital Cases, 18 N.Y.U. L. Rev. L. & Soc. Change 245 (1990-91); Stephen B. Bright, Counsel for the Poor: The Death Sentence Not for the Worst Crime but for the Worst Lawyer, 103 Yale L.J. 1835 (1994); Marcia Coyle, et al., Fatal Defense: Trial and Error in the Nation=s Death Belt, Nat'l L.J., June 11, 1990, at 30; Paul Duggan, Attorneys' Ineptitude Doesn't Halt Executions, Wash. Post, May 12, 2000, at A1; Ruth E. Friedman & Bryan A. Stevenson, Solving Alabama's Capital Defense Problems: It's a Dollars and Sense Thing, 44 Ala. L. Rev. 1 (1992); Gary Goodpaster, The Trial for Life: Effective Assistance of Counsel in Death Penalty Cases, 58 N.Y.U. L. Rev. 299 (1983); Dirk Johnson, Shoddy Defense by Lawyers Puts Innocents on Death Row, N.Y. Times, Feb. 5, 2000, at A1; Anthony Paduano & Clive Stafford Smith, The Unconscionability of Sub-Minimum Wages Paid Appointed Counsel in Capital Cases, 43 Rutgers L. Rev. 281 (1991); Albert v. Vreeland, II, Note, Statutory Fee Limitations and Ineffective Assistance of Counsel in Capital Litigation, 90 Mich. L. Rev. 626 (1991); other sources cited infra notes 946-61.

    942. Sources suggesting a link between being poor quality representation and an increased probability of a death sentence include, e.g., Associated Press, Ohio County Paying Defense Counsel the Least Sends the Most to Death Row, Feb. 3, 2000 ("Hamilton County [Cincinnati], which sends more people to death row than any other county in [Ohio], pays public defenders less to represent those people than" all but one of Ohio's other 88 counties; its hourly "rate of $30 an hourC57% below the state averageCis the same whether the case is a minor felony such as theft or a death penalty case"); Alan Berlow, Lethal Injustice, The American Prospect, Mar. 27-Apr. 10, 2000 (describing a study done for the Texas Judicial Council in the mid-1980s, which found that a defendant's chances of being convicted of murder in the state were 28% higher if his or her attorney was court-assigned and that 79% of capital defendants with appointed counsel were sentenced to death, compared to 55% of capital defendants with retained lawyers); William J. Bowers, The Pervasiveness of Arbitrariness and Discrimination under Post-Furman Capital Statutes, 74 J. Crim. L. & Criminology 1067, 1075-77 (1983) (describing interviews with Florida prosecutors revealing that the quality of the defense lawyer is an important factor in their decision whether to accept a noncapital plea in cases charged capitally); Dead Man Walking Out, The Economist, June 10, 2000, at 21 (describing a 1999 study finding that criminal defendants in Houston with court-appointed lawyers were twice as likely to go to prison as those who were able to hire their own counsel); Michael L. Radelet, Rejecting the Jury: The Imposition of the Death Penalty in Florida, 18 U.C. Davis L. Rev. 1409, 1414-15 (1985) (describing a Florida study, which found that 40% of defendants who had private counsel received life sentences at capital trials, compared to 23% of capital defendants with public defenders or appointed attorneys); Tina Rosenberg, Deadliest D.A., N.Y. Times, July 16 (1995), Magazine, at 46, 50 (comparing capital representation by appointed lawyers who handle close to 80% of Philadelphia capital cases for a flat fee of $1700 plus $400 for each day in court and $300 for an investigator, with an average cost to the county in 1995 of $2700 per capital case, to representation by the local public defender office in the one in five cases in which that elite office is permitted to represent capital defendants, which provides two attorneys, a mitigation specialist, an investigator, and access to a staff psychiatrist and expert witnesses; although a large proportion of Pennsylvania's death row is made up of Philadelphia defendants who were represented at trial by appointed attorneys, not a single defendant represented by the elite public defender office received the death penalty in 1993-95 [or, in fact, from 1993 to early 2000]). See also Editorial, Rising Doubts on Death Penalty, USA Today, Dec. 22, 1999, at 17A ("You are more likely to receive the death penalty if you are . . . poor.").

    943. A "common thread" in the cases of innocents freed from death row is "poorly financed, often incompetent defense lawyers who failed to uncover and present key evidence." Dirk Johnson, Shoddy Defense by Lawyers Puts Innocents on Death Row, N.Y. Times, Feb. 5, 2000. See, e.g., Ken Armstrong & Steve Mills, Inept Defenses Cloud Verdicts, Chi. Trib., Nov. 15, 1999 (four of 13 former death row inmates who were exonerated and released were represented at trial by lawyers who have had their licenses suspended or withdrawn); Benjamin Weiser, Judge Overturns Verdict in 1980 Murder, N.Y. Times, Nov. 12, 1999, at B4. A recent study revealed that ineffective assistance of counsel was a contributing factor in 27% of the American cases in which wrongful convictions were uncovered using DNA evidence. See Vivian Berger, Actual Innocence: Five Days to Execution, and Other Dispatches from the Wrongly Convicted (Book Review), N.Y. L.J., May 1, 2000.

    944. Unlike 9 out of 10 non-capital charges, which are settled after a few hours of plea bargaining, most capital cases require full investigations and a full-blown trial:

    For documentation of these points, see the authorities cited supra note 941.

    945. See sources cited infra note 947.

    946. See, e.g., American Bar Association, Toward a More Just and Effective System of Review in State Death Penalty Cases, 40 Am. U. L. Rev. 1, 16 & 69 (1990) (detailed study concluding that "inadequate compensation of counsel at trial" is one of the "principal failings of the capital punishment systems in the states today"); supra note 942 (Hamilton County, Ohio).

    947. See, e.g., Coyle et al., supra note 941, at 30 (AWholly unrealistic statutory fee limits on defense representationCsuch as Mississippi's flat, unwaiveable $1,000 cap [on compensation for capital defense lawyers], equivalent to a fee of about $5 per hour for many lawyers [a provision that was still in effect as late as March 2000, see Sara Rimer, Questions of Death Row Justice for Poor People in Alabama, N.Y. Times, Mar. 1, 2000, at "16Cact as disincentives to thorough trial investigation and preparation."); Friedman & Stevenson, supra note 941 (criticizing Alabama's built-in monetary disincentive--maximum compensation of $20 per hour for any work done out of court and $40 per hour for in-court activity, with a $1000 reimbursement capCagainst thorough representation at the trial level); Johnson, supra note 943 (documenting that capital defendants "are often represented by lawyers who are paid a few thousand dollars, or less, and spend only two days on a case" and that a proper defense in a death penalty case takes months of research and costs $250,000 or more); Jeb Phillips, Justice at 50 cents an Hour: Defending Death Row Case Drove Lawyer into Bankruptcy, Birmingham Post-Herald, Dec. 14, 2001; Rosenberg, supra note 942, at 21, 46, 50 (comparing capital representation by appointed lawyers who handle close to 80 percent of Philadelphia capital cases for a flat fee of $1700 plus $400 for each day in court and $300 for an investigator, with an average cost to the county in 1995 of $2700 per capital case, to the rare representation by a retained lawyer for whom the going rate in Philadelphia is $50,000 per case); Stan Swofford, A Reasonable Doubt: Are There Innocent People on North Carolina's Death Row, Greensboro News & Rec., Aug. 6, 2000, at A1 (comparing North Carolina's $85 per hour cap on compensation for defense attorneys appointed to represent indigent capital defendants, to the going rate of $200 or more per hour for such representation by experienced retained criminal defense lawyers in the state).

    948. See Jonathan Alter, The Death Penalty on Trial: Special Report, Newsweek, June 12, 2000 ("Texas provides only $2,500 for investigators and expert witnesses in death penalty cases (enough for one day's work, if that).").

    949. Bruce A. Green, Lethal Fiction: The Meaning of "Counsel" in the Sixth Amendment, 78 Iowa L. Rev. 433, 491-92 (1993). See also Rimer, supra note 947 (quoting an Alabama lawyer who was appointed to represent a capital defendant at trial at a maximum $20 per hour for out of court work, with a $1000 compensation cap, and who consequently ended up being reimbursed at about $5 an hour for the work he put in, vowing that "I will go to jail before I handle another capital case").

    950. James Coleman, Testimony in Support of Title II of the Innocence Protection Act of 2000, United States House of Representatives, June 21, 2000. In a recent report, the American Bar Association described a variety of disturbing techniques that under-funding has forced state appointing officials to use to provide defense representation in capital cases, including patronage selections off a general list of all local attorneys, regardless of capital, or even criminal, experience; contract systems under which all cases over a particular period go to the lowest bidder (with a flat fee bid covering all experts and other expenses), including complex and unanticipated capital cases that suddenly appear on the county's docket; reimbursement schemes that limit lawyers to, e.g., $2500 for the entire representation Aplus $50 for each motion . . . filed up to five motions--with the result that the number of motions filed in almost every case is exactly five--or $1000, including expenses for expert and investigative assistance; or what amounted to "$15 to $20 per hour" and "$11.84 per hour" to represent two innocent men who were sentenced to die but were eventually released for lack of evidence of guilt. Randall Coyne & Lyn Entzeroth, Report Regarding Implementation of the American Bar Association's Recommendations and Resolutions Concerning the Death Penalty, 4 Geo. J.F.P. 253 (1996) (hereinafter AABA Report"). For other sources who have comprehensively documented the limited resources available for capital defense, see, e.g., Coyle, supra note 941; Lefstein, supra note 937; Joe Margulies, Resource Deprivation and the Right to Counsel, 80 J. Crim. L. & Criminology 673 (1989); Michael D. Moore, Analysis of State Indigent Defense Systems and their Application to Death-Eligible Defendants, 37 Wm. & Mary L. Rev. 1617 (1996); Lise Olsen, The Death Penalty: Uncertain Justice, Seattle Post-Intelligencer, Aug. 6, 2001 ("Judges help create the problem [of "inept" capital defense lawyers] by appointing inexperienced local lawyers to capital cases instead of those recommended by the state. Counties often pay these defenders so poorly that they cut corners, putting convictions and sentences on shaky legal ground."); Douglas W. Vick, Poorhouse Justice: Underfunded Indigent Defense Services and Arbitrary Death Sentences, 43 Buff. L. Rev. 329 (1995).

    951. See, e.g., Barrett, supra note 941; Lawyer's Fast Work on Death Cases Raises Doubts About System, Wall St. J., Sept. 7, 1994; Mary Flood, What Price Justice? Gary Graham Case Fuels Debate Over Appointed Attorneys, Hous. Chron. July 1, 2000, at A1 (stating that in some Texas counties in the 1970s and 1980s, "courthouse appointment lists were often an informal string of each judge's friends and campaign contributors; some were competent and trained, others were not[; t]here were no requirements for experience, no classes, no tests").

    952. See, e.g., Armstrong & Mills, Inept Defenses, supra note 943 (finding that 33 defendants sentenced to die in Illinois were represented by a lawyer who has been disbarred or suspended); Berlow, supra note 942, at 82 (discussing Kentucky study showing that 25 percent of state's death row inmates had been represented by attorneys who had since been disbarred or had resigned to avoid disbarment, and Louisiana study showing that lawyers of inmates executed in state had bar discipline rate 68 percent higher than bar members as a whole); Liz Chandler, Lawyers, Inadequate Defense Cited in a Third of Death Case Reversals, Charlotte Observer, Sept. 11, 2000, at 1A ("Since 1977, when the Carolinas restored capital punishment, at least 15 death verdicts have been overturned because of poor lawyering at trial. And at least 16 other death row inmates-including three who were executed-were represented by lawyers who have been disbarred or disciplined for unethical or criminal conduct"); Defense Called Lacking for Death Row Indigents, Dallas Morning News, Sept. 10, 2000, at 1A (reporting, based on examination of 461 Texas capital cases, Athat nearly one in four condemned inmates has been represented at trial or on appeal by court-appointed attorneys who have been disciplined for professional misconduct at some point in their careers," in "about half" of which cases, "the misconduct occurred before the attorney was appointed to handle the capital case"); Lise Olsen, The Death Penalty: Uncertain Justice, Seattle Post-Intelligencer, August 6, 2001 ("Washington State authorizes the death penalty, but does little to ensure that defendants are represented equally. Nearly 1/5 of the men to face execution were represented by lawyers who had been, or were later, suspended or arrested. In the past 20 years, 84 people in this state have faced the possibility of capital punishment. Despite the gravity of the cases, some defendants were given inexperienced or inept counsel.").

    953. See, e.g., Liz Chandler, With Lives on Line, Attorney Turned to Drink, Charlotte Observer, Sept. 11, 2000, at 6A; Dan Malone & Steve McGonigle, Questions of Competence Arise in Death Row Appeal: Lawyer With History of Problems Defends Handling of Case, Dallas Morning News, Sept. 11, 2000, at 1A.

    954. See, e.g., Rimer, supra note 947 (discussing capital defense situation in Alabama, which "has no state-wide public defender system" and relies on "[c]ourt-appointed capital defense lawyers [who] are paid so little that many lawyers refuse the work[,] . . . leaving many of those charged with capital offenses with inadequate representation" by "lawyers who fail to do the most fundamental tasks, like investigating the crime and their clients' backgrounds and presenting closing arguments"); Henry Weinstein, a Sleeping Lawyer and a Ticket to Death Row, L.A. Times, July 15, 2000, at A1.

    955. See, e.g., Michele McQueen & Ted Koppel, Crime and Punishment, Poor Counsel, Nightline, ABC News, Feb. 3, 2000; Margulies, supra note 950, at 677-82 (citing burdensome workloads as a result of under-funding and under-staffing as the "single greatest obstacle to effective representation" in capital and other cases). See also Stephen Bright, Neither Equal Nor Just, 1997 Ann. Survey of Am. L. 783, 817 (1997) (documenting defense attorney caseloads in particular states).

    956. See, e.g., Rimer, supra note 947 (blaming low maximum hourly rates, and $1000 fee cap, for fact that (1) appointed capital defense lawyers in Alabama "often" limit themselves to 50 hours or less on each capital case, "even though adequate preparation . . . should take 500 to 1,000 hours," and (2) lawyers who attempt to put in the necessary hours end up working "at a rate of about $5 an hour").

    957. See, e.g., Jeanne Cummings, Bad Lawyers Tip the Scales of Justice Toward Death Row, Atlanta J. Const., Apr. 1, 1990, at A1; Paul Duggan, in Texas, Defense Lapses Fail to Halt Executions, Wash. Post, May 12, 2000, at A1 (reporting on instances in which defense lawyers in capital trials have slept through key testimony, failed to file crucial legal papers correctly or on time, or had been cited for professional misconduct repeatedly in their careers); Editorial: Rush to Death, St. Petersburg Times, Feb. 10, 2000, at 16A (discussing Florida case in which a "young, inexperienced lawyer who knew little about presenting a capital defense," had his fee capped at $2500 for pretrial and trial proceedings, "was paid the equivalent of $13 an hour" for what little he did, and failed to present any witnesses to describe his client's "long history of mental illness" because (as the lawyer later testified in court) "he could not afford to call witnesses"); Bill Freivogel, Bad Lawyers: Rx for Death, St. Louis Post-Dispatch, May 4, 2000; ; Johnson, supra note 943 (citing accounts from around the nation of defense lawyers who slept through trials, came to court drunk, and had never before tried a criminal case).

    958. Coopers & Lybrand Consulting, Report on Costs and Recommendations for the Control of Costs of the Defender Services Program, Jan.28, 1998 (reporting on interviews with a series of judges) (cited in Statement of Judge William Terrell Hodges, Chairman, Executive Committee, Judicial Conference of the United States, June 11, 1998); see also Editorial, No Money for Justice: State Will Help Prosecute Death-Penalty Cases but Falls Short in Ensuring a Fair Defense, Phil. Inquirer, Mar. 8. 2000, at A22 (discussing Pennsylvania Legislature's defeat of bill to increase money for defense of capital case in same session in which it appropriated additional funds to assist local prosecutors in such cases).

    The problems described here extend beyond cases of appointed counsel to ones in which impecunious defendants, who are fearful of the consequences of accepting an appointed attorney, retain private counsel at fire-sale prices, "often settl[ing], in desperation, for any lawyer who will take a case, regardless of reputation." Johnson, supra note 943.

    959. Testimony of Professor Lawrence C. Marshall, Director, Northwestern University Center on Wrongful Convictions, to Illinois legislative committee, Sept. 15, 1999, at 10. See also Armstrong & Mills, Inept Defenses, supra note 943 (reporting that appointed lawyers "in Illinois capital cases have included a tax lawyer who had never before tried a case, an attorney just two years out of law school and an attorney just 10 days off a suspension for incompetence and dishonesty"; the assigned lawyer two years out of law school handled the case himself, though he had never before tried a murder case, was carrying 100 other appointed criminal cases, and was denied funds for an investigator and experts (by contrast, "the local prosecutor received help trying the case from a lawyer in the Illinois attorney general's office, a common [practice]").

    960. Martinez-Macias v. Collins, 810 F. Supp. 782, 788-813 (W.D. Tex. 1991), aff'd, 949 F.2d 1067 (5th Cir. 1992).

    961. Martinez-Macias v. Collins, 949 F.2d at 1067 (emphasis added). For another example, see Olsen, Defense for Capital Crimes Often Done on the Cheap, Seattle Post-Intelligencer, Aug. 8, 2001 (discussing representation of James Leroy Brett by a contract attorney in Clark County, Washington, which caps pretrial compensation in capital cases at $12,500, or between "$12.50 to $25 an hour, based on the 500 to 1,000 hours of pretrial work experts say the cases usually require"; Brett's death sentence was subsequently overturned because of egregious ineffective assistance of counsel in failing to prepare the case properly, costing the state $254,889 in appellate expenses).

    962. Here again, we agree with the President of the Association of Government Attorneys in Capital Litigation, Joe Birkett, that states and counties should "[i]nsist on a quality defense" in capital cases, because "[s]potting problems early saves time and money on appeals years later." Olsen, Solutions, supra note 926. See also Editorial, Seattle Post-Intelligencer, Aug. 12, 2001 ("A system of death-penalty defense on the cheap not only disadvantages the person with the most to lose, it costs the public in the long run. Front-loading capital cases with experienced, decently paid attorneys with adequate budgets for pre-trial discovery helps ensure that fewer verdicts will be overturned because of bad lawyering. Those years of costly appeals also delay closure for victims of capital crimes and frustrate the public will.").

    963. See supra pp. 194-95, 202, 258-59, 263, 372-76.

    964. See supra pp. 65-67 & Figure 5, 254-46.

    965. See Gallegos, supra note 44:

    966.See supra pp. 377-82.

    967 See Editorial, Ryan's Tough Call on Executions, Chi. Trib., Aug. 28, 2001 (endorsing veto of legislation adding new aggravating circumstance to Illinois capital statute for "gang-related killings," in part because "[a]dding gang-related murder as a so-called 'aggravating factor' in sentencing would lengthen Illinois' long list of what makes a person eligible for the death penalty and invite more arbitrariness in the judicial system."; noting that "[w]hen Illinois restored capital punishment, it started out with a handful of factors that qualified a defendant" but "[n]ow there are 20 such factors"; recommending that the state "ought to narrow the eligibility list to something resembling the original"). See also Christi Parsons & Ray Long, House Won't Push Death Penalty Bill, Chi. Trib., Nov. 16, 2001 (AHouse Speaker Michael Madigan said Thursday there will be no attempt to override Gov. George Ryan's veto of a bill that would have made gang activity one of the aggravating factors that qualifies convicted murderers for the death penalty. . . . "I don't think there's any question the administration of the death penalty has not been well-done," Madigan said.").

    968 See Olsen, Solutions, supra note 926 (listing Birkett's reform proposals).

    969 See supra pp. 379-81.

    970 See supra pp. 14-35.

    971 See supra pp. 79-80 & Figure 9.

    table of contents