next | previous | table of contents

D. A System that Does Not Work, and Is Fraught with Costly and Serious Error

Throughout the 23—year study period—and evidently since—serious error has not been just an attribute of the death penalty system. It has been the system's defining trait:

  • Reversible error—the only error counted in our study—is serious error. By legal definition, reversible error affects the reliability of the outcome. In actual fact, where the facts are known, reversible error produces capital outcomes that do not hold up on retrial in over 4 cases out of 5, and are replaced by acquittals nearly 1 time in 10.218

  • But not all serious error is reversible error, so that by counting only reversible error, we under-count serious error. Courts often miss error, or spot it but ignore it as harmless, not prejudicial or waived—even when doing so approves innocent people to be executed.219

  • Even though we do not count all error, but only reversible error, and even given that our judicious methods substantially understate the amount of such error, state and federal courts still found reversible error in just under seven-tenths of all death verdicts imposed and reviewed in the 23-year study period.220

  • Reversible capital error is chronic across place and time. All but two states with at least one case that progressed through the entire review process had rates of reversible error greater than 50%. And error rates higher than 50% were discovered in death verdicts reviewed in all but two of the 23 study years.221

  • High reversal rates persist from the first to the last review stage, creating a high risk that the review process cannot catch all serious error that infects death verdicts.222

  • This risk is a reality: Not only have trial courts repeatedly sentenced innocent capital defendants to die, but all three stages of reviewing courts have repeatedly approved innocent capital defendants for execution.223 Still more defendants have been approved for execution though the law bars death as a penalty for their crimes.

Capital trials make too many serious errors, and capital appeals miss too many of those mistakes, to satisfy any reasonable definition of a system that works. The system is broken.

Typically, capital error is reversible only because it is serious, undermining the reliability of death verdicts. In addition, error is serious because it causes reversal. When, as has been true for decades, reversal is the rule, not the exception, it devastates the system. The most obvious way citizens and taxpayers measure the success of the death penalty that they for the most part support and pay dearly to operate224 is to ask how often imposed death sentences are carried out. As a result of chronic error, that number is lower than most citizens think, and lower than we have suggested so far:

  • Of death sentences fully reviewed during the 23—year study period, 68% were overturned and sent back for a retrial—meaning only 32% of the death verdicts fully reviewed during the period were found fit to be carried out.

  • But nothing like 32% of the nearly 6000 death verdicts imposed during the period were carried out.

    • The 32% applies only to death verdicts fully reviewed in the study period. But the arduous review process required by high error rates—and the stifling review burden on those courts—means that cases proceed through the courts slowly: over 9 years on average from death sentence to execution during the entire study period; just under 11 years on average in the latter half of that period; and around 12 years for executions occurring in the last few years. At any given time, therefore, most death verdicts are stuck in the review process, meaning the 32% of fully reviewed verdicts approved for execution are a much smaller proportion of all verdicts.

    • As we note above, we include in the 32% of so—called successful verdicts—ones approved for execution by all three levels of review—some that were subsequently never carried out because it was discovered by others that the prisoner was innocent.225 Thus, a significant number of even the 32% of fully reviewed verdicts that our study counts as error—free are in fact seriously flawed and never carried out.

  • Thus, 32% is not a valid measure of the system's success rate. Instead, our data show that:

    • Fewer than 8% of all death verdicts known to have entered the review process from 1973 to 1995—358 out of 4546—were approved for execution by all three sets of reviewing courts. The rest were overturned or got stuck in the system.

    • Only 6% of all death sentences imposed during that period—358 out of 5826— were approved for execution. The rest either were overturned, mired in the review process or waiting to enter that process.

    • Only 5% of all death verdicts imposed during the period—313 out of 5826— were carried out in the period. Figure 6, p. 73 below, compares the states on this most basic measure of success: Success is in black, failure in white. Figure 6 speaks for itself.

    • As Table 1 and Figures 7A—7C show (pp. 74—75 below), the result of a capital system that for decades sentenced between 250 and 325 people to die annually, that requires an elaborate and overburdened review process to catch its many mistakes, and that can make only a tiny fraction of those sentences stick, is a huge death row population (now about 3700), and an ability to execute at most only 2.7% of those inmates in a given year—with an average of only 1.2% per year since 1981, and with only 1.8% of death row being executed in 2001.

    • States with the death penalty executed about two-thirds of one percent of all their homicide offenders during the 23-year study period.

Figure 6. Percent Death Verdicts Carried Out (Nonconsensual Executions), 1973-95

Table 1: Death Row Population, Executions and Percent of Death Row Executed, 1973—2001

Year Death Row Population Executions Non-Consensual Executions % of Death Row Executed % of Death Row Executed,Non-Consensual Executions
1973 376 0 0 0 0
1974 283 0 0 0 0
1975 542 0 0 0 0
1976 721 0 0 0 0
1977 557 1 0 0.18 0
1978 610 0 0 0 0
1979 635 2 1 0.31 0.16
1980 769 0 0 0 0
1981 926 1 0 0.11 0
1982 1131 2 1 0.18 0.09
1983 1327 5 5 0.38 0.38
1984 1499 21 21 1.40 1.40
1985 1689 18 14 1.07 0.83
1986 1888 18 17 0.95 0.90
1987 2089 25 23 1.20 1.10
1988 2255 11 10 0.49 0.44
1989 2374 16 14 0.67 0.59
1990 2484 23 16 0.93 0.64
1991 2610 14 14 0.54 0.54
1992 2755 31 30 1.13 1.09
1993 2866 38 31 1.33 1.08
1994 3037 31 27 1.02 0.89
1995 3212 56 49 1.74 1.53
1996 3381 45 37 1.33 1.09
1997 3516 74 70 2.10 1.99
1998 3613 68 58 1.88 1.61
1999 3625 98 87 2.70 2.40
2000 3711 85 78 2.29 2.1
Sources: BJS 2001 Cap. Pun. Study (death row pop.); Death Row U.S.A., Fall 2001 (executions).

Figure 7A. Persons on Death Row by Year, 1974-2001

Figure 7B. Number of Executions by Year, 1974-2001

Figure 7C. Percent of Death Row Executed by Year, 1974-2001

The capital system's functional success rate thus is between 1 and 8%. Indeed, as Figures 8A and 8B further illustrate, the only possible sense in which the modern death penalty system can be said to "work" is as a costly Rube Goldberg contraption for making serious errors and then trying, but failing, to fix them. Although intended to move cases quickly and smoothly from death verdict to execution, the main momentum of the system is in the opposite direction: After moving slowly and haltingly through a review process clogged with error-laden cases, most cases flow backwards to retrials, then out of the system entirely as non-capital sentences and acquittals. Among the trickle of cases getting through the review process to the execution stage are some in which the condemned prisoner is innocent, and more where the prisoner is guilty of a crime but not one for which the law allows to the death penalty to be imposed.

Figure 8A: Outcomes Foll0wing Arrest for Homicides Committed from 1973 to 1995 in States with the Death Penalty

Figure 8B: Known Outcomes, Three Stages of State and Federal Court Review

Figure 9: Overall Error Rate and Percent of Death Verdicts Carried Out (Nonconsensual Executions), 1973-1995

The contribution of serious, reversible error to the capital system's dismal success rate is immense. Figure 9, p. 79 above, compares states with at least one fully reviewed death verdict based on (1) the percentage of death verdicts they imposed that were carried out during the study period (the dark grey line) and (2) their overall reversal rates (the light grey line). Figure 9 shows that:

  • More than 70% of the states executed fewer than 8% of the people they sentenced to die. All but one of those states had error rates over 60%; most of their error rates were over 70%; nearly half were 80% or more.


  • The two states with error rates below 30% (Delaware and Virginia) are the only two states that executed 18% or more of the men and women they condemned.


  • The correlation between high error rates and low rates of death sentences carried out is strong and significant. 226

It is hard to imagine another public or private operation being allowed to continue at all, much less for decades, with this record of error and failure. This is especially so, given the cost:

  • Nearly 100 men and women spent years on death row for crimes they did not commit, or for which they eventually were acquitted. Many others were condemned for crimes for which the law does not permit death as a punishment.

  • We cannot say how many innocent men and women have been executed, in part because officials with the information needed to answer the question won't release it. But the risk that this has happened, and will happen again absent reform, is high.

  • While innocent people have languished on death row, actual killers have gone free, in some cases raping and killing again.227
  • Murder victims' families have repeatedly had their expectations shattered, their grievous losses replayed, and their excruciating fears rekindled as a result of chronic reversals and retrials, altered outcomes and exonerations.228
  • Requiring the same courts and judges to review growing numbers of seriously flawed death verdicts has caused a drastic pile-up of cases awaiting review.229 During the 23-year study period, over 9 years passed on average between the death verdict and execution. In 1981, an average of about 5 years passed from death verdict to final habeas review. By 1995, the figure was about 12 years.230
  • Millions of dollars have been wasted on flawed trials and lengthy appeals.231

  • Retribution and deterrence have given way to uncertainty, mistake and delay.232

  • Public faith in the criminal justice system has plummeted.233

The death penalty system documented here does not work. Instead, it plays a cruel hoax on taxpayers, the judicial system, innocent defendants and families of murder victims.

E. A Research Imperative: Seeking the Causes of Reversible Error in Death Cases

Serious, reversible error permeates the existing death penalty system and threatens to destroy it. It puts innocent lives at risk, heightens the suffering of victims, leaves killers at large, wastes tax dollars, and fails citizens, the courts and the justice system. Deciding whether the system is salvageable and, if so, how requires a better understanding of the causes of reversible error in capital trials and verdicts. The statistical analyses discussed in the rest of this Report address that question. Part III discusses study methods and data. Parts IV-VI report study results. Part VII distills those results into overall conclusions that then are the basis for a set of policy options outlined in Part VIII.

next | previous | table of contents