The Part discusses three reasons to study factors related to serious capital error:

• A Broken System found that American capital verdicts were compromised by high numbers and rates of serious, reversible error throughout the 1973 to 1995 period.

• The problem of reversible error may be worse than A Broken System reports because its conservative methods probably undercount the numbers and rates of error.

• The errors A Broken System documents reveal a significant risk that the existing capital system is convicting, capitally sentencing and executing individuals who are innocent of any crime, or any for which the law allows a death sentence.

A. A System Overwhelmed by Error

At p. 9 above is an updated National Report Card documenting the functioning of the death penalty in the United States from 1973 to 1995.⁹⁰ Nationally, state and federal courts reversed, 68% of all capital verdicts imposed and fully reviewed between 1973 and 1995. Based on 5401 final state and federal court decisions in capital cases—4546 final direct appeal decisions, 257 final state post—conviction reversals, and 598 final federal habeas decisions—A Broken System found that:

• Substantial proportions of the death verdicts reviewed at each of three review stages were reversed: 41% of the verdicts reviewed by state supreme courts on direct appeal; at least (probably well more than) 10% of the remaining verdicts reviewed by state post—conviction courts;⁹¹ and 40% of the remaining verdicts reviewed by federal habeas courts.

• Serious, reversible error was the rule, not the exception, among states using the death penalty in the 1973—1995 period. As Figure 1A, p. 51 below, reveals, over 90% of the states with capital verdicts reviewed at all three review stages had overall error rates of 50% or more; 85% had error rates of 60% or more; three—fifths had error rates 70% or more.

• Serious, reversible error also was the rule throughout the study period. As Figure 2A, p. 55 below, shows, over half of all death verdicts reviewed were found seriously flawed in 17 of the last 18 study years. In 18 of the 23 study years, the overall error rate was 60% or more.

• Reversals occurred far more often in the capital cases studied than it does in non—capital cases. Data on non—capital cases is sparse, but our best estimate—revised since A Broken System was released—is that the reversal rate in non—capital cases is less than 10% and probably less than 5%. Capital verdicts are 7 to 14 times more likely to be reversed than non—capital ones.

Reversible error affected both phases of trial: 38% of state post—conviction reversals tainted verdicts reached at the first trial phase (sometimes loosely called the "guilt" phase); the rest affected verdicts at the second phase (sometimes called the "sentencing" phase).⁹² 43% of the errors found on federal habeas tainted first—phase verdicts, while 58% compromised second—phase verdicts.⁹³

As the above paragraph begins to show, reversible error undermines the reliability of many guilt—focused, as well as many sentencing, decisions. A more precise allocation of error between guilt and sentencing decisions is possible, however. In capital cases, the question of guilt or innocence includes the question of the level of offense the defendant committed: Was it manslaughter (as opposed to negligent or non—culpable homicide), second— or first—degree murder or capitally aggravated murder? The law allows the death penalty only for this last category of murder. A few death penalty states assign all these decisions to the first phase of trial. Most states, however (e.g., Alabama, Arizona, the Carolinas, Florida, Georgia, Illinois, Indiana, Mississippi, Pennsylvania, Tennessee and Washington) make the last of these decisions at the second phase of trial.⁹⁴ In these states, not only all error at the first phase of trials but also some error at the second phase undermines the decision that the defendant committed capitally aggravated murder.⁹⁵ Counting all error tainting the decision of the level of offense, if any, that the defendant committed reveals that at least half the serious capital error found on state post—conviction and federal habeas review was guilt—innocence—related error.⁹⁶

There are three reasons why all error requiring capital verdicts to be reversed and retried is serious, whether it affects the reliability of the finding that the defendant committed capital murder or the reliability of the finding that death is the proper penalty for the crime:⁹⁷

• The decision whether to take a person's life for a crime is as serious as any decision a state can make. As the Supreme Court said in the leading capital punishment decision of the modern era, the legal treatment of capital punishment in the United States:

  rests squarely on the predicate that the penalty of death is qualitatively different from a sentence of imprisonment, however long. Death, in its finality, differs more from life imprisonment than a 100—year prison term differs from one or only a year or two.⁹⁸

  The decision to impose the death penalty depends on two crucial conclusions—(1) that the defendant committed capital murder, and (2) that the law permits his execution for that offense given the relevant aggravating and mitigating factors. Because error undermining the reliability of either of these conclusions compromises the most serious of state decisions, all such error is a matter of serious policy concern that warrants study and reform.

• The decision to impose death is a matter of grave concern to all participants in the criminal justice system, not just the defendant. This is clear from large impact on the victim's family, the community, the taxpaying public, the media, and the trial jurors, lawyers and judge of a decision to impose death instead of a lesser sentence. That impact typically is far greater than the affect of a decision to impose life without parole rather than a term of years. So is the impact of a reviewing court's decision to overturn a death sentence—and of a retrial verdict rejecting that penalty—compared to the impact of decisions overturning other sentences.

• It is difficult and dangerous to study errors affecting the reliability of one part of a capital trial while ignoring errors affecting other parts. When a lawyer is incompetent or a prosecutor withholds material evidence, the harm nearly always affects all decisions the jury makes. A decision that reversible error occurred affecting any part of the capital verdict thus undermines the fairness and accuracy of the trial as a whole. In such situations, the typical reviewing decision overturning a death sentence but not the finding of capital murder holds that there was, say, a 45% or 60% or 90% probability the violation affected the sentencing decision but "only," say, a 30% probability it affected the guilt determination. To treat such a case as not serious is not only to treat decisions between life and death as unimportant, but also to ignore evidence compromising the verdict of guilt. This is particularly so given the consensus among most modern public and private enterprises that people and systems prone to error in any part of the process of making products or delivering services are probably prone to error in all parts of the process—so that the detection of serious error is as important for what it says about the safety and effectiveness of the enterprise as a whole as for what it says about the particular error.⁹⁹

For these reasons, our best judgment is that the most crucial public policy issue in the death penalty area is the reliability of verdicts choosing between life and death based on whether the defendant committed capitally aggravated murder and whether the accompanying circumstances allow death as a penalty. It is the reliability of that decision on which we focus.