Judicature volume 84 number 2 September-October 2000

          Copyright 2000 American Judiacture Society.  Article posted on the John Jay College Library web site with permission of the editor.
 
 

DEATH MATTERS: a reply to Latzer and Cauthen

by James S. Liebman, Jeffrey Fagan, and Valerie West


The imposition of death as a punishment drastically increases the importance of a reliable conviction and sentence.
 

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 of only a year or two.” 1  This predicate is among “the evolving standards of decency that mark the progress of a maturing society” and determine whether a punishment is “cruel and unusual” in violation of the Constitution. 2  Because “‘[f]rom the point of view of the defendant, [death] is different in both its severity and its finality,’ ” and “‘[f]rom the point of view of society, the action of the sovereign in taking the life of one of its citizens . . . differs dramatically from any other legitimate state action, it is’”—as the Supreme Court repeatedly has said—“‘of vital importance to the defendant and to the community that any decision to impose the death sentence be, and appear to be, based on reason rather than caprice or emotion.’” 3  The importance of assuring accuracy and avoiding mistakes thus extends to all facets of the decision to impose death, from the conviction of murder, to the determination that the offense is of the first degree and capitally aggravated, to the conclusion that no extenuating factors require a sentence less than death.

Additional points of view reveal the breadth of the consensus that the decision to take life is, by orders of magnitude, more important than other criminal verdicts.  Compare, for example, the intensity with which citizens and policy makers debate the proper parameters of the death penalty, to the relative invisibility of analogous discussions of the proper scope of murder as opposed to manslaughter, or of mandatory minimum terms versus life without parole.   Similarly, the agonizing of relatives of murder victims over whether to press for the death penalty—and their vehemence when their answer is “yes”—are rarely matched when the question is whether to charge first  as opposed to second degree murder, or to seek life without parole as opposed to a term of years. 4

The possibility of death as a punishment also drastically alters the criminal process. Because “the Eighth Amendment requires increased reliability of the process by which capital punishment may be imposed,” the Supreme Court has repeatedly defined the sentencing stage of a capital proceeding as a separate “trial” at which important due process protections apply that do not apply to other sentencing proceedings or even to guilt determinations at non capital trials. 5 As was recently explained by states’ attorneys who prefer to devote resources to prosecuting general crimes rather than to “‘extremely expensive’” capital cases:

By the very nature of the gravity of the case, defense lawyers and prosecutors spend more time on a capital case than a noncapital one.  It takes longer to pick a jury, longer for the state to present its case and longer for the defense to put on its witnesses.  There are also considerably greater expenses for expert witnesses, including psychologists and, these days, DNA experts. Then come the defendant’s appeals, which can be considerable, but are not the biggest cost of the case, prosecutors say. 6

And, as Felix Frankfurter noted, “[w]hen life is [put] at hazard in a trial, it sensationalizes the whole thing . . . .” 7

The belief that the decision to take life matters more than other criminal verdicts also has prodded centuries of Anglo American law to define murder ever more narrowly and precisely. 8  The goal of these efforts has not been to keep from misdirecting the onus of a conviction of that crime, but instead to keep from overusing death as the sentence for it.  When England divided its single, capital offense of taking the life of another (homicide) into two offenses (manslaughter and murder), it was not because in the natural scheme of things there are two distinct types of killing that the law must distinguish, but because death was recognized as too severe a penalty for many killings. Thus arose the distinction between killing with “malice,” for which the mandatory punishment was death, and killing without malice, for which death could not be imposed.

After the Revolution, American legislators concluded that even the narrowed category of malice murder was too broad to justify death across the board, and they divided it into “degrees,” with only murders of the “first degree” (e.g., premeditated ones) permitting the still mandatory penalty of death.  When, despite this innovation, jurors routinely refused to convict even clearly premeditated killers of first degree murder because they believed death was too severe a penalty, American states gave jurors discretion to spare the lives of defendants convicted of first degree murder.  When the exercise of that discretion proved arbitrary and discriminatory, the Supreme Court in the 1970s ordered states to distinguish more accurately and objectively among murders that do and do not warrant the death penalty, while still taking individualized factors into account.  This led states to bar the death penalty except (generally speaking) upon proof of (1) murder, (2) in the first degree, that is (3) capitally aggravated, and (4) more aggravated overall than mitigated. 9

For centuries, therefore, the recognition that death is a qualitatively more serious penalty than any other has driven Anglo American law to define capital murder ever more precisely—and to insist that defendants are convicted of it ever more accurately 10 —for the single, crucial purpose of assuring that all capital sentences that result are legally and factually deserved.  Capital crimes and convictions matter so much because of how much the resulting capital sentences matter.

Based on the premise that the accuracy and legality of verdicts of death matter most in assessing the capital system’s success,  we spent nine years collecting data on the validity of the 5,760 capital verdicts imposed by American states between 1973 and 1995 as judged by the capital system’s own judicial inspectors.  In A Broken System: Error Rates in Capital Cases, 1973 1995, we report that 68 percent of all verdicts fully reviewed in that period were found to be so seriously flawed that they had to be scrapped and retried.  Where outcomes are known (for state post conviction reversals), only 18 percent of retrials resulted in the reimposition of death.  Seventy five percent ended in a sentence less than death for murder, and seven percent ended in an acquittal.  “Our 23 years worth of results,” we concluded,
reveal a death penalty system collapsing under the weight of its own mistakes.  They reveal a system in which lives and public order are at stake, yet for decades has made more mistakes than we would tolerate in far less important activities.  They reveal a system that is wasteful and broken and needs to be addressed. 11

In Capital Appeals Revisited, Professors Barry Latzer and James N.G. Cauthen claim that these levels of error are tolerable, and they criticize A Broken System for raising an alarm.  Their principal claim is that our category of “serious error” should be divided into two types of error—so called “sentencing” and “conviction” error—and that only conviction error matters.  For them, errors that seriously compromise “only” the decision whether to impose “imprisonment or death” are not a matter of “major” policy concern.  Latzer and Cauthen make two further claims: (1) When their two types of error are analyzed separately based on a sample of 837 state court reversals of capital verdicts, the unimportant type—“sentencing” error—is more prevalent than the important type—“conviction” error—so that the “true” error rate is lower than we report.  (2) Because the sentence of death is qualitatively different from all other sentences, courts look harder for error in capital than in non capital murder cases, explaining why courts (as Latzer and Cauthen ultimately acknowledge) find substantially more serious error in capital than in non capital murder verdicts.

We make four points in reply:  (1) The error rates we calculate are a valid measure of the risk that our existing capital system mistakenly executes the wrong people,  i.e., ones who have not committed acts of sufficient culpability to make death a legal punishment.  (2) Latzer and Cauthen’s measure of  “conviction” error does not validly assess that risk—or anything else—because it systematically undercounts “conviction” error and arbitrarily assigns identical reversals to opposite categories.  (3) Their central claim—that policy makers and citizens ought not be concerned with mistaken decisions to take life, or with erroneous executions that result—is out of line with American criminal and constitutional law and the considered judgments of the people who adopted, operate, and rely upon the capital system.  (4) Although the “death is different” premise of Latzer and Cauthen’s last section is correct, the inference that capital verdicts are more carefully scrutinized for error is unjustified; the opposite may be true.
 

A measure of risk

Before explaining why high rates of reversible error in death verdicts can help measure the risk that our capital system condemns and executes the wrong people—ones who did not commit acts for which death is a legal penalty—we explain why other proposed measures are unworkable.

Innocence measures.  Some claim the best way to measure the risk that capital verdicts are wrong is to count the number of innocent people who have been executed. 12  To succeed, however, the innocence inquests that any such study would entail would require that experts conducting them have access to evidence in government files that bears on the guilt or innocence of executed people—DNA samples included.  No state conducts such inquests or requires officials to cooperate with them.  As a result, officials regularly refuse to disclose such information and often destroy it. 13

In addition, this is not a measure of risk but of already sustained harm.  Using it frustrates the goal of risk assessment—heading off harm before it occurs—by requiring, in effect, that the reactor explode before steps can be taken to avoid the calamity.  Given how hard it is to detect capital calamities that have occurred, and how unwilling officials are to aid in doing so, it is difficult to distinguish reliance on this measure of risk or harm from a means of covering it up.

There is a more fundamental problem.  Homicide is a capital offense only if it is (1) murder (2) in the first degree, that is (3) capitally aggravated (e.g., by the statutory aggravating factor that the defendant knowingly took the life of a peace officer), and overall is (4) more aggravated than mitigated. Suppose Jones is executed for his part, with Williams, in killing an undercover police officer, Smith, in the course of a drug deal.  Jones was not innocent.  He and Williams killed Smith, and their role in the drug deal made the killing culpable homicide.  But did Jones commit a crime for which death is a legal punishment? Or, in the Supreme Court’s phrase, was Jones  “innocent of the death penalty?” 14  Suppose the killing was accidental, so only second degree murder or manslaughter, but the witness who could prove it was never disclosed by the police or tracked down by Jones’s lawyer. Or suppose the killing was first degree murder but was not capitally aggravated because Williams (as he confessed in a statement never disclosed to the jury) did not tell Jones that Smith was a policeman.  Or suppose the offense was more mitigated than aggravated because Jones was mentally retarded, had a clean record and participated under duress by Williams, the triggerman, who received a life sentence—things Jones’s lawyer never told the jury.  In all these situations, a wrongful execution occurred.  But it is missed by an innocence based measure of risk.

In declaring a moratorium on executions in his state, Illinois Governor George Ryan recently cited a different measure of risk:  a comparison of the number of “death row exonerations” during the modern death penalty era (88 nationally; 13 in Illinois) to the number of executions in the same period (650 nationally; 12 in Illinois).  Better than 1 to 7.5 national odds that a death sentence will end in exoneration not execution (or better than 1 to 1 odds of the same thing in Illinois) certainly imply a substantial risk of erroneous capital verdicts and executions.  But like the prior one, this measure is impaired by inadequate information and is too narrow.  It takes about 10 years on average for capital verdicts to get through the overburdened review process.  Most death sentences revealed by a snapshot of the system (such as the one taken by an exonerations-to-executions analysis) have not, therefore, led to exonerations, executions, or any other conclusion. Instead, they are awaiting review.  If, on average, the review process leading to exonerations takes longer than the review process that precedes executions, 15 the number of each that have occurred as of a given moment will underestimate the exonerations to  execution odds once review finally occurs.  The reverse is true if the time to execution is usually longer than the time to exoneration.

The exonerations-to-executions measure may underestimate risk in other ways.  If an innocent person is executed, the measure is doubly skewed because it counts as an execution what should have been an exoneration.  Exonerations also exclude a number of probably innocent prisoners whose capital verdicts are reversed and who are set free following pleas to lesser offenses and “time served.”  For an innocent prisoner who has spent years on death row due to a mistaken trial verdict, an offer of immediate release on time served instead of risking another faulty trial verdict may be one he can’t refuse—even if the outcome is one an exonerations researcher can’t compute. 16

Finally, exonerations require outright acquittal of homicide or the defendant’s release upon a state attorney’s decision not to prosecute. They miss prisoners like our hypothetical Jones who receive lesser convictions or sentences after their death verdicts are reversed due to serious error, and who were wrongly condemned for homicides that were more mitigated than aggravated or not capitally aggravated at all, or were not first , or any, degree of murder.  An exonerations to executions metric misses many such reversals.  For every one capital state post conviction reversal between 1973 and 1995 that led to a not guilty finding, there were seven that led to other outcomes less than death.

Our measureA Broken System uses a different measure of the risk of erroneous capital verdicts and executions:  the rate at which “death sentences subjected to judicial inspection nationally and in . . . [each of the] death sentencing states [a]re found to be seriously flawed and [a]re reversed by the courts,” or put another way, the amount of “error that substantially undermines the reliability of the guilt finding or death sentence imposed at trial.” 17  This measure avoids the informational difficulties faced by the measures discussed above because it permits a dispositive answer for each of the thousands of capital verdicts that have been finally reviewed: during our 23 year study period,  68 percent were overturned because of a substantively or procedurally flawed determination that the defendant committed culpable enough acts to permit a death sentence.

To assess the usefulness of this figure as a measure of the risk of deadly mistakes, we first examined the types of legal error based on which courts may reverse capital verdicts, concluding that such errors almost always undermine the reliability of death verdicts.  Reversals are not the product of “technicalities” but of judicial findings that (1) the state never proved to the requisite degree of certainty that the defendant committed capitally punishable acts, or that (2) the procedures used to make that showing not only were illegal (and for the most part, unconstitutional) but also (a) were inherently unreliable and prejudicial, (b) were so demonstrably prejudicial that but for the faulty procedure there is a probability that the outcome would have been different, (c) had a “substantial and injurious” effect on the verdict, or (d) were not “harmless.”

We next examined the errors based on which courts at the state post conviction stage actually reversed death sentences, finding that 80 percent fall into four categories of serious error—egregiously incompetent defense lawyers, prosecutorial suppression of evidence and other misconduct, misinstruction of juries, and biased judges and juries.  The first three types of error (accounting for 76 percent of the reversals) require proof of a probability that the error changed the outcome of the trial; the last type (decision maker bias) is inherently prejudicial.

We then considered the possibility that the judges who found all these reliability threatening errors were, as a class, prone to find error or hostile to the states’ death penalty laws.  We discounted this hypothesis upon finding that high error rates were remarkably consistent over time and across a broad range of judicial decision makers in many states, the vast majority of whom were unlikely to be sympathetic to convicted killers or hostile to state law:  (1) Judges in office throughout the period 1977 to 1995 found overall error rates of more than 50 percent in each of those years save one.  (2) Judges with jurisdiction in all but two of the 28 states in which death sentences were imposed and finally reviewed in 1973 through 1995 found 50 percent plus rates of error.  (3) State trial and appellate judges, the vast majority of them elected, ordered 90 percent of the approximately 2500 capital verdict reversals we identified.  (4) Most of the remaining, federal reversals occurred when most federal judges were Reagan and Bush appointees.  (5) The rates of error found by federal judges on habeas review are almost identical to (and slightly lower than) those found by state high courts on direct review.

Finally, we followed up the state post conviction reversals in our study to see if outcomes changed when errors were cured on retrial—something that would occur only if the errors triggering the reversals were serious.  We found that 82 percent of the outcomes changed from a verdict that the defendant must die to a decision that he should live: 75 percent received sentences of imprisonment for some degree of homicide; 7 percent were acquitted.  The errors prompting these reversals were not trivial or technical.  They evidently led thousands of jurors to misjudge a defendant’s culpability by at least the margin of his or her life.
For these reasons, the decades of high reversal rates we documented nationally and in most capital states provide strong evidence that American death verdicts—decisions that men and women have committed sufficiently culpable acts that they legally may, and should, be executed—are generally unreliable.  There is, in sum, a high risk that many capital verdicts are wrong.
 

Intolerable risk

This risk is intolerable for two reasons.  First, even if appeals could catch all the serious, reliability impairing errors that occur, doing so requires a hugely expensive inspection process, which most death verdicts fail (68 percent between 1973 and 1995), followed by similarly costly repair efforts on retrial which also usually fail, resulting in a lesser sentence or acquittal (82 percent of the time following state post conviction reversals). As Latzer and Cauthen say about conviction
focused retrials, every retrial at which life is at stake “is costly to the criminal justice system, burdensome to the witnesses, and painful to the family of the murder victim.”  Moreover, because at any given time the vast majority of death verdicts are bottled up in a decade long review process, and most that exit the process are sent back, the rate of death sentences carried out from 1973 to 1995 was 5 percent. As would be true of any other government or private sector operation, a 68 percent failure rate upon inspection, an 82 percent scrap rate upon repair, and only a 5 percent rate of products working as planned is not cause for congratulations about a well oiled quality control machine but a good reason to overhaul a bankrupt production process.

Worse, there is reason to fear that a high risk of unreliable capital verdicts means a high risk of wrongful executions—appeals notwithstanding.  Unless the criteria for identifying flaws and the judges applying them are perfect, the 68 percent rate at which judges find serious error that impairs the reliability of capital verdicts implies some odds that they miss other unreliable verdicts.  The fact that state judges found serious error in 47 percent of the capital verdicts they reviewed, and yet federal judges found serious error in 40 percent of the verdicts that cleared the state inspections, shows that judges and review criteria are not perfect.  So does evidence that some states have faulty error detection systems. 18 And so especially do the cases of Ronald Monroe (Louisiana), Anthony Porter (Illinois), Don Paradis (Idaho), Lloyd Schlup (Missouri), and Earl Washington (Virginia), all of whom were proven innocent or probably so after their death verdicts passed a full set of judicial inspections. 19  The likelihood of “misses” has increased, moreover, as the number of death verdicts under judicial review has risen from 635 in 1979, to 3,752 in 1999.
 

A flawed measure

Latzer and Cauthen dispute our measure of risk and propose their own:  the number of capital “convictions,” as opposed to “sentences, ” that courts found seriously flawed.  They collect a sample of 837 published reversals of capital judgments by state supreme courts, divide them into so called “conviction” and “sentence” reversals, jettison the latter as unimportant, and come up with a “reversal rate in death penalty cases [that] is closer to 27 percent.”

Four methodological defects deprive Latzer and Cauthen’s study of validity.  First, they are forced to use phrases like “closer to 27 percent” when juxtaposing their number of “conviction” reversals to our 68 percent figure for all capital reversals because the comparison they purport to make is classically one of apples and elephants.  They derived their number by merely sampling reversals between 1990 and 1999.  We counted all reversals between 1973 and 1995.  Because the numbers are based on different sets of cases, they are not comparable.

Second, Latzer and Cauthen’s sampling method undercounts so called “conviction” reversals.  To begin with, Latzer and Cauthen generated their sample using search terms (“death penalty,” “capital murder,” “sentenced to death,” or “death sentence”; the word “conviction” was not used) that are nicely designed to capture most cases where a “death sentence” is reversed but are not well designed to capture cases where a “conviction” is overturned and in which there, accordingly, need be no mention of the “sentence” or its “capital” character. The sample also excluded reversals by state intermediate courts, even when they were the highest court to rule on, and reverse, death verdicts. We have no idea how many reversals Latzer and Cauthen’s sample misses because many of their cases are from years we did not study.  But we have good reason to think that most of the missed cases were “conviction reversals,” which, as a result, were undercounted. A quick search of only a subset of the years Latzer and Cauthen sampled revealed 40 captial reversals their  sampling method misses.  Eighty  three percent are “conviction” reversals. 20

Third, Latzer and Cauthen indiscriminately mix their biased but relatively large sample of direct appeal reversals with a less complete sample of state post conviction reversals (many of which occur in unpublished and lower court opinions that the authors did not count).  Without knowing the relative proportions of reversals that occurred at each review stage in their study and in the universe of all reversals, the convictions versus sentencing breakdown for reversals they discovered at each review stage, and the effect on each breakdown of the authors’ various sampling decisions, the stew of cases in their sample provides no basis for judging the actual rates of “conviction” and “sentencing” reversals at either review stage or at the two combined.

Most crucially, Latzer and Cauthen assign cases to their two categories—“conviction” and “sentence” reversals—based on an arbitrary distinction between identical reversals.  The result is a meaningless count of reversals in each category.  Although Latzer and Cauthen never explain their terms, they evidently define a “conviction” reversal as one requiring a new first, or “guilt,” phase of trial, and a “sentence ” reversal as one requiring a new second, or “sentencing,” trial.  What they fail to note, or even to realize, is that different states assign different functions to the two trial phases.

As noted, most capital sentencing states predicate death sentences on four findings: that the crime, beyond a reasonable doubt, was (1) murder (e.g., because the killing was intentional), (2) in the first degree (e.g., because it was premeditated), (3) capitally aggravated (e.g., because the victim was a peace officer); and was (4) more aggravated overall than mitigated.  State law varies, however, as to whether findings (2) and (3) are made at the first or the second phase of trial.  Some states make both findings at the first phase; others make both at the second phase; and still others make the first degree finding at the first phase and the capital aggravation finding at the second phase. 21

Latzer and Cauthen acknowledge that all serious errors in deciding whether an offense was first degree murder (finding 2) and capitally aggravated (finding 3) are cause for concern. 22 But they arbitrarily assign only some  of those errors to their “important” category of “conviction” error.  Only when the state in question happens to use the first trial phase to make the finding that was flawed do Latzer and Cauthen deem it “conviction” error.  When the relevant state makes the very same finding at the second phase—as is true for finding 2 in major capital states like Florida, Georgia, Mississippi, and Texas, and for finding 3 in nearly all states—Latzer and Cauthen relegate the identical error to the supposedly irrelevant category of “sentencing” reversals. If an erroneous “first degree” or “capital aggravation” finding is a serious defect when it occurs in California and Louisiana, there is no reason to treat the same error as irrelevant when it occurs in Florida or Texas. But Latzer and Cauthen do just that—in the process, seriously undercounting so called “conviction” error.

To justify dividing capital reversals into two categories based on their relative significance, Latzer and Cauthen must explain how the two types of reversals differ in an important and well specified way, and must accurately count the number of each.  Their theoretical and methodological hodge podge of “conviction” and “sentencing” reversals does neither.

Nor is any other division likely to be convincing, because it must inevitably assume—with Latzer and Cauthen—that some errors decisive of life or death ought not concern citizens and policy makers, no matter how often they occur. Herein lies our main disagreement with Latzer and Cauthen. For them, a verdict that wrongly decides “only” whether the facts and law permit a man’s life to be taken gets essentially everything right. For us, a verdict that is wrong by the margin of a person’s life is essentially wrong.
 

Deadly error

The fundamental purpose of any system of “capital punishment” is to decide accurately whether the defendant has committed acts of sufficient culpability that the law permits death as a penalty. Accurate assessments of the defendant’s guilt of  murder, first degree murder, capital aggravation, and net aggravation are all crucial to that decision.  If the system cannot make all those assessments accurately at least some substantial proportion of the time, it fails to achieve its fundamental purpose.  Any useful measure of risk applied to the system must assess the likelihood that its verdicts are accurate in all four ways.

Nor is it surprising that the fundamental object of the system cannot be described except as the accurate imposition of the “sentence” or “punishment” of death.  Because the penalty of death is qualitatively different from a sentence of imprisonment, however long, and because it is of vital importance to the defendant and community that any decision to impose the death sentence be based on reason rather than caprice or mistake, the fear of inaccurate life or death decisions has haunted American law, the evolving standards of decency, defendants, victims, policy makers, courts, citizens, and the press.

This emphasis on accurate death verdicts is of course driven by moral and legal judgments that Latzer and Cauthen may dispute. But neither their views, nor ours, on the point affect A Broken System’s conclusion that the capital system is collapsing under the weight of its own mistakes.  The audience for the report is the people who adopt, operate, depend upon and pay for that system.  By all accounts those people want it expeditiously to capitally convict and punish the right offenders, but not the wrong ones.  Yet as A Broken System reports, the consequence of two in three rates of error is that the existing system (1) does nothing expeditiously; (2) does little that, within its own terms, is right—never managing in any of the last 27 years to execute more than three percent of the people on death row; and (3) takes a big risk that a verdict that the defendant committed acts for which death is a legal penalty is mistaken and, as a result that men and women are being wrongly executed.  No matter how one slices up the ways in which capital verdicts go wrong, the system is failing.
 

More error, less cure.

In the end, Latzer and Cauthen admit that the relevant actors treat capital verdicts as more important than other decisions, and that there are “too many reversals in capital cases”—which even by their undercount occur three to five times more often than in non capital cases.  This prompts them to make a second search for reasons why some serious errors requiring the reversal of death verdicts are “less equal than others,” and thus why high error rates are (in their words) “not unacceptable.” This second effort is no more successful than the first.

Latzer and Cauthen’s primary argument goes as follows:  Death is different, so reliability is more important, so more protections are needed, so there are more ways capital verdicts can go seriously wrong, so more reliability impairing error occurs—so we should worry about it less.   This is classically the clock striking 13.  That capital verdicts so often fail to achieve the reliability we demand is a reason to be concerned, not sanguine.

They next say “the law governing ineffective counsel is more hospitable to claims aimed at the penalty phase” of capital trials.  They offer no evidence for this claim, and it is demonstrably false.  The Supreme Court insists that the same ineffective assistance standard apply to the capital sentencing as the guilt phase.  Because that standard gives the prisoner the burden of  proving that, but for attorney error, the trial outcome probably would have been different, any greater subjectivity in the “Is death deserved?” question than in the  “Is the defendant guilty?” question works against the prisoner by making it harder to prove that different lawyering would have made a difference. 23 Not surprisingly, although ineffective lawyering at both phases of trial is a common basis for reversal, the claim (which is made in most cases) fails seven times more often than it succeeds at the federal habeas stage, is no more likely to succeed than the average claim, and succeeds slightly more often at the guilt phase than at the sentencing phase.

Third, “overturning a capital sentencing trial,” as opposed to the guilt trial, is (as Latzer and Cauthen say) a “relatively ‘low cost’ reversal.”  It thus gives appellate courts a compromise outcome, short of reversing the conviction, that is not available in non capital cases and decreases the amount of “conviction” relief below what courts would grant if “half a loaf” were not an option. 24

Finally, while declining to dispute Professor Gross’s convincing showing that capital cases lead to more miscarriages than other criminal cases, 25 Latzer and Cauthen claim that high error rates in death cases are due instead to capital defendants’ “greater opportunity” to appeal and to more intense review.  Given Gross’s demonstration, Latzer and Cauthen’s assertion is unconvincing on its face.  They also fail to recognize (1) that the review process for capital cases that they describe at such length is the same in non capital cases (except that capital prisoners have shorter filing deadlines), and (2) that capital prisoners may file more appeals because there is more error affecting their verdicts, not the reverse.

Their assertion that appellate courts are “motivated by a distaste for the death penalty”—and a corresponding taste for reversing them—is implausible.  It ignores the fact that 90 percent of reversals are by mainly elected state judges in states where the death penalty is so popular and so often an election issue that an imagined taste for reversing death sentences implies a very unlikely taste for losing office.  The fact that high reversal rates have afflicted dozens of states, over decades, and at every level of court belies any assumption that a few judges out of step with their constituency account for chronically high reversal rates.  On the contrary, the fact that so much error is found and corrected against the political odds suggests that there is more, not less, error than meets the eye.

Latzer and Cauthen’s claim that high rates of error in capital cases are a distracting result of worrying too much about mistaken death verdicts and executions is disturbingly close to dismissing high rates of safety lapses in nuclear power plants because we scrutinize them more carefully than coal fired incinerators.  There is a reason to look harder for error that undermines the reliability of capital verdicts and to be more concerned when we find it.  Death matters.
 
 

Editors note: A final exchange on this topic between Professors Latzer and Cauthen and Professors Liebman and Fagen, and Ms. West, will appear in the November-December issue of Judicature.
 

 1. Woodson v. North Carolina, 428 U.S. 280, 305 (1976).
2. Trop v. Dulles, 356 U.S. 86, 101 (1958).  See Gregg v. Georgia, 428 U.S. 153, 173 (1976).
 3. Lankford v. Idaho, 500 U.S. 110, 125 (1991).
 4. See, e.g., Fryer, Family Weighs Death-Penalty Issue: Like Many Others, Slaying Victim’s Relatives Have Doubts About Executions, Seattle Times, May 14, 2000, at B3; Rice, Families Oppose Executions (Letter to the Editor), Chicago Trib., Feb. 26, 2000, at 24; Montgomery, ‘I Would Like Him to Suffer, but He Won’t’, Wash. Post, May 27, 2000, at B1; Romano, Mixed Feelings In a Murder Case, Wash. Post, May 6, 2000, at A3; AP, A Victim’s Dad Applauds Death Sentence, Houston Chron., Sept. 22, 2000, at 38.
 5. Herrera v. Collins, 506 U.S. 390, 405 (1993).  See, e.g., Ramdass v. Angelone, 120 S. Ct. 2113, 2117 (2000); Turner v. Murray, 476 U.S. 28 (1986); Caldwell v. Mississippi, 472 U.S. 320 (1985); Arizona v. Rumsey, 467 U.S. 203 (1984); Strickland v. Washington, 466 U.S. 668 (1984); Estelle v. Smith, 451 U.S. 454 (1981); Bullington v. Missouri, 451 U.S. 430 (1981).
6. Bonner and Fessenden, States with No Death Penalty Share Lower Homicide Rates, N.Y. Times, Sept. 23, 2000, at A1, A23.
 7. Frankfurter, Of Law and Men 81 (1956).
 8. See, e.g., McGautha v. California, 402 U.S. 183, 198 99 (1971).
 9. See Woodson v. North Carolina, 428 U.S. 280 (1976); Gregg v. Georgia, 428 U.S. 153 (1976); Furman v. Georgia, 408 U.S. 238 (1972).
 10. See, e.g., Powell v. Alabama, 287 U.S. 45, 71 (1932); Stein v. New York, 346 U.S. 156, 196 (1953); Beck v. Alabama, 447 U.S. 625, 637 38 (1980).
11. Liebman, Fagan, and West, “A Broken System: Error Rates in Capital Cases, 1973 1999,” <papers.ssrn.com/paper.taf?abstract_id= 232712>, at i, reprinted in part in Liebman, Fagan, West and Lloyd, Capital Attrition: Error Rates in Capital Cases 1973 1995, 73 Tex. L. Rev. 1862 (2000).
12. See, e.g., Cassell, We’re Not Executing the Innocent, Wall St. J., June 16, 2000 at A14; Wilson, Where’s the Proof Innocent Are Being Executed?, Houston Chron., July 14, 2000 at A17.
  13. See, e.g., Dwyer, Soft On Napping Lawyers, DNA, N.Y. Daily News, Mar. 5, 2000 at 8 (describing officials’ systematic destruction of rape kit evidence in Houston following exoneration of Kevin Byrd); Enzinna, Afraid of a Shadow of a Doubt, Wash. Post, May 7, 2000 at B8 (discussing post execution refusal to produce and destruction of DNA evidence by Virginia officials); Masters, DNA Testing in Old Cases Is Disputed, Wash. Post., Sept. 10, 2000, at A1.
 14. Sawyer v. Whitley, 505 U.S. 333, 334-35, 338-39, 343-49 (1992).
 15. See, e.g., Dieter, “Innocence and the Death Penalty: The Increasing Danger of Executing the Innocent,” <http://www.essential.org/dpic/inn.html> (discussing exonerations of Clarence Brandley, Henry Drake, Vernon McManus, and Rolando Cruz after 10 years; Patrick Croy and Gary Nelson (11 years), Randall Adams (12 years), Joseph Brown and John Knapp (13 years); Ricardo Guerra (15 years); Anthony Porter (16 years), James Richardson (21 years); Paris Carriger (22 years)).
 16. See, e.g., Brown, From Death Row to Halfway House, Phoenix Gazette, Jan. 24, 1995, at B1 (John Serna, released on time served after eight years); Dieter, supra n. 15, at 19 20 (Sonia Jacobs and Mitchell Blazak, same, after 16 and 20 years); Moore, Cloud of Doubt, Houston Chron., Sept. 12, 1999, at 18 (Kerry Cook, same, after 18 years); Death Penalty Information Center (DPIC), Additional Cases of Innocence and Possible Innocence, July, 1997,  <http://www. essential.org/dpic/dpicrecinnoc.html> (Lee Perry, Anthony Scire, and Andrew Mitchell, same, after 7, 9, and 18 years; Victor Jimenez, released on time served plus 18 months after nine years).
 17. Latzer and Cauthen claim incorrectly that “many in the national press” misread A Broken System as a study of reversals of capital “convictions.”  Of 12 national press stories on our report that we have identified, 11 accurately describe it as a study of capital “verdicts,” “judgments,” “convictions and sentences,” or “sentences.”  Even the one publication they cite gives an accurate account of the reversals we studied in all but the single paragraph they quote (and also described it accurately in an accompanying editorial) and, at our insistence, printed a correction of the one mistake.  See Metro. Desk, Corrections, N.Y. Times, June 14, 2000.
 18. See Liebman, Fagan, West, and Lloyd, supra  n. 11, at 1858 n.57 (discussing evidence of faulty error detection in Virginia).
 19. See Dieter, supra n. 15; DPIC, supra n. 16.
 20. A list of the 40 cases is on file with Judicature.
 21. See Lowenfield v. Phelps, 484 U.S. 231 (1988);  Zant v. Stephens, 462 U.S. 862 (1983).
 22. In Latzer and Cauthen’s view, it is only error in finding 4 that is unimportant.  Their explanation—that finding 4 reversals mean only that the reviewing judges “disagreed” with the sentencing jurors on the subjective question whether a death sentence was deserved—fundamentally misconceives capital appeals. Judges can reverse capital verdicts due to error in finding 4 only if they conclude that the procedures or instructions that led to that finding are erroneous in one of the same objective ways that lead to reversals of findings 1 3.  When courts reverse based on finding 4, moreover, they do not generally substitute a sentence of their choice, but (as with finding 1 3 reversals) remand for a new trial at which death can be imposed again.
 23. See Strickland v. Washington, 466 U.S. 668 (1984).
 24. See, e.g., State v. Brimmer , 1999 WL 612888 (Tenn. Ct. Cr. App. 1998) (reversing the death sentence, but not the conviction, of a defendant whose lawyer was just as strung out on alcohol and cocaine at the first as at the second phase of trial).
 25. See Gross, The Risks of Death: Why Erroneous Convictions Are Common in Capital Cases, 44 Buff. L. Rev. 469, 496 (1996); see also Liebman, The Overproduction of Death, 100 Colum. L. Rev.    (forthcoming December 2000).  Latzer and Cauthen do question Gross’s conclusion that less plea bargaining occurs in capital than in non capital cases, but they support the point with a capital case statistic that confuses non negotiated exercises of charging discretion with plea bargaining, and in any event proves Gross’s point because it is 50 percent lower than the non capital plea bargaining rate.