Executive Summary - Flaws in Americas Death Penalty System

• Feb. 22, 2016 • Locations: United States of America • Topics: Death Penalty

Download original document:
Brief thumbnail

This text is machine-read, and may contain errors. Check the original document to verify accuracy.

Executive Summary
There is a growing bipartisan consensus that flaws in America’s death-penalty system have
reached crisis proportions. Many fear that capital trials put people on death row who don’t belong
there. Others say capital appeals take too long. This report—the first statistical study ever undertaken
of modern American capital appeals (4,578 of them in state capital cases between 1973 and 1995)—
suggests that both claims are correct.
Capital sentences do spend a long time under judicial review. As this study documents,
however, judicial review takes so long precisely because American capital sentences are so
persistently and systematically fraught with error that seriously undermines their reliability.
Our 23 years worth of results 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.
Our central findings are as follows:
•

Nationally, during the 23-year study period, the overall rate of prejudicial error in the
American capital punishment system was 68%. In other words, courts found serious,
reversible error in nearly 7 of every 10 of the thousands of capital sentences that were
fully reviewed during the period.

•

Capital trials produce so many mistakes that it takes three judicial inspections to catch them
—leaving grave doubt whether we do catch them all. After state courts threw out 47% of
death sentences due to serious flaws, a later federal review found “serious error”—error

undermining the reliability of the outcome—in 40% of the remaining sentences.
•

Because state courts come first and see all the cases, they do most the work of correcting
erroneous death sentences. Of the 2,370 death sentences thrown out due to serious error,
90% were overturned by state judges—many of whom were the very judges who imposed
the death sentence in the first place; nearly all of whom were directly beholden to the
electorate; and none of whom, consequently, were disposed to overturn death sentences
except for very good reason. This does not mean that federal review is unnecessary. Precisely
because of the huge amounts of serious capital error that state appellate judges are called
upon to catch, it is not surprising that a substantial number of the capital judgments they
let through to the federal stage are still seriously flawed.

•

To lead to reversal, error must be serious, indeed. The most common errors—prompting a
majority of reversals at the state post-conviction stage—are (1) egregiously incompetent
defense lawyers who didn’t even look for—and demonstrably missed—important
evidence that the defendant was innocent or did not deserve to die; and (2) police or
prosecutors who did discover that kind of evidence but suppressed it, again keeping it
from the jury. [Hundreds of examples of these and other serious errors are collected in
Appendix C and D to this Report.]

•

High error rates put many individuals at risk of wrongful execution: 82% of the people whose
capital judgments were overturned by state post-conviction courts due to serious error were
found to deserve a sentence less than death when the errors were cured on retrial; 7% were
found to be innocent of the capital crime.

•

High error rates persist over time. More than 50% of all cases reviewed were found seriously
ii

flawed in 20 of the 23 study years, including 17 of the last 19. In half the years, including
the most recent one, the error rate was over 60%.
•

High error rates exist across the country. Over 90% of American death-sentencing states
have overall error rates of 52% or higher. 85% have error rates of 60% or higher. Threefifths have error rates of 70% or higher.

•

Illinois (whose governor recently declared a moratorium on executions after a spate of deathrow exonerations) does not produce atypically faulty death sentences. The overall rate of
serious error found in Illinois capital sentences (66%) is very close to—and slightly
lower than—the national average (68%).

•

Catching so much error takes time—a national average of 9 years from death sentence to the
last inspection and execution. By the end of the study period, that average had risen to 10.6
years. In most cases, death row inmates wait for years for the lengthy review
procedures needed to uncover all this error. Then, their death sentences are reversed.

•

This much error, and the time needed to cure it, impose terrible costs on taxpayers, victims’
families, the judicial system, and the wrongly condemned. And it renders unattainable
the finality, retribution and deterrence that are the reasons usually given for having a
death penalty.
Erroneously trying capital defendants the first time around, operating the multi-tiered

inspection process needed to catch the mistakes, warehousing thousands under costly death row

conditions in the meantime, and having to try two out of three cases again is irrational.
iii

This report describes the extent of the problem. A subsequent report will examine its causes
and their implications for resolving the death penalty crisis.

Table of Contents
A Broken System: Error Rates in Capital Cases, 1973-1995
Executive Summary ....................................................................................................................... i
Table of Contents .......................................................................................................................... iv
Figures, Flow Charts, Tables and Capital-Sentencing Report Cards ............................................. x
Acknowledgments .....................................................................................................................xviii
I.

Introduction ........................................................................................................................ 1

II.

Summary of Central Findings

III.

Confirmation from a Parallel Study ................................................................................. 14

IV.

Implications of Central Findings ...................................................................................... 15

The Capital Review Process ............................................................................................. 18

First Inspection: State Direct Appeal ................................................................... 18

Second Inspection: State Post-Conviction ........................................................... 19

Third Inspection: Federal Habeas Corpus ............................................................ 20

VI.

The Study ......................................................................................................................... 23

VII.

The National Capital Punishment Report Card ................................................................ 28
A.

Capital-Sentencing History .................................................................................. 31

Sentences and Executions .................................................................................... 32

Error and Success Rates ....................................................................................... 32

Length of Time of Review .................................................................................. 40

Capital-Sentencing and Execution Rates ............................................................. 41

Demographic Factors ............................................................................................ 42
v

G.
VIII.

Court Factors: The Context of State Court Decision Making .............................. 44

State Comparisons ............................................................................................................ 46
A.

Rates of Serious Error Found on State Direct Appeal .......................................... 46

Rates of Serious Error Found on State Post-Conviction ...................................... 49

Rates of Serious Error Found on State Direct Appeal and
State Post-Conviction ........................................................................................... 52

Rates of Serious Error Found on Federal Habeas Corpus .................................... 56

Rates of Serious Error Found by State Versus Federal Courts ............................ 62

Overall Rates of Error Found on State Direct Appeal, State PostConviction and Federal Habeas Corpus ............................................................... 67

Length of Time of Review ................................................................................... 73

Capital-Sentencing and Execution Rates, and the Two Compared ...................... 78

Demographic Factors ........................................................................................... 87

Court Factors ........................................................................................................ 93

IX.

Federal Circuit Court and Regional Comparisons ......................................................... 102

Conclusion: A Broken System; the Need for Research into Causes .............................. 109

End Notes ................................................................................................................................... 115

Appendices (separately bound)
Appendix A: State Capital Punishment Report Cards
Description of Information in State Capital Punishment
Report Cards ................................................................................................................... A-1
National Composite Capital Punishment Report Card.................................................... A-5
Table 30: State-by-State Comparisons of Direct Appeal,
vi

State Post-Conviction, and Federal Habeas Corpus Reversal
Rates and Overall Error Rates ........................................................................................ A-7
Alabama Capital Punishment Report Card .................................................................... A-9
Arizona Capital Punishment Report Card .................................................................... A-11
Arkansas Capital Punishment Report Card .................................................................. A-13
California Capital Punishment Report Card ................................................................. A-15
Delaware Capital Punishment Report Card ................................................................. A-17
Florida Capital Punishment Report Card ..................................................................... A-19
Georgia Capital Punishment Report Card .................................................................... A-21
Idaho Capital Punishment Report Card ........................................................................ A-23
Illinois Capital Punishment Report Card ..................................................................... A-25
Indiana Capital Punishment Report Card ..................................................................... A-27
Kentucky Capital Punishment Report Card ................................................................. A-29
Louisiana Capital Punishment Report Card ................................................................. A-31
Maryland Capital Punishment Report Card ................................................................. A-33
Mississippi Capital Punishment Report Card .............................................................. A-35
Missouri Capital Punishment Report Card .................................................................. A-37
Montana Capital Punishment Report Card ................................................................... A-39
Nebraska Capital Punishment Report Card .................................................................. A-41
Nevada Capital Punishment Report Card ..................................................................... A-43
North Carolina Capital Punishment Report Card ......................................................... A-45
Oklahoma Capital Punishment Report Card ................................................................ A-47
Pennsylvania Capital Punishment Report Card ............................................................ A-49
South Carolina Capital Punishment Report Card ......................................................... A-51
Tennessee Capital Punishment Report Card ................................................................ A-53
Texas Capital Punishment Report Card ....................................................................... A-55
Utah Capital Punishment Report Card ......................................................................... A-57
Virginia Capital Punishment Report Card ................................................................... A-59
Washington Capital Punishment Report Card ............................................................. A-61
Wyoming Capital Punishment Report Card ................................................................. A-63

vii

Appendix B: Federal Circuit Court and Regional Capital Punishment Report Cards
Description of Information in Circuit and Regional
Capital Punishment Report Cards .................................................................................. B-1
Third Circuit Capital Punishment Report Card .............................................................. B-4
Fourth Circuit Capital Punishment Report Card ............................................................ B-6
Fifth Circuit Capital Punishment Report Card ............................................................... B-8
Sixth Circuit Capital Punishment Report Card ............................................................ B-10
Seventh Circuit Capital Punishment Report Card ........................................................ B-12
Eighth Circuit Capital Punishment Report Card .......................................................... B-14
Ninth Circuit Capital Punishment Report Card ........................................................... B-16
Tenth Circuit Capital Punishment Report Card ........................................................... B-18
Eleventh Circuit Capital Punishment Report Card ...................................................... B-20

Appendix C: Incomplete List of Capital Judgments Reversed on
State Post-Conviction or Related Types of Review
Introduction and Key ...................................................................................................... C-1
Summary of National Composite Results ...................................................................... C-3
Alabama .......................................................................................................................... C-4
Arizona ........................................................................................................................... C-6
Arkansas ......................................................................................................................... C-8
California ........................................................................................................................ C-9
Delaware ....................................................................................................................... C-10
Florida .......................................................................................................................... C-11
Georgia ......................................................................................................................... C-17
Idaho ............................................................................................................................. C-20
Illinois ........................................................................................................................... C-21
Indiana .......................................................................................................................... C-24
Kentucky ...................................................................................................................... C-26
Louisiana ...................................................................................................................... C-27
Maryland ...................................................................................................................... C-28
Mississippi .................................................................................................................... C-30
Missouri ........................................................................................................................ C-31
Montana ........................................................................................................................ C-32
Nebraska ....................................................................................................................... C-33
Nevada .......................................................................................................................... C-34

viii

Appendix C, cont’d
North Carolina .............................................................................................................. C-35
Oklahoma ..................................................................................................................... C-37
Pennsylvania ................................................................................................................. C-38
South Carolina .............................................................................................................. C-39
Tennessee ..................................................................................................................... C-41
Texas ............................................................................................................................ C-44
Utah .............................................................................................................................. C-47
Virginia ......................................................................................................................... C-48
Washington ................................................................................................................... C-49
Wyoming ...................................................................................................................... C-50
Appendix D: Appendix D: Examples of Serious Error
Warranting Federal Habeas Corpus Relief

Appendix E: Tables and Figures
Table 1:

Overall Error Rates and Percent Death Sentences
Carried Out, by State, 1973-1995 ........................................................... E-2

Table 2:

Death Row Population, Executions, and Percent
Executed, 1973-1999............................................................................... E-3

Table 3:

State and Federal Reversals by Year, 1973-1995.................................... E-4

Table 9:

Overall Error Rates, by State, 1973-1995 Excluding
State Post-Conviction.............................................................................. E-5

Figure 11:

Combined Error Rates on State Direct Appeal and
Federal Habeas, 1973-95 ....................................................................... E-6

Table 11:

Years from First Death Sentence to First NonConsensual Execution, 1973-2000 ......................................................... E-7

Table 12:

Average Years from Death Sentence to Execution,
1973-95 ................................................................................................... E-8

Table 13:

Proportion of 1973-1995 Death Sentences Awaiting
State Direct Review as of 1995 ............................................................... E-9

Appendix E, cont’d
Table 14:

Death Sentences Per 1,000 Homicides, 1973-1995 ............................. E-10

Table 15:

Death Sentences Per 100,000 Population, 1973-1995 ......................... E-11

Table 16:

Death Sentences Per 1,000 Prison Population, 1973-1995 .................. E-12

Table 17:

Non-Consensual Executions Per 1,000 Homicides,
1973-1995 ............................................................................................. E-13

Table 18:

Non-Consensual Executions Per 100,000 Population,
1973-1995 ............................................................................................. E-14

Table 19:

Non-Consensual Executions Per 1,000 Prison
Population, 1973-1995 ......................................................................... E-15

Table 20:

Average Homicides Per 100,000 Population, 1973-1995 .................... E-16

Table 21:

Average Percent of Population that is Non-White,
1973-1995 ............................................................................................. E-17

Table 22:

Index of Political Pressure on State Courts .......................................... E-18

Table 23:

State Expenditures Per Capita on Courts, 1973-95 .............................. E-19

Table 24:

State Court Criminal Cases Per 1,000 Population,
1973-95 ................................................................................................. E-29

Table 26:

Death Sentences Per 100,000 Population by Region,
1973-1995 ............................................................................................. E-21

Table 27:

Non-Consensual Executions Per 100,000 Population
by Region, 1973-1995 .......................................................................... E-21

Table 28:

Percent of Death Sentences Carried Out (All
Executions), 1973-1995 ....................................................................... E-22

Table 29:

Percent of Death Sentences Carried Out (NonConsensual Executions), 1973-1995 .................................................... E-23

Figure 36:

Percent Death Sentences Carried Out (NonConsensual Executions), 1973-95 ........................................................ E-24
x

Figures, Flow Charts, Tables and Capital-Sentencing Report Cards
Figures
Figure 1:

Overall Error Rate and Percent of Death Sentences
Carried Out, 1973-1995 ........................................................................................ 11

Figure 2:

Persons on Death Row and Percent and Number
Executed, 1974-1999............................................................................................. 13

Figure 3:

Error Rates Detected on State Direct Appeal, Federal
Habeas, and the Two Combined, by Year, 1973-95 ............................................. 38

Figure 4:

Known State Post-Conviction Reversals, 1973-2000 .......................................... 39

Figure 5:

Percent of Death Sentences Reversed on State
Direct Appeal, 1973-95 ........................................................................................ 48

Figure 6:

Percent of Death Sentences Reversed on State
Direct Appeal or State Post-Conviction, 1973-95 ................................................ 54

Figure 7:

Percent of Death Sentences Reversed on Federal
Habeas, 1973-95 ................................................................................................... 58

Figure 8:

Percent of Capital Sentences Reversed on Federal
Habeas by State and Circuit, 1973-95 .................................................................. 61

Figure 9:

Percent of Capital Judgments Reversed on State
Direct Appeal and Federal Habeas, 1973-95 ........................................................ 63

Figure 10:

Percent of Capital Judgments Reversed on State
Direct Appeal or State Post-Conviction and on
Federal Habeas, 1973-95 ...................................................................................... 64

Figure 11:

Combined Error Rate on State Direct Appeal and
Federal Habeas, 1973-95 .................................................................................... E-6

Figure 12:

Combined Error Rate on State Direct Appeal, State
Post-Conviction and Federal Habeas, 1973-95 ..................................................... 69

Figure 13:

Overall Error Rate (Including State Post-Conviction)
and Reversal Rate by Type of Review, 1973-95 .................................................. 72

Figure 14:

Years from First Death Sentence to First
Non-Consensual Execution, 1973-2000................................................................ 74

Figure 15:

Average Years from Death Sentence to Execution,
1973-95 ................................................................................................................. 75

Figure 16:

Percent of 1973-95 Death Sentences Awaiting
Review as of 1995 ................................................................................................ 77

Figure 17:

Per Capita Death Sentencing Rates by State,
1973-95 ................................................................................................................. 79

Figure 18:

Per Capita Non-Consensual Executions by State,
1973-95 ................................................................................................................. 80

Figure 19:

Death Sentences and Executions per 1,000 Homicides
by State, 1973-95 .................................................................................................. 82

Figure 20:

Death Sentences and Executions per 100,000 Population
by State, 1973-95 .................................................................................................. 83

Figure 21:

Death Sentences and Executions per 1,000 Prisoners,
1973-95 ................................................................................................................. 84

Figure 22:

Per Capita Death Sentences and Percent Death
Sentences Carried Out, 1973-95............................................................................ 86

Figure 23:

Death Sentencing Rates Per 1,000 Homicides and
Per Capita Homicide Rates, 1973-95 ................................................................... 88

Figure 24:

Execution Rates and Homicide Rates, 1973-95 ................................................... 89

Figure 25:

Per Capita Death Sentencing Rates and Percent
Non-White Population, 1973-95 .......................................................................... 91

Figure 26:

Percent of Death Sentences Carried Out and
Non-White Population, 1973-95 .......................................................................... 92

Figure 27:

Political Pressure and Death-Sentencing Rate,
1973-95 ................................................................................................................. 94
xii

Figure 28:

Political Pressure and Percent of Death Sentences
Carried Out, 1973-95 ............................................................................................ 95

Figure 29:

Per Capita Spending on Courts and Per Capita Death
Sentencing Rates, 1973-95 ................................................................................... 97

Figure 30:

Per Capita Spending on Courts and Percent of
Death Sentences Carried Out, 1973-95 ................................................................ 98

Figure 31:

State Court Caseloads and Death Sentencing Rates
per 1,000 Homicides, 1973-95 ........................................................................... 100

Figure 32:

State Court Caseloads and Percent of Death Sentences
Carried Out, 1973-95 .......................................................................................... 101

Figure 33:

Rate of Error Detected on Habeas in Circuits with
Over 10 Cases, 1973-95 ...................................................................................... 104

Figure 34:

Per Capita Death Sentencing and Per Capita Execution
Rates by Circuit, 1973-95 ................................................................................... 108

Figure 35:

Percent Death Sentences Carried Out, All Executions,
1973-95 ............................................................................................................... 111

Figure 36:

Percent Death Sentences Carried Out, NonConsensual Executions, 1973-95 ..................................................................... E-34

Flow Charts
The Attrition of Capital Judgments ........................................................................ 7
The Capital Criminal Process: Trial Through State and
Federal Post-Conviction ....................................................................................... 22

Tables
Table 1:

Overall Error Rates and Percent Death Sentences
Carried Out, by State, 1973-1995 ....................................................................... E-2

Table 2:

Death Row Population, Executions, and Percent
xiii

Executed, 1973-1999........................................................................................... E-3
Table 3:

State and Federal Reversals by Year, 1973-1995................................................ E-4

Table 4:

Percent of Capital Judgments Reviewed on Direct
Appeal in Which Reversible Error Was Found,
1973-1995 ............................................................................................................. 47

Table 5:

Known State Post-Conviction Reversals, 1973-1995,
By State ................................................................................................................ 50

Table 6:

State-by-State Comparisons of Rates of Error Detected
by All State Courts (State Direct Appeal and State
Post-Conviction) ................................................................................................... 53

Table 7:

Percent of Capital Judgments Reviewed on Federal
Habeas Corpus in Which Reversible Error Was Found,
1973-1995 ............................................................................................................. 57

Table 8:

Error Rates, by Selected States, Found by Federal
Circuit Courts on Habeas Review, 1973-1995...................................................... 60

Table 9:

Overall Error Rates, by State, 1973-1995 Excluding
State Post-Conviction.......................................................................................... E-5

Table 10:

Overall Error Rates, by State, 1973-1995 Including
State Post-Conviction............................................................................................ 68

Table 11:

Years from First Death Sentence to First NonConsensual Execution, 1973-2000 ..................................................................... E-7

Table 12:

Average Years from Death Sentence to Execution,
1973-95 ............................................................................................................... E-8

Table 13:

Proportion of 1973-1995 Death Sentences Awaiting
State Direct Review as of 1995 ........................................................................... E-9

Table 14:

Death Sentences Per 1,000 Homicides, 1973-1995 ......................................... E-10

Table 15:

Death Sentences Per 100,000 Population, 1973-1995 ..................................... E-11

Table 16:
Table 17:

Death Sentences Per 1,000 Prison Population, 1973-1995 .............................. E-12
Non-Consensual Executions Per 1,000 Homicides,
xiv

1973-1995 ......................................................................................................... E-13
Table 18:

Non-Consensual Executions Per 100,000 Population,
1973-1995 ......................................................................................................... E-14

Table 19:

Non-Consensual Executions Per 1,000 Prison
Population, 1973-1995 ..................................................................................... E-15

Table 20:

Average Homicides Per 100,000 Population, 1973-1995 ................................ E-16

Table 21:

Average Percent of Population Non-White, 1973-1995 .................................. E-17

Table 22:

Index of Political Pressure on State Courts ...................................................... E-18

Table 23:

State Expenditures Per Capita on Courts, 1973-95 .......................................... E-19

Table 24:

State Court Criminal Cases Per 1,000 Population,
1973-95 ............................................................................................................. E-20

Table 25:

Error Rates Detected on Habeas Review and Overall
(State Direct Appeal and Federal Review Combined)
by Federal Circuit/Multi-State Regions ............................................................. 103

Table 26:

Death Sentences Per 100,000 Population by Region,
1973-1995 ......................................................................................................... E-21

Table 27:

Non-Consensual Executions Per 100,000 Population
by Region, 1973-1995 ...................................................................................... E-21

Table 28:

Percent of Death Sentences Carried Out (All Executions),
1973-1995 ......................................................................................................... E-22

Table 29:

Percent of Death Sentences Carried Out (NonConsensual Executions), 1973-1995 ................................................................ E-23

Table 30:

State-by-State Comparisons of Direct Appeal, State
Post-Conviction, Federal Habeas Corpus Reversal
Rates and Overall Error Rates ............................................................................ A-7

Tables in End Notes
Outcomes Following State Post-Conviction Reversals,
1973-April 2000 ............................................................................note 45 (text p.5)
Direct Appeal Reversal Rates in States in Which No
Capital Judgments Had Completed Federal Habeas
Review by End of Study Period ................................................note 147 (text p.33)
Total Number of Executions Compared to Executions
by Texas and Virginia ...............................................................note 218 (text p.85)
Rank in Death Sentences per 1,000 Homicides
Compared to Rank in Sentences Carried Out
Following Full Review ............................................................note 238 (text p.112)

Capital Punishment Report Cards
National Capital Punishment Report Card
National Composite Capital Punishment Report Card.............................................. 29, A-5
State Capital Punishment Report Cards
Alabama Capital Punishment Report Card ..................................................................... A-9
Arizona Capital Punishment Report Card..................................................................... A-11
Arkansas Capital Punishment Report Card................................................................... A-13
California Capital Punishment Report Card.................................................................. A-15
Delaware Capital Punishment Report Card .................................................................. A-17
Florida Capital Punishment Report Card ...................................................................... A-19
Georgia Capital Punishment Report Card..................................................................... A-21
Idaho Capital Punishment Report Card......................................................................... A-23
Illinois Capital Punishment Report Card ...................................................................... A-25

xvi

Indiana Capital Punishment Report Card...................................................................... A-27
Kentucky Capital Punishment Report Card .................................................................. A-29
Louisiana Capital Punishment Report Card .................................................................. A-31
Maryland Capital Punishment Report Card .................................................................. A-33
Mississippi Capital Punishment Report Card ............................................................... A-35
Missouri Capital Punishment Report Card ................................................................... A-37
Montana Capital Punishment Report Card.................................................................... A-39
Nebraska Capital Punishment Report Card................................................................... A-41
Nevada Capital Punishment Report Card...................................................................... A-43
North Carolina Capital Punishment Report Card.......................................................... A-45
Oklahoma Capital Punishment Report Card ................................................................. A-47
Pennsylvania Capital Punishment Report Card............................................................. A-49
South Carolina Capital Punishment Report Card.......................................................... A-51
Tennessee Capital Punishment Report Card ................................................................. A-53
Texas Capital Punishment Report Card ........................................................................ A-55
Utah Capital Punishment Report Card .......................................................................... A-57
Virginia Capital Punishment Report Card .................................................................... A-59
Washington Capital Punishment Report Card .............................................................. A-61
Wyoming Capital Punishment Report Card.................................................................. A-63
Circuits Capital Punishment Report Cards
Third Circuit Capital Punishment Report Card .............................................................. B-4
Fourth Circuit Capital Punishment Report Card ............................................................. B-6
xvii

Fifth Circuit Capital Punishment Report Card................................................................ B-8
Sixth Circuit Capital Punishment Report Card ............................................................. B-10
Seventh Circuit Capital Punishment Report Card......................................................... B-12
Eighth Circuit Capital Punishment Report Card ........................................................... B-14
Ninth Circuit Capital Punishment Report Card ............................................................ B-16
Tenth Circuit Capital Punishment Report Card ............................................................ B-18
Eleventh Circuit Capital Punishment Report Card ....................................................... B-20

xviii

Acknowledgments
This Report is the product of thousands of hours of labor by dozens of people, including
colleagues, administrative assistants, and many tireless and immensely resourceful and productive
student research assistants at Columbia Law School. Special thanks go to John Lundin, Randy
Hertz, Jon Lloyd, Garth Davies, Michael J. Balzano (for his terrific graphics assistance), Clare
Huntington, Julia Tarver, Honri Marcel, Janet Sabel, and especially Kate McGarry. We also express
our appreciation to Alicia Amezcua, Sandor Bak, Dan Canova, Molly Bishop, Kristin Buff, Sean
Feller, Kara Finck, Brandon Garrett, Brian Gifford, George Hatzmann, Justine Harris, Tricia
Hoefling, Nick Isaacson, Matt Kadusian, Alice Kim, Hyeon Lee, Jeremiah Marble, Stephan Murphy,
Stefan Schick, Yosef Rothstein, Troy Selvaratnam, Erika Smith, and Ernest Yakob.
Scores of people in death-sentencing states around the nation have helped us immensely over
the years—typically taking time away from important activities and withering schedules to provide
us with information about capital cases in their states. We are very grateful. In this regard, we would
especially like to remember Patsy Morris, and to thank Donna Gloekner and Professor Michael L.
Radelet.
From 1991 until June 1999, this research was funded entirely by the Columbia Law School through
a series of stipends for faculty research and student research assistance during the summer months.
We deeply appreciate that support. We also express our thanks to the Law and Society Program of
the Open Society Institute for a grant that supported the work of one of the authors, Valerie West,
during the spring and summer of 1999 and has helped defray the cost of distributing this Report to
the public.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

A Broken System: Error Rates in Capital Cases, 1973-1995

by James S. Liebman, Jeffrey Fagan & Valerie West
June 12, 2000

Introduction
A new debate over the death penalty is raging in the United States.1 Until now, the focus of

that debate has been the fairness of particular capital convictions and sentences. This Report
addresses a different and broader question: the reliability—indeed, the bare rationality—of the
death penalty system as a whole. It asks whether the mistakes and miscarriages of justice known
to have been made in individual capital cases2 are isolated, or common? The answer provided by our
study of 5,760 capital sentences and 4,578 appeals is that serious error—error substantially
undermining the reliability of capital verdicts— has reached epidemic proportions throughout our
death penalty system. More than two out of every three capital judgments reviewed by the
courts during the 23-year study period were found to be seriously flawed.
Americans seem to be of two minds about the death penalty.3 In the last several years,
executions have risen steeply, reaching a 50-year high.4 Two-thirds of the public support the
penalty.5
Two-thirds support, however, represents a steady decline from the four-fifths of the
population that supported the penalty only six years ago, leaving support for capital punishment at
a 20-year low.6 When life without parole is proposed as an alternative, support for the penalty drops
even more—often below a majority.7 Grants of executive clemency reached a 20-year high in 1999.8

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

In 1999 and 2000, Governors, attorneys general and legislators in Alabama, Arizona, Florida,
and Tennessee have fought high-profile campaigns to speed up and increase the number of
executions.9
In the same period, however:
•

The Republican Governor of Illinois, with support from a majority of the electorate, declared
a moratorium on executions in the state.10

•

The Nebraska Legislature did the same. Although the governor vetoed the legislation, the
Legislature appropriated money for a comprehensive study of the even-handedness of the
state’s exercise of capital punishment.11 Similar studies have since been ordered by the Chief
Justice, task forces of both houses of the state legislature and the Governor of Illinois,12 and
also the Governors of Indiana and Maryland and the Attorney General of the United States.13

•

Serious campaigns to abolish the death penalty are under way in New Hampshire16 and (with
the support of the Governor and a popular former Republican Senator) in Oregon.17

•

The Florida Supreme Court and Mississippi Legislature have recently acted to improve the
quality of counsel in capital cases,18 and bills aiming to do the same and to improve capital
prisoners’ access to DNA evidence have been introduced in both houses of the United States
Congress, with bipartisan sponsorship.19

•

Observers in the Wall Street Journal, New York Times Magazine, and Salon and on ABC This
Week see “a tectonic shift in the politics of the death penalty .”20 In April 2000 alone, George
Will21 and Rev. Pat Robertson—both strong death penalty supporters—expressed doubts

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

about the manner in which government officials carry out the penalty in the United States,
and Robertson advocated a moratorium on Meet the Press.22
Fueling these competing initiatives are two beliefs about the death penalty. One is that death
sentences move too slowly from imposition to execution, undermining deterrence and retribution,
subjecting our criminal laws and courts to ridicule, and increasing the agony of victims.23 The other
is that death sentences are fraught with error, causing justice too often to miscarry, and subjecting
innocent and other undeserving defendants—mainly, the poor and racial minorities— to execution.24
Some observers attribute these seemingly conflicting events and opinions to “America’s
schizophrenia—we believe in the death penalty, but shrink from it as applied.”25 These views may
not conflict, however, and Americans who hold both may not be irrational. It may be that capital
sentences spend too much time under review and that they are fraught with disturbing amounts of
error. Indeed, it may be that capital sentences spend so much time under and awaiting judicial
review precisely because they are so persistently and systematically fraught with alarming
amounts of error. That is the conclusion to which we are led by a study of all 4,578 capital
sentences that were finally reviewed by state direct appeal courts, 248 state post-conviction reversals
of capital judgments, and all 599 capital sentences that were finally reviewed by federal habeas
corpus courts between 1973 and 1995.26

II.

Summary of Central Findings
In Furman v. Georgia27 in 1972, the Supreme Court reversed all existing capital statutes and

death sentences. The modern death-sentencing era began the next year with the implementation of
4

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

new capital statutes designed to satisfy Furman. Unfortunately, no central repository of detailed
information on post-Furman death sentences exists.28 In order to collect that information, we
undertook a painstaking search, beginning in 1991 and accelerating in 1995, of all published state
and federal judicial opinions in the U.S. conducting direct and habeas review of state capital
judgments, and many of the available opinions conducting state post-conviction review of those
judgments. We then (1) checked and catalogued all the cases the opinions revealed, and (2) collected
hundreds of items of information about each case from the published decisions and the NAACP
Legal Defense Fund’s quarterly death row census, and (3) tabulated the results.29
Nine years in the making, our central findings thus far are these:
•

Between 1973 and 1995, approximately 5,760 death sentences were imposed in the U.S.30
Only 313 (5.4%; one in 19) of those resulted in an execution during the period.31

•

Of the 5,760 death sentences imposed in the study period, 4,578 (79%) were finally reviewed
on “direct appeal” by a state high court.32 Of those, 1,885 (41%; over two out of five) were
thrown out because of “serious error,” i.e., error that the reviewing court concludes has
seriously undermined the reliability of the outcome or otherwise “harmed” the defendant.33

•

Nearly all of the remaining death sentences were then inspected by state post-conviction
courts.34 Our data reveal that state post-conviction review is an important source of review
in states such as Florida, Georgia, Indiana, Maryland, Mississippi, North Carolina, and
Tennessee.35 In Maryland, at least 52% of capital judgments reviewed on state postconviction during the study period were overturned due to serious error; the same was true

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

of at least 25% of the capital judgments that were similarly reviewed in Indiana, and at least
20% of those reviewed in Mississippi.36
•

Of the death sentences that survived state direct and post-conviction review, 599 were finally
reviewed in a first habeas corpus petition during the 23-year study period.37 Of those 599,
237 (40%; two out of five) were overturned due to serious error.38

•

The “overall success rate” of capital judgments undergoing judicial inspection, and its
converse, the “overall error-rate,” are crucial factors in assessing the effectiveness of the
capital punishment system. The “overall success rate” is the proportion of capital judgments
that underwent, and passed, the three-stage judicial inspection process during the study
period. The “overall error rate” is the reverse: the proportion of fully reviewed capital
judgments that were overturned at one of the three stages due to serious error.39 Nationally,
over the entire 1973-1995 period, the overall error-rate in our capital punishment system
was 68%.40

•

“Serious error” is error that substantially undermines the reliability of the guilt finding
or death sentence imposed at trial.41 Each instance of that error warrants public concern.
The most common errors are (1) egregiously incompetent defense lawyering (accounting
for 37% of the state post-conviction reversals), and (2) prosecutorial suppression of
evidence that the defendant is innocent or does not deserve the death penalty
(accounting for another 16%—19%, when all forms of law enforcement misconduct are
considered).42 As is true of other violations, these two count as “serious” and warrant

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

reversal only when there is a reasonable probability that, but for the responsible actor’s
miscues, the outcome of the trial would have been different.43
•

The seriousness of these errors is also revealed by what happens on retrial, when the errors
are cured. In our state post-conviction study, an astonishing 82% (247 out of 301) of the
capital judgments that were reversed were replaced on retrial with a sentence less than
death, or no sentence at all.44 In the latter regard, 7% (22/301) of the reversals for serious
error resulted in a determination on retrial that the defendant was not guilty of the
capital offense.45

•

The result of very high rates of serious, reversible error among capital convictions and
sentences, and very low rates of capital reconviction and resentencing, is the severe
attrition of capital judgments. As is illustrated by the flow chart below:
1.

For every 100 death sentences imposed and reviewed during the study period, 41
were turned back at the state direct appeal phase because of serious error. Of the 59
that got through that phase to the second, state post-conviction stage, at least46
10%—meaning 6 more of the original 100—were turned back due to serious
flaws. And, of the 53 that got through that stage to the third, federal habeas
checkpoint, 40%—an additional 21 of the original 100—were turned back
because of serious error. All told, at least 68 of the original 100 were thrown out
because of serious flaws, compared to only 32 (or less) that were found to have
passed muster—after an average of 9-10 years had passed.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

And among the individuals whose death sentences were overturned for serious error,
82% (56 in our example) were found on retrial not to have deserved the death
penalty, including 7% (5) who were found innocent of the offense.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

•

High error rates pervade American capital-sentencing jurisdictions, and are geographically
dispersed. Among the 26 death-sentencing jurisdictions with at least one case reviewed in
both the state and federal courts and as to which information about all three judicial
inspection stages is available:
1.

24 (92%) have overall error rates of 52% or higher;

22 (85%) have overall errors rates of 60% or higher;

15 (61%) have overall error rates of 70% or higher.

Among other states, Maryland, Georgia, Alabama, Mississippi, Indiana, Oklahoma,
Wyoming, Montana, Arizona, and California have overall error rates of 75% or
higher.47

•

It sometimes is suggested that Illinois, whose governor declared a moratorium on executions
in January 2000 because of a spate of death row exonerations there,48 generates “uniquely”
flawed death sentences.49 Our data dispute this suggestion: The overall rate of serious error
found to infect Illinois capital sentences (66%) actually is slightly lower than the
nationwide average (68%).50

•

High error rates have persisted for decades. A majority of all cases reviewed in 20 of the 23
study years—including in 17 of the last 19 years—were found seriously flawed. In half of
the years studied, the error rate was over 60%. Although error rates detected on state direct
appeal and federal habeas corpus dropped some in the early 1990s, they went back up in
199551. The amount of error detected on state post-conviction has apparently risen

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

throughout the 1990s.52
•

The 68% rate of capital error found by the three stage inspection process is much higher
than the error rate of less than 15% found by those same three inspections in noncapital
criminal cases.53

•

Appointed federal judges are sometimes thought to be more likely to overturn capital
sentences than state judges, who almost always are elected in capital-sentencing states.54 In
fact, state judges are the first and most important line of defense against erroneous death
sentences. They found serious error in and reversed 90% (2,133 of the 2,370) capital
sentences that were overturned during the study period.55

•

Under current state and federal law, capital prisoners have a legal right to one round of direct
appellate, state post-conviction and federal habeas corpus review.56 The high rates of error
found at each stage—including even at the last stage—and the persistence of high error
rates over time and across the nation, confirm the need for multiple judicial inspections.
Without compensating changes at the front-end of the process, the contrary policy of cutting
back on judicial inspection makes no more sense than responding to the insolvency of the
Social Security System by forbidding it to be audited.

•

Finding all this error takes time. Calculating the amount of time using information in
published decisions is difficult. Only a small percentage of direct appeals decisions report
the sentence date. By the end of the habeas stage, however, a larger proportion of sentencing
dates is reported in one or another decision in the case. Accordingly, it is possible to get a

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

good sense of timing for only the 599 cases that were finally reviewed on habeas corpus.
Among those cases:
1.

It took an average of 7.6 years after the defendant was sentenced to die to complete
federal habeas consideration in the 40% of habeas cases in which reversible error was
found.

In the cases in which no error was detected at the third inspection stage and an
execution occurred, the average time between sentence and execution was 9
years. Matters did not improve over time. In the last 7 study years (1989-95), the
average time between sentence and execution rose to 10.6 years.57

•

High rates of error, and the time consequently needed to filter out all that error, frustrate the
goals of the death penalty system. Figure 1 below compares the overall rate of error detected
during the state direct appeal, state post-conviction, and federal inspection process in the 28
states with at least one capital case in which both inspections have been completed (the
orange line), to the percentage of death sentences imposed by each state that it has carried
out by execution (the red line).58 In general, where the rate of serious reversible error in
a state’s capital judgments reaches 55% or above (as is true for the vast majority of
states), the state’s capital punishment system is effectively stymied—with its proportion
of death sentences carried out falling below 7%.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 1. Overall Error Rate and
Percent of Death Sentences Carried Out, 1973-95
30

100
90

70
60
15

50
Overall Error Rate

Percent of Death Sentences Carried Out

10
30
20
5
10
0

Vir
gin
De
i
law a
are
Lo
uis *
ian
a
Ut
ah
Te
xa
s
Mi
ss
ou
ri
Ar
ka
ns
as
Ge
or
gia
Mo
nta
na
Ne
va
Wa
da
sh
ing
ton
*
Ala
ba
ma
Flo
rid
Ne
a
br
So
as
uth
Ca k a
ro
lin
a
Illi
no
i
Ind s
No
i
an
rth
a
Ca
ro
l
i
n
Ok
lah a
o
Wy ma
om
ing
Ar
izo
na
Ida
ho
Ma
ryl
an
Mi
d
ss
iss
Pe
i
pp
nn
i
sy
lva
n
i
Ca
a
lifo
rni
Ke
a
ntu
ck
Te
y
nn
es
se
e

Percent Reversed

State

Percent Death Sentences Carried Out

25
80

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

The recent rise in the number of executions59 is not inconsistent with these findings. Instead
of reflecting improvement in the quality of death sentences under review, the rising number of
executions may simply reflect how many more sentences have piled up for review. If the errorinduced pile-up of cases is the cause of rising executions, their rise provides no proof that a cure has
been found for disturbingly high error rates. To see why, consider a factory that produces 100
toasters, only 32 of which work. The factory’s problem would not be solved if the next year it made
200 toasters (or added 100 new toasters to 100 old ones previously backlogged at the inspection
stage), thus doubling its output of working products to 64. With, now, 136 duds to go with the 64
keepers, the increase in the latter would simply mask the persistence of crushing error rates.
The decisive question, therefore, is not the number of death sentences carried out each year,
but the proportion. And as Figure 2 below shows:60
•

In contrast to the annual number of executions (the middle line in the chart), the proportion
of death row inmates executed each year (the bottom line) has remained remarkably
stable—and extremely low. Since post-Furman executions began in earnest in 1984, the
nation has executed an average of about 1.3% of its death row inmates each year; in no
year has it ever carried out more than 2.6 percent—or 1 in 39—of those on death row.61

•

Figure 1 thus suggests that executions are increasing, not because of improvements in the
quality of capital judgments, but instead because so many more people have piled up
on death row that, even consistently tiny proportions of people being executed—because
of consistently prodigious error and reversal rates—are prompting the number of

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

executions to rise.62 As in our factory example, rising output does not indicate better
products, and instead seems to mask the opposite.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 2. Persons on Death Row and
Percent and Number Executed, 1974-99
100

3800
3600
Death Row Population

3400

3200

Executions

3000
2800

2400
60

2200
2000

50
1800
1600

1400
1200

1000
800

600
400

200

Year

0
19

Death Row Population

2600

Executions / Percent of Death Row Executed

Percent of Death Row Executed

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 1, p. 11 above, illustrates another finding of interest that recurs throughout this
Report: The pattern of capital outcomes for the State of Virginia is highly anomalous, given the
State’s high execution rate (nearly double that of the next nearest state, and 5 times the national
average) and its low rate of capital reversals (nearly half that of the next nearest state, and less
than one-fourth the national average). The discrepancy between Virginia and other capitalsentencing states on this and other measures63 presents an important question for further study: Are
Virginia capital judgments in fact half as prone to serious error as the next nearest state and 4 times
better than the national average?64 Or, on the other hand, are its courts more tolerant of serious error?
We will address this issue below and in a subsequent report.65

III.

Confirmation from a Parallel Study
Results from a parallel study by the U.S. Department of Justice suggest that our 32%, or

one-in-three, figure for valid death sentences actually overstates the chance of execution:
•

Included in the Justice Department study is a report of the outcome as of the end of 1998 of
the 263 death sentences imposed in 1989.66 A final disposition of only 103 of the 263 death
sentences had been reached nine years later.67 Of those 103, 78 (76%) had been overturned
by a state or federal court. Only 13 death sentences had been carried out.68 So, for every one
member of the death row class of 1989 whose case was finally reviewed and who was
executed as of 1998, six members of the class had their cases overturned in the courts.

•

Because of the intensive review needed to catch so much error, 160 (61%) of the 263 death
sentences imposed in 1989 were still under scrutiny nine years later.69
17

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

•

The approximately 3,600 people on death row today have been waiting an average of 7.4
years for a final declaration that their capital verdict is error-free—or, far more probably, that
it has to be scrapped because of serious error.70

•

Of the approximately 6,700 people sentenced to die between 1973 and 1999, only 598—less
than one in eleven—were executed.71 About four times as many had their capital judgments
overturned or gained clemency.72

IV.

Implications of Central Findings
To help appreciate these findings, consider a scenario that might unfold immediately after

any death sentence is imposed in the U.S. Suppose the defendant, or a relative of the victim, asks a
lawyer or the judge, “What now?”
Based on almost a quarter century of experience in thousands of cases in 28 death-sentencing
states in the U.S. between 1973 and 1995, a responsible answer would be: “The capital conviction
or sentence will probably be overturned due to serious error. It’ll take nine or ten years to find
out, given how many other capital cases being reviewed for likely error are lined up ahead of this
one. If the judgment is overturned, a lesser conviction or sentence will probably be imposed.”73
As anyone hearing this answer would probably conclude as a matter of sheer common sense,
all this error, and all the time needed to expose it, are extremely burdensome and costly:
•

Capital trials and sentences cost more than noncapital ones.74 Each time they have to be done
over—as happens 68% of the time—that difference grows exponentially.

•

The error-detection system all this capital error requires is itself a huge expense—apparently
18

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

millions of dollars per case.75
•

Many of the resources currently consumed by the capital system are not helping the public,
or victims,76 obtain the valid death sentences for egregious offenses that a majority support.
Given that nearly 7 in 10 capital judgments have proven to be seriously flawed, and given
that 4 out of 5 capital cases in which serious error is found turn out on retrial to be more
appropriately handled as non-capital cases (and in a sizeable number of instances, as nonmurder or even non-criminal cases),77 it is hard to escape the conclusion that large amounts
of resources are being wasted on cases that should never have been capital in the first place.

•

Public faith in the courts and the criminal justice system is another casualty of high capital
error rates.78 When most capital-sentencing jurisdictions carry out fewer than 6% of the death
sentences they impose,79 and when the nation as a whole never executes more than 2.6% of
its death population in a year,80 the retributive and deterrent credibility of the death penalty
is low.

•

When condemned inmates turn out to be innocent81—an error that is different in its
consequences, but is not evidently different in its causes, from the other serious error
discussed here82—there is no accounting for the cost: to the wrongly convicted;83 to the
family of the victim, whose search for justice and closure has been in vain; to later victims
whose lives are threatened—and even taken—because the real killers remain at large;84 to
the public’s confidence in law and legal institutions; and to the wrongly executed, should
justice miscarry at trial, and should reviewing judges, harried by the amount of error they

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

are asked to catch, miss one.85
If what were at issue here was the fabrication of toasters (to return to our prior
example), or the processing of social security claims, or the pre-takeoff inspection of
commercial aircraft—or the conduct of any other private- or public-sector activity—neither
the consuming and the taxpaying public, nor managers and investors, would for a moment
tolerate the error-rates and attendant costs that dozens of states and the nation as a whole have
tolerated in their capital punishment system for decades. Any system with this much error and
expense would be halted immediately, examined, and either reformed or scrapped.
The question this Report poses to taxpayers, public managers and policymakers, is
whether that same response is warranted here, when what is at issue is not the content and
quality of tomorrow’s breakfast, but whether society has a swift and sure response to murder,
and whether thousands of men and women condemned for that crime in fact deserve to die.
*****
The remainder of this Report more fully describes our findings. Part V describes the review
process for capital sentences. Part VI describes our study methodology. Parts VII, VIII and IX more
thoroughly document and display our findings about the frequency with which reversible error is
found in capital judgments in the United States between 1973 and 1995, and the time taken to find
those errors. Part VII examines relevant factors at the national level. Part VIII does so using
comparative analyses of the 28 capital-sentencing states in which at least one case had advanced
through the entire post-sentence inspection process. And Part IX does the same thing, comparing the

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

8 federal judicial circuits and corresponding regions into which they are divided. After presenting
a variety of information, Parts VII, VIII and IX preliminarily address the potential causes of so much
error in capital sentencing. Finally, Part X briefly describes the more sophisticated analyses we will
undertake in the next phase of our study (to be published in the Fall) to set the stage for proposed
reforms.

The Capital Review Process
This phase of our study asks what state and federal courts discovered when they inspected

capital convictions and sentences imposed during the 23-year study period. In a later phase, we will
consider some candidate causes of the evidently irrational patterns of error that those courts have
detected. In order to frame these questions, we first describe the capital-inspection process whose
results we are studying.
A.

First Inspection: State Direct Appeal

In Furman v. Georgia and later cases, the Supreme Court suggested that state high courts
were required to review all death sentences on direct review.86 As a consequence, the law of nearly
all states requires that capital judgments be automatically appealed.87 And as a matter of fact,
virtually all capital judgments are appealed.88 In all but two of our study states, that appeal ran
directly from the trial court to the highest court in the state with criminal jurisdiction, which is
typically the state supreme court or, as in Oklahoma and Texas, a “court of criminal appeals.”89 In
Alabama and Ohio, there were two rounds of appeals in the state direct review process—first to an

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

intermediate court of criminal appeals, and then to the state supreme court.90 Reversal of a capital
conviction or sentence on direct appeal requires a showing of “serious error” as defined earlier.91
In nearly all cases in which the direct appeal decision runs entirely against the defendant, he
or she seeks certiorari in the United States Supreme Court.92 Although in the vast majority of cases,
the Supreme Court denies review, it occasionally undertakes merits review and either affirms or
reverses.93 Certiorari proceedings are typically understood to be a part of the direct review, or prefinality, stage of a criminal case,94 and they are treated that way here. If the Supreme Court reversed
a capital conviction or sentence on direct review of the state high court’s decision, we counted that
decision as a direct-appeal finding of serious (indeed, in all such cases, federal constitutional95) error.
B.

Second Inspection: State Post-Conviction

In order to seek federal habeas review of a constitutional claim, the prisoner must have
exhausted at least one full round of state judicial remedies for the claim.96 There are certain kinds
of claims that cannot easily be exhausted at trial and on direct appeal because the defendant cannot
discover or adequately litigate the facts or the legal principles supporting the claims at trial or on
direct appeal.97 This sometimes occurs (1) because a police officer, prosecutor or other state actor
has suppressed the relevant facts (which may itself have violated the Constitution, as when the
suppressed facts show the defendant is innocent,98 or may keep the defendant from establishing the
violation of some other principle, as when police suppressed evidence that they coerced the
defendant into confessing, or when the prosecutor hid his efforts to keep African-Americans off of
criminal juries99); (2) because the agent of the violation was the defendant’s own trial or direct

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

appeal attorney (as in the case of ineffective assistance of counsel), thus preventing the defendant
from recognizing or fairly litigating the claim;100 (3) because the evidence establishing the claim was
not reasonably available to the defense at the time of trial or appeal for some other reason101 (as
when counsel later discovers that the trial judge was corrupt102 or biased,103 that a juror lied during
the jury selection process,104 or that the bailiff secretly lobbied the jury to convict or condemn105);
or (4) because the legal rule establishing the claim did not exist at the time of trial or appeal and the
rule applies “retroactively” to the prisoner’s case.106
Because the Supreme Court has suggested that states are constitutionally required to provide
adequate state post-conviction remedies for federal constitutional claims that cannot properly be
pursued at trial and on direct appeal,107 and because federal habeas law rewards states when they do
provide such remedies,108 all states now do so.109 State capital prisoners seeking to preserve their
access to federal habeas review accordingly are obliged to exhaust those remedies, and the
professional obligation of capital attorneys to subject their clients’ convictions and sentences to
searching scrutiny compels them to pursue state post-conviction review in nearly all capital cases.110
State post-conviction review takes a variety of forms under a variety of names (e.g., habeas
corpus, coram nobis, extraordinary motion for new trial, and state post-conviction procedures acts).
Traditionally, such proceedings have taken place after the completion of state direct appeal and have
entailed the filing of a petition for review with the judge who presided over the original trial, and the
appeal of any adverse rulings up to an intermediate state appellate court and then to the state high
court.111 More recently, an increasing number of states (1) have adopted “unitary appeal” procedures

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

that require direct appeal and state post-conviction proceedings to take place nearly
simultaneously,112 and/or (2)

have required prisoners to commence state post-conviction

proceedings in a state intermediate or high court that either can grant or deny state post-conviction
relief once and for all, or can remand the case to a trial court to take evidence. In most states, state
post-conviction review is limited to claims that were not and could not have been raised on direct
appeal and that arise under state or federal constitutional law.113
Most capital prisoners also seek U.S. Supreme Court review on certiorari of adverse state
post-conviction proceedings, which the Supreme Court (very) occasionally grants.114 In the event that
the Court does so, and grants relief, our classification scheme counts that decision as part of the state
post-conviction inspection phase.115
C.

Third Inspection: Federal Habeas Corpus

Because federal habeas corpus practice is controlled by federal statute,116 it is far more
uniform across states than are direct appeal and state post-conviction proceedings. Habeas
proceedings begin with the filing of a petition in a United States District Court in the state in which
the defendant was convicted and is incarcerated.117 If relief is denied, and if (but only if) the prisoner
can show that his petition presents a substantial constitutional claim, he may appeal the denial to a
federal circuit court,118 and if the district court opinion is affirmed and a stay of execution is
available, he may petition the Supreme Court for certiorari.119 Although habeas proceedings at the
district court level are a matter of statutory right, stays of execution are not, thus limiting capital
habeas proceedings to cases in which the prisoner can secure a federal stay of execution based on

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

a substantial constitutional claim.120 Habeas relief is limited to a category of “serious error” that is
even narrower than the analogous of category of “serious” direct-appeal error.121
A stylized depiction of the post-trial review process in capital cases that we are studying here
is set out below.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

VI.

The Study
This study began in 1991 when the Chair of the Senate Judiciary Committee asked the lead

author of this Report to calculate the frequency of relief in capital habeas corpus cases.122 Simply
identifying the relevant cases turned out to be a monumental task, because there is no single
repository of capital habeas corpus decisions either nationally or even (especially at the time) in most
death-sentencing states, and key-word searches of reported cases are substantially under-inclusive
(because some decisions that are capital are not identified as such) and over-inclusive (because many
cases in which a death sentence was not imposed either began as capital cases or refer to capital
cases). Working with volunteer law student assistants, therefore, the senior author undertook a
painstaking search for capital habeas cases relying on (1) the NAACP Legal Defense Fund’s (LDF’s)
quarterly death row census,123 (2) computerized and book research, and (3) a series of conversations
with staff members of state death penalty resource centers and other local death penalty lawyers who
were familiar with some of the cases and death row inmates in their states.
In late 1995, the study was expanded from a simple count of cases and their outcomes to a
search for information that might help explain why relief is granted in so many capital habeas cases.
In that year, a team with social scientific expertise was assembled, and began collecting
approximately 1300 items of information about each case—relating to defendants, victims, offenses,
evidence, lawyers, judges, timing, claims, defenses, court procedures, and the like. We soon
determined that the only reasonably accessible source of this kind of information was published
judicial decisions of federal habeas courts themselves and of state courts when they denied relief at

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

earlier inspection stages.124
During 1996, 1997 and 1998, the senior authors developed, tested and revised a study
instrument, developed and fine-tuned a set of research protocols, assembled and trained a series of
law student researchers to collect the information called for by the study information, periodically
checked and rechecked their completed forms, and in this way collected data on 599 initial federal
habeas corpus cases and 173 second or successive federal habeas corpus cases. The research protocol
called for researchers first to identify the “final federal habeas corpus decision” (the decision of the
last and highest federal court to finally resolve the merits of the habeas application), then to identify
all other available state and federal decisions addressing the same capital judgment (i.e., either the
capital conviction, sentence or both), and then to extract from each of those decisions a variety of
information that was then coded onto the research instrument. Beginning in 1997 and continuing
through 1999, the information on the study instrument in each case was entered into a data base and
again checked and rechecked.
We collected the results of all federal habeas corpus decisions that became “final” between
January 1, 1973 and October 2, 1995.125 By “final,” we mean that (1) the highest federal court to
which the case has been timely brought either by the filing of a petition or an appeal has finally ruled
on the validity of the capital judgment (meaning both the conviction and death sentence), (2) the time
for reconsideration or rehearing by that court has passed, and (3) the time for U.S. Supreme Court
review has passed without that Court’s choosing to review the decision or, if it did choose to review
it, with its own final merits decision having been rendered. Here again, a finding of “serious error”

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

is made only if the capital conviction, the capital sentence, or both were overturned due to
prejudicial, reversible error.126
Early on, it appeared that a major factor in determining outcomes in federal habeas cases was
the state that imposed the capital judgment under review. For example, although judges of the same
(Eleventh) federal circuit court reached nearly all of the final federal habeas decisions in cases from
Florida, Alabama and Georgia, their reversal rates in cases emanating from each of those three states
were quite different (respectively, 37%, 45%, and 65%), suggesting that there was something about
each particular state’s death sentences that made them more or less error-prone.127 To study this
possibility, we collected information (in 1997 through 1999) about how states differ in regard to their
demography, law, politics, judicial organization and funding, death-sentencing history and the like.
An early hypothesis in this regard was that the rate of error found by federal habeas
proceedings might be related to the rate of error found in state direct appeals—either because lax
state inspections might impose extra work on later federal ones (suggesting an inverse relationship
between error rates found at the two stages), or because excessive amounts of error might overwhelm
judges at the first checkpoint, permitting considerable remaining error to slip through and be caught
(if at all) by judges at a later checkpoint (suggesting a more direct relationship between error rates
found at the two stages).128 To test this hypothesis, we collected information about each state’s
capital direct appeal outcomes—prompting our second major study, covering the approximately
4,600 state direct appeal decisions during the 1973-1995 study period. Working back and forth from
the LDF death row census and computerized legal research data bases, we compiled a list of all

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

capital direct appeal decisions in the study period, then collected a small set of information about
each case from published opinions that our search identified.
We collected the results of all direct appeal decisions that became “final” between January
1, 1973 and December 31, 1995. By “final,” we mean that (1) the highest state court with jurisdiction
over the appeal had finally ruled on the validity of the judgment (meaning both the conviction and
death sentence), (2) the time for reconsideration by that court had passed, and (3) the U.S. Supreme
Court did not review the decision or, if it did review it, had rendered a final merits decision by the
end of 1995.129 A finding of “serious error” was made if reversible error was found and the capital
conviction, sentence or both were overturned.130
Substantially later in the process, we began collecting data on state post-conviction
outcomes. Those data are especially hard to find. Unlike state direct appeal decisions and appellatelevel federal habeas decisions, which almost always are published in capital cases, state postconviction decisions often are not published, even in capital cases. This is particularly so because
state post-conviction review often begins—and when it leads to reversal, ends—in trial courts that
almost never publish their decisions.131 Nor is there any central repository of information about when
and where capital state post-conviction petitions are pending, making it difficult to ascertain (1) the
number of state post-conviction cases that actually were decided at that stage during the study period
(as opposed to the number that were available for resolution at that stage, because they had “cleared”
state direct appeal) and, thus, (2) the proportion of actually decided cases in which “serious error”
was found.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

For these reasons, as is more fully described in the introduction to Appendix C, we limited
our collection of state post-conviction data to a list of known state post-conviction reversals of
capital judgments in the study states in which capital cases had progressed significantly beyond the
direct appeal stage by the end of 1995. This list, set out in full in Appendix C, enables us to derive
an interesting, though incomplete, picture of the rates of error detected by state post-conviction
courts in reviewing death sentences. To do so, we make three obviously inaccurate, but reliably
conservative, assumptions: First we assume that we have a complete list of capital state postconviction reversals due to serious error that occurred during the study period. In fact, our list is
incomplete, although it probably contains most such reversals. Second, we assume that every capital
case that was available for state post-conviction review because it had “cleared” direct appeal during
the 1973-1995 study period was finally decided on state post-conviction during that period. In fact,
many of the “available” cases were not finally decided and were still being litigated on state postconviction as of the end of 1995. Taken together, these two assumptions lead to a third assumption
—that every capital judgment that was available for state post-conviction review and is not known
to have been reversed due to serious error during the study period was affirmed. Calculating error
rates in this manner systematically underestimates those rates (and overstates success rates) by (1)
underestimating the numerator (the number of serious errors found, which we have undercounted)
and (2) overstating the denominator (the number of cases finally reviewed for serious error, for
which we have substituted the obviously larger number of cases available for review).132
Accordingly, our estimates of the rate of serious error found on state post-conviction review are

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

understated and conservative.
Analysis of the data collected in our habeas corpus and direct appeal studies began in earnest in mid1999 and continues at this writing, along with analyses of our newer, state post-conviction data. This
Report presents the findings of our initial analyses. These focus on the basic operation and outcomes
of the post-trial system for reviewing capital judgments: How many and what proportion of death
sentences were reviewed at each of the three inspection stages during the study period—nationally,
in each capital-sentencing state, and in each federal judicial circuit and corresponding geographic
region? How much error was found, and by whom? How long did the process take? How do states
compare in their sentencing and execution rates and along other dimensions that might help explain
differences in the frequency of capital-sentencing error?

VII. The National Capital Punishment Report Card
In this Part, we present a “national composite capital-sentencing report card.” The report card
describes a variety of information, including the error rates found to characterize, and the time
needed to review, death sentences in capital states during the study period. This Part also explains
the two-page report card format that we use to report state, federal judicial circuit and regional as
well as national data. (In Part VIII below, we present state-by-state comparisons of the information
on state report cards for the 28 death-sentencing states in which at least one final direct appeal and
federal habeas decision occurred during the 1993-1995 period.133 In Part IX below, we present
similar comparisons of information on the federal judicial circuit court/regional report cards.134)

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

The national capital-sentencing report card is set out below. It combines information about
the rates of error detected on direct appeal of capital judgments imposed in all 34 death-sentencing
states in which at least one state capital direct appeal was completed during the 1973-1995 study
period. Our 34-state cohort is: Alabama, Arizona, Arkansas, California, Colorado, Connecticut,
Delaware, Florida, Georgia, Idaho, Illinois, Indiana, Kentucky, Louisiana, Maryland, Mississippi,
Missouri, Montana, Nebraska, Nevada, New Jersey, New Mexico, North Carolina, Ohio, Oklahoma,
Oregon, Pennsylvania, South Carolina, Tennessee, Texas, Utah, Virginia, Washington and
Wyoming. Because capital cases in six states (Colorado, Connecticut, New Jersey, New Mexico,
Ohio and Oregon) had not advanced far, or at all, into the state post-conviction stage of review, and
no case from those states had completed federal habeas review, the bulk of the composite data—
those covering the state post-conviction and federal habeas stages and the “overall rates”—omit
these states and focus on what we call our 28-state cohort.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

National Composite Capital Punishment Report Card, 1973-1995
History (34 States)
1973
First Death Sentence
First Direct Appeal
First Consensual Execution
First Non-Consensual Execution

1973
1977
1979

Sentences and Executions (34 States)
Total Number of Death Sentences
Total Number of Executions
Percentage of Death Sentences Carried Out

5,760
313
5%

Error Rates
State Direct Appeal (34 States)
Number Reviewed on Direct Appeal
Number Reversed on Direct Appeal
Percentage Reversed on Direct Appeal
Number Awaiting Direct Appeal
Percentage Awaiting Direct Appeal
Number Forward to State Post-Conviction

4,578
1,885
41%
1,182
21%
2,693

State Direct Appeal (28 States)
Number Reviewed on Direct Appeal
Number Reversed on Direct Appeal
Percentage Reversed on Direct Appeal
Number Forward to State Post-Conviction

4,364
1,782
41%
2,582

State Post-Conviction (28 States)
Number Reviewed on Post-Conviction
Number Reversed on Post-Conviction
Percentage Reversed on Post-Conviction

Unknown
≥248
≥10%
34

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Number Forward to Federal Habeas Corpus

Unknown

State Direct Appeal and State Post-Conviction Combined (28 States)
Overall Rate of Error Found by State Courts

≥47%

Error Rates (Continued)
Federal Habeas Corpus (28 States)
Number Reviewed on Habeas
Number Reversed on Habeas
Percentage Reversed on Habeas

599
237
40%

Overall Rates Including [and Excluding] State Post-Conviction (28 States)
Overall Error Rate
Overall Success Rate

68% [64%]
32% [36%]

Time (28 States)
Time From First Death Sentence to First NonConsensual Execution
Average Time from Sentence to Execution
Average Time from Sentence to Final Federal Relief

6
9
7.6

Sentencing and Execution Rates (34 States )
Death States

Death Sentences per 1000 homicides
Death Sentences per 100,000 pop.
NC Executions per 1000 homicides

14.90
3.9
.68

Whole Nation

12
2.46
.54

Demographic Information (34 States)
Death States
35

Whole Nation

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Average Population
Average Homicides
Average Homicides Rate per 100,000 Population
Percentage Population Non-White

181,374,347
16,860
9.3
19%

237,905,964
21,197
9
20%

Sources: DRCen; Death Row U.S.A., Winter 2000; DADB; HCDB; Appendix C; UCRDB; USCen

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

The National Composite Capital Punishment Report Card contains six categories of
information for either the 34 capital-sentencing states in which at least one capital judgment had
been finally reviewed on state direct appeal during the study period, or for the 28 of those states in
which at least one capital judgment was finally reviewed on federal habeas during the study period.
Because the same six categories appear on all of the succeeding report cards—along with a seventh
category in the state report cards—this section describes the types of information that all seven
categories contain, then discusses the actual national composite results for each category.
A.

Capital-Sentencing History

In the “History” category of each report card is information about the years in which four
important capital-sentencing events occurred in the jurisdiction or set of jurisdictions in question
following the Supreme Court’s invalidation of all preexisting capital statutes and sentences in
Furman v. Georgia.135 The requisite information for the nation as whole (in this case comprised of
the 34-state cohort of capital-sentencing jurisdictions) as revealed by the top category in the National
Report Card is as follows:
•

The first post-Furman death sentences were meted out in 1973.136

•

The first post-Furman state direct appeal decision finally determining the legality of a
post-Furman death sentence also occurred in 1973.137

•

The first post-Furman execution of any sort (Gary Gilmore’s consented-to execution by the
State of Utah) was in 1977.138

•

The first “non-consensual” execution after Furman—i.e., the first time an American

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

jurisdiction carried out a post-Furman capital judgment that had passed inspection by all
available levels of judicial review—was in 1979, when Florida executed John Spenkelink.139
We focus on non-consensual, as well as all, executions because we are interested in error
rates, and only non-consensual executions reveal the inspection system’s conclusion that the
death sentence is free of “serious error” as defined above.140
B.

Sentences and Executions

This section reports the number of death sentences imposed, the number of executions
carried out, and executions as a proportion of death sentences in each jurisdiction or group of
jurisdictions. Nationally, during the years 1973-1995, 34 American jurisdictions imposed 5,760
death sentences141 and carried out 313 executions.142 In other words, only 5% of the death
sentences imposed were carried out.
C.

Error and Success Rates

The third section of the report cards identifies (1) the rates of serious, reversible error
discovered at each level of judicial inspection,143 (2) the overall error rate, meaning the proportion
of capital judgments undergoing judicial inspection that were thrown out before reaching the end of
the inspection process,144 and, conversely, (3) the overall success rate, meaning the proportion of
capital judgments found after full review to be free of serious error. In the “overall rates” category,
we give the error and success rates considering all three judicial inspections and also, in brackets,
the rates considering only the direct appeal and federal habeas inspections. Nationally, our data
reveal that:

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

•

Direct appeal. State courts in 34 capital-sentencing jurisdictions finally reviewed 4,578
death sentences on direct appeal during 1973-1995.145 Because 5,760 death sentences were
imposed during that period, this figure reveals that 1,182—or 21%—of the death sentences
were awaiting direct appellate review at the end of the study period.146 Of the 4,578
capital judgments finally reviewed on direct appeal, 1,885—or 41%—were overturned
based on serious error.147 This means that 2,693 death sentences148 from 34 jurisdictions
passed the first judicial inspection and were available to be reviewed at the second, state
post-conviction stage of review.149 Many of our subsequent analyses focus on the 28 capitalsentencing jurisdictions in which a full complement of review procedures took place in at
least one capital case between 1973 and 1995. On the national report card, therefore, we
calculate the direct appeal error rates a second time for just the 28 states. That analysis
reveals the same 41% rate of serious error detected at that stage (1,782 capital judgments
overturned due to serious error, out of 4,364 reviewed at that stage), and shows that 2,582
capital judgments from the 28-state cohort passed the first judicial checkpoint and were
available for state post-conviction review.150

•

State post-conviction. As is discussed above and in Appendix C, our state post-conviction
data include only known state post-conviction reversals during the 23-year study period; it
does not contain information about the number of state post-conviction proceedings that
actually were completed during that period.151 For that reason, each report card lists as
“Unknown” both the “Number Reviewed on Post-Conviction” (i.e., the number of capital

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

judgments that went forward to state post-conviction review and were finally reviewed
there), and also the “Number Forward [from State Post-Conviction] to Federal Habeas
Corpus.” What we are able to calculate is the known reversals152 as a proportion of the
number of capital judgments moving forward from state direct appeal to state post-conviction
in our 28-state cohort of capital-sentencing jurisdictions.153 Although we report this
calculation as the rate of error discovered on state post-conviction—i.e., as the “Percentage
Reversed on [State] Post-Conviction”—we in fact underestimate that error rate by a
substantial amount, because we take the known reversals as a percentage of the cases
available for review, rather than as a proportion of the cases actually reviewed, during the
study period.154 That underestimation accounts for our use of the “≥” symbol in this row of
each report card and our use of the phrase “at least” in discussing that row. Nationally, for
the relevant 28 study states, there were at least 248 state post-conviction reversals due to
serious error during the study period, so that serious error was found in more than—
probably significantly more than—10% of the cases reviewed at that stage.155 Although
state post-conviction proceedings are not generally thought to be major sources of postsentencing reversals of seriously flawed capital judgments, in fact there were more state
post-conviction findings of reversible error infecting American capital judgments (248)
than there were analogous federal habeas findings (237).
•

State direct appeal and state post-conviction combined. This item in the national report
card indicates the combined rates of error found at the two state court checkpoints.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Nationally, state courts as a whole found 47%—nearly one out of every two—capital
judgments they reviewed to be infected with serious error.156
•

Federal habeas corpus. Between 1973 and 1995,157 federal habeas courts with jurisdiction
over prisoners in capital-sentencing states around the nation finally reviewed158 599 death
sentences. They overturned 237—or 40%—of those sentences based on serious error.159

•

Overall rates: This portion of the report card gives the overall error (and success) rates,
meaning the proportion of capital judgments from the relevant jurisdiction or set of
jurisdictions that underwent full judicial inspection and were found to have (and to be free
of) serious error.160 Overall, between 1973 and 1995, less than one-third—32%—of all
death sentences passing through the nation’s state and federal judicial inspection
system were cleared of serious error. Conversely, over two-thirds—68%—were thrown
out because of serious error.161
The information presented thus far make this a useful place to discuss error rates over time.

Earlier, in discussing Figure 2, pp. 12-13 above, we touched on patterns of capital error and success
rates over time. There, we noted that, although executions have been on the rise since 1988, the
principal cause of that rise seems to be the steady increase in the number of individuals piled up on
death row who are potentially available to be executed, and not any sharp increase in the success rate
of capital judgments. Figures 3 and 4 below look at patterns over a longer period of time, beginning
in 1973.
Figure 3 below depicts the rates of error detected on state direct appeal, federal habeas

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

corpus, and in those two stages combined, by year, from 1973 to 1995.162 (The first two years for
which we plot the rate of error found on federal habeas review are 1978 and 1980, because no capital
habeas proceedings were completed before 1978 or during 1979.) Figure 3 reveals the following
about error rates detected over time during the first (state direct appeal) and third (federal habeas)
inspection stages:
•

From the 1970s through 1982, when relatively few cases were under review, rates of error
detected on state direct appeal and federal habeas review were extremely volatile and high.

•

As of 1983, as larger numbers of capital judgments came under review at both stages, error
rates stabilized, and they remained relatively stable throughout the remainder of the period.
Thus, during the final 13 study years (1983-1995):
1.

Capital-sentencing error rates found on state direct appeal across the nation
consistently remained within the 30% to 45% range.

With the exception of three years (two with a lower rate; one with a higher rate),
capital error rates found on federal habeas review stayed within that same 15point range.

With one exception, the combined error rates detected at those two stages stayed
consistently within the 54% to 69% range.

Broadly speaking: while the error rate found on federal habeas modestly dipped
during the 1987-1991 period, the error rate found on state direct appeal (affecting a
much larger pool of cases) modestly rose during that same period. Both rates dipped

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

some during the years 1992 through 1994, then rose sharply in 1995.
Figure 3 is incomplete because it does not contain rates of serious error found, by year, at the
state post-conviction stage, nor thus any overall reversal rate, by year, for the three inspection stages
as a whole. Rates of serious error detected during state post-conviction review cannot be calculated
because only data on the number of reversals— but not on the total number of cases decided, and
thus the reversal rate—are available by year for the state post-conviction stage.163 The next chart,
however —Figure 4—provides some information about state post-conviction error rates over time,
revealing that in the same years when a modest downward trend in federal habeas reversal rates was
occurring (1987-1994), a marked increase in state post-conviction reversals occurred.164 If we
assume (though we can’t know for certain) that the number of capital state post-conviction cases
finally decided during the 1985-1994 period was fairly steady, then an increase in the error rate
detected at the state post-conviction stage would have occurred and offset the decrease in the federal
habeas reversal rate. Making that assumption leads to an estimate of the overall rate of error
detected by all three judicial inspections during the 1988-1994 period of roughly 60-65%.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 3: Error Rates Detected on State Direct Appeal, Federal
Habeas, and the Two Combined, 1973-95
100
% Reversed on Direct Appeal

% Reversed on Habeas
Combined Reversal Rate

60
50
40
30
20
10

Year

95
19

94
19

93
19

92
19

91
19

90
19

89
19

88
19

87
19

86
19

85
19

84
19

83
19

82
19

81
19

80
19

79
19

78
19

77
19

76
19

75
19

74
19

0
19

Percent Reversed

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 4. Known State Post-Conviction
Reversals, 1973-2000
30

Year

0*
00

-2

95
19

93
19

92
19

89
19

85
19

81
19

77
19

74
19

0
19

Number of Reversals

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Length of Time of Review

This section of each report card provides information for the relevant jurisdiction or set of
jurisdictions on (1) the number of years that elapsed between the state’s first death sentence and its
first non-consensual execution (not necessarily in the same case); (2) the average number of years
it took death sentences to proceed through the three-stage inspection process to execution in the
small proportion of cases in which an execution took place; and (3) the average time from death
sentence to federal habeas reversal in the 10% of cases in which reversal occurred at the third
(federal habeas) checkpoint, as opposed to taking place at one of the earlier (state court) checkpoints.
The national report card reveals the following about the length of time required to identify the high
amounts of error described above:
•

Nationally, 6 years passed between the imposition of the first death sentence and the first
non-consensual execution.165

•

We don’t know how much time was required for judicial inspection of death sentences at the
direct appeal and state post-conviction stages. One of the report-card categories discussed
above—the percentage of death sentences awaiting direct appellate review as of the end of
1995—does, however, suggest the extent to which the direct appeal stage is a bottleneck in
the review process. Nationally, 21% of all death sentences imposed between 1973 and
1995 were still awaiting a state direct appeal decision as of 1995.166 That 21% (1,182
death sentences) represents close to five years’ worth of death sentences backed up at the
direct appeal stage as of the end of 1995, at the average annual rate of 250 death sentences

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

imposed per year.167 This suggests that, as of 1995, an average of about 5 years was elapsing
between imposition of a death sentence and the end of state direct appeal—and thus that
about half of the time required for the entire review process was being consumed by the
first, state direct appeal inspection.
•

In the minority of cases in which death sentences passed the three-stage inspection and
were carried out by execution, the average time, nationally, from sentence to execution
was 9 years. In the last 9 study years (1987-1995), by which point the pile-up on death row
was substantial (see Figure 2, p.12 above), the average time from sentence to execution had
increased to 10.6 years.168

•

In cases in which serious error was detected during the third, federal habeas review,
the average time from sentence to federal reversal was 7.6 years.169
E.

Capital-Sentencing and Execution Rates

The report cards next answer two questions. (1) How often does the relevant jurisdiction or
set of jurisdictions impose death sentences? To answer this question, we consider death sentences
as a proportion of three populations: per 1,000 homicides, per 100,000 population, and (in the state
report cards, but not the national one) per 1,000 incarcerated inmates in the jurisdiction. (2) How
often (relative to homicides, population and prison population) does the jurisdiction execute
offenders? Because we are interested in success rates, we consider only “non-consensual”
executions, i.e., ones based on capital judgments that have been fully reviewed and found to be free
of serious error.170 Because not all states have the death penalty, our national report card computes

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

these figures for the nation as a whole and for our 34-state cohort.
These numbers are most useful for the comparative purposes to which we put them below.171
Providing a national baseline for those comparisons, the capital-sentencing and execution rates for
the nation as a whole, and for the 34 death-sentencing states that decided at least one direct appeal
during the study period, are as follows:
•

For the 34 capital-sentencing states, an average of 14.9 death sentences were imposed for
every 1,000 homicides during the study period. For the same states, an average of 3.9 death
sentences were imposed for every 100,000 people during the same period.172

•

Because so few death sentences actually result in executions, the execution rates determined
by each of these population categories are much lower. During the study period, deathsentencing states carried out an average of: .68 executions for every 1,000 homicides; and
.15 executions for per 100,000 persons.173

•

Comparing the last two points reveals that during the study period, death states capitally
sentenced 22 times more defendants per 1,000 homicides than they executed. And they
sentenced 26 times more defendants per 100,000 population, than they executed.

•

For the whole nation during the study period, an average of: (1) 12 death sentences were
imposed for every 1,000 homicides; (2) 2.46 death sentences were imposed for every
100,000 people; and (3) .54 non-consensual executions were carried out for every 1,000
homicides.174

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Demographic Factors

The demographic information reported in the sixth report card category reveals the
population pools against which each jurisdiction’s number of death sentences and executions are
compared to determine sentencing and execution rates. They also provide bases for distinguishing
among states and thus, potentially, for explaining variations among states in terms of the capital error
rates detected on direct appeal and habeas corpus inspection. At the national level we again report
data for the 34 death states as well as for the nation at large.
“Average population” is the relevant jurisdiction’s yearly average population from 19731995. For the whole nation, the average population during the study period was 237,905,964. For
the 34 death states, it was 181,374,347.175
“Average homicides” are the total number of homicides from 1973-1995 divided by 23, the
number of years in our study . For the whole nation, the average number of homicides each year
during the study period was 21,197. For the 34 death penalty states, it was 16,860.176 Comparing this
and the last category reveals that death-sentencing states account for about 76% of the nation’s
population and about 80% of its homicides.
Homicides per population establishes a jurisdiction’s homicide rate. By “average
homicides/average population,” we mean the number of homicides per year for every 100,000
persons in the jurisdiction, averaged over the population during the study period. For the whole
nation, average homicides/average population during the study period was 9. For the 34 death

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

sentencing states, it was 9.3.177 This again reveals that homicide rates are slightly higher in deathsentencing than in nondeath-sentencing states.
“Average prison admissions” means the average number of persons admitted each year to the
state’s prisons during the study period.178
“Average prison population” means the jurisdiction’s average population over study
period.179
We also report here the percentage of each jurisdiction’s population during the study period
that was nonwhite, which for the nation as a whole was 20% and for the 28 study states was 19%.180
G.

Court Factors: The Context of State Court Decision Making

In the state (but not the national and circuit/regional) report cards, we report four measures
of the social and political contexts in which judges make decisions. Contextual measures such as
those analyzed here have been shown in empirical studies to help explain variation in sentencing
from county to county within states and across states.181 We consider them here to see whether they
can help explain state variations in capital-sentencing error rates, and also in capital-sentencing and
execution rates themselves.
The “political pressure” index measures the extent to which state judges are subject to
electoral scrutiny and discipline. Although nearly all the state judges in our study are subject to
election at some point if they wish to remain in office, the forms and frequency of elections differ
in ways that are likely to increase or decrease the extent to which judges are put at political risk
because of the capital outcomes produced in their courts (meaning, at the trial level, whether the

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

verdict was death or life and, at the appellate level, whether a death sentence under review was
affirmed or reversed). The index considers whether judges initially are elected or appointed, whether
judicial elections are partisan, the length of judges’ terms of office, and whether judges’ continuation
in office is determined by contested or retention elections.182
The “party competition index” is a composite of the vote share of each party in state
gubernatorial elections from 1968-1996.183
Our penultimate (“state court criminal caseload”) item reports the yearly average number of
criminal case filings in each jurisdiction from 1985-1994 per 1,000 people in the population.184 We
include this figure to test the hypothesis that high criminal caseloads may in some way affect the
quality of state-court capital judgments.
Finally, aiming to test a similar hypothesis having to do with available judicial resources, we
report each state’s average annual court-related expenditures during the fiscal years 1982-1992.185

VIII. State Comparisons
Appendix A to this Report presents capital punishment report cards for each jurisdiction in
our 28-state cohort, arranged alphabetically. Observers and policymakers in each state may find their
state’s report card to be interesting in and of itself. The report cards are especially informative,
however, when used comparatively. With the help of a number of tables and figures, this section
undertakes a variety of state-by-state comparisons.
A.

Rates of Serious Error Found on State Direct Appeal

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 4 and Figure 5 below compare the rates of capital error discovered on direct appeal
during the 23-year study period in each of the 28 study states to the rates in the other states and to
the national composite of 41%. Table 4 and Figure 5 show that at the first state inspection stage,
elected high court judges in a large majority (64%) of American capital-sentencing states
found that over a third of their states’ capital judgments were seriously flawed. In well over
half the study states, state high court judges found serious error in 40% of more of their capital
judgments. The error rate found on direct appeal was 50% or more in a quarter of American deathpenalty jurisdictions.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 4: Percent of Capital Judgments Reviewed on Direct Appeal
in Which Reversible Error Was Found, 1973-1995
State
National Composite
1. Wyoming
2. Mississippi
2. North Carolina
4. Alabama
5. South Carolina
6. Maryland
7. Kentucky
8. Florida
9. Oklahoma
10. Louisiana
11. Washington
12. Arizona
12. Idaho
12. Montana
15. Arkansas
16. Illinois
17. Georgia
17. Utah
19. Indiana
20. California
20. Texas
22. Nevada
23. Nebraska
23. Tennessee
25. Pennsylvania
26. Delaware
27. Missouri
28. Virginia

Percent Reversed on Direct Appeal
41
67
61
61
55
54
53
50
49
48
46
45
42
42
42
40
39
35
35
32
31
31
30
29
29
28
26
17
10

Source: DADB

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 5. Percent of Death Sentences Reversed
on State Direct Appeal, 1973-95
80

Percentage Reversed on Direct Appeal

Wy
om
Mi
ing
s
si
No
rth s s i p
pi
Ca
ro
lin
So Alab a
uth
am
a
Ca
ro
l
i
n
Ma
a
ryl
an
Flo d
Ke rida
ntu
ck
Ok
lah y
om
Lo
a
u
Wa isian
sh
a
ing
ton
Ar
izo
na
Ida
Mo ho
nt
Ar ana
ka
ns
as
Illi
no
Ge i s
or
gia
Ut
a
Ind h
ian
Ca
lifo a
rn
ia
Te
xa
Ne s
v
Ne ada
bra
s
Te
nn ka
Pe
es
se
nn
e
sy
lva
n
ia
De
law
a
re
Mi
ss
ou
Vir r i
gin
ia

Percent Reversed

National Reversal Rate

State

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 4 and Chart 5 identify two states whose records are so different from others as to raise
questions about why: Missouri’s high court finds error only 17% of the time—9 percentage points
less often than the next lowest state (after which the distribution of states becomes more continuous).
And the Virginia Supreme Court finds error only 10% of the time—7 percentage points below
Missouri, 16 percentage points below where the distribution becomes continuous, and 31 percentage
points below the national average. All other states range from two-thirds (67%) to just over 1.5 times
the national average of 41%; by contrast, Missouri’s rate is only 40%, and Virginia’s is less than
25%, of the national average. A question for further study is whether the fact that all other state high
courts discover serious error in anywhere from 26% to 67% of their capital judgments provides a
reason to question the care with which the Missouri and Virginia high courts screen for such error,
given that they find it only 17% and 10% of the time,186 or whether capital judgments in those states
are substantially less prone to error than capital judgments everywhere else.187
B.

Rates of Serious Error Found on State Post-Conviction

Table 5 below reveals what we know about the comparative amounts and rates of serious
capital error found during state post-conviction review proceedings. As we have noted, the available
data do not permit an accurate determination of the rates of error actually found in decided cases,
because there is no accurate count of those cases. The data do, however, enable us to derive a
systematically underestimated proxy for that state post-conviction reversal rate by taking the
(incomplete) number of state post-conviction reversals we have been able to identify as a proportion
of the cases that were available for state post-conviction review (whether or not they actually

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

completed that review) during the study period.188 Table 5 presents that (under)estimated rate of error
found on state post-conviction in each state.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 5: Known State Post-Conviction Reversals, 1973-1995, By State
State

Known Number of State PostConviction Reversals

National Composite
1. Maryland
2. Wyoming
3. Indiana
4. Utah
5. Mississippi
6. South Carolina
7. Florida
7. Tennessee
9. Nebraska
10. Georgia
11. Arizona
11. North Carolina
13. Alabama
13. Montana
15. Nevada
16. Illinois
16. Louisiana
18. Texas
19. Idaho
20. Arkansas
20. California
20. Missouri
23. Virginia
24. Oklahoma
25. Pennsylvania
26. Kentucky
Delaware
Washington

248
14
1
13
3
11
10
64
13
2
24
12
9
11
1
5
10
4
22
1
2
7
3
3
2
1
0
unknown
unknown

Reversals as % of Cases
Available for State PostConviction Review*
10
52
33
25
23
20
18
17
17
13
12
10
10
9
9
8
7
7
6
5
4
4
4
3
2
1
0
unknown
unknown

*This column does not report the proportion of capital judgments actually reviewed on state post-conviction that were
reversed due to serious error, because that information is not available. It instead reports the reversals known to have
occurred (despite the difficulty of collecting data) as a percentage of all of the capital judgments that were available to

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

be reviewed (almost all of which eventually complete state post-conviction review, but many of which had not completed
that review (i.e., they instead were awaiting final review) at the end of the study period. This table thus undercounts the
actual number and rate of reversals on state-post conviction . See infra Appendix C, pp. C-1 to C-2.
Source: Appendix C; DRCen; DADB

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 5 shows the following:
•

State post-conviction review is an important source of review in some states, including
Florida, Georgia, Indiana, Maryland, Mississippi, North Carolina, and Tennessee. In
Maryland, at least 52% of capital judgments reviewed on state post-conviction during the
study period were overturned due to serious error; the same was true of at least 25% of the
capital judgments that were similarly reviewed in Indiana, and at least 20% of those
reviewed in Mississippi.

•

Table 5 is especially revealing when the post-conviction reversal-rate rankings it assigns to
particular states are compared to their direct appeal reversal-rate rankings in Table 4 (p. 47).
That comparison identifies a number of states in which high error rates are found at both
state court review stages. Of particular interest are three southeastern states—South Carolina,
North Carolina and Maryland, all of which fall within the jurisdiction of the United States
Court of Appeals for the Fourth Circuit (“Fourth Circuit Court”), based in Richmond,
Virginia. All three of those states (and, most especially Maryland and South Carolina) rank
fairly high on both state direct appeal and state post-conviction reversal rates. In this regard,
they contrast sharply with the one remaining state within the jurisdiction of the federal Fourth
Circuit Court—Virginia—which falls in the very bottom cohort of states in regard to error
detection at both state review stages.

•

Other states in which relatively high rates of error manifest themselves at both the state direct
appeal and state post-conviction stage are Wyoming and Mississippi (both falling within the

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

top fifth of states in terms of capital error rates found at both state court inspection stages)
and Florida, which ranks seventh and eighth on the two error rates.
•

Falling in the bottom rank insofar as error detection by both sets of state courts is concerned,
in addition to Virginia, are California, Missouri, and Pennsylvania.

•

By contrast, in some states, close scrutiny at one state-review stage seems to compensate for
less exacting scrutiny at another. In Indiana and Tennessee, for example, relatively low errordetection rates on direct appeal (the states are ranked 19th and 23rd, respectively, in terms of
their reversal rates at that stage) are partly offset by high error-detection rates on state postconviction (where the states are ranked 3rd and 7th, respectively). Georgia, Nebraska and Utah
also fit this pattern.

•

The inverse pattern—high direct appeal, but low state post-conviction, error-detection rates—
characterizes states such as Kentucky and Oklahoma.
C.

Rates of Serious Error Found on State Direct Appeal and State Post-Conviction

Table 6 and Figure 6 below display the combined rates of error detected in the two statecourt inspection phases.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 6: State-by-State Comparisons of Rates of Error Detected
by All State Courts (State Direct Appeal and State Post-Conviction)
State

Percent Reversed in State Courts, Overall*
47%
78%
77%
69%
65%
62%
59%
58%
50%
50%
50%
50%
49%
48%
47%
44%
43%
43%
43%
41%
38%
35%
35%
33%
29%
20%
13%
unknown
unknown

National Composite
1. Wyoming
2. Maryland
3. Mississippi
4. North Carolina
5. South Carolina
6. Alabama
7. Florida
8. Kentucky
8. Louisiana
8. Utah
8. Oklahoma
12. Indiana
13. Arizona
13. Montana
15. Idaho
16. Arkansas
16. Georgia
16. Illinois
19. Tennessee
20. Nebraska
21. Nevada
21. Texas
23. California
24. Pennsylvania
25. Missouri
26. Virginia
Delaware
Washington

*This column does not report the proportion of capital judgments actually reviewed in state court that were reversed due
to serious error, because the post-conviction information needed to make that calculation is not known. Instead, it reports
the reversals known to have occurred (despite the difficulty of collecting state post-conviction data) as a percentage of
all of the capital judgments that were available to be reviewed on state direct appeal or state post-conviction (almost all

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

of which were eventually reviewed on state post-conviction but many of which were not finally reviewed (i.e., they were
as yet undecided and awaiting final review) at that stage at the end of the study period. The actual state court reversal
rate thus is higher in most or all instances. See infra Appendix C, pp. C-1 to C-2.
Source: DADP; Appendix C; DRCen

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 6. Percent of Death Sentences Reversed on
State Direct Appeal or State Post-Conviction, 1973-95
90

Percent Reversed
National Reversal Rate

W
yo
m
M in
ar g
M yl
N iss and
or i s
t
s
So h C i p p
a
ut
r i
h oli
C na
ar
o
A lina
la
ba
m
Fl a
or
K
en ida
t
Lo uck
ui y
si
an
a
O Ut
kl ah
ah
om
In a
di
a
A na
riz
M ona
on
ta
na
Id
a
A
rk ho
an
s
G as
eo
rg
I l l ia
Te i n
nn o i s
e
N sse
eb e
ra
s
N ka
ev
ad
Te a
C
x
Pe alif as
nn or
sy nia
lv
a
M nia
is
so
Vi u r i
rg
in
ia

Percent Reversed

State

National Average

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 6 and Figure 6 reveal the extent of serious error detected by state courts as a whole.
The results are remarkable:
•

Even before any federal courts become involved, state courts across the country find
serious error in close to half (at least 47%) of the capital judgments that reach their two
checkpoints.

•

State courts found capital error rates of 40% or more in five-sixths of the death-penalty
states. They found serious error in 60% or more of the capital judgments in a fifth of
those states.

•

A number of the states in the nation’s “death belt” (where most American death sentences
are imposed and the largest death rows exist) have some of the nation’s highest rates of
serious capital-sentencing error—by the lights of the states’ own elected judges: Florida at
58%; Alabama at 59%; South Carolina at 62%; North Carolina at 65%; Mississippi
at 69%; and Maryland at 77%.

•

As in other analyses, Virginia is a distinct anomaly. Its courts’ capital error-detection rate
during the study period was less than a third the national average, and 35% below the next
nearest state, Missouri—which itself has an error-detection rate 31% below the next lowest
state, after which the differences among states are small.
D.

Rates of Serious Error Found on Federal Habeas Review

Table 7 and Figure 7 below compare the rates of error detected on federal habeas corpus
review of death sentences in the 28 capital-sentencing jurisdictions with at least one completed

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

federal habeas proceeding during the study period. As discussed above, virtually all capital
judgments reviewed on federal habeas had previously been given two state court inspections: one
on state direct appeal (at which 41% of the judgments reviewed were thrown out) and a second on
state post-conviction (after which, the state courts together had thrown out 47% of the capital
judgments they reviewed).

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 7: Percent of Capital Judgments Reviewed on Federal Habeas Corpus
in Which Reversible Error Was Found, 1973-1995
State
National Composite
Kentucky*
Maryland*
Tennessee*
California
Montana
Mississippi
Idaho*
Georgia
Arizona
Indiana
Nevada
Oklahoma
Wyoming*
Arkansas
Alabama
Nebraska
Illinois
Pennsylvania
Florida
Utah*
Washington*
Louisiana
Texas
North Carolina
Missouri
South Carolina
Virginia
Delaware*

Percent Reversed on Habeas Corpus
40
100
100
100
80
75
71
67
65
60
50
50
50
50
48
45
43
40
40
37
33
33
27
26
18
15
14
6
0

* States with three or fewer completed federal habeas cases during the study period.
Source: HCDB

M
ar
y
K lan
en d
Te tuc
nn ky
e
C sse
al
ifo e
r
M nia
on
M
i s tan
si
ss a
ip
p
Id i
ah
G
eo o
rg
A ia
riz
W on
yo a
O min
kl
ah g
om
In a
di
a
N na
ev
A ad
rk
a
an
s
A
la as
b
N am
eb a
ra
sk
Pe Il a
nn lino
sy is
lv
an
ia
F
W lo
as rid
hi
a
ng
to
n
Lo Uta
ui h
si
an
N
a
or
T
th e x
a
C
ar s
ol
S o M ina
ut i s s
h
o
C ur
ar i
ol
i
Vi n a
rg
D ini
el
a
aw
ar
e

Percent Reversed

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 7. Percent of Death Sentences Reversed on Federal
Habeas Review, 1973-95

100

90
Percent Reversed on Habeas

80
Average Reversal Rate

State

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Although Table and Figure 7 include all 28 states, our narrative analysis here puts aside the
states (indicated with an asterisk) with three or fewer federal habeas cases during the study period.
Among the 20 capital-sentencing states that had a substantial number of their capital judgments
reviewed on federal habeas during the study period:
•

Federal courts found serious error189 in 40% of the capital judgments they reviewed at
this third inspection point.

•

In two-fifths of the study states, federal courts detected error rates of 50% or more at
this third inspection.

•

Virginia is again an anomaly in this analysis. The 6% error-detection rate among Virginia
capital habeas cases is well under half that of the next lowest state (South Carolina at
14%), and is exactly 15% of the national average.190
Table 7 and Figure 8 below reveal an important fact about federal habeas review, which

undermines two frequent, but contradictory, criticisms of federal judges. According to one criticism,
unelected federal judges tend to oppose the death penalty, prompting them to overturn capital
judgments whenever they can.191 According to the opposed view, federal judges—especially since
appointees of Presidents Reagan and Bush became a majority in the mid-1980s—are ideologically
“conservative” and prone to uphold state-imposed death sentences at every turn.192 Our data suggest
that federal judges are more discerning and sensitive to context than either view claims. Thus, the
same judges on the same federal circuit court often find very different rates of reversible error in
capital judgments they review depending on the state of origin of the judgments in question. This

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

suggests that factors specific to each states’ capital judgments have more of an effect of federal
judges’ behavior in capital habeas cases than the judges’ ideological dispositions. Table 8 and Figure
8 below compare the rates of error that 4 federal circuit courts found in capital judgments imposed
by states subject to their jurisdiction during the study period.

Table 8: Error Rates, by Selected States, Found by Federal
Circuit Courts on Habeas Review, 1973-1995
Circuit
State

% Capital Judgments
Reversed on Habeas

Fourth Circuit
North Carolina
South Carolina
Virginia

18
15
6

Fifth Circuit
Mississippi
Louisiana
Texas

71
27
26

Eighth Circuit
Arkansas
Nebraska
Missouri

48
43
15

Eleventh Circuit
Georgia
Alabama
Florida

65
45
37

Source: HCDB

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 8. PercentofC apitalSentences
R eversed on FederalH abeas by State and C ircuit,1973-95
80
70

Fourth
Circuit

Eleventh
Circuit

50
40
30
20
10

State and Circuit

Fl
or
id
a

Al
ab
am
a

G
eo
rg
ia

i
M
is
so
ur

sk
a
N
eb
ra

Ar
ka
ns
as

Te
xa
s

Lo
ui
si
an
a

M
is
si
ss
ip
pi

Vi
rg
in
ia

C
ar
ol
in
a

So
ut
h

C
ar
ol

in
a

N
or
th

Percent Reversed

Eighth
Circuit

Fifth
Circuit

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Rates of Serious Error Found by State Versus Federal Courts

Figures 9 and 10 below compare the rates at which state courts in each of the study
jurisdictions found serious error in that state’s capital judgments to the corresponding rates for
federal courts. Figure 9 compares the rate of serious capital error that was found for each state by its
courts on direct appeal to that found by federal courts on habeas review.193 Figure 10 makes a
similar comparison of the rate of serious capital error found for each state by its courts on both state
direct appeal and state post-conviction review to the corresponding serious-error rate found by
federal courts on habeas.194
Figures 9 and 10 arrange the states by the extent of the difference between the rates of serious
capital error found on state versus federal review. On the left side of each chart are states as to
which state courts found more serious capital error than federal courts. On the right side are states
as to which federal courts found more serious error than state courts. In between are states as to
which state and federal courts found similar rates of capital error.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 9. Percent of Capital Judgments Reversed on State
Direct Appeal and Federal Habeas, 1973-95
100
% Reversed on Federal Habeas
90

% Reversed on State Direct Appeal

Percent Reversed

70
60
50
40
30
20
10

t
So h C
u t aro
h
C lina
ar
o
D lin
el
a a
Lo w a
ui r e
s
W ian
yo a
m
in
F
W lo g
as ri
hi da
ng
A to
la n
ba
m
Te a
x
Vi a s
rg
M ini
is a
so
ur
i
U
ta
Il h
O lino
kl
ah i s
A om
rk
a
M an
is sa
Pe si
s
nn ssi
sy pp
lv i
N an
eb ia
ra
s
A ka
riz
o
In na
di
a
N na
ev
ad
Id a
a
G ho
eo
M rgi
on a
M tan
ar a
y
C lan
al
if d
K orn
en ia
Te tuc
nn ky
es
se
e

State

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 10. Percent of Capital Judgments Reversed on State
Direct Appeal or State Post-Conviction and on Federal
Habeas, 1973-95
100
% Reversed on Federal Habeas

% Reversed on State Direct Appeal or State Post-Conviction

Percent Reversed

70
60
50
40
30
20
10

om
a
Ar
ka
ns
as
Mi
ss
iss
Pe
ipp
nn
i
sy
lva
nia
Ne
bra
sk
a
Ar
izo
na
Ind
ian
a
Ne
va
da
Ida
ho
Ge
or
gia
Mo
nta
na
Ma
ryl
an
d
Ca
lifo
rn
ia
Ke
ntu
ck
Te
y
nn
es
se
e

lah
Ok

Illi

s
xa

gin
Vir

a
ian
a
Wy
om
ing
Flo
rid
a
Ala
ba
ma

lin

uis

ro
Ca

uth
So

rth

lin

State

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Especially when we put to one side the (asterisked) states that had too few capital habeas
reviews to permit analysis, both charts reveal a strong degree of similarity between the rates of
capital-sentencing error detected by state and federal courts in each state (i.e., in how close together
the two lines are). More important is the even stronger degree of similarity between state and
federal courts’ judgments about the various state’s comparative rates of capital-sentencing
error (i.e., in how closely each line’s upward and downward ticks as it moves from one state to the
next are paralleled by the other line’s upward and downward ticks). What Figures 9 and 10195 thus
suggest is that state and federal courts examining the same pools of capital judgments generally
find—and react similarly to—the same relative levels of serious capital-sentencing error. In plain
English: Where state courts find comparatively high, low or average rates of error in a
particular jurisdiction’s capital judgments relative to error rates found elsewhere, so do the
federal courts reviewing the same jurisdiction’s capital cases. Figures 9 and 10 thus refute the
notion that elected state judges as a group react differently to the possibility of error in capital cases
from the way that federal judges react as a group. In fact state and federal judges’ reactions to
capital error on both these measures of comparative amounts of error196 are very much in sync.
That said, it is interesting to consider the relatively small numbers of states that fall on the
left and the right edges of the chart where the state and federal error-detection lines diverge. In doing
so, we focus on Figure 10 (the more informative of the two charts197) and on the (non-asterisked)
states with sufficient numbers of federal habeas cases.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

One interpretation of Figure 10 is that the courts of North Carolina, South Carolina,
Louisiana, Florida and Alabama—the states on the left side of the chart—are doing the lion’s share
of error detection for capital judgments in those states, leaving significantly less error to be detected
by the relevant federal courts. Alternatively, the courts of those five states may have increased their
level of vigilance to compensate for what they perceive (based, e.g., on past experience and (more
probably) on information transmitted by lawyers) to be unusually lax error-detection by the federal
courts. This latter interpretation might explain the North and South Carolina courts’ robust errordetection in capital cases. Both states fall within the United States Court of Appeals for the Fourth
Circuit, which has by far the lowest capital-error detection rate of any federal judicial circuit in the
country.198
The corresponding interpretation for Georgia, Montana and California—on the right side of
the chart—is that federal courts have taken the lead error-detection role as to capital judgments from
those states to compensate for low state court error-detection.
The hypotheses offered in the preceding two paragraphs present important questions for
future research.
We conclude our discussion of Figure 10 by again noting a discrepancy between Virginia and
the other states. Unlike almost every other state (Missouri, again, and Texas are in an intermediate
category) Virginia’s state-review “square” and its federal-review “circle” are both located at the very
bottom of the chart. In this respect, the Virginia courts may be contrasted to those of the other states
in the Fourth Circuit, which are discussed on pp. 51 and 65 above: unlike the courts of the

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

neighboring states, there is no evidence that Virginia’s courts have tried to compensate for very low
error detection by the Fourth Circuit. Quite the contrary, Virginia courts have the lowest errordetection rates of the 28 study states. As a consequence of simultaneously low state and federal error
detection, the rate of error detected in Virginia capital judgments is both extremely, and unusually,
low.
F.

Overall Rates of Serious Error Found on State Direct Appeal, State PostConviction, and Federal Habeas Corpus

Tables 9 and 10, and corresponding Figures 11 and 12, compare the various study states
based on their overall rates of serious capital-sentencing error (i.e., the rates of serious error
found during full state and federal court review199). Table 9 and Figure 11 consider only the first
(state direct appeal) and third (federal habeas) review stages.200 A more comprehensive picture is
provided by Table 10 and Figure 12, which include, in addition, what we know about the second,
state post-conviction stage. For that reason, we display and discuss Table 10 and Figure 12 here.
Table 9 and Figure 11 are in Appendix E (pp. E-5 and E-6).

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 10: Overall Error Rates, by State, 1973-1995
Including State Post-Conviction
State
1. Kentucky*
1. Maryland*
1. Tennessee*
4. Mississippi
5. Wyoming*
6. California
6. Montana
8. Idaho
9. Georgia
10. Arizona
11. Alabama
12. Indiana
12. Oklahoma
14. Florida
15. North Carolina
16. Arkansas
17. Nevada
18. South Carolina
18. Utah*
20. Illinois
21. Nebraska
22. Louisiana
23. Pennsylvania
24. Texas
25. Missouri
26. Virginia
Delaware*
Washington*

Overall Error Rate, Including State Post-Conviction*
100%
100%
100%
91%
89%
87%
87%
82%
80%
79%
77%
75%
75%
73%
71%
70%
68%
67%
67%
66%
65%
64%
57%
52%
32%
18%
unknown
unknown

* States with three or fewer federal habeas cases.
Sources: DADB; Appendix C; DRCen; HCDB

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 12. Combined Error Rate on State Direct Appeal, State
Post-Conviction and Federal Habeas, 1973-95
100

100
Error Rate by State

K
en
tu
M cky
ar
*
Te yla
nn nd
*
e
M ss
ee
is
si
s *
W sip
yo pi
m
C ing
al
ifo *
r
M nia
on
ta
na
Id
a
G ho
eo
rg
A ia
riz
on
A
la a
ba
m
In a
di
O
a
k l na
ah
om
N
or Flo a
th
r
C ida
ar
o
A lin
rk
a
an
sa
So
s
N
ut ev
h
ad
C
ar a
ol
in
a
U
ta
h
Ill *
in
o
N
eb is
ra
Lo sk
Pe ui a
nn sia
sy na
lv
an
ia
Te
M xas
is
so
Vi u r i
rg
in
ia

Percent Reversed

Average Reversal Rate

State

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 10 and Figure 12 reveal that:
•

For the study states as a whole, the overall rate of serious error was 68%.201

•

Overall error rates vary enormously, from 18% in Virginia to 91% in Mississippi.202

•

All but two states (Virginia and Missouri) had overall error rates of 52% or higher.
All but four states (those two, plus Texas and Pennsylvania) had overall error rates of
64% or higher.

•

Put the other way around, only two states out of 26 produced capital judgments that
passed inspection for serious error more than half the time.

•

Numerous states, in all sections of the country—including Alabama, Arizona,
California, Georgia, Indiana, Mississippi, Montana and Oklahoma—had error rates
of three-quarters or more, with Mississippi’s being more than 9 out of 10 (a success rate
of less than 1 in 10).

•

As noted above, the Governor of Illinois cited evidence of high rates of serious error in
Illinois capital judgments, and particularly a spate of exonerations of innocent men released
from death row, as the reason for declaring a moratorium on executions there.203 This
prompted other policymakers, including the Governors of Florida and Texas, to suggest that
actions in Illinois are not relevant elsewhere, because high error rates are unique to Illinois.204
In fact, the rate of error detected by state and federal courts in Illinois capital sentences,
66%—while high in absolute terms—is not at all unique. On the contrary, the Illinois
error rate is very close to, and a bit lower than, the national average of 68%.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

•

As one would expect from our previous discussion, and as Figure 12 demonstrates, Virginia
is a distinct outlier here, falling almost literally “off the charts” on the low side of error
detection. Virginia’s overall rate of detected error is barely half that of the next closest state
(Missouri, which itself is much lower than all the other states), and barely a quarter the
national rate. In technical terms, Virginia’s overall-error detection rate is nearly 3 standard
deviations below the mean (2.88).
Figure 13 below plots (1) the combined state direct appeal and state post-conviction reversal

rates, (2) the federal habeas reversal rate, and (3) the overall error rate that is a composite of the other
two.205 Figure 13 illustrates three points that (for the most part) we have discussed above:
•

High overall error rates across most states.

•

Similar state and federal patterns of error detection in most states, with some exceptions
where high state court error detection compensates for low federal court error detection (e.g.,
North and South Carolina state courts compensating for the federal Fourth Circuit), and vice
versa (e.g, the federal Ninth and Eleventh Circuits compensating for low state court error
detection in California and Georgia, respectively).

•

Virginia’s outlier status.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 13. Overall Error Rate (Including State PostConviction) and Reversal Rate by Type of Review, 1973-95
100
Combined DA/PC and Habeas

% Reversed on State DA + PC
% Reversed on Federal Habeas

Percent Reversed

70
60
50
40
30
20
10

K
en
tu
M ck
ar y
Te yla *
nn nd
*
e
M ss
is
ee
si
s *
W sip
yo p
m i
C ing
al
ifo *
M rnia
on
ta
n
Id a
ah
G o*
eo
rg
A ia
riz
A ona
la
b
O am
kl
ah a
om
In a
di
an
N
or Fl a
th or
C ida
ar
o
A lin
rk
an a
s
N as
ev
So Lo ada
ut uis
i
h
C ana
ar
ol
in
a
U
ta
h
Ill *
i
n
N
o
Pe eb is
r
nn as
sy ka
lv
an
i
Te a
M xas
is
so
Vi uri
rg
in
ia

State

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Length of Time of Review

The multiple inspections needed to detect all this error take time—a 23-year average of about
9 years if the outcome is execution (with that figure rising to 10.6 years in the latter third of the study
period), and 7.6 years if the outcome is reversal on habeas corpus.206 Figures 14-16 below provide
a variety of perspectives on the length of time required to cleanse capital judgments of chronically
high rates of error.
Figure 14207 below compares states on the basis of how many years elapsed between each
state’s first death sentence and its first non-consensual208execution (not necessarily in the same case):
•

In 16 (57%) of the 28 study states, it took (or will take209) 15 or more years to get from
the state’s first death sentence to its first execution following full review.

•

In 71% of the states it took 10 or more years.
Figure 15210 compares the 23 study states in which at least one execution (consensual or non-

consensual) took place between 1973 and 1995 based on the amount of time that elapsed, on average,
between the same prisoner’s death sentence and execution. Subject to missing data, and the fact that
the table counts consensual executions, which causes it to understate the time needed for full
review,211 Figure 15 reveals that:
•

In the vast majority of states, executions took place on average 7 or more years after
death sentences during the study period.

•

In over two-thirds of the states, executions took place an average of 9 or more years
after the death sentence was imposed.

nn
es
se
e
Id

(2
0)

*
o
N
ev Mo **
ad nta
a
n
(1 a
99
6
Pe
n In )*
K ns dia
en y
n
tu lva a
ck ni
a
y
(1 **
99
N
eb 7)*
ra
M
sk
ar
A
a
yl
an rizo
d
n
(1 a
99
W
yo 7)*
m
in
g
W Illi
as no
hi is
n
A gto
rk
n
a
C nsa
al
ifo s
D rni
el
a a
O wa
kl
ah r e
om
a
U
t
G ah
eo
M rgia
i
M sso
is
si uri
ss
So A i p p
ut lab i
a
h
C ma
ar
ol
in
N
a
or
T
th e x
a
C
ar s
ol
i
Vi na
rg
in
Fl ia
o
L o rid
ui a
si
an
a

Years

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 14. Years from First Death Sentence
to First Nonconsensual Execution, 1973-2000

State

ta
na
A
riz
N ona
eb
ra
sk
O
kl
ah a
om
a
Pe Ill
i
n
nn
o
sy is
lv
a
W nia
yo
m
in
A
g
rk
an
sa
M
is s
so
ur
G
eo i
rg
N Ala ia
or
ba
th
m
C
So
ar a
ut
ol
h
i
C na
ar
ol
in
Fl a
or
id
a
U
ta
h
Te
x
C
al a s
ifo
rn
Vi ia
W rgi
n
as
hi ia
ng
D ton
el
aw
M
a
is
si re
ss
Lo i p p
i
ui
si
an
In a
di
a
N na
ev
ad
N
at Mar a*
io
na yla
l A nd
ve *
ra
ge

Years

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 15. Average Years from Death Sentence
to Execution, 1973-95

State

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 16212 below compares states based on the proportion of their 1973-1995 death
sentences that were awaiting direct review in 1995. As is discussed above, this comparison provides
a rough measure of the extent to which state direct appeal is a bottleneck in the inspection process.213
Nationally, 21% of capital sentences imposed between 1973 and 1995—about 5-years-worth of death
sentences—were awaiting direct appeal in 1995:
•

In over a third of the 28 states, 20% or more of all post-Furman death sentences were backed
up at the state direct appeal stage 23 years after Furman.

•

In three of the nations most prolific capital-sentencing states, Texas, Pennsylvania and
California, the 1995 log-jam of cases awaiting state direct appeal contained (respectively)
27%, 27% and 47% of the state’s post-1972 cases. In Washington and Wyoming, the 1995
logjam contained 45 and 70% of the post-Furman cases.

•

Of note, although the federal Ninth Circuit Court of Appeals is sometimes blamed for
holding up executions in the states within its jurisdiction, the three states in that circuit with
the largest death rows—California, Arizona and Washington—were all in the top cohort of
states as of 1995 in terms of the proportion of cases bottled up in the state courts awaiting
direct appeal.

W
yo
m
C
al ing
W ifo
as rn
hi ia
ng
to
n
Pe
nn Tex
sy as
lv
a
M nia
is
Te so
nn uri
es
se
N
e
or
th U
C tah
ar
ol
i
A na
riz
K on
en a
M tu
is ck
si
ss y
ip
I n pi
di
an
Ill a
in
So Lou ois
ut i s
i
h
C ana
ar
o
A lina
la
b
O am
kl
ah a
N om
eb a
ra
sk
Fl a
or
id
a
Id
ah
o
N
ev
ad
G
eo a
A rgi
rk
an a
s
Vi as
rg
M inia
on
M tan
ar a
y
N
at
D lan
e
d
io
na law
l A ar
ve e
ra
ge

Percent Waiting for Review

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 16. Percentage of 1973-95 Death Sentences
Awaiting Direct Review as of 1995

State

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Capital-Sentencing and Execution Rates, and the Two Compared

This section compares states to each other based on (1) how many death sentences they
impose “per capita” and (2) how many executions the carry out “per capita.” We use three different
per capita measures—sentences and executions per 1,000 homicides, per 100,000 population, and
per 1,000 prison population.214 The middle measure is particularly interesting, given the expectation
that the number of death sentences each jurisdiction imposes and carries out would be responsive
to the number of homicides committed there. This section also asks whether, as one would expect,
states that undertake to capitally sentence more offenders per capita than other states also execute
more people per capita.
Figure 17 below compares states based on their death sentencing rates per 1,000 homicides,
per 100,000 population and per 1,000 prisoners. Figure 18 below compares states based on their nonconsensual execution rates per the same three populations.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 17. Per Capita Death Sentencing Rates
by State, 1973-95

Death Sentences per 1,000 Homicides
Death Sentences per 100,000 Population

Death Sentences per 1,000 Prison Population

n
ui
s
Ke i a n a
nt
uc
k
Illi y
no
is
Vi
rg
in
I n ia
di
Mi a n a
s
T e sou
nn
r
es i
se
e
Pe
T
n n exa
sy
s
lva
n
ia
So Ark
an
ut
s
h
No Ca as
rth r o l
Ca i n a
ro
l
Ge ina
o
Ne r g i a
br
as
ka
Ut
Mi
a
ss
iss h
i
M o ppi
nt
Al ana
ab
am
a
Fl
o
D e rida
la
Ok w a r
lah e
om
a
Ar
izo
n
Ne a
va
da
Id
a
W
h
yo o
m
in
g

or
W

lif

lan

Sentencing Rate

State

Ma
ryl
Ke and
n
Te tuck
y
n
Pe ness
nn
syl ee
va
n
Ne ia
va
da
Ida
ho
Ca
lifo
rn
Ind ia
Wa
ia
sh na
ing
ton
No
Illin
rth
o
Ca is
rol
in
Ari a
z
on
M
So issis a
sip
uth
Ca pi
r
Ok olina
lah
o
Ne ma
bra
s
Ala ka
ba
m
Ge a
org
i
Flo a
ri
Mi da
sso
Mo uri
nt
Lo ana
uis
ian
a
Uta
Wy
h
om
ing
Te
Ark xas
an
sa
Vir s
g
De inia
law
are

Execution Rate
3.0

2.5

Figure 18. Per Capita Non-Consensual Executions
by State, 1973-95
Executions per 1,000 Homicides

Executions per 100,000 Population

Executions per 1,000 Prison Population

2.0

1.5

1.0

0.5

0.0

State

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figures 17 and 18 reveal huge variations among states in both their death-sentencing rates
and their execution rates measured per homicides and per population:215
•

Measured against both populations, some death-sentencing states have death-sentencing
rates that are 10 times those in other death sentencing states.

•

In Wyoming, for example, nearly 6% of all homicides result in a death sentence—over four
times the national average for death-sentencing states. In Maryland, less that six-tenths of 1%
of homicides lead to a death sentence.

•

Nevada condemns nearly 11 people out of every 100,000—about three times the national
average for death-sentencing states. Washington State does so to less than 1 person out of
every 100,000.

•

Similar disparities characterize the execution rates in the various death-sentencing
jurisdictions.

•

The disparities among states in death sentences and executions per 1,000 homicides are
particularly interesting, revealing the absence of what one would expect to be a consistent
relationship between homicides and capital punishment.
Figures 19-21 below consider whether high (or low) death-sentencing rates (per homicide,

per population or per prisoners) translate, as one would expect, into high (or low) execution rates.

ar
y
C a land
W lifor
a s nia
hi
n
Lo g t o n
ui
si
Ke ana
nt
uc
ky
Illi
no
V i is
rg
in
In ia
di
a
M na
iss
Te
o
n n uri
es
se
e
Pe
Te
nn
xa
sy
s
lva
n
S o A r k ia
ut an
s
h
No Ca as
rth rol
Ca ina
ro
li
G na
eo
N e rgia
br
as
ka
M Uta
iss
h
iss
i
p
M
p
on i
ta
A l na
ab
am
Fl a
o
De rida
la
O war
kla e
ho
m
Ar a
izo
N e na
va
da
Id
a
h
W
yo o
m
in
g

Death Sentences per 1,000 Homicides
70

50
2.0

40
1.5

20
1.0

10
0.5

0
0.0

State

91
Executions per 1,000 Homicides

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 19. Death Sentences and Executions
per 1,000 Homicides by State, 1973-95
3.0

Death Sentences per 1,000 Homicides

Executions per 1,000 Homicides
2.5

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 20. Death Sentences and Executions
per 100,000 Population by State, 1973-95
12

0.7

0.6

Executions per 100,000 Population

0.5
8
0.4
6
0.3
4
0.2

0.1

0.0

i a
e s o
s
a
a y
a
a is a
a
e
a g
a ri
a h
a
a
a
a
a
d
n
to lan ask Uta ian uck gini rni ino ani tan ou sse ian war nsa ah olin olin exa rgi in sipp om rid am zon vad
g
d
m
s
o
o
o
i
s
l
d
y
v
r
b
t
h
I
n
r
r
n
T e yo is la Fl la
e
In en Vir alif Il syl Mo Mis nne oui ela rka
Ar N
hi ar eb
G W iss k
Ca Ca
A
D A
K
e L
n
C
as M N
O
h rth
T
n
t
M
W
u o
Pe
So N

State

Executions per 100,000 Population

Death Sentences per 100,000 Population

M
W ary
as la
hi nd
ng
t
Vi on
rg
C
al inia
ifo
rn
In ia
d
D ian
el
a a
Lo w a
ui r e
s
So Ke ian
ut ntu a
h
C cky
ar
N olin
eb a
ra
M sk
is a
so
u
Ill ri
in
oi
s
U
ta
N
or
h
th T e
x
C
a as
T e rol
n n ina
e
A ss
Pe rk ee
nn ans
sy as
lv
a
G nia
eo
M rgi
o a
O nta
kl na
ah
o
A ma
r
M iz
is on
si
ss a
ip
Fl p i
or
id
Id a
A ah
la o
ba
m
N
ev a
W ad
yo a
m
in
g

Death Sentences per 1,000 Prisoners
40.0

35.0

30.0

25.0
1.5

20.0

15.0
1.0

10.0
0.5

5.0

0.0
0.0

State

93
Executions per 1,000 Prisoners

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 21. Death Sentences and Executions
per 1,000 Prisoners, 1973-95
2.5

Death Sentences per 1,000 Prison Population

Executions per 1,000 Prison Population

2.0

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Judging from figures 19-21, there is no relationship between death-sentencing and
execution rates. When states are arranged in order of their death sentences per capita, the line
representing their executions per capita fluctuates wildly and randomly:
•

Idaho, Nevada, Arizona and Oklahoma rank 2nd, 3rd, 4th, and 6th (and range from 3 to 4 times
the national average) when it comes to how often homicides result in death sentences. Those
same states, however, are tied for 23rd, tied for 24th, 17th, and 14th among 28 states (near
or well below the national average) when it comes to how often homicides result in
execution.

•

On the other hand, Texas, Virginia and Louisiana rank 18th, 22nd, and 25th in death sentences
per homicide (ranging from slightly above, down to two-thirds, the national average) but 4th,
2d, and 7th in executions per homicide (ranging from over twice to nearly four times the
national average).216 Thus, the three states most associated in the public’s mind with
executions— Louisiana, which was the nation’s execution capital in the late 1980s,217 and
Texas and Virginia which claimed that distinction in the 1990s218—did not attain that status
by sentencing disproportionately large numbers of people to death row. Instead, they have
done so by translating below-average death-sentencing rates into above-average execution
rates.
Figure 22 below asks a related question: Are states that are most likely to punish homicides

with death also most likely to translate death sentences into executions?

M
ar
y
C lan
al
W ifo d
as rn
hi ia
n
Lo gto
ui n
s
K ian
en a
tu
ck
Ill y
in
Vi o i s
rg
in
In ia
di
M ana
i
Te s s o
nn u r
es i
se
Pe
e
nn Tex
sy as
lv
So Ar ani
ut kan a
h
s
N
or Car as
th o l
C ina
ar
ol
G ina
eo
N rg
eb ia
ra
sk
a
M
U
is
t
s i ah
ss
M ipp
on i
t
A an
la a
ba
m
Fl a
or
D
e l ida
a
O wa
kl
ah r e
o
A ma
riz
o
N na
ev
ad
a
Id
W ah
yo o
m
in
g

Death Sentences per 1,000 Homicides
60

50
Percent Death Sentences Carried Out

40
20

30
15

20
10

10
5

0
0

State

95
Percent of Death Sentences Carried Out

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 22. Per Capita Death Sentences and
Percent Death Sentences Carried Out, 1973-95
30

Death Sentences per 1,000 Homicides
25

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 22 reveals no relationship between death sentencing and execution rates. Indeed, for
nearly half the states—Louisiana, Virginia, Missouri, and Texas (with comparatively low deathsentencing but high death-sentences-carried-out rates) and Wyoming, Idaho, Nevada, Arizona,
Oklahoma, Florida, Alabama, and Mississippi (with comparatively high death sentencing rates but
low death-sentences-carried-out rates), the relationship is the inverse: the more frequently states
sentence killers to die, the less frequently they execute them, and vice versa.
Overall, therefore, it seems clear that a powerful disposition to sentence offenders to die does
not go hand in hand with a strong capacity to carry out the death sentences that are imposed. Figuring
out why this is so is a question we will address in a subsequent report. Our analysis so far, however,
suggests one place to look for the source of the discrepancy: the distribuingly high rates of capitalsentencing error that we document above.
I.

Demographic Factors

This section considers two other possible explanations for the frequency with which states
sentence individuals to die, and the frequency with which they carry out the capital sentences they
impose. The first is violent crime—measured by each state’s homicide rate per 100,000
population.219 The second is race—based on the proportion of each state’s population that is nonwhite.220
Figures 23 and 24 below consider the relationship between homicide rates per 100,000
population and, respectively, capital-sentencing and execution rates.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 23. Death Sentencing Rates Per 1,000 Homicides and
Per Capita Homicide Rates, 1973-95
70

Death Sentences per 1,000 Homicides
Homicides per 100,000 Population

Rate

an
Te a
x
G as
M eor
is
si gia
ss
ip
N pi
ev
A ada
la
b
C am
al
ifo a
rn
So
i
ut Flo a
h
rid
C
ar a
o
M lin
ar a
Te yla
nn nd
es
se
N
or I l l e
i
th
n
C ois
ar
o
M lina
is
s
A ou
rk
r
an i
s
A as
riz
o
Vi na
rg
O
kl inia
ah
K om
en a
tu
c
Pe In ky
d
nn ia
sy na
lv
D ani
el
a a
W war
y
W om e
as
i
hi ng
ng
M ton
on
ta
na
Id
N ah
eb o
ra
sk
a
U
ta
h

State

an
Te a
G xa
M eo s
i s rg
s i ia
ss
i
N ppi
ev
A ad
la a
C bam
al
ifo a
rn
So
u t Fl ia
h or
C id
ar a
M olin
a
a
Te ryla
nn nd
es
s
N
or I l e e
th l i n
C oi
ar s
o
M lin
is a
A sou
rk
an ri
s
A as
riz
o
Vi na
O rgi
k l ni
ah a
K om
en a
tu
Pe In cky
nn di
sy an
lv a
D an
el ia
a
W wa
y
W om re
as
hi ing
ng
M to
on n
ta
n
Id a
N ah
eb o
ra
sk
a
U
ta
h

Executions per 1,000 Population
30

Homicides per 100,000 Population

20
10

15
8

10
6

5
2

0
0

State

Homicide and Execution Rates

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 24. Execution Rate and Homicide Rate, 1973-95
Nonconsensual Executions per 1,000 Homicides

Percent Death Sentences Carried Out

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

If there is any relationship at all between homicide and capital-sentencing rates (a matter
requiring more sophisticated analysis), Figure 23 suggests that it is weak and inverse. Figure 24 asks
whether variations in rates of serious crime, as measured by homicides per 100,000 population, can
explain variations in execution rates, or vice versa. Figure 24's decisive answer is that there is no
such relationship between a state’s serious crime rate and its willingness or capacity to execute its
citizens.
Turning to the issue of race, Figure 25 below compares capital-sentencing states’ relative
death-sentencing rates (per 1,000 homicide) to their percent nonwhite population.
Surprisingly, perhaps, this chart suggests that proportionately larger minority populations are
associated with somewhat lower death-sentencing rates, and vice versa. Figure 25 also reveals the
sharp variation among capital-sentencing states in terms of the proportion of their populations that
are nonwhite, ranging from 5% in Idaho (which, incidently, has a very high death sentencing rate per
homicide) to 37% in Mississippi (where the death-sentencing rate per homicide is relatively low).
Figure 26 below considers whether race influences execution, as opposed to deathsentencing, rates. Here, the relationship is weaker than in Figure 25, and runs in the opposite
direction: Although states with larger proportions of racial minorities tend to capitally sentence less
often than states with proportionately smaller minority populations, those same states tend to carry
out relatively more of the death sentences they impose.

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 25. Per Capita Death Sentencing Rate
and Percent Non-White Population
70
Death Sentences per 1,000 Homicides
60

Percent Non-White Population

a
lin
a
Ge
or
gia
Al
ab
am
a
Ma
ryl
an
d
Ca
lifo
rn
ia
No
Te
xa
rth
s
Ca
ro
lin
a
Vir
gin
ia
De
law
are
Flo
rid
a
Illi
no
is
Ar
ka
ns
as
Te
nn
es
se
e
Ar
izo
na
Ok
lah
om
a
Ne
va
da
Mi
ss
Pe
o
ur
nn
i
sy
lva
Wa
nia
sh
ing
ton
Ind
ian
a
Ke
ntu
ck
y
Wy
om
ing
Mo
nta
na
Ne
br
as
ka
Ut
ah
Ida
ho
ro

ian

uth

uis

iss

ipp

Rate / Percent

State

100

si
s
So Lo s i p
ut uis p i
h
C ian
ar a
o
G lina
eo
A rg
la ia
b
M am
ar a
C yla
al nd
ifo
rn
N
or
i
th T e a
x
C
ar a s
ol
Vi ina
r
D gin
e l ia
aw
a
Fl r e
or
i
I l l da
i
n
A
rk oi
Te an s
nn sa
es s
A see
r
O izo
kl n
ah a
o
N ma
ev
Pe Mi ada
nn sso
s
u
W ylv ri
as an
hi ia
ng
t
In on
di
K a
en na
t
W uc
yo ky
m
M in
on g
N tan
eb a
ra
sk
a
U
ta
Id h
ah
o

Percent

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 26. Percent of Death Sentences Carried Out
and Non-White Population

35
Percent Death Sentences Executed

Percent Non-White Population

State

101

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Court Factors

Here, we consider whether differences among states’ judicial systems account for the marked
variability in their capital-case error rates, death-sentencing rates, and execution rates. Relevant,
reliable, and comparable state-court contextual data are difficult to obtain. For purposes of this initial
report, we have developed three comparative measures: “political pressure” (the extent to which state
sentencing and appellate judges are subject to electoral discipline for actions they take as judges221),
judicial workloads (which we measure by comparing the various states’ criminal court caseloads per
1,000 persons during the relevant period) and judicial resources (comparing the dollars the respective
states spent on their courts per capita during the relevant period).222 The details of each of these
measures are described at pp. 44-45 above.223
Figure 27 and Figure 28 below consider the impact of political pressure on, respectively,
death-sentencing and execution (more specifically, death-sentences-carried-out) rates. Because error
rates and the rates at which death sentences are carried out are so highly correlated (see Figure 1,
supra p. 11), the latter chart is also a rough measure of the relationship between political pressure
and capital error rates.

102

A
la
ba
G ma
eo
rg
ia
Id
ah
In o
d
M ian
ar a
y
N lan
eb d
ra
s
N ka
ev
O
kl ada
ah
om
a
W
as Ut
a
hi
n h
W gto
yo n
m
A ing
ri
A zon
rk
an a
s
Fl as
o
K rid
en a
tu
c
N
or Mo ky
th n t a
C
ar n a
ol
in
Pe
a
T
nn ex
sy as
Te lva
nn nia
e
C sse
al
ifo e
rn
ia
M Illin
is
si ois
ss
M ipp
is
s i
So Lo o u
ut uis r i
h
i
C ana
ar
o
D lin
el
aw a
Vi a r e
rg
in
ia

Index of Political Pressure
9

8
Death Sentences per 1,000 Homicides

7
50

5
40

4
30

3
20

1
10

0
0

State

103
Death Sentences per 1,000 Homicides

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 27. Political Pressure and
Death Sentencing Rate, 1973-95
70

Political Pressure Index

A
la
ba
G ma
eo
rg
i
Id a
ah
In o
d
M ian
ar a
y
N lan
eb d
ra
s
N ka
ev
O
kl ad
ah a
om
a
W
as Ut
hi ah
n
W gto
yo n
m
A ing
riz
A on
rk
an a
s
Fl as
o
K rid
en a
tu
c
N
or Mo ky
th n t
C an
ar a
ol
in
Pe
a
T
nn e
sy xas
Te lva
nn nia
e
C ss
al e e
ifo
rn
i
M Illin a
is
o
si is
ss
M ipp
is
s i
So Lo o u
ut uis r i
h
i
C ana
ar
o
D lin
el
aw a
Vi a r e
rg
in
ia

Index of Political Pressure
9

5
15

3
10

2
5

0
0

State

104
Percent

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 28. Political Pressure and Percent of
Death Sentences Carried Out, 1973-95
30

Political Pressure Index

Percent of Death Sentences Carried Out

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figures 27 and 28 reveal a curious and potentially significant pattern: In general, the more
electoral pressure a state’s judges are under, the higher the state’s death-sentencing rate, but the
lower the rate at which it carries out its death sentences. Assuming a causal relationship, this
suggests that political pressure tends to impel judges—or to create an environment in which
prosecutors and jurors are impelled—to impose death sentences, but then tends to interfere with the
state’s capacity to carry out the death sentences that are imposed.
Whether it is fair to infer a causal relationship here and, if so, what might account for that
relationship is a question for further research. One hypothesis is suggested by possible relationships
between high death-sentencing rates and high error rates, and between the latter and low execution
rates: Public opinion may place a premium on obtaining death sentences.224 If so, a desire to curry
favor with voters may lead elected prosecutors and judges to cut corners in an effort to secure that
premium— simultaneously causing death-sentencing rates, and error rates, to increase. In that event,
high rates of reversible error would explain why high political-pressure states, after imposing so
disproportionately many death sentences—making so many errors in the process—end up carrying
out so disproportionately few of their death sentences. These are questions for further research.
Figures 29 and 30 below relate, respectively, states’ death-sentencing rates, and the rates at
which they carry out death sentences, to their per capita court expenditures.

105

N
ev
al ada
if
W orn
yo ia
D min
el
aw g
A are
riz
o
Fl na
or
M id
W ary a
a
la
Pe shi nd
n
nn g
sy ton
lv
M ani
on a
ta
I l l na
Lo i n o
ui i s
si
an
a
U
Vi t a h
rg
in
T e ia
A xa
la s
ba
m
Id a
K ah
en o
tu
G cky
eo
N rg
eb ia
ra
M sk
is a
T s
N en o u
or n
r
th e s i
C se
a
e
O roli
kl n
ah a
om
So
u t Ind a
h
C ian
ar a
A olin
r
M kan a
is s
si as
ss
ip
pi
C

Per Capita Spending on Courts
80

70
Death Sentences per 1,000 Homicides

State

106

60
50

50
40

30
30

20
20

Death Sentences per 1,000 Homicides

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 29. Per Capita Spending on Courts and Per Capita
Death Sentencing Rates, 1973-95
70

Per Capita Spending on Courts

N
ev
al ada
ifo
W rn
yo ia
m
D
el ing
aw
A are
riz
o
Fl na
or
M
id
W ary a
as la
Pe hi nd
nn ng
sy ton
lv
M ani
on a
ta
n
Ill a
Lo i n o
ui i s
si
an
a
U
t
Vi a h
rg
in
i
Te a
A xa
la s
ba
m
Id a
K ah
en o
tu
G cky
eo
N rg
eb ia
ra
M sk
is a
s
T
N en o u
or n
r
th es i
C se
e
a
O roli
kl n
ah a
om
So
ut Ind a
h
i
C ana
ar
A olin
rk
M an a
is sa
si
ss s
ip
pi
C

Per Capita Spending
80

70
25

50
20

40
15

30
10

20
5

0
0

State

107

Percent of Death Sentences Carried Out

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 30. Per Capita Spending on Courts and Percent of
Death Sentences Carried Out, 1973-95
30

Per Capita Spending on Courts

Percent of Death Sentences Carried Out

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

With some exceptions, Figure 29 appears to indicate that comparatively high expenditures
on courts are associated with relatively high death-sentencing rates. It is difficult to know what to
make of this relationship, especially because capital cases are themselves costly and thus may partly
account for high expenditures. It may be, however, that states whose courts have substantial amounts
of resources are more capable of handling capital cases—and thus do so more often—than states
with less well-funded courts.
As was the case when we looked at capital punishment and political pressure, the
relationship between capital punishment and spending reverses when we move from analyzing death
sentencing rates to rates of death sentences carried out: Figure 29 shows a direct relationship
between court expenditures and death sentencing (the higher the one is, the higher the other tends
to be); by contrast, Figure 30 shows a weak inverse relationship between court expenditures and
death sentences carried out—as states’ spending on their courts increases, the proportion of the death
sentences imposed that are carried out tends to decrease. The cause of that relationship (if any exists)
is unclear. If, however, it were the case that the processing of death cases is itself responsible for
significantly driving up court expenditures, then Figures 29 and 30 might suggest that spending
relatively large sums to secure relatively large numbers of death sentences has little pay off—and,
indeed, is counterproductive—when it comes to securing executions. If so, the policy alternative of
spending less by securing fewer death sentences225—each of which, however, is more likely to be
carried out—would be indicated.
Figures 31 and 32 below consider the relationship between state court caseloads and,

108

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

respectively, death sentencing rates and the rate of death sentences carried out.

109

rth
Ca

rol

Ida
ho

ina
Illin
ois
Vir
gin
ia
Lo
uis
ian
a
Ark
ans
So
uth
as
Ca
rol
ina
Mo
nta
na
Ok
lah
om
a
Ari
zon
a
Mis
sou
ri
Tex
a
s
De
law
are
Ne
bra
ska
Ke
ntu
cky
Ma
ryl
and
Ala
bam
a
Wy
om
ing
Flo
rid
a
Uta
Wa
h
shi
ng
t
o
Ten
n
nes
see
Ca
lifo
rni
a
Ge
org
ia
Ind
i
a
na
Mis
sis
sip
Pe
pi
nn
syl
van
ia
Ne
vad
a

Court Cases per 100,000 Population
90

80
Death Sentences per 1,000 Homicides

State

110

70
50

30
20

10
10

Death Sentences per 1,000 Homicides

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 31. State Court Caseloads
and Death Sentencing Rates per 1,000 Homicides
70

State Court Caseloads

Id
a
ar h o
ol
i
I l l na
in
Vi o i s
L o rgin
u i ia
So A sia
r
ut ka na
h ns
C
ar as
o
M lin
on a
O ta
kl n
ah a
o
A ma
riz
M on
is a
so
u
T ri
D exa
el
a s
N wa
eb re
K ras
en ka
t
M uck
ar y
y
A lan
la d
b
W am
yo a
m
i
Fl ng
or
id
a
W
as Ut
hi ah
Te ng
nn to
e n
C ss
al e
ifo e
G rnia
eo
rg
M Ind i a
i
i
Pe ssi ana
nn ss
sy ipp
lv i
a
N nia
ev
ad
a

Caseloads per 1,000 Population
90

State Court Caseloads

State

111

Percent of Death Sentences Carried Out

50
15

20
5

Percent of Death Sentences Carried Out

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 32. State Court Caseloads and Percent of Death
Sentences Carried Out, 1973-95
30

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Judging from Figure 31, there is no relationship between how many cases per capita state courts
handle and the rate at which those courts impose death sentences. Figure 32 does, however, suggest
a weak relationship between court caseloads and death sentences carried out: As per capita caseloads
drop, the rate of death sentences carried out also tends to drop. One might hypothesize that states
with smaller courts (ones with lower caseloads) are more likely to generate seriously flawed death
sentences at the trial level, thus depressing the rate at which their death sentences are carried out.
Alternatively, state appellate courts with lower caseloads may be superior error detectors, thus
(given high error rates across all states) accounting for lower rates of executions—or, in this
scenario, lower rates of flawed executions. Further research is called for.

IX.

Federal Circuit Court and Regional Comparisons
Appendix B contains report cards for the nine federal judicial circuits that conducted federal

habeas corpus review of state death sentences during the 1973-1995 study period.226 Those circuits
reviewed between 2 (Sixth Circuit) and 215 (Eleventh Circuit) death sentences in that period
Referring to these tables as Federal Circuit report cards is at times misleading, because much
of the information in them considers results generated by state courts or other state actors in the
states (noted at the top of on each report card) that are grouped in that circuit. For purposes of the
latter sorts of information, these are actually regional report cards, which aggregate the results of
actions by a variety of state actors in multiple states in particular segments of the nation. Only the
three items falling within the “Federal Habeas Corpus” category of each report (which we have

112

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

marked with a number sign (#)) report the results of actions exclusively by the federal courts in the
relevant circuit. An additional six rows of information (which we mark with a plus sign (+)) report
on a mixture of actions by the relevant state courts and the federal courts in the circuit.
In this section, we focus on information generated either by the federal courts alone, or by
them in conjunction with state courts.
Table 25 displays the rates of error detected on federal habeas review and overall (state and
federal review) by circuit. Figure 33 below compares the circuits’ error detection rates on habeas.

113

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 25: Error Rates Detected on Habeas Review
and Overall (State Direct Appeal and Federal Review Combined)
by Federal Circuit/Multi-State Regions
Circuit

Number Reviewed on Habeas

Number Reversed on Habeas

Error Rate Found
on Habeas

Overall Error
Rate (Region)

Sixth

100%

Ninth

62%

78%*

Eleventh

215

108

50%

77%

Tenth

47%

74%

Seventh

43%

68%

Fourth

15%

62%

Fifth

200

32%

61%

Third

29%

55%*

Eighth

33%

54%

National
Composite

599

237

40%

68%

* Does not include state post-conviction information for Washington (9th Cir.) or Delaware (3d Cir.)
Source: HCDB; DADB; Appendix C; DPCen

114

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 33. Rate of Errors Detected on Habeas
in Circuits with Over 10 Cases, 1973-95
70

Percent of Cases Reviewed

)

th
ur
Fo

fth

8)
th
gh
Ei

Se
ve
nt
h

at
N

115

(1
4)

7)
h

(2
15
)
El
ev
en
th

th
in
N

Circuit

)

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Table 25 and Figure 33 reveal that:
•

During the 23-year study period, 7 of the 9 federal death-penalty circuit courts (including
the three circuits with the most cases) found serious error in a third or more of the
death sentences they reviewed at the final (federal habeas) inspection stage—
notwithstanding that two state court inspections had already occurred.

•

Over half the circuits detected error 40% or more of the time.

•

The Eleventh Circuit—the nation’s most active capital reviewing federal court (with
jurisdiction over Alabama, Florida and Georgia capital judgments)—detected error in 50%
of the death sentences it reviewed.

•

Even after excluding the Sixth Circuit (which only reviewed two capital judgments), there
is much wider variation among the rates of serious error detected by the circuits on federal
habeas review alone (ranging from 15% to 62%) than the rates of error detected overall by
a combination of state and federal courts (which only range from 54% to 78%). This
indicates, as we have already suggested,227 that state and federal court review may somewhat
compensate for each other, tending to moderate variations that occur when the results of only
state court or federal court inspection is considered.

•

Although there is substantial variation among circuits, there also—as we already have noted
(see Table 8 and Figure 8, supra pp. 60, 61)—is substantial variation in federal habeas error
detection within circuits. The Fifth Circuit, for example, finds error in 71% of the Mississippi
death sentences it reviews but only 26% of the Texas death sentences it reviews—suggesting

116

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

that more is at issue in determining error detection rates than a federal court’s uniform
disposition with regard to error affecting capital sentences.
As did Figure 8 (p. 61) above, Table 25 and Figure 33 identify the Fourth Circuit—with
jurisdiction over Virginia, North Carolina, South Carolina and Maryland—as an outlier on the low
end of federal habeas corpus error detection. The Fourth Circuit finds error only half as often as the
next lowest circuit and just under a third as often as do the other circuits as a whole minus the
Fourth.228 Interestingly, though, as we already have noted (pp. 51, 65-66 above), state courts in three
of the four states within the Fourth Circuit—all those save the Virginia courts229—largely
compensate for the Fourth Circuit’s low error detection rate with unusually high direct appeal and
state post-conviction error detection rates of their own. Thus, although the Fourth Circuit is way
below the other circuits in error detection on habeas, when state and federal error detection are
combined, the overall rate of error detected in the Fourth Circuit region (62%) is higher than the
overall rate of error detected in three other regions (the Fifth, Third and Eighth Circuit) and not much
lower than the national average (68%). If Virginia (whose Supreme Court rarely detects error) is
excluded, the overall error rate for capital judgments from the other three states in the Fourth Circuit
region rises to 76%, significantly above the national average. The “double whammy” effect noted
earlier (p. 66) of distinctly lower error detection rates at the checkpoints operated both by the
Virginia Supreme Court and by the U.S. Court of Appeals for the Fourth Circuit thus is a unique
feature of Virginia capital judgments.230
In considering whether Virginia capital judgments are substantially less error prone than all

117

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

others in the nation or, on the other hand, whether laxer error detection takes place there, the deathsentencing states that surround Virginia and lie within its same federal judicial circuit—Maryland,
North Carolina and South Carolina—may be treated as partial “natural controls.”231 Insofar as
philosophical, cultural or historical factors—which probably do not vary much between Virginia and
its neighbors—are thought to be the main influences on the amount of expected error in capital
judgments, the fact that high capital error rates are consistently found in states bordering Virginia
casts doubt on the hypothesis that Virginia capital sentences are starkly less error-prone. For this
analysis to show convincingly that Virginia courts are laxer detectors of serious capital error than
courts in the surrounding states, there would have to be an explanation for that difference among
presumably similar states. One such explanation is the unusual extent to which the Virginia courts
limit review of capital judgments: (1) enforcing the region’s (and nation’s) strictest procedural
default doctrine (the rule permitting even egregious error to be ignored on appeal if it was not
objected to at trial); (2) often appointing substandard trial attorneys to represent the indigents who
make up 97% of the state’s death row, thus increasing the probability that necessary objections will
not be made at trial, and thus that appellate review will be cut off; (3) applying a very strict test for
reversing capital judgments based on incompetent lawyering (until the Supreme Court overturned
Virginia’s test earlier this year232); and limiting defendants’ ability to petition for a new trial based
on innocence to a 21-day period following conviction, the shortest such time-frame in the region
(and nation).233 These questions bear further study.
We close this section with a circuit comparison documenting the actions of state officials

118

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

within the states that are regionally grouped in the respective circuits. Figure 34 compares the
circuits based on their component states’ death sentencing rates (death sentences per 100,000
population) and execution rates (non-consensual executions234 per 100,000 population).235

119

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Figure 34. Per Capita Death Sentencing and
Per Capita Execution Rates by Circuit, 1973-95
0.60

Death Sentences per 100,000 Population

0.50

6
0.40
5
0.30
4
0.20
3
0.10
2

0.00

El
ev
en
th

fth
Fi

h
nt
Te

th
in
N

th
gh
Ei

Se
ve
nt
h

-0.10
ird

Circuit

120

Executions per 100,000 Population

Non-Consensual Executions per 100,000 Population

Death Sentences per 100,000 Population

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Like their state counterparts,236 the regional comparisons in Figure 34 show that relatively
high death-sentencing rates often go hand in hand with relatively low execution rates, and vice versa.
For example:
•

Alabama, Florida and Georgia (the states in the Eleventh Circuit region) impose nearly 60%
more death sentences per capita than Louisiana, Mississippi, and Texas (the states in the
Fifth Circuit region), but carry out 60% fewer executions.

•

The states in the Eleventh Circuit (Alabama, Florida and Georgia) likewise sentence nearly
three times as many people to death as Arkansas, Missouri, and Nebraska (in the Eighth
Circuit region), but the two regions’ execution rates are very similar.

As we already have suggested, the impulse to make frequent use of death sentences does not translate
into, and may even interfere in some way with, the capacity to do so reliably enough to permit death
sentences to pass judicial inspection for serious error and be carried out.

Conclusion: A Broken System; the Need for Research into Causes
Over the course of the 23-year study period, a large majority of death sentences subjected

to judicial inspection nationally and in nearly all death-sentencing states were found to be
seriously flawed and were reversed by the courts. The 60% and 70% rates of serious error that
have existed nationally and in the vast majority of states have obliged courts to provide, and have
obliged taxpayers to foot the bill for, a elaborate and lengthy judicial inspection process—one
that, even so, almost inevitably must fail to catch and correct some amount of the error that has
flooded the system. As an inevitable result of so many serious errors and the multi-tiered process
121

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

needed to catch them, it has taken nearly a decade—more recently, it has taken over a decade—
for the small number of death sentences that pass inspection to be carried out.
Very few death sentences succeed, and it takes years to cull out the majority of failures.
So far we have used the rate of serious error detected by state and federal courts as the
measure of the success or failure of our capital punishment system. But there is another important
measure that bears consideration. Presumably, the most immediate goal of a system of capital
punishment is the execution of capital sentences. In this light, the most obvious measure of the
“success” of our death penalty system—indeed, the most obvious measure of the system’s sheer
rationality—is its capacity to translate the death sentences it imposes into executions.
By this measure, the capital punishment system revealed by our 23-year study is not a
success, and is not even minimally rational. Figure 35 below plots the proportion of the death
sentences imposed at some point during the 23-year study period that had been carried out by the end
of that period—comparing the 28-state cohort of capital-sentencing jurisdictions and the national
average.237

122

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

123

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

As Figure 35 reveals:
•

Nationally, during the study period, the proportion of death sentences actually carried
out was a meager 5.4%, one in nineteen.

•

Given high error rates, and the painstaking review needed to catch it, well over half of all
American death-sentencing states that have been in the business the longest failed to
carry out 95% or more of their death sentences. Nearly half failed to convert more than
1 in 30 death sentences into executions. Three-quarters carried out fewer than 7% of their
death sentences. The vast majority (86%) carried out 15% or fewer.

•

Only 1 state, Virginia, managed to carry out more that a quarter of the death sentences
it imposed over the 23-year study period—and there is serious question whether it did so
only by dint of inferior error detection.238

*****
Through a variety of measures, our 23 years worth of findings reveal a capital punishment
system collapsing under the weight of its own mistakes. In so doing, they pose three principal
questions (and a host of subsidiary ones) that will be the subject of a second report later this year:
•

What has remained the same, and what has changed, since 1995? By all indications
examined here, the error-proneness and irrationality documented by our study of thousands
of cases reviewed by hundreds of state and federal judges, in three separate review
processes, in 34 states across the nation over the course of nearly a quarter century has
not somehow evaporated in the succeeding four years.239 In none of those four years, for
124

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

example, as in none of the preceding 23, has the nation managed to execute even 3% of its
death row inmates—and in 1996 and 1998, it executed fewer than 2% (about the same
proportion as it had executed in, e.g., 1984, 1987, 1993 and 1995).240 Indeed, if the recent
findings of a variety of media investigations across the nation are any indication, error rates
and the consequent confounding of the death penalty system may be getting worse.241 In this
regard, we hope to explore whether the surge of state and federal court reversals in the last
study year (1995) was a harbinger,242 and any other patterns that may appear.
•

What accounts for the generally high rates of serious error that state and federal
courts have detected in American capital judgments? In this Report, we have briefly
examined the types of errors that predominate (incompetent lawyering and prosecutorial
misconduct leading the way243); identified differences among the respective states and federal
courts—for example, disproportionately low error-detection by the Virginia courts and the
U.S. Court of Appeals for the Fourth Circuit; noted the relationship between high error rates
and low execution rates (especially rates of death sentences carried out); discovered some
potentially suggestive evidence that low execution rates (especially, low rates of death
sentences carried out) are associated with high death-sentencing rates; and considered the
effect on death-sentencing and execution rates of (1) some demographic factors (finding that
homicide rates seem to have no effect on death-sentencing and execution rates, and that the
size of nonwhite populations may be inversely related to death-sentencing rates but directly
related to execution rates) and (2) judicial-contextual factors (finding that political pressure

125

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

on state judges and that state expenditures on courts may be positively correlated with deathsentencing rates but negatively correlated with the rate at which death sentences are carried
out). These analyses represent our first steps towards the main goal or our next research
phase: Identifying the causes of the huge amounts of serious error infecting American capital
convictions and sentences.
•

What policy responses are called for? In advance of these additional efforts to explore the
causes of our capital system’s error-proneness and irrationality, we have the least to say here
about the policy implications of our findings. That, however, will be a third important focus
of our next phase of research.

126

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

END NOTES

*. An abridged version of this Report will be published in the Texas Law Review, October 2000.
1.
See, e.g., Rethinking the Death Penalty, ABC News Nightline, May 22, 2000
<http://abcnews.go.com/onair/nightline/transcripts/nl000522_trans.html> (“[A] lot of places are rethinking
the death penalty. Last week in New Hampshire, the state Legislature voted to abolish capital punishment,
although the governor there vetoed the measure. And around the country, people are asking new questions
about overzealous prosecutors, incompetent defense lawyers, and . . . DNA testing, which has cleared some
people on death row. Nightline[] . . . reports on an old issue, which is the focus of a whole new debate.”);
Jonathan Alter & Mark Miller, A Life or Death Gamble: A New Debate About the Fairness of a Death
Sentence, Newsweek, May 29, 2000 <http://newsweek.com/nw-srv/printed/us/na/a20098- 2000may21.htm>
(noting “a new debate about the fairness of the death penalty”).
2.
See, e.g., Ofra Bikel, The Case for Innocence, Frontline, PBS, Jan. 11, 2000
<http://www.pbs.org/wgbh/pages/frontline/shows/case/etc/script.html>; Barry Scheck, Peter Neufeld & Jim
Dwyer Actual Innocence: Five Days to Execution and Other Dispatches from the Wrongly Convicted (2000).
3. See, e.g., Paul Duggan, Rising Number of Executions Welcomed, Decried, Washington Post, Dec. 13,
2000, at A3, available in 1999 WL 30308153; Abraham McLaughlin, 98 Executions in ’99 Reignite a Capital
Debate, Christian Science Monitor, Dec. 27, 1999, at 1, available in 1999 WL 5384560.
4. From 1984 to 1991, an average of about 15 men and women were executed each year in the United States.
The average rose to about 30 a year between 1992 and 1994, to about 60 in the next four years, and to 98 (the
most in a single year since 1951) in 1999. See Linda Greenhouse, Death Penalty Gets Attention of High Court,
N.Y. Times, Oct. 30, 1999, at A1 (“[T]here were 82 executions in the first 10 months of [1999], a pace
unequaled since the early 1950's.”); NAACP Legal Defense and Educational Fund, Death Row U.S.A., Spring
2000, at 8 [hereinafter, Death Row U.S.A.]. Notably, however, two states, Texas and Virginia, have accounted
for half the executions in the United States during the last 15 years. See id. at 11-22; Frank Green, Virginia
Bucks Death Row Flow, Richmond Times-Dispatch, March 13, 2000; infra note 218. In contrast, several other
states with large death row populations—California, Ohio, Pennsylvania, Mississippi and Tennessee, for
example—rarely execute more than one person in any year. See Death Row U.S.A., supra at 11-22. Moreover,
since 1976, the number executed annually in the United States has never exceeded three percent of the nation’s
death row population, and stayed continuously within the one-half to two percent range from 1984 to 1998.
See infra Table 2, Appendix E, at E-3. The likelihood that any death row prisoner will be executed has been,
and remains, low.
5.
See Dalia Sussman, Split Decision on Death Penalty, ABCNEWS.com. Jan 10, 2000
<http://abcnews.go.com/sections/politics/dailynews/poll00019.html> (reporting that in an ABCNews.com poll,
64% of Americans say they support the death penalty for people convicted of murder); Frank Newport,
Support for Death Penalty Drops to Lowest Level in 19 Years, Although Still High at 66%, Gallup News
Service, Feb. 24, 2000 <http://www.gallup.com/poll/releases/pr000224.asp>. See also David Frum, The
Justice Americans Demand, N.Y. Times, Feb. 4, 2000, at A29, available in LEXIS, News Library, NYT File

127

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

(questioning the democratic legitimacy of a moratorium on executions in light of broad popular support for
the penalty); Frank Green, Falwell Opposes a Moratorium, Richmond Times-Dispatch, Apr. 11, 2000, at B4,
available in 2000 WL 5035053 (reporting Rev. Jerry Falwell’s disagreement with other conservative
evangelical Protestant figures who have called for a moratorium, and Falwell’s call for expedited executions);
Eugene H. Methvin, Death Penalty Is Fairer than Ever, Wall St. J., May 10, 2000, at A26, available in
2000WL-WSJ 3028765 (arguing that lengthy appeals and DNA testing make capital convictions more reliable
than ever).
6. Compare Phoebe C. Ellsworth & Samuel R. Gross, Hardening of the Attitudes: Americans' Views on the
Death Penalty, 50 J. Soc. Issues, Summer 1994, at 19, 20 fig. 1, 52 (discussing poll data indicating increasing
popular support for the death penalty from 1967 through 1994) with Mark Gillespie, Public Opinion Supports
the Death Penalty, Gallup News Service, Feb. 24, 1999 <http://www.gallup.com/poll /releases> (noting
support for the death penalty has declined about 10 percentage points over the past six years) and Newport,
supra note 5 (noting a 4 percentage point decline in support for the death penalty in 2000) and Kathy Walt,
Death Penalty Support Plunges to 30-year Low, Houston Chron., Mar. 15, 1998, at A1 (nationally, opposition
to the death penalty grew from 7 percent in 1994 to 26 percent in 1998; in Texas, “support for the death
penalty has slipped to its lowest point in more than three decades”) and Editorial, A Moratorium on Killing,
St. Louis Post Dispatch, Dec. 26, 1999, at B2, available in 1999 WO 3062342 (“Could this [rise in opposition
to the penalty] be a start of a shift in public opinion . . .?”).
For statewide trends in public opinion, see Kathy Barret Carter, 63% of Jerseyans Favor Death
Penalty— But Support Drops 9% Since ‘94 Polls Show, Star-Ledger (Newark), Oct. 10, 1999, at 25; Death
Penalty
Information
Center,
Recent
Poll
Findings
(visited
May
17,
2000)
<http://www.essential.org/dpic/po.html> [hereinafter DPIC] (reporting that polls reveal drops in support for
the death penalty in Illinois, Kentucky, Michigan, North Carolina, and Virginia); Death Penalty Support
Declines, New Chicago Tribune Poll, PR Newswire, Mar. 6, 2000 (reporting that public support for death
penalty in Illinois fell from 76% in August 1994, to 63% in Mar. 1999, to 58% in 2000); Greg Lucas, Poll
Takes Snapshot of Californians' Views, S.F. Chron., Jan. 14, 2000, at A20, available in 2000 WL 6472849
(“In California, only 49 percent of those surveyed favor the death penalty compared to 56 percent nationally);
Support for Death Penalty Slips in Minnesota, Associated Press, Mar. 21, 2000 (reporting that the number of
Minnesotans who favor the death penalty dropped from 73% in 1996 to 57% in 2000); Walt, supra (Texas
polls; discussed above).
7. See Sussman, supra note 5 (in a January 2000 ABCNEWS.com poll in which 64% of Americans said they
support the death penalty for murder, the number of supporters dropped to 48% when life without parole was
proposed as a sentencing option). State-specific polls reveal similar trends. See, e.g., Carter, supra note 6
(reporting that in New Jersey, 63% approval for capital punishment drops to 44% when life without parole is
a choice); DPIC, supra note 6 & Editorial, supra note 6 (noting that among Missouri residents—who, in the
abstract, “overwhelmingly support” the death penalty—support for the death penatly drops to 46% when life
without parole is an alternative); Lucas, supra note 6 (reporting on a recent California poll that asked
respondents to choose between death or life without parole as the appropriate punishment for murder: 49%
chose death and 47% chose life without parole); Eric Zorn, Prosecutors Deaf to Outcry Against Death Penalty,
Chi. Trib., Mar. 7, 2000, at 1, available in 2000 WL 3623214 (showing a 15-point drop in support for the
death penalty—from 58% to 43%—when life without parole is an option). Forty-two states (including most

128

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

of the eleven noncapital sentencing states) offer life without parole as a sentencing option. See Editorial,
Rising Doubts on Death Penalty, USA Today, Dec. 22, 1999, at 17A, available in 1999 WL 6861984.
8. See Jim Yardley, Texas’ Busy Death Chamber Helps Define Bush’s Tenure, N.Y. Times, Jan. 7, 2000, at
A1.
9. See Alabama Looks to Speed Up Executions, Associated Press, Mar. 3, 2000; Yoji Cole, Napolitano Wants
to Find Ways to Speed Executions, Ariz. Repub., Feb. 16, 2000, at B7; Sara Rimer, Florida Passes Bill to
Quicken Execution Pace, N.Y. Times, Jan. 6, 2000, at A1; John Shiffman, Tennessee May Limit Death Row
Appeals, Nashville Tennessean, Apr. 10, 2000; see also Most Oklahomans Favor Death Penalty–Poll Finds
Most Want Shorter Appeals Process, The Oklahoman, Mar. 13, 2000.
Among the states mentioned, only Florida actually adopted speed-up legislation, and it was
unanimously invalidated under the state constitution by the Florida Supreme Court. See David Cox, Court
Strikes Down GOP's Death Row Appeal Plan, Florida Sun-Sentinel, Apr. 15, 2000, at 1A (“In a major blow
to Gov. Jeb Bush and state Republican leaders, the Florida Supreme Court on Friday unanimously struck down
the Legislature's overhaul of the appeals process for Death Row inmates.”).
10. See William Claiborne, Ill. Governor, Citing Errors, Will Block Executions, Wash. Post, Jan. 31, 2000,
at A1, available in 2000 WL 2283005 (“Gov. George H. Ryan (R) has decided to effectively impose a
moratorium on the death penalty in Illinois [by indefinitely staying all proposed executions] until an inquiry
has been conducted into why more death row inmates have been exonerated than executed since capital
punishment was reinstated in 1977. . . . ‘There are innumerable opportunities along the way for serious errors,
and the governor wants to take a pause here,’ Ryan's press secretary, Dennis Culloton, said today.”); Dirk
Johnson, Illinois, Citing Faulty Verdicts, Bars Executions, N.Y. Times, Feb. 1, 2000, at A1, available in
LEXIS, News File (“Citing a ‘shameful record of convicting innocent people and putting them on death row,’
Gov. George Ryan of Illinois today halted all executions in the state, the first such moratorium in the nation).
See also Steve Mills & Ken Armstrong, Gov. George Ryan Plans to Block the Execution of Any Death Row
Inmate, Chi. Trib., Jan. 30, 2000, available in 2000 WL 3631638 (citing a March 1999 poll showing that
Illinois death row exonerations have prompted 54% of the state’s voters to favor and only 37% to oppose a
moratorium, notwithstanding majority support for the penalty in the abstract).
11. In 1999, Nebraska’s unicameral legislature passed a death penalty moratorium bill co-sponsored by
Republican Senator Kermit Brashear and Democrat Senator Ernie Chambers. Governor Mike Johanns vetoed
the bill, but the legislature unanimously overrode the veto as to a section of the bill that allocated $165,000
to study the issue. See Robynn Tysyer, Death Penalty Study OK’d, Omaha World-Herald, May 28, 1999,
available in Westlaw, News File, ALLNEWS database.
12. See, e.g., Ken Armstrong & Steve Mills, String of Exonerations Spurs Legislative, Judicial Panels to
Study Reforms, Chi. Trib., Nov. 16, 1999, at N8, available in 1999 WL 2932558 (noting that the Illinois
General Assembly and Illinois Supreme Court have created four committees to study the death penalty); Bob
Chiarito, House Panel Set to Consider Moratorium on Executions, Chicago Daily L. Bull. Jan. 26, 2000, at
3, available in Westlaw, News Library, ALLNEWS File; Ryan Keith, Task Force on Capital Cases Calls for
Videotaping of Suspects, Chi. Trib., Mar. 16, 2000, at M6, available in 2000 WL 3646355; Steve Mills & Ken
Armstrong, Illinois Prosecutors Under Glare at Death Penalty Hearing, Chi. Trib., Jan. 28, 2000, at N20,

129

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

available in LEXIS, News File; Evan Osnos & David Heinzmann, Death Penalty Still an Option, Chi. Trib.,
Jan. 31, 2000; Maurice Possley & Ken Armstrong, Revamp Urged in Handling of Capital Cases, Chi. Trib.,
Nov. 4, 1999, at N1, available in 1999 WL 2928957.
13. See Moratorium Now! (visited May 16, 2000) <http://www.quixote.org/ej>. See also Benjamin
Wallace-Wells, States Follow Illinois Lead on Death Penalty, Boston Globe, Feb. 9, 2000, at A3, available
in LEXIS, News File.
[End notes 14 and 15 are omitted]
16. See John DiStaso & John Toole, Death Penalty’s Fate Still Up in the Air in Senate, Manchester Union
Leader, Mar. 13, 2000 (noting that on Mar. 9, 2000, New Hampshire’s majority-Republican House of
Representatives voted to abolish the death penalty); Robert Anthony Phillips, N.H. Considers Abolishing
Unused Death Penalty <http://www.APBnews.com> (Mar. 10, 2000); Rethinking the Death Penalty
(Nightline), supra note 1 (noting abolition bill’s eventual passage by New Hampshire Legislature and veto by
Governor).
17. See Brad Cain, Two Oregon Titans Want Death Penalty Ended, The Columbian, Apr. 7, 2000, at B5
(“[Governor John] Kitzhaber and [former Senator Mark] Hatfield—two of Oregon’s most popular political
figures—have lent their names to an effort to ask voters to outlaw capital punishment in November.”).
18. See Jack Elliott Jr., Death Row-Defense Bills Move Through Legislature, Biloxi Sun Herald, Mar. 2,
2000, at A5, available in LEXIS, News Library, BILSUNH File (discussing proposals in the Mississippi
legislature to provide state money to assist smaller Mississippi counties to bear the expense of competent trial
and state post-conviction representation in capital cases); Carol Marbin Miller, State High Court Raises
Standard for Death Row Case Lawyers, Miami Daily Bus. Rev., Nov. 5, 1999, at B1 (discussing the Florida
Supreme Court’s adoption of rules setting minimum standards for defense attorneys in capital cases—including
at least 9 jury trials in serious or complex matters and at least 2 capital cases—and encouraging trial judges
to appoint two defense lawyers in each case). See also Possley & Armstrong, Revamp Urged, supra note 12,
at N1 (reporting that at least a dozen states “have established minimum standards for defense attorneys in
capital cases,” which typically “require that at least two attorneys be appointed in capital cases and that they
have a certain number of years of experience in trying criminal matters”).
19. See, e..g., Mike Dorning, Death Penalty Reforms Gain Backers in D.C., Chi. Trib., Mar. 31, 2000, at N1
(describing bipartisan support and sponsorship for a House bill—paralleling one proposed by Democratic Sen.
Patrick Leahy and Republican Sen. Gordon Smith in the Senate—aimed at improving capital defendants’
access to DNA evidence, adequate representation and other protective procedures); Mike Dorning, Senator
to Propose Death Row Safeguards, Chi. Trib. (Feb. 10, 2000), at N1 (describing federal legislation proposed
by Senator Patrick Leahy and others that would, inter alia, give states incentives to “require that indigent death
penalty defendants be allowed a team of at least 2 court-appointed attorneys” who “meet competency standards
set by the U.S. Administrative Office of the Courts”).
20.
ABC This Week (ABC television broacast, Apr. 9, 2000) <http://abcnews.go.com/onair/thisweek/ThisWeekIndex.html> (roundtable discussion among Sam Donaldson, Cokie Roberts, George

130

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Stephanopolous & George Will, focusing on George Will’s column, cited infra note 21, and Rev. Pat
Robertson’s expression of support for a death penalty moratorium, in which all four panelists agreed with
Robertson, prompting Stephanoplous to discern a “really a tectonic shift in the politics of the death penalty”).
See also John Harwood, Bush May Be Hurt by Handling of Death-Penalty Issue, Wall St. J., Mar. 21, 2000,
available in 2000 WL-WSJ 3022420 (noting the “remarkable . . . absence of public protest” when Governor
Ryan declared the Illinois moratorium on executions and discerning “a national shift in the politics of capital
punishment”);
Michael
Kroll,
Executioner’s
Swan
Song?
(Feb.
8,
2000)
<http://www.salon.com/news/feature/2000- /02/08/death_penalty/index.html> (concluding that Governor
Ryan’s decision to suspend the death penalty represents a “public shift”); Bruce Shapiro, Capital Offense, N.Y.
Times Mag., Mar. 26, 2000 (“But suddenly . . . death-row innocence cases have taken hold of the public mind,
and capital punishment itself seems to be approaching a political tipping point.”). See also Mark Hansen,
Death Knell for Death Row?, ABA J., June 2000, at 40; Steven A. Holmes, Look Who’s Questioning the
Death Penalty, N.Y. Times, Apr. 16, 2000, available in LEXIS, News Library, NYTIMES File (noting a
“conservative rethinking” of the death penalty); Johnson, supra note 10 (reporting that the issue of wrongful
executions “is gaining resonance around the nation, after many years in which it was seen as essentially a dead
letter in American politics”); Lucas, supra note 6 (stating that recent public opinion polls suggest “that
politicians need not be so rigid in their stance and their perception of the public’s opinion on the use of the
death penalty”); Clarence Page, Close Calls on Death Row Finally Prompting Second Thoughts, Dallas
Morning News, Apr. 16, 2000.
21. In an opinion column discussing a recently published book, see Scheck, Neufeld & Dwyer, supra note
2, George Will concluded:
You could fill a book with . . . hair-curling true stories of blighted lives and justice traduced [as a
result of the capital conviction of innocent defendants]. Three authors have filled one. It should
change the argument about capital punishment and other aspects of the criminal justice system.
Conservatives, especially, should draw this lesson from the book: Capital punishment, like the rest of
the criminal justice system, is a government program, so skepticism is in order.
George F. Will, Innocent On Death Row, Wash. Post, Apr. 6, 2000, at A23, available in 2000 WL 2295245.
22. See Brooke A. Masters, Pat Robertson Urges Moratorium On U.S. Executions, Wash. Post, Apr. 8, 2000,
at A1, available in 2000 WL 2295691 (quoting Robertson’s statement that “a moratorium would be
appropriate”); Meet the Press (NBC television broadcast, May 7, 2000) (interviewing Rev. Pat Robertson,
who explained his simultaneous support for the death penalty and for a moratorium on executions); Robertson
Backs ,Moratorium, Says Death Penalty Used Unfairly, Chi. Trib. Apr. 8, 2000, at N12, available in 2000 WL
3654070.
23. See the views expressed by Fallwell, Frum and Methvin, supra note 5 and accompanying text.
24. See the views expressed by Robertson, Will and others, supra notes 20-22 and accompanying text.
25. Kroll, supra note 20.

131

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

26. Much of the information reported here is contained in ten data bases. The authors generated four of those
data bases; the other six were generated at least in part by others.
The first (electronically stored) data base the authors generated—referred to herein as “DADB”—
contains information on all 4,578 state capital direct appeals that were finally decided between 1993 and 1995.
To be “finally decided” within that 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. See also infra pp. 25-26. (Capital judgments are overturned on direct appeal only on the basis of
“serious error,” as defined infra note 33; infra p.5 & nn.42, 43.) 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. DADB contains: the
sentencing state; the year; outcome; citation; and subsequent judicial history (rehearing, certiorari) of the
decision finally resolving the appeal; and information about the basis for reversal of the capital judgment
under review, if a reversal occurred.
The second (electronically stored) data base that the authors generated—referred to herein as “HCDB”
—contains information on all 599 initial (i.e., nonsuccessive) capital federal habeas corpus cases that were
finally decided between 1993 and 1995. 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 (a)
denied habeas corpus relief, thereby approving the capital judgment, or (b) granted habeas corpus relief from
the capital judgment (either the conviction or sentence) on one or more grounds; (2) if an appeal was timely
filed, a United State Court of Appeals must have taken or approved action (1)(a) or (1)(b); and (3) if certiorari
review was timely filed, the United States Supreme Court must have either (a) denied review or (b) granted
review and taken or approved action (1)(a) or (1)(b). See also infra p.24. (Federal habeas relief from capital
judgments is granted only on the basis of “serious error,” as defined infra notes 33, 38; infra p.5 & nn.42, 43.)
HCDB contains: the sentencing state; the timing of the habeas petition and its adjudication at the various
stages; the outcome at the various stages; information about the petitioner, lawyers, judges, courts, victim,
offense; the aggravating and mitigating circumstances found at trial; procedures used during the habeas review
process; and the asserted and the judicially accepted bases for and defenses to habeas relief from the capital
judgment was under review.
The third data set generated by the authors is laid out in full in Appendix C to this Report. It contains
an incomplete list of the capital cases in which state post-conviction relief was granted between 1973 and April
2000, and provides available information about citations or other identifying information, the basis for the
grant of relief, the outcome on retrial, and timing. A full description of that data set and of the manner in
which it was gathered, and its limitations, is set out infra Appendix C, pp. C-1 to C-2.
Our fourth and final author-generated data base, PolPres, collects information about the constitutional
and statutory law governing the selection and retention of judges in each of the 28 capital-sentencing states
that we study. It includes information on method of selection and retention of judges, length of judicial terms,
frequency of judicial elections, and types of judicial elections (e.g., selection, retention and recall elections).
The first of the data bases relied upon here that was generated at least in part by others —referred to
herein as “DRCen”—is a compilation of the information used to produce the NAACP Legal Defense Fund’s
quarterly death row census, Death Row U.S.A., supra note 4. This data base has the name of all individuals

132

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

who were on a state death row between 1973 and 1995, the state where their death sentence was imposed, and
the sentencing year. Death Row U.S.A. is also our source of information about executions: when and where
they occurred and whether they were consensual or non-consensual, as described infra notes 31, 208; infra
p.32 & n.140.
Three additional data sources used here contain information collected by the United States
Government.“USCen” is a compilation of information collected by the United States Census Bureau. In order
to estimate the racial composition of each state and circuit (region) in our study, we used Unpublished Census
data PE-19 1970-79 and three Census Bureau publications: State Estimates by Age, Sex, and Race; Estimates
of the Population of States by Age, Sex, Race and Hispanic Origin: 1981 to 1989; and Estimates of the
Population of State by Age, Sex, Race and Hispanic Origin: 1990 to 1998. (Figures for 1980 were estimated
by averaging 1979 and 1981). “UCRDB” is a compilation of information reported in U.S. Dep’t of Justice,
FBI Uniform Crime Reporting Program Data [United States]: County Level Arrest and Offense Data, for the
years 1973 through 1996. “PrisCen” is a compilation of information collected by the Bureau of Justice
Statistics and reported in the Sourcebook of Criminal Justice Statistics for the years 1977 through 1996.
Our penultimate data base—CtCaLd—has information for each state in our 28-jurisdiction cohort
about the state’s average annual criminal case filings per 1,000 persons in the population for years 1985-1994.
These data, and the underlying case load measure, are taken from Inter-University Consortium for Political and
Social Research, State Court Statistics 1985-1994 (ICPSR 9266, 1995). Our final data base—CtExpen—has
information for each of the same 28 states on its average annual court-related expenditures for fiscal years
1982-1992. These data, and the underlying measure, are taken from Expenditure and Employment Data for
the Criminal Justice System 1992 (ICPSR 6579, 1993).
27. 408 U.S. 238 (1972).
28. Although the Justice Department collects aggregate data on capital cases by state, its data (1) have only
37 variables, (2) contain no case- or event-specific information, (3) are derived from reports by prison officials
who lack information about some individuals under sentence of death who are incarcerated in local jails or for
some other reason are not physically located on death row, and (4) are derived from answers to questions about
outcomes that (a) do not distinguish between state and federal court reversals, and (b) provide no information
on the reason for a reversal. See, e.g., U.S. Dep’t of Justice Bureau of Justice Statistics, Capital Punishment
1998, at 1 <http://www.ojp.usdoj.gov/bjs/pub/pdf/cp98.pdf>
(Dec. 1999; NCJ 179012) (supporting
documentation is available on line) [hereinafter, BJS 1998 Report]. Likewise, although the NAACP Legal
Defense Fund’s quarterly death row census, see Death Row U.S.A., supra note 4, lists inmates on death row,
it provides very little information about each. See infra note 123 (discussing these and other limitations of the
data in Death Row U.S.A.).
29. We are now conducting complex multivariate statistical analyses to identify potential causes of those
results. We will report on those analyses later in the year.
30. Our study considers only state, not federal, death sentences.
31. DRCen; Death Row U.S.A., supra note 4, at 8-22. The figure in the text refers to all executions during
the study period. For the reasons discussed infra pp.32, 41, it often is sensible to consider only the executions

133

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

that were “non-consensual,” meaning that the prisoner availed himself of the full review process before he was
executed. The number of non-consensual executions between 1973 and 1995 was 273, or 4.7% of the total
number of death sentences.
32. DRCen; DADB. The state direct appellate process is described infra pp. 18-19.
33. DADB. In calculating error rates, we count only errors that result in reversal of a capital conviction or
sentence. To do so, the error must be “serious” in three respects that render our calculation of “error”
conservative. First, to be reversible, error must be prejudicial, either because the defendant has actually shown
that it probably affected the outcome of his case or because it is the kind of error that almost always has that
effect. See generally 2 James S. Liebman & Randy Hertz, Federal Habeas Corpus Practice and Procedure §§
32.1, 32.3, 32.4 (3d ed. 2000) (generally discussing the harmless error doctrine). The vast majority of error
that state appellate courts discover is deemed harmless and does not result in reveral. In Illinois, for example,
in addition to reversing half of the capital judgments it has reviewed, “the Illinois Supreme Court has upheld
scores of death sentences while forgiving trial errors that benefited prosecutors, dismissing the errors as
harmless.” Ken Armstrong & Steve Mills, Death Row Justice Derailed, Chi. Trib., Nov. 14, 1999, at 1,
available in 1999 WL 2932178. One such case was Anthony Porter’s case, in which the Illinois Supreme Court
based its harmlessness findings on the “‘overwhelming’” evidence of Porter’s guilt; Porter was later released
as innocent when another man confessed to his crime. Id. Another study of harmless error found that:
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 those cases—or nearly three out of four times—the reviewing court
affirmed the conviction, holding that the misconduct was “harmless.”
Ken Armstrong & Maurice Possley, Break Rules, Be Promoted, Chi. Trib., Jan. 14, 1999, at 1, available in
1999 WL 2834609. And in Oklahoma, although at least four convicted murderers have received new trials
“based upon appellate findings that [Oklahoma City’s District Attorney] broke the rules,” that same office has
been criticized by courts for similar misconduct in “at least 17 other” cases in which the errors were found to
be harmless. Ken Armstrong, ‘Cowboy Bob’ Ropes Wins—But at Considerable Cost, Chi. Trib., Jan. 10, 1999,
at N13.
Second, to be reversible, error generally must have been properly preserved. Most state direct appeal
courts will not grant relief based on error—no matter how egregious and prejudicial—that the defendant did
not properly preserve by way of (1) a timely objection at trial, (2) reiteration in a timely new trial motion at the
end of trial, and (3) timely and proper assertion on appeal. See 1 Liebman & Hertz, supra §§ 7.1a, at 276-77
& n.29, 26.1. This is true even in cases in which the failure to preserve the error was the fault of counsel, not
the defendant, and even in many instances in which the lawyer’s mistake resulted from inexperience,
incompetence or sheer stupidity, and not a valid exercise of professional judgment. See Stephen B. Bright,
Death by Lottery— Procedural Bar of Constitutional Claims in Capital Cases Due to Inadequate
Representation of Indigent Defendants, 92 W. VA. L. REV. 679, 683 (1990); Randall Coyne & Lyn Entzeroth,
Report Regarding Implementation of the American Bar Association’s Recommendations and Resolutions
Concerning the Death Penalty and Calling for a Moratorium on Executions, 4 GEO. J. ON FIGHTING POVERTY
3, 28-30 (1996). Numerous prisoners have been executed despite acknowledged prejudicial errors affecting

134

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

their convictions and sentences, because they failed to preserve their objections. Examples include the capital
prisoners in Gray v. Netherland, 518 U.S. 152, 162-70 (1996); Coleman v. Thompson, 501 U.S. 722, 747-49
(1991); Dugger v. Adams, 489 U.S. 401, 408 (1989); Smith v. Murray, 477 U.S. 527, 533-35 (1986), each
of whom had an evidently meritorious constitutional claim that he was capitally convicted or sentenced in
violation of the United States Constitution but nonetheless was denied relief in state (and then, as a
consequence, federal) court based on his failure to assert the claim at the time or in the manner required by
state law and was subsequently executed. See Death Row U.S.A., supra note 4, at 9-22.
Finally and most obviously, error—no matter how prejudicial—only results in reversal if it is
discovered. If it is not discovered, because, for example, the party responsible for it fails to disclose it, see, e.g.,
infra note 98, reversal will not occur and the error will not be deemed “serious” by our measure.
Hundreds of examples of “serious error” found in state post-conviction proceedings are collected in
Appendix C infra. Dozens of examples of the even narrower category of “serious error” that warrants federal
habeas relief are collected in Appendix D infra. See also cases cited infra notes 36, 44, 97-106.
34. The state post-conviction process is described infra pp.19-20.

35. Our post-conviction data are set out in Appendix C. For discussion of the incomplete nature of these data,
see infra n.39; infra pp. 26-27, 33-34; infra Appendix C, pp. C-1 to C-2.
36. Appendix C; Florida, Georgia, Indiana, Maryland, Mississippi, North Carolina and Tennessee Capital
Punishment Report Cards, infra Appendix A. We say “at least” in the text for the reasons set out infra note
39; infra pp.26-27 & n.132, 33-34 & n.152; Appendix C, infra pp. C-1 to C-2.
For the reasons stated in Appendix C, p. C-13 n.10, Georgia has used a variety of post-conviction
procedures to derail many more death sentences than we count as post-conviction reversals (e.g., by ordering
hearings on mental retardation (which poses a constitutional bar to execution in Georgia)—that very often
never take place, leaving the prisoner with a tacit life sentence).
The category of “serious error” that leads to state post-conviction reversal is narrower than “serious
error” at the direct appeal stage, cf. supra note 33, because, generally, only properly preserved state and federal
constitutional violations that (1) were not, and (2) could not have been raised on direct appeal can be the basis
for state post-conviction reversal. As at the direct appeal stage, moreover, error—no matter how egregious and
how much it undermines the accuracy of the capital verdict—never gets corrected at the state post-conviction
stage (and thus does not count as “serious error” in our analysis) unless it is discovered and litigated. See supra
note 33. And given the failure of a number of capital-sentencing states—Virginia, prominent among them—to
provide any lawyers or funding for them at all at the state post-conviction stage, the likelihood that serious
error will not be discovered and litigated in state post-conviction proceedings is often very high. See, e.g.,
American Civil Liberties Union of Virginia, Unequal, Unfair and Irreversible: The Death Penalty in Virginia
(Apr. 2000) <http://www.aclu.org/news/2000/n040700a.html> (visited Apr. 28, 2000) [hereinafter, Virginia
Report]
The United States Supreme Court itself occasionally grants relief in capital cases on review of state
direct review proceedings. See, e.g., Yates v. Evatt, 500 U.S. 391, 411 (1991) (overturning conviction due to
prejudicial jury instructions giving the defendant the burden of proof); Johnson v. Mississippi, 486 U.S. 578,
585-90 (1988) (overturning death sentence that state prejudicially based on unconstitutional and unreliable

135

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

aggravating circumstance); ; Tison v. Arizona, 481 U.S. 137, 158 (1987) (overturning two death sentences that
were imposed absent proof of the constitutional minimum level of criminal culpability required to impose
death); Truesdale v. Aiken, 480 U.S. 527 (1987) (overturning death sentence imposed after trial court forbade
defendant to inform jury of important aggravating information about his demonstrated prospects for
rehabilitation). We treat these Supreme Court cases reviewing state post-conviction decisions as findings of
serious (in all these cases, federal constitutional) error infecting capital sentences. For many additional
examples of “serious error” that was caught and corrected during state post-conviction proceedings, see infra
Appendix C. See also cases cited infra notes 97-106.
37. HCDB. “Final review” is defined supra note 26; infra pp.24-26.
38. HCDB. The definition of “serious error” that warrants reversal in federal habeas corpus proceedings is
even narrower than the analogous definitions at the direct appeal stage (which is set out supra note 33 and
accompanying text) and at the state post-conviction stage (see supra note 36). This is because error is only
reversible on habeas if it meets the three criteria for “seriousness” on direct appeal—the error must be (1)
prejudicial, (2) properly preserved and (3) discovered, see Liebman & Hertz, supra note 33, §§ 7.1a, 11.2b,
26.1, 32.1-32.5; supra note 33—and if, in addition, the error (4) violates the federal Constitution, see 28
U.S.C. §§2241(c)(3), 2254(a); (5) not arise the Fourth Amendment exclusionary rule (search and seizure
violations, that is, cannot be the basis for federal habeas relief), see Stone v. Powell, 428 U.S. 465, 495 (1976);
(6) in habeas cases litigated in 1989 and after, is not based on a “new rule” of federal law, see Teague v. Lane,
489 U.S. 288, 299 (1989), and (7) in habeas cases litigated in 1993 and after, meets an especially high standard
of prejudice or “harmful error,” see Brecht v. Abrahamson, 507 U.S. 619 (1993). See generally Liebman &
Hertz, supra note 33, §§ 9.1, 9.2, 25.1, 32.1 (discussing constraints (4)-(7) on habeas relief). Dozens of
examples of “serious error” warranting federal habeas relief from capital judgments imposed by nearly all of
the study states are collected in Appendix D infra. See also cases cited infra notes 97-106, 140.
39. The production-line/product-inspection analogy helps explain how these figures are calculated. The
“overall error rate” is the proportion of capital judgments thrown out during the first (state direct appeal)
inspection due to serious error, plus the proportion of the original judgments that survive the first inspection
but are thrown out at the second (state post-conviction) inspection, plus the proportion of the original
judgments that survive both state inspections but are thrown out at the final (federal habeas) stage. The “overall
success rate” is the converse. In note 40 infra, we use this method to calculate the national composite “overall
error rate.”
As we indicate by our use of the phrase “at least” in our narrative, and by our use of the “≥” symbol
in the national, state and circuit Report Cards, see Appendix A, the “overall error rates” calculated here are
in fact underestimates. Due to incomplete data, we assume that all death sentences that survived the direct
appeal inspection and are not known to have been reversed during the state post-conviction inspection passed
muster during that inspection. In fact, many capital judgments affirmed on direct appeal were pending in, but
had not yet been finally decided by, state post-conviction proceedings by the end of the study period. Inflating
the denominator in this way—i.e., using the class of cases available for review as a proxy for the cases that
actually underwent final review—leads us systematically to overestimate the success rate and underestimate
the error rate. See infra pp.26-27 & n.132, 33-34 & n.152; infra Appendix C, pp. C-1 to C-2.
40. DADB; Appendix C; HCDB. Because 41% of the capital judgments reviewed on state direct appeal were

136

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

found to be tainted by serious error, only 59% of those judgments were available for state post-conviction
review. Because at least 10% (this figure is probably higher, see supra note 39; infra Appendix C, pp. C-1 to
C-2) of that 59%—meaning at least 5.9% of the original pool (≥.10 x .59 = ≥.059)—failed this second, state
post-conviction inspection, the overall rate of error found by state courts is 47% (41% + 6%) of the original
pool. Then, of the 53% (100%-47% = 53%) of capital judgments that were available for federal habeas review,
40%—meaning 21% of the original pool (.40 x .53)—failed the federal inspection. The “overall error rate”
thus is at least 68% of the overall pool (41% +≥6%+ 21% = ≥68%). In other words: At least 68% of the capital
judgments that were fully inspected were found seriously flawed at some stage.
(We have simplified the above calculation by omitting fractions represented by numbers after the
decimal points. In computing overall rates in the various report cards, we included the numbers after the
decimal point until the error rate was obtained, at which point we applied the normal rounding convention.)
Our “overall error rate” is not the rate of error in the 5,780 death sentences imposed between 1973 and
1995. That number cannot be calculated because, at the end of 1995, many of those death sentences were
pending in some court awaiting review, but had not yet been finally resolved at one of the three inspection
stages. This rate instead uses the outcomes of the 4,578 cases in which state direct review occurred during the
study period, and the 599 of those cases in which subsequent federal habeas review occurred, together with
the 248 known state post-conviction reversals (taken as a proportion of the 2,693 capital judgments that had
“cleared” state direct appeal) to calculate the error rate found in capital judgments that were finally reviewed.
41. See supra notes 33, 36, 38; infra p.5.

42. The data in this Report on the types of “serious error” that led to the reversal of capital judgments come
from our study of state post-conviction reversals, set out in Appendix C. See State Post-Conviction National
Composite Results, infra Appendix C, p. C-3. A variety of prejudicial errors in the instructions given to
jurors— which by legal definition lead to reversal only if they probably affected the outcome of the trial, see
Boyde v. California, 494 U.S. 370, 380 (1994)—account for another 20% of the reversals, and, together with
lawyer incompetence and law enforcement misconduct, account for three-fourths of all state post-conviction
reversals. When reversals due to demonstrably prejudicial judicial or juror bias are added, the total for the four
types of claims discussed so far (ineffective assistance of counsel, prosecutorial misconduct, unconstitutional
jury instructions and judge/jury bias) reaches 80% of all reversals. See State Post-Conviction National
Composite Results, infra Appendix C, p. C-3.
43. See, e.g., Williams v. Taylor, 120 U.S. 1495, 1496 (2000); Strickler v. Green, 527 U.S. 263, 264 (1999).
44. See State Post-Conviction National Composite Results, infra Appendix C, p. C-3. If a capital conviction
is overturned on appeal or post-conviction review, the defendant may be (1) released for lack of evidence of
guilt (as, for example, in the Bowen/Oklahoma, Brown/Florida, Jimerson/Illinois, Nelson/Georgia and
Williamson/Oklahoma (among many other) cases summarized in Appendix C and Appendix D); (2) permitted
to accept a plea to a lesser offense or to the same offense but a lesser penalty (as in the Carriger/Arizona, Jent
& Miller/Florida cases in Appendix C); (3) retried and (a) acquitted (as in the Munson/Oklahoma case
summarized in Appendix C and in the Wallace/Georgia case summarized in Appendix D ), (b) released upon
the jury’s failure to agree on a verdict (as in the Kyles case summarized in Appendix D), (c) reconvicted of

137

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

a noncapital offense (as in numerous cases in Appendix C and Appendix D), (d) reconvicted of a capital
offense but awarded a lesser sentence (ditto), or (e) reconvicted and resentenced to die. If only the death
sentence was overturned, the defendant may be (1) offered and accept a plea or other arrangement resulting
in a lesser sentence; or (2) subjected to a new sentencing hearing at which the outcome is (a) a lesser sentence
or (b) a death sentence. For a listing of outcomes in recent North Carolina cases, see Stephen Dear, A Death
Penalty Cease-Fire for N.C., News & Observer (Raleigh), Apr. 16, 2000, at A31, available in 2000 WL
3924050:
Last May, a Superior Court [state post-conviction] judge overturned the murder conviction
and death sentence of Charles Munsey . . . because it was clear that he was innocent of murder, and
that the district attorney who prosecuted him . . . as well as other law officials withheld exculpatory
evidence. Tragically, Munsey died . . . awaiting a new trial.
Last summer, a Guilford County prosecutor told a [state post-conviction] hearing judge that
he “just plain forgot” about a credible independent witness who could have provided a solid alibi for
[death row inmate] Stephen Mark Bishop. Bishop is awaiting a second trial.
In November [1999], Alfred Rivera had been on North Carolina’s death row for two years for
a double murder . . . when, in a second trial, a jury acquitted him. The N.C. Supreme Court [on direct
appeal] had ordered the new trial, ruling that the trial judge should have allowed jurors to hear
testimony that Rivera had been framed by his co-defendants.
[Governor] Hunt commuted the death sentence of Wendell Flowers . . . in December over
doubts about his guilt . . . .

45. As revealed by the data collected in Appendix C, the post-reversal outcomes in our state post-conviction
study were as follows:
Outcomes Following State Post-Conviction Reversals, 1973-April 2000
Sentence
Less than
Death*

Not Guilty
of Capital
Crime*

Death
Sentence

247

Total,
Outcome
Known

Died
Awaiting
Retrial

301

Retrial
Pending as
of 4/2000
37

Outcome
Unknown

Total, All
Cases

342

*The “Not Guilty of Capital Crime” column, a subset of the “Sentence Less than Death” column,
includes individuals as to whom murder charges either were dropped by the prosecutor, dismissed by
the trial judge, or rejected by the jury. Individuals who were reconvicted of murder—even noncapital
degrees of murder—and were given a sentence other than the death penalty are included in the
“Sentence Less than Death” column but not the “Not Guilty of Capital Crime” column.

138

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

46. See supra note 39; infra pp.26-27 & n.132, 33-34 & n.152; infra Appendix C, infra pp. C-1 to C-2 (all
explaining why we say “at least”).
47. DADB; Appendix C; HCDB. See Table 10 and Figure 12, infra pp.68, 69. Recently, the regional press
has discovered the same patterns our study demonstrates, in a variety of states: California, Florida, Illinois,
Nevada, Tennessee, Utah and Washington. See infra note 241 (summarizing the journalists’ findings).
48. See supra note 10 and accompanying text.
49. See, e.g., Governor Says He Will Not Impose Moratorium on Executions, A.P. Newswires, Feb. 15, 2000
(quoting Florida Governor Jeb Bush as stating: “Illinois appears to have a unique problem with the
administration of capital punishment. Here in Florida, there is no competent evidence that suggests an innocent
person has been wrongly executed.”); Sara Rimer & Raymond Bonner, Bush Candidacy Puts Focus on
Executions, N.Y. Times, May 12, 2000, at A1 (quoting Texas Governor George W. Bush explaining on Meet
the Press that he did not consider events in Illinois relevant to Texas’s death penalty system because in Illinois,
but not in Texas, “‘they’ve had some problems in their courts . . . they’ve had some faulty judgments’”).
50. DADB; Appendix C; HADB. See National Composite and Illinois Report Cards, infra Appendix A, pp.
A-5, A-25; Figures 6-13 and Tables 4-10, infra pp.47, 50, 53, 54, 57, 58, 60, 61, 63, 64, 68, 69, 72, E-5, E-6
(state comparisons).
51. See Figure 3, Table 3, infra pp.38, E-4.
52. See Figure 4, Table 3, infra pp.39, E-4; infra pp.35-37.
53. Data on direct appeal and post-conviction outcomes in noncapital cases are sketchy, but suggest the
following conclusions: (1) At the direct appeal stage, serious, or reversible, error is detected in about 12 to 20%
of the noncapital criminal judgments that are appealed. (2) Noncapital criminal judgments that are appealed
make up only a small subset of the criminal convictions that are obtained. The vast majority of criminal
convictions are a result of bargained guilty pleas, and most convictions based on pleas are not appealed. (By
contrast, virtually every capital conviction and sentence is appealed. See infra notes 87-88 and accompanying
text.) (3) The best available evidence is that serious error is detected in about 3% of the noncapital federal
habeas corpus petitions that are filed, and that such petitions are filed by about 3 or 4 out of every 1,000 state
prisoners each year. (4) Although there are no similar data for noncapital state post-conviction proceedings,
most criminal lawyers believe noncapital error is detected less often there than on federal habeas corpus, and
that prisoners are no more likely to seek state post-conviction than federal habeas corpus review. (These
conclusions are based on evidence presented in James S. Liebman, The Overproduction of Death, 100 Colum.
L. Rev. (forthcoming 2000); Daniel J. Meltzer, Habeas Corpus Jurisdiction: The Limits of Models, 66 So.
Cal. L. Rev. 2507, 2524 (1993) (“of every thousand person convicted in state prosecutions and committed to
custody in any given year, only three to four actually file habeas corpus petitions challenging their custody”);
Brief Amicus Curiae of Benjamin Civiletti, et al., in Support of Frank R. West in Wright v. West, No. 91-542,
505 U.S. 277 (1992) (filed Mar. 4, 1992), at App. A, Table I & n.1 (providing data on the rate of relief granted
to state prisoners from 1963-1981).
Assume, very conservatively, that 70% of all criminal judgments are reviewed on direct appeal, among

139

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

which 20% (14% of the original pool) are found to contain serious error; that 10% of the cases that were
affirmed on direct appeal (i.e., 6% of the original pool) go on to state post-conviction review, at which stage
5% (.3% of the original pool) are found to contain serious error; and that 10% of the cases that were affirmed
on direct appeal and were not overturned on state post-conviction (another 6% of the original pool) go on to
federal habeas review, at which stage another 5% (.3% of the original pool) are found to contain serious error.
Even vastly overestimating the appeal and reversal rates in this way generates only a 15% (14% + .3% + .3%
= 14.6%) overall error rate.
54. See, e.g., 142 Cong. Rec. S3362 (daily ed. April 16, 1996) (Sen. Hatch) (“[O]ne of the biggest [Federal
habeas corpus] problems [is] looney judges in the Federal courts who basically will grant a habeas corpus
petition for any reason at all.”).
With two exceptions (Delaware and Maryland), all of the capital-sentencing states in the 28-state
cohort on which most of our analyses focus make their judges stand for election either by the public directly
(in 24 of the states) or periodically by the state legislature (in South Carolina and Virginia). See 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, 776-80 (1995).
55. DADB, Appendix C; HCDB. See National Composite Capital Punishment Report Card, infra Appendix
A, pp. A-5 o A-6. Because some post-conviction reversals are unknown, see supra note 39; infra Appendix
C, pp. C-1 to C-2, while all federal court reversals are known, the ratio of state to federal reversals is actually
higher. On the other hand, we count a handful of United States Supreme Court reversals on certiorari following
direct appeal and state post-conviction as, respectively, direct appeal and state post-conviction findings of error.
See, e.g., supra note 36; infra note 93 and accompanying text.
56. See, e.g., Slack v. McDaniel, 120 S. Ct. 1595, 1605 (2000); Williams v. Taylor, 120 S. Ct. 1495 (2000);
Stewart v. Martinez-Villareal, 523 U.S. 637, 644 (1998); Liebman & Hertz, supra note 33, §§ 2.4, 30.2.
57. HCDB. See National Composite Capital Punishment Report Card, infra p.30 and infra Appendix A, p.
A-6. Judicial review of the 120 individuals executed in the years 1989-1995 consumed a total of 1274.53 caseyears, meaning 10.6 years per case. HCDB. A Justice Department study concludes that the time from death
sentence to execution has increased over time to about 11 years for 1998 executions. See BJS 1998 Report,
supra note 28, at 1.
58. The data underlying Figure 1—taken from DRCen, DADB, Appendix C and HCDB—are displayed in
Tables 1, 10 and 28, infra p.68; infra Appendix E, pp. E-2, E-22.
59. Between 1984 and 1991, there were an average of 15 non-consensual executions each year; that number
rose to 27 between 1992 and 1994, to 53 in the succeeding four-year period and then to 88 in 1999. See Death
Row U.S.A. supra note 4, at 8-22. On the reasons for focusing on non-consensual execution, see supra note
31; infra p.32& n.140.

60. The data depicted in Figure 2—which are taken from BJS 1998 Study, supra note 28; Death Row U.S.A.,
supra note4, at 8-22—are displayed in Table 1, infra Appendix E, p. E-2.

140

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

61. See Table 2, infra Appendix E, at E-3.
62. The proportion of death row executed each year has moved up modestly during the 1990s—albeit at
nothing like the rate at which the number of executions has risen, and staying mainly within the 1.5% to 2.5%
range. See Table 2, infra Appendix E, p. E-3. Even this increase may be the result of swelling numbers of
prisoner piled up on death row—as overburdened judicial inspectors, faced with ever-expanding numbers of
cases under and awaiting their review, inadvertently miss more serious error, or become more tolerant of it and
more often let it through.
63. See infra note 190; infra pp.51, 59 & n.190, 65, 106-07.
64. For this view, see the statements by Virginia officials quoted in Brooke A. Masters, A Rush on Va.’s
Death Row, Wash. Post, Apr. 28, 2000, at A1, available in 2000 WL 19606141 (presenting the arguments of
Virginia officials who attribute the pronounced discrepancy between Virginia and other states to Virginia’s
prosecutorial restraint and narrow sentencing statutes).
65. For a report taking this position, see Virginia Report, supra note 36 (discussed infra p.107).
66. For this purpose, 1989 was an average year. See BJS 1998 Report, supra note 28, at 12, tbl. 12.
67. See id. at 13, app. tbl. 1.
68. See id.
69. See id.
70. Id. at 1, 14 & app. tbl. 2; Death Row U.S.A., supra note 4, at 1.
71. BJS 1998 Report, supra note 28, at 1, 14 & app. tbl. 2; Death Row U.S.A., supra note 4, at 1. (The 6,700
figure used here covers the 1973-1999 period, and includes an estimate of death sentences imposed in 1999,
which is not covered by the Justice Department’s 1998 report.)
Returning to our 1989 example, the 13 executions by 1998 of individuals sentenced to die in 1989
represent only 1 in 20 of the 263 people condemned in 1989.

72. See BJS 1998 Report, supra note 28, at 13, app. tbl. 1; id. at 6, tbl. 5.
73. See supra Part II, pp.3-14.
74. The trial, incarceration and execution of sentence in capital cases cost from $2.5 to $5 million dollars per
inmate (in current dollars), compared to less than $1 million for each killer sentenced to life without parole.
See, e.g., Aaron Chambers, Resources a Concern in Death Penalty Reform, Chi. Daily L. Bull., Apr. 24, 1999,
at 19, available in Westlaw, News Library, CHIDLB file (estimating that a capital case costs $5.2 million from
pretrial proceedings to execution); Margot Garey, The Cost of Taking a Life: Dollars and Sense of the Death

141

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Penalty, 18 U.C. DAVIS L. REV. 1268, 1268-70 (1985); Samuel R. Gross, The Romance of Revenge: Capital
Punishment in America, 13 STUDIES L., POL. & SOC. 71, 78 (1993) (reporting a $3.2 million cost per
execution in Florida, and Kansas’ rejection of the death penalty because of the cost); Paul W. Keve, The
Costliest Punishment—A Corrections Administrator Contemplates the Death Penalty, Federal Probation, Mar.
1992, at 11; Duncan Mansfield, The Price of Death Penalty? Maybe Millions, A.P. Newswires, Mar. 26, 2000,
available in Westlaw News Library, APWIRES file (estimated $1 to $2 million cost per Tennessee execution);
David Noonan, Death Row Cost Is a Killer: Capital Cases Can’t Be Handled Fairly and Affordably, Critics
Claim, N.Y. Daily News, Oct. 17, 1999, at 27, available in 1999 WL 23488045 (giving cost of prosecuting
and defending New York capital cases at the trial phase, in a period during which only five capital sentences
were imposed (from 1994 to 1999), as $68 million); A. Wallace Tashima, A Costly Ultimate Sanction, The
Los Angeles Daily J., June 20, 1991 (cost per execution to California taxpayers is $4 to $5 million).
75. When post-trial review costs are factored in, the cost comparison between capital and noncapital cases
is something like $24 million dollar per executed prisoner, compared to $1 million for each inmate serving a
sentence of life without possibility of parole. See S.V. Date, The High Price of Killing Killers, Palm Beach
Post, Jan. 4, 2000, at 1A, available in 2000 WL 7592885. See also Ken Armstrong & Steve Mills, Inept
Defenses Cloud Verdicts, With Their Lives at Stake, Chi. Trib., Nov. 15, 1999, at N1, available in 1999 WL
2932352 (“in Illinois, the resources rallied on appeal often dwarf those summoned to keep a defendant off
Death Row in the first place”); Armstrong & Mills, Justice Derailed, supra note 33, at N1 (discussing the
“staggering” costs of capital case reversals and exonerations in Illinois: “Taxpayers have not only had to
finance multimillion-dollar settlements to wrongly convicted Death Row inmates—[Dennis] Williams alone
received $13 million from Cook County—but also have had to pay for new trials, sentencing hearings and
appeals in more than 100 cases where a condemned inmate’s original trial was undermined by some
fundamental error.”).
76. Cf. Maurice Possley & Ken Armstrong, The Flip Side of a Fair Trial, Chi. Trib, Jan. 11, 1999, at N1 and
Maurice Possley & Ken Armstrong, Prosecution on Trial in DuPage, Chi. Trib., Jan. 12, 1999, at 1 (study of
effects of prosecutorial misconduct in Illinois homicide and capital cases, concluding that “the reversals exact
a toll on victims and their families who are forced to come back to court, reopening sometimes barely healed
emotional wounds”).
77. See supra p.4; National Composite State Post-Conviction Results, infra Appendix C, p. C-3.

78. For example, see Armstrong & Mills, Justice Derailed, supra note 33:
Capital punishment in Illinois is a system so riddled with faulty evidence, unscrupulous trial
tactics and legal incompetence that justice has been forsaken, a Tribune investigation has found. . .
.
The findings reveal a system so plagued by unprofessionalism, imprecision and bias that they
have rendered the state’s ultimate form of punishment its least credible.
79. See Figure 35, infra p. 111; Tables 28 and 29, infra Appendix E, pp. E-22, E-23.

142

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

80. See supra Figure 2, p.13.
81. See Dan Rather, Dead Wrong: Did the State of Texas Execute an Innocent Man?, CBS 60 Minutes II,
Apr. 12, 2000 <http://cbsnews.cbs.com/now/story/0,1597,182812-412.shtml> (visited May 17, 2000)
(contending that there is strong evidence that Jerry Lee Hogue, whom Texas executed in 1998, was innocent).
Between 1972 and the beginning of 1998, 68 people were released from death row on the grounds that their
convictions were faulty, and there was too little evidence to retry the prisoner. See Samuel R. Gross, Lost
Lives: Miscarriages of Justice in Capital Cases, 61 L. & CONTEMP. PROB. 125, 130-32 (1998); Michael L.
Radelet et al., Prisoners Released from Death Rows Since 1970 Because of Doubts About Their Guilt, 13
COOLEY L. REV. 907, 916 (1996) . As of this writing (May 2000), the number of inmates released from death
row as factually or legally innocent apparently has risen to 87, including nine released in 1999 alone. See Frank
Green, Question of Life or Death: Illinois Exonerations Spark a Debate, Richmond Times-Dispatch, Apr. 2,
2000, at A1, available in 2000 WL 503442.
82. See Scheck, Neufeld & Dwyer, supra note 2, at 172-92 (attributing the conviction of the innocent in large
part to incompetent lawyers and prosecutorial suppression of evidence—the two most common errors detected
in the reversals discussed in this study, see supra p. 5; National Composite State Post-Conviction Results, infra
Appendix C, p. C-3).
83. Cf. Ken Armstrong & Steve Mills, Flawed Murder Cases Prompt Calls for Probe, Chi. Trib., Jan. 24,
2000, at N1, available in 2000 WL 3629579 (reporting that Illinois paid $36 million to settle lawsuits by four
men who were wrongly convicted of murder, and two of whom were sentenced to die); Sasha Abramsky, Trial
by Torture, Mother Jones, March 3, 2000 ($1 million paid to civil rights plaintiffs who were tortured into
confessing to (and then being falsely convicted of) capital crimes); Laurie Goering, Florida Lets Speed Govern
Executions, Chi. Trib. Feb. 28, 2000, at 1, available in 2000 WL 3640614 (noting that Florida paid $1 million
in damages for falsely incarcerating two inmates on death row for 12 years); Paul M. Valentine, Maryland to
Give Cleared Man $300,000, Wash. Post, June 23, 1994, at B1, available in 1994 WL 2426459.
84. See Ken Armstrong & Maurice Possley, Trial & Error: How Prosecutors Sacrifice Justice, Chi. Trib. Jan.
13, 1999, at N1, available in 1999 WL 2834238 (detailing how, 12 years after the “Ford Heights 4" were
falsely convicted in Chicago (two capitally) of two rape-murders, and five years before the four were
exonerated following several judicial decisions ordering a new trial, one of the actual perpetrators still at large
suffocated a third woman to death in a vacant apartment near the scene of the earlier crimes); Brooke Masters,
Lucky Release from Behind Bars, Wash. Post, Apr. 28, 2000, at A23 (discussing David Vasquez’s
incarceration in Virginia for a capital murder he did not commit, and the murder spree on which the real killer
embarked in the meantime).
85. See supra note 21 (discussing George Will’s conclusion that innocent men and women have been
executed); supra note 33 (discussing how close the Illinois Supreme Court came to missing the miscarriage
of justice in Anthony Porter’s case).
All the implications of our findings that we discuss in text are poignantly illustrated by a recent article
in the Seattle Times about Seattle murder victim, Esther Vinikow. After prosecutors said they would consider
the views of the victims’ family before deciding whether to seek the death penalty against the alleged killer,
Robert Wentz, a reporter interviewed Ms. Vinikow’s children:

143

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Like most Americans, Esther Vinikow's children support the death penalty. But they say
Wentz, if found guilty, should not be executed. Not because whoever killed her doesn't deserve it, but
because it takes too long and costs too much.
To Jerome Vinikow, 58, Esther Vinikow's only son, the death penalty seems to only protract
the tragedy . . . . “As long as he's away permanently, I'm not sure . . .,” he trails off. “If he does get
the death penalty, and it's 10 to 12 years of waiting, I don't know what good that does.”
In many ways, the family's misgivings reflect a growing national impatience and unease about
capital punishment. In the aftermath of a tragedy, they have become drawn into a discussion
that provides no easy answers.
Superior Court trials cost taxpayers an average of $388,680. State and federal appeals of
death-penalty cases take an average of 11 years, according to a recent study by state Supreme Court
Justice Richard Guy. That's eroded public confidence in the justice system, Guy said.
But polls also suggest growing unease about capital punishment, particularly after several
death-row inmates in Illinois were released when new evidence proved their innocence.
The decades it takes to execute an inmate may have saved lives, notes Jerome Vinikow . . .
. That possibility should not be lost in the rush for justice. “I'm not against the death penalty. I used
to wonder why it took 10 or 12 years, but it's obvious when you see all the mistakes in Illinois, you
have to be careful,” he said.
....
At first, [the victim’s daughter, Dolores] Beck-Schwartz, 62, of Putnam Valley, N.Y., wanted
whomever a jury convicted to be put to death. It seemed an appropriate punishment for someone who
took the life of such a defenseless, gentle person, she said.
But Beck-Schwartz had second thoughts when she considered the years that pass between trial
and execution—if the sentence isn't overturned along the way. “If it happened within a year, I'm fine
with that. But if it dragged on year after year, it won't make it any easier,” she said. “It won't bring her
back. It won't make me feel better.”
Alex Fryer, Victim’s Family Wrestles Death-Penalty Issue, Seattle Times, May 14, 2000,
<http://archives.seattletimes.com/cgi-bin/texis/web/vortex/display?slug=deth14m&date=20000514&query=vin
ikow.
86. See, e.g., Zant v. Stephens, 462 U.S. 862, 874-80 (1983); Gregg v. Georgia, 428 U.S. 153, 204-06
(1976); Furman, 408 U.S. 238.

144

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

87. Whitmore v. Arkansas, 495 U.S. 149, 174 & n.1 (1990) (citing statutes).
88. See id. at 174-75 (“since the reinstitution of capital punishment in 1976, only one person, Gary Gilmore,
has been executed without any appellate review of his case”).
89. See generally Leigh B. Bienen, The Proportionality Review of Capital Cases by State High Courts After
Gregg: Only ‘The Appearance of Justice,’ 87 J. CRIM. L. & CRIMINOLOGY 130, 131-33 (1996); Penny J.
White, Can Lightening Strike Twice? Obligations of State Courts After Pulley v. Harris, 70 U. COLO. L. REV.
813, 816-17 (1999).
90. For a brief overview of the direct appellate process with citations to other works, see Liebman & Hertz,
supra note 33, § 3.4a, at 177-79.
91. See supra notes 33, 36, 38; supra p.5 & nn.42, 43.
92. See Liebman & Hertz, supra note 33, at 178 (recommending the filing of certiorari petitions, particularly
in capital cases). By making certiorari the prisoner’s last opportunity to raise novel federal claims, the Supreme
Court has strongly encouraged prisoners, especially ones under sentence of death, to file certiorari petitions.
See id., § 25.1, at 940-41.
93. See, e.g., Lilly v. Virginia, 527 U.S. 116, 119 S. Ct. 1887, 1898 (1999) (reversing capital conviction);
Schad v. Arizona, 501 U.S. 624, 631-32 (1991) (affirming capital conviction); Mu’Min v. Virginia, 500 U.S.
415, 421 (1991) (affirming capital sentence); Enmund v. Florida, 458 U.S. 782 (1982) (reversing capital
sentence).
94. See, e.g., Lambrix v. Singletary, 520 U.S. 518, 527 (1997); Stringer v. Balck, 503 U.S. 222, 226 (1992).
95. The Supreme Court’s certiorari jurisdiction is limited to federal questions, which in criminal cases almost
always means federal constitutional questions. See 28 U.S.C. § 1257.
96. See 28 U.S.C. §§ 2254(b), 2254(c); O’Sullivan v. Boerckel, 119 S. Ct. 1728, 1731-33 (1999).
97. See generally Liebman & Hertz, supra note 33, § 7.1b, at 290-92, § 7.2e, at 314-17 & n.87, §§ 20.3e,
26.3b (providing examples and citing other sections of the treatise with additional examples).
98. See, e.g., Kyles v. Whitley, 514 U.S. 419, 441-45 (1995) (overturning conviction based on prosecutorial
suppression of evidence demonstrating, among other things, that the eyewitnesses who confidently identified
petitioner at trial as the attacker had originally described a different perpetrator and had only focused on
petitioner as a result of suggestive photo arrays).
99. See, e.g., Amadeo v. Zant, 486 U.S. 214 (1988) (holding that prosecutor’s failure to make public his
instructions to the jury commissioner to under-represent African-Americans on the jury venire provided
“cause” for the habeas petitioner’s failure to make a jury challenge in a timely manner).

145

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

100. See, e.g., Williams v. Withrow, 507 U.S. 680, 688 (1993) (violations of the right to counsel “would
often go unremedied” if left to review at trial and on direct review”); other authority cited in Liebman & Hertz,
supra note 33, § 25.4, at 969-70 n.42.
101. See, e.g., Liebman & Hertz, supra note 33, § 26.3b, at 1093-94 & n.28.
102.

See, e.g., People v. Fields, 690 N.E.2d 999 (Ill. 1998).

103. See, e.g., Porter v. State, 723 So.2d 191 (Fla. 1998); Suarez v. State, 604 So.2d 488 (Fla. 1992); People
v. Fields, 690 N.E.2d 999 (Ill. 1998).
104. See, e.g., Williams v. Taylor, 120 S. Ct. 1495 (2000); State v. Freeman, 605 So.2d 1258 (Ala. Crim.
App. 1992).
105. See, e.g., Turpin v. Todd, 519 S.E.2d 678 (Ga. 1999); Simants v. State, 277 N.W.2d 217 (Neb. 1979).
106. See, e.g., Liebman & Hertz, supra note 33, § 26.3b, at 1090-92 & n.27.
107. See Case v. Nebraska, 381 U.S. 336 (1965) (granting certiorari to consider whether a constitutional right
to state post-conviction review exists, but dismissing the grant after Nebraska adopted a comprehensive state
post-conviction review scheme). But cf. Murray v. Giarratano, 492 U.S. 1, 10-11 (1989) (plurality opinion)
(“State collateral proceedings are not constitutionally required . . . .”); Pennsylvania v. Finley, 481 U.S. 551,
554-55 (1989) (plurality opinion) (similar). See generally Liebman & Hertz, supra note 33, § 7.1b.
108. See 28 U.S.C. §§ 2254(d), 2254(e)(1) (providing a laxer standard of review for certain kinds of claims
that were “adjudicated on the merits” in state court proceedings).
109. See Liebman & Hertz, supra note 33, § 3.5a, at 179-80, § 6.1 & n.1 (citing authority).
110. See McFarland v. Scott, 512 U.S. 1256, 1261 (1994) (Blackman, J., dissenting from the denial of
certiorari) (“State habeas corpus proceedings are a vital link in the capital review process, not the least because
all federal habeas claims first must be adequately raised in the state court ... [to avoid being denied in federal
court] as procedurally defaulted or waived . . . .”); Coleman v. Balkcom, 451 U.S. 949, 956-57 (1981)
(Rehnquist, J., dissenting from denial of certiorari) (describing typical post-trial course of proceedings in
capital cases, which includes a state post-conviction petition); Liebman & Hertz, supra note 33, §§ 6.4c, 7.1a,
7.1b, 7.2f (describing counsel’s legal and ethical obligations in regard to pursuing state post-conviction
remedies in capital cases).
111. See generally Liebman & Hertz, supra note 33, §§ 3.5a(6), 6.1, 6.2, 7.1.
112. See, e.g., Rimer, supra note 9, at A1, A9 (describing new state post-conviction procedures recently
adopted in Florida but then invalidated, see supra note 9, that, inter alia, gave capital prisoners 180 days after
the filing of their direct appeal brief to file a state post-conviction petition; barred all claims that were or could
have been raised at trial or on direct appeal; forbade extensions of time, even if delays were the result of the

146

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

state’s illegal withholding of exculpatory evidence or a court’s failure to compel legally required disclosure
of public records; barred successive petitions unless they were based on previously undiscoverable evidence
establishing a constitutional violation and the prisoner’s factual innocence; and imposed strict time limits on
the adjudication of state post-conviction and public records act petitions). See generally Liebman & Hertz,
supra note 33, § 3.3b nn.9-12 (discussing “unitary review” procedures).
113. See Liebman & Hertz, supra note 33, § 3.5a(6).
114. See id., § 6.4 & n.13.
115. See supra note 36.
116.

See 28 U.S.C. §§ 2241-2254.

117. See id. § 2243; Rules 2, 3 of the Rules Governing § 2254 Cases.
118. See 28 U.S.C. § 2253; Barefoot v. Estelle, 463 U.S. 880, 892-93 (1983); Liebman & Hertz, supra note
33, § 34.4.
119. See Liebman & Hertz, supra note 33, §§ 39.1, 39.3c.
120. See 28 U.S.C. § 2251; Liebman & Hertz, supra note 33, §§ 13.1, 13.2.
121. See supra note 33. Some state capital prisoners file, and in rare instances secure the stay of execution
needed to allow them to litigate, a second or “successive” federal habeas petition after their first petitions are
denied. See 28 U.S.C. § 2244; Liebman & Hertz, supra note 33, §§ 28.1-28.4. For the reasons given infra note
126, this study only considers error detected during initial federal habeas proceedings.
122. An early and very preliminary count of cases is reported in Memorandum to Senator Joseph F. Biden,
Chairman, Senate Judiciary Committee from James S. Liebman (July 15, 1991), reprinted in Statement of John
J. Curtin, Jr., President of the American Bar Association, and of James S. Liebman, Professor of Law,
Columbia University School of Law and Member, ABA Task Force on Death Penalty Habeas Corpus, on
behalf of the American Bar Association, Hearings before the Subcomm. on Civil and Constitutional Rights
of the Comm. on the Judiciary of the U.S. House or Rep. Concerning Fairness and Efficiency in Habeas
Corpus Adjudication, 102d Cong., 1st Sess (July 17, 1991).
123. See Death Row U.S.A., supra note 4. By combining the data on LDF reports produced periodically over
the period from 1973 to 1995, one can collect the name of and a small amount of information (e.g., race of
defendant and race and number of victims) about all individuals who have been incarcerated on death row for
at least some period of time between those dates. Although helpful, the LDF census did not narrow our casegathering task very much, because it contains nearly 6000 individuals who were on death row at some point
during the period, the vast majority of whom have never had their cases reviewed on federal habeas corpus
(many having received relief or still being in the process of seeking relief in the state courts), and because the
information—a name and a state, e.g., Charles Williams of Georgia— often leads to many false positives in

147

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

follow-up computer research. See also supra note 28.
124. See supra p.3 & nn.28-29.

125. HCDB. Habeas corpus cases typically become final upon the Supreme Court’s denial of a petition for
certiorari either by the prisoner or by the state challenging an adverse decision of a U.S. Court of Appeals.
Many more such denials are announced by the Court on the first Monday in October than on any other day,
because that is when the Court generally rules on cases that have accumulated over the summer months when
the Court is not in session. We accordingly chose the first Monday in October, 1995, as our termination point.
126. Although we collected data on the published outcomes of capital successive habeas litigation during our
study period, in addition to the outcomes of all initial federal habeas corpus petitions that were finally
adjudicated during the study period, our data on successive petitions are incomplete. (Many successive-petition
cases are never published, and they are difficult to find.) Our data indicate, however, that grants of habeas
review and relied based on successive petitions are rare, but not nonexistent. Grants of successive petitions
include Smith v. Singletary, 61 F.3d 815 (11th Cir. 1995); Aldridge v. Dugger, 925 F.2d 1320 (11th Cir.
1991); Booker v. Dugger, 922 F.2d 633 (11th Cir. 1991); Songer v. Wainwright, 769 F.2d 1488 (11th Cir.
1985); Schlup v. Bowersox, No. 4:92CV443 JCH (E.D. Mo. 1997). For these reasons, we only report here
the results of initial habeas corpus proceedings. In this respect, as well as others noted elsewhere, see supra
notes 33, 36, 39; infra pp.26-27 & n.132; infra Appendix C, pp. C-1 to C-2, C-13 n.10, our calculation of rates
of serious error is conservative and omits some judicial findings of even egregious error. (Because the
standards for successive habeas litigation have always been very stringent, see Liebman & Hertz, supra note
33, ch.28, it is only in the case of egregious error that relief is granted at this stage.)
127. See Table 8 and Figure 8, infra pp.60, 61.
128. See infra note 190; infra pp.62-66 & n.198 (presenting some data on this question).
129. See supra note 26.
130. DADB. See supra notes 33, 36, 38; supra p.5 & nn.42, 43 (defining “serious error”).
131. See cases collected in Appendix C infra. In some states, even appellate post-conviction decisions are
not generally published or available on line, as in Tennessee prior to 1985 and Nevada and Texas to this day.
132. It is possible to get a rough sense of how much we have overestimated the denominator (by treating all
cases available for review as if they actually were finally reviewed), by considering three facts. First, one out
of five cases available for state direct review during the study period was not finally decided at that stage
during that period. See infra pp. 32-33; National Composite Capital Punishment Report Card, infra pp.29 &
Appendix A, p. A-5. Second, cases often are pending for longer periods on state post-conviction review than
on state direct appeal, because the former, but not the latter, include evidentiary and multi-court proceedings.
Third, only 22% (599) of the 2,693 cases that cleared state direct appeal during the study period also cleared
state post-conviction and completed federal habeas review during the study period. See National Composite

148

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Capital Punishment Report Card, infra pp.29 & infra Appendix A, p. A-5.
If, say, 30% (i.e., 809) of the 2,693 cases available for state post-conviction review were not decided
during that period, which would leave a balance of 1,884 cases decided during the period, the state postconviction reversal rate, which we very conservatively estimate as 10%, would rise to 13% (still fairly
conservatively estimated), and the national overall rate of error would rise to 70%.

133. The state report cards themselves are collected in Appendix A, infra.
134. The federal judicial circuit/regional report cards are collected in Appendix B, infra.
135. DRCen. See supra p.3.
136. DADB. Georgia imposed the nation’s first post-Furman death sentence on Chester Thomas Akins in
early May 1973, about six weeks after Governor Jimmy Carter signed the state’s post-Furman death-penalty
statute into law. Six months later, the state supreme court overturned Akins’ death sentence. See Akins v. State,
202 S.E.2d 62 (Ga. 1973).
137. See Akins v. State, 202 S.E.2d 62 (Ga. 1973) (discussed supra note 136).
138. Death Row U.S.A., supra note 4, at 10.
139. See Ramsey Clark, Spenkelink’s Last Appeal, 229 Nation 385 (1979).
140. See supra notes 33, 36, 38; supra p.5 & nn.42, 43. Of the 313 executions between 1973 and 1995, 273
(87.2%) were non-consensual and 40 (12.8%) were consensual. See Death Row U.S.A., supra note 4, at 8-22.
One might hypothesize that individuals who contemplate ending their appeals and being executed do so in
large part because of a belief that their capital judgments are error free, hence that their appeals are fruitless.
If that were the actual motivation for consented-to executions, and if, in addition, death row inmates’
evaluations of their chances on appeal were accurate, it would make sense to treat non-consensual executions
the same as others. The available evidence is inconsistent with these conjectures, however. Numerous
examples exist of men who nearly were executed after they initially gave up their appeals, then changed their
minds and had their death sentences—in some cases, multiple death sentences—overturned. See, e.g., Potts
v. Kemp, 814 F.2d 1512 (11th Cir. 1987) (reinstating, in pertinent part, Potts v. Zant, 734 F.2d 526, 529-30,
535-35 (11th Cir. 1984) (overturning multiple capital convictions of prisoner who previously came within days
of being voluntarily executed, then decided at the last minute to pursue his appeals, based on trial court’s
failure to instruction the jury on essential elements of capital murder, and based on the prosecutor’s inaccurate
statements in closing argument that “prior decisions of the state supreme court mandated the imposition of the
death penalty in this case”)); Vickers v. Ricketts, 798 F.2d 369, 373 (9th Cir. 1986), cert. denied, 479 U.S.
1054 (1987) (overturning conviction of prisoner who came within days of being voluntarily executed, then
changed his mind, because the jury instructions at his trial kept the jurors from considering a lesser included
offense supported by evidence); Clark v. Louisiana State Penitentiary, 694 F.2d 75 (5th Cir. 1982) (overturning
capital conviction of prisoner who originally attempted to end his appeals, then changed his mind, because the
jury at his trial was instructed that he had the burden of proving a critical element of capital murder). See

149

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

generally Liebman & Hertz, supra note 33, § 4.2 (discussing factors other than likelihood of success on appeal
that lead condemned inmates to give up their appeals and ask to be executed).
141. DRCen. All of these death sentences were imposed by state courts.
142. See Death Row U.S.A., supra note 4, at 8-22.
143. The three levels of judicial inspection are described supra pp. Part V, pp.18-22.
144. See supra pp.4-5 & nn.39, 40.

145. DADB. See supra note 26; supra pp.25-26 (defining “final review”).
146. DRCen; DADB. Death sentences imposed (5760) - death sentences finally reviewed on direct appeal
(4578) = death sentences awaiting direct review (1182).
Death sentences awaiting direct review (1182) ÷ death sentences imposed (5760) = percentage
awaiting direct appeal (21%).
147. DADB. See supra note 33 (defining “serious error,” meaning in this context, only error that was
discovered, preserved and prejudicial).
Additional information on most of the direct appeal decisions discussed here is contained in the state
report cards in Appendix A infra. Appendix A contains state report cards for the 28 states with at least one
federal habeas corpus decision. Direct appeal information for the remaining 6 capital-sentencing states is as
follows:
Direct Appeal Reversal Rates in States in Which No Capital Judgments
Had Completed Federal Habeas Review by End of Study Period
State

Number of
Death Sentences

Number Reversed/ Number
Reviewed On Direct Appeal

Percent Reversed
on Direct Appeal

7/8

Connecticut

3/3

100

New Jersey

33/38

New Mexico

2/8

183

30/125

28/32

287

103/214

Colorado

Ohio
Oregon
Total

150

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

148. DADB. Number reviewed (4578) - number reversed (1885) = number carried forward to next inspection
stage (2693).
149. DADB. The vast majority of capital prisoners who remain alive seek state post-conviction review. See
supra note 100 and accompanying text. Some number of prisoners die of natural causes or foul play, see, e.g.,
BJS 1998 Report, supra note 28, or forgo state post-conviction review and volunteer to be executed, see supra
notes 31; supra p.32 & n.140 and accompanying text.
150. DADB. Number reviewed (4364) - number reversed (1782) = number carried forward to next inspection
stage (2582).
151. See supra note 39; supra pp. 26-27 & n.132; infra Appendix C, pp. C-1 to C-2.

152. The number of known reversals is set out in the “Number Reversed on Post-Conviction” row within the
“Error Rates/State Post-Conviction” section of each report card. Because it is not possible to obtain
information on all state post-conviction reversals, see supra note 39; supra pp. 26-27 & n.132; infra Appendix
C, pp. C-1 to C-2, these figures are reported with the “≥” symbol.
153. The number of capital judgments moving forward from state direct appeal to state post-conviction is
listed in the last row of the “Error Rates/State Direct Appeal” section of each report card.
154. See supra note 39; supra pp. 26-27 & n.132; infra Appendix C, pp. C-1 to C-2.
155. Appendix C; DRCen; DADB. See supra note 132.
156. Appendix C; DRCen; DADB. Following the same procedure used to (under)estimate the state postconviction reversal rate (in which we use the number of capital judgments available for state post-conviction
review as a rough proxy for the number of capital judgments actually reviewed at that stage), see supra note
39; supra pp. 26-27 & n.132, 33-34 & n.152; infra Appendix C, pp. C-1 to C-2, we calculate this figure by
taking the sum of the reversals at the state direct appeal and state post-conviction stages as a proportion of the
total number of capital judgments reviewed on direct appeal. In the national composite report card, we use the
figures for the 28-state cohort of states with cases furthest along in the review process: (1782 + 248) ÷ 4364
= .47.
157. Actually, the first Monday in October 1995. See supra note 125.
158. See supra note 26 and supra p.24 (defining “finally review”).
159. HCDB. See supra notes 33, 38 (defining “serious error,” meaning, in this context, that the error was:
discovered, preserved, prejudicial, not based on an invalid search and seizure, violated the U.S. Constitution,
and (in the post-1988 cases) not based on “new law”).

151

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

160. See supra pp. 4-5 & nn.39, 40 (discussing the calculation of these rates, and showing how the 68%
overall error rate for the nation was calculated). The error and success rates in brackets are for only the state
direct appeal and federal habeas stages; the nonbracketed numbers include state post-conviction reversals, as
well.
161. DADB; DRCen; Appendix C; HCDB. As is shown in brackets on the national report card, if only the
(first) state direct appeal and the (third) federal habeas stages are considered, the combined national error rate
was 64% and the combined success rate was 36%. Although our information on cases at those two stages is
more accurate than our information about the state post-conviction stage, the information that is available on
the intermediate stage provides a reliably conservative estimate of what took place there. See supra note 39;
supra pp. 26-27 & n.132; infra Appendix C, pp. C-1 to C-2. For this reason, we usually focus on the more
comprehensive, three-stage “overall” rates.
162. The data in Figure 3—drawn from DADB, HCDB—are presented in Table 3, infra Appendix E, p. E-4.
163. See supra note 39; supra pp. 26-27 & n.132; infra Appendix C, pp. C-1 to C-2.
164. The data in Figure 4, which are compiled from Appendix C, are also displayed in Table 3, infra
Appendix E, p. E-4.

165. This figure is likely to be more meaningful when only cases from a single state are considered.
166. DRCen; DADB; National Composite Capital Punishment Report Card, supra p.29 & infra Appendix
A, p. A-5.
167. See Table 2, infra Appendix E, p. E-3.
168. HCDB; Death Row U.S.A., supra note 4, at 8-22.
169. HCDB.
170. See supra note 31; supra p.32.
171. See infra notes pp. 78-87.
172. DRCen; UCRDB; USCen.
173. Death Row U.S.A., supra note 4, at 8-22; UCRDB; USCen.
174. DRCen; Death Row U.S.A., supra note 4, at 8-22; UCRDB; USCen.
175. USCen.

152

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

176. UCRDB.
177. UCRDB; USCen.
178. PrisCen. This category of information and the next are omitted from the national, but presented in the
state and regional, report cards.
179. PrisCen.
180. USCen.
181. See, e.g., James Eisenstein & Herbert Jacob, Felony Justice: An Organizational Analysis of Criminal
Courts (1977); Martha Myers and Suzette Talarico, The Social Contexts of Sentencing (1987).
182. See PolPres. See also Bright & Keenan, supra note 54, at 76-80 (describing types of judicial elections);
supra note 54 (listing study states with judicial elections); infra note 221 (political pressure on judges).
183. This measure is from Stanley & Niemi, Vital Statistics on American Politics 1997-1998 (1999).
184. CtCaLd.
185. CtExpen.
186. See Virginia Report, supra note 36 (taking this position in regard to Virginia).
187. See supra note 64 (newspaper article quoting Virginia law enforcement officials taking this view).

188. See supra note 39; supra pp. 26-27 & n.132; infra Appendix C, pp. C-1 to C-2. Data on the number of
cases available for state post-conviction review in each state is found in the “Number Forward to State PostConviction” category of each state’s report card, infra Appendix A. We derive that number from DRCen and
DADB. The number of state post-conviction reversals, also provided on each report card, is computed from
the data in Appendix C.
189. The narrow category of error sufficiently egregious to qualify as “serious” and “reversible” at the federal
habeas stage is described supra note 38.
190. On one interpretation, there are actually four anomalies among the non-asterisked states on Figure 7.
Although 16 of the 20 non-asterisked states fall in the range of two-thirds to 1.5 times the national 40% rate
of error, four states—North Carolina, Missouri, South Carolina and Virginia—are below half the national
average. (As we noted, however, even compared to other anomalies, Virginia is an anomaly, at 15% of the
national average.)
The status of Virginia and Missouri here may seem to support the hypothesis (see supra note 64 and
accompanying text) that both states have lower rates of serious capital error than other states, because low error

153

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

rates are detected at successive state and federal inspection points. Although possibly valid for Missouri, this
hypothesis is confounded as to Virginia by a striking fact about that state and the other federal habeas outlying
states besides Missouri: All are states in which the availability of federal habeas relief is largely controlled by
the United States Court of Appeals for the Fourth Circuit, which, as we show elsewhere, has markedly lower
error detection rates than the other federal circuit courts. See Figures 8 and 33, infra pp.61, 104; Table 25,
infra p.103; infra p.106. (By contrast, the Eighth Circuit Court of Appeals, which presides over Missouri
habeas cases, does not consistently detect low rates of serious capital error. Contrasting with the 15% rate of
serious error it finds in Missouri capital judgments is the 48% rate of serious error it finds in Arkansas
judgments.) Given the Fourth Circuit’s consistent and pronounced inclination to find low error rates in all
capital judgments it reviews—including capital judgments from states (Maryland, North Carolina, and South
Carolina) whose own courts find exceptionally high rates of serious error in those states’ capital judgments,
see Table 6, Figure 6, supra pp.53, 54; supra p.55; infra pp.66 & n.198, 106-07—the Fourth Circuit’s
discovery of low rates of serious error in Virginia cases provides little confirmation of the low-error-rate
hypothesis, and little disproof of the lax-error-detection hypothesis.
191. See supra note 54.
192. See Carol S. Steiker, Counter-Revolution in Constitutional Criminal Procedure? Two Audiences, Two
Answers, 94 Mich. L. Rev. 2466, 2470 (1996).
193. The data underlying Figure 9—compiled from DADB and HCDB—are displayed in Tables 4 and 7,
supra pp. 47, 57.
194. The data underlying Figure 10—compiled from DRCen, Appendix C and HCDB—are displayed in
Tables 6 and 7, supra pp.53, 57.
195. Figure 10 is the more informative of the two charts because it permits us to compare all relevant state
judicial behavior to all relevant federal judicial behavior. See supra note 161.
196. The two measures, again, are (1) how much error judges (here, state vs. federal judges) detect when
reviewing capital judgments from the same state; and (2) how much error judges (state vs. federal) find relative
to the amount of error found in capital judgments from other states.
197. See supra notes 161, 195.
198. See also supra p.51. The Fourth Circuit’s low rates of error detection in capital (and, especially, Virginia
capital) cases are well known. See, e.g., Green, Virginia Bucks Death Row Flow, supra note 4; Masters, A
Rush on Va.’s Death Row, supra note 64.
The courts of another state in the Fourth Circuit, Maryland, also have very high capital error-detection
rates. See Table 6, supra p.53; Figure 6, supra pp.51, 54. Although Maryland’s federal habeas reversal rate
appears to be high as well, the state had only a small number of habeas cases reviewed during the study period,
and all were decided at the federal district court level, with the Fourth Circuit court of appeals never becoming
involved. See HCDB.
In contrast to the courts of Maryland, North Carolina and South Carolina, it is less likely that the

154

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Louisiana, Florida and Alabama courts have ratcheted up their error detection to compensate for predictably
low error detection by the Fifth Circuit (in reviewing Louisiana capital judgments) and the Eleventh Circuit
(in reviewing Florida and Alabama capital judgments). Unlike the Fourth Circuit’s uniformly low errordetection, the Fifth Circuit and Eleventh Circuits error-detection rates vary state to state, and are quite high
for some states (respectively, Mississippi and Georgia). See Table 8, supra p.60; Figure 8, supra p.6. This
variance suggests that the Fifth and Eleventh Circuit courts are sensitive to differences in the amounts of error
infecting the cases they review, see supra pp.59-60, and thus that it is those two federal courts (and not the
state courts) that are doing the compensating, based on how relatively error-prone or error-free they find capital
judgments from each of the states within their jurisdiction.
199. See supra pp.4-5 & nn.39, 40.
200. See supra note 161, explaining why we sometimes report reversal rates for state direct appeal and federal
habeas corpus, excluding state post-conviction, and on other occasions report the overall rates for all three
stages.
201. Two states from our cohort of 28, Delaware and Washington, are omitted from this analysis because
state post-conviction information is not available for them. Both in any event have less than three federal
habeas cases, making them relatively unreliable targets of comparison.
202. Kentucky, Maryland and Tennessee have 100% error rates, but only small numbers of final federal
habeas cases (2, 3 and 1 respectively).
203. See sources cited supra note 10.

204. See supra note 49.
205. The data underlying Figure 13 are displayed in Tables 6, 7 and 10, supra pp.53, 57, 68.
206. See supra note pp.40-41; National Composite Capital Punishment Report Card, supra p.30 & infra
Appendix A, p. A-6.
207. The same information—taken from DRCen; Death Row U.S.A., supra note 4, at 8-22—is in Table 11,
infra Appendix E, p. E-7.

208. By non-consensual executions, we mean ones occurring after the prisoner insisted upon and received
full judicial review. For further explanation of the difference between consensual and non-consensual
executions and the reasons for looking at the latter, and for some data about the relative frequency of each type
of execution, see supra note 31; supra 0p.32 & n.140, 41.
209. Two of the study states (Idaho and Pennsylvania) have yet to have a post-1973 non-consensual
execution.

155

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

210. The same information—from DRCen and Death Row U.S.A., supra note 4, at 8-22—is in Table 12,
infra Appendix E, p. E-8.
211. See supra notes 31, 140, 208.
212. The same information—from DRCen and DADB—is in Table 13, infra Appendix E, p. E-9.
213. See supra pp.40-41.
214. The data underlying all the comparisons in this section—which come from DRCen, Death Row U.S.A.,
supra note 4, UCRDB, USCen, PrisCen—are displayed in Tables 14-19, infra Appendix E, pp. E-10 to E-15.
Tables 14, 15, and 16 compare states’ death sentencing rates, respectively, per homicides, population and
prison population. Tables 17, 18, and 19 then make the same comparisons of the respective states’ nonconsensual execution rates.
215. Variations are not quite as great per prison population, suggesting that some part of the variation in
death-sentencing and execution rates per homicides and population is due to variable punitiveness among the
states.
216. Similarly, Nevada and Idaho are among the top three states when it comes to the proportion of homicides
that result in death sentences, but both states are in the very bottom cohort of states when it comes to the
proportion of their death sentences that are validated on judicial review and result in executions. See also infra
note 238. (Nevada and Idaho are also among the top four states when it comes to the proportion of their prison
population under sentence of death, but they are in the very bottom category of states when it comes to
executions.) Conversely, Virginia and Louisiana are in the top four states when it comes to the proportion of
their prison population that they execute but in the bottom cohort of states when it comes to the proportion of
their prison population that is under sentence of death.
217. See Jason DeParle, Abstract Death Penalty Meets Real Execution, N.Y. Times, Sept. 28, 1991, § 4, at
2 (discussing a period in 1987 when Louisiana executed eight men in 11 weeks and was “so enthusiastic about
capital punishment that a legal newspaper dubbed it ‘Death Mill, U.S.A.’”).
(Notes continue on the next page)

218. During the 1990s, Texas and Virginia have consistently executed about as many individuals as all the
other states combined:
Total Number of Executions Compared to
Executions by Texas and Virginia
Year

Total Executions

TX, VA Executions

1991

156

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

1992

1993

1994

1995

1996

1997

1998

1999

Total

455

226 (49.7%)

Death Row U.S.A., supra note 4, at 11-19. See Green, Virginia Bucks Death Row Flow, supra note 4; Masters,

A Rush on Va.’s Death Row, supra note 64; supra note 4.
219. The relevant states’ average homicides rate per 100,000 population during the 23-year study period—
taken from UCRDB, USCen—are in Table 20, infra Appendix E, p. E-16. See supra pp.43 (explaining how
average homicide rates are calculated). As Table 20 demonstrates, average homicide rates varied greatly among
death-sentencing states during the study period, ranging from 3.28 per 100,000 population in Utah to 15.19
per 100,000 population in Louisiana.
220. Average percent nonwhite populations for our 28-state cohort during the 23-year study period—taken
from USCen—are set out in tabular form in Table 21, infra Appendix E, p. E-17.
221. A number of authorities (1) have noted instances in which elected judges’ careers were positively or
negatively affected by whether their prior actions on the bench had seemed (respectively) sympathetic to, or
skeptical about, capital punishment, and (2) have concluded that political pressure is likely to skew capital
decision making by state court judges. See, e.g., Harris v. Alabama, 513 U.S. 504, 519-20 & n.5 (1995)
(Stevens, J., dissenting) (“The ‘higher authority’ to whom present-day capital judges may be ‘too responsive’
is a political climate in which judges who covet higher office—or who merely wish to remain judges—must
constantly profess fealty to the death penalty. . . . The danger [is] that they will bend to political pressures
when pronouncing sentence in highly publicized capital cases.”); Bright & Keenan, supra note 54, at 760
(“Decisions in capital cases have increasingly become campaign fodder in both judicial and nonjudicial
elections. The focus in these campaigns has been almost entirely on the gruesome facts of particular murders,
not the reason for the judicial decisions. Judges have come under attack and have been removed
from the bench for their decisions in capital cases—with perhaps the most notable examples in states with
some of the largest death rows and where the death penalty has been a dominant political issue. Recent
challenges to state court judges in both direct and retention elections have made it clear that unpopular
decisions in capital cases, even when clearly compelled by law, may cost a judge her seat on the bench, or
promotion to a higher court.”); Coyne & Entzroth, supra note 33, at 13 (“The death penalty and politics . .

157

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

. are inseparable,” particularly because “the vast majority of judges who preside over capital cases must answer
to the electorate” and because “‘judges are far less likely to . . . take . . . tough action if they must run for
reelection or retention every few years’” (quoting ABA, Report of the Comm’n on Professionalism, 112 F.R.D.
243, 293 (1986)); Symposium, Politics and the Death Penalty: Can Rational Discourse and Due Process
Survive the Perceived Political Pressure?, 21 Fordham Urb. L.J. 239 (1994).
222. Tables 22, 23 and 24—set out infra Appendix E, pp. E-18 to E-20—compare the 28 study states in
regard to, respectively, electoral pressure on judges, court expenditures per capita, and court caseloads per
capita.
223. We developed the political pressure measurement ourselves, using statutory information about how
judges are elected and retained in the various states. See supra note 26. We are fairly confident about the
quality of the underlying data. The other measures come from state-self-reported data, see id., the accuracy and
computational-comparability of which we are less sure of.
224. See sources cited supra note 221.
225. This proposal (were it supported by the data) would not call for spending less on each death sentence
obtained. Rather, it would call for spending less overall, by seeking and securing fewer death sentences
overall. The spending on each death sentence that is obtained might actually increase.
226. Included are report cards on the Third, Fourth, Fifth, Sixth, Seventh, Eighth, Ninth, Tenth, and Eleventh
Circuits.
227. See supra note 190; supra pp.14, 51, 59 & n.190, 65-66 & n.198.
228. Not counting the Fourth Circuit, federal courts found serious error in 229 (42%) of 547 death sentences
reviewed.
229. See supra Table 6 and Figure 6, supra pp.53, 54; supra pp. 51, 65-66 & nn.190, 198.
230. Outside the Fourth Circuit, the only other state where there are relatively low state and federal error
detection rates—although not nearly as low (in either case) as in Virginia—is Missouri, which falls within the
jurisdiction of the U.S. Court of Appeals for the Eighth Circuit. Cf. supra note 190.
231. Without changing this analysis, one could expand it to the two other death-sentencing states that border
Virginia, but are not in the same federal judicial circuit: Kentucky and Tennessee. (West Virginia and the
District of Columbia do not have the death penalty.)
232. See Williams v. Taylor, 120 S. Ct. 1495 (2000).

233. See Virginia Report, supra note 36, at 11-37; Green, Virginia Bucks Death Row Flow, supra note 4;
Masters, A Rush on Va.’s Death Row, supra note 64.

158

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

234. See supra note 31; supra p.32 & n.140; supra note 208 (explaining the reasons for focusing on nonconsensual executions).
235. Tables 26 and 27—derived from DRCen and Death Row U.S.A., supra note 4, and set out infra
Appendix E, p. E-27—compare the federal circuit courts based on, respectively, their component states’ deathsentencing and execution rates per 100,000 population.
236. See Figures 19-22, supra pp.82-84, 86.
237. Figure 35 is based on the information—taken from DRCen and Death Row U.S.A., supra note 4—in
Table 28, infra Appendix E, p. E-22. Figure 35 and Table 28 look at all executions, both consensual and nonconsensual. For the reasons discussed supra note 31; supra pp.32 & n.140, 41, a better measure of success
might be the proportion of death sentences carried out non-consensually. For that information, in tabular and
graphic form, see Table 29 and Figure 36, infra Appendix E, pp. E-23 to E-24.
(Notes continue on the next page)

238. See supra pp.14, 51, 59 & n.190, 65-66 &n.198, 105-07.
Comparing Figure 35 to Figure 19, supra p. 82, helps confirm a point made above—that the path to
more executions is not, as one might expect, more death sentences. See supra pp.82-87. A comparison of
Figures 35 and 19 reveals that:
•

Six of the top 11 (of 28) states when it comes to death sentences per 1,000 homicides, including the
top 4 states, are in the bottom half of the states when it comes to percent of death sentences carried
out after full review:
State

Wyoming
Idaho
Nevada
Arizona
Oklahoma
Mississippi

•

Rank in Death Sentences
per 1,000 Homicides

1 (of 28)
2
3
4
6
11

Rank in Percent Death
Sentences Carried Out
Following Full Review
16 (of 28)
23
26
20
19
18

On the other hand, of the top 5 states when it comes to percent of death sentences carried out after full
review are in the bottom 11 states in regard to death sentences per 1,000 homicides:
State

Rank in Percent Death
Sentences Carried Out

159

Rank in Death Sentences
per 1,000 Homicides

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Following Full Review
Virginia
Louisiana
Texas
Missouri

1 (of 28)
2
3
5

22 (of 28)
25
18
20

239. See supra pp.12-13 & Figure 2, 35-37.
240. See Table 1, infra Appendix E, p. E-2.
241. States in which recent press accounts have linked high capital error rates and the state’s incapacity to
make its death penalty work in a rational fashion include:
California: See Paul Elias & Rinat Fried, A Failure to Execute, The Recorder, Dec. 15, 1999, at 1 (“Since
1978, when . . . California . . . reinstitut[ed] the death penalty, 647 men and women have been sentenced to
death. Only [seven] have been executed. [Over] four times as many California death row inmates have died
in San Quentin of causes other than execution. Fifty-seven sentences have been overturned.”); Howard Mintz,
Slow Death: The Capital Punishment Gridlock in California, San Jose Mercury News, Mar. 12, 2000, at A1,
available in Westlaw, News Library, SJMERCURY file (reporting that between 1992 and 2000, California’s
death row grew from 350 to about 550 inmates, but it only executed 7 men; in the same period, state courts
overturned approximately 10 death sentence, and federal courts overturned 13).
Florida: Rene Stutzman, High Court Puts Death Cases Back into Play: Errors Were Found in 10 of 12
Capital Punishment Cases Reviewed this Year, Orlando Sentinel, Aug. 24, 1999, at D1, available in 1999 WL
2829798 (in the first eight months of 1999, the Florida Supreme Court found trial errors requiring retrial,
resentencing, or imposition of a life sentence in 83% of the first-time death penalty appeals it has reviewed;
the figure for all of 1998 was 77% (20/26)).
Illinois:
An Illinois Supreme Court ruling on Friday pushed the number of death-penalty cases in Illinois that
have been reversed for a new trial or sentencing hearing to 130—exactly half the total of those capital
cases that have completed at least one round of [state] appeals, according to a Tribune analysis.
Ken Armstrong & Christi Parsons, Half of State’s Death-Penalty Cases Reversed, Chi. Trib., Jan. 22, 2000,
at 1, available in 2000 WL 3629108.
Nevada: See Sean Whaley, Nevada’s Death Row History Criticized, Las Vegas Rev.-J., Feb. 7, 2000, at 1B,
available in Westlaw, News Library, LV-RJ-C file (finding that since 1979, 8 Nevada Death Row inmates have
been executed (all but one consensually, i.e., in advance of full judicial review, see Death Row U.S.A., supra
note 4, at 8-22); since 1993, the same number, 8, have had their capital judgments reversed by the state and
federal courts, among whom 3 (as of this writing, 4, see Brendan Riley, Emotional Mazzan Released, Las

160

"Broken System: Error Rates in Capital Cases, 1973-1995"
Copyright 2000. James Liebman, Jeffrey Fagan and Valerie West
.

Vegas Rev.-J., May 7, 2000, at 1) were thereupon released from prison).
Tennessee: Duncan Mansfield, The Price of Death Penalty? Maybe Millions, AP Newswires, Mar. 26, 2000,
available in Westlaw, News Library, APWIRES file (“Tennessee, with 97 people on death row [who have
accumulated over at least 23 years] is [still awaiting] its first execution since 1960.”).
Utah: See Lee Davidson, Death Row the End?: Most Get Out Alive, Deseret News (Salt Lake City), Dec. 13,
1999, at B1, available in 1999 WL 26543645 (noting that since Utah reinstated the death penalty in 1973, 16
prisoners have left the state’s death row, 6 by execution and 10 (63%) because their convictions or sentences
were overturned by the courts).
Washington: See, e.g., Mike Carter, Court Orders Retrial in 1986 Kitsap Rape-Murder Case, Seattle Times,
July 15, 1999, at B1, available in 1999 WL 6282738 (noting that 7 Washington State capital sentences were
overturned in 8 years, at a time when there were a total of only 14 men on Washington’s death row, see BJS
1998 Report, supra note 28, app. tbl. 2).
242. See supra pp. 36-38 & Figure 3.
243. See supra p.5; State Post-Conviction National Composite Results, infra Appendix C, p.C-3..

161