next | previous | table of contents

E. Results of Analysis 18: County-Only Analysis of Florida, Georgia and Texas

Florida, Georgia and Texas were among the first states to reinstate the death penalty after 1972, when the Supreme Court overturned all pre-existing sentences.571 During the 1973-1995 study period, those states accounted for one-fourth of the 1002 counties that imposed at least one death verdict, one-third of the death verdicts imposed, 38% of the death verdicts reversed and 40% of the verdicts finally reviewed. Florida led the nation on all these measures, save the number of capital counties.572 Each of the three states, and particularly Florida, thus provides a potentially useful laboratory for testing the impact of county-level factors on county reversal rates in a setting in which comparative state effects are not relevant.

During the study period:

  • 53 Florida counties imposed at least one death verdict during the 23-year study period. The total number of counties and years with at least one death verdict is 394 (about 7.5 observations—i.e., years in which at least one death verdict was imposed—per capital county, on average).

  • 88 Georgia counties imposed at least one death verdict during the study period, and there were 251 county-years with at least one death verdict (just under 3 death-sentencing years per capital county, on average).

  • 105 Texas counties imposed one or more death verdicts during the study period, and there were 353 county-years with at least one death verdict (3.36 observations per county, on average).

On all these measures, it is possible to say that Florida and its counties led the nation in death sentencing and the number of reversals during the study period.

1. Except in Florida, too little county variance for informative analysis.

For each state, we began with both a binomial and Poisson regression analysis of variance among county reversal rates calculated as proportions of imposed death verdicts that were reversed. The analyses treated county and year as random effects and time trend as a fixed effect. Those factors comprised the baseline analysis. In two of the three states, we encountered a result that helps explain why variance in capital reversal rates by county and year is not as fruitful a focus of analysis as state-by-year variance. Although the Analysis 18 baseline inquiry for Florida left some minimal amount of county-to-county variance to be explained,573 the baseline analyses for Georgia and Texas left no significant variance to be explained beyond that expected as a result of random differences among counties.574 None of the baseline analyses left significant unexplained year-to-year variance beyond that expected as a result of random variation.

2. Factors related to higher Florida county reversal rates: higher death-sentencing rates, population density and homicide rates.

Given the above results, we limited Analysis 18 to a Poisson analysis of variance in Florida county reversal rates, with county and year as random effects. Those factors plus time trend comprised the baseline analysis. The diagnostic tests indicate that the results are useful. The reversal rates predicted by the two best analyses (18A and 18B) of specific explanatory factors fit actual reversal rates significantly better than the reversal rates predicted by the baseline analyses, and both analyses left less unexplained variance than the baseline analyses.575

Individually and as a set, the significant explanations for Florida county reversal rates largely track those identified in our main state analysis. Controlling for other factors:

  • Florida verdicts imposed in later years were less likely to be reversed than earlier verdicts—at least in part because delays in reviewing cases means that later verdicts were less likely to have finished being reviewed by the study's end point, and thus that errors were less likely to have been found, and faulty verdicts less likely to have been reversed, by that point.576 Effect size is small, however—a predicted decline of only 2 to 3% per year577suggesting that an upward tug from higher rates of reversible error being detected over time was largely neutralizing the downward pull of delayed review.

  • Florida counties with higher numbers of death verdicts awaiting review have lower reversal rates—another result, at least in part, of delays in review.578

  • Florida counties with higher death-sentencing rates have higher capital error rates.579 In given years, Florida counties had death-sentencing rates per 1000 homicides ranging from .43 to 400, averaging 16. For each doubling of the death-sentencing rate—e.g., from 2 to 4, to 8, to 16, to 32—Analysis 18 predicts that county reversals will rise by about 15%.

  • More heavily and densely populated Florida counties have higher capital error rates than more sparsely populated counties.580 Effect size is large with predicted reversal rates nearly quadrupling across the range of counties from the least to the most urbanized.

  • In one of the two best analyses, counties with higher homicide rates have higher capital error rates. The group of factors that includes homicide rates does not improve fit as much as the alternative set of factors, however.581 Contributing to a pattern noted earlier, homicide rates are significant when two usually important factors are not significant—the size of the black population, and the homicide risk to whites relative to blacks. The county-level versions of these factors were not significant predictors of reversal rates in Analysis 18. (The state-level versions—which are significant in nearly all of our state and county-state analyses—cannot be used in this county-only study of a single state.582) Effect size is modest—about a 10% increase in reversal rates for every doubling of homicide rates.

F. Principal Findings of County Analyses: The Influence of Previously Identified State-Level Explanations for Capital Reversal Rates, Plus County Death-Sentencing Rates and Population Structure

Together with the eight state analyses discussed above, the 10 county analyses described here reveal that state, not county, conditions are the more informative focus of analysis. This is mainly because there is more measurable variance to explain at the state than at the county level.583 A possible cause of lower county variability is the smaller number of death verdicts and reviewing decisions in each county and year than in each state and year. Or, this might occur because aggregate state data on factors related to reversal rates better captures the sum of local experiences than do less frequently available and less reliable county data.584 At least in part, however, the result seems to reflect the fact that it is state rather than local policies that are more strongly related to differences in reversal rates. This latter interpretation is especially likely to be true of two consistently significant factors that are almost entirely traits of states:

  • State laws establish the method for selecting state judges, which in turn determines how much pressure judges are under to conform their rulings to popular opinion.

  • Capital and general caseloads and delays mainly reflect traits of uniform state court systems.

As is noted above, other significant state-level conditions substantially reflect state policies:585

  • Rates of death verdicts per 1000 homicides are substantially a result of the narrowness or breadth of state laws defining the range of cases in which the death penalty may be used.

  • Rates of apprehension, conviction and incarceration, and the extent of the threat of crime influential citizens face, are strongly affected by states' laws and crime-fighting policies.

Second, analyses of county reversal rates and county explanations for those rates do not identify significant explanations for state or county capital reversal rates that were not previously identified by main state Analysis 1 and supporting Analyses 2-6, 14-15. Although the county analyses provide new information—chiefly, that some important factors measured at the state level are also sometimes related to reversal rates when measured at the county level—this result supplements, without altering or neutralizing, state study findings. With the exception of population structure in Analysis 8, there was no suggestion that the effect of any of the important factors when measured at the county level deprived the same factor of significance when measured at the state level. With that one exception, significant county factors simply add a county dimension to factors already identified as—and that remain—important when measured at the state level.

Third, as is clear when Table 8 below is compared to Table 7 above (p. 238), the eight analyses in which state-level as well as county-level conditions were studied point to virtually the same set of clearly important state-level factors as the state analyses had previously identified, supplemented by a smaller number of related but less consistently significant county-level factors.

Table 8: State-Level Factors That Were Sometimes Significant in County-State Analyses, and How Often They Were Significant586

Which state-level explanatory factor?

Significant in how many of 8 analyses?*

Significant in what types of analyses?H

% of Population that is African-American

8

all four types

White/Black Homicide Victimization Rate

8

all four types

Index 1, 2 of Political Pressure on Judges

8

all four types

-Prison Population/100 FBI Index Crimes

8

all four types

-Death Verdicts Awaiting Review, 3 Stages

8

all four types

Population Size and Density

7

all four types

Death Verdicts/1000 Homicides

6

(+ 2 just above .05)

cnty-in-st, predval, da st=comp/cntys

% of Population that is African-American x White/Black Homicide Victimization Rate

5

(+1 just above .05)

cnty-in-st, predval, da

Death Verdicts Awaiting Decision x 4-

Factor Measure of All State Court

Cases Per Capita Awaiting Decision

-4-Factor Measure of All State Court

Cases Per Capita Awaiting Decision

+4-Factor Measure of All State Court

Cases Per Capita Awaiting Decision

5

 

 

3

 

1

cnty-in-st, predval,

st=comp/cntys

 

cnty-in-st, predval

 

cnyt-in-st

County Reversal Rates Predicted by Full Set of Important Factors in Best State Analyses

2

predval

+Passage of Time (reliable)^

1^

da

-Passage of Time (unreliable)^

2^

(+1 just above .05)

predval

-Homicide Rate+

1

cnty-in-st

As is true of the state-only analyses, the county-state analyses catalogued in Table 8 identify seven factors that are significantly related to capital reversal rates in a large majority of the analyses, and two related factors that sometimes are important:

  • High numbers of capital verdicts awaiting review are related to low county rates of capital reversal—at least in part because they are related to low rates of review.587

    • High per capita rates of backlogged court cases (noncapital as well as capital) had the same effect in some analyses, but effect size is very small.

    • On the other hand, the interaction of both large capital and large general backlogs is associated with higher county reversal rates in a majority of analyses.

  • High state death-sentencing rates are associated with higher capital error rates at the county level.

  • High proportions of African-Americans in a state's population are associated with higher county capital error rates.

  • A high risk of homicide to whites in the state, relative to the risk to blacks, is associated with higher capital error rates at the county level.

    • When this and the preceding racial condition—the proportion of the state's population comprised of African-Americans—interact, reversal rates are especially high.

  • State judicial selection techniques that increase political pressures on judges are associated with high county capital error rates.

  • Low levels of arrest, conviction and incarceration of serious criminals, measured at the state level, are associated with high capital error rates at the county level.

  • Highly and densely populated states tend to have higher county reversal rates in analyses of reversal rates at all three review stages combined and at the direct appeal stage.

  • In addition, the infrequency with which the passage of time is significantly related to declining reversal rates over time, despite the delay-related downward force on reversal rates in most analyses, plus the strong increase in reversal rates over time in Analysis 10, suggest that, after controlling for other factors, error rates increased over time, neutralizing the downward effect on reversal rates of delay.

The county-state analyses serve two purposes. First, they test the power of each important state-level factor as an explanation for county-level reversal rates—finding them to be important explanations of county, as well as state reversal rates. County-state Analyses 11 and 12 also test the power of the entire set of state-level factors to predict county-level reversal rates. These latter analyses find that the county reversal rates predicted by the best set of explanations for state reversal rates are highly significant predictors of actual county reversal rates.

As Table 9, below, shows, two county-level factors are also significantly related to county capital reversal rates with some degree of consistency:

  • In six of the 10 county analyses, high county capital-sentencing rates are significantly associated with high county capital error rates, and that relationship fell just above the .05 significance level in a seventh analysis. In each analysis, the contribution this factor made towards explaining county capital error rates was in addition to the contribution made in the same analysis by high state-level death-sentencing rates.

  • In four analyses, highly and densely populated counties have higher reversal rates.

    One other county factor was significant often enough to deserve mention:

  • In two analyses—both, however, analyzing only county factors—counties with high homicide rates had high capital reversal rates. As is noted above, this factor seems to lose power when the relative size of states' black population, and states' rates of white relative to black homicide victimization, are included as possible explanations for reversal rates.588

Table 9: County-Level Factors That Were Sometimes Significant Explanations for County Reversal Rates,
and How Often They Were Significant
589

Which county-level explanatory factor?

Significant in how many of 10 analyses?*

Significant in what types of analyses?

Death Verdicts/1000 Homicides

6 (+ 1 just above .05)

cnty, cnty-in-st, da, fl

Population Size and Density

4 (+ 1 just above .05)

cnty, cnty-in-st, fl, (da)

Homicide Rate

2

cnty, fl

-Death Verdicts Awaiting Review, 3 Stages+

2

cnty, fl

-Passage of Time (unreliable)^

2^

cnty, fl

 

G. One Implication and One Illustration of the Relationship Between High County Capital Error Rates and High County Capital-Sentencing Rates

1. An additional explanation for fairly uniform capital error rates among counties in the same state.

The finding that county death-sentencing rates sometimes predict capital reversal rates above and beyond the effect of state death-sentencing rates may help explain why there is relatively little variation among study counties within each state. Recall that study counties are only those using the death penalty at least once during the 23-year study period. As Figures 42A and 42B (pp.248-49 above) demonstrate, however, not all counties in capital states are death-sentencing counties. On the contrary, nearly 60% of the counties in the 34 study states imposed no death verdicts during the 23-year study period. If a major factor in explaining county-to-county differences in capital error rates is the counties' death-sentencing rates per 1000 homicides, then an even more important difference among counties in capital states may be between those that never use the death penalty and those that use it sometimes. Because our analyses examine error rates among imposed or reviewed death verdicts, they exclude counties with no death verdicts. The entire set of counties in capital states thus may be more diverse than a study of only death-sentencing counties reveals.590

To test this hypothesis, we conducted a simple analysis of whether there is more variability in the use of the death penalty among counties in the 34 death-sentencing states than one would expect by chance (i.e., as a matter of random variation), if one assumes that (1) death-sentencing rates vary among states, but (2) within a state, there is a constant probability that any homicide will lead to a death sentence. Given these assumptions, one would expect that 1325 of the 2341 counties in the 34 death-sentencing states would not have imposed a death verdict during the 23-year study period. In fact, the number of counties without a death verdict was 1339—very close to what one would expect under the assumption of variability among states but constancy within them. This result tends to confirm that states and not counties are the main source of capital-sentencing variation, and thus that it is appropriate to focus our main analysis (Analysis 1) on explaining state, not county, differences in capital reversal rates. We intend to pursue this question further in a subsequent phase of our work.

2. County comparisons illustrating the relationship between high death-sentencing rates and high rates of serious error.

What are the top death-sentencing counties in the U.S.? And how how often do they make serious capital mistakes? This section addresses these questions with a series of tables comparing the capital-error profiles of comparable counties with higher and lower death-sentencing rates.

At least 1004 counties in the United States imposed one or more death verdicts during the 1973-1995 study period.591 To identify top death-sentencing counties among those jurisdictions, we used two criteria—number of death verdicts, and rate of verdicts per 1000 homicides. Counties in the 34 active death-sentencing states imposed anywhere from 0 to 190 death verdicts during the study period, with most imposing none or just 1 or 2.592 We began our study of high death-sentencing counties by excluding all with fewer than five death verdicts, believing that it is not reasonable to identify such counties as heavy users of the death penalty, no matter what their death-sentencing rates may be.

Although the regression analyses used elsewhere in this Report can do so reliably, it is difficult in a simple comparison of counties like that in this section to meaningfully compare the death-sentencing and error rates of counties with fewer than five death verdicts. A county with only one homicide leading to one death verdict during the study period has the highest possible death-sentencing rate. But if the homicide was highly aggravated—on a par with crimes for which counties that only rarely use the death penalty would have imposed it—this hypothetical county is not really similar to another county with the same 100% death-sentencing rate that sentenced 10 people to death for 10 homicides of widely varying degrees of aggravation. Moreover, the former county has only two possible error rates for its one death verdict—0 or 100%—not because it is a perfectly reliable or perfectly unreliable death-sentencer but because its single death verdict allows only a small set of possible reversal rate outcomes.593 By contrast, the latter county, with 10 death verdicts, has 11 possible reversal rates—0, 10%, 20%, etc. Its reversal rate thus is a more sensitive measure of its actual death-sentencing reliability. By including only counties with five or more death verdicts, we partially avoid the difficulties involved in comparing death-sentencing rates and error rates in counties with very low numbers of homicides and death verdicts.

Among the 1004 death-sentencing counties, 244 (24%) imposed five or more death verdicts during the study period.594 We ranked these counties by their rates of verdicts per 1000 homicides during the period when the state in which the county is located had a valid capital statute.595 Tables 10A and 10B below list the 15 counties in the U.S. with 50 or more death verdicts during the relevant part of the study period. Table 10A lists those counties in order of their number of death verdicts. Table 10B lists the same counties in order of their death-sentencing rates. The purpose of the different organizing principles is illustrated by the first two entries in Table 10A. Harris County (Houston), Texas imposed 190 death verdicts during the relevant period, compared to the next closest county, Los Angeles, which imposed 150. But because Houston had many fewer homicides than Los Angeles in the relevant period, the difference between them in death-sentencing rates—19 death verdicts per1000 homicides for Houston, versus only 8 per 1000 for Los Angeles (2.4 to 1)—is much greater than the raw number of verdicts would suggest (190 vs. 150, or only 1.27 to 1).

As is further developed below,596 Tables 10A and 10B reveal wide disparity in death-sentencing rates. For example, among cities in counties with 50 or more death verdicts, Pima County (Tucson), Arizona had the highest rate of death verdicts per 1000 homicides: 64. Pima County homicides were:

  • 21 times more likely to result in a death verdict than homicides in St. Louis, Missouri;

  • almost six times more likely to be punished by death than homicides in Dallas, Texas;

  • and from 5 to 13 times more likely to be punished capitally than homicides in Charlotte, North Carolina (14/1000), Austin, Texas (10/1000) and Richmond, Virginia (5/100)—all of which had about the same number of homicides as Tucson during the relevant period.

When the data are rearranged in order of death-sentencing rates, not numbers, an important pattern appears: Among counties that imposed at least 50 death verdicts during the study period, those imposing more death verdicts per 1000 homicides had appreciably higher rates of serious capital error—and condemned to die much larger proportions of people later shown to be factually or legally innocent—than counties with lower death-sentencing rates:

  • On average, counties with more than 30 death verdicts per 1000 homicides had combined capital reversal rates at the direct appeal and federal habeas stages597 that were 43% higher than the reversal rates of counties with 30 or fewer death verdicts per 1000 homicides. All six highest death-sentencing counties had overall capital reversal rates at the two review stages of 64% or more; five of the six had overall error rates over 70%.

  • Persons sentenced to die by the higher death-sentencing counties were about 67% more likely to be proven innocent thereafter than those sentenced to die by the lower death-sentencing counties.

  • The six highest death-sentencing counties account for about 11% of all demonstrably innocent people sentenced to die but are only about six-tenths of 1% of all death-sentencing counties, and account for only 7% of all death verdicts imposed in the study period.

Table 10A. The 15 Counties With 50 or More Death Verdicts, 1973-1995+

County (City), State

Death Verdicts

Homicides

Death Verdicts /
1000 Homicides

Error Rate†

# Not Guilty

% Not Guilty

Harris (Houston), TX

190

9,829

19

32%

2

1.1

Los Angeles, CA

150

17,998

8

37%

0

0

Cook (Chicago), IL

138

12,586

11

57%

8

5.8

Philadelphia, PA

127

4,698

27

25%

0

0

Maricopa (Phoenix), AZ

114

2,782

41

84%

5

4.4

Dade (Miami), FL

103

6,936

15

67%

1

.9

Clark (Las Vegas), NV

71

1,288

55

64%

2

2.8

Oklahoma (City), OK

68

1,361

50

75%

3

4.4

Hillsborough (Tampa), FL

67

1,839

36

72%

0

0

Duval (Jacksonville), FL

66

2,232

30

51%

0

0

Pima (Tucson), AZ

63

986

64

71%

1

1.6

Dallas, TX

61

5,682

11

67%

1

1.6

Jefferson (Birmingham), AL

55

2,161

25

55%

0

0

Broward (Ft. Lauderdale), FL

55

2,599

21

84%

2

3.6

Pinellas (St. Petersburg), FL

51

1,018

50

89%

0

0

All 15 Counties

1379

73,995

avg. 4,933

19

62%

(avg.)

25

1.8

_______________

+ Death verdicts, homicides and death-sentencing rates ((death verdicts/homicides) x 1000) are those occurring during the portion of the 1973-1995 period when the state in which the county is located had a valid post-Furman capital statute. See supra note 595.

† Error rates are the overall capital reversal rates at the state direct appeal and federal habeas stages. See supra n.597.

Sources: DRCen, DADB, HCDB, Vital Statistics.

Table 10B. The 15 Counties With 50 or More Death Verdicts, 1973-1995:+
High vs. Low Death-Sentencing Counties

County (City), State

Death Verdicts

Homicides

Death Verdicts / 1000 Homicides

Error Rate†

# Not Guilty

% Not Guilty

Pima (Tucson), AZ

63

986

64

71%

1

1.6

Clark (Las Vegas), NV

71

1,288

55

64%

2

2.8

Pinellas (St. Petersburg), FL

51

1,018

50

89%

0

0

Oklahoma (City), OK

68

1,361

50

75%

3

4.4

Maricopa (Phoenix), AZ

114

2,782

41

84%

5

4.4

Hillsborough (Tampa), FL

67

1,839

36

72%

0

0

All 6 Counties

434

9,274

47

avg:76%

11

2.5

-VS.-

Duval (Jacksonville), FL

66

2,232

30

51%

0

0

Philadelphia, PA

127

4,698

27

25%

0

0

Jefferson (Birmingham), AL

55

2,161

25

55%

0

0

Broward (Ft. Lauderdale), FL

55

2,599

21

84%

2

3.6

Harris (Houston), TX

190

9,829

19

32%

2

1.1

Dade (Miami), FL

103

6,936

15

67%

1

.9

Cook (Chicago), IL

138

12,586

11

57%

8

5.8

Dallas, TX

61

5,682

11

67%

1

1.6

Los Angeles, CA

150

17,998

8

37%

0

0

All 9 Counties

945

64,721

15

avg: 53%

14

1.5

_______________

+ Death verdicts, homicides and death-sentencing rates ((death verdicts/homicides) x 1000) are those occurring during the portion of the 1973-1995 period when the state in which the county is located had a valid post-Furman capital statute. See supra note 595.

† Error rates are the overall capital reversal rates at the state direct appeal and federal habeas stages. See supra n.597.

Sources: DRCen, DADB, HCDB, Vital Statistics.

Tables 11A and 11B in Appendix B list the 244 counties with at least five death verdicts in order of their death-sentencing rates. Table 11A lists the counties in the order of their ranking from 1st to 244th most frequent users of the death penalty. To assist readers in locating counties, Table 11B groups the same counties alphabetically by state and by counties within states, with a notation of each county's death-sentencing ranking from 1st to 244th. For comparative purposes Table 12, Appendix B, lists all 1004 counties with at least one death verdict during the study period, alphabetically by state and county. The 244 counties with five or more death sentences that are ranked by death-sentencing rates in Table 11A reflect the same pattern as in Table 10:

  • Among counties in the top quarter in terms of their death-sentencing rates, 23% had direct appeal/federal habeas error rates of 90% or more. One-fifth of these highest death-sentencing counties had reversal rates of 100%.

  • Among counties in the bottom quarter in terms of their capital-sentencing rates, only 10% had error rates of 90% or more.

We next examined three sets of counties that are similar to each other in terms of the number of homicides each had to select from in deciding when to use the death penalty during the study period.598 As above, we divided each set of comparable counties into two groups—those with relatively high, and those with relatively low, death-sentencing rates. We then compared the capital error rates of the two groups of counties. All three comparisons strongly confirm the pattern noted above: High death-sentencing counties have high capital error rates.

Among the top third of capital counties599 based on their death-sentencing rates are four with the most homicides: Pima (Tucson), Arizona; Clark (Las Vegas), Nevada; Pinellas (St. Petersburg), Florida; and Oklahoma (City), Oklahoma. These counties had from 986 to 1381 homicides during the study period. Among the bottom third of capital counties given their death-sentencing rates are 12 counties with the same range of (950 to 1400) homicides.600 Table 13A, p. 294 below, lists the death-sentencing and capital error rates for the four high death-sentencing counties. Table 13B, p. 295, has the same information for the 12 low death-sentencing counties. The two groups of counties have very similar homicide profiles. The high-sentencing counties averaged 1163 homicides; the low-sentencing counties averaged 1121 homicides. But the high death-sentencing counties imposed over five times more death verdicts per homicide than the low death-sentencing counties: 54 verdicts for every 1000 homicides for the high death-sentencing counties, compared to 10 verdicts per 1000 homicides for the low-sentencing counties. These large differences in per-homicide use of the death penalty correspond to large differences in capital error rates:

  • The average capital error rate for the high death-sentencing counties is 67% higher than the average capital error rate for the low death-sentencing counties.

  • Three of the four high-death sentencing counties have sentenced at least one innocent person to die—two have done so more than once. None of the 12 comparable low death-sentencing counties have been found to have done so.

One can also compare the top four death-sentencing counties, in Table 13A, to the bottom four death-sentencing counties in Table 13B. In this more compact comparison, see Table 13C, p. 295 below, the two subsets of counties are even more closely matched. The four highest death-sentencing counties in the group—encompassing Tucson, Las Vegas, St. Petersburg, and Oklahoma City—averaged 1163 homicides per county. The four lowest death-sentencing counties in the group—Little Rock, Arkansas; Nashville, Tennessee; Prince George's County (suburban Washington, D.C.), Maryland; and Richmond, Virginia—averaged 1156 homicides per county. Despite similar homicide profiles, the two subsets of four counties have very different capital-sentencing profiles: The four top death-sentencing counties in the group imposed nine times more death verdicts per 1000 homicides than the bottom four counties (54 vs. 6). Associated with these drastically different capital-sentencing profiles are, again, drastically different capital-error profiles:

  • The top four counties in terms of their death-sentencing rates had an average capital error rate nearly twice as high as the bottom four counties (75% vs. 39%).

  • Three of the top four death-sentencing counties condemned innocent people to death. None of the bottom four counties did so.

Table 13A. Capital Error Rates in Top-Third Death-Sentencing Counties*
With Highest Number of (986-1361) Homicides, 1973-1995+

County (City), State

Death Verdicts / 1000 Homicides

Homicides

Death Verdicts

Error Rate†

# Not Guilty

% Not Guilty

Pima (Tucson), AZ

64

986

63

71%

1

1.6

Clark (Las Vegas), NV

55

1,288

71

64%

2

2.8

Pinellas (St. Petersburg), FL

50

1,018

51

89%

0

0

Oklahoma (City), OK

50

1,361

68

75%

3

4.4

All 4 Counties

54

4,653

avg. 1,163

253

avg. 63

75%(avg)

6

2.4

 

Notes to Tables 13A-C

* Top-third counties are the 81 counties, among the 244 counties with five or more death verdicts, that have the highest rates of death verdicts to homicides. Bottom-third counties are the 81 counties, among the 244 counties with five or more death verdicts, that have the lowest rates of death verdicts to homicides.

+ Death verdicts, homicides and death-sentencing rates ((death verdicts/homicides) x 1000) are those occurring during the portion of the 1973-1995 period when the state in which the county is located had a valid post-Furman capital statute. See supra note 595.

† Error rates are the overall capital reversal rates at the state direct appeal and federal habeas stages. See supra note 597.

Sources: DRCen, DADB, HCDB, Vital Statistics.

Table 13B. Capital Error Rates in Bottom-Third Death-Sentencing Counties*
with Comparable Number of (950-1400) Homicides, 1973-1995+

County (City), State

Death Verdicts / 1000 Homicides

Homicides

Death Verdicts

Error Rate†

# Not Guilty

% Not Guilty

DeKalb (sub. Atlanta), GA

17

1,065

18

100%

0

0

Fresno, CA

14

1,256

18

40%

0

0

Mecklenburg (Charlotte), NC

14

1,013

14

64%

0

0

Santa Clara (San Jose), CA

13

1,161

15

22%

0

0

Jefferson (Louisville), KY

12

1,201

15

53%

0

0

Allegheny (Pittsburgh), PA

12

1,145

14

64%

0

0

Travis (Austin), TX

10

975

10

44%

0

0

Contra Costa, CA

9

1,015

9

0%

0

0

Pulaski (Little Rock), AR

7

1,157

8

60%

0

0

Davidson (Nashville), TN

6

1,323

8

29%

0

0

Prince George=s (sub. Washington), MD

6

1,074

6

50%

0

0

Richmond, VA

5

1,071

5

17%

0

0

All 12 Counties

10

13,456

avg. 1,121

140

avg. 12

45%

(avg)

0

0

 

Table 13C. Bottom-Four Death-Sentencing Counties*+

 

 

County (City), State

Death Verdicts / 1000 Homicides

Homicides

Death Verdicts

Error Rate†

# Not Guilty

% Not Guilty

Pulaski (Little Rock), AR

7

1,157

8

60%

0

0

Davidson (Nashville), TN

6

1,323

8

29%

0

0

Prince George=s (sub. Washington), MD

6

1,074

6

50%

0

0

Richmond, VA

5

1,071

5

17%

0

0

All 4 Counties

6

4,625

avg. 1,156

27

avg.7

39%

(avg)

0

0

 

The other identifiable groups of top-third and bottom-third death-sentencing counties that can be compared reliably are counties with 200 to 700 homicides. The eight counties with homicides in this range that are also in the top third of all counties based on death-sentencing rates are listed in Table 14A, p. 297. They averaged 409 homicides in the study period, and about 55 death verdicts per 1000 homicides. Figure 14B, pp. 298-99, lists the 36 counties with 200-700 homicides that are in the bottom third of counties based on death sentencing rates. The latter counties averaged 438 homicides but only about 15 death verdicts per 1000 homicides. Again, despite similar homicide profiles, the high and low death-sentencing counties have starkly different error profiles:

  • Capital error rates are about 63% higher in the high death-sentencing counties than in the comparable subset of low death-sentencing counties.

  • Half of the eight high death-sentencing counties sentenced someone to die who later proved to be innocent. Only 6% (2 out of 36) low death-sentencing counties did so.

  • People sentenced to die in the high death-sentencing counties were 3 1/2 times more likely to be shown to be innocent than those sentenced in the low death-sentencing counties.

We can also compare the eight top death-sentencing counties with 200 to 700 homicides to the bottom eight death-sentencing counties in the same group. The top eight, in Table 14A, averaged 409 homicides during the study period. The bottom eight, Table 14C, p. 299, averaged 566 homicides. The top eight death-sentencing counties imposed nearly 6 times more death verdicts per 1000 homicides than the bottom eight counties (55 vs 10), and had correspondingly greater tendencies towards capital error:

  • The top eight death-sentencing counties in the group had an average capital error rate nearly 80% higher than the bottom eight counties (78% vs. 44%).

  • Four of the top eight death-sentencing counties sentenced innocent people to die at least once. None of the bottom eight counties did.

Table 14A. Capital Error Rates in Top-Third Death-Sentencing Counties*
With Next Highest Number of (238-612) Homicides, 1973-1995+

 

County (City), State

Death Verdicts 1000 Homicides

Homicides

Death Verdicts

Error Rate†

# Not Guilty

% Not Guilty

Pasco (sub. Tampa-St. Petersburg), FL

72

279

20

100%

2

10

Robeson (Lumberton), NC

62

340

21

76%

0

0

Baltimore County (suburbs), MD

56

612

34

100%

1

2.9

Bay (Panama City), FL

55

238

13

83%

0

0

Escambia (Pensacola), FL

55

513

28

87%

1

3.6

Horry (Myrtle Beach), SC

54

261

14

82%

0

0

Brevard (Melbourne), FL

50

482

24

54%

1

4.2

Volusia (Daytona Beach), FL

49

546

27

44%

0

0

All 8 Counties

55

3,271

avg. 409

181

avg. 23

78%

(avg.)

5

2.8

Notes to Tables 14A-C

* Bottom-third counties are the 81 counties, among the 244 counties with five or more death verdicts, that have the lowest rates of death verdicts to homicides. Top-third counties are the 81 counties, among the 224 counties with five or more death verdicts, that have the highest rates of death verdicts to homicides.

+ Death verdicts, homicides and death-sentencing rates ((death verdicts/homicides) x 1000) are those occurring during the portion of the 1973-1995 period when the state in which the county is located had a valid post-Furman capital statute. See supra note 595.

† Error rates are the overall capital reversal rates at the state direct appeal and federal habeas stages. See supra note 597.

Sources: DRCen, DADB, HCDB, Vital Statistics.

Table 14B. Capital Error Rates in Bottom-Third Death-Sentencing Counties*
with Comparable Number of (200-700) Homicides, 1973-1995+

County (City), State

Death Verdicts / 1000 Homicides

Homcides

Death Verdicts

Error Rate†

# Not Guilty

% Not Guilty

Lauderdale, MS

20

246

5

80%

0

0

Lucas (Toledo), OH

20

498

10

17%

0

0

Lubbock, TX

20

609

12

60%

0

0

Buncombe (Asheville), NC

19

259

5

50%

0

0

Lafayette, LA

19

265

5

25%

0

0

Jefferson (Pine Bluff), AR

18

327

6

100%

0

0

Ventura, CA

18

545

10

14%

0

0

Brazoria, TX

18

273

5

33%

0

0

Cumberland (Fayetteville), NC

18

602

11

63%

1

9

Calcasieu (Lake Charles), LA

18

330

6

100%

0

0

Knox (Knoxville), TN

18

499

9

100%

0

0

Clayton (suburban Atlanta), GA

18

279

5

80%

0

0

Seminole (Orlando), FL

18

335

6

33%

1

17

Virginia Beach, VA

18

335

6

0%

0

0

St. Lucie, FL

18

395

7

71%

0

0

Wichita (Falls), TX

17

287

5

80%

0

0

Santa Barbara, CA

17

287

5

0%

0

0

Douglas (Omaha), NE

17

658

11

68%

0

0

Franklin (Columbus), OH

16

497

8

17%

0

0

Fayette (Lexington), KY

16

315

5

40%

0

0

Tulare, CA

16

515

8

25%

0

0

Bell (Killeen), TX

15

388

6

67%

0

0


Table 14B (cont'd). Capital Error Rates in Bottom-Third Death-Sentencing Counties*
with Comparable Number of (200-700) Homicides, 1973-1995+

County (City), State

Death Verdicts / 1000 Homicides

Homicides

Death Verdicts

Error Rate†

# Not Guilty

% Not Guilty

Gregg (Longview), TX

14

348

5

75%

0

0

Bibb (Macon), GA

13

595

8

56%

0

0

Fairfax (sub. Washington), VA

13

376

5

14%

0

0

Hidalgo (McAllen), TX

12

409

5

50%

0

0

Delaware (sub. Philadelphia), PA

12

491

6

0%

0

0

Greenville, SC

11

555

6

40%

0

0

Camden, NJ

11

559

6

100%

0

0

Guilford, NC

11

564

6

60%

0

0

Galveston, TX

11

664

7

44%

0

0

Richland (Columbia), SC

9

634

6

40%

0

0

Salt Lake, UT

8

655

5

20%

0

0

All 36 Counties

15

15,782

avg. 438

239

avg. 6.6

48%

(avg.)

2

.8

 

Table 14C. Bottom-Eight Death-Sentencing Counties*+

County (City), State

Death Verdicts / 1000 Homicides

Homicides

Death Verdicts

Error Rate

# Not Guilty

% Not Guilty

Hidalgo (McAllen), TX

12

409

5

50%

0

0

Delaware (sub. Philadelphia), PA

12

491

6

0%

0

0

Greenville, SC

11

555

6

40%

0

0

Camden, NJ

11

559

6

100%

0

0

Guilford, NC

11

564

6

60%

0

0

Galveston, TX

11

664

7

44%

0

0

Richland (Columbia), SC

9

634

6

40%

0

0

Salt Lake, UT

8

655

5

20%

0

0

All 8 Counties

10

4,531

avg. 566

47

avg. 6

44%

(avg.)

0

0

 

As a last comparison of this type, we considered all 16 capital counties with 1500 to 3000 homicides during the study period. Table 15, p. 301, compares the seven counties in the group with death-sentencing rates of 20 or more death verdicts per 1000 homicides to the nine counties with death-sentencing rates below 20 per 1000 homicides. (The death-sentencing rates in the former group ranged from 20 to 41 verdicts per 1000 homicides, with an aggregate rate of 28 per 1000 homicides; the corresponding rates in the latter group ranged from 3 to 16 death sentences per 1000 homicides, with an aggregate rate of 11. The former group of counties averaged 2201 homicides per year; the latter group averaged 2075 homicides per year.) Again, the high death-sentencing counties in the group have higher capital error and innocence rates than the low death-sentencing counties.

Table 15. Capital Error Rates in Counties with 1500-3000 Homicides, 1973-1995:
Low Versus High Death-Sentencing Counties*

County (City), State

Death Verdicts / 1000 Homicides

Homicides

Death Verdicts

Error Rate†

# Not Guilty

% Not Guilty

Maricopa (Phoenix), AZ

41

2,782

114

84%

5

4.4

Hillsborough (Tampa), FL

36

1,839

67

72%

2

5.6

Duval (Jacksonville), FL

30

2,232

66

51%

0

0

Jefferson (Birmingham), AL

25

2,161

55

55%

0

0

Cuyohoga (Cleveland), OH

22

2,053

45

24%

1

2.2

Broward (Ft. Lauderdale), FL

21

2,599

55

84%

3

4.5

Orange, CA

20

1,738

35

100%

0

0

All 7 Counties

28

15,404

avg. 2,201

437

avg. 62

67%

(avg.)

11

2.5

-VS.-

Tarrant (Fort Worth), TX

16

2,636

42

54%

0

0

Alameda (Oakland), CA

15

2,010

31

0%

0

0

San Bernadino, CA

15

1,950

30

100%

0

0

Lake, IN

15

1,500

23

100%

1

4.5

Shelby (Memphis), TN

14

2,219

32

23%

0

0

San Diego, CA

10

2,322

23

60%

0

0

Jackson (Kansas City), MO

6

1,827

11

33%

0

0

Essex, NJ (Jersey City)

4

1,905

8

88%

0

0

St. Louis City, MO

3

2,306

7

0%

0

0

All 9 Counties

11

18,675

avg. 2,075

207

avg. 23

51%

(avg.)

1

.5


* Only counties with five or more death sentences. Sources: DRCen, DADB, HCDB, Vital Statistics.

+ Death verdicts, homicides and death-sentencing rates ((death verdicts/homicides) x 1000) are those occurring during the portion of the 1973-1995 period when the state in which the county is located had a valid post-Furman capital statute. See supra note 595.

†Error rates are the overall capital reversal rates at the state direct appeal and federal habeas stages. See supra note 597.

Finally, Table 16, p. 304 below, compares the 73 counties with 600 or more homicides during the relevant portion of the study period, ranked by their overall capital error rates at the direct appeal and federal habeas stages.601 Death-sentencing rates are included so that reversal and death-sentencing rates may be compared. The first page of Table 16 lists the 37 counties with the highest reversal rates; the second page lists the 36 counties with the lowest reversal rates. Table 16 again reveals a strong association between high capital-error rates and high capital-sentencing rates:

  • Among these counties, Pima County (Tucson), Arizona has the highest death-sentencing rate: 64 death sentences per 1000 homicides. The City of St. Louis has the lowest rate: 3 death sentences per 1000 homicides.

  • The 37 counties with relatively high error rates include 91% of the counties that imposed more than 30 verdicts per 1000 homicides. The 36 counties with relatively low error rates include 76% of the counties with fewer than 10 verdicts for every 1000 homicides.

  • Five counties in Table 16 have death-sentencing rates of 50 or more per 1000 homicides during the study period, all with 34 or more death sentences and 600 or more homicides. These five top death-sentencing counties had overall capital reversal rates at the direct appeal and federal habeas stages of anywhere from 64% on the low end to 100% on the high end, with an average reversal rate of 80% at those stages.602 All but one of the counties put men on death row who later proved to be not guilty. The five counties are:

    • Pima County (Tucson), Arizona, with a death-sentencing rate per 1000 homicides of 64 and an overall direct appeal and federal habeas reversal rate in the study period of 71%; 1 death row prisoner later found not guilty.

    • Suburban Baltimore County, Maryland, with a death-sentencing rate of 56 and an overall reversal rate of 100%; 1 death row prisoner later found not guilty.

    • Clark County (Las Vegas), Nevada, with a death-sentencing rate of 55 and an overall reversal rate of 64%; 2 death row prisoners later found not guilty.

    • Pinellas County (St. Petersburg), Florida, with a death-sentencing rate of 50 and an overall reversal rate of 89%.

    • Oklahoma County (Oklahoma City), with a death-sentencing rate of 50 and an overall reversal rate of 75%; 3 death row prisoners later found not guilty.

  • Five counties in Table 16 have the lowest death-sentencing rates: San Francisco, California; Richmond, Virginia; Fulton County (Atlanta), Georgia; Essex County (Newark), New Jersey; and St. Louis City, Missouri. These five lowest death-sentencing counties had an average error rate of only 43% at the first and last review stages. None sentenced anyone to death who has since been found not guilty.

  • The 10 top death-sentencing counties with 600 or more homicides during the study period—in order, Pima County (Tucson), Arizona; suburban Baltimore County, Maryland; Clark County (Las Vegas), Nevada; Pinellas County (St. Petersburg), Florida; Oklahoma (City), Oklahoma; Maricopa County (Phoenix), Arizona; Hamilton County (Cincinnati), Ohio; Hillsborough County (Tampa), Florida; Polk County, Florida; and Muscogee County (Columbus), Georgia—have an average error rate at the direct appeal and habeas stages of 71%. 8 of the 10 counties put a total of 16 people on death row during the study period who were later found innocent.

    • In contrast, the 10 lowest death-sentencing counties with 600 or more homicides—San Francisco, California; Richmond, Virginia; Fulton County (Atlanta), Georgia; Essex County (Newark), New Jersey; St. Louis City, Missouri; Pulaski County (Little Rock), Arkansas; Bernalillo County (Albuquerque), New Mexico; Davidson County (Nashville), Tennessee; Jackson County (Kansas City), Missouri; and Prince George's County (suburban Washington), Maryland—had an average error rate at the direct appeal and habeas stages of 41%, and none sentenced anyone to death during the study period who was later found not guilty.

Table 16: Overall Error Rates and Death-Sentencing Rates
for All Counties With 600 or More Homicides, 1973-1995*

County (City), State

Overall Reversal Rate
(Dir. App. + Fed. Hab. Stage)

Death-Sentencing Rate
(for every 1000 homicides)

Homicides

Baltimore County (suburbs), MD

100%

56

612

Orange, CA

100%

20

1738

De Kalb (suburban Atlanta), GA

100%

17

1065

Tulsa, OK

100%

16

794

San Bernardino, CA

100%

15

1950

Lake, IN

100%

15

1500

Richmond (Augusta), GA

100%

10

705

Pinellas (St. Petersburg), FL

89%

50

1018

Multnomah (Portland), OR

88%

13

760

Essex (Newark), NJ

88%

4

1905

Chatham (Savannah), GA

85%

22

787

Maricopa (Phoenix), AZ

84%

41

2782

Broward (Ft. Lauderdale), FL

84%

21

2599

Hinds (Jackson), MS

81%

24

907

Polk, FL

78%

35

894

Oklahoma (City), OK

75%

50

1361

El Paso, TX

73%

18

734

Orleans (New Orleans), LA

73%

9

3126

Hillsborough (Tampa), FL

72%

36

1839

Fulton (Atlanta), GA

71%

4

3314

Pima (Tucson), AZ

71%

64

986

Orange (Orlando), FL

71%

32

1241

Douglas (Omaha), NE

68%

17

658

Dade (Miami), FL

67%

15

6936

Dallas, TX

67%

11

5682

East Baton Rouge, LA

67%

11

857

Muscogee (Columbus), GA

66%

33

607

Clark (Las Vegas), NV

64%

55

1288

Meckleburg (Charlotte), NC

64%

14

1013

Allegheny (Pittsburgh), PA

64%

12

1145

Cumberland (Fayetteville), NC

63%

18

602

Lubbock, TX

60%

20

609

San Diego, CA

60%

10

2322

Pulaski (Little Rock), AR

60%

7

1157

Cook (Chicago), IL

57%

11

12586

Jefferson, LA

56%

16

869

Mobile, AL

56%

28

1298

-VS.-

County (City), State

Overall Reversal Rate
(Dir. App. + Fed. Hab. Stage)

Death-Sentencing Rate
(for every 1000 homicides)

Homicides

Jefferson (Birmingham), AL

55%

25

2161

Tarrant (Ft. Worth), TX

54%

16

2636

Jefferson (Louisville), KY

53%

12

1201

Duval (Jacksonville), FL

51%

30

2232

Palm Beach, FL

50%

12

1461

St. Clair (Belleville), IL

50%

7

945

Prince George's (sub. Wash.), MD

50%

6

1074

Bexar (San Antonio), TX

48%

13

3275

Travis (Austin), TX

44%

10

975

Galveston, TX

44%

11

664

Fresno, CA

40%

14

1256

Richland (Columbia), SC

40%

9

634

San Francisco, CA

40%

5

1444

Los Angeles, CA

37%

8

17998

St. Louis County (suburbs), MO

37%

26

1387

Kern (Bakersfield), CA

36%

23

961

Jefferson (Beaumont), TX

35%

26

685

Jackson (Kansas City), MO

33%

6

1827

Harris (Houston), TX

32%

19

9829

Riverside, CA

31%

18

1477

Sacramento, CA

29%

22

1329

Davidson (Nashville), TN

29%

6

1323

Philadelphia, PA

25%

27

4698

Cuyahoga (Cleveland), OH

24%

22

2053

Shelby (Memphis), TN

23%

14

2219

Santa Clara (San Jose), CA

22%

13

1161

Nueces (Corpus Christi), TX

20%

20

770

Salt Lake, UT

20%

8

655

Bernalillo (Albuquerque), NM

20%

6

814

Marion (Indianapolis), IN

18%

10

1433

San Joaquin (Stockton), CA

17%

12

769

Richmond, VA

17%

5

1071

Hamilton (Cincinnati), OH

8%

40

727

Alameda (Oakland), CA

0%

15

2010

Contra Costa, CA

0%

9

1015

St Louis (City), MO

0%

3

2306.00


* Includes only counties with five or more death verdicts during the study period. Sources: DRCen, DADB, HCDB, Vital Statistics.

Tables 10-16 provide strong support for two conclusions. The first tracks consistent findings of the regression analyses discussed above. The second follows logically from the first:

  • Aggressive death sentencing is a recipe for high rates of reversible error that undermine the reliability of capital verdicts and the death penalty's effectiveness as a law enforcement tool.603

  • Jurisdictions with high capital-sentencing, and thus typically high capital-error, rates also have a dangerous propensity to convict and condemn the innocent.
next | previous | table of contents