Abstract:

This study tested the deterrence hypothesis in Texas, the most active execution jurisdication during the modern era.
 

Full Text:

This study tested the deterrence hypothesis in Texas, the most active execution jurisdiction during the modern era. Using monthly observations during 1984 through 1997, both the general relationship between executions and murder rates and the specific relationship between executions and felony murder rates were examined. An initial bivariate relationship between executions and murder rates proved to be spurious when appropriate control variables were included in regression models. Within a context so ideally suited for finding any potential deterrent effects, this study confirmed the results of previous ones that failed to find any evidence of deterrence resulting from capital punishment.

The dominant approach to determining the relationship between deterrence and the death penalty involves comparing the rate of homicide or some subset of homicide and either the legal status of the death penalty or the performance of actual executions within or across particular jurisdictions. The deterrence hypothesis is supported when lower homicide rates are found within time periods or jurisdictions where the death penalty has been available or in use. If homicide rates are higher in the presence of capital punishment, the alternative, or "brutalization hypothesis," is supported (Bowers and
Pierce 1980). The third possible outcome is that the death penalty is found to have no influence on homicide rates.

Empirical studies of deterrence and capital punishment are best classified by their research design. Cross-sectional designs compare homicide rates
across jurisdictions. The earliest deterrence studies of this kind simply compared rates of homicide in retentionist states that have statutory provisions
for the death penalty to the rates in abolitionist states without such provisions. Findings showed that retentionist states typically experienced
higher rates of homicide than did abolitionist jurisdictions (Sutherland 1925). However, examination of geographical, social, and economic dissimilarities
between abolitionist and retentionist states suggested that factors other than the death penalty could have influenced homicide rates.

Scholars then began making comparisons that are more specific between neighboring states, which were presumed to be comparable on such factors. These studies failed to support the deterrence hypothesis, finding that retentionist states most often experienced higher rates of homicide than did contiguous abolitionist states (Schuessler 1952; Sellin 1967). A new generation of cross-sectional studies has employed multiple regression analyses to predict the rate of homicide across jurisdictions while controlling for extraneous variables (Forst 1977; Passell 1975) and has consistently found executions to have no effect on murder rates (Cheatwood 1993; Peterson and Bailey 1988).

Longitudinal designs are used to study the influence of the death penalty in a single jurisdiction over time. The earliest of these studies examined
homicide rates in jurisdictions before and after a legislative change in the legal status of the death penalty that either abolished or reimplemented
capital punishment. The deterrence hypothesis would be supported if states experienced lower rates of homicide during retentionist periods and higher
rates of homicide during abolitionist periods. These studies failed to support the deterrence hypothesis (Bedau 1967; Sellin 1967) because they
found inconsistent changes in murder rates after legislative enactments.

More recently, the advent of sophisticated statistical techniques has influenced the methodology used in testing the deterrence hypothesis. Time series
analysis, introduced by economist Isaac Ehrlich in 1975, has proven to be a superior means of testing the deterrent effect of the death penalty
over time. Time series analyses typically concentrate on the effect of actual executions and enable the researcher to simultaneously control for
the influence of alternative explanatory variables. In his initial study, Ehrlich claimed that executions carried out during 1933 through 1969 had
resulted in a significant reduction in the number of homicides occurring throughout the United States. However, reanalyses of his work failed to
find support for the deterrence hypothesis; instead, researchers concluded that the reduction in homicides observed by Ehrlich was an artifact of
measurement error that resulted from inappropriate design specifications and faulty statistical analysis (Baldus and Cole 1975; Bowers and Pierce
1975; Klein, Forst, and Filatov 1978). Recent time series analyses have confirmed the findings of Ehrlich's critics; they have failed to find evidence
of a deterrent effect (Bailey 1983; Cochran, Chamlin, and Seth 1994; Decker and Kohfeld 1984).

Another recent advance in methodology has been to limit analyses to only those types of murder likely to be deterred by capital punishment. Because
only certain instances of murder can result in the death penalty, many researchers have disaggregated the universe of homicides to limit the dependent
variable to those that have death as a possible sentence. For example, Bailey and Peterson (1987) found that the likelihood of receiving a death
sentence was not related to the killing of police officers in the United States during 1973 through 1984. The findings supported those of earlier
studies that have failed to find a relationship between capital punishment and police killing (Cardarelli 1968; Sellin 1980). One limitation of these
studies was that the researchers were unable to consider the certainty of punishment because very few executions had been carried out during that
period.

In a later study that included a measure of the certainty of punishment, Peterson and Bailey (1991) analyzed the relationship between actual executions
and the monthly rates of felony murder throughout the United States from 1976 through 1987. The researchers found no consistent relationship between the number of executions, the level of television publicity of these executions, and the rate of felony murder. A study following Oklahoma's return to capital punishment disaggregated homicides into felony murders and murders involving strangers (Cochran et al. 1994). Using an interrupted time-series design, Cochran and colleagues found no change in the rate of felony homicides over the 68 weeks following this highly publicized execution, but observed an increase in the rate of stranger homicides. This brutalization effect was recorded in another study that found an increase in several types of homicide in metropolitan areas after Arizona's first execution in 29 years (Thompson 1997).

After thoroughly reviewing the empirical literature, Peterson and Bailey (1998) concluded that the lack of evidence for any deterrent effect of
capital punishment was incontrovertible. According to them, no credible empirical studies had ever been able to demonstrate that the severity,
certainty, or celerity of capital punishment reduced the rate of homicide. However, they did envision situations that might present unique opportunities
to engage the deterrence hypothesis.

One such opportunity presented itself in Texas in recent years. By far the most active death penalty state, Texas has accounted for more than
a third of all executions in the United States since the reimplementation of capital punishment in the years following Furman v. Georgia (1972).
In 1997 alone, Texas executed a record number of 37 capital murderers, accounting for half of the 74 U.S. executions in that year. Texas has provided
an ideal natural experiment to engage the deterrence hypothesis.

One study of the effect of capital punishment on homicide rates in Texas from 1933 through 1980 found no support for the deterrence hypothesis (Decker and Kohfeld 1990). Although this study did not include the effects of any post-Furman executions in Texas, an update of their research extended the period studied through 1986. Decker and Kohfeld then found that executions were actually followed by an increase in homicide rates, supporting the brutalization hypothesis. Their studies, however, included a limited number of control variables, an aggregate measure of homicide, and the use of years as the unit of analysis. Their updated study captured few of the executions that were to occur in the post-Furman era. The study reported here advances the work of Decker and Kohfeld by examining the deterrence hypothesis in Texas from 1984 through 1997, capturing the most active period of executions in a jurisdiction during the post-Furman period. It also simultaneously incorporates the methodological strengths of recent studies to provide one of the most compelling tests of the deterrence hypothesis completed thus far.

DATA AND METHODS

To examine the deterrence hypothesis during the modern era, data that spanned the years from 1984 through 1997 were collected from official sources.
The year 1984 was chosen as the beginning of the data collection period because of the availability of specific data on homicides and the onset
of executions.' Because no executions took place until December 1982, the period before the onset of executions was eliminated from our analyses
due to a lack of variance in the independent variable. Data collection was further limited as a result of the Houston Police Department's failure
to report information on homicides in 1983 for inclusion in the Supplemental Homicide Reports (SHR).2 Estimating the murder rate for 1983 would, to
some unknown degree, bias measures of the deterrent effect of the lone execution in December 1982, particularly because no further executions
were carried out until March 1984.3 Because of these potential sources of bias, data collection began with the year 1984.

The number of executions served as the independent variable. The number of executions was tabulated from ledgers provided by the Texas Department of Criminal Justice-Institutional Division. The dependent variables included rates of murder and rates of felony murder. Information on the number of murders was collected from the Texas Department of Public Safety-Uniform Crime Reporting Division. The murder rate was based on the number of murders and nonnegligent manslaughters occurring in Texas during the period studied. Excluded from this category were negligent manslaughters, accidental homicides, justifiable homicides committed by citizens and police officers, and executions performed by the state. Murders involving burglary, robbery, or sexual assault were coded as felony murders.4 Information on control variables that have most often been found to be related to homicide rates in previous studies was also collected. Information related to homicide in general, including the percentage of the state population living in metropolitan areas, the percentage of the population aged 18 through 34, and the unemployment rate, were culled from the Statistical Abstracts of the United States (see Land, McCall, and Cohen 1990; Peterson and Bailey 1991). The number of physicians per 100,000 residents was also coded from the Statistical Abstracts of the United States and is included as a proxy for the availability of emergency services, which could prevent an aggravated assault from turning deadly. Other variables available in the Statistical Abstracts of the United States that are typically included in homicide studies are the percentage of Blacks and the percentage of divorced individuals. They were excludedfrom this study because both were constant over the time period studied. Furthermore, these variables were not significant predictors of homicide rates in a recent deterrence study (Peterson and Bailey 1991).

Additional information was collected from alternate sources. The percentage of murders resulting in convictions was collected from the Annual Reports
of the Texas Judicial Council as an additional measure of the certainty of punishment. The rate of incarceration per 100,000 in the state was gathered from the Bureau of Justice Statistics. The incarceration rate was included to control for possible incapacitation effects resulting from a vast increase
in Texas's prison population during the time period studied. Information on the percentage of Texas residents who are on Aid to Families with Dependent Children (AFDC) was gathered from the Texas Department of Human Resources. The direction of its expected relationship to homicide is not specified herein. Although a direct relationship between welfare and homicide rates is typically expected, a recent study found that AFDC is an indicator of available resources that act to mitigate the harshness of poverty, thereby decreasing homicide rates (DeFronzo 1997).

Control variables were also calculated from the SHR data. The percentage of homicides resulting from gunshots was included as a proxy for the availability of firearms. Temporal variables were included to account for surges and lulls in the homicide rate. A high- and low-season variable specified months that were found to be significantly higher or lower in general homicide rates. High season included the months of July and August, whereas low
season included only the month of February. Because the state experienced a record number of executions in 1997, an indicator of that year was also
included as a control variable. Lagged-execution variables, Tl to T3, were also calculated.

Following Chamlin, Grasmick, Bursik, and Cochran (1992), the unit of analysis is the month. Although Chamlin and colleagues did not find significant
macro-level deterrent effects when their data were aggregated at longer time intervals, they did find deterrent effects when lagging data in shorter
temporal aggregations. A month was the shortest time interval available for which information about the dependent variables was recorded. Accordingly, although we aggregated the number of executions by month, control variables were typically observed on a yearly basis; thus, monthly figures were estimated using linear interpolation. These estimation procedures were appropriate because these variables were treated only as control variables, and not as alternative explanatory variables (Peterson and Bailey 1991).

ANALYSIS AND FINDINGS

Figure 1 provides an overview of execution and murder rates during the time period encompassed by the study. This figure illustrates the episodic
nature of executions. After the first execution, which was that of Charlie Brooks, was carried out in December 1982 (not included in Figure 1), the
next one did not take place until James Autry was executed in March 1984. A small wave of executions, which peaked at 10 in 1986, followed. A slump
in executions then occurred, with an average of four per year being carried out during 1988 through 1991. The ascendance of executions in 1992 signaled the beginning of a more substantial wave of executions, with an average of 15.5 executions per year during 1992 through 1995.

A legal challenge to Texas's procedures for speeding up the appellate processing of capital cases resulted in a moratorium on executions. With the exception of the voluntary execution of Joe Gonzalez in September 1996, executions were halted to await a decision of the Texas Court of Criminal Appeals on the legality of the new procedures. Executions resumed in February 1997, after the court's pronouncement that the expedited appellate procedures were constitutional. The next wave of executions in Texas would be of historical significance. In dispatching 37 backlogged cases, Texas reached a new record for the number of executions carried out in the state during a single year.

The rate of murder in the state from 1984 through 1991 showed no discernible trend in relation to the execution rate. Although there was a slight decrease in murder rates in 1987 through 1989 after the execution wave of the mid-1980s that could be attributed to a deterrent effect, the homicide rate only began to increase in 1990 and 1991, which was after a 2-year lull in executions during the late 1980s. Although the increase could be attributed to the earlier lull in executions and hence support the deterrence hypothesis, the considerable lag in its increase would suggest that any deterrent effect, or lack thereof, occurred only after a considerable time and is of limited significance.

[IMAGE GRAPH] Captioned as: Figure 1:

The greatest amount of support for the deterrence hypothesis is found when the decrease in murder rates is paired with the increase in executions
during the 1990s. During the execution wave of the 1990s, the murder rate declined substantially in the state. In the same year that the state reached
a historical high in executions, the murder rate fell below what was experienced in decades. Although this seems to provide a strong support for the deterrence hypothesis, the downward trend in homicide rates does not appear to be disturbed by the moratorium on executions in 1996, as the deterrence hypothesis would predict; instead, the downward trend continued. Although a bivariate regression model (not reported in tabular form) produced a significant equation, with executions explaining 7 percent of the variance in murder rates (b = -.046; t= -3.640; p < .001), the estimates were not reliable due to a high degree of serial correlation (Durbin-Watson = .558). Furthermore, murder rates have been declining throughout the United States during this same period, which suggest that factors unrelated to executions were responsible for this pattern.

To test for the influence of other causal factors, control variables were included along with executions and used to predict general murder rates
and felony murder rates across the monthly series of data from 1984 through 1997. In the first model, the general murder rates were regressed on executions and the control variables. An analysis of residuals from a preliminary ordinary least squares (OLS) regression model that was used to predict
murder rates indicated possible heteroscedasticity. In addition, serial autocorrelation found in the original OLS equation (Durbin-Watson = 1.594)
suggested the need for some type of correction. The equation was recalculated using the Newey-West variance estimator, which was specifically designed to correct for these violations of OLS assumptions (Newey and West 1987; StataCorp 1997). Because the Durbin-Watson test indicated problematic levels of autocorrelation up to the third lag, the model presented in Table 1 included Newey-West variance estimates based on a model with three lags.

As shown by the coefficients presented in Table 1, the number of executions was not related to murder rates over the 14-year period that was studied.
Control variables positively related to murder rates included the percentage of the population in metropolitan areas, the percentage of the population
age 18 to 34, the murder conviction rate, and the high season. The low season, February, was the only variable with a significant negative relation
to murder rates. Inclusion of these variables produced a high degree of fit to the data with an overall R^sub 2^ of .75.

The model presented in Table 2 limits the dependent variable to felony murders. Because the Durbin-Watson statistic did not indicate a high degree
of autocorrelation and because the residuals were more normally distributed, a simple OLS regression model was employed. Because models that were run with lags, T^sub 1^ to T^sub 3^, showed no difference in findings, only the nonlagged model is presented below.

Just as in the model that predicted murder rates in general, the rate of felony murder was not related to the number of executions. The same variables
found to be significantly related to murder rates in general were again found to be significant predictors of felony murder rates. The percentage
of the population in metropolitan areas, the percentage of the population age 18 to 34, the murder conviction rate, and the high season were positively
related to felony murder rate, whereas the low season was negatively related. One additional variable having a significant positive coefficient in the
felony murder model was the unemployment rate.

CONCLUSIONS

This study found that recent evidence from the most active execution state in the nation lent no support to the deterrence hypothesis. The number
of executions did not appear to influence either the rate of murder in general or the rate of felony murder in particular. At the same time, no
support was found for the brutalization hypothesis. Executions did not reduce murder rates; they also did not have the opposite effect of increasing
murder rates. The inability to reject the null hypothesis supports findings from the vast majority of studies on deterrence and capital punishment
(Peterson and Bailey 1998). From the data presented, it appears that other factors are responsible for the variations and trends in murder rates.

[IMAGE TABLE] Captioned as: TABLE 1:

Although appropriate methodology and statistics were employed in the analysis, a number of criticisms could be raised concerning our failure to find evidence of a deterrent effect. As noted elsewhere, using the SHR to measure dependent variables presents problems of reliability and validity (Maxfield 1989). Missing data are particularly troubling when data are disaggregated to calculate felony murder rates (Peterson and Bailey 1991). Cases that are missing data on the circumstances surrounding a homicide in the SHR are likely to turn into felony murders when their circumstances are finally
uncovered (Riedel, Zahn, and Mock 1985). Furthermore, it may be that the deterrent effects of executions on potential offenders can never be adequately ascertained (Van den Haag 1975), because measuring crimes that did occur is only a proxy for how many that were prevented. We also concede that the findings are limited to the sampled time period and the jurisdiction studied.

Considering these limitations, any research that makes a claim concerning the deterrence hypothesis should be treated with caution. However, because
we confirmed the findings of previous studies and because the Texas context was so uniquely suited for finding any potential deterrent effects, there
is little reason to question the findings.

[IMAGE TABLE] Captioned as: TABLE 2:

Once this argument is accepted, several implications can arise. Some may infer, for example, that these results suggest the repeal of the death
penalty because it fails to serve the penological function that is so often offered in its defense. Others would argue that various other goals must
also be taken into consideration before making this determination, such as whether the public supports its use, whether it serves the goals of
retribution, whether it saves money over life imprisonment, whether it serves to provide justice to the families of victims, and whether it serves
the interests of the criminal justice system in general. However, these justifications have also been challenged by research (Acker, Bohm, and
Lanier 1998; Bedau 1997). Along with the steady stream of consistent findings on the failure of capital punishment in all of these areas, this study
cannot help but support the abolitionist argument.

NOTES

1. Detailed information on homicides, a necessity in disaggregating felony murders from more general ones, has been routinely kept by the state since
1976 in the Supplemental Homicide Reports.

2. In addition to being the largest jurisdiction in Texas, Houston is the most significant contributor to the number of murders, particularly felony
murders, in the state.

3. Information from Houston would be crucial in estimating statewide murder rates, especially felony murder rates, for 1983, the year immediately following the first post-Furman execution.

4. Disaggregating murders into a felony murder category was imperative, because this type of murder was eligible for capital punishment. Felony-related
murders have also been the category of capital murders that have most often resulted in death sentences and eventual executions in Texas (Marquart,
Ekland-Olson, and Sorensen 1994). Noncapital murders, especially those occurring in the heat of passion, should not be expected to decline in
response to executions because they are not punishable by death. Death-eligible homicides, particularly those involving the premeditation of felony-related murders, should reasonably be expected to decrease in response to executions if the deterrence hypothesis is correct.

REFERENCES

Acker, James R., Robert M. Bohm, and Charles S. Lanier. 1998. America's Experiment with Capital Punishment: Reflections on the Past, Present, and
Future of the Ultimate Sanction. Durham, NC: Carolina Academic Press.

Bailey, William C. 1983. "Disaggregation in Deterrence and Death Penalty Research: The Case of Murder in Chicago." Journal of Criminal Law and Criminology 74:827-59.

Bailey, William C. and Ruth D. Peterson. 1987. "Police Killings and Capital Punishment: The Post-Furman Period." Criminology 25:1-25.

Baldus, David C. and James W. L. Cole. 1975. "A Comparison of the Work of Thorsten Sellin and Isaac Ehrlich on the Deterrent Effect of Capital
Punishment." Yale Law Journal 85:170-86.

Bedau, Hugo A. 1967.The Death Penalty in America Rev. ed. New York: Doubleday.
1997. The Death Penalty in America: Current Controversies. New York: Oxford University Press.

Bowers, William J. and Glenn Pierce. 1975. "The Illusion of Deterrence in Isaac Ehrlich's Research on Capital Punishment." Yale Law Journal 85:187-208.

.1980. "Deterrence or Brutalization: What is the Effect of Executions?" Crime & Delinquency 26:453-84.

Cardarelli, Albert P.1968. "An Analysis of Police Killed in Criminal Action: 1961-1963." Journal of Criminal Law, Criminology and Police Science 59:447-53.

Chamlin, Mitchell B., Harold G. Grasmick, Robert J. Bursik, Jr., and John K. Cochran. 1992. "Time Aggregation and Time Lag in Macro-Level Deterrence Research." Criminology 30:377-95.

Cheatwood, Derral.1993. "Capital Punishment and the Deterrence of Violent Crime in Comparable Counties." Criminal Justice Review 18:165-79.

Cochran, John K., Mitchell B. Chamlin, and Mark Seth.1994. "Deterrence or Brutalization? An

Assessment of Oklahoma's Return to Capital Punishment." Criminology 32:107-34. Decker, Scott H. and Carol W. Kohfeld.1984. "A Deterrence Study of the Death Penalty in Illinois, 1933-1980." Journal of Criminal Justice 12:367-77.

Decker, Scott H. and Carol W. Kohfeld. 1990. "The Deterrent Effect of Capital Punishment in the Five Most Active Execution States: A Time Series Analysis." Criminal Justice Review 15:173-91.

DeFronzo, James.1997. "Welfare and Homicide." Journal of Research in Crime & Delinquency 34:395-406.

Ehrlich, Isaac. 1975. "The Deterrent Effect of Capital Punishment: A Question of Life and Death.""American Economic Review 65:397-417.

Forst, Brian. 1977. "The Deterrent Effect of Capital Punishment: A Cross-Tabular Analysis of the 1960's." Minnesota Law Review 61:743-67.

Furman v. Georgia, 408 U.S. 238 (1972).

Klein, Lawrence R., Brian Forst, and Victor Filatov.1978. "The Deterrent Effect of Capital Punishment: An Assessment of Estimates." Pp. 331-60 in
Deterrence and Incapacitation: Estimating the Effects of Criminal Sanctions on Crime Rates, edited by A. Blumstein, J. Cohen, and D. Nagin. Washington, DC: National Academy of Sciences.

Land, Kenneth C., Patricia L. McCall, and Lawrence E. Cohen. 1990. "Structural Covariates of Homicide Rates: Are There Any Invariances Across Time and Social Space?" American Journal of Sociology 95:922-63.

Marquart, James W., Sheldon Ekland-Olson, and Jonathan R. Sorensen. 1994. The Rope, the Chair, and the Needle: Capital Punishment in Texas, 1923-1990.Austin: University of Texas Press.

Maxfield, Michael G. 1989. "Circumstances in Supplemental Homicide Reports: Variety and Validity." Criminology 27:671-95.

Newey, Whitley and Kenneth D. West.1987. "A Simple, Positive, Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix." Econometrica 55:703-8. Passell, Peter. 1975. "The Deterrent Effect of the Death Penalty: A Statistical Test." Stanford Law Review 28:61-80.

Peterson, Ruth D. and William C. Bailey.1988. "Murder and Capital Punishment in the Evolving Context of the Post-Furman Era." Social Forces 66:774-807.

. 1991. "Felony Murder and Capital Punishment: An Examination of the Deterrence Question." Criminology 29:367-95.

1998. "Is Capital Punishment an Effective Deterrent for Murder? An Examination of the Social Science Research." Pp. 157-82 in America's Experiment with Capital Punishment: Reflections on the Past, Present, and Future of the Ultimate Sanction, edited by J. R. Acker, R. M. Bohm, and C. S. Lanier.
Durham, NC: Carolina Academic Press. Riedel, Marc, Margaret A. Zahn, and Lois Mock. 1985. The Nature and Pattern of American Homicide. Washington, DC: Government Printing Office.

Schuessler, Karl F 1952. "The Deterrent Effect of the Death Penalty." The Annals of the American Academy of Political and Social Science 284:54-62.
Sellin, Thorsten. 1967. Capital Punishment. New York: Harper & Row. . 1980. The Penalty of Death. Beverly Hills, CA: Sage.

StataCorp.1997. Stata Statistical Software (Release 5.0) [Computer software]. College Station, TX: Author.

Sutherland, Edwin H. 1925. "Murder and the Death Penalty." Journal of Criminal Law and Criminology 15:522-9.

Thompson, Ernie. 1997. "Deterrence Versus Brutalization: The Case of Arizona." Homicide Studies 1:110-28.

Van den Haag, Ernest. 1975. "Why Capital Punishment?" Albany Law Review 54:501-14.