Over the past few years, the names of Amadou Diallo and Patrick Dorismond have been seared into the public consciousness of New York, in large part because they both were killed during a time when the public perception was that New York City was safer than ever before. The stories of these two men--innocent of any crime, yet gunned down by the police--seemed to emphasize a disturbing trend of suspect shootings by law enforcement officers. Former New York City Mayor Rudolph Giuliani has disputed this trend, arguing that in fact the police killed more citizens during his predecessor's reign. (1) Yet, as even the most junior student of econometrics could point out, Giuliani's predecessor also oversaw a more violent city.
Commentators have focused on a number of possible explanations for this apparent increase in the level of police-citizen violence--ranging from poorly trained police forces (2) and the recent arrival of a large number of African and West Indian immigrants, (3) to the personal leadership style of former Mayor Giuliani himself. (4) Yet none of the commentators has suggested that the apparent relative rise in gun violence between police and suspects could have resulted from, or be related to, what is truly the most dramatic change in New York's criminal justice system in the last decade: the reintroduction of capital punishment. (5)
This reintroduction of capital punishment could affect the likelihood of police-citizen violence in two ways. First, the possibility that the suspect will be subjected to capital punishment for an initial crime--such as a murder committed during an armed robbery--might lead a suspect to "fire away" upon encountering police rather than surrender peacefully. (6) In other words, the death sentence a suspect could face for the initial capital offense might assume greater immediate significance to the suspect than the risk of being sentenced to death for killing a police officer in a shootout while attempting to avoid arrest for the initial offense. Second, and conversely, the possibility of facing capital punishment for killing a police officer might be greater than the punishment for the initial crime, thus marginally deterring violence against police--i.e., the suspect may be deterred from resorting to gunfire in a police encounter by the risk of being subjected to a more severe punishment than that which the suspect could be subjected to for the initial offense.
This article examines the relationship between changes in capital punishment laws across the United States, and violence between police officers and suspects. It seeks to answer the question of whether the reintroduction of the death penalty makes it more or less likely that police officers will engage in violent conflicts while attempting to arrest suspects. My interest in this subject is motivated by three classes of issues: the empirical implications of capital punishment, the search for a theoretical model that could be used to predict possible outcomes for the relationship between capital punishment and police-citizen violehce, and the policy behind the need for such research. Part II of the article briefly describes past empirical research on the subject of police-citizen violence. This type of violence will be referred to as "shoot-outs." Part III describes the data and the empirical strategy employed in this essay; Part IV describes the results; and Part V provides a theoretical model of police-citizen violence. Finally, Part VI provides a conclusion.
A. Empirical Motivation
This article seeks to remedy the failure of scholars to establish empirically that the death penalty produces real behavioral effects. In the 1976 capital punishment case of Gregg v. Georgia, (7) Justice Potter Stewart wrote, "[s]tatistical attempts to evaluate the worth of the death penalty as a deterrent to crimes by potential offenders have occasioned a great deal of debate. The results simply have been inconclusive." (8) A review of the work of the countless scholars cited in Gregg, in addition to others, leads to the same conclusion that the Court reached: The death penalty cannot be conclusively said to change criminal behavior. (9) This inconsistency in the results of the statistical work on deterrence has persisted despite death penalty scholars' increasingly sophisticated empirical approaches and despite the application of the most advanced computer technology to their statistical studies. (10)
Since a death sentence differs so substantially from all other forms of punishment, (11) one would expect its use to have profound behavioral implications. According to past research, however, the death penalty does not appear to induce any changes in behavior.'2 As the Supreme Court stated in Gregg, `"[A]fter all possible inquiry, including the probing of all possible methods of inquiry, we do not know, and for systematic and easily visible reasons cannot know, what the truth about this `deterrent' effect may be....'" (13) It was also noted by Professors William Bailey and Ruth Peterson that "research indicates ... that the overall (general) homicide rate is not responsive to capital punishment. It is still possible that some forms of killing may be deterred by capital punishment, while other types of murder may be encouraged ... by the death penalty." (14)
There are several reasons why one would expect the incidence of police-citizen violence to depend more heavily on capital punishment laws than the incidence of homicide generally. First, killers do not always commit a capital crime. Those who commit non-intentional, non-premeditated homicides may not be subject to capital punishment. Limiting the focus of a statistical study of the deterrent effects of the death penalty to first-degree murderers, however, is an imperfect solution. Such an approach ignores the fact that many actual first-degree murderers plead guilty to lower degrees of homicide. The decision to engage in a violent conflict with police, however, always has at least one potential outcome: the murder of a police officer, which, where capital punishment is imposed, is almost always a capital crime. (15)
Second, unlike in many other homicides, where strong emotions and personal relationships play a role, (16) violence against police officers may be less likely to result from emotional impulses--and certainly is less likely to result from emotional attachments. For example, personal feelings of love and hate are not as likely to influence suspects' choices of whether or not to engage in violence, because the suspects rarely have any personal attachment to the officers that come knocking at the door. (17) The relatively minor role that strong emotions play in police-citizen conflicts suggests that a "rational actor" model is more appropriate in this context than, for example, in the context of spousal homicide.
Third, police-citizen violence might stem either from decisions by citizens or from decisions by police. Criminal suspects might decide to engage in violent behavior based on inaccurate information. For instance, a suspect might not have accurate information concerning the possibility and likelihood of capital punishment. Thus a suspect's behavior in a given situation might be obscured by his or her lack of information regarding the death penalty. In contrast, police officers are much more likely than suspects to have accurate information regarding capital sentencing in their jurisdiction. For example, if a particular District Attorney was more aggressive implementing the death penalty, police officers might assume that criminals, acting with this knowledge, would be more aggressive in their attempts to avoid capture. As a result of this assumption, police officers might adopt a posture toward suspects that would make the occurrence of shootouts more likely. Therefore, even if criminals rationally do not choose violence based on accurate information, police officers may choose violence by reasonably inferring that criminals operate within a rational and fully informed framework.
B. Theoretical Motivation
This article also seeks to craft a theoretical model for the possible relationship between capital punishment and police-citizen violence. The only previous study of this relationship covered a span of eleven years starting in 1976. It was conducted by two social scientists on an empirical examination of the effect of the death penalty on homicides committed against police officers. (18) This present study employs a distinct, and hopefully more effective, empirical strategy by capturing a period in which the death penalty was more prevalent. (190 It also offers a thorough theoretical exposition using a formal economic model of the relationship in question, whereas past studies have failed to construct such a model. (20)
Since scholars have failed to make use of the powerful analytical tool of a formal model, they cannot fully appreciate the dynamics of the relationship between capital punishment and police-citizen violence. The death penalty could make police-citizen violence more likely, or, alternatively, the death penalty could make police-citizen violence less likely. (21) Past studies sought simply to evaluate whether capital punishment in fact deters the killing of police officers. (22) The theoretical model elaborated in this article indicates, however, that capital punishment may be likely to induce a higher level of violence against police officers rather than a lower level. (23)
Statistical results indicating no clear positive link between the death penalty and police-citizen violence might indicate, as some earlier work concluded, that the death penalty has no effect. (24) In contrast, the theoretical model proposed in this article indicates that the deterrence of and the inducement to violence provided by capital punishment may in fact offset one another. The profound implication of this is that capital punishment might influence behavior, even though no significance is achieved in a statistical analysis of this behavior.
C. Policy Motivation
This article also seeks to provide meaningful answers to law enforcement policy-makers concerned with the effects of the death penalty on police safety. For the hypothetical social-utility-maximizing policy-maker, the question posed in this study is of great importance.
The death penalty might increase social welfare by deterring violent crime. (25) If suspects know for example, that they will face the death penalty in the event they choose to commit a violent crime, suspects might be deterred and the risk of violent crime might be reduced. Yet, if the death penalty increases the incentive to violently attack police officers, it might directly and indirectly reduce social welfare. For example, increasing the chances of suspects' harming police officers would reduce the social utility of having the death penalty, and thus would have a direct impact. On the other hand, fear of conflict with violent suspects could affect the way police officers treat citizens on a daily basis, and thus could have an indirect impact. Further, police officers more fearful of violence might be more likely to treat citizens with suspicion. This "chilling effect" on police-community relations might prevent implementing effective proactive law-enforcement strategies such as community policing. (26) In an extreme form, this "chilling effect" might EVEN lead to police officers' actually harming citizens--including innocent ones--because of the police officers' fear of being attacked. Any actual attacks would, of course, further contribute to hostility between police forces and members of the community. (27)
II. PAST RESEARCH ON POLICE-CITIZEN VIOLENCE
Since the 1950s, several scholars have examined the link between the killing of police officers and capital punishment. The author of the first of these studies surveyed police departments in seventeen death-penalty and six non-death-penalty states, and found that the rate of police homicides did not differ substantially between the two categories of jurisdictions. (28) This study was flawed in that it used police homicides per 100,000 persons as the dependent variable, ignoring the possibility that the number of killings of police officers could depend more on the number of officers than on the number of citizens they patrolled. A further study corrected this problem by using police homicides per thousand police officers as the dependent variable--it nonetheless obtained the same results. (29)
William Bailey and Ruth Peterson conducted the most recent empirical work on the subject. They examined whether changes in the probability of the execution of convicted murderers and changes in the probability of the execution of convicted police killers explained the variation in the rate of police killings that occurred over time. (30) The authors used the amount of media attention devoted to executions as an independent variable. (31) By using a national sample, however, the authors hopelessly doomed their results. The problem is that no individual potential police killer actually faces the nationally aggregated certainty of execution. They face an individual state's possibility of execution. Furthermore, the authors failed to explain why they expected there to be any variation from month to month in most of their exogenous variables. (32) Finally, the authors ignored--indeed, they tried to control for (33)--an incredible source of natural variation: changes in death penalty laws across the country. (34)
In a later study, Mona Margarita found that people who kill police officers tend to kill …