The Effects of Ethnic Disparities in. Violent Crime

Transcription

1 Senior Project Department of Economics The Effects of Ethnic Disparities in Police Departments and Police Wages on Violent Crime Tyler Jordan Fall 2015

2 Jordan 2 Abstract The aim of this paper was to analyze whether regression modeling could suggest a causal relationship between police officer race and resident race. Also, it aimed to suggest that police wages would decrease crime. The expected signs were consistent with theory and intuition and could possibly have very important implications for the future.

3 Jordan 3 Table of Contents I. Introduction Page 4 II. Literature Review and Theoretical Model Page 6 III. Econometric Model Page 12 IV. Data Page 13 V. Results Page 15 VI. Implication Page 17 VII. Limitations VII. Conclusion Page 18 Page 19 VIII. References Page 20

4 Jordan 4 Introduction The Washington Post recently reported that 2014 had the lowest number of reported violent crimes in decades in the U.S (Ingraham 2015). According to Evans and Owens (2007), a contributing factor may have been Community Oriented Policing Services grants, or COPS grants, awarded to local police departments by the federal government. Their results suggest that in some cities these COPS grants, grants created to increase the number of police officer in a cities police department, can negatively impact crime. The purpose of the COPS grants was to increase funding for the police departments that applied so they could hire more police officers. This would allow them to have more police officers patrolling the streets to either catch criminals or negate crime. The grant period lasts for 36 months and includes up to 75% of the new officer s salary and includes the benefits that the new officer would receive. The grant program was started to ensure that police departments could continue, or begin, community policing in areas that were high in crime. The idea is that community policing is an effective way to eventually decrease crime by detaining criminals using a police force with adequate manpower. Alternatively, community policing may stop crime from happening by the threat of apprehension appearing high to criminals. Police grants awarded to departments for the purpose of community policing is one way to examine police officers effects on crime is a way to deal with the department s endogeneity. However, the measurement of the grants awarded is not exactly a measurement of community policing, but he United States Department of Justice is defines the practice as, A philosophy that promotes organizational strategies that support the systematic use of partnerships and

5 Jordan 5 problem-solving techniques to proactively address the immediate conditions that give rise to public safety issues such as crime, social disorder, and fear of crime (COPS 2014). It goes on to break down community policing into three parts: Organizational Transformation, Problem Solving, and Community Partnership. Evans and Owens (2007) instrumental variable, COPSs grants, was a component of organizational transformation, the alignment of management, structure, and personnel for the implementation of community policing. Community partnership, the relationship between law enforcement and those that they protect, is a third part of this definition and warrants attention. In these trying times it is important to examine the context of situations. Thus, by examining the representation of the community by police departments we can begin to understand the relationships between police officers and crime. This paper will discuss the discrepancy between police departments and the residents in the community. In addition to examining the relationship between Black police officers and Black residents, it will examine another aspect of organizational transformation by scrutinizing police officers salaries. The relationship between violent crime and police officers with adequate resources is considered ambiguous. More police can result in more criminals caught in the act or apprehended after a crime is reported. This leads to the reported crime rate actually increasing instead of decreasing. There is also the possibility that more crime will cause the hiring of more police officers. This sequence of events seems to cause the opposite of what people would think a greater police presence would do. Alternatively, more police can lead to less criminal activity by their mere presence raising the probability of criminals being caught or the perception that more crimes are being committed since police can catch these criminals in the act. Both schools of thought are important when assessing the recurring, hot topic of crime. Finding the most effective way to utilize police funds is imperative to decreasing crime.

6 Jordan 6 However, one aspect that has not been analyzed yet is the effect of police wages on crime. Police officers, like all other forms of employment, respond to incentives, particularly pay. It is important to determine the effect that wage has on crime because more favorable policing outcomes are beneficial to society both politically and economically. According to the previously cited U.S. Department of Justice s report, personnel alignment is an integral part of community policing. Higher police wages would likely cause personnel alignment to increase productivity since police officers would be responsive to the incentive of wages. Police productivity has yet to be analyzed as workers who respond to incentives. Certain attributes necessary to be a competent police officer may be different from other forms of employment, but the occupation should still obey the theory that a higher wage will motivate workers, in this case police officers, to perform more productively. Also, theory says that employee compensation should be based on worker productivity. This means that more productive police officers will be compensated with higher wages. It can be inferred that when community policing is implemented police officers with high wage will align their goals with that of the police department that employs them. Previous literature examines the probability of being caught by police officers presence, the severity of punishment, and the probability of conviction by the criminal justice system, but none have examined police salary. This paper will analyze the effect of police officers wages on the violent crime rate in This paper will also have a focus on community policing by examining the impact of the gap between the percentage of police officers that are Black and the percentage of residents that are Black in 2007.

7 Jordan 7 Literature review Becker (1978), develops the theory that criminals use cost-benefit analysis when deciding to commit a crime. Their expected utility, how much the criminal plans to gain by committing the crime, is decided by the actual utility they would gain, how much income they would lose through conviction, the probability of being convicted, and the punishment per offense. He states that the number of offenses committed is a function of the probability of conviction, punishment per offense, and other influences. These other influences include the effectiveness of police officers. The author contends policing is a combination of manpower, technology, and capital. Becker argues that criminal behavior is rational, but, that is only true for property crime. He suggests that violent crime is likely to be irrational since income is potentially always nothing. The benefit of violent crime is implicit in the crime itself. However, since crime, whether it is violent crime or property is affected by police officers the hypothesis that policing affects violent crime still stands. Violent crimes may be affected by the number of police officers. Some literature that applied Becker s theory such as Vollaard (2006) and Levitt (1997) focused on how police officers affect crime. However, the ambiguous and endogenous relationship between the two needs to be controlled. Many analytical papers on the topic used instrumental variables to control the relationship. An instrumental variable is a variable that affects only the independent variable in the endogenous relationship. The instrumental variable estimates the causality of the independent variable on the dependent variable by showing the instrumental variable s effect on the dependent variable in the regression. One case is to use a two-stage least squares regression. In other words, an instrumental variable is used because it is ambiguous whether hiring more police officers is a response to crime or higher crime is a response to hiring more police officers.

8 Jordan 8 Levitt (1997) used gubernatorial and mayoral elections as instrumental variables while attempting to explain how the number of police officers affects violent crime. His regression suggested that an increase in the number of police officers significantly decreases violent crime, but has a smaller impact on property crime. Levitt used data from 59 cities from 1970 to Levitt reaffirms the theory that the number of police officers negatively affects crime. Vollaard (2006) uses Dutch victimization data instead of reported crimes used in Levitt (1997). His findings suggest that more police leads to less less crime is a conclusion that is supported by Becker s theory, but he also finds that in areas with a high police presence there are less protective measures taken by people. He finds that in areas with a high police presence violent crime and property crime decrease. Lin (2009) argues that state sales tax affects state tax revenue which in turn affects funding for law enforcement. State sales tax acts as the instrumental variable in his analysis. The revenue that state government gains from these sales taxes is then used to fund law enforcement. Then, the amount of funding that is granted to a police department or law enforcement agency determines how many police officers are employed to deter crime. Lin s findings suggest that increases in state sales tax deterred crime in both his OLS and his 2SLS by adding funding to police departments and law enforcement. Lin s findings also show that his findings are much more robust in the 2SLS than in the OLS. Lin s findings are not a very strong addition to literature. Evans and Owens (2007) use Community Oriented Policing Services grants as instrumental variables to explain the link between crime and police officers. Initially, they used a regression to demonstrate that the federal grants actually increased the size of police forces and

9 Jordan 9 were not used to simply supplement the salaries of current employees. This initial finding was consistent with the hypothesis. The authors use Universal Hiring Grants, funds granted to police departments for the sole purpose of hiring more police officers, in their 2SLS regression. With the dollar amount of grants per 10,000 people as their independent variable they found that the COPS grants negatively affected all violent crime rates. This study is also consistent with Becker s theory that increasing the number of police officer s, increasing the probability of capture, decreases crime. Mustard (2003) builds upon this not by using a 2SLS regression, but by using a weighted least squares regression to estimate the impact of conviction rates on violent crime. He reasons that unweighted least squares regressions would produce heteroskedasticity in the model because of the low crime rates in counties with low populations. The theory implies that even if a criminal is caught it is still possible for him to incur little or very little cost if he is not convicted of the crime. He aimed to find if the probability of conviction, like the probability that a police officer will catch a criminal, decreases crime. His results suggest that conviction rates, the number of convictions divided by arrests, negatively impact violent crime rates. This is consistent with Becker s theory that increasing the cost of crime will decrease crime. Instead of examining police officer s effect on criminals it examines the criminal justice system s effectiveness at convicting criminals. Although Mustard s conviction rates are dependent on arrests that police officers must make, his findings suggest that the criminal justice system also deter criminals from crime by suggesting to criminals that they will pay for their crimes if caught. The previously literature is consistent with Becker s theory that both violent crime and property crime are decreased by the probability that the criminal will be caught and convicted.

10 Jordan 10 However, they do not examine the severity of punishment as a deterrent to crime. Becker s theory considers this as another cost to criminals. Mathur (1978) uses data on major crimes at the state level to test his hypotheses that the severity of punishment for a crime and probability of punishment reduce crime. To account for the fact that some variables of interest may be endogenous, he estimates a two-stage least squares using data from 1960 and Mathur s findings suggest that the severity of punishment does deter criminals from committing violent crimes. His results also suggest that the probability of punishment is a much more potent deterrent of violent crime than the severity of punishment. These findings suggest that violent offenders find the probability of punishment is a larger cost than the severity of punishment. Katz (2003) examines incarceration from a different angle. He uses capital punishment and prison deaths as independent variables in his analysis. He uses the number of prisoners executed per 1,000 state prisoners to define his capital punishment variable. He considers every prison death per 1,000 state prisoners as a definition of prison deaths. In his series of weighted least squares regressions, Katz study suggests that deaths in prison, most of them due to illness, may decrease violent crime and property crime. In his analysis, he finds that when the number of inmates who die behind bars increases, violent crime decreases more than property crime.. Although Katz findings suggest that violent crime is deterred by deaths in prison he hypothesizes that his method of defining prison deaths has many limitations. These findings may suggest that the probability of punishment is the strongest deterrent to crime in Becker s theory of crime and is also the most feasible. Thus, finding a way to increase the probability of punishment is important to the expansion of literature.

11 Jordan 11 Levitt (1997), Evans and Owens (2007), and Lin (2009) all use instrumental variables when estimating the impact of police officers on crime. Levitt s use of elections was important to estimating how the number of police officer will affect crime. Evans and Owens use of COPS grants were a way to estimate how a growing number of police officers affect crime while at the same time analyzing how community policing can affect crime. Lin s paper did not add much to the literature, but the use of state sales tax as an instrumental variable was a creative instrumental variable. Mathur (1978), Katz (2003), and Mustard (2003), decided to use data after the arrest to test their respective hypotheses. Severity of punishment and incarceration rates are all valid variables, but since these variables were not relevant for the hypotheses they were omitted. Levitt (1997) and Evans and Owens (2007) used increasing the probability of punishment as a way to deter crime. They also used crimes reported in their regression analysis to align their hypotheses with Becker s theory. This would be important to the hypothesis of violent crime being decreased by community policing since community partnership largely deals with the relationship between people in the community and the police force in that designated area. So, community policing and incentives for community policing are potentially close ideas that can explain crime. Since Evans and Owens model is closely related to the focus of this paper, the models will be based on their model. I have omitted variables that I deemed not relevant to my analysis and added variables that I believe will impact the model. The papers in the past also used different means of defining crime. In this paper, violent crimes reported will be used because the hypotheses proposed have a focus on police officers deterring crime by having an active presence in the community that they have jurisdiction over. These crimes reported should decrease if community policing is effective. Specifically, crime should decrease if the police

12 Jordan 12 force s ethnicity accurately reflects the residents of the community. The second hypothesis is that paying police officers more will decrease crime by causing more effective police officers to be a part of the organizational transformation. More effective police officers will reduce crime. Many of the variables in Evans and Owens (2007) are not relevant to the proposed model so they will not be used. Unemployment rates found at the Bureau of Labor Statistics were used to help explain through the model by suggesting that the more people who have less opportunities to work will increase crime. The unemployed do not have potentially lost wages if they are apprehended for a crime. So, the potential cost of committing a crime decreases. The population growth rate and the percent change in police officers will not be used since a dynamic model is not being used. The average salary of jobs covered by unemployment insurance could not be found, so the average salary of all occupations was used as a proxy. The number of people employed divided by the number of jobs insured by the population will not be used since only a proxy for the average salary of jobs covered by unemployment insurance was used. What I will be doing differently in my model is using police salaries as an independent variable. Patrolman and sheriff s officer salary is the independent variable that is imperative for the hypothesis that is being tested. The sign for police salary is negative due in part to the idea that police officers will respond to higher hourly wages or a higher annual salary by being more productive. This productivity may result in more alert patrolman catching criminals. So far, it has been explained that community policing is the focus of this paper and it will be examined through community partnership and organizational transformation. Those are two of the three ways that community policing will affect violent crime. The first, community partnership will be defined using the racial gap between Black residents and Black police officers. The larger the gap the greater violent crime will be. The second is through

13 Jordan 13 organizational transformation. This is where personnel decisions are important. The wages of police officers will be used to test the hypothesis that police wages will decrease crime. Empirical Model (1) logvcphtm = - logpolicewage + logpopm + URm + Youngm + Blackm - logpipcm εm, m = Metropolitan statistical area This first model is used because estimating the effect of police officer s wage by metropolitan statistical area would ensure many large cities and the suburbs at the same time. Police wages is taken as a log. Police wages themselves are the dollar wage that a police officer or sheriff s officer earns an hour. The population is the population of the metropolitan statistical area. An area with a high population in such a small area is expected to increase violent crime because with a higher population density more people should victimize each other more than a rural area with not many people. The unemployment rate is the number of people unemployed divided by the number of people in the work force. It is a percentage. It should increase crime because the cost of potentially losing income is much lower for people without income. Young is the percentage of people that are young in the MSA. Young people tend to commit more crime since their potential income is much lower than a person who is older with more potential income. The percentage of people that are Black is included because the narrative that Black people commit crime is important for the second hypothesis. Personal income per capita is defined as how much on average a person earns in a year. The higher personal income will be, the less crime should take place since being caught and punished for a crime can decrease income. It is taken as a log.

14 Jordan 14 (2) VCPHTc = - logpolicewagec + URc Gapc + logpopc + logpolicec + εc, c = city This second equation is used to determine the effect of inequities in a police department s racial makeup in communities that are racially dissimilar from the police department that has jurisdiction in that geographic area. The gap variable is the difference in the percentage between police officers that are Black and the percentage of residents that are Black. This is a part of community partnership in the community policing theory that the Department of Justice applies. The greater the gap is, the less people will decrease crime. The data for the gap variable is not available on the MSA level so it will be used on a city level. So, the first model will focus on police wages and the second model will focus on the gap variable.

15 Jordan 15 Table 1 City MSA Variable Name Abbreviation Variable Definition Violent Crimes Per Hundred Thousand Personal Income Per Capita % of the population that is Black % of the population that is Young Log of the population Police Wage Unemployment Rate logvcpht The log of the number of violent crime offenses reported per 100,000 people Mean (SD).012 (.034) logpipc+ Personal Income Per Capita (24.29) Black+ Young+ logpop+ logpolicewage+ UR+ Percent of the population that is Black or of African descent Percent of the population between the ages of 15 to 24 years old The log of the population The average wage of Patrolman and Sheriff's Officers The unemployment rate Black Police BP- The percentage of police officers that are Black Total Police logpolice+ The total number of police officers Gap Gap+ The difference between the percentage of Black residents and the percentage of Black police officers (3.88) 8.79 (1.70) 9.23 (14.50) (765.35) (16.15) (16.15) Mean (SD).032 (.076) 36,996.2 (6,740.88) (10.83) (3.06) 7,611,170 (1,693,395) (5.69) 9.06 (3.00) Source

16 Jordan 16 The unit of observation for the first regression is by metropolitan statistical area and the data was recorded in The unit of observation was by city for the second regression and the data was recorded in The population variable will be used because areas with higher populations tend to have higher crime. Nolan (2004) points out that there is a direct relationship between population and crime. This explains the positive sign in the above equation. The personal income variable will be included because it will be used as a predictor of potential wages lost if a criminal commits a crime. The income of a metropolitan area may be indicative of the income of its inhabitants. Becker speculated that lost wages would serve as a cost to crime. Thus, it can be hypothesized that higher metropolitan income could have a negative impact on crime. The number of police officers will be included because police officers deter and catch criminals. This is an assertion that Becker s theory makes early on and is stated in the aforementioned academic papers. The sign before the variable is ambiguous because police officers may deter criminals or catch them. The former would decrease violent crime. The latter would increase violent crime. This will be used in the second model since it deals specifically with police officers numbers and the first model focuses on the impact of police officer wages. Also, the variable could not be accurately calculated on the MSA level. The difference between the percentage of police officers that are black and the percentage of residents are black is used to approximate the lack of community policing in the regression using cities as the unit of observation. The expected sign is positive because a larger gap is expected to be indicative of poor community policing. This variable is an indicator of

17 Jordan 17 community policing since it deals with community partnership, the relationship between police officers and their community. It shows the possible disconnect where community policing may be lacking. The log of violent crimes per 100,000 people is the dependent variable that was used. All variables that were not originally percentage were logged for a double logged regression for interpretation. The logged variables can be interpreted as percent changes in the estimates. For those variables that were already percentages, the estimate can be interpreted as a semi-log model. Data Table 1 has all of the variables used in the two models and has descriptive statistics for each on that is used. The first model uses data from 253 MSAs across the United States in all 50 states. The second model uses data from 78 cities in 10 states. Each of the variables has their mean above the standard deviation in parentheses. The variables that were exclusive to each model are included in the single table with their individual mean and standard deviation. The variables that did not have values for one of the models have been left blank.

18 Jordan 18 Results MSA Table 2 The log of violent crimes per hundred thousand Independent Variable Intercept logpolicewage logpipc UR logpop Parameter Estimate (Standard error in Parentheses) *** (1.69) -0.29** (0.13) (.167) (.0088) 1.095*** (0.027) PercentYoung PercentBlack (0.0083) 0.017*** (0.0024) R 2.89 N 253 Notes: Significant at 90% level (*) Significant at 95% level (**), Significant at 99% level (***) In the above table the log of police wages is found to be significant to the 95% significant level. We can interpret the parameter estimate as suggesting that when police wages increase by 1% violent crimes per hundred thousand in a metropolitan statistical area will increase by.29%.

19 Jordan 19 So, with 95% confidence when the violent crime rate is at the mean of.034 if police officer wages increase by 1% the violent crime rate will decrease between.16% and.42%. These results are consistent with the hypothesis, but they were expected to have a much larger effect. This shows that although there is a causal relationship between the two variables, it does not appear to be very strong in the regression. Like Lin (2009) the results are statistically significant, but suggest that the variable of interest is not very effective. The log of population was found to be significant at a 99% significant level. The parameter estimates suggest that when the population increases by 1% the number of violent crimes per hundred thousand people increases by 1.095%. This is consistent with the theory that places with high populations have more crime like in all other papers where population is used. The log of the percentage of residents that are Black is found to be significant at a 99% significant level. The parameter estimate suggests that when the percentage of the population is Black the number of violent crimes per hundred thousand increases by 0.017%. Like Levitt (1997) this variable statistically significant, but more importantly it can lead to implications in the second model s results about the racial gap between residents and police officers.

20 Jordan 20 City Table 3 The log of violent crimes per hundred thousand Independent Variable Intercept Gap logpolicewage UR logpop Parameter Estimate (Standard error in parentheses) *** (1.46) 0.015*** (.004).29 (.40) (.040) 0.62*** (0.13) logpolice 0.59*** (.13) R N 78 Notes: Signifcant at 90% level (*) Significant at 95% level (**) Significant at 99% level (***) In the above table, the difference between the percentage of the population that is Black and the percentage of the police force that is Black was found to be significant to a 99% level. The parameter estimate can have the interpreted meaning that when the gap between the percentage of residents that are Black and the percentage of police officers that are Black increases by 1% then the number of violent crimes per hundred thousand increases by 1.5%. We

21 Jordan 21 can say with 99% confidence that when the number of violent crime is at the mean of.012 if the racial disparity between police officers increases by 1% violent crime will increase between 1.1% and 1.9%. This is consistent with the hypothesis that an increase in the racial gap between police officers and residents will cause violent crime to increase. It is also important to note that in the last model where the percentage of the population that is Black is positive and statistically significant could possibly be influenced by the discrepancy in race. Like in the second table the log of the population was found to be significant to the 99% confidence level. So, it can be said that with 99% confidence that when the population increases by 1% the number of violent crimes per hundred thousand will increase by 0.62%. The log of the number of police officers is found to be significant to a 99% level. It can be suggested that when the number of police officers increases by 1% the number of violent crimes per hundred thousand increases by.59%. Implications One implication is that raising police officer wages does little to affect crime. It may decrease crime on an MSA level, but only an insignificant amount. Therefore, financial resources should not be routed into higher police officer salaries to encourage community policing, but finances could serve a purpose elsewhere with a larger effect on violent crime, possibly. A second implication is that a racial disparity between police officers and residents is a considerable problem. The disproportion could be a cause of civil unrest recently being publicized in the United States. So, it would be appropriate to suggest that a police department s ethnic diversity should reflect that of those that the police department is supposed to reflect.

22 Jordan 22 A third implication is that more police officers will increase crime. This is consistent with one of the outcomes that Becker proposed and other authors attempted to reject. No instrumental variable was used, but it is important to note that since the police officer variable was used in the model where the Gap variable was used it may be accurate since the police officer s ethnic makeup is important. Limitations There are limitations in the data. An issue that was found in both regression equations used is that the dependent variable, the log of the number violent crimes per hundred thousand, was violent crimes reported. Thus, it can only be suggested that the Gap variable measures that crime increases when there is a disparity between the diversity of a population and the police force in the community. A second issue is that the number of arrests or convictions of Black residents could not be found. If this variable could have been measured and used its implications could serve to be very interesting. The first limitation in the second regression was that police wages could not be found by city and the closest unit of observation, where the metropolitan statistical area was used. The second was that the number of police could have been endogenous with the number of violent crimes. A second limitation was not being able to find an instrumental variable for the effect of police officers on crime. Arrest rates would have been useful in both models, but the data was not available at either the MSA level or city level. In the future, it would be interesting to analyze how arrest rates would be affected by racial disparities in police departments and police officer wages.

23 Jordan 23 Conclusion In conclusion, police wages are an integral part of the criminal justice system, but the effect that they have on violent crime is very low. The complex matter of crime can be addressed in ways other than compensation for the police officers who begin the process of institutionalizing criminals. Furthermore, racial disparities between police officers and residents has a larger impact on violent crime. The regression analysis suggests that a racially homogenous police force may not be in the best interest of the people that a police department is tasked with serving. References Becker, Gary S. "Crime And Punishment: An Economic Approach." Journal of Political Economy, 1978, 169. Accessed February 11, Lin, Ming-Jen. "More Police, Less Crime: Evidence from US State Data." International Review of Law and Economics 29, no. 2 (2009): Accessed January 27, doi: /j.irle Mathur, Vijay K. "Economics of Crime: An Investigation of the Deterrent Hypothesis for Urban Areas." The Review of Economics and Statistics 60.3 (1978): 459. JSTOR. Web. Vollaard, Ben A. "Police Effectiveness: Measurement and Incentives." Diss. Pardee Rand Graduate School, Pardee RAND Graduate School, Jan Web. 27 Jan

24 Jordan 24 Levitt, Steven D. "Using Electoral Cycles in Police Hiring to Estimate the Effects of Police on Crime." The American Economic Review 87.3 (2002): JSTOR. Web. 27 Jan Mustard, David B. "Reexamining Criminal Behavior: The Importance of Omitted Variable Bias." Review of Economics and Statistics 85.1 (2003): Web. 27 Jan Katz, L. "Prison Conditions, Capital Punishment, and Deterrence." American Law and Economics Association 5.2 (2003): Web. 27 Jan Evans, William N., and Emily G. Owens. "COPS and Crime." Journal of Public Economics 91: Nolan, James J. "Establishing the Statistical Relationship between Population Size and UCR Crime Rate: Its Impact and Implications." Journal of Criminal Justice 32 (2004): Accessed October 28, "Community Policing Defined." Community Oriented Policing Services Accessed December 22,