DOES REPRESENTATIVE BUREACRACY IMPROVE PUBLIC SERVICE PERFORMANCE?: EVIDENCE FROM AN INSTRUMENTAL VARIABLE APPROACH

10.5465/AMBPP.2015.235

DOES REPRESENTATIVE BUREACRACY IMPROVE PUBLIC SERVICE PERFORMANCE?: EVIDENCE FROM AN INSTRUMENTAL VARIABLE APPROACH SOUNMAN HONG College of Social Sciences Yonsei University Seoul, Korea ABSTRACT This study finds that an increase in the share of ethnic minority officers in a given force is associated with a decrease in the number of crimes in the area under the force’s jurisdiction during the 10-year period. INTRODUCTION The theory of representative bureaucracy suggests that when the staff of a public organization is broadly representative, in demographic terms, of the communities it serves, the organization tends to pursue policies that benefit minorities. Representativeness in demographic characteristics may be particularly important among street-level bureaucrats such as police officers, teachers, and firefighters: these individuals frequently interact with the general public and thus represent the frontlines of government policy (Lipsky, 1980). Given the high degree of discretion inherent in their interactions with citizens, these frontline bureaucrats’ attitudes, values, and predispositions affect their decisions and the ensuing outcomes; these subjective factors, in turn, are often influenced by individuals’ backgrounds (Meier, 1993). For instance, police officers’ racial or ethnic backgrounds may affect their decisions to interrogate a black teenager about suspicious behavior—or even to escalate to more aggressive action. An example of such dynamics arose on August 4, 2011, when a 29-year-old black man, Mark Duggan, was shot and killed by police in North London. This incident raised tensions between members of the black community and police, resulting in public protests that escalated into one of the largest riots in modern English history. Three years later, August 9, 2014, a similar event occurred in the U.S. when an 18-year-old unarmed black teenager, Michael Brown, was fatally shot by a white police officer in Ferguson, Missouri. The disputed circumstances of the shooting sparked protests and civil unrest throughout the region and received considerable attention worldwide. These events illustrate that racial tensions continue to run deep even in advanced democracies; such tensions may be rooted in pronounced disparities in racial makeup between police departments and the communities they serve. Despite considerable public and media attention, however, relatively little quantitative empirical research has examined the impact of workforce representativeness or diversity on public service performance (though see Andrews, Ashworth, & Meier, 2014; Andrews, Boyne, Meier, O’Toole, & Walker, 2005; Meier, Wrinkle, & Polinard, 1999). Prior approaches have used self-reported survey data to test the relationship between diversity management and workgroup performance (Riccucci, Van Ryzin, & Lavena, 2014). Although this survey-based evidence has advanced collective understanding of the potential impact of workforce diversity, policymakers would be able to design diversity management policies with much greater

10.5465/AMBPP.2015.235

confidence if these findings were supplemented and supported by robust quantitative research using observational data. As such, this study contributes new evidence on the impact of police organizations’ ethnic minority representation on crime rates, a key indicator of police performance (Kelman & Hong, 2015; Kelman, Hong, & Turbitt, 2013). Police officers are typical examples of street-level bureaucrats who have a considerable level of discretion and can influence the quality of public service (Chaney & Saltzstein, 1998; Lipsky, 1980). Police organizations thus offer an excellent testing-ground for investigating the impact of representative bureaucracy (Thompson, 1976). By exploiting the large variation across areas and time periods in the relative share of ethnic minorities in the UK police workforce, this study estimates the impact of representative bureaucracy on organizational performance. In the UK, the Chief Constable of a given territorial police force1 is responsible for recruiting new officers; as such, myriad social, cultural, and political heterogeneities could affect his or her decision. When workforce-selection outcomes are examined econometrically, these background factors could also bias results if standard ordinary least squares (OLS) techniques are used. Our empirical strategy thus features an instrumental variable that constrained each force’s decision. In 1999, the UK central government set forcespecific 10-year targets for recruiting new police officers from ethnic minority backgrounds with the goal of increasing the share of ethnic minority officers on each English and Welsh police force to match the share of ethnic minorities in the population it served. We thus use this set of force specific-targets as an instrumental variable in our analysis of observational data from the 42 police forces of England and Wales from 2000 to 2010. Overall, our findings suggest that an increase in the share of ethnic minority police officers is associated with a significant decrease in the number of crimes within the associated police jurisdictional area. A 10 percent increase in the share of ethnic minority police officers during the studied period was associated with a 1.4 to 3.8 percent decrease in the number of crime incidents reported to the average police force. FRAMEWORK AND VARIABLES The instrumental variable (henceforth “instrument”) employed in this study is the increase in the share of ethnic minority police officers that was required to be achieved during the 10-year period for a given police force to become representative of its population in terms of the share of ethnic minorities. This is calculated as the percentage of ethnic minorities in the population minus the percentage of ethnic minority officers on the associated police force as of 1998 (Home Office, 1999). For the police forces that were already representative (or even overly representative) of minorities when the plan was formulated, this measure is set at zero. The instrument does not vary over time: the UK Home Office set these targets for each police force as the goals to be achieved at the end of the 10-year plan, rather than updating them annually (Home Office, 1999). Based on the targets set, police forces submitted their own 10year plans to the Home Office in 1999 and were required to follow them, rather than submitting yearly plans. Growth in the share of minority officers depended on rates of officers’ retirement, resignation, and dismissal; as these varied across forces, the incremental changes in minority representation did not follow a linear trend in most police forces. Thus, we use the time-invariant 10-year target as the instrument, rather than updating it annually, as the fixed target best describes the exogenous pressure the Home Office placed on the police forces.

10.5465/AMBPP.2015.235

The first-stage 2SLS regression thus tests the association between this instrument (the gap to close to reach the target) and the change in the treatment variable (the change in the share of minorities on the force). The dependent variable in the first stage (which becomes the treatment variable in the second stage) is first-differenced because the instrument corresponds to a change in the share of ethnic minorities. The first stage of the 2SLS is as follows. (1) where is the share of ethnic minorities employed by police force i in year t-1. is the instrument (the targeted increase in ethnic minority representation) for police force i, whereas and are the first-differenced treatment and control variables, respectively. is a year fixed effect, and is an error term. The model includes a number of control variables within the vector . First, the local population size is included to control for the widely different territorial sizes when comparing the number of recorded crimes. Second, the socioeconomic status of residents within the given area is controlled for, as the socioeconomic context may constrain the capacity or willingness of police forces to provide high-quality services. Specifically, we include the local unemployment rate and average income level. Third, each police force’s total budget is included, as expenditures (and therefore resources more broadly) may have affected crime rates during the study period. Fourth, Welsh police forces are identified with an indicator, as they may have different institutions or cultures from those in England. The last control variable is the size of the given police force, measured in terms of the number of officers. The second stage then tests the association between the change in the treatment variable and the change in the number of reported crimes, as follows. (2) where the dependent, treatment, and control variables (except for the Welsh police force indicator) are all first-differenced; is a year fixed effect and is an error term. is the dependent variable, the number of crimes reported in year t within police force i’s jurisdictional territory. The treatment variable now indicates the change in the share of ethnic minority police officers on the given force (in percentage points), whereas the dependent variable is the change in the number of crimes reported within the territory for which the force is responsible. The timeinvariant force-specific unobserved heterogeneity, , is eliminated. RESULTS We present the results of the 2SLS model in Table 1. Column 1 reports the 2SLS estimate with no control variables, while column 2 includes certain control variables but not police force size. The estimated impacts become smaller if police force size is controlled for, as reported in column 3. Results in column 3 suggest that, during the studied period, a 10 percent increase in the share of ethnic minority police officers was associated with a 3.8 percent decrease in the number of crime incidents reported to the average police force. Overall, the estimated impacts are larger under the 2SLS approach than under the OLS model (not reported).

10.5465/AMBPP.2015.235

-----------------------Table 1 about here -----------------------In column 4 of 5, we examine whether the negative association between police force ethnic representativeness and crime is moderated by police force size. Specifically, we test whether the impact of changes in ethnic minority shares is greater for forces that experienced large changes in size (column 4, Table 5). Results suggest that the interactions between ethnic minority representation and force size in terms of either level or change are not statistically significant. CONCLUSION The results of our study of representative bureaucracy within UK police forces suggest that, during the studied period, a 10 percent increase in the share of ethnic minority police officers was associated with a 1.4 to 3.8 percent decrease in the number of crime incidents as well as with a decrease in both the number and share of racist incidents. This supports the claim that ethnically representative bureaucracy improves the organizational performance of public services. Previous studies have largely focused on whether representative bureaucracy benefits minority groups. This study goes further to show that the benefits of representative bureaucracy may extend beyond the gains to minority groups to improve overall public service performance. ENDNOTES 1. In two territorial police forces (the Metropolitan Police and City of London Police), the chief officer holds the rank of Commissioner, rather than Chief Constable. REFERENCES Andrews, R., Ashworth, R., & Meier, K. J. 2014. Representative bureaucracy and fire service performance. International Public Management Journal, 17: 1-24. Andrews, R., Boyne, G. A., Meier, K. J., O’Toole, L. J., & Walker, R. M. 2005. Representative bureaucracy, organizational strategy, and public service performance: An empirical analysis of English local government. Journal of Public Administration Research and Theory, 15: 489-504. Chaney, C. K., & Saltzstein, G. H. 1998. Democratic and bureaucratic responsiveness: The police and domestic violence. American Journal of Political Science, 42: 745-68. Home Office. 1999. Dismantling Barriers to Reflect the Community We Serve. The Recruitment, Retention, and Progression of Ethnic Minority Officers: Targets. London: Home Office. Kelman, S., & Hong, S. 2015. This could be the start of something big: Linking early managerial choices with subsequent organizational performance. Journal of Public Administration

10.5465/AMBPP.2015.235

Research and Theory, 25: 135-164. Kelman, S., Hong, S., & Turbitt, I. 2013. Are there managerial practices associated with the outcomes of an interagency service delivery collaboration? Evidence from British crime and disorder reduction partnerships. Journal of Public Administration Research and Theory, 23: 609-630. Lipsky, M. 1980. Street-level Bureaucracy: Dilemmas of the Individual in Public Services. New York: Russell Sage Foundation. Meier, K. J. 1993. Representative bureaucracy: A theoretical and empirical exposition. In J. L. Perry (Ed.), Research in Public Administration, 2: 1-35. Meier, K. J., Wrinkle, R. D., & Polinard, J. L. 1999. Representative bureaucracy and distributional equity: Addressing the hard question. Journal of Politics, 61: 1025-1039. Riccucci, N. M., Van Ryzin, G. G., & Lavena, C. F. 2014. Representative bureaucracy in policing: Does it increase perceived legitimacy? Journal of Public Administration Research and Theory, 24: 537-551 Thompson, F. J. 1976. Minority groups in public bureaucracies: Are passive and active representation linked? Administration and Society, 8: 221-226.

10.5465/AMBPP.2015.235

TABLE 1. Ethnic Minority Representation in Police and Total Crime Incidents (2SLS) Dependent Variable:

Estimator N

(1) -39.70** (4.955)

(2) -40.00** (5.064)

(3) -21.27** (6.984) -1.874** (0.108)

2SLS 420

-1.299 (3.275) -3.297 (21.06) -0.249 (0.392) 0.011 (0.054) -0.698 (0.691) 2SLS 420

-1.491 (2.211) 9.377 (16.45) -0.273 (0.284) 0.020 (0.042) 0.131 (0.551) 2SLS 420

(4) -23.44** (9.064) -1.553* (0.884) -0.364 (0.919) -1.522 (2.329) 7.917 (17.61) -0.261 (0.294) 0.020 (0.043) 0.033 (0.604) 2SLS 420

Note:1. Robust standard errors clustered at the police forces in parentheses, * p < 0.10, ** p < 0.05; 2. All specifications include year fixed effect