New York Police Diversity Management

Ethnic diversity in policing: An application of quantile regression to the New York City Police Department

Salomon Alcocer Guajardo

Department of Public Management, John Jay College of Criminal Justice, The City University of New York, New York, New York, USA

ARTICLE HISTORY Received 8 June 2015 Revised 21 February 2016 Accepted 27 April 2016

ABSTRACT This study expanded on the use of Simpson’s diversity index (D D 1 - Pp2) to measure ethnic heterogeneity in policing and other criminal justice-related agencies. The index was applied to the police and civilian workforces of the New York City Police Department (NYPD) by using employment data for January 2015. Concomitantly, quantile regression was used to supplement the Tobit regression and to illustrate its ability to assess the effects of organization-based variables across the conditional distribution of the diversity scores. The quantile regression analysis found that ethnic diversity in the NYPD is affected differently by the type of workforce, the type of police and civilian positions, and by the number of employees in the police and civilian job titles and classifications. Statistically, the quantile regression detected significant relationships at various points of the conditional distribution while the Tobit regression analysis produced mixed results for the three models that were tested.

KEYWORDS Ethnicity; gender; race; quantitative research

Introduction

Recent research on ethnic and gender integration and diversity in the New York City Police Department (NYPD) has shown that minorities and women continue to be underrepresented in the police force and overrepresented in civilian positions (e.g., Guajardo, 2014a,b). However, the NYPD employment data for 2012 revealed that the level of diversity was generally higher within policing-based ranks in com- parison to the level of diversity within the civilian-based job classifications (e.g., Guajardo, 2014c). The recent research also shows that the employment disparity between men and women in policing-based supervisor1 (e.g., sergeant and lieuten- ant) and command2 (e.g., captain and bureau chief) positions has remained rela- tively stable despite an increase in the number of women officers and an increase in the number of command positions (e.g., Guajardo, 2015b,c). These findings are

CONTACT Salomon Alcocer Guajardo [email protected] Department of Public Management, John Jay College of Criminal Justice, The City University of New York, 445 West 59th Street, Room 3536 NH, New York, NY 10019, USA. © 2016 Taylor & Francis

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 2016, VOL. 14, NO. 4, 254–289 http://dx.doi.org/10.1080/15377938.2016.1187236

consistent with prior studies which have found that minority men and women have made marginal progress with respect to entering into policing and subse- quently ascending to supervisor and command positions (e.g., Bolton, 2003; Felkenes et al., 1993; Felkenes & Schroedel, 1993; Franklin, 2005; Martin, 1991, 2004; Polisar & Milgram, 1996; Raganella & White, 2004; Riccucci & Saldivar, 2014; Schroedel et al., 1996; Shelly et al., 2011; Stokes & Scott, 1996; Townsey, 1982; Warner et al., 1989; Zhao, Herbst, & Lovrich, 2001). Since few studies on eth- nic and gender integration into policing have utilized diversity indices to assess the overall level of workforce heterogeneity in police and civilian positions, the extent to which police agencies are diverse in terms of ethnicity and gender is not fully known (Guajardo, 2014a,b, 2015a). In addition, since the majority of the studies have used regression models (e.g., ordinary least squares [OLS] and Tobit regres- sion) that assess how independent variables affect the mean, the effect of the inde- pendent variables used in the analysis on different values of the dependent variable is not known. In a departure from prior research on ethnic and gender integration in policing, this study uses Tobit and quantile regression (QR) to assess whether organization-based variables (or covariates) have the same effect on the condi- tional distribution of the Simpson diversity scores calculated for the NYPD data for January 2015. Separate regression analyses also are performed for male and female cohorts to assess whether the covariates have similar statistical relationships in terms of the level of ethnic diversity that has been achieved by the NYPD.

This study has two main purposes. First, it expands on the application of the Simpson diversity index (D D 1 – Sp2) (Simpson, 1949) to employment data per- taining to the NYPD police and civilian workforces (e.g., Guajardo, 2015a). It does so by including the ethnicity and gender of employees jointly to obtain a more comprehensive measurement of workforce heterogeneity. In prior studies, the per- cent of Black, minority, or women officers in policing have served as dependent (or response) variables (e.g., Gustafson, 2013; Hochstedler & Conley, 1986; Kellough, 1990; Kim, 1993; Zhao, Herbst, & Lovrich, 2001). While these measures are useful for assessing ethnic or gender integration of specific populations into policing, they provide limited information in terms of the overall workforce diver- sity achieved by a police agency (e.g., Guajardo, 2015a). By contrast, the applica- tion of diversity indices (e.g., Simpson’s diversity index) allows for the assessment of the overall level of heterogeneity within and between police agencies by obtain- ing a single measure that captures the probability that individuals will be similar in terms of their social characteristics (e.g., Guajardo, 2015a; Simpson, 1949).

The second purpose of this study is to apply QR3 to policing data to illustrate its utility in assessing the relationships between a set of organizational-based covari- ates (i.e., predictors) and workforce diversity.4 While QR has been used in econom- ics extensively and in other disciplines since its introduction in 1978 (e.g., Davino, Furno, & Vistocco, 2014; Fitzenberger et al., 2002; Koenker, 2005; Talbert & Cade, 2013), few studies have used QR to analyze criminal justice-related issues (e.g., Berk, 2011; Britt, 2009; DeLisi et al., 2011). In contrast to prior studies which have

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 255

assessed ethnic and gender diversity in policing with mean-centered regression methods (e.g., Felkenes et al., 1993; Martin, 1991; Warner et al., 1989; Zhao & Lov- rich, 1998), this study uses QR to assess the effect of organization-based predictors across different values of diversity along the distribution of low to high scores. More specifically, QR is used to test whether the predictors in the theoretical framework have a consistent effect across the conditional distribution of the diver- sity scores as obtained by the Simpson index (see Figure 1). QR accomplishes this by examining the effect of the predictors on the response variables across specific quantiles5 of the conditional distribution (e.g., 10th, 25th, or 90th quantile). As DeLisi et al. (2011, p. 80) observe, QR permits the examination of “the relationship between a set of independent variables and the different parts of the distribution of the dependent variable.” This type of analysis is not possible with the use of OLS or other linear regression techniques which focus on the mean (e.g., Davino, Furno, & Vistocco, 2014; Davino, Romano, & Naes, 2015; Fitzenberger et al., 2002; Hao & Naiman, 2007; Koenker, 2005; Koenker & Hallock, 2001). QR also over- comes some of the disadvantages of regression methods centered on central ten- dencies (see Davino, Romano, & Naes, 2015; DeLisi et al., 2011; Hao & Naiman, 2007). In this study, QR complements the findings obtained by Tobit regression (TR) by providing a more in depth analysis of how the organization-based

Figure 1. Theoretical framework of workforce diversity in policing organizations. Source. Adapted from Guajardo, S. A. (2014c).

256 S. A. GUAJARDO

covariates affect ethnic diversity at different sections and values of the distribution of the Simpson scores. The use of QR in this study also provides statistical support with respect to whether the theoretical framework guiding the empirical analysis holds at the different quantiles of the diversity scores that are examined.

Theoretical framework

This study is based in the theoretical framework that factors within police agencies determine the extent to which they hire and promote minority men and women into and within their workforces (see Figure 1; e.g., Guajardo, 2014c, 2015c; Gustafson, 2013; Warner et al., 1989; Zhao, He, & Lovrich, 2006; Zhao, Herbst, & Lovrich, 2001). At the organizational level (i.e., macro level), internal organiza- tion-based covariates (e.g., the type of workforce and the size of the workforce) effect the rate to which minority men and women are hired into a police agency and effect the level of ethnic diversity (e.g., Guajardo, 2014c & 2015c; Gustafson, 2013; Warner et al., 1989; Zhao, He, & Lovrich, 2006; Zhao, Herbst, & Lovrich, 2001). Within the hierarchical structure of police agencies (i.e., micro level), inter- nal organization-based covariates also effect the integration of minority men and women into supervisor and command positions (e.g., Bolton, 2003; Guajardo, 2015c; Gustafson, 2013). As illustrated by Figure 1, this study examines whether ethnic diversity within the NYPD’s hierarchical structure is a function of the type of workforce within the agency, the types of police and civilian positions that exist in each workforce, or the size of each workgroup (i.e., total number of employees).

Type of police workforce

Few studies on diversity in policing have examined whether differences exist between police and civilian workforces in terms of whether demographic heteroge- neity can be attributed to the type of workforce. In contrast to the majority of the prior studies on workforce diversity in policing, Guajardo (2014b) examined changes in the level of ethnic heterogeneity in the NYPD’s police and civilian workforces from 2000 to 2012. When the level of ethnic diversity in the police and civilian workforces were compared, the police force had a greater level of ethnic heterogeneity in comparison to the civilian workforce. This study also found that the level of ethnic diversity in the police workforce increased by 24% over the 12- year period due to a decrease in White employment and an increase in minority employment. When the change in the level of ethnic diversity in the police force was analyzed by gender, ethnic heterogeneity among male police officers increased by 29% from 2000 to 2012; by contrast, ethnic heterogeneity among female police officers increased by 3% during the same time period. In the civilian workforce, the increase in ethnic diversity was substantially lower for the workforce (3%) as well as for the male (5%) and female (1%) civilian workers. A follow-up study examined employment data for 2012 with respect to whether the type of workforce (police vs. civilian workforce) was related significantly with demographic

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 257

heterogeneity in the NYPD (Guajardo, 2014c). The findings revealed that the level of ethnic diversity was related positively with the police force. Although ethnic diversity was found to be related positively with the type of workforce in the NYPD (Guajardo, 2014c), the research did not examine specifically whether the positive relationship between the type of workforce in the NYPD and ethnic diver- sity held true for both men and women because the scores were not analyzed sepa- rately by gender.

In this study, it is anticipated that the level of workforce diversity is related posi- tively with the police force when ethnicity and gender are used concurrently to measure demographic heterogeneity. Specifically, three hypotheses are examined. The first hypothesis relates to the level of ethnic diversity for men and women combined (i.e., agency-wide diversity). This hypothesis is as follows:

H1: Ethnic diversity is related positively with the NYPD police force when the level of het- erogeneity is compared to the civilian workforce.

The second hypothesis pertains to the level of ethnic diversity among male police officers and civilians. This hypothesis is as follows:

H1A: When the level of ethnic heterogeneity among male police officers and civilians is compared between the police and civilian workforces, ethnic diversity is related positively with the police force.

The hypothesis pertaining to the analysis of ethnic diversity among women offi- cers and civilians is as follows:

H1B: When the level of ethnic heterogeneity among women police officers and civilians is compared between the police and civilian workforces, ethnic diversity is related inversely with the police force.

Hierarchical structure

The hierarchical structure within policing agencies also affects the level of work- force diversity within and between police ranks and civilian job classifications (e.g., Bolton, 2003; Guajardo, 2014a). Specifically, studies on policing and ethnic and gender integration have found that the employment and promotion of minority men and women officers is related inversely with police ranks (e.g., Felkenes et al., 1993; Guajardo, 2014a, 2015b; Gustafson, 2013; Harrington & Lonsway, 2004; Martin, 2004; Schulz, 2004; Townsey, 1982). Gustafson (2013) observes that opportunities for advancement in policing are rare. In examining minority diver- sity in managerial police positions, Gustafson (2013) found that the promotion of minorities to higher positions is independent of the percentage of minorities in subordinate positions.6 Specifically, Gustafson (2013, p. 729) states that “the per- centage of minority line police is not consistently associated with minority promo- tion.” Gustafson also notes that the promotion of minorities to managerial police positions is independent of the ethnicity of the police chief. In addition, Gustafson

258 S. A. GUAJARDO

found that the number of Black police managers declines when the proportion of Latino police managers is greater. According to Gustafson (2013), the lack of minority police managers may be due to the small number of managerial positions that are available. Building on Gustafson’s research, Guajardo (2014c) examined the relationship between supervisor positions in the NYPD and ethnic diversity. The study found an inverse relationship between police and civilian supervisor positions and ethnic diversity.7

With respect to the promotion of women police officers to supervisor and com- mand positions, the research shows that women are underrepresented in the upper ranks of the police hierarchy. For instance, Townsey (1982) found an inverse rela- tionship between police rank and the number of White and Black female officers who ascended to supervisor or command positions within their agencies. Felkenes et al. (1993) observed that the Los Angeles Police Department (LAPD) had no female officers in the top 4 ranks from 1980 to 1990. They also observed that by 1990 2.5% of supervisors in the LAPD were women (Felkenes et al., 1993). Several recent studies also show that women officers are underrepresented in supervisor and command police positions in the NYPD (e.g., Guajardo, 2014a, 2015b,c).

Research specific to the NYPD has found that supervisor and command posi- tions within the police and civilian workforces are generally less ethnically diverse in comparison to lower-level positions (e.g., Guajardo, 2014a,c, 2015b). A signifi- cant inverse relationship was found between NYPD police and civilian job classifi- cations and ethnic diversity (e.g., Guajardo, 2014c). When the 2012 NYPD employment data for men and women police officers were combined, the police classification of detective second grade had a diversity score of 0.57; by contrast, the diversity score for the rank of lieutenant was 0.43 (Guajardo, 2014a). For the rank of captain, a diversity score of 0.32 was obtained. A similar trend was observed for men and women police officers when separate analyses where con- ducted by gender (e.g., Guajardo, 2014a, 2015b).

As presented above, the descriptive and inferential studies on ethnic diversity in policing have found that workforce heterogeneity is inversely related with the hier- archical structure of police agencies. Based on the findings of the extant research, it is hypothesized that ethnic diversity in the NYPD is inversely related to its hierar- chical structures in the police and civilian workforces. The hypotheses for the diversity scores based on ethnicity and gender for command, supervisor, and police officer positions are as follows:

H2: Ethnic diversity is related inversely with command positions in the NYPD.

H3: Ethnic diversity is related positively with supervisor positions in the NYPD.

H4: Ethnic diversity is related positively with noncommand and nonsupervisor police positions in the NYPD.

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 259

The hypotheses associated with the diversity scores for male police officers and civilians are the following one:

H2A: Ethnic diversity among male police officers and civilians is related inversely with command positions in the NYPD.

H3A: Ethnic diversity among male police officers and civilians is related positively with supervisor positions in the NYPD.

H4A: Ethnic diversity among male police officers and civilians is related positively with noncommand and nonsupervisor police positions in the NYPD.

The hypotheses for the women police and civilian workforces are the following ones:

H2B: Ethnic diversity among women police officers and civilians is related inversely with command positions in the NYPD.

H3B: Ethnic diversity among women police officers and civilians is related inversely with supervisor positions in the NYPD.

H4B: Ethnic diversity among women police officers and civilians is related positively with noncommand and nonsupervisor police positions in the NYPD.

Workgroup size

Finally, demographic diversity in police agencies is influenced by the size of the police agencies and the workgroups within their police and civilian workforces (e.g., Guajardo, 2014a–c, 2015a–c; Gustafson, 2013; Hochstedler & Conley, 1986; Martin, 1991; McCluskey & McCluskey, 2004; Poulos & Doerner, 1996; Warner et al., 1988). The rationale for using agency size as a covariate is based on the assumption that large organizations will generally have more job openings that may be filled by minority men and women in comparison to smaller agencies. The research has obtained mixed results for agency size and ethnic and gender integra- tion into policing. Hochstedler & Conley (1986), for instance, found that depart- ment size had a minor effect on the employment of Black women officers but had no effect on the employment of White women. Martin (1991) detected a positive relationship between department size and the percent of female officers. Poulos & Doerner (1996) obtained a similar result between department size and the percent of female officers. However, when McCluskey & McCluskey (2004) examined the relationship between department size and the percent of Hispanic officers in municipal police agencies, they failed to find a significant result among these varia- bles. By contrast, Gustafson (2013) found that large police agencies tended to have more minority managers. In a departure from prior studies, the size of police and civilian workgroups has been found to have a positive effect on integration and

260 S. A. GUAJARDO

heterogeneity (e.g., Guajardo, 2014c). Based on prior studies, it is hypothesized that the size of the workgroup (i.e., the total number of employees within a specific job title or classification) is related positively with ethnic diversity in the NYPD.

Briefly, the hypotheses associated with the total number of employees in each NYPD police and civilian job title or classification (i.e., workgroup) are the follow- ing ones:

H5: Ethnic diversity is related positively with the total number of employees in each job title or classification.

H5A: Ethnic diversity among male police officers and civilians is related positively with the total number of employees in each job title or classification.

H5B: Ethnic diversity among women police officers and civilians is related positively with the total number of employees in each job title or classification.

All of the hypotheses are tested at a D 0.05.

Statistical methods

The review of prior research shows that researchers have used logistic (e.g., McCluskey & McCluskey, 2004), OLS (e.g., Felkenes et al., 1993; Martin, 1991; Warner et al., 1989; Zhao & Lovrich, 1998), Poisson (e.g., Gustafson, 2013), or Tobit (e.g., Poulos & Doerner, 1996) regression models to assess the integration of minorities and women into policing. The use of logistic regression is a special case where McCluskey & McCluskey (2004) used a binary dependent variable. In the study conducted by Gustafson (2013), the use of Poisson regression was based on the use of count data and on assumption violations for the use of OLS regression. While Felkenes et al. (1993), Martin (1991), and Zhao & Lovrich (1998) used OLS regression to analyze the percent of minorities or women officers in police agen- cies, Poulos & Doerner (1996) used TR to analyze gender equality in policing. Their use of TR was due to the restricted distribution of the dependent variables which prohibited the use of OLS regression. As explained in the methods section, TR is used in this study because the Simpson diversity index produces scores with a restricted distribution which ranges from 0 to 1 (i.e., 0 < D < 1). To assess how the independent variables affect the diversity scores along its conditional distribu- tion, QR is used to supplement the Tobit analyses.

As stated above, there is a shortage of studies on diversity in policing which have used QR to assess whether the relationship among the predictors of ethnic hetero- geneity and the level of diversity holds across the distribution of values of the response variable (i.e., diversity scores). While a thorough technical discussion of QR is beyond the scope of this article, it is beneficial to discuss in general terms how QR differs from linear regression models (LRMs). First, QR models (QRMs) focus on how a set of covariates affect the conditional median and the values of the

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 261

response variable at various quantiles (t, where 0 < t < 1), whereas LRMs focus on how a set of predictors effect the conditional mean (e.g., Davino, Furno, & Vis- tocco, 2014; Hao & Naiman, 2007; Koenker, 2005; Talbert & Cade, 2013; Yu et al., 2003). Second, in contrast to LRMs, QRMs use least absolute distance (or value) estimation instead of least squares estimation (Davino, Furno, & Vistocco, 2014; Hao & Naiman, 2007; Koenker, 2005; Talbert & Cade, 2013; Yu et al., 2003). The absolute distance is defined as │Y - m│ where Y represents the values of the response variable and m represents the median (e.g., Hao & Naiman, 2007). Hao & Naiman (2007, p. 17) state that one can “measure how far Y is from m by the absolute distance │Y - m│ and measure the average distance in [a] population by the mean absolute distance E│Y - m│.” A similar process is followed for samples (e.g., Hao & Naiman, 2007; Yu et al., 2003). To obtain the value (q) that minimizes the mean absolute distance at a particular quantile, the following formula is used (Hao & Naiman, 2007): Y:E[qy(Y,q)]. With respect to obtaining the beta (b) coeffi- cients that minimize the sum of absolute residuals, QRMs utilize the following formula (e.g., Hao & Naiman, 2007): Eijyi – b0 – b1xij. In addition to these differ- ences, QRMs differ from LRMs with respect to assessing shifts in the slopes, shifts in the scale, and shifts in skewness (e.g., Davino, Furno, & Vistocco, 2014; Hao & Naiman, 2007; Koenker, 2005). Finally, in contrast to LRMs, which assume that the errors are normally distributed, QRMs are distribution free and account for the non-normal distribution of error terms (e.g., Britt, 2009; Hao & Naiman, 2007; Koenker, 2007). DeLisi et al. (2011, p. 81) summarize QR as a “robust regression technique that accounts for the non-normal distribution of error terms and heteroskedasticity.”

Similar to LRMs, QRMs produce intercepts and beta coefficients. The difference is that QRMs produce these coefficients for each quantile specified in the analysis (e.g., 10th, 50th, or 75th quantile). For each quantile, the intercept and beta coeffi- cients are interpreted in a manner similar to those produced by LRMs (Brit, 2011; Cade & Noon, 2003; Davino, Romano, & Naes, 2015; Hao & Naiman, 2007). As Brit (2011, 357) states, “The quantile regression coefficients represent the effect of a one-unit change in the independent variable on the dependent variable at the t quantile” (e.g., 5th, 25th, or 90th quantile). When a predictor is coded dichoto- mously (e.g., 0,1), the regression coefficient is interpreted as the change in the condi- tional quantile brought about by changing from one group to another (e.g., Hao & Naiman, 2007). In this study, the coefficient corresponding to command and other positions would be interpreted as the change in the conditional diversity score at t quantile brought about by changing from noncommand positions (0) to command positions (1). With respect to interpreting the intercepts, the intercepts represent the estimated conditional quantile function of the distribution of the response variable and its covariates at each quantile specified in the study (e.g., Koenker & Hallock, 2001).

Statistically, QRMs have several advantages over LRMs. First, the use of condi- tional medians and least absolute deviations make QRMs less sensitive to outliners

262 S. A. GUAJARDO

(e.g., Davino, Furno, & Vistocco, 2014; Hao & Naiman, 2007; Koenker, 2007). In addition, QRMs are more robust in analyzing distributions that are moderately to highly skewed in comparison to LRMs (e.g., Davino, Furno, & Vistoco,2014; Hao & Naiman, 2007; Koenker, 2007). Hao and Naiman (2007, p, 3) state, “[when] the distribution is highly skewed, the mean can be challenging to interpret while the median remains highly informative.” QRMs also have the ability to assess how a set of covariates effect the response variable at different points of the conditional distribution, whereas LRMs focus on how different values of the predictor effect the value of the conditional mean (e.g., Davino, Furno, & Vistocco, 2014; Hao & Naiman, 2007; Koenker, 2007). Davino, Furno, & Vistocco (2014) observe that QRMs are able to detect more effects than LRMs because the former do not restrict the analysis to the conditional mean, which, in turn, permits the entire conditional distribution of the response variable to be approximated. Another advantage of QRMs is that the various intercepts and coefficients (i.e., slopes) may be tested for equivalence (e.g., Britt, 2009; Davino, Furno, & Vistocco, 2014; Hao & Naiman, 2007; Koenker, 2003; Talbert & Cade, 2013). For instance, if the errors of the vari- ous predictors are homogeneous, the coefficients will be similar across the quan- tiles, but the intercepts will differ (e.g., Talbert & Cade, 2013). Britt (2009) notes that there are two tests available to determine the stability of the coefficients: loca- tion shift and location-scale. The location shift test assesses whether the coeffi- cients may be summarized by a single value; the location-scale test, on the other hand, assesses the extent to which the coefficients are affected by nonconstant errors (e.g., Hao & Naiman, 2007; Talbert & Cade, 2013). An assessment of the shift in the skewness of the distribution is also possible with the use of QRMs (e.g., Hao & Naiman, 2007).

Although QRMs are able to assess the relationships between predictors and a response variable across different quantiles, QRMs have several drawbacks. First, sample size influences the spread of the estimates and the distance between the quantiles examined in a study (e.g., Davino, Furno, & Vistocco, 2014). Davino, Furno, & Vistocco, (2014, 57) observe that estimates tend to be more scattered for small sample sizes (e.g., n D 30). In addition, the structure of the dataset affects the number of distinct quantile solutions (e.g., Davino, Furno, & Vistocco, 2014). A second drawback of QRMs is that the medians may be affected by atypical cases included in the sample. For instance, Maclean et al. (2014) found that the condi- tional QR (CQR) model produced inconsistent estimates at the lower quantiles when examining the impact of cigarette taxes on smoking consumption. Another weakness of QRMs is that the interpretation of the coefficients changes when dif- ferent sets of predictors (or covariates) are included in the model (Maclean et al., 2014). The interpretation of the coefficients also becomes more difficult when data transformations are performed on the raw scores (e.g., Hao & Naiman, 2007). In addition, some QRMs produce quantile curves that may cross one another (Yu et al., 2003). However, as noted above, QRMs are able to perform in depth statisti- cal analyses despite their drawbacks when contrasted with LRMs.

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 263

Measurement of ethnic and gender integration

With the exception of McCluskey & McCluskey (2004) and Gustafson (2013), eth- nic and gender integration have been measured quantitatively. Specifically, prior research has used a percentage of Black, minority, or women as dependent varia- bles (e.g., Felkenes et al., 1993; Martin, 1991; Warner et al., 1989; Zhao & Lovrich, 1998). While these operationalizations of the dependent variable are useful for assessing the integration of specific groups, they limit the ability to assess the over- all level of ethnic and gender integration and diversity in police agencies (e.g., Gua- jardo, 2015a). A recent study on ethnic diversity in policing suggested the use of diversity indices to assess whether the ethnic or gender composition of police agen- cies is changing (e.g., Guajardo, 2015a). This study departs from prior studies and applies the Simpson index of diversity (D D 1 – Sp2) to NYPD employment data for January 2015 to obtain several measures of workforce diversity. As stated above, Tobit and QR are applied to these dependent variables to illustrate the use- fulness of QR in analyzing ethnic diversity in policing and to test whether the rela- tionships among the covariates and workforce diversity hold across the conditional distribution of the heterogeneity scores obtained by the Simpson index.

Diversity in the NYPD

Table 1 summarizes the level of ethnic diversity in the NYPD police force for Janu- ary 2015. The Simpson scores for ethnicity and gender (DEG) suggest an inverse relationship between police rank and the level of ethnic heterogeneity. When eth- nicity and gender are considered jointly, the Simpson scores indicate that minority men and women are poorly represented at the command positions. The DE scores also suggest that police ranks tend to be more homogenous. Additionally, the level of ethnic homogeneity tends to be greater in the supervisor and command ranks. When the level of ethnic diversity among women officers is measured, the DFE scores reveal a relatively equal distribution of White and minority women across the police ranks; however, there are no women officers who serve as Bureau Chiefs, Assistant Chiefs, or Surgeons. In terms of the level of ethnic diversity among male officers, the DME scores indicate a greater level of ethnic homogeneity. Specifically, there are few minority male officers in comparison to White male officers. The diversity scores also indicate that there are fewer minority male officers propor- tionally in supervisor and command positions in comparison to minority women officers in similar positions. Put differently, the ethnic distribution of minority males in supervisor and command positions varies more in comparison to the eth- nic distribution of minority women in similar positions.

The Simpson scores obtained for the NYPD civilian workforce reveal that the job classifications tend to have similar levels of diversity with the exception of the skilled crafts category (see Table 2). In terms of the ethnic and gender diversity, the DEG scores indicate that the proportions of White and minority men and women are similar within and between job classifications with the exception at the

264 S. A. GUAJARDO

Ta b le 1.

Em p lo ym

en t d at a an d Si m p so n d iv er si ty sc or es

fo r th e N YP D p ol ic e fo rc e fo r Ja nu ar y 20 15 .

W hi te

Bl ac k

H is p an ic

A si an

N at iv e A m er ic an

Si m p so n d iv er si ty sc or es

Ra nk

M al e

Fe m al e

M al e

Fe m al e

M al e

Fe m al e

M al e

Fe m al e

M al e

Fe m al e

To ta l

D M al e s

D Fe m al e s

D Et h n ic it y

D G E

Bu re au

C hi ef s

6 1

2 0

1 0

1 0

0 0

11 0. 58 0

0. 00 0

0. 54 5

0. 29 0

A ss is ta nt

C hi ef s

18 3

0 0

2 0

0 0

0 0

23 0. 18 0

0. 00 0

0. 15 9

0. 09 0

D ep ut y C hi ef

42 3

6 3

4 0

0 0

0 0

58 0. 32 8

0. 50 0

0. 36 9

0. 41 4

In sp ec to r

97 2

3 1

5 2

1 1

0 0

11 2

0. 15 9

0. 72 2

0. 21 3

0. 45 5

D ep ut y In sp ec to r

13 1

6 7

3 21

1 3

0 0

0 17 2

0. 32 7

0. 54 0

0. 34 6

0. 43 4

C ap ta in

30 7

19 25

10 36

7 9

0 0

0 41 3

0. 32 3

0. 60 6

0. 35 8

0. 46 5

Su rg eo n

30 1

1 0

1 0

1 0

0 0

34 0. 17 1

0. 00 0

0. 16 6

0. 08 5

Li eu te na nt

(L T) C O D et ec ti ve

Sq ua d

11 0

3 4

2 15

2 1

0 0

0 13 7

0. 27 0

0. 65 3

0. 30 2

0. 46 1

LT Sp ec ia lA

ss ig nm

en t

65 9

7 5

12 2

2 0

0 0

10 2

0. 40 2

0. 57 0

0. 44 1

0. 48 6

Li eu te na nt

93 3

75 11 6

49 21 3

40 76

11 0

0 1, 51 3

0. 47 8

0. 68 2

0. 51 3

0. 58 2

Se rg ea nt

Su p er vi so r D et ec ti ve

20 9

10 41

10 55

16 13

0 0

0 35 4

0. 52 0

0. 64 8

0. 55 5

0. 58 4

Se rg ea nt

Sp ec ia lA

ss ig nm

en t

13 1

14 10

11 24

4 9

2 0

0 20 5

0. 40 8

0. 64 9

0. 46 8

0. 53 1

Se rg ea nt

2, 01 2

22 1

39 2

25 9

71 5

20 5

27 9

19 4

1 4, 10 7

0. 58 6

0. 68 1

0. 62 4

0. 63 4

D et ec ti ve

Fi rs t G ra d e

21 5

12 45

13 52

8 7

0 0

0 35 2

0. 49 9

0. 65 4

0. 52 8

0. 57 6

D et ec ti ve

Se co nd

G ra d e

54 0

29 11 1

41 18 5

39 22

1 2

0 97 0

0. 54 2

0. 66 6

0. 57 7

0. 60 4

D et ec ti ve

In ve st ig at or

1, 33 6

13 9

34 1

84 66 9

12 4

79 7

2 1

2, 78 2

0. 60 0

0. 66 8

0. 61 3

0. 63 4

(c on tin ue d )

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 265

Ta b le 1.

(C on tin ue d )

W hi te

Bl ac k

H is p an ic

A si an

N at iv e A m er ic an

Si m p so n d iv er si ty sc or es

Ra nk

M al e

Fe m al e

M al e

Fe m al e

M al e

Fe m al e

M al e

Fe m al e

M al e

Fe m al e

To ta l

D M al e s

D Fe m al e s

D Et h n ic it y

D G E

D et ec ti ve

Sp ec ia lis t

39 8

51 84

39 12 9

41 14

5 0

0 76 1

0. 53 3

0. 68 5

0. 57 5

0. 61 0

Po lic e O ffi ce r

9, 72 8

1, 17 7

2, 40 0

1, 29 6

5, 04 4

1, 69 8

2, 54 4

11 9

11 2

24 ,0 19

0. 66 0

0. 67 6

0. 65 1

0. 66 9

A ve ra g e

0. 42 0

0. 53 3

0. 44 5

0. 47 8

St an d ar d d ev ia ti on

0. 15 9

0. 25 2

0. 15 9

0. 17 1

266 S. A. GUAJARDO

executive level and the skilled crafts classification. The DE scores also reveal that the level of ethnic diversity is lower at the executive level in comparison to other civilian positions. In terms of the level of ethnic diversity among women employ- ees, the DFE scores show that the level of ethnic diversity is lower among supervisor and clerical positions. Among male civilians, the DME scores reveal that the level of ethnic diversity is significantly lower at the executive level in comparison to other civilian job classifications.

Research design

Similar to Carter (1986), Felkenes et al. (1993), Felkenes & Schroedel (1993), Guajardo (2014a–c, 2015a,b), Raganella & White (2004), and Winters (1991) who have analyzed a single police agency, this study examines how well minority men and women are integrated within the hierarchical structure of a single police agency. Specifically, this study examines how internal organization-based covari- ates influence ethnic diversity within the NYPD’s police and civilian workforces (see Figure 1). In addition, this study includes employment data for the NYPD’s police and civilian workforces and their job titles and classifications to assess whether the type of workforce affects the level of ethnic diversity within the organ- ization’s hierarchical structure (e.g., Guajardo, 2014b,c, 2015a). While this research design is consistent with prior research, this study differs slightly from previous research in that it analyzes ethnic diversity between the NYPD’s police and civilian workforces simultaneously (e.g., Guajardo, 2014c). In so doing, this study provides an opportunity to assess differences in levels of ethnic diversity between a bottom- up8 police workforce and a more fluid civilian workforce more directly.

A second reason for focusing solely on the NYPD is that the agency has amended its EEOC reports since 2014 by including more detailed employment data for its police and civilian workforces. In contrast to other municipal police agencies that do not provide detailed employment data by rank (e.g., Guajardo, 2015a,2015c; Riccucci & Saldivar, 2014), the NYPD’s EEOC reports contain data for each police rank and civilian job classification. Because there is a lack of com- patible policing-based employment data for major municipal police agencies in the United States (e.g., City of Chicago Police Department, Los Angeles Police Depart- ment), a comparative analysis of ethnic diversity in policing is not possible. As Riccucci & Saldivar (2014) observe, incomplete police employment data by rank makes it difficult to assess the progress made by women and people of color in police agencies. Methodologically, the lack of readily available detailed police and civilian employment data for large municipal police agencies prohibits an in depth comparative analysis of diversity between the NYPD and its peers. Despite the inability to conduct a comparative analysis for major police agencies in the United States, the use of police and civilian employment data by rank and job classification for the NYPD allows for an in depth analysis of how internal organization-based factors affect ethnic diversity between police and civilian workforces. In addition,

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 267

Ta b le 2.

Em p lo ym

en t d at a an d Si m p so n d iv er si ty sc or es

fo r th e N YP D ci vi lia n w or kf or ce

fo r Ja nu ar y 20 15 .

W hi te

Bl ac k

H is p an ic

A si an

N at iv e A m er ic an

Si m ps on

d iv er si ty sc or es

Jo b C la ss ifi ca ti on

M al e

Fe m al e

M al e

Fe m al e

M al e

Fe m al e

M al e

Fe m al e

M al e

Fe m al e

To ta l

D M al e s

D Fe m al e s

D Et h n ic it y

D G E

Ex ec ut iv es

15 9

3 3

0 2

1 0

0 0

33 0. 34 9

0. 52 0

0. 43 3

0. 43 5

A d m in is tr at or s

71 47

15 29

15 12

7 10

0 1

20 7

0. 52 5

0. 66 4

0. 60 6

0. 59 9

Pr of es si on al s

13 7

21 5

42 17 0

17 61

41 73

0 0

75 6

0. 59 9

0. 68 8

0. 67 1

0. 65 3

Te ch ni ci an s

10 3

32 36

33 23

10 46

10 0

0 29 3

0. 66 4

0. 68 0

0. 68 3

0. 67 9

Pr ot ec ti ve

Se rv ic es

Su p er vi so rs

36 23

12 3

21 8

38 52

69 11

2 0

57 2

0. 68 5

0. 44 9

0. 59 0

0. 56 8

Pr ot ec ti ve

Se rv ic es

30 8

82 7

15 41

4, 16 2

79 9

1, 82 4

68 2

18 5

5 16

10 ,3 49

0. 67 9

0. 56 6

0. 61 3

0. 62 3

Pa ra -p ro fe ss io na ls

8 4

5 8

0 1

3 2

0 0

31 0. 61 7

0. 62 2

0. 64 7

0. 62 9

O ffi ce

& C le ri ca lS up

er vi so rs

21 19 5

40 94 5

9 16 1

18 32

0 1

1, 42 2

0. 68 4

0. 46 2

0. 48 2

0. 57 3

O ffi ce

& C le ri ca ls

70 27 3

14 7

1, 77 6

44 42 4

54 95

2 2

2, 88 7

0. 68 8

0. 48 3

0. 51 3

0. 58 6

Sk ill ed

C ra ft s

34 9

0 55

1 71

0 47

0 3

0 52 6

0. 52 1

0. 00 0

0. 52 2

0. 26 0

Se rv ic e & M ai nt en an ce

68 12

11 2

95 78

83 27

6 2

0 48 3

0. 70 9

0. 58 1

0. 67 3

0. 64 5

A ve ra g e

0. 61 1

0. 51 9

0. 58 5

0. 56 8

St an d ar d d ev ia ti on

0. 10 9

0. 19 3

0. 08 5

0. 12 1

268 S. A. GUAJARDO

the use of the NYPD police and civilian employment data for January 2015 assists to assess the reliability of the findings obtained by Guajardo (2014a,c, 2015c). Although the examination of ethnic diversity at the organizational level provides information with respect to how well a police agency integrates minority men and women into its police and civilian workforces, an analysis at the macro level would provide limited information with respect to ethnic diversity between and within job titles and classifications. A macro analysis of ethnic diversity also limits the ability to assess how well men and women of different ethnic backgrounds are dis- tributed within and between workforces and their job titles and classifications.

The third reason for restricting the analysis to the NYPD is that the agency has undergone a number of internal organizational changes over the past several years. As discussed above, the NYPD has modified its EEOC reports by expanding the police and civilian job classifications by adding new titles such as surgeon and by disaggregating job classifications such as executives and administrators to distin- guish between the two. By expanding and disaggregating the job classifications and titles contained in its EEOC reports, more detailed employment data by rank and job classification are available for analyses. For instance, the disaggregation of the executive/administrators job classification into two categories (i.e., executives and administrators) expands the number of job classifications that are categorized as command and senior management positions. The creation of these two categories also helps to assess the level of workforce diversity within these job classifications which was not possible prior to 2015. Similarly, the creation of the protective serv- ices supervisor and office and clerical supervisor classifications expands the num- ber of supervisor positions in the civilian workforce and allows for assessing ethnic diversity within these categories. Although these reporting changes enrich the qual- ity of the employment data and facilitate conducting more sophisticated statistical analyses of the NYPD’s hierarchical structure and its impact on ethnic heterogene- ity, they also have the unintended consequence of changing the reporting structure so that the data are not compatible to prior years and further limit the types of comparative analyses that may be conducted with other municipal police agencies.

Labor changes in the NYPD’s police and civilian workforces which have occurred since 2000 also prohibit undertaking a comparative analysis of ethnic diversity with other municipal police agencies in the United States. Over the past 14 years, the total number of police positions have decreased as the number of positions for select police ranks have increased (e.g., Guajardo; 2014b, 2015b). With respect to the NYPD’s civilian workforce, the total number of civilian posi- tions has increased since 2000 as the total number of police positions has declined (e.g., Guajardo 2014b, 2015c). In addition, while the police force has shrunk and the civilian workforce has expanded, the number of minority hires increased as the number of White employees decreased (e.g., Guajardo, 2014b, 2015b). Further, the number of minority men and women who have been promoted to higher positions has increased since 2000 (e.g., Guajardo, 2014b, 2015b,c). While these changes in the NYPD labor force have changed the hierarchical structure and have increased

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 269

the demographic heterogeneity of the police and civilian workforces, the changes may be significant enough to question whether it is methodologically appropriate to compare the NYPD to other large municipal police agencies that have not expe- rienced similar labor changes and that have had relatively stable workforce envi- ronments during the past 14 years.

Data and method

The data used in this study are based on the NYPD EEOC report for January 2015, which contains demographic statistics for men and women in the police and civil- ian workforces (see Tables 1 and 2). Demographic data are available for each police rank and job title from police officer to bureau chief. In 2014, the NYPD began providing employment data pertaining to the position of surgeon in the police force. For the civilian workforce, data are provided for each job category from ser- vice and maintenance to executives (see Table 2). In 2014, the job classifications in the civilian workforce were expanded to include protective services and office and clerical supervisors. As stated above, the job classification of Executives/Adminis- trators was reclassified into two categories (executives and administrators) and took effect in 2014. These reporting changes provide more employment data and allow for assessing whether the level of diversity is associated positively or nega- tively with supervisor and command positions.

Independent variables

Similar to the studies cited above, this study uses organization-based independent variables (IVs). The first IV is the type of workforce (WF) where job titles and ranks in the police force are coded as 1, and those in the civilian workforce are coded as 0. This coding allows for assessing whether higher or lower levels of diver- sity are associated with the police or civilian workforce (e.g., Guajardo, 2014c). The second IV is command positions (CP). In the police force, the positions of captain to bureau chief are assigned a 1, and other positions are assigned a 0. Similarly, in the civilian workforce, executive and administer positions (CP) are coded as 1, and other positions are coded as 0. This coding allows for examining whether work- force diversity is associated positively or negatively with command positions. Third, supervisor positions (SP) in the police force (i.e., sergeants and lieutenants) are assigned a 1, and other positions are assigned a 0. In the civilian workforce, the job classifications of protective services supervisors and office and clerical supervi- sors are coded as 1, and other positions are coded as 0. Policing-based (PO) posi- tions (i.e., police officers, detectives, and protective services) are assigned a 1, and other police and civilian positions are assigned a 0. The final IV is workgroup size (WGS) which is measured by the total number of employees in each police and civilian job title or classification (e.g., Guajardo, 2014c). This IV allows for assess- ing whether the level of diversity increases with an increase in the total number of employees in police and civilian job titles or classifications.

270 S. A. GUAJARDO

Dependent variables

Three dependent variables (DV) are used in this study. The first DV is DEG and measures the level of ethnic and gender diversity jointly in the NYPD police and civilian workforces. The second DV is DME and measures the level of ethnic diver- sity among male officers and civilians. DFE is the third DV, and it measures the level of diversity among the women officers and civilians. Each of these DVs was obtained by applying the Simpson diversity index to the demographic data. The Simpson diversity index is discussed below.

Simpson’s diversity index

Although the majority of the previous research on ethnic and gender integration and diversity in policing have not applied an index of diversity to assess the level workforce heterogeneity between and within agencies, recent studies have done so at a macro and micro organizational level (e.g., Guajardo, 2014a–c, 2015a,b). More specifically, Guajardo (2015a) demonstrated how the Simpson diversity index may be applied to policing agencies to obtain heterogeneity scores for their police and civilian workforces as well as for job titles and classifications within each work- force. In regards to assessing ethnic diversity in the NYPD, the Simpson index has been used extensively to assess the level of heterogeneity between and within the police and civilian workforces (e.g., Guajardo, 2014a–c, 2015c,b). Simpson’s index of diversity also has been applied to male and females cohorts within the NYPD police and civilian workforces (e.g., Guajardo 2014a,b, 2015b). Since the Simpson index has been used in previous studies to analyze the level of workforce diversity in the NYPD, this study applies the same index to NYPD employment data for January 2015 to obtain measures of heterogeneity that are consistent with the pre- vious studies (e.g., Guajardo, 2014a–c, 2015c,b).

Simpson’s (1949, D D 1 – Sp2) diversity index obtains the “probability that two individuals chosen at random and independently from [a] population will be found to belong to the same group” (p. 688). The index produces scores that range from 0 to 1 (i.e., 0 < D < 1). A score of 0 indicates the absence of heterogeneity (or diversity), and a score of 1 indicates maximum heterogeneity9 (e.g., Blau, 1977, Guajardo, 2014c, 2015a,b; Lieberson, 1969; Simpson, 1949; Starks, 2009).

Diversity scores for the NYPD may be obtained by squaring the frequencies of each group included in the analysis (i.e., f2), summing the squared frequencies (i.e.,

P f2),

and then by dividing P

f2 by the total number of employees (i.e., F) in the organization squared (i.e., F2). This product is then subtracted from 1. Symbolically, the set of calcu- lations are represented as follows (see Guajardo, 2014c, 2015a, b):

D D 1 ¡ P

f

F2

2

Alternatively, the scores may be obtained by obtaining the proportions (i.e., p) for each group in the organization and then squaring each p (i.e., p2) and summing

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 271

the products to obtain P

p2. The final step is to subtract P

p2 from 1 (i.e., D D 1 ¡ P

p2; see Table 3). When the ethnic and gender characteristics of the officers and civilians in the NYPD are used simultaneously to obtain a more comprehensive set of diversity scores for the agency, the formula for the Simpson diversity index takes the following form:10

Dv D 1 ¡ P

p2n Vn

In this formula, V represents the number of characteristics that are used to cal- culate the diversity scores. Since ethnicity and gender are used concomitantly to obtain DEG, V equals 2.

Table 3. Calculation of Simpson diversity scores for the NYPD police force for January 2015.

A. Frequency for male and female police officers

Gender

Ethnicity Male Female Total

White 16,309 1,775 18,084 Black 3,595 1,826 5,421 Hispanic 7,183 2,189 9,372 Asian 2,061 165 2,226 Native American 19 4 23 Total 29,167 5,959 35,126

B. Proportion (P) of male and females

Gender

Ethnicity PMale PFemale Pethnicity

White 0.56 0.30 0.51 Black 0.12 0.31 0.15 Hispanic 0.25 0.37 0.27 Asian 0.07 0.03 0.06 Native American 0.00 0.00 0.00 Total 1.00 1.00 1.00

C. Simpson diversity scores

Gender

Ethnicity P2Male P 2 Female P

2 Ethnicity

White 0.3127 0.0887 0.2651 Black 0.0152 0.0939 0.0238 Hispanic 0.0606 0.1349 0.0712 Asian 0.0050 0.0008 0.0040 Native American 0.0000 0.0000 0.0000 Total 0.3935 0.3183 0.3641

Simpson scores

DMales DFemales DEthnicity DGE

0.607 0.682 0.636 0.644

272 S. A. GUAJARDO

Statistical methods and models

Because the Simpson diversity scores range from 0 to 1 (i.e., 0 < D < 1), Tobit regres- sion (TR) is used to analyze the relationship between the independent and dependent variables. As Ryan (2009) notes, TR is appropriate when the actual values of the inde- pendent and dependent variables have a restricted range. For each DV presented above, the lowest and highest values of the diversity scores are used as the censoring criteria (i.e., left- and right-censored observations). In addition, robust standard errors are used because the errors of the covariates and response variables are not independent and identically distributed (iid). For instance, the errors for the size of the workgroup (i.e., WGS) and the diversity scores (i.e., DEG, DME, and DFE) are correlated to some degree because the total number of employees in each job title is used to obtain the proportion of men and women by ethnicity. The assumption of normality for each variable also is likely to be violated because of the number of observations.

As discussed above, quantile regression (QR) is used to examine whether the set of independent variables have a constant effect across the conditional distribution of the diversity scores. The QR analysis is performed at the 10th, 25th, 50th, 75th, and 90th quantiles. Although additional quantiles may be included in the analysis, the sample size limits the number of quantiles that may be used (e.g., Davino, Furno, & Vistocco, 2014). Robust standard errors are used to be consistent with the Tobit analysis discussed above.

The Tobit and QR models for the DVs are as follows: Model 1: Ethnic and gender diversity for all employees (DEG)

TR D̂EG D b0 C b1WF C b2CP C b3SP C b4PO C b5WGS QR D̂EG.u/ D b0 uð Þ C b1 uð ÞWF C b2 uð ÞCP C b3 uð ÞSP

C b4 uð ÞPO C b5 uð ÞWGS

Model 2: Ethnic diversity for male employees (DME)

TR D̂ME D b0 C b1WF C b2CP C b3SP C b4PO C b5WGS QR D̂ME.u/ D b0 uð Þ C b1 uð ÞWF C b2 uð ÞCP C b3 uð ÞSP

C b4 uð ÞPO C b5 uð ÞWGS

Model 3: Ethnic diversity for women employees (DWE)

TR D̂FE D b0 C b1WF C b2CP C b3SP C b4PO C b5WGS C b6DME QR D̂FE.u/ D b0 uð Þ C b1 uð ÞWF C b2 uð ÞCP C b3 uð ÞSP

C b4 uð ÞPO C b5 uð ÞWGS C b6 uð ÞDME

For each model, the betas are tested at a D 0.05 (i.e., HO: bn D 0 at a D 0.05).

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 273

The first model assesses whether the level of ethnic diversity (DEG) is a function of the type of WF, police and civilian job titles and classifications (i.e., CP, SP, or PO), and the total number of employees in each job title or classification (WGS). Similarly, the second model examines whether the level of ethnic diversity among men (DME) in the NYPD is a function of the same set of IVs. The third model examines the level of ethnic diversity among women officers and civilians. In this model, DME is included as an IV to assess whether the level of ethnic diversity among men has a positive or negative effect on the level ethnic diversity among women.

Data limitations

Because the NYPD added additional job titles and classifications to its police and civilian workforces in 2014 and amended its EEOC reports from prior years, the annual employment data used in this study have several limitations which restrict the types of analyses that may be performed. One limitation of the data is that the rates of hire or promotion for men and women officers and civilians are not reported by the NYPD. In turn, this prohibits the analysis of the relationship between the rates of hire or promotion and ethnic diversity. Another limitation of the data is the absence of attrition rates for men and women officers and civilians by job title. This limits the ability to assess whether the promotion of women offi- cers and civilians to supervisor and command position is related directly to attri- tion rather than position growth. Finally, since this study analyzes annual employment data provided by the NYPD, the generalizability of the findings to other police agencies is limited. Despite these limitations, the data allow for assess- ing the relationships between the internal organization-based covariates and ethnic diversity as illustrated in Figure 1.

Results

Descriptive statistics for the three sets of Simpson diversity scores for the NYPD are presented in Table 4. The distribution for the diversity scores obtained by the concurrent use of ethnic and gender characteristics of police and civilian employ- ees is asymmetrical and negatively skewed. An average diversity score of 0.5122 with a standard deviation of 0.1583 was obtained for the distribution. The test for normality and skewness produced significant results so that the assumption of nor- mality is rejected. With respect to the conditional distribution of the diversity scores, the analysis produced a score of 0.2604 at the 10th quantile. A score of 0.4547 was obtained for the 25th quantile. For the median (i.e., 50th quantile), the value is 0.5763. The values for the upper tail of the conditional distribution are 0.6225 and 0.6533 for the 75th and 90th quantiles, respectively.

When the level of ethnic diversity among male officers and civilians is exam- ined, the descriptive statistics show that the distribution is symmetrical but that the assumption of normality is still violated (see Table 4). The average diversity

274 S. A. GUAJARDO

score for the male cohort is 0.4926 with a standard deviation of 0.1683. A value of 0.1800 was obtained for the 10th quantile. For the 25th quantile, the value is 0.3490. At the median, the value is 0.5250. The values for the 75th and 90th quan- tiles are 0.6172 and 0.6849, respectively.

The descriptive statistics for the female cohort show that women of different ethnic backgrounds are distributed more equally in comparison to the men; how- ever, the findings suggest that there is more variation in the level of diversity among men and women between and within the police and civilian workforces (see Table 4). An average of 0.5281 with a standard deviation of 0.2274 was obtained for the female cohort. The tests for normality and skewness reveal that the assumption of normality is violated. At the 10th quantile, the value is 0.0001, which indicates an absence of ethnic heterogeneity at the lower tail of the condi- tional distribution of the diversity scores. A value of 0.5000 was obtained for the 25th quantile. The value for the median is 0.6222. For the 75th quantile, the value is 0.6683. At the upper tail of the conditional distribution of the diversity scores, the value is 0.6849.

Table 5 summarizes the TR and QR results for the diversity scores based on eth- nicity and gender (DEG). The TR findings show that command (b D ¡0.014, p > 0.05), supervisor (b D 0.138, p > 0.05), and policing-based positions (b D 0.202, p > 0.05) are not related significantly with ethnic diversity. A nonsignificant find- ing was obtained for workgroup size (b D 0.001, p > 0.05) and the Simpson diver- sity scores. With respect to differences in the level of ethnic diversity between the police and civilian workforces, the findings show that the civilian workforce tends to have higher levels of ethnic heterogeneity; however, this relationship is not sig- nificant at the .05 level (b D ¡0.126, p > 0.05). These results indicate that the null hypotheses for each covariate should be retained.

Although the TR analysis failed to find any significant relationships between the covariates and DEG, the QR analysis shows that the effects of the predictors on DEG change along the conditional distribution of the diversity scores (see Table 5). At the 10th quantile, the findings shows a significant inverse relationship between

Table 4. Summary statistics for the Simpson diversity scores.

Simpson diversity scores

Descriptive Statistics DGE DMales Dfemales

10th quantile 0.2604 0.1800 0.0001 25th quantile 0.4547 0.3490 0.5000 50th quantile (Median) 0.5763 0.5250 0.6222 75th quantile 0.6225 0.6172 0.6683 90th quantile 0.6533 0.6849 0.6849 Mean 0.5122��� 0.4926� 0.5281���

Standard Deviation 0.1583 0.1683 0.2274 Skewness ¡1.4482�� ¡0.5803 ¡1.6823��� Kurtosis 4.4090 2.2005 4.4036 N 29 29 29

�Significant at a D .05; ��Significant at a D .01; ���Significant at a D .001.

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 275

Ta b le 5.

To b it an d q ua nt ile

re g re ss io n re su lt s fo r Si m p so n d iv er si ty sc or es

fo r g en d er an d et hn ic it y fo r Ja nu ar y 20 15

(S ta nd

ar d er ro rs ar e in p ar en th es es ).

Q ua nt ile

Re g re ss io n

To b it Re g re ss io n

10 th

Q ua nt ile

25 th

Q ua nt ile

50 th

Q ua nt ile

75 th

Q ua nt ile

90 th

Q ua nt ile

So ur ce

C oe ffi ci en t

t- st at is ti c

p >

jtj V al ue

D .2 60 4

V al ue

D .4 54 7

V al ue

D .5 76 3

V al ue

D .6 22 5

V al ue

D .6 53 3

W or kf or ce

(1 D

Po lic e fo rc e)

¡0 .1 56

¡1 .9 49

0. 06 3

¡0 .1 09

� ¡0

.0 82

¡0 .0 39

0. 00 8

¡0 .1 37

�� �

(0 .0 80 )

(0 .0 52 )

(0 .1 06 )

(0 .0 72 )

¡0 .0 28

¡0 .0 15

C om

m an d (1 D

C om

m an d p os it io ns )

¡0 .0 14

¡0 .1 23

0. 90 3

0. 00 5

0. 17 6

¡0 .1 76

¡0 .1 96

�� �

¡0 .0 78

�� �

(0 .1 14 )

(0 .0 66 )

(0 .1 34 )

(0 .0 91 )

¡0 .0 35

¡0 .0 19

Su p er vi so ry (1 D

Su p er vi so ry p os it io ns )

0. 13 79

1. 22 8

0. 23 2

0. 37 7�

�� 0. 30 9�

¡0 .0 60

¡0 .0 82

� 0. 04 6�

(0 .1 12 )

(0 .0 66 )

(0 .1 35 )

(0 .0 91 )

¡0 .0 35

¡0 .0 20

Po lic in g (1 D

Po lic e O ffi ce rs )

0. 20 2

1. 70 6

0. 10 1

0. 49 3

0. 39 8�

0. 01 2

¡0 .0 52

0. 06 3�

(0 .1 18 )

(0 .0 80 )

(0 .1 63 )

(0 .1 10 )

¡0 .0 43

¡0 .0 24

W or kg ro up

si ze

0. 00 1

1. 51 9

0. 14 2

¡0 .0 01

0. 00 1

0. 00 1

0. 00 1

0. 00 1�

��

(0 .0 01 )

(0 .0 01 )

(0 .0 01 )

(0 .0 01 )

¡0 .0 01

¡0 .0 01

In te rc ep t

0. 52 8

7. 08 8

0. 00 1

0. 19 5�

�� 0. 25 8�

� 0. 62 9�

�� 0. 65 1�

�� 0. 67 5�

��

(0 .0 75 )

(0 .0 43 )

(0 .0 88 )

(0 .0 60 )

¡0 .0 23

¡0 .0 13

Si g m a

0. 12 6

(0 .0 22 )

F 5 ,2 4

4. 04

��

Ps eu d o R2

¡0 .9 6

0. 49

0. 31

0. 33

0. 30

0. 33

N 29

29 29

29 29

29

� S ig ni fi ca nt

at a D

.0 5;

�� Si g ni fi ca nt

at a D

.0 1;

�� � S ig ni fi ca nt

at a D

.0 01 .

276 S. A. GUAJARDO

the type of workforce and ethnic diversity (b D ¡0.109, p < 0.05). More specifi- cally, this finding indicates that the level of diversity in the civilian workforce is greater at the 10th quantile in comparison to the level of heterogeneity in the police workforce at the same quantile. This relationship, however, is not significant at the 25th, 50th, or 75th quantiles. At the 90th quantile, the inverse relationship between the type of workforce and ethnic diversity holds (b D ¡0.137, p < 0.001), where the level of ethnic diversity in the police force is lower in comparison to the civilian workforce. Based on these findings, the null hypothesis for the type of workforce should be rejected at the lower and upper tails of the conditional distribution of the response variable.

With respect to the effect of command positions on ethnic diversity, the QR analysis reveals that the effect changes along the conditional distribution (see Table 5). At the 10th quantile, there is a nonsignificant relationship between com- mand positions and ethnic diversity (b D 0.005, p > 0.05). Similar findings were obtained for the 25th and 50th quantiles, respectively, b D 0.176 (p > 0.05) and b D ¡0.176 (p > 0.05). At the 75th quantile, a significant inverse relationship was obtained for command positions and ethnic diversity (b D ¡0.196, p < 0.001), indicating that the level of ethnic diversity is higher for noncommand positions. A similar finding was found at the 90th quantile (b D ¡0.078, p < 0.001). These find- ings indicate that the null hypothesis for command positions should be rejected at the upper tail of the conditional distribution of the response variable.

The QR analysis also indicates that supervisor positions have negative and posi- tive effects on ethnic diversity (see Table 5). At the 10th quantile, a positive rela- tionship exists between the supervisor positions and ethnic diversity (b D 0.377, p < 0.001), which indicates that supervisor positions have higher levels of ethnic heterogeneity in comparison to other positions. This positive relationship holds at the 25th quantile (b D 0.398, p < 0.05). At the 75th quantile, however, an inverse relationship exists between supervisor positions and ethnic diversity (b D ¡0.082, p < 0.05), indicating that nonsupervisor positions have higher levels of diversity in comparison to supervisor positions. This negative relationship does not hold at the 90th quantile where a positive relationship was obtained for the supervisor posi- tions (b D 0.046, p < 0.05). These findings indicate that the null hypothesis for supervisor positions should be rejected at the lower, central, and upper tails of the conditional distribution of the response variable.

While the TR analysis failed to detect a significant relationship between polic- ing-based positions and ethnic diversity (DEG), the QR findings show that polic- ing-based positions are more diverse in terms of ethnic composition in comparison to other positions (see Table 5). At the 10th quantile, a positive rela- tionship was obtained for policing-based positions and ethnic diversity (b D 0.493, p < 0.001). This relationship holds at the 25th quantile (b D 0.398, p < 0.05). At the 50th and 75th quantiles, nonsignificant findings were obtained, respectively, b D 0.012 (p > 0.05) and b D ¡0.052 (p > 0.05). A positive relationship was obtained at the 90th quantile (b D 0.063, p < 0.05). These findings indicate that

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 277

positions in the police and civilian workforces that are responsible for performing law enforcement activities are generally more ethnically diverse in comparison to administrative and support positions in the NYPD. Based on these findings, the null hypothesis for policing-based positions should be rejected at the lower and upper tails of the conditional distribution of the response variable.

Table 6 presents the TR and QR results for ethnic diversity for males. The TR findings reveal that the ethnic diversity among males in the NYPD (DME) is a func- tion of the type of workforce and size of the job titles and classifications. A signifi- cant inverse relationship was found between the type of workforce and ethnic diversity (b D ¡0.228, p < 0.01), which indicates that the civilian workforce has higher levels of ethnic heterogeneity in comparison to the police force. A signifi- cant positive relationship was detected between the size of the job title or classifica- tion and ethnic diversity among males (b D 0.001, p < 0.01), which shows that job titles and classifications with more positions have higher level of ethnic diversity in comparison to those with fewer positions. Table 6 also shows nonsignificant find- ings for command (b D ¡0.030, p > 0.05), supervisor (b D 0.059, p > 0.05), and policing-based (b D 0.126, p > 0.05) positions and ethnic diversity. Accordingly, the null hypotheses for command, supervisor, and policing-based positions should be retained for the data pertaining to males; however, the hypotheses for the type of workforce and the size of the workforce should be rejected.

The QR analysis for ethnic diversity among males in the NYPD (DME) produced mixed results (see Table 6). Table 6 reveals that the type of workforce has both pos- itive and negative effects on ethnic diversity. At the 10th and 25th quantiles, nega- tive relationships exists between the type of workforce and ethnic diversity, respectively b D ¡0.190 (p < 0.001) and b D ¡0.273 (p < 0.01). These findings indicate that the civilian workforce has higher levels of ethnic diversity in compari- son to the police force. However, this negative relation does not hold at the 50th and 75th quantiles, where positive relationships were found between the type of workforce and ethnic diversity, respectively b D 0.207 (p < 0.001) and b D 0.164 (p < 0.01). At the 90th quantile, a nonsignificant relationship was found (b D ¡0.109, p > 0.05). Similar to the TR findings for the data pertaining to males, the null hypothesis for the type of workforce should be rejected; however, it should be noted that the covariate has positive and negative effects across the conditional dis- tribution of the response variable.

The QR finding for command, supervisor, and policing-based positions are gen- erally consistent with the TR results (see Table 6). With the exception of the signif- icant inverse relationship between command positions and ethnic diversity at the 75th quantile (b D ¡0.181, p > 0.05), the level of ethnic diversity for command and other positions is statistically the same. For supervisor positions, the findings show nonsignificant relationships at each quantile. With the exception of the posi- tive relationship between policing-based positions and ethnic diversity at the 10th quantile (b D 3.28, p < 0.001), nonsignificant relationships were obtained for the other quantiles. The QR findings for the size of the job title or classification are

278 S. A. GUAJARDO

Ta b le 6.

To b it an d q ua nt ile

re g re ss io n re su lt s fo r Si m p so n d iv er si ty sc or es

fo r m al es

fo r Ja nu ar y 20 15

(S ta nd

ar d er ro rs ar e in p ar en th es es ).

Q ua nt ile

Re g re ss io n

To b it Re g re ss io n

10 th

Q ua nt ile

25 th

Q ua nt ile

50 th

Q ua nt ile

75 th

Q ua nt ile

90 th

Q ua nt ile

So ur ce

C oe ffi ci en t

t- st at is ti c

p >

jtj V al ue

D .1 80 0

V al ue

D .3 49 0

V al ue

D .5 25 0

V al ue

D .6 17 2

V al ue

D .6 84 9

W or kf or ce

(1 D

Po lic e fo rc e)

¡0 .2 28

¡3 .8 20

0. 00 1

¡0 .1 90

�� �

¡0 .2 73

�� 0. 20 7�

�� 0. 16 4�

� ¡0

.1 09

(0 .0 57 )

(0 .0 55 )

(0 .0 86 )

(0 .0 52 )

(0 .0 55 )

(0 .0 61 )

C om

m an d (1 D

C om

m an d p os it io ns )

¡0 .0 30

¡1 .2 38

0. 22 8

¡0 .0 11

¡0 .0 65

¡0 .0 93

¡0 .1 81

� ¡0

.0 18

(0 .0 80 )

(0 .0 70 )

(0 .1 09 )

(0 .0 66 )

(0 .0 70 )

(0 .0 78 )

Su p er vi so ry (1 D

Su p er vi so ry p os it io ns )

0. 05 96

0. 08 6

0. 39 9

0. 09 9

0. 15 7

0. 05 9

0. 00 9

¡0 .0 24

(0 .0 70 )

(0 .0 70 )

(0 .1 09 )

(0 .0 66 )

(0 .0 70 )

(0 .0 78 )

Po lic in g (1 D

Po lic e O ffi ce rs )

0. 12 6

1. 63 0

0. 11 6

0. 32 8�

�� 0. 25 2

0. 11 9

0. 02 8

¡0 .0 06

(0 .0 77 )

(0 .0 84 )

(0 .1 31 )

(0 .0 80 )

(0 .0 84 )

(0 .0 94 )

W or kg ro up

si ze

0. 00 1

3. 81 2

0. 00 1

¡0 .0 01

0. 00 1

0. 00 1

0. 00 1

0. 00 1

(0 .0 01 )

(0 .0 01 )

(0 .0 01 )

(0 .0 01 )

(0 .0 01 )

(0 .0 01 )

In te rc ep t

0. 60 2

14 .7 10

0. 00 1

0. 36 0�

�� 0. 51 7�

�� 0. 61 7�

�� 0. 67 2�

�� 0. 70 7�

��

(0 .0 41 )

(0 .0 46 )

(0 .0 71 )

(0 .0 43 )

(0 .0 46 )

(0 .0 51 )

Si g m a

0. 10 6

(0 .0 18 )

F 5 ,2 4

9. 27

�� �

Ps eu d o R2

¡2 .3 1

0. 51

0. 49

0. 49

0. 45

0. 31

N 29

29 29

29 29

29

� S ig ni fi ca nt

at a D

.0 5 �� Si g ni fi ca nt

at a D

.0 1 �� � S ig ni fi ca nt

at a D

.0 01 .

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 279

inconsistent with the significant results obtained by the TR analysis. These findings indicate that the null hypothesis for command positions should be retained at the lower and upper tails of the conditional distribution of the response variable for the data pertaining to males. With respect to the null hypothesis for supervisor positions, the null hypothesis should be retained. The findings also show that the null hypothesis for policing-based positions should be rejected at the central and upper tails of the conditional distribution.

Mixed results were obtained for ethnic diversity among female officers and civilians (see Table 7). The TR analysis failed to find any significant relationships between the covariates and ethnic diversity among women (DFE). These findings are consistent with the first model which examined ethnic and gender diversity jointly (DEG). The QR anal- ysis, however, found significant relationships between the predictors and ethnic diver- sity among women. At the 10th and 90th quantiles, the type of workforce is related positively with ethnic diversity, respectively b D 0.098 (p < 0.01) and b D 0.061 (p < 0.001). These findings show that the police force has higher levels of ethnic diversity among women than the civilian workforce. This relationship does not hold at the 25th, 50th, or 75th quantiles, respectively, b D 0.114 (p > 0.05), b D 0.019 (p > 0.05), and b D 0.088 (p > 0.05). For the data pertaining to women police officers and civilians, the TR findings indicate that the null hypotheses for each covariate should be retained; by contrast, the QR results show that the null hypothesis for the type of workforce should be rejected at the lower and upper tails of the conditional distribution of the response variable.

The findings for command, supervisor, and policing-based positions show that these variables have different effects on the level of ethnic diversity among women in the NYPD at different quantiles (see Table 7). At the 10th, 25th, and 50th quantiles, the QR analysis failed to detect significant relationships between the command positions and ethnic diversity. Significant negative relationships were obtained at the 75th and 90th quantiles, respectively b D ¡0.131 (p < 0.01), and b D ¡0.023 (p < 0.001). These results indicate that the level of ethnic diversity among women declines within com- mand positions at the upper end of the conditional distribution of the diversity scores. With respect to the supervisor positions, the QR findings show a positive relationship between supervisor positions and ethnic diversity among women at the lower end of the conditional distribution. For instance, at the 10th and 25th quantiles, significant positive relationships were obtained, respectively b D 0.547 (p < 0.001) and b D 0.474 (p < 0.05). Nonsignificant findings were obtained at the 50th and 75th quantiles, respectively b D 0.046 (p > 0.05) and b D ¡0.080 (p > 0.05). At the 90th quantile, a significant inverse relationship was detected for supervisor positions and ethnic diver- sity (b D ¡0.066, p < 0.001). Briefly, for the policing-based positions, the QR analysis found a significant positive relationship at the 10th quantile (b D 0.654, p < 0.001) and a significant negative relationship at the 90th quantile (b D ¡0.063, p < 0.001). Based on these findings, the null hypothesis for command positions should be rejected for the upper tail of the conditional distribution of the response variable but retained for the lower and central quantiles for the data pertaining to women police officers and

280 S. A. GUAJARDO

Ta b le 7.

To b it an d q ua nt ile

re g re ss io n re su lt s fo r Si m p so n d iv er si ty sc or es

fo r fe m al es

fo r Ja nu ar y 20 15

(S ta nd

ar d er ro rs ar e in p ar en th es es ).

Q ua nt ile

re g re ss io n

To b it Re g re ss io n

10 th

Q ua nt ile

25 th

Q ua nt ile

50 th

Q ua nt ile

75 th

Q ua nt ile

90 th

Q ua nt ile

So ur ce

C oe ffi ci en t

t- st at is ti c

p >

jtj V al ue

D .0 00 1

V al ue

D .5 00 0

V al ue

D .6 22 2

V al ue

D .6 68 3

V al ue

D .6 84 9

W or kf or ce

(1 D

Po lic e fo rc e)

¡0 .0 50

¡0 .2 63

0. 79 5

0. 09 8�

� 0. 11 4

0. 01 9

0. 08 8

0. 06 1�

��

(0 .1 89 )

(0 .0 30 )

(0 .2 13 )

(0 .1 87 )

(0 .0 46 )

(0 .0 05 )

C om

m an d (1 D

C om

m an d p os it io ns )

0. 10 9

0. 49 8

0. 62 3

0. 00 1

0. 41 4

¡0 .0 65

¡0 .1 31

�� ¡0

.0 23

�� �

(0 .2 19 )

(0 .0 30 )

(0 .2 13 )

(0 .1 87 )

(0 .0 46 )

(0 .0 05 )

Su p er vi so ry (1 D

Su p er vi so ry p os it io ns )

0. 23 94

1. 44 0

0. 16 3

0. 54 70

�� �

0. 47 4�

0. 04 6

¡0 .0 80

¡0 .0 66

�� �

(0 .1 66 )

(0 .0 30 )

(0 .2 10 )

(0 .1 84 )

(0 .0 45 )

(0 .0 05 )

Po lic in g (1 D

Po lic e O ffi ce rs )

0. 28 6

1. 67 4

0. 10 8

0. 65 4�

�� 0. 54 3

0. 06 3

¡0 .0 92

¡0 .0 63

�� �

(0 .1 71 )

(0 .0 37 )

(0 .2 62 )

(0 .2 30 )

(0 .0 56 )

(0 .0 06 )

W or kg ro up

si ze

¡0 .0 01

¡0 .2 62

0. 79 6

0. 00 1

¡0 .0 01

0. 00 1

¡0 .0 01

¡0 .0 01

(0 .0 00 )

(0 .0 01 )

(0 .0 01 )

(0 .0 01 )

(0 .0 01 )

(0 .0 01 )

M al e et hn

ic d iv er si ty

0. 23 3

0. 29 0

0. 77 4

0. 00 1

0. 14 4

¡0 .0 11

0. 09 0

0. 00 8

(0 .8 02 )

(0 .0 91 )

(0 .6 49 )

(0 .5 70 )

(0 .1 39 )

(0 .0 16 )

In te rc ep t

0. 28 2

0. 53 7

0. 59 6

0. 09 8

¡0 .0 75

0. 58 9

0. 62 0�

�� 0. 68 3�

��

(0 .5 24 )

(0 .0 57 )

(0 .4 08 )

(0 .3 58 )

(0 .0 87 )

(0 .0 10 )

Si g m a

0. 23 8

(0 .0 51 )

F 6 ,2 3

1. 21

Ps eu d o R2

0. 00

0. 47

0. 17

0. 12

0. 02

0. 07

N 29

29 29

29 29

29

� S ig ni fi ca nt

at a D

.0 5;

�� Si g ni fi ca nt

at a D

.0 1;

�� � S ig ni fi ca nt

at a D

.0 01 .

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 281

civilians. With respect to the null hypothesis for supervisor positions, it should be rejected at the lower and upper tails of the conditional distribution. Similarly, the null hypothesis for the policing-based positions should be rejected at the lower and upper tails of the conditional distribution.

Discussion

The TR and QR models show that the internal organizational-based covariates analyzed in this study have different effects on the level of ethnic diversity in the NYPD. When the level of ethnic diversity for the NYPD was analyzed, the TR anal- ysis failed to detect significant findings for each covariate. However, the QR model found significant relationships for several of the covariates across the conditional distribution of the response variable. In addition, the TR analysis for ethnic diver- sity among male police officers and civilians found that the type of workforce and the size of the workgroup were related significantly with the Simpson scores. By contrast, the QR analysis based on the same data revealed that the effects of the sig- nificant covariates varied across the conditional distribution of the response vari- able. When the Simpson scores for the data pertaining to women police officers and civilians were analyzed, the TR analysis found nonsignificant relationships between the organizational-based covariates and ethnic diversity; however, the QR analysis found significant relationships for several of the covariates.

The findings are generally consistent with prior studies with respect to the inte- gration of minorities and women into policing (e.g., Felkenes et al., 1993; Guajardo, 2014c; Harrington & Lonsway 2004; Martin, 2004; Schulz, 2004; Townsey, 1982). While the Tobit analysis failed to detect a significant relationship between type of workforce and ethnic diversity, the QR analysis found that the NYPD’s police force is less diverse in terms of ethnicity in comparison to the civilian workforce. This inverse relationship holds when the analysis is conducted for males in the police and civilian workforces. When the analysis is performed for women officers and civilians, the QR analysis detected positive relationships between type of workforce and ethnic diversity. Specifically, the level of ethnic diversity among women offi- cers is higher than the level of ethnic diversity of women in the NYPD’s civilian workforce. These findings reveal that the NYPD has been successful to some degree in hiring more minority women than men.

Consistent with prior research (e.g., Felkenes et al., 1993; Guajardo, 2014c; Harring- ton & Lonsway, 2004; Martin, 2004; Schulz, 2004; Townsey, 1982), ethnic and gender integration declines with an increase in police rank and job title. The QR analyses found that there are few minority men and women in command positions. When the level of ethnic diversity for supervisor positions was examined, the QR analyses found positive and negative relationships. At the lower end of the conditional distribution, supervisor positions have a positive effect on ethnic diversity, but an inverse relationship was found at the upper end of the conditional distribution. These positive and negative rela- tionships hold for women officers and civilians. In general, these findings indicate that

282 S. A. GUAJARDO

few minority women ascend to supervisor and command positions. There are multiple plausible explanations as to why few minority women ascend to supervisor and com- mand positions in the NYPD, such as: few women apply to take the promotional exams, few women meet time-in-rank or ¡grade requirements to seek promotional opportunities, and few women seek to overcome the institutional barriers which hinder female officers and civilians from advancing (e.g., a male-dominated organizational cul- ture and perceived gender discrimination; Guajardo, 2015c).

The findings for the total number of employees in police and civilian job tiles and classifications (i.e., workgroup size) are generally inconsistent with prior research which have found positive relationships between organizational size and ethnic and gender integration (e.g., Guajardo, 2014c; Gustafson, 2013; Hochstedler & Conley, 1986; Martin, 1991; Poulos & Doerner, 1996; Warner et al., 1988). While the Tobit analysis found a positive relationship between workgroup size and ethnic diversity for males, the QR analyses failed to find similar results across the condi- tional distribution of the diversity scores. When ethnic diversity among women officers and civilians was analyzed, the Tobit and QR analyses failed to detect any significant relationships between workgroup size and ethnic diversity. Within the NYPD, the number of positions in each police rank or job classification is not a determinant of ethnic diversity. This nonsignificant result may be due to the change in the reporting structure where a number of positions were disaggregated in 2014 (e.g., executives and administrators) and subsequently divided the total number of employees in the original job classifications so that the size of the work- group and the level of ethnic diversity were reduced concomitantly.

The level of ethnic diversity among males in the NYPD does not necessarily increase the level of ethnic diversity among women officers and civilians. As dis- cussed above, the Tobit analysis failed to find a significant relationship between the level of ethnic diversity among men and women. Similarly, the QR analyses failed to detect any significant relationships across the conditional distribution of ethnic diversity scores for women. These findings are inconsistent with prior research which has found a positive relationship between minority male and women integration in policing (e.g. Gustafson, 2013; Hochstedler & Conley, 1986; Kellough, 1990; Kim, 1993; Zhao, Herbst, & Lovrich, 2001; Zhao & Lovrich, 1998). One plausible explanation for this finding is that the male-dominated culture in the NYPD may discourage minority male supervisors and commanders from men- toring or encouraging minority women to seek similar positions. An alternative explanation for the nonsignificant finding is that a gender bias exists in the NYPD in terms of promoting women to supervisor and command positions.

Policy implications11

Although the number of minority men and women hired by the NYPD has increased over the past 14 years, the NYPD continues to face challenges with respect to minority men and women ascending to supervisor and command positions in the

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 283

police and civilian workforces (e.g., Guajardo, 2014a, 2015c). This study has found that the ethnic distribution of minority officers and civilians is uneven within the NYPD’s hierarchical structure and is effected by organizational-based factors. In order to increase the level of ethnic diversity in supervisor and command positions in the NYPD, its police chiefs and commissioners should assess the extent to which institutional barriers hinder minority men and women from seeking and advancing to supervisor and command ranks and job classifications (e.g., Polisar & Milgram, 1996). By doing so, the police chiefs and commissioners would obtain greater insight into how institutional barriers affect the advancement of minority officers and the longevity of their careers in the NYPD (e.g., Bolton, 2003). This assessment would also identify the extent to which institutional barriers create a hostile work environ- ment for minority men and women (e.g., Bolton, 2003). By identifying and address- ing the institutional barriers in the NYPD that negatively impact the advancement of minority men and women, the police chiefs and commissioners would enhance the promotion rates of minority officers and increase the level of ethnic diversity in the supervisor and command positions.

NYPD chiefs and commissioners should assess policies, practices, and processes pertaining to promotional exams to determine whether gender or racial biases exist and subsequently disqualify minority officers from advancement either inten- tionally or unintentionally. In so doing, this would ensure that promotional exams are unbiased, fair, and job related (e.g., Peffer, 2009). For upper-level jobs, the fea- sibility of administering a battery of written and oral promotional exams should be entertained (e.g., Riccucci & Saldivar, 2014). If advancement to an upper-level position requires a written and oral exam, the weight given to the exams should be assessed (e.g., Riccucci & Saldivar, 2014). In cases where oral exams are more important, Riccucci & Saldivar (2014) suggest that more weight should be given to the oral exam than the written exam. As part of the testing process, the NYPD should offer coaching and tutoring to all of its employees with respect to taking written and oral promotional exams so as to ensure that all employees are equally prepared for the exams, eliminate any unfair testing advantage that the dominant group may have over other groups, and diversify the candidate pool eligible for promotion (e.g., Riccucci & Saldivar, 2014).

Police chiefs and commissioners in the NYPD should also assess the extent to which gender and race attitudes discourage minority officers from seeking advancement. In undertaking this assessment, a critical question that should be addressed is the extent to which the dominant group blocks minority access to supervisor and command positions in an effort to protect its interests (e.g., Gustaf- son, 2013). This assessment should also examine the extent to which gender and race attitudes shape the organizational structure of the NYPD and contribute to the under-representativeness of minority men and women in supervisor and com- mand positions (e.g., Bolton, 2003). The effect of gender and race attitudes on minority officer training, job performance, and promotions should be assessed and addressed as well (e.g., Bolton, 2003). By examining how gender and race attitudes

284 S. A. GUAJARDO

contribute to low levels of ethnic diversity in supervisor and command positions, proactive steps should be taken to increase the number of minority officers apply- ing for promotion to upper-level positions.

Finally, NYPD police chiefs and commissioners should assess the percentages of minority officers who seek advancement and subsequently pass the promotional exams (e.g., Polisar & Milgram, 1996). This entails assessing whether the number of minority men and women applying for promotion is proportionate with the total minority work- force in the agency (e.g., Polisar & Milgram, 1996). If the number of minority applicants is not proportionate with the total minority workforce, the root causes for the low appli- cation rate should be identified and addressed. According to Polisar & Milgram (1996), one method for increasing the promotional rate among minorities is to encourage them to apply for supervisor and command positions. In addition, if the pass rate for minority candidates is lower than expected, the reasons for the low success rate should be identified and addressed. If the root cause is the lack of sufficient guidance, the estab- lishment of a formal mentoring program should be entertained so that minority men and women officers have a support system to advance their careers in the NYPD (e.g., Polisar & Milgram, 1996).

Conclusion

This study expanded on the application of Simpson’s diversity index to include the ethnicity and gender characteristics of officers and civilians in the NYPD concomi- tantly to obtain a more comprehensive measurement of heterogeneity. Simpson’s index also was applied separately to the employment data for men and women offi- cers and civilians in the NYPD. By doing so, it was possible to analyze how the level of ethnic diversity fluctuates across police ranks and job classifications. Fur- ther studies should utilize diversity indices to assess the level of social heterogeneity in policing and other criminal justice-related agencies (e.g., courts, correctional institutions, or probation).

In addition to applying of the Simpson diversity index to the NYPD employ- ment data, this study applied quantile regression to assess whether the effects of the organization-based covariates were consistent across the conditional distribu- tion of the diversity scores. The analyses produced findings which supplemented the Tobit analysis. When the Tobit analysis failed to detect significant relation- ships, quantile regression detected significant results at different points across the conditional distribution of the diversity scores. More importantly, the quantile regression analyses found that the effect of the organization-based covariates dif- fered at different quantiles. Future studies on diversity in policing and criminal jus- tice-related agencies should attempt to supplement their regression analyses with quantile regression to obtain a more in depth understanding of how independent variables affect the dependent variable across its distribution.

Despite being the largest municipal police agency in the United States (e.g., Raganella & White, 2004), the NYPD faces challenges with respect to hiring and

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 285

promoting minority men and women. Similar to minority male police officers of different ethnic backgrounds, few minority women ascend to supervisor and com- mand positions within the NYPD. Further, the presence of minority officers in supervisor and command positions does not necessarily increase the integration of minority women into similar ranks. To its credit, however, the level of ethnic diversity among women officers in the NYPD is greater than the level of ethnic diversity among women in the civilian workforce.

Because this study focused exclusively on the NYPD, future studies should ana- lyze the progress that large, midsize, and small policing agencies have made with respect to increasing ethnic diversity. In order to accomplish this, policing agencies need to make their employment data available. Future studies also should examine whether the level of ethnic diversity achieved by policing agencies differs signifi- cantly from other criminal justice-related agencies such as courts, corrections, pro- bation, etc. In addition, future research should increase the types of independent variables (e.g., attrition and hiring rates) used to analyze ethnic and gender integra- tion and diversity in policing and criminal justice-related agencies. As Riccucci & Saldivar (2014, 273) state, “[There] is still considerable room for improvement in the representation of women and people of color in the uniformed services.”

Notes 1. Supervisor positions refer to job titles and classifications in the NYPD police and civilian

workforces that are responsible for performing tasks and activities side-by-side with the individuals that hold subordinate positions and that have very limited ability to set or implement policies, manage financial and nonfinancial resources, or change operations within the agency. In the police workforce, the supervisor positions are sergeants and lieu- tenants; in the civilian workforce, the supervisor positions include clerk supervisors and protective service supervisors.

2. Command positions refer to job tiles and classifications in the NYPD police and civilian workforces that are responsible for setting and implementing policies, allocating and man- aging financial and nonfinancial resources, and making management decisions regarding agency operations and human resources management. In the police workforce, the com- mand positions span from captain to bureau chief; in the civilian workforce, the command positions are the job classifications of administrators and executives.

3. For a more thorough discussion on QR, see Berk (2001), Davino, Furno, & Vistocco (2014), Hao & Naiman (2007), and Koenker (2005).

4. This study omits Tobit and QR analyses for diversity scores based strictly on ethnicity and gender.

5. In this article, quantiles are defined as “points taken at regular intervals from the cumula- tive distribution function of a random variable” (Maclean et al., 2014, p. 194).

6. The analysis did not make a distinction between police and civilian positions that were supervisory or command in nature. The analysis also did not examine employment data for men and women separately.

7. Similar to Gustafson (2013), this study classified both supervisory and command positions in the police and civilian workforces as supervisory and rank-and-file positions as nonsu- pervisory. This study also did not conduct separate analyses for men and women officers and civilians.

286 S. A. GUAJARDO

8. Although the progression to supervisor and command positions in the NYPD comes from the bottom ranks generally, the appointment of an individual to the position of police sur- geon and above is based on the discretion of the commissioner and the police chiefs. Simi- larly, appointments to executive and administrator positions in the civilian workforce are made to individuals from outside of the NYPD. Incumbents of supervisor positions in the civilian workforce may also come from the outside and not necessarily from subordinate positions.

9. Guajardo (2014a, 2015a) provides guidelines on the interpretation of Simpson diversity scores.

10. Lieberson (1969) illustrates how to calculate diversity scores with Simpson’s index when two or more social characteristics are used simultaneously. Lieberson also provides exam- ples on how to obtain standardized scores.

11. Police chiefs and commissioners in jurisdictions outside of New York City should under- take similar steps to assist minority officers and civilians to obtain supervisory and com- mand positions in addition to assessing the extent to which organizational-based factors hinder the progress made by minority officers and civilians in terms of ascending to super- visor and command positions.

Acknowledgments I would like to thank the anonymous reviewers of JECJ and Philip Ragusa, MPA, The City Uni- versity of New York, for their comments and suggestions that helped to improve and strengthen this manuscript.

Notes on contributor Salomon Alcocer Guajardo, Ph.D., is an associate professor of public management in the Department of Public Management at John Jay College of Criminal Justice, The City University of New York. His recent research on ethnic and gender diversity and integration in policing has appeared in the International Journal of Police Science and Management, the Journal of Ethnicity in Criminal Justice, and Women & Criminal Justice.

References

Berk, R. (2011). Asymmetric loss functions for forecasting in criminal justice settings. Journal of Quantitative Criminology, 27(1), 107–123.

Blau, P. M. (1977). Inequality and heterogeneity. New York, NY: Free Press. Bolton, K., Jr. (2003). Shared perceptions: Black officers discuss continuing barriers in policing.

Policing: An International Journal of Police Strategies and Management, 26(3), 386–399. Britt, C. L. (2009). Modeling the distribution of sentence length decisions under a guidelines

system: An application of quantile regression models. Journal of Quantitative Criminology. 25(4), 341–370.

Davino, C., Furno, M., & Vistocco, D. (2014). Quantile Regression: Theory and applications. New York: John Wiley and Sons, Ltd.

Cade, B. S., & Noon, B. R. (2003). A gentle introduction to quantile regression for ecologists. Frontiers in Ecology and the Environment, 1(8), 412–420.

Davino, C., Romano, R., & Naes, T. (2015). The use of quantile regression in consumer studies. Food Quality and Preference. 40(Part A), 230–239.

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 287

DeLisi, M., Beaver, K. M., Wright, K. A., Wright, J. P., Vaughn, M. G., & Trulson, C. R. (2011). Criminal specialization revisited: A simultaneous quantile regression approach. American Journal of Criminal Justice. 36(2), 73–92.

Felkenes, G. T., & Schroedel, J. R. (1993). A case study on minority women in policing. Women & Criminal Justice. 4(2), 65–89.

Felkenes, G. T., Peretz, P., & Schroedel, J. R. (1993). An analysis of the mandatory hiring of females: The Los Angeles Police Department experience. Women & Criminal Justice. 4(2), 31–63.

Fitzenberger, B., Koenker, R., & Machado, J. A. F. (2002). Introduction. In B. Fitzenberger, R. Koenker, & J. A. F. Machado (Eds.), Economic applications of quantile regression (1–5). New York: Physica-Verlag.

Franklin, C. A. (2005). Male peer support and the police culture: Understanding the resistance and opposition of women in policing. Women & Criminal Justice. 16(3), 1–25.

Guajardo, S. A. (2014a). Workforce diversity: Ethnicity and gender diversity and disparity in the New York City Police Department. Journal of Ethnicity in Criminal justice. 12(2), 93–115.

Guajardo, S. A. (2014b). Workforce diversity: Downsizing in the NYPD and its effect on minor- ity integration. International Journal of Police Science and Management. 16(2), 155–167.

Guajardo, S. A. (2014c). Workforce diversity: Assessing the impact of minority integration on intra-group interaction. International Journal of Police Science and Management. 16(3), 205– 220.

Guajardo, S. A. (2015a). Measuring diversity in police agencies. Journal of Ethnicity in Criminal Justice. 13(1), 1–15.

Guajardo, S. A. (2015b). NYPD downsizing and its impact on female officer employment. Jour- nal of Ethnicity in Criminal Justice. 13(4), 255–282.

Guajardo, S. A. (2015c). Women in policing: A longitudinal assessment of female officers in management positions in the NYPD. Women & Criminal Justice. 26, 1–28.

Gustafson, J. (2013). Diversity in municipal police agencies: A national examination of minority hiring and promotion. Policing: An International Journal of Police Strategies & Management. 36(4), 719–736.

Harrington, P., & Lonsway, K. A. (2004). Current barriers and future promise for women in policing. In B. R. Price & N. J. Sokoloff (Eds.), The criminal justice system and women (pp. 495–510). New York, NY: McGraw-Hill.

Hao, L., & Naiman, D. Q. (2007). Quantile regression. Thousand Oaks, CA: Sage Publications Inc.

Hochstedler, E., & Conley, J. A. (2011). Explaining underrepresentation of Black officers in city police agencies. Journal of Criminal Justice. 14(4), 319–328.

Kellough, J. E. (1990). Integration in the public workplace: Determinants of minority and female employment in federal agencies. Public Administration Review. 50(5), 557–564.

Kim, P. S. (1993). Racial integration in the American federal government with special reference to Asian Americans. Review of Public Personnel Administration, 13(1), 52–66.

Koenker, R. (2005). Quantile regression. New York: Cambridge University Press. Koenker, R., & Hallock, K. F. (2001). Quantile regression. Journal of Economic Perspectives. 15

(4), 143–156. Lieberson, S. (1969). Measuring population diversity. American Sociological Review. 34(6), 850–

862. Martin, S. E. (1991). The effectiveness of affirmative action: The case of women in policing. Jus-

tice Quarterly. 8(4), 489–504. Martin, S. E. (2004). The interactive effects of race and sex on women police officers. In B. R.

Price & N. J. Sokoloff (Eds.), The criminal justice system and women (pp. 527–542). New York, NY: McGraw-Hill.

288 S. A. GUAJARDO

Maclean, J. C., Webber, D. A., & Marti, J. (2014). An application of unconditional quantile regression to cigarette taxes. Journal of Policy Analysis and Management. 33(1), 188–210.

McCluskey, C. P., & McCluskey, J. D. (2004). Diversity in policing: Latino representation in law enforcement. Journal of Ethnicity and Criminal Justice, 2(3), 67–81.

Peffer, S. L. (2009). Title VII and disparate-treatment discrimination versus disparate-impact discrimination: The Supreme Court’s decision in Ricci v. DeStefano. Review of Public Person- nel Administration. 29(4), 402–410.

Polisar, J., & Milgram, D. (1996). Recruiting, integrating, and retaining women police officers: Strategies that work. Police Chief. 65(10), 42–53.

Poulos, T. M., & Doerner, W. (1996). Women in law enforcement: The distribution of females in Florida policing agencies. Women & Criminal Justice. 8(2), 19–33.

Raganella, A. J., & White, M. D. (2004). Race, gender, and motivation for becoming a police offi- cer: Implications for building a representative police department. Journal of Criminal Justice. 32(6), 501–513.

Riccucci, N. M., & Saldivar, K. (2014). The status of employment discrimination suits in police and fire departments across the United States. Review of Public Personnel Administration. 34 (3), 263–288.

Ryan, T. P. (2009). Modern Regression Methods. Second edition. Hoboken, NJ: John Wiley & Sons, Inc.

Schroedel, J. R., Frisch, S., Hallamore, N., Petetsen, J., & Vanderhosrt, N. (1996). The joint impact of race and gender on police department employment practices. Women & Criminal Justice. 8(2), 59–77.

Schulz, D. M. (2004). Invisible no more: A social history of women in U.S. policing. In B. R. Price & N. J. Sokoloff (Eds.), The criminal justice system and women (pp. 483–494). New York, NY: McGraw-Hill.

Shelly, T. O., Morabito, M. S., & Tobin-Gurley, J. (2011). Gendered institutions and gendered roles: Understanding the experiences of women in policing. Criminal Justice Studies: A Criti- cal Journal of Crime, Law and Society. 24(4), 351–367.

Simpson, E. H. (1949). Measuring diversity. Nature. 163, 688. Starks, G. L. (2009). Minority representation in senior positions in U.S. federal agencies: A para-

dox of underrepresentation. Public Personnel Management. 38(1), 79–90. Stokes, L. D., & Scott, J. F. (1996). Affirmative action and selected minority groups in law

enforcement. Journal of Criminal Justice. 24(1), 29–38. Talbert, M. K., & Cade, B. S. (2013). User manual for Blossom statistical package for R. Reston,

VA: U.S. Geological Survey. Townsey, R. D. (1982). Black women in American policing: An advancement display. Journal of

Criminal Justice. 10(6), 455–468. Warner, R. L., Steel, B. S., & Lovrich, N. P. (1989). Conditions associated with the advent of rep-

resentative bureaucracy: The case of women in policing. Social Science Quarterly. 70(3), 562– 578.

Yu, K., Lu, Z., & Sandler, J. (2003). Quantile regression: Applications and current research areas. The Statistician. 52(Part 3), 331–350.

Zhao, J., & Lovrich, N. (1998). Determinants of minority employment in American municipal police agencies: The representation of African American officers. Journal of Criminal Justice. 26(4), 267–277.

Zhao, J., He, N., & Lovrich, N. (2006). Predicting the employment of minority officers in U.S. cities: OLS fixed-effect panel model results for African Americans and Latino officers for 1993, 1996, and 2000. Journal of Criminal Justice. 33(4), 377–386.

Zhao, J., Herbst, L., & Lovrich, N. (2001). Race, ethnicity, and the female cop: Differential pat- terns of representation. Journal of Urban Affairs. 23(3–4), 243–257.

JOURNAL OF ETHNICITY IN CRIMINAL JUSTICE 289

Copyright of Journal of Ethnicity in Criminal Justice is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.