Abstract

Scholars have suggested that the benefits of representative bureaucracy arise from bureaucrats acting in the interests of clients who share their characteristics, increased diversity encouraging even nonminority bureaucrats work to further the interests of minority clients, and/or the actions of clients that are more responsive to bureaucrats that share their characteristics. Despite decades of research, the literature has been unable to empirically disentangle these mechanisms, primarily because the vast majority of studies examine only organization-level data, and, at the aggregate level, they all produce identical findings. In contrast, this study makes use of data that allows us to observe the behavior of individual clients and bureaucrats, as well as the aggregate characteristics of the organizations in which they interact. Specifically, we make use of student-level data to predict differences in the probability that an elementary student is referred to gifted services by race. Our results suggest that black students are more likely to be referred to gifted services when taught by a black teacher but that increased presence of black teachers in the school other than the classroom teacher has little effect. We find some evidence that the classroom teacher effect is partially driven by teachers’ more positive views of own-race students. Our results do not suggest, however, that the positive impact of teacher-student race congruence on gifted assignment can be explained by differences in student test score performance or increased parental interaction with the teacher. Ученые предпологают, что выгоды от представительной бюрократии возникают из чиновников, действующих в интересах клиентов, которые разделяют их характеристики, из увеличения разнообразия среди сотрудников, которое поощряет даже чиновников не являющихся членами меньшинств к продвижению интересов меньшинств, либо из действий клиентов, которые более открыты к бюрократам разделяющим их характеристики. Несмотря на десятилетия исследований, литература не смогла эмпирически распутать эти механизмы, в первую очередь потому, что подавляющее большинство исследований рассматривает данные только на уровне организации, где на агрегированном уровне, все исследования производят одинаковые результаты. В отличие от этого, это исследование использует данные позволяющие нам наблюдать за поведением отдельных клиентов и чиновников в совокупности с характеристиками организаций, в которых они взаимодействуют. В частности, мы используем данные на ученическом уровне для предсказания различий в вероятности того, что ученик начальных классов направлен на услуги для одаренных детей по признаку расы. Наши результаты свидетельствуют о том, что черные ученики имеют больше шансов быть отнесены к услугам для одаренных, когда учитель является черным, но, что увеличение присутствия черных учителей в школе, кроме учителя в классе мало на что влияет. Мы нашли доказательства того, что эффект классного учителя частично зависит от более позитивных взглядов учителя об учениках собственной расы. Наши результаты не показывают, однако, что позитивное влияние конгруэнции расы между учителем и учеником на запись к услугам для одаренных можно объяснить различиями в баллах на ученических тестах или повышениями родительского взаимодействия с учителем.

Introduction

Despite decades of evidence that minority clients fare better as the percent of minority bureaucrats within an organization increases, the underlying causal mechanisms responsible for this observation remain unresolved. Scholars have suggested that bureaucratic representation is simply that: bureaucrats acting in the interests of clients who share characteristics. These actions may take the form of proactive advocacy on the behalf of individuals or simply a lower propensity to discriminate against demographically similar clients. Authors have also argued that increased diversity in organizations creates a culture in which even nonminority bureaucrats become more committed to furthering the interests of historically underserved clients. Finally, some have suggested that positive outcomes associated with representative bureaucracy are actually a result of the actions of clients rather than bureaucrats. This work argues that clients are more comfortable with, responsive to, and/or perform better for bureaucrats that share their characteristics, which explains the positive relationship between client outcomes and race congruence.

The failure to identify the causal mechanisms underlying representative bureaucracy arises primarily from the choice of empirical strategy that has dominated this literature. The vast majority of studies examine only aggregate relationships, modeling the mean change in outcomes for some group of clients served by an organization as a function of the percent of the organization’s employees from that same group. Using results from these analyses to “confirm” theoretical mechanisms fundamentally grounded in individual-level behavior obviously raises significant ecological inference concerns. More importantly, for our purposes, it also makes it impossible to empirically distinguish among the competing theoretical mechanisms because, at the aggregate level, all produce identical patterns.

This article offers a test of the empirical accuracy of the three dominant causal mechanisms assumed to underlie representative bureaucracy. To do so, it makes use of a data set that allows us to observe the behavior of individual clients and bureaucrats, as well as the aggregate characteristics of the organizations in which they interact. Specifically, we use nationally representative data from the Early Childhood Longitudinal Study, Kindergarten cohort (ECLS-K) to test for evidence of representation in the differential assignment of students to gifted programs by race. We make use of student-level data to predict the relative probability a student is referred to gifted services as a function of the student’s race, the race/ethnicity of the student’s classroom teacher, the racial/ethnic composition of the teachers at that school, and the interactions among these variables. Importantly, the ECLS-K data also allow us to account for students’ scores on cognitive assessments, which are likely to be highly associated with the probability of gifted referral.

Our results show that, even conditioning on current test scores, black students are assigned to gifted services at higher rates when their classroom teacher is black but that the presence of black teachers in the school other than the classroom teacher has little effect. Moreover, we find some evidence that higher referral probability associated with an own-race classroom teacher is partially mediated by more positive assessments of demographically similar students by those teachers—which, in turn, impacts the probability that the teacher exercises discretion on behalf of the student in the gifted referral process. In contrast, we find little evidence of mediation from a client-side response of either students or parents either in the form of improved achievement performance or differences in the probability that parents have contact with the teacher.

Representative Bureaucracy and Its Causal Mechanisms

The search for representation of diverse interests within public organizations is driven largely by the belief that bureaucracy will serve democratic principles better if it reflects the demographic characteristics of citizens (Rourke 1978). Representation of diverse groups, in other words, helps to ensure pluralism in the implementation of public policies and programs (Denhardt and deLeon 1995). Based on these normative suppositions, a large number of studies have investigated the degree to which the bureaucracy reflects the composition of the larger population (see, e.g., Hall and Saltzstein 1977; Kellough 1990; Nachmias and Rosenbloom 1973) and the factors that influence the prevalence of minority bureaucrats (see, e.g., Cornwell and Kellough 1994; Mladenka 1991; Welch, Karnig, and Eribes 1983).

A key motivation for the interest in passive representation is the expectation of a connection between the shared values and understanding that come with shared demographic characteristics and more equitable treatment of traditionally disadvantaged groups by bureaucrats from those groups (Mosher 1982; Pitkin 1967). Indeed, across a variety of public sector settings, research has shown that policy outputs and outcomes for minority clients of government services improve when the demographic composition of the organization providing those services better reflects that of the client population. These settings include the Equal Employment Opportunity Commission (Hindera 1993), public schools (Meier, Stewart, and England 1989; Meier, Wrinkle, and Polinard 1999; Nicholson-Crotty, Grissom, and Nicholson-Crotty 2011), police organizations (Meier and Nicholson-Crotty 2006; Wilkins and Williams 2008,1), Veteran’s Affairs (Sowa and Selden 2003), and many others.2

Despite decades of research and the significant body of evidence, however, the causal mechanisms underlying observed relationships between minority bureaucrats and improved outcomes for minority clients remain speculative. Typically, the relationship is tested empirically, but then ascribed to an assumed mechanism or set of mechanisms which remain untested.

Those mechanisms can, generally speaking, be lumped into three categories. The first of these is that bureaucrats act in the interests of clients who share their characteristics (Grissom et al. 2009; Keiser et al. 2002; Meier 1993; Meier and Nicholson-Crotty 2006; Nicholson-Crotty et al. 2011; Wilkins and Keiser 2006). This narrative is the oldest and most prominent in the literature, dating to early studies when scholars began to distinguishing between “passive” or descriptive representation and “active” representation—bureaucrats actively pursuing the interests of clients from demographically similar backgrounds—linked to client outcomes.

The second explanation for findings ascribed to representative bureaucracy turns on the organizational changes wrought by increased employee diversity. In this narrative, an increase in the presence of minority bureaucrats in the organization sensitizes their colleagues to the historic underrepresentation and/or unique needs of minority clients. As a result, everyone works to further the interests of those clients, whose outcomes can improve even if they are not paired directly with someone who shares their characteristics (Lim 2006). As an example of this case, Meier et al. (1999) suggest that minority students may show improved test scores when they are in schools with sufficient numbers of minority teachers because both minority and nonminority teachers will more actively represent their interests in that context.

Finally, the literature on representative bureaucracy has suggested that improved outcomes for minority clients may not be directly attributable to the actions of bureaucrats at all. Instead, authors have speculated that clients may be more responsive to bureaucrats who share their demographic characteristics and therefore change their own behaviors in ways that improve their outcomes, independently of any “active” representation (Lim 2006; Thielmann and Stewart 1996; Theobald and Haider-Markel 2009). Meier and Nicholson-Crotty (2006) appeal to this mechanism when they suggest that a correlation between female police officers and sexual assault arrests could be due to active representation or to the fact that female victims are more likely to report these crimes to a female officer.

We argue that the failure to empirically disentangle these theoretical mechanisms arises primarily from the choice of empirical strategy that has dominated this literature. Determining whether improved outcomes for minority clients are a function of active representation by minority bureaucrats, more equitable behavior among nonminority administrators, or changes in the behavior of clients themselves requires data that allow the analyst to observe the behavior of individual clients and bureaucrats, as well as the aggregate characteristics of the organizations in which they interact. Unfortunately, the vast majority of representative bureaucracy studies model only aggregate relationships. With the organization as the unit of analysis, they look for a mean change in outcomes for a subgroup of clients as a function of the percent of the organization’s employees from that same subgroup. By construction, this aggregate focus cannot tell researchers how an individual bureaucrat acted or why an individual client saw improved outcomes. All of the mechanisms offered in the literature turn, in part at least, on the behaviors of individuals within organizations, but the outcomes of these individual behaviors yield identical patterns at the organizational level.

To be fair, there are a few exceptions to this general rule, which have used individual-level data to make claims about the impact of bureaucratic representation. Bradbury and Kellough (2008) highlight individual-level research in criminal justice which has shown that black police officers are less likely to ticket black drivers and that black judges are less likely to incarcerate defendants of the same race. Similarly, Ouazad (2014) demonstrates that individual black teachers are more likely to provide positive subjective assessments of black students, even after controlling for objective measures of student performance. Examining the other side of the bureaucrat/client relationship, Gade and Wilkins (2012) demonstrate that veterans rate therapy experiences more positively when their individual counselor has similar military experiences to their own.

Although these studies offer an important validation of the concept of active representation, they do not satisfactorily answer questions about the causal mechanisms underlying representative bureaucracy. They do not tell us whether clients fare better under bureaucrats like them because of favoritism/lack of discrimination or improved performance on the part of the client. They also do not test the related argument that in the presence of sufficient organizational diversity, the need for individual bureaucrat-client congruence is ameliorated.

Before moving on, we want to make the argument that differentiating between these casual mechanisms is important for more than simply theoretical reasons. If we are ultimately interested in using research findings from representative bureaucracy to create additional public value for clients that have been historically underserved in the administrative process, then understanding which mechanism is actually driving observed correlations between minority bureaucrats and improved outcomes for minority clients is crucial. This is because these mechanisms suggest different management solutions for reducing disparities. For example, if the presence of more minority bureaucrats helps minority clients, regardless of specific bureaucrat-client congruence, then managers simply need to concentrate on building diverse organizations. Alternatively, if individual-level congruence is what matters, then proper assignment of clients becomes the appropriate focus.

An Empirical Test of the Mechanisms of Representative Bureaucracy

We test for the mechanisms discussed above in an individual-level analysis of student assignment to gifted programs in US elementary schools. Gifted (or gifted and talented) programs originated in the 1970s following a report by the federal government suggesting that 3%–5% of the nation’s children “by virtue of outstanding abilities…. require differential educational programs and/or services beyond those provided by the regular school program in order to realize their contribution to self and the society” (Marland 1972). Although an Office of Gifted and Talented Education was established within the Department of Education in 1988, criteria for assignment to and the funding for these programs remain the responsibility of the states. Thirty-two states mandate that school districts provide some form of gifted education, whereas the other 11 for whom data are available leave that decision to the discretion of individual districts (see www.nagc.org). Districts in all states retain significant discretion in identification and screening of, as well as curriculum for, gifted students.3

We choose gifted assignment for a test of the causal mechanisms of representative bureaucracy because it is an area in which we can reasonably expect all of them to manifest (see Grissom, Kern, and Rodriguez 2015). Virtually all gifted assignment systems, including those developed by states and local education authorities, allow teacher nomination or referral to initiate the identification process. The majority of placements are initiated by teacher referral, particularly in the case of minority students who typically lack other advocates for their participation in gifted programs (Donovan and Cross 2002). Referrals for further evaluation are thus one stage at which teachers might exercise discretion in ways that benefit demographically similar students. Moreover, districts often give teachers a formal role in evaluation decisions, incorporating teachers’ classroom observations and assessments alongside standardized test scores and other indicators in deciding whether a student should receive gifted services. For example, in the sample protocol provided as a guide to districts by the Minnesota Department of Education, classroom teachers not only fill out a “nomination form” that rates a potentially gifted student on several “exceptional behaviors” but later compile a portfolio that “contains evidence of students’ strengths in mathematics or language arts, creativity and any other areas of strength, such as artistic, musical, etc.” This involvement in evaluation creates another opportunity for teachers to actively represent the interests of students with whom they share characteristics.

Historically, cognitive ability or achievement has been the primary criterion on which giftedness is judged, but definitions of giftedness have expanded in recent decades in ways that likely enhance the importance of teacher discretion. For example, 19 states formally mandate that local districts use a “multiple criteria method” when assessing giftedness. The method is both “a broadened conception {of giftedness} that includes multiple criteria that might not be measured through an IQ test” and a method of identifying students that uses “several types of assessments, both qualitative and quantitative to determine what students’ strengths are and what curricula modifications are necessary to meet their needs” (Cramond 1997; NAGC 2010). The multiple criteria method emphasizes the use of diverse assessments a greater role for teachers in identifying students often missed by traditional identification methods.

The key role for teachers in the assignment process makes it feasible that active representation might occur when minority teachers recommend minority students for gifted screening at higher rates, based either on scores on cognitive exams or on subjective factors emphasized in “multiple criteria” approaches, or when they participate in gifted evaluation processes. Yet it is also possible that “passive representation” could produce higher referrals to gifted services if the prominence of minority teachers makes even nonminority teachers sensitive to the many factors other than test scores that may make a minority student a good candidate for enhanced programming. Finally, gifted services are a case in which the behavior of the client is undeniably responsible, at least in part, for benefits received from the organization, and there is evidence that black students perform better on a host of criteria when paired with a black teacher (Dee 2004; Grissom, Kern, and Rodriguez 2015). Thus, it is a case where any observed increase in gifted assignment under conditions of student-teacher race congruence could be due to changes made by the client.

Gifted assignment is also a useful place to test these mechanisms because it has been the focus of several recent studies of the impact of representative bureaucracy. For example, Grissom and coauthors (2009) report that assignment to gifted and talented programs for black students in the South increased as the proportion of black teachers in a school increased, findings that a 1 standard deviation increase in the number of black teachers at a school is associated with a 21% increase in the rate of black students present in gifted and talented programs. Rocha and Hawes (2009) also find that a higher percentage of African American or Hispanic teachers increases the odds that a black student will be placed in gifted. Finally, Nicholson-Crotty and colleagues (2011) find that an increase in the portion of black or Hispanic teachers is associated with an increase in the proportion of students in gifted who are black or Hispanic, respectively, particularly in schools in which students from those groups are underrepresented in gifted relative to their proportion in the school’s population. Each of these studies suggests a relationship between minority teachers and the assignment of minority students to gifted, but because all use aggregate data, they cannot identify or distinguish between the mechanisms responsible for that association. An additional study exploring the factors contributing to black and Hispanic student underrepresentation in gifted programs using student-level data finds evidence that black students are assigned at higher rates in classrooms with black teachers, but it does not test whether the racial composition of the faculty as a whole affects this relationship, nor does it consider possible mediating factors driving this result (Grissom and Redding 2016).

Data

We use individual-level data from the restricted-use version of the ECLS-K, a nationally representative sample of 21,260 kindergartners as of the 1998–99 school year (Tourangeau et al., 2006). In the 1998–99 school year, the study recruited a nationally representative sample of 21,260 kindergarteners. The National Center for Education Statistics (NCES) followed this cohort of students through eighth grade, collecting data during the fall and spring of kindergarten as well as first, third, fifth, and eighth grades, though we use only the data through fifth grade in this study, given that most gifted identification takes place in elementary school.

We further restrict the analysis sample to only include public schools with a gifted program. We consider schools to have a gifted program if the administrator describes the school as having a gifted program or a teacher reports any students as receiving gifted services. Across the survey waves, between 33% and 38% of public schools have no gifted program by this definition, which limits the analysis sample to a maximum of 10,640 students in kindergarten, 9,120 in first grade, 8,250 in third grade, and 7,000 in fifth grade, with further reductions due to missing data on any of the variables included in the models. We also limit the sample to only black and white students to simplify interpretation of the results. Longitudinal survey weights supplied by NCES are used in all analyses. Descriptive statistics for all variables are provided in table A1 in the appendix.

Dependent Variable

The dependent variable for this analysis comes from teacher’s reports of whether or not a student was in the school’s gifted program in either reading or mathematics. We assume that a student first identified for gifted services at time t does not begin receiving them until a subsequent time period. We believe that this assumption is supportable because, after referral, districts whose policies we can observe typically require a test designed to measure cognitive ability. Some districts administer this test two times a year, but many do so only once a year, often in the spring.4 Once the results from that test are in, districts may then require an identification committee to evaluate the totality of evidence regarding the multiple criteria on which giftedness is being assessed. If the student is identified as gifted, all districts we found then require parental consent to place the students. Given this process, we believe that it is unlikely that a student typically would begin receiving gifted services in same school year that the initial referral takes place.5

Thus, our analysis considers the probability of being in gifted services in the next survey period, conditional on not being in gifted services in the present period. Note that because ECLS-K does not collect data in students’ second- and fourth-grade years, we cannot directly observe if a student who is in a gifted program in third grade, but was not in first grade, was technically assigned in first or second grade. These gaps in the data should significantly bias the results against our expectations because we cannot directly observe student race congruence in second and fourth grade, which will cause us to underestimate the impact of that variable on assignment to gifted services. As shown in the appendix, approximately 6% of students are assigned to gifted across the sample.

Independent Variables

The main independent variables for this analysis focus on the race of students, their classroom teachers, and the teaching faculty of the school as a whole, plus interactions among these variables. Student and classroom teacher race are each coded as 1 if the student (or teacher) is black and 0 otherwise. Each is based on self-reports. The racial composition of the teaching faculty is measured as the proportion of the teachers in the school who are black. This value is reported as part of the response to the school administrator questionnaire each year.

According to appendix table A1, approximately 83% of students in the sample are white, and 17% are black. Teacher race characteristics are more extreme, with 93% of students in the sample taught by white teachers. Unsurprisingly given these figures, white students are much more likely to be taught by own-race teachers in the sample (about 95% are) than are black students (just 24%).

Control Variables

Models also include a host of student, teacher, and school characteristics that may influence the probability of assignment to gifted services (see Grissom and Redding 2016). Most importantly, most models include a lagged composite measure of math and reading performance on a criterion-referenced test conducted as part of ECLS-K data collection (Pollack et al., 2005), which allows us to test the impact of representative bureaucracy mechanisms holding academic achievement—a major factor in gifted assignment—constant. Student scores are standardized within each wave. Lagging this measure alleviates concerns about endogeneity that may arise if current-period test performance is influenced by having an own-race teacher, a potential mechanism that we revisit later.

We control for numerous other student characteristics, including gender, a standardized scale measure of socioeconomic status (SES), the parent’s report of the child’s health status at entry to kindergarten, and the child’s age in months as of September of 1998, approximately the start of kindergarten. The measure of SES is provided by NCES and combines mother and father’s education, the prestige of the mother and father’s occupation, and household income. The child health measure is a subjective measure rated on a 5-point scale (excellent to poor), which we reverse-coded and standardized.

Teacher characteristics include years of experience in their current school and indicators for whether or not the teacher has a Master’s degree or is certified. We also include the size of the teacher’s class. School characteristics include indicators for whether the school is suburban or urban (with rural as the omitted category), region of the country (four regions, East is the omitted category), school enrollment, the percentage of black students in the school, and the mean achievement level of the sampled students in the school, which we include to account for the possibility that the likelihood that a student is assigned to gifted services varies with the achievement levels of his or her peers.

Methods

We estimate a series of models of the probability that a student will be assigned to a gifted program in the next survey period for all students not currently in a gifted program, with controls for student, teacher, and school characteristics. The dependent variable in this analysis is a binary indicator set equal to 1 if a student is assigned to gifted services in survey wave t + 1 and 0 if not, dropping students from subsequent waves once they have been assigned. Equation (1) shows the general form of these models:

Pr ( g i f t e d ) i t + 1 = f ( B l a c k S t d n t i , B l a c k T c h r i t , B l a c k S t d n t i × B l a c k T c h r i t , C i t , T i t , S i t , γ t , ε i t ) (1)

where C it is a vector of the child’s characteristics for student i in year t, T it is a vector of characteristics of the student’s teacher, and S it is a vector of school characteristics in a given year. A wave fixed effect γ t is included to control for differences in students’ probabilities of being assigned to gifted services as they advance through different grades. Black Stdnt is an indicator for whether the student is black. Black Tchr is either an indicator for whether the student’s classroom teacher at time t is black or the proportion of all teachers in the school who are black, depending on the model. We also include the two-way interaction between Black Stdnt and each of the teacher race variables. A positive coefficient on the interaction term provides evidence of differential probability of assignment to gifted services of black and white students by the teacher race composition of his or her classroom or school.

Before moving on, it is important to recognizing the insights from older literature on representative behavior, which suggests that bureaucrats may only actively represent similar clients when they work in organizations that are sufficiently diverse and, thus, assumed to be more supportive of such behavior (Kanter 1977). In order to test this assertion, we present an additional model which includes a three-way interaction between black student, black teacher, and the percent black teachers within a school.6 Gifted assignment models are estimated using logistic regression given the dichotomous dependent variable. In all models, standard errors are clustered at the student level to account for multiple observations of the same student across time.

Mediation Analysis

In supplementary analysis, we also conduct tests of mediation to examine various mechanisms that may inform gifted assignment, following the Baron and Kenny (1986) approach. We suggest three plausible mechanisms that may mediate the influence of race congruence on gifted assignment: improved client performance, positive perceptions on the part of the bureaucrat, and positive client behaviors. We operationalize client performance as improved student performance on the ECLS-K math and reading standardized assessment. Bureaucrat perceptions include the teacher’s rating of students’ academic ability and classroom behaviors. The rating of academic ability comes from teachers’ reports of how well the child is performing in reading and mathematics. The rating of classroom behavior includes a teacher’s mean rating of their students across four domains: self-control, approaches to learning, internalizing problem behaviors (e.g., anxiety, withdrawal), and externalizing problem behaviors (i.e., acting out). These measures all come from the ECLS-K Social Rating Scale. We coded each set of items so that positive values indicated more favorable ratings. Client behaviors focus on how parents interact with the school, focusing on whether or not the parent attends regular meetings with the teacher.7 For each mediator, we run three regressions: (1) gifted assignment on student and teacher race and their interaction, (2) the mediating variable on student and teacher race and their interaction, and (3) gifted assignment on student and teacher race and their interaction, controlling for the mediating variable. We have evidence of partial mediation if, in the second stage, the interaction predicts the mediator and, in the final stage, the interaction decreases in substantive significance.

The evidence from this mediation analysis is meant to be suggestive rather than conclusive. Research in the psychological and biomedical sciences has described a number of shortcomings to the Baron and Kenny approach to mediation (Holmbeck 1997, 2002), although it remains common in practice. Moreover, the measures available to us capture bureaucrat and client perceptions and behaviors only on some relevant dimensions. We present these results as a step toward investigating possible mechanisms underlying the connection between student and teacher race and the gifted assignment process.

Main Results

The findings from the main models described above are presented in table 1. The coefficients are odds ratios, so values greater than 1 show an increase in the probability of assignment, whereas those below 1 suggest a negative impact for the variable. The first column contains the initial test of the passive representation mechanism, which suggests that greater organizational diversity will improve outcomes for individual minority clients. The controls perform largely as expected. All else equal, students from homes with higher SES, with higher test scores, and in smaller schools are more likely to be referred to gifted programs.

Table 1. (1) (2) (3) (4) Black student 0.46* (−2.18) 0.33** (−3.82) 0.38** (−2.89) 0.24** (−3.46) % Black teachers 7.07* (2.22) 6.35+ (1.94) 5.45 (1.57) Black student × % black teachers 0.29 (−1.10) 0.10 (−1.45) 1.99 (0.45) Black teacher 1.30 (0.80) 1.06 (0.16) 0.91 (−0.18) Black student × black teacher 2.88+ (1.76) 3.35 (1.45) 11.38* (2.39) Black teacher × % black teachers 2.49 (0.42) Black student × black teacher × % black teachers 0.004+ (−1.94) Test score (lagged) 3.33** (13.43) 3.47** (14.42) 3.36** (13.68) 3.35** (13.62) Female 0.95 (−0.43) 0.91 (−0.74) 0.96 (−0.30) 0.96 (−0.30) Socioeconomic status 1.58** (4.84) 1.56** (4.91) 1.60** (5.05) 1.61** (5.10) Parent’s health rating 0.87+ (−1.84) 0.93 (−1.07) 0.90 (−1.49) 0.90 (−1.47) Age in months at start of kindergarten 0.99 (−0.69) 0.99 (−0.56) 0.99 (−0.62) 0.99 (−0.67) Teacher experience 1.00 (0.29) 1.00 (−0.67) 1.00 (−0.12) 1.00 (−0.17) Master’s degree 1.20 (1.31) 1.23 (1.56) 1.22 (1.42) 1.23 (1.48) Certified 1.01 (0.02) 1.09 (0.39) 1.04 (0.17) 1.05 (0.22) Urban 1.00 (−0.02) 1.11 (0.67) 0.90 (−0.61) 0.89 (−0.67) Rural 0.80 (−1.39) 0.80 (−1.42) 0.76+ (−1.68) 0.76+ (−1.66) Midwest 1.40+ (1.69) 1.35 (1.49) 1.36 (1.53) 1.37 (1.56) South 1.32 (1.41) 1.23 (1.03) 1.29 (1.27) 1.30 (1.32) West 1.08 (0.30) 0.94 (−0.29) 0.98 (−0.06) 0.99 (−0.05) School size (100s) 0.91** (−2.90) 0.91** (−3.13) 0.90** (−3.25) 0.90** (−3.22) Class size 1.02 (1.14) 1.01 (1.01) 1.01 (0.83) 1.01 (0.84) Average school test score 0.77 (−1.22) 0.71+ (−1.65) 0.72 (−1.49) 0.73 (−1.44) Constant 0.10* (−2.00) 0.10* (−2.16) 0.12+ (−1.84) 0.12+ (−1.80) Observations 7,610 8,200 7,440 7,440 Pseudo-R2 0.197 0.203 0.203 0.205 (1) (2) (3) (4) Black student 0.46* (−2.18) 0.33** (−3.82) 0.38** (−2.89) 0.24** (−3.46) % Black teachers 7.07* (2.22) 6.35+ (1.94) 5.45 (1.57) Black student × % black teachers 0.29 (−1.10) 0.10 (−1.45) 1.99 (0.45) Black teacher 1.30 (0.80) 1.06 (0.16) 0.91 (−0.18) Black student × black teacher 2.88+ (1.76) 3.35 (1.45) 11.38* (2.39) Black teacher × % black teachers 2.49 (0.42) Black student × black teacher × % black teachers 0.004+ (−1.94) Test score (lagged) 3.33** (13.43) 3.47** (14.42) 3.36** (13.68) 3.35** (13.62) Female 0.95 (−0.43) 0.91 (−0.74) 0.96 (−0.30) 0.96 (−0.30) Socioeconomic status 1.58** (4.84) 1.56** (4.91) 1.60** (5.05) 1.61** (5.10) Parent’s health rating 0.87+ (−1.84) 0.93 (−1.07) 0.90 (−1.49) 0.90 (−1.47) Age in months at start of kindergarten 0.99 (−0.69) 0.99 (−0.56) 0.99 (−0.62) 0.99 (−0.67) Teacher experience 1.00 (0.29) 1.00 (−0.67) 1.00 (−0.12) 1.00 (−0.17) Master’s degree 1.20 (1.31) 1.23 (1.56) 1.22 (1.42) 1.23 (1.48) Certified 1.01 (0.02) 1.09 (0.39) 1.04 (0.17) 1.05 (0.22) Urban 1.00 (−0.02) 1.11 (0.67) 0.90 (−0.61) 0.89 (−0.67) Rural 0.80 (−1.39) 0.80 (−1.42) 0.76+ (−1.68) 0.76+ (−1.66) Midwest 1.40+ (1.69) 1.35 (1.49) 1.36 (1.53) 1.37 (1.56) South 1.32 (1.41) 1.23 (1.03) 1.29 (1.27) 1.30 (1.32) West 1.08 (0.30) 0.94 (−0.29) 0.98 (−0.06) 0.99 (−0.05) School size (100s) 0.91** (−2.90) 0.91** (−3.13) 0.90** (−3.25) 0.90** (−3.22) Class size 1.02 (1.14) 1.01 (1.01) 1.01 (0.83) 1.01 (0.84) Average school test score 0.77 (−1.22) 0.71+ (−1.65) 0.72 (−1.49) 0.73 (−1.44) Constant 0.10* (−2.00) 0.10* (−2.16) 0.12+ (−1.84) 0.12+ (−1.80) Observations 7,610 8,200 7,440 7,440 Pseudo-R2 0.197 0.203 0.203 0.205 View Large

Table 1. (1) (2) (3) (4) Black student 0.46* (−2.18) 0.33** (−3.82) 0.38** (−2.89) 0.24** (−3.46) % Black teachers 7.07* (2.22) 6.35+ (1.94) 5.45 (1.57) Black student × % black teachers 0.29 (−1.10) 0.10 (−1.45) 1.99 (0.45) Black teacher 1.30 (0.80) 1.06 (0.16) 0.91 (−0.18) Black student × black teacher 2.88+ (1.76) 3.35 (1.45) 11.38* (2.39) Black teacher × % black teachers 2.49 (0.42) Black student × black teacher × % black teachers 0.004+ (−1.94) Test score (lagged) 3.33** (13.43) 3.47** (14.42) 3.36** (13.68) 3.35** (13.62) Female 0.95 (−0.43) 0.91 (−0.74) 0.96 (−0.30) 0.96 (−0.30) Socioeconomic status 1.58** (4.84) 1.56** (4.91) 1.60** (5.05) 1.61** (5.10) Parent’s health rating 0.87+ (−1.84) 0.93 (−1.07) 0.90 (−1.49) 0.90 (−1.47) Age in months at start of kindergarten 0.99 (−0.69) 0.99 (−0.56) 0.99 (−0.62) 0.99 (−0.67) Teacher experience 1.00 (0.29) 1.00 (−0.67) 1.00 (−0.12) 1.00 (−0.17) Master’s degree 1.20 (1.31) 1.23 (1.56) 1.22 (1.42) 1.23 (1.48) Certified 1.01 (0.02) 1.09 (0.39) 1.04 (0.17) 1.05 (0.22) Urban 1.00 (−0.02) 1.11 (0.67) 0.90 (−0.61) 0.89 (−0.67) Rural 0.80 (−1.39) 0.80 (−1.42) 0.76+ (−1.68) 0.76+ (−1.66) Midwest 1.40+ (1.69) 1.35 (1.49) 1.36 (1.53) 1.37 (1.56) South 1.32 (1.41) 1.23 (1.03) 1.29 (1.27) 1.30 (1.32) West 1.08 (0.30) 0.94 (−0.29) 0.98 (−0.06) 0.99 (−0.05) School size (100s) 0.91** (−2.90) 0.91** (−3.13) 0.90** (−3.25) 0.90** (−3.22) Class size 1.02 (1.14) 1.01 (1.01) 1.01 (0.83) 1.01 (0.84) Average school test score 0.77 (−1.22) 0.71+ (−1.65) 0.72 (−1.49) 0.73 (−1.44) Constant 0.10* (−2.00) 0.10* (−2.16) 0.12+ (−1.84) 0.12+ (−1.80) Observations 7,610 8,200 7,440 7,440 Pseudo-R2 0.197 0.203 0.203 0.205 (1) (2) (3) (4) Black student 0.46* (−2.18) 0.33** (−3.82) 0.38** (−2.89) 0.24** (−3.46) % Black teachers 7.07* (2.22) 6.35+ (1.94) 5.45 (1.57) Black student × % black teachers 0.29 (−1.10) 0.10 (−1.45) 1.99 (0.45) Black teacher 1.30 (0.80) 1.06 (0.16) 0.91 (−0.18) Black student × black teacher 2.88+ (1.76) 3.35 (1.45) 11.38* (2.39) Black teacher × % black teachers 2.49 (0.42) Black student × black teacher × % black teachers 0.004+ (−1.94) Test score (lagged) 3.33** (13.43) 3.47** (14.42) 3.36** (13.68) 3.35** (13.62) Female 0.95 (−0.43) 0.91 (−0.74) 0.96 (−0.30) 0.96 (−0.30) Socioeconomic status 1.58** (4.84) 1.56** (4.91) 1.60** (5.05) 1.61** (5.10) Parent’s health rating 0.87+ (−1.84) 0.93 (−1.07) 0.90 (−1.49) 0.90 (−1.47) Age in months at start of kindergarten 0.99 (−0.69) 0.99 (−0.56) 0.99 (−0.62) 0.99 (−0.67) Teacher experience 1.00 (0.29) 1.00 (−0.67) 1.00 (−0.12) 1.00 (−0.17) Master’s degree 1.20 (1.31) 1.23 (1.56) 1.22 (1.42) 1.23 (1.48) Certified 1.01 (0.02) 1.09 (0.39) 1.04 (0.17) 1.05 (0.22) Urban 1.00 (−0.02) 1.11 (0.67) 0.90 (−0.61) 0.89 (−0.67) Rural 0.80 (−1.39) 0.80 (−1.42) 0.76+ (−1.68) 0.76+ (−1.66) Midwest 1.40+ (1.69) 1.35 (1.49) 1.36 (1.53) 1.37 (1.56) South 1.32 (1.41) 1.23 (1.03) 1.29 (1.27) 1.30 (1.32) West 1.08 (0.30) 0.94 (−0.29) 0.98 (−0.06) 0.99 (−0.05) School size (100s) 0.91** (−2.90) 0.91** (−3.13) 0.90** (−3.25) 0.90** (−3.22) Class size 1.02 (1.14) 1.01 (1.01) 1.01 (0.83) 1.01 (0.84) Average school test score 0.77 (−1.22) 0.71+ (−1.65) 0.72 (−1.49) 0.73 (−1.44) Constant 0.10* (−2.00) 0.10* (−2.16) 0.12+ (−1.84) 0.12+ (−1.80) Observations 7,610 8,200 7,440 7,440 Pseudo-R2 0.197 0.203 0.203 0.205 View Large

As expected, black students are significantly less likely to be referred than their white peers, even in the presence of controls for important student characteristics, such as prior achievement. Substantively, the results suggest that all else equal, being black reduces the odds of assignment to gifted programming by 54% (in a school with no black teachers). The predicted probability of being assigned to gifted programming is approximately three percentage points lower for black students than white students.

As model 1 shows, counter to the passive representation hypothesis, the percentage of teachers in a school that are also black does not moderate the negative correlation between being a black student and gifted assignment. The interaction between black student and percent black teachers is wrong-signed and falls far short of conventional levels of statistical significance.

The model in the second column of table 1 turns to the hypothesis that student outcomes improve when working directly with a bureaucrat who shares their characteristics. The interaction between black student and having a black teacher is positive and significant at the 0.10 level. Substantively, the impact is quite large, suggesting that the odds of being assigned to gifted programming are nearly three times greater when a black student has a black teacher rather than a white teacher.

Column 3 shows the result of a model that includes interactions between whether the student is black and the percentage of black teachers in the school and whether the student is black and whether the classroom teacher is black. Here, the magnitude of the latter interaction increases as compared to column 2, though the odds ratio is no longer statistically significant at a conventional level (p = .14). This loss of statistical significance is explained at least in part by multicollinearity among the race variables; for example, the Spearman correlation between having a black classroom teacher and the percentage of black teachers in the school is 0.42.

Column 4 shows results from a model that includes the three-way interaction between student race, classroom teacher race, and the racial composition of the teachers in the school overall. A significant interaction among these three variables indicates that the increase in the probability that a black student is assigned to gifted services when taught by a black teacher is moderated by the percentage of the school’s faculty that is black. The odds ratio on the interaction term in column 4 is indeed statistically significant but substantially smaller than 1, indicating that the correlation between classroom teacher and student race may be attenuated when that teacher works with a larger number of black teachers.8

To more accurately interpret the three-way interaction, however, we need to plot the predicted probabilities for black and white students with and without a same race teacher at different percentages of black teachers. Figure 1 presents the predicted probability of gifted assignment is depicted for schools in which 0%, 25%, and 50% of the teachers are black. In general, the confidence intervals for these estimates are wide, making the evidence inconclusive regarding the dependence of the relationship between congruence and assignment on the percentage of black teachers in a school. Still, these results do not support the idea that active representation by minority bureaucrats is more likely when they represent a larger share of the organization’s workforce.

Figure 1. View largeDownload slide Three-way interaction between student race, classroom teacher race, and racial composition of teachers in school.

Figure 1. View largeDownload slide Three-way interaction between student race, classroom teacher race, and racial composition of teachers in school.

Mediation Results

The prior discussion illustrated that it is racial congruence between a student and his or her own classroom teacher rather than with the teachers in the school overall that inform the probability that he or she is assigned to gifted services. Next, we turn to simple mediation analyses aimed at providing suggestive evidence regarding the mechanisms linking race congruence to gifted outcomes. We examine three possible mediators: improved student test score performance in the year he or she is taught by an own-race teacher, higher assessments of student academic performance and behavioral outcomes by own-race teachers, and differential parent involvement with own-race teachers. These variables provide a way to investigate how minority bureaucrats might improve outcomes for minority clients as well as whether outcomes attributed to active representation are, in part, a result of client behavior.

Table 2 shows the results, with two models corresponding to each of the four possible mediating variables (teacher assessments of academic performance and behavioral outcomes are shown separately). In each set, the odd-numbered column shows the result of regressing the potential mediator on indicators for whether the student is black, whether the teacher is black, and their interaction, plus control variables. A positive and significant coefficient on the interaction in this model would indicate that race congruence is associated with a higher value on the potential mediator, the first requirement for a mediating relationship in the Baron and Kenny (1986) paradigm. The even-numbered column shows the result of regressing gifted assignment on the same binary indicators for student and teacher race, their interaction, and control variables, adding the potential mediator as an explanatory variable. In this column, we look for evidence both that the potential mediator is associated with gifted assignment and that there is a meaningful reduction in the odds ratio on the black student-black teacher interaction when compared to the baseline model in column 3 of table 1 (odds ratio = 3.35). If these conditions are met, we conclude that there is some evidence that the mediating variable at least partially explains the association between student-teacher race congruence and gifted assignment.

Table 2. Mediator Test Performance Academic Rating Behavioral Rating Parent Attends Meetings Model outcome Mediator Gifted Assignment Mediator Gifted Assignment Mediator Gifted Assignment Mediator Gifted Assignment (1) (2) (3) (4) (5) (6) (7) (8) Black student −0.07** (−3.27) 0.38** (−3.20) −0.06+ (−1.68) 0.35** (−3.68) −0.25** (−4.12) 0.37** (−3.53) −0.66** (−4.64) 0.33** (−3.80) Black teacher −0.02 (−0.47) 1.23 (0.57) 0.04 (0.73) 1.27 (0.70) −0.10 (−0.85) 1.42 (1.18) 0.07 (0.21) 1.30 (0.81) Black student × black teacher 0.08 (1.23) 2.72 (1.58) 0.09 (1.09) 2.44 (1.44) 0.33* (2.31) 2.47 (1.55) 0.36 (0.91) 2.85+ (1.73) Mediator 5.59** (11.08) 2.45** (9.03) 1.42** (4.93) 0.89 (−0.49) Observations 8,190 8,190 8,170 8,170 8,190 8,190 7,980 7,980 Adjusted R2 0.7 0.45 0.17 0.13 Pseudo-R2 0.26 0.234 0.21 0.2 Mediator Test Performance Academic Rating Behavioral Rating Parent Attends Meetings Model outcome Mediator Gifted Assignment Mediator Gifted Assignment Mediator Gifted Assignment Mediator Gifted Assignment (1) (2) (3) (4) (5) (6) (7) (8) Black student −0.07** (−3.27) 0.38** (−3.20) −0.06+ (−1.68) 0.35** (−3.68) −0.25** (−4.12) 0.37** (−3.53) −0.66** (−4.64) 0.33** (−3.80) Black teacher −0.02 (−0.47) 1.23 (0.57) 0.04 (0.73) 1.27 (0.70) −0.10 (−0.85) 1.42 (1.18) 0.07 (0.21) 1.30 (0.81) Black student × black teacher 0.08 (1.23) 2.72 (1.58) 0.09 (1.09) 2.44 (1.44) 0.33* (2.31) 2.47 (1.55) 0.36 (0.91) 2.85+ (1.73) Mediator 5.59** (11.08) 2.45** (9.03) 1.42** (4.93) 0.89 (−0.49) Observations 8,190 8,190 8,170 8,170 8,190 8,190 7,980 7,980 Adjusted R2 0.7 0.45 0.17 0.13 Pseudo-R2 0.26 0.234 0.21 0.2 View Large

Table 2. Mediator Test Performance Academic Rating Behavioral Rating Parent Attends Meetings Model outcome Mediator Gifted Assignment Mediator Gifted Assignment Mediator Gifted Assignment Mediator Gifted Assignment (1) (2) (3) (4) (5) (6) (7) (8) Black student −0.07** (−3.27) 0.38** (−3.20) −0.06+ (−1.68) 0.35** (−3.68) −0.25** (−4.12) 0.37** (−3.53) −0.66** (−4.64) 0.33** (−3.80) Black teacher −0.02 (−0.47) 1.23 (0.57) 0.04 (0.73) 1.27 (0.70) −0.10 (−0.85) 1.42 (1.18) 0.07 (0.21) 1.30 (0.81) Black student × black teacher 0.08 (1.23) 2.72 (1.58) 0.09 (1.09) 2.44 (1.44) 0.33* (2.31) 2.47 (1.55) 0.36 (0.91) 2.85+ (1.73) Mediator 5.59** (11.08) 2.45** (9.03) 1.42** (4.93) 0.89 (−0.49) Observations 8,190 8,190 8,170 8,170 8,190 8,190 7,980 7,980 Adjusted R2 0.7 0.45 0.17 0.13 Pseudo-R2 0.26 0.234 0.21 0.2 Mediator Test Performance Academic Rating Behavioral Rating Parent Attends Meetings Model outcome Mediator Gifted Assignment Mediator Gifted Assignment Mediator Gifted Assignment Mediator Gifted Assignment (1) (2) (3) (4) (5) (6) (7) (8) Black student −0.07** (−3.27) 0.38** (−3.20) −0.06+ (−1.68) 0.35** (−3.68) −0.25** (−4.12) 0.37** (−3.53) −0.66** (−4.64) 0.33** (−3.80) Black teacher −0.02 (−0.47) 1.23 (0.57) 0.04 (0.73) 1.27 (0.70) −0.10 (−0.85) 1.42 (1.18) 0.07 (0.21) 1.30 (0.81) Black student × black teacher 0.08 (1.23) 2.72 (1.58) 0.09 (1.09) 2.44 (1.44) 0.33* (2.31) 2.47 (1.55) 0.36 (0.91) 2.85+ (1.73) Mediator 5.59** (11.08) 2.45** (9.03) 1.42** (4.93) 0.89 (−0.49) Observations 8,190 8,190 8,170 8,170 8,190 8,190 7,980 7,980 Adjusted R2 0.7 0.45 0.17 0.13 Pseudo-R2 0.26 0.234 0.21 0.2 View Large

Beginning with column 1, we see little evidence of a possible mediating relationship for test performance. The coefficient on the race interaction is very close to 0 and statistically insignificant by any standard, providing no evidence of an association between student test score performance in year t—controlling, importantly, for performance in the prior wave—and race congruence with the student’s teacher for this sample of students. Unsurprisingly, then, there is no reduction in the odds ratio corresponding to the interaction in column 2, and in fact the interaction becomes more positive.

The two teacher rating variables, in contrast, point more in the direction of a mediating relationship, though more so for the behavioral rating than the academic rating. For the academic rating, the coefficient on the race interaction in column 3 is positive but not statistically significant at traditional levels (p = .14).9 For the behavioral rating, the coefficient on the interaction is positive and statistically significant at the 0.05 level, indicating that teachers provide own-race students with more favorable ratings of such noncognitive outcomes as self-control, interpersonal skills, and problem behaviors. In the cases of both rating variables, adding them to the gifted model substantively reduces the odds ratio on the student-teacher race interaction (to about 2.5 in both columns 4 and 6), suggesting that students’ higher probabilities of gifted assignment under own-race teachers may result from higher assessments of those students by their teachers. Higher subjective assessments plausibly are positively related to the likelihood that a teacher refers the student for gifted evaluation or provides a positive recommendation in the evaluation process.

The final possible mediator we consider, whether or not the parents report attending formal meetings at the child’s school, shows no evidence of a mediating relationship. There is no evidence of an association with race congruence (column 7), and including it in the gifted model does not attenuate the odds ratio on the interaction (column 8). Moreover, although not statistically significant, the variable is negatively signed in the gifted model, which is inconsistent with a hypothesis that race congruence increases the likelihood of gifted assignment by making parents more likely to engage with the school, at least through formal meetings.

We also estimated models (not shown) to test for possible mediating relationships with other parent involvement measures, including parent reports of attending informal meetings with the child’s teacher, seeking special placements for their child, and seeking a specialist evaluation for their child, the latter two of which are perhaps more associated with special education placement than with gifted placement. In no cases was student-teacher race congruence meaningfully or statistically associated with the potential mediator.

Conclusions

For five decades, the literature has consistently demonstrated that, under the right circumstances at least, minority clients fare better when receiving public services from a bureaucrat that shares their characteristics. Despite the prevalence of this result, however, the mechanisms by which representation is translated into improved outcomes for individuals remain a mystery. As noted at the outset of this article, studies to date have been largely unable to attribute any particular finding of representative bureaucracy to one or more of these mechanisms because they have typically been conducted at the organizational level.

As an alternative, this study makes use of data that allows us to observe the behavior of individual clients and bureaucrats, as well as the aggregate characteristics of the organizations in which they interact. These data allow us to distinguish among the mechanisms of representative bureaucracy identified in the literature, at least in terms of their power to explain the assignment of black students to gifted programs. Generally speaking, analyses of those data suggest that black students are assigned to gifted services at higher rates when their classroom teacher is black and that the presence of black teachers in the school other than the classroom teacher has little effect. They suggest that one of the mechanisms underlying this active representation is more positive subjective assessments of students by teachers of the same race. Finally, the results do not indicate that changes in the behavior of students or their parents, both of whom can be considered clients in the context of education organizations, explain higher gifted assignment rates for black students by black teachers, though we acknowledge that measures of client behaviors more specific to the gifted assignment process would provide better evidence on this point.

These findings suggest a set of important implications for the delivery of public services. First, findings from the existing representative bureaucracy literature have been used to provide a normative argument for bureaucracies that mirror, as closely as possible at least, the larger population in terms of ethnic, racial, and gender composition. Our results confirm that, to the degree that such representativeness increases the likelihood of client-bureaucrat race congruence, it can improve outcomes for those clients. They also suggest, however, that after accounting for those individual-level effects, organizational composition may have no independent effect on the benefits the client receives from that organization.

The findings also validate the key assumption underlying the concept of active representation—that improved outcomes for clients are a function of the actions of the bureaucrat—in two ways. First, scholars have long speculated that minority bureaucrats are more sympathetic to the challenges faced by and unique skills of clients who share their background and characteristics. This study empirically confirms that better subjective evaluations of clients, regarding both skills and behavior, given by bureaucrats with similar characteristics are associated with increased benefits. Second, some have challenged the concept of active representation by suggesting that observed relationships are actually the result of client rather than bureaucratic behavior. Our study does not provide evidence that clients behave differently when working with a bureaucrat of the same race and, more importantly, its empirical results are not consistent with the idea that client behaviors substitute for the importance of active representation.

References

Baron Reuben M. Kenny David A. 1986 . Moderator-mediator variables distinction in social psychological research: Conceptual, strategic, and statistical considerations . Journal of Personality and Social Psychology 51 : 1173 – 82 . Bradbury , Mark D. Kellough J. Edward and 2008 . Representative bureaucracy: Exploring the potential for active representation in local government . Journal of Public Administration Research and Theory 18 : 697 – 714 . Cornwell Christopher Kellough J. Edward 1994 . Women and minorities in federal government agencies: Examining evidence from panel data . Public Administration Review 54 : 265 – 70 . Cramond B. Morelock M. Lewis G. Geiger R. 1997 . The highly gifted. Gifted education supplement no. 3 . Roeper Review: A Journal on Gifted Education 19 : A1 – A8 . Dee T. S 2004 . The race connection: Are teachers more effective with students who share their ethnicity? Education Next 4 : 52 – 60 . Denhardt Robert B. deLeon Linda 1995 . Greater thinkers in personnel management . In Handbook of public personnel administration , ed. Jack Rabin 21 – 41 . New York, NY : Marcel Dekker . . In, ed.et al. , Donovan M. Suzanne Cross Christopher T. 2002 . Minority students in special and gifted education . Washington, DC : National Academy Press . Gade Daniel M. Vicky M. Wilkins 2013 . Where did you serve? Veteran identity, representative bureaucracy, and vocational rehabilitation . Journal of Public Administration Research and Theory 23 : 267 – 88 . Grissom Jason A. Kern Emily C. Rodriguez Luis A. 2015 . The “representative bureaucracy” in education: Educator workforce diversity, policy outputs, and outcomes for disadvantaged students . Educational Researcher 44 : 185 – 92 . Grissom Jason A. Nicholson-Crotty Jill Nicholson-Crotty Sean 2009 . Race, region, and representative bureaucracy . Public Administration Review 69 : 911 – 9 . Grissom Jason A. Redding Christopher 2016 . Discretion and disproportionality: Explaining the underrepresentation of high-achieving students of color in gifted programs . AERA Open 2 : 1 – 25 . Grissom Jason A. Rodriguez Luis A. Kern Emily C. Teacher and principal diversity and the representation of students of color in gifted programs: Evidence from national data . Elementary School Journal . . Forthcoming. Hall Grace Saltzstein Alan 1977 . Equal employment opportunity for minorities in municipal government . Social Science Quarterly 57 : 864 – 72 . Hindera John J 1993 . Representative bureaucracy: Imprimis evidence of active representation in the EEOC district offices . Social Science Quarterly 74 : 95 – 108 . Holmbeck Grayson N 1997 . Toward terminological, conceptual, and statistical clarity in the study of mediators and moderators: Examples from the child-clinical and pediatric psychology literatures . Journal of Consulting and Clinical Psychology 65 : 599 – 610 . ——— 2002 . Post-hoc probing of significant moderational and meditational effects in studies of pediatric populations . Journal of Pediatric Psychology 27 : 87 – 96 . Kanter R. M 1977 . Some effects of proportions on group life . American Journal of Sociology 82 : 965 – 90 . Keiser Lael Wilkins Vicky M. Meier Kenneth J. Holland Catherine A. 2002 . Lipstick and logarithms: Gender, institutional context, and representative bureaucracy . American Political Science Review 96 : 553 – 64 . Kellough J. Edward 1990 . Integration in the public workplace: Determinants of minority and female employment in federal agencies . Public Administration Review 50 : 557 – 66 . Lim Hong Hai 2006 . Representative bureaucracy: Rethinking substantive effects . Public Administration Review 66 : 193 – 204 . Marland S. P. Jr Jr 1972 . Education of the gifted and talented, volume I: Report to the congress of the United States by the commissioner of education . Washington, DC : United States Government Printing Office . Meier , Kenneth J 1993 . Latinos and representative bureaucracy: Testing the Thompson and Henderson hypotheses . Journal of Public Administration Research and Theory 3 : 393 – 414 . Meier , Kenneth J. Bohte John and 2001 . Structure and discretion: Missing links in representative bureaucracy . Journal of Public Administration Research and Theory 11 : 455 – 70 . Meier Kenneth J. Nicholson-Crotty Jill 2006 . Gender, representative bureaucracy, and law enforcement: The case of sexual assault . Public Administration Review 66 : 850 – 60 . Meier Kenneth J. Stewart Joseph Jr. Jr. England Robert E. 1989 . Race, class and education: The politics of second-generation discrimination . Madison, WI : Univ. of Wisconsin Press . Meier Kenneth J. Wrinkle Robert D. Polinard J. L. 1999 . Representative bureaucracy and distributional equity: Addressing the hard question . Journal of Politics 61 : 1025 – 39 . Mladenka Kenneth R 1991 . Public employee unions, reformism, and Black employment in 1,200 American cities . Urban Affairs Quarterly 26 : 532 – 48 . Mosher , Frederick C 1982 . Democracy and the public service , 2nd ed. New York, NY: Oxford Univ. Press. , 2nd ed. New York, NY: Oxford Univ. Press. Nachmias David Rosenbloom David H. 1973 . Measuring bureaucratic representation and integration . Public Administration Review 33 : 590 – 7 . National Association for Gifted Children 2010 . http://www.nagc.org/index.aspx?id 574&ir (accessed ). Nicholson-Crotty Jill Grissom Jason A. Nicholson-Crotty Sean 2011 . Bureaucratic representation, distributional equity, and democratic values in the administration of public programs . Journal of Politics 73 : 582 – 96 . Ouazad Amine 2014 . Assessed by a teacher like me: Race and teacher assessments . Education Finance and Policy 9 : 334 – 72 . Pitkin , Hanna F 1967 . The concept of representation. Univ. of California Press. . The concept of representation. Univ. of California Press. Pollack Judith M. Narajian Michelle Rck Donald A. Atkins-Burnett Sally Hausken Elvira Germino 2005 . Early Childhood Longitudinal Study-Kindergarten Class of 1998–99 (ECLS-K), psychometric report for the fifth grade . Washington, DC : U.S. Department of Education, National Center for Education Statistics . Rocha Rene Hawes Daniel 2009 . Racial diversity, representative bureaucracy, and equity in multiracial school districts . Social Science Quarterly 90 : 326 – 44 . Rourke Francis E 1978 . Bureaucratic power in national politics: Readings . New York, NY: Little, Brown . ., ed.. New York, NY: Selden , Sally C 1997 . The promise of a representative bureaucracy: Diversity and responsiveness in a government agency . Armonk, NY: M.E. Sharpe. . Armonk, NY: M.E. Sharpe. Selden , Sally Jeffrey L . Brudney J. Edward Kellough and 1998 . Bureaucracy as a representative institution: Toward a reconciliation of bureaucratic government and democratic theory . American Journal of Political Science 42 : 717 – 44 . Sowa Jessica E. Selden Sally Coleman 2003 . Administrative discretion and active representation: An expansion of the theory of representative bureaucracy . Public Administration Review 63 : 700 – 10 . Theobald Nick A. Haider-Markel Donald P. 2009 . Race, bureaucracy, and symbolic representation: Interactions between citizens and police . Journal of Public Administration Research and Theory 19 : 409 – 26 . Thielmann Greg S. Stewart Joseph Jr Jr 1996 . A demand-side perspective on the importance of representative bureaucracy: AIDS, ethnicity, gender, and sexual orientation . Public Administration Review 56 : 168 – 73 . Tourangeau , A. E. Hall L. M. Doran D. M. Petch T. 2006 . Measurement of nurse job satisfaction using the McCloskey/Mueller Satisfaction Scale . Nursing Research 55 : 128 – 36 . Welch Susan Karnig Albert K. Eribes Richard A. 1983 . Changes in Hispanic local public employment in the Southwest . Western Political Science Quarterly 36 : 660 – 73 . Wilkins Vicky M. Williams Brian N. 2008 . Black or blue: Racial profiling and representative bureaucracy . Public Administration Review 68 : 654 – 64 . Wilkins , Vicky M. Keiser Lael R. and 2006 . Linking passive and active representation by gender: The case of child support agencies . Journal of Public Administration Research and Theory 16 : 87 – 102 .

Appendix

Table A1. Mean (SD) Students assigned to gifted 0.061 (0.24) Race congruence White student × white teacher 0.80 Black student × black teacher 0.04 Student characteristics Reading test (standardized) 0.11 (0.93) Math test (standardized) 0.14 (0.94) Female 0.49 White 0.83 Black 0.17 Socioeconomic status 0.00 (0.74) Parent’s health rating 0.03 (0.95) Age in months at start of kindergarten 66.54 (4.22) Teacher/classroom characteristics White teacher 0.93 Black teacher 0.06 Teaching experience (in current school) 9.85 (8.46) Master’s degree 0.40 Certified 0.94 Class size 20.75 (4.04) School characteristics Suburban 0.47 Urban 0.22 Rural 0.31 Northeast 0.13 Midwest 0.27 South 0.48 West 0.13 School size (100s) 5.37 (2.32) Average school test score 0.06 (0.44) Fraction white teachers 0.89 (0.19) Fraction black teachers 0.06 (0.13) Observations 8,970 Mean (SD) Students assigned to gifted 0.061 (0.24) Race congruence White student × white teacher 0.80 Black student × black teacher 0.04 Student characteristics Reading test (standardized) 0.11 (0.93) Math test (standardized) 0.14 (0.94) Female 0.49 White 0.83 Black 0.17 Socioeconomic status 0.00 (0.74) Parent’s health rating 0.03 (0.95) Age in months at start of kindergarten 66.54 (4.22) Teacher/classroom characteristics White teacher 0.93 Black teacher 0.06 Teaching experience (in current school) 9.85 (8.46) Master’s degree 0.40 Certified 0.94 Class size 20.75 (4.04) School characteristics Suburban 0.47 Urban 0.22 Rural 0.31 Northeast 0.13 Midwest 0.27 South 0.48 West 0.13 School size (100s) 5.37 (2.32) Average school test score 0.06 (0.44) Fraction white teachers 0.89 (0.19) Fraction black teachers 0.06 (0.13) Observations 8,970 View Large

Table A1. Mean (SD) Students assigned to gifted 0.061 (0.24) Race congruence White student × white teacher 0.80 Black student × black teacher 0.04 Student characteristics Reading test (standardized) 0.11 (0.93) Math test (standardized) 0.14 (0.94) Female 0.49 White 0.83 Black 0.17 Socioeconomic status 0.00 (0.74) Parent’s health rating 0.03 (0.95) Age in months at start of kindergarten 66.54 (4.22) Teacher/classroom characteristics White teacher 0.93 Black teacher 0.06 Teaching experience (in current school) 9.85 (8.46) Master’s degree 0.40 Certified 0.94 Class size 20.75 (4.04) School characteristics Suburban 0.47 Urban 0.22 Rural 0.31 Northeast 0.13 Midwest 0.27 South 0.48 West 0.13 School size (100s) 5.37 (2.32) Average school test score 0.06 (0.44) Fraction white teachers 0.89 (0.19) Fraction black teachers 0.06 (0.13) Observations 8,970 Mean (SD) Students assigned to gifted 0.061 (0.24) Race congruence White student × white teacher 0.80 Black student × black teacher 0.04 Student characteristics Reading test (standardized) 0.11 (0.93) Math test (standardized) 0.14 (0.94) Female 0.49 White 0.83 Black 0.17 Socioeconomic status 0.00 (0.74) Parent’s health rating 0.03 (0.95) Age in months at start of kindergarten 66.54 (4.22) Teacher/classroom characteristics White teacher 0.93 Black teacher 0.06 Teaching experience (in current school) 9.85 (8.46) Master’s degree 0.40 Certified 0.94 Class size 20.75 (4.04) School characteristics Suburban 0.47 Urban 0.22 Rural 0.31 Northeast 0.13 Midwest 0.27 South 0.48 West 0.13 School size (100s) 5.37 (2.32) Average school test score 0.06 (0.44) Fraction white teachers 0.89 (0.19) Fraction black teachers 0.06 (0.13) Observations 8,970 View Large

© The Author 2016. Published by Oxford University Press on behalf of the Journal of Public Administration Research and Theory, Inc. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.