- Split View
-
Views
-
Cite
Cite
Katie Vinopal, Understanding Individual and Organizational Level Representation: The Case of Parental Involvement in Schools, Journal of Public Administration Research and Theory, Volume 28, Issue 1, January 2018, Pages 1–15, https://doi.org/10.1093/jopart/mux036
- Share Icon Share
Abstract
With the substantive effects of racial and ethnic representation well-documented in the representative bureaucracy literature, scholars have increasingly turned their attention to better understanding the causal mechanisms underlying these effects. One explored but yet unresolved issue is whether the benefits of representation stem from individual (direct)- versus organizational (indirect)-level pathways, or both. The current study advances this conversation by testing the effect of both levels of representation on parental involvement in schools using a current, nationally-representative dataset: the Early Childhood Longitudinal Study, Kindergarten Class of 2010–2011 (ECLS-K: 2011). In doing so, this study also contributes to research on parental involvement, and how it may relate to racial and ethnic achievement gaps. Results indicate that both direct and indirect representation matter, though perhaps not equally. Direct representation has a larger association with parental involvement outcomes, especially for nonwhite students. After controlling for this direct representation, indirect representation shows statistically significant but smaller increases in parent-reported conference attendance.
Introduction
A large body of research over the last several decades has advanced the theory of representative bureaucracy, which posits that a public workforce gains legitimacy and improves outcomes for the public if it is reflective of the demographics of the population it serves (e.g., Dee 2005; Grissom, Kern, and Rodriguez 2015; Hong 2017; Krislov 1974; Meier 1993; Ouazad 2014; Riccucci and Meyers 2004; Rosenbloom and Featherstonhaugh 1977; Wilkins 2007). Many studies have applied representative bureaucracy theory to the education context, examining the effect of racial/ethnic representation among teachers on student outcomes, and have documented benefits of representation for minority students (e.g., Dee 2005; Ehrenberg et al. 1995; Favero and Molina, forthcoming; Gershenson, Holt, and Papageorge 2016; Meier 1993; Nicholson-Crotty et al. 2016; Ouazad 2014).
With the substantive effects of representation well-documented, scholars have increasingly turned their attention to better understanding the causal mechanisms underlying these effects. Although the literature is complete with speculative explanations of potential pathways, relatively few studies have investigated them empirically, and those that have yield mixed results. Of particular interest is whether the benefits of representation stem primarily from individual- versus organizational-level interactions (or both). Individual-level, or direct, pathways occur when direct contact between minority teachers and students results in positive effects for students. Organizational-level, or indirect, pathways occur when representation at some aggregate level within a sphere of influence (usually the school) results in positive effects for students (Favero and Molina, forthcoming; Nicholson-Crotty et al. 2016).
The current study advances this ongoing conversation regarding the mechanisms underlying the substantive effects of representation by testing the effect of both individual- and organizational-level representation on parental involvement in schools using a current, nationally-representative dataset: the Early Childhood Longitudinal Study, Kindergarten Class of 2010–2011 (ECLS-K: 2011). In doing so, this study also contributes to research on parental involvement in children’s schools, and how it may relate to racial and ethnic achievement gaps. By the start of kindergarten, black children are about one half of a standard deviation behind white peers on standardized math and reading tests (Burchinal et al. 2011) and this gap may be even larger between Hispanic and white children (Reardon and Galindo 2009). Some, but not all of these differences can be explained by differences in income and wealth (e.g., Orr 2003; Yeung and Conley 2008). Although schools play a role in closing the achievement gap, the fact that these gaps exist before school entry points to the contribution of family and household circumstances and practices in explaining children’s academic outcomes (Kraft and Rogers 2015). Additionally, the interactions between home and school environments—the two settings where children spend the vast majority of their time—may provide opportunities to improve child outcomes by increasing communication and consistency for children and bolstering parents’ efforts to provide academic and social support (Bergman 2015; Harackiewicz et al. 2012; Kraft and Rogers 2015; Kraft and Dougherty 2013).
Specifically, this study addresses two primary research questions:
Does having a same-race/ethnicity teacher affect parent and teacher reports of parents’ school-based involvement? Does this vary by student race or income?
Does school-level representation in the teacher workforce affect parents’ school-based involvement? Does this vary by student race or income?
Theory and Literature
Representative Bureaucracy
The theory of representative bureaucracy is concerned with how the demographic characteristics of bureaucrats are associated with implementation and policy outcomes of clients who share those characteristics. A robust literature has established a link between racial/ethnic and gender representation and favorable outcomes for minority or female clients.
This literature has traditionally distinguished between passive and active representation (Mosher 1968). An institution is passively representative to the extent that the background characteristics of its employees mirror those of the population they serve. Characteristics most often examined are race/ethnicity and gender, though LGBTQ (Lewis and Pitts 2011) and veteran status (Gade and Wilkins 2013) have also been explored. Early studies in this field documented the degree of (Meier 1975) and factors determining (Kellough 1990; Naff 1998; Riccucci and Saidel 1997) passive representation.
The study of active representation, on the other hand, is concerned with how this representation actually influences bureaucratic behavior and policy implementation (Mosher 1968). Much of the representative bureaucracy literature has been concerned with understanding the translation from passive to active representation; that is, the relationship between representation and client outcomes. This translation is assumed to stem from shared values, attitudes, and concerns between bureaucrats and clients with shared racial/ethnic or gender identities. Then, to the extent that a bureaucrat uses his or her discretion to act in alignment with his or her values, his or her actions will result in improved outcomes for clients with whom an identity is shared.
Other work has sought to better explain when and in what context passive translates into active representation. This research has identified three major conditions for this translation to take place. First, bureaucrats must have discretion (Sowa and Selden 2003). Second, bureaucrats must share values, attitudes, and interests with the social group to which they belong; that is, they must take on a representative role for that group. Several studies have found differences by race/ethnicity and gender in the priorities of bureaucrats within an organization, and congruence in attitudes between bureaucrats and clients of the same demographic group (Bradbury and Kellough 2008; Riccucci and Meyers 2004; Rosenbloom and Featherstonhaugh 1977; Wilkins 2007). However, others have found that organizational socialization can partially or fully overwhelm racial or ethnic identification in some contexts, eliminating the link between passive and active representation (Roch and Pitts 2012; Wilkins and Williams 2008; Wilkins and Williams 2009). Finally, the policy area or organization type being analyzed must be salient for the demographic group under question (Dolan 2000; Keiser et al. 2002; Kelly and Newman 2001; Meier 1993; Saidel and Loscocco 2005; Wilkins and Keiser 2006). For example, the policy area might benefit that group or have a history of discrimination against that group.1
Increasingly, however, scholars have recognized the capacity for passive representation to result in substantive outcomes without a conscious effort by bureaucrats (Atkins and Wilkins 2013; Fairlie, Hoffmann, and Oreopoulos 2014; Gade and Wilkins 2013; Meier and Nicholson-Crotty 2006; Riccucci, Van Ryzin, and Lavena 2014; Riccucci, Van Ryzin, and Li 2016; Shah 2009; Theobald and Haider-Markel 2009). Keiser et al. (2002) discussed how representation can change the nature of the relationship between the bureaucrat and the client due to shared experiences or backgrounds, and this idea was first explored empirically by Meier and Nicholson-Crotty (2006). Lim (2006) summarizes these ideas by arguing that the effects of representative bureaucracy can be divided into effects requiring partiality in behavior by bureaucrats and effects resulting from a change in behavior of other (majority) bureaucrats or the clients themselves. This symbolic representation has come to define a third category in representative bureaucracy theory and push researchers past the passive–active distinction. Symbolic representation occurs when simply the existence of representation changes the behavior of clients in ways that result in more favorable outcomes. In other words, as Riccucci and colleagues (2016) describe, “This symbolic representation hypothesis leads to the expectation that citizens will be more willing to cooperate and thus coproduce important outcomes when their gender, race, ethnicity, or even shared identity is represented in a government bureaucracy” (p. 121). Of course, obvious co-producers of educational outcomes are parents, and the opportunities for this coproduction are vast. Previous research suggests that the larger the opportunity for coproduction, the better representative bureaucracy performs (Andrews, Ashworth, and Meier 2014).
In sum, representation may result in substantive outcomes for represented groups through a number of passive, symbolic, and active mechanisms that may or may not be related to partiality by bureaucrats. Additionally, these pathways may be direct—at the individual level—or indirect—at the organizational level. The relative influence of these two broad types of representation remains under debate (Bradbury and Kellough 2011; Favero and Molina, forthcoming; Meier and Bohte 2001; Nicholson-Crotty et al. 2016; Weissberg 1978; Wilkins and Keiser 2006).
Two recent papers have tackled this question with different approaches, populations, and data—and reached different conclusions. First, Nicholson-Crotty and colleagues (2016) use nationally representative data from the 1998–99 Panel of the Early Childhood Longitudinal Study-Kindergarten Cohort to analyze the effect of representation on the probability that a black student is referred to gifted services. By modeling the effects at both the individual level (i.e., student–teacher racial match) and the organizational level (i.e., percent of representation among teachers at a school), the authors disentangle the influence of direct teacher effects versus indirect school effects. The results indicate strong support for the individual level hypothesis: having a black teacher increased the likelihood that a black student was referred to gifted services. After controlling for these individual-level effects, the authors found no additional effect of representation at the organizational level.
Favero and Molina (forthcoming) investigate a similar question: are there indirect, organizational-level effects of representation on minority clients? Using panel data on schools in Texas, the authors conduct a series of analyses comparing the effect of more direct, sub-organizational representation to the effect of indirect, organizational-level representation on Latino student test scores. The authors provide evidence that representation at both levels—sub-organizational and organizational—have independent, substantive effects on Latino student test scores. They argue that their results “suggest that the relationship between passive representation and client outcomes is driven as much by organizational-level phenomena as it is by individual bureaucratic preferences and decisions.” (p. 17).
The present study contributes to this ongoing theoretical and empirical investigation regarding the roots of the substantive effects of representation using current, nationally-representative data and an under-examined outcome: parental involvement. Parents are important co-producers of education for children, but their direct response to organizational factors at the school has received less attention than effects on children directly. There are at least two potential direct or individual-level mechanisms at play in the current context. First, a minority teacher is more likely to serve in a representative role; that is, engage in active representation by encouraging minority parents to become involved (Atkins, Fertig, and Wilkins 2014). Second, a same-race/ethnicity teacher may serve as a role model for students and parents, making school services and opportunities for involvement more attractive to minority parents. There are also at least two potential indirect, or organization-level mechanisms. First, minority teachers may induce changes in the behavior of majority teachers toward minority students and parents. Indeed, in a study of the Equal Employment Opportunity Commission (EEOC), Hindera and Young (1998) find that when black administrators constitute a plurality, there is a change in the behavior of white bureaucrats toward black clients, amplifying the effect of representation. Second, passive representation may cause shifts in the policy or priorities of the school in a way that is favorable for minority families. Although the current article cannot distinguish between mechanisms within direct versus indirect categories, it contributes to understanding whether or not both categories of mechanisms are at play. The analysis also investigates whether particular racial or socioeconomic status (SES) subgroups are more or less affected by these different types of representation.
Empirically, this analysis builds on a large and growing body of research by education and public administration scholars examining the effect of teacher–student racial/ethnic “match” (at either the individual or organizational level) on student outcomes (Grissom, Kern, and Rodriguez 2015). In the education context, many studies have documented a relationship between racial/ethnic and gender representation among teachers and positive student outcomes, usually for minority and female students, including test scores (Egalite, Kisida, and Winters 2015; Favero and Molina, forthcoming), teacher assessment of student performance (Dee 2005; Ehrenberg et al. 1995; Ouazad 2014), pass rates on state exams (Meier 1993; Meier and Bohte 2001), educational expectations and school connectedness (Atkins, Fertig, and Wilkins 2014; Gershenson, Holt, and Papageorge 2016), dropout rates (Meier 1984), assignment to special education (Meier 1993; Meier and Stewart 1992), assignment to gifted programs (Grissom, Nicholson-Crotty, and Nicholson-Crotty 2009; Grissom, Kern, and Rodriguez 2015; Meier and Stewart 1992; Nicholson-Crotty, Grissom, and Nicholson-Crotty 2011; Nicholson-Crotty et al. 2016; Rocha and Hawes 2009), rates of corporal punishment, suspension, expulsion, and other disciplinary measures (Grissom, Nicholson-Crotty, and Nicholson-Crotty 2009; Meier and Stewart 1992), and reproductive health (Atkins and Wilkins 2013). This work has also been extended to the community college context, demonstrating a relationship between having a same-race instructor and a reduction in the minority achievement gap (Fairlie, Hoffmann, and Oreopoulos 2014).
Relevant to the current study, a small group of papers has begun to explore the relationship between representation and parental involvement among racial or ethnic minority parents. Most related, Shah (2009) used a survey of 324 Latino parents in Chicago and found a positive association between their school involvement and the level of Latino representation measured at the classroom, administrator, and school council levels. Also analyzing Chicago, Marschall (2006) finds an association between the percentage of Latino members of the Local School Council, among other factors, and more school outreach to involve parents and, ultimately, parental involvement. Similarly, Marschall, Shah, and Donato (2012) find associations between minority principals and parent involvement programing in new immigrant destinations (areas of the country with small but rapidly growing foreign-born populations), and an association between proportion of minority teachers and parental involvement programing.
Parental Involvement in Schools
School reform policies have emphasized parental involvement in schools since the 1980s (Domina 2005; Wilder 2014). Such policies reflect lessons taken from decades of research documenting the importance of home and family life on children’s educational outcomes (Domina 2005; Epstein 1986). By working to incorporate parents across race/ethnicity and social class at the school, such parental involvement initiatives attempt to boost the social capital resources available to disadvantaged children, and reduce education inequality (Domina 2005).
Recent surveys show that parents’ involvement in and their knowledge of their children’s schools are lower than the levels desired by both schools and parents. Public Agenda (2012) found that 65% of parents polled wished they were doing more to be involved in their children’s education (Public Agenda 2012). Results from the 2012 National Household Education Survey (NHES) shows that although overall rates of general school meeting or parent–teacher organization meeting attendance are relatively high, with 87% of families with school-age children in the United States reporting attendance, only 43% of parents report volunteering in schools, and only 41% of parents report getting a phone call about their student (Noel et al. 2013).
Rates of parental involvement in schools vary by parents’ sociodemographic characteristics. Parents without a high school degree are much less likely to be involved than parents with a bachelor’s degree (19% versus 55% for volunteering; 64% versus 80% for conference attendance, for example) (Noel et al. 2013). Racial and ethnic differences also exist: black and Hispanic families are less likely than white families to attend events, volunteer, or fundraise at school (Noel et al. 2013). To understand these differential rates, researchers have investigated how, why, in what contexts parents choose to become involved in their children’s classrooms and schools, and how schools might encourage participation (Anderson and Minke 2007; Balli et al. 1998; Castro et al. 2004; Green et al. 2007; Heymann and Earle 2000; Lamb-Parker et al. 2001; Lawson 2003; Marschall 2006; Marschall and Shah, forthcoming; Marschall, Shah, and Donato 2012; Overstreet et al. 2005; Sheldon 2002; Waanders et al. 2007). One prominent theme in this literature is the impact of cultural and language barriers and perceived cultures of exclusion at the school that may decrease parental involvement, especially for low-SES, minority, and immigrant families (e.g., Garcia Coll et al. 2002; Huntsinger and Jose 2009; Marschall 2006; Marschall and Shah, forthcoming; Marschall, Shah, and Donato 2012; Pena 2000; Ramirez 2003; Turney and Kao 2009).
The gaps in parental involvement by educational attainment and race/ethnicity are concerning because of the generally positive relationship between parental involvement and children’s outcomes (e.g., Avvisati et al. 2013; Barnard 2004; Cheung and Pomerantz 2012; Domina 2005; Epstein and Sheldon 2002; Fan and Chen 2001; Hill and Stafford 1980; Houtenville and Conway 2008; Phillips 2011; Todd and Wolpin 2007; Zellman and Waterman 1998). Although many of these older studies are descriptive, several more recent experimental evaluations have documented a positive causal impact of increased parent–teacher communication on student outcomes (Avvisati et al. 2013; Bergman 2015; Harackiewicz et al. 2012; Kraft and Rogers 2015; Kraft and Dougherty 2013). For example, in a recent experimental design, Kraft and Rogers (2015) evaluate the impact of a teacher–parent communication intervention aimed at increasing parents’ support of children’s academic success. Parents in the treatment group at an at-risk urban school were sent weekly one-sentence text messages from teachers. Results indicate that this relatively simple intervention increased the probability that students earned credit for classes by decreasing the dropout rate—and this was attributed to parents talking with their children about what they needed to do to improve in school (Kraft and Rogers 2015). This and other experiments show the importance of leveraging parental involvement to improve student performance, and shed light on relatively low-cost interventions that may help narrow the achievement gap. Additionally, parents’ involvement in schools is an important topic of study in its own right, as involvement is an indicator of parents’ community engagement and social capital (Cruz 2009; Harper and Kelly 2003; Putnam 1995, 2001), can influence school resources, and plays a role in teacher self-perception and job satisfaction (Tschannen-Moran and Hoy 2007).
The current study also builds on recent sociological work on the role of race and class in home–school relationships that uses the lens of cultural capital, theorizing that families that are less familiar with the language and style of educational settings are less likely to be involved (Lareau 2003). Cultural capital also shapes how teachers perceive families, informing ideas about the appropriate “type” of parents (Abrams and Gibbs 2002; Lareau 2003; Lareau and Horvat 1999). Indeed, research has shown that the power dynamic between teachers and parents over child rearing is sometimes ambiguous, as cultural experiences shape how different groups see the relative roles of home and school, and can result in miscommunication and the mislabeling of parents as uncaring (Cooper 2010), ultimately resulting in exclusion and isolation (Delgado-Gaitan 1991).
Lewis and Forman (2002) argue that, especially in schools with a less diverse teaching staff, parental involvement is important in overcoming the lack of “sociocultural congruence” between teachers and students, which can be a barrier to academic achievement. However, it is precisely because of these “cultural discontinuities” that home–school relationships are more challenging in these settings (Lewis and Forman 2002). Several researchers have noted that race and class differences between school staff and parents could increase the “social distance” between these groups (Delgado-Gaitan 1991; Horvat, Curci, and Partlow 2011; Lewis and Forman 2002). For example, Lasky (2000) finds that teachers “were more comfortable with parents who filled a related set of expectations and who shared their value systems” (p. 857). Relatedly, Carlisle et al. (2005) suggest that a parent–teacher racial or language mismatch might lead to parents feeling uncomfortable and therefore avoiding involvement.
Finally, it should be noted that the findings of this analysis are especially relevant given the lack of diversity in the teacher workforce in the United States relative to the diversity of students. In the 2011–2012 school year, for example, only 17.3% of elementary and secondary teachers were of a minority racial or ethnic group, whereas minority children comprised over 44% of elementary and secondary students (Ingersoll 2013). The gap in representation exists for both black and Hispanic subgroups as well. For example, in that same school year, 14.4% of students were black and 21.1% of students were Hispanic. However, only 6.4% of teachers were black, and only 7.5% of teachers were Hispanic (Ingersoll 2013).
Data and Empirical Approach
Data for this analysis come from the ECLS-K: 2011, collected by the National Center for Education Statistics (NCES). The ECLS-K follows a nationally representative sample of kindergarteners in fall and spring of kindergarten in 2010, and spring of 1st and 2nd grades.2 The primary sample uses one observation per child, per year: the spring of kindergarten, 1st, and 2nd grade waves of data. Surveys document lower rates of parental involvement for older children (Noel et al. 2013), which is in line with previous literature that shows that both rates of parental involvement in schools and the effectiveness of that parental involvement declines as children age (e.g., Catsambis and Garland 1997; Muller 1998). This pattern, combined with the fact that most parental involvement policies are targeted at younger children, makes a sample of early elementary school children the most ideal age range for the current analysis (Domina 2005).
The analytic sample has a number of other restrictions. First, I limit the sample to only white, black, and Hispanic children with white, black, or Hispanic teachers, as other racial groups (e.g., American Indian or Alaska Native, Asian, Native Hawaiian or other Pacific Islander) have too small a sample size to generate reliable estimates. For research question one, I further restrict the sample to classrooms that have at least two students participating in the ECLS-K study, as my identification strategy relies on classroom fixed effects (described further below).3 These restrictions combined with missing data result in a sample size that varies by dependent variable, from about 10,330 to 13,530 observations. Sample sizes for each model are reported in the results tables.
I consider five binary measures of school-based parental involvement as outcome variables: teacher-reported parent-initiated contact with teacher, teacher-reported parent attends meetings at school, teacher-reported parent returns calls or emails, teacher-reported parent attends scheduled conferences, and parent-reported scheduled conference attendance.
The primary independent variable for research question one is a binary indicator of “nonmatch”; that is, a variable that equals one if the teacher and student are different races/ethnicities, and equals zero if they are the same race/ethnicity. All racial/ethnic determinations are based on the parent-reported race/ethnicity of the child and teachers’ self-reported race/ethnicity. In order to disentangle which groups are driving results, additional analyses replace this nonmatch indicator with a series of binary variables denoting teacher–student racial/ethnic combinations: nonwhite teacher–nonwhite student, white teacher–nonwhite student, and nonwhite teacher–white student, with white teacher–white student as the reference group. The primary independent variable for research question two is the percentage of teachers at the school that are the same race/ethnicity as the student. To observe heterogeneity by race/ethnicity, I also include terms that interact this “percent match” variable with student race/ethnicity.
Finally, I include a number of control variables to reduce the potential for bias and increase the precision of estimates. At the student level, I include parent-reported child race/ethnicity, gender, whether English is the primary language spoken at home, whether the child is a twin, age, disability status, and whether the child has an individualized education plan (IEP), as previous research has shown that parents of children who are doing well in school may have different patterns of involvement than those that have a learning disability (e.g., Hornby and Lafaele 2011). At the family level, I control for household size, maternal depression (measured at the first wave),4 whether the parent is married, and employment status (full time, part time, not in labor force, and unemployed) as these factors likely affect the amount of time parents have to be involved (e.g., Heymann and Earle 2000). I also use an indicator for whether the parent believes their child will attend college (measured at the first wave), and a series of dummy variables to control for the highest level of education in the household (measured at first wave: less than high school, high school diploma, some college, and a college degree or more), as previous research shows that more educated parents are generally more involved in their children’s education (e.g., Bianchi 2000; Kalil, Ryan, and Corey 2012). Similarly, I include a control for whether the family receives Temporary Assistance for Needy Families (TANF), and categorical dummies for family income, as lower-SES families generally have lower levels of school-based parental involvement (e.g., Lareau 2003). For research question two, which relies on school-level variation, in addition to the controls described above, I include several teacher-level controls, including teacher gender, whether the teacher has a bachelor’s degree, whether the teacher has education beyond a bachelor’s degree, teacher race/ethnicity, and teacher years of experience.
Although ordinary least squares (OLS) regression estimation of the effect of different-race/ethnicity teachers on parental involvement can identify descriptive differences by racial/ethnic alignment, because of unobserved omitted variables, results do not isolate a causal relationship in such an approach. If such omitted variables are correlated both with parental involvement and the way teachers and students are assigned to one another—that is, if teachers and students are non-randomly matched through, for example, parents asking for particular teachers or moving to areas with certain school or teacher types (e.g., Holme 2002)—these omissions may confound the effects of interest by conflating same-race/ethnicity effects with some other parent–teacher selection criteria.
Therefore, my identification strategy for research question one relies on within-classroom diversity to compare the parental involvement of different race/ethnicity students with the same teacher. That is, I use a classroom fixed effects model to control for unobserved teacher factors in a given year and classroom, and therefore produce within-classroom effects of different-race/ethnicity teacher on parental involvement.5,6 Similarly, for research question two, I rely on within-school-and-year diversity to compare the parental involvement of different race/ethnicity students with the same school-level teacher racial/ethnic representation using a school-by-year fixed effects model. Within both of these models, I control for a myriad of student and parent characteristics (and teacher characteristics, in the school-by-year fixed effects model) that are theoretically related to both teacher or school selection and parental involvement, described in above. All analyses use ECLS-K provided sampling weights and robust standard errors, clustered at the school level to make statistical inference robust to serial correlation within schools.
For research question one, I separately analyze the practical and statistical significance of an indicator of representative bureaucracy—parent–teacher racial/ethnic mismatch—on each of the five dependent variables separately in a series of OLS regressions. Linear probability models (LPM) are preferred to logit models for binary outcome variables first because the proposed model contains a large number of dummy variables, which may bias maximum likelihood estimates (Wooldridge 2013), and second because interaction terms, described below, are more easily interpretable when using LPMs.7 Specifically, I estimate regressions of the form:
where Y is the parental involvement outcome for student i in classroom j in year t; NONMATCH is a binary indicator that student i has a different-race/ethnicity teacher; X is a vector of student- and family-level controls; α is a classroom fixed effect; and ε is an idiosyncratic error term.
To examine heterogeneity by racial/ethnic group for research question one, I replace nonmatch in the primary specification with a series of dummies indicating the racial/ethnic composition of each teacher–student dyad. Specifically, I estimate regressions of the form:
where Y, i, j, t, X, α, and ε are the same as in equation (1), NWNW equals one if both the teacher and student are nonwhite, and zero otherwise; WNW equals one if the teacher is white and the student is nonwhite, and zero otherwise; and NWW equals one if the teacher is nonwhite and the student is white, and zero otherwise.
Finally, to better understand heterogeneity by race/ethnicity and income, I augment equation (1) with a triple interaction term: NONMATCH×NONWHITE×LOWINCOME, where LOWINCOME equals one if the family income is less than $20,000/year.
For research question two, I analyze the practical and statistical significance of a school-level indicator of representative bureaucracy—the proportion of school teaching staff that is of the same race/ethnicity as the parent—above and beyond individual level matching on each of the five dependent variables separately in a series of OLS regressions. Specifically, I estimate regressions of the form:
where Y is the parental involvement outcome for student i in school s and year t, X now includes teacher-level control variables and nonmatch, along with the student-level controls described above, α is a school-by-year fixed effect, and ε is an idiosyncratic error term. To examine heterogeneity by racial/ethnic group for research question two, I augment equation (3) with two interaction terms: PercentMatch×black and PercentMatch×Hispanic. Finally, to examine heterogeneity by race/ethnicity and income, as in research question one, I augment equation (3) with a triple interaction term: PercentMatch×NONWHITE×LOWINCOME.8
Results
Table 1 presents summary statistics for the analytic sample overall, by race/ethnicity, and by match status. Overall rates of parental involvement range from about 79% of parents initiating contact with teachers to about 96% of parents returning calls and emails from teachers. In general, parents of white students have higher rates of involvement than those of nonwhite students. An expected pattern also emerges across match/nonmatch categories: parents of students with same-race/ethnicity teachers have higher rates of involvement than parents of students with different-race/ethnicity teachers. About 64% of the sample is white, 12.6% is black, and 23.4% is Hispanic. Because the majority of teachers are white, black and Hispanic students are much more likely to have a different-race/ethnicity teacher than white students. Maternal depression, household size, and TANF receipt is higher among nonwhite parents, and income, parental education, and marriage rates are generally lower.
. | Full Sample . | White . | Nonwhite . | Match . | Nonmatch . |
---|---|---|---|---|---|
Contact | 78.7% | 82.6% | 71.7% | 81.4% | 71.5% |
Meeting | 87.8% | 91.1% | 82.8% | 89.8% | 83.0% |
Return call | 95.5% | 97.6% | 91.3% | 96.7% | 91.9% |
Conference, teacher report | 93.5% | 96.1% | 88.6% | 94.5% | 90.4% |
Conference, parent report | 92.6% | 94.2% | 89.6% | 93.1% | 91.0% |
Nonmatch | 27.4% | 4.1% | 68.8% | 0.0% | 100.0% |
Black | 12.6% | 0.0% | 34.9% | 5.0% | 32.7% |
Hispanic | 23.4% | 0.0% | 65.1% | 10.5% | 57.7% |
White | 64.0% | 100.0% | 0.0% | 84.5% | 9.6% |
Female student | 48.7% | 48.4% | 49.3% | 48.6% | 49.1% |
English not spoken at home | 0.7% | 0.4% | 1.1% | 0.4% | 1.5% |
Parent expects child to attend college | 80.3% | 77.9% | 84.5% | 79.7% | 82.0% |
Maternal depression | 12.6% | 10.3% | 16.8% | 11.3% | 16.3% |
Child has IEP | 9.1% | 9.3% | 8.8% | 8.9% | 9.8% |
Age (in months) | 84.7 | 85.2 | 83.7 | 85.0 | 83.7 |
Disability | 17.3% | 18.8% | 14.7% | 17.8% | 16.2% |
Income less than 20k | 16.6% | 8.1% | 31.8% | 12.8% | 26.7% |
Income 20k–35k | 17.3% | 11.8% | 27.2% | 14.5% | 25.0% |
Income 35k–75k | 31.6% | 32.1% | 30.7% | 31.2% | 32.6% |
Income 75k+ | 35.5% | 46.7% | 15.4% | 41.2% | 20.2% |
Parent education less than HS | 6.4% | 1.3% | 15.5% | 4.5% | 11.7% |
Parent education HS degree | 18.9% | 12.7% | 29.9% | 15.5% | 27.8% |
Parent education some college | 34.0% | 34.0% | 33.8% | 33.5% | 35.3% |
Parent education college + | 40.7% | 51.9% | 20.8% | 46.6% | 25.2% |
Parent one FT employed | 44.3% | 44.4% | 44.0% | 44.2% | 44.5% |
Parent one PT employed | 23.1% | 26.5% | 17.2% | 24.7% | 19.0% |
Parent one unemployed | 5.5% | 3.4% | 9.1% | 4.3% | 8.6% |
Parent one not in labor force | 27.1% | 25.6% | 29.6% | 26.8% | 27.9% |
Parent two FT employed | 69.3% | 77.5% | 54.7% | 74.0% | 56.9% |
Parent two PT employed | 4.7% | 4.4% | 5.2% | 4.6% | 5.0% |
Parent two unemployed | 2.3% | 1.8% | 3.3% | 2.1% | 3.1% |
Parent two not in labor force | 3.6% | 3.0% | 4.7% | 3.2% | 4.8% |
Parents married | 75.3% | 83.5% | 60.5% | 80.1% | 62.3% |
Household size | 4.6 | 4.5 | 4.8 | 4.6 | 4.7 |
TANF receipt | 4.2% | 2.0% | 8.2% | 3.1% | 7.3% |
Observations | 12,980 | 9,060 | 3,920 | 10,020 | 2,960 |
. | Full Sample . | White . | Nonwhite . | Match . | Nonmatch . |
---|---|---|---|---|---|
Contact | 78.7% | 82.6% | 71.7% | 81.4% | 71.5% |
Meeting | 87.8% | 91.1% | 82.8% | 89.8% | 83.0% |
Return call | 95.5% | 97.6% | 91.3% | 96.7% | 91.9% |
Conference, teacher report | 93.5% | 96.1% | 88.6% | 94.5% | 90.4% |
Conference, parent report | 92.6% | 94.2% | 89.6% | 93.1% | 91.0% |
Nonmatch | 27.4% | 4.1% | 68.8% | 0.0% | 100.0% |
Black | 12.6% | 0.0% | 34.9% | 5.0% | 32.7% |
Hispanic | 23.4% | 0.0% | 65.1% | 10.5% | 57.7% |
White | 64.0% | 100.0% | 0.0% | 84.5% | 9.6% |
Female student | 48.7% | 48.4% | 49.3% | 48.6% | 49.1% |
English not spoken at home | 0.7% | 0.4% | 1.1% | 0.4% | 1.5% |
Parent expects child to attend college | 80.3% | 77.9% | 84.5% | 79.7% | 82.0% |
Maternal depression | 12.6% | 10.3% | 16.8% | 11.3% | 16.3% |
Child has IEP | 9.1% | 9.3% | 8.8% | 8.9% | 9.8% |
Age (in months) | 84.7 | 85.2 | 83.7 | 85.0 | 83.7 |
Disability | 17.3% | 18.8% | 14.7% | 17.8% | 16.2% |
Income less than 20k | 16.6% | 8.1% | 31.8% | 12.8% | 26.7% |
Income 20k–35k | 17.3% | 11.8% | 27.2% | 14.5% | 25.0% |
Income 35k–75k | 31.6% | 32.1% | 30.7% | 31.2% | 32.6% |
Income 75k+ | 35.5% | 46.7% | 15.4% | 41.2% | 20.2% |
Parent education less than HS | 6.4% | 1.3% | 15.5% | 4.5% | 11.7% |
Parent education HS degree | 18.9% | 12.7% | 29.9% | 15.5% | 27.8% |
Parent education some college | 34.0% | 34.0% | 33.8% | 33.5% | 35.3% |
Parent education college + | 40.7% | 51.9% | 20.8% | 46.6% | 25.2% |
Parent one FT employed | 44.3% | 44.4% | 44.0% | 44.2% | 44.5% |
Parent one PT employed | 23.1% | 26.5% | 17.2% | 24.7% | 19.0% |
Parent one unemployed | 5.5% | 3.4% | 9.1% | 4.3% | 8.6% |
Parent one not in labor force | 27.1% | 25.6% | 29.6% | 26.8% | 27.9% |
Parent two FT employed | 69.3% | 77.5% | 54.7% | 74.0% | 56.9% |
Parent two PT employed | 4.7% | 4.4% | 5.2% | 4.6% | 5.0% |
Parent two unemployed | 2.3% | 1.8% | 3.3% | 2.1% | 3.1% |
Parent two not in labor force | 3.6% | 3.0% | 4.7% | 3.2% | 4.8% |
Parents married | 75.3% | 83.5% | 60.5% | 80.1% | 62.3% |
Household size | 4.6 | 4.5 | 4.8 | 4.6 | 4.7 |
TANF receipt | 4.2% | 2.0% | 8.2% | 3.1% | 7.3% |
Observations | 12,980 | 9,060 | 3,920 | 10,020 | 2,960 |
Note: All differences between white and nonwhite are statistically significant except female, IEP, income 35k–75k, and parent education some college. All differences between match and nonmatch are statistically significant except female, twin, disability, income 35k–75k, parent education some college, parent one full-time employment, and parent two part-time employment. Sample sizes vary by dependent variable; see table 2 for analytic sample size by dependent variable. Descriptive statistics for control variables were calculated on the analytic sample for the dependent variable Contact, though remain qualitatively similar across analytic samples. All sample sizes are rounded to the nearest 10 to comply with private-use ECLS-K rules.
. | Full Sample . | White . | Nonwhite . | Match . | Nonmatch . |
---|---|---|---|---|---|
Contact | 78.7% | 82.6% | 71.7% | 81.4% | 71.5% |
Meeting | 87.8% | 91.1% | 82.8% | 89.8% | 83.0% |
Return call | 95.5% | 97.6% | 91.3% | 96.7% | 91.9% |
Conference, teacher report | 93.5% | 96.1% | 88.6% | 94.5% | 90.4% |
Conference, parent report | 92.6% | 94.2% | 89.6% | 93.1% | 91.0% |
Nonmatch | 27.4% | 4.1% | 68.8% | 0.0% | 100.0% |
Black | 12.6% | 0.0% | 34.9% | 5.0% | 32.7% |
Hispanic | 23.4% | 0.0% | 65.1% | 10.5% | 57.7% |
White | 64.0% | 100.0% | 0.0% | 84.5% | 9.6% |
Female student | 48.7% | 48.4% | 49.3% | 48.6% | 49.1% |
English not spoken at home | 0.7% | 0.4% | 1.1% | 0.4% | 1.5% |
Parent expects child to attend college | 80.3% | 77.9% | 84.5% | 79.7% | 82.0% |
Maternal depression | 12.6% | 10.3% | 16.8% | 11.3% | 16.3% |
Child has IEP | 9.1% | 9.3% | 8.8% | 8.9% | 9.8% |
Age (in months) | 84.7 | 85.2 | 83.7 | 85.0 | 83.7 |
Disability | 17.3% | 18.8% | 14.7% | 17.8% | 16.2% |
Income less than 20k | 16.6% | 8.1% | 31.8% | 12.8% | 26.7% |
Income 20k–35k | 17.3% | 11.8% | 27.2% | 14.5% | 25.0% |
Income 35k–75k | 31.6% | 32.1% | 30.7% | 31.2% | 32.6% |
Income 75k+ | 35.5% | 46.7% | 15.4% | 41.2% | 20.2% |
Parent education less than HS | 6.4% | 1.3% | 15.5% | 4.5% | 11.7% |
Parent education HS degree | 18.9% | 12.7% | 29.9% | 15.5% | 27.8% |
Parent education some college | 34.0% | 34.0% | 33.8% | 33.5% | 35.3% |
Parent education college + | 40.7% | 51.9% | 20.8% | 46.6% | 25.2% |
Parent one FT employed | 44.3% | 44.4% | 44.0% | 44.2% | 44.5% |
Parent one PT employed | 23.1% | 26.5% | 17.2% | 24.7% | 19.0% |
Parent one unemployed | 5.5% | 3.4% | 9.1% | 4.3% | 8.6% |
Parent one not in labor force | 27.1% | 25.6% | 29.6% | 26.8% | 27.9% |
Parent two FT employed | 69.3% | 77.5% | 54.7% | 74.0% | 56.9% |
Parent two PT employed | 4.7% | 4.4% | 5.2% | 4.6% | 5.0% |
Parent two unemployed | 2.3% | 1.8% | 3.3% | 2.1% | 3.1% |
Parent two not in labor force | 3.6% | 3.0% | 4.7% | 3.2% | 4.8% |
Parents married | 75.3% | 83.5% | 60.5% | 80.1% | 62.3% |
Household size | 4.6 | 4.5 | 4.8 | 4.6 | 4.7 |
TANF receipt | 4.2% | 2.0% | 8.2% | 3.1% | 7.3% |
Observations | 12,980 | 9,060 | 3,920 | 10,020 | 2,960 |
. | Full Sample . | White . | Nonwhite . | Match . | Nonmatch . |
---|---|---|---|---|---|
Contact | 78.7% | 82.6% | 71.7% | 81.4% | 71.5% |
Meeting | 87.8% | 91.1% | 82.8% | 89.8% | 83.0% |
Return call | 95.5% | 97.6% | 91.3% | 96.7% | 91.9% |
Conference, teacher report | 93.5% | 96.1% | 88.6% | 94.5% | 90.4% |
Conference, parent report | 92.6% | 94.2% | 89.6% | 93.1% | 91.0% |
Nonmatch | 27.4% | 4.1% | 68.8% | 0.0% | 100.0% |
Black | 12.6% | 0.0% | 34.9% | 5.0% | 32.7% |
Hispanic | 23.4% | 0.0% | 65.1% | 10.5% | 57.7% |
White | 64.0% | 100.0% | 0.0% | 84.5% | 9.6% |
Female student | 48.7% | 48.4% | 49.3% | 48.6% | 49.1% |
English not spoken at home | 0.7% | 0.4% | 1.1% | 0.4% | 1.5% |
Parent expects child to attend college | 80.3% | 77.9% | 84.5% | 79.7% | 82.0% |
Maternal depression | 12.6% | 10.3% | 16.8% | 11.3% | 16.3% |
Child has IEP | 9.1% | 9.3% | 8.8% | 8.9% | 9.8% |
Age (in months) | 84.7 | 85.2 | 83.7 | 85.0 | 83.7 |
Disability | 17.3% | 18.8% | 14.7% | 17.8% | 16.2% |
Income less than 20k | 16.6% | 8.1% | 31.8% | 12.8% | 26.7% |
Income 20k–35k | 17.3% | 11.8% | 27.2% | 14.5% | 25.0% |
Income 35k–75k | 31.6% | 32.1% | 30.7% | 31.2% | 32.6% |
Income 75k+ | 35.5% | 46.7% | 15.4% | 41.2% | 20.2% |
Parent education less than HS | 6.4% | 1.3% | 15.5% | 4.5% | 11.7% |
Parent education HS degree | 18.9% | 12.7% | 29.9% | 15.5% | 27.8% |
Parent education some college | 34.0% | 34.0% | 33.8% | 33.5% | 35.3% |
Parent education college + | 40.7% | 51.9% | 20.8% | 46.6% | 25.2% |
Parent one FT employed | 44.3% | 44.4% | 44.0% | 44.2% | 44.5% |
Parent one PT employed | 23.1% | 26.5% | 17.2% | 24.7% | 19.0% |
Parent one unemployed | 5.5% | 3.4% | 9.1% | 4.3% | 8.6% |
Parent one not in labor force | 27.1% | 25.6% | 29.6% | 26.8% | 27.9% |
Parent two FT employed | 69.3% | 77.5% | 54.7% | 74.0% | 56.9% |
Parent two PT employed | 4.7% | 4.4% | 5.2% | 4.6% | 5.0% |
Parent two unemployed | 2.3% | 1.8% | 3.3% | 2.1% | 3.1% |
Parent two not in labor force | 3.6% | 3.0% | 4.7% | 3.2% | 4.8% |
Parents married | 75.3% | 83.5% | 60.5% | 80.1% | 62.3% |
Household size | 4.6 | 4.5 | 4.8 | 4.6 | 4.7 |
TANF receipt | 4.2% | 2.0% | 8.2% | 3.1% | 7.3% |
Observations | 12,980 | 9,060 | 3,920 | 10,020 | 2,960 |
Note: All differences between white and nonwhite are statistically significant except female, IEP, income 35k–75k, and parent education some college. All differences between match and nonmatch are statistically significant except female, twin, disability, income 35k–75k, parent education some college, parent one full-time employment, and parent two part-time employment. Sample sizes vary by dependent variable; see table 2 for analytic sample size by dependent variable. Descriptive statistics for control variables were calculated on the analytic sample for the dependent variable Contact, though remain qualitatively similar across analytic samples. All sample sizes are rounded to the nearest 10 to comply with private-use ECLS-K rules.
The main results for research question one—Does having a same-race/ethnicity teacher affect parents’ school-based parental involvement?—are presented in table 2.9 For two parental involvement outcomes, there is statistically significant evidence that a difference in race/ethnicity between teacher and student reduces the likelihood a parent will engage in that form of involvement. Parents of a student with a different-race teacher are 6.2 percentage points less likely to initiate contact with the teacher (7.8%) and 5.1 percentage points less likely to self-report parent–teacher conference attendance (5.5%).
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Nonmatch | −0.062** (0.031) | −0.034 (0.027) | −0.023 (0.017) | −0.008 (0.025) | −0.051* (0.026) |
Black student | −0.018 (0.044) | −0.063* (0.036) | 0.001 (0.027) | −0.058* (0.033) | 0.055* (0.031) |
Hispanic student | 0.032 (0.033) | −0.015 (0.030) | 0.008 (0.017) | −0.008 (0.025) | 0.034 (0.027) |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.598 | 0.659 | 0.610 | 0.643 | 0.606 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Nonmatch | −0.062** (0.031) | −0.034 (0.027) | −0.023 (0.017) | −0.008 (0.025) | −0.051* (0.026) |
Black student | −0.018 (0.044) | −0.063* (0.036) | 0.001 (0.027) | −0.058* (0.033) | 0.055* (0.031) |
Hispanic student | 0.032 (0.033) | −0.015 (0.030) | 0.008 (0.017) | −0.008 (0.025) | 0.034 (0.027) |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.598 | 0.659 | 0.610 | 0.643 | 0.606 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on classroom fixed effects and the full set of controls described in the text and summarized in table 1. The estimated coefficients on these variables are reported in Supplementary Appendix Table 1.
*p < .1, **p < .05.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Nonmatch | −0.062** (0.031) | −0.034 (0.027) | −0.023 (0.017) | −0.008 (0.025) | −0.051* (0.026) |
Black student | −0.018 (0.044) | −0.063* (0.036) | 0.001 (0.027) | −0.058* (0.033) | 0.055* (0.031) |
Hispanic student | 0.032 (0.033) | −0.015 (0.030) | 0.008 (0.017) | −0.008 (0.025) | 0.034 (0.027) |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.598 | 0.659 | 0.610 | 0.643 | 0.606 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Nonmatch | −0.062** (0.031) | −0.034 (0.027) | −0.023 (0.017) | −0.008 (0.025) | −0.051* (0.026) |
Black student | −0.018 (0.044) | −0.063* (0.036) | 0.001 (0.027) | −0.058* (0.033) | 0.055* (0.031) |
Hispanic student | 0.032 (0.033) | −0.015 (0.030) | 0.008 (0.017) | −0.008 (0.025) | 0.034 (0.027) |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.598 | 0.659 | 0.610 | 0.643 | 0.606 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on classroom fixed effects and the full set of controls described in the text and summarized in table 1. The estimated coefficients on these variables are reported in Supplementary Appendix Table 1.
*p < .1, **p < .05.
Table 3 presents results from models that consider the various combinations of the teacher–student racial/ethnic dyad: nonwhite teacher–nonwhite student, white teacher–nonwhite student, and nonwhite teacher–white student, with white teacher–white student as the reference group. Table 3 reveals that compared to white students paired with white teachers, parents of nonwhite students paired with nonwhite teachers are 38.3 percentage points (41%) more likely to report conference attendance, though there are no statistically significant differences across the other involvement outcomes. Further, parents of nonwhite students paired with white teachers are less likely to be involved across a number of measures than parents of white students with white teachers (between 3.0 and 6.5 percentage points less likely). Finally, compared to parents of nonwhite students with a white teacher, parents of nonwhite students with a nonwhite teacher are 39.2 percentage points (42%) more likely to report conference attendance.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
White teacher, white student (omitted) | |||||
Nonwhite teacher, nonwhite student | 0.082 (0.217) | −0.007 (0.131) | 0.121 (0.178) | 0.052 (0.089) | 0.383* (0.212) |
White teacher, nonwhite student | −0.046** (0.021) | −0.065*** (0.022) | −0.017 (0.011) | −0.030** (0.012) | −0.009 (0.012) |
Nonwhite teacher, white student | 0.010 (0.215) | −0.007 (0.132) | 0.102 (0.175) | 0.077 (0.093) | 0.313 (0.216) |
δ(nonwhite teacher, nonwhite student) − δ(white teacher, nonwhite student) | 0.128 | 0.058 | 0.138 | 0.082 | 0.392* |
p value on F-test | 0.558 | 0.668 | 0.441 | 0.364 | 0.063 |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.598 | 0.659 | 0.611 | 0.642 | 0.607 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
White teacher, white student (omitted) | |||||
Nonwhite teacher, nonwhite student | 0.082 (0.217) | −0.007 (0.131) | 0.121 (0.178) | 0.052 (0.089) | 0.383* (0.212) |
White teacher, nonwhite student | −0.046** (0.021) | −0.065*** (0.022) | −0.017 (0.011) | −0.030** (0.012) | −0.009 (0.012) |
Nonwhite teacher, white student | 0.010 (0.215) | −0.007 (0.132) | 0.102 (0.175) | 0.077 (0.093) | 0.313 (0.216) |
δ(nonwhite teacher, nonwhite student) − δ(white teacher, nonwhite student) | 0.128 | 0.058 | 0.138 | 0.082 | 0.392* |
p value on F-test | 0.558 | 0.668 | 0.441 | 0.364 | 0.063 |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.598 | 0.659 | 0.611 | 0.642 | 0.607 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on classroom fixed effects and the full set of controls described in the text and summarized in table 1.
*p < .1, **p < .05, ***p < .01.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
White teacher, white student (omitted) | |||||
Nonwhite teacher, nonwhite student | 0.082 (0.217) | −0.007 (0.131) | 0.121 (0.178) | 0.052 (0.089) | 0.383* (0.212) |
White teacher, nonwhite student | −0.046** (0.021) | −0.065*** (0.022) | −0.017 (0.011) | −0.030** (0.012) | −0.009 (0.012) |
Nonwhite teacher, white student | 0.010 (0.215) | −0.007 (0.132) | 0.102 (0.175) | 0.077 (0.093) | 0.313 (0.216) |
δ(nonwhite teacher, nonwhite student) − δ(white teacher, nonwhite student) | 0.128 | 0.058 | 0.138 | 0.082 | 0.392* |
p value on F-test | 0.558 | 0.668 | 0.441 | 0.364 | 0.063 |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.598 | 0.659 | 0.611 | 0.642 | 0.607 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
White teacher, white student (omitted) | |||||
Nonwhite teacher, nonwhite student | 0.082 (0.217) | −0.007 (0.131) | 0.121 (0.178) | 0.052 (0.089) | 0.383* (0.212) |
White teacher, nonwhite student | −0.046** (0.021) | −0.065*** (0.022) | −0.017 (0.011) | −0.030** (0.012) | −0.009 (0.012) |
Nonwhite teacher, white student | 0.010 (0.215) | −0.007 (0.132) | 0.102 (0.175) | 0.077 (0.093) | 0.313 (0.216) |
δ(nonwhite teacher, nonwhite student) − δ(white teacher, nonwhite student) | 0.128 | 0.058 | 0.138 | 0.082 | 0.392* |
p value on F-test | 0.558 | 0.668 | 0.441 | 0.364 | 0.063 |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.598 | 0.659 | 0.611 | 0.642 | 0.607 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on classroom fixed effects and the full set of controls described in the text and summarized in table 1.
*p < .1, **p < .05, ***p < .01.
Results examining heterogeneity by race and income simultaneously are presented in table 4. The estimates show that nonwhite, low income students with a teacher of a different race/ethnicity are much less likely to initiate contact with a teacher, attend a meeting with the teacher, or return a teacher’s calls or emails compared to matched, higher-income, white students. In other words, having a teacher of a different race/ethnicity reduces parental involvement more for nonwhite, low-income students compared to white students and higher-income, nonwhite students.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Nonmatch × nonwhite × low income | −0.381** (0.149) | −0.344** (0.164) | −0.153** (0.070) | −0.099 (0.135) | 0.020 (0.085) |
Nonmatch × nonwhite | 0.068 (0.124) | 0.063 (0.142) | −0.027 (0.141) | −0.042 (0.164) | −0.173 (0.192) |
Nonmatch × low income | 0.306** (0.134) | 0.257* (0.154) | 0.042 (0.032) | 0.059 (0.124) | 0.028 (0.066) |
Nonwhite × low income | 0.127** (0.057) | 0.090* (0.054) | 0.072 (0.053) | 0.026 (0.045) | −0.006 (0.050) |
Nonmatch | −0.104 (0.071) | −0.071 (0.077) | 0.003 (0.066) | 0.018 (0.083) | 0.034 (0.093) |
Nonwhite | −0.010 (0.070) | −0.053 (0.076) | 0.016 (0.078) | −0.001 (0.087) | 0.123 (0.105) |
Low income | −0.182*** (0.037) | −0.065* (0.039) | −0.044* (0.026) | −0.040 (0.024) | −0.008 (0.026) |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.601 | 0.660 | 0.613 | 0.643 | 0.607 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Nonmatch × nonwhite × low income | −0.381** (0.149) | −0.344** (0.164) | −0.153** (0.070) | −0.099 (0.135) | 0.020 (0.085) |
Nonmatch × nonwhite | 0.068 (0.124) | 0.063 (0.142) | −0.027 (0.141) | −0.042 (0.164) | −0.173 (0.192) |
Nonmatch × low income | 0.306** (0.134) | 0.257* (0.154) | 0.042 (0.032) | 0.059 (0.124) | 0.028 (0.066) |
Nonwhite × low income | 0.127** (0.057) | 0.090* (0.054) | 0.072 (0.053) | 0.026 (0.045) | −0.006 (0.050) |
Nonmatch | −0.104 (0.071) | −0.071 (0.077) | 0.003 (0.066) | 0.018 (0.083) | 0.034 (0.093) |
Nonwhite | −0.010 (0.070) | −0.053 (0.076) | 0.016 (0.078) | −0.001 (0.087) | 0.123 (0.105) |
Low income | −0.182*** (0.037) | −0.065* (0.039) | −0.044* (0.026) | −0.040 (0.024) | −0.008 (0.026) |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.601 | 0.660 | 0.613 | 0.643 | 0.607 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on classroom fixed effects and the full set of controls described in the text and summarized in table 1.
*p < .1, **p < .05, ***p < .01.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Nonmatch × nonwhite × low income | −0.381** (0.149) | −0.344** (0.164) | −0.153** (0.070) | −0.099 (0.135) | 0.020 (0.085) |
Nonmatch × nonwhite | 0.068 (0.124) | 0.063 (0.142) | −0.027 (0.141) | −0.042 (0.164) | −0.173 (0.192) |
Nonmatch × low income | 0.306** (0.134) | 0.257* (0.154) | 0.042 (0.032) | 0.059 (0.124) | 0.028 (0.066) |
Nonwhite × low income | 0.127** (0.057) | 0.090* (0.054) | 0.072 (0.053) | 0.026 (0.045) | −0.006 (0.050) |
Nonmatch | −0.104 (0.071) | −0.071 (0.077) | 0.003 (0.066) | 0.018 (0.083) | 0.034 (0.093) |
Nonwhite | −0.010 (0.070) | −0.053 (0.076) | 0.016 (0.078) | −0.001 (0.087) | 0.123 (0.105) |
Low income | −0.182*** (0.037) | −0.065* (0.039) | −0.044* (0.026) | −0.040 (0.024) | −0.008 (0.026) |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.601 | 0.660 | 0.613 | 0.643 | 0.607 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Nonmatch × nonwhite × low income | −0.381** (0.149) | −0.344** (0.164) | −0.153** (0.070) | −0.099 (0.135) | 0.020 (0.085) |
Nonmatch × nonwhite | 0.068 (0.124) | 0.063 (0.142) | −0.027 (0.141) | −0.042 (0.164) | −0.173 (0.192) |
Nonmatch × low income | 0.306** (0.134) | 0.257* (0.154) | 0.042 (0.032) | 0.059 (0.124) | 0.028 (0.066) |
Nonwhite × low income | 0.127** (0.057) | 0.090* (0.054) | 0.072 (0.053) | 0.026 (0.045) | −0.006 (0.050) |
Nonmatch | −0.104 (0.071) | −0.071 (0.077) | 0.003 (0.066) | 0.018 (0.083) | 0.034 (0.093) |
Nonwhite | −0.010 (0.070) | −0.053 (0.076) | 0.016 (0.078) | −0.001 (0.087) | 0.123 (0.105) |
Low income | −0.182*** (0.037) | −0.065* (0.039) | −0.044* (0.026) | −0.040 (0.024) | −0.008 (0.026) |
Observations | 12,980 | 10,330 | 11,820 | 13,170 | 13,530 |
R2 | 0.601 | 0.660 | 0.613 | 0.643 | 0.607 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on classroom fixed effects and the full set of controls described in the text and summarized in table 1.
*p < .1, **p < .05, ***p < .01.
Primary results for research question two are presented in table 5.10 Unlike the teacher-level results, overall school-level representation does not have a significant effect on parental involvement across any outcome. In other words, having a higher percentage of teachers of the same race/ethnicity as the student at that school is not associated with the likelihood that students’ parents are involved at the school.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race | 0.000 (0.001) | −0.001 (0.001) | 0.000 (0.000) | −0.000 (0.000) | −0.001 (0.001) |
Nonmatch | −0.011 (0.041) | −0.036 (0.032) | −0.019 (0.019) | −0.031* (0.016) | −0.012 (0.020) |
Black student | 0.003 (0.060) | −0.065 (0.065) | −0.020 (0.030) | −0.071* (0.039) | −0.042 (0.062) |
Hispanic student | −0.001 (0.053) | −0.038 (0.060) | 0.009 (0.023) | −0.032 (0.032) | −0.101 (0.062) |
Observations | 5,730 | 4,420 | 5,230 | 5,810 | 5,960 |
R2 | 0.310 | 0.349 | 0.337 | 0.345 | 0.363 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race | 0.000 (0.001) | −0.001 (0.001) | 0.000 (0.000) | −0.000 (0.000) | −0.001 (0.001) |
Nonmatch | −0.011 (0.041) | −0.036 (0.032) | −0.019 (0.019) | −0.031* (0.016) | −0.012 (0.020) |
Black student | 0.003 (0.060) | −0.065 (0.065) | −0.020 (0.030) | −0.071* (0.039) | −0.042 (0.062) |
Hispanic student | −0.001 (0.053) | −0.038 (0.060) | 0.009 (0.023) | −0.032 (0.032) | −0.101 (0.062) |
Observations | 5,730 | 4,420 | 5,230 | 5,810 | 5,960 |
R2 | 0.310 | 0.349 | 0.337 | 0.345 | 0.363 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on school-by-year fixed effects and the full set of controls described in the text and summarized in table 1. The estimated coefficients on these variables are reported in Supplementary Appendix Table 3.
*p < .1.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race | 0.000 (0.001) | −0.001 (0.001) | 0.000 (0.000) | −0.000 (0.000) | −0.001 (0.001) |
Nonmatch | −0.011 (0.041) | −0.036 (0.032) | −0.019 (0.019) | −0.031* (0.016) | −0.012 (0.020) |
Black student | 0.003 (0.060) | −0.065 (0.065) | −0.020 (0.030) | −0.071* (0.039) | −0.042 (0.062) |
Hispanic student | −0.001 (0.053) | −0.038 (0.060) | 0.009 (0.023) | −0.032 (0.032) | −0.101 (0.062) |
Observations | 5,730 | 4,420 | 5,230 | 5,810 | 5,960 |
R2 | 0.310 | 0.349 | 0.337 | 0.345 | 0.363 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race | 0.000 (0.001) | −0.001 (0.001) | 0.000 (0.000) | −0.000 (0.000) | −0.001 (0.001) |
Nonmatch | −0.011 (0.041) | −0.036 (0.032) | −0.019 (0.019) | −0.031* (0.016) | −0.012 (0.020) |
Black student | 0.003 (0.060) | −0.065 (0.065) | −0.020 (0.030) | −0.071* (0.039) | −0.042 (0.062) |
Hispanic student | −0.001 (0.053) | −0.038 (0.060) | 0.009 (0.023) | −0.032 (0.032) | −0.101 (0.062) |
Observations | 5,730 | 4,420 | 5,230 | 5,810 | 5,960 |
R2 | 0.310 | 0.349 | 0.337 | 0.345 | 0.363 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on school-by-year fixed effects and the full set of controls described in the text and summarized in table 1. The estimated coefficients on these variables are reported in Supplementary Appendix Table 3.
*p < .1.
However, estimates for models that examine heterogeneity in this relationship by race/ethnicity, presented in table 6, show a more nuanced story. Percent same race/ethnicity differentially increases parent-reported parent conference attendance for parents of black and Hispanic students. For example, if the percentage of black teachers at a school increases by 10 percentage points, black parents are about 9 percentage points more likely to report they attended a parent–teacher conference. This effect is the same for Hispanic students. Also, parents of white students are less likely to report attending a parent–teacher conference as the percentage of white teachers increases. Finally, table 7 presents estimates for models that examine heterogeneity in the relationship between percent match and parental involvement by race/ethnicity and income simultaneously. Unlike the results in the heterogeneity analysis of direct representation in research question one, here there is no evidence that school-level representation differentially affects parent involvement by race and income.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race × black | 0.001 (0.004) | −0.001 (0.005) | −0.000 (0.002) | 0.001 (0.003) | 0.009*** (0.004) |
Percent same race × Hispanic | 0.001 (0.004) | 0.003 (0.005) | −0.001 (0.001) | 0.003 (0.002) | 0.009** (0.004) |
Percent same race | −0.000 (0.002) | −0.002 (0.003) | 0.000 (0.001) | −0.001 (0.001) | −0.005*** (0.002) |
Nonmatch | −0.011 (0.041) | −0.036 (0.031) | −0.020 (0.019) | −0.031* (0.016) | −0.009 (0.018) |
Black student | −0.032 (0.151) | −0.143 (0.233) | 0.007 (0.052) | −0.149 (0.097) | −0.445** (0.178) |
Hispanic student | −0.038 (0.155) | −0.149 (0.243) | 0.039 (0.052) | −0.123 (0.101) | −0.494*** (0.183) |
Observations | 5,730 | 4,420 | 5,230 | 5,810 | 5,960 |
R2 | 0.310 | 0.350 | 0.337 | 0.345 | 0.366 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race × black | 0.001 (0.004) | −0.001 (0.005) | −0.000 (0.002) | 0.001 (0.003) | 0.009*** (0.004) |
Percent same race × Hispanic | 0.001 (0.004) | 0.003 (0.005) | −0.001 (0.001) | 0.003 (0.002) | 0.009** (0.004) |
Percent same race | −0.000 (0.002) | −0.002 (0.003) | 0.000 (0.001) | −0.001 (0.001) | −0.005*** (0.002) |
Nonmatch | −0.011 (0.041) | −0.036 (0.031) | −0.020 (0.019) | −0.031* (0.016) | −0.009 (0.018) |
Black student | −0.032 (0.151) | −0.143 (0.233) | 0.007 (0.052) | −0.149 (0.097) | −0.445** (0.178) |
Hispanic student | −0.038 (0.155) | −0.149 (0.243) | 0.039 (0.052) | −0.123 (0.101) | −0.494*** (0.183) |
Observations | 5,730 | 4,420 | 5,230 | 5,810 | 5,960 |
R2 | 0.310 | 0.350 | 0.337 | 0.345 | 0.366 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on school-by-year fixed effects and the full set of controls described in the text and summarized in table 1.
*p < .1, **p < .05, ***p < .01.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race × black | 0.001 (0.004) | −0.001 (0.005) | −0.000 (0.002) | 0.001 (0.003) | 0.009*** (0.004) |
Percent same race × Hispanic | 0.001 (0.004) | 0.003 (0.005) | −0.001 (0.001) | 0.003 (0.002) | 0.009** (0.004) |
Percent same race | −0.000 (0.002) | −0.002 (0.003) | 0.000 (0.001) | −0.001 (0.001) | −0.005*** (0.002) |
Nonmatch | −0.011 (0.041) | −0.036 (0.031) | −0.020 (0.019) | −0.031* (0.016) | −0.009 (0.018) |
Black student | −0.032 (0.151) | −0.143 (0.233) | 0.007 (0.052) | −0.149 (0.097) | −0.445** (0.178) |
Hispanic student | −0.038 (0.155) | −0.149 (0.243) | 0.039 (0.052) | −0.123 (0.101) | −0.494*** (0.183) |
Observations | 5,730 | 4,420 | 5,230 | 5,810 | 5,960 |
R2 | 0.310 | 0.350 | 0.337 | 0.345 | 0.366 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race × black | 0.001 (0.004) | −0.001 (0.005) | −0.000 (0.002) | 0.001 (0.003) | 0.009*** (0.004) |
Percent same race × Hispanic | 0.001 (0.004) | 0.003 (0.005) | −0.001 (0.001) | 0.003 (0.002) | 0.009** (0.004) |
Percent same race | −0.000 (0.002) | −0.002 (0.003) | 0.000 (0.001) | −0.001 (0.001) | −0.005*** (0.002) |
Nonmatch | −0.011 (0.041) | −0.036 (0.031) | −0.020 (0.019) | −0.031* (0.016) | −0.009 (0.018) |
Black student | −0.032 (0.151) | −0.143 (0.233) | 0.007 (0.052) | −0.149 (0.097) | −0.445** (0.178) |
Hispanic student | −0.038 (0.155) | −0.149 (0.243) | 0.039 (0.052) | −0.123 (0.101) | −0.494*** (0.183) |
Observations | 5,730 | 4,420 | 5,230 | 5,810 | 5,960 |
R2 | 0.310 | 0.350 | 0.337 | 0.345 | 0.366 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on school-by-year fixed effects and the full set of controls described in the text and summarized in table 1.
*p < .1, **p < .05, ***p < .01.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race × nonwhite × low income | 0.004 (0.003) | 0.001 (0.002) | 0.000 (0.002) | 0.002 (0.002) | 0.001 (0.002) |
Percent same race × nonwhite | 0.001 (0.001) | −0.001 (0.001) | −0.000 (0.001) | −0.000 (0.001) | 0.001 (0.001) |
Percent same race × low income | −0.004 (0.003) | −0.000 (0.002) | −0.000 (0.002) | −0.001 (0.002) | −0.001 (0.001) |
Nonwhite × low income | −0.306 (0.320) | −0.031 (0.193) | 0.012 (0.170) | −0.099 (0.151) | −0.071 (0.151) |
Nonwhite | −0.036 (0.046) | −0.010 (0.038) | −0.005 (0.021) | −0.008 (0.020) | −0.019 (0.021) |
Low income | 0.247 (0.317) | 0.006 (0.193) | −0.005 (0.174) | 0.049 (0.148) | 0.097 (0.148) |
Nonmatch | −0.016 (0.040) | −0.036 (0.029) | −0.021 (0.019) | −0.028* (0.016) | 0.001 (0.017) |
Observations | 5,820 | 4,500 | 5,320 | 5,900 | 6,050 |
R2 | 0.313 | 0.350 | 0.337 | 0.345 | 0.361 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race × nonwhite × low income | 0.004 (0.003) | 0.001 (0.002) | 0.000 (0.002) | 0.002 (0.002) | 0.001 (0.002) |
Percent same race × nonwhite | 0.001 (0.001) | −0.001 (0.001) | −0.000 (0.001) | −0.000 (0.001) | 0.001 (0.001) |
Percent same race × low income | −0.004 (0.003) | −0.000 (0.002) | −0.000 (0.002) | −0.001 (0.002) | −0.001 (0.001) |
Nonwhite × low income | −0.306 (0.320) | −0.031 (0.193) | 0.012 (0.170) | −0.099 (0.151) | −0.071 (0.151) |
Nonwhite | −0.036 (0.046) | −0.010 (0.038) | −0.005 (0.021) | −0.008 (0.020) | −0.019 (0.021) |
Low income | 0.247 (0.317) | 0.006 (0.193) | −0.005 (0.174) | 0.049 (0.148) | 0.097 (0.148) |
Nonmatch | −0.016 (0.040) | −0.036 (0.029) | −0.021 (0.019) | −0.028* (0.016) | 0.001 (0.017) |
Observations | 5,820 | 4,500 | 5,320 | 5,900 | 6,050 |
R2 | 0.313 | 0.350 | 0.337 | 0.345 | 0.361 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on school-by-year fixed effects and the full set of controls described in the text and summarized in table 1.
*p < .1.
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race × nonwhite × low income | 0.004 (0.003) | 0.001 (0.002) | 0.000 (0.002) | 0.002 (0.002) | 0.001 (0.002) |
Percent same race × nonwhite | 0.001 (0.001) | −0.001 (0.001) | −0.000 (0.001) | −0.000 (0.001) | 0.001 (0.001) |
Percent same race × low income | −0.004 (0.003) | −0.000 (0.002) | −0.000 (0.002) | −0.001 (0.002) | −0.001 (0.001) |
Nonwhite × low income | −0.306 (0.320) | −0.031 (0.193) | 0.012 (0.170) | −0.099 (0.151) | −0.071 (0.151) |
Nonwhite | −0.036 (0.046) | −0.010 (0.038) | −0.005 (0.021) | −0.008 (0.020) | −0.019 (0.021) |
Low income | 0.247 (0.317) | 0.006 (0.193) | −0.005 (0.174) | 0.049 (0.148) | 0.097 (0.148) |
Nonmatch | −0.016 (0.040) | −0.036 (0.029) | −0.021 (0.019) | −0.028* (0.016) | 0.001 (0.017) |
Observations | 5,820 | 4,500 | 5,320 | 5,900 | 6,050 |
R2 | 0.313 | 0.350 | 0.337 | 0.345 | 0.361 |
. | (1) . | (2) . | (3) . | (4) . | (5) . |
---|---|---|---|---|---|
Contact . | Meeting . | Return Call . | Conference, Teacher Report . | Conference, Parent Report . | |
Percent same race × nonwhite × low income | 0.004 (0.003) | 0.001 (0.002) | 0.000 (0.002) | 0.002 (0.002) | 0.001 (0.002) |
Percent same race × nonwhite | 0.001 (0.001) | −0.001 (0.001) | −0.000 (0.001) | −0.000 (0.001) | 0.001 (0.001) |
Percent same race × low income | −0.004 (0.003) | −0.000 (0.002) | −0.000 (0.002) | −0.001 (0.002) | −0.001 (0.001) |
Nonwhite × low income | −0.306 (0.320) | −0.031 (0.193) | 0.012 (0.170) | −0.099 (0.151) | −0.071 (0.151) |
Nonwhite | −0.036 (0.046) | −0.010 (0.038) | −0.005 (0.021) | −0.008 (0.020) | −0.019 (0.021) |
Low income | 0.247 (0.317) | 0.006 (0.193) | −0.005 (0.174) | 0.049 (0.148) | 0.097 (0.148) |
Nonmatch | −0.016 (0.040) | −0.036 (0.029) | −0.021 (0.019) | −0.028* (0.016) | 0.001 (0.017) |
Observations | 5,820 | 4,500 | 5,320 | 5,900 | 6,050 |
R2 | 0.313 | 0.350 | 0.337 | 0.345 | 0.361 |
Note: Robust SEs, clustered at school level, in parentheses; All regressions are weighted by ECLS-K provided weights and condition on school-by-year fixed effects and the full set of controls described in the text and summarized in table 1.
*p < .1.
Discussion and Conclusion
Although the substantive effects of racial/ethnic representation have been consistently demonstrated empirically, the mechanisms underlying these effects remain under debate. In particular, there is little agreement about whether benefits derive primarily from direct interaction with a bureaucrat of one’s same race/ethnicity or from indirect representation at the organizational level, or both. This study contributes to the ongoing disentangling of this theoretical question by analyzing the substantial effects of representation at both levels in the school context. In particular, I ask whether teacher and/or school level racial/ethnic representation is associated with parental involvement in schools.
Results indicate that both types of representation matter, though perhaps not equally. Direct representation has a larger association with parental involvement outcomes, especially for nonwhite students. For example, compared to parents of nonwhite students with a white teacher, parents of nonwhite students with a nonwhite teacher are 39.2 percentage points (42%) more likely to report conference attendance. After controlling for this direct representation, indirect representation shows statistically significant but smaller increases in parent-reported conference attendance. A 10 percentage point increase in the number of black teachers at a school is associated with a 9 percentage point increase the likelihood that a black parent attends a parent–teacher conference. The results are the same for Hispanic students.
Work in representative bureaucracy, then, should continue to consider the effects of representation at both the individual and the aggregate levels. Researchers should also consider the possibility that mechanisms are context dependent. The organization type here—a school—provides a particular pattern of interactions between clients (children and parents) and bureaucrats (teachers), along with specific structures of socialization and input from teachers. One can imagine that different types of organizations have patterns of interaction and socialization that enable a different direct versus indirect representation dynamic for clients. Future research should explore organization level characteristics that might explain such contingencies.
These findings also have implications for the study of parental involvement in education, especially given that the results are largest and most robust for parents of nonwhite, low-income students compared to students in higher-income households and white students. Parental involvement is not a silver bullet for closing the achievement gap; there are many other factors, including income, wealth, segregation, and poor quality schools, to name a few, that contribute to existing racial/ethnic and SES-based achievement gaps. Still, given recent experimental evidence connecting parental involvement and children’s academic outcomes, understanding barriers to involvement for minority and lower-income families continues to be an important area of exploration. The results presented here—paired with past research—suggest that creating more representative and inclusive school environments may improve home–school relationships, ultimately benefiting children and their families.
In terms of public policy, there are at least three potential recommendations. First, like other representative bureaucracy research, these results imply that more underrepresented teachers be recruited, particularly given the changing demographics of children in American schools. In a recent paper, Putman and colleagues (2016) finds obstacles to having more black and Hispanic teachers at every point in the pipeline: majoring in education, getting hired as a teacher, and staying in teaching year after year. To increase representation, therefore, every point along this pipeline must be addressed. A second, more immediate-term policy implication is more teacher professional development that emphasizes cultural competency and effective outreach to parents. Such an approach would ease the potentially negative effect of nonmatch. Finally, especially in schools with low levels of representation, parental involvement policies and programs should be designed to be more sensitive to culture. In particular, programs could be better designed to empower and encourage parents who come from a variety of backgrounds with a variety of understandings about what constitutes parental involvement.
Although the findings here are suggestive of a relationship between representation and parental involvement in schools, results should be interpreted within the study’s limitations. First, the analysis relies on teacher- and parent-reported parental involvement. These tend to have high rates of positive values, thus limiting variability. Further, for various social and memory reasons, these measures may be inaccurate and/or inflated. If parents or teachers nonrandomly misreport involvement by race, the results will be biased. There may be evidence of this, given some differences in the significance of results for teacher- versus parent-reported parent-conference attendance. This measurement error may be exacerbated by the use of a fixed effects model. Second, because of a lack of variation in the race/ethnicity of teachers in the sample, I am only able to look at three racial/ethnic groups: white, black, and Hispanic. There are obviously many groups missing from this analysis. Further, racial/ethnic identity is complicated and may not be appropriately captured in these simple categories. Third, although my method allows me to control for within-classroom effects, and I have many important child- and family-level control variables, I am not perfectly controlling for all child or parent characteristics. If these unobserved attributes are correlated both with classroom assignment and parental involvement, my results will be biased. Further, although classroom fixed effects models allow me to control for unobserved teacher and school factors, it relies on within-classroom diversity to generate results. This may reduce the external validity of my findings, as I am identifying off of a sample of children in classrooms with at least one student of a different race/ethnicity. It is possible the results may not apply to children in segregated schools and classrooms. This is important given that rates of school racial segregation are high and increasing in recent years (Orfield and Frankenberg 2014).
Supplementary Material
Supplementary material is available at the Journal of Public Administration Research and Theory online (https://academic.oup.com/jpart).
References
———.
———.
———.
———.
——.
Footnotes
Relatedly, the translation from passive to active may be especially strong in regions where race is especially salient (Grissom, Nicholson-Crotty, and Nicholson-Crotty 2009).
These represent currently available waves of data; collection will continue through the Spring of the students’ 5th grade year.
Similarly, for research question two I limit the sample to schools with at least two students, as that identification strategy relies on school-by-year fixed effects.
In the ECLS-K, mothers’ depressive symptoms were measured using a modified, 12-item version of the Center for Epidemiologic Studies Depressive Scale (CES-D) (Radloff 1977). Following other research, a threshold score of 9 or higher was used to define mothers with moderate or severe depressive symptoms (Paulson, Dauber, and Leiferman 2006).
For the purposes of the current analysis, student- and parent-level are the same.
Student fixed effects models yield qualitatively similar overall results.
Supplementary Appendix Tables 2 and 4 present fixed effect logit estimates (Chamberlain 1980) and estimates of the primary LPM for the restricted sample of classrooms (for research question one) and schools (for research question two) that experienced variation in each dependent variable. The FE-logit coefficient magnitude cannot be directly compared to the LPM coefficients, nor can precise average partial effects (APE) be computed because values of the FE are unobserved. However, scaled coefficients that are approximations of APE, which can be compared to the LPM coefficients, can be approximated using the product of the sample average of Pr (Y = 1) and one minus this probability as a scaling factor. The resulting APEs are qualitatively similar to the LPM results.
Unfortunately, due to data limitations, this study cannot test the effect of SES-based representation; that is, I cannot identify whether students and teachers are from the same socioeconomic background. However, it is still useful to understand whether more or less disadvantaged students are driving the results.
Full results are presented in Supplementary Appendix Table 1.
Full results are presented in Supplementary Appendix Table 3.