Choice,Preferences,and Constraints

Abstract

Does ‘‘choosing a home’’ still matter for ‘‘choosing a school,’’ despite implementation of school choice policies designed to weaken this link? Prior research shows how the presence of such policies does little to solve the problems of stratification and segregation associated with residentially based enrollment systems, since families differ along racial/ethnic and socioeconomic lines in their access to, and how they participate in, the school choice process. We examine how families’ nearby school supply shapes and constrains their choices. Drawing on a unique dataset consisting of parents’ ranked preferences from among one urban district’s full menu of public schools, we find that Hispanic, white, and black parents share a strong preference for academic performance, but differences in their choices can be traced to variation in nearby supply. Our findings illustrate how the vastly different sets of schools from which parents can choose reproduce race-based patterns of stratification.

Keywords

school choice parent preferences stratification race/ethnicity

Across the United States, the momentum for parents to take more control over their children’s education has been building. School choice policies are a major—if controversial—component of this movement. On the one hand, giving parents the option to choose schools for their children has the potential to reduce racial and socioeconomic segregation and enhance equity in public education (Archbald 2004; Betts 2005). Such opportunities for school choice have long been available to families who could afford to move to areas serviced by high-performing schools or enroll their students in private schools. Choice policies aim to extend those opportunities to all families regardless of socioeconomic status or where a family happens to live in a district (Berends and Zottola 2009; Coons and Sugarman 1978; Goyette 2014; Hoxby 1998; Schneider, Teske, and Marschall 2002; Weininger 2014). On the other hand, school choice can lead to more segregated schools (Kimelberg and Billingham 2013; Saporito and Sohini 2007). Prior research has explored a number of reasons why this has been the case—namely, the restricted access to exercising choice among nonwhite and lower-income families (Holme 2002; Lauen 2007; Saporito and Lareau 1999), and white families’ tendency to avoid schools with higher proportions of nonwhite families (Bifulco, Ladd, and Ross 2009a; Saporito 2003; Saporito and Sohini 2007).

In addition to the ability to exercise choice and the preferences of those who do, the efficacy of school choice policies to positively affect students’ opportunities to access higher-performing public education environments rests on the availability of those options. The removal or loosening of barriers, such as residentially determined attendance boundaries, is not sufficient. Families must also have reasonable and realistic access to a variety of alternatives to their default or neighborhood public schools. As a result of longstanding and persistent patterns in residential segregation, the supply of schools around nonwhite and lower-income families—and thus the public school options from which they are likely to choose— tends to be lower performing than the supply near white and higher-income families (Jargowsky 2014; Kozol 2005; Lubienski, Gulosino, and Weitzel 2009).

To better evaluate the equity potential of school choice, we need to understand how the substance of parents’ choices is shaped by these contextual supply dynamics. We address this question by capitalizing on a unique dataset generated by parents’ applications to attend nonassigned schools from among any of the public options in Denver Public Schools (DPS). Specifically, we focus on parents making enrollment decisions at two key points in their children’s educational trajectories: the transitions from elementary to middle school (i.e., from fifth to sixth grade) and from middle to high school (i.e., from eighth to ninth grade). Denver provides a useful context given its high level of choice participation and its robust availability of choice options. Our analyses shed light on the school attributes parents value when they select schools for their children, how school choices differ by race/ethnicity, and how Hispanic, white, and black parents’ school choices compare to and are shaped by the supply of schools around them.

Background

For much of the twentieth century, students’ school assignments have been largely based on the neighborhoods in which their families decided to live (Lareau and Goyette 2014; Lauen 2007). More advantaged families have long been able to exercise school choice by virtue of their ability to buy homes in neighborhoods with higher-performing schools, and research indicates that these families do indeed take the reputation and quality of neighborhood schools into account when deciding where to live (Lareau 2014). Not only are lower-income and nonwhite families unable to move into many of the neighborhoods serviced by schools in higher demand, but their residential moves tend to be motivated by reasons other than educational opportunities or resources (Jargowsky 2014; Rhodes and DeLuca 2014). This means that students of different racial/ethnic and socioeconomic backgrounds attend different schools, and their schools are unequal in terms of performance (Logan, Minca, and Adar 2012).

The expansion of school choice policies, however, aims to extend the opportunity to make school enrollment decisions to families regardless of socioeconomic resources and racial/ethnic background (Berends and Zottola 2009; Coons and Sugarman 1978; Hoxby 1998; Schneider et al. 2002). In other words, proponents argue that school choice decouples “choosing a school” from “choosing a home” (Lareau and Goyette 2014). School choice policies have their roots in the desegregation movement of the 1960s, when mandatory and voluntary school reassignment policies—like busing and magnet schools—weakened the longstanding relationship between students’ home addresses and their assigned public schools (Lauen 2007). Since then, many districts across the country have adopted some form of school choice, from allowing a student to transfer from his or her local school to another school within (or, in some cases, outside of) his or her own district to providing vouchers to private schools and opening new public options like charter schools (Betts 2005).

By allowing more parents to act on their preferences and choose where to send their children within the public school system, school choice policies in their many forms are expected to improve students’ educational opportunities and outcomes. Proponents contend that these policies “liberate” nonwhite and lower-income families from the residential constraints they face by opening up school options outside their neighborhoods (Archbald 2004). Although residential segregation has declined in recent decades, many metropolitan areas remain quite segregated by race/ethnicity and economic status (Jargowsky 2014; Logan 2013; Massey and Denton 1993; Reardon and Bischoff 2011). Racial and socioeconomic segregation is particularly high in metropolitan areas fragmented by numerous school districts, as more advantaged parents’ residential preferences reflect district-level resources (Bischoff 2008; Clotfelter 1999; Reardon et al. 2008). These segregation patterns are reproduced in public schools when districts assign students to schools based on their neighborhoods: lower-income and minority children are more likely than their more affluent, white counterparts to be trapped in lower-performing schools within a given district or in lower-performing (typically urban) school districts within a city (Bischoff 2008; Fiel 2013; Goyette, Farrie, and Freely 2012; Logan et al. 2012; Saporito and Sohini 2007).

According to the liberation model of school choice, giving all parents the ability to choose their children’s schools has the potential to weaken the link between residential neighborhood and public school assignment and thereby reduce racial/ethnic and socioeconomic school segregation (Archbald 2004; Bifulco, Ladd, and Ross 2009b). By decoupling school preferences from residential preferences, parents across racial/ethnic and socioeconomic groups can make choices that need not necessarily reflect the supply of schools in their immediate neighborhoods. Instead, parents have the opportunity to select schools they believe will provide good academic environments for their children, regardless of where they live (Coons and Sugarman 1978; Schneider et al. 2002).

But realizing this hypothesized equity potential of school choice is far from automatic. Prior research highlights two ways in which school choice policies fall short of alleviating the problems of segregation and stratification associated with residentially based school assignment policies. First, not every family has equal opportunity to participate in the school choice process. School choice systems are often overly complex, decentralized, and opaque (Bell 2009; Roda and Wells 2013). The exercise of choice requires parents to “be informed about their children’s school options, to make choices about a range of options, and to follow school registration procedures and timelines” (Archbald 2004:284). These procedures can be particularly difficult to navigate for nonwhite and lower-income families, who are less likely to have access to the kinds of information and social networks that could help them participate in the choice process and select schools (Holme 2002; Lareau 2014; Mickelson and Southworth 2005). Decentralized choice systems also leave open the possibility that schools, in managing their own application and admission procedures, find wider latitude for counseling out certain students or being more selective than is strictly allowed under law (Jennings 2010). As a result of these barriers, white, more affluent families are generally more likely than minority and lower-income families to participate in school choice (Bifulco et al. 2009b; Holme 2002; Lauen 2007; Saporito 2003; Saporito and Lareau 1999; Saporito and Sohini 2007).

Second, given the opportunity to participate, if parents choose schools based not on educational principles (e.g., a school’s ability to improve student achievement) but instead on other nonacademic criteria (e.g., a school’s socioeconomic or racial/ethnic composition), then school choice could further reify and exacerbate patterns of stratification and segregation in public education (Bifulco et al. 2009b; Renzulli and Evans 2005; Saporito and Sohini 2006). Parents generally report a preference for academic quality (Buckley and Schneider 2003; Phillips, Hausman, and Larsen 2012; Reback 2008), but they are also sensitive to a school’s racial and socioeconomic composition (Goyette et al. 2012; Henig 1996; Renzulli and Evans 2005; Saporito and Lareau 1999; Saporito and Sohini 2006; Sohini and Saporito 2009). Patterns of out-group avoidance are particularly prevalent among white and more affluent families (Bifulco et al. 2009b). White families’ propensity to avoid schools with higher proportions of nonwhite students persists even when accounting for other school traits that often serve as proxies for racial composition, such as school test scores and safety (Billingham and Hunt 2016; Saporito 2003). Saporito and Lareau (1999) find that white families appear to engage in a two-step decision-making process, in which they first rule out schools based on this single criterion (the percentage of black families in a school)—even if this means dropping relatively high-performing and safe schools from consideration—and then they go on to consider other criteria. White parents may also use new public school options, like charter schools, to “flee” district schools that have higher proportions of nonwhite students (Renzulli and Evans 2005; Saporito and Sohini 2006).

Given these dynamics in terms of who exercises school choice and which school attributes they select on when they do, school choice policies can increase inequity and segregation by race/ethnicity and socioeconomic status compared to the outcome if students went to their neighborhood-assigned schools (Bifulco et al. 2009b; Renzulli and Evans 2005; Saporito and Sohini 2006).

The degree to which school choice is effective in achieving the hypothesized outcomes of the liberation model—namely, the relative increase in access to higher-performing schools across racial/ethnic and socioeconomic groups—also depends on the extent to which families have viable alternatives to their default or neighborhood public schools. It is not sufficient that districts remove barriers such as residentially determined attendance zones or that parents participate in the choice process, particularly when minority and lower-income families are more likely to live near lower-performing schools and to have inadequate access to transportation to and from potentially desirable schools (Hastings, Kane, and Staiger 2006; Jargowsky 2014; Kozol 2005; Logan et al. 2012; Lubienski et al. 2009). In this article, we evaluate the equity potential of school choice by examining how such contextual supply dynamics shape and constrain parents’ preferences. Prior research compellingly shows that parents’ choices differ along racial/ethnic lines— for example, in terms of parents’ sensitivity to the racial composition of hypothetical schools (Billingham and Hunt 2016) or the schools they choose to enroll in or leave (Renzulli and Evans 2005; Saporito and Sohini 2006). We compare parents’ choices across racial/ethnic groups and to the supply of schools nearby and within the district at large. To carry out this research objective, we need (1) data that reflect parents’ actual preferences and choices and (2) a research setting in which a large number and wide variety of alternatives are at least ostensibly made available from which parents can choose.

Leveraging Public School Applications

Much of what we know about parents’ school preferences comes from surveys about parents’ ideal or hypothetical schools (e.g., Billingham and Hunt 2016; Goyette 2008; Kleitz et al. 2000; Roda and Wells 2013; Schneider et al. 1998; Tedin and Weiher 2004; Weiher and Tedin 2002), from where a student ultimately enrolled (e.g., Phillips et al. 2012), or from whether families decided to move a student to a different school (e.g., Rich and Jennings 2015). This represents a real limitation to our understanding of what parents want from and value in schools, and thus the implications for the equity potential of school choice policies remain unclear. Parents may feel compelled to give socially desirable responses on surveys, and surveys do not require parents to make real trade-offs in school attributes (Jacob and Lefgren 2007; Weiher and Tedin 2002). Furthermore, inferring preferences from where parents ultimately enroll their children—be it a district-run, charter, magnet, or other school of choice—assumes that parents considered only that choice, or at least that the school was preferable to all other schools they could have chosen. It also conflates access with preference; factors other than wanting to enroll a child in a particular school (e.g., a school’s capacity) affect enrollment.

Some innovative work capitalizes on parents’ revealed preferences. For example, Schneider and Buckley (2002) examine the search behavior of parents who accessed a website that provides comprehensive data on all public schools in Washington, D.C. Most similar to our own approach is research that uses data from public school applications (e.g., Hastings et al. 2006; Hastings, Kane, and Staiger 2009; Phillips, Larsen, and Hausman 2015; Saporito 2003; Saporito and Lareau 1999), although these studies have limitations as well. Some look only at a single choice listed on an application (Saporito 2003; Saporito and Lareau 1999), at applications to specific programs like magnet schools (Saporito 2003), or at enrollment application processes that exclude important school options such as charter schools (Hastings et al. 2006, 2009; Phillips et al. 2015). In short, studies generally do not take into account the full range of public school options—including district-run, charter, and magnet schools—available to parents.

To carry out our evaluation of the extent to which school choice truly liberates families from the constraints of a residentially determined enrollment policy, we examine parents’ preferences as revealed by their applications to one urban district’s central office. Parents could select and rank up to five of any of the district-run, charter, and magnet public schools. This feature of our data is especially useful, because exploring “the school characteristics that drive preferences requires observing the choices of families with a wide variety of schools to choose from” (Saporito and Lareau 1999:420). From these applications, we know not only whether a school was a parent’s top choice but also what other choices they considered from among all the available public options and the relative ranking of those choices.

The Choice Context in DPS

In the fall of 2012, DPS enrolled roughly 84,000 students in nearly 200 schools. Parents have a high degree of choice at their disposal: Not only can they apply to any district-run school, but the district has a large number of charter and magnet schools. About 40 charter schools enrolled 14 percent of the student population. Parents can also select from around 20 magnet or specialized programs, such as dual-language, Montessori, and expeditionary learning.¹

Both DPS and the city of Denver are coterminous with Denver County, and the demographics of the district generally reflect those of the county. DPS is a majority-minority district; about 60 percent of its students are Hispanic, one fifth are white, and 14 percent are black. Denver County’s population under the age of 18 is 50 percent Hispanic, 32 percent white, and 11 percent white (data come from the five-year 2010 to 2014 American Community Survey). Some of this discrepancy may be due to the rather high number of Denver residents who attend private schools (about 11,000 in 2012) and to Colorado’s interdistrict choice policy (begun in 1994). In 2012, about 5,000 students who lived outside the district attended DPS schools, and a little more than 7,000 students who lived within the DPS boundary attended schools in other districts (data are available from the Colorado Department of Education [http://cde.state.co.us]). Littleton, for example, a school district on DPS’s southern border serving 16,000 students, enrolled about 900 students who lived within DPS’s boundaries; 74 percent of Littleton’s students were white, and only 17 percent were Hispanic. This provides some evidence of white flight to suburban districts (Logan and Burdick-Will forthcoming; Renzulli and Evans 2005; Saporito and Sohini 2007). Other districts neighboring DPS, however, had quite a bit of variation in their racial/ethnic makeup. In the Adams-Arapahoe school district, which is just east of Denver and enrolled nearly 700 Denver residents in 2012, 54 percent of the 40,000 students were Hispanic, and 19 percent were white.

Parents’ participation in school choice, particularly among parents whose children are transitioning from elementary to middle school (i.e., from fifth to sixth grade) and from middle to high school (i.e., from eighth to ninth grade), appears to be relatively more equitable in Denver than prior research has found elsewhere (Holme 2002; Lauen 2007; Phillips et al. 2015; Saporito and Lareau 1999). Table 1 presents summary statistics for the population of DPS students enrolled in fifth and eighth grades during the 2011–2012 and 2012–2013 school years, this study’s population of interest, by their racial/ethnic group and whether their parents submitted a school choice application in spring 2012 or spring 2013 for enrollment in the following academic year. White parents were slightly more likely to submit a choice application (69 percent) than were Hispanic (65 percent) or black (64 percent) parents, but the overall large portion of parents submitting applications suggests that parents across racial/ethnic groups exercised school choice. Indeed, the overall percentage of students in the district who were Hispanic (60 percent), white (19 percent), and black (14 percent) closely mirrors the racial/ethnic breakdown of students whose parents submitted a choice application. Within-race participation rates were similarly even among this population of students. For instance, about 90 percent of Hispanic students whose parents did and did not apply were eligible for free or reduced-price lunch (FRL), and 50 percent of Hispanic applicants were English-language learners (ELLs), compared to 48 percent overall.

Table 1.

Descriptive Statistics.

	Overall		Hispanic Families		White Families		Black Families
Category	Apps	All	Apps	All	Apps	All	Apps	All
School choice participation
Submitted choice application		0.66		0.65		0.69		0.64
Number of choices listed	3.09		3.11		2.67		3.56
Student characteristics
Hispanic	0.59	0.60
White	0.19	0.19
Black	0.14	0.14
Other race	0.07	0.07
Female	0.50	0.49	0.50	0.49	0.50	0.50	0.51	0.50
Free or reduced-price lunch	0.74	0.75	0.90	0.90	0.23	0.28	0.83	0.83
English-language learner	0.34	0.33	0.50	0.48	0.05	0.05	0.11	0.10
Special education	0.10	0.12	0.10	0.13	0.08	0.09	0.14	0.16
Standardized math test score	−0.24	−0.32	−0.45	−0.51	0.61	0.50	−0.73	−0.75
In grade 5	0.56	0.54	0.54	0.54	0.63	0.56	0.51	0.50
In grade 8	0.44	0.46	0.46	0.46	0.37	0.44	0.49	0.50
Current school characteristics
Average School Performance Framework rating^a	3.38	3.33	3.17	3.13	3.98	3.93	3.28	3.23
Average proportion Hispanic or black	0.72	0.72	0.83	0.82	0.43	0.45	0.73	0.73
N student-year observations	15,195	23,134	8.998	13,800	2,972	4,305	2,119	3,319

Note: Apps = school choice applicants. Data consist of the population of students enrolled in fifth and eighth grade who attended school in Denver Public Schools in 2011–2012 and 2012–2013; data are pooled across the two school years. The sample in the second and third columns contains students in racial/ethnic groups other than Hispanic, white, or black (N = 1,710, 1,106—or 65 percent—of whom submitted choice applications). Apps applied in the spring of 2012 and 2013 for enrollment in the next school year. Unless otherwise noted, values are proportions.

We calculated average School Performance Framework by assigning each rating a numerical value (5 = distinguished, 4 = meets expectations, 3 = on watch, 2 = on priority watch, 1 = on probation) and excluding unrated schools.

Furthermore, Hispanic parents tended to apply to about one half more schools, on average, than did white parents, and black parents applied to about one additional school compared to white parents. Other researchers have found that when disadvantaged (i.e., minority and lower-income) families do exercise school choice, their choices tend to be less frequent than those made by more advantaged (i.e., white and higher-income) families (Bifulco et al. 2009b; Goyette 2008), but this appears to be less of a concern in Denver.

One reason for the relative parity in the exercise of choice may be the particularly robust choice environment in Denver, which aligns well with many of the conditions underlying school choice theory (Archbald 2004; Betts 2005). In addition to the large number of “consumers” (i.e., parents making choices, as evidenced by the high participation rates among parents of rising sixth and ninth graders [see Table 1]), the district has a large number of “suppliers,” that is, the number and variety of public schools from which parents can choose. As in other districts, Denver’s intradistrict transfer policy means school choice options are available across all neighborhoods and social contexts within the district (Phillips et al. 2012; Phillips et al. 2015). Furthermore, Denver made a recent strategic decision to broaden its school options and the accessibility of its choice system. In spring 2012, the district implemented a centralized enrollment process to alleviate many of the barriers associated with the exercise of school choice. The expansion of magnet, citywide, and charter schools in districts across the country means that in many cities, individual schools operate their own enrollment procedures, creating a complicated patchwork of processes for parents to navigate. Many of the problems with school choice can be traced to the resulting complex responsibilities foisted on parents (Archbald 2004).

Under Denver’s centralized enrollment system, parents have to submit only a single application form, on which they specify up to five choices from among any of the public schools in the district for which they are eligible. Application submissions and assignment announcements across all schools occur on a common timeline with a single date for each step in the process, and all student-school matches are made through a single, agreed-upon algorithm. This centralization lowers sector differences between schools, compared to districts where applying to a charter, district-run, or magnet school involves disparate processes. Furthermore, parents do not have to reapply each year; when children are admitted to a school, they may stay at that school until they reach the end of its grade-span without taking further action. Many other districts have intradistrict choice options, and some—such as New Orleans, New York, Boston, and San Francisco—also have centralized enrollment systems that feature common applications, but their systems often exclude certain schools (e.g., several of New Orleans’ high-performing schools opt out of participation in the centralized process) or are limited to district-run schools. Denver’s system, on the other hand, encompasses all of its public schools, and thus the complete set of public alternatives from which parents can choose, including district-run schools, magnet or specialized programs, and public charter schools.

Alongside its centralized enrollment process, the district has built a robust information system, including an interactive website where parents can explore their school options. The district also produces printed and online guidebooks that contain a common set of metrics on each school in the district, including information about school programs, performance, and student populations. These guidebooks, which focus group and interview data suggest are utilized by parents (Gross, DeArmond, and Denice 2015), are available in English and Spanish. The district also hosts a series of school fairs and sponsors a number of resource centers around the city to provide support for parents as they choose schools.

These features of Denver’s choice environment—numerous consumers and suppliers, an accessible and transparent process, and available information—suggest that participation in school choice should be relatively diffused among a wide swath of Denver’s families, advantaged and disadvantaged alike. The participation metrics in Table 1 generally bear this out.

At the same time, the realities of geography in Denver, as in many other urban districts, mean that schools are unevenly distributed across the city, which likely has implications for the content of and constraints on parents’ choices. This can be seen in Figure 1, which illustrates the racial/ethnic makeup of the city and the location of middle schools and high schools rated in the lowest two categories of the district’s School Performance Framework (SPF; “on probation” and “on priority watch”) in the first year in which parents could submit an application through the district’s centralized enrollment system. Roughly a quarter of schools were rated in these lowest two categories in 2012, and they were clustered in census tracts in which the majority of the population was Hispanic or black. Indeed, about one fifth of Hispanic and black students attended low-rated schools in 2012, compared to only about 4 percent of white students. At the other end of the spectrum, just 4 percent of Hispanic and black students, but 10 percent of white students, attended schools in the top SPF category (schools rated “distinguished”).

Figure 1.

Spatial distribution of schools serving sixth and ninth grade in Denver, 2012

The context of open, centralized enrollment; racial/ethnic residential segregation; and clustering of underperforming schools in Denver provide useful conditions for evaluating the extent to which the supply of schools in a neighborhood shapes and constrains the choices families make for their children’s schooling.

Data and Methods

We draw on two years of student-level enrollment and school application data provided by DPS. These data include a record for every student enrolled in each of the 2011–2012 and 2012–2013 academic years, as well as an application from every parent who applied in the spring 2012 and spring 2013 choice processes. As such, we analyze population data pooled across two years. We selected these two years because 2011–2012 was the first year in which parents could submit an application through the district’s centralized choice system and because these are the most recent years for which we have enrollment and student testing information.

For each student in the district, our data include the school in which a student was presently enrolled; demographic characteristics such as race/ethnicity, gender, a student’s eligibility for FRL, whether the student was enrolled in special education (SPED), and whether the student was an ELL; and the student’s scores on state math and reading assessments, which are standardized by grade, subject, and year. For students whose parents submitted a school choice application, these data also include home addresses and the schools ranked by their parents. School-level information—such as schools’ average standardized test scores, ratings on the district’s SPF, student composition (the proportion of students who are Hispanic or black), and physical address—is linked to students’ current and (for students whose parents submitted an application) requested schools. The district’s SPF is a rating system that assigns schools to one of five categories based on a wide variety of measures, including academic proficiency and growth, student engagement, and parent satisfaction. In descending order, the ratings are “distinguished,”“meets expectations,”“on watch,”“on priority watch,” and “on probation;” a sixth category captures schools that are “unrated” due to lack of sufficient data or having just recently opened. Schools are further categorized as public charter schools, magnet schools, or other specialized programs. Indicators for charter and magnet schools come from the National Center for Education Statistics’ Common Core of Data, and we rely on a list of specialized schools, provided to the authors by the district, that includes dual-language, expeditionary, and Montessori programs. We combined magnet and the district’s specialized schools into a single indicator because both refer to programs with a particular focus or curricular approach. Finally, although we lack a direct measure of school safety, we operationalized the relative safety of a school’s campus by the crime rate in the area surrounding the school. We aggregated incident-level crime data from Denver’s Open Data Catalog (http://data.denvergov.org) to each school’s block-group for the year during which parents were making their selections (2012 and 2013). We then calculated the crime rate per 1,000 residents using population estimates from the five-year 2010 to 2014 American Community Survey.

A small amount of test score data was missing. For nine schools across the two years, we imputed missing average standardized test scores using multivariate regression that included controls for school-level student composition (proportions of students eligible for FRL or in SPED, who are ELL, and by race); whether the school was a charter, magnet, or other specialized program; the crime rate per 1,000 residents of a school’s block-group; the school’s SPF rating; and year fixed effects. A small proportion of students (1.3 percent) was also missing standardized test scores. We imputed scores for these students with a multivariate regression model controlling for students’ race; gender; FRL, SPED, and ELL statuses; grade level; and school and year fixed effects. Sensitivity analyses show no substantive differences in descriptive and multivariate results if we exclude student and school records that had missing testing information, or if we instead imputed missing information using the sample means.

We focus on the choices made by the population of parents whose children currently attended a DPS school and were applying for enrollment in the sixth and ninth grades the next fall. We do this for two reasons. First, substantially more parents submit an application on behalf of their children entering the sixth and ninth grades (nearly 70 percent) than for other elementary, middle, and high school grade levels (less than 10 percent).² Second, the transition into middle and high school grade levels represents a natural decision point that has important implications for students’ schooling (Lauen 2007), as rising sixth and ninth graders are often required to switch schools or academic programs as they leave the elementary or middle school grades behind. Parents submitting applications in other, nontransitional grade levels—which, again, tends to be quite rare—may be doing so for particular reasons. For instance, they may be considering a mid-grade-span switch due to discontent with their current schooling or due to a residential move that takes them too far from their current school. Choices made by parents of students transitioning to sixth grade (middle school) and ninth grade (high school) are likely to be more general. Results from supplemental analyses that include students entering all grade levels, and from models estimated separately for students entering sixth and ninth grades, are substantively similar to those reported here and are available from the authors upon request.

Analytic Strategy

In addition to presenting descriptive comparisons between the schools parents selected and the supply of schools nearby (i.e., within a two-mile radius) and in the district at large, we examine parents’ preferences for particular school attributes using rank-ordered logit models (Allison and Christakis 1994). Researchers in various disciplines have used these models to analyze how decision makers combine attributes of alternatives into overall evaluations. Sociologists have used these models to estimate preferences for job and occupational characteristics (Halaby 2003; Kalleberg and Marsden 2013; Zhou 2005) as well as intergroup friendship choices (Zeng and Xie 2008).

The rank-ordered logit model is based on the conditional logit model, and it allows for estimation of differences among individuals’ preferences for the ranked alternatives. Each parent $i$ assigns a unique integer rank $Y_{ij}$ from 1 (the highest rank) to $J$ (the lowest rank, which in our case is 5) to each school. Based on a random utility model, we assume that parent $i$ has a certainty utility $U_{ij}$ for each school $j$ . Each $U_{ij}$ is the sum of a systematic component $μ_{ij}$ and a random component $ε_{ij}$ :

U_{ij} = μ_{ij} + ε_{ij},

where $μ_{ij}$ represents the extent to which parent $i$ prefers school $j$ to all the other schools, and $ε_{ij}$ is identically and independently distributed with an extreme-value distribution. The systematic component $μ_{ij}$ can be represented as a linear combination of a set of school attributes $X_{j}$ :

μ_{ij} = β X_{j},

where coefficients $β$ express the direction and weight of the value parents place on the school attributes. In the same way that the ordered logit model is similar to repeated applications of the usual logit model, the rank-ordered logit model can be thought of as sequential applications of the conditional logit model. Whereas the coefficients in a conditional logit model reflect the change in the log-odds of the alternative being chosen versus not chosen given a unit change in an explanatory variable, the coefficients in a rank-ordered logit model describe the change in the log-odds that alternative $j$ is ranked ahead of alternative $k$ .

Our analysis aims to capture a wide variety of attributes parents might value in a school. We include attributes describing a school’s performance (measured with school-level standardized math test scores), proximity (measured by the straight-line distance in miles between a student’s home address and the school’s address), and student composition (measured by the proportion of a school that is Hispanic or black, and its squared term) as well as indicators for remaining at the student’s current school, enrolling in a charter school, or enrolling in a magnet or specialized program (measured with dummy variables). We also control for parents’ likely concern about school safety, using the crime rate per 1,000 residents in the school’s census block-group.

In setting up the data to estimate our rank-ordered logit models, we matched every student whose parent submitted a choice application to every public school in the district—not only the schools ranked by their parents on the application—that was open in the next year, contained the student’s next grade level, and made no obvious restrictions to their student body (e.g., boys are not matched to the city’s all-girl school, the Girls Athletic Leadership School). Across the two years of data, students entering 6th grade were matched to an average of 78.5 schools, and rising 9th graders were matched to an average of 62.6 schools. It was not uncommon that students were matched to, and their parents applied to, their current schools. Of the 109 schools that students in our data could have been matched to, 24 were elementary/middle schools (with grade distributions inclusive of grades in both the PK–5 and 6–8 ranges), 28 were middle/high schools (with grades spanning the 6–8 and 9–12 ranges), and 5 served students in grade levels across the K–5, 6–8, and 9–12 ranges. Roughly 6 percent of parents of rising 6th and 9th graders listed their children’s current schools on their applications.

Schools left unranked by parents are retained in the analysis and treated by the model as less desirable than all of the selected schools (Allison and Christakis 1994). Practically, this means that unranked schools are assigned the next highest choice number. For instance, if a parent ranked four schools, all other schools in the district are given the rank of five. When parents filled in all five rankings, all other schools are given the rank of six. In the latter case, the model assumes that parents have a clear ranking for schools one through five (e.g., school one is the most preferred of all schools, school two is the most preferred of all remaining schools, and so on); we know that all schools left unranked are less desirable than the ranked schools, but the preference ordering among those unranked schools is not known.

A standard rank-ordered logit model, as represented by Equation 2, assumes that all respondents—in this case, parents choosing schools for their children—use the same valuation function (Allison and Christakis 1994). In other words, parents apply the same decision weights when ranking schools. This is likely an unreasonable assumption for our study. Prior research shows that parents make systematically different choices by race, socioeconomic status, and context (Bell 2009; Burgess et al. 2015; Hastings et al. 2006; Schneider et al. 2002; Teske and Schneider 2001). We accommodate this heterogeneity in two ways. First, we allow the coefficients of the school’s average math test score to vary with a student’s own math test score by interacting the two. This allows for the possibility that parents will seek to maximize school fit by finding a school that better matches—or possibly exceeds—their children’s own ability. Second, we estimate separate models for the three primary racial/ethnic groups in Denver: Hispanics, whites, and blacks. Much prior literature demonstrates how access to school choice and the choices parents make are conditioned by race and socioeconomic status (e.g., Phillips et al. 2015). Given the high correlation between race/ethnicity and socioeconomic status (as measured by FRL status) in Denver (see Table 1), and to conserve space, our focus is on race/ethnicity (results from rank-ordered logit models estimated separately by whether a child is FRL eligible are available upon request).

In a manifestation of the independence from irrelevant alternatives assumption of traditional multinomial logit models, rank-ordered logit models assume that the relative preference for any two alternatives is invariant to all other features of the choice set (e.g., which schools are in the current choice set, which other schools a parent has already chosen, the number of schools already chosen, or the order in which parents selected schools). It is not difficult to imagine scenarios where this assumption fails to hold (e.g., parents might select schools by different criteria depending on whether they are making their first or their fifth choice), but Allison and Christakis (1994) argue that this invariance assumption provides a reasonable approximation of complex realities. As a sensitivity test, we compared results from our rank-ordered logit models to results from conditional logit models (in which we model the likelihood that a school is selected as a parent’s top choice), and we found few substantive differences. This suggests that parents’ preferences across their rankings generally adhere to a similar underlying random utility model (Chapman and Staelin 1982; Hausman and Ruud 1987).

Findings

Whereas Table 1 shows relative parity in parents’ participation in school choice, Table 2 reveals stark differences in both the substance of parents’ choices and the availability of alternatives by parents’ race/ethnicity. The schools white parents selected as their top choices had substantially higher standardized test scores and SPF ratings than did the schools chosen by Hispanic and black parents. Column 4 of Table 2 suggests a possible reason: within a two-mile radius of white parents’ homes (which represents the median distance between parents’ residence and their selections on the choice application), the average school’s standardized math test score is –.26, compared with average scores of –.50 and –.44 near the homes of Hispanic and black families, respectively. Although white parents selected higher-performing schools than did Hispanic and black parents, Table 2 also indicates that, overall, parents chose better-performing schools than the average school within a two-mile radius of their homes (particularly for their top choice). This suggests that across racial/ethnic groups, parents are using school choice in an attempt to access better schools than those to which their children would otherwise be assigned.

Table 2.

Characteristics of Parents’ Choice Sets.

Category	Top Choice	All Choices	All Schools within Two Miles	All Schools in DPS
Standardized math test score				−0.47(0.55)
All applicants	−0.15(0.46)	−0.19(0.35)	−0.44(0.34)
Hispanics	−0.24(0.42)	−0.28(0.30)	−0.50(0.28)
Whites	0.10(0.44)	0.05(0.37)	−0.26(0.42)
Blacks	−0.13(0.48)	−0.16(0.33)	−0.44(0.36)
Average SPF rating^a				3.20(1.26)
All applicants	3.74(0.94)	3.68(0.71)	3.24(0.65)
Hispanics	3.63(1.00)	3.57(0.75)	3.15(0.62)
Whites	3.96(0.75)	3.92(0.60)	3.42(0.71)
Blacks	3.89(0.84)	3.82(0.56)	3.36(0.62)
Proportion Hispanic or black				0.78(0.21)
All applicants	0.72(0.23)	0.73(0.19)	0.78(0.17)
Hispanics	0.81(0.18)	0.81(0.15)	0.83(0.13)
Whites	0.50(0.21)	0.53(0.17)	0.62(0.19)
Blacks	0.67(0.21)	0.68(0.16)	0.77(0.17)
Proportion charter schools				0.29(0.45)
All applicants	0.31(0.46)	0.29(0.32)	0.24(0.17)
Hispanics	0.36(0.48)	0.34(0.34)	0.26(0.17)
Whites	0.15(0.36)	0.16(0.25)	0.20(0.17)
Blacks	0.30(0.46)	0.30(0.29)	0.21(0.16)
Proportion magnet or specialized schools				0.10(0.30)
All applicants	0.04(0.20)	0.04(0.12)	0.07(0.11)
Hispanics	0.04(0.18)	0.04(0.11)	0.07(0.11)
Whites	0.06(0.24)	0.06(0.15)	0.09(0.13)
Blacks	0.04(0.19)	0.04(0.11)	0.06(0.09)
Average distance (in miles)^b				7.01(4.32)
All applicants	2.41(2.27)	2.66(1.92)
Hispanics	2.17(2.05)	2.47(1.77)
Whites	2.44(2.01)	2.53(1.66)
Blacks	3.39(3.06)	3.70(2.46)

Note: DPS = Denver Public Schools; SPF = School Performance Framework. Data consist of the population of rising sixth and ninth graders who attended a school in DPS in either 2011–2012 or 2012–2013, whose parents submitted a public school application to DPS in spring of 2012 or 2013, and who are classified into one of DPS’s three main racial/ethnic groups (all applicants = 14,089, Hispanic = 8,998, white = 2,972, black = 2,119). There are 217 records for schools serving sixth or ninth grade across the two years. Standard deviations are in parentheses. The second column describes parents’ first-ranked schools; the third column describes all the schools ranked by parents (parents could rank up to five schools on their applications). The fourth column shows the characteristics for all schools (ranked and unranked) within a two-mile radius of a family; this is around the 50th percentile of distances between all students in the sample and the schools to which their parents applied. The fifth column describes all the schools in Denver to which the sample of parents could have applied.

Gives the average School Performance Framework rating of schools, obtained by assigning each rating a numeric value (5 = distinguished, 4 = meets expectations, 3 = on watch, 2 = on priority watch, 1 = on probation); unrated schools are excluded from the calculation.

Gives the average straight-line distance in miles between a student’s residence and his or her top choice (Column 2), all choices (Column 3), and all the schools in the district to which his or her parents could have applied (Column 5). We do not calculate the average distance for Column 4 given the set distance parameter.

Owing at least partially to the patterns of residential segregation evident in Figure 1, the schools selected by, and located close to the homes of, white parents tend to enroll lower proportions of Hispanic and black students than do the schools selected by and located near Hispanic and black families. Charter schools also tend to be located closer to Hispanic families, constituting about one quarter of their nearby school options, compared to about one fifth for white and black families. This fits with research finding that charter schools in other major cities are more likely to be located in areas with poorer-performing existing public schools and higher proportions of nonwhite residents (Burdick-Will, Keels, and Schuble 2013; Henig and MacDonald 2002). White families, however, have a higher share of magnet and other specialized programs within a two-mile radius of their homes. Compared to white parents, a higher proportion of Hispanic and black parents chose charter schools. A higher proportion of white parents, however, chose schools with a specialized curricular focus. Black students were willing to (or had to) travel the furthest to get to their parents’ preferred school. On average, Hispanic parents’ top choice was located more than one mile closer to home than was the top choice of black parents.

Because parents consider multiple dimensions of schools when making choices (Kimelberg 2014), we estimate rank-ordered logit models to assess how parents combine schools’ attributes into overall assessments of their options. Table 3 presents results from these models, estimated separately for each racial/ethnic group to examine the potential heterogeneity in the valuation of school attributes.

Table 3.

What Parents Want from Schools: Estimates from Rank-ordered Logit Models.

	Hispanic Families	White Families	Black Families
School average standardized math score	2.14	1.62	1.56
School average standardized math score	(0.03)	(0.07)	(0.06)
Distance (in miles)	−0.52	−0.54	−0.32
Distance (in miles)	(0.01)	(0.01)	(0.01)
School Score × Distance	−0.03	−0.01	−0.03
School Score × Distance	(0.01)	(0.01)	(0.01)
Proportion Hispanic or black	8.42	14.24	14.36
Proportion Hispanic or black	(0.19)	(0.34)	(0.40)
Proportion black or Hispanic squared	−4.17	−13.15	−12.06
Proportion black or Hispanic squared	(0.15)	(0.32)	(0.36)
Last year’s school (1 = yes)	0.28	1.18	1.04
Last year’s school (1 = yes)	(0.07)	(0.09)	(0.11)
Charter school (1 = yes)	−0.36	−0.69	−0.19
Charter school (1 = yes)	(0.02)	(0.05)	(0.04)
Magnet or specialized school (1 = yes)	−1.09	−1.37	−1.12
Magnet or specialized school (1 = yes)	(0.03)	(0.05)	(0.07)
Crime rate in school’s block-group (per 1,000)	−0.002	−0.001	−0.003
Crime rate in school’s block-group (per 1,000)	(0.0001)	(0.0002)	(0.0002)
School Score × Student Score	0.28	0.39	0.29
School Score × Student Score	(0.02)	(0.03)	(0.03)
N observations	628,239	212,680	146,626
N students	8,998	2,972	2,119

Note: Results, expressed in log-odds units, come from rank-ordered logit models estimated separately for each of the three main racial/ethnic groups in Denver. The analytic sample is the population of rising sixth and ninth graders who attended a school in Denver Public Schools (DPS) in either 2011–2012 or 2012–2013 and whose parents submitted a public school application to DPS in spring 2012 or spring 2013. Although we show the standard errors (in parentheses) to provide a sense of the variation around the coefficients, and to aid in comparing coefficients across racial/ethnic groups, we do not present significance levels of the coefficients themselves.

Overall, we see that school performance (as measured by a school’s average standardized math test score) matters to all families. In fact, performance appears to matter relatively more for Hispanic families. Furthermore, parents are more likely to include a higher-scoring school if their own child scores relatively high on standardized tests. Similarly, for all groups, proximity is important. The farther away a school is, the less likely a parent will be to select that school. Parents also indicate a strong preference for children’s current schools. This makes intuitive sense; why move students if the school appears to be serving them well, at least according to the parents’ perspective? All parents also react to the relative safety of the area in which schools are located, as measured by the crime rate per 1,000 residents. Additionally, across the board, parents are less likely to apply to a charter, magnet, or other specialized school versus a district-run school, possibly because these school types constitute the minority in DPS. But we do see differences by parents’ race/ethnicity. Holding all other school attributes constant at their means, the predicted probability that a black parent selects a charter school as one of his or her choices is .42, compared to .14 among Hispanic parents and .05 among white parents.

White parents’ preferences are sensitive to the presence of Hispanic and black families in a school, even though we also control for attributes such as school performance and safety that often serve as proxies for racial composition (Billingham and Hunt 2016; Goyette et al. 2012; Saporito 2003). To better understand this sensitivity, we calculated the predicted probability that a parent applies to a school using the model results from Table 3, varying the proportion of a school’s student body that is Hispanic or black, and holding all other school attributes at their means. The predicted probability that a white parent applies to a school rises to a peak of around .26 when the school is 50 percent Hispanic or black, and then drops sharply to .13 at 80 percent and .06 at 90 percent Hispanic or black (both scenarios are not uncommon in the district). Conversely, Hispanic parents have a predicted probability of just .04 of applying to a school in which 25 percent of its students are either Hispanic or black, and a relatively consistent 20 percent or higher chance of applying to schools where the percentage of students who are Hispanic or black is at least 80 percent. These differences are likely related to the residential segregation evident in Figure 1 and what prior research has termed “out-group avoidance” on the part of white parents (Bifulco et al. 2009b; Saporito 2003).

Given descriptive evidence from Table 2, that white, Hispanic, and black families do not have the same level of school performance in their nearby choice sets as a function of where in the city they reside, we explore further the intersection between proximity and school performance. Figure 2 shows the predicted probabilities that parents include a school on their choice application based on their own racial/ethnic group, the school’s average standardized math test score, and the distance between the school and the home. Panel a illustrates the predicted probabilities of parents choosing a relatively high-scoring school with an average standardized math test score of .25 (which is between the 50th and 75th percentiles of all schools in the district serving sixth and ninth grades). Unsurprisingly, families of all three racial/ethnic groups would like to send their children to high-performing schools relatively close to their homes. Across the board, parents are nearly 100 percent likely to apply to a higher-scoring school within 1 mile of their home. Black families, however, appear more likely than Hispanic and white families to travel a bit farther to access higher-scoring schools. Even at 10 miles away from home, a school with a relatively high average standardized math test score has a roughly 50/50 chance of inclusion on black parents’ choice applications.

Figure 2.

Predicted probabilities that parents apply to a school by student race/ethnicity and the school’s performance and proximity

The trade-off, however, comes in the supply of high-performing schools close to where people live, and, as we saw earlier, this varies systematically by race. Panel b of Figure 2 shows the predicted probabilities of parents choosing a relatively low-scoring school with an average standardized math test score of –.85 (which is around the percentile of all schools in the district serving sixth and ninth grades). Hispanic and black families are markedly more likely to sacrifice performance for proximity. This is due, in large part, to how much farther away they live from highly rated schools and to the very real constraints on how to get their children to and from school. Lower-performing schools very near to families’ homes have a predicted probability of appearing on Hispanic parents’ applications of about .80, and .85 among black parents, compared to around .60 for white parents. This probability gap holds fairly constant between Hispanic and white families as the distance increases to 5 miles. In other words, Hispanic parents are more likely than white parents to list a lower-performing school on their applications, even as the distance from their home to the school increases. The gap in the predicted probability of applying to a poorly performing school between black and white parents, and between black and Hispanic parents, widens through the 5-mile mark, before closing again slightly as the distance parents are willing to travel increases to 10 miles.

Despite the intended use of school choice policies to break the link between students’ residence and school, students remain largely separated from one another along racial lines, and nonwhite students remain trapped in lower-performing schools. These patterns are evident in Table 4, which shows simple simulations to illustrate how the distribution of sixth- and ninth-grade students in schools would change if all students whose parents submitted a school choice application were assigned to their top-choice school. The majority—about 79 percent—of students whose parents submit an application are assigned to their top choice, yet the one-fifth who are not leaves open the possibility of considering a best-case scenario in which all schools had the capacity to accommodate every student who sought a seat. That is, how would enrollment by race/ethnicity be distributed as a result of parents’ choices if school capacity were not an issue? Columns 2 through 5 of Table 4 compare the actual enrollment of students in 2012–2013 in low- and high-rated schools to enrollment across schools had all parents received their top choice on their spring 2012 application. We see virtually no differences. Much higher proportions of Hispanic and black students (29 and 23 percent, respectively) than white students (6 percent) are enrolled in low-rated schools regardless of whether we look at actual enrollments or a situation where every school a parent listed as his or her top choice had the capacity to accommodate that student. Additionally, we find no differences in terms of the level of segregation between white and Hispanic students, and between white and black students, when we look at actual versus counterfactual enrollment (as measured by a dissimilarity index in columns 6 and 7 in Table 4). This suggests that Hispanic and black families— those concentrated in areas of the city disproportionately served by low-rated schools—are choosing from a different set of schools than are white families. We find similar results when we examine concentration in lower- and higher-rated schools and dissimilarity by socioeconomic status (using eligibility for FRL; available upon request).

Table 4.

Distribution of Students in Schools, Actual and Counterfactual, 2012–2013.

	Enrolled in Low- rated School (%)^a		Enrolled in High- rated School (%)^b		Dissimilarity Index (vs. Whites)^c
Racial/Ethnic Group	Actual	Counterfactual	Actual	Counterfactual	Actual	Counterfactual
Hispanic	28.85	28.50	7.39	7.25	0.63	0.63
White	5.78	5.73	19.44	19.39
Black	22.77	22.48	6.65	3.36	0.57	0.57

Note: Table 4 compares the actual distribution of students across schools in the 2012–2013 school year to a counterfactual distribution had every student whose parent submitted a school choice application been assigned to their top choice school. Data come from the population of students enrolled in sixth and ninth grades in Denver Public Schools in 2012–2013. Counterfactuals are based on the applications submitted by parents in the spring of 2012.

The percentage of students enrolled in low-rated schools is based on the percentage of students enrolled in the two lowest School Performance Framework (SPF) categories: ‘‘on probation’’ and ‘‘on priority watch.’’

The percentage of students enrolled in high-rated schools is based on the percentage of students enrolled in the highest SPF category: ‘‘distinguished.’’

Dissimilarity indices between Hispanic and white students, and between black and white students, are calculated as $D = 0.5 \sum_{j = 1}^{N} | \frac{c_{j}}{C} - \frac{w_{j}}{W} |$ , where c_j is the number of Hispanic (or black) students in school j, C is the total number of Hispanic (or black) students in the district, w_j is the number of white students in school j, and W is the total number of white students in the district. D is the percentage of Hispanic (or black) students who would need to enroll in another school so that white and Hispanic (or black) students would be evenly distributed across schools.

Conclusions

School choice policies are intended to liberate families from the constraints of residentially based enrollment systems by extending to all families, regardless of socioeconomic background or where in the city they live, the opportunity to enroll their children in public schools of their choosing—a choice that more advantaged families have long had by virtue of their ability to move to neighborhoods with desirable schools (Archbald 2004; Bifulco et al. 2009b). Much prior research illustrates the logistical barriers to exercising school choice (Holme 2002; Lareau 2014; Mickelson and Southworth 2005), as well as white families’ tendency to use choice policies to avoid schools with high proportions of nonwhite families (Billingham and Hunt 2016; Renzulli and Evans 2005; Saporito and Sohini 2006). However, given the persistent realities of racial/ethnic residential segregation, coupled with the uneven distribution of relatively high-performing schools that tends to correspond with patterns of segregation (see Figure 1), it is important to also consider how the available supply of neighborhood schools shapes and constrains the choices families make.

In this article, we drew on data generated by the population of parents who submitted applications to the DPS central office during the 2011–2012 and 2012–2013 school years to enroll their students in a public middle or high school other than the one assigned to them based on their neighborhood. To the extent that high-scoring options are available, parents choose schools with high academic performance. Some research suggests that minority and lower-income families place less emphasis on academics when choosing schools (Harris and Larsen 2015; Hastings et al. 2009), but our analyses indicate this is not the case, at least in Denver with its particularly open choice system. To the extent that Hispanic and black families do sacrifice academic performance in our study, it appears to be the result of a relative dearth of higher-scoring schools nearby. Nonwhite families simply do not have the same set of school options, on average, that white families do.

Some scholars suggest that the expansion of school choice in urban districts will have little impact on overall school inequality because white and higher-income parents will choose schools in their relatively privileged neighborhoods (Goyette 2008; Schwartz and Stiefel 2014), and this is borne out by our results. Hispanic and black parents do appear to be selecting better schools than the overall stock around them, but as evident by comparing columns 2 and 3 to column 4 in Table 2, this is in part because they are willing to travel relatively farther to access them. White parents’ selections, particularly their top-choice schools, tend to be closer to their homes than the choices of Hispanic and black families. But even if all parents were to receive their top-choice school, regardless of schools’ capacities, the constraints placed on Hispanic and black families’ choices—due to residential segregation and the uneven distribution of higher-rated schools around the city—mean that nonwhite students remain largely trapped in lower-performing schools.

This study yields a useful contribution to the school choice literature, but we do not suggest that the findings will generalize to all urban districts in the United States. Denver has a particularly open and robust brand of school choice. This is due, in part, to its location in Colorado, which has been at the fore of the choice movement since passing intra- and interdistrict transfer legislation in the early 1990s. Colorado was also the third state in the country to enact charter school legislation. The district itself has implemented policies aimed at increasing options and access to choice. Families in DPS can apply to multiple types of public schools (including district-run, charter, and magnet schools), and they can do so via a single application form and process.

At the same time, features of the Denver context do hold relevance for other districts. For example, as in Denver, Hispanic students now constitute a majority in many urban districts nationwide (Billingham and Hunt 2016). Relatedly, segregation of Hispanics from whites, although lower than black-white segregation, is on the rise (Jargowsky 2014). Research on how families experience school choice needs to continue to move beyond the more typical black-white comparisons that characterize much of the extant literature (Haynes, Phillips, and Goldring 2010). Additionally, Denver could be considered part of a population of relatively large, urban districts that are taking stock of their choice policies to better manage issues of access and supply. Like Chicago in research by Lauen (2007) and Rich and Jennings (2015), and Durham, North Carolina, in work by Bifulco and colleagues (2009b), Denver has a largely disadvantaged and segregated student population, as well as an evolving and expanding set of schools from which families can choose.

It is also important to keep in mind that we cannot observe the choices made by parents of the roughly 8 percent of Denver’s students enrolled in public schools outside the city and district. This choice mechanism is not randomly distributed among the public school parent population. Prior work finds that white parents are more likely than Hispanic parents, but less likely than black parents, to enroll their children in a district in which they do not reside (Holme and Richards 2009; Lavery and Carlson 2015). Furthermore, wealthy parents are more likely than their less-affluent counterparts to opt into other districts, reflecting the demands on time, information, and other resources necessary to take advantage of this choice mechanism (Holme and Richards 2009). At the same time, the factors weighed by parents looking outside their home districts generally mirror those found in this study of parents choosing a school within DPS. Interdistrict transfers are driven primarily by a desire for higher-achieving schools and by structural characteristics such as distance (e.g., an out-of-district school may be closer to a family than a school in the district in which they reside) (Carlson, Lavery, and Witte 2011).

Even under an open-enrollment choice system that includes all public school options and removes many of the logistical barriers to exercising choice found in other settings—that is, a best-case scenario for realizing the ideals of liberating minority and disadvantaged students from underperforming schools—parents’ school choices are still largely determined by where they live. Saporito (2003) contends that laissez-faire school choice policies, which allow the unrestricted movement of students across schools, may further deteriorate the educational conditions for those left behind in their residentially assigned public schools. We did not find evidence of a worsening situation, but our results do suggest that the availability of school choice on its own is not a sufficient condition for equity. Simply allowing parents to choose their children’s schools does not directly address the fact that high-scoring schools are unevenly distributed across cities in the United States, so many families remain unable to access good schools. In this way, Saporito’s point is well taken: Choice policies on their own will not eliminate enrollment gaps without further improvements in other areas of the public school system. In addition to being able to choose schools, parents need viable alternatives from which to choose.

Footnotes

Acknowledgements

Previous versions of this article were presented at the American Sociological Association meetings in Chicago (August 2015) and at the Center for Studies in Demography and Ecology seminar series at the University of Washington (October 2014), and the authors thank participants for their helpful questions and suggestions. The authors also thank staff at Denver Public Schools, especially Brian Eschbacher and Yu-lu Hsiung, for providing access to the data. Finally, the authors are grateful to the editor and anonymous reviewers for their useful comments. All views and errors are the authors’ own.

Research Ethics

The research in this study, which relied on analysis of secondary data provided to the authors by a data-sharing agreement with Denver Public Schools, has been approved by the Institutional Review Board at the University of Washington and was performed according to the ethical standards laid out in their approval.

Notes

Author Biographies

Patrick Denice is a postdoctoral fellow in sociology at Washington University in St. Louis as well as an affiliated researcher at the Center on Reinventing Public Education (CRPE) at the University of Washington Bothell. His research examines stratification in education and the labor market. He earned his PhD in sociology from the University of Washington, where his dissertation focused on nontraditional pathways through postsecondary education.

Betheny Gross is a senior research analyst and research director at the Center on Reinventing Public Education (CRPE) at the University of Washington Bothell. Her research examines urban school district reform and alternative school governance models including charter schools and policies to bridge charter and district schools. She co-authored the book Strife and Progress documenting school reforms in several urban districts. She earned her PhD in educational policy studies from the University of Wisconsin-Madison.

References

Allison

Paul D.

Christakis

Nicholas A.

1994. “Logit Models for Sets of Ranked Items.”Sociological Methodology 24:199–228.

Archbald

Douglas A.

2004. “School Choice, Magnet Schools, and the Liberation Model: An Empirical Study.”Sociology of Education 77(4):283–310.

Bell

Courtney A.

2009. “All Choices Created Equal? The Role of Choice Sets in the Selection of Schools.”Peabody Journal of Education 84(2):191–208.

Berends

Mark

Zottola

Genevieve C.

2009. “Social Perspectives on School Choice.” Pp. 35–53 in Handbook of Research on School Choice, edited by Berends

Springer

M. G.

Ballou

Walberg

H. J.

New York, NY: Routledge.

Betts

Julian R.

2005. “The Economic Theory of School Choice.” Pp. 14–39 in Getting Choice Right: Ensuring Equity and Efficiency in Education Policy, edited by Betts

J. R.

Loveless

Washington, DC: Brookings Institution Press.

Bifulco

Robert

Ladd

Helen F.

Ross

Stephen L.

2009a. “The Effects of Public School Choice on Those Left Behind: Evidence from Durham, North Carolina.”Peabody Journal of Education 84(2):130–49.

Bifulco

Robert

Ladd

Helen F.

Ross

Stephen L.

2009b. “Public School Choice and Integration Evidence from Durham, North Carolina.”Social Science Research 38(1):71–85.

Billingham

Chase M.

Hunt

Matthew O.

2016. “School Racial Composition and Parental Choice: New Evidence on the Preferences of White Parents in the United States.”Sociology of Education 89(2):99–117.

Bischoff

Kendra

. 2008. “School District Fragmentation and Racial Residential Segregation: How Do Boundaries Matter?”Urban Affairs Review 44(2):182–217.

10.

Buckley

Jack

Schneider

Mark

. 2003. “Shopping for Schools: How Do Marginal Consumers Gather Information about Schools?”Policy Studies Journal 31(3):121–45.

11.

Burdick-Will

Julia

Keels

Micere

Schuble

Todd

. 2013. “Closing and Opening Schools: The Association between Neighborhood Characteristics and the Location of New Educational Opportunities in a Large Urban District.”Journal of Urban Affairs 35(1):59–80.

12.

Burgess

Simon

Greaves

Ellen

Vignoles

Anna

Wilson

Deborah

. 2015. “What Parents Want: School Preferences and School Choice.”Economic Journal 125(587):1262–89.

13.

Carlson

Deven

Lavery

Lesley

Witte

John F.

2011. “The Determinants of Interdistrict Open Enrollment Flows: Evidence from Two States.”Educational Evaluation and Policy Analysis 33(1):76–94.

14.

Chapman

Randall G.

Staelin

Richard

. 1982. “Exploiting Rank Ordered Choice Set Data within the Stochastic Utility Model.”Journal of Marketing Research 14:288–301.

15.

Clotfelter

Charles T.

1999. “Public School Segregation.”Land Economics 75(4):487–504.

16.

Coons

John E.

Sugarman

Stephen D.

1978. Education by Choice: The Case for Family Control. Berkeley: University of California Press.

17.

Fiel

Jeremy E.

2013. “Decomposing School Resegregation: Social Closure, Racial Imbalance, and Racial Isolation.”American Sociological Review 78(5):828–48.

18.

Goyette

Kimberly A.

2008. “Race, Social Background, and School Choice Options.”Equity and Excellence in Education 41(1):114–29.

19.

Goyette

Kimberly A.

2014. “Setting the Context.” Pp. 1–24 in Choosing Homes, Choosing Schools, edited by Lareau

Goyette

K. A.

New York: Russell Sage Foundation.

20.

Goyette

Kimberly A.

Farrie

Danielle

Freely

Joshua

. 2012. “This School’s Gone Downhill: Racial Change and Perceived School Quality among Whites.”Social Problems 59(2):155–76.

21.

Gross

Betheny

DeArmond

Michael

Denice

Patrick

. 2015. Common Enrollment, Parents, and School Choice: Early Evidence from Denver and New Orleans. Seattle: Center on Reinventing Public Education, University of Washington.

22.

Halaby

Charles N.

2003. “Where Job Values Come From: Family and Schooling Background, Cognitive Ability, and Gender.”American Sociological Review 68(2):251–78.

23.

Harris

Douglas N.

Larsen

Matthew

. 2015. What Schools Do Families Want (and Why)? School Demand and Information Before and After the New Orleans Post-Katrina School Reforms. New Orleans, LA: Education Research Alliance for New Orleans.

24.

Hastings

Justine S.

Kane

Thomas J.

Staiger

Douglas O.

2006. “Parental Preferences and School Competition: Evidence from a Public School Choice Program.” NBER Working Paper No. 11805, National Bureau of Economic Research, Cambridge, MA.

25.

Hastings

Justine S.

Kane

Thomas J.

Staiger

Douglas O.

2009. “Heterogeneous Preferences and the Efficacy of Public School Choice.” NBER Working Paper No. 11805, National Bureau of Economic Research, Cambridge, MA.

26.

Hausman

Jerry

Ruud

Paul

. 1987. “Specifying and Testing Econometric Models for Rank-ordered Data.”Journal of Econometrics 34:83–104.

27.

Haynes

Katherine Taylor

Phillips

Kristie J. R.

Goldring

Ellen B.

2010. “Latino Parents’ Choice of Magnet School: How School Choice Differs across Racial and Ethnic Boundaries.”Education and Urban Society 42(6):758–89.

28.

Henig

Jeffrey R.

1996. “The Local Dynamics of Choice: Ethnic Preferences and Institutional Responses.” Pp. 95–117 in Who Chooses? Who Loses? Culture, Institutions, and the Unequal Effects of School Choice, edited by Fuller

Elmore

R. F.

New York: Teachers College Press

29.

Henig

Jeffrey R.

MacDonald

Jason A.

2002. “Locational Decisions of Charter Schools: Probing the Market Metaphor.”Social Science Quarterly 83(4):962–80.

30.

Holme

Jennifer Jellison

. 2002. “Buying Homes, Buying Schools: School Choice and the Social Construction of School Quality.”Harvard Educational Review 72(2):177–205.

31.

Holme

Jennifer Jellison

Richards

Meredith P.

2009. “School Choice and Stratification in a Regional Context: Examining the Role of Inter-district Choice.”Peabody Journal of Education 84(2):150–71.

32.

Hoxby

Caroline M.

1998. “What Do America’s Traditional Forms of School Choice Teach Us about School Choice Reforms?”Economic Policy Review 4(1):47–59.

33.

Jacob

Brian A.

Lefgren

Lars

. 2007. “What Do Parents Value in Education? An Empirical Investigation of Parents’ Revealed Preferences for Teachers.”Quarterly Journal of Economics 122(4):1603–37.

34.

Jargowsky

Paul A.

2014. “Segregation, Neighborhoods, and Schools.” Pp. 97–136 in Choosing Homes, Choosing Schools, edited by Lareau

Goyette

K. A.

New York: Russell Sage Foundation.

35.

Jennings

Jennifer L.

2010. “School Choice or Schools’ Choice? Managing in an Era of Accountability.”Sociology of Education 83(3):227–47.

36.

Kalleberg

Arne L.

Marsden

Peter V.

2013. “Changing Work Values in the United States, 1973–2006.”Social Science Research 42(2):255–70.

37.

Kimelberg

Shelley McDonough

. 2014. “Middle-class Parents, Risk, and Urban Public Schools.” Pp. 207–236 in Choosing Homes, Choosing Schools, edited by Lareau

Goyette

K. A.

New York: Russell Sage Foundation.

38.

Kimelberg

Shelley McDonough

Billingham

Chase M.

2013. “Attitudes toward Diversity and the School Choice Process: Middle-class Parents in a Segregated Urban Public School District.”Urban Education 48:198–231.

39.

Kleitz

Bretten

Weiher

Gregory R.

Tedin

Kent

Matland

Richard

. 2000. “Choice, Charter Schools, and Household Preferences.”Social Science Quarterly 81(3):846–54.

40.

Kozol

Jonathan

. 2005. The Shame of the Nation: The Restoration of Apartheid Schooling in America. New York: Crown.

41.

Lareau

Annette

. 2014. “Schools, Housing, and the Reproduction of Inequality.” Pp. 169–206 in Choosing Homes, Choosing Schools, edited by Lareau

Goyette

K. A.

New York: Russell Sage Foundation.

42.

Lareau

Annette

Goyette

Kimberly A.

2014. Choosing Homes, Choosing Schools. New York: Russell Sage Foundation.

43.

Lauen

Douglas Lee

. 2007. “Contextual Explanations of School Choice.”Sociology of Education 80(3):179–209.

44.

Lavery

Lesley

Carlson

Deven

. 2015. “Dynamic Participation in Interdistrict Open Enrollment.”Educational Policy 29(5):746–79.

45.

Logan

John R.

2013. “The Persistence of Segregation in the 21st Century Metropolis.”City and Community 12(2):160–68.

46.

Logan

John R.

Burdick-Will

Julia

. Forthcoming. “School Segregation, Charter Schools, and Access to Quality Education.”Journal of Urban Affairs.

47.

Logan

John R.

Minca

Elisabeta

Adar

Sinem

. 2012. “The Geography of Inequality: Why Separate Means Unequal in American Public Schools.”Sociology of Education 85(3):287–301.

48.

Lubienski

Christopher

Gulosino

Charisse

Weitzel

Peter

. 2009. “School Choice and Competitive Incentives: Mapping the Distribution of Educational Opportunities across Local Education Markets.”American Journal of Education 115(4):601–47.

49.

Massey

Douglas S.

Denton

Nancy A.

1993. American Apartheid: Segregation and the Making of the Underclass. Cambridge, MA: Harvard University Press.

50.

Mickelson

Roslyn Arlin

Southworth

Stephanie

. 2005. “When Opting out Is Not a Choice: Implications for NCLB’s Transfer Option from Charlotte, North Carolina.”Equity and Excellence in Education 38(3):249–63.

51.

Phillips

Kristie J. R.

Hausman

Charles

Larsen

Elisabeth S.

2012. “Students Who Choose and the Schools They Leave: Examining Participation in Intradistrict Transfers.”Sociological Quarterly 53:264–94.

52.

Phillips

Kristie J. R.

Larsen

Elisabeth S.

Hausman

Charles

. 2015. “School Choice and Social Stratification: How Intra-district Transfers Shift the Racial/Ethnic and Economic Composition of Schools.”Social Science Research 51:30–50.

53.

Reardon

Sean F.

Bischoff

Kendra

. 2011. “Income Inequality and Income Segregation.”American Journal of Sociology 116(4):1092–1153.

54.

Reardon

Sean F.

Matthews

Stephan A.

O’Sullivan

David

Lee

Barrett A.

Firebaugh

Glenn

Farrell

Chad R.

Bischoff

Kendra

. 2008. “The Geographic Scale of Metropolitan Racial Segregation.”Demography 45(3):489–514.

55.

Reback

Randall

. 2008. “Demand (and Supply) in an Inter-district Public School Choice Program.”Economics of Education Review 27(4):402–16.

56.

Renzulli

Linda A.

Evans

Lorraine

. 2005. “School Choice, Charter Schools, and White Flight.”Social Problems 52(3):398–418.

57.

Rhodes

Anna

DeLuca

Stefanie

. 2014. “Residential Mobility and School Choice among Poor Families.” Pp. 137–206 in Choosing Homes, Choosing Schools, edited by Lareau

Goyette

K. A.

New York: Russell Sage Foundation.

58.

Rich

Peter M.

Jennings

Jennifer L.

2015. “Choice, Information, and Constrained Options: School Transfers in a Stratified Educational System.”American Sociological Review 80(5):1069–98.

59.

Roda

Allison

Wells

Amy Stuart

. 2013. “School Choice Policies and Racial Segregation: Where White Parents’ Good Intentions, Anxiety, and Privilege Collide.”American Journal of Education 119(2):261–93.

60.

Saporito

Salvatore

. 2003. “Private Choices, Public Consequences: Magnet School Choice and Segregation by Race and Poverty.”Social Problems 50(2):181–203.

61.

Saporito

Salvatore

Lareau

Annette

. 1999. “School Selection as a Process: The Multiple Dimensions of Race in Framing Educational Choices.”Social Problems 46(3):418–39.

62.

Saporito

Salvatore

Sohini

Deenesh

. 2006. “Coloring outside the Lines: Racial Segregation in Public Schools and Their Attendance Boundaries.”Sociology of Education 79:81–105.

63.

Saporito

Salvatore

Sohini

Deenesh

. 2007. “Mapping Educational Inequality: Concentrations of Poverty among Poor and Minority Students in Public Schools.”Social Forces 85(3):1227–53.

64.

Schneider

Mark

Buckley

Jack

. 2002. “What Do Parents Want from Schools? Evidence from the Internet.”Educational Evaluation and Policy Analysis 24:133–44.

65.

Schneider

Mark

Marschall

Melissa

Teske

Paul

Roch

Christine

. 1998. “School Choice and Culture Wars in the Classroom: What Different Parents Seek from Education.”Social Science Quarterly 79(3):489–501.

66.

Schneider

Mark

Teske

Paul

Marschall

Melissa

. 2002. Choosing Schools: Consumer Choice and the Quality of American School. Princeton, NJ: Princeton University Press.

67.

Schwartz

Amy Ellen

Stiefel

Leanna

. 2014. “Linking Housing Policy and School Reform.” Pp. 295–314 in Choosing Homes, Choosing Schools, edited by Lareau

Goyette

K. A.

New York: Russell Sage Foundation.

68.

Sohini

Deenesh

Saporito

Salvatore

. 2009. “Mapping School Segregation: Using GIS to Explore Racial Segregation between Schools and Their Corresponding Attendance Areas.”American Journal of Education 115(4):569–600.

69.

Tedin

Kent

Weiher

Gregory R.

2004. “Racial/Ethnic Diversity and Academic Quality as Components of School Choice.”Journal of Politics 66(4):1109–33.

70.

Teske

Paul

Schneider

Mark

. 2001. “What Research Can Tell Policymakers about School Choice.”Journal of Policy Analysis and Management 20(4):609–31.

71.

Weiher

Gregory R.

Tedin

Kent L.

2002. “Does Choice Lead to Racially Distinctive Schools? Charter Schools and Household Preferences.”Journal of Policy Analysis and Management 21:79–92.

72.

Weininger

Elliot B.

2014. “School Choice in an Urban Setting.” Pp. 268–94 in Choosing Homes, Choosing Schools, edited by Lareau

Goyette

K. A.

New York: Russell Sage Foundation.

73.

Zeng

Zhen

Xie

. 2008. “A Preference-opportunity-choice Framework with Applications to Intergroup Friendship.”American Journal of Sociology 114(3):615–48.

74.

Zhou

Xueguang

. 2005. “The Institutional Logic of Occupational Prestige Ranking: Reconceptualization and Reanalyses.”American Journal of Sociology 111(1):90–140.