Diving into School Choice Data
With the implementation of school choice, parents are no longer confined to any one school, based on location. Parents are empowered to choose the best education for their children. However, strong academics are not the only thing that parents look for when choosing a school for their child.
We sat down with Steve Glazerman, a Senior Fellow at Mathematica Policy Research. Steve is a national expert on evaluating school choice systems and recently completed a study with his colleague, Dallas Dotter, using data from Washington, DC’s common application and lottery system. The study and an associated policy brief are available here.
What were you hoping to find with your recent study of DC school choice data?
We wanted to understand what parents value when they choose schools, and what kinds of outcomes might result from expanding school choice. We also wanted to try and predict the consequences of different policy changes. We think this research is helpful for individual schools to deciding where to locate, and charter authorizers or district leaders who are expanding schools, authorizing new schools, or making decisions about programmatic offerings. Application data from a unified enrollment system provides a rich dataset that can be used by policymakers trying to make these types of informed decisions.
What did you find in the data?
Analyzing data from DC’s school lottery in 2014, in which parents rank-ordered their school choices, we were able to quantify the rate at which parents trade-off distance with different indicators of academic quality and other attributes of the school when making school selections. Not surprisingly, parents generally prefer schools close to where they live. However, we found that parents also prioritize the academic performance of a school, and the characteristics of its student body, including the proportion of low-income students and the racial composition of the school. For example, the data suggests that parents of middle school students would be willing to send their child to a school 1.2 miles farther away for a ten-point increase in proficiency. And for a jump from 10% to 20% in the proportion of students with the same race or ethnicity as their child, parents would be willing to go 2.1 miles farther.
Interestingly, we also found that different indicators of academic quality registered in varying ways for different groups of choosers. For instance, low-income parents were more responsive to proficiency rates at a school, whereas higher income parents were more responsive to a school’s overall accountability rating. That might have something to do with how school information was made available – it’s all available online, but if one set of information takes more clicks or a deeper understanding of the data to interpret, that could have a differentiated impact on different groups.
Were you faced with any limitations during your research?
An important limitation to our research is that we were only able to quantify the relative importance of information that parents had access to electronically through the My School DC website. We also analyzed public data on demographics and crime. But we didn’t have any information on informal channels that parents use, such as what they talk to their friends about at church or at the supermarket, or what they learn when they tour a school. All of those sources of information could be just as influential, or even more so.
You can read more about the study and our findings here.
What was the most surprising thing you found?
We found some puzzling results in terms of the impact of neighborhood characteristics on preferences for schools. For example, the data indicated that parents preferred schools located in areas that had higher reported violent crimes, which does not make intuitive sense. However, we used objective crime data, from the Uniform Crime Reports, whereas parents are likely making decisions based on their perception of crime. So there’s more to explore there.
What do you plan to study next?
We are aiming to replicate this research in other contexts and create tools for policymakers that use the model we built to predict the effects of all kinds of policy changes on the way students sort themselves across schools.
In addition, we’re currently running a really interesting experiment, sponsored by the U.S. Department of Education, to help policymakers learn how to select and present school choice data most effectively. We’re studying how different ways of displaying and sorting information about schools influences parents’ choice behavior and enhances their understanding of schools. Using an online survey, we present school profiles that we constructed using hypothetical data to emphasize different attributes of a school, and ask parents to rank these schools in their preferred order. We randomly assign parents to see one of 72 different versions of the school profiles, so we can measure the impact of different design decisions. The survey respondents are low-income parents of school-aged children across the United States. Based on the results of the experiment, we will develop a guide for school districts and policymakers who are designing online tools to share information about schools with parents.
- Dive into your unified enrollment data. Exploring how parents selected schools and what choices parents made can help cities and schools better understand the needs of their families.
- Distance matters. Parents prefer schools that are closer to their home.
- Other factors also play a role in parents’ school selection including academic performance of a school, and the characteristics of its student body, such as the proportion of low-income students and the racial composition of the school.
- Think about your community. Although groups of parents value different school attributes, some of these differences may be based on access to information. Make sure access to learning about schools is easy and equitable.