Over the River and through the Woods: The Role of Distance in Participation in Rural School Choice

ABSTRACT

Although there exists a large body of literature concerning the impacts of school choice policies, few studies focus on the choices of rural students. Using a unique dataset that includes administrative records, geocoded addresses, and commute times for Michigan public school students over 6 years, I describe who participates in interdistrict choice in rural Michigan, where 15% of rural students attend a nonresident district. In particular, I examine the roles of commute time and school closures in choice decisions – two factors that may be particularly salient in rural communities. I find that gaps in participation in interdistrict choice between rural and non-rural districts exist in kindergarten and persist across grades. Furthermore, I provide evidence that students who live farther away from their assigned school have lower opportunity costs to participate in interdistrict choice especially in rural districts. Also, school closures may induce students living farther away from their assigned school to attend a nonresident district in remote rural areas.

KEYWORDS:

• School choice
• rural education
• student commutes

Acknowledgments

I would like to thank Josh Cowen for his helpful comments. This research did not receive specific grant funding.

Disclosure statement

To my knowledge, I have no conflicts of interest.

Notes

1. In Michigan, students are considered economically disadvantaged if they qualify for free and reduced lunch, receive food (SNAP) or cash assistance (TANF), or they are homeless, migrant, or in foster care.

2. CEPI reports a resident district for each student and has an indicator for attending a nonresident district. However, 0.7% of observations have conflicting reported resident districts and nonresident indicators. Furthermore, 2.6% of observations had no address within their reported district. Because of the conflicts and my use of student addresses to construct commute time to nearest school, my variable of interest, I use students’ addresses and district boundaries to define resident districts and non-residency status. However, my definition may not reflect the number of students who are considered nonresidents, regardless of actual address, for funding purposes. My nonresident definition matches the reported nonresident definition for 98% of observations in my main sample. As a specification check, I estimate my main models using the reported nonresident definition. Results are similar and are available by request.

3. Additionally, I exclude 5,851 observations of students living in districts that close during the panel, 5,154 observations of students living in all charter school districts but attend a TPS, 2,468 observations of students who are reported in a grade that is higher than the reported terminal grade of the school, 3,210 observations that I was unable to calculate a driving distance for and 606 observations of students who lived in districts that did not have at least 10 students take a state exam that school year. In my analytic sample of kindergartners, I also drop 40 students who were reported in any grade the year prior. About 5% of student-year observations have multiple residences. To deduplicate addresses, I first drop observations with residences in a district that do not match the reported resident district from CEPI when the student has an address that does match the reported resident district. Second, I drop observations of residences in another district when the student has a residence in their attended district. Next, I drop excess observations that have commute times that are over an hour to their attended school when the student has one that is closer. Then, I choose a residence at random since I have no information concerning when during the school year the student lived there.

4. NCES determines district locales by enrollment weighting the school locale assignments of the schools operated by the district and uses the majority locale. NCES determines school locales using the location of its physical address in U.S. Census defined urbanized and rural areas. By U.S. Census definitions, rural areas are any territories that are not considered an Urbanized Area or Cluster. An Urbanized Area has more than 50,000 people. Urbanized Areas are comprised of cities and suburbs. An Urban Cluster, which NCES classifies as a town, has between 2,500 and 50,000 people. Therefore, a rural district may have a population of more than 2,500 people depending on its land size (Geverdt, 2019). I note that this definition may not be generalizable to large county wide districts commonly found in the South that encompass cities, suburbs, and rural areas. As a specification check, I use the NCES locale code of the student’s nearest school as an alternative definition of rural in my models. 90% of observations in my analytic sample have a nearest school with the same rural classification as their resident district. Results are similar and available by request.

5. A fringe rural area is a rural territory that is less than or equal to 5 miles from an Urbanized Area or less than or equal to 2.5 miles from an Urban Cluster. Since Urbanized Areas include suburban areas, fringe rural areas have low populations and are located outside of Metropolitan and Micropolitan Statistical Areas. A distant rural area is more than 5 miles but less than or equal to 25 miles from an Urbanized Area or more than 2.5 miles but less than or equal to 10 miles from an Urban Cluster. A remote rural area is more than 25 miles from an Urbanized Area and also more than 10 miles from an Urban Cluster (Geverdt, 2019).

6. I also estimate the model represented by Equation 1 with a resident district by year fixed effect. Results are similar and are available by request.