How parents choose schools: Evidence from the high-school admissions process in Romania

The intervention takes place at an information session at middle schools to which parents of middle school children are invited. In the control group, no information is provided, and parents complete only a baseline survey. In the treatment group, parents first complete the baseline survey and are subsequently presented with information about the concept of value added, and told the within-town rank of value added of their high school track choices. After receiving the information, parents in the treatment group are asked again to rank their track choices.

Since Friedman (1955), economists have asked if school choice improves educational outcomes. This likely depends on whether households prefer schools that are more effective in the sense of having a greater impact on human capital (Hanushek 1981; Rothstein 2006; Abdulkadiroglu et al 2017). If households value effectiveness, schools will compete on this trait. Whether this is the case remains a question, in part, because the evidence on the causal impact of choice is mixed. One possibility often raised is that households simply are unable to observe value added. In this project we estimate the value added for high schools in Romania (validating it with regression-discontinuity-based estimates). We will experimentally explore if distributing this information alters households’ school choice. This is feasible because in Romania school selection is centralized. Thus we can observe school choices among treatment and control groups, and then see if information on value added affects the resulting choices.

In Romania, there is a centralized mechanism to allocate middle-school students to tracks within high schools. (Note that students select tracks within schools, rather than schools.) Students take an exam at the end of middle school. They submit a ranking of their school choices and are then allocated based on a serial dictatorship. Under this allocation mechanism, students’ preferences are considered in the order in which they score on an admissions index, which is the combination of the admissions exam and the middle-school GPA. As a result, each school has a minimum threshold above which students can gain admission. Students are assigned to the track that they most prefer among those to which they gain admission.

We calculate value added with respect to passing the baccalaureate exam. For each year, we regress an indicator for the student passing the exam on the student’s covariates and a track fixed effect. Student covariates include whether the student is female; cubics in the student’s middle-school GPA and in scores on each component of the high-school admissions exam; interactions between female and the middle-school GPA and admissions exam scores; a cubic in the number of students in the student’s middle school; an interaction of female and the number of students in the student’s middle school; and averages and standard deviations of middle-school GPA and admissions exam scores in the student’s middle school. We interpret the track fixed effect as the track’s value added. Since this measure depends on scores on the 12th grade baccalaureate exam, it is not observable to households at the time of track choice. Consequently, we predict it using a local linear random forest (Friedberg, Tibshirani, Athey, and Wager 2019) with covariates that are available to households at the time of track choice. These covariates include the track’s academic specialty and past values of value added and peer quality that are observable at the time of track choice. We then calculate the within-town rank of predicted value added for the 2019 entering cohort. We provide this rank to treated households in the information session.

2019-06-03

2019-08-31

Perceived value added of high school tracks, high school track choice ranks

Parents are asked to rate each high school track's value added with respect to their child passing the baccalaureate exam on a 1 to 5 scale, with 1 representing a low perceived value added, and 5 a high perceived value added. Parents report how they ranked their available high school track choices, from most to least preferred.

Perceived value added of high school tracks in a follow-up survey

In a follow-up survey we will again ask parents to rank their child's high school track choices by value added.

The intervention takes place at an information session at middle schools to which parents of middle school children are invited. In the control group, no information is provided, and parents complete only a baseline survey. In the treatment group, parents first complete the baseline survey and are subsequently presented with information about the concept of value added, and told the within-town rank of value added of the high school track choices. After receiving the information, parents in the treatment group are asked again to rank their track choices.

Randomization by computer

Middle schools

Yes

The 228 middle schools corresponding to 280 individual classes. We expect non-participation at the school level of 10% to 30%.

We expect an interview rate of 15 to 20 students per class, so approximately 3630 to 4840 students. We expect a follow-up interview rate of 50% to 60%.

We study 228 middle schools. Of these, 114 are randomly assigned into the treatment group and 114 into the control group.