Principal investigator: Eduarda Miller de Figueiredo
Original Article Title: Does School Integration Generate Peer Effects? Evidence from Boston's Metco Program
Authors: Joshua D. Angrist and Kevin Lang
Location of the Intervention: Boston
Sample Size: 6.000 students
Sector: Education
Primary Variable of Interest: Score on educational achievement tests
Type of Intervention: Sending students to schools in other districts
Methodology: Instrumental Variables
Summary
Literature has shown that policies integrating different schools through the transportation of students from one district to another have positive results due to increased racial integration and economic gains for Black students. However, there had been no study of the impacts of this type of program on the students who "received" students from other districts. To complete this column, the authors of this article seek to provide an analysis of this type of program through regressions on the performance tests of students in the Metco program in Boston. Through OLS regressions and the use of instrumental variables, the authors found little evidence of significant social or statistically significant effects.
- Policy Problem
The social consequences of school integration have been debated since the year the article discussed here was developed – and continue to this day – with policymakers and researchers debating the impact of students being bused to schools in other school districts for the purpose of racial balance. According to Welch and Light (1987) and Rossel and Armour (1996), integration policies have been partially successful because they have increased the likelihood of white and black students studying together and have resulted in economic gains for blacks with the end of segregation.
Due to the Metco program, Black students were transported by bus to schools that only had white students, and vice versa. However, there were no studies analyzing the impacts of this school integration on the white students, who remained in the schools to which the Black students were transported by bus.
Therefore, the study sought to provide an initial overview of the impact of Metco by analyzing average school-level test scores in the population of the districts that received the program and their surrounding areas.
- Implementation and Evaluation Context
The Metco program sends students – mostly Black – from the Boston district to schools in neighboring suburban districts, which are predominantly white. Metco is voluntary on the part of the families of the students being transported and the school districts receiving the students, as parents interested in enrolling their children in the program must join a waiting list.
In 1970, four years before the decision that mandated bus use in Boston's district, 29 districts that received the program enrolled nearly 1.400 students. In the 2000-2001 school year, nearly 3.200 Metco students attended schools in 32 suburban districts.
- Policy/Program Details
Candidates are selected from the waiting list for participation in Metco on a first-come, first-served basis. Parents in Bostan cannot choose a suburban district, but they can refuse placement, which becomes more difficult as the student progresses through the grades.
The state provides funding to districts that accept Metco students, which currently amounts to around $2.800 per student, plus transportation costs. Brookline has approximately 6.000 public school students attending eight neighborhood elementary schools with grades from kindergarten through eighth grade and a single high school, where students consistently perform well on national tests, have low dropout rates, and a high likelihood of attending school. Furthermore, 10% of Brookline students are Black (including those in the program).
Under the contract with Metco, Brookline enrolls 300 students in the program each year, which represents about 5% of total enrollments.
- Assessment Method
The scores from the fourth grade onwards were analyzed because, in addition to being the only grade where math and English proficiency is tested, there are many more elementary schools than high schools in each district. This, according to the authors, facilitated a conditional analysis of district effects. Performance data from the ITBS were used.[1] for the school years 1994-2000, from a sample of districts operating within or along Interstate 195[2].
Binary variables were added to the model for districts within or along Route 128.[3], variables for the number of students tested and binary variables for the type of school (elementary or middle school).
The authors point out that preliminary results may be biased due to omitted variables and, therefore, they also perform estimates using the instrumental variables (IV) method. The IV estimates exploit the fact that Metco students are enrolled in Brookline schools, in part based on a space constraint. When classes are expected to be small, the Metco office is notified that space is available for program students. Knowing this, the authors modeled the Metco enrollment process by allocating one program student per classroom if the expected enrollment is less than 23 (considering classes of up to 25 students).
- Main results
A significant racial disparity was observed for Brookline residents, with Metco students scoring significantly lower than Brookline residents. However, Black students in the program scored very similarly to Black Brookline students.
The differences in average performance between Brookline residents and Metco students are large enough that the presence of students participating in the program reduces average test scores in Brookline. Therefore, the presence of Metco students has a marked negative impact on average performance. Increasing the number of program students by ten percentage points in Brookline (about two per class) reduces average scores by almost 2,5 percentage points.
The authors point out that previous research suggests a strong positive correlation between individual performance and peer performance levels in the classroom (Lazear, 2001), which may indicate a peer effect. Thus, the effect of the Metco fraction on average scores is large enough that increases in the program fraction could induce a negative peer effect.
Estimating the effect of Metco students on the performance of non-Metco students, through a regression of the national average percentile ranking of non-Metco students and another regression from microdata controlling for student characteristics, shows small positive effects in the first approach and a negative effect in the second approach, but both approaches are insignificant. This suggests that the proportion of Metco students has no effect on non-Metco students.
Estimates using the instrumental variables method show almost negligible effects of class size on Metco achievement in Brookline.
- Lessons in Public Policy
Although Metco students have noticeably lower scores than Brookline students, the authors find little evidence of statistically significant social or blending effects. This is because most studies of peer effects and student mixing are confounded by omitted variable bias.
References
ANGRIST, Joshua D.; LANG, Kevin. Does school integration generate peer effects? Evidence from Boston's Metco Program. American Economic Review, v. 94, no. 5, p. 1613-1634, 2004.
[1] Iowa Test of Basic Skills.
[2] A road around Boston in its western part, known as the Boston Inner Beltway.
