AERA 2018 Annual Meeting: “The Dreams, Possibilities, and Necessity of Public Education”
April 13-17, 2018
Title: The Role of CSRP and Early Environmental Factors in School Selection
Authors: Deanna A. Ibrahim, Tyler Watts (PhD Alumnus), C. Cybele Raver
Reardon and colleagues (2016) elucidated the national pattern of inequality in educational outcomes at the district level using a new dataset that includes district-level information on race-based achievement gaps across the country. We leveraged this new dataset to investigate whether the Chicago School Readiness Project (CSRP) influenced students’ eventual selection into school districts that contained more or less opportunity to succeed. These analyses also allowed us to examine the early environmental influences on school selection.
This study utilized student-level data from the CSRP study’s 11-year follow-up, in which 463 students reported the school they attend (most were enrolled in high school). We merged this data with data from the Stanford Education Data Archive (SEDA; 407 students merged), which contained district-level information regarding academic achievement, race-based achievement gaps, and neighborhood indicators of disadvantage. Further, during the Head Start year, parents reported their home addresses, which allowed us to use geocoded data from the US census tract to incorporate early neighborhood characteristics. The majority of students attended a school in the Chicago Public School (CPS) district (74%), while 26% attended a school outside of CPS. In general, CPS schools were more disadvantaged than non-CPS schools, as CPS schools had higher race-based achievement gaps (p < 0.001) and lower levels of achievement (p < 0.001). In general, we found that student-level characteristics were balanced between the preschool treatment and control groups, but neighborhood and Head Start characteristics tended to favor the control group. Given that we found vast differences in markers of disadvantage between the CPS and non-CPS schools, we measured selection into districts as a binary indicator for whether a student was enrolled in a CPS (coded “0”) or non-CPS school (coded “1”).
Preliminary results suggest that the students assigned to the treatment group were approximately 10-15% more likely to attend a non-CPS school in high school. This effect was robust to the inclusion of personal demographic characteristics, neighborhood and Head Start site information. Further, we found that Black and Hispanic students were much less likely to attend a non-CPS school in high school, however, this result was completely explained by the racial makeup of children’s early childhood neighborhoods. Results were robust to alternative modeling specifications (i.e., logistic and probit models).
Our results indicate that early childhood environmental factors influence eventual selection into high school environments. Further, we found evidence that the program influenced students to attend schools with less markers of inequality over 10 years after the program ended. Future work will model outcomes at the school level in order to exploit variation in school-level differences.