Examining the achievement test score gap between urban and suburban students

Abstract

Data from the National Longitudinal Survey of Labor Market Experience for Youth (1997 cohort) are used to examine the urban school achievement gap. Specifically, we use the Blinder–Oaxaca technique to decompose differences in Armed Services Vocational Aptitude Battery scores for students who attended urban and suburban schools. We find that approximately 75% of the gap in this achievement measure is explained by the high concentration of disadvantaged students in urban schools. Broken down further, 36% of the gap can be attributed to differences in family background. The lower income of urban families alone explains 25% of the gap. Differences in measures of school quality, such as small classes, large schools, and private school attendance, explain very little of the gap. While current policy focuses on schools and school reform, our results are a reminder that meaningful efforts to improve performance in urban schools must address socioeconomic conditions in urban areas.

Keywords:

• test score gap
• urban students
• decomposition

Notes

1. There are a high number of missing values, particularly among the family background variables, because only 88% of the parent surveys were completed and personal questions have a high refusal rate. If family background measures are excluded, our sample size is 4951 (compared with 1955 used in the current study with these measures included). However, missing values do not substantially alter the sample distribution. For example, 49% of the overall NLSY97 survey respondents (8984) are female, 21% are Hispanic, and 26% are black. Our subsample of 1955 respondents is 48% female, 22% Hispanic, and 20% black. While the representation of black respondents is relatively lower in our subsample (20 versus 26%), this minority remains over‐sampled compared with the national rate used in the NLSY97 representative portion of the sample (12%).

2. The point elasticity is 0.135 when evaluated at the variable means for the combined sample of urban and suburban students. Mean for income as a percent of poverty is 309.53, and the mean test score is 48.16.

3. Jann’s (2008) illustration is based on the pooled model decomposition employed by Oaxaca and Ransom (1994) and Nuemark (1988), but is applicable to our specification with separate equations.