Why the gap between Black and White performance in school?: A report on the effects of race on student achievement in the St. Louis public schools

Journal of Negro Education, The, Summer 1997 by Trent, William T

This report, which was submitted to the court during the 1996 Liddell hearings, has been divided into two parts and revised to incorporate points raised during Trent's testimony on March 19, 1996. Part one analyzes the effects of race on student achievement and is based on data supplied by the St. Louis Public School District, including its student master file for 1994-95, cohort files covering four cohorts of students, and call-in enrollment data files. Part two draws on national survey data, and focuses on the effects of race on education, employment, and attitudes. The policy implications of both sets of findings are discussed.

My analysis of the St. Louis school district data suggests conclusions that are consistent with the research literature, including my own research on the subject. I found a persistent "race effect," by which I mean that Black students perform less well on the Stanford Achievement Test in both reading and mathematics than do their White peers. Using regression analysis, I attempted to see if the difference between Black and White students' scores could be explained in part by differences in the students` family backgrounds. For this analysis, I considered factors that are often associated with differences in achievement, and found that the race effect persists even after introduction of several control measures. Among these controls were student background (age, sex, and socioeconomic status [SES], as represented by eligibility for free or reduced-price lunches), prior test scores (the earliest test score available in the same subject in kindergarten or first grade), and school characteristics (including school size-that is, total enrollment, and poverty concentration).

In a separate analysis of the above data, I looked at poverty concentrations in the neighborhood where students live, as measured by U.S. Census tract. Again, the persistent, negative consequences of student race-or more precisely, being Black-remained, even after controlling for poverty concentration where students live. More important for educational policymakers, however, I found a consistently negative effect of high poverty concentrations in school on students' educational attainment. Black students are more likely to attend schools with higher concentrations of economically disadvantaged students than are White students. Thus, they are more likely to experience a quality of educational treatment that reduces their scores on the Stanford Achievement Test, even after factoring out the effects of other possible causes. Tables I and II present the evidence for these conclusions.1 For the benefit of nonsocial scientists, I will try to explain the meaning of these numbers. My analysis of the St. Louis school district data entailed merging the school system's master student file, four cohort files, and its call-in enrollment data to create a working file of 4,096 cases for individual students. I approached these data by means of regression analyses. To try to explain why there exists a gap between the achievement scores of Black and White students in the St. Louis area, I employed students' achievement test scores in reading and mathematics as dependent variables. I next introduced several independent variables to determine their effect on the difference. These factors included students race, age, sex, socioeconomic status (eligibility for free or reduced-price school lunches), prior test scores in the same subject, and school characteristics.2 Some of the independent variables were found to yield no difference, while others made a real difference. Some of these differences were statistically significant-that is, they evidenced a high likelihood that the difference resulting from a variable could not have occurred by chance.

First, I determined, by use of regression analysis, the differences in achievement scores of Black and White students. To walk through Table I, the value for the unstandardized coefficient associated with race ("B") is -24.09, which represents how much being African American is associated with reducing a student's full scale score3-the "effect" of racewith a statistically significant "t" value of -17.703.(4) The adjusted R^sup 2^, .073, is a measure of how much of the variance in test scores is explained by each factor. As shown in the final line of this table, 7% of the difference in test scores can be explained by race alone.

I next introduced and controlled for factors known to be associated with race such as student background, prior test scores (a measure of the ability of students at the earliest school grade), and poverty level of the school. As shown in Table II, by introducing SES, the effect of race was reduced by approximately seven points (24.090 minus 17.712), lowering the unstandardized coefficient to 17.712. Thus, approximately 23% (.234) of the variation in achievement test outcomes can be explained by race and SES, suggesting that background factors add an additional 16% (.234 minus .073). Adding prior test scores produced very little change in the coefficient for race, less than two-tenths of a point. Poverty concentration in the school, on the other hand, reduced the negative effect of race by an additional seven points (17.531 minus 10.017). Thus, these variables have explained about 35% of the variance in reading achievement test scores. Race remained statistically significant and meaningful after the inclusion of all the independent variables in the analysis.