Review: Inequality at the Starting Gate by Lee & Burkam

Inequality at the Starting Gate: Social Background Differences in Achievement as Children Begin School

A Review

 Dry ahistorical statistical analysis of child inequality

Lee and Berkham use data from the U.S. Department of Education's Early Childhood Longitudinal Study, Kindergarten Cohort (ECLS-K) to show that various variables affect `cognitive skills' of children, such as everything from socioeconomic context to television habits, particularly as children first enter school. They contend that of all influences, socioeconomic status has the most impact on child learning, and propose several directions for policy based on their analysis, including beginning out-of-home daycare/school at an earlier age and increasing school resources in low SES communities.

Why only a "2" star rating? The book is dry and ahistorical: there is no analysis of *why* or *how* we find ourselves in the dire situation of today. Instead, Lee and Burkam act as if we could disentangle race and class through hierarchical modeling, and construct the problem as a simple one of addition: race contributes this much to learning, class this much, school resources this much, and so on. In fact, historically this is absurd, particularly in the Unites States where class formation occurred along the lines of race institutionalized in slavery. There is no simple number that will extricate race from class.

There are many other problems with the work, even within its own paradigm. For example, Lee and Burkam use a hierarchical least squares regression model to disentangle the effects of (in order): race; social class; child demographics; home demographics; education expectations and pre-K care; at-home activities; outside-home activities (p. 49-56)

However, this model has a bias that is left unexamined in the report: the order in which the variables are used in the hierarchical analysis matters. This model is mathematically hierarchical; when it is applied to social science situations, it is generally used to study phenomena that are naturally hierarchically structured. For example, the first level might divide students by state, then by district, then by school, and so on (in naturally nested subunits of students). However, students are not naturally divided into race, social class, and so on in the same naturally hierarchical way. In fact, Lee and Burkam rank their chosen variables in order of what they believe most characterize students: first, at the top as the most significant characteristic, race, then, social class, and so on. Their analysis would yield different numbers/correlations, in other words, if they had ranked their variables differently. Their 'results', then, that SES matters more than class, are in fact invalidated by their own construction of the problem.

Other question marks: This research is published by EPI, according to them, a non-profit thinktank (turns out EPI is funded by labor unions and similar organizations). Of note is that this research is published `in-house'. Also: Lee and Burkam do not critically question the hierarchical framework of testing and of their own quantitative framework, do not analyze their assumptions, and so on.

This research could have been published as one article--certainly it does not merit a book.