Association for Public Policy Analysis & Management

A rich body of literature has examined the effect of summer vacation on test score gaps by socioeconomic status (SES) and race/ethnicity.  Early studies in geographically circumscribed areas (Heyns, 1978; Entwisle & Alexander, 1992) found that summer vacation had a detrimental effect on SES-based achievement gaps.  Initial analyses of the nationally representative Early Childhood Longitudinal Study, Kindergarten Class of 1998-1999 (ECLS-K:99) corroborated these early results, but recent re-analyses by von Hippel (2014) call some findings into question.  Findings on the black-white gap have been similarly contested (Heyns, 1987; Fryer & Levitt, 2004). 

In this study, I use newly-released nationally representative data from the National Center for Education Statistics’ ECLS-K: 2011 to replicate and extend research on seasonal gap trends.  I go beyond past research in 2 ways.  First, seasonal gap trend studies have focused almost exclusively on absolute group mean differences in learning rates by season.  While this is an important perspective, it does not reveal how the overlap of groups’ test score distributions may change by season, which is a relevant perspective on inequality.  I therefore examine differences in learning rates as well as changes in distributional overlap.  Second, previous studies have not decomposed seasonal gap changes into within and between-school changes.  Comparing summer and school year gap change decompositions can provide important evidence about whether within school or between school factors contribute to widening (or shrinking) of gaps.  In this study, I ask the following research questions: 

1)      What are the racial/ethnic and socioeconomic test score gap trends in math, reading, and working memory over kindergarten, summer, and first grade?

• How do differences in learning rates and changes in test score variance contribute to inequality dynamics?

2)      To what extent do gaps change within versus between schools over kindergarten, summer, and first grade?

I use OLS regression to estimate changes in racial/ethnic gaps and differences in changes (Ho, 2009; Quinn, 2015); to estimate socioeconomic gap trends, I obtain the model-implied 10^th -90^th percentile SES gaps (based on a continuous composite of parental income, education, and occupational prestige). I use methods described by Reardon (2008) and Page, Murnane, and Willett (2008) to decompose gap changes into within- and between-school components.

My results show a general pattern in which social groups that begin kindergarten with lower academic skills tend to either receive an initial learning boost over kindergarten or equalization of learning rates (compared to their more advantaged reference group).  After this initial school boost, however, the trend is toward between-school factors helping advantaged groups pull further ahead of their less advantaged peers.  When operationalizing gap changes as changes in the relative difference between groups over time, gaps trends are amplified by the fact that test score variance shrinks over the school year and expands over the summer.  Finally, racial/ethnic gaps tend to widen primarily between schools over the school year, but narrow between schools over the summer (controlling for SES).  This finding is consistent with between-school differences in school quality leading to widening academic achievement gaps.