Indiana University SPEA Edward J. Bloustein School of Planning and Public Policy University of Pennsylvania AIR American University

Panel Paper: Preschool Growth in Mathematics and Long-Run Achievement: An Instrumental Variables Approach

Thursday, November 12, 2015 : 8:30 AM
Brickell South (Hyatt Regency Miami)

*Names in bold indicate Presenter

Tyler Watts1, Greg J. Duncan1, Douglas H. Clements2, Julie Sarama2, Christopher B. Wolfe3 and Mary Elaine Spitler2, (1)University of California, Irvine, (2)University of Denver, (3)St. Leo University
An accumulating body of research has found that school-entry mathematical skills strongly predict later measures of mathematics achievement.  These well-controlled, correlational, studies have been viewed as evidence that interventionists should target school entry academic skills, as spurring gains in early academic competencies could lead to long-run changes in academic trajectories.  However, recent experimental evidence suggests that such intervention impacts do not last, as most effects fade substantially in the years following the end of treatment.  Due to these discrepant findings, the current paper investigated the causal relation between preschool mathematics gains and late elementary school mathematics achievement. We used treatment assignment to an early mathematics intervention as an instrument for preschool mathematics growth using a two-stage least squares (2SLS) design.

  Data were drawn from the TRIAD (Technology-enhanced, Research-based, Instruction, Assessment, and professional Development) study, which evaluated the scale-up of Building Blocks (BB), a preschool mathematics curriculum.  Thirty schools were recruited from low-income neighborhoods in two Northeastern states, and participating schools were rank ordered based on their state-achievement test scores.  The ranked schools were then organized into 8 blocks, and schools were randomly assigned to either treatment or control within block.  We used block and treatment interactions as instruments for preschool mathematics gains, measured by a high-quality mathematics test administered in the fall (baseline) and spring (post-treatment) of preschool.  Dependent variables were measures of mathematics achievement in the fall and spring of fourth grade, and spring of fifth grade.

 The treatment and block interactions produced a large effects on preschool mathematics gains (F(8) =41.46, p < .001), and analytical checks suggested that baseline characteristics were balanced at random assignment. Further, the instruments had no detectable effect on later mathematics achievement except through preschool mathematics gains.  Results from 2SLS models indicated that a standard-deviation of intervention produced growth during preschool had an average impact of approximately .25 standard deviations on measures of achievement in fourth and fifth grade.  To our surprise, we found that most of this effect was driven by the impact of instrumented preschool gains on achievement in fifth grade, as preschool gains had no significant impact on achievement in fourth grade.  The average .25 impact across fourth and fifth grade was approximately half the size of the effect produced by OLS models relating later achievement to preschool math growth in the current data, and was approximately 30%-40% smaller than the effect reported by highly-controlled regression models from other datasets.

Our results indicate that although gains made in preschool academic skills do appear to have a causal effect on later achievement, this impact is not as large or stable as the correlational literature would suggest.  Thus, early academic interventions should be pursued with reasonable expectations, as spurring even large gains during preschool is probably insufficient to make a sizable and permanent impact on long-run academic trajectories.

Full Paper: