Child Care Choices and Children’s Cognitive Achievement: The Case of Single Mothers

We evaluate the effect of childcare vs. maternal time inputs on child cognitive achievement using the single mothers from the National Longitudinal Survey of Youth (NLSY79). To deal with non-random selection of children into childcare, we exploit the (plausibly) exogenous variation in welfare policy rules facing single mothers. In particular, the 1996 Welfare Reform, and earlier State level policy changes, generated substantial increases in their work/childcare use. Thus, we construct a comprehensive set of welfare policy variables, and use them (along with local demand conditions) as instruments to estimate child cognitive ability production functions. Because welfare rules are complex, we need many variables to characterize them. Thus, we face a “many instrument problem” (i.e., 2SLS severely biased toward OLS). We deal with this problem both by using LIML, and by using factor analysis to condense the instrument set. Results from the two approaches are very similar, and quite different from OLS. Using LIML along with factor analysis of the instruments leads to an efficiency gain (i.e., smaller standard errors) relative to using LIML alone. In our baseline specification, we estimate that a year of childcare reduces child test scores by 2.1% (.114 standard deviations). This estimate is quite robust across a wide range of specifications and instrument sets. But we find important interactions with type of care, maternal education and child gender. Indeed, only informal care leads to significant reductions in cognitive outcomes. Formal center-based care does not have any adverse effect. In addition, the value of the maternal time input is greater for more educated mothers, and girls are more adversely affect by childcare than boys. We do not find differential effects by child age or race/ethnicity. 1 Web appendices available: http://economia.uniandes.edu.co/es/profesores/planta/bernal_raquel/documentos_de_trabajo 2 We acknowledge support from Australian Research Council grants FF0561843 and DP0774247, and Bernal acknowledges financial support from “The Searle Fund.” Corresponding author: [email protected].