A large literature in developmental psychology and economics examines the impact of non-parental child care on measures of child cognitive and behavioral development. Yet for a number of reasons this literature has yielded mixed results.  This lack of consistency may be due in part to the challenge of surmounting the endogeneity of child care participation. Children who enroll in non-parental care are likely to differ from those who do not in ways that are difficult to capture using even rich specifications of the child production function. If these unobserved child or family differences influence measures of child well-being, then the estimated effect of non-parental care will be biased.

This paper proposes a novel empirical strategy to attempt to estimate the causal effect of non-parental child care. In particular, it uses plausibly exogenous variation in child care use generated by the participation “dip” that occurs during the summer months. To measure children’s enrollment status during the summer and non-summer months, I use the Birth cohort of the Early Childhood Longitudinal Study (ECLS-B), a nationally representative longitudinal survey of approximately 11,000 children born in 2001. A unique feature of this survey is that, during Waves 1 and 2 (when children are 9- and 24-months old, respectively), parental interviews and child assessments were conducted on a rolling basis throughout the calendar year. These survey waves are merged with state-specific information on school start and end dates in order to construct a measure of whether a given ECLS-B interview was conducted during a summer month.

Under the presumption that the timing of ECLS-B interviews are (i) correlated with non-parental child care use, but (ii) uncorrelated with measures of child well-being, this paper uses the “summer interview” variable as an instrument for child care participation. Regarding the first criterion, I find that child care enrollment rates in the ECLS-B vary substantially over the calendar year, with children experiencing a pronounced participation “dip” during the summer months. I buttress this finding with monthly establishment-level data on child care labor supply, which reveals similarly large employment dips between June and September of each year. Regarding the second criterion, the “summer interview” instrument would be invalidated if the timing of ECLS-B interviews is correlated with child well-being. Fortunately, the ECLS-B strictly controlled the timing of interviews in Waves 1 and 2 such that families were interviewed as close to each child’s 9- and 24-month birthday as possible. Therefore, conditional on controls for child age and month-of-birth, the “summer interview” instrument should be uncorrelated with the unobserved determinants of child well-being.

I examine a rich set of outcomes: body mass index, overweight and obesity propensity, Bayley Short Form mental and motor skills scores, and parental reports of child health. My (very) preliminary instrumental variables estimates suggest that non-parental child care use has varying impacts on child well-being, depending on the outcome and type of child care assessed.