Essays on social interactions and the long-term effects of early-life conditions

Dissertation presented at Uppsala University to be publicly examined in Hörsal 2, Ekonomikum, Kyrkogårdsgatan 10, Uppsala, Friday, December 11, 2009 at 14:15 for the degree of Doctor of Philosophy. The examination will be conducted in English. NILSSON, Peter, 2009, Essays on Social Interactions and the Long-term Effects of Early-life Conditions; Department of Economics, Uppsala University, Economic studies 120, 180 pp, ISBN 978-91-85519-27-9; ISSN 0283-7668, urn:nbn:se:uu:diva-109671(http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-109671)

This thesis consists of four self-contained essays.

Essay 1: From the late 1970s through mid 1990s blood-lead levels decreased drastically in Swedish children due to the sharp phase-out of leaded gasoline. Exploiting the distinct geographical variation in early childhood lead exposure induced by the regulations together with micro data on all children in nine birth cohorts I show that reduced lead exposure early in life improves scholastic performance, cognitive ability, and labor market outcomes among young adults. At the relatively low levels of exposure considered, the analysis reveals a nonlinear relationship between local air lead levels in early childhood and adult outcomes, indicating the existence of a threshold below which further reductions no longer improve adult outcomes. Importantly, the effect is greater for children of lower socioeconomic status (SES), suggesting that pollution is one mechanism through which SES affects long-term economic outcomes and that environmental policies potentially can reduce the intergenerational correlation in economic outcomes.

Essay 2: During a policy experiment in two Swedish regions in 1967 alcohol availability increased sharply, particularly for people under age 21. The policy experiment was abruptly ended after only 8.5 months due to a sharp increase in alcohol consumption. I exploit the distinct temporal, spatial and age-specific changes in alcohol availability induced by the policy experiment to estimate the long-term effects on those exposed to it in utero. I find that children in utero during the short period of increased alcohol availability have significantly lower educational attainments, earnings and increased welfare dependency rates at age 30 in comparison with the surrounding cohorts. Any direct effects of the increased availability on birth-cohort composition (e.g. through an increase in unplanned pregnancies) are not driving the results as the richness of the data allows for a focus on exposed children conceived before the policy experiment started. The results provide compelling evidence that investments in early-life health can yield large effects on outcomes later on in life.

Essay 3: We utilize a large-scale randomized social experiment to identify how co-workers affect each other’s effort as measured by work absence. The experiment altered the work absence incentives for half of all employees living in Göteborg, Sweden. Using administrative data we are able to recover the treatment status of all workers in more than 3,000 workplaces. We first document that employees in workplaces with a high proportion treated co-workers increase their own absence level significantly. We then examine the heterogeneity of the treatment effect in order to explore what mechanisms are underlying the peer effect. While a strong effect of having a high proportion of treated co-workers is found for the non-treated workers, no similar effects are found for the treated workers. These results suggest that pure altruistic social preferences can be ruled out as the main motivator for the behavior of a non-negligible proportion of the employees in our sample.

Essay 4: We examine the influence that co-workers’ have on each other’s fertility decisions. Using linked employer-employee panel data for Sweden we show that female individual fertility increases if a co-worker recently had a child. The timing of births among co-workers of the same sex, educational level and co-workers who are close in age are even more influential. Consistent with models of social learning we find that the peer effect for first time mothers is similar irrespective of the birth order of the co-worker’s child, while for higher order births within-parity peer effects are strong but crossparity peer effects are entirely absent. A causal interpretation of our estimates is strengthened by several falsification tests showing that neither unobserved common shocks at the workplace level, nor sorting of workers between workplaces are likely to explain the observed peer effect. We also provide evidence suggesting that peers not only affect timing of births but potentially also completed fertility, and that fertility peer influences spills over across multiple networks. Our results suggest that social interactions could be an important factor behind the strong inter-temporal fluctuations in total fertility rates observed in many countries.