Abstract: This paper explores the determinants of children's cognitive outcomes using novel panel data from two cohorts of children in India. As in Todd and Wolpin's study (2007), I do not find evidence supporting restrictive models that assume test scores depend only on contemporaneous inputs. Rather, the results show that lagged inputs affect the production of current skills. In models where past inputs are not observed or imperfectly measured, past nutritional status and/or past test scores turn out to be a good proxy-indicator of this variable, which is evidence for the 'sufficiency' assumption. I allow for the endogeneity of nutrition using an instrumental variable approach and find that a 1 standard deviation increase in height-for-age z-scores at the age of 1 leads to cognitive test scores that are about a quarter of a standard deviation higher at age 5. Results for the Older Cohort have a similar magnitude. I also study the behaviour of home inputs and find that parents seem to 'reinforce' children for early favourable outcomes rather than 'compensate' them for adverse scores; and they do so more in lower caste families and particularly with boys.

Keywords: cognitive outcomes, health, nutrition, children and test score gap