The empirical content of season-of-birth effects : an investigation with Turkish data

Although the season of birth variable is often used as an instrumental variable to estimate the rate of returns to schooling in the labor economics literature, there is an emerging consensus that the season of birth is systematically associated with later outcomes in life such as the educational and labor market success; thus, it is highly likely non-random. Using a large micro-level data set from Turkey, we argue that the degree of this non-randomness can be even larger in a developing-country context. Specifically, we show that around 20 percent of all individuals in Turkey have January as their month of birth due to a combination of geographical, seasonal, institutional, and idiosyncratic factors that lead to misreporting. We further document that being January-born strongly predicts worse socio-economic outcomes in later life. We show that this can be a serious problem in evaluating policies that define eligibility based on the month of birth - such as compulsory schooling and compulsory military service laws that set the eligibility birth date cutoff as the January 1st. We confirm the validity of this concern based on a series of regression discontinuity design exercises. We conclude that, in a developing-country context, additional caution should be exercised when using the season-of-birth variable as a statistical tool.