Forced to be Rich? Returns to Compulsory Schooling in Britain

Abstract

Researchers using changes in compulsory schooling laws as instruments have typically estimated very high returns to additional schooling that are greater than the corresponding OLS estimates. Given that the first order source of bias in OLS is likely to be upward as more able individuals tend to obtain more education, such high estimates are usually rationalized as reflecting the fact that the group of individuals who are influenced by the law change have particularly high returns to education. That is, the Local Average Treatment Effect (LATE) is larger than the average treatment effect (ATE). However, studies of a 1947 British compulsory schooling law change that impacted about half the relevant population (so the LATE approximates the ATE) have also found very high IV returns to schooling (about 15%), suggesting that the ATE of schooling is greater than OLS estimates would suggest. This constitutes a puzzle: How can the OLS return to schooling be a significantly downward biased estimate of the ATE when the primary source of OLS bias should be upward? We utilize a source of earnings data, the New Earnings Survey Panel Data-set (NESPD), that is superior to the datasets previously used and conclude that there is no such puzzle: the IV estimates are small and much lower than OLS. In fact, there is no evidence of any return for women and the return for men is in the 4-7% range. We do, however, find that men benefit from greater schooling through a reduction in earnings variability.