Erratum

In attempting to replicate the results on this investigation, Doug Almond and Bhashkar Mazumder have found a programming error in the original code Adriana Lleras‐Muney wrote (mean education levels for 1960 and 1970 were calculated inadvertently using the 1970 and 1980 data). Although the qualitative results and conclusions of the paper are unaffected, this error affects the aggregate OLS and TSIV results. The author of the paper is grateful to Almond and Mazumder for pointing out this error. The OLS value reported in Table 3 (page 208) was − 0·017(0·004), the corrected coefficient should be − 0·036(0·0043). The IV value reported in Table 4 (page 210) was − 0·051(0·026), the corrected coefficient is − 0·063(0·024). Since Larry Katz also uncovered some minor typographical errors in the state data variables used as controls in the paper (the old and corrected data sets are posted at the author’s web site at www.princeton.edu/~alleras ) the author has taken this opportunity to report results without any state controls for future reference. The OLS coefficient without state controls is − 0·036(0·0043) and the IV without state controls − 0·058(0·023). In table 5 (page 211), the corrected IV coefficients are as follows: Panel A (quarter of birth): − 0·048(0·023); Panel B (by gender): males: − 0·057(0·034); females − 0·037(0·033); Panel C (exclude 40, 50 and 60): − 0·058(0·024); Panel D (interact cohort dummies and 1970): − 0·066(0·025); Panel E (by regions): north − 0·036(0·066), south − 0·057(0·035), west − 0·057(0·058), midwest − 0·032(0·047).