Lower and upper bounds of returns to schooling: An exercise in IV estimation with different instruments

Several recent studies based on 'exogenous' sources of variation in educati onal outcomes show IV estimates of returns to schooling that are substantia lly higher than the corresponding OLS estimates. Card (1995a, Earnings, sch ooling, and ability revisited. Research in Labor Economics 14, 23-48) sugge sts that these results are explained by the existence of heterogenity in in dividual returns and by the fact that these studies are based on instrument s that influence only the educational decision of individuals with high mar ginal returns due to either liquidity constraints or to high ability. This conclusion is consistent with the local average treatment effect (LATE) int erpretation of IV (Imbens and Angrist, 1994, Identification and estimation of local average treatment effects. Econometrica 62, 467-475) according to which IV identifies only the average returns of those who comply with the a ssignment-to-treatment mechanism implied by the instrument. We show evidenc e for Germany suggesting that returns to schooling are heterogeneous, instr uments matter and the LATE interpretation of IV makes sense. With an approp riate choice of instruments we also show how IV can be used to approximate the range of variations of returns to schooling in Germany. (C) 1999 Elsevi er Science B.V. All rights reserved. JEL classification: J31; I21, J24; C25 .