IV Estimation with weak instrument s: an application to the determinants of school attainment in Peru

Abstract

The instrumental variables (IV) method has recently become widely popular in applied econometrics. However, recent work in the literature has shown that if instruments do not hold the relevance and exogeneity conditions at least in a «strong» way, the estimation results could be quite different from the population parameters, and an interpretation of the results would be meaningless. This paper presents this problem and shows some recent tests to verify the validity of the instruments employed. Additionally, I apply those tests to the estimation of the effects of child labor and household work on school attainment, by using a sample of 913 children from the Peruvian Living Standard Measurement Survey (LSMS) of 1997 and 2000. In this estimation, the instruments selected passed all the tests of exogeneity and relevance, so we can trust the consistency of these results. They show that household work performed by children has a positive impact on the children’s observed educational performance three years later. In addition, the impact of child labor is negative, but it is not significantly different from zero.