In a recent article Newey and Windmeijer (Generalized method of moments with many weak moment conditions. Econometrica 2009; 77(3): 687-719) propose a new variance estimator for generalized empirical likelihood. In Monte Carlo examples they show that t-statistics based on the new variance estimator have nearly correct size. I have replicated their Monte Carlo simulations and in addition used the new variance estimator to re-estimate Angrist and Krueger's (Does compulsory school attendance affect schooling and earnings? Quarterly Journal of Economics 1991; 106(4): 979-1014) returns to education.
