This article considers empirical likelihood in the context of efficient semiparametric estimators of average treatment effects. It shows that the empirical likelihood ratio converges to a nonstandard distribution, and proposes a corrected test statistic that is asymptotically chi-squared. A small Monte Carlo experiment suggests that the corrected empirical likelihood ratio statistic has competitive finite sample properties. The results of the article are applied to estimate the environmental effect of the World Trade Organisation.