LEEBOUNDS: Stata module for estimating Lee (2009) treatment effect bounds

leebounds computes treatment effect bounds for samples with non-random sample selection/attrition as proposed by Lee (Review of Economic Studies, 2009). The lower and upper bound, respectively, correspond to extreme assumptions about the missing information that are consistent with the observed data. As opposed to parametric approaches to correcting for sample selection bias, such as the classical Heckman (1979) estimator, Lee (2009) bounds rest on very few assumptions, i.e. random assignment of treatment and monotonicity. Monotonicity means that the treatment status affects selection in just one direction. That is, receiving a treatment makes selection either more or less likely for any observation. In technical terms, the approach rests on a trimming procedure. Either from below or from above, the group (treatment, control) that suffers less from sample attrition is trimmed at the quantile of the outcome variable that corresponds to the share of 'excess observations' in this group. Calculating group differentials in mean outcome yields the lower and the upper bound, respectively, for the treatment effect depending on whether trimming is from below or above. leebounds assumes that it is unknown, a priori, which group (treatment, control) is subject to the higher selection probability and estimates this from data (see Lee, 2009:1090).