A Potential Outcome Approach to Dynamic Programme Evaluation : Nonparametric Identification

This paper approaches the problem of an econometric evaluation of dynamic programme sequences from an potential outcome perspective. The identifying power of several different assumptions about the connection between the dynamic selection process and the potential outcomes of different programme sequences is discussed. The assumptions invoke different types of randomisation compatible with different selection regimes. Parametric forms are not involved. When participation in the sequences is decided every period depending on the success in the past, the resulting endogeneity problem destroys nonparametric identification for many parameters of interest, so that several dynamic versions of the average treatment effects on the treated parameter are not identified. However, some interesting dynamic forms of the average treatment effect are still identified. We also present a bounds analysis to learn from the data as much as possible, even when parts of the identifying assumptions are violated