Contents: Heckman's sample selection model is undoubtedly one of the most important works in 20th century program evaluation. His seminal work triggered both tremendous discussion on modeling selection bias as well as the development of new statistical models addressing the problem of nonignorable treatment assignment. Heckman's key contributions to program evaluation are summarized as follows: (a) he provided a theoretical framework that emphasized the importance of modeling the dummy endogenous variable; (b) his model was the first attempt that estimated the probability (i.e., the propensity score) of a participant being in one of the two conditions indicated by the endogenous dummy variable, and then used the estimated propensity-score model to estimate coefficients of the regression model; (c) he treated the unobserved selection factors as a problem of specification error or a problem of omitted variables, and corrected for bias in the estimation of the outcome equation by explicitly using information gained from the model of sample selection; and (d) he developed a creative two-step procedure by using the simple algorithm of least-squares.
In this workshop, I will use examples from social work research to review Heckman's model, particularly the four contributions listed above. Two examples pertaining to social work research are used to illustrate Heckman model. One example is the application of Heckman model to evaluating a school-based intervention that aimed to increase children's social competence. The challenge in this program evaluation was selection bias triggered by a quasi-experimental design. The application of Heckman's sample selection model shows efficiency and robustness of controlling for selection bias via a two-stage process. The second example involves evaluation of the effects of caregivers' use of substance abuse services on children's psychological well-being using survey data (i.e., National Survey of Child and Adolescent Well-Being). Common pitfalls in establishing causality using observational data and strategies to overcome them are illustrated via this example.
Pedagogical Techniques: Teaching methods include lecture, PowerPoint presentation, and computer demonstration. I will demonstrate estimation of Heckman model using two Stata programs called heckman and treatreg.