Duke 2008 | ASHE | Equity and Efficiency in Health and Healthcare

Most empirical research in economics is conducted with the goal of providing scientific evidence that will serve to inform current and future policy. Such policy analytic studies typically use nonexperimental (survey) data and focus on a particular variable (the policy variable, x_p) that is at present, or will in the future be, under the control of a policy decision-making entity. The goal though often not clearly articulated, is the estimation of the effect that a prospective exogenous change in x_p would have on a targeted outcome of interest (y). This goal is often difficult to achieve using observational data because the policy variable and the outcome are endogenously sampled.

In this article, we offer a rigorous but practical modeling framework for the analysis of such marginal (continuous x_p) and incremental (x_p binary or discrete) policy effects in the presence of endogenous sampling. When x_p is continuous, its marginal effect is measured in terms of the partial p derivative of the conditional mean of y given the relevant controls. If x_p is binary or discrete, its incremental effect measures the change in the conditional mean of y that would be caused by a change in x_p from one exogenously specified policy relevant value of x_p to another. A consistent estimator for such marginal and incremental effects is proposed. The correct formulation of the asymptotic standard error of the estimator is derived. The modeling framework and estimator are applied to the analysis of the effect of substance abuse on the probability of full-time employment.