Working Paper: NBER ID: w9130
Authors: Dahlia K. Remler; Joshua Graff Zivin; Sherry A. Glied
Abstract: Estimates of the costs and consequences of many types of public policy proposals play an important role in the development and adoption of particular policy programs. Estimates of the same, or similar, policies that employ different modeling approaches can yield widely divergent results. Such divergence often undermines effective policy-making. These problems are particularly prominent for health insurance expansion programs. Concern focuses on predictions of the numbers of individuals that will be insured and the costs of the proposals. Several different simulation modeling approaches are used to predict these effects, making the predictions difficult to compare. In this paper, we do the following: (1) We categorize and describe the different approaches used; (2) we explain the conceptual and theoretical relationships between the methods; (3) we demonstrate empirically an example of the (quite restrictive) conditions under which all approaches can yield quantitatively identical predictions; and (4) we empirically demonstrate conditions under which the approaches diverge and the quantitative extent of that divergence. All modeling approaches implicitly make assumptions about functional form that impose restrictions on unobservable heterogeneity. Those assumptions can dramatically affect the quantitative predictions made.
Keywords: health insurance; modeling approaches; public policy; cost estimates
JEL Codes: I1; H0; C0
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| different modeling approaches for health insurance expansions (G52) | divergent predictions regarding the number of individuals insured (G52) |
| different modeling approaches for health insurance expansions (G52) | divergent predictions regarding associated costs (F12) |
| assumptions about functional forms and unobservable heterogeneity (C51) | divergent predictions regarding the number of individuals insured (G52) |
| assumptions about functional forms and unobservable heterogeneity (C51) | divergent predictions regarding associated costs (F12) |
| restrictive conditions (F55) | identical predictions from all modeling approaches (C59) |
| common circumstances (Z00) | significant divergence in predictions from modeling approaches (C59) |
| differences in model assumptions regarding participation, drug usage, and cost projections (H51) | discrepancies in estimates of costs for similar policy proposals (H59) |