Where Did the Moral Hazard Go? Decomposing Expenditure into Quantity, Intensity, and Coding

Extend semiparametric approaches to decompose spending changes after insurance into three channels: (i) patient quantity (visits/procedures), (ii) treatment intensity (within-visit resource use), and (iii) coding/severity inflation. Leverage provider audits, RVU weights, and DRG upcoding flags; use quasi-experimental shocks to provider incentives or audits to isolate supply-induced demand. This challenges the common assumption that utilization is the sufficient statistic for moral hazard, explicitly testing whether moral hazard manifests on non-quantity margins—intensity and coding—especially when provider behavior interacts with insurance. The project integrates demand-side disentangling literature with supply-induced demand, clarifying when null findings on utilization are offset by hidden margins. It can explain conflicting findings across contexts, such as small utilization effects but rising spending, by quantifying which margin is moving. Methodologically, the decomposition is portable to multiple datasets and settings. The impact includes refining welfare evaluations of insurance generosity and utilization management, and helping regulators target audit and coding policies without bluntly restricting access.

References:

1. Moral hazard, adverse selection, and health expenditures: A semiparametric analysis. Patrick Bajari, C. Dalton, H. Hong, Ahmed Khwaja (2014).
2. Disentangling Adverse Selection, Moral Hazard and Supply Induced Demand: An Empirical Analysis of the Demand for Healthcare Services. V. Atella, A. Holly, A. Mistretta (2016).