Employment-based health insurance: implications of the sampling unit for policy analysis.

One of the least understood aspects of using employer data for the health reform debate concerns important differences between enterprise- and establishment-level surveys of employers. We demonstrate that the choice of sampling unit affects the size distribution of employees between large and small firms, as well as the estimated proportion of firms offering health insurance. Because health insurance decisions in multi-establishment enterprises generally are made for the entire enterprise rather than individual establishments, we conclude that enterprise surveys are most appropriate for collecting information on the factors affecting the decision to provide health insurance coverage. Nevertheless, an establishment-level survey may be preferred for evaluating decisions made at the state, regional, or industry level. But establishment-level surveys will underestimate the impact of an employer mandate on the unit that ultimately makes the decision to offer health insurance--the enterprise. Given the enormous human and financial stakes involved in health care reform, we urge policymakers and analysts to reflect on the strengths and limitations of alternative sampling units, and to consider the implications for interpreting survey data and formulating policy decisions.