Data impediments to empirical work on health insurance markets

We compare four datasets that researchers might use to study competition in the health insurance industry. We show that the two datasets most commonly used to estimate market concentration differ considerably from each other (both in levels and in changes over time), and reflect implausibly high volatility in market shares. By comparison, market share volatility is much lower in a private dataset gathered by a leading investment bank, and in state-level hospital discharge data. We also demonstrate that the outcome of regressions using these data vary considerably by the source used. We conclude that researchers should be cautious about using available data and recommend a new source be developed for public use.