This paper attempts to improve our understanding of why many small private employers in the US choose not to offer health insurance to their employees. We develop a theory model, simulate its predictions, and assesses whether the model helps explain empirical patterns of firm decisions to offer insurance. Our theory model provides an explanation for why many small firms do not offer health insurance to their employees even when it may seem attractive to firms, employees and insurers to do so. Small firms have relatively large between-firm variability in expected employee health care costs, and job turnover rates for young and old employees go down differentially when firms offer health insurance. This heterogeneity and differential change in turnover rates mean that expected health costs will increase once health insurance is offered. State regulations on annual rates of premium change, or insurer reluctance to publicly increase premiums rapidly mean that coverage is only offered to small firms at high premiums, those above initial expected costs. The resulting separating equilibrium is one in which some firms face high initial premiums, choose not to offer health insurance, and tolerate higher turnover rates than if offering insurance at lower premiums were feasible. High administrative costs of offering insurance by small firms exacerbate this dynamic selection problem. We examine the predictions of this model using data from the 1997 Robert Wood Johnson Foundation’s Employer Health Insurance Survey (EHIS), which contains establishment data on employees and their offerings of health insurance. We show that turnover rates are systematically higher for in industries not offering insurance. Consistent with previous studies, the EHIS data confirm that small firms are more heterogeneous in their age distribution, income, other health-related variables than large firms. Rather than interpreting this as causing small firms to choose not to offer insurance, we see this as partial evidence in support of our theoretical model that such heterogeneity is partly the consequence of whether health insurance is offered. We then use MEDSTAT MarketScan data from 1998-99 which has individual health care costs of 890,000 adult employees and their dependents. We develop predictive models of health care spending, and simulate distributions of firm-level expected health costs repeatedly for each firm by merging the MEDSTAT and EHIS samples by age, gender, and industry code. Small firms have a great deal of heterogeneity in expected costs. Even if employees are highly risk averse, many small firms will find it unattractive to offer insurance given with high administrative costs even when large subsidies are provided. Moreover, high turnover rates make it easy for firms to quickly change the expected costs, making it difficult for insurers to commit to constant premiums when offering insurance.
