The demand for health insurance coverage by low-income workers: can reduced premiums achieve full coverage?

Health Services Research, Oct, 1997 by Michael Chernew, Kevin Frick, Catherine G. McLaughlin

One serious problem that plagues all studies on this topic is the endogeneity that arises because individuals have some control over their place of employment and hence have some control over the employee contribution that they face. The endogeneity is exacerbated by insurer requirements that a minimum percentage of workers participate in any employer-sponsored plan. Our sample consists of workers who have chosen to work for firms that sponsor health insurance plans for their employees. These individuals are likely to have a higher demand for coverage than other workers. As a result, we believe that our estimates tend to overestimate the coverage rates that would be associated with any given premium. Upper-bound estimates such as ours are important because they illustrate that, even under a generous set of assumptions, universal coverage will be difficult if not impossible to achieve with purely voluntary approaches.

RESULTS

DEMAND FOR INSURANCE

The point estimates suggest that, within the range of observed premiums, the probability of participation is higher in firms that have relatively high total premiums (see Table 2). These findings may indicate that the premium is measuring the quality of the plan, with a higher total premium representing a higher-quality plan.

As Feldman et al. (1989) suggest, the relationship between the premium and the quality of plan may influence the parameter estimates, leading to underestimates of demand elasticities. Therefore, we re-estimated the model including measures of the coinsurance rate and deductible as explanatory variables.(12) Inclusion of these variables does not change the findings; participation rates are generally unresponsive or weakly positively correlated with insurance premiums.

When we change the measure of price to capture only the employee contribution, we do detect a responsiveness to price (Table 2).(13) The likelihood of participation is negatively related to the employee contribution and is statistically significant.(14) The quadratic relationship between price and participation probability suggests that the effect of price is greater at lower employee contributions and then declines (i.e., moving from a "free" plan to requiring a $10 per month contribution has a greater effect at depressing participation rates than moving from a $150 to a $160 contribution). The participation of low-income individuals is discouraged by even low prices. Marginal reductions in substantial prices have relatively small effects.(15)

The magnitude of the estimates suggests that if insurance were perceived by all workers in our sample to be free, participation rates for single workers would be 92.9 percent. This compares to 81.7 percent if the premium costs were perceived to be 50 percent of the area average premium.(16) The arc elasticity, computed between a premium perceived to be 25 percent of the area average and a premium perceived to be free, is -.033. The elasticity rises to -.095 if it is computed between 25 and 50 percent of the area average premium. The relatively low elasticities reflect, in part, the high base probabilities of participation.