With respect to estimating the elasticity of switching, the endogeneity of premiums is once
again a concern. Since much of the health plan choice literature is based on cross-sectional data,
allowing premium variation to be observed across plans; typically the quality of the plans
(benefit design and the provider network) is unobservable. Because it is likely that these
unmeasured plan attributes are correlated with premium, the resulting estimates of price
elasticity probably are biased. In addition, knowing what plan options actually are available to
employees is critical. Feldman et al. (1989) demonstrated that including employees who had
different (but unobserved) insurance options produced biased estimates of price elasticity.
The unobserved cost of switching plans also complicates the estimation of price elasticity.
Because consumers who switch plans may incur significant costs unrelated to price—such as
uncertainty about the quality of the new plan and the possible change of providers—some may
prefer to remain in their current plan even if the price of another plan is a lower. The failure to
observe switching costs probably biases estimates of price elasticity downward. To account for
switching costs that may be related to certain personal characteristics (such as age and health
status), some researchers have developed multivariate models that interact premium variables
with employee characteristics. These interacted variables allow for a heterogeneous response to
price related to potential differences in switching costs.
Finally, people with higher than average expected health care expenditures may enroll
disproportionately in a more generous plan option. Subsequently, adverse selection may cause
premiums to increase and enrollees with low expected expenditures to respond by switching out
of the plan. As a result, adverse selection may cause the long-run elasticity of switching in
response to price to exceed the short-run elasticity.