health plans. Lacking such comprehensive data, researchers have sometimes linked data from
multiple sources to assemble the information they need for modeling (e.g., simulating workforce
characteristics in order to understand employer offer behaviors observed in employer surveys).
Because each data source typically has a different sample design as well as differences in
variable definitions and other features, linking them is a challenging task in itself, requiring
advanced statistical techniques. Even so, the error introduced by linking data that are not
designed for this purpose may be substantial. Some expansion of current surveys, carefully
considering both the sample design and the definition of variables with a view toward linking the
data for modeling purposes, could substantially help to improve elasticity estimates.
Consideration of Available Options
Many recent studies have failed to model the options available to individuals in their
decisions to purchase health insurance or health care services, although the availability of
demand for health insurance and money wages. Even studies of plan choice often compare plans
with similar benefit designs and premiums, offering little insight about how consumers might
choose when confronted with very different benefit designs and a large difference in price.
Efforts to address such issues in data and analysis—for example, obtaining information about
eligibility and premiums for coverage from a spouse’s plan when not taken—could greatly
improve price elasticity estimates.
26 For example, Honig (2005) found that potential coverage by spouses plays a large role in the decisions of both husbands and wives regarding take-up of their own coverage.