Information Technology and Consumer Search for Health Insurance

Abstract

We explore the impact of information technology on the level of premiums paid for individual health insurance by asking which kinds of buyers will have larger gains from the use of new technology. We compare ‘asking price’ data posted on an electronic insurance exchange with survey data on premiums actually paid before the advent of exchange and examine whether the pattern of differences between asking prices and transactions prices can be explained using a simple search theory. We hypothesize that older consumers, expecting to pay higher premiums for a given policy, had engaged in more intensive search than younger consumers, given the same distribution of prices and search costs. Therefore, the introduction of an electronic exchange that lowers the cost of search should have a larger effect on decreasing the level of premiums paid for those who previously searched less (i.e., younger consumers). We find evidence consistent with this hypothesis.

Key Words:

• Consumer Search
• Price Dispersion
• Information Technology
• Health Insurance
• JEL Classifications: L1, D83, I11

Notes

1. Indeed, the largest nationwide individual insurer, Mutual of Omaha, has announced plans to exit this market because of low profitability relative to other types of insurance it sells.

2. While their unadjusted data show a positive association between Internet searchers and shorter unemployment spells, Kuhn and Skuterud (2004) show that observable characteristics correlated to both using the internet and worker ‘quality’ (e.g., education) appear to explain this positive relationship. They also find some evidence for a negative effect of using the internet on the length of unemployment spells (upon controlling for observable characteristics) and suggest that there may be certain unobservable characteristics related to both the use of the internet and poor worker quality (e.g., unhappy or mobile workers).

3. Stigler actually said that dispersion would be lower for commodities that form a larger share of total household consumption spending, whether or not the per‐unit price was high or low. Sorensen (2000) more recently has argued that ‘search‐induced price variation will be independent of scale’ because ‘dispersion is a function of search costs, which are generally modeled as independent of prices.’ We reconcile Rothschild’s conjecture with (at least as far as insurance is concerned) Sorensen’s proposition in our discussion.

4. Phelps’ (1997) textbook notes that ‘the price of insurance is L, the “loading fee” of the insurance above expected benefits’ (p. 343).

5. Thus, in Sorensen’s (2000) terminology, although search is assumed to be independent of the loading (as a unit price), it is dependent on the premium (as a measure of total annual spending, or price times quantity).

6. One possible explanation is that, in contrast to the usual search model in which a seller can credibly quote a price only by having a costly ‘store,’ additional insurance premium postings require no additional resource costs.

7. We later provide a sensitivity analysis of the implications of using different estimates of growth in premiums over time. Individual insurance data from the CTS (across all types of plans; e.g., FFS, PPO, POS, HMO) show a similar growth rate for this period of about 52%. The small (and thus imprecise) sample of individual health insurance plans in the MEPS Household Component data (which likewise can not account for changes in the type of plans over time) shows a lower growth rate of about 26%.

8. Indicator variables for location are included in this regression to serve as control variables for the geographic variation in premiums rather than relevant measures of risk.

9. To obtain 10th decile estimates for the ‘composite’ premium sample, we used our regression model for Internet premiums on plan characteristics. On the basis of this regression, we selected premiums whose residuals were in the bottom decile of all Internet premium residuals. With this sub‐sample of Internet premiums, a regression was run on these observations to estimate a ‘10th decile’ premium for each CTS observation.

Additional information

Notes on contributors

Mark V. Pauly

This research was supported, in part, by the Robert Wood Johnson Foundation’s Changes in Health Care Financing and Organization (HCFO) Initiative.