The results of our analysis suggest that state insurance regulation has some impacts on health insurance markets. However, these effects differ in size and direction in the group and individual markets when the analysis controls for other circumstances, and they often have only weak statistical significance (90 percent). Moreover, our analysis indicates (in the one specification in which we tested dummy regulatory variables) that analyses that fail to account for the strength of state reforms are more likely to detect significant regulatory impacts than analyses that do account for them. In all tests, state effects are highly significant. Typically, year effects (usually over the two-year span of the data) are also significant. These results for the group and individual markets, respectively, are reported in Tables 11 and 12, and summarized below.

A. Group Market Effects of Regulation

The following sections describe our results with respect to three areas of regulation: guaranteed issue and renewal, limits on preexisting condition exclusions, and limits on rating.

Guaranteed issue. We find no evidence that guaranteed issue in the group health insurance – required by HIPAA but implemented earlier in some states – forces insurers from markets. To the contrary, markets with guaranteed issue of all products had more insurers than states that did not, and because they were less concentrated (the largest five insurers held less market share), arguably were more competitive. These significance of even these effects is weak (90-95 percent), suggesting that if guaranteed issue of all products has any impact at all on market structure, it appears to support greater competition — perhaps by “leveling the playing field” for all insurers.

Not surprisingly, more modest reform of this type had no significant impact on most measures of market structure. Guaranteed issue of some but not all products drove a modest increase in the Herfindahl index (with weak significance), suggesting that even modest guaranteed issue reforms may force some concentration of markets via the exit of some small insurers. Similarly, guaranteed renewal forced an increase in both the Herfindahl index (and, with very weak significance, the market share held by the largest insurers in the state). To the extent that small insurers experience strong economies of scale (that is, lower average cost at a higher volume of business), greater market concentration via small insurers leaving the market or merging with others may be a positive outcome of even modest guaranteed issue reforms, if it reduces these insurers’ average cost and if lower insurer cost translates into a lower price for coverage.

Preexisting condition exclusions. Limits on waiting periods for coverage of preexisting conditions in group markets also drove greater concentration in these markets, measured either by the Herfindahl index or by the largest insurers’ market share. The highly significant increase in the Herfindahl index, in particular, suggests that preexisting condition exclusions favor somewhat larger firms over very small firms. In response, some very small firms may exit the market (although this effect was statistically insignificant) or (more likely, over the short period of our study) they may respond to shortened preexisting condition exclusions by raising price and forfeiting market share.

Rate reforms. Statutes that constrain insurers’ composite rates (or, in states without a composite rate band, limit rating both on age and on health) have the clearest impact on the structure of group health insurance markets. On the whole, states that constrain overall rate variation have more insurers than states that do not, and this effect is large: a state with a very narrow composite rate band, all else being equal, would have 30 more insurers than a state with a 2:1 composite rate band. However, also imposing a narrow rate band on health fully neutralizes this impact: the coefficient on health rating is approximately the same size but has the opposite sign (and of the same statistical significance, 95 percent). On net, a state with pure community rating had only slightly fewer insurers than a state with no constraints on rating. States with constraints on both health and age rating also had about the same number of insurers as states that impose limits only on age rating. In practice, only one state – Arizona – has imposed composite rate bands without also imposing separate rate bands on health; however, the composite rate band in Arizona is broad (4:1) and, therefore, any positive impact on the number of insurers probably was small.

Perhaps ironically, states that impose constraints only on age rating or only on health rating, but not both (and also do not limit insurers’ composite rates) may cause greater disruption in their group health insurance markets than states that limit both age and health rating and/or limit composite rates. In all specifications of our model, tighter limits on health rating alone drove a significant reductions in the number of insurers (but had no impact on any other measure of market structure). In one specification, tighter limits on age rating also had this effect, but this result appears to be nonrobust; it disappears entirely in alternative specifications of the model.

It is notable that, when we did not account for the narrowness of allowable rate variation (that is, when health and age rating were entered as a dummy variables), the impacts of age rating and health rating on the number of insurers writing coverage in the state were both more significant and negative. Considering our results, we conclude that analyses which fail to account both for the narrowness of allowable rate bands and the presence of composite or multiple rate bands are likely to bias upwards their estimates of the impact of rate regulation.

Finally, we tested each of the above forms of regulation on insurers’ average loss ratios – potentially a leading indicator of future change in market structure. However, the measurement of loss ratios, in itself, is problematic. Among insurers that wrote group health lines other than major medical, we were unable to discern their loss ratio for major medical coverage from their loss ratio on other health lines. In addition, the likely accounting differences between the medical loss ratios of commercial insurers or BCBS organizations, and those of HMOs is well documented (Robinson, 1997). Recognizing these problems of measurement, we tested the impact of regulation on arguably the most sensitive measure of insurers’ loss ratios available to us: the average (statewide) loss ratio only among commercial insurers whose major medical business comprised at least 85 percent of their total earned premiums in the group market. We found no significant impact of any regulatory variable on insurers’ loss ratios measured this way. We constructed the same type of loss ratio in the individual market, and also finding no significant result, omitted it from the individual-market specifications described below.

B. Individual Market Effects of Regulation

Guaranteed issue. Federal law does not require guaranteed issue in the individual market, as it does in the group market. As a result, only 7 states had all-product guaranteed issue in any year between 1995 and 1997, and only 5 states required guaranteed issue of some products. Neither of these provisions had a significant impact on the number of insurers writing coverage in the state.

However, guaranteed issue provisions may affect the relative market share of different types of insurers. The imposition of some-product guaranteed issue increased commercial insurer market share (with high significance), and (with lower significance) reduced the market share of both BCBS plans and HMOs. The imposition of all-product guaranteed issue raised BCBS market share and reduced commercial insurers’ market share, although these latter results are statistically weak (90 percent).

These results suggest some interaction between the dominance of one type of insurer in the individual market and the state’s adoption of guaranteed issue rules (although their simple correlation is low). Historically, many BCBS organizations have community rated, and in some states where they are dominant, they still do so. In such states, the BCBS plans have complained about adverse selection as a result of commercial insurers underwriting aggressively. In states where BCBS is dominant, they may be more likely to obtain legislative relief through all-product guaranteed issue rules (all else being equal) than in states where they are less dominant. Conversely, in states where commercial insurers have a very strong presence, they are more likely to be engaged in public policy discussions and therefore to drive reforms more acceptable to the commercial insurance industry – such as some products guaranteed issue, but not all-products guaranteed issue.

In contrast to the web of impacts associated with guaranteed issue in the individual market, guaranteed renewal – now required by HIPAA – had no impact on any measure of market structure.

Preexisting condition exclusions. Limits on waiting periods for coverage of preexisting conditions in individual insurance markets had no significant impact on any measure of market structure. This result suggests that, where the states have shortened the maximum waiting period for coverage of preexisting conditions, these reforms did not drive substantial adverse selection by encouraging individuals to drop and seek coverage as their health care needs changed.

Rate reforms. Constraints on either age rating or on health rating in the states had no significant impact on the number of insurers writing coverage in the state. However, as with guaranteed issue provisions, restrictions on health rating may have an effect on the relative market share of commercial insurers. Narrower constraints on health rating in the individual market reduced commercial insurers’ market share in favor of greater BCBS market share, and increased market concentration (measured by the Herfindahl index). This pattern of effects (specifically, the absence of a significant impact on the largest insurers’ market share) suggests that narrower constraints on health rating may have caused the smallest commercial insurers either to abandon the individual market or to merge in order to gain market share.

High risk pools. In general, one would expect that high risk pools might favor the presence of small insurers, supporting a larger number of insurers in the market, greater competition, and potentially lower insurance prices. Indeed, at least one recent study (Sloan and Conover, 1998) found evidence consistent with that expectation: a significant (and positive) effect on coverage from the presence of a high risk pool, despite the very small size of the risk pools in all but two states (California and Minnesota) and their small size relative to population in all states but one (Minnesota). However, we find no measurable impact of the presence of risk pools on any measure of market structure.