Prior studies of health insurance regulation and coverage have examined either the impact of small-group reforms on employer offer of coverage or the impact on coverage among workers or other segments of the population under age 65. Of course, all of these studies examine impacts of state reform prior to HIPAA. As a result, they are able to observe effects related to differences in some types of regulation (for example, guaranteed issue) that for groups of 2 to 50 are now uniform among the states.
All studies of the impacts of state regulation on either group or individual insurance coverage have measured various types of reform as simple dummy variables. Because few states legislated only one type of reform, use of dummy variables to measure reforms creates serious problems of multicollinearity, leading some researchers to “bundle” reforms for empirical analysis. However, studies that bundle reforms have been able to identify the effects of various reforms only in combination; they cannot parse whether these effects may be due to a particular reform or to all reforms in combination.2
Among the studies that examine employer offer, one examines specific reforms (Jensen and Morrissey, 1999); two others consider bundles of reforms (Hing and Jensen, 1999; Monheit and Schone, 1998). Each of these studies discovered effects on employer offer across all firm sizes, despite the fact that the states’ reforms typically applied only to firms of fewer than 50 lives. Also, each found patterns of results that contradicted expectations.
Jensen and Morissey concluded that, among reforms considered alone and in combination (but excluding rating reforms, due to multicollinearity), only limits on preexisting condition exclusion periods affected employers’ decisions to offer coverage. The imposition of limits on preexisting condition exclusions raised the likelihood of employer offer. Early results from work conducted by Monheit and Schone (1999) indicate that community rating (but not limits on preexisting condition exclusions) raised the probability of employer offer.3 Finally, while Hing and Jensen concurred that the probability of employer offer (aggregating firms of all sizes) was greater in reform states ( and especially in “recent” reform states rating constraints were imposed, but not guaranteed issue), they found that employee enrollment was lower in these states, suggesting that employee contributions for coverage in these states may have increased.4
A larger number of studies have addressed the impact of state insurance on the probability of insurance coverage, typically considering the probability of group coverage or any private coverage (group and individual) as well as the probability of being insured at all. Some studies also have considered the impact on public coverage. Three studies examined market reforms individually, and found different impacts in the small-group market than in the individual market. All used the same basic data (the March CPS), and while their results were generally consistent, they varied across subsets of the population.
- Sloan and Conover (1998) examined the impact of state reforms (considered individually) on the probability of coverage, estimated at the person level with CPS data. They concluded that small-group community rating laws had no impact on private coverage among general population, but raised coverage among older adults. In the individual market, regulation requiring community rating reduced the likelihood of being insured.
- Zuckerman and Rajan (1999) considered the impact state reforms both individually and in combination on state-level rates of health insurance coverage. They detected no impact of small-group reform (alone or in combination) on average rates of employer coverage. However, in the individual market, Zuckerman and Rajan concluded (consistent with Sloan and Conover) that comprehensive individual market reforms (in any combination) reduced both private coverage and total (private plus public) coverage. Individually, only guaranteed issue in the individual decreased total coverage (but, oddly, it had no significant impact on the rate of private coverage).
- Browne and Frees (2000) considered the impact of small-group rating reforms, and then any restriction on underwriting (including guaranteed issue or renewal) on coverage among single adults (aged 15-64) and then on all adults employed in small firms (of less than 10). They concluded that constraints on rating affected coverage among the segments of the working population would most likely be affected. Specifically, constraints on health rating reduced group coverage among workers who reported no work-related disability, but raised group coverage among those with work-related disability. Constraints on gender rating increased coverage among women and reduced coverage among men. Constraints on age rating increased coverage among older workers (but had no discernable impact on younger workers). Browne and Frees found no significant effects from underwriting restrictions in the individual market.
Three other studies have attempted to consider the impact of bundles of reforms. These studies use different population survey data and also different data identifying state regulation. Probably due to differences in both data and methods, they found inconsistent results.
- Using March CPS data, Marsteller et al. (1998) concluded that some small-group reforms (any combination of guaranteed issue, guaranteed renewal, limits on preexisting condition exclusions or portability) raised rates of total (private plus public) coverage, but only when all were present did they raise the rate of private coverage. The combination of guaranteed issue and rating reforms in the individual market reduced rates of total coverage but, oddly, their impact on private coverage was weak.
- Also using CPS data but a different database describing state reforms, Simon (1999a) considered three “bundles” of small-group reforms — states with full reform, partial reform, or only a waiver of mandated benefits.5 Simon concluded that full group reforms (but in general, not partial reforms) decreased employer coverage among small- firm workers (especially among demographically low-risk populations), even when age and gender were allowed rating factors.6 Constraints on rating for age and gender magnified the negative effects of other group reforms. In a later paper using a somewhat different estimation method, Kaestner and Simon (2000) concluded that even partial reforms reduced coverage among groups “vulnerable” to insurance loss (low-educated employees and young, unmarried employees without children).
- Using employer-survey data to examine the effects of state regulation on the price of group insurance to employers and employees, Simon (1999b) concluded that full reform increased premiums (by about 4 percent), and that most of the increase was passed on to workers as higher employee contributions for coverage. Presumably responding to higher contribution levels, full reform did not affect the rate of employer offer but decreased the rate of employer coverage (implicitly, employee take-up). Again, partial reform had no significant impact on either employer offer or coverage.
Appendix 1 provides a summary of the measures and methods used in each of these studies.
Differences in these studies’ models and the populations they consider appear to drive many of the inconsistencies in their findings. However, Simon (1999a and 199b) — arguably the most reliable investigations of the impacts of regulation7 — discovers the same general result using different databases and methods in each of her studies. Specifically, some reforms in combination apparently discourage employer coverage, probably because insurers respond by raising insurance prices and employers then increase employee contributions. Among the several reforms most states had enacted before HIPAA compliance, guaranteed issue in combination with rating reform may be the most likely cause of a price-contribution-coverage sequence: constraints on rating with guaranteed issue, may force broader increases in premiums (affecting more groups and individuals) than guaranteed issue alone.
Finally, these studies rarely consider individual market reforms. Those that do consider individual market reforms combine them in a general analysis of population coverage (employer coverage and individual coverage) and find a contradictory pattern of effects — for example, an impact on total (private plus public) coverage but no significant impact on private coverage. Others have noted that most of these studies suffer from specification problems which may produce both biased and inefficient estimates of effects (Nichols, 1999).