In order to more carefully examine the sources of variation in second-lowest cost silver plan premiums among rating areas, we applied statistical models to obtain estimates of the association between second-lowest cost silver plan premiums for selected ages and a number of Marketplace characteristics. Our primary indicator of competition is the number of issuers in a rating area. We also examine the percent of all issuers that were defined as “established,” meaning that they issued a policy in the private individual market within the state during 2012 and 2013. Such issuers may have greater knowledge of the area or have established provider markets that allow them to charge a lower premium; or, on the other hand, they may have a loyal customer base that
is willing to accept higher premiums. We also included a variable to reflect a specific type of issuer—the consumer operated and oriented plan issuers (CO-OP). The consumer operated and oriented plan program was established to foster the creation of qualified nonprofit health insurance issuers to offer competitive health plans in the individual and small group markets. We expect the presence of a CO-OP in a rating area to have a negative association with the second-lowest cost silver plan premium.
In addition, we use a measure of hospital market concentration, the Herfindahl-Hirschmann Index (HHI),29 in our statistical models. Since more concentrated hospital markets could result in higher prices for hospital services, insurance premiums may be higher in these rating areas relative to those with less concentrated hospital markets. Since we focus on premiums within age bands, a variable was included to denote market areas in New York and Vermont, which are the only two states that do not permit setting premium rates based on age. Other market characteristics included an indicator of a Federally-facilitated Marketplace, an indicator of a Medicaid expansion state, the percent of the rating area population that is uninsured, the log of state health care expenditures, and the log of the rating area population density. We used three different model specifications in order to investigate the association between premiums and both an alternative measure of health expenditures and the exclusion of health expenditure measures from the model.30
Results indicate that the premiums are negatively correlated with the number of issuers (see Table 7). Specifically, an increase of one issuer in a rating area is associated with a decrease of approximately 4 percent in the second-lowest cost silver plan premium for a 27-year-old individual.31 These results are consistent with recent findings using a somewhat different approach that also found that greater competition reduced second-lowest cost silver plan
premiums in 2014.32 In addition, a greater percent of established issuers in a state is associated with lower premiums—approximately a 2 percent reduction in second-lowest cost silver premiums for each 10 percentage point increase in the percent of all issuers that were established issuers. However, this finding was not statistically significant for all model specifications shown in Table 7. The hospital market concentration did not have a statistically significant association with the second-lowest cost silver plan premium.
TABLE 7 Linear Regression Model Results of the Association Between Second-Lowest Cost Silver Plan Premiums, the Number of Issuers, and Other Marketplace Characteristics, by Rating Area, 2014 Health Insurance Marketplace
|Log of the SeconFord-Lowest Cost Silve a 27-Year-Old (N=r Plan Premiums 494)|
|Model 1||Model 2||Model 3|
|Market Characteristics by Rating Area||Coefficient (P-Value)||Coefficient (P-Value)||Coefficient (P-Value)|
|Number of Issuers||-0.04 (<0.001)||-0.04 (<0.001)||-0.04 (<0.001)|
|Percent of Established Issuers||-0.19 (0.03)||-0.22 (0.11)||-0.14 (0.19)|
|CO-OP (1,0)||-0.03 (0.48)||-0.05 (0.30)||-0.02 (0.63)|
|FFM State (1,0)||-0.09 (0.23)||-0.05 (0.51)||-0.08 (0.27)|
|Medicaid Expansion State (1,0)||0.00 (0.83)||0.01 (0.81)||-0.01 (0.85)|
|Full Community Rating State (1,0)||0.55 (<0.001)||0.50 (<0.001)||0.56 (<0.001)|
|Log of Hospital Market Concentration (HHI)||-0.01 (0.78)||0.01 (0.73)||-0.00 (0.98)|
|Percent Uninsured Population||0.18 (0.57)||0.13 (0.72)||0.09 (0.79)|
|Log of State-Level Health Care Expenditures||0.54 (0.001)||NA||NA|
|Log of State-Level Small Group Premiums||NA||0.56 (0.01)||NA|
|Log of Population Density||-0.01 (0.41)||-0.01 (0.63)||-0.00 (0.80)|
|Constant||1.32 (0.28)||0.94 (0.59)||5.73 (<0.001)|
|F-Statistic||60.57 (<0.001)||114.93 (<0.001)||105.66 (<0.001) 0.26|
Source: ASPE computations of plan and premium data from the following publicly available sources: Healthcare.gov, state rate filings (where available), and State-based Marketplace websites.
NOTE: Other model specifications included using the second-lowest cost silver plan premiums for 35, 40, 50, and 60-year-olds as the dependent variable, respectively. Results were consistent across different specifications.
29 The HHI refers to the Herfindahl-Hirschmann Index which is the standard measure used in economic analysis of market competition and is computed as the sum of squared market shares in the market. The HHI ranges from 0 indicating perfect competition to 10,000 indicating monopoly. The Department of Justice and the Federal Trade Commission guidelines define a market as “highly concentrated” if the HHI exceeds 2500.
30 Variations of the final model use as the dependent variable the logged values of the second-lowest cost silver premiums for ages 27, 35, 40, 50, and 60 with consistent results across these different model specifications. The model is a multivariate linear regression model utilizing the cluster option in Stata to produce robust standard errors that take into account the potential that premiums in rating areas within a state may not be independent of each other. State-level health care expenditures were estimated using Truven MarketScan Commercial Claims and Encounters Database for 2012 and the average state-level small group premiums were taken from the Agency for Healthcare Research and Quality, Center for Financing, Access and Cost Trends, 2012 Medical Expenditure Panel Survey-Insurance Component, Table II.C.1, “Less than 50 Employees.”
31 The observed association may also reflect other factors which we could not currently measure, such as the extensiveness of provider networks. Our cross-sectional analysis implies association and not causality.
32 Leemore Dafny, Jonathan Gruber, and Christopher Ody. “More Insurers Lower Premiums: Evidence from Initial Pricing on the Health Exchanges.” NBER Working Paper No. 20140. May 2014.