Published on August 10, 2015. Updated October 07, 2015. Views: 17517. Downloads: 5078. Suggestions: 15.
Larger Issuers, Larger Premium Increases: Health insurance issuer competition post-ACA
Eugene Wang and Grace Gee
The Patient Protection and Affordable Care Act (ACA) has substantially reformed the health insurance industry in the United States by establishing health insurance marketplaces, also called health exchanges, to facilitate the purchase of health insurance. The ACA has increased transparency in insurance pricing and in issuer pricing behavior. Using 2014 and 2015 Unified Rate Review (URR) data, this study examines changes in health insurance premiums made by individual health insurance issuers in 34 federally facilitated and state-partnership health insurance exchanges.
Results summary: Our study shows that the largest issuer in each marketplace had a 75% higher premium increase from 2014 to 2015 compared to other same-state issuers (p=0.03, one-tailed paired t-test). On average, the largest issuers raised rates by 23.9%, while the other issuers only raised rates by 13.7%. Moreover, the largest issuers’ premium increase affects a larger proportion of plans (p=0.008, one-tailed paired t-test) and do not seem justified from the standpoint of incurred claims-to-premium ratio (p=0.31, one-tailed paired t-test for higher claims ratio) in the reported experience period of 2013. Projected Index Rate from the rate review process is used as a summary of an issuer’s premiums across different plans and Projected Member Months as a proxy for on-exchange market share. Our findings suggest that even after the Affordable Care Act, the largest on-exchange issuers may be in a better position to practice anti-competitive pricing compared to their same-state counterparts.
Health insurance has always been a highly consolidated industry. In a 2000-2003 study on HMO and PPO plans for insured and self-insured employer funding arrangements, Robinson finds that the top insurer controlled at least one-third of the market in 38 of the 48 states . Even as recently as 2013, an American Medical Association study showed that in 45 states, the top two health insurers held more than 50% of the market share. Using merger guidelines issued by the U.S. Department of Justice and Federal Trade Commission, the study also finds the insurance industry in 71% of metropolitan areas to be "highly concentrated" . As with any industry with dominant players, consumers are often susceptible to adverse pricing behavior by health insurance carriers (hereafter "issuers").
However, thanks to the Patient Protection and Affordable Care Act (ACA), the health insurance industry has undergone substantial reforms with increased transparency in issuer premium pricing behavior. Specifically, premium changes must now be reviewed through Rate Filing Justification documents (45 CFR § 154.215) and claims, utilization, taxes/fees, and other expenses must be disclosed by the issuers to justify their premium changes. Beyond the rate review process, issuers must also ensure that a minimum proportion of the collected premiums, known as the medical loss ratio (MLR), must be spent on medical expenditure. Premiums in excess of the minimum must be rebated to consumers (45 CFR § 158.240). These changes are intended to ensure fair premium pricing, even among issuers with dominant market shares. So, how is ACA doing?
The introduction of the federal and state marketplaces as well as the creation of the Consumer Operated and Oriented Plan (CO-OP) Program are intended to allow smaller health insurance issuers to compete against the major players. By 2014, 36 new issuers, previously absent from the individual insurance market, joined the exchanges. However, at the same time, many large issuers refused to join the exchanges to avoid competitive pressures on the marketplaces. National issuers such as Aetna and Humana only participated in a subset of exchanges while UnitedHealthcare chose to sell only off-exchange plans. Using different measures of market competitiveness, including the Herfindahl–Hirschman Index (HHI), a Kaiser Family Foundation study finds that 7 state-based exchanges showed a variety of changes in market competitiveness. While California and New York appear more competitive, Connecticut and Washington are less competitive than their individual markets in 2012 .
Are exchanges keeping anticompetitive pricing behavior of dominant issuers in check?
Since the ACA was only enacted in March 2010 and the first on-exchange plans were only made available in 2014, research on the pricing behavior of marketplace issuers is relatively scarce. One of the rare pertinent studies is by the Department of Health and Human Services. Using the number of issuers as a measure of competition, the study finds that in each of the 36 federally facilitated and state-partnership exchanges, an increase of one issuer is associated with a 4% premium decrease for the second-lowest cost silver plan. This is based on a regression model built upon 2014 premiums data  However, it is unclear if the absolute number of issuers is a good measure of competitiveness in the insurance market. After all, the ACA CO-OP Program brought about the creation of many smaller issuers, but many faltered at meeting their targeted market share. A 2015 study by the Government Accountability Office finds that 14 out of 22 participating CO-OPs failed to reach their enrollment target . As long as other issuers cannot threaten the market share of incumbents, the number of issuers in a state is immaterial in determining market competitiveness.
Dafny et al. conducted a similar study, except they used rating areas as their units of study instead of states. Rating areas are state-demarcated regions where a plan’s premiums are only allowed to vary according to family structure, age and tobacco use. Using HHI as a measure of competitiveness, Dafny et al. finds that the 2014 on-exchange marketplaces are less competitive than the insurance marketplace in 2011 because a major carrier, UnitedHealthcare, decided not to take part in the 2014 federal marketplace . This decrease in competitiveness, according to a regression model, raised the second-lowest-cost silver plan premiums in each rating area by 5.4% on average. One of the caveats of the study is that the market share of issuers in 2014 was estimated from their 2011 market share. This can be problematic because many issuers have undergone mergers and new issuers have sprung up, thanks to the ACA. Additionally, it may not be fair to compare premiums between plans before and after the ACA given that prior to 2014, plan benefits and designs were not uniform across the board. Starting in 2014, all plans must contain Essential Health Benefits (EHB) (45 CFR § Part 156, Subpart B) and plans on the marketplace must meet minimum Qualifying Health Plan (QHP) standards (45 CFR § Part 156, Subpart C). Most importantly, prior to the ACA, plans could exclude coverage for pre-existing conditions, resulting in lower prices for 2011 plans. Therefore the lower premiums may not be directly related to the more competitive 2011 market that included UnitedHealthcare’s participation.
To address the limitations of previous studies and add to the growing literature on issuer competition in the post-ACA regime, we compared the pricing behavior of the biggest individual health insurance issuers of each state under federally facilitated and state-partnership marketplaces with that of other issuers from the same state. Unlike previous work, this analysis examines issuers as the unit of study. Instead of making interstate or inter-rating area comparisons, we compared premium changes from 2014 to 2015 among same-state issuers. This ensures that both interstate and intrastate differences are accounted for.
This paper also introduces novelties in premium and market share determination in an attempt to circumvent some of the caveats noted in previous work. For instance, instead of choosing the premium of a single plan (e.g. second lowest silver plan), we utilized the Projected Index Rate used by issuers in the new premium setting process to summarize the plan premiums of an issuer in a single number for year-over-year comparison. Also, by comparing only changes in the Projected Index Rate, we minimize the effects of plan-specific confounding variables such as differences in healthcare delivery costs and network size when making inter-issuer comparisons.
For market share determination, this paper uses Projected Member Months in the rate setting process as a proxy for market share. This is advantageous as the data is more recent and is a more accurate approximation of on-exchange market share. Unlike other studies that measure number of issuers as a proxy for competitiveness [4, 7], our measure takes into account the relative enrollment shares of same-state issuers in determining market sizes.
This research is especially relevant in light of recent market consolidation attempts by large issuers. While these attempted takeovers have sparked antitrust concerns, proponents of consolidation argue that a dominant issuer can better negotiate medical costs among network providers, passing the cost reductions to consumers . Research on pricing behavior of dominant issuers under the ACA is imperative to evaluate the merits of this argument.
The study units in this paper are issuers in different states. For the purpose of this study, an issuer whose coverage area spans multiple states is treated as a distinct entity for each state examined. This is consistent with 45 CFR § 156.80, under which all the individual health insurance plans issued by an issuer in a state have to be considered as a single risk pool.
Only issuers in 34 states with federally facilitated and state-partnership marketplaces are considered since their premiums data is easily available through public Center for Medicare and Medicaid Services (CMS) files. Other states running a state-based marketplace or federally supported state-based marketplace are omitted due to a general lack of data accessibility or consistency.
This study uses two datasets:
The CMS releases a QHP Medical Landscape file each year showing plans by county with cost-sharing scenarios and premium data .
Before each open enrollment period, issuers must submit Rate Filing Justification documents for all their plans products stating the new premiums for the upcoming plan year. The documents are composed of several parts, I-III. We will use Part I, which is the Unified Rate Review and must include historical and projected claims data, utilization trend projections and other data (45 CFR § 154.215 (b)) that are posted publicly on the CMS website (45 CFR § 154.215 (h)).
Projected Index Rate as a summary measure for an issuer’s premiums across different plans
Before quantifying changes in an issuer’s premium, it is important to formulate a measure that summarizes an issuer’s premiums across all its plans. This is especially challenging as premiums within the same issuer show wide variations with different plan cost-sharing designs, geographical areas, as well as individual-calibrated factors like age and tobacco use. To complicate things, an issuer may terminate some plans and introduce others during a new plan year.
Most methods in the literature only attempt to measure premium changes for a state marketplace or rating areas, aggregated across all issuers. Dafney et al., for instance, used the second lowest-cost silver plan to represent the premium of a rating area . Another study by the Kaiser Family Foundation used the second lowest-cost silver plan and lowest-cost bronze plan aggregated across issuers as a measure of premiums for each state .
The closest study that formulates a measure to summarize premiums for an issuer is by the Government Accountability Office (GAO). In "Private Health Insurance", GAO uses a simple average for premiums across an issuer’s plans . However, this measure assumes that different plans and metal tiers are equally enrolled. This is problematic as enrollment is heavily skewed with 69% selecting a silver marketplace plan in 2015 .
Another plausible measure is a weighted average of all plan premiums for an issuer using enrollment data specific to the plans. However, this is generally not attempted because plan-level enrollment data is not required to be published. The closest available dataset on enrollment, from the CMS Newsroom, is a state-level snapshot of enrollment with all the issuers combined . Even with the plan enrollment data, we would still require data on the average premium each enrollee pays, since these premiums can vary with age, rating areas (geography) and tobacco usage.
To find a satisfactory summary measure for an issuer’s premiums, we turn to the premiums setting process for issuers under 45 CFR § 156.80. The figure below shows the process of premium determination released by the CMS .
For each issuer there is only one Experience Period Index Rate, Projected Index Rate and Market Adjusted Index Rate. Since the Plan Adjusted Index Rate and Consumer Adjusted Premium Rates are specific to plan and consumer respectively, each issuer has many of them. As such, Plan Adjusted Index Rate and Consumer Adjusted Premium Rates cannot be used as a summary for premiums across an issuer’s plans.
Market Adjusted Index Rate would be ideal as a benchmark for comparison, since it is the rate that saw the most adjustments and is closest to the final premiums shown to consumers. However, as explicitly stated in "2016 Unified Rate Review Instructions", the Market Adjusted Index Rate is not available in the URR PUF . Theoretically, it can be calculated by dividing Plan Adjusted Index Rate by the Actuarial Value Pricing Value. However, for 2014 data, such a calculation does not provide a consistent number for each issuer within a state.
We chose the next best rate, the Projected Index Rate, as the ideal measure of an issuer’s premiums. The rate is available for both 2014 and 2015 URR PUFs as "Index Rate for Projection Period", allowing year over year changes to be calculated. To determine if the Projected Index Rate is a good representation of an Issuer’s premiums over all its plans, we carried out the following multilinear regression:
Premiums of all plans in different rating areas across US (95,546 data points) ~ Projected Index Rate of Issuer + Metal Tier + Plan Network Type (e.g. PPO, HMO) + State
The target variable is the premium that consumers eventually see in their own rating area. We chose the premium of a 21-year-old individual for the target variable as the premiums across different ages are now fixed by a federal or state-specific age curve (PHS Act section 2701(a)(1)(A)(iii)), and all premiums for ages 21 and above can be calculated by multiplying the premium of a 21-year-old by a fixed age-dependent factor. From Table 1, the regression model suggests that a simple combination of the Projected Index Rate, two other plan-specific parameters, and the state the plan is offered in can account for 76.5% of all variation in 95,546 premiums across the 34 exchanges in the 2015 Medical Landscape File. This is remarkable since rating areas are not being taken into account. The coefficient of Projected Index Rate, highlighted, is statistically significant at p-value < 0.00001. All these together with the theoretical underpinnings suggest that the Projected Index Rate is a good representation of all the premiums of an issuer across its plans.
Hereafter, the term "premium change" will refer to the year-over-year change in Projected Index Rate for an issuer.
Table 1. Regression results using Projected Index Rate, metal tier, network type and state to predict premiums
Projected Member Months as a measure of 2015 on-exchange market share
Quantifying an issuer’s market share would be easy if marketplace enrollment data for each issuer were available. Since the closest data set published by the CMS is aggregated across all issuers in each state, an innovative measure of market share must be used .
The first decision we made is that market share should be a reflection of the number of enrollees for each issuer, rather than total premiums collected. Since this paper is interested in the effect of market share on premiums, it follows that our measure of market share should not be commingled with premiums.
The second decision was to measure the issuers’ market share only on the federally facilitated or state-partnership marketplaces. Since our measure of premiums for each issuer is based entirely on their on-exchange products, the market share measure should also be the on-exchange market share to be consistent.
The last requirement for market share measure is that it has to be current as of 2015. Our measure of premium changes correspond to changes from 2014 to 2015 (the only two years with federal marketplaces in existence). During this period, many issuers merged, and many new issuers appeared as a result of various government programs promoting competition, such as the CO-OP programs. Any market share datasets before 2014, such as the authoritative National Association of Insurance Commissioners  and Kaiser Foundation data , are obsolete because their data does not correspond to issuers’ market share on federal marketplaces.
Given these requirements, we turned to the 2015 URR PUF for the most recent enrollment numbers.
The first possible choice of a market share measure is the "Experience Period Member Months", which corresponds to the number of enrollees for an issuer in a state in 2013 multiplied by the number of months each enrollee is insured. Unfortunately, many issuers omitted data in this field by citing reasons such as unrepresentative data. The experience period data also do not correspond to the issuers’ first year experience on the federal exchanges.
As a compromise, our chosen measure of market share is the "Projected Member Months" in the 2015 URR. To confirm the accuracy of our market share measure, we compared the data with the most recently published Kaiser Foundation database of issuers’ market share in 2013 . Table 2 shows each state’s top issuer back in 2013 and their 2015 calculated state market share using Projected Member Months in all 34 states with federally facilitated or state-partnership marketplaces.
Table 2. Top issuer by state according to Kaiser  and their market share (%) calculated with Projected Member Months. Highlighted entries are states where the issuer is no longer the largest in the 2015 marketplace.
For 26 out of the 34 states, the largest issuer is consistent between the 2013 Kaiser database and our calculated measure. There may be some discrepancy in the market share percentage since our calculated market share only corresponds to market share in the federal marketplace, whereas the Kaiser market share was calculated from the life-years of those enrolled in major medical and mini-med plans. There are 8 states (highlighted in Table 2) where the largest issuers do not line up. In these cases the states’ largest issuers in 2013 chose not to participate in the federal exchange. For example, in South Dakota and Iowa, Wellmark Inc. decided not to participate in the 2014 and 2015 enrollment cycle [16, 17].
Table 3. Top 3 issuers by state and their market share (%) calculated using Member Months projected for 2015. Authors’ calculation from 2015 URR data set.
Using Projected Member Months as a measure of market share is advantageous as the data comes from the same information source used in premium change calculation and is much more fine-grained and recent than data from other commercial databases.
We calculated the market share of each issuer within each state by simply dividing Projected Member Months by the total Projected Member Months for all issuers with the same state. Hereafter, the term "market share of an issuer" will refer to this calculation.
Data Wrangling Steps (for reproducibility)
The following steps were taken to preprocess and analyze the 2014 and 2015 URR and the 2015 QHP Medical Landscape file:
Figure 1. Number of issuers in each state
Figure 2. Frequency distribution of issuer market share
There is premium change data for 141 issuers (down from 228 issuers for market share data) across 34 states. This means that 87 issuers recently joined the exchanges, so that their pricing data is absent from the 2014 URR. From Figure 3, the average premium change from 2014 to 2015 (not market share weighted) is an increase of 17.7%, and the distribution of premium change is symmetrically distributed around the mean. Figure 4 shows a scatterplot of the premium change against the state market share ranking of each issuer. Almost all of the largest issuers in each state (those with a ranking of 1) raised rates, whereas a number of smaller issuers (ranking above 10) decreased rates.
Figure 3. Histogram showing distribution of premium changes from 2014 to 2015 across all states
Figure 4. Premium changes against state market share ranking for each issuer. Rank 1 is the largest issuer in the corresponding state.
The largest issuer in each state raised premiums on average 75% higher than other same-state issuers. On average, the largest issuers raised rates by 23.94%, while the other issuers only raised rates by 13.68%. This result is statistically significant with a one-tailed paired t-test.
To analyze the impact of market share on premium increases, the premium change of the largest issuer (ranking 1) in each state is compared against the average premium change (weighted by market share) of the rest of the issuers (ranking 2 and above) in the same state. It is essential to compare only issuers within the same state to ensure that state-dependent factors, such as health care cost variations, will not interfere with the analysis. In statistical parlance, we conducted a one-tailed paired t-test with null hypothesis defined below that states the mean difference between the premium change from the largest issuers in each state and from the smaller issuers in each state is less than 0
Where represents the premium change of the largest issuer and represents the average premium change of all other issuers in the same state weighted by market share. For states where the largest issuer in 2015 was not present on the exchange in 2014, the data for the state is omitted. There are 10 such states, so that the final test statistic was calculated from 24 states. Figure 5 shows a distribution of
with mean 10.3 percentage points, suggesting that on average, the rate increase by the largest issuer in each state was 10.3 percentage points higher than that of other issuers in the same state. Since the t-statistic is 1.9704 and the corresponding p-value is 0.03047, we reject the H0 as not true and conclude at 5% significance level that the largest issuers in each state raised premiums more than or as much as other issuers in the same state. The assumptions of t-test are largely valid, since the distribution is approximately normal (the Shapiro-Wilk test fails to reject the null hypothesis of normality) and the premium changes made by issuers are largely independent (most issuers submit their rates around the same time in the last week before the deadline). Figure 6 shows the average premium increase for the largest issuers within each state across the country and that for the other issuers. On average, the largest issuers raised rates by 23.94%, while the other issuers only raised rates by 13.68%. This corresponds to a 75% higher premium increase by the largest issuers compared to the other issuers.
Figure 5. Histogram of the difference in premium changes between a state’s largest issuer and other issuers within the same state. On average, the largest issuer’s premium increases were 10.3 percentage points greater than other same-state issuers.
Figure 6. Average premium increase for the largest issuers in their corresponding states across the US vs. the premium increase for the other issuers. On average, the largest issuers raised rates by 23.94%, while the other issuers only raised rates by 13.68%.
The largest issuer of each state raised premiums on a larger proportion of plans compared to other same-state issuers. This result is also statistically significant with a one-tailed paired t-test.
To extend the analysis further, we used the same statistical technique to investigate the distribution of premium increases across plans for the largest issuers and the other issuers. This analysis addresses a common question: whether the premium increases of large issuers are concentrated within a few plans. After all, if only a few plans of the largest issuers experience marked rate increases, the premium increase will not affect as many consumers because they can switch to other cheaper plans provided by the same issuer. To carry out the analysis, the premium change for each plan is calculated using
In the 2015 URR file, the Plan Adjusted Index Rate is found under column DD named "Plan Section 4 - Plan Adjusted Index Rate", while in the 2014 URR file, the rate is also found under column DD but is named "Plan Section 4 - Average Rate PMPM". The proportion of plans that have a positive premium change (rate increase) is calculated for each issuer. Denoting this as Prop+, the following one-tailed paired t-test is carried out with null hypothesis below that states that the mean proportion of plans that increased their premiums from the largest issuers in each state minus the proportion of plans that increased their premiums from the smaller issuers in each state is less than 0.
Again, the proportions of plans with increased premiums for the other issuers were averaged using their market shares as weights. Figure 7 shows the distribution of the difference in proportions for the largest issuer and the other issuers in the same state. The test statistic is 18.9 percentage points, which means that on average, the largest issuer raised rates on a much larger proportion of its plans than other issuers. The t-statistic is 2.584, and the p-value for the one-tailed test is 0.008295. We reject the H0 and conclude at 5% significance level that the largest issuers in each state not only raise premiums higher, but also raise premiums on more of their plans than other issuers in the same state (here equality is also rejected since the corresponding p value for the two-tailed test is 0.01659 < 0.05).
Figure 7. Histogram of the difference between the proportion of plans with rate increases for the largest issuer and the other issuers within the same state. On average, the proportion of plans that the largest issuer raises rates on is 18.9 percentage points higher than other issuers.
The largest issuers’ incurred claims-to-premium ratio is not found to be statistically significantly higher than other same-state issuers.
But can the largest issuers justify their higher premium increases in 2015 by claiming that they had higher incurred claims relative to collected premiums in the experience period, compared to other issuers? To test this hypothesis, the following ratio, R, is calculated using information from the 2015 URR file:
Here the experience period refers to 2013. By the time of filing between May and June 2014, issuers only have a few months of claims data for 2014 and they do not qualify as experience period data, since the CMS requires one whole calendar year of data to be present. Note that this is very similar to the Medical Loss Ratio (MLR) as formulated in 45 CFR §158.221, which is used to calculate the MLR rebates issuers have to give to consumers if the MLR falls below 0.80 (i.e. they spent too little on medical expenses) in the experience period. But the MLR calculation would require the denominator to include only premiums collected in the experience period net of taxes, licensing and regulatory fees, which data is unavailable in the URR. To test if the largest issuers in each state have higher incurred claims relative to premiums in the experience period, the following null hypothesis was tested using a paired one-tailed t-test:
Again was calculated as a market share weighted average. After omitting states where data on the largest issuer was not available (the largest issuers having only joined the exchange in 2015) and omitting issuers without experience data (issuers can choose not to input experience period claims and premiums data if they deem that the data they have collected is not substantially representative), the test statistic is eventually calculated from 20 data points. Figure 8 shows the distribution
with a mean of 0.02, which means that the largest issuer has a higher claims ratio of only 0.02. Figure 9 shows that the claims-to-premium ratio for the largest issuers is 0.85 while that for the other issuers is 0.83. Since the test statistic is 0.50139 with a one-tailed p value of 0.31092, we conclude at 5% significance level that there is insufficient evidence to reject the null hypothesis in states where this ratio could be calculated. The claims-to-premium ratio of the largest issuer is not statistically significantly higher than other same-state issuers. Even if it were significant, a 2.5% higher claims-to-premium ratio is unlikely to be sufficient to justify the 75% higher premiums increase that the largest issuer incurred.
Figure 8. Histogram of the difference between the claims-to-premium ratio for the largest issuer and other issuers from the same state.
Figure 9. Ratio of incurred claims to premium for the largest issuers in their corresponding states vs. other issuers. On average, the largest issuers have a claims-to-premium ratio of 0.85 while the other issuers have a ratio of 0.83.
Relative to other same-state issuers, this study finds that the average largest issuer of each state had a 75% higher premium increase from 2014 to 2015 and that their rate increases affects a larger proportion of plans. Yet the largest issuers’ higher premium raises do not seem justified from the standpoint of incurred claims-to-premium ratio.
One possible caveat in the study is the lack of actual 2014 claims data in the calculation of incurred claims-to-premium ratio. Because CMS requires the experience period to be a whole calendar year starting from January 1, the experience period data used for pricing 2015 plans corresponds to calendar year 2013, which is prior to the first year issuers offered ACA plans . As such, it may be possible that larger issuers experience much higher medical loss ratio for the first few months of 2014 and have to adjust their premiums higher.
Besides, there may be other possible justifications for the higher premium increase that are not covered within this study’s scope. Perhaps the largest issuers in each state expanded their network faster than other issuers, so that premium increases reflect more accessible or higher quality medical care. To validate this claim, data on the size and quality of plan networks would be required. Future research can also examine the change in enrollee demography for the largest issuer. The higher premium increases may be due to older enrollees for the largest issuer compared to other same-state issuers. Older people may tend to be more conservative in selecting their insurance and choose a larger, more recognized issuer, whereas young, healthy people may be more receptive to smaller, new issuers. These older enrollees may have higher variance in their medical utilization costs and issuers have to raise premiums to compensate for additional risks.
Beyond the results, this paper also contributes two novel measures of premiums and market share. The first is using the Projected Index Rate as a representation of an issuer’s premium across all its plans. The second is using the Projected Member Months in determining an issuer’s on-exchange market share. Both numbers are available through the URR PUF and are updated for each plan year. There may be some caveats related to these measures. For instance, the Projected Index Rate measures anticipated allowed claims for EHBs (PMPM) and can increase from year to year if a worsening of enrollee demography is anticipated. The most objective measure of premium should correct for enrollee demography, and this should fall under the purview of future research once plan-specific enrollee demographic data is available.
Overall, this study is important in light of recent market consolidation efforts in the health insurance industry. This paper provides evidence that even after ACA, the largest on-exchange issuers may be in a better position to practice anti-competitive pricing compared to their same-state counterparts. This evidence should be prudently considered in any antitrust debate.
Updated to version 2 on 10/8/2015
The revised version of this paper includes a correction for the issuer's experience period used to calculate the issuers’ incurred claims-to-premiums ratio. The original version stated that the experience period corresponds to 2014. While a small minority of issuers provided 2014 data for the experience period, a majority of them reported data from 2013. Parts of the abstract, results and conclusion sections have been updated to clarify this.
See Version 1 published on August 11, 2014
Eugene Wang graduated top 24 from Harvard (Junior 24 Phi Beta Kappa) with a B.A. in Applied Mathematics. In 2013, he was mentioned by the Federal Reserve for winning the collegiate Federal Reserve Challenge.
Grace Gee graduated Magna Cum Laude with a B.A. in Applied Mathematics from Harvard. She is a 20 Under 20 Thiel Fellow.
After graduation, the duo founded HoneyInsured.com, a health insurance startup that helps individuals find their ideal health plan through data and visualizations. HoneyInsured is also an authorized partner of Healthcare.gov.
Wang E, Gee G. Larger Issuers, Larger Premium Increases: Health insurance issuer competition post-ACA. Technology Science. 2015081104. August 10, 2015. Version 2. https://techscience.org/a/2015081104/
Wang E, Gee G. Replication Data for: Larger Issuers, Larger Premium Increases: Health insurance issuer competition post-ACA. Harvard Dataverse. August 5, 2015. http://dx.doi.org/10.7910/DVN/GFTHAN
Enter your recommendation for follow-up or ongoing work in the box at the end of the page. Feel free to provide ideas for next steps, follow-on research, or other research inspired by this paper. Perhaps someone will read your comment, do the described work, and publish a paper about it. What do you recommend as a next research step?
Suggestion #1 | August 23, 2015
It is easy to confuse experience periods, especially in 2013 and 2014 as it related to issuers' ACA business or the issuers' ACA experience. I would like to see a study that includes 2015 data. RESPONSE FROM THE AUTHORS: Thanks for the suggestion. We have updated the study to reflect the correct experience period year. See further response to Suggestion 7.
Suggestion #2 | August 23, 2015
I am not sure the loss ratio comparison is the best way to analyze the issuers' ACA business. I would like to see a study that used other mechanisms.
Suggestion #3 | August 24, 2015
Is the experience period referred to in the paper 2013, and not 2014? This is related to Suggestion 4. RESPONSE FROM THE AUTHORS: Thanks for the suggestion. Because CMS requires the experience period to be a whole calendar year starting from January 1, the experience period data used for pricing 2015 plans corresponds to calendar year 2013 for the majority of issuers. We have updated the study to reflect the correct experience period year.
Suggestion #4 | September 02, 2015
I have worked for the federal govt for 35 1/2 yrs and I am only a GS-7, making just under 50K a year. I just found out Fri that my Healthnet HO premiums are going up 43 percent in Jan and my co-pays are going up to. So, I now pay $151 a month just for me and they want $214 from me so I am cancelling in open season. Am I suppose to move to the exchanges and no longer look at my employer options? Do health insurance exchanges now offer better plans than employers? I would like to see a study about comparing options across these two systems or a study of the impact of ACA on employer options.
Suggestion #5 | September 07, 2015
All healthcare is local and I'm sure some smaller health plans in each local area could broaden your research.
Suggestion #6 | September 28, 2015
Go back and look at the URRT file you used to pull the 2015 data. The experience period is shown in the file. In almost all cases it's calendar year 2013. [In response to this suggestion, the authors did revisit the data files used and updated the paper accordingly.]
Suggestion #7 | August 11, 2015
I would like to see someone redo the study for later years to see if you get similar results, especially in light of mergers happening in the insurance industry. [Editors]
Suggestion #8 | August 18, 2015
Please analyze ex-premium data, i.e., deductibles, co-pays, co-insurance to the extent that such data is available in the datasets you have used or that are otherwise available. The purpose would be twofold (1) to illuminate the actual cost to ACA consumers (premiums are a poor proxy of actual costs), (2) the impact of the ACA on increases of ex-premium costs to insureds as a whole and also (3) the effect of insurer consolidation on Year 1 ex-premium actual costs and then year-over-year increases.
Suggestion #9 | August 20, 2015
I wonder whether dominant issuers got that way by underpricing in 2014, and they increased premiums in 2015 to make up for that under-pricing. I would like to see a study that would investigate this possibility. RESPONSE FROM AUTHORS: We ran a one-tail hypothesis test with null hypothesis that initial premiums of the largest state issuers in 2014 are larger than others and failed to reject the null at 95% confidence level. Because we cannot conclude that the premiums of the largest issuers are initially underpriced, it is unlikely the reason for the subsequent larger premium hikes from 2014 to 2015 found in this paper.
Suggestion #10 | August 21, 2015
This is an important study and an important message for policy-makers. I would like to see a study that includes the complete market, not just those reporting in the national insurance databases.
Suggestion #11 | August 24, 2015
I am confused. Look at Iowa. Did CHC (an insurer that we Iowans haven't heard of) have 99% of the market? Iowa's dominant insurer BY FAR is Wellmark Blue Cross. RESPONSE FROM THE AUTHORS: The paper is factually correct with Iowa. The marketshare we are measuring corresponds to the marketshare of the insurance carriers on the health insurance marketplace as mentioned and rationalized in the paper... The second decision was to measure the issuers’ market share only on the federally facilitated or state-partnership marketplaces. Since our measure of premiums for each issuer is based entirely on their on-exchange products, the market share measure should also be the on-exchange market share to be consistent... Wellmark Inc., while the largest in Iowa, chose not to join the exchanges in 2014 and 2015. We specifically mentioned the reason why Wellmark Inc. was missing in the paper in reference to Table 2...the states' largest issuers in 2013 chose not to participate in the federal exchange. For example, in South Dakota and Iowa, Wellmark Inc. decided not to participate in the 2014 and 2015 enrollment cycle [16, 17]...Our external reference  further corroborates that Iowans only have the choice of CHC if they wish to purchase plans on the Health Insurance Marketplaces. As shown in Table 2 of the paper, the case where the largest state health insurance carrier is missing from the exchanges only affects 3 out of 34 states in the study, out of which two involves Wellmark Inc. Despite this, for the vast majority of states, the on-exchange marketshare corresponds closely to the marketshare of the issuers before ACA when we compared them with a 2013 Kaiser Foundation database. Perhaps another approach we could have taken is to just omit the three states, IA, MS and SD, where the largest state issuer did not join the health insurance exchanges. Out of curiosity, we ran the same pair t-test and there is almost no change to our conclusion. Without the 3 states, the largest state issuers raised rates by 24.4% while the others raised rates by 14.2%. The result is significant once again with one-tail p-value of 0.033.
Suggestion #12 | August 26, 2015
Nice work - I was an insurance commissioner in a state (state based exchange) and made the calls on rates when aca was implemented. Your focus on the index rate is ABSOLUTELY RIGHT and should be followed by others. See no methodological problems with market share calculation. One conceptual issue is defining the market for individual insurance. In some cases dominant insurer chose not to play in exchange because it determined it did not have to (wellmark in IA, apparently). In others, it could participate in the exchange market half- heartedly, with limited product offerings or less aggressive pricing - it takes vigorous regulatory oversight, even with ACA safeguards, to ensure comparability between between on and off exchange strategies. This leads to your findings. You are capturing pricing behavior and documenting that the largest insurers price more cautiously I am not sure that is surprising - 2015 rates were due with at most three months' of 2014 claims experience. Nobody had much information to go on and larger insurers had more to lose if they missed the mark. I look forward to further studies that draw lessons for market consolidation using future years of data.
Suggestion #13 | November 19, 2015
I found the paper to be really interesting and well written. I have a question, though. Are you guys doing a paired t-test with data that are percentages? If that is the case, it may not be theoretically correct. You should use a McNemar test. I doubt the results would change significantly though.
Suggestion #14 | November 08, 2016
Greetings! Very useful advice within this article! It is the little changes that make the most important changes. Many thanks for sharing!
Suggestion #15 | November 08, 2016
Thanks for this article. I might also like to say that it can end up being hard if you find yourself in school and merely starting out to create a long credit rating. There are many learners who are only trying to endure and have long or good credit history are often a difficult issue to have.