Do Premiums Increase After Health Insurance Mergers? A Reassessment of Guardado et al. s Findings

Transcription

1 Do Premiums Increase After Health Insurance Mergers? A Reassessment of Guardado et al. s Findings Robert C. Bourke, Mark Israel, Ben Wagner, and David A. Weiskopf 1 January 15, 2017 Abstract Guardado, Emmons, and Kane (2013) found that premiums in Nevada increased by 13.7 percent following the merger of United Healthcare and Sierra Health Services in Using methods similar to those used by Guardado et al., we study a series of more recent mergers to see if their results generalize, particularly given recent changes in the health insurance industry. We find that the results of their analysis do not generalize, with no pattern of higher premiums following recent mergers and, in fact, lower premiums following the mergers with the highest combined shares, a result consistent with lower medical costs or other efficiencies. 1 Weiskopf: Compass Lexecon and Johns Hopkins University; Bourke, Israel, and Wagner: Compass Lexecon. The opinions expressed are those of the authors, and do not necessarily represent the views of any institution. We would like to acknowledge Chris Rybak for his excellent research assistance. 1

2 I. INTRODUCTION Economists have conducted retrospective studies to evaluate the effect of health insurance mergers on premiums. One such recent study is the work of Guardado, Emmons, and Kane (2013) (hereinafter, Guardado ). 2 These retrospective studies have been have been relied on by testifying experts in the competitive analysis of health insurance mergers. 3 In particular, despite the different conditions that exist today as a result of the Affordable Care Act ( ACA ) and other marketplace changes (and despite the fact that not all mergers are likely to have the same effect on premiums), retrospectives from mergers occurring long before the ACA have been highlighted as important evidence that health insurance mergers are generally associated with higher prices. In this paper, we study whether such an inference is proper: that is, whether the results of Guardado and, by implication, other retrospective studies of health insurance mergers, particularly those that occurred many years ago, generalize to more recent mergers. Guardado analyzed the 2008 merger of UnitedHealth Group ( United ) and Sierra Health Services ( Sierra ). The study found that average premiums charged to small groups by health insurers increased by 13.7 percent in the year following the merger. However, even Guardado warns that the main results are based on a case study, [and thus] it is unclear whether they have external validity. In this paper, we formally assess whether the Guardado result generalizes to more recent health insurance mergers. We find that the Guardado result does not generalize to other, more recent mergers. In particular, using a methodology very similar to Guardado s, we match their results for United Sierra very closely. However, for a set of five additional health insurance mergers that occurred after the United Sierra merger we find no pattern of higher health insurance premiums following the merger, and, in fact, evidence of lower premiums following the mergers with the highest combined shares, a result consistent with lower medical costs or other such efficiencies. The remainder of the paper is organized as follows. In Section II we provide an overview of the merger retrospective literature. In Section III we provide a brief summary of the Guardado study and describe our replication of their results. In Section IV we present our findings applying our methodology to five health insurance mergers that occurred after the United Sierra merger studied by Guardado. Section V concludes. 2 3 Dafny et al. (2012) is another retrospective analysis which evaluates the 1999 merger of Aetna and Prudential. We are unable to evaluate this study given that they employ a proprietary database that is unavailable to us and that alternative, publicly available data going back to circa 1999 are unavailable. We view the present study as addressing the general question of whether these retrospective analyses generalize to other, more recent health insurance mergers. Testimony of Leemore S. Dafny, Ph.D, Before the Senate Committee on the Judiciary, Subcommittee on Antitrust, Competition Policy, and Consumer Rights, on Health Insurance Industry Consolidation: What Do We Know from the Past, Is It Relevant in Light of the ACA, and What Should We Ask?, September 22,

3 II. OVERVIEW OF THE MERGER RETROSPECTIVE LITERATURE A. RETROSPECTIVE STUDIES OF HEALTH INSURANCE PREMIUMS As discussed above, Guardado finds that average premiums rose by 13.7 percent in markets affected by the United Sierra merger relative to a set of control markets. Other studies of health insurance mergers have also found a positive effect on premiums as a result of increased concentration. Dafny et al. (2012) study the impact of the 1999 merger of Aetna and Prudential Healthcare, two large national insurers, and find that the increase in local market concentration from due to the merger resulted in a seven percent increase in premiums for large group plans. Notably, Manne and Sperry (2016) have criticized the application of the Dafny results to current mergers, arguing that it [is] impossible to draw inferences from statistical results based on the effects of a 17- year-old merger, primarily because the health insurance market is so dramatically different than at the time of the Aetna Prudential merger. Other studies have also addressed the effect of increased health insurance concentration, although not resulting from mergers per se, on premiums: Dafny et al. (2015) exploit variation in the number of insurers participating in Health Insurance Marketplaces across federally delineated ratings areas established by the ACA and find that higher concentration in a ratings area is associated with higher premiums. Trish & Herring (2015) attempt to disentangle the competing effects of increased health insurer concentration on premiums. They find that, on the one hand, increased market power allows health insurers to charge higher premiums to consumers, while on the other hand increased concentration strengthens health insurers bargaining position when negotiating with providers, which can provide cost savings that are then passed on to consumers. The authors find that increased concentration leads to higher premiums as a result of increased market power, but that there are offsetting effects due to the increased negotiating power of the combined entity. B. MERGER RETROSPECTIVES USING DIFFERENCE-IN-DIFFERENCES ( DID ) METHODOLOGY The DiD methodology is a common technique used generally in merger retrospective studies. For example, in the retail supermarket industry, Hosken, Olson, and Smith (2012) analyze the price effects of a series of mergers by comparing prices in markets experiencing a merger to prices in similar markets not experiencing a merger (or any other change in market structure). They find that in general prices in highly concentrated 3

4 markets tended to increase and prices in less concentrated markets tended to decrease after a merger. 4 The DiD methodology is also commonly used to study the price effects of consummated mergers in the airline industry. Carlton et al. (2016) study recent consolidation in the airline industry that has reduced the number of legacy airlines from six to three. They find an increase in output (passengers) on overlap routes relative to control routes and no statistically significant fare effects, which leads them to conclude that the mergers were pro-competitive. 5 In the retail gasoline industry, Hastings (2004) analyzes the effect of the conversion of 260 independent gasoline stations to branded stations in Southern California. The author compares the change in price at stations that lost an independent competitor to the change in price at stations that experienced no change in ownership of nearby stations and finds that the removal of an independent competitor was associated with an increase in retail gasoline prices. 6 III. SUMMARY OF GUARDADO S ANALYSIS AND REPLICATION OF THEIR RESULTS A. SUMMARY OF GUARDADO S ANALYSIS Guardado employs a two-stage technique known as a local linear difference-indifferences (DID) propensity score matching estimator. In the first stage of this technique, the probability (known as a propensity score) that a market was affected by the merger is estimated. Based on these probabilities, the treatment markets are then matched to control markets. After selecting control MSAs using the propensity score method, Guardado then used a DiD model that calculates the change in premiums in the control markets relative to the treatment markets See also Allain et al. who analyze the impact of a 1999 French grocery store merger between the second and fifth largest national players on food prices. The merger affected concentration in local markets differentially in some local markets both grocery stores were present, while in other local markets only one or none of the grocery stores were present. They find that on average the merging firms raised their prices after the merger, more so nationally than locally, and that in some circumstances competitors prices increased as well. Also see Kim and Singal (1993) who analyze a series of fourteen airline mergers that occurred from They compare fare changes on routes where mergers reduced the number of competitors (overlap routes) to fare changes on similar routes (e.g. similar distance) and find that fares increased on the overlap routes relative to the control routes. See also Hastings and Gilbert (2005) who use a DiD approach to study the effect of vertical integration on wholesale gasoline prices. They find evidence that vertical mergers can impact wholesale prices by increasing competitors input costs. Additionally, Taylor and Hosken (2007) analyze the impact of a joint venture between two petroleum companies on both rack and retail gasoline prices by comparing changes in prices in cities likely affected by the joint venture to changes in gasoline prices in similar cities unlikely to be affected by the joint venture. They find that prices for rack gasoline increased in the two cities experiencing the largest increase in concentration from the joint venture, but that retail prices did not increase. 4

5 Using a proprietary database on plan characteristics and premiums obtained from Mercer, 7 Guardado estimated the impact of the United Sierra merger on average premiums in Reno and Las Vegas, Nevada, the only metropolitan areas where the merging companies had significant share overlap pre-merger. Post-merger changes in premiums in Reno and Las Vegas were compared to changes in premiums in a control group of MSAs. 8 Guardado found that premiums charged to individual employees enrolled in fully-insured, commercial health insurance plans in the Nevada MSAs rose by 13.7 percent more than premiums in the control MSAs. 9 B. REPLICATION OF GUARDADO S RESULTS Because the Mercer data on plans and premiums used by Guardado are not available to us, we use publicly available data on plans and premiums as an alternative. The primary dataset we use is the Medical Expenditure Panel Survey Insurance Component ( MEPS ), which is based on a survey of employers providing data on average, fullyinsured premiums for commercial plans, by state, selected MSAs, and year. 10 The MEPS data include fully-insured plans and self-insured plans. Those employers with a selfinsured plan were asked to estimate the cost of switching to a fully-insured plan. Therefore, the survey aims to capture fully-insured premiums or the fully-insured premium equivalents, not simply administrative services only ( ASO ) fees for selfinsured plans. 11 As in Guardado, we use a DiD methodology to estimate the predicted percentage change in premiums following the United Sierra merger. In addition to the different data sources, our analysis differs from Guardado s in a number of other respects. First, our Additional data sources used by Guardado included HealthLeaders InterStudy share data, Medicare Parts A and B data from the Centers for Medicare and Medicare Services (CMS), and demographic, unemployment, and income data compiled by the U.S. Census Bureau, U.S. Bureau of Labor Statistics, and U.S. Bureau of Economic Analysis, respectively. The data we use to replicate Guardado (discussed in more detail below) only include the Las Vegas-Paradise, NV, MSA and no MSA for Reno, NV. Using an alternative control group of all MSAs in the U.S., Guardado found that average premiums in Nevada increased by 14.7 percent. The MEPS data includes information on the 20 largest MSAs nationwide and at least one MSA within each state. See site accessed March 25, Our data includes 74 unique MSA-state combinations. For each state the MEPS data also includes information on all areas outside the selected MSAs aggregated into a Remainder of state category. We do not use this remainder of state data in our analysis. This approach may introduce some statistical noise in the dependent variable; however, there is no reason to believe any error in the dependent variable would be correlated with the merger under study, and thus there is no bias in our estimated coefficients, as confirmed by our close match of the Guardado results. 5

6 methodology does not use propensity scores to match treatment and control markets. Rather, we limit control MSAs to the same census region as the MSA in which the merger occurred. We also include sensitivity tests where we include all MSAs and control for differences between treatment and control markets that could affect health insurance premiums, including differences in demographics, income, unemployment, Medicare expenditures, average copays, average coinsurance rates, and average deductibles. Second, Guardado defines the post-merger period as starting the year after the merger (so the year of the merger is part of the pre-period). We use a longer time period than Guardado, specifically two years before (including the year of the merger) and two years after the merger, to account for possible dissipating effects. 12 To ensure that differences between the findings of Guardado and our findings for other mergers are not an artifact of differences in data or model specification, we first replicate the Guardado results for the United Sierra merger, using our data and model. We obtain very similar results to those of Guardado for the United Sierra merger, demonstrating that the difference in technique does not have a material effect on the results. More specifically, as reported in Table 2 below, we find that premiums increased 12.7 percent more in Nevada than in the control group, very close to the 13.7 percent found by Guardado. This result is not statistically significantly different from the 13.7 percent increase reported by Guardado. IV. APPLYING THE METHODOLOGY USED TO REPLICATE GUARDADO S RESULTS TO ADDITIONAL MERGERS Having established that our data and methodology generate results that are very similar to those of Guardado, we next apply the same methodology to a set of more recent health insurance mergers that occurred after the United Sierra merger. Using a set of criteria described in the Appendix, we select five candidate health insurance mergers for analysis. These criteria ensure, for example, that the merging parties had meaningful pre-merger shares in treatment geographies. Specifically, we require that each merging party have at least five percent of the fully-insured, commercial group enrollment share in the MSA and the combined entity s share exceeds 20 percent in the MSA. The selection process results in five mergers for study: 1. Aetna Coventry (2013) 2. Coventry Preferred (2010) 3. Coventry Mercy (2010) 4. United HealthNet (2009) 5. Anthem M-Plan (2008) As is standard in DiD studies, we include MSA and year fixed effects in our regression (to control for the effect of factors other than the merger on premiums). Technically, the MEPS data is at the MSA-state level (there are a few MSAs present in multiple states) so we include MSA-state level fixed effects in our regressions. When we refer to MSA throughout the rest of this paper we are technically referring to MSA-state. The Anthem-M-Plan event was not technically a merger. Rather, M-Plan wound down operations in Indiana and recommended that its members migrate to Anthem. Anthem paid M-Plan s former 6

7 Table 1 below lists the United Sierra merger plus the five other mergers as well as premerger shares in the MSAs in which the merging parties met the threshold. 14 Two of the mergers (Aetna Coventry and United HealthNet) had significant overlap in multiple MSAs present in the MEPS data, and the other three mergers had significant overlap in one MSA present in the MEPS data. Two of the five mergers we study had substantially higher pre-merger shares than the others: Coventry Preferred (2010) and Anthem M Plan (2008). All else equal, MSAs with larger increases in concentration would have a greater likelihood of resulting in price effects post-merger. On the other hand, a larger combined insurer post-merger will have more bargaining power with providers and a greater likelihood of achieving cost savings that could be passed on to consumers. Thus the results from these two mergers will help us determine which of the two effects predominates. Table 1: List of Mergers and Pre-Merger Shares Pre-Merger Estimated Share Acquiring Acquired Merger MSA (available in MEPS-IC data) Pre-Merger Year Party Party Combined United - Sierra Λ Las Vegas-Paradise, NV 2007 Λ 16% 35% 51% Philadelphia-Camden-Wilmington, PA-NJ-DE-MD % 11% 25% Aetna - Coventry Miami-Fort Lauderdale-Miami Beach, FL % 13% 24% Atlanta-Sandy Springs-Marietta, GA % 10% 21% Coventry - Preferred Wichita, KS % 69% 75% Coventry - Mercy St. Louis, MO-IL % 7% 29% United - HealthNet Bridgeport-Stamford-Norwalk, CT % 13% 23% New Haven-Milford, CT % 12% 21% Anthem - M-Plan Indianapolis-Carmel, IN % 23% 70% Source: HealthLeaders-InterStudy data. Λ Reported shares are shares in the merger year per Guardado et al. (2013), Table 1. Our standard DiD econometric model is: 15 owner (Clarian Health System) commissions for each M-Plan member that migrated to Anthem. Although this technically was not a merger, it did involve one carrier leaving the market and a consolidation of share; therefore, we include it in our set of mergers to study. See: J.K. Wall, Anthem increases its hold on Indiana, Indianapolis Business Journal, March 24, 2008, available at: site accessed February 29, We exclude from the analysis MSAs with a modest amount of overlap that do not meet our selection criteria described in the appendix. Specifically, we exclude MSAs that do not meet our selection criteria but where both merging parties have at least two percent share. As discussed in more detail below, we also include sensitivity tests that include more terms in the DiD model: percentage of the MSA population over 64, percent male, percent Hispanic, percent black, one year lag of the unemployment rate, per-capita income, average MSA-level coinsurance rates, copays, and deductibles, and the one-year lag of log Medicare costs per capita. 7

8 log where the unit of observation is MSA i in a particular year t. The primary independent variable of interest is an indicator variable ( ) that takes a value of 1 for the treatment MSA (or MSAs if the merger affected multiple MSAs see Table 1 above) in the post-merger period, and zero otherwise. Specifically, the indicator variable is the interaction of two indicator variables, one that takes the value of 1 for the treatment MSA(s) (i.e. ) and one that takes the value of 1 for the postmerger years (i.e. ), resulting in our primary independent variable of interest that takes the value of 1 when both 1 and 1. Because we also include MSA fixed effects (δ i ) and year fixed effects (θ t ), this variable reflects the incremental effect of the merger on premiums in treatment MSA(s) relative to control MSAs. Once we run the DiD regression per above, we then convert β 1 from natural logs to percentage change using the formula: 1. The results of our DiD econometric analysis are presented in Table 2 below. As discussed above, for the United Sierra merger, we estimate the merger caused a 12.7 percent increase in premiums in the treatment MSA(s) relative to control MSAs. By contrast, for the Aetna Coventry merger in 2013, the most recent merger we analyze, we find no statistically significant effect of the merger on average, MSA-level premiums. We also find no statistically significant effect on premiums as a result of the Coventry Mercer (2010) merger or the United HealthNet (2009) merger, although we find small, positive coefficients on the premium effect variable for both mergers. Interestingly, for two of the mergers we analyze, we find a statistically significant decrease in average MSA-level premiums. That these are the two mergers with the highest estimated combined shares, namely Coventry Preferred (2010) and Anthem M Plan (2008), is a notable result. One explanation for the reduction in premiums postmerger for these two mergers is that they both might have resulted in substantial medical cost savings from the ability of the two firms to negotiate better discounts with providers. This conclusion would be consistent with the results presented by Trish and Herring (2015). 8

9 Table 2: Estimated Effect of Merger on Premiums Merger 1 Census Region Time Period United-Sierra (2008) West Aetna-Coventry (2013) South Coventry-Preferred (2010) Midwest Coventry-Mercy (2010) Midwest United-HealthNet (2009) Northeast Anthem-M-Plan (2008) Midwest Estimated Coefficient Measuring Effect of Merger on Premiums 0.120*** (0.015) (0.019) *** (0.018) (0.018) (0.086) *** (0.015) Standard errors shown in parentheses; standard errors are clustered at MSA-state-level. *** p<0.01, ** p<0.05 Notes: [1] Year merger was consummated indicated in parentheses. [2] Pre-period includes [3] Pre-period includes [4] Pre-period includes [5] Pre-period includes [6] Pre-period includes We also run two sensitivity tests to confirm our findings described in Table 2 above. The first test, in the second column of Table 3 below, removes the restriction that control MSAs be in the same Census region as the treatment MSA(s). In this specification we find a statistically significant price effect from the Coventry Mercy merger, but there are no other notable differences from Table 2 above (also reproduced in the first column of Table 3 for reference), specifically no consistent pattern of higher premiums following the mergers, with negative and significant effects for the two cases involving the highest combined share. Hence, this sensitivity test increases our confidence in our findings. The second test, presented in the third column of Table 3, removes the restriction that control MSAs be in the same Census region as the treatment MSA(s) and also includes as explanatory variables other factors that might affect insurance premiums in an MSA. These include (all at the MSA level): percentage of the population over 64, percent male, percent Hispanic, percent black, one year lag of the unemployment rate, per-capita income, average coinsurance rates, copays, and deductibles, and the one-year lag of log Medicare costs per capita. The results of this sensitivity test are similar to the specification presented in the second column, meaning that adding the control variables does not change any of our conclusions. 9

10 Table 3: Sensitivity Tests Merger Λ Census region without explanatory variables All MSAs without explanatory variables All MSAs with explanatory variables ΛΛ United-Sierra (2008) 0.120*** 0.125*** 0.132*** (0.015) (0.009) (0.024) Aetna-Coventry (2013) (0.019) (0.016) (0.026) Coventry-Preferred (2010) *** *** *** (0.018) (0.008) (0.018) Coventry-Mercy (2010) *** 0.060*** (0.018) (0.008) (0.017) United-HealthNet (2009) (0.086) (0.076) (0.074) Anthem-M-Plan (2008) *** *** *** (0.015) (0.009) (0.018) Standard errors shown in parentheses; standard errors are clustered at MSA-state-level. *** p<0.01, ** p<0.05 Notes: Λ Year merger was consummated indicated in parentheses. ΛΛ All MSAs with explanatory variables version for Aetna-Coventry uses premium and other information due to data limitations. V. CONCLUSION Guardado estimated an increase in premiums following the United-Sierra merger, a finding we replicate. However, we show that this result is reversed for other, more recent mergers. In particular, using a methodology very similar to Guardado s, we match their results for United Sierra very closely. However, for the recent Aetna Coventry merger, and for a set of four other, more recent mergers, we find no systematic evidence that health insurance mergers result in higher premiums, with premium reductions for the two mergers resulting in the highest combined share. 10

11 APPENDIX: MERGER SELECTION METHODOLOGY We use a list of horizontal insurance mergers from the Health Care Services Acquisition Reports, published by Irving Levin Associates. This is a comprehensive source of mergers in the health care industry and was the source that Guardado relied on to identify potential mergers for study. Because we want to study mergers that were no older than United Sierra and we want to allow for a two year post-merger period, we select the 83 mergers that were announced between 2008 and For each of these 83 mergers, at least one of the parties was a health insurer. We add one merger that was not listed in the Health Care Services Acquisition Reports, the Cigna acquisition of Great West in We then exclude certain mergers because at least one of the merging parties was less focused on the commercial health insurance market and/or was not primarily a health insurer; these exclusions include companies that: Focused mostly/exclusively on Medicare Advantage or Medicaid; Focused mostly/exclusively on vision or dental plans; Focused mostly/exclusively outside of U.S.; Are not primarily health insurers (e.g., providers, third-party administrators, pharmacy benefits managers). Next, we implement stand-alone and combined share criteria in order to focus on those mergers where we would be more likely to find a premium effect. We require that each company have at least a five percent share of the commercial group health insurance market in at least one state pre-merger, and that the two merging parties combined had a pre-merger share of at least 20 percent. Table 4 below lists the 84 mergers we start with, and shows the criteria by which certain mergers were eliminated. 11

12 Table 4: Potential Mergers Excluded Based On Year Announced Acquiring Party Target Party Focus 5/5/20 Share 2007 CIGNA Corp. Great-West Healthcare Y 2008 Anthem M-Plan 2008 UnitedHealth Group, Inc. Unison Health Plans Y 2008 AmeriHealth Mercy Family of Companies Behavioral HealthCare Network of Pennsylvania Y 2008 MCCI Group Holdings, LLC Palm Beach Health Associates Y 2008 DentaQuest Atlantic Dental Y 2008 Centene Corporation Celtic Group, Inc. Y 2008 Humana, Inc. OSF HealthPlans Y 2008 Medical Mutual of Ohio Premier Health Systems Y 2008 WellPoint, Inc. Resolution Health Y 2008 WellPoint, Inc. DeCare Dental Y 2008 Humana, Inc. Metcare Health Plans, Inc. Y 2008 Humana, Inc. Cariten Health Y 2008 Molina Healthcare, Inc. Florida NetPASS, LLC Y 2008 Blue Cross Blue Shield of Florida Florida Health Care Plans Y 2008 Health Care Service Corporation TMG Health, Inc. Y 2008 Capital BlueCross Dominion Dental Y 2008 Centene Corporation Amerigroup Community Care Y 2008 Humana, Inc. Las Vega Medicare business Y 2009 Express Scripts, Inc. NextRx Y 2009 Peoples Health Network Membership interest Y 2009 PacificSource Health Plans Primary Health Y 2009 Triple-S Management Corporation La Cruz Azul de Puerto Rico, Inc. Y 2009 Dental Network of America, LLC DenteMax Y 2009 UnitedHealth Group Health Net 2009 PacificSource Health Plans Commercial membership Y 2009 Aetna, Inc. Horizon Behavioral Services, LLC Y 2009 SeeChange Health (division of Trivers, LLC) Central Benefits National Life Insurance Y 2009 Blue Care Network of Michigan Physicians Health plan of Mid-Michigan Y 2009 Coventry Health Care, Inc. Preferred Health Systems 2009 PacificSource Health Plans Clear One Health Plans, Inc. (fka Clear Choice Health Plans) Y 2010 Hawaii Medical Assurance Association Summerlin Life & Health Insurance Y 2010 Centene Corporation Carolina Crescent Health Plan Y 2010 Simply Healthcare Plans Total Health Choice Y 2010 Coventry Health Care, Inc. Mercy Health Plans 2010 Centene Corporation Citrus Health Care assets Y 2010 HealthSpring, Inc. Bravo Health, Inc. Y 2010 CIGNA Corp. Vanbreda International, NV Y 2011 Catalyst Health Solutions, Inc. Walgreen Health Initiatives Y 2011 Aetna, Inc. Prodigy Health Group Y 2011 WellPoint, Inc. CareMore Health Group Y 2011 Aetna, Inc. Medicare supplement business (Genworth Financial) Y 2011 Express Scripts, Inc. Medco Health Solutions, Inc. Y 2011 Independence Blue Cross AmeriHealth Mercy Family of Cos. Y 2011 Partners HealthCare System Neighborhood Health Plan Y 2011 Humana, Inc. Arcadian Management Services Y 2011 Humana, Inc. MD Care Y 2011 CIGNA Corp. FirstAssist Insurance Services Y 2011 CIGNA Corp. HealthSpring, Inc. Y 2011 Amerigroup, Inc. Health Plus Y 2011 Coventry Health Care, Inc. Children's Mercy's Family Health Partners Y 2011 UnitedHealth Group XLHealth Corp. Y 2012 Highmark BCBS of Delaware Y 2012 Universal American Corp APS Healthcare Inc. Y 2012 UnitedHealth Group Medica Healthcare Plans Y 2012 UnitedHealth Group Preferred Care Partners Y 2012 Riverside Company DentalPlans.com Y 2012 HealthSmart Holdings Innovative Benefit Administrators Y 2012 Nautic Partners, LLC Superior Visions Holdings, Inc. Y 2012 Guardian Life Insurance MasterCare DENTS Y 2012 McLaren Health Plan CareSource Michigan Y 2012 CIGNA Corp. Great American Supplemental Benefits Y 2012 Towers Watson & Co. Extend Health Inc. Y 2012 SelectAccount HAS Business Y 2012 Visiting Nurse Service of NY SelectHealth Y 2012 Prime Health Services Casualty Management Network Y 2012 CIGNA Corp. Medicare Advantage Plans Y 2012 WellPoint, Inc. Amerigroup Y 2012 WellCare Health Plans, Inc. Arcadian Health's Arizona assets Y 2012 The Principal Financial Group First Dental Health Y 2012 Aetna, Inc. Coventry 2012 WellCare Health Plans, Inc. Easy Choice Health Plan Inc. Y 2012 Health Care Service Corporation BCBS of Montana Y 2012 Inova Amerigroup VA Y 2012 UnitedHealth Group Amil Participacoes S.A Y 2012 Detroit Medical Center ProCare Health Plan, Inc. Y 2012 WellCare Health Plans, Inc. United HealthCare's Medicaid Business Y 2012 Torchmark Corp. Family Heritage Life Insurance Company Y 2012 Humana, Inc. Metropolitan Health Networks Y 2013 WellCare Health Plans Missouri Care, Inc. Y 2013 Citrus Universal Healthcare, Inc. Universal Health Care Group Y 2013 MVP Health Care Hudson Health Plan Y 2013 WellCare Health Plans Windsor Health Group Y 2013 Health Care Service Corporation Lovelace Health Plan Y 2013 Aetna, Inc. InterGlobal Y Source: Health Care Services Acquisition Reports, published by Irving Levin Associates; Cigna Website; 12

13 Finally, we examine year-to-year changes in state-level HHIs to identify any potential mergers that were not captured above. Using a 20 percent change in the HHI, we identify only one additional event, Anthem-M-Plan. This was not a merger; instead, M-Plan wound down its operations and recommended that its members migrate to Anthem. Anthem paid M-Plan's former owner (Clarian Health System) a commission for each M- Plan member that migrated to Anthem. 16 Although this technically was not a merger, it did involve one carrier leaving the market and a consolidation of share; therefore, we include it in our set of mergers to study. 16 J.K. Wall, Anthem increases its hold on Indiana, Indianapolis Business Journal, March 24, 2008, available at: site accessed February 29,