Monopoly, monopsony and contestability in health insurance: a study of Blue Cross plans.

Foreman, Stephen Earl ; Wilson, John Anderson ; Scheffler, Richard M. 等

I. INTRODUCTION

Blue Cross and Blue Shield Plans (the "Plans") have historically occupied a unique niche in health care. Pauly [1987] notes that the Plans serve both as sellers of health insurance and as buyers of health services. Marmor [1991] described the Plans' development of health insurance concepts in the 1930s. Their role as innovators permitted them to occupy a dominant market share position in most geographic areas. As a result, Blair, Ginsburg and Vogel [1975] believe they have traditionally received favorable regulatory treatment including non-profit organization status and exemption from taxation.

Potentially, the Plans have dual market power(1): monopoly power in the market for health insurance and monopsony power in the market for health services. This may result in operating inefficiencies and may produce adverse welfare consequences. Monopsony theory posits that the Plans will reduce payments to providers and may stifle innovation in the market for health services. This occurs as the result of the Plans' dominant market shares (generated from the fact that they were often the first in the market for insurance) and from their ability to enforce a credible threat to reduce providers' business should providers refuse to provide reduced cost services. The Plans' dominant market share and their advantage in the market for health services permit them to maintain a competitive advantage in the market for health insurance.(2) Monopoly theory suggests that the Plans' advantage (in the market for medical services due to regulatory and tax advantages) will cause them to (i) raise premium prices, (ii) sell only a limited range of health insurance policies, (iii) offer policies with more complete coverage than consumers demand,(3) (iv) reduce innovation in the health insurance market and (v) demonstrate unduly high administrative costs.

In the 1970s empirical work supported these predictions. However, the 1980s has been an era of growth of competition and the potential for even more competition in the market for health insurance. Baumol, Panzar and Willig [1982a] noted that a credible threat of entry can produce "competitive" market conditions in a concentrated market, resulting in reduced prices, increased quantity and more efficient administrative cost structures. If Pauly is correct, that there are no inherent scale economies in the health insurance market, and if the market for health insurance is contestable, the Plans will pass advantages along to subscribers as reduced premiums and will tolerate fewer administrative cost inefficiencies.

Accordingly, the Plans present an opportunity to study market structure and its impact in both the market for health care and the market for health care insurance. To date, no study has evaluated this interplay. Greater understanding of these interactions is important not only to deal with the phenomena of the Plans (for instance, more appropriate regulatory policies) but also to refine the single-payer debate and to consider operating standards for Health Alliances and Health Plans included in the Administration's health care reform package outlined by the White House Domestic Policy Council [1993]. Section II of this paper evaluates prior studies of the Plans. Section III motivates the model for Plan behavior. Section IV describes the results of our empirical study of the Plans' premium pricing, provider payment and administrative costs. Section V contains our conclusions.

II. PRIOR STUDIES OF PLAN BEHAVIOR

A number of prior studies have evaluated the Plans' unique position in the market for health insurance and in the market for health care. Most focus on the Plans' monopsony power and payments to providers. Most important, they evaluate markets that have gone through substantial change as health maintenance organizations and preferred provider organizations (HMOs and PPOs) have replaced fee-for-service plans. Several assess market power and administrative costs. We have been unable to find any that consider the relationship between market share and premiums.

Monopsony Power

Surprisingly, previous studies fail to reach consensus on whether the Plans have monopsony power or, given its existence, how the Plans use it. Adamache and Sloan [1983] found that the Plans exercised monopsony power by increasing the discounts that they were able to demand from hospitals while a subsequent study of Plan behavior in Indiana by Staten, Dunkelberg and Umbeck [1987] contests the Plans' monopsony power.(4)

Feldman and Greenberg [1981] considered the inverse question, the effect of discounts on the Plans' market share. While discounts seemed to increase market share, this did not necessarily generate greater discounts. They speculated that the Plans' dissipated their market power as "administrative slack."(5) Feldman and Greenberg [1982] found that while economies of scale (Plan size) correlated with cost control measures, market share did not. Frech [1988] reviewed Plan behavior in Massachusetts. In order to reduce costs to subscribers, the Plan banned balance billing. This suggests that it held and used monopsony power.

Commentators also hold different views on the welfare implications of the Plans' monopsony power. Pauly [1987] and Frech [1988] suggest that monopsonists will maintain price at inefficiently low levels (restricting supplier output) and will engage in inefficient administrative practices. However, Adamache and Sloan [1983] believe that use of power to reduce costs can be beneficial since it reduces health care prices to a more competitive level.

Only a few studies have considered premiums. These review the effect of premiums on market share rather than the effect of share on premiums. For example, Frank and Welch [1985] found that premium differences explained 89 percent of the variation in enrollment. Even if the Plans use their power to reduce payments for health services, the literature does not indicate whether they pass along any part of the reductions.

Administrative Slack

The traditional view is that monopsonists and monopolists will have little motivation to control administrative costs.(6) Holtzmann [1983] has suggested that nonprofit firms may not control their administrative costs either, and the Plans are also nonprofit organizations. Accordingly, on two levels there is the potential for the Plans to have inefficiently high administrative costs. Prior empirical studies reach differing conclusions on this issue. One study found that the greater the tax benefit received by the Plan the lower the Plans' administrative costs. Kass and Paultler [1981] determined that the Plans seemed to use their competitive advantage to reduce administrative costs. Blair, Ginsburg and Vogel [1975] detected no economies of scale among the Plans; if the Plans have scale economies they disappear in managerial slack. Adamache and Sloan [1983] discovered evidence of economies of scale but only tentative support for dissipation of their competitive advantage in managerial slack. Vogel [1977] evaluated the relationship between tax rate and the Plans' administrative costs. In Vogel's study tax increases were associated with reduced administrative costs. Perhaps Plans eliminated inefficiency to offset the effect of increased taxes.(7)

Competition: Changes in the Plans' Markets

Prior studies of the Plans' response to their market power and nonprofit status consider a state of the world that has changed. Frech and Ginsburg [1978] found that Blue Cross and Blue Shield dissipated their cost advantages on more complete insurance, administrative slack and inefficiency However, Feldman, Kralewski and Dowd [1989] note that for the past fifteen years Plans have faced increasing competition. Anecdotal evidence provided by Goldberg and Greenberg [1985] suggests that they have responded to this threat. Frech [1988] updated his 1977 study of competition in the health insurance industry and found that insurers were far more willing to compete by controlling costs.

In many areas the Plans seem to have managed to retain a high market share. How have the changes in the market for health insurance impacted the Plans? Entry into the market by commercial insurers and by health maintenance organizations has been relatively easy.(8) Government and industry action have removed entry barriers.

In short, the Plans may or may not have market power. The existence of this power might prompt the Plans to operate inefficiently, to reduce price and quantity in the market for health service and to increase premiums and reduce quantity in the market for health insurance. However, the insurance market may have become more competitive. If so, we would expect to see evidence of the Plans' increased operating efficiency through reduced administrative costs, reduction of amounts paid to providers and reductions of premiums charged to subscribers. Existing studies have not evaluated the effect of market share or premiums or the impact of changes in the market on premiums and administrative costs. This paper addresses these topics, using empirical data on plan performance.

III. THE MODEL

Feldstein and Wickizer [1992] break down health insurance premiums (Premium) into two components: claims experience (payments to providers or Paym't) and the loading charge (profit and administrative costs or Loading). Claims experience is the true actuarial cost of insurance. If the Plans exercise some form of monopsony power, we would expect the Plans' payments to providers (their claims experience) to be a function of (i) the benefits provided to patients (Benefit), (ii) the Plans' market power (implied by market share or Share),(9) (iii) any economies of scale in the market for health services (Size), (iv) the characteristics of the patients who receive care (Dem), (v) cost containment measures (Contain), (vi) the cost structure of hospitals and physicians (primarily wages or Wage), (vii) the level of competition in the market for health services (reflected by proportions of Plan members who are members of HMOs and PPOs) and (viii) rate setting and other regulation (Reg). Equation (1) demonstrates this relationship:

(1) Paym't = f(Benefit, Share, Size, Dem, Contain, Wage, HMO, PPO, Reg).

The loading charge reflects the Plans' overhead and profit. Since the Plans are nonprofit we concentrate on the overhead component. Loading costs are a function of administrative costs. These, in turn, include (i) the wages paid to Plan employees (Wageins), (ii) economies of scale enjoyed by the Plans (Size), administrative activity as reflected by claims processed (Claims), (iv) competition (the HMO/PPO share surrogate), (v) cost containment activities requiring administrative time (Contain) and (vi) the degree to which the Plans' market power impacts administrative costs (Share). If monopsony and monopoly power prompt the Plans to engage in administrative slack, we can expect the Plans' market shares to have a positive impact on overhead. If other factors are at work market share could have a neutral or negative impact.

(2) Admin = g(Share, Size, Wageins, Claims, Contain, HMO, PPO).

Accordingly, since premiums are a function of claims and loading:

(3) Premium = j(Paym't, Admin) = j(f(Benefit, Share, Size, Dem, Contain, Wage, HMO, PPO, Reg) + g(Share, Size, Wageins, Claims, Contain, HMO, PPO)).

We consider the impact of changes in the Plans' market share on loading costs, on payments to providers and on premiums. For example, holding other variables constant, we consider the changes in premiums with respect to market share:

(4) [Delta]Premium/[Delta]Share = [Delta]j / [Delta]f x df / dShare + [Delta]j/[Delta]g x dh/dShare.

The term [Delta]j/[Delta]f represents the change in the premium attributable to changes in payments to providers. Since increases in payments to providers will usually result in greater health insurance premiums, we assume that this expression is positive. The term df/dShare reflects the effect on subscriber premiums produced by the Plans' market shares (monopsony power) operating through payments to providers. If this value is positive, the Plans use their power to increase premiums.

The term [Delta]j/[Delta]g is the change in premiums associated with changes in loading (administrative) costs. Again, we assume that this expression is positive since increased loading costs will usually translate into premium increases. We assume that this expression is also positive since increased administrative costs translate into increased loading costs. Accordingly, the measure of interest is dh/dShare, a measure of the change in administrative costs produced by changes in the Plans' market share. If this expression is positive, increased market share results in greater administrative costs, implying that the Plans exhibit managerial inefficiencies as a result of their market power. If it is negative, it would appear that the Plans are somehow able to use their market power to reduce their administrative costs (perhaps by transferring costs to the providers).

Data and Variables

We obtained the data for the study from The Blue Cross and Blue Shield Association and from the Area Resource File. Table I contains a summary of study variables, a description of their derivation and the source of data for each variable. Table II contains summary statistics for the variables.

Our unit of observation is individual Plans for the years 1986 through 1988. We have sufficient data for the study from forty-seven of the fifty-three plans reporting.(10) Each of the Plans included in our study is a merged Blue Cross and Blue Shield Plan. Payments to providers include both hospital and physician payments. The regional mix of our sample reflects the regional mix of all of the Plans. Forty percent of the Plans in our sample are from the Southern region, 25 percent from the Northeast, 21 percent from the North-central and 14 percent from the West. Thirty-seven percent of all Plans are from the Southern region, 26 percent are from the Northeast, 22 percent from the north-central and 15 percent from the West. The average Plan penetration rate for the sample is 29 percent. This is similar to the penetration rate for all of the Plans (28 percent). For each of the dependent variables studied, we rejected the null hypothesis that the means of the forty-seven-Plan sample were significantly different from the means of the entire fifty-three-Plan population.(11) In short, we conclude that our sample is representative of the entire population of Plans.

Dependent Variables

Our three dependent variables are: (i) health insurance premiums charged by the Plans (Premium) as expressed in equation (3) above, (ii) payments to providers for providing health services to Plan members (Paym't) as in equation (2) and administrative costs for each Plan (Admin) as in equation (1). We compute premiums in terms of subscription revenue per thousand members. The usual premiums measure [TABULAR DATA FOR TABLE I OMITTED] is subscription revenue per subscriber. Members include the subscribers and each of the subscriber's dependents. Accordingly, the study's premium measure is smaller than the usual premium measure. In effect, the use of members instead of subscribers adjusts for family size. We evaluate payments in terms of payments per thousand members. Administrative costs are each Plan's administrative costs divided by thousands of members in the Plan. We deflated each of the administrative cost variables using the consumer price index and the payment and premium variables by the index of medical care prices (U.S. Department of Commerce [1992]).(12) The natural log form [TABULAR DATA FOR TABLE II OMITTED] of the dependent variables provided the best model fit, so we used the log form of Admin, Paym't and Premium as the study variables of interest.

Market Structure Variables

Our main interest relates to the monopoly and monopsony conduct of the Plans. Accordingly, the crucial variable is the Plans' market share (Share) in the geographic area. The geographic area is that area defined by each Plan as its market area.(13) The variable Share is the proportion of Plan members to area population. In addition, we separate the effect of market share from the effect of Plan size as measured by thousands of members enrolled in the Plan (Size). Use of the size variable will determine whether there is any economy of scale effect that differs from the market share impact.

The types of competitors also determine the market structure: fee for service insurance plans and managed care organizations. While we do not have a measure of private HMO and PPO penetration in each Plan's market, we do have measures of the proportion of each Plan's members enrolled in HMOs and PPOs. This may give some reflection of managed care activity and competition in the Plan's market.

Regulatory and Cost Containment Variables

Our model includes several "regulatory" variables: a qualitative independent variable indicating (i) whether a Plan is subject to active rate review (generally limited to individual businesses) (Review), (ii) rate filing (Filing), (iii) a requirement for Plans to file contracts with regulators (Contract) and (iv) whether a Plan is subject to legislation mandating a number of benefits the Plan must provide (Mandate). The Plans are also subject to a number of different types of taxes. The model includes the subscription taxes incurred by the Plans (Subscript), domestic taxes in their principal place of business (Domestic), and taxes imposed by other states (Foreign). The model also includes the total for all taxes paid (Taxes). In addition, many Plans have implemented cost containment programs. The supply-side cost containment variables include retroactive and concurrent utilization review (Retroact and Concurrent).

Demographic Variables

Cultural variables and income often impact the demand for health insurance and for health care. Accordingly, we included a number of demographic variables in our equations such as (i) per capita income (Percapita) (per capita income deflated by the CPI), the proportion of the population that is white (White) and the proportion of the employed population in a Plan area that belongs to unions (Union). Further, there are regional variations in medical practice and in medical care costs.

Wage Variables

Finally, the wages paid by the Plans and by providers impact premiums, payments and costs. Wage variables include insurance company wages (Wageins), hospital wages (Wagehos) and physician wages (Wagephys). We deflated insurance industry wages by the CPI and hospital and physician wages by the medical care component of the CPI.(14)

Time Considerations

The dependent variables are subject to change over time. We used a time trend variable (Trend) to measure the effect of time. In addition, quarterly data implies some seasonality. We control for this by including qualitative independent variables for quarters (Qtr1, Qtr2, Qtr3 and Qtr4).

In sum, our three-part model reflects the models used in a number of prior studies, such as Adamache and Sloan [1983], particularly in our equations for administrative costs and payments to providers.

Estimation

Because the basic models illustrate autocorrelation and heteroscedasticity we used the Fuller-Battese time-series cross-section pooling method to estimate the equations. This "error components" model separates error terms into three portions: (i) time-series error, (ii) cross-section error and (iii) random error. The method corrects for autocorrelation and heteroscedasticity so that the error remaining in the model reflects only random effects.

Frank and Welch [1985] found that market share can impact premiums, payments to providers and administrative costs. Premiums also impact market share. Because payments to providers and administrative costs impact premiums, they indirectly affect market share as well. Accordingly, there is some likelihood that the Plans' market share variable is endogenous.(15) In order to avoid this problem we created an instrumental variable for market share. We used an instrument to predict market share using a linear regression equation that models share as a function of plan size (to isolate differences based on economies of scale from market share effects), time, geographic indicators, union membership, per capita income, HMO and PPO membership and various regulation (rate setting) variables. The predicted market share becomes the value for Share as described in equation (1) above.

(5) Share = f(Trend, Union, Percapita, Review, Filing, Subscript, Domestic, Foreign, White, Size, Sizesq, HMO, PPO).

The equation for the instrumental variable has good predictive power and the coefficients for most of the exogenous variables are significant (see Table III).(16) Per capita income is a good predictor of market share as is the proportion of the population that is white. Employed workers may prefer the Plans to other types of health insurance. Market shares increase where there is active rate review and diminish when Plans merely file their rate schedules. This finding may be attributable to a number of factors. The Plans may be using rate review to shield themselves from competition. Alternatively, this may reflect a greater use of regulation where the Plans have larger market shares. Market share declines for the Plans as taxes increase, supporting the assertion that much of the Plans market advantage derives from favorable tax and regulatory treatment. Market share increases with Plan size. The HMO and PPO variables are significant and negative suggesting that these reflect increased competition (from other plans and from self-insured employer plans) and that increased competition will reduce the Plans' market shares.

We use the predicted value of market share (Share) as a proxy for market share in the equations for administrative costs, payments and premiums to avoid problems with endogeneity. The best fit for each of the dependent variables occurs with the natural log specification. The equations for administrative costs, payments to providers, and premiums, estimated using Fuller-Battese time-series cross-section pooling, are as follows:

(6) Paym't = f(Share, Size, Mandate, Percapita, Mandate, Review, Filing, Contract, Wagehos, Wagephy, HMO, PPO).

(7) Admin = f(Share, Size, Percapita, Retroact, Concurrent, Qtr1, Qtr2, Qtr4, Taxes, HMO, PPO, Community, Experience).

(8) Premium = f(Share, Size, Qtr2, Qtr3, Qtr4, Community, Experience, Mandate, Percapita, Retroact, Concurrent, Review, Filing, Contract, Wagehos, Wagephy, HMO, PPO).

TABLE III

Instrumental Variable for Market Share

Dependent Variable: Share

 Parameter Standard T for H0:
Variable Estimate Error Parameter = 0

Intercept 22.856 6.810 3.356
Trend -0.105 0.141 -0.745
Union 0.038 0.041 0.944
Percapita 0.000 0.000 3.812
Review 14.467 1.184 12.209
Filing -13.100 1.845 -7.098
Subscript -8.024 0.978 -8.201
Domestic -0.944 0.568 -1.662
Foreign -5.130 1.331 -3.854
White 0.161 0.049 3.282
Size 0.000001 0.000 6.532
Sizesq -2.248E-8 0.000 -3.269
HMO -0.468 0.077 -6.075
PPO -0.242 0.025 -9.607

F-Value = 275.200 Adj [R.sup.2] = 0.9048

Equation (6) provides the empirical model for payments to providers that operationalizes the theoretic model specified in equation (1). Equation (7) operationalizes the model for administrative costs described in equation (2). Equation (8), the empirical model for premiums developed by combining equations (6) and (7), operationalizes equation (3).

We examined a number of the more important independent variables for endogeneity using a Hausman test. We found market share to be endogenous and used an instrumental variable to deal with the potential bias that such endogeneity might produce. We did not find statistically significant endogeneity for the other independent variables.(17) Moreover, even if there was an endogeneity problem with respect to these variables, the lack of correlation between them and the instrument for market share indicates that the model conclusion about the impact of market share is not subject to bias.

IV. EMPIRICAL RESULTS

Payments to Providers

Table IV reports the estimates relating the impact of predicted market share on payments to providers.(18) An increase in the Plans' market share is significantly associated with reduced payments to providers. A 10 percent increase in share is associated with an 11.6 percent reduction in payments. The Plans seem to be using market share to hold down health care costs in the form of payments to providers. Thus, the Plans have and are exercising some monopsony power. This is not attributable merely to size since size is significant as well. A 10 percent increase in size is correlated with a 5 percent reduction in payments. The Plans are able, as Pauly [1987] predicted, to use their advantages to enforce their buying power. If simply reducing payments to providers is a policy goal, providing purchasing size and power to the Plans seems to meet this goal.(19)

TABLE IV

Fuller-Battese Model for Administrative Cost

Dependent Variable: Admin

 Parameter Standard T for H0:
Variable Estimate Error Parameter = 0

Intercept 3.192636 0.446 7.147
Share -0.025625 0.008 -3.075
Size -4.807E-8 2.139E-8 -2.247
Percapita 0.000111 4.000E-5 2.640
Retroact -0.092097 0.150 -0.610
Concurrent -0.318566 0.077 -4.132
Qtr1 -0.972262 0.032 -30.196
Qtr2 -0.355555 0.029 -12.108
Qtr4 0.269310 0.029 9.173
Taxes 1.117E-5 8.900E-6 1.244
HMO 0.001556 0.006309 0.246
PPO -0.007058 0.003484 -2.025
Community 1.79E-7 6.572E-8 2.724
Experience 2.99E-7 8.371E-8 3.570

Variance Component Estimates

SSE 13.14727 DFE 362
MSE 0.036318 Root MSE 0.190574

Variance Component for Cross Sections 0.237536
Variance Component for Time Series 0.000028
Variance Component for Error 0.032366

Other significant variables in the provider payment equation include rate regulation, hospital wage rates, and HMO and PPO shares. Rate regulation increases payments to providers. Payments are 0.79 percent higher in states that have rate review. Greater HMO and PPO membership proportions reduce payments to providers, reflecting the incentives for Plans facing competition to hold down their costs. A 10 percent increase in the HMO enrollment proportion diminishes payments by 1.22 percent while a similar increase in the PPO enrollment proportion reduces payments by 12.8 percent. Increases in hospital wage rates raise payments to providers. Hospitals with greater wage rates have incentives to negotiate greater plan payments.

Administrative Costs

Table V provides the results for the study of the impact of market share on administrative costs.(20) Increased market share is associated with a small but significant decline in administrative costs, as are increases in the Plans' size. A 10 percent increase in share produces a 6.9 percent reduction in administrative costs. A 10 percent increase in size is associated with a 1.66 percent reduction in administrative costs. Increases in plan size can produce economies of scale. Increases in share can coexist with reduced administrative costs for marketing and for managing HMO and PPO products.

TABLE V

Fuller-Battese Model for Payments to Providers

Dependent Variable: Paym't

 Parameter Standard T for H0:
Variable Estimate Error Parameter = 0

Intercept 6.195 0.999 6.198
Share -0.043 0.017 -2.553
Size -1.66E-7 2.9E-8 -5.641
Mandate 0.120 0.213 0.560
Review 0.794 0.383 2.071
Filing -0.275 0.565 -0.487
Contract -0.122 0.882 -0.138
Wagehos 0.001 9.47E-4 1.971
Wagephy 7.3E-5 2.69E-4 0.271
HMO -0.021 0.008 -2.546
PPO -0.009 0.004 -2.142

Variance Component Estimates

SSE 3.675743 DFE 365
MSE 0.010071 Root MSE 0.100352

Variance Component for Cross Sections 0.431698
Variance Component for Time Series 0.001084
Variance Component for Error 0.009163

The model produces a number of additional significant results of interest. As the per capita income of members increases Plan administrative costs rise. A 10 percent increase in per capita income gives an 11.9 percent increase in administrative costs. Wealthier groups may require a wider range of specialized services, resulting in increased overhead for providing service to them. Administrative costs rise each quarter during the year. Plans may defer expenditures to later periods during the year. Concurrent and retrospective utilization review each reduce administrative costs, although only concurrent review does so significantly. A 10 percent increase in concurrent review activities reduces administrative costs by 3.2 percent. As more of the Plan's business consists of PPO contracts, the Plan's administrative costs drop. Discounted contracts may be easier to administer. Greater proportions of the Plan's costs represented by PPO contracts may correlate with reduced numbers of providers. Having to work with fewer providers may ease the Plan's claims burden and, therefore, reduce administrative costs. Greater numbers of PPO contracts may reflect greater competition and incentives for the Plans to manage their administrative costs in a more efficient manner. Finally, as expected, increases in claim volume translate into increases in administrative costs.

TABLE VI

Fuller-Battese Model for Premiums

Dependent Variable: Premium

 Parameter Standard T for H0:
Variable Estimate Error Parameter = 0

Intercept 3.491 0.794 4.391
Share -0.023 0.010 -2.139
Size -1.2E-8 1.78E-8 -0.681
Qtr1 -0.901 0.070 -12.765
Qtr2 -0.325 0.068 -4.724
Qtr4 0.226 0.068 3.288
Community 1.4E-7 6.5E-8 2.105
Experience 2.2E-7 8.4E-8 2.595
Mandate 0.078 0.111 0.703
Percapita 1.6E-5 4.0E-5 0.394
Retroact -0.036 0.152 -0.238
Concurrent -0.358 0.076 -4.659
Review 0.341 0.234 1.454
Filing -0.447 0.310 -1.443
Contract 0.293 0.513 0.572
Wagehos 0.003 0.001 2.163
Wagephy 0.001 0.000 4.710
HMO -0.009 0.006 -1.369
PPO -0.006 0.003 -2.068

Variance Component Estimates

SSE 12.8224 DFE 357
MSE 0.035917 Root MSE 0.189518

Variance Component for Cross Sections 0.100077
Variance Component for Time Series 0.003580
Variance Component for Error 0.032881

Premiums

Table VI delineates the results for premium payments.(21) Increases in predicted market share are associated with a decline in premiums. The Plans seem to be passing reductions in administrative costs and payments along to employer groups as reduced premiums. A 10 percent increase in share results in a 6.2 percent premium reduction. However, size does not significantly reduce premiums.

Other significant variables in the premium equation include seasonality (premium revenue rises by quarter during the year), claims volume (increase premiums), hospital and physician wages (reflecting the ability to pass along the expense of increased payments to providers in premium increases), PPO member proportion (increased PPO enrollment holds down premium increases, perhaps reflecting competition), and concurrent review (reduces premiums). A 10 percent increase in hospital wages produces a 1.16 percent increase in premiums.

Summary of Results

The study results indicate that the Plans have used increased market shares to reduce their administrative costs, to reduce payments to providers and to reduce premiums charged to subscribers. Contrary to theory and unlike the findings of previous studies, we find that the Plans' market power is not translating into operating inefficiency.(22) In accordance with a number of prior studies, the Plans are still using monopsony power to reduce payments to providers. Finally, despite traditional monopoly theory, the Plans have not used their increased shares in the market for health insurance to increase premiums. To the contrary, they have translated their market shares into premium reductions. This indicates (as predicted by Pauly) that the Plans may have purchasing power in the market for health care services without any concurrent selling power in the market for health insurance.

V. CONCLUSIONS

These results are generally consistent with the direction of changes in the world of health insurance over the past fifteen years. Along with the market for health insurance, Plans appear to have become more competitive. This has required them to become more efficient. Moreover, the fact that the Plans seem to be using economies of scale (greater size) and monopsony power to reduce their costs and that they seem to be passing the reductions along to their customers suggests that contestability theory (or the expectation of competition) may well be operating in the market for health insurance. Contestability in the market for health insurance, coupled with monopsony power in the market for health services and administrative services predicts all three aspects of the Plans' market conduct that we observe in this study: increased administrative efficiency, reduction of payments to providers and more competitive pricing in the sale of health insurance.

Indeed, the evolution of competition in the market for health insurance is well documented, as is the incidence of decline in the Plans' market shares. If we extend the concept of contestable markets to the concept of "expectation of competition" the Plans' rational response would clearly take the form that we observe here. Faced with a belief that competition has grown and will continue to grow, a rational Plan manager will use market power to enhance the Plan's competitive position through enhanced operating efficiency and reductions in provider payment, using these advantages to reduce premiums and maintain market share. Moreover, while we have not been able to compare the Plans' response (as nonprofit firms) to the strategic response of proprietary providers of health insurance, Holtzmann's [1983] theory suggests that the nonprofit response will not differ.

These findings suggest major policy implications for dealing with the Plans as well as for the ongoing debate concerning reform of health care. As Train [1991] has noted, a contestable market will exhibit competitive efficiencies without the need for regulatory intervention. At the present time many states pervasively regulate the Plans. This regulation extends to provider payments, Plan operating costs and premium rates. Our study suggests that steps designed to enhance the contestability of their markets may be a better response.

The Administration's now defunct health reform plan promulgated by the White House Domestic Policy Council [1993] appeared to be moving in the direction of treating health care as a regulated bilateral monopoly or a monopsony/oligopoly industry. A key element of the Clinton health reform called for establishment of Regional Health Alliances (a single alliance would serve each geographic area) and Corporate Health Alliances that would enter into contracts or franchise arrangements with a limited number of health plans. The health plans would provide care to beneficiaries pursuant to standard benefit packages in accordance with terms negotiated with the Alliances. A specific antitrust exemption encouraged states to replace competition with rate regulation.

Alternatively, a number of scholars and politicians advocate a single-payer system. This would produce a health care system based on a negotiation model rather than on a competition model. Pauly [1987] and Scherer and Ross [1990] have described the inefficiencies that often characterize negotiated arrangements. Numerous scholars including Olson [1981] have commented about the problems produced by health care regulation. The results of our study indicate that a contestable market (even a rationally expected or potentially contestable market) for health insurance characterized by the expectation of competition may actually produce a superior result.

Our conclusions are, as is usual in a study of this nature, limited. We considered only Blue Cross and Blue Shield Plans and, at that, only a sample of the Plans. Care must be exercised before generalizing its results to the entire market for health insurance. Detailed study of recent responses by all of the Plans as to their expectation of competition, as well as the response of commercial insurers and health maintenance organizations to the expectation of competition would be necessary before broader generalizations would be in order. However, prior to radical reform of the market for health insurance we might well consider undertaking such a study in order to determine whether continued evolution of contestability might produce a more optimal result than regulated monopoly power.

1. Pauly [1987] notes that market share is a necessary but not sufficient condition for market power. Ease of entry and exit can prevent a firm that holds a large market share from converting the share into market power.

2. The Plans have another advantage that is seldom discussed: many of the Plans act as payment intermediaries for the Medicare program. In that role the Plans review provider cost reports. The Plans can use this information to enforce monopsonist threat behavior.

3. See Frech [1980].

4. Pauly [1987] criticized this result.

5. Headen [1982] has questioned portions of these results on theoretic and empirical grounds.

6. However, the presumption that firms maximize profits contradicts this theory.

7. Lower administrative costs may not be beneficial per se. If the Plans incur greater administrative costs in order to hold down payments to providers (e.g., engage in cost containment activities), there may be a net gain. For example, in their review of the relationship between market share and cost control activities, several reviews have concluded that Plans with greater market shares were more likely to engage in cost control activities like utilization review as in Feldman and Greenberg [1981] and Scheffler, Sullivan and Ko [1991].

8. Managed care growth has been explosive. in 1980 there were few enrollees in managed care plans. The Health Insurance Association of America [1991] estimates that by the end of 1990 enrollment had reached sixty million.

9. As described in greater detail below, we measure market share in terms of the Plans' penetration rate, the proportion of the insured population served by the Plans.

10. Enough data was missing for six of the plans to prompt their deletion from the study's data base.

11. For the deflated log of administrative costs (Admin) the mean of the study sample was 8.187, the mean for the entire Plan population was 8.024. The difference between the means was, therefore, -0.163. The pooled variance was 1.78. We had 150 observations for the sample and 227 for the population. Accordingly, the t-value is 0.89, leading us to reject the hypothesis that the means are significantly different. Similarly, the t-value for the difference between the mean of the sample and the population for premiums is 0.06 and the t-value for the difference between means for provider payments is 0.04. In each case, we conclude that our sample was not significantly different from the population.

12. The index of medical care prices measures consumers' medical care cost increases. It fails to reflect increases paid by insurance companies and government payers that exhibit even greater inflationary trends. Compared to the general consumer price index the medical care price index is noisy and is biased upward. The consumer price index is equally problematic. We used them as the best of a less than optimal alternative. Given that our study only includes two years, the use of a deflator had little effect in any event.

13. Self reporting of each Plan's market and market share may provide a source of bias and variability. To the extent that the Plans' reported market shares are biased upward from actual (the usual tendency in self-reported market share) and contain additional variance, the study's results are weaker than a "true" measure would provide.

14. Wages are a large part of a hospital's costs and a large portion of the costs in practices (and hospitals) that employ physicians. Accordingly, we deflated hospital and physician wages by the medical care component of the CPI. We have also evaluated the model with hospital and physician wages deflated by the CPI. The results do not change.

15. Indeed, tests for endogeneity using Wu's [1974] test and Hausman's [1978] test produce the conclusion that market share is endogenous. The Hausman test is weak in the absence of really good instruments.

16. The regression's adjusted [R.sup.2] is 0.905 and the F-value is 275.2.

17. This may be due to the predictive power of the instruments used.

18. The predictive power of this equation in linear regression form is also good: [R.sup.2] = 0.96 and F = 1008.95.

19. As Pauly [1987] points out in his critique of Staten, welfare is maximized at an optimal level of payment to providers, not merely at a minimum level. At some point the monopsonist may go too far, resulting in the danger of welfare loss as providers fail to obtain necessary payment and patients fail to gain access to necessary treatment.

20. In linear regression form this equation is a good predictor of administrative costs with an adjusted [R.sup.2] of 0.88 and F-value of 205.91.

21. Again the equation predicts well: [R.sup.2] = 0.91 and R = 217.21.

22. It is possible that there are unmeasured advantages enjoyed by the Plans (industry experience, customer lists or other regulatory advantages) that increase premiums and are endogenous with premiums. If those variables are collinear with market share (Plan advantages would presumably lead to increased share), the coefficients for market share in the model would be biased. However, since the advantages would, presumably, lead to increased premiums, the coefficient for market share would be biased upward (in a positive direction). If we were able to correct for this bias, the finding that market share reduces payments, administrative costs and premiums would be even more pronounced.

REFERENCES

Adamache, Killard W., and Frank A. Sloan. "Competition between Non-Profit and For-Profit Health Insurers. Journal of Health Economics, December 1983, 225-43.

Baumol, William J., John C. Panzar, and Robert D. Willig. Contestable Markets and the Theory of Industrial Structure. San Diego: Harcourt, Brace, Jovanovich, 1982.

Blair, Roger D., Paul B. Ginsburg, and Ronald J. Vogel. "Blue Cross - Blue Shield Administrative Costs: A Study of Non-Profit Health Insurers." Economic Inquiry, June 1975, 237-51.

Busche, Kelly, and Peter Kennedy. "On Economic Belief in the Law of Small Numbers." Economic Inquiry, October 1984, 602-3.

Feldman, Roger, and Warren Greenberg. "Blue Cross Market Share, Economies of Scale and Cost Containment Efforts." Health Services Research, Summer 1981, 175-83.

-----. "The Relationship between the Blue Cross Market Share and the Blue Cross Discount" on Hospital Charges: Reply." Economic Inquiry, December 1982, 643-51.

Feldman, Roger, John Kralewski, and Bryan Dowd. "Health Maintenance Organizations: The Beginning or the End?" Health Services Research, June 1989, 191-211.

Feldstein, Paul J., and Thomas Wickizer. "The Rise in Private Health Insurance Premiums, 1985-91: An Exploratory Analysis." Unpublished Working Paper, University of California, Irvine, 1992.

Frank, Richard G., and W. P. Welch. "The Competitive Effect of HMOs: A Review of the Evidence." Inquiry, Summer 1985, 148-61.

Frech, H. E., III. "Blue Cross, Blue Shield and Health Care Costs: A Review of the Evidence," in National Health Insurance: What Now, What Later, What Never, edited by Mark V. Pauly. Washington: American Enterprise Institute, 1980, 250-63.

-----. "Monopoly in Health Insurance: The Economics of Kartell v. Blue Shield of Massachusetts," in Health Care in America, edited by H. E. Frech III. San Francisco: Pacific Research Institute, 1988, 293-322.

Frech, H. E., III, and Paul B. Ginsburg. "Imposed Health Insurance in Monopolistic Markets: A Theoretic Analysis." Economic Inquiry, March 1975, 55-70.

-----. "Competition Among Health Insurers," in Competition in the Health Care Sector, edited by W. Greenberg. Washington: Federal Trade Commission, 1978, 210-37. Reprinted, Germantown: Aspen Systems, Inc., 1978.

-----. "Competition Among Health Insurers Revisited." Journal of Health Politics, Policy and Law, 13(2), 1988, 279-91. Reprinted in Competition in the Health Care Sector: Ten Years Later, edited by W. Greenberg. Durham: Duke University Press, 1988, 57-70.

Goldberg, Laurence G., and Warren Greenberg. "The Dominant Firm in Health Insurance." Social Science and Medicine, 20(7), 1985, 719-24.

Hausman, J. A. "Specification Tests in Econometrics." Econometrica, November 1978, 1251-71.

Headen, Alvin E., Jr., "The Relationship between the Blue Cross Market Share and the Blue Cross 'Discount' on Hospital Charges." Economic Inquiry, December 1982, 634-47.

Health Insurance Association of America. Sourcebook of Health Insurance Data. Washington, D.C.: Health Insurance Association of America, 1991.

Holtzmann, A. G. "A Theory of Non-Profit Firms." Economica, November 1983, 439-49.

Kass, David I., and Paul A. Pautler. "The Administrative Costs of Non-Profit Insurers." Economic Inquiry, July 1981, 515-21.

-----. "On Economic Belief in the Law of Small Numbers: A Reply." Economic Inquiry, October 1984, 604-5.

Marmor, Theodore R. "New York's Blue Cross and Blue Shield, 1934-1990: The Complicated Politics of Nonprofit Regulation." Journal of Health Politics, Policy and Law, Winter 1991, 761-92.

Olson, Mancur. A New Approach to the Economics of Health Care. Washington: American Enterprise Institute for Public Policy Research, 1981.

Pauly, Mark V. "Competition in Health Insurance Markets." Law and Contemporary Problems, Spring 1988a, 237-71.

-----. "Market Share/Market Power Revisited A New Test for an Old Theory: Reply." Journal of Health Economics, March 1988b, 85-87.

-----. "Monopsony Power in Health Insurance: Thinking Straight while Standing on Your Head." Journal of Health Economics, March 1987, 73-81.

Scheffler, Richard M., Sean D. Sullivan, and Timothy Haochung Ko. "The Impact of Blue Cross and Blue Shield Plan Management Utilization Programs, 1980-88." Inquiry, Fall 1991, 263-75.

Scherer, Frederick M., and David Ross. Industrial Market, Structure and Economic Performance, 3rd ed. Boston: Houghton Mifflin Co., 1990.

Sheperd, William G. "Contestability vs. Competition." American Economic Review, September 1984, 572-87.

Staten, Michael, William Dunkelberg, and John Umbeck. "Market Share and the Illusion of Power: Can Blue Cross Force Hospitals to Discount?" Journal of Health Economics, March 1987, 43-58.

Staten, Michael, John Umbeck, and William Dunkelberg. "Market Share/Market Power Revisited: A New Test for an Old Theory." Journal of Health Economics, March 1988, 73-83.

Train, Kenneth E. Optimal Regulation: The Economic Theory of Natural Monopoly. Cambridge: The MIT Press, 1991.

U.S. Department of Commerce, Bureau of the Census. Statistical Abstract of the U.S., 112th ed. Washington: U.S. Government Printing Office, 1992, table number 151.

Vogel, Ronald J. "The Effects of Taxation on the Differential Efficiency of Nonprofit Health Insurance." Economic Inquiry, October 1977, 605-9.

White House Domestic Policy Council. The President's Health Security Plan. New York: Times Books, 1993.

Woolley, J. Michael. "The Competitive Effects of Horizontal Mergers in the Hospital Industry." Journal of Health Economics, December 1983, 271-91.

Wu, De-min. "Alternative Tests of Independence between Stochastic Regressors and Disturbances: Finite Sample Results." Econometrica, May 1974, 529-46.

Stephen Earl Foreman is Assistant Professor of Health Policy and Administration at the Pennsylvania State University, University Park, Penn., John Anderson Wilson is Senior Research Analyst, The CNA Corporation, Alexandria, Va., and Richard M. Scheffler is Professor of Health Economics and Public Policy, University of California, Berkeley, Berkeley, Calif. The authors wish to thank H. E. Frech III, Theodore Keeler, coeditor Rodney T. Smith and an anonymous referee for their helpful review and comments.