Sympathy for the Insurance Companies
When assessing the effects of market concentration on competition, antitrust authorities look not just at how many competitors there are in a market, but also how they compete. In one standard model, called Cournot competition, producers set levels of output, and prices adjust in relation to supply. In another model, called Bertrand competition, producers set prices, and produce quantities sufficient to meet demand at the prices they set. In pure Cournot competition, prices vary substantially based on the number of competitors in the market. In a Cournot duopoly (a market with two competitors setting output), the producers will between them extract half of a monopoly profit. But pure Bertrand competition pushes prices down toward marginal costs no matter how many competitors there are.
While the differences between Cournot and Bertrand competition are fundamental concepts of antitrust economics, they seem to have been lost in the race to pin responsibility for rising premiums on concentration in the insurance industry. While the details of competition among insurers is complex and does not strictly follow any simple model, at a basic level they appear to engage in Bertrand competition for the business of insureds, i.e., they set premiums, and write as many policies as corresponding demand requires. In this case, we should expect them to compete prices down toward their costs regardless of market concentration. And we in fact see that insurers’ profit margins are relatively modest despite a longstanding trend of consolidation in the industry.
Providers, on the other hand, seem to engage (again, at a basic level) in Cournot competition. They produce a given amount of capacity — building hospital beds, hiring physicians — and then negotiate with insurers over the price to use it. If this is correct, then concentration in the provider market should have a significant impact on the prices that providers can charge insurers, and that insurers ultimately pass on to insureds. By the same token, concentration in the insurer market should allow insurers to pay lower prices to providers because they are also engaging in Cournot competition on the buy side — they have a certain amount of demand that is set by the needs of their insureds, and they bargain with providers over the price of fulfilling it. But because insurers are engaged in Bertrand competition on the sell side, they will tend to pass these savings on to their insureds notwithstanding their market concentration.
In this simple model at least, the ideal situation for insureds is to have a concentrated insurance market that approaches Bertrand duopoly on the sell side and Cournot duopsony (two buyers who set demand and take prices) on the buy side, with a competitive market for providers. Of course, there are complexities — such as product differentiation, switching costs, and homing asymmetries — that can both attenuate and amplify the effect of this basic market structure for consumers, and I hope to address some of them in later posts. But other things being equal, the implication is clear: concentration in the insurance industry does not necessarily harm and may well benefit consumers, while concentration among providers is a prescription for higher costs.
Health Care Costs and Market-Theory: Provider Network Externality
Sick of reading posts about market theory and health care costs? If so, skip this one and read this alternative post from the archives. It has nothing to do with health care costs or market theory.
Meanwhile, I’m obsessed with what market theory can tell us about health care costs. It’s a good thing because it’s my job (or part of it). Anyway, let’s revisit my favorite graph, which you might recall from prior posts on this topic. It illustrates how I think insurer market concentration relates to premiums (or health care costs in general), holding health care provider concentration fixed (the curve shifts up (down) for higher (lower) provider concentration).
While this graph explains a lot, it doesn’t explain everything. In particular it abstracts away the role played by what the insurers actually provide for the premium charged. The curve is for the average standardized premium, which means all the variations in benefit packages offered by insurers are taken into account or controlled for, adjusting the premium accordingly.
Of course benefits matter for enrollment, which is the source of insurer market power. The greater number of enrollees, the greater the power of the insurer. Most benefits are of the form “such-and-such is covered for a copayment of thus-and-so.” But one benefit is of a different flavor than all the rest. That benefit is the provider network.
The provider network is the set of providers to which policyholders have access through the insurer. Go outside that network and you’re not covered; you have to pay for the service entirely out of pocket. Of course the establishment of the provider network of an insurer is relevant to the market power of providers and the insurer. The negotiation over rates–payments from the insurer to providers for their services–includes the possibility of exclusion from the network. For a sufficient rate of pay providers will want to be in the network to have access to the volume of patients the insurer can deliver. At the same time, the insurer is only willing to pay so much to a given provider: more for a dominant provider, less for a small player.
What distinguishes the provider network from other benefits is that it exerts a positive externality on policyholders. This type of inter-group network externality is the defining feature of a two-sided market. The greater the number of providers in an insurer’s network, the more a consumer will pay that insurer for access. All other things equal, consumers like access to more providers. They’ll pay a higher premium for such access.
This (two-sided) market-theoretic aspect of health care costs is not reflected in the figure above, nor is it reflected in most theoretical motivation or empirical specification of analysis of health care markets in the literature. This is not a surprise. The foundational work on this is brand new.
Antitrust After Health Reform: Are Economists Ready?
In my post yesterday on the 2009 AcademyHealth Annual Research Meeting I mentioned accountable care organizations and antitrust. There is a relationship between the two and implications for health economics.
First of all, what is an accountable care organization? An ACO is a group of health care providers and institutions that are collectively responsible for (and held accountable to measures of) the health of a population. An ACO has an organizational structure that permits the encouragement of improvements in quality and lower costs through payment incentives.
That’s a little hard to grasp so let’s get concrete. The Veterans Health Administration (VHA) is the largest ACO in the U.S. It is a clearly defined set of providers and facilities that are responsible for the care of groups of veterans enrolled in the VHA. Quality performance measurement and low cost are routine within the VHA and it is not hard to imagine a variety of possible payment systems that would (or currently do) promote quality improvement and/or cost efficiency. Other currently existing ACOs include Kaiser Permanente, Geisinger Health Systems, and the Mayo Clinic. In one way or another, from a collective pool of revenue physicians in an ACO that are associated with higher quality get paid more than those associated with lower quality.
(For further reading on ACOs see Creating Accountable Care Organizations: The Extended Hospital Medical Staff (Fisher et al, Health Affairs, December 2006) and the January 27, 2009 press release from The Dartmouth Institute for Health Policy and Clinical Practice and the Engelberg Center for Health Care Reform at Brookings.)
Almost by definition an ACO implies a degree of collusion between providers. After all, under an ACO structure a set of providers must agree on how to distribute a bundled payment. Thus, if the U.S. health care system moves toward an ACO model we will see greater provider consolidation, all other things being equal. Provider groups’ greater market power to negotiate higher payments will be a countervailing force against the promise of the ACO model to provide higher quality care at lower cost. Therefore, there would seem to be a set of antitrust issues associated with ACOs. I’m not an expert in such matters but I am willing to bet that considerations of antitrust will either be explicit in legislation that defines and authorizes ACOs or be raised latter in legal battles.
While ACOs suggest greater market power on the provider side, there has been considerable rhetoric about increased market power on the insurer side. One of the purposes of an insurer is to bargain provider payments downward. The greater the insurer’s market power (i.e. the more policyholders it represents) the better prices it can negotiate, all other things being equal. This, in fact, is one of the functions of a public plan and one of the reasons Medicare, the VHA, and other large purchasers can negotiate (or set) relatively low prices.
Recent publications suggest a high degree of market concentration among health insurers already exists, at least in certain markets (see Robinson, Consolidation and the Transformation of Competition in Health Insurance, Health Affairs 23(6) and Dafny, Are Health Insurance Markets Competitive?, NBER working paper 14572). Will health reform, should we get any, lead to a more or less competitive insurance market? More critically, what degree of market power among insurers is ideal? Greater market power may provide leverage to negotiate lower provider prices but it also makes it less likely the insurer will pass those lower prices on to consumers through lower premiums.
There must be an optimum degree of insurer power given the degree of provider power, one that maximizes benefits to the consumer. This is a complicated two-sided market problem. So far I have not found much evidence that economists are able to tackle such a complex problem (I’m still looking). Platform competition in two-sided markets is hard to model and this area of specialty is relatively new.
I worry that some issues will be settled and constraints imposed via antitrust lawsuits before the economics community can makes sense of things. Consumers will likely be harmed if courts do not have the tools necessary to make rational judgments. It wouldn’t be the first time.
Two-Sided Markets, Part II: Health Insurance
This post is cross-posted on The Finance Buff.
This is the second of a two-post series on two-sided markets. The first post introduced basic concepts that I will apply to the discussion of health insurance markets below. I am aware of only a handful of studies that make explicit use of two-sided market theory in analysis of health insurance markets. There are many other studies that make ad hoc reference to elements of its two-sided nature.
Two very similar papers by Bardey and Rochet (this and that), consider health insurance plans as platforms serving two groups: policyholders and health care providers. For this to be a genuine two-sided market framework we must be able to identify either (1) inter-group network externalities or (2) a sensitivity of transaction volume to how total price is split between the groups. These were the two essentially equivalent ways of defining a two-sided market discussed in my prior post.
That the first definition (inter-group network externalities) applies is not hard to see. In general, health insurance policyholders prefer greater access to more providers. Thus, the greater the number of providers contracting with the insurance plan the more valuable that plan will be to policyholders. That’s a positive inter-group network externality. Likewise, providers prefer greater patient volume. Thus, all other things equal, more policyholders make contracting providers better off, another positive inter-group network externality.
It is not immediately clear how to see that the second definition of a two-sided market (transaction volume sensitive to price allocation) also applies. What are the prices? What are the transactions? Policyholders pay a price: the premium. Do providers pay a price to contract with the insurer? They do. The price is the per-service discount they’re willing to provide the insurer for the expected patient volume.
In the Bardey-Rochet model, a transaction is a unit of provider service. A sick individual consumes one unit of service and a well individual consumes none. The transaction price is the sum of premium and provider discount. Transaction volume is sensitive to the tradeoff between premium and discount. At the extreme of a zero premium, the discount would need to be so high that no providers would participate and transactions would go to zero. (This is similar in spirit to the pricing tradeoff faced by Dude in my prior post.) Very briefly (because it is not the focus of this post), Bardey and Rochet go on to use a two-sided market set-up to show that adverse selection can lead to higher, not lower, insurer profits due to the negotiating leverage the additional health care utilization provides with respect to provider discounts. Such a conclusion cannot be drawn with a one-sided view of the market as it is an inter-group phenomenon.
The other paper I am aware of that explicitly uses two-sided market concepts in discussing the health insurance market is by Howell. Here a completely different type of two-sidedness is introduced. Insurance policyholders consist of two groups. One group consists of those who are healthy and not receiving any health care services, and the other includes those who are sick and are receiving health services (there are no preventative services in this model).
The only way this set-up can be viewed as a two-sided market is if one can identify the transaction between the sick and healthy and if one can show that the transaction volume is sensitive to relative prices or that there are inter-group network externalities. So, where’s the transaction between the healthy and the sick policyholders? Howell’s argument is that one can interpret the health insurer’s role as balancing the interests of the healthy and the sick. In effect, the collective risk sharing arrangement establishes implicit contracts between the groups whereby the healthy provide financial resources and the sick spend them. A healthy individual is willing to enter such an implicit contract because there is a non-zero probability he will fall ill.
But this doesn’t exactly pin down the transaction in a way that permits enumeration. My own interpretation is that the transaction is the health insurance policy itself, interpreted as a contract between the current (presumed) healthy state and the potential future unhealthy state.
From this perspective it is not hard to see transaction sensitivity to allocation of total price between groups. If the sick pay relatively more (e.g. higher copayments relative to premium) then relatively more healthy will participate leading to a higher number of transactions (sold policies). If the healthy pay more (e.g. higher premiums relative to copayments) then relatively fewer healthy will participate, lowering the transaction volume. Alternatively, the inter-group network externalities are also easy to see. A larger group of healthy participants leads to a lower premium for all policyholders (favorable selection) while a greater number of sick raises it (adverse selection).
Howell goes on to (rhetorically, not mathematically) embed this two-sided market model in the one involving policyholders and providers discussed previously. Another two-sided market can be found in the relation between sponsors who provide insurance subsidies (e.g. employers or the public) and policyholders. Howell calls this monstrously complex tangle of competing interests a “four-sided” market.
It is beyond the scope of this post to lay out all the price sensitivities and inter-group network externalities for the four-sided model. It is worth noting, however, that such a model is something like what actually exists. That is, health insurers mediate an enormous number of competing interests many of which are opaque to one-sided analysis.
Two-Sided Markets, Part I: Gender-Based Price Discrimination at the Nightclub
This post originally appeared on The Finance Buff.
This is the first of a two-post series on two-sided markets, a relatively new idea in the economic theory of markets (developed circa 2000). Two-sided market theory more realistically models certain types of markets. In such markets empirical evidence is often inconsistent with conventional (one-sided) market theory but can be shown to reflect rational firm behavior with a two-sided perspective. This post explains the term and related concepts. The second post will explore the extent to which the idea can be applied to the market for health insurance.
We begin with a bit of plausible fiction. Dude, the proprietor of the new nightclub Dude’s Club, charges men and women the same price (parity pricing) to enter his establishment. One day, in going over his records, he noticed something. On the rare nights for which the numbers of women and men attending were roughly equal, he made a tremendous profit in bar sales. But typically the number of women was far below the number of men and he lost money.
Being a profit-maximizing sort of guy, Dude decided to try to attract more women to his club so that the typical night became profitable. After experimenting with the level of entry fee, he had a winning idea: let women enter for free and charge men twice the original price. This generated a gender mix that was roughly equal. Unbeknownst to Dude, he had discovered that his was a two-sided market. (In his (quite accessible) 2004 Review of Network Economics paper titled “One-sided Logic in Two-sided Markets“, Julian Wright also uses the heterosexual nightclub market to explore concepts in two-sided markets. He then proceeds to apply the concepts to credit card schemes.)
A two-sided market is one in which the volume of transactions is sensitive to how the total transaction price is allocated between two distinct groups of market participants (see definition 1 of this paper by Rochet and Tirole). Dude discovered that the true nature of his business was to match one-for-one members of one group of market participants (men) with those of another (women). Further, he discovered that the number of matches (transactions) was sensitive to how the total transaction price (the total price for one match, i.e. the entry price for one man and one woman) was allocated across genders. Specifically, holding the total price of a match constant, if women’s price decreased while men’s price increased then more women showed up relative to men and matches increased as did his profit.
To understand the phenomenon more deeply, Dude conducted some market studies. Using his background in polling research and applied statistics, Dude conducted and analyzed exit interviews. He learned that men’s satisfaction with their Dude’s Club experience increased as the number of women in attendance increased (holding the number of men constant). Conversely, women’s satisfaction increased as the number of men in attendance decreased (holding the number of women constant). Since the number of men was never less than the number of women these conclusions were consistent with the idea that overall satisfaction (and Dude’s profit) was maximized when the number of men and women were equal. Apparently, satisfied nightclub attendees buy more booze.
That the satisfaction of individuals in one group (e.g. men) depends on the number of individuals in the other group (e.g. women) is an example of an inter-group network effect (or network externality). A network externality is a favorable or unfavorable outcome that is related to the total number of participants. That your broadband service is frustratingly slower if more users are online is another example of a network externality. This is an example of a same-group externality because we’re only speaking of one group, all broadband users. The “inter-group” nature of the externality discovered by Dude is that the externality operates between groups: men’s numbers affect women’s satisfaction and vice versa. (There are likely also same-group externalities at play at Dude’s Club but they’re not the focus here.)
It is the very existence of these inter-group network externalities at Dude’s Club that gave rise to the fact that transaction volume was sensitive to relative group pricing. In fact, the existence of inter-group externalities is another definition of a two-sided market (see, again, Rochet and Tirole, Section 5.1.4, p. 21). Relative to the old parity pricing system, if women pay less than men to enter Dude’s Club, they are both more willing to attend and have a better time. While men are relatively less likely to attend as their price increases, this effect is somewhat offset by the anticipation of greater satisfaction with a 50-50 gender mix experience. Dude effectively “priced away” the negative inter-group externalities caused by an imbalance between men and women.
Dude’s new pricing scheme puts him in the role as mediator of a subsidy. Men pay twice the old rate and cross-subsidize women who pay nothing. Two-sided markets often involve cross-subsidies but subsidization is not a necessary feature. Some obvious two-sided markets with subsidies include broadcast TV, newspapers, and Google services (ad revenue subsidizes viewers, readers, and users, respectively). Many other examples are found in this Rochet and Tirole paper. Not mentioned there are health insurance plans. In at least two non-obvious ways one can view the market for health insurance as a two-sided one. That will be the subject of the next post in this series.
Subscribe
