Recently, Peter Orszag of Bloomberg View published two columns on Medicare reform. In his first column, which we’ve discussed previously, he offers a forceful critique of Ryan’s 2011 premium support proposal, drawing on a CBO report which found that its premium support payments would not keep up with the cost of providing the current Medicare benefit. In his second and most recent column, he focuses on the 2012 Wyden-Ryan plan, which is modeled on the Domenici-Rivlin plan devised by the Bipartisan Policy Center.
This is a long post, so I’ve put most of it below the fold. I’ve written it to raise questions rather than to offer conclusive answers, and (as always) I’m open to other thoughts and interpretations.
Ryan 2011 vs. Ryan 2012
In Orszag’s view, the differences between the two proposals are trivial:
The difference in the new version of the Ryan plan is that traditional Medicare would coexist with private plans. To suggest that this would change everything is to make an odd argument: Moving entirely to private competition would not generate big savings, but moving partially would.
Note that Orszag did not say that this is the “only” difference in the new version of the Ryan plan, as that would not be true. There are at least two other important differences. The first is that the growth rate for premium support is subject to a higher cap of GDP +0.5%, which is identical to the global budget proposed by the Obama White House. The second is that the 2012 proposal uses competitive pricing to determine the level of premium support, i.e., the capitation payment paid to insurance plans, rather than a fixed administrative formula.
Given that Orszag’s previous column was very concerned about the growth rate for premium support payments, this seems at least somewhat important. It undermines the argument that the Ryan plan shifts risks from taxpayers to beneficiaries. Of course, one could believe that the Obama administration and Paul Ryan are overly optimistic regarding the potential for cost containment, but both are working from the same global budget. Regardless, Ryan embraced a Medicare reform proposal in 2012 that was considerably more expensive than his 2011 proposal.
As for the odd argument Orszag invokes, there appears to be some confusion. The 2012 plan explicitly generates less in the way of savings than the 2011 plan, as the 2012 plan generates savings in a way that risked subjecting Medicare beneficiaries to more risk. In moving from a defined contribution system to what the Bipartisan Policy Center has called “competitive bidding on a budget,” a hybrid of defined benefit and defined contribution, Ryan has decided to accept a lower level of savings for taxpayers in order to better protect the interests of Medicare beneficiaries. That is, we know that moving entirely to a defined contribution with a tightly constrained budget per Ryan’s 2011 plan will generate big savings for taxpayers, as Orszag acknowledges in his previous column, e.g.:
CBO’s analysis, not surprisingly, confirms that federal expenditures under the Ryan Medicare plan would be reduced sharply. Federal payments for a typical beneficiary by 2030, for example, would be more than 20 percent lower than current projections.
It happens that Paul Ryan decided to allow for the possibility of smaller savings by embracing a less tightly constrained budget, in part due to the concerns raised by the CBO analysis of his 2011 plan. Preserving Medicare FFS is (mostly) orthogonal to the question of generating savings.
So as far as I can tell, neither Paul Ryan nor the Bipartisan Policy Center is making the odd argument Peter Orszag has in mind. Rather, the argument is that preserving Medicare FFS as a “public option” will serve a number of functions, which we will revisit at greater length. Briefly, there is at least some reason to believe that Medicare FFS will prove more cost-effective than private alternatives in regions in which medical providers have a great deal of market power. Over time, we can imagine that vigorous antitrust enforcement, the emergence of hospital chains and medical tourism, and other business model innovations might change this dynamic. But in the short to medium term, the most effective way to restrain the power of providers in some regions, particularly rural regions, might be for Medicare FFS to use its leverage on behalf of Medicare beneficiaries.
The lessons of Medicare Advantage
The president’s former budget director continues by drawing on the experience of the Medicare Advantage program:
In any case, we already have a system similar in some ways to the revised Ryan proposal. Almost 30 percent of Medicare beneficiaries are covered by private insurers through the Medicare Advantage program, which exists alongside traditional Medicare. So what can we learn about the potential impact of the Ryan proposal from our experience to date with Medicare Advantage plans?
There is, however, a crucial difference between competition between Medicare FFS and Medicare Advantage plans at present and under a system of competitive pricing. At present, payments to Medicare Advantage plans are based on costs in Medicare FFS. For example, the cost of treating a diabetic patient in Medicare FFS informs the payment Medicare will provide to a Medicare Advantage HMO to care for a diabetic patient. The obvious problem, of course, is that Medicare FFS in a given region might do a very inefficient job of caring for a diabetic patient, and so a Medicare Advantage HMO that does a relatively good job of caring for a diabetic patient will, if compensated on the basis of costs in the FFS sector, make a quite large profit.
Competitive pricing, in contrast, will encourage Medicare Advantage HMOs to offer bids based on their own costs rather than costs in the FFS sector, and these bids will form the basis of their capitation payment.
Orszag acknowledges that Medicare Advantage plans have been bidding below Medicare FFS to provide the standard Medicare benefit. Yet he argues that these lower bids are a mirage on the grounds that they reflect adverse selection rather than greater efficiency.
In Orszag’s view, Medicare Advantage plans are effectively cherry picking “good risks,” i.e., Medicare beneficiaries who will generate low expenditures relative to their risk score, while Medicare FFS has been left with “bad risks,” i.e., Medicare beneficiaries who will generate high costs relative to their risk score.
Much depends on how well the Medicare system calibrates its risk adjustment system in the first place. In theory, a well-designed risk score will leave relatively little room for this kind of gaming, and differences in the underlying cost of delivering care (not the price, which depends on the benchmark used to determine the capitation payment) will reflect underlying differences in efficiency. Orszag maintains that the risk adjustment system used by Medicare is not working well.
To see why, imagine two beneficiaries. One has medical expenses amounting to $150 and the other, $50. The average cost is $100. Now imagine that a private plan bids $90 to cover beneficiaries, so it looks to be about 10 percent cheaper than traditional Medicare. That plan, however, while it is designed to be very attractive to the $50 beneficiary, isn’t appealing to the $150 one, so that person stays in traditional Medicare.
One potential complication with Orszag’s story is that competitive pricing is designed to ensure that all plans are offering benefit packages with the same actuarial value, but of course this doesn’t preclude the possibility that some will be structured in ways that are more appealing to good risks than to bad risks, e.g., very sick people might be reluctant to switch medical providers, and a Medicare Advantage HMO or PPO plan might not offer the same suite of providers as Medicare FFS.
The result is that total costs rise from $200 ($150 for the expensive beneficiary plus $50 for the inexpensive one) to $240 ($150 for the expensive beneficiary plus $90 for the inexpensive one). So even though the plan “looks” like it saves money, it doesn’t. It overpays to cover the $50 beneficiary. (And that’s not even taking into account another factor: that if Medicare’s purchasing power is splintered, its negotiating leverage will be reduced. So the prices it must pay could rise. That would drive up the cost of covering the $150 beneficiary, pushing the total above $240.)
This is an intriguing scenario.
Price competition is a dynamic process
Let’s think through a slightly different scenario, in which we assume that there are two time periods, Year 0 and Year 1. Imagine that the beneficiary who has medical expenses amounting to $150 under Medicare FFS could be treated for $90 by a Medicare Advantage HMO, as the latter provides more efficient, better-integrated care. This implies that Medicare FFS has “wasted” $60 in Year 0. In the absence of price competition, however, Medicare FFS has very weak incentives to spend less in Year 1, as it is getting paid $150 regardless, in part due to the political power of local medical providers.
Indeed, the incentives may well run in the opposite direction, as Medicare beneficiaries are relatively insulated from the cost of wasteful spending of this kind while local medical providers find that it pads their margins, allowing them to increase compensation for medical professionals, either through higher wages and more generous benefits or through featherbedding, and to invest in new medical technologies that may or may not prove cost-effective.
Premium support is designed to strengthen price competition, as beneficiaries who choose Medicare FFS when it bids more than the second-lowest bidder will have to pay the difference out-of-pocket. This is part of why it is very important that Medicare FFS be given the flexibility it needs to increase efficiency.
Now let’s say the private plan (A) that bid $90 to cover a beneficiary who in fact cost $50 makes a substantial profit in Year 0. This might leads its competitors B and C to bid less than $90 in Year 1 in an effort to attract more Medicare beneficiaries. If A is undercut by B and C in Year 1, it will presumably react in kind by improving its underlying efficiency or going out of business. This dynamic process might eventually yield a change in underlying medical expenses, i.e., a beneficiary with medical expenses of $50 in Year 0 might generate medical expenses of $45 or $35 in Year (0 + X) as providers rely more heavily on nurse practitioners, physician assistants, and other medical professionals, and as they combine technology and labor in productivity-enhancing ways.
To be sure, this scenario depends on the creation of opportunities for entry. This is very difficult in light of the cartel-like power of physicians, which is why supply-side reforms are an important and neglected part of the larger challenge of encouraging greater efficiency in the health system.
The dangers of adverse selection
Rather than emphasize the dynamic nature of price competition, Orszag focuses on the dangers of adverse selection:
To counteract the selection effect on Medicare Advantage plans, a risk-adjustment process is used. The system has improved over time, but evidence suggests it still does not work very well. The models used to adjust payments can account for only about 10 percent of subsequent cost variation; even the most optimistic estimates suggest they could account for only 20 percent to 25 percent of the variation. This gap allows plans that can better predict beneficiary costs to game the system by selecting beneficiaries who are expected to cost much less than their risk-adjusted payments. (Plans do not always want the least-expensive beneficiaries, but rather those who are the least expensive compared with their risk-adjusted payment. The implication is the same, though: Plans can beat the risk adjustment, and be overpaid.) [Emphasis added]
To Orszag’s great credit, he is willing to acknowledge that the risk-adjustment process has improved considerably over time, which should be encouraging to those who believe that the adverse selection problem can be alleviated over time.
He does, however, suggest that the selection effect in Medicare Advantage at present is “huge”:
How big is this selection effect in Medicare Advantage? The evidence suggests it’s huge. The most careful analysis was reported in a 2011 National Bureau of Economic Research paper by Jason Brown of the Treasury Department, Mark Duggan of the University of Pennsylvania, Ilyana Kuziemko of Princeton and William Woolston of Stanford University. In 2006, Medicare Advantage plans were overpaid by more than $3,000 per beneficiary because they were able to select beneficiaries who cost less than their risk-adjusted payments. About $1,000 of that overpayment reflects what the plans were paid, rather than what they bid. So relative to their bids, the plans were overpaid by $2,000 per beneficiary — or roughly 25 percent of the bid, on average.
That overpayment rate, furthermore, is likely to be higher for the second-lowest bid in each county, and it is therefore likely to be larger than the much-touted 9 percent discount, which doesn’t appropriately account for the selection effects.
It is worth noting that Brown et al. have constructed a model that rests on a number of assumptions. The following is drawn from the as yet unpublished Brown et al. working paper:
While an MA plan must be open at the same price to all individuals in the plan’s geographic area of operation, the model assumes that, as shown in earlier work, plans have at least some scope to selectively enroll individuals, whether through targeted advertising, designing benefit and cost-sharing parameters that appeal to some groups but not others, or more readily facilitating the enrollment of certain individuals. As such, we do not model the consumer side of the enrollment decision and instead focus on firms’ decision to incur the costs associated with these screening activities in return for enrolling a selected subsample from the Medicare population. In our model, firms have an incentive to target individuals for whom the difference between capitation payments and expected costs is the greatest, and risk adjustment changes this set of individuals by changing how capitation payments are calculated.
There is a compelling logic to the idea that good risks will be attracted to Medicare Advantage HMOs while bad risks will be attracted to Medicare FFS. Sicker people are less price elastic, as they are intensive consumers of medical services and they tend to want as much choice as possible. Healthier people, in contrast, are more price elastic, as they consume little in the way of medical services and they mainly care about paying as little as possible for coverage.
Over time, however, there is reason to believe that the risk pool of Medicare Advantage HMOs and Medicare FFS would start to look more and more alike, as Bryan Dowd has explained. To understand why, consider that we are dealing with an iterative game. If there are ten beneficiaries, who will cost an insurer between $1 and $10, the Medicare Advantage HMO might start out with the patient who costs $1, but over time it will attract those who cost $2, $3, and so on in ascending order. So the average cost of Medicare FFS will shoot up over time — but so will the average cost of the Medicare Advantage HMO. The average across the two groups, however, will remain the same until the Medicare Advantage HMO has acquired even the most high-cost Medicare beneficiaries. The key question is whether the Medicare Advantage HMO can create real efficiencies.
The Brown et al. premise is, crudely, that the Medicare Advantage HMO in our example will stop at beneficiaries 1, 2, and 3 and leave Medicare FFS with beneficiaries 4, 5, 6, 7, 8, 9, and 10, with average costs of $2 and $7 respectively. Yet the Medicare Advantage HMO will be paid a number much closer to $7 than $2. The beauty of competitive pricing is that the bigger the wedge between $7 and $2, the greater the incentive for entry to compete away these large profits.
Note that the parameters that Brown et al. see as driving selection are subject to regulation. The active recruitment of good risks is already illegal, but of course the Medicare system could go further. For example, Congress could restrict or even ban targeted advertising, limiting Medicare beneficiaries to comparing plans on the basis of information provided by regional Medicare exchanges.
Limiting the ability of Medicare Advantage plans to design benefit and cost-sharing parameters might be problematic, as it might impair the ability of the plans to embrace productivity-enhancing, and consumer-welfare-enhancing, innovations. But if there is some reason to believe that differences in cost-sharing parameters are exacerbating adverse selection, Congress will have a much stronger political incentive to address the problem under competitive pricing than it does at present, when the benchmark is based on the cost of Medicare FFS.
Why is that? Consider the following line from Brown et al.:
If risk adjustment works perfectly–so that in expectation capitation payments are equal to an individual’s FFS costs–then whether an enrollee switches between FFS and MA should have no effect on his total Medicare expenditure levels.
That is, if risk adjustment works perfectly, Medicare Advantage HMOs will capture any cost differences with Medicare FFS as profits, as capitation payments will be tied to the individual’s Medicare FFS costs. Under competitive pricing, in contrast, capitation payments or premium support will be tied to the second-lowest bid. One of the virtues of tying capitation payments to the second-lowest bid, incidentally, is that it encourages plans to bid lower than they would if they were tied to the lowest bid, as tying payments to the lowest bid would increase anxiety about underbidding.
And so competitive pricing should help at least alleviate the overpayment problem Orszag identifies, as he implicitly suggests. In Orszag’s 2006 example, competitive pricing immediately cuts overpayments by one-third, hardly a trivial sum. Assuming that Orszag does not intend to abolish the Medicare Advantage program — that could be his larger goal, but he never makes it explicit — it is hard to think of many other reforms that would yield such significant savings so quickly.
And of course systematic overpayment represents, as we noted above, a profit-making opportunity for new entrants. That is, if Medicare Advantage plans are generating enormous profits by virtue of savvy risk selection, new Medicare Advantage plans will have a strong incentive to offer lower bids.
The fact that Medicare Advantage “better predict beneficiary costs” at present partly reflects higher “coding intensity,” as Brown et al. observe. Medicare Advantage plans are aware of the importance of the risk score of Medicare beneficiaries, and so they tend to take more careful and detailed records concerning medical conditions and other factors that might yield a higher risk score. This presumably reflects the fact that Medicare beneficiaries prefer not to switch from plan to plan every year, and that a higher risk score will yield a higher risk-adjusted premium.
One interpretation of higher “coding intensity” is that Medicare Advantage plans are essentially gaming the system. Another interpretation, however, is that Medicare FFS is relatively inattentive to genuine risk factors, in part due to the fact that price competition is muted. In an environment in which Medicare FFS faces more vigorous price competition, it is easy to imagine that its “coding intensity” will increase, and indeed that attentiveness to the needs of Medicare beneficiaries along other dimensions will also increase.
So what are we arguing about?
Orszag concludes as follows:
The bottom line is that, if anything, Medicare Advantage bids are above, not below, traditional Medicare — once you do the analysis correctly, on an apples-to-apples basis. So regardless of whether you use the CBO analysis of Ryan 1.0, or the evidence to date with Medicare Advantage to analyze Ryan 2.0, the conclusion is the same.
We don’t want to put all our chips down on the health-care competition tooth fairy.
If Orszag is correct and a more accurate risk-adjustment process would eliminate the apparent cost advantage of Medicare Advantage, the implication is that the problem can essentially be solved applying a mathematical discount designed to account for hidden risk selection. Medicare FFS will consistently be the lowest bidder under competitive pricing once it becomes clear that risk scores have been gamed. Medicare Advantage plans or Medicare FFS might continue to attract beneficiaries by offering features they are willing to pay for out-of-pocket, but the resulting cost differential will be borne by beneficiaries. It is thus hard to see what we lose by, in effect, ending overpayments to Medicare Advantage plans and to Medicare FFS.
There is much more to say about risk selection. For now, it is worth reflecting on where we stand. The main political argument against premium support has been that it will lead to cost-shifting. The 2012 Ryan proposal, based on the work of the Bipartisan Policy Center, addresses this concern by using competitive bidding on a budget — the same global budget used by the Obama administration.
So now we are left with the argument that risk selection raises the cost of the current Medicare system in which Medicare beneficiaries are given a modicum of choice relative to a counterfactual world in which Medicare beneficiaries are have no choice but to be part of Medicare FFS, an approach that presumably entails risks of its own. One of the reasons this argument is a potent one is that it is very hard to imagine a scenario in which Congress will have the political will to force bad risks to pay substantially more than they do at present. We’ve gone from arguing about whether Ryan and other advocates of competitive pricing aim to destroy the health safety net for retirees — they’re not — to whether they’ve stumbled on the best, most realistic formula for containing rising Medicare costs — we’re not really sure. That’s not much of a 30-second ad.
Orszag warns us against putting all our chips down on the health-care competition tooth fairy, which is fair enough. But surely it is unwise to put all our chips down on the power of administrative price-setting and congressional resolve to restrain spending.