Improving U.S. Health Care Policy Through Economic Research
Presented to the American Society of Health Economists
Dean Douglas Elmendorf
June 14, 2016
Thank you. I am delighted to be here at ASHE’s biennial conference and honored to have been invited to speak to you.
Dan Polsky and Mark Pauly asked me to talk about the role of economic research in improving health care policy. That is a terrific assignment because economic research—your research—has played such a central role in health care policymaking and will undoubtedly continue to do so. Given my background, I will focus on federal policy. I want to begin by talking about health insurance coverage, which is an area where the significant policy changes of the past several years were built on decades of research and, in my view, have now established a sensible system of subsidies and rules. Then I will talk about health care spending, which is an area where policymakers and analysts are still groping for a sensible system and where much research has been done but much more will be needed. I will close with a few suggestions about how you can help economic research to have the largest positive impact on health care policy.
Health Insurance Coverage
I will begin with insurance coverage. As you know, the Affordable Care Act (ACA) significantly expanded federal subsidies for health insurance for lower-income people and made important changes to the rules governing insurance markets. The design of the subsidies and rules in the ACA drew directly on economic research as well as on other factors. In my view, our new system of rules and subsidies is a sensible system in its fundamental elements, and therefore future policy changes should occur within that system. Let me elaborate.
The Members and staff of the Congressional committees that developed the Affordable Care Act were acutely aware of previous failures to establish universal, or nearly universal, health insurance coverage in this country. The most recent failure occurred during the Clinton Administration, whose proposal failed to win enough political support to become law. I think the developers of the ACA, with that history in mind, believed that they needed to satisfy at least three political constraints to succeed: The first was that their plan needed to raise insurance coverage to nearly 100 percent, or the supporters of universal coverage would not be satisfied. The second was that their plan could not significantly reduce employer-provided insurance, or the many people who already had such insurance would be worried about what they might lose. And the third was that the plan could not increase the federal budget deficit, or moderate Democrats would not support it.
To satisfy those constraints, the developers of the ACA needed a structure of subsidies and rules that would substantially increase insurance coverage at the lowest possible budgetary cost and with the smallest possible disruption to existing insurance. In other words, the political constraints generated a specific economic problem. The ACA’s developers solved that problem by consulting with economic researchers—in person, by reading their published work, and through intermediaries like the Congressional Budget Office (CBO).
CBO is a crucial intermediary between researchers and policymakers because CBO uses results from economic research to estimate the effects of proposed policies. In the half-dozen years leading up to 2009, CBO had built models to estimate the effects of various policy changes on health insurance coverage and the federal budget. In 2009 and early 2010, that model building continued, as proposals brought to CBO involved policy levers that had not been constructed previously. During that period, CBO and the staff of the Joint Committee on Taxation (JCT) produced estimates of dozens of alternative comprehensive changes to insurance subsidies and market rules.
That model building and those estimates drew on a very large number of research studies that my colleagues had read, evaluated, and extracted information from. Unfortunately, CBO never had time to write a report fully describing the methodology used for those estimates, so no complete reference list exists. But you can see partial lists in some reports my colleagues did write, including a 2007 paper titled “CBO’s Health Insurance Simulation Model: A Technical Description”; a 2010 paper titled “Will Health Insurance Mandates Increase Coverage? Synthesizing Perspectives from the Literature in Health Economics, Tax Compliance, and Behavioral Economics”; and a 2012 report titled “CBO and JCT’s Estimates of the Effects of the Affordable Care Act on the Number of People Obtaining Employment-Based Health Insurance.”
The Congressional committees used their own understanding of economic research and estimates from CBO and JCT based on that research to develop an insurance system that would satisfy the three political constraints I listed and to meet other objectives. At the same time, opponents of the ACA used their understanding of economic research as well as estimates from CBO and JCT to uncover disadvantages of the ACA approach and to explore alternatives to it. So, economic research was central to the policy development process.
Moreover, the estimates made by CBO and JCT based on that research have turned out to be pretty good estimates, at least so far. Obviously, the outcomes have not lined up precisely with the estimates—nor did I ever expect they would, given the tremendous uncertainty about the effects of such fundamental, out-of-sample changes in policy. But we were right that the number of uninsured people would fall sharply and that employment-based insurance would not fall much, and the cost of the coverage expansion has been very much in the ballpark of our estimate—all points that were hotly disputed in 2010 and later. So, our careful and informed use of economic research led to estimates that were much better than the guesses of many people who did not take the research literature as seriously. Research not only guided the process of reforming the health insurance system, it did so in a fairly accurate way.
Let me also explain why I think our new health insurance system is fundamentally sensible. The first part of my reasoning is a value judgment: I think we should bear the cost of achieving nearly universal coverage. I think that is especially important because the people who were uninsured were disproportionately people of modest means whose market incomes have benefited only a little from the growth of total output and income in this country during the past few decades.
The second part of my reasoning is an analytic judgment: There are no alternatives to the ACA framework that would achieve nearly universal insurance coverage at notably lower budgetary cost or with notably smaller changes in rules. This year, more than 10 million people are insured through exchanges and more than 10 million others are covered by Medicaid because of the ACA; nearly all of them would otherwise be uninsured. Indeed, the share of the population under age 65 without health insurance is falling nearly in half, from about 20 percent to about 10 percent. The costs of that increase are about $5,000 of direct federal subsidy per newly insured person in 2016 and a collection of significant changes in insurance rules. (Including not only the direct subsidy but also the tax revenue lost indirectly from a reduction in labor supply would raise that figure a little.) That is a significant amount of money but not surprising given the costs of health care in this country. In addition, the ACA subsidies will be less than 10 percent of total federal subsidies for health insurance over the next decade: About half of total subsidies will go to Medicare, about 20 percent to Medicaid apart from the ACA, about 20 percent to tax expenditures for employer-sponsored insurance, and the remaining less than 10 percent to the ACA’s subsidies.
Moreover, there is no way to achieve high levels of insurance coverage while spending much less on subsidies or imposing much looser rules on insurance markets. Because health care is expensive, raising coverage without imposing an excessive burden on lower-income people requires substantial subsidies. Because health care spending varies widely among people in ways that can be partly predicted and because society provides emergency care for those who need it, health insurance markets cannot function entirely like markets for most goods and services. Nothing I saw at CBO in six years of analyzing alternative policy changes suggests that there are less costly ways to significantly expand coverage than the ACA. That conclusion is confirmed by the fact that, six years after passage of the ACA, there is no alternative on the table with a reliable, independent estimate showing comparable effects on insurance coverage.
Health Care Spending
Let me turn now to health care spending—an area where, as I said, policymakers and analysts are still groping for a sensible system and where much research has been done but much more will be needed.
We do not have a sensible system today because we have neither centralized control of health care spending nor appropriate markets and incentives for decentralized decisions. One consequence is that we spend much more on health care per person than other countries do. Another consequence is that we do not get as much good health for our health-care dollars as one would hope. To establish a more sensible system, we should either strengthen the role of government in centrally managing the health care system or strengthen the ability of market forces to produce reasonable outcomes—or perhaps adopt a combination of those strategies. Given many Americans’ distrust of the government’s involvement in health care, as we saw in their reactions to the ACA, my own view is that greater centralized control is a nonstarter at this point, and I support using market forces to achieve better decentralized decisions.
That strategic choice leads me to favor four types of policy changes, which I will describe briefly. Then I will spend more time giving some examples of the role of economic research in informing efforts to use market forces in health care more effectively.
In terms of policy changes, the first type I favor is to stop providing open-ended public subsidies for health care. Because employers’ payments for insurance they sponsor are fully excluded from taxable income, insurers have less incentive to use deductibles, coinsurance, utilization management, and firm negotiations with providers to restrain spending. The ACA limited this subsidy by imposing the so-called Cadillac tax on plans with premiums above a threshold, and repealing that tax would be a serious mistake.
Second, I think we should move Medicare to a premium support system in which all beneficiaries would choose between competing insurers and bear the entire difference in cost between their choice and a benchmark. That approach would be similar to the subsidies in the ACA’s exchanges and could be viewed as an expansion of the Medicare Advantage program that would strengthen price competition between private insurers and strengthen competition between private insurers and fee-for-service Medicare.
Third, I think we should restructure payments to Medicare’s fee-for-service providers so that we are not just rewarding more care and more complex care, but instead rewarding care that is more appropriate, cost-effective, or of higher quality. Ongoing efforts by the Department of Health and Human Services to implement alternative payment models should proceed with as much vigor and urgency as possible. Changes in incentives faced by providers in Medicare would lead to higher-value care not only for Medicare patients but also for patients with private insurance, which often follows Medicare’s policies.
Fourth, I think we should aim to maintain as much competition in the underlying markets for health care and insurance as possible. That means limiting further consolidation by health insurers and by hospitals and other health care providers. It means relaxing scope-of-practice laws. And it means providing people with information and tools to help them be more-effective buyers of insurance and users of care.
Rather than discussing those policies at length, however, I want to offer some examples of the impact of economic research on policymakers’ efforts to make better use of incentives and market forces in health care. Not surprisingly, the examples are drawn from my experience at CBO. I chose them to illustrate the sorts of research and policy applications that we need much more of in the years ahead.
The first example is what would happen if competitive pressure was increased in Medicare by adopting a premium support system. Such a system would have significant risks and disadvantages. Some beneficiaries who remained in fee-for-service Medicare would pay substantially more than under the current system. In addition, setting benchmarks that are too low relative to market premiums or making benchmarks grow more slowly than premiums over time could increase beneficiaries’ burdens even more. Moreover, risk adjustment of payments to insurers would be much more challenging than in the current system. However, a premium support system in Medicare would have significant benefits as well. As we have seen with Medicare Part D and the ACA’s insurance exchanges, competition restrains costs. Also, beneficiaries’ choices in such a system would provide a useful signal of how much they value the benefits of additional spending for health care.
To quantify those considerations and others, CBO did a thorough analysis of premium support a few years ago. My colleagues estimated that Medicare spending under two illustrative premium support systems would be lower in the short term than under current law and would grow slightly more slowly in the long term. Those estimates were based on models constructed and parameterized using the research that you and others have done. That research addressed bidding behavior by insurers, risk selection and risk adjustment, variation in provider payment rates between insurers, beneficiaries’ sensitivity to premiums in choosing among insurance plans, inertia in switching between insurers over time, spillovers between Medicare Advantage and fee-for-service Medicare, and more. Indeed, the report presenting CBO’s estimates included a substantial appendix describing key elements of the methodology and listing some of the research papers on which the analysis drew. CBO’s report has played an important role in subsequent public and private discussions of premium support by policymakers.
The second example I want to describe is what would happen if people had stronger incentives to avoid unhealthy behavior. The possibility of reducing health care spending by improving people’s health is tempting for obvious reasons, and Members of Congress regularly pressed CBO to estimate the budgetary effects of improvements in health that might result from certain policies. Therefore, we undertook a large-scale effort to assess the effects on health and the federal budget of an increase in the excise tax on cigarettes.
A higher tax on cigarettes would reduce smoking, which would make some people healthier, reduce their annual health care costs, and enable them to live longer. My colleagues estimated that federal spending would be lower for a while because of lower per capita health care spending but ultimately would be higher because greater longevity would increase the number of Social Security and Medicare beneficiaries. The improvement in health would also make some people more productive and enable them to earn more income. My colleagues estimated that federal revenue would ultimately increase for that reason by about half as much as the direct increase in excise tax revenue. Putting those pieces together, CBO estimated that the effects of a higher cigarette tax in improving health would reduce the budget deficit for a number of decades, as Members of Congress expected, but not by much. That “not by much” finding is the reason this policy was not on the list of four key policies I recommended a few minutes ago.
CBO’s estimates were based on the agency’s own empirical analysis as well as extensive bodies of research about the effects of cigarette taxes on smoking, the effects of smoking on health status and longevity, the effects of smoking on health care costs, and the effects of smoking on labor earnings. CBO’s lengthy report on this analysis included references to dozens of studies that had been especially important in the work. And CBO’s findings have had important effects on policymakers’ thinking about this subject.
A third example of economic research being critical to understanding the effects of changes in incentives involves the design of health insurance coverage. The specific case I want to highlight is Medicare’s Part D prescription drug benefit. The ACA closed the so-called “coverage gap” in that benefit slowly over time. One consequence is to lower the price of certain prescriptions for certain beneficiaries, which will raise consumption of prescription drugs by those beneficiaries. An important question is whether that additional use of drugs will affect beneficiaries’ use of medical services such as hospitalizations and visits to physicians.
In 2009 and 2010, when CBO assessed the budgetary effects of closing the coverage gap, our estimate incorporated no change in the use of medical services. However, my colleagues later reviewed dozens of papers—many of them quite recent—that studied the effects of changes in the use of prescription drugs on the use of medical services. They concluded that researchers had, in fact, demonstrated such a connection, and in 2012 we changed our methodology. We published a report describing the relevant research and explaining that future CBO estimates of changes in Medicare’s drug policies would include an estimated change in the use of medical services. That research-based improvement in CBO’s methodology has affected the way Congressional policymakers think about changes in drug policies.
The fourth and final example I want to mention is the effect of changing the legal process regarding medical malpractice. This example is another one where CBO revised its methodology significantly in response to new research. Before 2009, CBO interpreted the evidence about the effects of tort reform on the use of health care services as mixed, so CBO’s estimates for proposed tort reforms incorporated changes in malpractice premiums but not changes in health care services. In 2009 my colleagues reviewed the latest evidence and concluded that significant tort reform would lead to small reductions in the use of health care services. We wrote a report explaining the research underlying our new approach, and we applied that approach in subsequent estimates. That research-based improvement in CBO’s methodology has affected Congressional debate about medical malpractice.
Helping Economic Research to Have the Largest Impact on Health Care Policy
In sum, economic research has had very substantial effects on changes in health care policy that have been enacted into law and on changes that have been discussed but not enacted. The connection between research and policy is a close one, and you should be pleased by that close connection and proud of your role in it. However, I want to close with three short suggestions about how you can help economic research to have the largest positive impact on health care policy in the future.
In offering these suggestions, I should be clear that I do not view improving policy as the only worthy goal of research. Advancing our understanding of the world is entirely appropriate for its own sake, and answers to abstract questions today can turn out to be valuable for policymaking later. But I hope the following suggestions are helpful for people who are interested in affecting policy in the near term.
My first suggestion is to tackle policy-relevant questions, even if they are harder to answer than some other questions. As the old joke has it, economists tend to look for lost keys under streetlamps, even if they know the keys were lost somewhere else. You should resist that tendency and look for keys where you think they are, even if the light is dimmer there. I do not mean that everyone should write papers addressing the broadest and most nebulous questions about our health care system or possible reforms to it. The many research papers to which I have alluded in my remarks addressed specific questions with persuasive empirical strategies. But the specific questions were ones that people formulating policy needed to know the answers to.
My second suggestion is to present your findings in policy-relevant ways. Partly this means writing abstracts, introductions, and conclusions of papers in clear and straightforward language that can be understood by people who are not on the cutting edge of research—even if the bodies and appendixes of papers can be understood only by people who are on the cutting edge. Presenting findings in policy-relevant ways also means being direct about the limitations of your work. People outside the research community will sometimes apply their desired spin to your findings, but you should not.
Presenting findings in policy-relevant ways further means focusing on the economic magnitude of estimated coefficients and on confidence intervals around those point estimates, rather than focusing on whether the point estimates are statistically different from zero with 95-percent confidence. Remember that standard tests of statistical significance are designed to reduce the probability of falsely rejecting a null hypothesis that one believes to be true and is actually true—but the null hypotheses in our papers are often chosen for convenience rather than because we believe them, and we should be as concerned about accepting false null hypotheses as rejecting true ones. In addition, confidence intervals that are described in economic as well as statistical terms are often more helpful in designing policies.
My third suggestion is to support the people who are intermediaries between researchers and policymakers. Intermediaries are needed because policymakers and their staffs often do not have the time or background to read and evaluate research. And researchers often do not have the time or background to engage with policymakers and their staffs on their terms. So, substantive experts who work full-time for government agencies, researchers who spend time visiting government agencies, serious think tanks, and the research communications teams at think tanks and universities play a very valuable role. When my colleagues and I at CBO would call friends—or strangers—in the research community with questions we faced, we were gratified by how readily people offered to help us. We depended on that help, because applying the informed professional consensus about some aspect of health economics to a specific policy issue was rarely straightforward. So when people from the policy world call you with questions, or try to hire your students, or invite you to visit, please say “yes”!
Thank you.