Federal Policies for Graying America

Keynote Speech, Second Annual Conference of the MIT Center for Finance and Policy

Douglas W. Elmendorf
The Brookings Institution
September 17, 2015

Thank you. I am delighted that the MIT Center for Finance and Policy has attracted such terrific researchers and is supporting such interesting and important research. I am honored to have this chance to talk with all of you this evening.

I want to give you my perspective on federal policies for our graying population. Older Americans are supported by a combination of public programs and private resources, and we need to make significant changes in federal policies toward both.

Four Developments That Deserve Emphasis

Before discussing specific policies, though, we should set the stage. There are four developments in this country that matter tremendously for the available policy options and, in my view, for the appropriate policy choices. These developments will not be surprises to you, but they deserve emphasis.

First, we are at the beginning of a surge of people into retirement. By 2025, there will be roughly 70 million beneficiaries of Social Security and Medicare, about 60 percent more than the number in 2007, just before the first baby boomers became eligible for early retirement benefits. Those 70 million beneficiaries will be about one-fifth of the U.S. population. Moreover, that share will increase further, albeit more slowly, after 2025. We are not facing a temporary graying of America but a long-lasting demographic transition. The key implication for federal policies is that getting policies for the aging population right is hugely important in both economic and social terms.

The second crucial development is that the incomes of people across most of the income distribution have risen quite slowly during the past few decades, while incomes at the high end of the distribution have risen very rapidly. According to estimates by the Congressional Budget Office (CBO), between 1979 and 2011, market incomes—that is, before taking account of taxes and transfers—rose 16 percent in the bottom quintile of the distribution, 9 percent in the middle quintile, about 60 percent in the top quintile apart from the top 1 percent, and 174 percent in the top 1 percent of the distribution. We should not put too much weight on those specific numbers because measuring income growth is difficult, and even though CBO’s methodology resolves some of the problems in other time series, measurement issues remain. Still, it is clear that most people in the country have gained very little directly from the growth of total output and national income in the past few decades. Tax and transfer policies have benefited people in the bottom part of the income distribution. Again from CBO, over the 1979-2011 period, income after taxes and transfers rose by at least 30 percent in each of the bottom four quintiles. Still, the fastest growth in income after taxes and transfers has been in the top quintile. The key implication, to my mind, is that changes in federal policies should protect people of modest means by imposing most of the burden of those changes on affluent people.

The third development is that changes in labor markets are significantly reducing the role of traditional employer-provided retirement benefits. There has been a flurry of interest recently in the so-called “gig” economy, in which people do specific tasks for companies without forming ongoing employee-employer relationships, but the extent of the gig economy is unclear. The bigger issue is people who are employed in traditional relationships but are much less likely than their predecessors of a few decades ago to receive traditional benefits, such as defined-benefit pensions. This change has two implications for federal policies: One is that portability of benefits between employers is becoming more important. The other is that the other sources of support in retirement—namely, personal saving and federal benefits—are becoming more important.

The fourth development I want to highlight is that federal debt relative to the size of the economy is very high by historical standards and will increase over the next few decades without significant policy changes. Federal debt is now 74 percent of gross domestic product (GDP), compared with an average of 38 percent over the past 50 years, and CBO’s long-term baseline shows debt exceeding 100 percent of GDP in 25 years. Economists do not know what the optimal amount of debt is, but one can make a strong case that we should plan to reduce the debt-GDP ratio gradually over the next few decades rather than letting it increase. And since benefits for older Americans account for the largest portion of federal spending today and the lion’s share of the increase in spending that will occur under current law, the fiscal challenge runs squarely up against the challenge of supporting people in retirement. The key implication is that federal resources are even more scarce than usual and need to be focused on their most valuable uses.

With those four developments in mind, let me explain the sorts of changes I would make in three key areas of federal policy—Social Security, Medicare, and tax incentives for personal saving. Then I will touch briefly on other policy areas where changes are needed as well.

Social Security

For Social Security, we should make the trust funds sustainable for 75 years primarily by reducing benefits for high-income beneficiaries and raising payroll taxes on workers with high earnings, but not by making significant across-the-board cuts in benefits or using general revenues.

We should make the program sustainable without general revenues because financing government programs with dedicated sources of funds has great advantages in terms of political economy. From a pure economics perspective, focusing on Social Security in isolation from the rest of the federal government’s activities is not sensible. Moreover, including trust funds that accumulate balances in a budget that tracks annual cash flows is inconsistent from an accounting perspective and leads to confused debates about double-counting and whether the government is truly prefunding benefits or not. So, the many analysts who focus on the overall federal budget have the economics right.

But this is not just a matter of economics. Linking certain taxes to certain spending improves people’s understanding of where their taxes are going, which facilitates a more constructive public debate about levels of taxes and spending. That understanding is crucial, especially at a time when people’s misunderstanding of basic facts about the federal budget has led to a terribly unconstructive debate. I admit that dedicated funding is not a panacea for effective governance: If it were, we would quickly agree to raise the gasoline tax to its inflation-adjusted level of 20 years ago as a straightforward user fee for highway construction that almost everyone seems to want—and we have not done that. Nonetheless, linking taxes to the benefits they finance offers a fighting chance for the kind of public debate we should have in this country, and I support it for that reason.

The specific goal of sustainability for 75 years is a little arbitrary, but it is the traditional planning horizon for Social Security, and I think it is in the right ballpark in terms of the effect on the overall budget. Suppose that our goal was to bring the debt-to-GDP ratio back down to its long-term average over the next 40 years. Then the changes needed to achieve 75-year sustainability for the Social Security trust fund—if phased in slowly, as would almost certainly be the case—would amount to about one-quarter of the total budgetary changes needed.

We should reach this target without significant across-the-board cuts in benefits because such cuts would significantly reduce total retirement income for many lower-income and middle-income people. Imposing that burden on people of modest means would be wrong, in my view, because those people have been experiencing such slow income growth. It would also be wrong for three other reasons: First, a significant number of people are already not saving enough for retirement, given current-law Social Security benefits. Second, according to the Center on Budget and Policy Priorities, 44 percent of elderly Americans have incomes apart from Social Security benefits that would leave them below the poverty line, while only 9 percent remain below the poverty line after accounting for those benefits. With smaller benefits, some people would save more on their own, but there would still be a risk of a substantial increase in poverty among older Americans. Third, Social Security benefits have some distinctive characteristics that are becoming even more important given the decline in defined-benefit pension plans:, Social Security benefits protect against longevity risk by being annuities for workers and their survivors, they impose no financial-market risk, and they do not require planning or self-control.

To avoid significant across-the-board cuts in Social Security benefits, some of the standard arrows in Social Security reformers’ quiver, such as increasing the eligibility age for full retirement benefits, should be a last resort rather than a first resort. Instead, we should focus on reducing benefits for high-income beneficiaries and raising payroll taxes on workers with high earnings. For example, CBO’s 2010 report on Social Security policy options includes an option to lower benefits for workers in the top half of the lifetime earnings distribution but leave benefits for those in the bottom half of the distribution unchanged. That report also includes multiple options for taxing income above the current-law taxable maximum. Taken together, such targeted reductions in benefits and increases in taxes could eliminate much of the estimated 75-year shortfall in the system.

I do not want to pretend that this approach offers a free lunch. Making Social Security more progressive would weaken the connection between an individual’s taxes and benefits, which could undermine the earned-benefit character of the system if taken too far. Also, people with higher incomes would face both higher taxes and lower benefits, which would reduce their incentives to work and save. But I think this is the best of the available alternatives for putting Social Security on a sustainable path without unduly burdening those Americans who have fared the worst in economic terms in recent years.

Before leaving Social Security, I should note that I am not going to talk about disability insurance. That issue deserves more attention from both policymakers and researchers than it has received so far. But disability insurance is not a direct consideration for people over the age of 65, so I will leave it aside tonight.


Let me turn now to Medicare. For Medicare, we should make two sorts of changes: increase the role of means testing, and accelerate ongoing transformations in how the program pays for health care.

In contrast with Social Security, I do not have a financial target to recommend for Medicare, for a variety of reasons. As some of you know, the Medicare trust fund that is projected to be exhausted in about 15 years covers only payments to hospitals and some similar providers. Payments to physicians and most other providers, as well as payments for prescription drugs, are made out of a different Medicare trust fund that is automatically replenished with general revenues as needed and therefore cannot be exhausted. Indeed, the system currently collects as much in general revenues as in taxes. This split structure is becoming increasingly odd as more Medicare payments go to private insurance plans that cover services from all types of providers or go to groups of providers providing bundles of services. Moreover, the open-ended use of general revenues greatly weakens the linkage between the system’s benefits and its dedicated funding sources. In addition, predicting the evolution of the health care financing and delivery systems is very difficult, so projections for 75 years are just not meaningful. For all of those reasons, I will talk about the sorts of changes I would make in Medicare without offering a financial target.

The rationale for greater means-testing for Medicare is the same as for Social Security: We should narrow the projected gap between federal benefits paid and taxes received, and given the relative income growth of people higher and lower in the income distribution during the past few decades, we should impose most of this additional burden on higher-income people—despite the distortions to work and saving that would result.

I would significantly expand what are known as “income-related premiums” for Medicare Part B, which covers payments to doctors and most other providers except hospitals. Today, about 6 percent of Medicare beneficiaries pay more than the standard premium for Part B, and under current law that figure will rise over time but only by about half a percentage point per year. I would boost that share substantially—maybe doubling it. I would also raise payroll tax rates for workers with higher earnings. The Medicare tax rate is now 2.9 percent for most people but 3.8 percent for people with earnings above certain thresholds. Those thresholds are not indexed in current law, so more people will pay the higher rate over time, but I would lower the thresholds to collect additional tax revenue from a larger number of high-income people.

In terms of Medicare’s payments for health care, about 30 percent of Medicare beneficiaries today are enrolled in private insurance plans through a system known as Medicare Advantage, and CBO projects that this share will rise to about 40 percent by 2025 under current law. Although Medicare Advantage involves competition between private insurers, the structure of the system does not maximize the intensity of that competition or create a level playing field between private insurers and traditional fee-for-service Medicare.

A system known as “premium support” could be viewed as an expansion of Medicare Advantage. Under premium support, beneficiaries would purchase insurance from either a private insurer or the government-run fee-for-service program, and they would pay 100 percent of any premium amount above a benchmark or receive 100 percent of any savings relative to the benchmark. With that arrangement, price competition between private insurers, and between private insurers and the fee-for-service program, would be stronger than it is under current law.

I would establish a premium support system with two specific characteristics: I would maintain fee-for-service as one of the competing options, both because some people are more comfortable with fee-for-service and because maintaining the fee-for-service system would lower overall costs for various reasons described in CBO’s report on the subject a few years ago. Also, I would set the benchmark premiums based on the average bid in each region rather than a lower bid, as is sometimes suggested; using the average bid would limit the savings for the federal government but would also limit the shift of additional costs onto beneficiaries.

According to CBO’s estimates, a premium support system with those characteristics would lower federal costs for affected beneficiaries by 4 percent and would lower beneficiaries’ total payments (premiums plus out-of-pocket payments) by 6 percent relative to what would happen under current law. Those numbers are fairly small in part because payments to Medicare providers are scheduled to increase quite slowly, and private insurers will have difficulty lowering costs much further. Moreover, some beneficiaries would pay significantly higher premiums than under current law, with the biggest additional burden borne by people who stayed in the fee-for-service program in regions with high fee-for-service costs. Still, there are important advantages to such a system. In addition to the savings I just mentioned, the heightened competition would probably lower long-term spending growth a little by reducing demand for costly procedures and increasing demand for cost-saving procedures. Also, beneficiaries’ choices in a premium support system would provide an important signal of how much they value additional spending for health care.

Even under a premium support system, however, a substantial share of Medicare beneficiaries would enroll in fee-for-service. In addition, private insurers sometimes follow the policies adopted by Medicare fee-for-service. Thus, it is important that the Medicare fee-for-service program improves the way it pays health care providers. We should support the work of the Center for Medicare & Medicaid Innovation, which is experimenting with alternative payment models. One important direction for change is to “bundle” more payments, so that Medicare makes one payment for all of the care associated with a particular episode of care rather than paying each provider separately. Such bundling can reduce Medicare’s costs and improve the coordination of care that patients receive because providers that are being paid together will work together more closely. Medicare is already expanding its use of bundled payments, and I think it should do as much as possible as fast as possible along those lines.

Tax Incentives for Personal Saving

That brings me to the third policy area I want to discuss, which is tax incentives for personal saving. We should overhaul those incentives both to increase saving by people who are under-saving today and to reduce the federal cost of saving incentives.

Studies suggest that a significant minority of Americans is not saving enough now. One factor that will make saving enough more difficult in the future is that the rate of return on savings will probably be lower than in the past because the aging of the population, slow productivity growth, and other factors will hold down the return to capital. In CBO’s estimates, the interest rate on Treasury debt will average about 2 percent in real terms going forward, compared with about 3 percent in the period before the financial crisis, and most of the factors that will depress Treasury rates will also depress the return on private assets. So, we need to encourage additional saving, especially by people who are not saving enough now.

At the same time, reducing federal subsidies for saving is important because the subsidies are expensive, go overwhelmingly to higher-income people, and may not have a large impact on people’s saving. For example, the exclusion from taxation of pension contributions and earnings costs the government about $150 billion per year now, including revenue lost from both income taxes and payroll taxes. About 85 percent of that lost revenue goes to people in the top two quintiles of the income distribution, and many of those people probably reduce other saving to offset much of their pension contributions.

Fortunately, economists know ways to increase personal saving that do not rely on substantial federal subsidies. As Raj Chetty discussed in his Ely lecture this year, there is a growing body of evidence that tax incentives for saving have smaller effects than “nudges” such as defaulting people into savings plans. Various proposals have built on this evidence in suggesting reforms to federal policies with three main elements: First, proposals would reduce tax subsidies for large savers by lowering the limits on contributions to tax-favored accounts or lowering the tax rates at which those contributions reduce tax liability. Such a reduction would improve the budget outlook and free up resources to increase retirement saving through other means—without significantly lowering overall saving or raising the share of people who are under-saving. Second, proposals would strengthen incentives for employers to arrange for retirement savings plans for their employees, to automatically enroll employees in those plans and establish default contributions, and to increase annuitization in those plans. Today, only about half of private-sector employees report that their employer even sponsors a retirement plan, but providing additional incentives to employers would probably raise that figure notably. Third, proposals include other changes to cajole people into doing more retirement saving, such as establishing refundable tax credits for contributions to retirement accounts or consolidating people’s retirement savings into single, portable accounts. Implementing a set of reforms along those three lines would make good sense.


Before I finish, let me mention two other policy areas where changes are warranted. One, the Pension Benefit Guaranty Corporation and many state and local pension plans are underfunded. It would be good to address that underfunding sooner rather than later. Two, private long-term care insurance is very limited. Medicaid pays for long-term care for people who have essentially no income or wealth. That benefit is costly for governments—the federal portion will exceed $100 billion a year in less than a decade—but still leaves people bearing significant financial risk. Generating a substantial increase in private coverage would be difficult—in part because of the safety net that Medicaid provides—but very valuable for society, and we should try.

In sum, older Americans are supported by a combination of public programs and private resources. To ensure that this support will be sufficient in coming decades for the growing number of older Americans, we should make significant changes in Social Security, Medicare, and tax incentives for private saving. These changes also would go a long way toward putting the federal budget on a more sustainable trajectory. And given the trends in income growth in recent decades, those changes should protect people of modest means by imposing most of the burden on affluent people. That is a feasible agenda, and we should get to it. Thank you.