Policy Review Banner

Why Spending Has Got to Give

Saturday, April 1, 2006

Federal spending rose from about 18.5 percent of gross domestic product (gdp) at the end of the Clinton administration to 20.3 percent by the end of George W. Bush’s first term — during the watch, that is, of a Republican president and a Republican Congress. Of course, much of this increase is in defense spending and homeland security. But President Bush has not chosen guns at the expense of butter: He has opted for both. He did not veto even a single spending bill. And real (that is, inflation-adjusted) domestic discretionary spending, not counting homeland security, rose by an annual average of 4.8 percent over his first four years in office.1 Someone who favors relatively small government could get awfully depressed looking at these numbers.

One does not get less depressed contemplating the spending increases that are projected over the next 45 years. Credible estimates from the Congressional Budget Office and from independent budget analysts show federal spending doubling as a percent of gdp by the middle of the twenty-first century, reaching about 40 percent of gdp (see Figure 1). Yet some historical constants and some facts about Americans’ views on taxes incline this observer to believe that federal spending will come nowhere close to 40 percent of gdp by mid-century.

figure 1
A Scenario for Total Federal Spending and Revenues, Percentage of GDP

Source: Congressional Budget Office.
Note: cbo categorized this scenario as one of high spending and lower revenues. The scenario is explained in detail in cbo, the Long-Term Budget Outlook (December 2003), 6–12.

Before we turn to the good news, let us consider the bad news: the news on spending. Defense spending rose by $161 billion between fiscal years 2001 and 2005, an increase of $117 billion in 2001 dollars. This was large, and yet it took defense spending from a postwar low of 3.0 percent of gdp in 2001 (tied with 1999 and 2000) to 3.8 percent of gdp. Even if the U.S. government maintains a strong military presence in the world, which seems likely, it can do so with less than 4 percent of gdp. To see where spending is projected to grow substantially as a percentage of gdp, therefore, we must look elsewhere. The three programs accounting for most of this increase are projected to be Medicare, Medicaid, and Social Security. Medicare and Medicaid spending together are credibly predicted to be about 21 percent of gdp by 2050, and Social Security spending is expected to equal about 6 percent of gdp by 2050. All three are driven by demographics — the aging of the U.S. population — and the first two are also driven, ironically, by improvements in health care.

Consider Social Security first. Social Security spending, which is now about 4.2 percent of gdp, is likely to be 6.2 percent of gdp by the middle of the twenty-first century unless changes are made in the program. This is due to two main factors: 1) the retirement of the baby-boom generation and 2) the increasing life expectancy of the elderly. The two factors together mean that the fraction of people aged 65 or older will rise from 12 percent of the population in 2000 to 19 percent in 2030. The working-age population, by contrast, is projected to fall from 59 percent to 56 percent.2 Based on this, the Social Security trustees project that the number of workers per Social Security recipient will decline from about 3.3 in the early 2000s to 2.2 in 2030.3 Of course, substantially increased immigration of younger people or a significant decline in life expectancy of the elderly could slow this trend but absent that, these population numbers are fairly firm. And absent a change in that ratio, absent policy changes in Social Security (more on that later), and absent a substantial increase in the growth of productivity, the increase in Social Security to about 6 percent of gdp is also fairly firm.

The scarier numbers are in Medicare, the federal government’s socialized medicine program for the elderly, and Medicaid, the program for the poor and near-poor: Not only is the number of people enrolled in these programs increasing, but spending per person has also increased and will likely continue to do so.

Since 1967, the first full year of Medicare spending, spending has risen from 0.2 percent of gdp to about 2.3 percent in fiscal year 2004. Medicaid spending rose from 0.3 percent of gdp in 1970 to 1.5 percent in 2003, a quintupling of its share of output. Between 1970 and 2003, Medicare spending per person rose annually by 3 percentage points more than the growth of per capita gdp. Over approximately the same period, Medicaid spending per person rose annually by 2.7 percentage points more than the growth of per capita gdp. The spending growth comes from the combination of Medicare and Medicaid beneficiaries spending other people’s money plus the incentive thereby created to develop technological improvements allowing doctors and hospitals to do more. The spending is valuable. That’s not the problem. The problem is that Medicare and Medicaid recipients are spending other people’s money and therefore do not restrain their spending as much as they would if they were spending their own. What would otherwise have been an individual decision by someone trading off between health care and other goods becomes, instead, society’s problem because the government has made it into society’s problem. And because incremental dollars spent are partly paid for by taxpayers, people trade off at a different point than they would if they were spending their own money. Specifically, they buy medical care that they would not be willing to purchase on their own.

This is not to say that medical spending would not be rising as a percentage of gdp if there were no Medicare or Medicaid. Economists Robert E. Hall and Charles I. Jones argue that because health care adds years to our lives, people will voluntarily spend a higher fraction of their income on health care as their real incomes grow.4 Hall and Jones project, in fact, that overall health care spending could be as much as 33 percent of gdp by mid-century (up from about 15 percent today) and argue that there is nothing wrong with that. They are right. The problem, as noted above, comes when people spend other people’s money.

Based on past growth in spending per person and assuming no changes in policy, the Congressional Budget Office projects that by 2050, Medicare and Medicaid spending could be as much as 21 percent of gdp (see Figure 2). Together with the growth in Social Security spending, and assuming that other spending doesn’t fall as a percentage of gdp, this would put federal government spending in 2050 at about 40 percent of gdp, or twice its share of gdp today.

Tax increases aren’t enough

That’s the bad news. Now the good news. The Congressional Budget Office said it best:

In the past half-century, total revenues have ranged from 16.1 percent to 20.8 percent of gdp, with no obvious trend over time. On average their share of gdp has hovered around 18.5 percent.5

This is about as close to a historical constant as one finds in public-sector economics. U.S. experience differs dramatically from that of other countries. This is probably because of the political equilibrium we have reached in the United States due not only to our particular demographics but also, and more important, to division of powers and a republican rather than a parliamentary political system. Clearly, unless the deficit takes up a lot of the slack (which is highly unlikely for reasons that are argued below), something’s gotta give. Which will it be: taxes or government spending? The odds-on favorite is government spending.

figure 2
Total Federal Spending for Medicare and Medicaid, Percentage of GDP

Source: Congressional Budget Office. See The Long-Term Budget Outlook (December 2003).

It’s true that taxes as a share of gdp did rise in the late 1970s from 18 percent in fiscal year 1977 to 19.6 percent in fiscal 1981. But there are three things to note about this. First, the high inflation of the 1970s drove people into higher tax brackets, and the indexing of tax brackets that Ronald Reagan and Congress put into the 1981 tax law, effective in 1985, means that inflation alone can no longer put people into higher tax brackets.6 Inflation, therefore, cannot be the income-tax-revenue generator for government that it was in the late 1970s. Second, the increase in tax revenue generated by inflation was one of the main factors that led to popular support for Reagan’s 1981 cut in income tax rates at all income levels. Third, this increase in tax revenue as a percent of gdp still kept the number within the 20-percent upper limit.

Ample polling data also support the view that Americans, whatever their other positions, are against higher taxes on themselves. Various polling organizations asked Americans their views on taxes in 1938 and in most years since 1947. The percentage who thought their taxes were too low was always between 0 and 2, except for 2003, 2004, and 2005, when the number hit 3 percent. And in every year but 1949 and 2003, the percentage who thought their taxes were too high exceeded the percentage who thought their taxes were about right, usually by a wide margin.7

It’s not surprising, of course, that people would think their own taxes are too high. But as the following evidence shows, most Americans are even against taxing the highest-income people more and, indeed, favor taxing them less. A Roper Center/Reader’s Digest poll in October 1995 asked people what they thought was the highest percentage of income governments at all levels should take in taxes of all forms. The higher the hypothetical income, of course, the higher was the percentage people found to be fair. What was striking, though, was how low this percentage was. Even for the highest-income family asked about, one making $200,000 a year, the mean percentage that people found to be fair was 27 and the median percentage found to be fair was 25. To put this in perspective, a real family making $200,000 a year or more at about that time paid about 28.7 percent of its income in federal taxes of all forms.8 Adding in their state and local taxes would take this number well above 30 percent and close to 35 percent. In other words, high-income people were already paying well above the median level that the Americans surveyed thought fair.

Nor did the Bush tax cuts change this much. A 2002 study by economists William G. Gale and Samara R. Potter found that, taking account of the 2001 Bush tax cut, people in the top 1 percent of the income distribution paid 31.3 percent of their income in taxes of all forms.9,10 To add more perspective, consider the fact that left-wing politician Al Sharpton, who advocates higher taxes on “the rich,” when asked by abc News reporter John Stossel what percentage of their income “the rich” should pay in federal income taxes, answered “around fifteen percent.”11 This is well below the approximately 20 percent, in income taxes alone, that they now pay. It’s true that Sharpton can get away with advocating tax increases on high-income people because few people know just how much high-income people pay in taxes, but various economists and reporters will certainly get these facts out to the public whenever a tax increase becomes a serious threat.

The bottom line is that it would be extremely difficult for the federal government to raise taxes by more than a few percentage points of gdp. Even if federal taxes were to rise, say, to 25 percent of gdp, this would imply a 35-percent increase in the federal tax share of gdp over its historical average share of 18.5 percent and would still require government spending to “give” much more than taxes.

The implications of these facts about government spending and taxation are huge. First, the dominant problem of domestic economic policy for the next 40 years will be how to rein in the growth of government spending. The president and Congress may have the luxury of escaping it for the next few years, but by sometime in the next decade (and possibly even during this one), reining in spending will be paramount. Second, many of the proposals now being pushed by pro-free-market economists­ that are not treated seriously by politicians in Washington will be taken very seriously very soon. These include cashing out Medicare by giving every recipient straight cash or a health care voucher, raising the age to receive full Social Security benefits to 70 and then indexing it to life expectancy, raising the age to qualify for Medicare to accord with the age to qualify for Social Security, capping Social Security benefits in real terms so that they no longer grow,12 requiring Medicare and Medicaid recipients to pay substantial co-payments for medical care, and privatizing the disability insurance component of the Social Security program.13 All of these, I predict, will be on the bargaining table. They may not yet be politically feasible but will quickly become so once they become politically necessary.


Why worse will be better

Consider, for example, one way the political dynamics of Social Security might change as more and more baby boomers (those born between 1946 and 1964) start collecting benefits. One of the most important insights from “public choice” economics is what I like to call “the importance of being unimportant.” Translation: If an interest group is suitably small, its members can be relatively cohesive. They can get together to obtain a special subsidy, regulation, or import barrier that costs members of a much larger group a little each in order to give members of the small group a lot each. One reason the farm lobby has been so successful at getting subsidies for its members is that there are so few farmers. This means that each farmer can get a substantial gain at the expense of consumers and taxpayers (in the form of higher prices and higher taxes, respectively) and that the consumers and taxpayers don’t bother organizing to fight the wasteful farm policies because each pays a much smaller amount than the gain per farmer. This “importance of being unimportant” explains a phenomenon that has surprised many observers: Even as the farm population has shrunk, the lobbying success of the farm lobby has grown.

But the reverse also holds. All other things being equal, the larger the interest group becomes relative to the size of those paying for its special privileges, the bigger becomes the loss to the payers. The case of Social Security and Medicare now becomes relevant. One reason there has been relatively little resistance by the working population to increased subsidies to the elderly is that for a few decades there have been about three to four workers for every elderly beneficiary. But as the number of workers per beneficiary falls to 2.2 by 2030, as noted above, the resistance among workers will grow because the cost per worker will grow. This could imply a new political equilibrium in which the amount of benefit per elderly person would not grow as quickly as planned and might even fall somewhat. Readers of Malcolm Gladwell’s The Tipping Point might think that that is what I’m describing here. But it’s not necessarily a tipping point. Rather, increased political pressure by the relatively young will lead, along a continuum, to a different outcome.

Now, the above does not mean that there would be an across-the-board reduction in benefits or in the growth of benefits for all the elderly. Instead, a new political coalition might form between the working-age population and, say, the oldest Social Security beneficiaries to rein in benefits for the relatively young seniors who can most afford to give some up. Or it may be a coalition between the working-age population and the younger elderly. Which coalition comes about is difficult to predict — that there will be a coalition seems likely.

This might sound heartless, but note two things: First, it’s not heartless to recognize that the cupboard is bare and to prepare for it. Second, it’s important to remember why we will be in this fix — it’s because Franklin Delano Roosevelt, as president, purposely structured Social Security to put us there. In a famous statement about the Social Security payroll taxes, fdr said:

[T]hose taxes were never a problem of economics. They are politics all the way through. We put those payroll contributions there so as to give the contributors a legal, moral, and political right to collect their pensions. . . . With those taxes in there, no damn politician can ever scrap my Social Security program.14

If heartlessness is to be ascribed to anyone, therefore, it is to that Machiavellian man who set things up.

Of course, not all ways of cutting the growth of benefits are created equal. Some are fairer than others, and some are more efficient than others. Here’s where the insights of economists who recognize the incentive effects of policies will be important. One problematic way to cut Social Security and Medicare benefits, or the growth of benefits, is means testing; that is, cutting benefits more for the higher-income elderly than for the lower-income elderly, which is what President Bush and many other politicians have proposed. There is one and only one merit to this: Those with more will be better able to take the hit. But the proposal runs up against a tough philosophical problem and a tough practical/economic problem.

The philosophical problem is this: Consider two people, Smith and Jones, who are the same age and have the same time profile of earned income. Smith saves a high percent of his income and invests it in stocks, bonds, and other investments. Jones goes to Europe every few years and saves little. When they retire, Smith gets substantial income from his investments while Jones gets little. So, under some proposals, Smith would get less than Jones. This is unfair. Smith is being penalized for saving. Many of us have read to our children the story of the ant and the grasshopper — the ant who stores up food for the winter and the grasshopper who doesn’t. We tell our children that it’s unjust for the grasshopper to have a free ride. Unless we think we are teaching our children an unjust ethic, therefore, the injustice of this way of cutting benefits should be apparent.

This gets us to the practical economic problem. When governments penalize saving and high incomes, they can expect there to be less saving and lower incomes. Cutting benefits more for higher-income people discourages people from saving as much as they might otherwise, so the income from their investments is not as large as it might otherwise be. That’s one part of the problem. The other is what means testing would do to people — especially the elderly — who have saved and put themselves in a higher-income category who, therefore, have a great deal of discretion over how much income to earn in their old age. With means testing, the implicit marginal tax rates they will face when they earn more than a relatively modest income can easily be over 70 percent.

The reason marginal tax rates would be so high is that they are already substantial. Means testing would drive them higher. Consider, for example, a single 66-year-old man whose earnings from taxable bonds are $22,000 per year. (This could be based on a 4-percent interest rate on bonds worth $550,000.) Assume the man’s annual Social Security benefits are $10,000 a year and that he is working part-time for $12,000 a year. Thus, his total income is $44,000 a year. Assume for simplicity that he has no tax-exempt interest income. If he takes the standard deduction and has no dependents, this man will be in the 25-percent federal tax bracket. But for every dollar he earns, an additional 85 cents of his Social Security benefit will be taxed at his federal income tax rate. Thus, every dollar he earns will bring not one dollar, but $1.85 under taxation at his marginal rate of 25 percent. Imagine he is trying to decide whether to work a little harder, producing goods and services that others value, for an extra $1,000 annually. By making that extra $1,000, he would incur a tax liability of $250 plus 25 percent of $850 of his Social Security income, for an additional federal tax liability of $462.50. And that’s just his federal income tax. To that we must add Social Security and Medicare taxes equal to 7.65 percent of the additional $1,000 in earnings, or $76.50. Then, if he lives in a state with an income tax, he could well pay 4 or 5 percent of this $1,000 in state income taxes. Even if it’s just 4 percent, this is an additional $40. So, on that extra $1,000 in earnings he’s considering, he would pay an additional $579 in taxes. His marginal tax rate, therefore, is 57.9 percent.

Now introduce even a modest means-tested Social Security where, for example, he gives up $150 in Social Security benefits for every additional $1,000 in earned income. The implicit extra marginal tax rate on this $1,000, then, is 15 percent. Add this to the 57.9 percent earlier computed and our relatively modest-income “upper-income” elderly person now faces a whopping marginal tax rate of 72.9 percent. He may well decide, along with millions of other elderly people, not to earn that extra $1,000 producing goods and services that others value. That is the practical problem.15

There’s a related practical problem. Because many people in this man’s situation may decide not to work as much as they would have, the government gives up the income and payroll tax revenue that it would have collected on that extra income. This means that the static estimate of the Social Security spending that the government estimates it would save thanks to means testing will overstate the net positive impact on the government’s finances. Offsetting this saving from means testing will be a reduction in revenues that would otherwise have been collected, and this reduction might well be large.

One might argue that the reduction in work that I predict above will be modest because few people will be aware of these high implicit marginal tax rates. It is true that they will be implicit rather than explicit and that, therefore, many elderly people will be unaware of them. At first. But is it really a good idea to implement a policy based on the notion that people won’t find out, especially when there will be countless books, magazine articles, and websites telling people how to map out their income-earning, spending, and sheltering strategies?

A way to cut benefits or the growth of benefits that avoids these incentive problems and that, in ant/grasshopper terms, is more fair is to raise the age for Medicare in line with Social Security (phased in to give people time to plan), raise both ages to 70 (again phased in to give people time to plan), and cut the real growth of benefits. Interestingly, this was done in 1983, when the age for receipt of full Social Security benefits was raised in stages to 67, effective in 2027. So we have a precedent for it. Nor was it done at the behest of either the 1983 Greenspan commission on Social Security or the Reagan administration. The real heroes in this were a Democratic congressman from Texas named Jake Pickle, who attached it to the House bill implementing the Greenspan commission’s recommendations, and the House of Representatives generally, then under Democratic control. The fact that it was done then with so little fanfare suggests that it could be done again.



I  said earlier that the deficit is unlikely to take up a lot of the slack. Why do I believe that? The reason is that other than during World War ii, when federal spending on the war alone was well over 40 percent of gdp, there has been no period in our history when the deficit was over 5 percent of gdp in a nonrecession year. Even a few years of deficits at 5 percent or more of gdp — to put this in perspective, that would be a budget deficit of over $600 billion — would raise the debt-to-gdp ratio dramatically. This seems unlikely, if only because such a debt level would be increasingly difficult to finance.

The budget numbers are such that various market-based reforms will be looked at seriously — soon and for a long time. We must not give up on these reforms because they are not politically popular today. What reformers should do, instead, is keep honing their proposals for reining in government spending and keep their powder dry.

1See Stephen Slivinski, “The Grand Old Spending Party: How Republicans Became Big Spenders,” Cato Institute Policy Analysis 543 (May 3, 2005).

2The data in this paragraph are from Congressional Budget Office, The Long-Term Budget Outlook (December 2003), 4.

3Why so low, given that the ratio of 56 percent to 19 percent is almost 3? The reason is that the working-age population includes many people who are not working and who receive Social Security benefits, either disability benefits or Social Security benefits for those who retire before age 65. This both reduces the numerator (the number of workers) and increases the denominator (the number of Social Security recipients).

4Robert E. Hall and Charles I. Jones, “The Value of Life and the Rise in Health Spending,” National Bureau of Economic Research Working Paper 10737 (August 2004).

5Congressional Budget Office, Long-Term Budget Outlook, 43.

6This is slightly inaccurate. Neither the tax brackets for the Alternative Minimum Tax nor the brackets for taxation of Social Security benefits are indexed. Also, capital gains are not indexed. Nor is the corporate income tax indexed to inflation. All these factors do give the government some incentive to inflate in order to increase tax revenues. But countering this incentive is the fact that excise taxes on alcohol, cigarettes, and gasoline are legislated in pennies per unit and, therefore, fall as a percent of gdp if the government inflates.

7See Karlyn H. Bowman, “Public Opinion on Taxes,” aei Studies in Public Opinion (April 15, 2005).

8Joint Committee on Taxation, “Distributional Effects of the ‘Taxpayer Relief Act of 1998,’” Pamphlet jcx-63-98 (Government Printing Office, September 15, 1998). Referenced in Harvey S. Rosen, Public Finance, sixth edition (McGraw-Hill Irwin, 2002), 279.

9William G. Gale and Samara R. Potter, “An Economic Evaluation of the Economic Growth and Tax Relief Reconciliation Act of 2001,” National Tax Journal 40 (March 2002). Referenced in Harvey S. Rosen, Public Finance, seventh edition (McGraw-Hill Irwin, 2005), 300.

10To be in the top 1 percent, a family’s income had to be greater than $373,000.

11Quoted in John Stossel, “Myths, Lies and Downright Stupidity,” abc News (January 23, 2004).

12Built into the Social Security benefit formula are real benefit increases as long as real wages increase. In 1996, economists Martin Feldstein and Andrew Samwick estimated that the average annual benefit (in 1995 dollars) would rise from $7,510 in 1995 to $8,790 in 2016 and $9,290 in 2023. Simply eliminating this growth in real benefits prospectively, to give people a few years to adjust, would make Social Security benefits substantially lower as a percentage of gdp than otherwise. See Martin Feldstein and Andrew Samwick, “The Transition Path in Privatizing Social Security,” National Bureau of Economic Research Working Paper 5761 (September 1996).

13For reasons beyond the scope of this article, the case for privatizing the federal disability insurance program could well be stronger than the case for privatizing the retirement part of Social Security.

14From Arthur M. Schlesinger, Jr., The Age of Roosevelt, vol. 2, The Coming of the New Deal (Houghton Mifflin, 1959), 309–310, referenced in Martha Derthick, Policymaking for Social Security (Brookings Institution, 1979), 230.

15Indeed, we see this problem writ large in a contrast between the United States, Canada, and Japan on the one hand and Germany, France, and Italy on the other. Edward Prescott, co-winner of the 2004 Nobel prize in economics, has estimated that the reason the average working-age person in Germany, France, and Italy works about one-third fewer hours than the average working-age person in the United States, Canada, and Japan is that marginal tax rates in Germany, France, and Italy are about 60 percent versus “only” 40 percent in the United States, Canada, and Japan. Edward C. Prescott, “Why Do Americans Work So Much More than Europeans,” Federal Reserve Bank of Minneapolis Quarterly Review 28:1 (July 2004).