Everybody is aware that a crisis in the Social Security system is looming. The enormous baby boom generation will bankrupt the Social Security system simply because there are so many of them relative to younger generations, who will be paying into the system. This article critically examines one of the proposed solutions: private investment accounts. It is taken from my recent book Cycles in American Politics, which contains a more extensive treatment of this issue.

In the absence of steadily rising real wages, keeping social security solvent as it is currently structured would require tax increases, cuts in benefits, or both. Conservatives argue that private savings are the preferred way to provide old age insurance. Their argument is the pay-as-you-go nature of social security lacks the investment aspects of retirement programs based on private savings. In pensions or 401(k) accounts, retirement savings are used to produce an investment return that funds retirement instead of relying on future payers into the system. If individuals invested their social security contributions, they would get much more retirement income than they do under social security as presently structured.

To present this argument more explicitly, I will start by considering the problem of a worker making the median wage saving for retirement. I will put everything in real (constant dollar) terms to avoid having to deal with inflation. Since real wages have not risen in the last 30 years, I will assume that the real income of the worker saving for retirement is constant over the saving period. The question becomes: how much must one save and for how long, in order to provide a given retirement income until death? This question can be formalized a follows. Assume a worker invests S constant dollars per year for T₁ years at a real return of I% and expects to live T₂ years after retirement. What retirement income (P) can he expect?

A solution to this problem is given in Figure 1. Here it is assumed that saving starts at age 40 and retirement assets are planned to last until age 90. In this case, T₁ is the period from age 40 to the retirement age and T₂ is the period from retirement to age 90. Thus, selecting retirement age fixes both T₁ and T₂. Retirement payments (P) are directly proportional to the amount saved (S). Thus, what one needs to know is the ratio P/S, which appears on the vertical axis of Figure 1. Retirement age is on the horizontal axis. The only parameter left is the investment return (I%) which is expressed as a series of curves for 2% through 10% real returns.

As an example, consider a worker who expects to retire at age 65 and invests his money in safe government bonds paying a real return of 2%. Moving along the 2% line in Figure 1 gives P/S of 1.6 for retirement at age 65. Let us assume the worker plans to use his savings to provide income equal to 50% of his working income after saving (this is the income he currently uses to live on and pay his taxes). Any additional income will have to come from part-time employment. If he saves 24% of his current income (S=24%), he will obtain a retirement income of 38% (P=1.6 x 24%) of his income. This 38% is one-half of his after-savings income of 76% (=100%-24%). Thus, the solution to the problem of how much to save to meet this 50% income requirement in 25 years starting at age 40, is to save 24% of one's income. This assumes a 2% real return on savings.

Figure 1. Pension as function of real rate of return and retiring age (saving starts at 40)

Most workers would balk at the idea of saving 24% of their income. This is, after all, nearly twice their and their employer's contribution to social security. One way to reduce the savings requirement would be to retire later. If a retirement age of 70 is chosen, P/S becomes 2.5. In this case, he would have to save only 17% of his income. This is better, but still not very attractive.

Another thing he can do is start saving earlier. Figure 2 shows the same type of graph except saving is assumed to begin at age 25. In this case, retirement at age 65 would require a savings rate of 14%, while retiring at 70 would require only 10%. Another way to reduce savings requirements would be for the worker to pay into a retirement fund with an insurance feature that paid an inflation-adjusted income upon retirement until death based on fixed contributions while working. In this case, he need only save enough to cover retirement until the average age of death (the insurance feature protects him if he happens to live longer than the average). On the other hand, if he dies early, his heirs receive none of his savings, these go to pay the benefits of those who do live longer than average. Life expectancy at age 65 is about 83. Using 83 as the age of death, 25 as the age when contributions begin, and assuming a fee of 0.5% from the insurance company to run the program, retirement at age 65 would require a 12.5% savings rate, about the same as the 12.4% for social security. In order to do better than social security one must invest in something having a higher rate of return.

Figure 2. Pension as function of rate of return and retirement age (saving starts at 25)

Stocks provide the best historical return, close to 7% in real terms since 1802.¹ Using 6% as a more conservative estimate for expected real return from stock investments, Figure 1 shows a P/S ratio of 4.5 for starting saving at 40 and retiring at 65. With this ratio, a savings rate of only 10% would be required. This result is clearly superior to the deal one gets from social security and serves as the basis for comparisons between private savings and social security.

Stocks are risky and do not produce a 6% real return over all time periods. One can deal with risk by employing a balanced portfolio consisting of both stocks and bonds. A 50% mix of each would produce an expected return of about 3.5%. The same 10% savings, if begun at age 28, would provide the same benefit as the all-stock strategy. So it seems that private savings soundly beat the public pension program and that the conservatives are right. Social security should be scrapped in favor of individual savings accounts invested in a broad-based stock index and a diversified mix of bonds.

There is a problem with this analysis. Although perfectly valid for an individual, it is not valid for the entire population. If the entire population attempted to save for retirement in this way, the investment return of stock and bonds would fail to deliver much above a 2% real return. In other words, any broad-based program of mandatory individual savings accounts to fund retirement would create a situation much like the first example of a worker using government bonds as an investment vehicle. That is, a social security return is a realistic return for a universal retirement program.

Why an investment-based universal retirement system cannot work

The reason for this counterintuitive result is subtle. To show this I will construct a hypothetical investment-based retirement system for the entire population. Using this example, I will gradually consider all the elements that would be needed in such a system and then show how this system provides an inferior result to the current system.

The first thing to realize is that investment is not the same thing as savings. Savings refers to the accumulation of income not spent. Savings are accumulated in the form of highly liquid vehicles such as bank accounts and money market funds. These types of assets have shown a real return of about 1% in real terms over the course of the twentieth century.² This return is even lower than the 2% return from social security and so are completely unsuitable as an investment vehicle for retirement.

In order to obtain a higher return, the retirement saver must purchase investment products. This investment product can be a variety of things: stocks, bonds, real estate, commodities, collectibles, etc. Investment products provide return in two ways. Some pay an income directly to their owner. Examples include most bonds, some stocks, and rental property. Investment products can also produce a capital gains (loss) return through increases (decreases) in the price of the investment between purchase and sale.

A universal retirement system employing personal savings would feature two populations: workers and retirees. Workers accumulate investment products. They are net buyers of investment products; both through their own savings and by reinvesting their investment income. Retirees spend their investment income and are net sellers of investment products (for the capital gains).

The income received by retirees comes from capital gains and investment income. Capital gains income comes from the sale of an investment product. The net retirement income obtained through capital gains for retirees as a group comes from the net purchase of the associated investment products by workers as a group. It is not difficult to see that retirement income produced via the capital gains mechanism represents a net transfer of income from workers to retirees. If there are many workers looking to accumulate assets for retirement relative to the number of retirees selling them, the price of the asset will be high and the retiree will get a large income from his sales. If there are few workers per retiree, the assets will fetch a low price and the retiree will obtain a small income from his sales.

Using capital gains to finance retirement involves a transaction similar to that of social security. Retirement income is provided directly from worker incomes. To the extent that retirees are relying on capital gains returns, they are relying on a pay as you go system. The big difference is that under social security, workers are required to pay into the system. Under a saving system, workers can chose not to buy the investment products retirees are selling at the price the retirees need to obtain to get a decent capital gains return. This makes relying on capital gains more risky than social security.

Over the long run, the amount of return that can be achieved via capital gains will be limited by the increase in the ability of the buyers to afford assets. As the population gets richer, as measured by growth in real GDP per person, the price of assets should rise in tandem. Indeed, the average capital gains return from stocks over the last 200 years has been a bit under 2% in real terms, about the same as the growth rate in GDP per person.³

Pay-as-you-go social insurance programs generate this same return by the growth of taxable income over time (which increases contributions per worker) and by investing surpluses in government bonds (which relies on future increases in GDP that will yield increased tax revenues in the future). Both social security and capital gains returns assume that future contributors (buyers) will be richer and so able to pay more in taxes or more for assets.

But the private savings mechanism also makes use of investment income. Unlike capital gains income, investment income is paid regularly; it does not depend on the willingness of workers to buy assets. Table 1 shows a breakdown of personal income by source. Three income sources in Table 1 correspond to investment income: interest, dividend and rental. In total, they amount to 1.67 trillion dollars. The income received by working people is roughly the sum of employee compensation, social insurance and self-employed, which totals to $6.97 trillion. Investment income is then equal to 24% of working income. If we assume that retirees only require 70% of the income of working people, existing investment income is sufficient to support a retired population equal to 34% of the working population, or about three working people per retiree (less than today's ratio of 3.4 but more than the 2.1 expected in several decades).

Table 1 Sources of personal income for year ending September 30, 2003

To this investment income, I now add capital gains income. Significant capital gains income would be realized from the gradual sale of retiree investment products to workers. An important point needs to be made here. In the example I am developing here, it is assumed that the retirement assets of a particular retiree are not "owned" by that retiree. That is, his heirs do not inherit upon the death of the retiree. Upon the death of a retiree, the income from his retirement assets would go to his spouse or if he has no living spouse, to the retirement trust. That is, retirement assets would be analogous to social security benefits, which can go to a spouse after death of the primary earner, but not to his heirs. The way this would work is that upon retirement, a worker would buy an annuity that would provide his and his spouse's retirement income for the remainder of their lives. So the investment and capital gains income that I refer to as being received by the retiree would actually go to the annuity rather than the individual retirees. The general results I obtain form the subsequent arguments will still be the same. The annuity is still conducting investment activities on the behalf of retirees and with their accumulated assets.

Since capital gains income is pay as you go, it would work (on average) much like social security does today. Equation 1 presents an expression that gives retiree capital gains income (CGI) as a function the ratio of workers to retirees (R) and the fraction of worker's income used to purchase assets for retirement (T). T plays the same role as the social security tax rate.

1. CGI = R T

Assuming a worker's contribution equal to the current social security tax rate of 10.6% (this figure does not include contributions for disability), capital gains income can be expected to provide about 22% of median income to the median retiree. Subtracting this from the 70% requirement gives 48% of median income to be provided from investment income. Investment income is then enough to provide the rest of the income for a population of retired people equal to 50% (=24%/48%) of the working population, that is, a ratio of workers to retirees (R) of two. Thus, this investment program could work if it held all of the income-producing assets.

This analysis assumes that all investment income goes solely to the retiree’s annuity trust. Strictly speaking, this is not be the case because a fraction of income-producing assets would necessarily be owned by workers still saving for retirement. However, since workers are accumulating assets, all investment income generated by their assets goes to purchase additional assets from retirees, who are net sellers of retirement assets. That is, all investment income produced by worker’s assets provides additional capital gains income to retirees beyond that accounted for by equation 1. The net effect is that all investment income goes to retirees.

What I have ignored so far is that investment products already have owners. In order to use this investment income to fund a universal retirement system, it is necessary for workers to initially purchase all of the investment products from the current owners so that they end up in the annuity assets when the workers retire. This means that additional funds will have to be raised through general taxation to provide retirement income to existing retirees while this new investment-based system is being implemented. Let us assume that the necessary taxes are raised and used to provide this income until retirees who have participated in the investment-based system during their working years now fill most of the ranks of the retired. Will this sort of investment-based global pension program work?

The answer is no. To see this we must consider the impact of an investment-based global retirement program on the price level of investment products. The price of investment income (which is what workers are buying for their retirement) is simply the inverse of the associated product's yield. For stocks, the relevant yield is the dividend yield, which has averaged 4.4% since 1900. For bonds the yield is the interest rate, which has averaged about 5.3% since 1900 for a mix of corporate and government bonds. This means that the average price for the dividend income in Table 7.1 would be 10.6 trillion dollars, while that for the interest income would be $20.4 trillion. Assuming a yield from rental property similar to that for bonds, an estimate of $2.3 trillion for rental income is obtained. In total, assuming investment products sold at their historically average price, the price of all the investment products underlying the $1.67 trillion of investment income is about $33 trillion. Our hypothetical retirement system would have to buy all of these assets. The current outlay of $770 billion for social security is enough to buy 1/43^rd of these assets at their average prices every year.

As the economy grows the amount of investment income it generates will rise and so will the value of investments behind the income. The outlay for social security will grow in tandem, however, and should continue to be about 1/43^rd of the total value of investment products at their average historical price.

So it would seem that after an initial 43 year period, during which time general tax revenues would be used to provide retirement income to those grandfathered in under the old system, the same social security contributions presently provided by workers would buy the investment assets needed start up this hypothetical investment-based retirement approach. Afterward, sufficient investment returns would then be generated in our economy to fund retirement for everyone using this investment-based system.

There is still a problem. Although 43 years of public savings are sufficient to purchase all the investment assets (at their historic average price) needed for the system to work, in practice, if any trust fund attempted to do this, they would find that they could not buy the needed assets. The reason is that these investments have current owners who would receive the $770 billion of worker contributions each year through sale of their assets. The sellers of investment products would not spend the proceeds of the sale, the fact that they owned the investments means that they have already chosen not to spend the money invested in them. Instead, they would re-invest the proceeds. As a result, most of the $33 trillion of private wealth currently in investment products would stay in investment products. An additional $33 trillion from workers would go into investment products as well, with the result that after 43 years of contributions, the trust fund will only own about half of the investment products, and will receive about half of the investment income. There will be twice as much money tied up in investments, meaning that their average price will have doubled. What has happened is simply that a new player (workers retirement savings) has entered into the markets for investment products, driving up demand for investment income. The price simply rises in response.

So in reality, the requirement to buy the investment products means that the global retirement program I am discussing here will end up with roughly half the investment products that it set out to purchase. It will obtain half of the capital gains income as well. This would be enough to fund retirement for a retired population less than 26% of the working population, that is, the ratio of workers to retirees (R) would be about 3.8, higher than today. A savings based program requires a larger ratio of workers to retirees to work than does Social Security. The present system does a better job than the hypothetical savings-based system does without the transition costs.

The key idea here is that an investment-based universal retirement system provides no benefit relative to pay-as-you-go, as long as there is a requirement to buy the investment products. Indeed, the result of such buying would be to drive up prices of investment products, reducing yield. This would create a highly stimulatory environment that would make formation of financial bubbles more frequent and possibly destabilize the financial system. We have already seen the effect of such a program on a small scale in the 1990's stock market. Contributions to 401(k) accounts invested primarily in stocks swelled stock valuations to lofty levels in 2000. As a result, stocks bought in 2000 will return less than money market funds for at least 20 years.³ Should individual retirement accounts funded through social security contributions become reality, long-term stock returns will likely become permanently depressed (or at least as long as the program is in place).

The reason why private savings fails for a universal system of retirement, but succeeds for individuals, is this competition for assets. The existing social security system provides a significant fraction of all retirement income. It operates completely outside of the financial markets and so it does not distort market prices upwards. Of course, social security does not provide all retirement income. Many people receive pensions from their former employers and many others have private savings such as 401(k) accounts. These alternative retirement systems do compete for investment products. If the above analysis is correct, the growth of tax-favored individual retirement accounts in recent decades should have driven up the price of investment products, particularly stocks (the assets with the highest historical return). This would show up as reduced dividend yields, which indeed has been the case as shown in Table 2. In this table, dividend yields from 1871 to 2003 were broken into seven successive 19-year periods and the average for each period calculated. The yield from the most recent period clearly stands out as 3.3 standard deviations lower than the average value.

Table 8.2 Average dividend yield over successive 19-year period

Advocates of investment-based retirement generally do not advocate a broad-based government-operated retirement program that makes use of investments. What is typically suggested is that individuals be allowed to place a portion of their social security contributions into self-directed individual investment accounts. The above analysis still holds for such a scheme. Diversion of social security funds into financial markets would still increase prices and reduce yields. The average result will be the same as that for a global system, inferior to the current approach. Already, the proliferation of 401(k)s has probably reduced future long-term real stock returns by about a third. As workers inject more funds into financial products, the lower their future returns (as a group) will fall. Some individuals will do well (above average) and retire early, requiring that others do very poorly (below average) and be forced to work until they die. It is hard to see any advantage of such a system.

The current system is not just better for those who currently rely on social security for the bulk of their retirement. It is also better for those (such as myself) who are using private savings for retirement. By keeping social security funds out of the markets, returns are higher than they would otherwise be, and it is possible for individuals willing to save to obtain good investment returns on average and so to retire earlier than those relying on social security. Social Security rewards thrift by providing attractive returns to those willing to forego current spending. Diverting social security funds into the markets would create a situation like that in Japan, where workers save a huge fraction of their incomes and investment returns are poor.

The key idea here is the economy can only produce so much investment income on a per worker basis. Table 1 suggests that about one fifth of the income deriving from the output of workers is available to investors; the other four-fifths goes to the workers. As the economy grows, investment income rises, but so does other income. And with this rising income come increased expectations for retirement income. Thus the pie is fixed and must be split between workers saving for retirement and private investors. Currently, the majority of workers do not save anywhere near the amount required for retirement, most will rely on social security to provide a substantial fraction of their retirement income. Social security, because of its pay-as-you go nature, does not and has never competed with private investors for investment products, which has resulted in the high average yields for investments cited earlier.

Should workers now be required to invest their social security payments in investment goods a new competitor for investment goods will enter the market, one with substantial buying power. The result will be rising prices (and falling yields) which will depress investment returns. Today, with social security out of the investment business, workers are told to expect 10% returns from stocks. Given a 3% inflation rate, there is a zero probability that stocks will produce a 10% return over the 2000-2020 period.³ Even from today's lower market levels, such a return is almost certainly not going to happen. The reason for this is that stock prices rose (and yields fell) so much over the past couple of decades. Should social security funds start to be invested in stocks and bonds at any meaningful level, dismal investment returns will extend further into the future, possibly becoming permanent.

Should investment of social security funds begin, there will be winners. Existing owners of financial assets will see a one-time bonanza as the prices of their assets are bid up. Even bigger winners will be those, such as corporate executives, who can create new financial assets at will (e.g. incentive stock options) which can then be sold to workers trying to accumulate investment assets for retirement. This was seen on a moderate scale during the 1990's boom when vast quantities of shares were created by corporate insiders and sold for billions to a gullible public. With social security funds flowing into the financial markets an opportunity for much larger transfers of wealth from the public to insiders exists. Other winners would include those who sell investment products such as brokers, investment bankers and securities dealers. It is not surprising such a policy would have a number of strong advocates. Those who advocate for investment of social security funds into investment products often stand to gain.

References:

1. Siegel, Jeremy, Stocks for the Long Run: The Definitive Guide to Financial Market Returns and Long-term Investment Strategies, New York: McGraw-Hill, 1998.

2. Dimson, Elroy, Paul Marsh and Mike Stanton, Triumph of the Optimists: 101 Years of Global Investment Returns, Princeton NJ: Princeton University Press, 2002.

3. Alexander, Michael A. Stock cycles: Why Stocks Won't Beat Money Markets over the Next Twenty Years, Writers Club Press, 2000.