Individual Accounts: What Rate of Return Would They Earn?
CRS Report for Congress
Individual Accounts: What Rate of Return
Would They Earn?
July 25, 2005
Brian W. Cashell
Specialist in Quantitative Economics
Government and Finance Division
Specialist in Macroeconomics
Government and Finance Division
Congressional Research Service ˜ The Library of Congress
What Rate of Return Would They Earn?
It has been proposed to add individual accounts to Social Security in which
investors could hold private securities. Calculations that project the earnings of
individual accounts typically presume that they will earn a rate of return equal or
close to the historical rate of return. But is there evidence that future rates of return
will differ from history in predictable ways?
Since the mid-1990s, equity (stock) prices have been consistently above the
historical norm in relation to corporate earnings and dividends. This implies that
future rates of return would be below the historical average unless future earnings
and dividends grow more quickly than they have been. One reason why current stock
prices may be higher and future rates of return might be lower is because of a decline
in the equity premium, which is the difference in rates of return between stocks and
bonds that investors require to be willing to bear the additional risk of equities.
Increasing rates of stock ownership and accessibility for the average investor have
suggested to some that the equity premium may have fallen.
The projected decline in the growth of the labor force would reduce the
economic growth rate, all else equal. Slower growth in the economy is likely to
mean slower growth in corporate earnings, which could lower the rate of return.
Ultimately, rates of return are determined by the equilibrium between saving and
investment. Many have predicted that the saving rate will decline as the baby
boomers begin to draw down their savings to finance their retirement. But at the
same time, investment demand is likely to grow more slowly because of the decline
in the growth rate of the labor force. If the growth in saving declines more quickly
than the growth in investment, rates of return would rise. But the opposite is equally
plausible and would lead to lower rates of return.
Under current policy, budget deficits would become much larger in the future,
and this would be expected to increase rates of return as deficits “crowd out” private
investment. Individual accounts could also influence the national saving rate. If they
are deficit-financed, then public saving would fall and private saving would rise,
leaving public saving — and, hence, rates of return — unchanged. If they are
financed through higher taxes or lower spending today, then national saving would
rise, and rates of return would fall. However, even if individual accounts left
national saving and overall rates of return unchanged, they could push up the demand
for stocks, thereby lowering their rate of return, while increasing the supply of U.S.
Treasuries, thereby raising their rate of return.
Comparing the rates of return of equities and other securities directly is
analogous to comparing apples to oranges. Standard risk adjustment techniques
would set the rate of return on equities equal to U.S. Treasuries. This is the only way
that a risky return can be directly compared to the risk-free Social Security benefit
offsets that accompany the individual accounts. This report will not be updated.
What is the Historical Rate of Return?.............................1
Effects of a Changing Equity Premium.............................3
The P/E Ratio, P/D Ratio, and Future Rates of Return.................6
Economic Growth and Rates of Return............................10
Capital Income Growth Is Unlikely to Exceed GDP Growth.......10
The Rate of Economic Growth Is Expected to Fall...............12
Slower GDP growth and rates of return........................15
Demography, Saving, and Investment.............................16
Effect of Future Government Budget Deficits on Rates of Return.......24
Effect of Individual Accounts on National Saving and Rates of Return...25
Effect of Individual Accounts on Relative Rates of Return.............28
List of Figures
Figure 1: Price/Earnings Ratio and Price/Dividends Ratio, 1946-2004........8
Figure 2. Capital and Land Share of National Income.....................12
Figure 3. Demographics and Equity Prices.............................18
Figure 4: Change in Unified Budget Balance From Introducing Individual
Accounts in Plan 2 (67% Participation Rate).......................27
List of Tables
Table 1: Real Rates of Return for Equities Over Different Historical Periods...2
Table 2: National Income, 2004......................................11
Table 3: Productivity Growth, Selected Periods.........................14
Table 4: Long-Run Real GDP Growth Projections.......................15
Table 5: Percentage of Families Who Saved, by Age of Family Head, 2001...19
Table 6: Household Wealth by Age Class, 2001.........................19
Table 7: Percent of Net Worth by Percentile Group, 2001 .................21
Individual Accounts: What Rate of Return
Would They Earn?
A recurring question in the debate on adding individual accounts (IAs) to Social
Security is how large the typical account would be upon retirement.1 Making this
calculation requires assumptions about the contribution rate, length of time the
account was held, and rate of return earned on the account (net any administrative
fees and taxes). Because of the power of compound interest, changing the rate of
return on the account leads to greatly different outcomes. For example, $1 invested
every month for 40 years accumulates to $1,526 if earning 5%, $2,625 if earning 7%,
and $4,681 if earning a 9% rate of return. Many estimates of how large the IAs
would be have assumed that assets in the account would earn a rate of return equal
to the historical average. For example, in the official actuarial estimates of various
IA plans, the Social Security Administration actuaries assume that equities will
average a rate of return of 6.5% and government bonds will earn 3%. Some
proposals assume that the accounts would hold equities (stocks), others bonds, and
others some combination of the two. Since stocks historically have a higher rate of
return than bonds on average, the rate of return earned by the IA’s would increase as
the fraction held in equities increased.
But is the assumption that future rates of return will mirror historical rates of
return a valid one? Or are there ways in which the future may differ from the past
that will have a predictable effect on the rate of return? This report will analyze
demographic, economic, and financial reasons why the assumption might, or might
not, be valid.
What is the Historical Rate of Return?
Before exploring factors that might influence the future rate of return, it is useful
to discuss what the rate of return has been historically, and how much confidence we
can have that the past is a useful guide to the future.2 As can be seen in Table 1, the
historical real rate of return is sensitive to the time period under consideration. The
average rate of return over the past 20 years was quite a bit higher than over the past
1 For more information on individual accounts, see CRS Issue Brief IB98048, Social
Security Reform, by Dawn Nuschler; CRS Report RL31498, Social Security Reform:
Economic Effects, by Jane Gravelle and Marc Labonte.
2 This report considers only factors that would influence gross rates of return.
Administrative costs, annuitization fees, taxes, and so on, would make the net rate of return
on the individual account lower than the gross rates of return measured by the historical data
calculating an average return.3 For example, in the 1990s, the real rate of return on
equities averaged an impressive 14.8%, but in the 1970s, it averaged only 1.4%. Not
only are rates of return highly volatile, but they can be highly persistent: gains or
losses can persist for several years, and take several years longer to be reversed. This
implies that the retirement income of an individual whose IA was invested in equities
could change significantly based on the market’s performance in the period before
retirement. For example, a worker who invested $1 in 1959 and withdrew it upon
retirement in 1999 would have had a balance of $16.55, while a worker who invested
$1 in 1962 and retired in 2002 would have had a balance of $8.05. In other words,
average rates of return mask the wide range of returns that actual individual accounts
(holding the same assets) would earn over time.4
Table 1: Real Rates of Return for Equities Over Different
Period:Average Annual Real Rate of Return:
Source: Ibbotson’s; Haver Analytics
Notes: December 1925 is earliest available data. Data are calculated from December to December
for each year.
While basing projections on 50 years of historical data or more may seem to
imply a high degree of certainty, this evidence is actually less robust than one might
expect. Because of the significant volatility in the annual data, the standard error
associated with the data is very large. For example, Cochrane estimates that from
1946 to 1996, the 95 percent confidence interval for the average annual return would
be between 3% and 13%. Over longer holding periods, the volatility diminishes, but
then there are fewer observations from which to make statistical inference, which
3 The meaning of “long run” depends on the context. In some cases, it may simply refer to
events that are independent of the fluctuations of the business cycle. In the case of capital
stock adjustments the long-run may be as long as 30 years. With respect to technological
progress it may be even longer. In this report, the long run is defined by the 75-year
projections of the Social Security trustees.
4 For examples of how much IA performance would have varied historically, see CRS
Report RL31324, Social Security Reform: The Effect of Economic Variability on Individual
Accounts and Their Annuities, by Geoffrey Kollman, Dawn Nuschler, and Patrick Purcell.
increases uncertainty. For example, for the rate of return over 20-year holding
periods, there are only five statistically independent observations over 100 years of
Cochrane points out two other problems with extrapolations made from
historical data. First, economic outcomes were much better here than abroad. A
long-term average return based on a number of countries would be much lower than
the average for the United States. In this sense, investors in U.S. stocks were lucky
because U.S. stocks turned out to be less risky than in most other countries. That
does not guarantee that the U.S. stocks will be luckier than foreign stocks in the
future. Second, the data suffer from “non-normality” over the past 50 years. During
that time, the United States has not experienced any rare calamities such as economic
depression, natural disaster, or government breakdown that would bring down the
average returns on equities. These rare events may not occur in the future either, but
basing future projections on a historical experience that is not comprehensive of the
entire range of possible experiences may not be suitably representative of the full
range of possible outcomes that IA holders could face.
Effects of a Changing Equity Premium
There may be reasons to believe that the rate of return on equities relative to
other assets could be different in the future than it has been in the past. (For a
discussion of relative rates of return, see the text box on the next page.) The equity
premium refers to the difference in average rate of return between equities and other
assets. Siegel calculates that between 1802 and 1998, the ex-post equity premium
averaged 3.5 percentage points using data for equities and long-term government
bonds, and between 1946 and 1998, the premium rose to 6.5 percentage points.6
5 John Cochrane, “Where Is the Market Going? Uncertain Facts and Novel Theories,”
Economic Perspectives, Federal Reserve Bank of Chicago, Nov./Dec. 1997, p. 3.
6 Jeremy Siegel, “The Shrinking Equity Premium,” Journal of Portfolio Management, Fall
1999, p. 12. The rise in the later period was largely because bond returns fell over that
period while equity returns remained stable. The ex-post equity premium is measured by
comparing actual equity returns to actual bond returns. Ideally, we would like to measure
the ex-ante equity premium, the premium that investors expected before the fact when they
made their investment, rather than the ex-post premium, but that is impossible.
Comparing Asset Returns Through Risk Adjustment
The fact that equities have a higher rate of return than bonds on average begs
the question of why. If an IA held in equities would grow faster than an IA held
in bonds, then why would anyone hold bonds? For that matter, why wouldn’t
investors holding bonds realize this, and trade all of their bonds for equities
tomorrow? If market participants are rational, as economists believe, some factor
other than rate of return must be holding investors back from converting all of
their bonds into equities.
Economists explain the different rates of return offered by different financial
instruments in terms of a risk-return tradeoff. Equities are more risky than
corporate bonds because a bondholder is promised a specific rate of return, and
bondholders are paid off before equity holders if a firm enters bankruptcy.
Investors are willing to hold the securities of well-established firms at lower rates
of return than “fly-by-night” firms because investors are more certain that well-
established firms will meet the financial targets that they set. Investors will hold
bonds of the federal government (U.S. Treasuries) at a lower rate of return than
corporate bonds because only the federal government can levy compulsory taxes
or print money to service its debt.
Thus, comparing the rate of return on different financial instruments, such as
equities and bonds, is like comparing apples and oranges, since rates of return
alone tell us nothing about risk. To make an apples to apples comparison, rates
of return must somehow be adjusted to cancel out differences in risk between
different financial instruments. This leads to an important insight: if individuals
are rational, then the higher rate of return on equities than bonds is exactly equal
to the loss in utility (happiness) that the average investor suffers from taking on
the additional risk of holding equities rather than debt. If that is true, then all
financial instruments earn the same rate of return on average after adjusting for
risk. In other words, the average stockholder is no better off than the average
bondholder even though the former can expect to receive a higher rate of return
than the latter. If it were not true, and the utility enjoyed from earning a higher
rate of return on equities exceeded the utility lost from the higher risk of holding
equities, then investors would sell bonds, pushing their rate of return up, and buy
equities, pushing their rate of return down, until the discrepancy no longer existed.
For that reason, in its analysis of individual accounts, the Congressional
Budget Office (CBO) has assumed that individual accounts, regardless of what
types of assets they hold, earn a rate of return equal to government bonds. This
is the only way that the earnings on the account, which are uncertain, can be
meaningfully compared to the benefit offsets proposed by the Administration and
others, which are always the same no matter what the IA earns. (The
Administration has proposed that Social Security benefits be offset by the amount
of taxable payroll diverted to individual accounts, compounded at a rate of 3% a
There is a vast literature in financial economics on an “equity premium puzzle.”7
This literature finds that the difference in average rate of return between equities and
bonds is too great to be explained by any theoretical representation of risk aversion.
In other words, the increased riskiness of equities compared to bonds is simply not
great enough to merit the return premium on equities compared to bonds. Of course,
economists use highly simplified models to quantify human behavior which nobody
knows how to properly quantify, so the problem could be with the model, and not
with human behavior.8 But if these models are correct, it suggests that investors’ fear
of equities is irrationally high, and, by extension, standard risk-adjustment techniques
are undercounting the gains to be had from holding equities if the rate of return on
those equities is adjusted to equal the government bond rate.9
Regardless of what gains could be had from holding equities compared to
bonds, historically investors have been indifferent between the two at prevailing
interest rates, as irrational as that decision may have been. If they continue to be
indifferent between the two in the future, it is difficult to see how the government can
claim to “know better” than the individual about his investment decisions and eschew
standard risk-adjustment techniques on the grounds that investors are irrational. But,
again, this line of reasoning assumes that financial markets will operate the same in
the future as they have in the past. What would be the implications for rates of return
if the equity premium were smaller in the future, or had already become smaller, than
it had been in the past?
Using historical rates of return in projections of IA earnings assumes that the
historical equity premium will remain constant in the future. Yet it is well known
that ownership of equities has changed markedly in past decades. Equities have
become more attractive to small investors through innovations such as discount
brokers and mutual funds, and the rise of defined-contribution pensions. All of these
innovations have reduced the transaction costs associated with buying and selling
equities for the average investor.10 In addition, there is at least superficial evidence
from popular culture that the average investor’s fears of equities may have lessened
(at least before the stock market crash of 2000). In 1989, 31.6% of households
owned equities; by 2001, the figure had risen to 51.9%. Equities increased from
7 For a survey, see Narayana Kocherlakota, “The Equity Premium: It’s Still a Puzzle,”
Journal of Economic Literature, vol. 34, no. 1, Mar. 1996, p. 42.
8 As Cochrane explains, if the economic models are calibrated to so that people are risk
averse enough to generate the historical equity premium, then other variables in the model,
such as interest rates, would fluctuate wildly, contrary to historical experience. John
Cochrane, “Where is the Market Going? Uncertain Facts and Novel Theories,” Economic
Perspectives, Federal Reserve Bank of Chicago, Nov./Dec. 1997, p. 3.
9 It also implies that the government could improve economic efficiency if the current share
of wealth held in equities is inefficiently low at present, and individual accounts increase
that share. This outcome is not a given, however, since no current proposal requires that a
proportion of individual accounts be held in equities. Individuals may prefer to continue to
hold the same share of their IA wealth in equities as they currently hold privately.
10 Some economists have argued that once transaction costs are netted out, the equity
premium is lower than it first appears. If transaction costs have fallen, then gross returns
relative to bonds would also fall.
33.7% of financial wealth in 1992 to 56.0% in 2001.11 The generation that lived
through the Great Depression may have become overly averse to investing in
equities, but generations since then do not base their decisions on first hand
knowledge of that experience. If equities have become relatively more attractive to
investors, this could imply a lower equity premium moving forward.
If the equity premium has fallen, then the absolute rate of return on equities can
be expected to fall and the absolute rate of return on less risky assets to rise, all else
equal.12 Were the equity premium to disappear entirely, the fall in equity returns, at
the extreme, would be significant. The average annual return on equities from 1871
to 1993 was 3.9 percentage points higher than short-term commercial paper (a form
of corporate debt), and in the absence of an equity premium it has been predicted that
it would be only 0.1 percentage points higher.13
Is there any empirical evidence that the equity premium has fallen? Looking at
average rates of return in recent years is not particularly useful because the stock
market has gone through a sharp fall, followed by a relatively smaller (but still large)
rebound. These short-term fluctuations probably give little evidence of longer term
trends. What is needed is some reliable estimate of future expected returns. One
possibility would be to see what professional forecasters are predicting for future
returns, although few forecasts extend more than a few years. A recent Wall Street
Journal article reported that a survey of 10 financial economists predicted that the
average real return on equities from now to 2050 would decline to 4.8%.14 Global
Insight, a private forecasting firm, provides a 25-year projection of equities, and
predicts that they will rise 6.4% annually in real terms.15
The other possibility is to look at current valuations and attempt to estimate
what they imply about investors’ views on future valuations. This strategy will be
explored in the next section.
11 Board of Governors of the Federal Reserve, 2001 Survey of Consumer Finances, Jan.
12 In the transition period to a lower equity premium, rates of return would be expected to
temporarily be unusually high because prices would rise. Prices would rise because less-
cautious investors would now be willing to pay more for a share of the same future earnings
and dividends. At the time, many financial analysts argued that this could explain why rates
of return were unusually high in the 1990s. If this transition period were included in the
historical average, then the difference between the historical average and future returns
would be even greater in the presence of a lower equity premium.
13 Nicholas Barberis and Richard Thaler, “A Survey of Behavioral Finance,” in G.M.
Constantides et al., Handbook of the Economics of Finance, Elsevier Science, 2003, Ch. 18.
14 Mark Whitehouse, “Social Security Overhaul Plan Leans on a Bullish Market,” Wall
Street Journal, Feb. 28, 2005, p C1. Of the 10 economists surveyed, eight were private
sector economists and two were academic economists.
15 Global Insight, The U.S. Economy: The 25-Year Focus, First Quarter 2005.
The P/E Ratio, P/D Ratio, and Future Rates of Return
Standard financial theory defines current equity prices as equal to the present
discounted value of future price appreciation and dividend payments. In other words,
someone purchasing a stock today is purchasing the expected future capital gains and
dividend payments associated with that stock, discounted to the present. Therefore,
today’s prices should reflect investors’ expectations of future rates of return.
An important measure in determining future rates of return is future corporate
earnings, since earnings can be used by the firm to either pay dividends, buy back its
stock, reduce its debt, or increase the firm’s capital stock; any of which could cause
its stock price to rise. Unfortunately, there is no way to accurately predict the path
of future dividends or future earnings — the higher one believed them to be, the
higher the expected rate of return would be for a given stock price.
For analytical purposes, many financial analysts assume that future dividends
and earnings will be equal to current dividends and earnings.16 By dividing share
price by earnings per share, they calculate a price-earnings (P/E) ratio, and by
dividing share price by dividends per share, they calculate a price-dividends (P/D)
ratio.17 If the P/E ratio or P/D ratio rises today, and earnings or dividends remain
constant in the future, then future rates of return will be lower.
16 This assumption would be unreasonable in a recession, since earnings are typically below
average when aggregate demand is low. But since the economy has been growing robustly
since 2003, it may be a reasonable assumption now.
17 A corporation can transfer earnings to an investor through either a dividend payment or
by buying back stock, and the effect on the investor’s rate of return should be exactly the
same. Therefore, some analysts have argued that the P/D ratio will be incomplete if
buybacks are not taken into account. If buybacks are constant over time, it would not be
problematic to make historical comparisons using the P/D ratio. But many analysts argued
that buybacks increased in the late 1990s, and this was one reason why the P/D ratio was
rising. Unfortunately, historical data on the P/D ratio that controls for buybacks are difficult
Figure 1: Price/Earnings Ratio and Price/Dividends Ratio, 1946-2004
1946 1949 1952 1955 1958 1961 1964 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003
P/E RatioP/D Ratio
Source: CRS calculations based on data from Haver Analytics
Notes: Ratios are calculated for Standard & Poor’s 500 index stocks. Earnings are measured as four-quarter
As can be seen in Figure 1, during the late 1990s, the P/E and P/D ratios were
extremely high by historical standards. At the time, some skeptics argued that this
meant that the stock market was experiencing a “bubble,” where stock prices could
not be justified by any reasonable expectation of future dividends and earnings, and
were unlikely to persist. In fact, the stock market fell 46% from peak to trough, and
a strong case can be made that the skeptics were correct. However, as can be seen
from the chart, although the P/E and P/D ratios are lower than they were in the late
1990s, they are still high compared to historical averages.18 Since the P/D ratio is
also affected by firms’ decisions whether to pay out their earnings as dividends, and
dividend payouts have fallen over time, the analysis will focus on the implications
of a high P/E ratio. There are a number of possible explanations for the currently
high P/E ratio:
!Future earnings will be higher than current earnings, perhaps
because productivity growth is now higher than it was in the past.
This implies that equity returns could continue to match historical
levels despite the higher P/E ratio.
!There is still a bubble in stock prices, and prices will have to fall
further (or stay flat for a time) before prices reflect economic
fundamentals. This implies below average equity returns for a
18 The Federal Reserve estimates a ratio of the market value of equities to a firm’s net worth
(assets less liabilities). This is similar to the economic concept known as Tobin’s Q. A high
value of the ratio would be associated with an overvalued market and imply low future rates
of return, and vice versa. This ratio follows the same pattern over time as the P/E ratio, and
prices are still relatively high by this measure as well, although a long data series is not
available. From 1986 to 1990, the ratio equaled between 44.1% and 52.5%. In the 1990s,
it rose rapidly, reaching 185.1% in 1999. In the last quarter of 2004, it stood at 96.4%.
period until the P/E ratio returns to historical levels, followed by
equity returns equal to historical averages.
!Investors require a lower risk premium than they did in the past and
will therefore pay more for a share of unchanged earnings. If so,
future equity returns would be lower than historical returns.19
Clearly, these three possibilities have very different implications for the future
earnings of individual accounts. The first possibility implies that the common
assumption that individual accounts will earn the historical rate of return is the
appropriate one. What this argument fails to explain is why current earnings are not
already as high as the present discounted value of future earnings since productivity
growth accelerated nearly a decade ago, unless one assumes that it will accelerate
even further in the future. If current earnings are already growing as fast as future
earnings will grow, then the P/E ratio should not be higher than its historical average.
From 2001:4-2005:1, real earnings per share have grown by 4.5% per quarter —
roughly three times the 1988-2005 average. It is difficult to see why earnings would
grow even more rapidly in the future.
The second possibility suggests that while stock returns may equal the historical
averages in the more distant future, in the short term returns could be well below
average — although it is impossible to predict accurately how and when those below
average returns will manifest themselves. Campbell and Shiller present evidence that
when the P/E and P/D ratios have been high historically, it has typically been falling
prices (the second scenario), not rising earnings or dividends (the first scenario) that
has brought the ratios back to their historic average over the longer run.20 This
implies that early generations of IA holders may experience low or negative rates of
return, while IAs would do well in the more distant future.
The third assumption suggests the historical average rate of return is higher than
should be expected in the future, because attitudes toward risk are fundamentally
different now than they were historically.
It is possible to derive a rough estimate of what the higher P/E ratio implies for
future earnings. From 1950 to 1999, stocks averaged a real rate of return of 10.33%.
Reichenstein breaks this return down into three parts — a dividend yield of 3.84%,
earnings growth of 3.04% and an increase in the P/E ratio of 3.41%. This implies
that if the P/E ratio remains constant in the future and dividends and earnings equal
their levels from 1950 to 1999, the rate of return would equal 6.92%. This is similar
to the Social Security actuaries’ assumption that the rate of return will equal 6.5%.
But dividend yields have been significantly lower than 3.84% in recent years. Unless
earnings growth exceeds the historical average, then Reichenstein argues that at
19 This argument is made, for example, in Eugene Fama and Kenneth French, “The Equity
Premium,” Journal of Finance, vol. LVII, no. 2, April 2002, p. 637.
20 John Campbell and Robert Shiller, “Valuation Ratios and the Long-Run Stock Market
Outlook,” Journal of Portfolio Management, Winter 1998, p. 11. The P/D and P/E ratios
have not been accurate predictors of the movement of prices over short time periods in the
current dividend yields, the rate of return will fall to 4.78%, which he finds consistent
with the view that there is now a lower equity premium. If the P/E ratio is to return
to a level more in line with historical experience, however, the rate of return would
be even lower. For example, if the P/E ratio were to fall from its recent average of
approximately 20 to its 60-year average of approximately 16 over the next 10 years,
it would reduce rates of return by 2.2%, based on Reichenstein’s calculation.21
The Social Security Administration has explicitly linked the currently high P/E
ratio and a lower equity premium to their decision to assume a rate of return in
projections that is a little lower than the historical average:
...the long-term ultimate average annual real yield assumed for equities is 6.5
percent. This is somewhat lower than the historical real equity yield over the last
several decades. A consensus is forming among economists that equity pricing,
as indicated by price-to-earnings ratios, may average somewhat higher in the
long-term future than in the long-term past. This is consistent with broader access
to equity markets and the belief that equities may be viewed as somewhat less
“risky” in the future than in the past. Equity pricing will vary in the future as in
the past. Price-to-earnings ratios were very high through 1999, and are now
lower. The average ultimate real equity yield assumed for estimates in this
memorandum is consistent with an average ultimate level of equity pricing22
somewhat above the average level of the past.
Using the P/E and P/D ratios to project future returns depends crucially on the
relationship between present and future earnings and dividends. Future earnings and
dividends, and hence returns, will depend directly on future economic growth. The
possibility of predictable changes in future growth is discussed in the next section.
Economic Growth and Rates of Return
Capital Income Growth Is Unlikely to Exceed GDP Growth. The
future long-run rate of return on equity may be affected, in part, by the long-run rate
of economic growth. Slower growth in national income would lead to slower growth
in income from capital unless capital’s share of income rises, which is equivalent to
a corresponding decline in labor’s share of income.
The historical record suggests that is unlikely. Estimates of income shares
accruing to labor and capital vary slightly because of the ways in which income is
attributed to human capital, but remain relatively stable over long time periods. Most
mathematical models designed to explain economic growth assume that labor’s and
capital’s income shares do not vary much over time.
The “functional distribution of income” describes how income is shared among
the different factors of production; labor, capital, and land. If the argument that
21 William Reichenstein, “What Do Past Stock Market Returns Tell Us About the Future?”
Journal of Financial Planning, vol. 15, no. 7, July 2002, p. 72.
22 Stephen Goss, Chief Actuary, “Estimated Financial Effects of The Progressive Personal
Account Plan,” memorandum, Social Security Administration, Dec. 1, 2003.
companies are increasing profits at the expense of labor is correct, then that should
be reflected in the way income is distributed among these factors of production.
The income for each of these factors of production is determined by the quantity
employed and the “price” for their use (for labor, wages; and for capital, the rate of
return).23 Relative income shares may vary either because of changes in the relative
quantities employed, or in their prices. In the national income accounts published by
the Bureau of Economic Analysis, income is accounted for as shown in Table 2.
Table 2: National Income, 2004
Billions of DollarsPercent of Total
T otal 8,841.0 100.0
Compensation of employees6,289.071.1
Rental income of persons153.81.7
Source: Department of Commerce, Bureau of Economic Analysis.
The amounts in Table 2 account for all of the payments to the three factors of
production; labor, capital, and land. Labor income is measured as compensation.
Profits, rent, and interest measure payments for the use of capital and land.
Proprietors’ income reflects payments to all three factors. Figure 2 shows estimates
of the property (capital and land) share of income since 1945. This measure assumes
that one-third of proprietor income is attributable to property. Income attributable
to proprietorships has declined relative to total national income over the long run.
23 Asset price appreciation is not included in BEA’s measure of capital income. This makes
a direct comparison to rates of return somewhat difficult.
Figure 2. Capital and Land Share of National Income
19 45 19 50 19 55 19 60 19 65 19 70 19 75 19 80 19 85 19 90 19 95 20 0020
Source: Department of Commerce, Bureau of Economic Analysis.
If there is a discernable trend in these data, it is one of modest decline since the
late 1940s, although there has been substantial variation and the share has risen from
relatively low levels in the mid-1970s.24 But capital’s share of income has always
been between 20% and 25%. This suggests that the effect of any lower GDP growth
on capital income is unlikely to be offset by a rise in capital’s share of income.
The Rate of Economic Growth Is Expected to Fall. Over the long run
economic growth can be accounted for by growth in the labor force, changes in the
hours worked by each worker, and the productivity (output per hour) of each worker.
Over the very long run, say more than 50 years into the future, any economic
projection might be characterized as speculation. In the nearer term, however there
is at least some basis for making an educated guess what the prospects are for
In the first place, the growth rate of the labor force can be estimated based on
population growth trends and trends in labor force participation. For the next 20
24 Poterba also finds little long-run change in the relative income shares of labor and capital,
see James M. Poterba, The Rate of Return to Corporate Capital and Factor Shares: New
Estimates Using Revised National Income Accounts and Capital Stock Data, NBER working
paper 6263, Nov. 1997, 45 pp.
25 Since the relevant time period with respect to Social Security is a very long one, short-
term cyclical variations in the unemployment rate and its effects on output can be ignored.
Long-run economic projections generally assume that the economy reaches and sustains full
employment, on average, over the forecast period.
years or so, short of any dramatic change in immigration rates or in average
retirement age, labor force growth can be predicted based on the current population
and its age distribution. Beyond that, labor force growth will depend not just on
current numbers of youth, but also on birth rates, mortality rates, and the chances that
there will be other changes in immigration rates and retirement age increase.
Historically, variations in the growth rate of the labor force have been
substantial. For example, from 1950 to 1960 the labor force grew at an annual rate
of 1.1%. Between 1970 and 1980, it grew at an annual rate of 2.6%. The large
increase was due both to the influx of baby boomers into the labor market, and also
to a large rise in the labor force participation rate of women. Since then, labor force
growth has slowed to a 0.8% rate between 2000 and 2004.
For the foreseeable future, labor force growth is expected to continue to decline,
in large part because of the aging and retirement of the baby-boom generation. The
intermediate projections of the Social Security trustees show the labor force growth
rate falling to 0.5% by 2015. In the very long run, the labor force is expected to grow
at a 0.2% rate between 2050 and 2080. Other things being equal, that would mean
a decline of 0.6 of a percentage point in the rate of economic growth.
If workers worked more hours, that would offset some of the effects of slower
growth in the labor force. But that is not likely to happen. Average weekly hours
vary by much less than the size of the labor force and have a much smaller effect on
variations in output growth. Between 1950 and 1980, average weekly hours fell at
an average annual rate of 0.4%. After 1980, hours fell at an average annual rate of
0.1%. There is less basis for reliably projecting average hours than is the case with
the labor force, but the Social Security trustees’ intermediate projection assumes that
average hours do not change, and have no effect on projected economic growth.
The third factor accounting for economic growth is average labor productivity.
Simply put, average labor productivity is the quantity of output divided by the
number of hours required to produce it (output per hour). Higher productivity growth
could potentially offset the lower growth rate of the labor force, and keep GDP
growth from slowing.
In order to make long-run projections of output, forecasters must make some
estimate of what productivity growth is likely to be over the period being forecast.
But the study of productivity has not advanced to the point where it can be projected
based on what is known now about economic conditions. Most forecasts project
productivity to continue at its current trend rate of growth; as discussed below, the
Social Security trustees are a notable exception.
The trend rate of productivity growth at any given time can be difficult to
discern because it is influenced in the short run by the business cycle. This makes
it difficult to tell whether a change in the rate of productivity growth is temporary or
indicative of a change in the long-run trend. Analysis of long-term trends indicates
that in 1973 productivity growth slowed and that in 1995 it accelerated. The reasons
for these changes are poorly understood, but it is widely believed that advances in the
production of computers and related equipment, as well as increased investment in
computers have contributed to the post-1995 acceleration. Table 3 shows how the
trend rate of growth of productivity has varied over time.
Table 3: Productivity Growth, Selected Periods
Annual Growth Rate of
Output per Labor Hour
1947 to 19732.8
1973 to 19951.4
1995 to 20043.1
Source: Department of Labor, Bureau of Labor Statistics.
Although it is now clear that productivity growth picked up beginning in 1995,
it was some time before this was thought to be part of a potentially durable shift in
the long-run rate of productivity growth. Most economic forecasters, including the
Congressional Budget Office (CBO) and the Office of Management and Budget
(OMB), underpredicted real economic growth by an average of about 2 percentage26
points in each of the four years after 1995.
Now that the acceleration in productivity has survived the end of the last
expansion and seems to be continuing into the present one, there is growing
confidence that it will persist. Whether or not that confidence is misplaced, only time
will tell. This is arguably the most important question for long-term growth
projections. Higher productivity growth means higher real incomes, which in
combination with progressive income tax rates yields higher federal revenues. As
long as Social Security operates on a pay-as-you-go basis, higher productivity growth
also extends the date of reckoning as far as the trust fund balances are concerned
because the incomes of those paying Social Security taxes would grow more rapidly
than the benefits. For example, if real wage growth increased from 1.1% to 1.6%,27
it would delay trust fund insolvency for six years. Whether or not productivity
growth continues at the rate it has since 1995 is a critical element in the trustees’28
projections of economic growth over the next 75 years.
The major difficulty in projecting productivity growth remains an imperfect
understanding of the causes of past variations. Some of the sources of labor
productivity growth are clearly understood. Increased investment and a growing
capital stock (discussed in the next section) raise labor productivity. Increased
education and training also contribute. But aside from the contributions of human
and physical capital, much less is certain. To a great extent, projections of
26 CRS Report RL30239, Economic Forecasts and the Budget, by Brian W. Cashell.
27 Board of Trustees, Federal Old Age and Survivors Insurance and Disability Insurance
Trust Funds, Annual Report, Mar. 2005, Table VI.D4.
28 Paul W. Blauer, Jeffrey L. Jensen, and Mark E. Schweitzer, “Productivity Gains, How
Permanent?” Federal Reserve Bank of Cleveland Economic Commentary, Sept. 1, 2001.
productivity still reflect the optimism or pessimism of the forecaster in the rate of
The Social Security trustees’ intermediate cost projection assumes that
productivity will not continue to grow at the same rate as it has since 1995. They
project that productivity growth will decline until about 2010, after which it will
average 1.6% through the rest of the projection period ending in 2080.
The trustees’ intermediate projections of no change in average hours, labor force
growth falling to 0.2% and productivity growth falling to 1.6% imply long-run
growth of 1.8% after 2050. If productivity growth continues at the pace of the last
decade, that projection is likely to prove overly pessimistic. However, even if
productivity growth remains at the post-1995 rate, growth would be lower than in the
last decade because of slower growth in the labor force.
Table 4 compares the trustees’ long-run projections with those of CBO, and of
Global Insight, a private economic forecasting firm. The trustees publish three
alternative projections. The intermediate-cost projection represents the trustees’ best
estimate of likely future demographic and economic conditions. The low-cost
projection assumes more rapid growth, and the high-cost projection assumes the
opposite. Global Insight projects out only 25 years, but over that period theirs is the
most optimistic forecast. The CBO forecast is not as recent as that of the trustees, but
it is similar to the intermediate cost projection of the trustees. All three forecasters
expect that the rate of economic growth is going to decline over the very long run.
Table 4: Long-Run Real GDP Growth Projections
Social Security Trustees
2005 to 20103.03.52.43.13.2
2010 to 2022.214.171.124.33.0
2020 to 20301.82.41.42.02.8
Sources: The 2005 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors
Insurance and Disability Insurance Trust Funds; Congressional Budget Office, The Long-Term
Budget Outlook, Dec. 2003; Global Insight, The U.S. Economy The 25-Year Focus, First Quarter
Slower GDP growth and rates of return. Lower GDP growth will lead
to slower capital income growth unless capital’s share of income rises, which is
unlikely to persist indefinitely. Since capital income consists of some of the same
factors — corporate earnings that are paid out in dividends or reinvested — that
determine the rate of return on equities, it seems likely that slower capital income
growth would also result in lower rates of return. Real earnings growth and GDP
growth have moved closely together historically.29
Baker, DeLong, and Krugman argue that if the growth rate of the economy
declines, as it is expected to, the return on stocks is likely to fall below the long-term
historical rate.30 Using theoretical models describing the connection between
economic growth and assets returns, the authors argue that only under unlikely
circumstances (e.g., a large shift in the distribution of income from labor to capital,
or a shift from the substantial current account deficit to a surplus) would a drop in
economic growth (caused by slower labor force growth) not cause asset returns to
Baker, DeLong, and Krugman explain the rate of return on equity as a sum of
the dividend yield, the rate of stock buybacks (companies buying their own stock
back from shareholders), and the rate of growth of dividends. Assuming the first two
remain constant, the authors consider it unlikely that a slowdown in capital income
growth would not be reflected in the growth rate of dividends, otherwise the capital
share of income would rise. If slower economic growth results in slower growth in
income from capital and slower growth in dividends, they say the rate of return is
also likely to fall.
Demography, Saving, and Investment
With given levels of technological progress and labor, most economic models
assume that the return to capital declines as the capital stock grows. In other words,
each new machine employed by a firm yields a smaller increase in production than
the last one. In the jargon of economics there is a diminishing marginal product of
capital. Thus the share of output that is saved and invested might be expected to
have some effect on the rate of return on capital, and therefore individual accounts.
This section examines the potential effects of upcoming demographic changes on the
outlook for saving and investment.
One among many potential economic consequences some fear may result from
the retirement of the baby-boom generation is a substantial decline in the stock
market and, more generally, in the prices of all financial assets. This concern is the
result of a presumption that the surge in retirements will result in a substantial sell-
off of stocks and bonds as retirees draw down their wealth in order to maintain living
standards in retirement.
29 In a bivariate regression from 1998:2-2005:1, a one percentage point increase in GDP
growth is predicted to lead to a 5.65 percentage point increase in real earnings. The result
is statistically significant at the 1% confidence level and has an R-squared of 0.18. Data are
based on S&P 500 earnings, inflation adjusted using the GDP deflator, from Haver
30 Dean Baker, J. Bradford DeLong, and Paul Krugman, “Asset Returns and Economic
Growth,” Brookings Papers on Economic Activity 2005:1, Forthcoming 2005.
From a theoretical standpoint, two common presumptions might support that
conclusion. First, at the center of most economic models of personal saving is the
assumption that individuals seek to avoid substantial fluctuations in their living
standards over the course of their lifetimes. Second is the observation that individual
incomes tend to rise during the course of their working lives and then decline after
retirement. Taken together, these two considerations lead to the prediction that most
people will tend to save relatively less when they are young and when they are
retired, and save more when they are in their peak earning years.
In this simple model, as baby boomers reach their peak earning (and peak
saving) years, it might be reasonable to expect an increase in the demand for financial
assets. But on retirement, they would gradually draw down their holdings of assets
to finance consumption in order to maintain their standard of living. Thus, when
retirements surge as the baby-boom generation ages, the increased sales of financial
assets might be expected to drive down their prices substantially.31
Figure 3 plots the proportion of the overall population aged 45 to 64 since
1950. Also shown is the Standard and Poor’s index of 500 stock prices, adjusted for
inflation. The unadjusted S&P 500 index is set at 10 for 1941 to 1943, and here it
is adjusted for inflation using the price index for gross domestic product. Anyone
looking for evidence that there is a relationship between the two might not be
disappointed. As the figure shows, it appears that, at times, the two series move32
31 See, for example, Diane Macunovich, “Discussion,” Social Security Reform, Federal
Reserve Bank of Boston, Conference Series 41, June 1997, p. 64. See also Brooks, who
derives a calibrated theoretical overlapping generations model in which the rate of return
on equities and bonds falls about 100 basis points between 2010 and 2020, before rising
about 50 basis points between 2020 and 2030. In this model, there are no bequests and
retirees consume their savings before death. Robin Brooks, “Asset Market Effects of the
Baby Boom and Social Security Reform,” American Economic Review, vol. 92, no. 2, May
32 The correlation coefficient between the two series is 0.603. As Poterba points out, this
evidence is less statistically robust than it may initially appear to be: “...even if we have 80
years of very reliable returns data on equity markets in the United States, we do not really
have 80 observations on demography and stock returns. We have one big baby boom that
has made its way through the financial markets. So the effective amount of information in
this case is much less than standard statistical procedures would suggest.” James Poterba,
“Impact of Population Aging on Financial Markets in Developed Countries,” Federal
Reserve Bank of Kansas City, Economic Review, 2004:4, p. 50.
Figure 3. Demographics and Equity Prices
1400Real S&P Index
8002145 to 64 age group
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 20000 18
Source: Department of Commerce; Standard and Poor’s.
Recent developments give reason to doubt a unique connection between the age
distribution and the saving rate. If those aged between 45 and 64 are, generally
speaking, in their peak saving years, then a rise in the proportion of the population
accounted for by that group might be expected to lead to an increase in the overall
saving rate. Between 1990 and 2003, the share of those aged 45 to 64 increased by
about five percentage points. Over that same period, the personal saving rate fell
from 7.0% to 1.4%. It is clear that knowledge of prospective changes in the age
distribution alone is insufficient to make predictions about saving behavior or asset
There are, however, other factors that may affect saving in addition to the desire
to smooth consumption over the course of one’s lifetime. Some may save for
precautionary reasons, for example, in order to insure against events less predictable
than the drop in income experienced at retirement. To the extent that people
accumulate precautionary savings, if they have a run of bad luck, or are faced with
unexpected out-of-pocket medical expenses, those expenses need not force a cut in
consumption. Those with children may have the additional motive of leaving a
bequest. If these two motives have a substantial influence on saving behavior, wealth
holdings might not be expected to decline after retirement.
Saving data by age are available from the Survey of Consumer Finances (SCF),
conducted by the Board of Governors of the Federal Reserve. Among the questions
asked in this survey is whether or not the responding family saved during the
previous year. Table 5 shows the results for the 2001 survey. These data also
confirm the expected hump in the cross-section, with fewer of those families at the
ends of the age distribution saving than those in the middle of working age.
Table 5: Percentage of Families Who Saved,
by Age of Family Head, 2001
less than 3535 to 4445 to 5455 to 6465 to 7475 and over
52.9 62.3 61.7 62.0 61.8 55.5
Source: Board of Governors of the Federal Reserve System.
While the data in Table 5 seem to confirm the main idea of the life-cycle model
which is that saving declines in retirement, more than half of families in the 75 and
over age group still report that they are saving. If the reason for saving is to
accumulate wealth, either to finance retirement or as a rainy day fund, then changes
in wealth may be a more useful measure of saving. Table 6 shows estimates of net33
worth by age group based on the results of the 2001 SCF. The mean is the average
wealth holding of each age class, and the median represents the mid-point of the
wealth distribution for each age class, with half of each group having more than the
median and half having less.
Table 6: Household Wealth by Age Class, 2001
Net Financial AssetsNet Worth
Age of Head of
Household Mean M edian M ean Median
20 to 24$26,330$-340$44,075$3,300
25 to 2911,6495052,28211,895
30 to 3432,80694088,51420,500
35 to 3946,5046,300122,71237,000
40 to 4475,09913,540204,48868,711
45 to 4999,24014,000240,27374,301
50 to 54181,18130,130369,670103,700
55 to 59210,90833,450455,729134,130
60 to 64207,84824,000421,902109,700
65 to 69156,28828,525346,338119,790
70 to 74205,07732,800409,932133,840
75 and up174,30827,835310,900114,000
33 These data are from: James Poterba, The Impact of Population Aging on Financial
Markets, NBER Working Paper 10851, Oct. 2004, 48 pp.
As with any survey, there are errors associated with these estimates, the more
so because of size of the age cohorts in the sample. Nonetheless, the figures show
that both net financial assets and total net worth rise up through the group aged 55
to 59 and then drop off somewhat. What may be more significant, though, is that the
drop-off in wealth for those groups that have largely retired is not as large as might
have been expected. There is a sizable drop between the 60-to-64 and 65-to-69
cohorts, but the increase in net worth increases for the 70-to-74 age group suggests
that some of that apparent drop may be due to sampling errors.
These data lend support to the idea that retirees finance their retirement, in part,
by drawing down their wealth. But the more remarkable characteristic of these data
may be that the decline in wealth of the very elderly is not as great as might be
expected. That suggests that bequests and precautionary saving may be as important
motives to save as retirement. That would make it less likely that there would be a
massive sell-off of assets as the baby-boom generation retires.34 On the contrary,
Rogers et al. show that the decline in the share of people under age 34, who have the
highest dis-saving rates, more than offsets the effect on saving of the increase in the
share of people over the age of 64 between now and 2040. Therefore, demographic
factors alone — assuming that future age cohorts had the same saving rate as today’s
age cohorts — would suggest that the national saving rate will rise between now and
The data in Table 6 also hint at a wealth distribution that is significantly
unequal. The ratio of mean to median is an indication of the degree of inequality in
a given distribution. An increase in the share of total wealth held by the richest
households in the distribution will increase the ratio of the mean to the median. In
the case of net financial assets the ratio is 6 or 7 for most of the age classes, while for
total net worth it is 3 or 4. That indicates that the distribution of net financial assets
is more unequal than the distribution of total net worth. Part of the reason for that
difference is the equalizing effect of including home equity.
Wealth is more unequally distributed than income. Table 7 presents data from
the 2001 SCF showing the degree to which the distribution of wealth is concentrated.
The data show that those households in the top 1% of the wealth distribution account
for nearly one-third of all wealth. The top 10% accounts for over two-thirds of total
34 Abel argues that, on the contrary, even if the baby boomers do not rapidly draw down
their assets in retirement, asset prices could still fall. See Andrew Abel, “Will Bequests
Attenuate the Predicted Meltdown in Stock Prices When Baby Boomers Retire?,” National
Bureau of Economic Research, working paper 8131, Feb., 2001.
35 Diane Lim Rogers, Eric Toder, and Landon Jones, “Economic Consequences of an Aging
Population,” Urban Institute, The Retirement Project Occasional Paper 6, Sep. 2000. Their
unorthodox prediction that demographic change will cause private saving to rise between
now and 2040 leads to surprising conclusions that, holding government saving equal,
interest rates will fall and the growth rate of investment spending per capita will rise from
current levels. However, the growth rate of income per capita still falls in their projection
between now and 2040 because of the decline in the worker-dependent ratio.
wealth. According to Poterba, the top 1% accounts for nearly half of all equity
Table 7: Percent of Net Worth by Percentile Group, 2001
0 to 49.950 to 89.990 to 94.995 to 98.999 to 100
2.8 27.4 12.1 25.0 32.7
Source: Board of Governors of the Federal Reserve System.
It would seem plausible that households at the top of the distribution might not
have to sell a significant fraction of their assets in order to maintain their standards
of living upon retirement, and are motivated to save largely by factors other than the
life-cycle factor. The more that is the case, the less reason there would be to expect
a large increase in sales of assets as the baby boom generation begins to retire.
Likewise, there is a significant minority of households without substantial financial
asset holdings who would have little effect on financial markets when they retire.
Thus far, only cross-section data have been presented. Those data indicate how
saving and wealth vary by age group at a single point in time. But they may not be
sufficient to project how individuals will behave as they move up the age
distribution. There may be important reasons other than age that determine how
individuals approach the decision of how much to save and how much wealth to
accumulate and hold over the course of their lives. For example, it may be that those
who grew up during the Great Depression beginning in 1929 developed a different
attitude about the importance of thrift from those who make up the baby-boom
generation. If that is true it would be unreasonable to use the behavior of the current
elderly population to predict the future behavior of baby boomers.
There is at least one reason to expect at least some asset sales as retirements
increase that does not require predictions of individual behavior. A substantial
number of workers are covered by defined benefit pension plans. A defined benefit
plan is one under which pension payments are determined by such variables as age
at retirement and length of employment. Firms that manage these pensions must
maintain a fund from which to pay future pension benefits. As the baby-boom
population entered the work force and the number of workers covered by these plans
grew the funds accumulated more and more assets. As those baby boomers who are
covered begin to retire the funds will need to increase sales of assets to meet their
pension obligations. Scheiber and Shoven project that the pension system will begin
to dissave in 2024 and dissave at a rate equal to 4% of payroll in 2065.37
36 James Poterba, “Impact of Population Aging on Financial Markets in Developed
Countries,” Federal Reserve Bank of Kansas City, Economic Review, 2004:4, p. 49.
37 Sylvester Schieber and John Shoven, “The Consequences of Population Aging for Private
Pension Fund Saving and Asset Markets,” in Michael Hurd and Naohiro Yashiro, eds., The
Economic Effects of Aging in the United States and Japan, University of Chicago Press,
(Chicago: 1997), p. 111.
A lower household saving rate would tend to raise interest rates and reduce
capital spending, all else equal. This would raise the rate of return on individual
accounts (after any decline in equity prices caused by the initial baby-boom “sell-
off.”) But in this case, all else would not be equal. At the same time that
demographic change was reducing the supply of saving, it might also affect the
demand for investment spending by firms. As discussed above, the aging of the U.S.
population means that the labor force will grow less rapidly, slowing the growth in
output as well. This means that the U.S. economy would not require the capital stock
to grow as rapidly as it had previously, but it is difficult to speculate what will
happen to investment growth relative to growth in the labor supply.
If firms want to keep the capital-labor ratio unchanged, then the growth in
investment spending would decline as much as the growth in the labor force, and
labor productivity growth would stay constant. If investment spending as a
percentage of output remains constant, that would also result in slower growth in the
capital stock, but the capital-labor ratio would rise. If the saving rate were constant,
as in the Solow model, the capital-labor ratio and productivity would rise.38 If there
is not a strong relationship between saving rates and age, as the previous evidence
suggested, then labor force growth may decline more quickly than the growth in
saving, and the capital-labor ratio and productivity growth would rise and interest
rates would fall.
Different general equilibrium models, which explicitly model endogenous
saving and investment rates that are functions of the labor force, make different
predictions about what happens to saving and investment rates when labor force
growth declines. In a literature survey, Bosworth et al. conclude, “At this stage in the
research, conclusions about the relative magnitude of the changes in rates of saving
and investment remain very sensitive to modest changes in the research design and
the data employed.”39
Higgins, for example, suggests that the aging of the baby boom could cause the
rate of investment to fall by more than the saving rate.40 Higgins argues that
investment demand has a cycle much like the life cycle plays a role in determining
saving rates. But while the peak ages for saving may be in the late 40s and early 50s,
the peak effect on investment demand is in the 20s when workers first enter the labor
force. If that is the case, the decline in saving as the baby boomers age may lag the
decline in investment demand. He estimates that national saving could fall by 1%-
2% of GDP, while investment could fall 5%-6% of GDP. In that case, the changing
age distribution would lead to a reduction (increase) in any current account deficit
(surplus). Cutler et al. also argue that investment demand will fall sharply, and argue
38 George Akerlof, “Comments and Discussion,” Brookings Papers on Economic Activity
39 Barry P. Bosworth, Ralph C. Bryant, and Gary Burtless, The Impact of Aging on Financial
Markets and the Economy: A Survey, Center for Retirement Research at Boston College,
40 Matthew Higgins, Demography, National Savings and International Capital Flows,
Federal Reserve Bank of New York Staff Report no. 34, Dec. 1997, 39 pp.
that the optimal response to demographic change is to lower, rather than raise, the
national saving rate.41
Interest Rates. Interest rates (the return on financial assets) will depend on42
the equilibrium between saving and investment. If saving falls more than
investment demand, then interest rates would rise, and vice versa. Thus, even if one
were relatively certain that saving would fall as the baby boomers enter retirement,
one cannot say which direction interest rates would move based on that information
alone. If investment stayed constant as a share of GDP, then saving would fall more
than investment and interest rates would rise. In the simple Solow model, where
saving is constant, interest rates would fall.
The effect on interest rates will also depend on international capital flows. In
an open economy, with international flows of both goods and capital, it is not
necessary that any decline in domestic saving be exactly matched by a decline in
domestic investment. By importing savings from abroad, domestic investment can
exceed domestic saving. When the nation is a net importer of saving from abroad
there is a corresponding gap between imports and exports of goods and services. The
extent to which domestic investment exceeds domestic saving determines by how
much imports of goods and services exceed exports. In other words, when
investment demand is greater than domestic saving there is a current account, or
trade, deficit. Since both saving and investment demand may fall when the baby
boomers retire, it is not clear whether demographic change would lead to a larger or
smaller trade deficit.
International capital flows do not depend on the United States alone. At the
same time that the United States will be experiencing the retirement of the baby
boomers, other industrialized nations will be experiencing a more drastic
demographic shift. The other G-8 countries are projected to experience a larger
decline in the worker-dependency ratio than the United States. Moreover, Germany,
Japan, and Italy are projected to experience an absolute decline in their populations
over the next 50 years.43 Thus, assuming foreigners respond to the life-cycle model
in the same way as Americans, the changes in saving and investment caused by
demographic change would be global in nature, and not limited to the United States.
This means that the effects of demographic change on our interest rates and
investment spending are unlikely to be canceled out by an unlimited source of
international capital flows from the rest of the industrialized world. It is harder to
predict how much capital will flow to or from the developing world in the distant
41 David Cutler et al., “An Aging Society: An Opportunity or Challenge?,” Brookings
Papers on Economic Activity 1, 1990, p. 1.
42 Unless otherwise noted, discussions of the movement in “interest rates” or “rates of
return” assumes that the absolute rate of return on all assets (including equities) will change
equally. In other words, relative rates of return will remain constant.
43 Richard Johnson, “Economic Implications of World Demographic Change,” Federal
Reserve Bank of Kansas City, Economic Review, 2004:1, p. 39.
Interest rates, and thus the return on individual accounts, will be affected by the
savings-investment balance. But it is uncertain what effect the aging of the baby-
boom generation will have on saving or investment. If there is a decline in domestic
saving, it might prompt an increase in the inflow of foreign capital leaving domestic
interest rates unchanged.
This section has considered the effects of demographic change on private saving
and investment. But national saving consists of both public and private saving, and
demographic change is projected to affect public saving as well. The next section
Effect of Future Government Budget Deficits on Rates of
Under current policy, spending on Social Security, Medicare, and Medicaid is
projected to rise rapidly, and there is not projected to be any corresponding rise in
revenues. As a result, budget deficits are forecast to rise very rapidly to44
unprecedently high levels. Unless health care spending grows no faster than GDP
and other spending falls as a share of GDP (the low cost scenario), CBO projects that45
deficits will reach 6.1%-12.3% of GDP by 2030 and 14.3%-34.5% of GDP by 2050.
Standard economic theory suggests that deficits this large would push up interest
rates significantly (until they ultimately became unsustainable and led to debt default
or accelerating inflation). That is because when the government borrows, it must
compete with private companies trying to finance capital investment out of the pool
of national saving. Government borrowing increases the demand for that pool of
saving, and interest rates rise as a result. If the nation borrows from abroad in
response to the budget deficits, that would take some upward pressure off interest
rates by reducing the demands put on national saving. However, it is unlikely that
deficits of the magnitude being discussed could be financed solely out of greater
foreign borrowing. In that case, at least some of the effect of the deficit would be
manifested in higher interest rates.
When interest rates rise, firms cut back on capital investment because it
becomes more expensive to finance. They cut back to the point where the marginal46
product of capital matches the higher interest rate. This means that the rate of
return on future capital investments would be higher than it would have been if
interest rates had not risen. The rate of return on equities must also rise if they are
to continue to be attractive to investors, now that alternative investments, such as
bonds and physical capital, are more attractive (because of higher interest rates). Or,
44 For detailed analysis, see CRS Report RL32747, Social Security and Medicare: The
Economic Implications of Current Policy, by Marc Labonte.
45 Data from Congressional Budget Office, The Long-Term Budget Outlook, December
46 Economists assume that each additional unit of capital is less productive than the last
because of the law of diminishing returns. Thus, when interest rates rise, firms will reduce
their capital investment until the point when investment becomes profitable again.
more directly, equity returns rise because shareholders are owners of the firm, and the
firm is now enjoying a higher rate of return on its capital investments. Thus, the
long-term rate of return on equities, and hence individual accounts, would likely rise
as budget deficits rise. But in the short run, the result would be the opposite —
higher budget deficits would cause equity prices to fall. In theoretical terms, this
would be because future earnings are now discounted at a higher rate, and therefore
discounted more heavily. The short-term decline in equity prices allows the firm’s
equity to earn a higher rate of return when it makes additions to the capital stock in
the future. This means that people holding individual accounts when interest rates
rise and equity prices fall would be worse off, but people who begin contributing to
individual accounts after the rise in interest rates would be better off than if interest
rates had not risen.
The Social Security Administration’s projections do not allow feedback between
budgetary effects and the macroeconomy, and thus do not adjust for these
possibilities. Moreover, it is highly likely that policy would be changed before the
“worst-case scenario” occurred.47 In that case, the effects on rates of return that
current policy implies would not come to pass. Nevertheless, it may be useful to
keep the implications of current policy in mind when forecasting rates of return for
two reasons. First, analyses of Social Security are based on current policy, and that
is the baseline by which proposed modifications to the program are compared. If
current policy is the baseline for policy analyses, then an argument can be made that
the implications of current policy should be incorporated in macroeconomic baseline
projections. Second, reducing future budgetary shortfalls would require someone to
be made worse off. Compared to current policy, either program beneficiaries would
be made worse off by large benefit cuts or taxpayers would be made worse off by
large tax increases. This suggests that reducing the future fiscal imbalance would be
politically difficult. Therefore, it is not unreasonable to assume for political economy
reasons that there is a very real possibility that, even after policy changes, budget
deficits would be larger than they have been historically. If so, this would affect the
rate of return on all assets.
Effect of Individual Accounts on National Saving and Rates
The previous section considered the effect of current policy on national saving,
and hence rates of return. But the individual accounts themselves may also affect
rates of return through the same channels. Proponents of individual accounts claim
that the existence of IAs that individuals can invest in U.S. private securities will
boost economic growth and the stock market. If true, this would increase the
resources available to the nation as a whole, making it easier to finance entitlement48
benefits for the baby boomers. Opponents point to the effects of the transition costs
of setting up the IAs on the budget deficit and claim that IAs will be detrimental to
47 In the worst case scenario, where current policy was pursued until the point where the
government defaulted on its debt, economic conditions would likely deteriorate to the point
where equities would be expected to fall in value precipitously.
48 See CRS Report RL30708, Social Security, Saving, and the Economy, by Brian W.
economic growth and investment rates. Which view is accurate? The method for
financing the transition costs turns out to be crucial in determining the effects on the
economy and the rate of return.
The creation of so-called carve out individual accounts increases private saving
by diverting part of the payroll tax into accounts that individuals are required to save.
But that is not the end of the story. The payroll taxes that would be diverted to the
IAs are currently financing Social Security and other government spending, and that
spending would still need to be financed after the accounts are created. This is the
essence of the transition costs. Thus, individual accounts can be financed in three
ways: by reducing government spending, raising taxes, or borrowing (increasing the
If the transition costs are financed through higher taxes or lower spending, the
creation of the IAs would not affect the budget deficit (government saving). Since
private saving would rise and government saving would remain the same, national
saving would rise and interest rates would fall. As a result, either private investment
would rise or borrowing abroad (the trade deficit) would fall. If private investment
rises, economic growth and future living standards would rise.49 The future profits
of corporations would rise since national saving would rise, interest rates would fall,
and investment would rise. In that case, stock prices would immediately rise to
reflect that change, causing the rate of return on equities to initially rise (assuming
firms did not respond by issuing more equity). However, the rate of return would
eventually decline below the original level to match the decline in the user cost of
capital (interest rates) that results from a higher national saving rate. From the firm’s
perspective, when the saving rate increases, the cost of capital declines and the firm
responds by increasing its capital investment. As a result, its output rises. The rate
of return from that investment falls because there are diminishing marginal returns
to investment for the firm.
If the rise in national saving is instead matched by a decline in borrowing
abroad, then there would be no change in interest rates. With no change in interest
rates, the rate of return on equities and IA holdings would not be expected to change.
Whether higher national saving results in lower interest rates or lower borrowing
abroad depends on how sensitive capital flows are to changes in interest rates, as
opposed to other factors that determine capital flows.
If the transition costs are instead financed through borrowing, the budget deficit
would rise. For example, Figure 4 illustrates the increase in the deficit (a decline in
the surplus) caused by individual accounts in the President’s Commission Plan 2 with
67% participation. (To isolate the effects of the IAs, it does not include the effects
of unrelated benefit cuts on the deficit.) The rise in private saving would be
completely offset by the decline in government saving, and national saving would
remain constant.50 There is no reason to think that the rise in private saving would
49 These positive effects on the economy are not unique to the creation of IAs; any policy
option that reduced the budget deficit would be expected to have the same results.
50 In the long run, national saving would increase because, all else equal, the benefit offsets
exceed the fall in government saving since, under current proposals, the IAs cannot
be “topped up” by extra voluntary contributions from the account owners.51 In that
case, private investment, foreign borrowing, interest rates, and economic growth
would be unaffected by the creation of individual accounts. (Usually, budget deficits
reduce national saving because they finance consumption; the result is different here
because they are financing private saving.) Because interest rates are not changing,
the rate of return on equities and earned by IAs would not be expected to change.
Figure 4: Change in Unified Budget Balance From Introducing
Individual Accounts in Plan 2 (67% Participation Rate)
2004 2014 2024 2034 2044 2054 2064 2074
Source: Chief Actuaries of Social Security Administration, Memorandum to President’s Commission to
Strengthen Social Security, December 2001, pp. 53-54.
Notes: Estimates based on Commission’s Option 2 reform plan with 66.7% participation rate. Estimate includes
only the budgetary effects of the individual accounts and benefit offsets; estimate does not include the effects of
any other changes to Social Security. It assumes accounts equal to 4% of payroll are introduced in 2004. A
negative value connotes a larger deficit.
would increase public saving, but these offsets would be negligible as a share of public
saving for the first few decades after the individual accounts were created.
51 This assumes private individuals react to the creation of individual accounts by not
altering their private saving behavior (since the individual accounts are offset by lower
Social Security benefits, and therefore do not increase net wealth). If individuals instead
react by saving less outside of Social Security, either because they incorrectly believe their
net wealth has increased or because they believe they will receive higher lifetime benefits
because the reform package reduces Social Security’s insolvency, then national saving could
If the IAs were designed as “add ons” instead of “carve outs,” the
macroeconomic effects would remain the same. Add-on accounts would be financed
from general revenues rather than by diverting payroll taxes. If the general revenues
used to finance the accounts were raised through higher taxes or lower government
spending, then national saving would rise because private saving would rise and
government saving would remain the same. This would cause interest rates, and the
return on equities, to fall. If the general revenues were instead raised by increasing
the budget deficit, the increase in private saving would be offset by the decrease in
government saving. This would leave interest rates and rates of return unchanged.
Effect of Individual Accounts on Relative Rates of Return
Although borrowing to finance individual accounts would have no effect on
aggregate saving and investment, and therefore overall interest rates, it is likely to
affect relative rates of return within financial markets. Higher government borrowing
would increase the supply of U.S. Treasury securities on the market. Unless the
demand for Treasuries is infinitely elastic, a higher interest rate would be required for
investors to be willing to hold more Treasuries. As investors shift into Treasuries
and out of other assets, the rate of return on other assets would fall, leaving aggregate
rates of return the same.
Just as the greater supply of Treasuries would affect their relative rates of return,
the creation of individual accounts would be expected to change relative rates of
return on any private assets that could be purchased by account-holders, such as
equities and corporate bonds. For example, if the demand for equities rose, stock
market prices would rise. The effect of higher equity prices on future rates of return
would depend on whether or not the accounts were deficit financed. If they were,
then saving would simply shift from the public to the private sector and investment
and economic growth would remain unchanged. In that case, the future profitability
of corporations and their ability to pay dividends would likely remain unchanged.
Given an increase in equity prices, and an unchanged future stream of dividends, the
rate of return on equities would be lower.
This is what one would expect in any market if demand rises and conditions on
the supply side of the market remained unchanged. Again, these results assume that
the demand for private securities is not perfectly elastic (meaning that the overall
change in demand for equities does not fall in proportion to the increase in their
price). Otherwise, investors outside of IAs could offset any change in relative rates
of return by selling private securities and buying Treasuries, in which case current
asset prices would not be affected by the creation of IAs.
The Social Security Administration does not adjust its projections to take these
effects into account, although they acknowledge that the effects may not be
It should be noted that the precise effects on implementing a plan that would
result in a large demand for equities and corporate bonds on the yields of these
securities is not clear. This demand would likely be at least partially offset by
reductions in demand for other investment mechanisms. For the purpose of these
estimates, it is assumed that there will be no net dynamic feedback effects on the52
economy or on the financial markets.
Estimates of how much individual accounts would earn typically assume that
financial securities will earn a rate of return equal to the historical average. Behind
this assumption lies the implicit belief that the future is likely to be similar to the
past. This report concludes that there are economic variables that are likely to be
different in the future from the past, and it reports that changes in those variables
would have a predictable effect on rates of return. It finds:
!Today’s high price/earnings and price/dividends ratio suggest that
future rates of return will likely be lower. Dividends and earnings
would have to grow unusually rapidly in the future to avoid this
!Historically, the equity premium that investors demand to hold
equities rather than bonds has been inexplicably high. This premium
may be smaller in the future, particularly since equity ownership is
much more widespread and transaction fees lower than in the past.
This would reduce equity returns.
!A lower future economic growth rate (as a result of slower labor
supply growth) is likely to reduce corporate earnings, unless
capital’s share of income rises continually. This could reduce the
rate of return on equities.
!A lower saving rate, as the baby boomers finance their retirement,
would raise the rate of return. Asset prices would likely decline
initially in order to achieve higher rates of return in the future. This
effect is likely to be smaller than popularly imagined because of
precautionary and bequest motives for saving, the unequal
distribution of wealth, and fewer young dissavers in the future.
!At the same time saving is falling, a slower growing labor force
would reduce investment demand, lowering the rate of return. It is
uncertain which of these two effects would dominate, or how much
both could be offset by international capital flows.
!Public policy directly alters the national saving rate through the
budget deficit. Larger budget deficits lead to higher rates of return
across all assets. Under current policy, budget deficits would
become unsustainably large; for that reason, current policy may not
be a good predictor of the future. However, the budgetary pressures
52 Stephen Goss, Chief Actuary, “Estimated Financial Effects of The Progressive Personal
Account Plan,” memorandum, Social Security Administration, Dec. 1, 2003.
facing policymakers in the future are large enough that an increase
in budget deficits seems a predictable outcome.
!Deficit-financed individual accounts would leave the national saving
rate unchanged while individual accounts financed through higher
taxes or lower spending would raise the national saving rate. The
current political debate is leaning toward individual accounts that
would be deficit financed, at least initially. In any case, the increase
in national saving that could result from IAs would, at best, be small
compared to future budget deficits.
!While deficit-financed IAs would not change overall rates of return,
it could change relative rates of return. The issuance of large
additions to the public debt in order to purchase equities could alter
the relative demand for those two assets, raising the rate of return on
U.S. Treasuries and lowering the rate of return on equities.
None of these results is certain. On balance, they probably point to a lower rate
of return in the future, but some factors point in the opposite direction, such as the
effect of potentially higher budget deficits on the national saving rate. At the least,
the results suggest that estimates of the earnings of IAs should be presented as a
range of estimates rather than point estimates, and the range of plausible scenarios
should be wide. Even if estimates were only based on historical data, the statistical
certainty surrounding that data is much less robust than is typically realized, and so
the range of plausible outcomes is relatively wide.
Estimates of individual account earnings can be misused to suggest that there
is a “free lunch” if they stress the high returns of equities without acknowledging that
ownership comes at the cost of higher risk. (Although investors’ appetite for the risk
associated with equities seems irrationally low to many economists, their beliefs are
notable.) In some contexts, risk adjustment that eliminates differences in rates of
return between different types of securities is the only way to make meaningful
comparisons. For example, in Social Security reform proposals in which individual
accounts are offset by Social Security benefit reductions, the earnings of IAs can be
meaningfully compared to the benefit offsets only if the rate of return on the IA is set
to the risk-free rate of return since the benefit offset is, in essence, “risk free.”