|
|
|
IV. LABOR MARKETS, HUMAN
RESOURCES, AND INTERNATIONAL
REFEREED PAPER
The Link between the Stock Market and
Retirement Income
Christian E. Weller
Economic Policy Institute
Abstract
This paper analyzes the effect of wealth
fluctuations on retirement income adequacy between 1992 and 2000. In addition,
the paper estimates how financial wealth relative to income may develop
in the medium to long-term. If a fixed real level of consumption is considered
for retirement income adequacy, the average household was more likely
to be inadequately prepared for retirement, even after wealth increased
dramatically in the late 1990s. Moreover, on average, households can expect
to reach their peak wealth to income levels only after a period of 30-50
years. To address the likely shortfalls in retirement income adequacy
for many households, public policy choices that could help to raise private
savings should be considered.
Introduction
Beginning in March 2000, the stock market
fell precipitously. With the slide in stock prices, household financial
wealth declined by 16 percent, and the combined housing and financial
wealth dropped by 13 percent--the largest nominal decline since 1952--by
March 2001.
While stock prices and housing wealth
declined, public policy debates focused on increasing the reliance of
households on private markets in saving for retirement. As more private
accounts, either as add-ons or carve-outs to Social Security, may become
more popular, the question arises whether and how the recent stock market
gyrations affect retirement income security. Consequently, this paper
analyzes the link between stock market fluctuations and retirement income,
and its adequacy, since 1992.
The Stock Market and Household Wealth in the 1990s
Within one year, from March 2000 to March
2001, direct equity holdings declined by $3.5 trillion, financial assets
fell by $4.3 trillion, and financial wealth dropped by $4.9 trillion (Board
of Governors 2001). Moreover, by March 2001, financial wealth to income
fell to 281 percent, the same level as in March 1997. What did this mean
for retirement income security?
To see how the adequacy of retirement
wealth changed during the stock market rise and fall, a target wealth-to-income
ratio is estimated and then compared to the actual average wealth-to-income
ratio. It is important to keep in mind, though, that this paper's focus
is on the average, rather than the median, household; however, because
retirement wealth is increasingly unequally distributed (Wolff 2002),
the median household's retirement income adequacy has also fallen further
behind that of the average household.
Several studies have focused on adequate
retirement wealth. Two general conclusions can be draw from the prior
research (Weller 2001). First, households had, on average, inadequate
savings in 1992. Second, the distribution of household wealth differs
widely, so a large minority of households fell far short of adequate retirement
savings.
To calculate a target ratio of adequate
retirement savings, this paper uses the estimates from Gustman and Steinmeier
(1999). Thus, both a nominal and a real replacement ratio are used. In
particular, Gustman and Steinmeier (1999) calculate a nominal replacement
ratio of 86 percent and a real replacement ratio of 60 percent for 1992.
These ratios are increased by one-fifth, to account for the fact that
retirees have lower consumption needs than workers (Gustman and Steinmeier
1999), resulting in an effective nominal replacement ratio of 103 percent
and an effective real replacement ratio of 72 percent. Further, the original
replacement ratios included housing and Social Security wealth. Housing
wealth amounted to 16 percent of total wealth, and Social Security wealth
to another 23.7 percent. Assuming that the shortfall of 28 percent of
the real replacement ratio would have to be covered by financial savings
only, financial wealth was 47 percent of what it should have been. If
it is assumed that the shortfall should have been covered by financial
and housing wealth, wealth was 67 percent of what it should have been
in 1992. Hence, the effective wealth-to-income ratio for 1992 was either
100 percent, 67 percent (for the real ratio, including housing wealth),
or 53 percent (for the real ratio, excluding housing wealth) of its target.
To calculate the target values after 1992,
the ratio for 1992 is adjusted for demographic changes, such as average
age, life expectancy at 65, and the share of the population over 65.
The target wealth-to-income ratio, W/T,
is equal to the adjusted wealth-to-income ratio of the previous period.
It is adjusted for the percent increase in the age of the average worker,
age, which itself is adjusted by an interest factor, ±. A
higher age means that fewer years are left to retirement, therefore requiring
more wealth relative to income. Not only does the worker have to compensate
for fewer working years, but also for the loss of compounded interest
over those years. Equation (1)' shows that age is adjusted for
the loss of long-term interest, rLT, compounded for
the number of years that the average age of workers, AGE, has changed.
Also, the target wealth-to-income ratio increase with life expectancy
at age 65, 65plust, because a longer life expectancy
requires more wealth to maintain the same retirement income. But a household
can increase its wealth partially due to compounded interest. Thus, the
increase in the wealth-to-income ratio is reduced by a discount factor.
Equation (1)'' shows that the adjustment is the long-term interest rate
compounded over the additional years, LE65. Another adjustment
is the change in the share of population over 65, 65plust,
because, by definition, retirees are dissavers and thus a larger share
of the elderly would imply fewer aggregate savings relative to income.
For the calculations, actual changes of
the average age of workers, of life expectancy at age 65 (Social Security
Administration 2000, 2001), and of the share of the population over 65
(International Database 2001) are chosen. Table 1 shows that the financial
wealth-to-income ratio was 24 percent above its target in 2000 and that
the financial and housing wealth-to-income ratio was 16 percent above
its target in 2000, starting from 100 percent adequacy, or at the same
levels as in 1997. Starting from adequacy ratios of less than 100 percent,
the average household never reached its target and the average household
was 22-34 percentage points below its target in 2000 (Table 1).
To evaluate what the future may hold,
this paper uses a regression-based simulation. The regression model considers
empirically relevant determinants of household wealth-to-income. Wealth
relative to income rose annually by 3.3 percent from 1992 to 2000, although
the stock market grew by 13.9 percent and income by 5.2 percent. Most
household assets were not allocated in corporate equities, because households'
direct and indirect equity holdings never amounted to more than 50 percent
of financial assets (Board of Governors 2001). Also, as wealth grew, so
did income. Faster income growth required faster wealth growth to maintain
the same level of retirement savings adequacy, and more wealth also provided
households with more collateral to borrow. Moreover, more wealth provided
households with more resources to increase their consumption, thereby
reducing savings. On average, households contributed about 8 percent of
personal disposable income (PDI) to their financial assets in the 1990s,
below the averages of all previous postwar business cycles. Also, the
wealth effect seems asymmetric. When holding gains were positive, households
added on average 8.5 percent of their PDI to their financial assets, and
when holding gains were negative they added only 6.5 percent (Board of
Governors 2001, Table F.100).
To study the relative importance of each
factor determining the wealth-to-income ratio, the following equation
is estimated
where the wealth-to-income ratio, W/Y, depends on the real value
of the S&P 500, on real income (both deflated by the consumer price
index, CPI), on the share of equities out of financial assets, E/A,
on the ratio of debt relative to income, L/Y, and on the savings
rate out of personal disposable income, S/PDI. To control for demographic
changes, the life expectancy at age 65, the average age of workers, and
the share of the population over 65 are included. Also, µ is a normally
distributed random error term. A logarithmic specification is used for
each variable.
The expected signs of the explanatory
variables are straightforward. The S&P 500, the savings rate, the
equity share in households' portfolios, average age of workers, and longevity
at 65 should all be positively related to the wealth-to-income ratio.
By contrast, real income, liabilities, and the population over the age
of 65 should have negative signs.
All economic variables are compiled from
the Flow of Funds Statistics for the United States (Board of Governors
2001), except the (seasonally adjusted) CPI, which is from the Bureau
of Labor Statistics. The data for life expectancy are from the National
Center for Health Statistics' Life Expectancy by Race, Sex, 1970 to
1998 (at Birth, Age 65 and Age 85), and the 2001 Social Securities
Trustees' Report (Social Security Administration 2001). Average age
is calculated as a weighted average of workers covered by Social Security
(Social Security Administration 2000). Missing demographic data are interpolated.
For the regression, a few adjustments
are made. The ratio of liabilities to income is nonstationary and hence
differenced once. As the savings rate may be endogenously related to the
dependent variable, it is instrumented by regressing it on itself lagged
once and on all other explanatory variables. To correct for autocorrelation,
a Corchran-Orcutt regression is used.
The estimated coefficients have the expected
signs or are insignificant (Table 2). The determinants of financial wealth-to-income
may vary over time. Thus, the sample is separated in 1982, which marks
the beginning of 401(k) plans. The results show a significant effect of
the stock market in the later period, but not in the earlier period. The
regression results seem robust, however, and all explanatory variables
are significant determinants at one time or another. The estimated coefficients
are consequently used to simulate future wealth-to-income ratios.
The target levels for the wealth-to-income
ratios are simulated using equation (1) as the basis for Monte Carlo simulations.
The target levels are calculated for the period from 2000 to 2050 using
Monte Carlo simulations, using 1,000 random observations for each input
variable, based on historic distributions. Also, there are two separate
starting points for each of two wealth-to-income ratios for the relative
adequacy levels in 1992: 100 percent and 53 percent for the financial
wealth-to-income rate, and 67 percent and 100 percent for the financial
wealth plus net housing wealth-to-income rate.
To simulate the actual levels, the estimated
coefficients for the full sample from Table 2 are inserted in Equation
(2). Monte Carlo simulations are used to simulate actual wealth-to-income
ratios based on 1,000 random values, which in turn are based on historic
distributions, for each variable for each of the next 50 years. The results
are used to calculate the chance of reaching the last peak wealth-to-income
ratio or of falling below the target level in any given year.
The results in Table 3 illustrate the
risks associated with using private wealth as a vehicle of retirement
income provision. The chance of remaining below the peak levels of March
2000 stays above 50 percent for the next 30 years. Further, if the starting
adequacy levels are lower, the chance of staying below the target level
is above 90 percent. The fact that the average household has no discernible
possibility to reach its target level should not be surprising in light
of the recent experience. Despite an unprecedented increase in the stock
market, the average household did not reach adequate wealth-to-income
ratios in the late 1990s.
It may be that the parameter estimates
for the wealth-to-income ratios for the full sample may not adequately
reflect the determinants of household wealth. Instead, the coefficient
estimates for the period after 1982 may more accurately reflect the importance
of each variable. Using the parameter estimates for the period after 1982,
however, the results in Table 4 suggest a similar future. The chance of
not reaching peak levels remains above 60 percent for the financial wealth-to-income
ratio and close to 100 percent for the financial and housing wealth-to-income
ratio. Similarly, when starting points of 53 percent and 67 percent of
adequate wealth-to-income are assumed, the chances of staying below adequate
levels stay above 90 percent and 100 percent, respectively.
Conclusion
This paper studies the consequences of
the stock market ups and downs for retirement income adequacy. The results
suggest that the average household's wealth-to-income ratio was 22-34
percentage points below its target by the end of 2000. Further, the chances
for the average household to recover the lost wealth and to reach adequate
retirement savings are very low. It will take the average household more
than 30 years, in the best-case scenario, to have a greater than 50 percent
chance of reaching its previous peak wealth-to-income level again. Moreover,
for the average household, the chances of reaching its target wealth-to-income
ratio remain below 50 percent, and often close to 100 percent, for the
next 50 years. This should not come as a surprise. The past years saw
an unprecedented boom on the stock market combined with a proliferation
of indirect stock ownership of households. Yet, households on average
fell short of adequate retirement savings.
References
Board of Governors. 2001. Flow of Funds Accounts
of the United States. Washington, DC: Board of Governors of the Federal
Reserve System.
Gustman, A., and T. L. Steinmeier. 1999. "Effects of
Pensions on Savings: Analysis with Data from the Health and Retirement
Study." Carnegie-Rochester Conference Series on Public Policy 50, No.
99, pp. '1-324.
International Database. 2001. Mid-Year Population
by Age for the United States. Washington, DC: U.S. Bureau of the Census.
Social Security Administration. 2001. 2000 Trustees
Report of Old Age, Survivorship and Disability Insurance Trust Funds.
Washington, DC: Social Security Administration.
Social Security Administration. 2000. 2000 Annual
Statistical Supplement. Washington, DC: Social Security Administration.
Weller, C. 2001. "Can Workers Afford a Higher Retirement
Age?" EPI Technical Paper No. 255. Washington, DC: Economic Policy Institute.
Wolff, E. 2002. "Retirement Income Security: How Many
Have it and How Has it Changed?" Presented at the annual meetings of the
American Economic Association, January 2002, Atlanta.
|
