Text of Greenspan speech (out at 10:00 A.M.)
federalreserve.gov
Remarks by Chairman Alan Greenspan
Opening remarks
At a symposium sponsored by the Federal Reserve Bank of Kansas City, Jackson Hole, Wyoming
August 31, 2001
The rapid technological innovation that spurred the advancement of the "information economy" has
resulted in some dramatic capital gains and losses in equity markets in recent years. These remarkable
developments have attracted considerable attention from economists and from macroeconomic
policymakers. At the same time, movements in the prices of some other assets in the economy--changes
in house prices, for example--have been steadier, less dramatic, but perhaps no less significant.
There can be little doubt that sizable swings in the market values of business and household assets have
created important challenges for policymakers. After having been relatively stable for a number of
decades, the aggregate ratio of household net worth to income rose steeply over the second half of the
1990s and reached an unprecedented level by early last year. That ratio has subsequently retraced some
of its earlier gains.
But we must ask whether the aggregate ratio of net worth to income is a sufficient statistic for
summarizing the effect of capital gains on economic behavior or, alternatively, whether the distribution of
capital gains across assets and the manner in which those gains are realized also are significant
determinants of spending. To answer these questions, we need far more information than we currently
possess about the nature and the sources of capital gains and the interaction of these gains with credit
markets and consumer behavior.
Analysts have long factored changing asset values into models that seek to explain consumption and
investment. Indeed, in recent years, household wealth variables have become increasingly important
quantitatively in endeavors to track consumer spending. The importance of household balance sheet
variables for explaining consumption and the possibility that not all these variables influence spending
identically suggest the need for greater disaggregation than is typically employed in most models.
Observing that, over the past half century, consumer spending has amounted to about 90 percent of
income, it might appear that income is largely sufficient to explain consumption. However, econometric
evidence suggests that such numbers may be deceptive. Wealth by itself now appears to explain about
one-fifth of the total level of consumer outlays, according to the Board's large-scale econometric model,
leaving disposable income and other factors to explain only four-fifths of consumption. Indeed, if capital
gains have any effect on consumption, the propensity of households to spend out of income must be less,
possibly much less, than 90 percent.
If income and wealth moved tightly together over time, the distinction between them might not be
meaningful for predicting the future path of consumption. And, over very long periods of time, capital
gains on physical assets are not independent of the trends in disposable income. But the relationship of
wealth to income is demonstrably not stable over time spans relevant for the conduct of policy. As a
consequence, a statistical system that augments income as a determinant of consumer spending with
information about wealth can significantly assist our understanding of this key economic relationship.
Conventional regression analysis suggests that a permanent one-dollar increase in the level of household
wealth raises the annual level of personal consumption expenditures approximately 3 to 5 cents after due
consideration of lags. Arguably, it would not be important to draw distinctions among various types of
wealth if all assets were engendering similar rates of capital gains. Owing to collinearity in such instances,
all wealth proxies would produce similar estimates of overall wealth effects on consumer spending.
At times, however, the rates of change in key asset prices have diverged. For example, over the past year
and a half, home values have appreciated, whereas equity values have contracted significantly. In such
circumstances, differences in the propensities to consume out of the capital gains and losses on different
types of assets could have significant implications for aggregate demand.
Assuming that the underlying propensities are, in fact, stable and given enough time-series data with
sufficient variation, standard regression procedures should be able to extract reasonably robust estimates
of any differential in spending propensities--for example, out of stock market wealth and home wealth.
But, in practice, these circumstances do not prevail. As a consequence, we at the Federal Reserve Board
are in the process of developing balance-sheet disaggregations that should help us infer the propensities to
spend out of capital gains across different classes of assets.
In carrying out this analysis, we have been especially mindful of the possibility that the amount by which
a capital gain affects spending may well be a function of whether or not the gain has been realized. On the
buyer's side, when an asset is transferred, the acquisition cost is its new book value and, by definition, its
market value. On the seller's side, the proceeds from the sale are available for asset accumulation, debt
repayment, and consumption. In this way, a capital gain is realized and made liquid, with the potential to
affect spending, assets, or debt. The capital gain in the process disappears as an element in the
householder's balance sheet.
Unrealized gains, to be sure, can be borrowed against, and the proceeds of the loan can be spent or used
for repayment of other debt. Alternatively, the unrealized gain could induce households to finance
additional outlays by selling other assets or by reducing their saving out of current income. But unless, or
until, this gain is realized or is extinguished by a fall in market price, it will remain on the asset side of the
householder's balance sheet, exposed to price change and uncertainty.
Equity extraction through realized gains creates liquid funds with certain value. Indeed, a significant
proportion of sellers do not purchase another home. In contrast, extraction of unrealized gains does not
reduce the householders' uncertainty about their net worth or their exposure to market price changes.
This suggests that the propensity to spend out of realized gains is likely to be greater than the propensity
to spend out of unrealized gains.
Although our asset-class analysis of detailed disaggregated data is still at an early stage, preliminary
examination finds that the data are consistent with the hypothesis of differential spending propensities by
asset type and by whether or not capital gains have been realized. For example, purchasers of existing
homes, on average, appear to take out mortgages about twice the size of the unamortized mortgage that
the typical seller cancels on sale. After accounting for closing expenses, the remaining unencumbered
cash is available for debt repayment, acquisition of financial and nonfinancial assets, and spending.
We have no direct evidence, of which I am aware, on the way that such funds are used. However, we
can make use of several surveys that have explored how cash-outs associated with mortgage refinancing
and home equity loans are expended. Typically, these surveys indicate that households allocate so-called
cash-outs--that is, the amount by which a refinanced mortgage exceeds the pre-refinanced outstanding
debt--to repayment of nonmortgage debt, acquisition of financial assets, outlays for home improvement,
and personal consumption expenditures in roughly equal proportions.
Our interest, of course, is primarily on spending; extracting home equity to repay debt or to purchase
financial assets merely reshuffles balance sheets and, at least immediately, does little to affect economic
activity. If these survey results are taken at face value and are applied to the case in which the home
changes hands--as distinct from, say, a refinancing-- the amount of personal consumption expenditures
generated from realized capital gains on the sale of homes, financed through the mortgage market,
represents approximately 10 to 15 cents on the dollar. 1
Of course, in addition to realized capital gains from the turnover of existing homes, there is a considerable
amount of cash that is extracted from home equity without a home sale, principally from refinancing
cash-outs and from home equity loans. Both types of equity extraction have risen considerably in recent
years, in line with the marked rise in unrealized capital gains on homes. Some preliminary calculations
suggest that the total of equity extractions from unrealized capital gains on homes that is spent on
consumer goods and services per dollar of capital gains is a fraction of the spending engendered by the
gains realized through the sale of a home. 2 3 This difference occurs, to a large extent, because the net
extraction of equity is much higher among homes that have turned over than among those that have not.
While data on home mortgage debt and house turnover can be used to analyze the particular channels
through which capital gains on homes spur consumer outlays, the financing linkages between stock
market capital gains and consumer spending are less clear. Homeowners typically own one home, which
they hold, on average, for nearly a decade. Financing is almost exclusively through the mortgage market,
and equity extractions for spending, accordingly, are readily identified. Stocks, in contrast, tend to be held
in portfolios that have far greater rates of turnover than homes, and financing sources are much more
diverse and changeable. Moreover, although gains in defined contribution plans, IRAs, and other
tax-deferred accounts almost surely affect consumer spending, the complicated tax treatment and
restrictions on the use of those funds make the connections between capital gains in these accounts and
spending quite indirect.
Nonetheless, even setting aside all pension-type assets, household capital gains on directly held equities
and mutual funds in recent years have been two to four times the size of overall gains on homes. The
sheer size of such gains suggests that capital gains on equities have been a more potent factor in
determining spending than gains on homes. In fact, if we accept a total net wealth effect on consumption
of 3 to 5 cents on the dollar, and if further analysis supports the larger net spending propensities from
capital gains on homes suggested by mortgage and survey data, then the propensity to spend out of each
dollar of stock market gains would be less than the propensity to spend out of a dollar from gains on
homes, but still larger in overall dollar magnitude.
Of course, these quantitative magnitudes are tentative, and a great deal of additional work will be
necessary to better understand and to confirm the nature and magnitudes of the relationships between
capital gains on houses and stocks--realized and unrealized--and consumer spending.
* * *
No matter how one differentiates the effects on consumer spending of capital gains on stock market and
housing wealth, it is clear that the massive increase in capital values over the past five years had a
profound impact on output and income. The influence of capital gains on economic behavior also is likely
to be of substantial consequence for the prospective performance of the economy.
That influence also can be seen in our national income and product accounts (NIPA). By design, these
accounts measure the market value of the output of goods and services and its distribution to the factors
of production. As such, they exclude capital gains and losses. This exclusion is especially relevant for
personal saving, where our accounting conventions result in capital gains having a large effect on the
published figures. In part, the reason is that the NIPA deduct taxes paid on realized capital gains from
personal income and treat them, in effect, as a transfer to the government sector, even though the capital
gains that generated those taxes are excluded from income. 4 This issue is not trivial. As best we can
determine, of the 4.6 percentage point decline in the personal saving rate between 1995 and 2000, a full
percentage point is attributable to the increase in federal and state capital gains taxes paid over that period.
Capital gains have also significantly influenced the measured personal saving rate as a result of the NIPA
treatment of the pension fund sector. In particular, because defined-benefit pensions are considered part
of the "personal sector," employer contributions to such plans are included in disposable income, as are
the interest, dividend, and rental incomes received by these plans. In contrast, benefit payments to
individuals are not part of personal income because they are considered intrasectoral transfers.
Neither households nor corporations, however, are likely to view their own financial activities in that
manner. Surely, for defined-benefit pensions, it is the benefit payments to retirees rather than the
employer inflows into the pension sector that individuals perceive as personal income. For their part,
businesses have often viewed defined-benefit pension plans, in effect, as business-sector profit centers
because capital gains affect corporate defined-benefit pension contributions and, hence, earnings.
This consideration is relevant in the measurement and interpretation of the personal saving rate. In recent
years, contributions to private defined-benefit plans have declined significantly as an increasing part of
these plans' accrued benefit liabilities have been met through a rise in the market value of their equity
holdings. Offsetting this decline, to some extent, has been an increase in dividend and other capital
income.
If private and state and local defined-benefit pension plans had been separated from the personal sector,
the personal saving rate would have fallen about 3/4 percentage point less from 1995 to 2000, all else
being equal.
All told, if households viewed taxes on capital gains as a subtraction from those gains and not from
income and, further, if households viewed benefit payments received from defined-benefit plans as
income rather than their employers' contributions (as well as the investment income of the plans),
perceived disposable income in 2000 would have been higher as would the personal saving rate.
In short, roughly two-fifths of the measured decline in the personal saving rate since 1995 reflects the
foregoing NIPA income-accounting conventions.
I should emphasize that any accounting adjustments made to personal saving because of changes in the
definition of disposable income are exactly offset in business and government saving so that national
saving is unaffected. The increment to personal saving associated with a treatment of the private
defined-benefit pension sector as a business profit center would be offset by a decline in corporate profits
and business saving. In addition, a designation of taxes on capital gains as capital transfers (in a manner
similar to estate and gift taxes) would raise measured personal saving and lower overall tax receipts and,
hence, government saving. 5 Thus, while total national saving would be unaffected by these specific
accounting adjustments for capital gains, the distribution of NIPA saving among households, businesses,
and governments would be significantly influenced.
One must recognize that no single way to array information on income, production, and capital gains is
best. The particular array employed depends on the specific purposes to which the data set is to be
applied. The treatment of capital gains in the NIPA, for example, is intended to allow the accounts to
most accurately attribute national saving to the various sectors in the accounts. Indeed, when that is the
objective, the removal of capital gains is essential. For analysis of issues related to consumer spending,
though, the NIPA personal saving rate presents an incomplete picture of the financial state of the
household sector in the aggregate, and an adjustment along the lines previously suggested may be
informative.
In addition to the effect of income-accounting conventions, of course, we must consider the real
economy influence of capital gains on the level of consumption. The estimates of the effect of household
capital gains on consumer spending of 3 to 5 cents on the dollar suggest that, directly and indirectly,
capital gains easily account for the remainder of the measured five-year decline in the saving rate.
Obviously, this is not to say that had asset prices been flat for an extended period the personal saving rate
would have been unchanged, on net, over the past five years. If asset prices had not risen, real incomes
would surely have been altered, and the vast array of secondary and tertiary effects of asset-price
changes would have been different. Nonetheless, this exercise fosters additional important insights into
the dynamics of household behavior and the relationships among asset prices, income, and consumption.
The complexity of these relationships underscores the potential usefulness of developing separate sets of
accounts to track capital gains. These accounts could supplement the income and product accounts, the
flow of funds accounts, and the balance of payments accounts. The last two currently exhibit, in part,
the effect of capital gains and can be separated into special accounts. A supplementary set of detailed
tables on capital gains exclusions from the national income and product accounts also would be a useful
addition to our overall system of economic accounts.
* * *
This morning I have not endeavored to discuss the effects of capital gains, other than peripherally, on
investment in plant and equipment, home improvement, tax revenues, and government surpluses, and
their obvious significance in tracking international economic flows. Clearly, these also are relevant to any
evaluation of macroeconomic events and warrant further study.
* * *
In closing, accounting systems are not ends in themselves. We construct them because they have a
function in aiding our understanding of some particular aspect of a business operation at a company level
or for an economy as a whole. As we endeavor to better understand how changes in the level and
composition of wealth affect economic behavior, new accounting systems may be required to supplement
those that have long served us so well. Technology has facilitated the production of information at a far
faster rate than at any time in the past. But in the information economy, it remains up to us to organize
and use that information in ways that improve the quality of decision making.
Footnotes
1. The realized capital gain on a home sale in recent years has engendered a net increase in the mortgage
debt (that is, net equity extraction) on that home averaging nine-tenths of the capital gain. Of the net
equity extraction, almost half has been expended on closing and related expenses. The remainder, we
assume, is distributed as indicated by the consumer surveys. Return to text
2. However, the consumption financed through mortgage debt extension somewhat overestimates the
net influence of housing capital gains on consumption. Debt must be repaid, and presumably,
consumption is reduced as a consequence of the repayment. In the absence of capital gains, borrowing
merely moves up a purchase rather than augmenting total purchases through time.
However, in the presence of increased capital gains, unrealized but still perceived as permanent, debt
capacity and levels are likely to rise. The consequently lowered debt repayment relative to debt
extensions suggests that the rate of offset to the initial consumption expenditures at the time of
repayment is also likely to be a good deal less. Our preliminary estimates, in fact, suggest that such
subtractions from the gross effects on spending are modest. Return to text
3. The time sequence of the emergence of capital gains and their effect on consumer spending is a
function of the channel through which equity is extracted from homes. For sales of existing homes,
equity extraction is generally concurrent with a realization of a capital gain. Presumably, however, the
cash extracted influences consumer spending only over time. Unrealized gains can build up over time
without any obvious effect on spending. But a cash-out refinancing or a home equity loan is presumably
initiated for a specific current purpose. Thus, the lags between the emergence of a capital gain and
spending may be a function of the degree of gains realization and the particular mortgage vehicle
employed for equity extraction. Another means of equity extraction of unrealized gains for which data
are scarce outside of decennial censuses are long-term first lien mortgages on residences previously
free of debt. Return to text
4. Capital gains, however, have not been fully stripped from personal income. The capital gains
embedded in exercised stock options, for example, are included in compensation of employees (and as a
charge against profits) in the NIPA. These gains are taxed as regular income. Return to text
5. This is not done in the NIPA owing, in part, to a desire by the Bureau of Economic Analysis (BEA) to
conform with international standards for national accounts. Return to text |
