Abstract
This Selected Issues paper analyzes the United State’s (U.S.) household savings role in supporting the U.S. recovery; and focuses on the market for single-family housing, and the importance for household balance sheets. It discusses the underfunding of corporate pension plans, macroeconomic, and policy implications; the U.S. fiscal position, and reviews the causes of the fiscal crisis. It examines the impact of energy shocks, energy policy, and the taxation role. It analyzes the growth in linkages between the United States and other G-7 countries, and the regional and bilateral trade links issues.
I. U.S. Household Saving: Long-run Determinants and Short-term Risks1
1. The recent drop in household wealth has raised concern regarding the extent to which consumption can continue to support the U.S. recovery. Household net worth has fallen sharply since 2000, from about 6¼ times personal disposable income to 4½ times personal disposable income. While personal saving rates have risen somewhat, the increase seems smaller than what would be implied by a casual inspection of the relationship between wealth, income, and saving. Indeed, Figure 1 suggests that the personal saving rate would need to increase by around 2½–3 percentage points to match the current levels of household wealth.
Household Wealth and Saving Rates
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
Household Wealth and Saving Rates
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
Household Wealth and Saving Rates
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
2. However, the analysis presented in this chapter suggests that the saving rate is only about 1 percentage point below its long-run trend.2 This unexpectedly modest differential reflects several factors. First, the effect on the saving rate of the decline in equity wealth has been partially offset by the effect of increases in other forms of wealth, including real estate and relatively liquid wealth such as demand deposits. Second, the U.S. saving rate has exhibited a secular, downward trend over the past decade—due, in part, to financial innovation.
A. Recent Trends in Household Wealth
3. The sharp increase in household wealth during the past decade has reflected gains in the equity and residential markets, as well as other forms of wealth. Partly owing to the strength of housing prices and stock market gains, net housing and net equity wealth rose by $3¼ trillion and $2½ trillion, respectively, between end-1995 and 2003Q1 (Figure 2). However, net holdings of non-equity financial wealth rose by even more—by $6¼ trillion over the same period, reaching $21¾ trillion in 2003Q1 and exceeding the value of equities at the height of the stock market in early 2000.3
Components of Household Net Wealth
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
Components of Household Net Wealth
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
Components of Household Net Wealth
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
4. Since the collapse of the equity bubble, net non-equity financial wealth has become increasingly concentrated in liquid holdings.4 Most of the recent rise in non-equity financial wealth reflected an increase in demand deposits and money market shares—spurred by the stock market decline that led households to redirect savings into less risky assets and to lock in capital gains. During 2000–2003, the share of non-equity financial wealth held in the form of demand deposits rose from around 25 percent to nearly 30 percent. At the same time, holdings of corporate and municipal bonds rose rapidly.
5. To a large extent, the recent shift in the composition of household wealth represents a return to more traditional portfolio allocations. Between 1969–1996, equity wealth accounted for just under 20 percent of household net wealth, and non-equity financial wealth accounted for almost 60 percent; residential wealth comprised the balance. During the latter half of the 1990s, the share of equity wealth rose sharply—to almost 45 percent at the height of the stock market bubble, but has since fallen sharply to 25 percent, with the proportion of other assets returning to roughly their long-run averages (Figure 3).
Components of Household Net Wealth
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
Components of Household Net Wealth
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
Components of Household Net Wealth
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
B. Long-run Determinants of Consumption and Saving
6. A long-run consumption model is estimated in order to gauge the extent to which the saving rate has deviated from its long-run equilibrium The basic framework is based on the permanent income hypothesis (PIH), which posits that household consumption and saving decisions reflect expected levels of permanent income and wealth.5 Following Lettau and Ludvigson (2001), a long-run relationship is estimated between (the logs of) real consumption services (ct), real household assets (at), and real labor income (
where assets are decomposed into residential (ar), equity (ae), and financial non-equity wealth (an).6 The results are reported in Table 1.
Estimates of the Consumption Cointegrating Equation
Estimates of the Consumption Cointegrating Equation
| Wealth Aggregation | |||||
|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | ||
| Constant | 0.009 | -0.432 | -0.311 | -0.079 | |
| (0.150) | (0.148) | (0.160) | (0.168) | ||
| Time | 0.043 | 0.020 | 0.026 | 0.027 | |
| (0.005) | (0.005) | (0.005) | (0.005) | ||
| Residential (ar) | 0.113 | 0.057 | |||
| (0.015) | (0.025) | ||||
| Equity (ae) | 0.056 | 0.059 | |||
| (0.005) | (0.005) | ||||
| Non-equity (an) | 0.264 | ||||
| (0.065) | |||||
| Composite Wealth | 0.255 | 0.156 | 0.287 | ||
| (0.035) | (0.017) | (0.024) | |||
| Income | 0.630 | 0.818 | 0.651 | 0.577 | |
| (0.087) | (0.059) | (0.046) | (0.067) | ||
| Wealth Effect | 3.5¢/$ | 2.0¢/$ | 3.1¢/$ | 2.3¢/$ | |
| Adjusted R2 | 0.996 | 0.997 | 0.997 | 0.997 | |
| Akaike Info. Criterion | -6.448 | -6.715 | -6.879 | -6.764 | |
| Wald test of: | |||||
| Income & Wealth sum to 1 | 3.955 | 2.779 | 0.015 | 0.771 | |
| p-value | 0.050 | 0.068 | 0.985 | 0.514 | |
| Sample period: 1970Q1–2002Q4 | |||||
Estimates of the Consumption Cointegrating Equation
| Wealth Aggregation | |||||
|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | ||
| Constant | 0.009 | -0.432 | -0.311 | -0.079 | |
| (0.150) | (0.148) | (0.160) | (0.168) | ||
| Time | 0.043 | 0.020 | 0.026 | 0.027 | |
| (0.005) | (0.005) | (0.005) | (0.005) | ||
| Residential (ar) | 0.113 | 0.057 | |||
| (0.015) | (0.025) | ||||
| Equity (ae) | 0.056 | 0.059 | |||
| (0.005) | (0.005) | ||||
| Non-equity (an) | 0.264 | ||||
| (0.065) | |||||
| Composite Wealth | 0.255 | 0.156 | 0.287 | ||
| (0.035) | (0.017) | (0.024) | |||
| Income | 0.630 | 0.818 | 0.651 | 0.577 | |
| (0.087) | (0.059) | (0.046) | (0.067) | ||
| Wealth Effect | 3.5¢/$ | 2.0¢/$ | 3.1¢/$ | 2.3¢/$ | |
| Adjusted R2 | 0.996 | 0.997 | 0.997 | 0.997 | |
| Akaike Info. Criterion | -6.448 | -6.715 | -6.879 | -6.764 | |
| Wald test of: | |||||
| Income & Wealth sum to 1 | 3.955 | 2.779 | 0.015 | 0.771 | |
| p-value | 0.050 | 0.068 | 0.985 | 0.514 | |
| Sample period: 1970Q1–2002Q4 | |||||
7. The results illustrate substantial differences in the “wealth effect,” depending on how the three asset categories are aggregated. The estimated wealth effect—i.e., the impact on consumption of a $1 dollar increase in wealth—ranges from 2–3½ cents. The smallest wealth effects are found if residential wealth is included separately, rather than being combined with another wealth component. In the disaggregated equation, the impact of shocks to equity and residential wealth are relatively small, with the largest effect coming from the non-equity financial wealth term. This large effect could be due to the relatively liquid nature of much of non-equity financial wealth (in the form of demand deposits), which could yield a non-pecuniary return in terms of lower transactions costs, or the possibility that more liquid assets reduce the need for precautionary saving. The importance of this term in the results reported here raises the question whether other studies that have shown a larger wealth effect from residential wealth may have been biased by the omission of non-equity financial wealth.7
8. The regressions where wealth is disaggregated appear to fit the PIH model better than the traditional regression that consolidates all of the wealth terms into one variable. The PIH model used to derive this cointegrating relationship predicts that the coefficients on the income and wealth terms should sum to one. A Wald test of this hypothesis is rejected for the model where all wealth is aggregated into one consolidated term (column 1), but fails to reject the hypothesis for the more disaggregated models. Further, the Akaike Information Criterion (AIC) indicates that the model with consolidated wealth terms fits the data the least well, out of all of the models considered here. Instead, AIC prefers the model that has residential and non-equity financial wealth combined into a composite wealth term.
9. The results also suggest that consumption has exhibited an exogenous upward trend, possibly related to improvements in household access to credit. This result is consistent with the study by Cerisola and De Masi (1999), which focused on the long-run properties of the personal saving rate and identified a downward trend in the saving rate. In their study, the longer-term decline in the saving rate is explained by improved household access to credit.
10. The consumption equation indicates that the personal saving rate is about 1 percentage point lower than the rate determined by fundamentals (Figure 4).8 Since saving is the residual between disposable income and consumption, the fitted values for consumption from the preferred equation (column 3) imply an equation for the saving rate.9 The fitted value of the personal saving rate is calculated at 4½ percent at end-2002, compared with the actual rate of 3½ percent, suggesting that the saving rate was only modestly out of line with prevailing levels of household income and wealth.
Personal Saving Rates
Actual and estimated rates from long-run relationships
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
Personal Saving Rates
Actual and estimated rates from long-run relationships
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
Personal Saving Rates
Actual and estimated rates from long-run relationships
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
11. The estimates from the long-run equation also confirm that the sharp retrenchment in equity prices has had a substantial effect on the saving rate (Figure 5). After declining steadily since the early 1990s, the personal saving rate rose by around 2 percentage points during 2000–2003. The model parameters suggest that the decline in equity wealth accounted for all of the increase, since it helped to push the saving rate up by 2 percentage points.10 At the same time, strong income growth also continued to exert upward pressure on the saving rate. These factors more than offset the dampening effect on the saving rate of the strong growth of residential and non-equity financial wealth.
Changes in the Personal Saving Rate
3-year change in selected components of the saving rate
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
Changes in the Personal Saving Rate
3-year change in selected components of the saving rate
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
Changes in the Personal Saving Rate
3-year change in selected components of the saving rate
Citation: IMF Staff Country Reports 2003, 245; 10.5089/9781451839623.002.A001
C. Does Saving Pose a Risk to the Outlook?
12. Despite the moderate shortfall in personal saving relative to levels determined by income and wealth, any upward adjustment could weigh on the short-term outlook. If saving were to adjust by the full 1 percentage point of disposable income necessary to return the actual rate to its fitted value, the effect would be to reduce GDP growth by around ¾ percentage points.
13. Furthermore, a larger correction in the saving rate remains a risk if the economic recovery disappoints. Because the rise in housing and other forms of wealth has been relatively strong, a slower-than-expected recovery could lead to a continued decline in household wealth and weaker labor market conditions. Net housing wealth, in particular, is vulnerable because a slowing economy could depress house prices—a potential danger because households have borrowed extensively against the value of their homes.11
14. Finally, the trend decline in the saving rate may not continue into the future. As Cerisola and De Masi (1999) highlight, financial innovation has likely allowed households to save less and still achieve the same level of wealth and consumption. If the pace of innovation has slowed or has been nearly completed, the saving rate is unlikely to continue its trend decline.
References
Attanasio, O.P. 1999. “Consumption Demand,” in: J.B. Taylor and M. Woodford (eds.), Handbook of Macroeconomics (vol. 1b), New York: Elsevier.
Case, K., J. Quigley and R. Shiller, 2001, “Comparing Wealth Effects: The Stock Market Versus the Housing Market,” NBER Working Paper No. 8606.
Cerisola, M., and P. De Masi, 1999, “Determinants of the U.S. Personal Saving Rate,” United States: Selected Issues, International Monetary Fund Staff Country Report No. 99/101.
Davis, M.A. and M.G. Palumbo, 2001, “A Primer on the Economics and Time Series Econometrics of Wealth Effects,” Federal Reserve Board, Finance and Economics Discussion Series, Working Paper No. 2001-09.
Hamilton, J., 1994, Time Series Analysis, Princeton: Princeton University Press.
IMF (International Monetary Fund), 2002, “Is Wealth Increasingly Driving Consumption?,” World Economic Outlook, Washington, D.C. (April 2002).
Lettau, M., and S. Ludvigson, 2001, “Consumption, Aggregate Wealth, and Expected Stock Returns,” The Journal of Finance, Vol. 56, No. 3, pp. 815–49.
Mead, C.I., C.P. McCully, and M.B. Reinsdorf, 2003, “Income and Outlays of Households and of Nonprofit Institutions Serving Households,” Survey of Current Business, Washington, D.C.: Department of Commerce (April).
OECD (Organization for Economic Cooperation and Development), 2001, OECD Productivity Manual: A Guide to the Measurement of Industry-Level and Aggregate Productivity Growth, Paris.
Rudd, J., and K. Whelan, 2002, “A Note on the Cointegration of Consumption, Income, and Wealth,” Federal Reserve Board, mimeo.
Stock, J.H., and M. Watson, 1993, “A Simple Estimator of Cointegrating Vectors in Higher Order Integrated Systems,” Econometrica, Vol. 61, No. 4, pp. 783–820.
Ucello, C.E, 2001, “Are Americans Saving Enough for Retirement?,” Center for Retirement Research, Issue In Brief No. 7.
APPENDIX: Details on Data Construction
Consumption services
This chapter provides a direct estimate of household consumption services, which is then used to estimate long-run cointegration equations. While Lettau and Ludvigson (2001) assume that the flow of consumption services can be proxied by nondurables and services consumption, Rudd and Whelan (2002) note that this approximation is poor. Indeed, when measured properly (below), consumption services is not a constant multiple of nondurable and services consumption—a critical assumption in Lettau and Ludvigson (2001). Nevertheless, the consumption services series is very similar to the personal consumption expenditures, because nondurables and services account for nearly 90 percent of both series.
The service flow of consumer durables is imputed from the consumer durables stock (from annual BEA data)—following the methodology suggested in OECD (2001, pp. 54–69). It is taken as the user cost of capital times the stock of durables for each of the 13 major categories of consumer durables.
The user cost of capital requires a measure of depreciation and interest rates for each variable. Depreciation is available from the BEA; the interest rates are calculated assuming that the rates correspond to the interest rates on auto loans, home mortgages, or personal loans (the 13 cost of capital terms use one of these three rates).
The nominal and real data for the stock of consumer durable assets are interpolated to a quarterly frequency, using the quarter’s share of nominal spending on consumer durables in the yearly total to allocate the yearly change in the capital stock into each quarter. Thus, the level of the capital stock in the fourth quarter corresponds to the yearly capital stock data, which is measured on a year-end basis. An implicit price deflator is derived from the nominal and real service flow data.
Once consumption services are available for the 13 categories (in nominal and real dollars), the 13 categories are chain-weighted together to form the flow series for durables consumption. The service flow of consumer durables is then chain-weighted with the consumption data for nondurables and services to form a chain aggregate.
Household net wealth
The Federal Reserve’s Flow of Funds provides a detailed breakdown of the net wealth position of households and nonprofits. This chapter uses three, broad classifications: net residential wealth, net equity wealth, and net non-equity financial wealth:
Net residential wealth is the difference between owner-occupied real estate assets and household mortgages.
Net equity wealth is calculated as the difference between broad equity assets and security credits. Broad equity assets are defined in Davis and Palumbo (2001, p. 46) as the sum of: (a) household holdings of corporate equities (Flow of Funds, B.100 line 24); (b) corporate equities held by private pension funds (L.119, line 14); (c) corporate equities held by state and local government retirement funds (L.120, line 13); (d) corporate equities held by bank personal trusts and estates (L.116, line 14); (e) corporate equities held by closed-end funds (L.123, line 6); (f) corporate equities held by mutual funds (L.122, line 10); and (g) corporate equities held by life-insurance companies (L.117, line 13), multiplied by the ratio of reserves of life insurance companies (L.117, lines 18 and 19) to the total financial assets of life insurance companies (L. 117, line 1).
Net non-equity financial wealth is the residual of net financial wealth (defined by the Flow of Funds) less net equity wealth (as measured above). Non-equity financial assets include: demand deposits, Treasury securities, corporate securities, municipal bonds, the imputed equity value of noncorporate businesses (e.g., sole proprietorships), and non-profits wealth. Non-equity financial liabilities include: consumer credit, bank loans, and nonprofit liabilities (commercial mortgages and trade payables).
Net wealth of non-profits is included in non-equity financial wealth, even though a significant component of their wealth is in real estate. Since the saving patterns of nonprofits and households are very different (Mead, McCully, and Reinsdorf, 2003), including non-profit real estate wealth in residential wealth would result in potentially biased results for the net housing wealth coefficient.
Because the service flow of durables is included in consumption services, consumer durable assets are not included in the measure of net household wealth.
Household labor income
Labor income is taken as personal disposable income less: proprietors’ income (with inventory valuation and capital consumption adjustments); rental income (with capital consumption adjustments); dividend income; and interest income.
Household normalization
The consumption, wealth, and income terms are all scaled by the number of U.S. households to accurately gauge the effect of household formation rates on the long-run equilibrium relationships.
While other authors have used per capita consumption, this chapter takes households as the most important unit for consumption and saving decisions. Household formation rates are an important determinant of residential wealth accumulation, which subsequently affects consumption of durables. Additionally, many services are also consumed at the household, and not individual, level—such as utilities and homeowner’s rent.
Prepared by Christopher M. Faulkner-MacDonagh (WHD).
However, this does not mean that U.S. households are saving “enough” for retirement or other goals. Indeed, survey evidence in Ucello (2001) suggests that 5–25 percent of U.S. households are not likely saving enough for retirement; most of these are poor, unmarried households.
Non-equity financial wealth comprises paper, bonds, deposits, and equity in noncorporate business (the imputed value of small businesses and sole proprietorships).
Not all non-equity financial wealth is liquid. A sizable portion (just under one-quarter, or $5 trillion) is in the form of equity in noncorporate businesses, which could be difficult to sell quickly.
The literature on consumption behavior is large and reviewed in Attanasio (1999).
Stock and Watson (1993) suggest including the leads and lags, which are not reported in Table 1, to address the endogeneity bias in OLS estimates of cointegrating relationships. The Akaike Information Criterion is used to choose the lag length (k=5). The standard errors are corrected as suggested in Hamilton (1994), p. 610.
For example, Case, Quigley, and Shiller (2001) and IMF (2002) use house prices to proxy for residential wealth and include a measure of equity wealth, but do not include a measure of non-equity financial wealth. Indeed, when residential and non-equity financial wealth are combined into a single variable (column 3, Table 1), the composite wealth term is about 5 times the coefficient on equity wealth.
The FM-OLS results from Cerisola and De Masi (1999) were duplicated using data from this paper, with the personal saving rate a function of: wealth to income ratios, the general government fiscal balance, and Social Security and Medicare transfers. These results suggest an “equilibrium” saving rate of 5 percent, only somewhat higher than estimated by this paper.
Technically, equation (1) provides an estimate for consumption services. To derive the fitted value of consumption, the fitted value of consumption services is adjusted by the difference between actual consumption and consumption services.
The components in Figure 5 may not sum to the total changes in the personal saving rate because the equation in (1) does not lend itself to a close-form solution for the saving rate, and because the trend term and residual terms are not included in the chart.
See Chapter II for a discussion of real estate market developments.
