This chapter assesses the prospects for U.S. personal saving in light of the sharp contraction of consumer spending during the financial crisis. Using two alternative econometric approaches, the paper finds that the saving rate may increase somewhat from current levels to about 4¾–5½ percent of disposable income in the medium term.


This chapter assesses the prospects for U.S. personal saving in light of the sharp contraction of consumer spending during the financial crisis. Using two alternative econometric approaches, the paper finds that the saving rate may increase somewhat from current levels to about 4¾–5½ percent of disposable income in the medium term.

This chapter assesses the prospects for U.S. personal saving in light of the sharp contraction of consumer spending during the financial crisis. Using two alternative econometric approaches, the paper finds that the saving rate may increase somewhat from current levels to about 4¾–5½ percent of disposable income in the medium term.

A. Introduction

1. During the early stages of the financial crisis, consumers sharply curtailed their spending. The saving rate jumped from 1 percent of disposable income in Q1/2008 to 5½ percent in Q2/2009 on the back of slumping asset prices, unprecedented uncertainty, and rapidly tightening credit conditions. Real personal consumption expenditures fell for two consecutive years during 2008–09—the first time since the Great Depression. However, with the diminishing tail risks and higher financial asset values, consumers have become somewhat less cautious in recent months. The saving rate has fallen to around 3–4 percent of disposable income, supporting a tentative recovery in consumer spending. Against this background, this Chapter examines the prospects for the U.S. household saving rate.

B. Experience of Nordic Economies

2. The historical record of Finland, Norway and Sweden provides a cautionary tale. Similarly to the United States, these economies experienced joint asset price and banking busts in the late 1980s. In the aftermath of the crises, the personal saving rate tended to remain elevated for many years (Figure 1). The trough-to-peak increase in saving was considerable—between 5–10 percentage points of disposable income.

Figure 1.
Figure 1.

U.S. Household Saving Rate Adjustment Could Be Protracted

Citation: IMF Staff Country Reports 2010, 248; 10.5089/9781455206759.002.A002

Note: Fiscal policy = contribution of the budget deficit; usiness cycle = contribution of the unemployment rate; demographics = contribution of population aged over 65 years, wealth effects = contribution of wealth-toincome ratio and wealth-to-income ratio interacted with mortgage market index.

3. In the United States, the run up in the saving rate has so far been smaller than in the Nordic economies. This may reflect a variety of factors. The extraordinary policy response in the U.S. quickly eliminated tail risks, and asset prices stabilized much faster than in Nordic economies, after about a year. The initial imbalances in the household sector were also by some measures less pronounced in the United States—the pre-crisis saving rates were negative in Nordic economies, between -2 and -4 percent of disposable income.2

4. However, the historical experience of Nordic countries also suggests that consumer deleveraging can take a long time—between 6 and 8 years. One notable feature of the U.S. developments so far is that household indebtedness has fallen little from its historical peak (Figure 1), while lending standards remain very tight. This would point to the need to maintain the U.S. household savings at a much higher level than during the pre-crisis period.

C. Cross-Country Models of the Saving Rate

5. To assess prospects for household saving, staff estimated two alternative econometric models—panel regressions and a state-space system. The panel regressions link personal saving to a range of fundamentals such as wealth, corporate saving, interest rates, fiscal policy, and demographics (Table 1).3

Table 1.

Household Saving Rate: Baseline Regression Results

article image
Notes: †These series are expressed in percent of household disposable income. Sample: 1985-2007 (2009 for the U.S.). Regressions include fixed and time effects. The mortgage market index ranges from 0.23 (France) to 0.98 (United States). The index captures institutional differences in national mortgage markets such as availability of mortgage equity withdrawal, typical loan-to-value ratio and maturity. See IMF (2008) for details.Robust standard errors in parentheses.

significant at 10 percent;

significant at 5 percent;

significant at 1 percent.

6. The panel regressions suggest that the personal saving rate will remain elevated and could under plausible assumptions increase to about 4¾–5½ percent of disposable income in the medium term.4

  • Wealth. With future asset price growth likely subdued (and continued downside risks to house prices), consumers will need to save to rebuild their net wealth, which at 500 percent of disposable income remains well below the average of the past 20 years (533 pct of disposable income). This factor contributes over 2½ percentage points to the predicted net increase in the saving rate from 1¾ percent in 2007 to 4¾–5½ percent in 2018 when the economy returns back to potential (Figure 1).

  • Fiscal policy and interest rates. Persistent large fiscal deficits will affect consumer behavior in two ways—higher interest rates will stimulate saving, while expectations of future tax increases may encourage some consumers to spend less relative to their current income. The higher prospective deficits and interest rates contribute over 2¼ percentage points to the predicted increase in the saving rate.

  • Demographics. In contrast, retirement of the baby-boomer generation over the next several years could reduce the U.S. saving rate by more than 2 percentage points. That said, the demographic effects could be somewhat weaker in the current environment, as some employees will need to work beyond their normal retirement age to replenish the wealth lost from their defined-contribution pension plans.

  • Cyclical effects. Cyclical factors such as the unemployment rate do not play a large role when decomposing the predicted changes in the saving rate between 2007 and 2018.5 However, the cyclical effects should boost the saving rate between now and the end of forecast period, as the unemployed who find a job in the future will be able to save more.

7. Uncertainty around the baseline saving rate projection is significant. For example, asset prices could rebound more strongly than expected, while robust foreign demand for the Treasury bonds could keep the yields and lending rates low. The households would then rebuild their balance sheets mostly through higher asset prices and the saving rate could be lower than under the baseline (Figure 1). Alternatively, ambitious fiscal consolidation could also reduce the household saving rate by limiting Ricardian effects and putting a ceiling on future interest rate increases.

D. What is the New Optimal Wealth Level?

8. Besides the uncertainty about future asset values, there is also little clarity about what consumers currently consider to be a “normal” level of assets, which provides them with an acceptable insurance against unexpected events. Staff therefore estimated an alternative model of U.S. “target wealth”, in which consumers tend to save more whenever the actual wealth is below the “target”, and vice versa (see Appendix). The target wealth is identified using a state-space model where the target wealth depends on the real interest rate and a statistical measure of uncertainty. Besides the deviation of actual wealth from the target, the household saving rate is also assumed to be affected by credit conditions and interest rates.

9. The model suggests that the target wealth has increased during the financial crisis. In the immediate pre-crisis period, the target net wealth was unusually low—well below 500 percent of disposable income—given the low interest rates and “great moderation”, i.e., an unusual drop in macroeconomic volatility. The large gap between actual and target wealth combined with loosening credit conditions explain why the saving rate fell so low during the 2000s (Figure 1). Since the onset of the crisis, the target wealth has increased considerably given large uncertainties, and could stabilize around 540–550 percent of disposable income once interest rates normalize. The medium-term saving rate predicted by the model is about 5¼ percent.

E. Conclusions

10. Despite the recent decline of the personal saving rate, households are not likely to return to their pre-crisis spending habits. While uncertainty has diminished and financial conditions have eased, the structural need to rebuild balance sheets, high fiscal deficits, and higher interest rates could raise the household saving rate in the medium term. The recent policy proposals to automatically set-up retirement plans and provide tax breaks for matching employer contributions could also boost saving (Economic Report of the President, 2010).

11. That said, the near-term dynamics of the saving rate is uncertain for a variety of reasons. In addition to the factors discussed above, the planned tax increases for upperincome households could reduce disposable income growth, thereby putting downward pressure on aggregate saving (but also on consumption). On the other hand, corporate profitability has remained strong and higher dividend payments could facilitate more saving in the near term. All in all, a significant decline in the saving rate appears unlikely and, given the expectations of slow recovery in the labor market, consumption growth will remain sluggish relative to the trends during previous recoveries.

Appendix 1. State-Space Model For Saving Rate and Wealth

In this specification, the saving rate depends on the deviation of household wealth from its (unobserved) target m-m*, credit conditions CC (approximated by banks’ willingness to extend consumer credit), and real interest rate r:


Consistent with the precautionary saving literature, the unobserved target wealth is modeled as a function of uncertainty δσ (measured by the Bloom’s index6) and real interest rates:


It is assumed that the gap between actual and target wealth can be highly persistent:


Realistically, the measured variables track the “true” underlying uncertainty and expected real interest rates (denoted by stars) only imperfectly:


The resulting state-space system is estimated using the U.S. quarterly data during 1966–2009. The estimated coefficients are reported in the table below. Figure on the next page plots the estimated path for target household wealth.

Table 1

State-Space Model: Coefficient Estimates

article image
Note: Standard errors are reported in parentheses. Star notation is the same as in Table 1.


  • Bloom, Nicholas, 2009, “The Impact of Uncertainty Shocks,” Econometrica, 77(3), p. 623– 685.

  • Economic Report of the President, 2010, “Saving and Investment”, Chapter 4 (Washington, D.C.).

  • Deutsche Bank, 2009, “Consumer Balance Sheet Adjustment: Half Way Done” (New York, NY).

  • International Monetary Fund, 2008, “The Changing Housing Cycle and The Implications for Monetary Policy”, Chapter 3 of World Economic Outlook (Washington, D.C.)

    • Search Google Scholar
    • Export Citation
  • Lee, Jaewoo, Pau Rabanal, and Damiano Sandri, 2010, “U.S. Consumption after the 2008 Crisis”, IMF Staff Position Note 10/01, January (Washington, D.C.).

    • Search Google Scholar
    • Export Citation

Prepared by Martin Sommer with contributions from Jirka Slacalek.


In addition, the Nordic economies continued to be pummeled by external shocks (economic transition in the CEECs, exchange rate volatility), their inflexible economies went through a period of labor and product market liberalization, and domestic asset prices fell for at least 4 consecutive years.


The sample consists of the G-7 data over 1985–2007 (2009 for the United States).


Deutsche Bank (2009) and Lee, Rabanal, and Sandri (2010) also find that the personal saving rate may settle above present levels in the medium term.


The unemployment rate is projected at close to 5 percent in both years.


The Bloom (2009) index combines information about stock market volatility, distribution of firm-level and industry-level growth rates, unemployment, and other relevant variables.