## _wp1242

## Source details

**Canonical URL:** [_wp1242](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp1242.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp1242.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp1242.pdf.json)

---

### Introduction and key question
- The Great Recession featured a striking increase in uncertainty following the Great Moderation, with repeated and sizeable downward revisions of growth projections (Figure 1).
- Empirical observation: between 2007 and 2009 consumption growth declined nearly ubiquitously across the sample, and household saving rates rose (Figure 2).
- Paper's aim: examine whether heightened uncertainty raised precautionary saving at the country level using an unbalanced panel of 27 advanced countries (1980–2010).

### Theoretical model of precautionary savings
- Framework: infinitely lived agent with constant relative risk aversion utility, stochastic labor income and stochastic returns on saving.
- Labor income unemployment specification: binary distribution with unemployment rate u, replacement rate ζ, correction factor λ so that expected income remains Y.
- Investment return specification: binary distribution with expected return r and shock size η.
- Calibration parameters used in examples:
  - risk aversion = 2
  - replacement rate ζ = 50 percent
  - expected return r = 2 percent
  - discount factor β set to 0.98 (more precisely 0.97933 to generate equilibrium wealth-to-income ratio of 2)
  - baseline unemployment u = 5 percent
  - baseline return shock η = 1 percent
- Model implications tested empirically:
  - An increase in unemployment risk raises the saving rate (illustrated by a permanent increase from 5 to 10 percent unemployment).
  - The effect of higher investment risk on the saving rate is ambiguous and generally muted (example: η from 1 to 3 percent can lower saving rate under calibration because risk of capital losses counteracts precautionary motive).
  - A reduction in wealth raises the saving rate as agents rebuild precautionary assets (example: an exogenous 10 percent wealth reduction lowering wealth-to-income from 2 to 1.8 leads to higher saving rate).

### Econometric approach
- Dependent variable: household net saving rate from the OECD.
- Main regressors include:
  - contemporaneous unemployment rate (UR_t)
  - lead of real disposable income growth (1 ln+Δ DI_t)
  - real short-term deposit rate (R_t)
  - GDP volatility V_t (instantaneous time-varying standard deviation of per capita real GDP growth from a first-order GARCH)
  - stock market volatility SM_t (standard deviation of daily changes over a year)
  - lagged household financial net worth to disposable income (W_t−1)
- Baseline regression specification:
  - s_t = α1 UR_t + α2 Δ ln DI_{t+1} + α3 R_t + α4 V_t + α5 SM_t + α6 W_{t-1} + ... + ε_t
- Data: annual panel data; country fixed effects used to control for unobserved heterogeneity. Sample: unbalanced panel of 27 advanced economies (1980–2010).

### Core empirical findings on uncertainty and savings
- Unemployment risk:
  - Country fixed-effect regression shows a positive, highly significant relationship.
  - Representative unemployment rate coefficients include 0.55***, 0.49***, 0.34*** (t-statistics [5.15], [4.43], [3.23]) in fixed effects and 0.48***, 0.30** in random effects.
  - Interpretation example: "A one percent increase in unemployment is associated with a half percent increase in the saving rate."
- GDP volatility (V_t):
  - GDP volatility is highly significant with expected sign; estimates include 0.96***, 1.05***, 0.77***, 0.91*** (t-statistics [4.67], [4.83], [3.91], [4.28]).
  - Paper reports: "an increase in income uncertainty by 1 percent is associated with a higher household saving rate by about 1 percentage point."
- Stock market volatility (SM_t):
  - Market volatility shows no clear influence; coefficient estimates include 0.23, -0.02, 0.21 with t-statistics ([0.71], [-0.07], [0.64]) indicating insignificance.
  - Consistent with model: investment-return uncertainty has ambiguous and generally muted impact on aggregate saving.
- Controlling for GDP volatility strengthens the coefficient on expected income growth.

### Wealth effects and housing
- Lagged financial net worth (scaled by disposable income):
  - Statistically significant negative correlation with saving rates. Coefficient examples: -2.17*** (t-stat [-5.09]) and -1.53*** (t-stat [-3.78]).
  - Interpretation: higher financial wealth lowers household saving rates; negative wealth shocks raise saving.
- Change in house prices (proxy for housing wealth):
  - Change in house prices scaled by disposable income has strongly negative coefficients: -21.43*** and -21.81** (t-stats [-7.69], [-7.70]).
  - Due to sample limitations, housing variable excluded in some specifications; financial net worth interpreted as partly capturing housing effects.

### Other determinants (robustness and extensions)
- Interest rate:
  - Real short-term deposit rate positively correlated with saving; examples include coefficients 0.74***, 0.75*** (t-stats [9.36], [9.53]) and 0.77***, 0.54*** in extended specifications.
- Lead of disposable income growth:
  - Negative association with saving; examples include -0.18**, -0.19***, -0.20*** and -0.31*** in different tables (t-statistics shown).
- Fiscal balance (structural balance % of potential GDP):
  - Coefficient -0.21** (t-stat [-2.25]); interpretation: widening government deficit by 1 percentage point of GDP raises household saving rate by around 0.2 percentage points (consistent with Ricardian effects).
- Demographics:
  - Old dependency ratio (%) negatively correlated with saving: coefficients include -0.75***, -0.63*** (t-stats [-6.79], [-5.26]).
  - Young dependency ratio (%) positively correlated with saving: coefficients 0.27**, 0.23** (t-stats [2.34], [2.01]).
- Credit conditions (private sector credit % of GDP):
  - Negative relationship with saving: coefficient -0.02*** (t-stat [-5.78]); interpretation: tighter credit conditions associated with higher saving rates.
- Robustness:
  - Results robust to subsamples and exclusion of individual countries/years (Table 5). Using 3-year averages weakens some significance due to smoothing of within-country variation.

### Global factors
- Global variables considered: world real GDP growth, copper-to-gold price ratio, TED spread.
- Table 6 results with global variables:
  - Unemployment rate coefficients remain significant: examples 0.28***, 0.27***, 0.21** (t-stats [2.84], [2.71], [2.11]).
  - Copper price / gold price: highly significant negative coefficients -7.69***, -9.33*** (t-stats [-2.85], [-3.42]); interpretation: expectation of stronger future growth reduces need for saving.
  - TED spread: positive and significant 0.76**, 1.02*** (t-stats [2.13], [2.85]); interpretation: stress in interbank market associated with higher saving.
  - World real GDP growth contemporaneous coefficient reported as -0.12 (t-stat [-1.14], not significant).
- Conclusion: global conditions affect saving rates, but domestic uncertainty measures remain salient.

### Quantitative assessment of the Great Recession
- Fitted values from regressions (column (2) of Table 4 and column (5) of Table 6) track average saving rate dynamics well (Figures 7–8).
- Decomposition of predicted change in saving rate between 2007 and 2009 (Figure 9):
  - Without global factors: increase in unemployment and GDP volatility together contribute more than 50 percent of the predicted increase in the saving rate.
  - With global factors included: unemployment and GDP volatility still account for over two fifths of the increase.
  - Financial net worth fell sharply in dataset (text notes financial net worth as share of disposable income fell between 2005–2009 average and crisis years), and asset value reductions also likely contributed to precautionary savings.
- Model tends to mildly over-predict the rise in saving rates; potential explanations include habits in consumption or lagged awareness of worsening conditions.

### Key findings on drivers of household saving during the Great Recession
- Precautionary savings played a central role during the Great Recession, driven by increased economy-wide uncertainty that affected labor income more than investment returns.
- More than two fifths of the increase in saving rates can be directly related to the increase in unemployment risk and GDP volatility.
- Financial wealth losses contributed considerably to the rise in saving rates; those wealth losses may themselves have been caused by heightened uncertainty.
- The decline in expected income growth contributed only moderately to the rise in saving rates.
- Fiscal stimulus during the crisis induced increased household saving rates.
- Monetary easing, through falling real deposit rates, created an offsetting decline in the saving rate.
- Household saving rates increased in response to the deteriorating outlook for global growth and tightening financial conditions.
- Consumption habits and continued macroeconomic uncertainty are likely to limit a rapid return of household saving rates to pre-crisis levels in the short-run.

### Policy implications and recommendations
- Sustaining the recovery will require finding new sources of demand or strong and coordinated policy actions geared to restore confidence and reduce uncertainty.
- Policymakers must pay particular attention to the consistency of their statements to establish the credibility of their actions; inconsistent signals risk perpetuating the high-uncertainty environment.
- Because higher uncertainty and lower growth can become a “bad” equilibrium, restoring confidence is critical to reduce precautionary saving and revive consumption.

### Data coverage and definitions (selected)
- Country samples and periods include examples: Canada 1980-2010; France 1980-2009; Germany 1995-2009; Italy 1990-2010; Japan 1996-2008; U.K. 1995-2009; U.S. 1980-2010.
- Household net saving rate: percent of disposable income (OECD’s National Accounts).
- Unemployment rate: IMF’s World Economic Outlook.
- Real household net disposable income growth: OECD’s National Accounts.
- Real short-term deposit rate: IMF’s World Economic Outlook.
- GDP Volatility: Instantaneous time-varying standard deviation of quarterly data on year-on-year growth of real GDP per capita based on a GARCH (1,1) estimation. Real GDP per capita calculated using data from the IMF’s World Economic Outlook on real GDP and population, extended by data from Haver Analytics.
- Stock market volatility: Standard deviation of daily changes in Morgan Stanley stock market index over the period of one year. Global Insight database.
- Household financial net worth: OECD’s Financial Annual Accounts.
- Housing prices: Average annual house price index of all types of houses and vintages across the whole economy with listed country-specific exceptions; indices rescaled to 2005=100.
- General government structural balance in percent of potential GDP: IMF’s World Economic Outlook.
- Share of young (0-14 years) and old (65+ years), in percent of total working-age population: World Bank’s World Development Indicators.
- Domestic credit to private sector (% of GDP): World Bank’s World Development Indicators.
- World real GDP growth (%): IMF’s World Economic Outlook.
- Copper-to-gold price: IMF’s World Economic Outlook.
- TED spread: Haver Analytics.

*Source: _wp1242 - References (IMF working paper content supplied).*

### References .............................................................................................................

### _wp1242 - References

### Introduction and key question
- The Great Recession featured a striking increase in uncertainty following the Great Moderation, with repeated and sizeable downward revisions of growth projections (Figure 1).
- Empirical observation: between 2007 and 2009 consumption growth declined nearly ubiquitously across the sample, and household saving rates rose (Figure 2).
- Paper's aim: examine whether heightened uncertainty raised precautionary saving at the country level using an unbalanced panel of 27 advanced countries (1980–2010).

### Theoretical model of precautionary savings
- Framework: infinitely lived agent with constant relative risk aversion utility, stochastic labor income and stochastic returns on saving.
- Labor income unemployment specification: binary distribution with unemployment rate u, replacement rate ζ, correction factor λ so that expected income remains Y.
- Investment return specification: binary distribution with expected return r and shock size η.
- Calibration parameters used in examples:
  - risk aversion = 2
  - replacement rate ζ = 50 percent
  - expected return r = 2 percent
  - discount factor β set to 0.98 (more precisely 0.97933 to generate equilibrium wealth-to-income ratio of 2)
  - baseline unemployment u = 5 percent
  - baseline return shock η = 1 percent
- Model implications tested empirically:
  - An increase in unemployment risk raises the saving rate (illustrated by a permanent increase from 5 to 10 percent unemployment).
  - The effect of higher investment risk on the saving rate is ambiguous and generally muted (example: η from 1 to 3 percent can lower saving rate under calibration because risk of capital losses counteracts precautionary motive).
  - A reduction in wealth raises the saving rate as agents rebuild precautionary assets (example: an exogenous 10 percent wealth reduction lowering wealth-to-income from 2 to 1.8 leads to higher saving rate).

### Econometric approach
- Dependent variable: household net saving rate from the OECD.
- Main regressors include:
  - contemporaneous unemployment rate (UR_t)
  - lead of real disposable income growth (1 ln+Δ DI_t)
  - real short-term deposit rate (R_t)
  - GDP volatility V_t (instantaneous time-varying standard deviation of per capita real GDP growth from a first-order GARCH)
  - stock market volatility SM_t (standard deviation of daily changes over a year)
  - lagged household financial net worth to disposable income (W_t−1)
- Baseline regression specification (as expressed in paper):
  - s_t = α1 UR_t + α2 Δ ln DI_{t+1} + α3 R_t + α4 V_t + α5 SM_t + α6 W_{t-1} + ... + ε_t
- Data: annual panel data; country fixed effects used to control for unobserved heterogeneity. Sample: unbalanced panel of 27 advanced economies (1980 earliest, 2010 latest for panel coverage details).

### Core empirical findings on uncertainty and savings
- Unemployment risk:
  - Table 1: country fixed-effect regression shows a positive, highly significant relationship.
  - Representative estimates: unemployment rate coefficients reported across specifications include 0.55***, 0.49***, 0.34*** (t-statistics [5.15], [4.43], [3.23]) in fixed effects and 0.48***, 0.30** in random effects.
  - Interpretation example in prose: "A one percent increase in unemployment is associated with a half percent increase in the saving rate."
- GDP volatility (V_t):
  - Table 2: GDP volatility is highly significant with expected sign; estimates include 0.96***, 1.05***, 0.77***, 0.91*** (t-statistics [4.67], [4.83], [3.91], [4.28]).
  - Paper reports: "an increase in income uncertainty by 1 percent is associated with a higher household saving rate by about 1 percentage point."
- Stock market volatility (SM_t):
  - Table 2: market volatility shows no clear influence; coefficient estimates shown include 0.23, -0.02, 0.21 with t-statistics indicating insignificance ([0.71], [-0.07], [0.64]).
  - Consistent with model: investment-return uncertainty has ambiguous and generally muted impact on aggregate saving.
- Controlling for GDP volatility strengthens the coefficient on expected income growth.

### Wealth effects and housing
- Lagged financial net worth (scaled by disposable income):
  - Table 3 and Table 4: statistically significant negative correlation with saving rates. Coefficient examples: -2.17*** (t-stat [-5.09]) and -1.53*** (t-stat [-3.78]) in Table 3 specifications.
  - Interpretation: higher financial wealth lowers household saving rates; negative wealth shocks raise saving.
- Change in house prices (proxy for housing wealth):
  - Table 3: change in house prices scaled by disposable income has strongly negative coefficients: -21.43*** and -21.81** (t-stats [-7.69], [-7.70]).
  - Due to sample limitations, housing variable excluded in some specifications; financial net worth interpreted as partly capturing housing effects.

### Other determinants (robustness and extensions)
- Interest rate:
  - Real short-term deposit rate positively correlated with saving; examples include coefficients 0.74***, 0.75*** (t-stats [9.36], [9.53]) in Table 1 and 0.77***, 0.54*** in extended specifications.
- Lead of disposable income growth:
  - Negative association with saving; examples include -0.18**, -0.19***, -0.20*** in Table 1 and -0.31*** in Table 2 (t-statistics shown).
- Fiscal balance (structural balance % of potential GDP):
  - Table 4: coefficient -0.21** (t-stat [-2.25]); interpretation: widening government deficit by 1 percentage point of GDP raises household saving rate by around 0.2 percentage points (consistent with Ricardian effects).
- Demographics:
  - Old dependency ratio (%) negatively correlated with saving: coefficients include -0.75***, -0.63*** (t-stats [-6.79], [-5.26]) — consistent with life-cycle theory.
  - Young dependency ratio (%) positively correlated with saving: coefficients 0.27**, 0.23** (t-stats [2.34], [2.01]).
- Credit conditions (private sector credit % of GDP):
  - Negative relationship with saving: coefficient -0.02*** (t-stat [-5.78]); interpretation: tighter credit conditions associated with higher saving rates.
- Robustness:
  - Results robust to subsamples and exclusion of individual countries/years (Table 5). Using 3-year averages weakens some significance due to smoothing of within-country variation.

### Global factors
- Variables considered: world real GDP growth (contemporaneous), copper-to-gold price ratio (forward-looking growth proxy), TED spread (LIBOR minus 3-month US T-bill).
- Table 6 results (baseline with global variables added):
  - Unemployment rate coefficients remain statistically significant: examples 0.28***, 0.27***, 0.21** across specifications (t-stats [2.84], [2.71], [2.11]).
  - Copper price / gold price: highly significant negative coefficient examples -7.69***, -9.33*** (t-stats [-2.85], [-3.42]); interpretation: expectation of stronger future growth reduces need for saving.
  - TED spread: positive and significant (0.76**, 1.02***, t-stats [2.13], [2.85]); interpretation: stress in interbank market associated with higher saving.
  - World real GDP growth contemporaneous coefficient reported as -0.12 (t-stat [-1.14], not significant).
- Conclusion: global conditions affect saving rates, but domestic uncertainty measures remain salient.

### Quantitative assessment of the Great Recession
- Fitted values from regressions (column (2) of Table 4 and column (5) of Table 6) track average saving rate dynamics well (Figures 7–8).
- Decomposition of predicted change in saving rate between 2007 and 2009 (Figure 9):
  - Without global factors: increase in unemployment and GDP volatility together contribute more than 50 percent of the predicted increase in the saving rate.
  - With global factors included: unemployment and GDP volatility still account for over two fifths of the increase.
  - Financial net worth fell sharply in dataset (text notes financial net worth as share of disposable income fell between 2005–2009 average and crisis years), and asset value reductions also likely contributed to precautionary savings.
- Model tends to mildly over-predict the rise in saving rates; potential explanations include habits in consumption or lagged awareness of worsening conditions.

*Source: _wp1242 - References (IMF working paper content supplied).*

### 1.7 to 1.2—also was an important spur to increased saving. It is possible that the reduction in

### _wp1242 - 1.7 to 1.2—also was an important spur to increased saving. It is possible that the reduction in

### Key findings on drivers of household saving during the Great Recession
- Precautionary savings played a central role during the Great Recession, driven by increased economy-wide uncertainty that affected labor income more than investment returns.
- More than two fifths of the increase in saving rates can be directly related to the increase in unemployment risk and GDP volatility.
- Financial wealth losses contributed considerably to the rise in saving rates; those wealth losses may themselves have been caused by heightened uncertainty.
- The decline in expected income growth contributed only moderately to the rise in saving rates.
- Fiscal stimulus during the crisis induced increased household saving rates.
- Monetary easing, through falling real deposit rates, created an offsetting decline in the saving rate.
- Household saving rates increased in response to the deteriorating outlook for global growth and tightening financial conditions.
- Consumption habits and continued macroeconomic uncertainty are likely to limit a rapid return of household saving rates to pre-crisis levels in the short-run.

### Econometric and empirical evidence
- Aggregate savings increase in the face of economy-wide uncertainty, particularly when uncertainty affects labor income rather than investment returns.
- The analysis applies econometric estimates to decompose the steep increase in saving rates during the Great Recession and attributes specific shares of the increase to unemployment risk, GDP volatility, and wealth losses.
- The paper notes an ongoing “new normal” of elevated uncertainty reflected in increased reference to the words “volatility” and “uncertainty” in press coverage (Figure 10).

### Policy implications and recommendations
- Sustaining the recovery will require finding new sources of demand or strong and coordinated policy actions geared to restore confidence and reduce uncertainty.
- Policymakers must pay particular attention to the consistency of their statements to establish the credibility of their actions; inconsistent signals risk perpetuating the high-uncertainty environment.
- Because higher uncertainty and lower growth can become a “bad” equilibrium, restoring confidence is critical to reduce precautionary saving and revive consumption.

### Data coverage and definitions (selected)
- Country samples and periods include (examples from the dataset): Canada 1980-2010; France 1980-2009; Germany 1995-2009; Italy 1990-2010; Japan 1996-2008; U.K. 1995-2009; U.S. 1980-2010.
- Household net saving rate: percent of disposable income (OECD’s National Accounts).
- Unemployment rate: IMF’s World Economic Outlook.
- Real household net disposable income growth: OECD’s National Accounts.
- Real short-term deposit rate: IMF’s World Economic Outlook.
- GDP Volatility: Instantaneous time-varying standard deviation of quarterly data on year-on-year growth of real GDP per capita based on a GARCH (1,1) estimation. Real GDP per capita calculated using data from the IMF’s World Economic Outlook on real GDP and population, extended by data from Haver Analytics.
- Stock market volatility: Standard deviation of daily changes in Morgan Stanley stock market index over the period of one year. Global Insight database.
- Household financial net worth: OECD’s Financial Annual Accounts.
- Housing prices: Average annual house price index of all types of houses and vintages across the whole economy with listed country-specific exceptions; indices rescaled to 2005=100.
- General government structural balance in percent of potential GDP: IMF’s World Economic Outlook.
- Share of young (0-14 years) and old (65+ years), in percent of total working-age population: World Bank’s World Development Indicators.
- Domestic credit to private sector (% of GDP): World Bank’s World Development Indicators.
- World real GDP growth (%): IMF’s World Economic Outlook.
- Copper-to-gold price: IMF’s World Economic Outlook.
- TED spread: Haver Analytics.

*Source: Excerpt from IMF working paper content provided in the PDF chapter text.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp1242.pdf_