## _wp1011

## Source details

**Canonical URL:** [_wp1011](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1011.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1011.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1011.pdf.json)

---

### I. Introduction — scope and motivation
- Private consumption constitutes the largest share of GDP: in the United States, private consumption has, in recent years, been 70 percent of GDP; in Germany, private consumption constitutes almost 60 percent of GDP.
- Paper objective: use an econometric specification of consumption growth for G-7 economies (Canada, France, Germany, Italy, Japan, the U.K, and the U.S.) to project consumption growth and assess effects on global imbalances.
- Two empirical observations guiding the model:
  - Savings rates vary over time within countries.
  - Large cross-country differences in savings persist (e.g., U.S. savings lower than Germany’s for nearly each of the last 40 years).
- Innovation: inclusion of a country’s economic volatility (standard deviation of forecast error of GDP growth) as a determinant of consumption growth.

### II. Data, sample, and methodology
- Sample and coverage:
  - Sample start: in principle 1984 but starting dates differ by country due to missing data; analysis ends in 2007.
  - G-7 Countries coverage: Canada 1984-2007; France 1984-2007; Germany 1992-2007; Italy 1984-2007; Japan 1984-2007; U.K. 1984-2007; U.S. 1984-2007.
  - Additional countries for robustness: Australia, Austria, Belgium, Finland, Ireland, the Netherlands, Norway, Spain, and Sweden with varying coverage.
  - Observations per country: between 16 and 24 observations per country.
- Data frequency and sources:
  - Annual data used. Real consumption growth: Annual log change in volume index for private final consumption (OECD). Real disposable income growth: Annual log change in net disposable income (OECD). Household wealth: Financial net worth of households and nonprofit institutions (OECD). Unemployment rate and real interest rates: OECD’s Monthly Economic Indicators. Population shares: World Bank’s World Development Indicators. GDP Volatility: Standard Deviation of forecast error of a cross-country VAR regression of real GDP growth (Stock and Watson (2005), extended by Carare and Mody).
- Econometric approach:
  - Panel assumption: all variables influence consumption growth in the same way across countries; cross-country differences arise from differences in explanatory variables.
  - Panel techniques with country and year fixed effects used in reported regressions.
  - Volatility measure Vt: standard deviation of the forecast error of real GDP growth estimated via a “factor-structural” VAR (FSVAR) following Stock and Watson (2005); estimates extended to 2007 by Carare and Mody (forthcoming).
  - Error-correction via lagged consumption as share of disposable income c_{t-1}.
  - Potential reverse causality acknowledged; results presented as descriptive rather than causal.

### III. Basic estimating equation
- Reported specification (as presented in the source):
  tttttt VUERyccεααααα + +Δ+Δ++=Δ
  -432110
  lnln.
- Dependent variable: Δlnc_t, the log change in real consumption.
- Explanatory variables:
  - c_{t-1}: lagged consumption as share of disposable income (error correction / mean reversion).
  - Δlny_t: log change in real disposable income.
  - ΔUER_t: change in unemployment rate from the previous year (precautionary-risk proxy).
  - V_t: GDP volatility (standard deviation of forecast error of real GDP growth).

### IV. Main empirical findings
- Error-correction and income effects:
  - The error correction term (c_{t-1}) generally has a negative and significant sign.
  - A rise in real disposable income is associated with increased consumption. The coefficient is in the range of 0.2, suggesting a marginal propensity to consume of 20 percent (consistent with Carroll (2001)).
- Precautionary behavior — unemployment and volatility:
  - ΔUER (change in unemployment rate) is always solidly statistically significant: a rise in unemployment is associated with a clear reduction in consumption growth.
  - V (GDP volatility) is robustly associated with a decline in consumption growth across specifications.
  - Magnitudes reported:
    - A rise in the unemployment rate by one percentage point reduces consumption growth by ¾ of a percentage point.
    - A one percentage point increase in the standard deviation of the forecast error is associated with about 0.9 percentage point decrease in consumption growth.
- Country-group differences and averaging:
  - Error-correction effect weaker in non-G-7 economies and weaker with 3-year averages.
  - Income effect somewhat weaker in non-G-7 economies and in 3-year averages.
  - Effect of ΔUER rises in 3-year averages.
  - Influence of V somewhat smaller in non-G-7 economies but similar across annual and 3-year averaged data.
- Wealth and long-run factors:
  - Households set target levels of wealth as buffers; annual consumption variation from this mechanism is usually limited because year-to-year wealth changes are small.
  - Large losses of financial wealth (as in the crisis) have larger effects on consumption and can have persistent cross-country effects.
  - Long-term tendency for financial instability (e.g., volatile equity premium) is associated with lower average consumption growth.
- Demographics:
  - Working-age population supporting more dependents (young and old) is associated with lower consumption growth; dependency rates contribute to long-term cross-country differences.

### V. Robustness checks: sample and sensitivity
- Appendix Tables A and B report the basic regression when we drop one country at a time.
  - Table A: full sample of 16 countries; Table B: drop-one-country from the G-7 economies.
  - Coefficients change somewhat but results remain robust.
- Dropping one year at a time confirms results are not driven by any particular year.
- Results with three-year averages are reported for select analyses; with 54 observations adding extra variables reduces degrees of freedom but primary results remain similar.

### VI. Wealth effects (Section III) — detailed findings
- Real interest rate does not appear to matter (consistent with Campbell and Mankiw (1989) and Carroll and Slacalek (2009)).
- Wealth variable w_{t-1}: lagged stock of households’ net financial wealth relative to disposable income (Gruber and Kamin (2008)).
- Fixed-effects regression:
  - Wealth has expected positive sign but is not statistically significant.
  - Non-linearity: wealth term negative, square term positive and significant.
  - Marginal effect: above a ratio of wealth-to-disposable income of about 2½ (close to the median), a rise in wealth is associated with a rise in consumption growth; below that cut-off the marginal effect is negative but becomes less negative as wealth approaches the threshold.
  - Overall marginal effects relatively flat and not statistically significant in much of observed wealth range.
- Random effects model:
  - Wealth effect clearer and stronger; non-linear specification shows much lower threshold, often making a simple linear term sufficient: countries with larger household wealth experience more rapid consumption growth.
  - The coefficient on V falls in half but remains highly significant.
- Data limitation: household housing wealth data are sparse (subsample data for 1995 and 2000 only).

### VII. Does financial (stock) volatility matter? (Section IV)
- Fixed-effects regression (Column 1, Table 4):
  - Stock market volatility (standard deviation of daily changes in stock market index over a year, 1984–2007) appears to have no influence on consumption growth.
  - Alternative specifications excluding V or excluding square of wealth: stock market volatility never approaches significance in fixed effects.
  - Possible reason: short-term market volatility effects are short-lived and hard to capture in annual data.
- Random effects regression:
  - Stock market volatility differentiates countries in the long run: more volatile countries exhibit lower consumption growth.
- Long-run equity premium volatility:
  - Barro and Ursúa (2008) long-run volatility of the equity premium soaks up the influence of contemporary stock market volatility and is statistically significant: greater long-run volatility is associated with lower consumption growth.
  - With Barro and Ursúa index added, marginal influence of V is reduced although V remains highly significant.
- Supporting literature: Sandri (2009); Weitzman (2008); Barro and Ursúa (2008).

### VIII. Demographics and working-age population (Section V)
- Two demographic variables: share of young population (0–14) and share of old population (65+).
  - Presumption: higher shares of dependents reduce working population share, lowering standards of living and consumption growth rates.
- Fixed effects (short-term within-country annual variation):
  - Increases in both young and old population shares reduce consumption growth in the short run.
  - Young dependents act as more of a short-term brake (possibly reflecting withdrawal of mothers from the workforce).
- Random effects (longer-term cross-sectional):
  - Longer-term effects of a younger population are more benign, although a larger share of the young continues to restrict consumption growth.
  - Share of older population has a smaller longer-run effect relative to its short-run impact, but the longer-run effect of the old is greater than the dampening by the young.
- Interaction evidence:
  - Where an older population benefits from social security arrangements, the buffeting due to higher GDP volatility is somewhat muted (Volatility * population 65+ coefficient: 1.60*** in specified model).

### IX. Interpreting results, projections, and implications for recovery and global imbalances
- Projection methodology:
  - Use Column 4 of Table 5 to identify key associates of consumption growth for Germany, Japan, and the U.S.; project consumption growth using WEO projections for RHS variables and assumptions about V.
- Assumptions for V in projections (Figure 6):
  - V = twice its 2007 value in 2008.
  - V = 2¼ times its 2007 value in 2009.
  - V = 1¾ times its 2007 value in 2010.
- Model performance:
  - Model tracks consumption growth closely for Germany and the U.S.; predicts anemic consumption growth for Japan (actual somewhat higher than predicted in recent years).
  - Germany and Japan weighed down by relatively high GDP volatility, high long-term volatility of the equity premium, and aging population; Germany also pulled down by low net financial wealth.
  - U.S. benefits from consumption growth between 2½ and 3¼ percent a year since 1992, relatively low GDP volatility (advantage eroding), low long-run equity premium volatility, low share of the old, and larger accumulated financial wealth enabling higher consumption.
- Specific projection highlighted:
  - U.S. consumption growth projected at 1¾ percent in 2010 after a shrinking in 2009.
- Implications:
  - Precautionary motives and high economic uncertainty are significant dampeners on consumption growth; G-7 consumption unlikely to be the engine for global growth recovery.
  - Global imbalances narrower in 2009 primarily because the U.S. experienced an earlier sharp rise in unemployment and larger losses in financial wealth; this narrowing could reverse quickly if U.S. recovery is earlier than elsewhere.
  - Longer-term forces—differences in fear factor (equity premium volatility) and demographics—are persistent and not easily reversed.
  - Policy implication: placing full burden of adjustment on savings and consumption is unrealistic; changes in investment behavior will also be required to address global imbalances and support sustained recovery.

### X. Key empirical coefficients and statistics (selected exact estimates)
- Table 1. Precautionary Behavior (Dependent Variable: Consumption Growth)
  - Lagged consumption (share of disposable income): -48.19***; -97.64***; -93.19***; -70.29***
  - Real disposable income growth: 0.29***; 0.17***; 0.12***
  - Change in unemployment rate: -0.82***; -0.75***
  - GDP volatility: -0.88***
  - Observations: 330; 321; 321; 290
  - R-squared: 0.361; 0.473; 0.593; 0.610
  - Number of countries: 18; 18; 18; 16
- Table 2. G-7 vs Non-G-7 (1-Year and 3-Year Horizons)
  - 1-Year Horizon (G-7, Non-G-7):
    - Lagged consumption: -70.02***; -34.81
    - Real disposable income growth: 0.15**; 0.07
    - Change in unemployment rate: -0.77***; -0.74***
    - GDP volatility: -0.98***; -0.79*
    - Observations: 160; 130
    - R-squared: 0.664; 0.635
    - Number of countries: 7; 9
  - 3-Year Horizon (G-7, Non-G-7):
    - Lagged consumption: -55.97**; 0.59
    - Real disposable income growth: 0.11; 0.12
    - Change in unemployment rate: -1.08***; -0.83**
    - GDP volatility: -1.05***; -0.77
    - Observations: 55; 47
    - R-squared: 0.757; 0.684
    - Number of countries: 7; 9
- Table 3. Wealth Effects (Fixed and Random Effects)
  - Lagged consumption: -58.19**; -49.28*; -55.48**; -15.22; -14.35
  - Real disposable income growth: 0.13*; 0.17**; 0.16**; 0.21***; 0.20***
  - Change in unemployment rate: -0.73***; -0.83***; -0.82***; -0.66***; -0.62***
  - GDP volatility: -1.09***; -1.01***; -0.99***; -0.49***; -0.51***
  - Lagged household wealth (share of disposable income): 0.17; -2.81***; -2.60**; -0.56; 0.63***
  - Lagged household wealth squared: 0.49***; 0.47***; 0.24*
  - Observations: 154
  - R-squared: 0.669; 0.691; 0.688; 0.605; 0.596
  - Number of countries: 7
- Table 4. Financial Volatility (G-7)
  - Lagged consumption: -56.62**; -21.99; -48.24**
  - Real disposable income growth: 0.17**; 0.19***; 0.13*
  - Change in unemployment rate: -0.86***; -0.60***; -0.64***
  - GDP volatility: -0.96***; -0.54***; -0.40***
  - Lagged household wealth (share of disposable income): -2.92***; 0.61***; 0.59***
  - Lagged household wealth squared: 0.52***
  - Standard deviation of daily MS Stock Market Index growth: 0.57; -0.66*; -0.10
  - Equity premium volatility: -7.94***
  - Observations: 154
  - R-squared: 0.695; 0.593; 0.622
  - Number of countries: 7
- Table 5. Working Age Population
  - Lagged consumption: -91.81***; -93.83***; 2.05; -16.22; -36.00*
  - Real disposable income growth: 0.18**; 0.12*; 0.09; 0.10; 0.15**
  - Change in unemployment rate: -0.87***; -0.74***; -0.80***; -0.75***; -0.73***
  - GDP volatility: -1.00***; -1.08***; -0.69***; -0.69***; -3.12***
  - Lagged household wealth (share of disposable income): -3.31***; 0.45; 0.48***; 0.51***; 0.39**
  - Lagged household wealth squared: 0.63***
  - Population 65+ years (percent): -0.42***; -0.35**; -0.32***; -0.34***; -0.68***
  - Population 0-14 years (percent): -0.70***; -0.58***; -0.09; -0.20**; -0.36***
  - Equity premium volatility: -5.77**; -12.44***
  - Volatility * population 65+: 1.60***
  - Observations: 154
  - R-squared: 0.742; 0.707; 0.721; 0.730; 0.746
  - Number of countries: 7

### XI. Appendix robustness (selected highlights from Appendix Table A and B)
- Appendix Table A (full sample drop-one-country regressions) — Lagged consumption across columns (1)–(17): coefficients range include -70.29***, -77.20***, -69.07***, -48.80**, -72.84***, -48.88***, -71.28***, -70.40***, -70.13***, -73.53***, -70.31***.
- Appendix Table A — Real disposable income growth generally ~0.11*** to 0.14*** across exclusions.
- Appendix Table A — Change in unemployment rate consistently ~ -0.71*** to -0.78*** across exclusions.
- Appendix Table A — GDP volatility consistently negative and significant, e.g., -0.88***, -0.75***, -1.01***, -0.97*** across exclusions.
- Appendix Table B (G-7 drop-one-country) — Lagged consumption coefficients include -70.02***, -74.08***, -52.14 (not significant), -34.37 (not significant), -83.49***, -89.19***, -70.13***.
- Appendix Table B — Real disposable income growth varies across exclusions: 0.15**; 0.11*; 0.13**; 0.13**; 0.18***; 0.21***; 0.24***; 0.10.
- Appendix Table B — GDP volatility coefficients range from -0.82*** to -1.28*** across exclusions.

*Source: _wp1011 - References (content and tables reproduced from the supplied PDF).*

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

### _wp1011 - References .............................................................................................................

### I. Introduction — scope and motivation
- Private consumption constitutes the largest share of GDP: in the United States, private consumption has, in recent years, been 70 percent of GDP; in Germany, private consumption constitutes almost 60 percent of GDP.
- Prior to the crisis there were concerns about widening global imbalances (U.S. current account deficit financed by German, Japanese, and Chinese surpluses). The crisis instead reflected U.S. housing bubble, subprime lending, financial market opacity, and bank balance-sheet discounting.
- This paper uses an econometric specification of consumption growth for G-7 economies (Canada, France, Germany, Italy, Japan, the U.K, and the U.S.) to project consumption growth and to assess effects on global imbalances.
- Two empirical observations guiding the model:
  - Savings rates vary over time within countries.
  - Large cross-country differences in savings persist (e.g., U.S. savings lower than Germany’s for nearly each of the last 40 years).
- Innovation: inclusion of a country’s economic volatility (standard deviation of forecast error of GDP growth) as a determinant of consumption growth.

### II. Data, sample, and methodology
- Sample start: in principle 1984 but starting dates differ by country due to missing data; analysis ends in 2007.
- Panel assumption: all variables influence consumption growth in the same way across countries; cross-country differences arise from differences in explanatory variables.
- Additional countries for robustness in preliminary section: 9 other OECD countries (Australia, Austria, Belgium, Finland, Ireland, the Netherlands, Norway, Spain, and Sweden).
- Data frequency: annual data used (quarterly disposable income and household wealth judged unreliable); panel techniques applied.
- Observations per country: between 16 and 24 observations per country.
- Volatility measure Vt: standard deviation of the forecast error of real GDP growth estimated via a “factor-structural” VAR (FSVAR) following Stock and Watson (2005); estimates extended to 2007 by Carare and Mody (forthcoming).
- Concerns and econometric choices:
  - Mean reversion term c_{t-1} (lagged consumption as share of disposable income) used as error-correction.
  - Possible correlation of c_{t-1} with error term noted but mitigated by time-series length.
  - Potential reverse causality (consumption → disposable income and unemployment) acknowledged; lagged RHS variables examined as instruments but no strong instruments adopted; results presented as descriptive rather than causal.
  - Fixed effects: country and year fixed effects used in reported regressions.

### III. Basic estimating equation
- Reported specification (as presented in the source):
  tttttt VUERyccεααααα + +Δ+Δ++=Δ
  -432110
  lnln.
- Dependent variable: Δlnc_t, the log change in real consumption.
- Explanatory variables:
  - c_{t-1}: lagged consumption as share of disposable income (error correction / mean reversion).
  - Δlny_t: log change in real disposable income.
  - ΔUER_t: change in unemployment rate from the previous year (precautionary-risk proxy).
  - V_t: GDP volatility (standard deviation of forecast error of real GDP growth).

### IV. Main empirical findings
- Error-correction and income effects:
  - The error correction term (c_{t-1}) generally has a negative and significant sign.
  - A rise in real disposable income is associated with increased consumption. The coefficient is in the range of 0.2, suggesting a marginal propensity to consume of 20 percent (consistent with Carroll (2001)).
- Precautionary behavior — unemployment and volatility:
  - ΔUER (change in unemployment rate) is always solidly statistically significant: a rise in unemployment is associated with a clear reduction in consumption growth.
  - V (GDP volatility) is robustly associated with a decline in consumption growth across specifications.
  - Magnitudes reported:
    - A rise in the unemployment rate by one percentage point reduces consumption growth by ¾ of a percentage point.
    - A one percentage point increase in the standard deviation of the forecast error is associated with about 0.9 percentage point decrease in consumption growth.
- Differences across country groups and averaging:
  - Table 2: G-7 versus non-G-7 results and results using 3-year averages to reduce short-term noise.
  - Error-correction effect weaker in non-G-7 economies and weaker with 3-year averages.
  - Income effect somewhat weaker in non-G-7 economies and in 3-year averages.
  - Effect of ΔUER rises in 3-year averages (possibility of reverse causality noted).
  - Influence of V somewhat smaller in non-G-7 economies but similar across annual and 3-year averaged data.
- Wealth and long-run factors (summary of findings reported later in paper):
  - Households set target levels of wealth as buffers; annual consumption variation from this mechanism is usually limited because year-to-year wealth changes are small.
  - Large losses of financial wealth (as in the crisis) have larger effects on consumption and can have persistent cross-country effects.
  - Long-term tendency for financial instability (e.g., volatile equity premium) is associated with lower average consumption growth.
- Demographics:
  - Working-age population supporting more dependents (young and old) is associated with lower consumption growth; dependency rates contribute to long-term cross-country differences.

### V. Interpretation and implications
- Precautionary motives are important: households emphasize precautionary savings and base “permanent income” on expected very near-term income.
- Short-run outlook:
  - High economic uncertainty is a significant dampener on consumption growth.
  - Because uncertainty is broad-based, G-7 consumption is unlikely to be the engine for global growth recovery.
- Global imbalances:
  - U.S. experienced a sharper rise in unemployment and greater loss of financial wealth, contributing to a relative rise in the U.S. savings rate and helping narrow global imbalances.
  - Caution: an earlier U.S. recovery could make this narrowing short-lived; long-term differences in volatility and demographics have been important sources of imbalances and may reassert themselves.
- Robustness:
  - Results robust to full country sample, G-7 sample, dropping countries one at a time, and to use of 3-year averages (reported in appendix Tables A and B and main Tables 1–5 and Figures 1–6).

*Source: _wp1011 - References .............................................................................................................*

### Appendix Tables A and B report the basic regression when we drop one country at a time.

### _wp1011 - Appendix Tables A and B report the basic regression when we drop one country at a time.

### Robustness checks: sample and time sensitivity
- Appendix Tables A and B report the basic regression when we drop one country at a time.
  - Table A: full sample of 16 countries.
  - Table B: drop-one-country from the G-7 economies.
  - The size of the coefficients changes somewhat, but the results are robust.
- Dropping one year at a time confirms that the results are not driven by any particular year.
- Results with three-year averages are not reported here because with 54 observations adding extra variables reduces degrees of freedom; however, the reported results remain very similar.

### Wealth effects (Section III)
- Real interest rate does not appear to matter (consistent with Campbell and Mankiw (1989) and Carroll and Slacalek (2009)).
- Wealth variable w_{t-1} follows Gruber and Kamin (2008): lagged stock of households’ net financial wealth relative to disposable income.
  - In fixed-effects regression:
    - Wealth has expected positive sign but is not statistically significant.
    - Non-linearity tested via square of wealth: wealth term negative, square term positive and significant.
    - Marginal effect: above a ratio of wealth-to-disposable income of about 2½ (close to the median), a rise in wealth is associated with a rise in consumption growth; below that cut-off the marginal effect is negative but becomes less negative as wealth approaches the threshold.
    - Overall marginal effects are relatively flat and not statistically significant in much of the observed wealth range.
  - In random effects model:
    - Wealth effect is clearer and stronger.
    - Non-linear specification shows a much lower threshold, encompassing virtually the entire sample (Figure 3b); a simple linear term suffices: countries with larger household wealth experience more rapid consumption growth.
    - The coefficient on V falls in half but remains highly significant, consistent with possible short-term overshooting in response to volatility.
- Note on data limitations: household housing wealth data are sparse (Slacalek (2006), ECB (2003) provide subsample data for 1995 and 2000 only).

### Does financial volatility matter? (Section IV)
- Question: does market (stock) volatility affect consumption behavior, and is it complementary or subsuming the relationship between consumption growth and GDP volatility?
- Fixed-effects regression (Column 1, Table 4):
  - Volatility of domestic stock market (standard deviation of daily changes in stock market index over a year, 1984–2007) appears to have no influence on consumption growth.
  - Tried alternative specifications (excluding V, excluding square of wealth) — stock market volatility never approaches significance in fixed effects.
  - Possible reasons: short-term variations carry little information; effects of changes in market volatility are short-lived and hard to capture in annual data.
- Random effects regression:
  - Stock market volatility shows more action: long-term differences in average volatility differentiate countries, with more volatile countries exhibiting lower consumption growth.
- Interpretation and supporting evidence:
  - Sandri (2009): higher stock market volatility needs to persist for a few years to have substantial impact because it depresses the portfolio interest rate.
  - Weitzman (2008) and Barro and Ursúa (2008): long-run perceptions of tail risks and catastrophic event probabilities raise equity premium and long-run stock market volatility and are associated with lower real returns on risk-free assets in some countries.
  - Barro and Ursúa (2008) long-run volatility of the equity premium (introduced in Column 3, Table 4) soaks up the influence of contemporary stock market volatility and is statistically significant: greater long-run volatility is associated with lower consumption growth.
  - With the Barro and Ursúa index added, the marginal influence of V is reduced although V remains highly significant.

### Working-age population and demographics (Section V)
- Two demographic variables introduced: share of young population (ages 0–14) and share of old population (65+).
  - Presumption: higher shares of dependents reduce the working population share, lowering standards of living and consumption growth rates.
- Fixed effects (short-term, within-country annual variation; Columns 1–2, Table 5):
  - Increases in both young and old population shares reduce consumption growth in the short run.
  - Young dependents act as more of a short-term brake (possibly reflecting withdrawal of mothers from the workforce).
- Random effects (longer-term, cross-sectional; Columns 3–4, Table 5):
  - Longer-term effects of a younger population are more benign, although a larger share of the young continues to restrict consumption growth.
  - Share of older population has a smaller longer-run effect relative to its short-run impact, but the longer-run effect of the old is greater than the dampening by the young.
- Interactions and further exploration:
  - Evidence (Column 5, Table 5) is suggestive that where an older population benefits from social security arrangements, the buffeting due to higher GDP volatility is somewhat muted; requires more careful country-level differentiation to draw definitive conclusions.
- Conclusion: demographic factors are important but do not negate prior conclusions on volatility, wealth, and consumption links.

### Interpreting results, projections, and implications for recovery and global imbalances (Sections VI and projections)
- Summary regression: use Column 4 of Table 5 to identify key associates of consumption growth for Germany, Japan, and the U.S.; then project consumption growth using WEO projections for RHS variables and assumptions about V.
- Assumptions for V in projections (Figure 6):
  - V = twice its 2007 value in 2008.
  - V = 2¼ times its 2007 value in 2009.
  - V = 1¾ times its 2007 value in 2010.
- Model performance:
  - Left-hand panel of Figure 5: model tracks consumption growth closely for Germany and the U.S.; predicts anemic consumption growth for Japan (actual somewhat higher than predicted in recent years).
  - Right-hand panel of Figure 5: decomposition of consumption growth differentials relative to G-7 average using country differences multiplied by coefficients from Column 4, Table 5.
    - Germany and Japan weighed down by:
      - Relatively high GDP volatility (Germany recently rising).
      - High long-term volatility of the equity premium ("fear" of financial tail risks).
      - Aging population (large share of the old outweighs advantage of a small share of the young).
    - Germany also pulled down by low net financial wealth.
    - U.S. benefits from:
      - Consumption growth between 2½ and 3¼ percent a year since 1992.
      - Relatively low GDP volatility (advantage eroding toward end of sample).
      - Low long-run equity premium volatility.
      - Low share of the old.
      - Larger accumulated financial wealth enabling higher consumption.
- Medium-term projections and fit with WEO (Figure 6):
  - Reasonable correspondence between model projections and WEO for 2008–2010 using WEO estimates for RHS variables and assumed V path; fit particularly good for the U.S. and Italy.
  - Model predicts lower consumption growth for Germany and Japan than WEO in 2009–2010.
  - Specific projection highlighted: U.S. consumption growth projected at 1¾ percent in 2010 after a shrinking in 2009.
  - Immediate weakness driven by rise in unemployment, expectation that GDP volatility remains high in 2010, and destruction of financial wealth.
- Implications for recovery and global imbalances:
  - Consumption is unlikely to lead the recovery if it bears a stable relation to income and share of consumption is falling (Keynes (1937) argument).
  - Investor reluctance and deep uncertainty dampen investment, reinforcing weak recovery.
  - Global imbalances narrower in 2009 primarily because:
    - U.S. experienced an earlier sharp rise in unemployment and larger losses in financial wealth (more U.S. households held risky assets).
    - Other countries experienced slower unemployment rises due to labor market structures and public support.
  - If projections hold, unwinding of global imbalances could reverse quickly in 2010 as dynamics change.
  - Longer-term forces—differences in fear factor and demographics—are persistent and not easily reversed.
    - High savings in countries like Germany and Japan have not necessarily produced large wealth accumulation (households invest in low-risk, low-return assets), keeping consumption growth low.
    - Low domestic consumption growth implies exports-driven growth, increasing GDP volatility—creating a reinforcing low-consumption, high-volatility regime.
- Policy implication:
  - It may be unrealistic to place the full burden of adjustment on savings and consumption; changes in investment behavior will also be required to address global imbalances and support a sustained recovery.

*Source: _wp1011 - Appendix Tables A and B report the basic regression when we drop one country at a time.*

### REFERENCES

### _wp1011 - REFERENCES

### References
- Akerlof, George A., and Robert J. Shiller, 2009, “Animal Spirits: How Human Psychology Drives the Economy, and Why it Matters,” Princeton University Press.
- Barro, Robert J., and José F. Ursúa, 2008, “Macroeconomic Crises since 1870,” NBER Working Paper No. 13940.
- Bloom, Nicholas, 2009, “The Impact of Uncertainty Shocks,” Econometrica, Vol. 77, No. 3, pp. 623-685.
- Campbell, John Y., and N. Gregory Mankiw, 1989, “Consumption, Income, and Interest Rates: Reinterpreting the Time Series Evidence,” NBER Macroeconomics Annual, Vol. 4, pp. 185-216.
- Carare, Alina, and Ashoka Mody, forthcoming, “International Spillovers of Domestic Shocks: Will they Counteract the “Great Moderation”? Washington D.C.
- Carroll, Christopher D., 1992, “The Buffer-Stock Theory of Saving: Some Macroeconomic Evidence,” Brookings Papers on Economic Activity, Vol. 2, pp. 62-156.
- Carroll, Christopher D., 2001, “A Theory of the Consumption Function, with and without Liquidity Constraints,” Journal of Economic Perspectives, Vol. 15, No. 2, pp. 23-45.
- Carroll, Christopher D., and Jiri Slacalek, 2009, “The American Consumer: Reforming, or just Resting?” http://econ.jhu.edu/people/ccarroll/papers/ReformingOrResting.pdf.
- European Central Bank, 2003, “Structural Factors in the EU Housing Markets,” report, European Central Bank, available at http://www.ecb.eu/pub/pdf/other/euhousingmarketsen.pdf.
- Feldstein, Martin, 2008, “Resolving the Global Imbalance: The Dollar and the U.S. Saving Rate,” Journal of Economic Perspectives, Vol. 22 No. 3, pp. 113-125.
- Friedman, Milton, 1963, “Windfall, the ‘Horizon,’ and Related Concepts in the Permanent-Income Hypothesis,” in: Measurement in Economics, Stanford University Press.
- Glick, Reuven, and Kevin Lansing, 2009, “U.S. Household Deleveraging and Future Consumption Growth,” FRBSF ECONOMIC LETTER, Number 2009-16, May 15.
- Gruber, Joseph H., and Steven B. Kamin, 2008, “Do Differences in Financial Development Explain the Global Pattern of Current Account Imbalances?” FRB International Finance Discussion Paper, No. 923.
- Keynes, John Maynard, 1937, “The General Theory of Employment,” The Quarterly Journal of Economics, Vol. 51, No. 2, pp. 209-223.
- Sandri, Damiano, 2009, “The Impact of Uncertainty Shocks on Consumption: the 2008 U.S. Crisis,” unpublished manuscript, International Monetary Fund.
- Slacalek, Jiri, 2006, “What Drives Personal Consumption? The Role of Housing and Financial Wealth,” mimeo, available at http://www.slacalek.com/research/sla06whatDrivesC/sla06whatDrivesC.pdf.
- Stock, James H., and Mark W. Watson, 2005, “Understanding Changes in International Business Cycle Dynamics,” Journal of the European Economic Association, Vol. 3, No. 5, pp. 968-1006.
- Summers, Lawrence, and Christopher D. Carroll, 1987, “Why is U.S. National Savings so Low?” Brookings Papers on Economic Activity, Vol. 1987 No. 2, pp. 604-642.
- Weitzman, Martin L., 2008, “Subjective Expectations and Asset Return Puzzles,” American Economic Review, Vol. 97 No. 4, pp. 1102-1130.

### Data Appendix — Country Coverage and Periods
- G-7 Countries:
  - Canada 1984-2007
  - France 1984-2007
  - Germany 1992-2007
  - Italy 1984-2007
  - Japan 1984-2007
  - U.K. 1984-2007
  - U.S. 1984-2007
- Non-G-7 Countries:
  - Austria 1996-2006
  - Belgium 1996-2006
  - Netherlands 1984-2006
  - Norway 1993-2007
  - Sweden 1994-2006
  - Finland 1984-2006
  - Ireland 2003-2006
  - Spain 2001-2006
  - Australia 1984-2007

### Data Definitions and Sources
- Real consumption growth: Annual log change in volume index for private final consumption. OECD’s Detailed Tables for Annual National Accounts.
- Real disposable income growth: Annual log change in net disposable income of households and nonprofit institutions. OECD’s Simplified Non-Financial National Accounts. Earlier data for 1985-1995 for Japan and for 1985-1987 for the U.K. from Kamin and Gruber (2007). Deflated by consumer price index from the IMF’s International Finance Statistics.
- Household wealth: Financial net worth of households and nonprofit institutions. OECD’s Financial Annual Accounts. Earlier data for France and Italy from 1985-1995 and for the U.K. from 1985-1987 from Kamin and Gruber (2007).
- Real interest rates: Long-term interest rates as reported in the OECD’s Monthly Economic Indicators.
- Unemployment rate: OECD’s Monthly Economic Indicators.
- Share of working age population (15-64 years), young (0-14 years) and old (65+ years), in percent of total population: World Bank’s World Development Indicators.
- Gross private national savings (percent of GDP): Gross national savings excluding gross public savings. IMF’s World Economic Outlook.
- GDP Volatility: Standard Deviation of forecast error of a cross-country VAR regression of real GDP growth, as in Stock and Watson (2005), as extended by Carare and Mody (forthcoming).
- Stock market volatility: Standard deviation of daily changes in Morgan Stanley stock market index over the period of one year. Global Insight database.
- Equity premium volatility: Standard deviation of equity premium over T-bill returns. Table 5, Barro and Ursúa (2008).

### Key Empirical Findings (Tables summary — selected coefficients and statistics preserved exactly)
- Table 1. Precautionary Behavior (Dependent Variable: Consumption Growth)
  - Lagged consumption (share of disposable income): -48.19***; -97.64***; -93.19***; -70.29*** (coefficients across specifications).
  - Real disposable income growth: 0.29***; 0.17***; 0.12***.
  - Change in unemployment rate: -0.82***; -0.75***.
  - GDP volatility: -0.88***.
  - Observations: 330; 321; 321; 290.
  - R-squared: 0.361; 0.473; 0.593; 0.610.
  - Number of countries: 18; 18; 18; 16.
  - Significance notation: Robust t-statistics in brackets. *** p<0.01, ** p<0.05, * p<0.1.

- Table 2. Consumption Growth in G-7 and Non-G-7 Countries: 1-Year and 3-Year Horizons
  - 1-Year Horizon (G-7, Non-G-7):
    - Lagged consumption: -70.02***; -34.81.
    - Real disposable income growth: 0.15**; 0.07.
    - Change in unemployment rate: -0.77***; -0.74***.
    - GDP volatility: -0.98***; -0.79*.
    - Observations: 160; 130.
    - R-squared: 0.664; 0.635.
    - Number of countries: 7; 9.
  - 3-Year Horizon (G-7, Non-G-7):
    - Lagged consumption: -55.97**; 0.59.
    - Real disposable income growth: 0.11; 0.12.
    - Change in unemployment rate: -1.08***; -0.83**.
    - GDP volatility: -1.05***; -0.77.
    - Observations: 55; 47.
    - R-squared: 0.757; 0.684.
    - Number of countries: 7; 9.

- Table 3. Wealth Effects (specifications: Fixed Effects and Random Effects)
  - Lagged consumption: -58.19**; -49.28*; -55.48**; -15.22; -14.35.
  - Real disposable income growth: 0.13*; 0.17**; 0.16**; 0.21***; 0.20***.
  - Change in unemployment rate: -0.73***; -0.83***; -0.82***; -0.66***; -0.62***.
  - GDP volatility: -1.09***; -1.01***; -0.99***; -0.49***; -0.51***.
  - Lagged household wealth (share of disposable income): 0.17; -2.81***; -2.60**; -0.56; 0.63***.
  - Lagged household wealth squared: 0.49***; 0.47***; 0.24* (where reported).
  - Observations: 154 across specifications.
  - R-squared: 0.669; 0.691; 0.688; 0.605; 0.596.
  - Number of countries: 7.

- Table 4. Does Financial Volatility Matter?—G-7 Economies
  - Lagged consumption: -56.62**; -21.99; -48.24**.
  - Real disposable income growth: 0.17**; 0.19***; 0.13*.
  - Change in unemployment rate: -0.86***; -0.60***; -0.64***.
  - GDP volatility: -0.96***; -0.54***; -0.40***.
  - Lagged household wealth (share of disposable income): -2.92***; 0.61***; 0.59***.
  - Lagged household wealth squared: 0.52*** (in one specification).
  - Standard deviation of daily MS Stock Market Index growth: 0.57; -0.66*; -0.10.
  - Equity premium volatility: -7.94*** (in one specification).
  - Observations: 154 across specifications.
  - R-squared: 0.695; 0.593; 0.622.
  - Number of countries: 7.

- Table 5. Working Age Population
  - Lagged consumption: -91.81***; -93.83***; 2.05; -16.22; -36.00*.
  - Real disposable income growth: 0.18**; 0.12*; 0.09; 0.10; 0.15**.
  - Change in unemployment rate: -0.87***; -0.74***; -0.80***; -0.75***; -0.73***.
  - GDP volatility: -1.00***; -1.08***; -0.69***; -0.69***; -3.12***.
  - Lagged household wealth (share of disposable income): -3.31***; 0.45; 0.48***; 0.51***; 0.39**.
  - Lagged household wealth squared: 0.63*** (reported in one specification).
  - Population 65+ years (in percent of total population): -0.42***; -0.35**; -0.32***; -0.34***; -0.68***.
  - Population 0-14 years (in percent of total population): -0.70***; -0.58***; -0.09; -0.20**; -0.36***.
  - Equity premium volatility: -5.77**; -12.44*** (in specified models).
  - Volatility * population 65+: 1.60*** (in specified model).
  - Observations: 154 across specifications.
  - R-squared: 0.742; 0.707; 0.721; 0.730; 0.746.
  - Number of countries: 7.

*Source: _wp1011 - REFERENCES (data, tables, and citations reproduced from the supplied PDF content).*

### Appendix Tables

### Appendix Tables

### Appendix Table A. Robustness to Full Country Sample (columns 1–9)
- Lagged consumption (share of disposable income):
  - (1) All Included: -70.29***
  - (2) Excl. Australia: -77.20***
  - (3) Excl. Austria: -69.07***
  - (4) Excl. Belgium: -70.04***
  - (5) Excl. Canada: -73.25***
  - (6) Excl. Finland: -90.62***
  - (7) Excl. France: -69.22***
  - (8) Excl. Germany: -74.37***
  - (9) Excl. Ireland: -70.11***
  - Robust t-statistics (bracketed) example: [-4.24], [-4.74], [-4.14], [-4.19], [-4.33], [-5.22], [-3.35], [-4.49], [-4.24]

- Real disposable income growth:
  - (1) 0.12***
  - (2) 0.13***
  - (3) 0.12***
  - (4) 0.12***
  - (5) 0.11***
  - (6) 0.14***
  - (7) 0.12***
  - (8) 0.11***
  - (9) 0.12***
  - Robust t-statistics (bracketed) example: [3.28], [3.47], [3.16], [3.13], [2.87], [3.63], [3.13], [3.10], [3.23]

- Change in unemployment rate:
  - (1) -0.75***
  - (2) -0.78***
  - (3) -0.75***
  - (4) -0.76***
  - (5) -0.77***
  - (6) -0.71***
  - (7) -0.74***
  - (8) -0.76***
  - (9) -0.76***
  - Robust t-statistics (bracketed) example: [-7.20], [-7.33], [-7.14], [-7.19], [-7.27], [-5.80], [-6.54], [-7.29], [-7.35]

- GDP volatility:
  - (1) -0.88***
  - (2) -0.75***
  - (3) -0.88***
  - (4) -0.88***
  - (5) -0.83***
  - (6) -0.96***
  - (7) -0.89***
  - (8) -0.78***
  - (9) -0.86***
  - Robust t-statistics (bracketed) example: [-4.59], [-3.71], [-4.60], [-4.55], [-3.85], [-4.86], [-4.23], [-3.93], [-4.49]

- Constant:
  - (1) 10.11***
  - (2) 10.54***
  - (3) 10.10***
  - (4) 10.15***
  - (5) 10.07***
  - (6) 11.66***
  - (7) 10.23***
  - (8) 10.29***
  - (9) 9.98***
  - Robust t-statistics (bracketed) example: [7.42], [8.11], [7.34], [7.41], [7.39], [8.63], [6.50], [7.59], [7.33]

- Observations:
  - (1) 290
  - (2) 266
  - (3) 279
  - (4) 279
  - (5) 266
  - (6) 267
  - (7) 266
  - (8) 274
  - (9) 286

- R-squared:
  - (1) 0.610
  - (2) 0.629
  - (3) 0.614
  - (4) 0.610
  - (5) 0.603
  - (6) 0.563
  - (7) 0.603
  - (8) 0.618
  - (9) 0.619

- Number of countries:
  - (1) 16
  - (2) 15
  - (3) 15
  - (4) 15
  - (5) 15
  - (6) 15
  - (7) 15
  - (8) 15
  - (9) 15

- Notes:
  - Robust t-statistics in brackets. *** p<0.01, ** p<0.05, * p<0.1. Coefficients for country and year dummies not reported.

### Appendix Table A. Robustness to Full Country Sample (concluded: columns 10–17)
- Lagged consumption (share of disposable income):
  - (10) Excl. Italy: -48.80**
  - (11) Excl. Japan: -72.84***
  - (12) Excl. Netherlands: -48.88***
  - (13) Excl. Norway: -71.28***
  - (14) Excl. Spain: -70.40***
  - (15) Excl. Sweden: -70.13***
  - (16) Excl. U.K.: -73.53***
  - (17) Excl. U.S.: -70.31***
  - Robust t-statistics (bracketed) example: [-2.49], [-4.21], [-3.00], [-4.23], [-4.25], [-4.13], [-4.09], [-4.16]

- Real disposable income growth:
  - (10) 0.13***
  - (11) 0.13***
  - (12) 0.11***
  - (13) 0.13***
  - (14) 0.12***
  - (15) 0.13***
  - (16) 0.14***
  - (17) 0.10***
  - Robust t-statistics (bracketed) example: [3.20], [3.25], [2.80], [2.94], [3.28], [3.40], [3.46], [2.76]

- Change in unemployment rate:
  - (10) -0.81***
  - (11) -0.70***
  - (12) -0.73***
  - (13) -0.73***
  - (14) -0.75***
  - (15) -0.76***
  - (16) -0.70***
  - (17) -0.75***
  - Robust t-statistics (bracketed) example: [-7.54], [-6.25], [-7.10], [-6.57], [-7.17], [-7.26], [-6.37], [-7.02]

- GDP volatility:
  - (10) -0.87***
  - (11) -0.92***
  - (12) -1.01***
  - (13) -0.88***
  - (14) -0.88***
  - (15) -0.88***
  - (16) -0.88***
  - (17) -0.97***
  - Robust t-statistics (bracketed) example: [-4.58], [-3.92], [-5.58], [-4.56], [-4.59], [-4.56], [-4.50], [-4.63]

- Constant:
  - (10) 8.26***
  - (11) 10.55***
  - (12) 8.72***
  - (13) 10.11***
  - (14) 10.12***
  - (15) 10.08***
  - (16) 10.41***
  - (17) 10.15***
  - Robust t-statistics (bracketed) example: [4.90], [6.73], [6.34], [7.23], [7.42], [7.21], [6.96], [7.47]

- Observations:
  - (10) 266
  - (11) 266
  - (12) 267
  - (13) 275
  - (14) 284
  - (15) 277
  - (16) 266
  - (17) 266

- R-squared:
  - (10) 0.627
  - (11) 0.604
  - (12) 0.644
  - (13) 0.622
  - (14) 0.610
  - (15) 0.616
  - (16) 0.610
  - (17) 0.619

- Number of countries:
  - (10)–(17): 15 (each column)

- Notes:
  - Robust t-statistics in brackets. *** p<0.01, ** p<0.05, * p<0.1. Coefficients for country and year dummies not reported.

### Appendix Table B. Robustness to G-7 Country Sample (columns 1–8)
- Lagged consumption (share of disposable income):
  - (1) All Included: -70.02***
  - (2) Excl. Canada: -74.08***
  - (3) Excl. France: -52.14
  - (4) Excl. Germany: -75.28***
  - (5) Excl. Italy: -34.37
  - (6) Excl. Japan: -83.49***
  - (7) Excl. U.K.: -89.19***
  - (8) Excl. U.S.: -70.13***
  - Robust t-statistics (bracketed) example: [-3.78], [-4.00], [-1.54], [-4.06], [-1.53], [-3.62], [-5.02], [-3.82]

- Real disposable income growth:
  - (1) 0.15**
  - (2) 0.11*
  - (3) 0.13**
  - (4) 0.13**
  - (5) 0.18***
  - (6) 0.21***
  - (7) 0.24***
  - (8) 0.10
  - Robust t-statistics (bracketed) example: [2.58], [1.69], [2.03], [2.23], [2.92], [2.95], [4.02], [1.56]

- Change in unemployment rate:
  - (1) -0.77***
  - (2) -0.84***
  - (3) -0.74***
  - (4) -0.81***
  - (5) -0.95***
  - (6) -0.66***
  - (7) -0.56***
  - (8) -0.79***
  - Robust t-statistics (bracketed) example: [-4.76], [-4.51], [-3.95], [-4.61], [-5.76], [-3.85], [-3.63], [-4.33]

- GDP volatility:
  - (1) -0.98***
  - (2) -0.82***
  - (3) -1.08***
  - (4) -0.85***
  - (5) -1.07***
  - (6) -1.28***
  - (7) -0.90***
  - (8) -1.11***
  - Robust t-statistics (bracketed) example: [-4.59], [-2.97], [-4.16], [-4.02], [-5.43], [-4.21], [-4.18], [-4.74]

- Constant:
  - (1) 10.08***
  - (2) 9.70***
  - (3) 9.29***
  - (4) 10.31***
  - (5) 7.35***
  - (6) 11.92***
  - (7) 11.34***
  - (8) 9.96***
  - Robust t-statistics (bracketed) example: [6.76], [6.72], [3.98], [6.98], [3.98], [5.84], [7.83], [6.73]

- Observations:
  - (1) 160
  - (2) 136
  - (3) 136
  - (4) 144
  - (5) 136
  - (6) 136
  - (7) 136
  - (8) 136

- R-squared:
  - (1) 0.664
  - (2) 0.656
  - (3) 0.656
  - (4) 0.678
  - (5) 0.732
  - (6) 0.678
  - (7) 0.674
  - (8) 0.687

- Number of countries:
  - (1) 7
  - (2)–(8) 6 (each column)

- Notes:
  - Robust t-statistics in brackets. *** p<0.01, ** p<0.05, * p<0.1. Coefficients for country and year dummies not reported.

### Figures (1–6): captions, sources, and notes
- Figure 1: Long-Term Differences in Savings Rates
  - Source: OECD, World Economic Outlook.
  - Note: Data on household savings rates only available from 1995.
  - Panels shown: Gross National Savings of the Private Sector (Percent of GDP) with years 1960–2004 and Net Household Savings Rate (Percent) with years 1995–2007 for Germany, US, Japan.

- Figure 2: Volatility: Standard Deviation of GDP Growth
  - Note: (Standard deviation of forecast error of GDP growth)
  - Time series axis shown from 1984 through 2005 for US, Germany, Japan.
  - Source: Carare and Mody (forthcoming).

- Figure 3: Effective Coefficient on Volatility in Panel Regression on Log Change in Real Consumption
  - Source: Authors' estimates.
  - Note: Dashed lines show 95 percent confidence interval of effective coefficient estimate.
  - Panels: a) Fixed effects regression; b) Random effects regression; x-axis labeled Lagged Household Wealth (share of disp.income).

- Figure 4: Long-Run Equity Premium, Real Treasury Bill Yields, and Stock Market Volatility
  - (Percent)
  - Source: Barro (2008); and authors' calculations.
  - Notes:
    - 1/ Defined as standard deviation of stock market rate of return, starting in 1870-1927 depending on the country and ending in 2006 (Table 5 of Barro and Ursúa, 2008).
    - 2/ Defined as difference between average rate of return on stock market index and treasury bills (Table 5 of Barro and Ursúa, 2008).
    - 3/ Defined as average bond rate of return, starting in 1870 depending on the country and ending in 2006 (Table 5 of Barro and Ursúa, 2008).
  - Panels include Stock Market Volatility, Equity Premium, and Real Return on Treasury Bills across listed countries.

- Figure 5: Random Effects Regression: Real Consumption Growth
  - (Percent)
  - Source: Authors' estimates.
  - Note: 1/ Combined contribution of population 65+ years and population 0-14 years.
  - Panels show Predicted and Actual Real Consumption Growth (Percent) and Contribution to Predicted Real Consumption Growth (Deviation from G7 average, in percent) for Germany, US, Japan with series labeled Population structure 1/, Lagged household wealth, Real disposable income growth, Change in unemployment rate, Equity premium volatility, GDP Volatility.

- Figure 6: Projections of Real Consumption Growth
  - (Percent)
  - Sources: World Economic Outlook October 2009 (WEO); and authors' estimates.
  - Panels display historical and projected series for US, Germany, Japan, UK, France, Italy, Canada with lines labeled WEO and Regression prediction.

*Source: _wp1011 - Appendix Tables (PDF chapter/section).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1011.pdf_