## _wp16251 - References

## Source details

**Canonical URL:** [_wp16251 - References](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16251.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16251.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16251.pdf.json)

---

### I. Introduction and Purpose
- Study period and sample: 62 countries between 1980 and 2013.
- Objective: Examine forecasting power of financial variables (private sector credit growth, stock prices, house prices, bond yields, deposit and lending rates) for macroeconomic variables (GDP growth, private consumption growth, private investment growth, CPI inflation).
- Modeling approach:
  - Simple, easily replicable specifications estimated country-by-country and in pooled cross-country panels.
  - Baseline specification uses only private sector credit growth and two policy controls (government consumption and policy rate).
  - Augmented specifications add stock prices, house prices, deposit and lending rates; for a subset add sovereign and corporate bond yields.
- Motivation for financial predictors:
  - Financial market imperfections link credit conditions to future macro outcomes.
  - Asset prices are forward-looking and incorporate information not yet reflected in macro outcomes.
  - Contemporaneous financial variables can assist nowcasting when macro data are lagged.

### II. Empirical Model and Data
- Forecasting equation (fixed-effects OLS FE):
  - Dependent variable: annualized growth rate of macro indicator Δh+ (annualized using scaling constant 400z).
  - Predictors: lagged dependent variable(s) (lags chosen by AIC), vector X of financial and policy controls, country fixed effects; Newey-West standard errors used for overlapping observations.
- Key macro variables: GDP growth, private consumption growth, private investment growth, CPI inflation.
- Financial predictors: private sector credit growth, stock prices, house prices, bank prime loan rate, deposit rate; extended sample adds sovereign bond yields and corporate bond yields for some advanced economies.
- Data frequency and coverage:
  - Quarterly data availability varies substantially by country and variable.
  - Average quarterly observations: advanced economies ~87, emerging economies ~41.
  - As of end-2015, IMF IFS has private sector credit annually for over 180 economies and quarterly for over 120 economies.
- Seasonal adjustment: X-12-ARIMA for non-seasonally adjusted raw data.
- Variables converted to real terms using country-specific GDP deflator.
- Sample winsorized at 2 percent for summary statistics.

### III. Main Empirical Findings
- General result: Incorporating financial variables significantly improves macroeconomic forecasting accuracy up to four quarters horizon.

A. Credit Growth: predictive power and magnitudes
- Credit growth is significantly associated with GDP growth, consumption growth, investment growth, and inflation (negative association) in many specifications.
- Baseline (nowcasting, simple model with credit growth):
  - A one standard deviation increase in credit growth (i.e. a 24 percentage point increase in annualized rate) is associated with a 1.79 percentage point increase in annualized GDP growth (nowcasting).
  - At one quarter ahead horizon: 1.15 percentage point increase in GDP growth.
  - At four quarters ahead horizon: 0.46 percentage point increase in GDP growth.
- Augmented model (with stock and house prices):
  - Nowcasting effects: 0.87 percentage point and 0.96 percentage point increases in GDP growth for a one standard deviation increase in credit growth (two augmented specs reported).
  - One quarter ahead: 0.65 percentage point and 0.43 percentage point increases.
  - Credit growth loses statistical significance at four quarters ahead in some augmented specifications.
- Investment growth sensitivity to credit:
  - Nowcasting: 6.9 percentage point increase in investment growth for one standard deviation increase in credit growth.
  - One quarter ahead: 4.2 percentage point increase.
  - Four quarters ahead: 2.0 percentage point increase.
- Subsample differences:
  - Coefficients on credit growth tend to be larger among emerging market and low-income countries than among advanced economies.
  - In the subsample including sovereign and corporate bond yields, credit growth sometimes loses statistical significance.

B. Equity and Housing Prices: predictive power and magnitudes
- Stock prices and house prices predict GDP growth, consumption growth, and investment growth in most specifications, conditional on credit growth.
- GDP growth effects (one standard deviation increase):
  - One quarter ahead: stock price index → 0.95 percentage point increase; house price index → 0.95 percentage point increase.
  - Four quarters ahead: stock price index → 0.73 percentage point; house price index → 0.85 percentage point.
  - Nowcasting: stock price index → 0.59 percentage point; house price index → 1.01 percentage point.
- Consumption:
  - House price increases have a stronger impact on consumption growth than stock price increases (housing wealth large share of household wealth).
- Investment:
  - Both stock and house prices have large and significant impacts on investment growth in most specifications.
- Inflation:
  - Neither stock prices nor house prices significantly predict inflation in most specifications.

C. Interest Rates, Bond Yields, and Other Financial Variables
- Deposit and lending rates:
  - Conditional on other financial variables and policy controls, deposit and lending rates generally do not significantly predict GDP growth.
  - Deposit rate tends to predict lower consumption growth and higher investment growth in some specifications.
  - Lending rate is strongly positively associated with future investment growth in several specifications.
- Corporate and sovereign bond yields:
  - Corporate bond yields show a negative association with GDP growth in some baseline specifications for advanced economies (e.g., corporate bond yield → roughly -0.7 and -0.75 percentage point reductions in GDP growth at one and four quarter horizons in baseline).
  - Effects of bond yields are not robust when other financial variables are included.
  - In the advanced-economies-with-bond-yields subsample, sovereign bond yield shows significance for inflation in some horizons.

D. Robustness and Group Results
- Findings robust across groups: advanced economies (AE), emerging markets (EM), and low-income countries (LIC).
- Low-income country results are based on a very small sample (LIC regression sample consists of only 3 countries in some specifications).
- Adjusted R-squared (selected ranges, all countries, four-quarter ahead horizon):
  - Investment and GDP growth: 0.25 to 0.3.
  - Consumption and inflation: 0.5 to 0.6.
- Goodness-of-fit comparisons:
  - Comparable goodness of fit for GDP growth in AE and EM.
  - Consumption growth fit much better in AE than EM (0.62 vs 0.18 reported).

### IV. Panel vs. Country-Specific Models: Forecast Performance
- In-sample vs out-of-sample patterns:
  - In-sample: individual country regressions have better fit than pooled panel regressions.
  - Out-of-sample: panel regressions outperform individual country regressions in forecasting accuracy.
- Out-of-sample forecasting exercise:
  - Full-sample fit period: 1980-2013.
  - Out-of-sample forecasts: models estimated on 1980-1999, four-quarter ahead forecasts computed for 2000-2007 (excluding crisis period).
  - Forecast errors measured by root mean squared errors (RMSE).
  - Three models evaluated:
    - AR model with lags chosen by AIC (up to 7 lags).
    - Financial model augmenting AR with credit growth and two policy controls (policy rate and government consumption).
    - Expanded financial model augmenting AR with credit growth, house price growth, stock price growth, and two policy controls.
- Advanced economies (selected results):
  - In-sample: panel model outperforms individual model in 4 percent of cases (Model 1) and 0 percent (Model 2) for the set reported.
  - Out-of-sample: panel model outperforms individual model in 65 percent of cases for Model 1 and 83 percent for Model 2 (83*** reported).
  - Large RMSE reductions in some countries (example statements: Belgium and Japan RMSE reductions by factor of 3 in some cases).
- Emerging markets:
  - In-sample: panel model outperforms individual in 6 percent (Model 1) and 0 percent (Model 2) of cases reported.
  - Out-of-sample: panel model outperforms individual in 100 percent of reported emerging economies for both Model 1 and Model 2 (100*** for both).
  - RMSE reductions in out-of-sample can be substantial (e.g., RMSE divided by about 2 for Turkey and Brazil; by about 4 for Colombia and Peru).
- Interpretation: pooled international information (common coefficients) improves out-of-sample forecast performance despite worse in-sample fit.

### V. Comparison with IMF WEO Forecasts
- Real-time out-of-sample exercise:
  - Four-quarter ahead quarterly GDP growth forecasts for each country from 2004 Q1 to 2013 Q1 (19 quarters).
  - Rolling 20-year estimation window prior to each forecasting period.
  - WEO forecast vintages: initial forecasts (April and September/October each year) compiled to compute errors.
- Model comparisons:
  - Simple financial model (Model 2 = AR + credit growth + policy controls).
  - Expanded financial model (Model 3 = AR + credit growth + stock prices + house prices + policy controls).
- Results:
  - Model 2 outperforms WEO forecasts for 69 percent of countries (Model 2 outperforms WEO (%) 69**).
  - Model 3 outperforms WEO forecasts for 85 percent of countries (Model 3 outperforms WEO (%) 85**).
- Example country-level comparisons (selected):
  - United States: WEO RMSE 3.127 (No. of obs 12) vs Model 2 RMSE 2.321 (No. of obs 19) → Model 2 outperforms.
  - Turkey: WEO RMSE 9.021 (No. of obs 5) vs Model 2 RMSE 5.017 (No. of obs 19) → Model 2 outperforms.
  - India: WEO RMSE 12.193 (No. of obs 4) vs Model 2 RMSE 4.135 (No. of obs 13) → Model 2 outperforms.
  - United States (Model 3): WEO RMSE 3.127 (No. of obs 12) vs Model 3 RMSE 1.774 (No. of obs 9) → Model 3 outperforms.
- Caveats:
  - RMSE comparisons sometimes based on differing numbers of forecasts (No. of obs differ across WEO and model series).
  - Models do not produce predictions when one or more predictors are missing; some country RMSEs based on less than 19 forecasts.
  - Forecast years limited to pre-2007 window in one exercise to exclude global financial crisis; full WEO comparison uses 2004-2013 rolling windows including later periods.

### VI. Policy-Relevant Implications and Recommendations
- Financial variables (credit growth, stock prices, house prices) contain economically meaningful information useful for nowcasting and short-term forecasting (up to four quarters).
- Practical recommendations:
  - Incorporate readily available financial indicators—especially private sector credit growth and, when available, stock and house price indices—into forecasting toolkits to improve accuracy relative to macro-only AR benchmarks and to standard IMF WEO initial forecasts.
  - Use simple baseline models relying on private sector credit growth and policy controls where other financial series are unavailable.
  - Consider pooled panel estimation to improve out-of-sample forecast accuracy for individual countries, as cross-country information can reduce forecast errors.
  - For advanced-economy forecasting where bond yield data are available, consider including sovereign and corporate bond yields cautiously; effects may be informative but are sometimes not robust when other financial variables are present.

### VII. Conclusion and Research Directions
- Main conclusion: Financial variables—both quantity measures like credit growth and price measures like equity and housing—help forecast macroeconomic variables up to four quarters ahead across advanced, emerging, and low-income country groups.
- Modeling trade-off: Country-specific regressions yield better in-sample fit, while pooled panel regressions yield better out-of-sample forecasts.
- Suggested further research:
  - Investigate reasons why panel models outperform country-specific models out-of-sample.
  - Extend comparisons to other forecasting benchmarks beyond WEO and explore behavior during financial crises (tail events).

### Key Summary Statistics (selected, exact values as reported)
- Sample counts and central tendency (All countries):
  - GDP growth (%) All: No. of Obs 6276; Mean 3.076; Std. Dev 8.247; Min -18.330; Median 2.968; Max 22.633.
  - Consumption growth (%) All: No. of Obs 6359; Mean 3.180; Std. Dev 10.615; Min -29.891; Median 3.117; Max 33.672.
  - Investment growth (%) All: No. of Obs 6337; Mean 2.699; Std. Dev 25.652; Min -74.258; Median 3.139; Max 73.203.
  - Inflation (%) All: No. of Obs 18900; Mean 10.233; Std. Dev 17.272; Min -10.293; Median 5.226; Max 88.776.
  - Credit growth (%) All: No. of Obs 17943; Mean 6.691; Std. Dev 24.343; Min -67.367; Median 6.470; Max 73.953.
  - Stock price growth (%) All: No. of Obs 6290; Mean 10.802; Std. Dev 54.356; Min -130.262; Median 11.976; Max 155.151.
  - House price growth (%) All: No. of Obs 4946; Mean 2.632; Std. Dev 12.292; Min -29.403; Median 2.133; Max 38.173.
  - Deposit rate (%) All: No. of Obs 17935; Mean 9.736; Std. Dev 10.529; Min 0.809; Median 6.713; Max 59.828.
  - Lending rate (%) All: No. of Obs 16714; Mean 16.693; Std. Dev 12.254; Min 4.250; Median 13.400; Max 67.770.
  - Sovereign Bond yield (%) All: No. of Obs 1834; Mean -5.928; Std. Dev 39.968; Min -100.151; Median -8.559; Max 94.837.
  - Corporate Bond yield (%) All: No. of Obs 714; Mean -5.606; Std. Dev 42.703; Min -111.946; Median -3.891; Max 83.888.
  - Government consumption growth (%) All: No. of Obs 6752; Mean 12.068; Std. Dev 24.500; Min -41.569; Median 7.496; Max 107.523.
  - Policy rate (%) All: No. of Obs 14940; Mean 10.686; Std. Dev 8.961; Min 0.500; Median 8.500; Max 45.000.
- Model performance vs WEO:
  - Model 2 (AR+Credit+policy) outperforms WEO in 69 percent of countries (Model 2 outperforms WEO (%) 69**).
  - Model 3 (AR+Credit+Equity+Housing+policy) outperforms WEO in 85 percent of countries (Model 3 outperforms WEO (%) 85**).

*Italic: Source — _wp16251 - References (PDF content unit provided)*

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

### _wp16251 - References

### I. Introduction and Purpose
- Study period and sample: 62 countries between 1980 and 2013.
- Objective: Examine forecasting power of financial variables (private sector credit growth, stock prices, house prices, bond yields, deposit and lending rates) for macroeconomic variables (GDP growth, private consumption growth, private investment growth, CPI inflation).
- Modeling approach: Simple, easily replicable specifications estimated country-by-country and in pooled cross-country panels; baseline specification uses only private sector credit growth and two policy controls (government consumption and policy rate); augmented specifications add stock prices, house prices, deposit and lending rates, and for a subset sovereign and corporate bond yields.
- Motivation for financial predictors:
  - Financial market imperfections link credit conditions to future macro outcomes.
  - Asset prices are forward-looking and incorporate information not yet reflected in macro outcomes.
  - Contemporaneous financial variables can assist nowcasting when macro data are lagged.

### II. Empirical Model and Data
- Forecasting equation (fixed-effects OLS FE):
  - Dependent variable: annualized growth rate of macro indicator Δh+ (annualized using scaling constant 400z).
  - Predictors: lagged dependent variable(s) (lags chosen by AIC), vector X of financial and policy controls, country fixed effects; Newey-West standard errors used for overlapping observations.
- Key macro variables: GDP growth, private consumption growth, private investment growth, CPI inflation.
- Financial predictors: private sector credit growth, stock prices, house prices, bank prime loan rate, deposit rate; extended sample adds sovereign bond yields and corporate bond yields for some advanced economies.
- Data frequency and coverage notes:
  - Quarterly data availability varies substantially by country and variable.
  - Average quarterly observations: advanced economies ~87, emerging economies ~41.
  - As of end-2015, IMF IFS has private sector credit annually for over 180 economies and quarterly for over 120 economies.
- Seasonal adjustment: X-12-ARIMA for non-seasonally adjusted raw data.
- Variables converted to real terms using country-specific GDP deflator.
- Sample winsorized at 2 percent for summary statistics.

### III. Main Empirical Findings
- General result: Incorporating financial variables significantly improves macroeconomic forecasting accuracy up to four quarters horizon.

A. Credit Growth: predictive power and magnitudes
- Credit growth is significantly associated with GDP growth, consumption growth, investment growth, and inflation (negative association) in many specifications.
- Baseline (nowcasting, simple model with credit growth):
  - A one standard deviation increase in credit growth (i.e. a 24 percentage point increase in annualized rate) is associated with a 1.79 percentage point increase in annualized GDP growth (nowcasting).
  - At one quarter ahead horizon: 1.15 percentage point increase in GDP growth.
  - At four quarters ahead horizon: 0.46 percentage point increase in GDP growth.
- Augmented model (with stock and house prices):
  - Nowcasting effects: 0.87 percentage point and 0.96 percentage point (two augmented specs reported) increases in GDP growth for a one standard deviation increase in credit growth.
  - One quarter ahead: 0.65 percentage point and 0.43 percentage point increases.
  - Credit growth loses statistical significance at four quarters ahead in some augmented specifications.
- Investment growth sensitivity to credit:
  - Nowcasting: 6.9 percentage point increase in investment growth for one standard deviation increase in credit growth.
  - One quarter ahead: 4.2 percentage point increase.
  - Four quarters ahead: 2.0 percentage point increase.
- Subsample differences:
  - Coefficients on credit growth tend to be larger among emerging market and low-income countries than among advanced economies.
  - In the subsample including sovereign and corporate bond yields, credit growth sometimes loses statistical significance.

B. Equity and Housing Prices: predictive power and magnitudes
- Stock prices and house prices predict GDP growth, consumption growth, and investment growth in most specifications, conditional on credit growth.
- GDP growth effects (one standard deviation increase):
  - One quarter ahead: stock price index → 0.95 percentage point increase; house price index → 0.95 percentage point increase (table statements).
  - Four quarters ahead: stock price index → 0.73 percentage point; house price index → 0.85 percentage point.
  - Nowcasting: stock price index → 0.59 percentage point; house price index → 1.01 percentage point.
- Consumption:
  - House price increases have a stronger impact on consumption growth than stock price increases (housing wealth large share of household wealth).
- Investment:
  - Both stock and house prices have large and significant impacts on investment growth in most specifications.
- Inflation:
  - Neither stock prices nor house prices significantly predict inflation in most specifications.

C. Interest Rates, Bond Yields, and Other Financial Variables
- Deposit and lending rates:
  - Conditional on other financial variables and policy controls, deposit and lending rates generally do not significantly predict GDP growth.
  - Deposit rate tends to predict lower consumption growth and higher investment growth in some specifications.
  - Lending rate is strongly positively associated with future investment growth in several specifications (notably where robust).
- Corporate and sovereign bond yields:
  - Corporate bond yields show a negative association with GDP growth in some baseline specifications for advanced economies (e.g., corporate bond yield → roughly -0.7 and -0.75 percentage point reductions in GDP growth at one and four quarter horizons in baseline).
  - Effects of bond yields are not robust when other financial variables are included.
  - In the advanced-economies-with-bond-yields subsample, sovereign bond yield shows significance for inflation in some horizons.

D. Robustness and Group Results
- The findings are robust across country groups: advanced economies (AE), emerging markets (EM), and low-income countries (LIC).
- Low-income country results are based on a very small sample (noted sample may be unrepresentative: LIC regression sample consists of only 3 countries in some specifications).
- Adjusted R-squared (selected ranges, all countries, four-quarter ahead horizon):
  - Investment and GDP growth: 0.25 to 0.3.
  - Consumption and inflation: 0.5 to 0.6.
- Goodness-of-fit comparisons:
  - Comparable goodness of fit for GDP growth in AE and EM.
  - Consumption growth fit much better in AE than EM (0.62 vs 0.18 reported).

### IV. Panel vs. Country-Specific Models: Forecast Performance
- In-sample vs out-of-sample patterns:
  - In-sample: individual country regressions have better fit than pooled panel regressions.
  - Out-of-sample: panel regressions outperform individual country regressions in forecasting accuracy.
- Out-of-sample forecasting exercise:
  - Full-sample fit period: 1980-2013.
  - Out-of-sample forecasts: models estimated on 1980-1999, four-quarter ahead forecasts computed for 2000-2007 (excluding crisis period).
  - Forecast errors measured by root mean squared errors (RMSE).
  - Three models evaluated:
    - AR model with lags chosen by AIC (up to 7 lags).
    - Financial model augmenting AR with credit growth and two policy controls (policy rate and government consumption).
    - Expanded financial model augmenting AR with credit growth, house price growth, stock price growth, and two policy controls.
- Advanced economies (selected results):
  - In-sample: panel model outperforms individual model in 4 percent of cases (Model 1) and 0 percent (Model 2) for the set reported.
  - Out-of-sample: panel model outperforms individual model in 65 percent of cases for Model 1 and 83 percent for Model 2 (statistical significance indicated for Model 2 as 83*** in table).
  - Large RMSE reductions in some countries (example statements: Belgium and Japan RMSE reductions by factor of 3 in some cases).
- Emerging markets:
  - In-sample: panel model outperforms individual in 6 percent (Model 1) and 0 percent (Model 2) of cases reported.
  - Out-of-sample: panel model outperforms individual in 100 percent of reported emerging economies for both Model 1 and Model 2 (100*** for both).
  - RMSE reductions in out-of-sample can be substantial (e.g., RMSE divided by about 2 for Turkey and Brazil; by about 4 for Colombia and Peru).
- Interpretation: pooled international information (common coefficients) improves out-of-sample forecast performance despite worse in-sample fit.

### V. Comparison with IMF WEO Forecasts
- Real-time out-of-sample exercise:
  - Four-quarter ahead quarterly GDP growth forecasts for each country from 2004 Q1 to 2013 Q1 (19 quarters).
  - Rolling 20-year estimation window prior to each forecasting period.
  - WEO forecast vintages: initial forecasts (April and September/October each year) compiled to compute errors.
- Model comparisons:
  - Simple financial model (Model 2 = AR + credit growth + policy controls).
  - Expanded financial model (Model 3 = AR + credit growth + stock prices + house prices + policy controls).
- Results:
  - Simple financial model (Model 2) outperforms WEO forecasts for 69 percent of countries in the sample (Model 2 outperforms WEO (%) = 69**).
  - Expanded financial model (Model 3) outperforms WEO forecasts for 85 percent of countries in the sample (Model 3 outperforms WEO (%) = 85**).
  - Example country-level comparisons (selected):
    - United States: WEO RMSE 3.127 (No. of obs 12) vs Model 2 RMSE 2.321 (No. of obs 19) → Model 2 outperforms (1).
    - Turkey: WEO RMSE 9.021 (No. of obs 5) vs Model 2 RMSE 5.017 (No. of obs 19) → Model 2 outperforms (1).
    - India: WEO RMSE 12.193 (No. of obs 4) vs Model 2 RMSE 4.135 (No. of obs 13) → Model 2 outperforms (1).
    - Model 3 examples: United States WEO RMSE 3.127 (No. of obs 12) vs Model 3 RMSE 1.774 (No. of obs 9) → Model 3 outperforms (1).
- Caveats:
  - RMSE comparisons sometimes based on differing numbers of forecasts (No. of obs differ across WEO and model series).
  - Models do not produce predictions when one or more predictors are missing; some country RMSEs based on less than 19 forecasts.
  - Forecast years limited to pre-2007 window in one exercise to exclude global financial crisis; full WEO comparison uses 2004-2013 rolling windows including later periods.

### VI. Policy-Relevant Implications and Recommendations
- Financial variables (credit growth, stock prices, house prices) contain economically meaningful information useful for nowcasting and short-term forecasting (up to four quarters).
- Practical recommendation: Incorporate readily available financial indicators—especially private sector credit growth and, when available, stock and house price indices—into forecasting toolkits to improve accuracy relative to macro-only AR benchmarks and to standard IMF WEO initial forecasts.
- For forecasting practice across many countries (including those with limited data):
  - Use simple baseline models relying on private sector credit growth and policy controls where other financial series are unavailable.
  - Consider pooled panel estimation to improve out-of-sample forecast accuracy for individual countries, as cross-country information can reduce forecast errors.
- For advanced-economy forecasting where bond yield data are available, consider including sovereign and corporate bond yields cautiously; effects may be informative but are sometimes not robust when other financial variables are present.

### VII. Conclusion and Research Directions
- Main conclusion: Financial variables—both quantity measures like credit growth and price measures like equity and housing—help forecast macroeconomic variables up to four quarters ahead across advanced, emerging, and low-income country groups.
- Modeling trade-off: Country-specific regressions yield better in-sample fit, while pooled panel regressions yield better out-of-sample forecasts.
- Suggested further research:
  - Investigate reasons why panel models outperform country-specific models out-of-sample.
  - Extend comparisons to other forecasting benchmarks beyond WEO and explore behavior during financial crises (tail events).

### Key Summary Statistics (selected, exact values as reported)
- Sample counts and central tendency (All countries):
  - GDP growth (%) All: No. of Obs 6276; Mean 3.076; Std. Dev 8.247; Min -18.330; Median 2.968; Max 22.633.
  - Consumption growth (%) All: No. of Obs 6359; Mean 3.180; Std. Dev 10.615; Min -29.891; Median 3.117; Max 33.672.
  - Investment growth (%) All: No. of Obs 6337; Mean 2.699; Std. Dev 25.652; Min -74.258; Median 3.139; Max 73.203.
  - Inflation (%) All: No. of Obs 18900; Mean 10.233; Std. Dev 17.272; Min -10.293; Median 5.226; Max 88.776.
  - Credit growth (%) All: No. of Obs 17943; Mean 6.691; Std. Dev 24.343; Min -67.367; Median 6.470; Max 73.953.
  - Stock price growth (%) All: No. of Obs 6290; Mean 10.802; Std. Dev 54.356; Min -130.262; Median 11.976; Max 155.151.
  - House price growth (%) All: No. of Obs 4946; Mean 2.632; Std. Dev 12.292; Min -29.403; Median 2.133; Max 38.173.
  - Deposit rate (%) All: No. of Obs 17935; Mean 9.736; Std. Dev 10.529; Min 0.809; Median 6.713; Max 59.828.
  - Lending rate (%) All: No. of Obs 16714; Mean 16.693; Std. Dev 12.254; Min 4.250; Median 13.400; Max 67.770.
  - Sovereign Bond yield (%) All: No. of Obs 1834; Mean -5.928; Std. Dev 39.968; Min -100.151; Median -8.559; Max 94.837.
  - Corporate Bond yield (%) All: No. of Obs 714; Mean -5.606; Std. Dev 42.703; Min -111.946; Median -3.891; Max 83.888.
  - Government consumption growth (%) All: No. of Obs 6752; Mean 12.068; Std. Dev 24.500; Min -41.569; Median 7.496; Max 107.523.
  - Policy rate (%) All: No. of Obs 14940; Mean 10.686; Std. Dev 8.961; Min 0.500; Median 8.500; Max 45.000.

- Model performance vs WEO:
  - Model 2 (AR+Credit+policy) outperforms WEO in 69 percent of countries (Model 2 outperforms WEO (%) 69**).
  - Model 3 (AR+Credit+Equity+Housing+policy) outperforms WEO in 85 percent of countries (Model 3 outperforms WEO (%) 85**).

*Italic: Source — _wp16251 - References (PDF content unit provided)*

### REFERENCES

### _wp16251 - REFERENCES

### References
- Bernanke, Ben, Mark Gertler, and Simon Gilchrist, 1996, “The Financial Accelerator and the Flight to Quality,” Review of Economics and Statistics, Vol. 78(1): 1-15.
- Claessens, Stijn, Ayhan, Kose, and Marco Terrones, 2008, “What Happens During Recessions, Crunches and Busts?”, IMF WP/08/274
- ECB, 2004, European Central Bank Monthly Bulletin, December 2004, European Central Bank.
- Arturo Estrella and Frederic S. Mishkin, 1998, “Predicting U.S. Recessions: Financial Variables as Leading Indicators,” Review of Economics and Statistics, Vol 80(1): 45-61.
- Garcia-Ferrer, A., R. A. Highfield, F. Palm and A. Zellner, 1987, “Macroeconomic Forecasting Using Pooled International Data,” Journal of Business & Economic Statistics Vol. 5(1): 53-67
- Gilchrist, Simon and Egon Zakrajšek, 2012, “Credit Spreads and Business Cycle Fluctuations,” American Economic Review, 102(4): 1692-1720.
- Hoogstrate AJ, FC Palm, GA Pfann2000, “Pooling in dynamic panel-data models: An application to forecasting GDP growth rates,” Journal of Business & Economic Statistics Vol. 5(1): 53-67
- Iacoviello, Matteo, 2012, “Housing Wealth and Consumption,” International Encyclopedia of Housing and Home, pp. 673-678, Elsevier.
- IMF, 2016, “2016 Handbook of IMF Facilities for Low-Income-Countries,” International Monetary Fund.
- Leamer, Edward, 2007. "Housing is the business cycle," Proceedings - Economic Policy Symposium - Jackson Hole, Federal Reserve Bank of Kansas City, pages 149-233
- Philippon Thomas, 2009. “The Bond Market's q,” The Quarterly Journal of Economics, Vol. 124(3): 1011-1056.
- Stock, James H. and Mark W. Watson, 2003, “Forecasting Output and Inflation: The Role of Asset Prices,” Journal of Economic Literature, Vol.(XLI): 788-829.

### Appendix — Table A1. Data Source
- Variable — Definition — Source
- GDP — Gross domestic product, volume — IFS
- Private consumption — Household consumption expenditure — IFS
- Private investment — Gross fixed capital formation — IFS
- CPI — Consumer price index — IFS
- Credit — Claims on private sector — IFS
- Stock price — Share price index — Bloomberg
- House price — House price — OECD and BIS
- Deposit rate — Deposit rate — IFS
- Lending rate — Lending rate — IFS
- Sovereign bond yield — Sovereign credit default swap, 10 year maturity — Bloomberg
- Corporate bond yield — Barclay global aggregate corporate bond index — Bloomberg
- Government consumption — Government final consumption expenditure — IFS
- Short term interest rate — Central bank policy rate or discount rate — IFS
- GDP deflator — Gross domestic product deflator index — IFS
- WEO forecasts
- Real GDP growth — computed from gross domestic product (current prices) — WEO

### Appendix — Table A2. Country List
- Group 1 (Advanced Economies) Full sample
  - Australia, Austria, Belgium, Canada, Cyprus, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Israel, Italy, Japan, Korea, Latvia, Malta, Netherlands, New Zealand, Norway, Portugal, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, United Kingdom, United States
- Subsample with bond yields
  - Australia, Canada, France, Germany, Italy, Japan, Netherlands, Spain, Sweden, Switzerland, United Kingdom, United States
- Group 2 (Emerging markets)
  - Albania, Armenia, Botswana, Brazil, Bulgaria, Chile, Colombia, Costa Rica, Croatia, Ecuador, Egypt, Georgia, Guatemala, Hungary, India, Indonesia, Iran, Kazakhstan, Macedonia, Malaysia, Mauritius, Mexico, Paraguay, Peru, Philippines, Romania, Russia, South Africa, Thailand, Turkey
- Group 3 (Low-income countries) 1/
  - Bolivia, Kyrgyz Republic, Moldova

- Source: IMF WEO, IMF (2016)
- Notes: Bolivia graduated from PRGT-eligibility on October 16, 2015 but is included in our low-income countries regression.

* _wp16251 - REFERENCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16251.pdf_