## _wp09280

## Source details

**Canonical URL:** [_wp09280](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09280.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09280.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09280.pdf.json)

---

### I. INTRODUCTION — study focus and approach
- Objective: Explain cross-country differences in the impact of the 2008–09 global financial crisis on developing countries and emerging markets by analyzing revisions in GDP growth forecasts for 2009 (pre- and post-crisis).
- Rationale for methodology:
  - Use forecast revisions to avoid confounding from differences in levels of development, cyclical positions, or anticipated non-crisis adjustments.
  - Forecast revisions incorporate expected short-term policy effects and allow for flexible transmission lags.
- Conceptual transmission channels considered: structural characteristics, initial vulnerabilities, macroeconomic policies, and financial/trade linkages with advanced economies.
- Empirical approach: descriptive evidence and cross-country regressions across a broad set of explanatory variables.

### II. DATA — forecasts and explanatory variables
- Forecast datasets:
  - Consensus Forecasts (monthly) — baseline dataset; uses averages for January–June 2008 and January–June 2009.
  - IMF World Economic Outlook (WEO) forecasts (revised April and October) — used to broaden country coverage for robustness.
- Dependent variable:
  - Change in Consensus Forecast for 2009 (January–June 2009 average minus January–June 2008 average).
- Explanatory variable groups (2007 values used to avoid endogeneity):
  - (i) Trade linkages: trade openness, trade composition (share of commodities vs. manufactured products), direction of trade (share with advanced economies).
  - (ii) Financial linkages: measures of financial openness, capital account restrictions, stock of bank lending from advanced economies (relative to GDP), remittances share.
  - (iii) Underlying vulnerabilities and financial structure: current account balance, international reserves, indebtedness, credit growth, leverage, complexity of financial structures.
  - (iv) Policy and institutional framework: exchange-rate regime flexibility, inflation, volatility of reserves and exchange rates, inflation-targeting dummy, primary fiscal gap, institutional quality measures.

### III. DESCRIPTIVE EVIDENCE — broad patterns
- Sample for descriptive analysis: core sample of 43 emerging markets (Consensus Forecasts).
- Range of growth forecast revisions in this sample: - 18 percent to - 1.5 percent.
- Average difference in impact between more and less affected groups: about 5 percentage points.
- Regional patterns:
  - Eastern Europe and Central Asia (EECA): largest and most volatile growth collapses.
  - Latin America: on average much more contained impact.
- Trade channel:
  - Little correlation between share of food or manufacturing in total exports and growth impact across the whole sample.
  - Latin America has higher food commodity shares, consistent with relative resilience when soft commodity prices held up.
- Financial linkages:
  - Most affected countries had liabilities to banks in advanced countries averaging about 66 percent of GDP; less affected countries averaged 19 percent of GDP.
  - EECA relied more on foreign credit — borrowing roughly double that of other regions in 2007.
- Vulnerabilities:
  - Positive correlation between leverage (credit-to-deposit ratio) and cumulative bank credit growth (2005–07) with severity of growth impact.
  - EECA exhibited higher vulnerabilities; average cumulative credit growth in EECA was about four times that in other regions; Latin America had the lowest rate.
- External position and buffers:
  - Less affected countries on average had current account surpluses; EECA recorded larger deficits relative to other regions.
  - Higher international reserves tended to be associated with smaller growth revisions, but the relationship is weak.
- Policy frameworks:
  - More flexible exchange rates appeared to buffer the shock; countries with pegs were hit particularly strongly. On average, EECA had the least flexible regimes.
  - Countries with larger downward revisions tended to have weaker fiscal positions (primary fiscal gap measure); Latin America exhibited the most favorable position on average.
  - Little correlation found between institutional variables and size of output impact (though institutional transparency indices correlated with extent to which financial markets were hit during peak turbulence).

### IV. REGRESSION RESULTS — baseline for emerging markets
- Baseline sample: 40 emerging market countries (dependent variable: Change in Consensus Forecast for 2009 between January–June 2009 and January–June 2008).
- Main empirical findings:
  - Financial vulnerabilities (leverage and cumulative credit growth) and exchange rate policy explain a large share of cross-country variation in growth forecast revisions.
  - Leverage effect stronger in EU accession countries (interaction with EU accession dummy).
  - Weak evidence that a stronger pre-crisis fiscal position shielded countries (primary fiscal gap).
  - Institutional variables and many policy-quality measures generally not statistically significant in a consistent way.
- Selected baseline regression coefficients and statistics (standard errors in parentheses; significance levels: *** p<0.01, ** p<0.05, * p<0.1):
  - Specification (1):
    - Leverage: -0.04** (0.01)
    - Exchange rate peg dummy: -1.94** (0.87)
    - Cumulative credit growth: -0.01* (0.00)
    - Constant: 0.72 (1.30)
    - Observations: 40
    - R-squared: 0.57
  - Specification (2):
    - Leverage: -0.04*** (0.01)
    - Exchange rate peg dummy: -1.82** (0.81)
    - Cumulative credit growth: -0.00* (0.00)
    - EU accession dummy: -2.29** (0.88)
    - Constant: 0.44 (1.21)
    - Observations: 40
    - R-squared: 0.64
  - Specification (3):
    - Exchange rate peg dummy: -4.27*** (1.05)
    - Primary gap: 0.48*** (0.15)
    - Constant: -7.06*** (0.98)
    - Observations: 32
    - R-squared: 0.46
- Quantitative interpretation highlighted:
  - An increase in leverage of ten percentage points is associated with a reduction in growth forecasts of 0.4 percentage points.
  - (Text truncated in source for the exact numeric interpretation of a ten-percentage-point-higher cumulative credit growth.)

### V. FINANCIAL VULNERABILITIES: leverage and credit growth (detailed)
- Leverage quartet comparisons:
  - Most-levered quartile average leverage: 185 percent of GDP.
  - Least-levered quartile average leverage: 83 percent of GDP.
  - If most-levered quartile had leverage of least-levered quartile, growth revisions would have been, on average, 4.1 percentage points smaller.
- Credit-growth quartet comparisons:
  - Fastest cumulative credit growth quartile average: almost 350 percent.
  - Slowest cumulative credit growth quartile average: 14 percent.
  - If fastest-growth quartile had credit growth of slowest quartile, growth revisions would have been 3.3 percentage points smaller.
- Attribution:
  - Leverage explains:
    - virtually all of the growth revision for the least affected countries in the sample;
    - roughly two thirds of the revision for the average country;
    - slightly more than half of the revision for those countries most affected by the crisis.
  - Credit growth explains a significant share of the growth revision for the average country and those most affected.

### VI. MONETARY, EXCHANGE-RATE, AND FISCAL FRAMEWORKS
- Monetary and exchange-rate frameworks:
  - Exchange rate flexibility matters more than other monetary-policy measures tested (dummy for inflation targeters, inflation levels, inflation volatility).
  - Moving from a peg toward a more flexible regime concentrated the benefits; distinguishing between crawls and floats does not improve fit.
  - In most regressions, pegged exchange rates experienced, on average, larger downward growth revisions (in excess of two percentage points) compared to more flexible regimes.
  - Stock of international reserves (share of GDP, exports, or short-term debt) did not have a statistically significant effect; possible nonlinear or threshold effects may exist.
- Fiscal policy and buffers:
  - In certain specifications (Equation 3), the primary fiscal gap is positively associated with better growth performances.
  - The quartile with the largest primary gaps had on average growth revisions 4.2 percentage points less negative than the quartile with the smallest primary gaps.
  - Other fiscal variables do not appear to matter once other factors are controlled for.
  - Public debt enters significantly in certain specifications but with a counterintuitive sign (higher debt associated with better growth performances), likely reflecting private-sector imbalances in some hardest-hit countries or high domestic savings/credible frameworks in some better-performing countries.

### VII. TRADE LINKAGES, BROADER SAMPLE RESULTS, AND INTERACTIONS
- WEO sample (126 countries) findings:
  - Share of commodities (both food and overall) in total exports associated with smaller downward growth revisions.
  - Share of manufacturing products in total exports correlated with worse growth performance for both advanced and developing countries.
- Emerging-market interaction:
  - Interaction terms for trade measures and an emerging-market dummy enter with coefficients of similar magnitude but opposite sign, implying an overall effect of zero for emerging market countries.
- Financial linkages in broader sample:
  - Larger stock of lending from advanced countries contributed to a more severe downward revision of the growth forecast.

### VIII. ROBUSTNESS AND ADDITIONAL OBSERVATIONS
- Robustness checks:
  - Results largely robust to changing periods of Consensus Forecasts changes (moving from changes in averages to April or to August).
  - Using cumulative credit growth revisions from April 2008 to 2009 loses statistical significance in one test, but economic significance remains similar.
  - Regressions with WEO forecast changes for the same countries reduce statistical significance of leverage due to differences in forecast revisions for Eastern European countries between Consensus and WEO datasets.
- Reported R-squared values across various tables and specifications:
  - Table 2: 0.240, 0.255, 0.258, 0.264, 0.338, 0.345, 0.391, 0.450.
  - Table 3: 0.508, 0.550, 0.423, 0.474, 0.558, 0.462.
  - Table 4: 0.489, 0.529, 0.482.
- Additional empirical nuances:
  - EU accession countries appear to have been hit particularly hard, possibly due to stronger trade and financial linkages with EU member countries at the core of the crisis.
  - Currency mismatch measures and share of foreign-currency deposits are significant in some specifications but weaken once leverage is controlled for.
  - Many countries with credit booms ran sizable external current account deficits; correlation between credit growth and the current account balance (excluding two oil-fueled credit-boom outliers) is -67.
  - Lending from advanced economies did not enter significantly in the Consensus sample once other factors are controlled for, but did in the broader WEO sample.

### IX. POLICY IMPLICATIONS
- Core policy recommendations:
  - Reduce domestic financial vulnerabilities by monitoring and addressing leverage and credit booms through prudential regulation and supervision.
  - Maintain exchange-rate flexibility as a shock absorber.
  - Build fiscal space in good times to enable countercyclical fiscal policies during shocks.
  - Consider trade composition: reliance on advanced-manufacturing exports increases vulnerability to demand collapses in advanced economies, whereas commodity exposure (food/overall commodities) can be relatively less damaging in certain episodes.

### X. CONCLUSION
- A relatively small set of variables—leverage, cumulative credit growth, and exchange-rate pegs—can explain much of the difference in countries’ prospects after the financial crisis intensified in September 2008.
- Results are robust across a wide variety of specifications and country samples, with evidence that trade linkages matter especially for non-emerging market developing countries.
- Further research is needed as more data become available and the global economy enters recovery to better understand policy responses and other institutional and structural factors affecting recession duration and recovery speed/size.

*Italic: Source — content from _wp09280 - References (PDF) as provided.*

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

### _wp09280 - References .............................................................................................................

### I. INTRODUCTION — study focus and approach
- Objective: Explain cross-country differences in the impact of the 2008–09 global financial crisis on developing countries and emerging markets by analyzing revisions in GDP growth forecasts for 2009 (pre- and post-crisis).
- Rationale for methodology:
  - Use forecast revisions to avoid confounding from differences in levels of development, cyclical positions, or anticipated non-crisis adjustments.
  - Forecast revisions incorporate expected short-term policy effects and allow for flexible transmission lags.
- Key conceptual points:
  - Transmission depends on structural characteristics, initial vulnerabilities, macroeconomic policies, and financial/trade linkages with advanced economies.
  - Analysis uses descriptive evidence and cross-country regressions across a broad set of explanatory variables.

### II. DATA — forecasts and explanatory variables
- Forecast datasets:
  - Consensus Forecasts (monthly) — baseline dataset; uses averages for January–June 2008 and January–June 2009.
  - IMF World Economic Outlook (WEO) forecasts (revised April and October) — used to broaden country coverage for robustness.
- Dependent variable:
  - Change in Consensus Forecast for 2009 (January–June 2009 average minus January–June 2008 average).
- Explanatory variable groups (2007 values used to avoid endogeneity):
  - (i) Trade linkages: trade openness, trade composition (share of commodities vs. manufactured products), direction of trade (share with advanced economies).
  - (ii) Financial linkages: measures of financial openness, capital account restrictions, stock of bank lending from advanced economies (relative to GDP), remittances share.
  - (iii) Underlying vulnerabilities and financial structure: current account balance, international reserves, indebtedness, credit growth, leverage, complexity of financial structures.
  - (iv) Policy and institutional framework: exchange-rate regime flexibility, inflation, volatility of reserves and exchange rates, inflation-targeting dummy, primary fiscal gap (difference between actual primary balance and balance consistent with constant debt-to-GDP), institutional quality measures.
- Appendix materials (listed in source): Appendix Table A.1 (variables and expected signs) and Table A.2 (countries in consensus sample).

### III. DESCRIPTIVE EVIDENCE — broad patterns
- Sample for descriptive analysis: core sample of 43 emerging markets (Consensus Forecasts).
- Range of growth forecast revisions in this sample: - 18 percent to - 1.5 percent.
- Average difference in impact between more and less affected groups: about 5 percentage points.
- Regional patterns:
  - Eastern Europe and Central Asia (EECA): largest and most volatile growth collapses.
  - Latin America: on average much more contained impact.
- Trade channel:
  - Little correlation between share of food or manufacturing in total exports and growth impact across the whole sample.
  - Latin America has higher food commodity shares, consistent with relative resilience when soft commodity prices held up.
- Financial linkages:
  - Most affected countries had liabilities to banks in advanced countries averaging about 66 percent of GDP; less affected countries averaged 19 percent of GDP.
  - EECA relied more on foreign credit — borrowing roughly double that of other regions in 2007.
- Vulnerabilities:
  - Positive correlation between leverage (credit-to-deposit ratio) and cumulative bank credit growth (2005–07) with severity of growth impact.
  - EECA exhibited higher vulnerabilities; average cumulative credit growth in EECA was about four times that in other regions; Latin America had the lowest rate.
- External position and buffers:
  - Less affected countries on average had current account surpluses; EECA recorded larger deficits relative to other regions.
  - Higher international reserves tended to be associated with smaller growth revisions, but the relationship is weak.
- Policy frameworks:
  - More flexible exchange rates appeared to buffer the shock; countries with pegs were hit particularly strongly. On average, EECA had the least flexible regimes.
  - Countries with larger downward revisions tended to have weaker fiscal positions (primary fiscal gap measure); Latin America exhibited the most favorable position on average.
  - Little correlation found between institutional variables and size of output impact (though a note: there is significant correlation between institutional transparency indices and extent to which financial markets were hit during peak turbulence).

### IV. REGRESSION RESULTS — baseline for emerging markets
- Baseline sample: 40 emerging market countries (dependent variable: change in Consensus Forecast for 2009 between January–June 2009 and January–June 2008).
- Main empirical findings:
  - Financial vulnerabilities (leverage and cumulative credit growth) and exchange rate policy explain a large share of cross-country variation in growth forecast revisions.
  - Leverage effect stronger in EU accession countries (interaction with EU accession dummy).
  - Weak evidence that a stronger pre-crisis fiscal position shielded countries (primary fiscal gap).
  - Institutional variables and many policy-quality measures generally not statistically significant in a consistent way.
- Table 1. Baseline Regression Results (specifications reported):
  - Specification (1):
    - Dependent variable: Change in consensus forecast
    - Leverage: -0.04** (standard error (0.01))
    - Exchange rate peg dummy: -1.94** (standard error (0.87))
    - Cumulative credit growth: -0.01* (standard error (0.00))
    - Constant: 0.72 (standard error (1.30))
    - Observations: 40
    - R-squared: 0.57
  - Specification (2):
    - Leverage: -0.04*** (0.01)
    - Leverage * EU accession dummy: - (reported in table header but coefficient not shown in source extract)
    - Exchange rate peg dummy: -1.82** (0.81)
    - Cumulative credit growth: -0.00* (0.00)
    - EU accession dummy: -2.29** (0.88)
    - Constant: 0.44 (1.21)
    - Observations: 40
    - R-squared: 0.64
  - Specification (3):
    - Leverage: - (not reported in extract)
    - Exchange rate peg dummy: -4.27*** (1.05)
    - Primary gap: 0.48*** (0.15)
    - Constant: -7.06*** (0.98)
    - Observations: 32
    - R-squared: 0.46
  - Note: Standard errors in parentheses. Significance levels: *** p<0.01, ** p<0.05, * p<0.1.
- Quantitative interpretation highlighted in text:
  - An increase in leverage of ten percentage points is associated with a reduction in growth forecasts of 0.4 percentage points.
  - A ten-percentage-point-higher cumulative credit growth during 2004–2007 predicts a growth cut of around (text is truncated in source at this point).

### V. SYNTHESIS — substantive findings and policy-relevant messages
- Substantive findings:
  - Financial vulnerabilities were central to the severity of the crisis impact: higher leverage and faster prior credit growth correlate with larger downward revisions in growth forecasts.
  - Exchange-rate flexibility helped buffer the shock; pegged regimes fared significantly worse.
  - For emerging markets, the financial channel dominated the trade channel; for a broader set of developing countries the trade channel also mattered, especially for exporters of advanced manufacturing goods versus food exporters.
  - Pre-crisis fiscal strength shows limited evidence of mitigating impact, possibly via capacity for countercyclical fiscal policy (primary fiscal gap result).
  - Institutional quality and many macro-policy quality indicators did not show consistent explanatory power in the regressions.
- Policy implications implied by analysis:
  - Reducing domestic financial vulnerabilities (monitoring leverage and credit booms) can reduce exposure to global financial shocks.
  - Maintaining exchange-rate flexibility can serve as a shock absorber.
  - Building fiscal space prior to shocks may improve the ability to respond countercyclically.
  - Trade composition matters for exposure; reliance on advanced-manufacturing exports increases vulnerability to demand collapses in advanced economies.

*Italic: Source — content from _wp09280 - References (PDF) as provided.*

### 0.1 percentage points (Equation 1). Put another way, if the countries in the most-levered

### _wp09280 - 0.1 percentage points (Equation 1). Put another way, if the countries in the most-levered

### Financial vulnerabilities: leverage and credit growth
- If the countries in the most-levered quartile (with average leverage of 185 percent of GDP) had had the same leverage ratios as the countries in the least-levered quartile (83 percent), their growth revisions would have been, on average, 4.1 percentage points smaller.
- If the quartile of countries with the fastest cumulative credit growth (with average growth of almost 350 percent) had had the same credit growth as the countries in the slowest credit growth quartile (with average growth of only 14 percent), their growth revisions would have been 3.3 percentage points smaller.
- Leverage explains:
  - virtually all of the growth revision for the least affected countries in the sample;
  - roughly two thirds of the revision for the average country;
  - slightly more than half of the revision for those countries most affected by the crisis.
- Credit growth explains a significant share of the growth revision for:
  - the average country; and
  - those most affected.

### Monetary and exchange-rate frameworks
- Exchange rate flexibility matters more than other monetary-policy measures tested (dummy for inflation targeters, inflation levels, inflation volatility).
- Countries with more flexible exchange rates as measured under the Fund’s classification system tended to experience smaller growth revisions.
- The benefits are concentrated in moving from a peg toward a more flexible regime; distinguishing between crawls and floats does not improve the fit.
- In most regressions, countries with pegged exchange rates experienced, on average, larger downward growth revisions (in excess of two percentage points) compared to countries with more flexible exchange rates.
- The stock of international reserves—measured as a share of GDP, exports, or short-term debt—did not have a statistically significant effect on growth revisions in the sample; a possible nonlinear relationship or threshold effect may explain this result.

### Fiscal policy and buffers
- In certain specifications (Equation 3), the primary fiscal gap is positively associated with better growth performances.
- The quartile of countries with the largest primary gaps had on average growth revisions that were 4.2 percentage points less negative than the quartile with the smallest primary gaps.
- Other fiscal variables (various measures of the fiscal balance, size of government) do not appear to matter once other factors are controlled for.
- Public debt enters significantly in certain specifications but with a counterintuitive sign (higher debt associated with better growth performances), likely reflecting private-sector imbalances in several hardest-hit countries or high domestic savings/credible frameworks in some better-performing countries.

### Trade linkages and composition of exports
- Using the WEO dataset for 126 countries, trade composition matters:
  - The share of commodities (both food and overall) in total exports is associated with smaller downward growth revisions.
  - The share of manufacturing products in total exports is correlated with worse growth performance for both advanced and developing countries.
- Interaction with an emerging-market dummy:
  - Interaction terms for trade measures and an emerging-market dummy enter with coefficients of similar magnitude but opposite sign, implying an overall effect of zero for emerging market countries.
- Financial linkages in this broader sample:
  - A larger stock of lending from advanced countries contributed to a more severe downward revision of the growth forecast.

### Regression and robustness highlights
- A simple specification of leverage, cumulative growth in credit, and controlling for exchange-rate pegs explains more than half the variation in the growth revisions (explanatory power unmatched by other policy variables analyzed).
- Robustness tests:
  - Results are largely robust to changing periods of Consensus Forecasts changes (moving from changes in averages to April or to August).
  - Cumulative credit growth using revisions from April 2008 to 2009 loses statistical significance in one test, but economic significance remains the same.
  - Regressions with WEO forecast changes for the same set of countries reduce the statistical significance of leverage (owing to differences in forecast revisions for Eastern European countries between Consensus and WEO datasets).
- Selected reported regression metrics (sample results):
  - Table 2 R-squared values reported across specifications: 0.240, 0.255, 0.258, 0.264, 0.338, 0.345, 0.391, 0.450.
  - Table 3 R-squared values reported across specifications: 0.508, 0.550, 0.423, 0.474, 0.558, 0.462.
  - Table 4 R-squared values reported across specifications: 0.489, 0.529, 0.482.

### Key empirical observations and nuances
- EU accession countries appear to have been hit particularly hard, possibly because of stronger trade and financial linkages with EU member countries at the core of the crisis.
- Currency mismatch measures (foreign assets minus foreign liabilities over GDP) and share of foreign currency deposits among total deposits are significant in some specifications but their effects are weaker or lose significance once leverage is controlled for.
- Many countries with credit booms ran sizable external current account deficits; correlation between credit growth and the current account balance (excluding two oil-fueled credit-boom outliers) is -67.
- Lending from advanced economies did not enter significantly in the Consensus sample once other factors are controlled for, but did in the broader WEO sample.

### Policy recommendations (preliminary)
- Exchange-rate flexibility is crucial to dampen the impact of large shocks.
- Prudential regulation and supervision should aim to prevent the build-up of vulnerabilities associated with credit booms.
- A solid fiscal position during “good times” creates buffers to conduct countercyclical fiscal policies during shocks.

### Conclusion
- A relatively small set of variables—leverage, cumulative credit growth, and exchange-rate pegs—can explain much of the difference in countries’ prospects after the financial crisis intensified in September 2008.
- Results are robust across a wide variety of specifications and country samples, with some evidence that trade linkages matter, especially for non-emerging market developing countries.
- Further research is needed as more data become available and the global economy enters recovery to better understand policy responses and other institutional and structural factors affecting recession duration and recovery speed/size.

*Source: IMF working paper content (excerpt provided).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09280.pdf_