## _wp10237 — Appendix 3. Reserve Robustness with Alternative Control Variables

## Source details

**Canonical URL:** [_wp10237 — Appendix 3. Reserve Robustness with Alternative Control Variables](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10237.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10237.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10237.pdf.json)

---

### I. Purpose and context
- Context:
  - Global growth fell by 6 percentage points from its pre-crisis peak to its trough in 2009, "the largest straight fall in global growth in the post-war era."
  - Real output in EMs fell about 4 percent between 2008Q3 and 2009Q1.
  - Sample: around 50 emerging market economies (EMs); overall country sample coverage noted as 57 economies in Appendix 1 (actual regressions vary by data availability).
- Purpose:
  - Explore channels and factors shaping heterogeneous crisis impact across EMs, with emphasis on the role of international reserves.
  - Focus on external sector vulnerabilities prior to the crisis: current account deficits, reserve holdings, external debt levels, and an IMF staff external vulnerability index.

### II. Measurement and preferred impact metric
- Alternative output loss metrics used for robustness:
  - Growth 1: percent change in seasonally adjusted real GDP between 2008Q3 and 2009Q1.
  - Growth 2: log difference between actual 2009Q1 GDP and a linear-trend projection (2003–2007).
  - Growth 3: country-specific peak-to-trough percent change in quarterly seasonally adjusted real GDP (preferred measure).
  - Growth 4: log difference between actual GDP at country’s trough and projected GDP based on 2003−2007 trend.
- Rationale for Growth 3:
  - Captures heterogeneity in timing of peaks and troughs across countries.
  - Highly correlated with Growth 1 and Growth 4 (strongest correlation with Growth 4).

### III. Core empirical findings on reserves and vulnerabilities
- Pre-crisis external vulnerabilities:
  - External sector vulnerability index (range 0 - 1): coefficient -6.40 ** (standard error (3.04)) in Table 2a; -7.64 ** (3.67) in Table 2b.
  - The least vulnerable EMs contracted 6½ percentage points less than the most vulnerable EMs (Table 2a).
  - Using Growth 4, the least vulnerable EMs contracted 7½ percentage points less than the most vulnerable EMs (Table 2b).
- Trade and financial linkages:
  - Trade linkages: EMs experienced an additional 1½ percentage point reduction in real output for every percentage point fall in domestic demand of their advanced economy trading partners.
  - Foreign bank claims (percent of GDP, log): negative associations with output. Selected coefficients:
    - Table 2a: -1.7 * (0.92), -1.94 ** (0.81), -1.85 ** (0.83), -1.92 * (1.00).
    - Table 2b: -1.89 * (1.11), -2.30 ** (0.99), -1.87 * (0.96), -2.10 * (1.16).
- Current account and external debt:
  - Current account balance (percent of GDP): 0.17 * (0.08) in Table 2a; 0.24 ** (0.10) in Table 2b.
  - External debt (percent of GDP): -0.09 ** (0.04) in Table 2a; -0.11 ** (0.04) in Table 2b.

### IV. Reserves: mitigating effect and non-linearity
- Main conclusion:
  - Higher pre-crisis reserve holdings reduced external vulnerability and moderated output collapse, but with diminishing returns—"at very high levels of reserves there is little discernable evidence of their moderating impact on output collapse."
- Reserve coverage definition and role:
  - Reserve coverage defined as Reserves / (short-term debt at residual maturity + current account deficit).
  - Reserve coverage reduced external vulnerabilities at low coverage levels; marginal gains largely negligible at very high reserve levels.
- Sample regressions (Figure 7 illustrations):
  - Full sample: y = 3.37ln(x) - 21.254, R² = 0.25
  - Excluding outliers: y = 3.66ln(x) - 22.399, R² = 0.2199
- Interpretation relative to other studies:
  - Blanchard and others (2010) emphasize short-term debt; this paper finds reserves significant at low reserve levels and demonstrates non-linearity whereby short-term debt magnifies output losses for high-reserve countries.

### V. Reserve-coverage regression highlights (selected coefficients)
- Table 3 — Full sample (column 1):
  - Reserves/GDP: 0.02 (0.07)
  - Short-term Debt/GDP: -0.32 *** (0.08)
  - Current account balance (percent of GDP): 0.04 (0.12)
  - Observations: 40; R-squared: 0.50
- Split by Reserves/Short-term Debt:
  - <100 (column 2): Reserves/GDP 0.60 ** (0.07); Short-term Debt/GDP 0.00 (0.23); Current account balance 0.87 *** (0.26); Observations: 14; R-squared: 0.74
  - >100<150 (column 3): Reserves/GDP -0.02 (0.05); Short-term Debt/GDP -0.31 *** (0.11); Current account balance -0.09 (0.08); Observations: 26; R-squared: 0.35
  - >150 (column 4): Reserves/GDP 0.45 *** (0.15); Short-term Debt/GDP -0.23 (0.14); Current account balance 0.48 * (0.24); Observations: 17; R-squared: 0.59
- Appendix 3 robustness tables (selected patterns):
  - Table A1 (Peak to Trough Growth): Reserves/GDP positive and significant in <100 samples (e.g., 0.645*** (0.149) in column (1)); negative/insignificant in >100 samples (e.g., -0.066 (0.058) in column (2)).
  - Table A2 (Deviation from trend, growth_4): Reserves/GDP 0.956*** (0.179) in column (1) and -0.141** (0.060) in column (2), illustrating non-linearity across splits.
  - Table A3 (Growth 2008Q4–2009Q1): Reserves/GDP 0.387** (0.134) in column (1); -0.039 (0.058) in column (2).
  - Table A4 (growth_4, All EMs): Reserves/GDP 0.027 (0.080) in column (1); Reserves/GDP 0.860*** (0.227) in <100 sample (column (2)); Reserves/GDP -0.043 (0.065) in >100 sample (column (3)).
  - Table A5 (growth_1, All EMs): Reserves/GDP 0.008 (0.058) in column (1); Reserves/GDP 0.377** (0.164) in <100 sample (column (2)).

### VI. Reserve usage and numerical examples
- Average reserve losses:
  - On average, countries lost around 7 percent of their GDP equivalent of international reserves (peak to trough) to protect currencies or balance sheets (Figure 9).
- Notable country example:
  - Russia spent more than $200 billion of its reserves, representing 13 percent of 2008 GDP.

### VII. Impact on sovereign spreads and credit
- Sovereign spreads:
  - External vulnerability, Spring 2007 coefficient 2.20 ** (1.05) in Table 4, column 1 — the most externally vulnerable country experienced about 220 basis points greater widening in spreads than the least vulnerable.
  - Reserve cover of ST debt at RM and CA deficit: -0.93 ** (0.40) (Table 4, column 2), indicating mitigation of spread increases.
  - Change in CPI 2003–2007: 0.13 *** (0.02) in column 1; 0.11 *** (0.02) in column 2.
  - Inflation targeters: -1.24 (0.75) in column 1; -1.80 *** (0.63) in column 2.
- Credit developments:
  - Pre-crisis credit booms—often foreign-financed—tended to end in credit and output busts.
  - Cross-border BIS claims on EMs ranged up to around 37 percent of GDP; average ~7 percent of GDP.
  - A country with double the average BIS claims experienced an additional 1¼ percentage points in output reduction (Table 2a).
  - Determinants of peak-to-trough real credit growth (Table 5, selected):
    - 2007 credit to GDP ratio: -0.21 *** (-0.03) in column 1; -0.004 *** (-0.001) in column 2.
    - Change in money market rate from August 2008 to peak: -0.59 *** (-0.18) in column 1; 0.004 (-0.009) in column 2.
  - Observations: 37 (column 1), 40 (column 2); R-squared: 0.76 and 0.22 respectively.

### VIII. Banking sector risks and heterogeneity
- Banking sector risks:
  - Greater initial fiscal vulnerabilities and dried-up interbank liquidity associated with higher banking sector risks (measured by change in Moody’s EDF from trough to peak).
  - Example relationships (Figure 12):
    - Change in EDF ~ Public sector vulnerability index: y = 6.2721x + 0.7041, R² = 0.3683.
    - Change in EDF ~ Average money market rate, 2008Q3–2009Q2: y = 0.3162x - 0.1857, R² = 0.3705.
- Country heterogeneity:
  - Empirical framework explains about half of cross-country variation; substantial country-specific variability remains.
  - Philippines: actual fall from peak to trough 2¼ percent; counterfactual based on 2005 vulnerabilities would have implied a 4 percent fall.
  - Russia: large output fall ~11 percentage points from peak to trough driven by oil price collapse, terms-of-trade shock, capital outflows, and credit crunch.
  - Indonesia and other large domestic-market economies (China, Egypt, India) had smaller adverse impacts due to large domestic demand shares (Indonesia: around 90 percent of real GDP in 2007).

### IX. Policy-relevant syntheses and implications
- Summary of policy-relevant findings:
  - EMs with smaller initial vulnerabilities entered recession later, exited earlier, and suffered considerably smaller declines in output during the first stage of the crisis.
  - Stronger external linkages—greater dependence on AE demand or larger exposure to foreign bank claims—led to sharper falls in output.
  - Pre-crisis credit booms (often foreign-financed and more pronounced under fixed exchange rate regimes) were associated with sharper output falls.
  - Reserves helped dampen crisis impact up to a limit:
    - Higher pre-crisis reserve cover associated with less deterioration in sovereign spreads and output.
    - Diminishing returns: little additional benefit for reserves in excess of short-term debt plus current account deficit.
- Areas for further research identified in the source:
  - Determinants of the fall in stock market prices.
  - The role of capital flows and their impact on growth during the crisis.

*Source: IMF staff analysis in "Appendix 3. Reserve Robustness with Alternative Control Variables" (excerpted from _wp10237).*

### References.................................................................................................25

### _wp10237 - References.................................................................................................25

### Box
- Box 1. Assessments of Underlying Vulnerabilities in Emerging Market Countries ...............27

### Tables
- 1a. Impact of the crisis 5
- 1b. Correlations - Real and financial variables 7
- 2a. Regressions for percent change in real output between peak and trough (growth_3)   14
- 2b. Regressions for percent change in real output between peak and trough (growth_4)   14
- 3. Dependent variable: Real output peak to trough (growth_3) 17
- 4. Determinants of the change in spreads from trough to peak (in percent) 19
- 5. Determinants of peak-to-trough real credit growth 20

### Figures
- 1. Pre-crisis external vulnerabilities 4
- 2. Median stock market indices 8
- 3. Capital flows to emerging markets 9
- 4. Impact of the crisis on output 10
- 5. Financial and corporate vulnerabilities 11
- 6. Alternative growth metrics 12
- 7. Reserve coverage and output collapse 15
- 8. Initial reserves and reserve use 17
- 9. Fall in reserves (peak to trough, percent of 2008 GDP) 18
- 10. Credit developments 21
- 11. Deleveraging 21
- 12. Banks' probability of default 22
- 13. Impact of crisis: Country variability in outcomes 23

### Appendices
- Appendix 1. Country sample ...................................................................................................28
- Appendix 2. List of variables 29

*Source: _wp10237 - References.................................................................................................25; https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10237.pdf*

### Appendix 3. Reserve Robustness with Alternative Control Variables 30

### Appendix 3. Reserve Robustness with Alternative Control Variables 30

### I. Introduction
- Context and sample:
  - Global growth fell by 6 percentage points from its pre-crisis peak to its trough in 2009, "the largest straight fall in global growth in the post-war era."
  - Real output in EMs fell about 4 percent between 2008Q3 and 2009Q1.
  - Sample: around 50 emerging market economies (EMs).
- Purpose:
  - Explore channels and factors shaping heterogeneous crisis impact across EMs.
  - Focus on external sector vulnerabilities prior to the crisis: current account deficits, reserve holdings, external debt levels, and an IMF staff external vulnerability index.
- Primary findings:
  - Countries more open to trade and financial linkages were more affected by the crisis.
  - Countries with better pre-crisis fundamentals and reduced vulnerabilities experienced less severe output contractions and less widening of sovereign spreads.
  - Higher international reserves holdings helped buffer the impact of the crisis, but with diminishing returns: "at very high levels of reserves there is little discernable evidence of their moderating impact on output collapse."
- Novelty/contributions:
  - First paper to find a significant and robust non-linear relationship between EMs’ reserve holdings and the actual decline in growth experienced during the crisis.
  - Uses a preferred impact measure estimated as the country-specific peak-to-trough decline in quarterly real GDP (captures heterogeneity in timing across countries).
  - Systematic analysis of crisis impact on financial variables: sovereign spreads and credit growth.

### II. A broad view of the impact of the crisis
- Aggregate and distributional outcomes (Table 1a, measured as percent change from peak to trough unless otherwise noted):
  - Output collapse 1/
    - Median-4.9-4.5
    - 25th percentile-8.4-6.6
    - 75th percentile-2.0-2.9
  - Stock market collapse 1/
    - Median-57.1-55.4
    - 25th percentile-72.0-64.1
    - 75th percentile-45.2-49.0
  - Rise in sovereign spreads 2/
    - Median462465
    - 25th percentile287...
    - 75th percentile772...
  - Notes: 1/ Measured as percent change from peak to trough. 2/ Measured as increase in basis points from trough to peak. For AEs, table reports rise in spreads on US corporates rated BBB.
- Stylized phases of financial transmission (median stock indices, August 9, 2007 = 100):
  - Decoupling (start to pre-Lehman): low & medium pre-crisis vulnerability EMs ≈ 15 percent below Aug-2007; AEs and high vulnerability EMs ≈ 30 percent below.
  - Re-coupling (Lehman collapse): widespread uniform fall across EMs.
  - Re-decoupling (post-stabilization): gap emerges—high vulnerability EMs remain depressed; low & medium recover toward pre-crisis levels.
- Capital flows:
  - Gross portfolio flows fell in 2008Q4 for all EMs; from 2009Q1 low/medium vulnerability EMs began to see positive flows while high vulnerability EMs continued gross outflows.
  - Regional contrast: Emerging Europe saw flows only return from 2009Q2; Asia experienced little decline in flows.
- Timing and magnitude of real activity collapse:
  - By 2008Q3 the majority of high or medium pre-crisis vulnerability countries were contracting; low vulnerability EMs held out longer.
  - Magnitude: low and medium vulnerability EMs suffered much smaller real output collapses and recovered faster than high vulnerability EMs and AEs.
- Comparative note with past crises:
  - Highly vulnerable EMs experienced a smaller initial fall in output during this crisis than in past EM crises.
  - Contributing factors: global coordinated response, rapid financing by international institutions (including IMF), absence of widespread banking or currency crises in many EMs.
  - For weak-fundamental countries, the crisis could be more protracted with U-shaped recoveries.

### III. Econometric analysis

#### A. Measuring impact: Alternative metrics
- Four alternative measures of output loss used to test robustness:
  - Growth 1: percent change in seasonally adjusted real GDP between 2008Q3 and 2009Q1 (applies equal timing to all countries).
  - Growth 2: log difference between actual seasonally adjusted real GDP in 2009Q1 and a counterfactual projection for 2009Q1 based on a linear trend estimated over 2003-2007.
  - Growth 3: country-specific peak-to-trough percent change in quarterly seasonally adjusted real GDP (preferred measure; unique to this study).
  - Growth 4: log difference between actual seasonally adjusted real GDP at the country’s trough and projected real GDP based on a linear trend estimated over 2003−07.
- Rationale for preferring Growth 3:
  - Captures heterogeneity in timing of peaks and troughs across countries.
  - More accurate and easier to interpret than imposing common peak-trough dates.
  - Growth 3 is highly correlated with Growth 1 and Growth 4; correlation strongest with Growth 4.
- Comparable peak-to-trough measures estimated for EMBI spreads and credit growth.

#### B. Impact on growth (regression framework and key controls)
- Primary regression specification explains fall in real output as function of:
  - Pre-crisis vulnerabilities (external vulnerability index, VEE Spring 2007).
  - Trade connectedness (percent change in domestic demand of AE trading partners weighted by trade shares, computed over a period similar to each EM’s peak-to-trough change).
  - International financial integration (consolidated stock of claims of BIS reporting banks on EMs, immediate borrower basis, in percent of the EM’s GDP in December 2007).
- Other controls tested (regional dummies, exchange rate regime, alternate indicators) did not affect central findings.
- Empirical conclusions (summarized from preceding sections and regression setup):
  - Pre-crisis external vulnerabilities are a significant determinant of output collapse, controlling for trade and financial linkages.
  - Trade and financial openness amplified crisis impact.
  - Higher reserve holdings reduced external vulnerability and moderated output collapse, but with diminishing returns at very high reserve levels.

*Source: IMF staff analysis in "Appendix 3. Reserve Robustness with Alternative Control Variables" (excerpts provided).*

### Appendix 2).

### Appendix 2

### C. Determinants of Output Collapse — Key Findings
- The least vulnerable EMs, on average, contracted 6½ percentage points less than the most vulnerable EMs (Table 2a).
- When using the alternative impact measure (“growth 4”), the least vulnerable EMs, on average, contracted 7½ percentage points less than the most vulnerable EMs (Table 2b).
- Trade linkages: EMs experienced an additional 1½ percentage point reduction in real output during the crisis for every percentage point fall in domestic demand in their advanced economy trading partners.
- External vulnerability index (individual factor results, Table 2a and 2b):
  - External sector vulnerability index (ranged 0 - 1): coefficient -6.40 ** (standard error (3.04)) in Table 2a; -7.64 ** (3.67) in Table 2b.
- Other sectoral vulnerability indicators—fiscal, financial, corporate—do not stand out as significant factors in explaining the output decline.
- Foreign bank claims (percent of GDP, expressed in logs) show negative associations with output:
  - Table 2a (selected columns): -1.7 * (0.92), -1.94 ** (0.81), -1.85 ** (0.83), -1.92 * (1.00).
  - Table 2b (selected columns): -1.89 * (1.11), -2.30 ** (0.99), -1.87 * (0.96), -2.10 * (1.16).
- Current account balance and external debt findings:
  - Current account balance (percent of GDP): 0.17 * (0.08) in Table 2a; 0.24 ** (0.10) in Table 2b.
  - External debt (percent of GDP): -0.09 ** (0.04) in Table 2a; -0.11 ** (0.04) in Table 2b.
  - External debt (percent of exports): not statistically significant in reported columns (e.g., -2.40 (1.59) and -3.01 (1.88)).

### C. The role of reserves — Findings and Non-linearity
- Pre-crisis reserve holdings were associated with positive but diminishing returns in mitigating output collapse.
- Reserve coverage defined as Reserves / (short-term debt at residual maturity + current account deficit) reduced external vulnerabilities at low levels of coverage, but marginal gains became largely negligible at very high levels of reserves.
- Figure 7 sample regressions (reserve coverage and output collapse):
  - For full sample: y = 3.37ln(x) - 21.254, R² = 0.25
  - After excluding outliers: y = 3.66ln(x) - 22.399, R² = 0.2199
- Robustness and interpretation:
  - Blanchard and others (2010) find short-term debt important; this paper finds reserves significant at low reserve levels but shows non-linearity: for high-reserve countries short-term debt magnifies output losses.
  - Results robust to controls including pre-crisis (2007) per capita GDP, GDP adjusted for PPP, pre-crisis credit growth (2003-2007), size of credit to GDP ratio, and size of domestic demand relative to GDP.
  - Alternative growth measures (“growth1”, “growth 4”) do not change results; “growth 4” intensifies significance of positive linear relationship between growth and reserves for low reserve coverage.

### Reserve-coverage Regression Results (Table 3) — Selected coefficients
- Full sample (column 1):
  - Reserves/GDP: 0.02 (0.07)
  - Short-term Debt/GDP: -0.32 *** (0.08)
  - Current account balance (percent of GDP): 0.04 (0.12)
  - Observations: 40; R-squared: 0.50
- Split by Reserves/Short-term Debt:
  - <100 (column 2): Reserves/GDP 0.60 ** (0.07); Short-term Debt/GDP 0.00 (0.23); Current account balance 0.87 *** (0.26); Observations: 14; R-squared: 0.74
  - >100<150 (column 3): Reserves/GDP -0.02 (0.05); Short-term Debt/GDP -0.31 *** (0.11); Current account balance -0.09 (0.08); Observations: 26; R-squared: 0.35
  - >150 (column 4): Reserves/GDP 0.45 *** (0.15); Short-term Debt/GDP -0.23 (0.14); Current account balance 0.48 * (0.24); Observations: 17; R-squared: 0.59
  - <median and >median splits reported with respective coefficients and R-squared values (see Table 3).

### Reserve usage during crisis
- Countries with more reserves going into the crisis made greater use of them to avoid sharp depreciations and protect balance sheets.
- On average, countries lost around 7 percent of their GDP equivalent of international reserves (peak to trough) either to protect the currency or the balance sheets (Figure 9).
- Example: Russia spent more than $200 billion of its reserves (representing 13 percent of 2008 GDP).

### D. Impact on Sovereign Spreads
- Pre-crisis external vulnerabilities help explain the rise in sovereign spreads during the crisis.
  - The country considered most externally vulnerable in Spring 2007 experienced a greater widening in spreads than the least vulnerable by around 220 basis points (Table 4, column 1: External vulnerability, Spring 2007 coefficient 2.20 ** (1.05)).
- Reserve coverage mitigated spread increases:
  - Reserve cover of ST debt at RM and CA deficit: -0.93 ** (0.40) (Table 4, column 2).
- Other determinants of spreads:
  - Change in CPI from 2003 to 2007 (inflation in 2003-2007 period): 0.13 *** (0.02) in column 1; 0.11 *** (0.02) in column 2.
  - Inflation targeters: -1.24 (0.75) in column 1; -1.80 *** (0.63) in column 2.
- Pre-crisis fiscal vulnerabilities do not have a significant impact on change in spreads in these regressions.

### E. Impact on Credit
- Pre-crisis credit booms—often funded from abroad—generally ended in credit and output busts.
- Cross-border claims of BIS reporting banks on EMs on the eve of the crisis ranged from close to zero to around 37 percent of GDP across countries in the sample.
- A country with double the average level of BIS claims (average ~7 percent of GDP) experienced an additional 1¼ percentage points in output reduction (Table 2a).
- Determinants of peak-to-trough real credit growth (Table 5, selected):
  - 2007 credit to GDP ratio in percent of 2003 ratio: coefficient -0.21 *** (-0.03) in column 1; -0.004 *** (-0.001) in column 2.
  - Change in money market rate from August 2008 to peak: -0.59 *** (-0.18) in column 1; 0.004 (-0.009) in column 2.
  - Observations: 37 (column 1), 40 (column 2); R-squared: 0.76 and 0.22 respectively.
- Credit busts were associated with sharp increases in money market rates (symptom of credit crunch).
- Regional patterns: Emerging Europe experienced particularly pronounced credit contractions due to heavy cross-border lending and global deleveraging; countries funded primarily by domestic deposits were better able to sustain credit growth.

### Banking sector risks
- Countries with greater initial fiscal vulnerabilities and where interbank liquidity dried up experienced higher banking sector risks (measured by change in Moody’s EDF from trough to peak).
- Relationship examples (Figure 12):
  - Change in EDF, trough to peak ~ positively related to Public sector vulnerability index (y = 6.2721x + 0.7041, R² = 0.3683).
  - Change in EDF, trough to peak ~ positively related to Average money market rate, 2008Q3-2009Q2 (y = 0.3162x - 0.1857, R² = 0.3705).

### F. Further issues and Country Heterogeneity
- The empirical framework explains about half of observed cross-country variation in outcomes; important country-specific variability remains.
- Examples of heterogeneity:
  - Philippines: had a smaller output fall (actual fall from peak to trough of 2¼ percent) than would have occurred had external vulnerabilities been as in 2005 (counterfactual output would have fallen by 4 percent based on Table 2 estimates). Contributing factors included remittances (~11 percent of GDP in 2009), lack of prior overheating, and strengthened financial sector.
  - Russia: large output fall (~11 percentage points of GDP from peak to trough) despite lower external vulnerabilities and foreign bank claims, driven by the collapse in oil prices, terms of trade shock, capital outflows, and credit crunch.
  - Indonesia and other large domestic-market economies (China, Egypt, India) experienced smaller adverse impacts because domestic demand formed large shares of aggregate demand (Indonesia: around 90 percent of real GDP in 2007).
- Size and openness variables when formally included did not perform well in regressions; simple measures may not capture contribution of external demand to domestic value added.

### IV. Conclusions — Policy-relevant Syntheses
- EMs with smaller initial vulnerabilities entered recession later, exited earlier, and suffered considerably smaller declines in output during the first stage of the crisis.
- Stronger external linkages—higher dependence on demand from AEs or larger exposure to foreign bank claims—led to sharper falls in output.
- Pre-crisis credit booms (often foreign-financed and more pronounced under fixed exchange rate regimes) were associated with sharper output falls, though less so than in past crises.
- Reserves helped dampen the crisis impact up to a limit:
  - Higher pre-crisis reserve cover associated with less deterioration in both sovereign spreads and output.
  - Effect exhibited diminishing returns: little additional benefit for reserves in excess of short-term debt plus current account deficit.
- Areas for future research identified:
  - Determinants of the fall in stock market prices (not covered here).
  - The role of capital flows and their impact on growth during the crisis.

*Source: Appendix 2 of the IMF working paper (_wp10237 - Appendix 2). Source materials: VEE Spring 2007; IFS; WEO; BIS; Haver; Bloomberg; Moody’s KMV; Fund staff calculations.*

### References

### _wp10237 - References

### References cited
- Abiad, A., R. Balakrishnan, P. Koeva Brooks, D. Leigh, I. Tytell, 2009, “What’s the Damage? Medium-term Output Dynamics after Banking Crises,” IMF Working Paper No. 09/245 (Washington: International Monetary Fund).
- Aisen, A., and M. Franken, 2010, “Bank Credit during the 2008 Financial Crisis: A Cross-Country Comparison,” IMF Working Paper No. 10/47 (Washington: International Monetary Fund).
- Bakker, Bas Berend and Anne-Marie Gulde, 2010, “The Credit Boom in the EU New Member States: Bad Luck or Bad Policies?” IMF Working Paper No. 10/130, (Washington: International Monetary Fund).
- Bank for International Settlements, 2009a, Quarterly Review-International Banking and Financial Market Developments (Basel).
- Bank for International Settlements, 2009b, Capital Flows and Emerging Market Economies, CGFS Paper No. 33 (Basel).
- Berglӧf, E., 2009, Transition Report 2009: Transition in Crisis? (London: European Bank for Reconstruction and Development).
- Berkmen, P., G. Gelos, R. Rennhack, and J. Walsh, 2009, “The Global Financial Crisis: Explaining Cross-Country Differences in the Output Impact,” IMF Working Paper No.09/280 (Washington: International Monetary Fund).
- Blanchard, O., H. Faruqee, and M. Das, 2010, “The Initial Impact of the Crisis on Emerging Market Countries,” Brookings Papers on Economic Activity (Washington: Brookings Institution).
- Cecchetti, S., 2008, “Monetary Policy and the Financial Crisis of 2007−08,” CEPR Policy Insight No. 21 (London: Center for Economic Policy Research).
- Freund, C., 2009, “The Trade Response to Global Downturns: Historical Evidence,” Policy Research Working Paper Series No. 5015 (Washington: The World Bank).
- International Monetary Fund, 2009a, World Economic Outlook: From Recession to Recovery: How soon and How Strong (Washington).
- International Monetary Fund, 2009b, World Economic Outlook: What’s the Damage? Medium-term Output Dynamics after Financial Crises (Washington).
- International Monetary Fund, 2010, How Did Emerging Markets Cope in the Crisis? (Washington).
- Kamil, H., and K. Rai, 2010, “The Global Credit Crunch and Foreign Banks’ Lending to Emerging Markets: Why Did Latin America Fare Better?” IMF Working Paper No.10/102 (Washington: International Monetary Fund).
- Lane, Philip R., and G. M. Milesi-Ferretti, 2010, “The Cross-Country Incidence of the Global Crisis” IMF Working Paper No.10/171 (Washington: International Monetary Fund).
- Ramakrishnan, U., and J. Zalduendo, 2006, “The Role of the IMF Support in Crisis Prevention,” IMF Working Paper No. 06/75 (Washington: International Monetary Fund).
- Reinhart, C., and K. Rogoff, 2008, “This Time is Different: A Panoramic View of Eight Centuries of Financial Crises,” NBER Working Paper No. 13882 (Cambridge: National Bureau of Economic Research).
- Rodrik, D., 2006, “The Social Cost of Foreign Exchange Reserves,” NBER Working Paper No. 11952 (Cambridge: The National Bureau of Economic Research).
- Rose, A., and M. Spiegel, 2009a, “Cross-Country Causes and Consequences of the 2008 Crisis: Early Warning,” NBER Working Paper No. 15357 (Cambridge: National Bureau of Economic Research).
- Rose, A., and M. Spiegel, 2009b, “Cross-Country Causes and Consequences of the 2008 Crisis: International Linkages and American Exposure,” NBER Working Paper No. 15358 (Cambridge: National Bureau of Economic Research).
- Taylor, J., and J. Williams, 2008, “A Black Swan in the Money Market,” NBER Working Paper Series No. 13943 (Cambridge: National Bureau of Economic Research).
- Tong, H., and S. Wei, 2009, “The Composition Matters: Capital Inflows and Liquidity Crunch During a Global Economic Crisis,” NBER Working Paper No. 15207 (Cambridge: National Bureau of Economic Research).
- Zettelmeyer, J., E. Berglöf, Y. Kormiyenko, and A. Plekhanov, 2009, “Understanding the Crisis in Emerging Europe” (London: European Bank for Reconstruction and Development).

### Box 1 — Assessments of Underlying Vulnerabilities in Emerging Market Countries
- Purpose:
  - Provide staff assessments of pre-crisis vulnerabilities for emerging market countries with access to international capital markets.
- History and updates:
  - Exercise established in 2001 and updated since.
  - Assessments conducted semi-annually.
- Methodology steps:
  - Compile data on vulnerabilities—flow and stock measures—derived from previous IMF studies and academic literature on empirical early warning system models for external, fiscal, corporate, and financial sectors.
  - Compare each indicator against a database of realized capital account crises to derive thresholds that minimize combined percentages of missed calls and false alarms.
  - Use differences in discriminatory powers (minimum sum of errors) to guide weights for each indicator in the index for each sector.
  - Combine sector indices into an aggregate index using judgment-based sector weights.
  - Vet analysis in area departments and apply country-specific considerations for final assessments.
- Rationale and usage:
  - Indices provide a summary statistic of vulnerabilities and address high correlations among variables that preclude simultaneous inclusion in regression analyses.
  - A 2007 paper updating the VEE methodology generally uses the model-based external vulnerability index measure and allocates external sector indicators the highest weight: 45 percent.

### Appendix 1 — Country Sample
- Coverage:
  - 57 economies are considered.
  - No strict ex ante qualification or exclusion criteria; sample subject to data availability.
  - Some economies classified as Advanced Economies (AEs) in the World Economic Outlook are included if present in S&P IFCI or S&P Frontier BMI emerging market stock indices.
  - Country samples differ by regression depending on data availability.
- Data sources:
  - Staff calculations using data from the Spring 2010 World Economic Outlook and internal staff assessment of vulnerabilities.
  - Additional public sources: BIS, CEIC, Haver, IFS, Moody's, and others.
- Country lists by region:
  - Asia (11): China, India, Indonesia, Korea, Malaysia, Mongolia, Pakistan, Philippines, Sri Lanka, Thailand, Vietnam
  - Emerging Europe and CIS (22): Albania, Belarus, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Georgia, Hungary, Kazakhstan, Latvia, Lithuania, Macedonia, Montenegro, Poland, Romania, Russia, Serbia, Slovak Republic, Slovenia, Turkey, Ukraine
  - Western Hemisphere (17): Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Jamaica, Mexico, Panama, Paraguay, Peru, Uruguay, Venezuela
  - Middle East and Africa (7): Egypt, Israel, Jordan, Lebanon, Morocco, South Africa, Tunisia

### Appendix 2 — List of variables
- Pre-Crisis Policy Fundamentals (2007):
  - 1 Exchange rate regime, MCM classification, March-07
  - 2 FX regime, Reinhart Rogoff classification
  - 3 Inflation targeting framework
  - 4 Debt stabilizing primary balance
  - 5 Primary gap
  - 6 External public sector debt
  - 7 Short-term public debt at residual maturity
  - 8 Public sector debt linked to FX
  - 9 Primary balance, % GDP
  - 10 Cyclically adjusted primary balance
  - 11 External debt, total, % GDP
  - 12 External debt, ST, % GDP
  - 13 External debt, ST at RM, % GDP
  - 14 General gov't debt to GDP
  - 15 Fiscal impulse
  - 16 Change in primary balance to GDP
  - 17 Financial connectedness (Foreign assets + liabilities)/GDP
- Measures of pre-crisis overheating:
  - 1 Real GDP growth between 2003 & 2007
  - 2 Real domestic credit growth between 2003 & 2007
  - 3 Percent change in CPI between 2003 & 2007
  - 4 Credit to GDP 2007
  - 5 Regional dummies
- Measures of trade linkages:
  - Exports to advanced economies
  - Exports to US, % GDP
  - Manufactures exports
  - Openness (X+M)/GDP
  - Export earnings: non-fuel primary commodities
  - Oil exports, % GDP
  - Fuel exporter dummy
- Measures of financial linkages:
  - Total external financing requirements, % GDP
  - Foreign currency loans (% of total loans)
  - Loan to deposits ratio
  - Claims on private sector, % GDP
  - Total capital inflows
  - Foreign ownership in % of total assets 2007
- Other controls:
  - Population
  - Per capita GDP
  - PPP valuation of country GDP
  - NEER peak to trough percent change
- Notes:
  - Each one of these indicators was tried in addition to the three core indicators mentioned in Table 2 in the text to check the robustness of results presented in the text.
  - See also Berkmen and others (2009) for a further list of possible explanatory variables.

*Source: _wp10237 - References*

### Appendix 3 – Reserve Robustness with Alternative Control Variables  Table A1 - Dependent Variable: Pe

### _wp10237 - Appendix 3 – Reserve Robustness with Alternative Control Variables  Table A1 - Dependent Variable: Pe

### Table A1 — Dependent Variable: Peak to Trough Growth (percent)
- Regression columns: (1) <100, (2) >100, (3) <100, (4) >100, (5) <100, (6) >100, (7) <100, (8) >100, (9) <100, (10) >100.
- Reserves/GDP coefficients (standard errors):
  - (1) 0.645*** (0.149)
  - (2) -0.066 (0.058)
  - (3) 0.710*** (0.158)
  - (4) -0.003 (0.062)
  - (5) 0.640** (0.200)
  - (6) -0.008 (0.059)
  - (7) 0.605*** (0.142)
  - (8) -0.046 (0.041)
  - (9) 0.683*** (0.190)
  - (10) -0.012 (0.059)
- Short-term Debt/GDP coefficients (standard errors):
  - (1) -0.136 (0.292)
  - (2) -0.315*** (0.092)
  - (3) -0.197 (0.178)
  - (4) -0.371*** (0.102)
  - (5) -0.108 (0.203)
  - (6) -0.361*** (0.104)
  - (7) -0.010 (0.214)
  - (8) -0.088 (0.121)
  - (9) -0.170 (0.193)
  - (10) -0.383*** (0.103)
- Current Account/GDP coefficients (standard errors):
  - (1) 0.834** (0.324)
  - (2) -0.001 (0.079)
  - (3) 0.581* (0.287)
  - (4) -0.033 (0.146)
  - (5) 0.825** (0.297)
  - (6) -0.040 (0.105)
  - (7) 0.966** (0.302)
  - (8) -0.007 (0.064)
  - (9) 0.857** (0.355)
  - (10) -0.007 (0.079)
- Other included controls (selected columns):
  - Credit/GDP: (3) not reported except (3) shows 0.035* (0.019) in one placement.
  - Dom demand/GDP: appears in some specifications: -0.248 (0.204) and -0.017 (0.073).
  - Ln PPP GDP: 1.000 (2.157) and 0.243 (0.657) in a specification.
  - Ln per capita GDP: -4.375 (2.859) and -2.920*** (0.782).
  - Credit growth: 0.002 (0.010) and -0.007*** (0.002).
- Constant terms (standard errors):
  - (1) -8.075* (3.967)
  - (2) -0.957 (2.066)
  - (3) 14.848 (20.288)
  - (4) 2.042 (7.173)
  - (5) -15.162 (11.754)
  - (6) -1.115 (3.398)
  - (7) 29.171 (25.141)
  - (8) 23.141*** (5.887)
  - (9) -9.573 (5.455)
  - (10) 1.258 (2.129)
- Sample and fit:
  - Observations: 14 (columns with <100) and 26 (columns with >100) across columns.
  - R-squared: 0.701, 0.368, 0.721, 0.298, 0.708, 0.303, 0.756, 0.551, 0.700, 0.366.
- Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

### Table A2 — Dependent Variable: Deviation of Peak to Trough Growth from Pre-Crisis Trend Growth (percent)
- Reserves/GDP coefficients (standard errors):
  - (1) 0.956*** (0.179)
  - (2) -0.141** (0.060)
  - (3) 1.007*** (0.201)
  - (4) -0.031 (0.073)
  - (5) 0.901*** (0.239)
  - (6) -0.037 (0.068)
  - (7) 0.907*** (0.183)
  - (8) -0.074 (0.051)
  - (9) 0.972*** (0.231)
  - (10) -0.050 (0.058)
- Short-term Debt/GDP coefficients (standard errors):
  - (1) -0.243 (0.349)
  - (2) -0.353** (0.127)
  - (3) -0.286 (0.219)
  - (4) -0.456*** (0.141)
  - (5) -0.127 (0.199)
  - (6) -0.437*** (0.139)
  - (7) -0.108 (0.254)
  - (8) -0.187 (0.174)
  - (9) -0.258 (0.225)
  - (10) -0.478*** (0.130)
- Current Account/GDP coefficients (standard errors):
  - (1) 1.044** (0.382)
  - (2) -0.003 (0.085)
  - (3) 0.798* (0.356)
  - (4) -0.078 (0.206)
  - (5) 1.032** (0.337)
  - (6) 0.003 (0.162)
  - (7) 1.171** (0.364)
  - (8) -0.015 (0.109)
  - (9) 1.049** (0.422)
  - (10) -0.013 (0.077)
- Other controls and coefficients (selected):
  - Credit/GDP: -0.011 (0.111) and 0.060** (0.025) in some specs.
  - Dom demand/GDP: -0.243 (0.226) and -0.044 (0.083) in some specs.
  - Ln PPP GDP: 2.187 (2.329) and -0.173 (0.729) in a column.
  - Ln per capita GDP: -4.143 (3.048) and -2.645** (1.134).
  - Credit growth: 0.001 (0.011) and -0.014*** (0.003).
- Constant terms (standard errors):
  - (1) -14.956*** (4.304)
  - (2) -3.630 (2.804)
  - (3) 8.037 (22.940)
  - (4) 2.968 (7.579)
  - (5) -28.513** (11.719)
  - (6) -0.524 (4.674)
  - (7) 20.835 (26.732)
  - (8) 19.195** (8.333)
  - (9) -15.606** (6.575)
  - (10) 0.497 (2.539)
- Sample and fit:
  - Observations: 14 and 26 across columns.
  - R-squared: 0.739, 0.429, 0.752, 0.297, 0.770, 0.296, 0.771, 0.428, 0.738, 0.478.
- Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

### Table A3 — Dependent Variable: Growth in 2008Q4 – 2009Q1 (percent)
- Reserves/GDP coefficients (standard errors):
  - (1) 0.387** (0.134)
  - (2) -0.039 (0.058)
  - (3) 0.428** (0.133)
  - (4) -0.020 (0.053)
  - (5) 0.394** (0.160)
  - (6) 0.000 (0.056)
  - (7) 0.352** (0.123)
  - (8) -0.032 (0.042)
  - (9) 0.389** (0.148)
  - (10) -0.001 (0.054)
- Short-term Debt/GDP coefficients (standard errors):
  - (1) -0.114 (0.184)
  - (2) -0.321*** (0.082)
  - (3) -0.147 (0.131)
  - (4) -0.301*** (0.098)
  - (5) -0.124 (0.172)
  - (6) -0.353*** (0.095)
  - (7) -0.025 (0.151)
  - (8) -0.139 (0.130)
  - (9) -0.120 (0.137)
  - (10) -0.350*** (0.103)
- Current Account/GDP coefficients (standard errors):
  - (1) 0.596** (0.222)
  - (2) -0.097 (0.063)
  - (3) 0.367** (0.157)
  - (4) 0.043 (0.122)
  - (5) 0.595** (0.218)
  - (6) -0.076 (0.088)
  - (7) 0.677** (0.212)
  - (8) -0.100 (0.064)
  - (9) 0.587** (0.248)
  - (10) -0.103 (0.068)
- Other controls (selected):
  - Credit/GDP: -0.004 (0.057) and 0.022 (0.019) in some specs.
  - Dom demand/GDP: -0.227 (0.136) and 0.108 (0.077).
  - Ln PPP GDP: -0.062 (1.662) and -0.221 (0.637) in a specification.
  - Ln per capita GDP: -2.645 (2.175) and -2.267*** (0.797).
  - Credit growth: -0.001 (0.007) and 0.001 (0.002).
- Constant terms (standard errors):
  - (1) -4.763 (3.129)
  - (2) 0.325 (1.982)
  - (3) 17.068 (13.339)
  - (4) -9.915 (7.909)
  - (5) -4.597 (9.912)
  - (6) 2.378 (3.010)
  - (7) 18.239 (19.524)
  - (8) 18.865*** (5.652)
  - (9) -4.864 (3.584)
  - (10) 1.035 (2.162)
- Sample and fit:
  - Observations: 14 and 26 across columns.
  - R-squared: 0.700, 0.409, 0.734, 0.408, 0.700, 0.383, 0.738, 0.542, 0.700, 0.378.
- Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

### Table A4 — Dependent Variable: Deviation of Peak to Trough Growth from Pre-Crisis Trend Growth (growth_4)
- Samples/columns: (1) All EMs, (2) <100, (3) >100, (4) <150, (5) >150, (6) <median, (7) >median.
- Reserves/GDP coefficients (standard errors):
  - (1) 0.027 (0.080)
  - (2) 0.860*** (0.227)
  - (3) -0.043 (0.065)
  - (4) 0.689*** (0.170)
  - (5) -0.100** (0.038)
  - (6) 0.670* (0.360)
  - (7) 0.052 (0.177)
- Short-term Debt/GDP coefficients (standard errors):
  - (1) -0.424*** (0.111)
  - (2) 0.012 (0.186)
  - (3) -0.411*** (0.145)
  - (4) -0.373** (0.177)
  - (5) -0.878*** (0.110)
  - (6) -0.562*** (0.135)
  - (7) -0.443** (0.198)
- Current Account/GDP coefficients (standard errors):
  - (1) 0.059 (0.154)
  - (2) 1.101*** (0.277)
  - (3) -0.055 (0.141)
  - (4) 0.546* (0.314)
  - (5) 0.155*** (0.046)
  - (6) 0.007 (0.291)
  - (7) 0.072 (0.134)
- Log(Population):
  - (1) 0.358 (0.648)
  - (2) 3.154 (2.322)
  - (3) 0.304 (0.495)
- Constant terms (standard errors):
  - (1) -4.550 (3.640)
  - (2) -27.060*** (6.216)
  - (3) -2.762 (3.390)
  - (4) -13.045*** (3.797)
  - (5) 3.755** (1.341)
  - (6) -9.316* (4.389)
  - (7) -4.114 (7.209)
- Sample and fit:
  - Observations: 40, 14, 26, 23, 17, 19, 21.
  - R-squared: 0.493, 0.802, 0.302, 0.612, 0.828, 0.534, 0.497.
- Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

### Table A5 — Dependent Variable: Growth in 2008Q4 – 2009Q1 (growth_1)
- Samples/columns: (1) All EMs, (2) <100, (3) >100, (4) <150, (5) >150, (6) <median, (7) >median.
- Reserves/GDP coefficients (standard errors):
  - (1) 0.008 (0.058)
  - (2) 0.377** (0.164)
  - (3) -0.004 (0.054)
  - (4) 0.259** (0.119)
  - (5) -0.002 (0.043)
  - (6) 0.368 (0.249)
  - (7) 0.021 (0.103)
- Short-term Debt/GDP coefficients (standard errors):
  - (1) -0.309*** (0.059)
  - (2) -0.082 (0.211)
  - (3) -0.335*** (0.098)
  - (4) -0.138 (0.098)
  - (5) -0.677*** (0.124)
  - (6) -0.245*** (0.064)
  - (7) -0.410*** (0.101)
- Current Account/GDP coefficients (standard errors):
  - (1) -0.009 (0.094)
  - (2) 0.603** (0.210)
  - (3) -0.127* (0.066)
  - (4) 0.400** (0.156)
  - (5) -0.048 (0.071)
  - (6) 0.067 (0.101)
  - (7) -0.091 (0.081)
- Log(Population) coefficients (standard errors) where included:
  - (1) -0.071 (0.481)
  - (3) 0.445 (1.829)
  - (5) 0.201 (0.429)
- Constant terms (standard errors):
  - (1) 0.903 (1.886)
  - (2) -6.598 (7.002)
  - (3) 0.289 (1.971)
  - (4) -4.473* (2.487)
  - (5) 3.396** (1.275)
  - (6) -3.992 (3.250)
  - (7) 1.091 (3.941)
- Sample and fit:
  - Observations: 40, 14, 26, 23, 17, 19, 21.
  - R-squared: 0.488, 0.703, 0.383, 0.599, 0.725, 0.417, 0.621.
- Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

*Appendix 3 tables and regression outputs as presented in the source PDF.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10237.pdf_