## _wp07114

## Source details

**Canonical URL:** [_wp07114](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07114.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07114.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07114.pdf.json)

---

### Overview and motivation
- Recent research highlights adverse effects of high output volatility on economic growth, welfare, and poverty.
- Declining volatility documented in industrial and developing countries in recent decades; some developing countries experienced large welfare losses from extreme volatility—large output drops—in the 1980s and 1990s.
- Paper combines two strands:
  - international business cycles using latent factor models;
  - cross-country regression analysis of determinants of volatility and large output drops.

### Data coverage and methods
- Sample: 87 countries including 67 emerging market and developing countries.
- Time periods: 1970–2004, subperiods 1970–86 and 1987–2004 (some regressions reference 1970–2003).
- Methods:
  - Bayesian dynamic latent factor model decomposing output into global, regional, and country-specific components (building on Kose, Otrok, and Whiteman, 2003).
  - Cross-country regressions using volatility and worst drops of the country-specific component from the dynamic factor analysis as dependent variables.
- Output measure: per capita real GDP growth from Penn World Tables, Version 6.1 (1970–2000) extended to 2004 with WEO data.

### Key empirical findings
- Volatility and severity of worst output drops have declined in most emerging market and developing countries in recent years, but remain well above industrial countries.
- Large regional differences:
  - Sub-Saharan Africa shows much higher output volatility and severity of drops than Asia.
  - South Asia and China, MENA, and CFA franc zone countries: sustained decline in volatility.
  - Latin America: volatility remained constant at a relatively high level.
  - East Asia: volatility increased since 1997.
- Size of worst output drops highly correlated with overall volatilities.
- Decomposition summary:
  - Significant global component exists, but country-specific factors play a greater role in emerging market and developing countries than in industrial countries.
  - Declines in volatility over 1987–2003 mainly due to less volatile country-specific factors rather than increased stability of global/regional factors.
  - Worst output drops in emerging market and developing countries have been mainly associated with country-specific events; their contribution increased in 1987–2003 versus 1970–86.
  - For industrial countries, worst drops in 1970–86 were largely global-driven; regional factor became more influential in 1987–2003.
- Regression evidence:
  - Volatility of discretionary government spending, terms of trade volatility, and the interaction of terms of trade volatility with exchange regime flexibility are key determinants of output volatility and worst drops.
  - Volatility in discretionary fiscal spending influences output volatility and the size of worst drops; effect larger in IV than OLS.
  - Terms of trade volatility raises output volatility and worst drops; interaction term implies fixed regimes amplify these effects compared with flexible regimes.
  - Exchange rate flexibility helps moderate worst drops and volatility driven by external shocks.
  - Fiscal procyclicality is not found significant for worst drops.
  - Financial sector development, trade openness, relative income, and capital account restrictions show mixed or sample-dependent effects.

### Contributions and extensions
- Integrates latent factor decomposition with cross-country regressions to identify drivers of volatility and large drops.
- Extends dynamic latent factor estimates to 1970–2004 and subperiods to analyze changes over time.
- Demonstrates that reductions in volatility in emerging market and developing countries are primarily due to less volatile country-specific factors.
- Links discretionary fiscal policy volatility to output volatility and severity of worst drops.
- Provides evidence that exchange rate flexibility mitigates impact of terms-of-trade shocks on volatility and worst drops.

### Decomposition: fraction of variance in real per capita output growth attributable to each factor (averages for each group, in percent)
- Sub-Saharan Africa — CFA countries: Global 5.7 / Regional 18.2 / Country 76.1
- Sub-Saharan Africa — Non-CFA countries: Global 6.8 / Regional 10.2 / Country 82.1
- Middle East and North Africa: Global 3.8 / Regional 15.9 / Country 80.3
- Latin America: Global 12.6 / Regional 13.7 / Country 73.7
- South Asia and China: Global 15.6 / Regional 20.6 / Country 63.8
- East Asia: Global 11.0 / Regional 41.8 / Country 47.2
- East Asia (1970–96): Global 18.0 / Regional 15.8 / Country 66.3
- Developing countries (group average): Global 9.3 / Regional 16.9 / Country 73.8
- Industrial countries (group average): Global 24.3 / Regional 21.7 / Country 54.0
- Key pattern: country-specific factor explains more than 60 percent of output volatility in about 90 percent of emerging market and developing countries, compared with only 40 percent of industrial countries.

### Contribution of factors to the decline in volatility (1970–86 versus 1987–2004) — averages for countries with a decline in volatility (in percent)
- Sub-Saharan Africa — CFA countries: Global -40.0 / Regional -2.1 / Country -5.8 / (Total) -32.1
- Sub-Saharan Africa — Non-CFA countries: Global -53.2 / Regional -6.9 / Country -8.3 / (Total) -38.0
- Middle East and North Africa: Global -60.7 / Regional -0.5 / Country -17.4 / (Total) -42.7
- Latin America: Global -12.1 / Regional -3.4 / Country -2.8 / (Total) -5.9
- South Asia and China: Global -12.0 / Regional -0.9 / Country -2.9 / (Total) -8.2
- Developing countries (average): Global -34.2 / Regional -3.2 / Country -6.8 / (Total) -24.2
- Industrial countries: Global -4.1 / Regional -1.1 / Country -0.8 / (Total) -2.3
- Note: Table reports only countries where volatility declined.

### Contribution of factors to the worst drop in real per capita output growth (averages for each group; in percent) — 1970–86 → 1987–2003 (Global / Regional / Country)
- Sub-Saharan Africa — CFA countries: 4.9 / 27.7 / 67.5  → 2.6 / 8.9 / 88.5
- Sub-Saharan Africa — Non-CFA countries: -1.8 / 11.1 / 90.7 → -14.7 / -1.8 / 116.4
- Middle East and North Africa: 6.0 / 16.9 / 77.1 → 9.9 / 15.1 / 74.9
- Latin America: 24.1 / 11.5 / 64.5 → 1.2 / 10.2 / 88.6
- South Asia and China: 35.0 / 30.5 / 34.6 → -113.5 / 10.6 / 202.9
- East Asia: 50.0 / 120.5 / -70.5 → 7.9 / 104.5 / -12.4
- Emerging market and developing countries (aggregate): 15.0 / 27.5 / 57.6 → -9.6 / 18.5 / 91.1
- Industrial countries: 71.6 / 46.1 / -17.8 → 20.7 / 59.5 / 19.8
- Observation: in emerging market and developing regions the contribution of the country-specific factor to worst drops increased from 1970–86 to 1987–2003.

### Cross-section regressions: determinants and magnitudes
- Dependent variables: standard deviation of real per capita GDP growth (1970–2003) and worst drops (1970–2003); alternative specifications use country-specific component measures from dynamic factor model.
- Explanatory variables: volatility of discretionary government spending; fiscal procyclicality; terms of trade volatility; interaction of terms of trade volatility with exchange rate flexibility; relative income; trade openness; financial sector development; capital account restrictions.
- Main empirical magnitudes and examples:
  - A one standard deviation increase in terms of trade volatility implies a 1.5 percentage point increase in output volatility in the region with the least exchange rate flexibility (the CFA franc zone) compared with a 0.3 percentage point increase in the region with the greatest exchange rate flexibility (non-CFA countries).
  - From worst-drop IV regression of the country-component: a one standard deviation increase in terms of trade volatility implies a 5 percentage point increase in the worst output drop in the CFA franc zone compared with a 3 percentage point increase in sub-Saharan Africa’s non-CFA countries.
  - Simulation from IV estimates: reducing volatility of cyclically adjusted government spending to east Asia levels would reduce output volatility of the country-component by 3.8 percentage points for CFA franc zone countries and by 3.1 percentage points for sub-Saharan Africa’s non-CFA countries (equivalent to about 50 and 60 percent of the country-specific output volatility respectively).

### Estimation approach, instruments, and robustness
- Regressions use OLS and IV; IV instruments include political/institutional indicators, Frankel and Romer predicted trade-to-GDP, legal-origin indicators, life expectancy, and political assassination counts.
- Hansen J tests do not reject IV specifications where reported.
- Results robust to using country-specific component measures; some exchange-rate-flexibility effects on worst drops sensitive to outliers (Burundi and Nicaragua).

### Policy implications and recommendations
- Domestic policy matters: much of output volatility and worst drops in emerging market and developing countries are driven by country-specific factors.
- Fiscal policy:
  - Volatility of discretionary government spending increases output volatility and severity of worst drops.
  - Policy measures: contain fiscal volatility via greater expenditure restraint during cyclical upturns and strengthen budgetary institutions.
  - Fiscal responsibility laws exist in several countries; their effectiveness remains an empirical issue.
- Exchange rate and external shocks:
  - Terms of trade volatility raises output volatility and worst drops; exchange rate flexibility can cushion these impacts.
- Financial development and openness:
  - Role of financial sector development not robust once fiscal policy volatility is included.
  - Trade openness may be associated with higher volatility in some samples (excluding industrial countries); net effects depend on trade composition and country circumstances.
- Overall conclusion: stabilizing discretionary fiscal policy and, where appropriate, adopting exchange rate flexibility can materially reduce output volatility and incidence of large output drops, improving long-term growth and welfare.

### Appendix I — Regional groupings and country coverage (selected lists)
- G-7: Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States.
- Other industrial countries: Australia, Austria, Belgium, Denmark, Finland, Greece, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, and Switzerland.
- Central America and the Caribbean: Costa Rica, Dominican Republic, El Salvador, Guatemala, Guyana, Haiti, Honduras, Mexico, Nicaragua, Panama, and Trinidad and Tobago.
- South America: Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Paraguay, Peru, Uruguay, and Venezuela.
- MENA: Algeria, Egypt, Iran, Israel, Jordan, Morocco, Syria, Tunisia, and Turkey.
- Sub-Saharan Africa — CFA franc zone: Burkina Faso, Cameroon, Republic of Congo, Côte d’Ivoire, Gabon, Guinea-Bissau, Mali, Niger, Senegal, and Togo.
- Sub-Saharan Africa — Non-CFA: Botswana, Burundi, Ethiopia, Gambia, Ghana, Kenya, Madagascar, Malawi, Mozambique, Namibia, Nigeria, Sierra Leone, South Africa, Tanzania, Uganda, and Zambia.
- South Asia and China: Bangladesh, China, India, Pakistan, and Sri Lanka.
- East Asia: Korea, Indonesia, Malaysia, the Philippines, Singapore, and Thailand.

### Appendix II — Dynamic factor model (method summary)
- Model decomposes each country’s output growth into global, regional, and country-specific components; factors and country-specific terms follow autoregressive processes.
- Estimation: Bayesian techniques; innovations variance normalized; model estimated allowing for a break in 1986 and estimated separately over subperiods to compute factor contributions to changes in volatility.

### Appendix III — Data definitions and selected descriptive statistics (1970–2003 averages unless otherwise stated)

- Variable definitions (selected):
  - Volatility of discretionary fiscal policy: standard deviation of cyclically-adjusted government spending (1960–2000) from Fatás and Mihov (2003).
  - Terms of trade volatility: standard deviation of the annual change in the terms of trade over 1970–2003 (IMF WEO).
  - Exchange rate regime flexibility index: average over 1970–2001 with values 1 = fixed, 2 = intermediate, 3 = free float (Reinhart and Rogoff (2004) de facto classification).
  - Financial sector development: average ratio of private sector credit to GDP over 1970–2003 (Beck, Demirgüç-Kunt, and Levine (1999)).
  - Trade openness: (imports + exports of goods and services) / GDP (World Development Indicators).

- Appendix III. Table 1 — Volatility and worst-drop statistics (averages and standard deviations by region)
  - G-7 countries
    - Average: Volatility of output growth 2.2; Volatility of country-component of output growth 1.4; Worst drop of output growth 3.1; Worst drop of country-component of output growth 1.0
    - Standard Deviation: 0.3; 0.3; 1.3; 0.7
  - Other industrial countries
    - Average: 2.5; 1.8; 4.2; 2.7
    - Standard Deviation: 0.7; 0.6; 2.7; 2.2
  - Latin America
    - Average: 4.9; 4.2; 10.5; 9.3
    - Standard Deviation: 2.1; 2.0; 5.1; 4.6
  - Middle East and North Africa
    - Average: 5.7; 4.8; 11.4; 9.6
    - Standard Deviation: 2.6; 1.8; 7.3; 5.4
  - Sub-Saharan Africa — CFA franc zone countries
    - Average: 7.3; 6.3; 15.0; 13.5
    - Standard Deviation: 3.6; 4.0; 7.4; 7.2
  - Sub-Saharan Africa — Non-CFA countries
    - Average: 6.6; 5.9; 16.3; 15.3
    - Standard Deviation: 2.3; 2.2; 8.5; 8.5
  - South Asia and China
    - Average: 3.3; 2.6; 6.0; 3.7
    - Standard Deviation: 0.8; 0.9; 3.5; 2.5
  - East Asia
    - Average: 3.7; 2.2; 8.0; 2.2
    - Standard Deviation: 0.5; 0.6; 3.4; 3.0

- Appendix III. Table 2 — Selected descriptive statistics on explanatory variables (averages and standard deviations by region)
  - G-7 countries
    - Volatility of discretionary fiscal policy: Average 0.8; Standard Deviation 0.2
    - Financial sector development: Average 84.7; Standard Deviation 25.2
    - Trade openness: Average 41.8; Standard Deviation 15.6
    - Relative income: Average 0.8; Standard Deviation 0.1
    - Terms of trade volatility: Average 4.6; Standard Deviation 1.6
    - Exchange rate regime flexibility: Average 2.2; Standard Deviation 0.5
    - Exchange regime in year prior to worst drop: Average 2.1; Standard Deviation 0.7
    - Capital account restrictions: Average 0.3; Standard Deviation 0.3
  - Latin America
    - Volatility of discretionary fiscal policy: Average 2.3; Standard Deviation 0.5
    - Financial sector development: Average 24.6; Standard Deviation 11.0
    - Trade openness: Average 58.6; Standard Deviation 33.4
    - Relative income: Average 0.2; Standard Deviation 0.1
    - Terms of trade volatility: Average 13.2; Standard Deviation 5.7
    - Exchange rate regime flexibility: Average 1.8; Standard Deviation 0.3
    - Exchange regime in year prior to worst drop: Average 2.2; Standard Deviation 0.8
    - Capital account restrictions: Average 0.6; Standard Deviation 0.3
  - Sub-Saharan Africa — CFA franc zone countries
    - Volatility of discretionary fiscal policy: Average 2.7; Standard Deviation 0.4
    - Financial sector development: Average 17.1; Standard Deviation 7.9
    - Trade openness: Average 68.1; Standard Deviation 25.6
    - Relative income: Average 0.1; Standard Deviation 0.1
    - Terms of trade volatility: Average 15.6; Standard Deviation 7.6
    - Exchange rate regime flexibility: Average 1.2; Standard Deviation 0.4
    - Exchange regime in year prior to worst drop: Average 1.2; Standard Deviation 0.7
    - Capital account restrictions: Average 1.0; Standard Deviation 0.0
  - Sub-Saharan Africa — Non-CFA countries
    - Volatility of discretionary fiscal policy: Average 2.6; Standard Deviation 0.6
    - Financial sector development: Average 16.1; Standard Deviation 13.9
    - Trade openness: Average 59.5; Standard Deviation 28.0
    - Relative income: Average 0.1; Standard Deviation 0.1
    - Terms of trade volatility: Average 20.8; Standard Deviation 10.4
    - Exchange rate regime flexibility: Average 1.9; Standard Deviation 0.3
    - Exchange regime in year prior to worst drop: Average 2.3; Standard Deviation 0.5
    - Capital account restrictions: Average 0.9; Standard Deviation 0.1
  - South Asia and China
    - Volatility of discretionary fiscal policy: Average 2.3; Standard Deviation 0.8
    - Financial sector development: Average 38.2; Standard Deviation 34.4
    - Trade openness: Average 36.6; Standard Deviation 20.0
    - Relative income: Average 0.1; Standard Deviation 0.0
    - Terms of trade volatility: Average 10.4; Standard Deviation 5.9
    - Exchange rate regime flexibility: Average 1.8; Standard Deviation 0.1
    - Exchange regime in year prior to worst drop: Average 1.8; Standard Deviation 0.5
    - Capital account restrictions: Average 1.0; Standard Deviation 0.0
  - East Asia
    - Volatility of discretionary fiscal policy: Average 2.1; Standard Deviation 0.2
    - Financial sector development: Average 58.3; Standard Deviation 23.3
    - Trade openness: Average 123.2; Standard Deviation 108.1
    - Relative income: Average 0.3; Standard Deviation 0.2
    - Terms of trade volatility: Average 7.1; Standard Deviation 4.2
    - Exchange rate regime flexibility: Average 1.8; Standard Deviation 0.3
    - Exchange regime in year prior to worst drop: Average 2.2; Standard Deviation 0.4
    - Capital account restrictions: Average 0.6; Standard Deviation 0.4

*Source: Appendix III, _wp07114 - Appendix III.*

### 1.     Volatility of Output Growth......................................................................................

### 1. Volatility of Output Growth

### Overview and motivation
- Recent research highlights adverse effects of high output volatility on economic growth, welfare, and poverty.
- Declining volatility has been documented in industrial and developing countries in recent decades, but some developing countries experienced large welfare losses from extreme volatility—large output drops—in the 1980s and 1990s.
- Two complementary strands of literature are combined: (1) international business cycles using latent factor models; (2) cross-country regression analysis of determinants of volatility and large output drops.

### Data coverage and methods
- Sample: 87 countries including 67 emerging market and developing countries.
- Time periods analyzed: 1970–2004, and subperiods 1970–86 and 1987–2004; some regression analyses reference 1970–2003.
- Methodologies:
  - Bayesian dynamic latent factor model to decompose output fluctuations into global, regional, and country-specific components (building on Kose, Otrok, and Whiteman, 2003).
  - Cross-country regression techniques using as dependent variables the volatility and worst drops of the country-specific component derived from the dynamic factor analysis.

### Key empirical findings
- Volatility and severity of worst output drops have declined in most emerging market and developing countries in recent years.
- Large regional differences persist: output volatility and severity of drops are much higher in sub-Saharan Africa than in Asia.
- Levels of volatility and worst drops in emerging market and developing countries remain well above those in industrial countries, indicating scope for further reduction.
- The size of the worst output drops (largest drops in annual per capita output growth between 1970 and 2003) is highly correlated with overall volatilities.
- Decomposition results:
  - A significant global component exists in individual countries' output fluctuations.
  - Country-specific and idiosyncratic factors play a greater role in explaining output volatility in developing countries than in industrial countries.
  - The observed declines in output volatility in emerging market and developing countries over 1987–2003 are mainly due to less volatile country-specific factors rather than increased contribution of more stable global and regional factors.
- Worst output drops:
  - Worst drops in emerging market and developing countries have been mainly associated with country-specific events.
  - Country-specific factors explain the size of the worst output drops in emerging market and developing countries to a greater extent in 1987–2003 than in 1970–86.
  - Country-specific events (worst drops of the country components from the dynamic factor analysis) were more benign in 1987–2003 compared with 1970–86.
  - For industrial countries, worst drops in 1970–86 appear largely driven by global events, with the regional factor becoming more influential in explaining worst output drops during 1987–2003.

### Regression evidence on determinants
- Using volatility of the country-specific component and worst drop of the country-specific component as dependent variables yields results broadly similar to using overall volatility and worst overall drops, because the country-specific factor explains the lion’s share of overall volatility and worst drops in emerging market and developing countries.
- Fiscal policy:
  - Volatility in discretionary fiscal spending (not procyclical fiscal policy per se) influences output volatility, consistent with Fatás and Mihov (2003) and Mody and Schindler (2005).
  - Volatility in discretionary fiscal spending is an important determinant of the size of the worst output drops, supporting the hypothesis that excessively expansionary government spending can culminate in severe output declines.
- Exchange rate regime and external shocks:
  - Evidence supports the view that exchange rate flexibility helps moderate the size of the worst output drops and reduces output volatility driven by external shocks, consistent with Meade (1960) and related literature.

### Contributions of the paper
- Integrates methodologies from latent factor and regression literatures to analyze drivers of volatility and size of large output drops.
- Extends dynamic latent factor estimates to 1970–2004 and subperiods to examine changes over time.
- Demonstrates that reductions in volatility in emerging market and developing countries are primarily due to less volatile country-specific factors.
- Links volatility in discretionary fiscal policy to both output volatility and the severity of worst output drops.
- Provides evidence that exchange rate flexibility mitigates worst output drops and volatility from external shocks.

*IMF Working Paper: “Volatility of Output Growth” (excerpt).*

### Section IV provides a summary and conclusions.

### _wp07114 - Section IV provides a summary and conclusions

### Stylized facts on volatility and worst output drops across regions
- The volatility of per capita output growth in emerging market and developing countries has declined over the past three decades (Figure 1), but remains considerably higher than in industrial countries.
- Regional patterns:
  - South Asia and China, the Middle East and North Africa (MENA), and CFA franc zone countries in sub-Saharan Africa: volatility has shown a sustained decline (Figure 2).
  - Latin America: volatility has remained constant at a relatively high level.
  - East Asia: volatility has increased since 1997.
  - On average over 1970–2003: countries in Asia have had the lowest volatility; countries in sub-Saharan Africa the highest.
- Table 1 (1970–86 versus 1987–2003):
  - Sub-Saharan Africa, south Asia and China, and MENA: decline in severity of worst output drops from the first subperiod to the second.
  - East Asia: severity of worst drops increased.
  - Latin America: severity of worst drops remained more or less the same.
- Cross-country correlation:
  - A plot of each country’s average volatility (excluding the worst drop) against the magnitude of the worst output drop confirms a strong correlation (Figure 3).

### Data and decomposition methodology
- Output growth definition and data:
  - Output growth is per capita real GDP growth measured using Penn World Tables (PWT), Version 6.1, covering 1970–2000; WEO data used to extend series to 2004.
- Dynamic factor model decomposition:
  - The model decomposes output fluctuations into three factors:
    - global factor: captures shocks affecting all countries (Global_t f),
    - regional factor: captures shocks common to countries in a particular region,
    - country-specific factor: residual capturing shocks affecting an individual country (and other unmodeled influences).
  - The global factor captures major peaks and troughs in global GDP growth 1970–2004, including oil price shocks in the 1970s, recessions in the early 1980s and 1990s, and the high-tech bust in the early 2000s (Figure 4).
  - Regional factors capture known cyclical events (e.g., 1997–98 east Asia crisis; South America’s debt crisis and late-1980s inflation; CFA swings from droughts and terms-of-trade shocks) (Figure 5).

### Contributions of factors to output volatility (Table 2; averages for each group, in percent)
- Fraction of variance in real per capita output growth attributable to each factor (Global / Regional / Country):
  - Sub-Saharan Africa — CFA countries: 5.7 / 18.2 / 76.1
  - Sub-Saharan Africa — Non-CFA countries: 6.8 / 10.2 / 82.1
  - Middle East and North Africa: 3.8 / 15.9 / 80.3
  - Latin America: 12.6 / 13.7 / 73.7
  - South Asia and China: 15.6 / 20.6 / 63.8
  - East Asia: 11.0 / 41.8 / 47.2
  - East Asia (1970–96): 18.0 / 15.8 / 66.3
  - Developing countries (group average): 9.3 / 16.9 / 73.8
  - Industrial countries (group average): 24.3 / 21.7 / 54.0
- Key patterns:
  - Output fluctuations in emerging market and developing countries are driven more by country-specific factors than in industrial countries.
  - The country-specific factor explains more than 60 percent of output volatility in about 90 percent of emerging market and developing countries, compared with only 40 percent of industrial countries.
  - The global factor explains less than 10 percent of output fluctuations for more than 60 percent of emerging market and developing countries; by contrast, for nearly half of industrial countries it explains between 10 and 20 percent.
  - In all emerging market and developing country regions except east Asia, at least 60 percent of output volatility is attributable to country-specific factors.
  - In emerging market and developing country regions, the regional factor explains a greater fraction of volatility than the global factor, unlike in industrial countries.
  - East Asia’s large regional-factor contribution mainly reflects the east Asian financial crisis; estimating the model over 1970–96 raises the role of the global factor for east Asia.

### What explains the decline in volatility across regions? (Table 3)
- The dynamic factor model estimated over two periods (1970–86 and 1987–2004) indicates:
  - Declines in output volatility in emerging market and developing country regions are mainly due to less volatile country-specific factors.
  - In all regions except Latin America, at least 70 percent of the decline in the variance of output growth is explained by the country-specific factor.
  - By contrast, for industrial countries the corresponding share is about 50 percent.
- Table 3 — contribution of each factor to the decline in the variance of output growth (averages for each group; in percent). Only countries with a decline in volatility are included:
  - Sub-Saharan Africa — CFA countries: Global -40.0 / Regional -2.1 / Country -5.8 / (Total) -32.1
  - Sub-Saharan Africa — Non-CFA countries: Global -53.2 / Regional -6.9 / Country -8.3 / (Total) -38.0
  - Middle East and North Africa: Global -60.7 / Regional -0.5 / Country -17.4 / (Total) -42.7
  - Latin America: Global -12.1 / Regional -3.4 / Country -2.8 / (Total) -5.9
  - South Asia and China: Global -12.0 / Regional -0.9 / Country -2.9 / (Total) -8.2
  - Developing countries (average): Global -34.2 / Regional -3.2 / Country -6.8 / (Total) -24.2
  - Industrial countries: Global -4.1 / Regional -1.1 / Country -0.8 / (Total) -2.3
- Note: Table 3 reports results for countries where volatility declined; this differs from Figure 2 which shows average volatility for all countries in a region.

### Drivers of worst output drops (Table 4)
- Worst output drops are mainly associated with country-specific events in emerging market and developing country regions; exception:
  - East Asia: the regional factor is the main contributor to the worst drops.
- The contribution of the country-specific factor to the worst output drops in emerging market and developing country regions increased from the first (1970–86) to the second (1987–2003) subperiod.
- Table 4 — contribution of each factor to the worst drop in real per capita output growth (averages for each group; in percent):
  - 1970–86 / 1987–2003 (Global / Regional / Country)
  - Sub-Saharan Africa — CFA countries: 4.9 / 27.7 / 67.5  → 2.6 / 8.9 / 88.5
  - Sub-Saharan Africa — Non-CFA countries: -1.8 / 11.1 / 90.7 → -14.7 / -1.8 / 116.4
  - Middle East and North Africa: 6.0 / 16.9 / 77.1 → 9.9 / 15.1 / 74.9
  - Latin America: 24.1 / 11.5 / 64.5 → 1.2 / 10.2 / 88.6
  - South Asia and China: 35.0 / 30.5 / 34.6 → -113.5 / 10.6 / 202.9
  - East Asia: 50.0 / 120.5 / -70.5 → 7.9 / 104.5 / -12.4
  - Emerging market and developing countries (aggregate): 15.0 / 27.5 / 57.6 → -9.6 / 18.5 / 91.1
  - Industrial countries: 71.6 / 46.1 / -17.8 → 20.7 / 59.5 / 19.8
- Additional observations:
  - The worst drop of the country component of output growth declined on average from the first subperiod to the second in all emerging market and developing country regions except Latin America (Table 1).
  - For industrial countries, the global factor was the main contributor to the worst output drops in 1970–86 (about 70 percent on average), while the regional factor became the main contributor in 1987–2003.

### Cross-section regressions and explanatory variables
- Dependent variables in cross-section regressions:
  - Standard deviation of real per capita GDP growth (1970–2003) and worst drops in real per capita output growth (1970–2003).
  - Alternative specification: standard deviation of the country-specific component of real per capita GDP growth (1970–2003) and the worst drop in that country-specific component (1970–2003) from the dynamic factor analysis.
- Explanatory variables used (consistent with prior literature):
  - indicator of the volatility of discretionary government spending,
  - indicator of the procyclicality of fiscal policy,
  - terms of trade volatility,
  - interaction of exchange rate flexibility with terms of trade volatility,
  - relative income,
  - trade openness,
  - indicator of financial sector development,
  - indicator of capital account restrictions.
- Data sources and variable definitions are discussed in the paper (methodology and data section).

*Italic: Source — _wp07114 - Section IV provides a summary and conclusions.*

### Appendix III.

### Appendix III

### Estimation approach and data
- Objective: Identify effects of fiscal policy and terms of trade volatility, in conjunction with exchange rate policy, on output volatility and worst output drops.
- Sample and periods:
  - Explanatory variables are averages over the 1970–2003 period.
  - Output growth volatility defined as the standard deviation of real GDP per capita growth rate over 1970–2003.
  - Volatility of country-component of output growth defined as the standard deviation of the country-specific component of per capita GDP growth between 1970–2003 (from a dynamic factor model).
  - Regressions use data from 46 developing countries and 20 industrial countries for volatility regressions; worst-output-drop regressions use data from 43 (44, in some cases) emerging market and developing countries and 20 industrial countries.
- Estimation methods:
  - Ordinary least squares (OLS) and instrumental variables (IV) estimations are reported.
  - IV regressions instrument for potential endogeneity of fiscal policy volatility (and for financial sector development and trade-to-GDP variables) using political and institutional instruments (e.g., nature of electoral system, nature of political system), Frankel and Romer predicted trade-to-GDP, legal-origin indicators, life expectancy, and political assassination counts.
  - Hansen J tests do not reject the IV specifications where reported.

### Key determinants (main empirical findings)
- Three variables are identified as the key determinants of output volatility and worst output drops:
  - Volatility of discretionary government spending.
  - Terms of trade volatility.
  - Interaction of terms of trade volatility with exchange regime flexibility.
- Volatility of discretionary fiscal policy:
  - Has a strong positive and significant effect on both measures of output volatility.
  - Effect is larger in the IV regressions than in OLS, suggesting a strong causal effect of discretionary fiscal policy on output volatility and on the country-specific component of output growth.
  - The volatility of discretionary spending variable is also a significant determinant of worst output drops in both OLS and IV regressions.
  - Fiscal procyclicality is not found to have a significant effect on worst output drops.
- Terms of trade volatility and exchange rate flexibility:
  - Terms of trade volatility has a positive and significant effect on output volatility and on worst output drops.
  - The exchange rate regime variable by itself is generally not significant.
  - The interaction of terms of trade volatility with the index of exchange regime flexibility is negative and significant, implying the association between terms of trade shocks and output volatility/worst drops is more pronounced under fixed than flexible exchange rate regimes.
  - Example (volatility regressions): a one standard deviation increase in terms of trade volatility implies a 1.5 percentage point increase in output volatility in the region with the least exchange rate flexibility (the CFA franc zone) compared with a 0.3 percentage point increase in the region with the greatest exchange rate flexibility (non-CFA countries).
  - Example (worst-drop IV regression of country-component): a one standard deviation increase in terms of trade volatility implies a 5 percentage point increase in the worst output drop in the region with the least exchange rate flexibility (the CFA franc zone) compared with a 3 percentage point increase in the region with the greatest exchange rate flexibility (sub-Saharan Africa’s non-CFA countries).
- Other variables:
  - Financial sector development: sign and significance vary; not robust once fiscal policy volatility is included. Positive but not significant in IV regressions for the entire sample; negative and sometimes significant in regressions excluding industrial countries (mostly in OLS).
  - Trade openness and capital account restrictions: generally positive but not statistically significant for the full sample. Trade openness is positively and significantly associated with country-specific and overall output volatility only in regressions that exclude industrial countries.
  - Relative income and other controls: largely insignificant in most specifications.
- Robustness and diagnostics:
  - Results using the country-specific component from the dynamic factor model are broadly the same in magnitude and significance as those using overall output volatility.
  - Hansen J tests of overidentifying restrictions reported for IV estimations do not reject the validity of the instruments where shown.
  - Some exchange-rate-flexibility effects on worst drops are sensitive to outliers (defined as Burundi and Nicaragua, the two countries with the most severe worst output drops).

### Regional patterns and simulated impacts
- Cross-sectional averages (1970–2003):
  - Sub-Saharan African countries had the highest volatility of output growth, the most severe worst drops, the highest volatility of discretionary government spending, and the highest terms of trade volatility.
  - East Asia had the lowest volatility of the country-component of output growth, the smallest worst drop, the lowest volatility of discretionary government spending, and smaller terms of trade shocks.
  - CFA franc zone countries had the lowest exchange regime flexibility on average; sub-Saharan Africa’s non-CFA countries had the highest exchange regime flexibility.
- Simulation: reducing discretionary fiscal volatility to east Asia levels
  - From IV estimates of the volatility of the country-component of output growth for emerging market and developing countries:
    - Reducing the volatility of cyclically adjusted government spending to the level in east Asia would have reduced output volatility by 3.8 percentage points for countries in the CFA franc zone.
    - The same reduction would have reduced output volatility by 3.1 percentage points for sub-Saharan Africa’s non-CFA countries.
    - These reductions are equivalent to about 50 and 60 percent of the country-specific output volatility respectively.
  - Countries in other regions, notably Latin America, would also gain from more stable fiscal policy.

### Policy implications and conclusions
- Domestic policy matters:
  - Much of output volatility and the worst output drops in emerging market and developing countries is driven by country-specific factors, emphasizing the role of domestic policies.
- Fiscal policy:
  - Volatility of discretionary government spending increases both output volatility and severity of worst output drops. Containing fiscal volatility—through greater expenditure restraint during cyclical upturns and strengthening budgetary institutions—would reduce output volatility.
  - Fiscal responsibility laws have been adopted by several countries but their effectiveness remains an empirical issue.
- Terms of trade and exchange rate regime:
  - Terms of trade volatility raises output volatility and worst output drops; exchange rate flexibility can cushion this impact.
- Financial sector development and openness:
  - The previously documented role of financial sector development is not robust to including fiscal policy volatility.
  - Trade openness may be associated with higher volatility in some samples (excluding industrial countries), but the net implications depend on trade composition and country circumstances.
- Overall conclusion:
  - Reducing volatility—especially by stabilizing discretionary fiscal policy and, where appropriate, adopting exchange rate flexibility—can materially lower output volatility and the incidence of large output drops, contributing to improved long-term growth and welfare.

*Source: Appendix III, _wp07114 - Appendix III.*

### Appendix I. Regional Groupings and Country Coverage

### Appendix I. Regional Groupings and Country Coverage

### Regional groupings and rationale
- Latin America subdivided into: (1) Central America and the Caribbean and (2) South America.
- Industrial countries subdivided into: (1) G-7 and (2) other industrial countries.
- Groupings justified by:
  - Geographic proximity implying shared shocks (weather-related shocks, terms of trade shocks).
  - Trade and financial linkages and policy coordination (example: peg of the CFA franc zone countries).
  - For industrial countries, grouping reflects size of economies, stage of economic development, and quality of institutions rather than geography.

### Country coverage by region
- G-7 countries: Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States.
- Other industrial countries: Australia, Austria, Belgium, Denmark, Finland, Greece, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, and Switzerland.
- Central America and the Caribbean: Costa Rica, Dominican Republic, El Salvador, Guatemala, Guyana, Haiti, Honduras, Mexico, Nicaragua, Panama, and Trinidad and Tobago.
- South America: Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Paraguay, Peru, Uruguay, and Venezuela.
- Middle East and North Africa (MENA): Algeria, Egypt, Iran, Israel, Jordan, Morocco, Syria, Tunisia, and Turkey.
- Sub-Saharan Africa
  - CFA franc zone countries: Burkina Faso, Cameroon, Republic of Congo, Côte d’Ivoire, Gabon, Guinea-Bissau, Mali, Niger, Senegal, and Togo.
  - Non-CFA countries: Botswana, Burundi, Ethiopia, Gambia, Ghana, Kenya, Madagascar, Malawi, Mozambique, Namibia, Nigeria, Sierra Leone, South Africa, Tanzania, Uganda, and Zambia.
- South Asia and China: Bangladesh, China, India, Pakistan, and Sri Lanka.
- East Asia: Korea, Indonesia, Malaysia, the Philippines, Singapore, and Thailand.

_This content unit: Appendix I. Regional Groupings and Country Coverage (plus Appendices II–III content excerpt)_

### Appendix II. The Dynamic Factor Model

### Purpose and structure
- Dynamic factor models posit that a few unobservable common factors drive fluctuations in large cross sections of macroeconomic time series; these factors serve as indexes of common activity (global or regional).
- The model decomposes each observable variable (e.g., output growth in a country) into:
  - a global component,
  - a regional component,
  - a country-specific (idiosyncratic) component.

### Model specification (region/block structure)
- The system contains a block of equations for each region; each country’s output growth (Y) equation includes a global factor, a regional factor, and a country-specific term.
- Example block form for the first region (G-7) is presented with country-specific factor loadings (b) multiplying the global factor (Global) and a country-specific component (c):
  - (Example notation shown in source for US, Japan, France with factor loadings b and country-specific terms c; same form is repeated for each of the nine regions.)

### Dynamics and estimation
- Factors (f s) and country-specific terms (c) are (independent) autoregressive processes:
  - Global factor AR representation: f_Global,t = φ(L) f_Global,t−1 + u_t where φ(L) is a lag polynomial and u_t is normally distributed.
- All factor loadings (b s) and lag polynomials are independent of each other.
- The model is estimated using Bayesian techniques as described in Kose, Otrok, and Whiteman (2003) and Otrok, Silos, and Whiteman (2003).
- The innovation variance of the factors (error term in the factor autoregressive equation) is normalized (normalization based on variance of underlying series).
- The model is estimated using deviations from mean output growth data allowing for a break in 1986.

### Variance decompositions and attribution of volatility changes
- Variance decompositions express the volatility of output growth into components due to each factor; factors are orthogonal so no cross-product terms appear.
- Example variance decomposition for the US first equation:
  - var(Y_US) = b_US^2 var(f_Global) + ... + var(c_US).
- The variance in real per capita output growth attributable to the global factor is:
  - (b_US^2 var(f_Global)) / var(Y_US)  (expressed in source as Global: b_US^2 var(f_Global) / var(Y_US)).
- To analyze changes in volatility between subperiods 1970–86 and 1987–2004, the model is estimated separately over the two periods and each factor’s contribution to the change in overall volatility is calculated.
- Example contribution of global factor to decline in US variance:
  - [ (b_US,1987–04^2 var(f_Global,1987–04)) − (b_US,1970–86^2 var(f_Global,1970–86)) ] contributing to var(Y_US,1987–04) − var(Y_US,1970–86).

### Appendix III. Data Definitions, Sources, and Descriptive Statistics

### Variable definitions and data sources
- Volatility of discretionary fiscal policy: standard deviation of cyclically-adjusted government spending over the 1960–2000 period from Fatás and Mihov (2003).
- Fiscal policy procyclicality: country-by-country regressions (annual) for two subperiods estimating:
  - CG_t = α_i + β_i CGDP_t + ε_it
  - CG = cyclical component of real government expenditure (log deviation from Hodrick-Prescott trend).
  - CGDP = cyclical component of real GDP (log deviation from Hodrick-Prescott trend).
  - β_i measures cyclicality: positive = procyclical; β_i > 1 = more than proportionate response.
  - Data source: IMF’s WEO database.
- Terms of trade volatility: standard deviation of the annual change in the terms of trade over 1970–2003; source: IMF’s WEO database.
- Exchange rate regime flexibility (for output volatility regressions): average over 1970–2001 of index with values:
  - 1 = fixed regime,
  - 2 = intermediate regime,
  - 3 = free float.
  - Classification based on Reinhart and Rogoff (2004) de facto “natural classification.”
  - For worst output drop regressions, index equals Reinhart and Rogoff exchange rate classification in the year prior to the worst drop (values 1–3).
  - Instances classified as “free fall” replaced with secondary classification reported in Reinhart and Rogoff (2004).
- Trade openness: (imports + exports of goods and services) / GDP; source: World Bank’s World Development Indicators.
- Financial sector development: proxied by average ratio of private sector credit to GDP over 1970–2003; source: Beck, Demirgüç-Kunt, and Levine (1999).
- Capital account restrictions: index = 1 if country has capital account restrictions (per IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions) for 1970–2003, = 0 otherwise.
- Relative income: level of real per capita income relative to the United States; real per capita GDP in constant 1996 prices from Penn World Tables, Version 6.1.

### Key descriptive statistics (1970–2003 averages unless otherwise stated)

- Appendix III. Table 1 — Volatility and worst-drop statistics (averages and standard deviations by region)
  - G-7 countries
    - Average: Volatility of output growth 2.2; Volatility of country-component of output growth 1.4; Worst drop of output growth 3.1; Worst drop of country-component of output growth 1.0
    - Standard Deviation: 0.3; 0.3; 1.3; 0.7
  - Other industrial countries
    - Average: 2.5; 1.8; 4.2; 2.7
    - Standard Deviation: 0.7; 0.6; 2.7; 2.2
  - Latin America
    - Average: 4.9; 4.2; 10.5; 9.3
    - Standard Deviation: 2.1; 2.0; 5.1; 4.6
  - Middle East and North Africa
    - Average: 5.7; 4.8; 11.4; 9.6
    - Standard Deviation: 2.6; 1.8; 7.3; 5.4
  - Sub-Saharan Africa — CFA franc zone countries
    - Average: 7.3; 6.3; 15.0; 13.5
    - Standard Deviation: 3.6; 4.0; 7.4; 7.2
  - Sub-Saharan Africa — Non-CFA countries
    - Average: 6.6; 5.9; 16.3; 15.3
    - Standard Deviation: 2.3; 2.2; 8.5; 8.5
  - South Asia and China
    - Average: 3.3; 2.6; 6.0; 3.7
    - Standard Deviation: 0.8; 0.9; 3.5; 2.5
  - East Asia
    - Average: 3.7; 2.2; 8.0; 2.2
    - Standard Deviation: 0.5; 0.6; 3.4; 3.0

- Appendix III. Table 2 — Descriptive statistics on explanatory variables by region (1970–2003 averages unless otherwise stated)
  - G-7 countries
    - Volatility of discretionary fiscal policy: Average 0.8; Standard Deviation 0.2
    - Financial sector development: Average 84.7; Standard Deviation 25.2
    - Trade openness: Average 41.8; Standard Deviation 15.6
    - Relative income: Average 0.8; Standard Deviation 0.1
    - Terms of trade volatility: Average 4.6; Standard Deviation 1.6
    - Exchange rate regime flexibility: Average 2.2; Standard Deviation 0.5
    - Exchange regime in year prior to worst drop: Average 2.1; Standard Deviation 0.7
    - Capital account restrictions: Average 0.3; Standard Deviation 0.3
  - Other industrial countries
    - Volatility of discretionary fiscal policy: Average 1.0; Standard Deviation 0.3
    - Financial sector development: Average 70.3; Standard Deviation 29.4
    - Trade openness: Average 68.5; Standard Deviation 27.6
    - Relative income: Average 0.7; Standard Deviation 0.2
    - Terms of trade volatility: Average 5.6; Standard Deviation 4.0
    - Exchange rate regime flexibility: Average 1.7; Standard Deviation 0.4
    - Exchange regime in year prior to worst drop: Average 1.8; Standard Deviation 0.4
    - Capital account restrictions: Average 0.5; Standard Deviation 0.3
  - Latin America
    - Volatility of discretionary fiscal policy: Average 2.3; Standard Deviation 0.5
    - Financial sector development: Average 24.6; Standard Deviation 11.0
    - Trade openness: Average 58.6; Standard Deviation 33.4
    - Relative income: Average 0.2; Standard Deviation 0.1
    - Terms of trade volatility: Average 13.2; Standard Deviation 5.7
    - Exchange rate regime flexibility: Average 1.8; Standard Deviation 0.3
    - Exchange regime in year prior to worst drop: Average 2.2; Standard Deviation 0.8
    - Capital account restrictions: Average 0.6; Standard Deviation 0.3
  - Middle East and North Africa
    - Volatility of discretionary fiscal policy: Average 2.1; Standard Deviation 0.1
    - Financial sector development: Average 34.7; Standard Deviation 18.8
    - Trade openness: Average 60.6; Standard Deviation 28.1
    - Relative income: Average 0.2; Standard Deviation 0.1
    - Terms of trade volatility: Average 14.5; Standard Deviation 11.1
    - Exchange rate regime flexibility: Average 1.8; Standard Deviation 0.3
    - Exchange regime in year prior to worst drop: Average 2.0; Standard Deviation 0.0
    - Capital account restrictions: Average 0.9; Standard Deviation 0.1
  - Sub-Saharan Africa — CFA franc zone countries
    - Volatility of discretionary fiscal policy: Average 2.7; Standard Deviation 0.4
    - Financial sector development: Average 17.1; Standard Deviation 7.9
    - Trade openness: Average 68.1; Standard Deviation 25.6
    - Relative income: Average 0.1; Standard Deviation 0.1
    - Terms of trade volatility: Average 15.6; Standard Deviation 7.6
    - Exchange rate regime flexibility: Average 1.2; Standard Deviation 0.4
    - Exchange regime in year prior to worst drop: Average 1.2; Standard Deviation 0.7
    - Capital account restrictions: Average 1.0; Standard Deviation 0.0
  - Sub-Saharan Africa — Non-CFA countries
    - Volatility of discretionary fiscal policy: Average 2.6; Standard Deviation 0.6
    - Financial sector development: Average 16.1; Standard Deviation 13.9
    - Trade openness: Average 59.5; Standard Deviation 28.0
    - Relative income: Average 0.1; Standard Deviation 0.1
    - Terms of trade volatility: Average 20.8; Standard Deviation 10.4
    - Exchange rate regime flexibility: Average 1.9; Standard Deviation 0.3
    - Exchange regime in year prior to worst drop: Average 2.3; Standard Deviation 0.5
    - Capital account restrictions: Average 0.9; Standard Deviation 0.1
  - South Asia and China
    - Volatility of discretionary fiscal policy: Average 2.3; Standard Deviation 0.8
    - Financial sector development: Average 38.2; Standard Deviation 34.4
    - Trade openness: Average 36.6; Standard Deviation 20.0
    - Relative income: Average 0.1; Standard Deviation 0.0
    - Terms of trade volatility: Average 10.4; Standard Deviation 5.9
    - Exchange rate regime flexibility: Average 1.8; Standard Deviation 0.1
    - Exchange regime in year prior to worst drop: Average 1.8; Standard Deviation 0.5
    - Capital account restrictions: Average 1.0; Standard Deviation 0.0
  - East Asia
    - Volatility of discretionary fiscal policy: Average 2.1; Standard Deviation 0.2
    - Financial sector development: Average 58.3; Standard Deviation 23.3
    - Trade openness: Average 123.2; Standard Deviation 108.1
    - Relative income: Average 0.3; Standard Deviation 0.2
    - Terms of trade volatility: Average 7.1; Standard Deviation 4.2
    - Exchange rate regime flexibility: Average 1.8; Standard Deviation 0.3
    - Exchange regime in year prior to worst drop: Average 2.2; Standard Deviation 0.4
    - Capital account restrictions: Average 0.6; Standard Deviation 0.4

_This content unit: Appendix I. Regional Groupings and Country Coverage (with Appendices II and III excerpts)._

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07114.pdf_