## _wp0711 - 1. Summary Statistics by Industry

## Source details

**Canonical URL:** [_wp0711 - 1. Summary Statistics by Industry](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0711.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0711.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0711.pdf.json)

---

### Introduction: context and motivation
- "Sudden Stops" or reversals of capital inflows and the subsequent withdrawal of international capital are identified as triggers of several recent crises, notably Thailand and South Korea.
- Thailand experienced exceptionally large capital inflow reversals:
  - cumulative inflows as a percent of GDP of approximately 51.5 percent between 1988 and 1994;
  - reversals of 26 percent between 1996 and 1997.
- Sudden stops typically require offsetting by reserve losses or lower current account deficits, and often produce:
  - contractions in output via large and unexpected swings in relative prices;
  - increased financial vulnerability through reserve losses;
  - higher incidence of nonperforming loans (NPLs) due to higher interest burdens and currency mismatches;
  - bank lending retrenchment and a "credit crunch", especially affecting small- and medium-size firms.

### Research objective and scope
- Aim: characterize factors that exacerbated financial constraints—proxied by adjustment of fixed assets—experienced by Thai nonfinancial firms during sudden stop episodes.
- Focus on firm-level and balance-sheet characteristics that induced firms to meet debt obligations through adjustment of fixed assets.
- Specific firm characteristics considered:
  - level of internal resources;
  - firm size;
  - issuance of American Depository Receipts (ADRs);
  - tradable vs. nontradable commodity production;
  - degree of foreign ownership;
  - multinational status;
  - degree of macroeconomic instability;
  - industry affiliation;
  - balance sheet deterioration (profitability and debt maturity structure);
  - short-term debt exposure and fluctuations in domestic demand.

### Data description and sample
- Primary data source: DataStream (annual corporate balance sheet and income statement data).
- Sample: 284 nonfinancial Thai firms publicly listed on the local stock market between the years 1992 and 2001.
  - Because of data limitations, the sample of firms is limited to those that remained in business (bankrupt/delisted firms are not included).
- Sector classification:
  - Tradable sectors: primary commodities, manufactures, household products and food.
  - Nontradable sectors: services and real estate.
- Sectoral composition and characteristics:
  - Service sector: 68 firms (largest sector).
  - Real estate sector: 28 firms (smallest sector).
  - Primary product sector: highest level of sales on average.
  - Household sector: lowest average sales.
  - Real estate and primary product sectors: tend to be largest in terms of size.
  - Food and household product sectors: smallest in terms of size.
  - After-tax profits: food sector has the highest profitability levels.
  - Tradable sector: characterized by having higher profits than the nontradable sector and by being more exposed to short-term debt.
  - Nontradable firms: tend to be of relatively larger size.

### Key descriptive findings and time patterns
- Fixed assets and capital flows:
  - Thai nonfinancial firms suffered a significant decline in their fixed assets throughout the capital outflow period.
  - Table III reveals the sudden stop episode led to a significant decline—of close to 50 percent—in the average growth of firms’ fixed assets.
  - Graph A in Figure II depicts the decline beginning in 1996 and not reversing until mid-1999; by mid-1999 the trend gradually reverses with signs of slower growth.
- Macroeconomic and sectoral demand:
  - Sector consumption growth averaged 6.2 percent during the 1990s except during the sudden stop episode when it declined to an average growth rate of -9 percent.
- Debt structure and tradable sectors:
  - Tradable sectors were taking significant amounts of short-term debt—about 80 percent of total debt prior to the crisis period—which significantly worsened their balance sheets once capital inflow reversals took place.
- Cross-country context on leverage (from cited literature):
  - Debt-equity ratios by region as cited: U.S. firms averaged 90 percent by end-1996, Latin American firms averaged 31 percent, while Thai firms averaged 155 percent.

### Key empirical patterns and macro outcomes
- Average growth in exports and sectoral inflation rose during the sudden stop because of the devaluation of the Thai baht:
  - export growth increased from an average of 2.5 percent during tranquil periods to 4.5 percent after the devaluation;
  - sectoral inflation rates increased from an average of 2.8 percent during tranquil times to 5.9 percent after the devaluation.
- Consumption and aggregate firm outcomes:
  - Average consumption growth in Thailand averaged 6.2 percent during the 1990s except during the sudden stop episode when it declined to an average negative growth rate of -9 percent.
  - A prior nonsudden stop episode average growth reported as 5.6 percent, but declined to an average growth rate of -4.2 percent during the sudden stop period.
- Aggregate decline in firms’ fixed assets during the sudden stop: a decline of approximately 30 percent.

### Firm-level sample and descriptive statistics
- Sample coverage: data from 284 nonfinancial firms listed in the Thai stock market between 1992 and 2001.
- Actual number of firms per year in the dataset:
  - 1992=150, 1993=183, 1994=225, 1995=249, 1996=275, 1997=278, 1998=275, 1999=273, 2000=266, and 2001=250.
- BIS lending (alternative measure to capture capital outflow):
  - Lending by BIS-reporting banks to the Thai nonfinancial private sector reached a peak of approximately US$40 billion during the second quarter of 1996 and then declined without recovering, but stabilized at US$15 billion.

### Firm financial distress and observable indicators
- Interest Coverage Ratio:
  - Average Interest Coverage Ratio declined from 21.9 during good times to 7.3 during the sudden stop period.
  - Note: the coefficients reflecting the interest coverage ratio between tranquil and sudden stop episodes are not statistically significant at conventional levels.
- Observed firm behavior consistent with fire sales and liquidity-driven asset disposals:
  - Substantial declines in the growth of firms’ fixed assets occurred around capital inflow reversals and abrupt devaluation, implying sales of assets at a discount (fire sales) driven by uncertainty and financial constraints.

### Heterogeneity across firm characteristics (descriptive and graphical)
- Debt maturity:
  - Firms with a relatively shorter debt-maturity structure suffered a steeper decline in their fixed assets during the sudden stop.
  - A shorter-term debt maturity structure led Thai firms during the sudden stop episode to a 0.2 percent decrease in their annual fixed asset growth.
- Multinationals:
  - Multinational firms adjusted their fixed assets downward drastically starting in 1997 and did not show clear recovery during the sample period.
  - Being multinational leads to a 0.1 percent decline in firms' fixed assets growth during tranquil times in the sample; not a significant factor during the sudden stop period in aggregate regressions, though effects differ by sector.
- Firm size:
  - Small firms suffer from a steeper decline in growth of fixed assets than larger ones.
  - When a firm becomes on average larger than the median, its fixed assets tend to grow by 0.1 percent (effect significant even during sudden stop episodes).
- ADR issuance and external financing:
  - Issuance of ADRs allows firms to have a higher growth rate of fixed assets during tranquil times and a faster recovery during downturns.
- Tradable vs. nontradable sectors:
  - Nontradable firms suffered a more pronounced and longer decline in the growth of their fixed assets after 1997 relative to tradable firms.
  - Plausible mechanism: tradable firms could partially offset domestic demand declines by selling abroad and gaining foreign exchange during devaluation episodes, reducing need to sell fixed assets.

### Main regression findings (random effects model; reduced form investment equation)
- Model structure:
  - Dependent variable: I_ijt / K_ijt-1 (adjustment of fixed assets of firm i, sector j, time t).
  - Regressors include lagged investment, profitability, financing costs (shocks to balance sheet), sectoral consumption, and firm characteristics.
  - Sudden stop dummy (SS) identifies 1997 and 1998 and is interacted with regressors to capture differential behavior during the sudden stop.
- Broad conclusions:
  - Shocks to net worth—profitability and maturity structure—matter for fixed assets adjustment, but in different periods: profitability matters in tranquil periods; maturity structure matters during the sudden stop.
  - During tranquil periods, fixed asset fluctuations depend primarily on profitability, size, multinational status, and ADR issuance.
  - During sudden stop episodes, characteristics that reduce likelihood of investment postponement or forced asset sales include holding long-term maturity debt and being a tradable goods producer.
- Quantitative effects and statistical results:
  - During tranquil periods a unit increase in profitability leads to a 0.4 percent increase in fixed assets growth in the following period.
  - Being multinational leads to a 0.1 percent decline in fixed assets growth during tranquil times (not significant during sudden stop in aggregate sample).
  - Firm size: being larger than the median leads to a 0.1 percent increase in fixed assets growth (significant in tranquil and sudden stop).
  - Short-term maturity interaction: during the sudden stop, short-term debt structure associated with a 0.2 percent decrease in annual fixed asset growth.
  - For tradable goods producers, sectoral export growth does not matter for fixed asset growth in tranquil or sudden stop periods in these regressions.

### Tradable versus nontradable producers (differential regression results)
- Profitability vs. maturity:
  - Profitability matters primarily during tranquil periods; exposure to short-term maturity debt is significant during sudden stop periods.
  - Impacts of profitability and maturity structure are larger for nontradable sector firms.
- Domestic demand sensitivity:
  - For nontradable goods producers, annual percentage increases in consumption lead to increases in growth of firm fixed assets of 1.8 percent during tranquil periods and 2.0 percent during sudden stop periods.
  - These effects are roughly four times larger than for tradable firms.
- Sector-specific firm characteristics:
  - For tradable firms, having a high degree of Thai ownership helped increase fixed assets growth by 0.1 percent during the sudden stop period.
  - For nontradable firms, being multinational reduces growth of fixed assets during tranquil times (multinational negative effect concentrated in nontradable sector).

### Debt structure: long-term vs. short-term maturity holders
- Model split by debt maturity structure yields:
  - Firms holding longer-term debt are less financially constrained and less likely to engage in constrained adjustment of physical capital during crises.
  - For short-maturity holders, firm size and tradable-producer status matter positively for fixed asset growth in both good and bad times.
  - For long-maturity holders, during tranquil times additional factors such as being multinational or an ADR issuer matter for fixed assets; during crisis times fixed assets adjustments of long-term debt holders depend mainly on firm size.
- Significance notes:
  - Size remains significant for both short- and long-term debt holders (though significance levels differ).
  - ADR issuance matters in tranquil times for long-maturity holders (reduces likelihood of fixed asset adjustment).

### Robustness and sensitivity checks
- Market-to-book value: found to be insignificant and excluded from benchmark specification.
- Firm age: one year of aging decelerates firms’ fixed asset growth by 0.01 percent during tranquil times (insignificant during sudden stop); sample size reduced to 413 observations when included.
- Quadratic specification for consumption: main conclusions unaffected; impact of quadratic consumption growth about half that of benchmark but maturity structure remains significant during sudden stop.
- Alternative continuous measure for capital outflow: BIS lending to Thai private sector used instead of binary sudden stop dummy—results show very little difference in significance and magnitude of coefficients; slight increase in magnitude of significant coefficients when using BIS lending interactive variable.

### Policy-relevant implications and interpretation
- Financial constraints mediated by debt maturity structure critically influence firms’ asset adjustment during sudden stops; policies that ease rollover risk or extend debt maturities could mitigate forced asset sales.
- Tradable sector exposure and access to foreign exchange can cushion firms during devaluation episodes; policies facilitating export channels or foreign-currency revenues may reduce fire-sale pressures.
- Firm-level factors—size, ADR access, multinational affiliation, and ownership structure—alter vulnerability; targeted financial support or liquidity backstops could be prioritized for smaller, nontradable, short-maturity firms during sudden stops.
- Future research recommended: directly measure forced investment adjustments with price pressure to capture fire sale dynamics, and test findings in other regions/markets that experienced similar episodes.

*Source: _wp0711 - 1. Summary Statistics by Industry (excerpt).*

### 1. Summary Statistics by Industry.......................................................................................

### _wp0711 - 1. Summary Statistics by Industry.......................................................................................

### Introduction: context and motivation
- "Sudden Stops" or reversals of capital inflows and the subsequent withdrawal of international capital are identified as triggers of several recent crises, notably Thailand and South Korea.
- Thailand experienced exceptionally large capital inflow reversals:
  - cumulative inflows as a percent of GDP of approximately 51.5 percent between 1988 and 1994;
  - reversals of 26 percent between 1996 and 1997.
- Sudden stops typically require offsetting by reserve losses or lower current account deficits, and often produce:
  - contractions in output via large and unexpected swings in relative prices;
  - increased financial vulnerability through reserve losses;
  - higher incidence of nonperforming loans (NPLs) due to higher interest burdens and currency mismatches;
  - bank lending retrenchment and a "credit crunch", especially affecting small- and medium-size firms.

### Research objective and scope
- Aim: characterize factors that exacerbated financial constraints—proxied by adjustment of fixed assets—experienced by Thai nonfinancial firms during sudden stop episodes.
- Focus on firm-level and balance-sheet characteristics that induced firms to meet debt obligations through adjustment of fixed assets.
- Specific firm characteristics considered:
  - level of internal resources;
  - firm size;
  - issuance of American Depository Receipts (ADRs);
  - tradable vs. nontradable commodity production;
  - degree of foreign ownership;
  - multinational status;
  - degree of macroeconomic instability;
  - industry affiliation;
  - balance sheet deterioration (profitability and debt maturity structure);
  - short-term debt exposure and fluctuations in domestic demand.

### Data description and sample
- Primary data source: DataStream (annual corporate balance sheet and income statement data).
- Sample: 284 nonfinancial Thai firms publicly listed on the local stock market between the years 1992 and 2001.
  - Because of data limitations, the sample of firms is limited to those that remained in business (bankrupt/delisted firms are not included).
- Sector classification:
  - Tradable sectors: primary commodities, manufactures, household products and food.
  - Nontradable sectors: services and real estate.
- Sectoral composition and characteristics:
  - Service sector: 68 firms (largest sector).
  - Real estate sector: 28 firms (smallest sector).
  - Primary product sector: highest level of sales on average.
  - Household sector: lowest average sales.
  - Real estate and primary product sectors: tend to be largest in terms of size.
  - Food and household product sectors: smallest in terms of size.
  - After-tax profits: food sector has the highest profitability levels.
  - Tradable sector: characterized by having higher profits than the nontradable sector and by being more exposed to short-term debt.
  - Nontradable firms: tend to be of relatively larger size.

### Key descriptive findings and time patterns
- Fixed assets and capital flows:
  - Thai nonfinancial firms suffered a significant decline in their fixed assets throughout the capital outflow period.
  - Table III reveals the sudden stop episode led to a significant decline—of close to 50 percent—in the average growth of firms’ fixed assets.
  - Graph A in Figure II depicts the decline beginning in 1996 and not reversing until mid-1999; by mid-1999 the trend gradually reverses with signs of slower growth.
- Macroeconomic and sectoral demand:
  - Sector consumption growth averaged 6.2 percent during the 1990s except during the sudden stop episode when it declined to an average growth rate of -9 percent.
- Debt structure and tradable sectors:
  - Tradable sectors were taking significant amounts of short-term debt—about 80 percent of total debt prior to the crisis period—which significantly worsened their balance sheets once capital inflow reversals took place.
- Cross-country context on leverage (from cited literature):
  - Debt-equity ratios by region as cited: U.S. firms averaged 90 percent by end-1996, Latin American firms averaged 31 percent, while Thai firms averaged 155 percent.

### Regression and comparative findings (high-level)
- Determinants of fixed asset adjustment vary by period:
  - Tranquil period: adjustment of fixed assets can be mostly explained by lagged variables, profitability, domestic consumption demand, and firm size.
  - Sudden Stop period: two additional firm characteristics help explain adjustment in firms’ favor:
    - being a tradable goods producer;
    - having a longer-term maturity structure of debt.
  - Profitability is significant in tranquil times but no longer matters in Sudden Stop periods.
- Additional empirical findings:
  - Thai ownership in tradable sectors helps firms be less financially constrained during sudden stop episodes.
  - Multinational firms decelerate their fixed asset growth during tranquil times, but only for firms producing nontradable output.
  - Growth of fixed assets of nontradable output firms is more affected by domestic consumption growth than that of tradable firms.
  - ADR issues play in a firm’s favor but only during tranquil times.
- Relation to literature:
  - Findings align with and extend work such as Aguiar (2004) and Bleakley and Cowan (2004) on balance sheet effects and investment adjustments; highlight the role of short-term foreign currency debt exposure in slowing investment during sudden stops.

### Structure of the rest of the paper (as provided)
- Section II: describes the data and provides summary statistics.
- Section III: discusses the empirical evidence.
- Section IV: concludes.
- Appendix: detailed definitions of variables used and their sources.

*Source: _wp0711 - 1. Summary Statistics by Industry (excerpt).*

### 5.6 percent during the nonsudden stop episode, but declined to an average growth rate of

### _wp0711 - 5.6 percent during the nonsudden stop episode, but declined to an average growth rate of

### Key empirical patterns and macro outcomes
- Average growth in exports and sectoral inflation rose during the sudden stop because of the devaluation of the Thai baht: export growth increased from an average of 2.5 percent during tranquil periods to 4.5 percent after the devaluation; sectoral inflation rates increased from an average of 2.8 percent during tranquil times to 5.9 percent after the devaluation.  
- Consumption and aggregate firm outcomes:
  - Average consumption growth in Thailand averaged 6.2 percent during the 1990s except during the sudden stop episode when it declined to an average negative growth rate of -9 percent.
  - A prior nonsudden stop episode average growth reported as 5.6 percent, but declined to an average growth rate of -4.2 percent during the sudden stop period.  
- Aggregate decline in firms’ fixed assets during the sudden stop: a decline of approximately 30 percent.

### Firm-level sample and descriptive statistics
- Sample coverage: data from 284 nonfinancial firms listed in the Thai stock market between 1992 and 2001.
- Actual number of firms per year in the dataset: 1992=150, 1993=183, 1994=225, 1995=249, 1996=275, 1997=278, 1998=275, 1999=273, 2000=266, and 2001=250.
- BIS lending (alternative measure to capture capital outflow):
  - Lending by BIS-reporting banks to the Thai nonfinancial private sector reached a peak of approximately US$40 billion during the second quarter of 1996 and then declined without recovering, but stabilized at US$15 billion.

### Firm financial distress and observable indicators
- Interest Coverage Ratio:
  - Average Interest Coverage Ratio declined from 21.9 during good times to 7.3 during the sudden stop period.
  - Note: the coefficients reflecting the interest coverage ratio between tranquil and sudden stop episodes are not statistically significant at conventional levels.
- Observed firm behavior consistent with fire sales and liquidity-driven asset disposals:
  - Substantial declines in the growth of firms’ fixed assets occurred around capital inflow reversals and abrupt devaluation, implying sales of assets at a discount (fire sales) driven by uncertainty and financial constraints.

### Heterogeneity across firm characteristics (descriptive and graphical)
- Debt maturity:
  - Firms with a relatively shorter debt-maturity structure suffered a steeper decline in their fixed assets during the sudden stop.
  - A shorter-term debt maturity structure led Thai firms during the sudden stop episode to a 0.2 percent decrease in their annual fixed asset growth.
- Multinationals:
  - Multinational firms adjusted their fixed assets downward drastically starting in 1997 and did not show clear recovery during the sample period.
  - Being multinational leads to a 0.1 percent decline in firms' fixed assets growth during tranquil times in the sample; not a significant factor during the sudden stop period in aggregate regressions, though effects differ by sector.
- Firm size:
  - Small firms suffer from a steeper decline in growth of fixed assets than larger ones.
  - When a firm becomes on average larger than the median, its fixed assets tend to grow by 0.1 percent (effect significant even during sudden stop episodes).
- ADR issuance and external financing:
  - Issuance of ADRs allows firms to have a higher growth rate of fixed assets during tranquil times and a faster recovery during downturns.
- Tradable vs. nontradable sectors:
  - Nontradable firms suffered a more pronounced and longer decline in the growth of their fixed assets after 1997 relative to tradable firms.
  - Plausible mechanism: tradable firms could partially offset domestic demand declines by selling abroad and gaining foreign exchange during devaluation episodes, reducing need to sell fixed assets.

### Main regression findings (random effects model; reduced form investment equation)
- Model structure:
  - Dependent variable: I_ijt / K_ijt-1 (adjustment of fixed assets of firm i, sector j, time t).
  - Regressors include lagged investment, profitability, financing costs (shocks to balance sheet), sectoral consumption, and firm characteristics.
  - Sudden stop dummy (SS) identifies 1997 and 1998 and is interacted with regressors to capture differential behavior during the sudden stop.
- Broad conclusions:
  - Shocks to net worth—profitability and maturity structure—matter for fixed assets adjustment, but in different periods: profitability matters in tranquil periods; maturity structure matters during the sudden stop.
  - During tranquil periods, fixed asset fluctuations depend primarily on profitability, size, multinational status, and ADR issuance.
  - During sudden stop episodes, characteristics that reduce likelihood of investment postponement or forced asset sales include holding long-term maturity debt and being a tradable goods producer.
- Quantitative effects and statistical results (exact reported estimates and significance statements):
  - During tranquil periods a unit increase in profitability leads to a 0.4 percent increase in fixed assets growth in the following period.
  - Being multinational leads to a 0.1 percent decline in fixed assets growth during tranquil times (not significant during sudden stop in aggregate sample).
  - Firm size: being larger than the median leads to a 0.1 percent increase in fixed assets growth (significant in tranquil and sudden stop).
  - Short-term maturity interaction: during the sudden stop, short-term debt structure associated with a 0.2 percent decrease in annual fixed asset growth.
  - For tradable goods producers, sectoral export growth (sectoral exports) does not matter for fixed asset growth in tranquil or sudden stop periods in these regressions (contrary to Aguiar (2004) who uses firm-level exports).

### Tradable versus nontradable producers (differential regression results)
- Profitability vs. maturity:
  - Profitability matters primarily during tranquil periods; exposure to short-term maturity debt is significant during sudden stop periods.
  - Impacts of profitability and maturity structure are larger for nontradable sector firms.
- Domestic demand sensitivity:
  - For nontradable goods producers, annual percentage increases in consumption lead to increases in growth of firm fixed assets of 1.8 percent during tranquil periods and 2.0 percent during sudden stop periods.
  - These effects are roughly four times larger than for tradable firms.
- Sector-specific firm characteristics:
  - For tradable firms, having a high degree of Thai ownership helped increase fixed assets growth by 0.1 percent during the sudden stop period.
  - For nontradable firms, being multinational reduces growth of fixed assets during tranquil times (multinational negative effect concentrated in nontradable sector).

### Debt structure: long-term vs. short-term maturity holders
- Model split by debt maturity structure yields:
  - Firms holding longer-term debt are less financially constrained and less likely to engage in constrained adjustment of physical capital during crises.
  - For short-maturity holders, firm size and tradable-producer status matter positively for fixed asset growth in both good and bad times.
  - For long-maturity holders, during tranquil times additional factors such as being multinational or an ADR issuer matter for fixed assets; during crisis times fixed assets adjustments of long-term debt holders depend mainly on firm size.
- Significance notes:
  - Size remains significant for both short- and long-term debt holders (though significance levels differ).
  - ADR issuance matters in tranquil times for long-maturity holders (reduces likelihood of fixed asset adjustment).

### Robustness and sensitivity checks
- Market-to-book value: found to be insignificant and excluded from benchmark specification.
- Firm age: one year of aging decelerates firms’ fixed asset growth by 0.01 percent during tranquil times (insignificant during sudden stop); sample size reduced to 413 observations when included.
- Quadratic specification for consumption: main conclusions unaffected; impact of quadratic consumption growth about half that of benchmark but maturity structure remains significant during sudden stop.
- Alternative continuous measure for capital outflow: BIS lending to Thai private sector used instead of binary sudden stop dummy—results show very little difference in significance and magnitude of coefficients; slight increase in magnitude of significant coefficients when using BIS lending interactive variable.

### Policy-relevant implications and interpretation
- Financial constraints mediated by debt maturity structure critically influence firms’ asset adjustment during sudden stops; policies that ease rollover risk or extend debt maturities could mitigate forced asset sales.
- Tradable sector exposure and access to foreign exchange can cushion firms during devaluation episodes; policies facilitating export channels or foreign-currency revenues may reduce fire-sale pressures.
- Firm-level factors—size, ADR access, multinational affiliation, and ownership structure—alter vulnerability; targeted financial support or liquidity backstops could be prioritized for smaller, nontradable, short-maturity firms during sudden stops.
- Future research recommended: directly measure forced investment adjustments with price pressure to capture fire sale dynamics, and test findings in other regions/markets that experienced similar episodes.

*Source: IMF working paper (_wp0711), analysis of 284 Thai nonfinancial firms listed 1992–2001, including Tables IV–VII, Figures I–II, and Data Appendix.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0711.pdf_