## _wp09164

## Source details

**Canonical URL:** [_wp09164](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09164.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09164.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09164.pdf.json)

---

### I. Introduction and objectives
- Financial globalization can bring capital, knowledge, and discipline, but empirical evidence is mixed.
- A specific downside is increased vulnerability to a financial crisis, especially when capital inflows are skewed toward non-FDI types such as bank lending and portfolio flows.
- Paper objectives:
  - Assess whether manufacturing firms in emerging economies experienced a liquidity crunch (beyond falling demand) during the 2007-2009 crisis.
  - Examine whether the pre-crisis volume and composition of capital inflows systematically affect the severity of the credit crunch across countries.
- Crisis context: global crisis started in August 2007 in the United States and evolved into a global financial and economic crisis through 2008–2009.

### II. Data and sample
- Sample and coverage:
  - Data on 3823 manufacturing firms in 24 countries (manufacturing sectors = U.S. SIC 3-digit codes 200–399).
  - Stock price log differences for manufacturing firms from 24 emerging countries and 20 developed economies over end-July 31, 2007 to end-December 31, 2008.
- Key sample statistics:
  - Log difference of stock price index among emerging economies: 81.8% mean; Standard deviation 66.7%.
  - Aggregate sample totals (Table 1 bottom line): Total obs: 7,911; Median change in stock price (log): -77.45; Mean change in stock price (log): -84.95; Std Dev: 73.98; Min: -764.01; Max: 264.45.
  - Table 2a selected: Change in stock price (log): Obs# 3,823; Median -77.8; Mean -81.8; Std Dev 66.7; Min -347.2; Max 55.4.
  - DEF_INV: Obs# 3,796; Median 0.2; Mean 0.2; Std Dev 0.3; Min 0.0; Max 1.0.
  - DEF_WK: Obs# 3,823; Median 86.8; Mean 88.5; Std Dev 28.5; Min 22.3; Max 169.2.
  - Sector-level demand sensitivity: Obs# 3,819; Median 1.4; Mean 1.5; Std Dev 0.7; Min -1.1; Max 4.3.
  - Company size: Obs# 3,823; Median 14.5; Mean 15.0; Std Dev 2.7; Min 9.0; Max 25.1.

### III. Key variables and empirical specification
- Firm-level financial dependence measures:
  - DEF_INV (investment dependence) constructed following Rajan and Zingales (1998): (capital expenditures - cash flow) / capital expenditures; cash flow = cash flow from operations + decreases in inventories + decreases in receivables + increases in payables.
  - DEF_WK (working-capital dependence) based on cash conversion cycle: 365 * (inventories / cost of goods sold + accounts receivable / total sales - account payables / total sales); sector medians from U.S. firms (1990–2006) used as benchmarks. Median = 71 days; mean = 71 days; standard deviation = 41 days.
- Basic cross-sectional specification (equation (1)):
  - StockReturn_{i,k,j} = country fixed effects + β * FinancialDependence_k + Control_{i,k,j} + ε_{i,k,j}
  - i = company, k = sector, j = country; key regressors pre-determined (2006).
- Country-level interaction (equation (2)):
  - β_j = β_1 + β_2 * Pattern_of_Capital_Flow_j
  - Pattern_of_Capital_Flow_j measured by either total pre-crisis capital inflow volume (2002–2006 average) or composition (FDI v. non-FDI).
- Controls:
  - Fama-French three factors: firm size (log assets), market-to-book ratio, beta (five-year correlation of firm stock and country market).
  - Momentum factor (Jan 31, 2007 to Jun 30, 2007 return).
  - Sector demand-sensitivity index (Tong and Wei (2008)); four-factor variables following Whited and Wu (2006).

### IV. Main empirical findings — extent of financial constraint (stock-price evidence)
- DEF_WK effects:
  - DEF_WK consistently negative and statistically significant across specifications.
  - Representative coefficients (Table 3): -0.156**, -0.154**, -0.139**, -0.123**, -0.136***, -0.130** with standard errors [0.0627], [0.0645], [0.0618], [0.0545], [0.0510], [0.0516].
  - Economic magnitude: increase in DEF_WK from 25th to 75th percentile (from 35 to 95 days) → extra decline in stock price by 9.3 percentage points.
- DEF_INV effects:
  - DEF_INV estimates vary; some negative but often statistically insignificant in baseline (example Table 3 Column 1: -2.893; other entries vary).
  - Correlation between DEF_INV and DEF_WK = 0.04.
- Additional firm characteristics:
  - Beta*Market Return: positive and highly significant (e.g., coefficients 0.326***, 0.310***, 0.303*** with standard errors [0.0440], [0.0440], [0.0426]).
  - Momentum: negative and highly significant (e.g., -0.145*** with [0.0399]).
  - Demand sensitivity: strong negative effect (e.g., -9.350*** [2.062]).
  - Leverage: large negative effect where included (e.g., -35.44*** [4.453]; -36.89*** [4.605]).
  - Firm size: positive in some specifications (e.g., 2.643** [1.093]; 2.842** [1.090]).
  - Market-to-book: negative in some specifications (e.g., -1.166* [0.672]; -1.250* [0.669]).
- Interpretation:
  - Evidence of a worsening credit crunch in emerging market economies in 2008: firms with top-quartile DEF_WK experienced a greater decline in stock prices by at least nine percentage points relative to bottom-quartile DEF_WK firms during the crisis period.
  - Average effects statistically significant but not quantitatively overwhelming relative to total fall in stock prices (more than half).

### V. Role of pre-crisis exposure to international finance — volume and composition
- Pre-crisis exposure measure:
  - Annual inflow of capital over GDP averaged 2002–2006 (de facto); de jure AREAER measures used in robustness checks.
- Volume (aggregate inflow) results (Table 5):
  - Interactions of total capital inflow volume with DEF_INV and DEF_WK mostly insignificant; weak indication of DEF_WK × Inflow significance at 10% in some specs (examples: -0.00778* [0.00468], -0.00846* [0.00479]).
  - Conclusion: volume alone does not strongly explain cross-country variation in liquidity crunch.
- Composition results (disaggregated into FDI, FPI, foreign loans; Table 6):
  - Time-series pattern: all components rose pre-crisis; reversal sharpest for international bank loans; FDI comparatively stable with milder reversal.
  - Cross-country suggestive evidence: higher pre-crisis FDI share associated with smaller magnitude of capital reversal; slope coefficient = 2.64 (standard error 1.76) (Figure 2).
  - Formal regressions — Representative interaction coefficients (Table 6):
    - DEF_INV × FDI: positive and sometimes significant (examples: 3.375** [1.627]; 3.240* [1.661]; 3.480** [1.606]).
    - DEF_INV × FPI: negative and sometimes significant (examples: -1.626* [0.909]; -1.503* [0.789]; -1.387* [0.799]).
    - DEF_INV × Foreign loan: negative but less precisely estimated (examples: -2.531 [1.651]; -2.491 [1.670]).
    - DEF_WK × FDI: positive and sometimes significant (examples: 0.0441** [0.0216]; 0.0407* [0.0226]).
    - DEF_WK × FPI: negative and statistically significant (examples: -0.0219*** [0.00817]; -0.0218** [0.00862]).
    - DEF_WK × Foreign loan: negative and statistically significant (examples: -0.0555*** [0.0172]; -0.0585*** [0.0195]).
  - Magnitudes: foreign loans coefficient more than twice that on FPI (consistent with faster reversal/non-renewal of international loans).
- Mechanisms highlighted:
  - Domestic banks’ reliance on international wholesale funding transmits capital reversals to domestic credit supply (Figure 3; Korea example cited).
  - Multinational internal capital markets via FDI can alleviate subsidiary liquidity constraints.
  - Withdrawal of portfolio capital raises rollover costs and seasonal equity offering constraints, tightening firm liquidity.

### VI. Robustness checks and extensions
- Domestic financial development:
  - Interaction of sector finance dependence with private credit/GDP end-2006 not significant; adding these controls does not change capital flow results (Table 7).
- Pre-crisis window variations:
  - Extending pre-crisis window to include 2007 strengthens some results: DEF_INV × FDI positive and significant at 1%; FPI significantly negative at 1%; foreign loans negative at 5% (Table 7 Column 3).
- De jure openness (AREAER 2006):
  - De jure and de facto measures positively correlated but imperfect: correlations = 0.38 (direct investment), 0.25 (portfolio), 0.37 (loans).
  - Regressions with de jure: pre-crisis FDI openness alleviates constraint for DEF_INV; pre-crisis openness to FPI worsens constraint for DEF_WK (Table 7 last column).
- Contemporaneous betas and alternative LHS:
  - Contemporaneous beta (weekly returns Jul 31, 2007–Dec 31, 2008) coefficient ≈ 0.93 (t-stat = 11.42) (Table 8 Column 1); does not alter DEF_WK significance (~ -0.12) or composition results.
  - Alternative LHS normalization ([logP_dec08 – logP_july07] / (½)[logP_dec08 + logP_july07]) yields same qualitative patterns (Table 8 Columns 3–4).
- Weighted regressions and sample restrictions:
  - Weighted least squares (weights = inverse sqrt(number of manufacturing stocks in country)) preserves negative significance of DEF_WK × pre-crisis FPI and foreign loans (Table 8 Column 5).
  - Restricting to countries with ≥ 25 manufacturing stocks (19 countries) yields same sign patterns; significant interactions include FDI × DEF_INV and FPI/foreign loans × DEF_WK (Table 8 Column 6).
- Sample extensions and demand channels:
  - Expanding to all non-financial firms increases sample by 50%; sign patterns unchanged though significance weaker (Table 9).
  - Interacting demand-sensitivity proxies with capital flows: FDI × pro-cyclicality dummy positive and significant; other interactions insignificantly negative; DEF_INV/DEF_WK results unaffected.
- Controlling for global growth opportunities:
  - USGrowth (median real annual growth rate by US SIC 3-digit sector, 1990–2006) interactions not significant (p-values > 0.4); liquidity-crunch patterns remain.

### VII. Placebo and event-study tests
- Placebo test (capital flows 2002–2005 vs. stock returns Jan 1, 2006–Jun 30, 2007):
  - No significant average effect for DEF_INV or DEF_WK; volume and most component interactions insignificant.
  - One weak result: FDI × DEF_INV significant at 10% in a specification but becomes insignificant with sector fixed effects.
  - Interpretation: baseline patterns are features of crisis periods, not normal times.
- Lehman Brothers event study (Sep 12–16, 2008; Table 11):
  - Short-window results (with sector fixed effects and firm controls):
    - Pre-crisis FDI × DEF_INV: significantly positive at 1% (examples: 0.332** [0.129]; 0.316*** [0.117]).
    - Pre-crisis non-FDI × DEF_INV: negative.
    - FPI and foreign loans × DEF_WK: significantly negative in several specifications.
  - Confirms earlier findings: FDI alleviates constraints; pre-crisis reliance on non-FDI tightens constraints during acute crisis episodes.

### VIII. Conclusion — policy-relevant insights
- Aggregate capital inflow volume alone does not robustly predict the severity of firm-level liquidity crunches during 2007–09; composition matters.
- Clear compositional pattern:
  - Higher pre-crisis exposure to foreign portfolio investment (FPI) and foreign loans → more severe liquidity shocks for firms, especially those dependent on external finance for working capital (DEF_WK).
  - Higher pre-crisis exposure to foreign direct investment (FDI) → less severe liquidity shocks; FDI acts as a stabilizing factor.
- Policy implication: analyses and policy design should disaggregate capital flows (FDI vs. non-FDI) when assessing vulnerability to liquidity crunches in crises.
- Caveat / future research: not a comprehensive welfare assessment; understanding longer-run effects in tranquil times and broader welfare consequences of flow composition remains an open research area.

*Source: IMF Working Paper (conclusion and empirical results sections of _wp09164).*

### Conclusion..............................................................................................................

### Conclusion

### I. Introduction and objectives
- Financial globalization can bring capital, knowledge, and discipline, but empirical evidence is mixed.
- A specific downside is increased vulnerability to a financial crisis, especially when capital inflows are skewed toward non-FDI types such as bank lending and portfolio flows.
- The paper has two objectives:
  - Assess whether manufacturing firms in emerging economies experienced a liquidity crunch (beyond falling demand) during the 2007-2009 crisis.
  - Examine whether the pre-crisis volume and composition of capital inflows systematically affect the severity of the credit crunch across countries.
- Context: the global crisis started in August 2007 in the United States and evolved into a global financial and economic crisis through 2008–2009.

### II. Data and sample
- Sample: data on 3823 manufacturing firms in 24 countries (manufacturing sectors = U.S. SIC 3-digit codes 200–399).
- Table coverage: stock price log differences for manufacturing firms from 24 emerging countries and 20 developed economies over the period from end-July 31, 2007 to end-December 31, 2008.
- Stock price change statistics among emerging economies:
  - Log difference of stock price index: 81.8% on average.
  - Standard deviation: 66.7%.
  - Large cross-country dispersion: Poland and Russia experienced the largest decline; Mexico and Thailand the smallest.
- Firm-level constructed indices use historical U.S. firm data from 1990–2006 and SIC 3-digit sectors expanded to 253 sectors (versus 36 in Rajan and Zingales (1998)).

### III. Key variables and empirical specification
- Two firm-level measures of intrinsic dependence on external finance:
  - DEF_INV: Intrinsic dependence on external finance for investment, constructed following Rajan and Zingales (1998) with formula:
    - Dependence on external finance for investment = (capital expenditures - cash flow) / capital expenditures
    - Cash flow = cash flow from operations + decreases in inventories + decreases in receivables + increases in payables.
    - Calculated using U.S. firms as the benchmark for least likely to suffer financing constraints.
  - DEF_WK: Intrinsic dependence on external finance for working capital, modified from Raddatz (2006) using 1990–2006 data.
- Basic cross-sectional specification (equation (1)):
  - StockReturn_{i,k,j} = country fixed effects + β * FinancialDependence_k + Control_{i,k,j} + ε_{i,k,j}
  - i = company, k = sector, j = country; key regressors are pre-determined (2006).
- Country-level interaction (equation (2)):
  - β_j = β_1 + β_2 * Pattern_of_Capital_Flow_j
  - Pattern_of_Capital_Flow_j measured by either total volume of pre-crisis capital inflows or composition (FDI v. non-FDI).
- Control variables guided by asset-pricing models:
  - Fama-French three factors: firm size (log assets), market-to-book ratio, beta.
  - In some specifications, a momentum factor from Lakonishok, Shleifer, and Vishny (1994).
  - Sector-level intrinsic sensitivity to demand contraction (Tong and Wei (2008)).
  - Four-factor variables entered directly following Whited and Wu (2006).

### IV. Main empirical findings
- Evidence of a worsening credit crunch in emerging market economies in 2008:
  - Firms with DEF_WK in the top quartile experienced a greater decline in stock prices by at least nine percentage points relative to firms with DEF_WK in the bottom quartile during the crisis period.
  - While average effects are statistically significant, they are not quantitatively overwhelming relative to the total fall in stock prices (more than half).
- Role of country-level exposure to capital flows:
  - Total volume of pre-crisis capital inflows is not systematically related to the severity of the credit crunch.
  - Composition matters: large pre-crisis exposure to non-FDI capital inflows tends to be associated with a more severe credit crunch during the crisis.
  - Pre-crisis exposure to FDI does not exacerbate the credit crunch.
  - Interpretation: different types of capital flows bring different benefits and costs; non-FDI flows (bank lending and portfolio flows) are more prone to reversal and link to domestic liquidity provision via foreign borrowing to domestic banks.

### V. Interpretation, robustness, and related literature links
- DEF_INV and DEF_WK are statistically significant with expected signs in most regressions; control variables reduce the magnitude of DEF_INV but have little impact on DEF_WK.
- DEF_WK may capture aspects of firm risk during the crisis not fully explained by the three-factor or four-factor asset-pricing models.
- The analysis is connected to two literatures:
  - Credit crunches and their origins/consequences (e.g., Bernanke and Lown, 1991; Borensztein and Lee, 2002; and others).
  - Benefits and costs of financial globalization and the differing effects of capital flow composition on volatility, persistence, and crisis vulnerability.
- The 2007–2009 crisis differs from typical balance-of-payments or home-grown crises studied in prior literature, providing a unique opportunity to test composition effects in cross-country transmission of liquidity shocks.

### VI. Empirical approach and robustness checks (overview)
- Cross-sectional prediction exercise using ex ante firm-level classifiers of liquidity dependence (pre-determined in 2006) to forecast ex post stock performance from July 31, 2007 to Dec 31, 2008.
- Controls include size, market-to-book, beta, momentum, and sector demand sensitivity.
- Findings hold after accounting for these controls and after exploring country-level interactions with measures of pre-crisis capital flow patterns.

*Source: IMF Working Paper (conclusion section of _wp09164).*

### 2006. Third, we calculate the sector-level median from firm ratios for each SIC 3-digit sector

### _wp09164 - 2006. Third, we calculate the sector-level median from firm ratios for each SIC 3-digit sector

### Construction of sector-level finance-dependence indices
- Rajan-Zingales (RZ) index: sector-level median of firm ratios for each SIC 3-digit sector containing at least 5 firms; winsorized to range between 0 and 1 to capture percentage of capital expenditure financed externally.
- Intrinsic dependence on external finance for working capital (DEF_WK):
  - Built from the "cash conversion cycle": 365 * (inventories / cost of goods sold + accounts receivable / total sales - account payables / total sales).
  - Inventories, accounts receivable, and accounts payable are year-end numbers; costs of goods and sales are aggregated over the year.
  - For U.S. firms (1990–2006, Compustat USA Industrial Annual): compute firm-level cycle, take SIC 3-digit sector median, then extrapolate sector medians to other countries.
  - Median = 71 days; mean = 71 days; standard deviation = 41 days.

### Control variables and summary statistics
- Firm-level controls (from Worldscope and Datastream) include:
  - Three Fama-French factors: firm size (log of book assets), market asset to book asset ratio, and beta (firm-level market beta based on correlation of monthly firm stock price and country-level market index over past five years).
  - Momentum factor: stock return from January 31, 2007 to June 30, 2007.
  - Domestic beta preferred (Griffin, 2002).
- Sector-level demand-sensitivity index:
  - Constructed from mean change in log stock price for US firms in each SIC 3-digit sector from September 10, 2001 to September 28, 2001 (Tong and Wei, 2008).
  - Excluding financial sector firms yields 361 3-digit sectors.
- Tables reported: Table 2a (summary statistics) and Table 2b (pair-wise correlations).

### Empirical analysis — extent of financial constraint (stock-price evidence)
- Dependent variable: percentage change in stock price (difference in log price) from July 31, 2007 to December 31, 2008 for manufacturing firms in 24 emerging countries.
- Key findings:
  - DEF_INV (dependence for external finance for investment) alone: negative but statistically insignificant (Column 1, Table 3).
  - DEF_WK alone: negative and significant at the 5% level (Column 2).
  - DEF_WK retains magnitude and sign when DEF_INV added (Column 3); correlation between DEF_INV and DEF_WK = 0.04.
  - Economic magnitude: increase in DEF_WK from the 25th to the 75th percentile (from 35 to 95 days) leads to an extra decline in stock price by 9.3 percentage points.
  - Adding beta*market return: positive and significant coefficient (firms with smaller beta experience smaller reduction).
  - Firm size: positive (larger firms may have better credit access).
  - Market-to-book ratio: firms with high market-to-book experience greater decline.
  - Sector demand sensitivity: significantly negative when added (Column 5).
  - Pre-crisis leverage: leveraged firms suffered greater stock price declines (Column 6).
  - Trade-sensitivity index (constructed by regressing firm annual return on annual % change of 3-digit SIC sector exports, 1992–2006): coefficient not statistically significant (0.05 with standard error 1.76); reclassifying negative values to zeros yields negative but still insignificant coefficient. Adding trade sensitivity does not alter DEF_WK results.

### Role of pre-crisis exposure to international finance — volume and composition
- Pre-crisis exposure measure: annual inflow of capital over GDP averaged 2002–2006 (de facto); alternative de jure measures from AREAER used in robustness checks.
- Aggregate volume results (Table 5):
  - Interactions of total capital inflow volume with DEF_INV and DEF_WK mostly insignificant; weak indication of significance for capital flow × DEF_WK at 10% in some specifications.
  - Conclusion: volume alone does not strongly explain cross-country variation in liquidity crunch.
- Composition results (disaggregating inflows into FDI, FPI, and foreign loans; IMF IFS definitions; Figures and Tables referenced):
  - Time-series pattern (Figure 1): all components rose pre-crisis; reversal sharpest for international bank loans; FDI comparatively stable and reversal milder (reversal started in 2008).
  - Cross-country suggestive evidence (Figure 2): higher pre-crisis FDI share associated with smaller magnitude of capital reversal; slope coefficient = 2.64, statistically significant at the 15% level.
  - Formal regressions (Table 6) — interactions of flow components with DEF_INV and DEF_WK:
    - DEF_INV × FPI: significantly negative (Column 1).
    - DEF_WK interactions (Column 2): FDI coefficient positive and significant at 5%; FPI and foreign loans coefficients negative and significant at 1%.
    - Magnitudes: foreign loans coefficient more than twice that on FPI (consistent with faster reversal/non-renewal of international loans).
    - When summing component interactions with DEF_WK plus coefficient on DEF_WK itself (–0.153), overall effect still aligns with earlier finding that higher DEF_WK associates with greater decline in stock prices.
  - Sector fixed effects (Column 3) sharpen asymmetry across flow components; firm-level controls (Columns 4–6) preserve patterns.
  - Leverage interactions: leverage ratio significantly negative on its own; interacting leverage with components yields positive coefficient for FDI and negative for FPI and foreign loans; does not alter DEF_INV/DEF_WK results.
- Mechanisms and examples:
  - Domestic banks’ reliance on international wholesale funding transmits capital reversals to domestic credit supply (Figure 3, Korea example, HSBC report Sep 09, 2008 cited).
  - Multinational internal capital markets can alleviate subsidiary liquidity constraints; pre-crisis cash-rich U.S. firms (Bates, Kahle, Stulz, 2007) may have helped via FDI in crisis.
  - Withdrawal of portfolio capital raises cost of rollover and seasonal equity offerings, tightening financial constraints.

### Robustness checks and extensions
- Domestic financial development interactions (private credit/GDP end-2006 and alternative sum proxy): interaction with sector finance dependence not significant; adding these controls does not change capital flow results (Table 7).
- Pre-crisis window variations:
  - Extending pre-crisis window to include 2007 strengthens results (Column 3, Table 7): DEF_INV × FDI positive and significant at 1%; FPI significantly negative at 1%; foreign loans move to significantly negative at 5%.
- De jure openness (AREAER 2006) disaggregated into FDI, FPI, foreign loans:
  - De jure and de facto measures positively correlated but imperfectly: correlations = 0.38 (direct investment), 0.25 (portfolio), 0.37 (loans).
  - Regressions with de jure: pre-crisis FDI openness alleviates constraint for DEF_INV; pre-crisis openness to FPI worsens constraint for DEF_WK (Table 7, last column).
  - Authors place more weight on de facto measures.
- Contemporaneous betas:
  - Weekly returns July 31, 2007–December 31, 2008 used to construct contemporaneous beta × local market return; coefficient ≈ 0.93, t-stat = 11.42 (Table 8 Column 1).
  - Contemporaneous beta does not alter DEF_WK significance (DEF_WK still –0.12) nor composition results (Table 8 Column 2).
- Alternative LHS definitions, weighted regressions, sample restrictions:
  - Alternative LHS normalization ([logP_dec08 – logP_july07] / (½)[logP_dec08 + logP_july07]) yields same qualitative patterns (Table 8 Columns 3–4).
  - Weighted least squares (weights = inverse sqrt(number of manufacturing stocks in country)) preserves negative significance of DEF_WK × pre-crisis FPI and foreign loans (Table 8 Column 5).
  - Restricting to countries with ≥ 25 manufacturing stocks (19 countries) yields same sign patterns; significant interactions include FDI × DEF_INV and FPI/foreign loans × DEF_WK (Table 8 Column 6).
- Sample extensions and demand channels:
  - Expanding to all non-financial firms increases sample by 50%; sign patterns unchanged though significance weaker.
  - Interacting demand-sensitivity proxies (FTSE/JSE pro-cyclicality dummy and Tong & Wei 2008 index) with capital flows: FDI × pro-cyclicality dummy positive and significant; other interactions insignificantly negative; DEF_INV/DEF_WK results unaffected.
- Controlling for global growth opportunities:
  - Construct USGrowth (median real annual growth rate by US SIC 3-digit sector, 1990–2006), Winsorize at 1%, interact with capital flow components.
  - USGrowth and its interactions not significant (p-values > 0.4) and do not affect DEF_INV interactions; liquidity crunch remains more serious for DEF_INV in countries with high exposure to foreign loans.

### Placebo and event-study tests
- Placebo test (capital flows 2002–2005 vs. stock returns Jan 1, 2006–June 30, 2007):
  - No significant average effect for DEF_INV or DEF_WK; volume and most component interactions insignificant.
  - One weak result: FDI × DEF_INV significant at 10% in Column 3 but becomes insignificant with sector fixed effects.
  - Interpretation: baseline patterns are features of crisis periods, not normal times.
- Lehman Brothers event study (Sept 12–16, 2008):
  - In the short window, with sector fixed effects and firm controls (Table 11 last column):
    - Pre-crisis FDI × DEF_INV: significantly positive at 1%.
    - Pre-crisis non-FDI × DEF_INV: negative.
    - FPI and foreign loans × DEF_WK: significantly negative.
  - Confirms earlier findings: FDI alleviates constraints; pre-crisis reliance on non-FDI tightens constraints during a crisis.

### Conclusion — policy-relevant insights
- Aggregate capital inflow volume alone does not robustly predict the severity of firm-level liquidity crunches during 2007–09; composition matters.
- Clear compositional pattern:
  - Higher pre-crisis exposure to foreign portfolio investment (FPI) and foreign loans → more severe liquidity shocks for firms, especially those dependent on external finance for working capital.
  - Higher pre-crisis exposure to foreign direct investment (FDI) → less severe liquidity shocks; FDI acts as a stabilizing factor.
- Implication: analyses and policy design should disaggregate capital flows (FDI vs. non-FDI) when assessing vulnerability to liquidity crunches in crises.
- Caveat / future research: this paper is not a comprehensive welfare assessment; understanding longer-run effects in tranquil times and broader welfare consequences of flow composition remains an open research area.

*Source: _wp09164 - 2006. Third, we calculate the sector-level median from firm ratios for each SIC 3-digit sector (IMF working paper excerpt).*

### REFERENCES

### REFERENCES

### Reference list
- Aguiar, Mark, and Gita Gopinath, 2005, “Fire-Sale Foreign Direct Investment and Liquidity Crises,”  Review of Economics and Statistics Vol. 87(3), pp 439-52.
- Almeida, Heitor,  Murillo Campello, Bruno Laranjeira, and Scott Weisbenner, 2009, “Corporate Debt Maturity and the Real Effects of the 2007 Credit Crisis,” NBER WP 14990.
- Bates, Thomas W., Kathleen M. Kahle, and René M. Stulz, 2007, “Why do U.S. Firms Hold So Much More Cash Than They Used To?” NBER Working Paper No. 12534
- Berg, Andrew, Eduardo Borensztein, and Catherine Pattillo, 2004, “Assessing Early Warning Systems: How Have They Worked in Practice?” IMF Working Paper 04/52.
- Bernanke, Ben S,  2008, Semiannual Monetary Policy Report to the Congress, February 27,

*Source: _wp09164 - REFERENCES*

### 2008. http://www.federalreserve.gov/newsevents/testimony/bernanke20080227a.htm

### _wp09164 - 2008 (excerpted empirical results and tables)

### Main empirical findings on firm-level stock returns during the 2007–08 crisis
- Dependent variable: change of stock price (log) from July 31, 2007 to December 31, 2008 (manufacturing firms in 24 emerging economies unless otherwise noted).
- Sample sizes reported:
  - Total observations (Table 1 summary): 7,911 firms.
  - Table 2a: Change in stock price (log) sample: 3,823 observations.
  - Table 3–8: typically 3,743–3,796 observations depending on specification; Table 9 (non-financial firms) up to 6,030 observations; Table 11 (Lehman week) up to 3,802 observations.
- Aggregate sample summary (Table 1, bottom line):
  - Total obs: 7,911
  - Median change in stock price (log): -77.45
  - Mean change in stock price (log): -84.95
  - Std Dev: 73.98
  - Min: -764.01
  - Max: 264.45

### Cross-country and firm-level summary statistics (Table 2a, Table 4 highlights)
- Table 2a summary (selected):
  - Change in stock price (log): Obs# 3,823; Median -77.8; Mean -81.8; Std Dev 66.7; Min -347.2; Max 55.4.
  - DEF_INV: Obs# 3,796; Median 0.2; Mean 0.2; Std Dev 0.3; Min 0.0; Max 1.0.
  - DEF_WK: Obs# 3,823; Median 86.8; Mean 88.5; Std Dev 28.5; Min 22.3; Max 169.2.
  - Demand sensitivity: Obs# 3,819; Median 1.4; Mean 1.5; Std Dev 0.7; Min -1.1; Max 4.3.
  - Company size: Obs# 3,823; Median 14.5; Mean 15.0; Std Dev 2.7; Min 9.0; Max 25.1.
  - Market/book: Obs# 3,823; Median 1.5; Mean 2.4; Std Dev 2.8; Min 0.3; Max 23.6.
  - Beta: Obs# 3,778; Median 0.64; Mean 0.71; Std Dev 0.65; Min -1.42; Max 3.45.
  - Momentum: Obs# 3,823; Median 20.77; Mean 26.45; Std Dev 37.54; Min -178.39; Max 331.42.
- Table 4: Pre-crisis exposure to capital inflows (% of GDP; averaged 2002–2006) — selected country-level values:
  - Argentina: Total Inflow 1.00; FDI 2.29; FPI -3.21; Foreign Loans 1.92; developed 0
  - Hong Kong (HK): Total Inflow 24.31; FDI 15.53; FPI -6.42; Foreign Loans 15.20; developed 0
  - Malaysia: Total Inflow 20.07; FDI 3.05; FPI 22.73; Foreign Loans -5.71; developed 0
  - Singapore: Total Inflow 30.45; FDI 14.11; FPI 3.89; Foreign Loans 12.46; developed 0
  - Ireland: Total Inflow 151.06; FDI 2.89; FPI 93.81; Foreign Loans 54.36; developed 1
  - United Kingdom (UK): Total Inflow 39.56; FDI 4.00; FPI 8.89; Foreign Loans 26.67; developed 1

### Estimated average effect of liquidity crunch and firm characteristics (Table 3)
- DEF_WK (external financial dependence for working capital):
  - Consistently negative and statistically significant across specifications.
  - Representative reported coefficients: -0.156**, -0.154**, -0.139**, -0.123**, -0.136***, -0.130** with standard errors [0.0627], [0.0645], [0.0618], [0.0545], [0.0510], [0.0516] respectively.
- DEF_INV (external financial dependence for investment):
  - Coefficient estimates vary across specifications (some negative large values noted but variable significance). Example values in Table 3: -2.893 (Case 1), -1.832 (Case 2) and other less negative/positive values in other cases; t-statistics/standard errors shown in brackets (e.g., [10.02], [8.276], [8.014], [7.164], [6.809], [7.152] for different cases).
- Beta*Market Return:
  - Positive and highly significant: reported coefficients 0.326***, 0.310***, 0.303***, 0.310*** with standard errors [0.0440], [0.0440], [0.0426], [0.0439].
- Momentum:
  - Negative and highly significant: e.g., -0.145***, -0.144***, -0.132***, -0.128*** with standard errors [0.0399], [0.0399], [0.0397], [0.0411].
- Demand Sensitivity:
  - Strong negative effect: e.g., -9.350***, -8.876***, -8.735*** with standard errors [2.062], [2.059], [2.204].
- Leverage:
  - Large negative effect where included: e.g., -35.44*** and -36.89*** with standard errors [4.453] and [4.605] respectively.
- Firm size and Market/Book show mixed effects; firm size positive in some specifications (e.g., 2.643**, 2.842** with [1.093], [1.090]); Market/Book negative in some (e.g., -1.166*, -1.250* with [0.672], [0.669]).

### Role of pre-crisis exposure to capital inflows — volume and composition effects
- Volume effect (Table 5):
  - Interaction DEF_WK*Inflow: negative and in some specifications statistically significant (e.g., -0.00778* [0.00468], -0.00846* [0.00479]).
  - DEF_INV*Inflow: coefficients reported 0.329, 0.442, 0.576 with standard errors [0.492], [0.455], [0.424] (not uniformly significant).
  - R-squared increases across specifications from 0.145 to 0.239 when adding industry fixed effects.
- Composition effect (Table 6):
  - DEF_INV*FDI: generally positive and sometimes statistically significant (examples: 2.859; 3.375** [1.627]; 3.240* [1.661]; 3.480** [1.606]; 3.610** [1.653]).
  - DEF_INV*FPI: negative and often significant (examples: -1.626* [0.909]; -1.503* [0.789]; -1.387* [0.799]; -1.499* [0.783]; -1.582* [0.814]).
  - DEF_INV*Foreign loan: negative but less precisely estimated (examples: -2.531 [1.651], -2.491 [1.670], -2.076 [1.798], -2.267 [1.768], -2.38 [1.839]).
  - DEF_WK interactions:
    - DEF_WK*FDI: positive and sometimes significant (e.g., 0.0441** [0.0216]; 0.0407* [0.0226]).
    - DEF_WK*FPI: negative and statistically significant (e.g., -0.0219*** [0.00817]; -0.0218** [0.00862]; -0.0198** [0.00816]).
    - DEF_WK*Foreign loan: negative and statistically significant (e.g., -0.0555*** [0.0172]; -0.0585*** [0.0195]; -0.0508*** [0.0192]).
  - Interpretation implied by coefficients: pre-crisis composition mattered — higher shares of FDI soften adverse returns associated with financial dependence, while higher shares of FPI or foreign loans exacerbate adverse returns for externally dependent firms.

### Robustness checks and subgroup analyses
- Robustness checks (Table 7 and Table 8):
  - DEF_INV*FDI remains positive and often significant across specifications (examples: 3.384* [1.724]; 3.861** [1.786]; 4.186** [1.751]).
  - DEF_INV*FPI often negative and sometimes significant (examples: -1.404* [0.821]; -1.329 [0.850]; -1.543** [0.612]).
  - DEF_WK*FPI and DEF_WK*Foreign loan remain negative and significant across many robustness specifications.
  - Alternative dependent variables, contemporary betas, weighted regressions, and alternative price-change measures preserve the main pattern: working-capital dependence (DEF_WK) associated with larger negative returns during the crisis, and the protective role of FDI vs harmful role of FPI/foreign loans in interactions.
- Non-financial firms (Table 9):
  - DEF_INV*FDI positive and often significant (examples: 2.732** [1.487]; 2.662** [1.437]; 2.994** [1.282]; 3.044** [1.240]).
  - DEF_INV*FPI negative and significant (examples: -1.272** [0.496]; -1.153** [0.570]; -1.549*** [0.548]; -1.395** [0.552]; -1.464*** [0.558]; -1.632*** [0.615]).
  - DEF_WK negative and significant in some specifications (e.g., -0.117** [0.0572]; -0.0990** [0.0456]).
  - DEF_WK*FDI sometimes positive and marginally significant (e.g., 0.0244* [0.0129]); DEF_WK*Foreign loan negative and sometimes significant.
- Placebo test (Table 10):
  - Using stock returns from Jan 1, 2006 to June 30, 2007 shows no consistent pattern of the same interactions as in crisis window. e.g., DEF_WK coefficients small and statistically insignificant in placebo windows; Beta*market index remains positive and significant (e.g., 0.143** [0.0603]).
- Lehman Brothers bankruptcy week (Table 11; Sep 12–16, 2008):
  - DEF_INV*FDI positive and statistically significant (e.g., 0.332** [0.129]; 0.284** [0.115]; 0.291** [0.116]; 0.316*** [0.117]; 0.330*** [0.121]).
  - DEF_INV*FPI negative and borderline/significant in some cases (e.g., -0.144* [0.0784]; other estimates -0.118 [0.0814], -0.0708 [0.0915]).
  - DEF_INV*Foreign loan negative in some specifications (e.g., -0.255* [0.134]).
  - DEF_WK interactions small during the Lehman week; several DEF_WK* components small but some DEF_WK*FPI and DEF_WK*Foreign loan negative and significant in panel variants (e.g., DEF_WK*Foreign loan -0.00373** [0.00169]).

### Figures — patterns in capital flows and banking stock responses
- Figure 1 (Capital Flow to Emerging Economies, US$ Billions) — series by instrument (Direct Investment, Portfolio Investment, Foreign Loans) plotted from 1999 to 2009; sample includes 24 emerging economies (source: IMF’s World Economic Outlook database).
- Figure 2 (Extent of capital reversal vs initial share of FDI):
  - Vertical axis: log(capital inflow/GDP) in 2009 – log(capital inflow/GDP) in 2007.
  - Horizontal axis: share of FDI inflow in total inflow in 2007.
  - Slope coefficient reported: 2.64 with standard error 1.76.
- Figure 3 (Change in log banking stock prices vs pre-crisis international bank loans — conditional scatter):
  - Vertical axis: change in log bank-sector stock price from July 1, 2007 to December 31, 2008.
  - Horizontal axis: initial international bank loans (log) — fitted values plotted by country.

### Key interpretation and policy-relevant implications drawn from empirical patterns (as reflected in tables)
- Firms with greater external dependence for working capital (DEF_WK) experienced significantly larger declines in stock prices during the crisis (negative DEF_WK coefficients across specifications).
- Pre-crisis capital inflow composition mattered:
  - Higher pre-crisis shares of FDI in a country’s capital inflows mitigated adverse stock-return effects for externally dependent firms (positive DEF_INV*FDI and DEF_WK*FDI interactions in many specifications).
  - Higher pre-crisis shares of portfolio investment (FPI) and foreign loans tended to exacerbate adverse stock-return outcomes for externally dependent firms (negative DEF_INV*FPI, DEF_WK*FPI, and DEF_WK*Foreign loan interactions).
- Leverage and demand-sensitivity amplified negative stock-return outcomes during the crisis (large negative leverage coefficients; strong negative demand sensitivity coefficients).
- Results are robust across multiple specifications, alternative dependent variables, weighted regressions, placebo period checks, and during specific crisis episodes (e.g., Lehman week), supporting the inference that both firm-level financial dependence and country-level capital-flow composition shape crisis vulnerability.

*Source: _wp09164 - 2008. (Tables and figures as provided in the supplied PDF excerpt.)*

### 2008.  On  the  horizontal  axis  is  the  pre-crisis  inflow  of  loans/GDP  averaged  over  2002-2006.  This  partial

### _wp09164 - 2008

### Pre-crisis inflow scatter plot and regression
- Horizontal axis: pre-crisis inflow of loans/GDP averaged over 2002-2006.
- Scatter plot conditioned on pre-crisis foreign direct investments and portfolio investments over GDP.
- Slope coefficient: -6.38
- Standard error: 3.24
- Source data: IMF’s WEO database and Datastream.

### Appendix Table 1. De Jure Financial Openness for Year 2006 — Key entries
- Table fields: Stocks | Bonds | Commercial Credit | Financial credit | FDI
- Argentina: 0 | 0 | 1 | 0 | 0
- Brazil: 0 | 1 | 1 | 1 | 0
- Chile: 1 | 1 | 1 | 1 | 1
- China: 0 | 0 | 0 | 0 | 0
- Colombia: 0 | 0 | 0 | 0 | 0
- Czech: 0 | 1 | 1 | 1 | 0
- Egypt: 1 | 1 | 1 | 1 | 0
- HK: 1 | 1 | 1 | 1 | 1
- Hungary: 1 | 1 | 1 | 1 | 1
- India: 0 | 0 | 0 | 0 | 0
- Indonesia: 0 | 0 | 0 | 1 | 0
- Israel: 1 | 1 | 1 | 1 | 1
- Korea: 1 | 1 | 1 | 1 | 0
- Malaysia: 1 | 1 | 0 | 0 | 0
- Mexico: 0 | 1 | 1 | 0 | 0
- Pakistan: 1 | 1 | 1 | 1 | 0
- Peru: 1 | 1 | 1 | 1 | 1
- Philippines: 1 | 0 | 0 | 0 | 1
- Poland: 1 | 0 | 1 | 0 | 0
- Russia: 0 | 0 | 1 | 0 | 0
- Singapore: 1 | 1 | 1 | 1 | 1
- South Africa: 1 | 1 | 1 | 0 | 1
- Thailand: 0 | 0 | 0 | 1 | 1
- Turkey: 1 | 1 | 0 | 0 | 1

### Data source for Appendix Table 1
- Source: The IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions in 2006.

*Source: IMF’s WEO database and Datastream; The IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions in 2006.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09164.pdf_