## wp18158

## Source details

**Canonical URL:** [wp18158](https://www.imf.org/-/media/files/publications/wp/2018/wp18158.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18158.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18158.pdf.json)

---

### Introduction — key findings
- Dataset: over 18,000 business loans with bank, borrower, loan, and relationship information from 50 developed and developing countries over 2004-2011, with the last 4 years 2008-2011 identified as the crisis period.
- Main empirical results:
  - Publicly-listed borrowers experienced slight increases rather than declines in lending quantities, but suffered large increases in interest rate spreads.
  - Privately-held borrowers experienced substantial quantity rationing, but minimal increases in loan prices.
  - Interest rate spreads became more sensitive to borrower leverage and credit rating during the crisis.
  - Foreign banks decrease lending quantities more and, among private borrowers, increase spreads less than domestic banks.
  - U.S. banks appear to engage in more quantity rationing than banks from other nations.
- Interpretation:
  - Interest rate is a more informative measure of credit market conditions for publicly-listed firms, whereas lending quantity is more informative for privately-held firms.
  - Evidence consistent with a multiple-contract equilibrium: loan markets for public borrowers continue to use price rationing during the crisis, while loan markets for private borrowers largely shut down.

### Contributions relative to existing literature
- Micro-level loan analysis using loan-level information with borrower characteristics to better disentangle loan demand and supply effects.
- Joint study of loan price (spreads) and loan quantity; finds price rationing for public borrowers and quantity rationing for private borrowers.
- Dataset covers both public and private borrowers, and spans 50 countries allowing U.S. vs non-U.S. comparisons.
- Documents that borrower composition and sensitivities of loan spreads to borrower characteristics change during the crisis.

### Research questions
- (1) How do loan quantities change between normal times and the recent crisis for foreign and domestic banks and for public and private firms?
- (2) How do loan spreads differ between foreign and domestic banks and for public and private firms, and how do they change between normal times and the recent crisis?
- (3) How do the effects of the financial crisis on loan quantities and spreads differ between U.S. and non-U.S. banks and between U.S. and non-U.S. borrowers?
- (4) How do the effects of the financial crisis on loan quantities and spreads differ between relationship and non-relationship borrowers?

### Data and sample construction — overview
- Time period and coverage: Annual dataset for 2004-2011 covering all banks available in 50 countries (25 developed and 25 developing).
- Bank data: BankScope and Claessens and Van Horen (2014) for bank ownership; Foreign Bank dummy = 1 if bank and borrower are from different countries; a bank’s country based on nationality of its owner.
- Loan and relationship data: DealScan as main loan source; dependent variables are loan quantities and spreads (total fees and interest paid over LIBOR per dollar drawn down); contract terms include Loan Size (natural log of deal amount in constant 2005 U.S. Dollars), Maturity (months), Collateral (dummy), and four purpose dummies; Relationship = dummy if lead bank on the deal was also a lead bank in another deal with the same borrower during t-5 to t-1.
- Borrower data: Orbis/Osiris for public and private firm financials; borrower variables include Size (natural log of total assets), Profitability (return on assets), Leverage (total liabilities divided by total assets), Asset tangibility (dummy); IPO and delisting dates from Osiris determine listing status at time of loan.
- Merging: manual and fuzzy matching across BankScope, DealScan, and Orbis/Osiris; matching results: over 33,000 observations with bank and loan information, and over 18,000 observations that additionally have borrower information.
- Data processing: winsorize all variables at 1%.
- Caveats: matching procedures may create selection biases; owner nationality determines bank country.

### Key methodological notes
- Crisis defined primarily as 2008-2011; alternative definition 2008-2009 used for robustness.
- Spreads regressions run separately for normal times and crisis subsamples and separately for public and private borrowers.
- Quantity regressions estimated in a Probit framework: Prob(Loan during Crisis | Loan before Crisis) with controls for borrower, loan, bank characteristics, and relationship.

### Major empirical findings — loan quantities (aggregate and multivariate)
- Definition of periods:
  - Normal times: 2004 to 2007.
  - Crisis: 2008 to 2011.
- Aggregate percentage changes (Table 2 summary):
  - Total number of loans dropped by 18% between normal times and the crisis.
  - Foreign banks contracted lending by 37%.
  - Domestic banks contracted lending by 12%.
  - Loans to private borrowers dropped by 38%.
  - Loans to public borrowers increased by 9%.
  - Within private borrowers:
    - Foreign banks contracted lending by 50%.
    - Domestic banks contracted lending by 34%.
  - Within public borrowers:
    - Foreign banks: -9% change in loans.
    - Domestic banks: +12% change in loans.
- Relationship lending:
  - Number of loans to relationship borrowers dropped by 9%.
  - Number of loans to non-relationship borrowers dropped by 23%.
- Credit-rating composition (Table 2):
  - A rated: number dropped by 28%.
  - B rated: number dropped by 48%.
  - C rated: number dropped by 51%.
  - D rated: number increased from 21 to 26 loans.
- U.S. vs non-U.S. loan count patterns (Table 2 Panels D and E):
  - U.S. banks: All Loans Normal Times 4,653; Crisis Years 2,568; Percentage Change -44.81.
  - Non-U.S. banks: All Loans Normal Times 5,622; Crisis Years 5,893; Percentage Change 4.82.
  - Foreign Banks (U.S.): 300 → 110; Percent Change -63.33.
  - Foreign Banks (Non-U.S.): 1,942 → 1,296; Percent Change -33.26.
  - Domestic Banks (U.S.): 4,353 → 2,458; Percent Change -43.53.
  - Domestic Banks (Non-U.S.): 3,680 → 4,597; Percent Change 24.92.
  - U.S. borrowers: All Loans Normal Times 5,067; Crisis Years 2,934; Percentage Change -42.1.
  - Non-U.S. borrowers: All Loans Normal Times 5,208; Crisis Years 5,527; Percentage Change 6.13.
- Multivariate probit results (Table 3 — selected marginal effects and significance):
  - Foreign Bank: -0.436 (11.13)*** (leading specifications show negative and highly significant effects).
  - Public Borrower: 0.431 (14.62)*** (public borrowers more likely to continue receiving loans).
  - U.S. Bank: -0.574 (8.07)*** (U.S. banks less likely to continue lending).
  - Relationship: 0.735 (23.26)*** (relationship borrowers much more likely to continue receiving loans).
  - Borrower Size: 0.036 (6.23)***; Profitability: 0.22 (6.74)***; Leverage: -0.456 (7.68)***.
  - Pseudo R-squared progression across specifications: 0.01, 0.04, 0.15, 0.16, 0.17, 0.18, 0.19.
  - Sample size N: 10,275 (most specifications); 9,897 in final specification.

### Major empirical findings — loan spreads (price) and heterogeneity
- Baseline patterns (Table 4, Figures):
  - Foreign banks charge higher spreads to public borrowers relative to domestic banks; this sensitivity increases during the crisis.
  - Foreign bank spreads to private borrowers are lower than domestic banks’ during the crisis in some specifications.
- Combined quantity and price changes (Figure 1 interpretation):
  - Foreign bank loans to public borrowers: Number of loans -9%; Average spread increases by 100 basis points.
  - Domestic bank loans to public borrowers: Number of loans +12%; Average spread increases by 89 basis points.
  - Foreign bank loans to private borrowers: Number of loans -50%; Average spread increases by 21 basis points.
  - Domestic bank loans to private borrowers: Number of loans -34%; Average spread increases by 66 basis points.
- Full specification coefficients (Table 4 Panel B — selected exact estimates):
  - Foreign Bank dummy estimated coefficients:
    - Normal times:
      - Public firms: 16.56.
      - Private firms: 8.16.
    - Crisis:
      - Public firms: 31.85.
      - Private firms: -35.01.
  - Panel A illustrative estimates (no controls) for comparison:
    - Foreign Bank Normal/Public: 15.808 (2.27)**; Crisis/Public: 27.1 (3.21)***.
    - Foreign Bank Normal/Private: -0.263 (0.03); Crisis/Private: -45.805 (3.50)***.
- Borrower characteristics and spreads:
  - Normal times: Larger, more profitable, less leveraged, and more tangible-asset-rich borrowers have lower spreads.
  - Crisis: coefficients on leverage and low ratings increase dramatically; spreads increase more for riskier borrowers.
  - Collateral associated with higher spreads (e.g., Collateral: 51.169 (10.03)*** Normal/Public; 71.644 (7.55)*** Normal/Private).
  - Rating effects: A Rated associated with large negative spread coefficients (e.g., A Rated Normal/Public: -54.829 (5.62)***; A Rated Normal/Private: -125.41 (5.59)***); C and D rated associated with large positive spreads in crises (e.g., C Rated Crisis/Private: 382.775 (6.55)***; D Rated Crisis/Public: 255.781 (6.95)***).
- U.S.-focused heterogeneity (Tables 5–6):
  - U.S. observations: approximately 40%.
  - For subsamples involving U.S. banks or U.S. borrowers, mixed results with notable large interaction coefficients:
    - Table 6 Panel B example: Foreign Bank x U.S. Borrower Crisis/Private: 174.142 (6.55)***.
    - Table 5 Panel B example: Foreign Bank x U.S. Bank Normal/Private: 72.11 (2.67)***; Crisis/Public: -77.852 (2.54)**.
  - Interpretation: interest-rate insurance particularly relevant for domestic lending in the U.S.; U.S. public firms often receive lower spreads during the crisis.
- Relationship vs non-relationship borrowers (Table 7):
  - Loan spreads are much lower for relationship than non-relationship borrowers.
  - Prior lending relationship lowers spread by:
    - Foreign banks: 58.19 (27.30) basis points in normal times (crisis period).
    - Domestic banks: 65.81 (40.16) basis points in normal times (crisis period).
  - In Panel B with controls: Foreign Bank (Normal/Relationship) 37.379 (3.49)***; Crisis/Relationship 6.901 (0.42).
  - Interpretation: relationships reduce spreads; relationships with foreign banks have less value than with domestic banks; foreign bank lending to non-relationship borrowers suffers the smallest increase in spread consistent with quantity rationing.
- Bank characteristics (Table 8):
  - Bank Size = natural log of Gross Total Assets; Equity = Total Equity / Gross Total Assets; Profitability = Pre-Tax Profits / Gross Total Assets; Liquidity = Liquid Assets / Gross Total Assets; NPLs = NPLs / Gross Total Assets.
  - Larger and better-capitalized banks grant loans with lower spreads.
  - During normal times, more profitable banks have higher-spread loans; during the crisis, more profitable banks have lower-spread loans.
  - Foreign Bank coefficients with bank controls (selected): Normal/Public 52.409 (4.66)***; Normal/Private 79.927 (5.01)***; Crisis/Public 44.693 (3.62)***; Crisis/Private -18.927 (0.97).
- Developed vs developing countries (Table 9):
  - Developed Countries All Loans: 9,710 → 7,869; Percentage Change -18.96.
  - Developing Countries All Loans: 565 → 592; Percentage Change 4.78.
  - Foreign Banks in Developed: 1,872 → 1,150; Percentage Change -38.57.
  - Domestic Banks in Developing: 195 → 336; Percentage Change 72.31.
  - Spreads: Foreign Bank x Developed Country Crisis/Private 176.942 (3.88)*** (Table 9 Panel B).
- Robustness and fixed effects:
  - Country and industry fixed effects, country×year fixed effects, alternative crisis window (2008–2009) yield qualitatively similar findings.
  - Alternative crisis-window Foreign Bank coefficients: Alternative Crisis/Public 33.01; Alternative Crisis/Private -43.56 (Table 11 comparison).
  - Results stronger among well-capitalized banks; undercapitalized foreign banks more likely to use both quantity and rate rationing for private borrowers.
  - Standard errors clustered at country and industry levels — main coefficients remain statistically significant.

### Conclusions — synthesis and policy implications
- Main empirical conclusions:
  - Public borrowers: slight increases in lending quantities, large increases in interest rate spreads — rationed more by price.
  - Private borrowers: substantial decreases in lending quantities, minimal increases in interest rate spreads — rationed more by quantity.
  - Interest rate spreads become significantly more sensitive to borrower leverage and credit rating during the crisis.
  - Foreign banks decreased lending quantities more; among private borrowers, foreign banks increase spreads less than domestic banks.
  - Declines in loan quantities were larger for U.S. banks than non-U.S. banks; U.S. banks experienced greater capital losses, profit decreases, more non-performing loans, and more loan loss provisions during the crisis.
- Policy implications:
  - A robust domestic banking sector may help protect private borrowers during financial crises, especially where banking sectors are dominated by foreign banks.
  - Ensuring viability of foreign banks (e.g., cross-border cooperation) is relevant.
  - Developing robust capital markets with opportunities for firms to go public may protect firms from credit rationing.
  - Improving transparency of private borrowers could reduce information gaps relative to public borrowers.
  - Significant capital pressure on banks during crises can lead to substantially more credit rationing of borrowers, disproportionately harming private firms.
  - Findings lend support to countercyclical capital requirements to protect the economy from financial crises.

### Key quantitative points and exact reported numbers (selected)
- Crisis period primary definition: 2008-2011; alternative: 2008-2009.
- Aggregate percentage changes (Panel A, Table 2):
  - All Loans: -17.65
  - Foreign Banks: -37.29
  - Domestic Banks: -12.17
  - Public Borrowers: 8.51
  - Private Borrowers: -38.23
  - Foreign Banks/Public Borrowers: -8.68
  - Foreign Banks/Private Borrowers: -50.36
  - Domestic Banks/Public Borrowers: 11.68
  - Domestic Banks/Private Borrowers: -33.8
- U.S. vs Non-U.S. counts (Table 2 Panel D examples):
  - U.S. Banks Normal Times 4,653; Crisis Years 2,568; Percentage Change -44.81.
  - Non-U.S. Banks Normal Times 5,622; Crisis Years 5,893; Percentage Change 4.82.
- Rating changes (Table 2 Panel B):
  - A Rated: 824 → 591; Percent Change -28.28.
  - B Rated: 2,443 → 1,273; Percent Change -47.89.
  - C Rated: 114 → 56; Percent Change -50.88.
  - D Rated: 21 → 26; Percent Change 23.81.
- Relationship spread reductions (normal times (crisis period)):
  - Foreign banks: 58.19 (27.30) basis points.
  - Domestic banks: 65.81 (40.16) basis points.
- Foreign Bank dummy crisis-period coefficients (alternative crisis window 2008-2009):
  - Public firms: 33.01
  - Private firms: -43.56
- Comparison coefficients (Table 4 Panel B):
  - Public firms: 31.85
  - Private firms: -35.01
- Probit marginal effects (Table 3 selected):
  - Foreign Bank: -0.436 (11.13)***; Public Borrower: 0.431 (14.62)***; U.S. Bank: -0.574 (8.07)***; Relationship: 0.735 (23.26)***.
- Bank capitalization note:
  - Annual average Bank Equity/Gross Total Assets series plotted for 2004–2011 shows U.S. banks had lower capital ratios and they fell more during the crisis (chart axis ticks shown between 10 and 13).

_Italicized source attribution: IMF Working Paper wp18158 — content unit._

### 1. Introduction

### 1. Introduction

### Key findings
- Dataset: over 18,000 business loans with bank, borrower, loan, and relationship information from 50 developed and developing countries over 2004-2011, with the last 4 years 2008-2011 identified as the crisis period.
- Main empirical results:
  - Publicly-listed borrowers experienced slight increases rather than declines in lending quantities, but suffered large increases in interest rate spreads.
  - Privately-held borrowers experienced substantial quantity rationing, but minimal increases in loan prices.
  - Interest rate spreads became more sensitive to borrower leverage and credit rating during the crisis.
  - Foreign banks decrease lending quantities more and, among private borrowers, increase spreads less than domestic banks.
  - U.S. banks appear to engage in more quantity rationing than banks from other nations.
- Interpretation:
  - Interest rate is a more informative measure of credit market conditions for publicly-listed firms, whereas lending quantity is more informative for privately-held firms.
  - Evidence is consistent with a multiple-contract equilibrium (Calomiris and Hubbard (1990)): loan markets for public borrowers continue to use price rationing during the crisis, while loan markets for private borrowers largely shut down.

### Contributions relative to existing literature
- Micro-level loan analysis:
  - Uses loan-level information with borrower characteristics to control for individual borrower demand and better disentangle loan demand and supply effects, in contrast to studies using aggregate- or portfolio-level data.
- Price and quantity:
  - Studies both loan price (spreads) and loan quantity; finds distinct mechanisms across borrower types (price rationing for public borrowers, quantity rationing for private borrowers).
- Public and private borrowers:
  - Dataset covers both public and private borrowers, important because bank credit is key to private firms without access to public equity financing.
- Firm heterogeneity:
  - Rather than relying on borrower fixed effects, studies firm heterogeneity directly and documents that borrower composition and sensitivities of loan spreads to borrower characteristics change during the crisis.
- International breadth:
  - Employs data from a large number of countries (50), allowing comparisons between U.S. and non-U.S. crisis responses and broader international conclusions.

### Research questions (structured to answer four questions)
- (1) How do loan quantities change between normal times and the recent crisis for foreign and domestic banks and for public and private firms?
  - Examine percentage change in total loans, loans by foreign and domestic banks, and loans to public and private borrowers; investigate changes in quantities by S&P credit rating categories.
- (2) How do loan spreads differ between foreign and domestic banks and for public and private firms, and how do they change between normal times and the recent crisis?
  - Analyze loan pricing across subsamples and investigate whether the crisis affects how borrower characteristics are priced; spread regressions do not restrict coefficients to be similar across subsamples.
- (3) How do the effects of the financial crisis on loan quantities and spreads differ between U.S. and non-U.S. banks and between U.S. and non-U.S. borrowers?
  - Divide observations into U.S. and non-U.S. subsamples for quantities; regress spreads on Foreign Bank dummy, U.S. dummy, and their interaction for pricing analysis.
- (4) How do the effects of the financial crisis on loan quantities and spreads differ between relationship and non-relationship borrowers?
  - Divide sample into borrowers with and without prior lending relationships (lead bank on the deal was also a lead bank on another deal with the same borrower during the past five years, from t-5 to t-1) and examine spreads controlling for borrower listing status.

### Data and sample construction — overview
- Time period and coverage:
  - Annual dataset for 2004-2011 covering all banks available in 50 countries (25 developed and 25 developing).
- Bank data:
  - Sources: BankScope and Claessens and Van Horen (2014) for bank ownership data.
  - Classification: Foreign Bank dummy = 1 if bank and borrower are from different countries, 0 otherwise. A bank’s country is based on the nationality of its owner (e.g., Citibank Thailand classified as a U.S. bank).
  - Bank characteristics: size, profitability, liquidity, and non-performing loans (NPLs/ Gross Total Assets).
- Loan and relationship data:
  - Main loan source: Thomson’s Loan Pricing Corporation’s DealScan (contains information on over $2 trillion of corporate and middle market commercial loans).
  - Dependent variables:
    - Loan quantities.
    - Spreads defined as total fees and interest paid over LIBOR per dollar drawn down.
  - Contract terms from DealScan: Loan Size (natural log of deal amount in constant 2005 U.S. Dollars), Maturity (months), Collateral (dummy), and four purpose dummies: Acquisition and Recapitalization; Capital Expenditure; Refinancing; Back-Up Line.
  - Relationship variable: dummy if lead bank on the deal was also a lead bank in another deal with the same borrower during the past five years (t-5 to t-1).
- Borrower data:
  - Main source: Bureau van Dijk’s Orbis (financial statements on over 10 million public and private firms worldwide).
  - Constructed borrower variables:
    - Size: natural log of total assets.
    - Profitability: return on assets.
    - Leverage: total liabilities divided by total assets.
    - Asset tangibility: dummy if Orbis reports the value of tangible assets.
  - Data processing: winsorize all variables at 1%.
  - Public versus private: dummy for listing status; IPO and delisting dates from Bureau van Dijk’s Osiris used to determine listing status at time of loan.
  - Additional dummies from DealScan: sector dummies (Corporation, Government, Financial Industry, Manufacturing, High-Tech) and S&P senior debt rating categories A, B, C, and D (omitted category = unrated).
  - Industry definitions:
    - Manufacturing Industry dummy = 1 if borrower’s SIC code is between 2000 and 3999.
    - High-Tech dummy = 1 if borrower is in high-tech industry according to the American Electronics Association (comprising 47 four-digit SIC codes in the two-digit industries 28, 35, 36, 38, 48, and 73).
- Merging across sources:
  - No common identifier across BankScope, DealScan, and Orbis/Osiris; manual and approximate matching used.
  - Banks in BankScope matched to lead arrangers in DealScan by bank names and nationalities; for multiple lead arrangers assign largest lead arranger as the lender.
  - Firms in Orbis/Osiris matched to DealScan borrowers by firm name and nationality using a fuzzy matching algorithm:
    - Standardize names by Anglicizing words and eliminating common suffixes.
    - Compute edit distance to measure similarity.
    - Retain perfect matches; discard non-matches; manually review borderline observations (similarity scores between 90% and 100%).
  - Matching results: over 33,000 observations with bank and loan information, and over 18,000 observations that additionally have borrower information.
- Data quality and caveats:
  - Matching procedures may create selection biases in favor of certain types of lenders or borrowers; authors argue such bias is unlikely to affect main results.
  - Winsorization at 1% to reduce influence of outliers or data errors.
  - For bank nationality classification, owner nationality determines bank country.

*Source: wp18158 - 1. Introduction*

### conclusions.  We focus on the changes between normal times and the crisis period, so as long

### Conclusions

### Changes in Bank Lending Quantities between Normal Times and the Financial Crisis
- Definition of periods:
  - Normal times: 2004 to 2007.
  - Crisis: 2008 to 2011.
- Aggregate loan quantity changes (Table 2 summary):
  - Total number of loans dropped by 18% between normal times and the crisis.
  - By bank nationality:
    - Foreign banks contracted lending by 37%.
    - Domestic banks contracted lending by 12%.
  - By borrower listing status:
    - Loans to private borrowers dropped by 38%.
    - Loans to public borrowers increased by 9%.
  - Within private borrowers:
    - Foreign banks contracted lending by 50%.
    - Domestic banks contracted lending by 34%.
  - Within public borrowers:
    - Foreign banks: -9% change in loans.
    - Domestic banks: +12% change in loans.
- Interpretation and mechanisms:
  - Findings consistent with credit rationing theories and increased information asymmetry during crisis.
  - Private borrowers are more opaque and riskier; their lending contracted more.
  - Domestic banks may provide liquidity insurance via relationship lending (references to Petersen and Rajan, 1994; Gatev and Strahan, 2006; Berger, Bouwman, and Kim, 2017).
- Relationship vs non-relationship borrowers (footnote):
  - Number of loans to relationship borrowers dropped by 9%.
  - Number of loans to non-relationship borrowers dropped by 23%.
- Borrower credit rating (Table 2 Panel B and C):
  - Number of loans changes by rating:
    - A rated: number dropped by 28%.
    - B rated: number dropped by 48%.
    - C rated: number dropped by 51%.
    - D rated: number increased from 21 to 26 loans.
  - Distributional shifts and bank types:
    - During normal times, 40% of foreign loans are to A and B rated borrowers; 30% of domestic loans are to A and B rated borrowers.
    - During the crisis these fractions drop to 36% (foreign) and 19% (domestic).
    - Public borrowers during normal times: 43% of loans are A and B rated.
    - Private borrowers during normal times: 23% of loans are A and B rated.
    - During the crisis, fraction of A and B rated loans drops to 28% for public borrowers and 14% for private borrowers.
  - Interpretation: riskiest private borrowers are quantity-rationed during the crisis.
- U.S. versus non-U.S. banks and borrowers (Table 2 Panels D and E):
  - U.S. banks show larger declines in loan quantities than non-U.S. banks; U.S. borrowers also show larger declines than non-U.S. borrowers.
  - Among U.S. banks:
    - Foreign banks (U.S. banks lending overseas): -63% change in number of loans.
    - Domestic banks (U.S. banks lending at home): -44% change in number of loans.
  - Among non-U.S. banks:
    - Foreign banks (lending in U.S. and elsewhere outside home): -33% change.
    - Domestic banks (lending at home): +25% change.
  - Among non-U.S. borrowers:
    - Foreign banks: -39% change.
    - Domestic banks: +25% change.
  - Public vs private by bank origin:
    - Among U.S. banks:
      - Public borrowers: -36.05% change.
      - Private borrowers: -55% change.
    - Among non-U.S. banks:
      - Public borrowers: +63% change.
      - Private borrowers: -28% change.
    - Among U.S. borrowers:
      - Public borrowers: -34% change.
      - Private borrowers: -52% change.
    - Among non-U.S. borrowers:
      - Public borrowers: +77% change.
      - Private borrowers: -29% change.
  - Notable exception:
    - For U.S. borrowers, foreign banks: -33% vs domestic banks: -44% — U.S. banks lending domestically cut back more than foreign banks operating in the U.S.
  - Suggested explanations:
    - Crisis originated in the U.S.; U.S. banks suffered greater capital losses, profit decreases, non-performing loans, and loan loss provisions.
    - Possible stricter enforcement of capital requirements in the U.S. (e.g., more stringent stress tests).

### Multivariate Analysis of Lending Quantities (Probit framework)
- Objective:
  - Estimate Prob(Loan during Crisis | Loan before Crisis) for each pre-crisis loan i.
- Key explanatory variables:
  - Foreign Bank, Public Borrower, U.S. Bank, U.S. Borrower, Relationship.
- Controls (Other Borrower Characteristics and loan terms):
  - Borrower characteristics: Size, Profitability, Leverage, Asset Tangibility; indicators for corporate vs government-owned; industry dummies (financial, manufacturing, high-tech); S&P senior debt rating dummies (A, B, C, D).
  - Loan contract terms: Loan Size, Maturity, Collateral.
  - Loan purposes: Acquisition, Capital Expenditure, Refinancing, Backup Line.
- Main multivariate findings (Table 3):
  - Foreign banks lending during normal times are less likely to continue lending during the crisis compared to domestic banks.
  - Public borrowers are more likely to continue receiving loans during the crisis.
  - U.S. banks are less likely than banks in other countries to continue to lend.
  - Relationship borrowers are more likely to continue to receive loans during the crisis.
  - In the full specification, coefficients on Foreign Bank, Public Borrower, U.S. Bank, and Relationship dummies are statistically significant at the 1% level.
  - Coefficient on U.S. Borrower is not statistically significant after controlling for borrower rating.
- Controls’ effects:
  - Larger borrowers, higher profitability, and lower leverage are associated with higher probability of continued lending; Size, Profitability, and Leverage coefficients are significant at the 1% level.
  - Borrowers in high-tech industries are less likely to receive loans during the crisis.
  - Borrowers with larger loans and longer maturities in normal times are less likely to borrow again during the crisis.
  - Borrowers with pre-crisis refinancing loans are more likely to borrow again during the crisis.
  - Overall: riskier borrowers are more likely to be quantity rationed during the crisis.

### Loan Spreads during Normal Times and the Financial Crisis
- Regression setup:
  - Dependent variable: all-in loan spread.
  - Key explanatory variable: Foreign Bank dummy; controls include borrower characteristics, Loan Size, Maturity, Collateral, and loan purpose dummies.
  - Estimations run separately for normal times and crisis subsamples, and separately for public and private borrowers.
- Baseline illustrative results (Table 4 Panel A and Figure 1):
  - Foreign banks charge higher spreads to public borrowers relative to domestic banks; this sensitivity increases during the crisis.
  - Foreign bank spreads to private borrowers are lower than domestic banks’ during the crisis.
- Combined quantity and price changes (Figure 1 interpretation from coefficients and Table 2):
  - Foreign bank loans to public borrowers:
    - Number of loans: -9%.
    - Average spread: increases by 100 basis points.
  - Domestic bank loans to public borrowers:
    - Number of loans: +12%.
    - Average spread: increases by 89 basis points.
  - Foreign bank loans to private borrowers:
    - Number of loans: -50%.
    - Average spread: increases by 21 basis points.
  - Domestic bank loans to private borrowers:
    - Number of loans: -34%.
    - Average spread: increases by 66 basis points.
  - Interpretation:
    - Public borrowers experienced flight to quality and increases in spreads; domestic banks increased their lending to public borrowers.
    - Private borrowers faced significant quantity rationing and limited spread increases, especially for foreign banks.
- Full specification results (Table 4 Panel B):
  - Direction of Foreign Bank dummy effects similar to baseline; magnitudes differ.
  - Estimated coefficients on Foreign Bank dummy:
    - Normal times:
      - Public firms: 16.56.
      - Private firms: 8.16.
    - Crisis:
      - Public firms: 31.85.
      - Private firms: -35.01.
  - Interpretation: increase in lending rates to private borrowers was much smaller for foreign banks (and negative in full specification during crisis), while public firms faced large spread increases for foreign banks during the crisis.
- Borrower characteristics and spreads:
  - Normal times:
    - Larger, more profitable, less leveraged, and more tangible-asset-rich borrowers have lower spreads.
    - Loans to banks have lower spreads than to governments and corporates.
    - Manufacturing borrowers have lower spreads; high-tech borrowers have higher spreads.
    - Better credit ratings generally associated with lower spreads.
  - Crisis vs normal times:
    - Coefficients on leverage and low ratings change dramatically; spreads increase more for riskier borrowers.
    - Statistical tests (Panel C) verify coefficients on borrower risk (leverage and credit ratings) change significantly between normal times and the crisis.
  - Loan contract terms:
    - Larger loans viewed as safer.
    - Longer maturities associated with higher spreads (term premium and/or higher risk).
    - Collateral correlated with higher spreads (consistent with collateral literature, though mixed in prior work).
    - Acquisition-purpose loans have higher spreads than capital expenditure, refinancing, and backup line loans.
- Overall synthesis:
  - Stark contrast between public and private borrowers:
    - During the crisis, aggregate number of loans went up for public borrowers (flight to quality) and spreads for public borrowers rose substantially.
    - For private borrowers, number of loans fell significantly and spread increases were minimal—consistent with intensive quantity rationing.
  - Effects are stronger for foreign banks, indicating information asymmetry strongly influences quantity and price outcomes in crises.

_Italicized source attribution: IMF Working Paper wp18158 — conclusions._

### 7. Spread of Loans Involving Banks and Borrowers from the United States

### 7. Spread of Loans Involving Banks and Borrowers from the United States

### Overview and motivation
- Approximately 40% of observations are from the U.S.
- Rationale for separate U.S. analysis:
  - The crisis originated in the U.S. and U.S. banks faced stronger capital pressure.
  - U.S. financial markets are the largest and most developed.
  - U.S. financial institutions are noted for innovation and operational efficiency.
  - U.S. borrowers (especially public firms) may be more transparent due to stricter regulation and certified-audit financial statements.
- Aim: examine whether spreads of banks and borrowers from the U.S. differ from the rest of the world, across normal times and crisis periods, and across public and private borrower subsamples.

### Empirical specifications
- Equation (3) (bank-focused): Spread_i = γ0 + γ1 Foreign Bank_i + γ2 U.S. Bank_i + γ3 (Foreign Bank_i × U.S. Bank_i) + γ4 Public Borrower_i + Γ5 Other Borrower Characteristics_i + γ6 Loan Size_i + γ7 Maturity_i + γ8 Collateral_i + γ9 Acquisition_i + γ10 Capital Expenditure_i + γ11 Refinancing_i + γ12 Backup Line_i + εγ_i
- Equation (4) (borrower-focused): Spread_i = δ0 + δ1 Foreign Bank_i + δ2 U.S. Borrower_i + δ3 (Foreign Bank_i × U.S. Borrower_i) + δ4 Public Borrower_i + Γ5 Other Borrower Characteristics_i + δ6 Loan Size_i + δ7 Maturity_i + δ8 Collateral_i + δ9 Acquisition_i + δ10 Capital Expenditure_i + δ11 Refinancing_i + δ12 Backup Line_i + εδ_i
- Estimation approach:
  - First estimate without control variables (Panels A of Tables 5 and 6).
  - Then estimate full specifications with control variables (Panels B of Tables 5 and 6).

### Loan contract terms (deal size, maturity, collateral)
- During normal times:
  - Foreign banks grant larger loans with longer maturities relative to domestic banks.
  - Foreign banks require collateral more often.
  - Interpretation: foreign banks lend larger/longer to borrowers with more severe information problems but secure loans with collateral.
- During the financial crisis:
  - Average loan size rises more for foreign banks (interpreted as quantity rationing of lower-quality borrowers who tend to have smaller deal sizes).
  - Foreign banks reduce maturity more and increase collateral usage more than domestic banks.
  - Comparing public vs private borrowers: increases in loan size and maturity contraction by foreign banks are stronger among private borrowers.
- Overall interpretation: evidence consistent with credit rationing theories — banks choose contract terms to accommodate borrower information problems.

### Theoretical ambiguity on spread changes
- Two competing mechanisms:
  - Subsamples with greater contraction in number of loans may have smaller increases in spreads because riskier borrowers are quantity-rationed.
  - Alternatively, banks that reduce credit supply most may raise interest rates and quantity ration simultaneously, producing larger increases in spreads.
  - Greater contraction in number of loans may indicate weaker lender-borrower relationships, producing higher spread increases.

### Main empirical patterns (U.S. vs non-U.S. and public vs private)
- For non-U.S. banks or non-U.S. borrowers:
  - Between foreign and domestic banks, spreads generally increase more for domestic banks.
  - Between public and private borrowers, spreads generally increase more for public borrowers.
- For subsamples involving U.S. banks or U.S. borrowers: mixed results.
  - Notable: for U.S. borrowers, the increase in spreads is smaller among domestic banks and among public borrowers.
  - Interpretation: interest-rate insurance is particularly relevant for domestic lending in the U.S.; U.S. public firms have valuable relationships with domestic banks and receive lower spreads during the crisis.

### Effects of adding controls
- When adding control variables from Section 6 (Panel B of Tables 5 and 6):
  - Most key coefficients retain the same sign.
  - Statistical significance is often diminished due to inclusion of additional variables.
  - Coefficients on control variables similar to Panel B of Table 4.

### Relationships vs non-relationship borrowers (Section 8 summary)
- Definition: relationship borrowers have prior lending relationships producing borrower-specific durable and reusable information.
- Findings:
  - Loan spreads are much lower for relationship than non-relationship borrowers.
  - Relationships with foreign banks have less value to borrowers than relationships with domestic banks.
  - Prior lending relationship lowers spread by:
    - 58.19 (27.30) basis points for foreign banks during normal times (the crisis period).
    - 65.81 (40.16) basis points for domestic banks during normal times (the crisis period).
  - Comparing normal times and crisis:
    - Loan spreads increase substantially during the crisis.
    - Foreign bank lending to non-relationship borrowers suffers the smallest increase in spread.
    - Interpretation: consistent with significant quantity rationing of non-relationship borrowers who are likely the riskiest.
- Robustness:
  - Controlling for Public dummy and other borrower/loan characteristics (Table 7 Panel B) still shows relationships with foreign banks are less valuable than with domestic banks.
  - Accounting for borrower characteristics accentuates the crisis effects on foreign bank coefficients among relationship borrowers.

### Bank characteristics and their roles
- Bank characteristics used in controls (Table 8):
  - Size = natural log of Gross Total Assets
  - Equity = Total Equity / Gross Total Assets
  - Profitability = Pre-Tax Profits / Gross Total Assets
  - Liquidity = Liquid Assets / Gross Total Assets
  - Non-Performing Loans (NPLs) = NPLs / Gross Total Assets (NPLs defined as loans past due for more than 90 days)
  - Note: pre-tax profits used to eliminate effect of differing taxation across countries.
- Key results with bank characteristics included:
  - During normal times, foreign banks have higher spreads than domestic banks.
  - Increases in spreads during the crisis are smaller for foreign banks, especially among private borrowers.
  - Coefficients on bank characteristics:
    - Larger and better-capitalized banks tend to grant loans with lower spreads.
    - During normal times, more profitable banks have higher-spread loans.
    - During the crisis, more profitable banks have lower-spread loans — consistent with lending to risky borrowers becoming less profitable during the crisis.

### Developed versus developing countries
- Aggregate quantity changes (Table 9 Panel A):
  - Number of loans dropped more for foreign banks and private borrowers during the crisis.
  - Effect is generally stronger in developing countries, supporting the notion that information problems are less severe in developed countries.
- Spread behavior (Table 9 Panel B):
  - Among private borrowers, spreads generally increase more for domestic banks than foreign banks.
  - This finding holds in both developed and developing country subsamples.

### Fixed effects, alternative crisis definition, and further robustness
- Country and industry fixed effects (Table 10 Panel A):
  - Including full country and industry fixed effects yields similar results: foreign banks have higher spreads; among private borrowers, increases in spreads during the crisis are generally smaller for foreign banks.
  - Replacing Foreign Bank dummy with bank fixed effects yields qualitatively similar results.
- Country × Year fixed effects (Table 10 Panel B):
  - Results similar to main specifications; crisis-period spread increases generally smaller for foreign banks and driven by loans to private borrowers.
- Alternative crisis definition (Table 11):
  - Defining crisis as 2008-2009 yields similar findings; increases in spreads during the crisis are generally smaller for foreign banks and driven by private borrowers.
  - Coefficients on Foreign Bank dummy during the crisis under alternative definition:
    - 33.01 for public firms
    - -43.56 for private firms
  - Compare to Table 4 Panel B: 31.85 for public firms and -35.01 for private firms.
- Capitalization comparison (U.S. vs non-U.S. banks):
  - Annual average Bank Equity/Gross Total Assets plotted for 2004–2011 (Figure 3): U.S. banks had lower capital ratios and they fell more during the crisis; lowest in 2009, pick up in 2010 and 2011 but still below 2007 level.
  - Unreported regressions: Bank Loans/Gross Total Assets and ROA are lower during the crisis; Bank NPL/Gross Total Assets and Reserve/Gross Loans are higher during the crisis. Effects stronger in the U.S.
- Additional robustness checks:
  - Splitting sample by lender capitalization: the finding that crisis-period spread increases are smaller for foreign banks’ lending to private borrowers is stronger among well-capitalized banks.
  - Undercapitalized foreign banks more likely to use both quantity and interest rate rationing to limit exposure to private borrowers during the crisis.
  - Standard errors clustered at country and industry levels — main coefficients remain statistically significant.

### Key quantitative points (preserve reported numbers exactly)
- U.S. observations: approximately 40%.
- Crisis period primary definition: 2008-2011.
- Alternative crisis definition analyzed: 2008-2009.
- Relationship spread reductions (normal times (crisis period)):
  - Foreign banks: 58.19 (27.30) basis points.
  - Domestic banks: 65.81 (40.16) basis points.
- Foreign Bank dummy crisis-period coefficients (alternative crisis window 2008-2009):
  - Public firms: 33.01
  - Private firms: -43.56
- Comparison coefficients (Table 4 Panel B):
  - Public firms: 31.85
  - Private firms: -35.01
- Bank characteristic definitions and measures preserved as presented.

### Conclusions and policy implications
- Main empirical conclusions:
  - Public borrowers: slight increases in lending quantities, large increases in interest rate spreads — rationed more by price.
  - Private borrowers: substantial decreases in lending quantities, minimal increases in interest rate spreads — rationed more by quantity.
  - Interest rate spreads become significantly more sensitive to borrower leverage and credit rating during the crisis.
  - Foreign banks decreased lending quantities more; among private borrowers, foreign banks increase spreads less than domestic banks.
  - Declines in loan quantities were larger for U.S. banks than non-U.S. banks; U.S. banks experienced greater capital losses, profit decreases, more non-performing loans, and more loan loss provisions during the crisis.
- Importance of borrower characteristics:
  - Public vs private status materially affects how banks treat borrowers during crises.
  - Borrower characteristics have significant explanatory power for loan spreads and their sensitivities change between normal and crisis periods.
- Policy implications:
  - A robust domestic banking sector may help protect private borrowers during financial crises, especially where banking sectors are dominated by foreign banks.
  - Ensuring viability of foreign banks (e.g., cross-border cooperation) is relevant.
  - Developing robust capital markets with opportunities for firms to go public may protect firms from credit rationing.
  - Improving transparency of private borrowers could reduce information gaps relative to public borrowers.
  - Significant capital pressure on banks during crises can lead to substantially more credit rationing of borrowers, disproportionately harming private firms (engines of innovation and growth).
  - Findings lend support to countercyclical capital requirements to protect the economy from financial crises.

*Source: "7. Spread of Loans Involving Banks and Borrowers from the United States" (wp18158).*

### References

### References

### Key empirical findings on loan volumes during the financial crisis
- Percentage change in number of loans (Panel A, Table 2):
  - All Loans: -17.65
  - Foreign Banks: -37.29
  - Domestic Banks: -12.17
  - Public Borrowers: 8.51
  - Private Borrowers: -38.23
  - Foreign Banks/Public Borrowers: -8.68
  - Foreign Banks/Private Borrowers: -50.36
  - Domestic Banks/Public Borrowers: 11.68
  - Domestic Banks/Private Borrowers: -33.8
- U.S. vs Non-U.S. bank loan counts (Panel D, Table 2):
  - All Loans: U.S. Banks Normal Times 4,653; Crisis Years 2,568; Percentage Change -44.81
  - All Loans: Non-U.S. Banks Normal Times 5,622; Crisis Years 5,893; Percentage Change 4.82
  - Foreign Banks (U.S.): 300 → 110; Percent Change -63.33
  - Foreign Banks (Non-U.S.): 1,942 → 1,296; Percent Change -33.26
  - Domestic Banks (U.S.): 4,353 → 2,458; Percent Change -43.53
  - Domestic Banks (Non-U.S.): 3,680 → 4,597; Percent Change 24.92
- U.S. vs Non-U.S. borrower loan counts (Panel E, Table 2):
  - All Loans: U.S. Borrowers Normal Times 5,067; Crisis Years 2,934; Percentage Change -42.1
  - All Loans: Non-U.S. Borrowers Normal Times 5,208; Crisis Years 5,527; Percentage Change 6.13
  - Public Borrowers (Non-U.S.): 1,739 → 3,074; Percentage Change 76.77
  - Private Borrowers (Non-U.S.): 3,469 → 2,453; Percentage Change -29.29

### Credit-rating composition shifts (Table 2, Panels B and C)
- Number of loans by rating (Panel B):
  - A Rated: 824 → 591; Percent Change -28.28
  - B Rated: 2,443 → 1,273; Percent Change -47.89
  - C Rated: 114 → 56; Percent Change -50.88
  - D Rated: 21 → 26; Percent Change 23.81
- Distribution changes (Normal Times → Crisis, Panel C examples):
  - Unrated loans: 66.88% → 77.00%; Change 10.12%
  - B Rated (All): 23.78% → 15.05%; Change -8.74%
  - A Rated (Foreign banks, Normal → Crisis): 16.09% → 12.59%; Change -3.51%

### Bank capitalization (Figure 3)
- Annual average Bank Equity/Gross Total Assets for U.S. versus Non-U.S. banks (sample covers 2004–2011, BankScope):
  - Chart scale indicates values between 10 and 13 (axis ticks shown: 10, 10.5, 11, 11.5, 12, 12.5, 13) for years 2004–2011.
  - Series labeled: U.S Banks and Non-U.S Banks (exact annual values presented in figure).

### Loan availability determinants (Table 3: probit regressions; Y = Probability(Loan after Crisis | Loan before Crisis))
- Key estimated marginal effects (selected columns; z statistics in parentheses; significance as indicated in table):
  - Foreign Bank: -0.436 (11.13)***; subsequent specifications: -0.371 (9.32)***, -0.542 (11.82)***, -0.573 (11.89)***, -0.589 (11.93)***, -0.611 (11.35)***, -0.546 (9.65)***
  - Public Borrower: 0.431 (14.62)***; alternative specs up to 0.491 (13.09)***
  - U.S. Bank: -0.574 (8.07)***; alternative specs down to -0.623 (7.94)***
  - U.S. Borrower: -0.179 (2.65)***; later specs: -0.142 (2.01)**, -0.108 (1.48), -0.014 (0.18), 0.011 (0.14)
  - Relationship (lead bank relationship): 0.735 (23.26)***; alternate specs: 0.701 (21.81)***, 0.702 (21.67)***, 0.688 (21.08)***, 0.625 (18.38)***
  - Borrower Size: 0.036 (6.23)***; alternatives: 0.032 (5.22)***, 0.024 (3.85)***, 0.032 (4.74)***
  - Profitability: 0.22 (6.74)***; alternatives: 0.218 (6.66)***, 0.208 (6.38)***, 0.208 (6.06)***
  - Leverage: -0.456 (7.68)***; alternatives: -0.451 (7.75)***, -0.427 (7.53)***, -0.414 (7.12)***
  - Maturity: -0.005 (8.66)*** (included in later specifications)
  - Deal Size: -0.024 (1.93)* (included in later specifications)
- Model performance:
  - Pseudo R-squared progression across specifications: 0.01, 0.04, 0.15, 0.16, 0.17, 0.18, 0.19
  - Sample size N: 10,275 (most specifications); 9,897 in final specification

### Effects on loan spreads — foreign bank and crisis interactions (Table 4, Panels A–C)
- Panel A (Foreign Bank effect on Spread):
  - Normal Times Public: 15.808 (2.27)** 
  - Normal Times Private: -0.263 (0.03)
  - Crisis Public: 27.1 (3.21)***
  - Crisis Private: -45.805 (3.50)***
- Panel B (with controls — selected coefficients):
  - Foreign Bank (Normal/Public): 16.56 (2.68)***
  - Foreign Bank (Normal/Private): 8.161 (0.81)
  - Foreign Bank (Crisis/Public): 31.85 (4.14)***
  - Foreign Bank (Crisis/Private): -35.014 (2.57)**
  - Collateral: 51.169 (10.03)*** (Normal/Public); 71.644 (7.55)*** (Normal/Private); 57.263 (8.99)*** (Crisis/Public); 24.989 (1.97)** (Crisis/Private)
  - A Rated: -54.829 (5.62)*** (Normal/Public); -125.41 (5.59)*** (Normal/Private); -76.827 (5.93)*** (Crisis/Public); -86.233 (3.12)*** (Crisis/Private)
  - C Rated and D Rated carry large positive spreads (e.g., C Rated Crisis/Private: 382.775 (6.55)***; D Rated Crisis/Public: 255.781 (6.95)***)
  - Deal Size: negative coefficients (e.g., Crisis/Private: -13.177 (2.95)***)
- Panel C (difference Crisis − Normal; selected entries and significance):
  - Foreign Bank, Public Difference: 15.290 (1.39)
  - Foreign Bank, Private Difference: -43.175 (2.29)**
  - Leverage differences: Public 46.341 (3.66)***; Private 37.513 (1.99)**
  - C Rated difference (Private): 285.792 (2.98)***
  - Backup Line difference (Private): -77.846 (3.02)***

### Heterogeneity: U.S. vs Non-U.S. banks and borrowers (Tables 5–6)
- Table 5 Panel A (U.S. Bank interactions, Panel A without controls):
  - Foreign Bank: Normal/Public 40.736 (3.21)***; Normal/Private -52.739 (4.19)***; Crisis/Public 85.873 (7.92)***; Crisis/Private -45.836 (2.64)***
  - U.S. Bank: Normal/Public 21.054 (1.87)*; Normal/Private -66.109 (5.99)***; Crisis/Public 64.215 (7.45)***; Crisis/Private -3.146 (0.2)
  - Foreign Bank x U.S. Bank: Normal/Public -80.395 (3.28)***; Normal/Private 120.393 (4.71)***; Crisis/Public -201.614 (6.35)***; Crisis/Private -27.182 (0.6)
- Table 5 Panel B (with controls; selected):
  - Foreign Bank x U.S. Bank, Normal/Private: 72.11 (2.67)***; Crisis/Public: -77.852 (2.54)**

- Table 6 Panel A (U.S. Borrower interactions, Panel A without controls):
  - Foreign Bank: Normal/Public -25.679 (1.71)*; Normal/Private -49.738 (3.78)***; Crisis/Public 10.449 (0.76); Crisis/Private -125.238 (6.76)***
  - Foreign Bank x U.S. Borrower: Normal/Public 72.926 (4.24)***; Normal/Private 83.134 (3.94)***; Crisis/Public 49.919 (2.89)***; Crisis/Private 209.02 (7.81)***
- Table 6 Panel B (with controls; selected):
  - Foreign Bank x U.S. Borrower, Normal/Public 40.916 (2.75)***; Crisis/Private 174.142 (6.55)***

### Relationship lending and spreads (Table 7)
- Panel A (no controls):
  - Foreign Bank effect among Relationship borrowers (Normal): 18.46 (1.71)*; Non-Relationship: 10.839 (1.40)
  - Crisis: Relationship 5.496 (0.33); Non-Relationship -7.361 (0.94)
- Panel B (with controls):
  - Foreign Bank (Normal/Relationship): 37.379 (3.49)***; Normal/Non-Relationship: 1.012 (0.13)
  - Crisis: Relationship 6.901 (0.42); Non-Relationship 0.521 (0.07)

### Bank- and borrower-characteristics in spread regressions (Table 8)
- Foreign Bank coefficients with bank controls:
  - Normal/Public: 52.409 (4.66)***
  - Normal/Private: 79.927 (5.01)***
  - Crisis/Public: 44.693 (3.62)***
  - Crisis/Private: -18.927 (0.97)
- Bank characteristics (selected coefficients):
  - Bank Size: -8.082 (2.84)*** (Normal/Public); -10.149 (1.77)* (Crisis/Private)
  - Bank Profitability: 55.889 (6.09)*** (Normal/Public); 61.833 (4.01)*** (Normal/Private); -25.663 (3.88)*** (Crisis/Public)
  - Bank NPL: 23.395 (2.80)*** (Normal/Public); -23.458 (1.78)* (Normal/Private)
  - Bank Equity: -2.526 (2.57)** (Normal/Public)

### Developed vs. Developing countries (Table 9)
- Loan counts (Panel A):
  - Developed Countries All Loans: 9,710 → 7,869; Percentage Change -18.96
  - Developing Countries All Loans: 565 → 592; Percentage Change 4.78
  - Foreign Banks in Developed: 1,872 → 1,150; Percentage Change -38.57
  - Domestic Banks in Developed: 7,838 → 6,719; Percentage Change -14.28
  - Domestic Banks in Developing: 195 → 336; Percentage Change 72.31
- Spreads (Panel B, interaction results):
  - Foreign Bank x Developed Country coefficients: Normal/Public 70.781 (2.50)**; Normal/Private 71.444 (1.78)*; Crisis/Public 67.168 (2.19)**; Crisis/Private 176.942 (3.88)***

### Fixed-effect and robustness specifications (Tables 10–11)
- Country and industry fixed effects (Table 10 Panel A):
  - Foreign Bank coefficients: Normal/Public 15.657 (2.39)**; Normal/Private 27.719 (2.38)**; Crisis/Public 29.649 (3.61)***; Crisis/Private 3.699 (-0.25)
  - R-squared: 0.33 (Normal/Public), 0.19 (Normal/Private), 0.31 (Crisis/Public), 0.26 (Crisis/Private)
- Country x Year fixed effects (Table 10 Panel B):
  - Foreign Bank coefficients: Normal/Public 21.615 (3.29)***; Normal/Private 31.783 (2.74)***; Crisis/Public 29.562 (3.91)***; Crisis/Private -8.902 (0.61)
  - R-squared up to 0.4 (Crisis/Public)
- Alternate crisis period (Table 11, crisis defined as 2008–2009):
  - Foreign Bank coefficients: Normal/Public 16.56 (2.68)***; Normal/Private 8.161 (0.81); Alternative Crisis/Public 33.083 (3.36)***; Alternative Crisis/Private -43.563 (2.59)***

### Cross-cutting analytical conclusions from tables and figures
- Crisis-driven loan quantity declines were sizable overall (-17.65) with much larger reductions for foreign-bank lending and private borrowers (e.g., Foreign Banks -37.29; Private Borrowers -38.23; Foreign Banks/Private Borrowers -50.36).
- Public borrowers experienced increased loan counts during the crisis (8.51) while private borrowers suffered large contractions (-38.23).
- Relationship lending strongly increased the probability of continued lending through the crisis (Relationship coefficient 0.735 (23.26)***), indicating a protective effect of prior bank-borrower relationships.
- Foreign bank status negatively affected loan availability (Foreign Bank coefficients in Table 3 are negative and highly significant across specifications).
- Effects on loan spreads are heterogeneous:
  - Foreign bank presence is associated with higher spreads for public borrowers during crisis periods (e.g., Foreign Bank Crisis/Public 31.85 (4.14)*** in Table 4 Panel B).
  - For private borrowers, foreign bank effects on spreads are negative during crisis in some specifications (e.g., Crisis/Private -35.014 (2.57)** in Table 4 Panel B), indicating complex heterogeneity by borrower type, borrower/location, and bank origin.
- U.S. bank and U.S. borrower interactions with foreign bank status produce large and significant heterogeneous spread effects (Tables 5–6), including very large positive interaction coefficients in some crisis/private contexts (e.g., Foreign Bank x U.S. Borrower Crisis/Private 174.142 (6.55)*** in Table 6 Panel B).
- Bank-level characteristics (size, profitability, NPL, equity, liquidity) matter for spreads: e.g., Bank Size negative; Bank Profitability positive in normals; Bank Liquidity positive in some specs.

*Italic source attribution: References and tables/figures as presented in wp18158 - References (source PDF: wp18158 - References).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18158.pdf_