## _wp0538

## Source details

**Canonical URL:** [_wp0538](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp0538.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp0538.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp0538.pdf.json)

---

### Introduction: scope and literature
- Many countries experienced significant credit slowdowns; research asks whether declines reflect depressed demand or diminished bank lending capacity.
- International examples cited:
  - Finland: banking system credit-to-GDP ratio shrank by over 44 percentage in 1992–97.
  - Thailand: banking system credit-to-GDP ratio shrank by some 36 percentage points in 1998–2000.
- Two strands of literature:
  - Post-crisis cases (Scandinavia, East Asia, Japan) generally find demand contractions rather than supply-driven credit crunches (Pazarbasioglu (1997); Ghosh and Ghosh (1999); Woo (1999)).
  - United States early 1990s: banking system reduced credit by an amount equal to 13 percentage points of GDP between 1990 and 1993; mixed evidence on “capital crunch” and bank-lending channel (Bernanke and Lown (1991); Berger and Udell (1994); Peek and Rosengren (1995, 1997, 2000)).

### Credit slowdowns in Latin America: patterns and magnitudes
- Examples and magnitudes:
  - Bolivia: bank credit-GDP ratio decline approaching 16 percentage points.
  - Mexico: bank credit-GDP ratio decline approaching 20 percentage points.
  - Panama: 16 percentage point decline over the last two years (noted as a notable exception on an average annual basis).
- Typical average annual declines: most tended to be in the range of 1–3 percentage points.
- Argentina: credit slowdown (1999–2000) versus crisis-related decline (2001–03) driven largely by substantial deposit outflows.
- Barajas and Steiner (2002) findings for eight LAC cases (Argentina, Bolivia, Brazil, Chile, Colombia, Mexico, Peru, Venezuela):
  - Both reductions in sources of funds and increases in alternative uses of funds occurred during credit declines.
  - Mexico: supply contraction swamped by demand-side contraction.
  - Colombia: demand and supply contractions roughly similar magnitude.
  - Regulatory and risk-related variables (loan loss provisions to nonperforming loans; nonperforming/total loans) significant in explaining credit supply contraction.

### Basel I adoption: database and timing
- Database constructed primarily from a mid-2003 survey with IMF desk economists; supplemented by World Bank bank regulation database, central bank websites, IMF MFD regulatory database, and research dating adoptions (e.g., Chiuri, Ferri and Majnoni (2002)).
- Evidence of Basel adoption in LAC beginning in 1991 and continuing throughout the 1990s; most countries adopted between 1993 and the late 1990s.
- Appendix Table A.1 adoption counts by year (LAC / Rest of World / Total):
  - 1988: 0 / 16 / 16
  - 1990: 0 / 1 / 1
  - 1991: 1 / 2 / 3
  - 1993: 4 / 8 / 12
  - 1994: 3 / 5 / 8
  - 1995: 3 / 6 / 9
  - 1996: 1 / 4 / 5
  - 1997: 4 / 7 / 11
  - 1998: 2 / 3 / 5
  - 1999: 1 / 6 / 7
  - 2000: 1 / 6 / 7
  - 2001: 0 / 4 / 4
  - 2002: 0 / 0 / 0
- Totals:
  - Total dated: 20 / 71 / 91
  - Total undated: 2 / 32 / 34
  - Total adopting countries: 22 / 103 / 125
  - Total nonadopting countries: 2 / 9 / 11

### Data and sample construction
- Regional sample: 24 countries in the region, 22 of which have adopted Basel and 2 that have not; adoption dates available for 20 adopting countries.
- Rest of world: 103 adopting countries and 9 nonadopting countries; adoption dates available for 71 rest-of-world countries.
- Bank-level dataset: 2,893 banks from over 150 countries; average of 19 banks per country.
- Time span: up to 14 years, 1987–2000, producing a total of 20,102 potential observations.
- Average number of banks per country: 132.3 on average for each country.
- Number of usable observations by variable:
  - Equity-Total Asset ratio: 20,102 observations.
  - Total Capital Ratio: 6,718 observations.
  - Tier 1 Ratio: 4,157 observations.
- Outlier treatment for loan-growth regressions: observations with annual real growth rate of loans greater than 100% or less than -50% were removed.

### Banking indicators and descriptive regional comparisons (sample means)
- Key indicators and world/LAC comparisons (1987–2000 sample means; number of observations indicated in Appendix Table A.1):
  - Net Loans/Total Assets (World average): 15.82 (Std Dev 17.07); Latin America & Caribbean: 17.06 (Std Dev 16.35); observations: 6,718.
  - Net Loans/Total Deposits & Borrowing (World average): 10.50 (Std Dev 13.61); Latin America & Caribbean: 12.72 (Std Dev 17.58); observations: 20,102.
  - Liquid Assets/Total Deposits & Borrowing (World average): 50.57 (Std Dev 21.85); Latin America & Caribbean: 51.39 (Std Dev 20.89); observations: 19,860.
  - Overhead/Total Assets (World average): 24.54 (Std Dev 69.57); Latin America & Caribbean: 20.61 (Std Dev 63.89); observations: 15,770.
  - Loan loss provisioning relative to net interest income: LAC 20.6% vs. world average 24.5%.
  - Loan loss reserves relative to total loans: LAC 5.3% vs. world average 5.8%.
  - Loan loss reserves relative to nonperforming loans: LAC 117.5%; United States and Canada 269%.
- Pre- vs. post-Basel patterns (BASELYR1):
  - Capitalization increased by regulatory measures and by average equity-asset ratio in aggregate.
  - Profitability, interest spreads, and average overhead declined in aggregate.
  - Net loans increased while liquid assets declined in the post-Basel period — no aggregate indication of Basel-induced credit contraction.
  - Regional variation:
    - LAC increased capital by more than world average and increased share of loans in assets by 5 percentage points vs. 1 for world average.
    - Europe: highest average capitalization before Basel; only region where both capitalization and lending ratios declined between periods.

### Mechanisms: how Basel I can affect bank behavior
- Three channels to increase regulatory or risk-weighted capital ratio:
  1. Increase capital (numerator).
  2. Decrease total assets (denominator).
  3. Shift composition toward lower risk-weight assets and away from higher risk-weight assets.
- Regulatory capital arbitrage (RCA): securitization and off-balance-sheet activities to decrease risk-weighted assets without raising costly capital; possible outcome: move safe loans off balance sheet and leave riskier loans on.
- Tradeoffs and incentives:
  - Deposit insurance and contagion concerns motivate capital requirements to reduce moral hazard.
  - Higher required capital increases value of holding capital, affecting intertemporal lending decisions.
  - Market contestability and financial market development influence incentives and feasibility of arbitrage.
  - Shifting away from credit-risk activities may increase exposure to market risk, interest rate risk, operational risk.

### Theoretical insights
- Chami and Cosimano (2001): bank capital endogenous; Basel-like constraints can alter optimal capital and future credit extension; in imperfectly competitive markets, constraints can increase cooperative/monopolistic behavior affecting loan supply.
- Regulatory dialectic (Kane, 2000): regulation induces evasion, prompting new regulation (e.g., Basel II responding to Basel I evasion).

### Empirical approach and estimation strategy
- Baseline means-test regression (equation (1)): Vijt = α0 + α1 BASELYRjt, where Vijt ∈ {EQTA, TOTCAPRAT, NLOANTA, RLOANGROWTH}.
- Four regression specifications: (i) OLS across countries and time; (ii) Fixed Effects with bank-specific intercepts; (iii) Regression including country dummies; (iv) Regression including FINDEV and its interaction with BASELYR.
- FINDEV: ratio of banking system credit to GDP in 1995 relative to worldwide maximum (Switzerland at 168%); FINDEV = 1 for Switzerland; examples: Mexico = 0.15, Korea = 0.32, United States = 0.41.
- Loan supply specification (equation (2)) includes lagged risk RISKjt-1 and interactions RISKjt-1 × BASELYRjt; risk measures used: EQTA and NPFRAT; macro controls Xjt include RGDPG and DM3GDP (annual change in inverse money velocity).
- Estimation methods: OLS, Fixed Effects, Instrumental Variables with Fixed Effects (instrumenting lagged bank risk variables).

### Main empirical findings (full sample)
- Capitalization:
  - Basel adoption associated with an average increase in capital and lending activities globally and in LAC.
  - Significant increase in average equity-asset ratio captured in most specifications except fixed effects.
  - Regulatory capital ratios increased significantly after Basel.
- Lending (NLOANTA and RLOANGROWTH):
  - Little evidence that loan-asset ratio or average growth rate of loans declines after Basel I adoption on average.
  - Some specifications show a significant increase in loan-asset ratio after Basel.
  - Means tests after removing outliers point to a decline in loan growth after Basel; when controlling for FINDEV, decline occurs only after a certain FINDEV level.
  - Cutoff implications from point estimates:
    - For NLOANTA, cutoff point would be 30% of maximum banking development, equivalent to 1995 private sector credit-GDP ratio of 50% or greater; countries above this (example list) would experience contraction in loans following Basel.
    - For RLOANGROWTH, a country with a credit-GDP ratio of over 64% would tend to experience a decline in lending growth.
- Risk retrenchment:
  - Some evidence of greater sensitivity of loan growth to certain risk factors after Basel consistent with “risk retrenchment.”
  - For full sample, loan growth became more sensitive to past shortfalls in equity in some specifications.
  - Instrumenting strengthens estimated responses: stronger response of loan-asset ratio to NPFRAT changes and loan growth to equity changes.
- Financial development heterogeneity:
  - More financially developed economies tend to have a contraction in loan growth after Basel; effect mainly driven by advanced European countries.
  - FINDEV often shows large, significant coefficients and FINDEV · BASELYR interactions frequently modify Basel-related effects substantially.
  - Threshold effects observed: threshold generally lies in the 40–50% FINDEV range for credit slowdown; evidence robust for RLOANGROWTH and remains after excluding early adopters.

### Latin America & Caribbean (LAC) specific findings
- Aggregate means:
  - Since Basel implementation LAC banks tended to hold a capital-asset ratio that is 4% more than the world average, and a loan-to-asset ratio about 1% over the world average.
  - Comparing pre- and post-Basel periods, return on equity decreased in LAC by about 7%, while the world average decreased by 3½ %.
  - LAC provisioning is 20.6% of net interest income vs. world average 24.5%.
- LAC loan-equation estimates:
  - Little evidence of a negative intercept shift (overall decline in NLOANTA or RLOANGROWTH) after Basel, with one exception in a loan growth regression.
  - Loan growth fits the empirical model better than loan-asset ratio: responds positively and significantly to demand-for-money (DM3GDP); loan-asset ratio does not.
  - FINDEV is positively related to loan-asset ratio but negatively related to loan growth.
  - Country-specific characteristics meaningfully increase explanatory power (R2).
- Risk retrenchment in LAC:
  - Stronger evidence that loan growth became more sensitive to past equity after Basel: interaction BASELYR1 × EQTA(-1) positive and significant across specifications.
  - No significant change in sensitivity of loan growth to NPFRAT in LAC.
  - Threshold for credit decline in LAC emerges only in one specification for NLOANTA with point estimate of 34% FINDEV (private sector credit-GDP ratio of 58%); only Panama (74%) reaches higher levels in LAC.

### Tables — selected empirical coefficients (representative examples)
- Table 5 (EQTA means test):
  - BASELYR OLS: 0.489 (t-ratio 1.65) *
  - BASELYR1 OLS: 3.218 (t-ratio 9.43) ***
  - FINDEV OLS: 6.688 (t-ratio 3.96) ***
  - Observations: 16,683
- Table 7 (NLOANTA means test):
  - BASELYR OLS: 0.850 (t-ratio 2.21) ***
  - BASELYR1 OLS: 1.022 (t-ratio 2.38) **
  - FINDEV OLS: 35.504 (t-ratio 16.28) ***
  - BASELYR · FINDEV: -31.959 (t-ratio 13.70) ***
  - Observations: 16,462
- Table 8 (RLOANGROWTH examples):
  - BASELYR1: 12.244 (t-ratio 3.59) ***
  - EQTA(-1) examples: 0.085 (t-ratio 6.24) ***
  - NPFRAT(-1) example: -0.41298 (t-ratio 2.57) ***
  - EQTA(-1) · BASELYR1: -0.154 (t-ratio 3.53) ***
  - DM2GDP example: 0.179 (t-ratio 5.67) ***
  - Observations: up to 14,041
- Table 9 (IV - Fixed Effects examples):
  - RLOANGROWTH BASELYR1: 8.564 (t-ratio 9.22) ***
  - NLOANTA FINDEV example: 23.059 (t-ratio 9.97) ***
  - Observations: examples 14,041 / 4,850
- Tables 13–15 (Latin America examples):
  - BASELYR: 11.800 (t-ratio 4.38) ***
  - FINDEV: 43.951 (t-ratio 8.26) ***
  - FINDEV · BASELYR1: -34.183 (t-ratio 4.89) ***
  - Observations: 2,908 / 1,439 (specs)

### Overall assessment
- Basel I adoption across 125 countries associated on average with:
  - Expansion of equity capital.
  - Increased lending on average.
  - Decline in return on equity, with decrease in net interest margin greater than decline in overhead expenses.
- No widespread, robust evidence of a Basel-induced credit crunch in Latin America:
  - No general loan-supply intercept shift across all banks and countries after Basel.
  - Some evidence of risk retrenchment: loan growth became more sensitive to lagged equity-asset ratio, especially in LAC.
  - Loan growth measures more consistent with credit-crunch-like behavior and macro controls than loan-asset ratios, but results are not conclusive.
- Role of financial development:
  - Findings consistent with regulatory arbitrage hypothesis: greater FINDEV associated with greater ability to circumvent regulations, altering estimated lending responses.
  - Emerging markets either used non-market arbitrage or were forced to raise costly capital; estimations suggest costly capital raising did occur with partial non-market arbitrage.

### Policy implications and projections for Basel II
- Basel I effects:
  - Increased banks’ regulatory capital and made banks more sensitive to changes in capital ratios.
  - Did not robustly increase sensitivity to credit risk (NPFRAT).
- Expectations for Basel II:
  - Basel II’s heavier reliance on risk-based capital and IRB approaches expected to increase loan sensitivity to risk.
  - Potential consequence: increased procyclicality as loan supply becomes more sensitive to risk factors varying with the business cycle.
- Policy caution:
  - Results do not imply additional and permanent credit declines in Latin America from Basel II adoption, but suggest heightened procyclicality risk.
  - Monitoring and complementary policies recommended to mitigate potential procyclicality.

### Recommendations for further research
- Use more disaggregated data to:
  - Explore heterogeneous effects across bank types (small vs. large).
  - Investigate changes in loan composition even if aggregate lending did not decline.
- Systematically assess role of other reforms (financial liberalizations; introduction of explicit deposit insurance) by combining Basel database with information on other financial sector reforms.

*Source: IMF working paper section "1. Credit Slowdowns in Latin America" and related excerpts from _wp0538.*

### 1. Credit Slowdowns in Latin America............................................................................28

### 1. Credit Slowdowns in Latin America

### Introduction: scope and literature
- Many countries have experienced significant credit slowdowns in recent years; research examines whether declines reflect depressed demand or diminished bank lending capacity (credit crunches).
- Notable international cases cited:
  - Finland: banking system credit-to-GDP ratio shrank by over 44 percentage in 1992–97.
  - Thailand: banking system credit-to-GDP ratio shrank by some 36 percentage points in 1998–2000.
- Two strands of the literature:
  - Post-crisis cases (Scandinavia, East Asia, Japan): studies by Pazarbasioglu (1997), Ghosh and Ghosh (1999), Woo (1999) largely find demand contractions rather than supply-driven credit crunches.
  - United States early 1990s: banking system reduced credit by an amount equal to 13 percentage points of GDP between 1990 and 1993; studies explore linkage to Basel Accord adoption and find mixed evidence on a "capital crunch" and bank-lending channel (Bernanke and Lown (1991); Berger and Udell (1994); Peek and Rosengren (1995, 1997, 2000)).

### Credit slowdowns in Latin America: patterns and magnitudes
- A variety of Latin American countries experienced notable credit slowdowns during the past 20 years, many in the 1990s.
- Examples and magnitudes from Table 1 (as described in text):
  - Bolivia: bank credit-GDP ratio decline approaching 16 percentage points.
  - Mexico: bank credit-GDP ratio decline approaching 20 percentage points.
  - Panama: 16 percentage point decline over the last two years (noted as a notable exception on an average annual basis).
- Typical average annual declines: most tended to be in the range of 1–3 percentage points.
- Even when declines ended before 2003, more recent credit growth tended to be sluggish as well.
- Argentina: differentiated periods—credit slowdown (1999–2000) versus crisis-related decline (2001–03) driven largely by substantial deposit outflows.

### Findings from Barajas and Steiner (2002) on eight LAC cases
- Countries analyzed: Argentina, Bolivia, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela.
- Detailed analysis of three recent episodes: Colombia, Mexico, and Peru—estimated separate supply and demand functions.
- Three major common elements across countries:
  - Both reductions in sources of funds and increases in alternative uses of funds occurred during the credit decline.
  - The three detailed cases contained some degree of contraction in credit supply:
    - Mexico: supply contraction was swamped by demand-side contraction.
    - Colombia: demand and supply contractions were of roughly similar magnitude.
  - Regulatory and risk-related variables were significant in explaining credit supply contraction.
    - Proxy variables used: ratio of loan loss provisions to nonperforming loans (proxy for severity of regulations on bank risk-taking); credit risk measured as ratio of nonperforming to total loans.

### Potential role of Basel Accord adoption
- Basel Accord (Basel I), originally negotiated in 1988, established risk-based minimum capital requirements and defined forms of capital and risk weights by asset category.
- Many Latin American countries adopted the Basel Accord, imposing risk-based minimum capital requirements.
- Hypothesis: Basel adoption may be related to recent credit slowdowns in Latin America as a regulatory factor not previously included in some studies.

### Data on Basel adoption: database construction and sources
- The authors constructed a cross-country database on Basel adoption primarily based on a mid-2003 survey conducted with IMF desk economists.
- Supplementary sources included:
  - World Bank database on bank regulation and supervision.
  - Various central bank websites.
  - IMF Monetary and Financial Systems Department (MFD) regulatory database.
  - Research papers dating Basel adoption, notably Chiuri, Ferri and Majnoni (2002).
- Findings summarized in Table 2 (described in text):
  - Evidence of Basel adoption in LAC beginning in 1991 and continuing throughout the decade, with most countries adopting Basel between 1993 and the late 1990s.

*Source: IMF working paper section "1. Credit Slowdowns in Latin America" (text and tables referenced in original PDF).*

### 1997. In our sample, we obtained a total of 24 countries in the region, 22 of which have

### _wp0538 - 1997. In our sample, we obtained a total of 24 countries in the region, 22 of which have

### Sample and Basel I adoption statistics
- Regional sample: 24 countries in the region, 22 of which have adopted Basel and 2 that have not.
- For the adopting countries in the region: specific adoption dates available for 20 countries.
- Rest of world: 103 adopting countries and 9 nonadopting countries.
- Adoption dates available in rest of world: 71 countries.
- Bank-level dataset: 2,893 banks drawn from over 150 countries; average of 19 banks per country.
- Time span for bank observations: up to 14 years, spanning from 1987 to 2000, producing a total of 20,102 potential observations.

### Empirical evidence and prior literature on Basel I effects
- Overall empirical evidence on Basel I’s impact on bank stability is mixed and often weak:
  - Barth, Caprio and Levine (2004): using cross-section regressions on a sample of close to 50 countries, find minimum capital requirements are positively related to stability (associated with a lower probability of crisis) but this result is not robust across specifications when other regulatory variables are included.
- Cross-country and single-country studies show varied results:
  - BIS (1999) survey of early adopters (G-10) showed capital ratios increased generally via a combination of banks raising new capital and reducing risk-weighted assets; also evidence of significant and growing regulatory arbitrage via securitization, particularly in the United States.
  - Dionne and Harchaoui (2003): relationship between securitization, regulatory capital ratios, and risk in Canada.
  - Konishi and Yasuda (2004): in Japan, risk measures based on bank stock prices showed risk was reduced after Basel I adoption.
  - Van Roy (2003): G-10 study found reduced credit risk-taking.
  - Chiuri et al. (2002): in 16 emerging markets, Capital Adequacy Ratio may have contributed to credit crunches in countries that experienced a financial crisis; effect greatest for banks initially not well capitalized; results support procyclicality of risk-based capital requirements.
- Recent literature on Basel II highlights continued heavy reliance on risk-based capital:
  - Of the 163 pages of the Basel Committee’s 2003 Consultative Paper, 132 pages are devoted to minimum capital requirements (“Pillar 1”) (Von Thadden, 2004).
- Concerns about Basel II and procyclicality:
  - IRB approach requires banks to assign default probabilities that depend on current information (equity price, book liabilities) affected by business cycle.
  - Studies (Ayuso, Perez and Saurina (2004); Caterineu-Rabell et. al. (2003); Kashyap and Stein (2004); Estrella (2004)) show banks may implement procyclical loan rating schemes, exacerbating business cycle effects.

### Mechanisms through which Basel I can affect bank behavior
- Three channels for increasing regulatory or risk-weighted capital ratio:
  1. By increasing capital (the numerator).
  2. By decreasing total assets.
  3. By shifting composition of assets toward lower risk-weight assets and away from higher risk-weight assets to lower the denominator (example: reduce commercial loans (100% risk weight) in favor of securities (zero risk weight)).
- Regulatory capital arbitrage (RCA) behavior:
  - Securitization and off-balance-sheet activities used to decrease risk-weighted assets without raising costly capital.
  - Example outcome: move safe loans off the balance sheet and leave riskier loans on.
- Tradeoffs and incentives:
  - Deposit insurance and contagion concerns motivate capital requirements to reduce moral hazard and probability of default, protecting taxpayers.
  - Higher required capital increases the value to banks of holding capital (modeled as a call option), influencing intertemporal lending decisions.
  - Market contestability affects incentives: greater contestability reduces the value of holding capital and may increase incentive to engage in regulatory arbitrage.
  - Level of financial market development and strength of legal/institutional frameworks affects banks’ ability to arbitrage (e.g., securitization more feasible in well-developed markets).
  - Moving away from credit-risk activities to activities with lower regulatory capital may lead to exposures to other risks (market risk, interest rate risk, operational risk).

### Theoretical insights drawn from models
- Chami and Cosimano (2001):
  - Treat bank capital as endogenous; show how regulatory changes and exogenous factors (market structure, economic activity) affect optimal capital and future credit extension.
  - In imperfectly competitive banking markets, Basel-like constraints can increase bank cooperation/monopolistic behavior, affecting loan supply and strategic interactions.
  - Higher required capital increases the value of holding capital; banks may hold excess capital to avoid future constraints and ensure ability to lend.
- Regulatory dialectic (Kane, 2000): regulation induces evasion, prompting new regulation (e.g., Basel II addressing banks’ attempts to avoid Basel I rules).

### This study’s methodological approach and distinctions from prior work
- Uses new dataset on Basel I adoption combined with a 2,893-bank dataset from over 150 countries to examine whether Basel I caused banks to reduce lending.
- Tests hypotheses for the world overall and for the Latin America and Caribbean (LAC) region in particular.
- Advantages over some prior studies:
  - Able to examine impact of Basel I over time as well as across countries, unlike Barth et al. (2004) who use a single time observation for capital regulations.
  - Utilizes Berger and Udell (1994) framework to test for structural change in banks’ overall loan supply and sensitivity to risk, requiring sufficient pre- and post-Basel observations for long-run impact assessment.
  - Contrasts with Peek and Rosengren (1995) framework and Chiuri et al. (2002), which test short-run impact with limited pre- and post-observations.

### Main empirical findings of this study
- Basel Accord association:
  - Associated with an average increase in capital and lending activities in Latin America as well as throughout the world.
- Credit crunch hypothesis:
  - Little evidence that either the loan-asset ratio or the average growth rate of loans declines after Basel I adoption.
  - Some evidence of greater sensitivity of loan growth to certain risk factors consistent with “risk retrenchment” under greater regulatory scrutiny.
  - In Latin America, observed primarily as greater sensitivity of loan growth to past shortfalls in equity.
- Financial development and loan growth:
  - More financially developed economies tend to have a contraction in loan growth after Basel.
  - This result is mainly driven by more advanced countries in Europe; the effect weakens considerably when earlier adopters of Basel are excluded or when focusing only on countries in Latin America and the Caribbean.

### Structure of the remainder of the paper (as indicated)
- Next sections:
  - Summary of existing theoretical work on impact of capital constraints on optimal bank behavior.
  - Description of the banking dataset and presentation of main empirical results.
  - Conclusion and policy implications.

*Source: _wp0538 - 1997. In our sample, we obtained a total of 24 countries in the region, 22 of which have adopted Basel and 2 that have not (excerpt).*

### 132.3 on average for

### _wp0538 - 132.3 on average for

### Data and sample
- Sample period constructed from combined Bankscope CD’s to build the 1987–2000 time period.
- Average number of banks per country: 132.3 on average for each country.
- Number of usable observations varies by variable:
  - Equity-Total Asset ratio: 20,102 observations.
  - Total Capital Ratio: 6,718 observations.
  - Tier 1 Ratio: 4,157 observations.
- Dataset expansion underway to include more recent observations and a greater number of banks per country.
- Early adopters: 14 countries adopted the Basel Accord in 1988; some specifications exclude these early adopters.
- Outlier treatment for loan-growth regressions: observations with annual real growth rate of loans greater than 100% or less than -50% were removed.

### Banking indicators and definitions used
- Four indicator groups in Table 3: capitalization; profitability and operational costs; asset composition and liquidity; loan loss provisioning.
- Capitalization measures:
  - Ratio of equity to total assets (EQTA).
  - Total Capital Ratio (TOTCAPRAT) and Tier 1 Capital Ratio (T1RAT), where denominators equal risk-weighted assets.
- Profitability and costs:
  - Return on assets.
  - Net interest margin.
  - Ratio of overhead to assets.
- Asset composition and liquidity:
  - Net loans to total assets (NLOANTA).
  - Net loans to deposits (LOANDEP).
  - Liquidity: coverage of short-term liabilities (deposits) by short-term assets.
- Loan loss provisioning:
  - Provisioning cost as percentage of interest revenue.
  - Stock of accumulated reserves as percentage of nonperforming and gross loans.
- Basel timing variables constructed:
  - BASELYR: year of legal adoption of Basel.
  - BASELYR1: implementation year (where distinguishable).

### Descriptive regional comparisons (selected findings)
- Latin America and the Caribbean (LAC) vs. world:
  - Capitalization and profitability: LAC banks appear similar or slightly higher than world averages.
  - Interest margins and overhead costs: LAC banks have decidedly higher interest margins and overhead costs; they are the highest of any region.
  - Lending and liquidity: LAC banks concentrate more of their activities in lending and less in liquid assets.
  - Loan loss provisioning relative to net interest income: LAC provisioning is 20.6% vs. world average of 24.5%.
  - Loan loss reserves relative to total loans: LAC is 5.3% vs. world average of 5.8%.
  - Loan loss reserves relative to nonperforming loans: LAC banks have reserves at 117.5% of nonperforming loans.
  - United States and Canada: reserves at 269% of nonperforming loans.
- Interpretation notes:
  - Lower provisioning ratios in LAC partly reflect higher interest spreads and larger size of lending activities.
  - High interest revenues allow provisioning to be less of a financial burden in LAC.
  - Moral hazard and weak regulation/supervision are cited as potential issues, but other studies find measurable market discipline even where explicit deposit insurance exists.

### Pre- vs. post-Basel comparisons (aggregate patterns)
- Using BASELYR1 to separate pre- and post-Basel periods, Table 4 findings:
  - Capitalization increased as measured by regulatory measures and the average equity-asset ratio.
  - Profitability, interest spreads, and average overhead declined.
  - Equity-to-total-assets increased less than regulatory capital measures.
  - Bank lending (scaled by total assets or short-term liabilities):
    - Net loans increased while liquid assets declined in the post-Basel period — no indication of a Basel-induced credit contraction in aggregate.
    - Post-Basel period coincided with increased bank intermediation activities and potentially greater vulnerability to deposit outflows; overall default risk presumably declined due to increased capitalization.
- Regional variation:
  - LAC: increased capital by more than world average and increased share of loans in assets by 5 percentage points vs. 1 for world average.
  - Europe: highest average capitalization rates before Basel; only region where both capitalization and lending ratios declined between periods.
  - Provisioning changes:
    - Increased in Asia, Latin America, and most notably the United States and Canada.
    - Decreased in Europe and the Middle East.
    - Africa: provisioning declined relative to interest revenue, but reserves-to-nonperforming-loans increased.

### Estimation approach — tests of Basel effects (full sample)
- Baseline means-test regression (equation (1)):
  - Vijt = α0 + α1 BASELYRjt
  - Dependent variables Vijt: EQTA, TOTCAPRAT, NLOANTA, RLOANGROWTH.
  - Indices i (bank), j (country), t (year).
- Four regression specifications run:
  - (i) OLS across countries and time.
  - (ii) Fixed Effects with bank-specific intercepts.
  - (iii) Regression including country dummies.
  - (iv) Regression including FINDEV and its interaction with BASELYR.
- FINDEV definition:
  - FINDEV = ratio of banking system credit to GDP in 1995 relative to worldwide maximum (Switzerland at 168%).
  - FINDEV = 1 for Switzerland; examples: Mexico = 0.15, Korea = 0.32, United States = 0.41.
- Additional robustness:
  - Tests run for full sample and excluding early adopters.
  - Also substituted BASELYR1 for BASELYR to examine implementation effects.

### Main estimation findings (capitalization and lending)
- Capitalization:
  - Significant increase in average equity-asset ratio captured in all except the fixed effects regressions.
  - Regulatory capital ratios increased significantly after Basel.
  - Banks in more financially developed countries had lower capitalization rates on average prior to Basel but tended to increase capitalization by a greater amount after Basel.
- Loan-asset ratio (NLOANTA):
  - Simple OLS and fixed effects estimates do not show a shift away from loans after Basel; instead, there is evidence of a significant increase in the loan-asset ratio after Basel in some specifications.
  - Interaction with FINDEV: the increase in loan-asset ratio after Basel tended to be smaller as countries were more financially developed.
  - Cutoff implication from point estimate in Table 7:
    - Cutoff point would be 30% of the maximum level of banking development, equivalent to a 1995 ratio of private sector credit to GDP of 50% or greater.
    - Countries such as Australia (73%), Korea (53%), and the United Kingdom (115%) would experience a contraction in loans following Basel; Chile (49%) would experience virtually no change.
- Real growth rate of loans (RLOANGROWTH):
  - Means tests after removing outliers point to a decline in loan growth after Basel.
  - Controlling for relative financial development shows the decline occurs only after a certain level of FINDEV.
  - Cutoff from point estimate in Table 8:
    - A country with a credit-GDP ratio of over 64% would tend to experience a decline in lending growth.

### Risk retrenchment hypothesis and loan-supply specification
- Purpose: test whether Basel made banks more sensitive to risk (risk retrenchment) or whether non-risk factors caused loan-supply shifts.
- Loan supply specification (equation (2)):
  - Vijt = α0 + α1 BASELYRjt + α2 RISKjt-1 + α3 (RISKjt-1 × BASELYRjt) + α4 Xjt + α5 BASELYRjt × Xjt
  - Dependent variables V: NLOANTA and RLOANGROWTH.
  - Risk measures used: capital ratio (EQTA) and nonperforming loans ratio (NPFRAT).
  - Macroeconomic controls X:
    - RGDPG: real growth rate of GDP (loan demand proxy).
    - DM3GDP: annual change in inverse money velocity (captures shifts in public demand for bank deposits and structural financial changes).
- Interpretation:
  - If Basel increased sensitivity to risk, increases in risk would be associated with larger contractions in lending post-Basel (consistent with Berger and Udell’s “risk retrenchment”).
  - DM3GDP expected to capture deposit demand shifts and large structural financial changes (e.g., financial liberalization lowering income velocity through higher deposit demand).

### Summary of overall conclusions from the analyzed sections
- Adoption of Basel I globally is associated with:
  - Expansion of equity capital.
  - Increased lending on average.
  - On average, return on equity declined, with the decrease in net interest margin greater than that in overhead expenses.
  - Europe is an exception, characterized by relatively high banking development and different post-Basel patterns.
- Evidence is nuanced and heterogeneous across regions and by level of financial development; some countries with high FINDEV experienced lending contractions or reduced loan growth after Basel, while many others experienced increased lending activity alongside higher capitalization.

*Source: Excerpt from IMF working paper _wp0538 - 132.3 on average for (PDF)._*

### introduction of a deposit insurance scheme, which makes bank deposits safer in the eyes of

### _wp0538 - introduction of a deposit insurance scheme, which makes bank deposits safer in the eyes of

### Empirical approach and identification
- Estimated effects of Basel I on bank lending using panel regressions at the bank level with macro controls.
- Key dependent variables:
  - NLOANTA (loan-asset ratio).
  - RLOANGROWTH (loan growth rate).
- Key bank risk variables:
  - EQTA (equity-asset ratio).
  - NPFRAT (nonperforming loan ratio).
- Macro controls included real GDP growth and the absolute annual change in the ratio of money and quasi-money to GDP (DM3GDP).
- Three estimation methods presented: OLS, Fixed Effects, and Instrumental Variables with Fixed Effects (instrumenting lagged bank risk variables with their lags and macro controls).
- Basel impact measured by coefficient α3 on interaction term between BASELYR (or BASELYR1) and each risk variable; α1 captures a direct intercept shift following Basel.

### Main findings on banks’ response to risk (full sample)
- Banks respond to risk in expected directions:
  - Loan supply contracts after a drop in capital (EQTA).
  - Loan supply contracts after increases in nonperforming loans (NPFRAT).
  - Real GDP growth is positively associated with a greater proportion of bank assets devoted to loans.
- Instrumenting tends to strengthen estimated responses:
  - Stronger estimated response of loan-asset ratio to changes in nonperforming loans.
  - Stronger estimated response of loan growth to changes in equity.

### Effects of Basel I on lending behavior (full sample)
- Loan-asset ratio (NLOANTA):
  - OLS: Basel adoption associated with a direct increase in loans (intercept shift).
  - Once bank-specific determinants (fixed effects) are included, the direct positive effect weakens.
  - Interaction coefficients generally did not show increased sensitivity to EQTA or NPFRAT after Basel; in many cases signs were not as expected and were not statistically significant.
  - In several regressions at least one coefficient signaled a decline in sensitivity toward risk after Basel (i.e., reduced prudential response).
- Loan growth (RLOANGROWTH):
  - Results are generally stronger and somewhat more consistent with a Basel-induced credit contraction.
  - Loan growth responses to bank fundamentals and macro controls were consistently of expected sign and generally statistically significant.
  - Direct intercept shift after Basel was usually positive.
  - Some indication of risk retrenchment: greater sensitivity of lending to nonperforming loans (negative interaction coefficient), though often not statistically significant; occasionally greater sensitivity to equity.

### Cross-country heterogeneity and financial development
- Regressions including financial development indicator (FINDEV) and its interaction with BASELYR1:
  - Lending shrinks in response to greater risk.
  - Direct and risk-related impacts of Basel adoption become more robust across specifications and samples when accounting for country heterogeneity.
  - Adoption appeared to increase loan supply directly in virtually all regressions with FINDEV included.
- Three key differences between NLOANTA and RLOANGROWTH:
  - Sensitivity of loan-asset ratio to risk tended to decline after Basel across specifications, while sensitivity of loan growth generally increased after Basel.
  - Loan growth responds more to macro controls (economic growth and change in money demand proxied by inverse velocity of M3); loan-asset ratio does not respond as well to these macro controls.
  - Level of banking development is highly correlated with loan-asset ratio but is not a significant explanatory variable for loan growth.
- Financial development moderates Basel effects:
  - Direct positive impact of Basel on lending is smaller the more financially developed the country is.
  - Evidence of a threshold beyond which Basel may have led to a credit slowdown; threshold generally lies in the 40–50% range (financial development index). This effect is particularly evident for RLOANGROWTH and remains after excluding early adopters.
  - For NLOANTA the effect can become positive once early adopters are excluded.

### Country-dummy specifications
- Replacing FINDEV with country dummies produces overwhelming joint significance and substantially increases R2.
- Direct increase of loan growth after Basel becomes less clear with country dummies, but there is evidence of increased sensitivity of loans to risk after Basel (e.g., loan growth may have become more sensitive to past equity).

### Latin America (LAC) specific results
- Means tests and region-dummy regressions:
  - After Basel, LAC banks increased capitalization rates, the proportion of assets devoted to loans, and loan growth rates on average.
  - Since Basel implementation LAC banks tended to hold a capital-asset ratio that is 4% more than the world average, and a loan to asset ratio about 1% over the world average.
  - Comparing pre- and post-Basel periods, return on equity decreased in LAC by about 7%, while the world average decreased by 3½ %.
  - Financial development in LAC reduced interest spreads and overhead costs between pre- and post-Basel periods, but spreads and costs remain high by world standards.
- Regional comparisons:
  - Prior to Basel, LAC capitalization rates were about average; Asia and USCANADA had lower levels, Europe higher.
  - After Basel, Asian banks increased capital by less than LAC; Europe appears to have reduced capitalization in absolute terms.
  - USCANADA increased regulatory capital by more than LAC but their average equity by less.
  - Impact of Basel on loan growth greater in USCANADA than in LAC, and less in Asia and Europe; Europe may have lowered loan growth on average after Basel.
  - Africa and MIDEAST had capitalization and lending growth rates similar to LAC before and after Basel.
- LAC loan-equation estimates (subsample regressions):
  - Little evidence of a negative intercept shift (overall decline in NLOANTA or RLOANGROWTH) after Basel, with one exception in a loan growth regression (Table 15).
  - Loan growth fits the empirical model better than loan-asset ratio: responds positively and significantly to demand for money; loan-asset ratio does not.
  - FINDEV is positively related to loan-asset ratio but negatively related to loan growth.
  - Country-specific characteristics meaningfully increase explanatory power (R2).
- Risk retrenchment in LAC:
  - Stronger evidence that loan growth became more sensitive to past equity after Basel: interaction between BASELYR1 and EQTA(-1) is positive and significant across specifications.
  - No significant change in sensitivity of loan growth to the nonperforming loan ratio in LAC.
  - Within LAC, the effect of Basel on loans does not seem to depend robustly on level of financial development; a threshold effect for credit decline arises only in a single specification for NLOANTA (first column of Table 14), where the point estimate suggests a threshold of 34% for the financial development indicator, or a private sector credit-GDP ratio of 58% — this level is achieved by only one country in the LAC region, Panama (74%).

### Overall assessment and summary
- Coverage: Basel I adoption during the 1990s eventually covered a total of 125 countries.
- Main conclusions:
  - Empirical evidence is weak for a widespread Basel-induced credit crunch in Latin America.
  - No general loan supply shift (intercept shift) across all banks and countries after Basel is found.
  - Some evidence of risk retrenchment: loan growth became more sensitive to the lagged equity-asset ratio, especially in LAC.
  - For the full sample, risk retrenchment is present but weaker than in LAC.
  - Loan growth measures tend to be more consistent with credit-crunch-like behavior and macro controls than loan-asset ratios, though results remain far from conclusive.
- Role of financial development:
  - Findings are consistent with the regulatory arbitrage hypothesis: greater financial development is associated with banks’ greater ability to circumvent costly regulations (e.g., reduce risk-weighted assets), leading to an estimated negative effect of FINDEV on lending in some specifications.
  - Emerging markets either used non-market arbitrage or were forced to raise costly capital to comply; estimations suggest costly capital raising did occur, with some partial non-market arbitrage.

### Policy implications and projections for Basel II
- Basel I effects:
  - Basel I succeeded in increasing banks’ regulatory capital and made banks more sensitive to changes in their capital ratios.
  - Basel I did not appear to increase banks’ sensitivity to credit risk (NPFRAT) in a robust way.
- Expectations for Basel II:
  - Basel II is expected to incorporate a wider range of weights applied to different risk types and a more accurate approximation of credit risk.
  - Anticipated outcome: loan sensitivity to risk is expected to increase under Basel II.
  - Potential consequence: Basel II might cause credit to become more procyclical as loan supply becomes more sensitive to risk factors that vary with the business cycle.
- Policy caution:
  - Results do not give cause for concern of additional and permanent credit declines in Latin America from Basel II adoption, but they do suggest increased procyclicality risk.
  - Monitoring and complementary policies may be warranted to mitigate potential procyclicality.

### Recommendations for further research
- Use more disaggregated data to:
  - Explore heterogeneous effects across different bank types (e.g., small vs. large banks).
  - Investigate changes in loan composition even if aggregate lending did not decline.
- Explicitly and systematically assess the role of other reforms:
  - Financial liberalizations that may have stimulated bank competition and lending.
  - Introduction of explicit deposit insurance systems, which may have provided incentives for banks to expand credit.
- Combine Basel database with information on different aspects of financial sector reforms to incorporate these policy changes more explicitly.

*Source: IMF working paper content provided.*

### Appendix Table A.1. Adoption and Implementation Dates—Various Countries

### Appendix Table A.1. Adoption and Implementation Dates—Various Countries

### Adoption of the Basel Accord (Table 2)
- Number of countries adopting by year (Latin America & Caribbean / Rest of the World / Total):
  - 1988: 0 / 16 / 16
  - 1989: 0 / 0 / 0
  - 1990: 0 / 1 / 1
  - 1991: 1 / 2 / 3
  - 1992: 0 / 3 / 3
  - 1993: 4 / 8 / 12
  - 1994: 3 / 5 / 8
  - 1995: 3 / 6 / 9
  - 1996: 1 / 4 / 5
  - 1997: 4 / 7 / 11
  - 1998: 2 / 3 / 5
  - 1999: 1 / 6 / 7
  - 2000: 1 / 6 / 7
  - 2001: 0 / 4 / 4
  - 2002: 0 / 0 / 0
- Totals:
  - Total dated: 20 / 71 / 91
  - Total undated: 2 / 32 / 34
  - Total adopting countries: 22 / 103 / 125
  - Total nonadopting countries: 2 / 9 / 11

### Selected Banking Indicators by Region, 1987–2000 (Table 3 — sample means)
- Net Loans/Total Assets (Mean / Standard Deviation):
  - Africa: 20.24 / 30.31
  - Asia: 13.98 / 14.73
  - Europe: 16.84 / 14.62
  - Middle East: 20.38 / 28.52
  - US and Canada: 11.63 / 10.43
  - Latin America & Caribbean: 17.06 / 16.35
  - World average: 15.82 / 17.07
  - Number of observations for this variable: 6,718
- Net Loans/Total Deposits & Borrowing (Mean / Standard Deviation):
  - Africa: 30.31 / 12.10
  - Asia: 9.12 / 14.33
  - Europe: 10.22 / 11.68
  - Middle East: 10.02 / 12.97
  - US and Canada: 8.29 / 5.49
  - Latin America & Caribbean: 12.72 / 17.58
  - World average: 10.50 / 13.61
  - Number of observations: 20,102
- Liquid Assets/Total Deposits & Borrowing (Mean / Standard Deviation):
  - Africa: 48.77 / 22.57
  - Asia: 58.41 / 16.85
  - Europe: 45.86 / 23.71
  - Middle East: 45.15 / 17.12
  - US and Canada: 64.56 / 16.64
  - Latin America & Caribbean: 51.39 / 20.89
  - World average: 50.57 / 21.85
  - Number of observations: 19,860
- Overhead/Total Assets (Mean / Standard Deviation):
  - Africa: 25.83 / 54.20
  - Asia: 30.22 / 88.12
  - Europe: 23.37 / 64.90
  - Middle East: 26.41 / 78.25
  - US and Canada: 24.01 / 59.58
  - Latin America & Caribbean: 20.61 / 63.89
  - World average: 24.54 / 69.57
  - Number of observations: 15,770
- Capitalization and other indicators (selected):
  - Total Capital Ratio (World average): 50.57 (Mean) / 21.85 (Std Dev) — number of observations 19,860
  - Tier 1 Ratio (World average): 61.48 (Mean) / 34.29 (Std Dev) — number of observations 9,835
  - Equity/Total Assets (World average): 34.94 (Mean) / 37.60 (Std Dev) — number of observations 8,536

### Pre- and Post-Basel Comparison by Region (Table 4 — sample means)
- World average (selected indicators, Pre-Basel / Post-Basel):
  - Net Loans/Total Assets: 14.52 / 15.20 (number of observations: Pre-Basel 1,221; Post-Basel 4,582)
  - Net Loans/Total Deposits & Borrowing: 12.32 / 9.12 (number of observations: Pre-Basel 4,635; Post-Basel 11,625)
  - Liquid Assets/Total Deposits & Borrowing: 50.58 / 51.24 (number of observations: Pre-Basel 4,748; Post-Basel 11,714)
  - Overhead/Total Assets: 23.55 / 25.42 (number of observations: Pre-Basel 3,597; Post-Basel 9,760)
- Note: "For the subset of countries for which Basel adoption and/or implementation dates are available, this table shows the means of the above variables in each region before (BASELYR1=0) and after (BASELYR1=1) Basel Accord implementation or adoption. Because of missing values, values between the two periods are not strictly comparable."

### Means Tests and Regression Results — Key Coefficients and Findings (Tables 5–15)
- Table 5 — Means Test for the Ratio of Equity to Assets (EQTA):
  - BASELYR (Adoption date) OLS coefficient: 0.489 (t-ratio 1.65) *
  - BASELYR1 (Implementation date) OLS coefficient: 3.218 (t-ratio 9.43) ***
  - FINDEV (financial development variables) OLS coefficient: 6.688 (t-ratio 3.96) ***
  - Number of observations: 16,683 (various specifications)
  - F-Tests for joint significance: Fixed effects 7.20 (0.00); Country dummies 96.79 (0.00)
- Table 6 — Means Test for Bank Total Capital Ratio (TOTCAPRAT):
  - BASELYR (Adoption date) OLS coefficient: -1.593 (t-ratio 1.64)
  - BASELYR1 (Implementation date) OLS coefficient: 1.513 (t-ratio 1.45)
  - FINDEV OLS coefficient: -1.967 (t-ratio 0.36)
  - BASELYR · FINDEV interaction: -10.201 (t-ratio 1.82) *
  - Number of observations: 5,803 across specifications
  - F-Tests: Fixed effects 3.84 (0.00); Country dummies 4,350.9 (0.00)
- Table 7 — Means Test for Bank Net Loans as a Percentage of Assets (NLOANTA):
  - BASELYR (Adoption date) OLS coefficient: 0.850 (t-ratio 2.21) ***
  - BASELYR1 (Implementation date) OLS coefficient: 1.022 (t-ratio 2.38) **
  - FINDEV OLS coefficient: 35.504 (t-ratio 16.28) ***
  - BASELYR · FINDEV interaction: -31.959 (t-ratio 13.70) ***
  - Number of observations: 16,462
  - F-Tests: Fixed effects 27.97 (0.00); Country dummies 239.96 (0.00)
- Table 8 — Means Test for Real Loan Growth (RLOANGROWTH):
  - BASELYR1 coefficient examples (various specifications): 12.244 (t-ratio 3.59) ***
  - Bank risk variables:
    - EQTA(-1) coefficient examples: 0.010 (t-ratio 0.29); 0.085 (t-ratio 6.24) ***
    - NPFRAT(-1) coefficient examples: -0.41298 (t-ratio 2.57) ***
  - Interaction terms:
    - EQTA(-1) · BASELYR1: -0.154 (t-ratio 3.53) ***
    - NPFRAT(-1) · BASELYR1: 0.036 (t-ratio 0.22)
  - Macro controls:
    - Real GDP growth, DRGDP examples: 0.014 (t-ratio 0.26); 0.365 (t-ratio 4.47) ***
    - Change in inverse velocity, DM2GDP examples: 0.179 (t-ratio 5.67) ***
  - Number of observations: up to 14,041 depending on specification
  - R-squared examples: 0.007 / 0.070 / 0.076 (various specs)
- Table 9 — Effects of the Basel Accord on Bank Loans (IV - Fixed Effects, dependent variable RLOANGROWTH and NLOANTA):
  - RLOANGROWTH (IV - Fixed Effects) BASELYR1: 8.564 (t-ratio 9.22) ***
  - NLOANTA (IV - Fixed Effects) FINDEV examples: 23.059 (t-ratio 9.97) ***
  - FINDEV · BASELYR1 interactions show large negative and positive coefficients in different specs (examples: -21.998 (t-ratio 9.97) ***, 6.024 (t-ratio 1.79) *)
  - Number of observations examples: 14,041 / 4,850 / 4,850 (various columns)
  - R-squared examples: 0.016 / 0.112 / 0.112 (various specs)
- Table 10–11 — OLS Regressions (Full sample, excluding early adopters, and with country dummies):
  - RLOANGROWTH BASELYR examples: 11.684 (t-ratio 3.06) *** ; 13.447 (t-ratio 3.14) ***
  - NLOANTA BASELYR examples: 1.233 (t-ratio 1.00) ; -0.325 (t-ratio 0.24)
  - With country-specific dummies, F-test for joint significance of country dummies: 234.58 (0.00) and other large values
  - Number of observations examples: 14,120 / 4,882 / 4,882; R-squared examples: 0.291 / 0.290 / 0.291
- Table 12 — Means Tests for Capitalization and Lending — Latin America and Other Regions:
  - BASELYR (Adoption date) examples: 3.647 (t-ratio 5.28) *** ; BASELYR1: 3.683 (t-ratio 5.39) ***
  - Region dummy examples (coefficients and t-ratios listed for AFRICA, EUROPE, ASIA, USCANADA, MIDEAST)
  - Interaction terms (BASELYR·Region) show substantial negative coefficients for EUROPE and ASIA (examples: BASELYR·EUROPE -5.920 (t-ratio 5.98) ***; BASELYR·ASIA -6.173 (t-ratio 7.18) ***)
  - Number of observations: 16,793 / 5,803 / 5,803 / 16,572 / 13,141
  - R-squared examples: 0.025 / 0.031 / 0.092 / 0.008
- Tables 13–15 — Effects of the Basel Accord on Bank Loans in Latin America (OLS specs, including country dummies):
  - Table 13 (Latin America OLS examples):
    - BASELYR: 11.800 (t-ratio 4.38) ***
    - FINDEV: 43.951 (t-ratio 8.26) ***
    - FINDEV · BASELYR1: -34.183 (t-ratio 4.89) ***
    - Number of observations: 2,908 / 1,439 / 1,439 / 2,591 / 1,308
    - R-squared examples: 0.051 / 0.135 / 0.142
  - Table 14 (Latin America OLS with other specs):
    - BASELYR examples: 13.012 (t-ratio 1.75) *
    - Interaction and bank risk coefficients reported with statistical significance (e.g., EQTA(-1) · BASELYR1 -0.285 (t-ratio 2.12) **)
    - Number of observations: 2,908 / 1,439 / 1,439; R-squared examples: 0.226 / 0.249 / 0.255
  - Table 15 (Latin America OLS with country-specific dummy variables):
    - BASELYR examples: 13.012 (t-ratio 1.75) *
    - F-Test for joint significance of country dummies: 50.78 (0.00); 91.13 (0.00); 81.64 (0.00)
    - Number of observations and R-squared values match Tables 13–14 specifications

### Overarching empirical patterns reported in these tables
- Adoption and implementation of the Basel Accord is associated in many specifications with statistically significant changes in capitalization measures (EQTA, TOTCAPRAT) and lending behavior (NLOANTA, RLOANGROWTH), with effects varying by:
  - whether the date used is adoption or implementation (BASELYR vs BASELYR1),
  - country coverage (full sample vs excluding early adopters),
  - region (interaction terms: BASELYR·Region show heterogeneous effects),
  - level of financial development (FINDEV) and its interaction with Basel-related dummies.
- Financial development (FINDEV) often shows large and significant coefficients, and the FINDEV · BASELYR1 interaction frequently modifies the Basel-related effect substantially (examples: FINDEV 35.504 (t-ratio 16.28) ***, FINDEV · BASELYR -31.959 (t-ratio 13.70) *** in Table 7; FINDEV 43.951 (t-ratio 8.26) ***, FINDEV · BASELYR1 -34.183 (t-ratio 4.89) *** in Table 13).
- Bank risk variables (EQTA(-1), NPFRAT(-1)) and their interactions with Basel dummies are statistically significant in multiple specifications, indicating that pre-existing bank risk and capitalization interact with Basel adoption/implementation in determining lending outcomes and loan growth.

*Source: _wp0538 - Appendix Table A.1. Adoption and Implementation Dates—Various Countries (IMF PDF content provided).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp0538.pdf_