## 2.  Regression Results Under Benchmark Specification

## Source details

**Canonical URL:** [2.  Regression Results Under Benchmark Specification](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1151.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1151.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1151.pdf.json)

---

### Overview
- Section title: "Regression Results Under Benchmark Specification"
- Location in document: page 9
- Purpose: Present benchmark panel regression estimates explaining banking sector private credit growth (Credit Growth_i,t) for an unbalanced panel of EMEs over 2002Q1–2010Q1, and interpret the roles of domestic deposit growth, non-resident liability growth, macroeconomic conditions, and monetary policy.

### Benchmark regression (full sample 2002Q1-2010Q1)
- Model features:
  - Dependent variable: banking sector private credit (Credit Growth_i,t).
  - Key explanatory variables and interactions:
    - Deposit growth × Shdepo_i,t-4 (Deposit Growth_i,t weighted by share of deposits in total credit four quarters earlier).
    - Non-resident liability growth × Shforeignlia_i,t-4 (Non-resident Liability Growth_it weighted by share of liabilities to non-residents in total credit four quarters earlier).
    - Inflation (π_it).
    - Lagged GDP growth (G_i,t-1).
    - Lagged deposit rate (Deposit rate_i,t-1) as proxy for monetary stance.
    - Change in US federal fund rate (Fed Fund Rate Change_i,t).
  - Country fixed effects included; time fixed effects not included in the benchmark.
  - Growth rates and changes are 4-quarter growth rates or 4-quarter differences unless specified otherwise.
  - Observations: 1084; Adjusted R-sq: 0.637.

- Estimated coefficients (Table 2, first column; standard errors in parentheses):
  - Deposit growth × deposit/credit: 0.508*** (0.0531).
  - Non-res liability growth × non-res liab/credit: 0.506*** (0.121).
  - Inflation: 0.367* (0.189).
  - Lagged GDP growth: 0.763*** (0.281).
  - Lagged deposit rate: -0.414** (0.159).
  - Change in fed fund rate: -0.548* (0.285).
  - Constant: 5.484*** (1.317).

### Main interpretations from the benchmark
- Symmetry of funding channels:
  - Domestic deposits and foreign liabilities have remarkably similar point estimates (~0.5), suggesting banks treat increases in these funding sources similarly when extending credit.
- Inflation and real credit:
  - Inflation has a positive coefficient (0.367*) on nominal credit growth, but the coefficient is less than 1, indicating inflation is detrimental to real credit growth.
- Real activity and credit demand:
  - Lagged GDP growth is strongly positive (0.763***), indicating stronger economic growth raises demand for credit and leads to higher credit growth.
- Monetary conditions:
  - Higher lagged deposit rates (tighter domestic monetary conditions) reduce credit growth (-0.414**).
  - Looser US monetary policy (decline in Fed funds rate) is associated with higher credit growth in EMEs (change in fed fund rate coefficient -0.548* implies an easing fed funds rate contributes positively to EME credit growth).

### Robustness checks and alternative specifications (summary)
- Controls and alternative measures:
  - Exchange rate changes: change in exchange rate positive and significant in specifications (e.g., 0.133*** (0.0409)), consistent with valuation effects from foreign-currency loans.
  - Initial credit-to-GDP ratio: negative and significant in some specifications (e.g., -0.361*** (0.0968)), consistent with faster credit growth in less financially deep countries.
  - Initial non-performing loan (NPL) ratio: negative and significant (e.g., -1.035*** (0.372)), indicating less healthy banking sectors extend less credit.
  - US M2 growth used as alternative global liquidity measure: results broadly similar.
  - Time dummies: inclusion (Table 4) leaves main findings intact and often raises adjusted R-sq.
- Pre-crisis sample (2002Q1-2007Q4; Appendix):
  - Results broadly intact when estimated on the pre-crisis period alone.

### Selected numerical ranges across alternative specifications
- Deposit growth coefficients: ranged from 0.475*** to 0.583*** across specifications.
- Non-res liability growth coefficients: ranged from 0.394** to 0.510***.
- Lagged GDP growth: consistently positive and significant, e.g., from 0.678*** to 0.889*** in alternative specifications.
- Lagged deposit rate: negative and often significant; example values include -0.953*** in one specification.
- Change in fed fund rate: negative and sometimes significant; example -0.599* in one specification.
- Pre-crisis benchmark coefficients (2002Q1-2007Q4, Appendix examples):
  - Deposit growth × deposit/credit: 0.439*** (0.0807); 0.444*** (0.0765) in alternative pre-crisis specifications.
  - Non-res liab growth × non-res liab/credit: 0.551*** (0.125); 0.522*** (0.124).
  - Lagged GDP growth: 1.121*** (0.367); 1.119*** (0.385).
  - Change in exchange rate: 0.139** (0.0551); 0.151** (0.0557).
  - Lagged credit-to-GDP ratio: examples -0.134* (0.0745); -0.318*** (0.117).
  - Adjusted R-sq range across pre-crisis specifications: e.g., 0.475 to 0.789 (without time dummies) and 0.789 to 0.917 (with time dummies).
  - Observations in many pre-crisis regressions: commonly 756; some specifications fewer (e.g., 270 when NPL or other variables used).

### Implications emphasized by the authors (linked to regression evidence)
- Role of foreign capital:
  - Countries heavily reliant on foreign borrowing experienced the largest swings pre- and post-crisis, implying vulnerability of banking sectors dependent on foreign capital.
  - Macro-prudential vigilance is warranted regarding foreign-capital-fueled credit booms that can reverse quickly.
- Importance of domestic deposits:
  - A robust domestic deposit base supports sustained and stable credit growth; jurisdictions with stable deposit growth experienced less deceleration during the crisis.
- Macroeconomic fundamentals:
  - Strong growth and low inflation support credit growth; policies improving fundamentals and lowering inflation facilitate credit expansion.
- Banking sector health:
  - Better-capitalized and healthier banks extend more credit; higher NPL ratios are associated with lower credit growth.

### Caveats and limitations
- Larger residuals observed in many cases during the post-crisis period.
- Significant heterogeneities across countries; regressions capture average sample behavior.
- Lack of direct data on cross-border lending acknowledged; cross-border lending is not directly accounted for in the estimations.
- Estimation results should be treated with caution given sample and data limitations.

*Source: _wp1151 - 2.  Regression Results Under Benchmark Specification (page 9) — IMF staff estimates and authors' calculations as presented in the supplied content.*

### 2.  Regression Results Under Benchmark Specification ....................................................9

### 2.  Regression Results Under Benchmark Specification ....................................................9

### Overview
- Section title: "Regression Results Under Benchmark Specification"
- Location in document: page 9

### Neighboring sections (document structure)
- 3. Regression Results Under Alternative Specifications .................................................11
- 4. Regression Results Under Alternative Specifications with Time Dummies ...............12

### Figures cited elsewhere in the document (for context)
- Figure 1. Average Credit Growth Before and After the Crisis .....................................................4
- Figure 2. Decomposition of Credit Growth in EU EMEs ...........................................................13
- Figure 3. Decomposition of Credit Growth in other European EMEs ........................................14
- Figure 4. Decomposition of Credit Growth in Middle East and Africa EMEs ...........................14
- Figure 5. Decomposition of Credit Growth in Asia EMEs .........................................................15
- Figure 6. Decomposition of Credit Growth in Central America EMEs ......................................16
- Figure 7. Decomposition of Credit Growth in South America EMEs ........................................16

### Appendix listing
- Pre-Crisis Sample Regressions ................................................................................................18

*Source: _wp1151 - 2.  Regression Results Under Benchmark Specification ....................................................9*

### References .............................................................................................................

### _wp1151 - References .............................................................................................................

### I. INTRODUCTION
- Research questions:
  - What determines credit growth in EMEs?
  - What were the main drivers of pre-crisis credit booms and post-crisis busts?
  - Are booms and busts caused by the same underlying factors?
  - Why were there regional differences in credit growth before and after the crisis?
- Main approach and focus:
  - Investigation of a large panel of EMEs over the last decade, covering pre-crisis and post-crisis periods.
  - Identification of demand and supply factors affecting credit growth, with emphasis on the supply side.
- Key summary findings (from introduction):
  - Domestic deposit growth and non-resident liability growth contribute positively and symmetrically to credit growth.
  - Stronger economic growth increases demand for credit and leads to higher credit growth.
  - High inflation increases nominal credit but is detrimental to real credit growth.
  - Loose monetary conditions, domestic or global, result in more credit.
  - A healthy banking sector tends to extend more credit than an unhealthy one.
  - Results are robust under various specifications and sample periods.
  - The same set of factors helps explain both time-series and cross-sectional variations in credit growth.

### II. STYLIZED FACTS AND LITERATURE REVIEW
- Sample-wide credit growth statistics:
  - Pre-crisis period: bank credit grew on average at around 24 percent; maximum 59 percent; minimum 6 percent.
  - Post-crisis period: average annual growth 8 percent; maximum 32 percent; minimum -3 percent.
- Regional and country highlights:
  - EU EMEs: pre-crisis average 33 percent; post-crisis average 4 percent; Baltic States registered declines.
  - Other European EMEs: pre-crisis average 36 percent; post-crisis slowed to 10 percent.
  - Middle East and Africa EMEs (sample of 4 countries): pre-crisis average 14 percent; sharp post-crisis slowdown in Egypt, Jordan, South Africa; Morocco nearly stable.
  - Asian EMEs: mixed outcomes — Indonesia, Korea, Thailand slowed; China, Malaysia, Philippines accelerated; Vietnam post-crisis ~30 percent.
  - Central America EMEs: pre-crisis average 17 percent; post-crisis slowed to 5 percent.
  - South America EMEs: pre-crisis average 21 percent; post-crisis average 10 percent.
- Literature synthesis (selected findings):
  - Takáts (2010): supply shock main determinant of slowdown in cross-border lending to emerging markets during the crisis.
  - Bakker and Gulde (2010): external factors (bad luck) main reason for credit booms and busts in new EU members.
  - Aisen and Franken (2010): pre-crisis boom and slowdown in partner countries were main determinants of credit growth during the crisis.
  - Kamil and Rai (2010): sources of funding (external vs. internal) mattered; externally reliant countries suffered more.
  - Barajas, Chami, Espinoza, and Hesse (2010): bank-level fundamentals—capitalization and loan quality—helped explain differences in credit growth across MENA countries.
- Contribution of this paper:
  - Broad coverage of EMEs where data are available; sample period 2002Q1–2010Q1.
  - Simultaneous consideration of demand and supply factors, with emphasis on supply side.
  - Use of a longer period to capture full boom-bust cycle.

### III. DATA AND METHODOLOGY
- Data scope:
  - Quarterly data from IMF databases: International Financial Statistics (IFS), World Economic Outlook (WEO), and Global Financial Stability Report (GFSR).
  - Period: first quarter of 2001 to the second quarter of 2010.
  - Countries covered (38): Argentina, Brazil, Bulgaria, Chile, China, Colombia, Costa Rica, Croatia, Czech Republic, Egypt, El Salvador, Estonia, Georgia, Guatemala, Hungary, Indonesia, Israel, Jamaica, Jordan, Korea, Latvia, Lithuania, Malaysia, Mexico, Morocco, Panama, Peru, Philippines, Poland, Romania, Russia, Serbia, South Africa, Thailand, Turkey, Ukraine, Venezuela, and Vietnam.
  - Note: Czech Republic and Korea included though reclassified later; Israel grouped with European EMEs.
- Dependent and explanatory variables:
  - Dependent: banking sector private credit (Credit Growth_i,t).
  - Key explanatory variables:
    - Growth rate of deposits (Deposit Growth_i,t), weighted by Shdepo_i,t-4 (share of deposits in total credit to private sector four quarters ago).
    - Growth rate of non-resident liabilities (Non-resident Liability Growth_it), weighted by Shforeignlia_i,t-4 (share of liabilities to non-residents in total credit to private sector four quarters ago).
    - Inflation (π_it).
    - Lagged GDP growth (G_i,t-1).
    - Lagged deposit rate (Deposit rate_i,t-1) as proxy for monetary stance.
    - Change in US federal fund rate (Fed Fund Rate Change_i,t).
    - Additional controls used in alternative specifications: exchange rate, US M2, non-performing loan ratio, initial credit-to-GDP ratio.
- Model specifics:
  - Benchmark regression broadly follows literature; controls for country fixed effects.
  - Unless otherwise specified, growth rates and changes are 4-quarter growth rates or 4-quarter differences.
  - Panel is unbalanced due to data availability gaps.
  - Acknowledged limitation: lack of data on cross-border lending means cross-border lending not directly accounted for.

### IV. ESTIMATION RESULTS AND ROBUSTNESS CHECKS
- Benchmark results (full sample 2002Q1-2010Q1; Table 2 first column):
  - Deposit growth × deposit/credit: 0.508*** (standard error 0.0531).
  - Non-res liability growth × non-res liab/credit: 0.506*** (0.121).
  - Inflation: 0.367* (0.189).
  - Lagged GDP growth: 0.763*** (0.281).
  - Lagged deposit rate: -0.414** (0.159).
  - Change in fed fund rate: -0.548* (0.285).
  - Constant: 5.484*** (1.317).
  - Country fixed effects included; time fixed effects not included.
  - Observations: 1084; Adjusted R-sq: 0.637.
- Key interpretations from benchmark:
  - Domestic deposits and foreign liabilities have remarkably similar point estimates (~0.5), suggesting symmetric treatment by banks.
  - Inflation increases nominal credit but coefficient less than 1 indicates inflation dampens real private credit growth.
  - Higher GDP growth predicts higher credit growth.
  - Higher deposit rate (tighter monetary conditions) reduces credit growth.
  - Looser US monetary policy (lower fed funds rate) associated with higher credit growth in EMEs.
- Robustness checks and alternative specifications (Table 3 and Table 4):
  - Controlling for exchange rate depreciation/appreciation: change in exchange rate positive and significant (e.g., 0.133*** (0.0409) in one specification), indicating valuation effects from foreign-currency loans.
  - Controlling for initial credit-to-GDP ratio: negative sign (e.g., -0.361*** (0.0968) in one specification) consistent with faster growth in less financially deep countries.
  - Controlling for initial NPL ratio: negative and significant (e.g., -1.035*** (0.372)), indicating less healthy banking sectors extend less credit.
  - Using US M2 growth as alternative global liquidity measure: results broadly similar.
  - Use of time dummies (Table 4) leaves main findings intact; adjusted R-sq values often higher when time dummies included.
  - Pre-crisis sample regressions (2002Q1-2007Q4; Appendix Tables) show results broadly intact when using only pre-crisis data.
- Selected numerical results from alternative specifications (Table 3 example):
  - Deposit growth coefficients ranged: 0.475*** to 0.583*** across specifications.
  - Non-res liab growth coefficients ranged: 0.394** to 0.510***.
  - Lagged GDP growth consistently positive and significant (e.g., 0.678*** to 0.889***).
  - Lagged deposit rate negative and often significant (e.g., -0.953*** in one specification).
  - Change in fed fund rate negative and sometimes significant (e.g., -0.599*).

### V. FINDINGS AND POLICY IMPLICATIONS
- Country- and region-level decompositions (using full-sample estimations) yield the following patterns:
  - EU EMEs:
    - Pre-crisis: Romania, Latvia, Bulgaria, Lithuania, Estonia >30 percent credit expansion; Hungary, Poland, Czech Republic moderate.
    - Heavy reliance on foreign funding in many countries (e.g., Latvia, Lithuania, Estonia, Hungary); foreign borrowing contribution sometimes larger than domestic deposits.
    - Post-crisis: retrenchment of foreign funding widespread; countries relying heavily on foreign funding experienced sharpest declines; post-crisis economic slowdown also important.
  - Other European EMEs:
    - Pre-crisis: almost all double-digit growth; Ukraine close to 60 percent.
    - Foreign borrowing a major contributor, but domestic deposits contribution larger than in EU EMEs.
    - Post-crisis: substantial slowdown with retrenchment in both domestic deposits and foreign borrowing; economic decline notably important in Ukraine, Russia, Croatia, Georgia.
  - Middle East and Africa EMEs (4 countries):
    - Pre-crisis: solid expansion driven by domestic deposits and economic growth.
    - Post-crisis: slowdown in all except Morocco; decline largely due to economic slowdown and weaker domestic deposits; Jordan and South Africa also saw declines in foreign borrowings.
  - Asia EMEs:
    - Pre-crisis: wide variance; Philippines, Thailand, Malaysia, Korea moderate; Vietnam, Indonesia, China 15 to 33 percent.
    - Post-crisis: China, Malaysia, Philippines accelerated; Vietnam maintained ~30 percent. Major determinants: domestic deposits and economic growth; foreign borrowing played role in Korea, Malaysia, Philippines.
  - Central American EMEs:
    - Pre-crisis: strong expansion (Jamaica and Costa Rica ~30 percent; Panama <10 percent).
    - Pre-crisis drivers: domestic deposits and economic growth; Jamaica and Panama relied more on foreign borrowings.
    - Post-crisis (data for 3 of 5): decline in domestic deposit contribution; foreign borrowing contribution turned negative in Costa Rica and Jamaica but increased in Mexico; economic activity provided negative contribution.
  - South America EMEs:
    - Pre-crisis: moderate credit growth except Venezuela.
    - Determinants: domestic deposits and economic activity; Argentina had negative contribution from foreign borrowing pre-crisis.
    - Post-crisis: moderate slowdown in most countries except Peru; slowdown driven by domestic deposits and economic activity.
- Policy lessons (enumerated):
  - Foreign capital is a mixed blessing:
    - Countries heavily reliant on foreign borrowing (notably some European EMEs) experienced the largest swings pre- and post-crisis.
    - Banking sectors dependent on foreign capital are vulnerable to external shocks and boom-bust cycles.
    - Macro-prudential policies should be vigilant about foreign-capital-fueled credit booms that can reverse quickly.
  - Building a robust domestic deposit base is key to sustained and stable credit growth:
    - Countries with robust/stable domestic deposit growth experienced little or no deceleration of credit during the crisis.
  - Strong growth and low inflation support credit growth:
    - Policies that improve fundamentals and lower inflation boost credit growth and economic activity.
  - Banking sector health matters:
    - A banking sector with a healthy balance sheet supports financial stability and credit growth.
- Cautions and caveats:
  - Larger residuals in many cases during the post-crisis period.
  - Significant heterogeneities across countries; regressions describe average behavior in the sample.
  - Estimation results should be treated with caution.

### Appendix: Pre-Crisis Sample Regressions (selected numerical highlights)
- Pre-crisis benchmark coefficients (2002Q1-2007Q4; Table 1 / Appendix):
  - Deposit growth × deposit/credit: 0.439*** (0.0807) in one specification; 0.444*** (0.0765) in another.
  - Non-res liab growth × non-res liab/credit: 0.551*** (0.125) in one specification; 0.522*** (0.124) in another.
  - Lagged GDP growth: 1.121*** (0.367) in one specification; 1.119*** (0.385) in another.
  - Lagged deposit rate: -0.380 (0.245) in one specification; -0.336 (0.288) in another.
  - Change in exchange rate: 0.139** (0.0551) in one specification; 0.151** (0.0557) in another.
  - Lagged credit-to-GDP ratio: negative in several specifications (e.g., -0.134* (0.0745); -0.318*** (0.117)).
  - Adjusted R-sq range across specifications: e.g., 0.475 to 0.789 (without time dummies) and 0.789 to 0.917 (with time dummies).
- Observations in pre-crisis regressions: commonly 756; some specifications with fewer observations (e.g., 270 when NPL or other variables used).

*Source: IMF staff estimates and authors' calculations as presented in the supplied content.*

### REFERENCES

### _wp1151 - REFERENCES

### References

- Aisen, Ari and Michael Franken, 2010, “Bank Credit During the 2008 Financial Crisis: A Cross-Country Comparison,” IMF Working Paper 10/47 (Washington: International Monetary Fund).

- Bakker, Bas and Anne-Marie Gulde, 2010, “The Credit Boom in the EU New  
MemberStates: Bad Luck or Bad Policies?” IMF Working Paper 10/130 (Washington: International Monetary Fund).

- Barajas, Adolfo, Ralph Chami, Raphael Espinoza, and Heiko Hesse, 2010, “Recent Credit Stagnation in the MENA Region: What to Expect? What Can Be Done?” IMF Working Paper 10/219 (Washington: International Monetary Fund).

- Barajas, Adolfo, Giovanni Dell’Ariccia, and Andrei Levchenko, 2007, “Credit  
Booms: The Good, the Bad, and the Ugly,” (unpublished: International Monetary Fund).

- International Monetary Fund, 2009, “Will International Banks Transmit the                             
Global Credit Crunch to Latin American and Caribbean Countries?” in Regional 
Economic Outlook: Western Hemisphere, April 2009 (Washington: International 
Monetary Fund).

- Kamil, Herman and Kulwant Rai, 2010, “The Global Credit Crunch and Foreign Banks’ 
Lending to Emerging Markets: Why Did Latin America Fare Better?” IMF Working 
Paper 10/102 (Washington: International Monetary Fund).

- McGuire, Patrick, and Nikola Tarashev, 2008, “Bank Health and Lending to Emerging 
Markets,” BIS Quarterly Review (Basel: Bank for International Settlements, 
December).

- Rosenberg, Christoph B., and Marcel Tirpák, 2008, “Determinants of Foreign      
Currency Borrowing in the New Member States of the EU,” IMF Working Paper 
08/173(Washington: International Monetary Fund).

- Takáts, Előd, 2010, “Was it credit supply? Cross-border bank lending to emerging market 
economies during the financial crisis,” BIS Quarterly Review (Basel: Bank for 
International Settlements, June).

*_wp1151 - REFERENCES_*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1151.pdf_