## wp1813

## Source details

**Canonical URL:** [wp1813](https://www.imf.org/-/media/files/publications/wp/2018/wp1813.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp1813.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp1813.pdf.json)

---

### 1. Introduction — motivation, laboratory, and context
- Motivation and research question:
  - Banking crises often follow periods of strong bank credit growth financed with foreign liquidity and are accompanied by credit crunches with significant real effects.
  - Research gap: lack of systematic evidence on whether macroprudential policies smooth local household credit cycles or mitigate spillovers from the global financial cycle.
  - Central question: How do macroprudential policies affect local bank credit to households and how do they interact with international factors such as FX lending, foreign bank funding, and global risk appetite?
- Laboratory and data:
  - Country and period: Romania, 2004-2012, quarterly frequency.
  - Comprehensive household credit register: close to 3,000,000 individual loans from 42 commercial banks.
  - Loan-level characteristics: loan volume and rate, loan-to-value ratio (LTV), debt-service-to-income ratio (DTI), etc.; matched with supervisory bank balance-sheet data.
- Institutional and macroeconomic context (selected facts):
  - Average GDP growth of 7.3% between 2004 and 2008.
  - Total bank credit grew at an average of 23% in real terms during 2004-2008.
  - Credit-to-GDP ratio reached 40% of GDP in 2008.
  - Real GDP fell by 7.8% during 2009-2010 and averaged 1.5% in 2011-2012.
  - Post-crisis bank loan claims at end-2012 were only four-fifths of their pre-crisis peak.
  - Sample: 42 banks (30 private commercial banks, 2 state-owned and development banks, 10 branches of foreign banks); largest five banks account for almost 80% of total banking sector assets.
  - Average capital ratios (Tier 1 percent of total assets): 7.4%.
  - Foreign bank assets account for 79.2% of total banking sector assets over 2004-2012 (54% in 2004 peaking at 89% at end of the boom).
  - Reliance on foreign funding at almost 20% of total assets.
  - Household credit composition: household credit claims about half of total private credit; between 2005 and mid-2008 household debt increased at an average annual rate of 77%.
  - Loan composition: 10% residential mortgages and almost 90% consumer loans; consumer loans 60% of total credit volume and mortgages 40%.
  - Mortgages currency composition: 81% EUR, 7% CHF; about 20% of consumer loans in forex (mainly EUR).
  - In credit register, 4% of loans originated during 2004-2012 were restructured or rescheduled and 9.7% were non-performing.

### 2. Macroprudential policies — adoption, index construction, and data
- Policies deployed by the National Bank of Romania (NBR) during 2004-2012:
  - Changes in minimum reserve requirements on local and foreign currency deposits (including foreign bank funding).
  - Limits on credit exposures in foreign currencies.
  - Changes in capital requirements and provisioning rules.
  - Time-varying ceilings on LTV and DTI ratios for household loans.
- Index construction (Macroprudential Policy index — MPP):
  - Each introduction or change coded as tightening (+1) or loosening (−1); simultaneous multiple measures coded as +2 or +3.
  - MPP defined as the running (cumulative) sum of these values starting 2004:Q1; higher values indicate a tightening.
  - Sub-indices:
    - Bank-based MPP: quantitative restrictions on institutions’ balance sheets (reserve requirements on FX liabilities, ceilings on FX credit exposures to unhedged borrowers, provisioning and loan classification rules).
    - Borrower-based MPP: instruments aimed at reducing household indebtedness risk (time-varying LTV and DTI limits).
- MPP summary statistics and properties:
  - Index range: between 0 and 12.
  - Mean: 5.943.
  - Standard deviation: 3.581.
  - Maximum attained in 2006:Q4.
  - Minimum attained in 2011:Q2-Q3.
  - Bank-based MPP: Mean 2.333; Median 1.500; St. Dev. 3.593.
  - Borrower-based MPP: Mean 2.25; Median 2.000; St. Dev. 1.052.
- Data and microdata:
  - Loan-level dataset: Total individual loans in final dataset over 2004-2012: 2,965,479 (Consumer loans: 2,654,962; Mortgages: 310,517).
  - Foreign currency loans: 996,240 (868,745 in EUR, 118,007 in CHF, remainder OTHER).
  - Unique individual borrowers: 1,437,059 (282,364 mortgage borrowers).
  - Loans extended by 42 banks.
  - Average loan amounts: Mortgages ~ USD 44,000; Consumer loans ~ USD 11,000.
  - Credit register observes loans larger than RON 20,000.
  - Loan and bank variables winsorized at the 1% level.
- Macro variables used:
  - European VIX: Euro Stoxx 50 Volatility Index (lower VIX = lower uncertainty and higher investor risk appetite).
  - EONIA: Euro OverNight Index Average.
  - Domestic controls: NBR 7-day repo rate, Real GDP growth (seasonally-adjusted), CPI inflation.

### 3. Empirical identification and strategy
- Empirical challenges:
  - Macroprudential policy responds to local financial/economic conditions (endogeneity).
- Identification strategy:
  - Exploit cross-sectional heterogeneity across loans, borrowers, and banks using microdata; emphasis on double and triple interactions:
    - Loan characteristics: FX vs local currency.
    - Borrower characteristics: more vs less risky (DTI at origination).
    - Bank characteristics: more vs less dependent on foreign funding.
  - Controls for domestic and global macroeconomic and financial variables and their interactions.
  - Fixed effects: bank×period and borrower county×period; loan-type×period fixed effects.
  - Main estimator: OLS with standard errors clustered on bank; macro and bank-level variables averaged over past two quarters in baseline.

### 4. Main empirical findings — aggregate effects and composition
- Aggregate and composition effects:
  - Tighter macroprudential conditions associated with a decline in household credit volume, especially for foreign (vs. local) currency loans.
  - Estimates statistically and economically significant for all FX loans, with similar magnitudes across EUR and CHF loans.
  - A tightening of macroprudential policy by one standard deviation reduces average foreign currency loan volume by 12% more than it does local currency loan volume (reported elsewhere as 3.581 units × (−0.0500) = −0.1224 or 17.9% in specific calculation contexts; see borrower-risk results for decomposition).
- Key coefficients (Table 2 highlights):
  - Macroprudential policy (column 1): -0.0531*** (0.018).
  - Macroprudential policy × FX (column 2): -0.0500*** (0.018).
  - FX loan coefficients across columns: 1.6617*** (0.109); 2.1067*** (0.241); 1.9483*** (0.211); 2.5370*** (0.250).
  - Observations: 2,753,494 (col 1); 2,965,459 (cols 2-4). R-squared up to 0.2630.

### 5. Differential effects — borrower riskiness and substitution
- Borrower riskiness (DTI) and currency composition (Table 3):
  - MPP×DTI (col 1): -0.0002 (0.005) (insignificant).
  - Spline interaction results (col 2):
    - MPP×DTI×FX: -0.0323*** (0.007).
    - MPP×DTI×RON: 0.0177*** (0.004).
  - Interpretation: tightening MPP shifts credit composition away from riskier FX lending toward less risky local-currency lending for borrowers with the same DTI.
  - Quantified effect (using column 2 estimates): a tightening of MPP by one SD (3.581 units) reduces FX loan growth by 11.6% and increases RON loan growth by 6.3% on average, for a differential of 17.9 percentage points.
    - Calculations reported: 3.581×(−0.0323) = 11.6% for FX loans and 3.581×(−0.0177) = 6.3% for RON loans (differential 17.9%).
  - Robustness: no effects for mortgage lending using LTV; no effects using income as borrower risk measure.

### 6. Bank heterogeneity — exposure to foreign funding
- Foreign funding facts:
  - Foreign funding measure: share of non-resident foreign currency deposits in banks’ total assets, with period-specific maturity definitions.
  - Almost 90% of banks’ foreign funding comes from parent banks; about 5% from international development banks.
  - During 2009-2016 about 43% of non-resident deposits were short term (<2 years); >70% in EUR and close to 20% in RON.
  - Foreign banks: average foreign funding ratio 25% vs. 9% for domestic banks.
- Triple interaction MPP×FX×ForeignFunding (Table 4):
  - Column 2 coefficient on MPP×FX×ForeignFunding: −0.025.
  - Reported calculation: (−0.025×34.64)−(−0.025×2.01) = 8.15% — the dampening effect of MPP on FX loan volume is larger by 8.2 percentage points for a bank at the 10th percentile (2.01%) vs. the 90th percentile (34.64%) of foreign funding exposure.
  - Standardized coefficients indicate bank-based MPP has a dampening effect on FX credit about 5 times larger than borrower-based MPP for a given level of foreign funding reliance (standardized coefficients −0.0012 and −0.0061 using SDs: bank-based MPP SD = 3.593, borrower-based MPP SD = 1.052).

### 7. Spillovers — global risk appetite (VIX) and foreign monetary policy (EONIA)
- Spillovers from global uncertainty and risk appetite (Table 5):
  - With MPP=0, a reduction of the VIX by one unit raises household credit by 6.5%; this effect is undone by an increase in MPP by 3.7 units (reported calculation 0.0655/0.0177 = 3.7 units; SD of MPP index is 3.581).
  - Macroprudential policy × VIX (col 2): 0.0177*** (0.004).
  - Macroprudential policy × VIX × FX (col 3): 0.0189*** (0.005); Macroprudential policy × VIX × RON (col 3): 0.0174*** (0.005).
  - Four-term spline results: MPP reduces FX loan growth relatively more during low VIX (high risk appetite) and also reduces FX loan growth during high VIX; coefficients on MPP×LowVIX×FX are statistically larger in absolute value than other spline terms.
  - Loan-type decomposition (Table A7): when risk appetite is high (VIX low), elasticity of loan growth w.r.t. MPP is larger for consumer loans than for mortgages; triple interaction MPP×LowVIX×FX larger for consumer loans.
- Spillovers from foreign monetary policy (Table 6):
  - EONIA×FX interactions indicate periods of low Eurozone policy rates associated with higher FX lending.
  - Triple interaction EONIA×MPP×FX is positive and statistically significant, indicating MPP reduces the spillover from foreign monetary conditions to FX household credit.
  - Columns show heterogeneous effects with borrower-based measures often stronger in these spillover mitigations.

### 8. Disaggregated instrument effects and policy implications
- Disaggregated instrument effects:
  - Borrower-side measures generally more effective at dampening risky lending decisions than bank-side measures in many specifications.
  - Bank-side measures are relatively more effective for banks with greater reliance on foreign funding and in mitigating spillovers from foreign funding exposure.
  - Currency-specific effects (Table A5): similar coefficient magnitudes for EUR and CHF loans; larger magnitudes for OTHER currencies (USD, GBP, JPY) though OTHER make up less than 0.5% of sample.
- Policy implications:
  - Macroprudential instruments can materially dampen risky forms of lending and mitigate international spillovers on local household credit growth by shifting lending composition away from FX and toward local-currency loans.
  - Effectiveness of MPP is stronger in tranquil periods of low VIX (economic booms) and when foreign monetary policy rates are low, suggesting greater potency of macroprudential policy during booms.

### 9. Key descriptive statistics (selected exact figures)
- Dataset counts and shares:
  - Total individual loans: 2,965,479.
  - Consumer loans: 2,654,962.
  - Mortgages: 310,517.
  - Foreign currency loans (FX): 996,240 (868,745 EUR; 118,007 CHF; OTHER 0.5% total).
  - Unique individual borrowers: 1,437,059 (282,364 mortgage borrowers).
- Loan-level moments (Table 1):
  - Loan amount (RON): Obs 2,965,479; Mean 68,173; Median 37,216; St. Dev. 208,357.
  - Log(loan amount, RON): Obs 2,965,479; Mean 9.80; Median 10.53; St. Dev. 2.78.
  - Borrower age: Obs 2,965,479; Mean 39.17; Median 38.02; St. Dev. 10.01.
  - DTI: Obs 2,139,977; Mean 61.66; Median 42.65; St. Dev. 56.49 (trimmed at max 300%).
  - Foreign currency loan (FX) dummy: Mean 0.336; St. Dev. 0.472.
  - Mortgage (MGG) dummy: Mean 0.104; Consumer loan (CONS) dummy: Mean 0.896.
- Macro moments (Table 1):
  - MPP Overall: Mean 5.943; Median 7.000; St. Dev. 3.581.
  - Monetary policy rate (domestic): Mean 8.136; St. Dev. 2.373.
  - GDP growth: Mean 4.367; St. Dev. 4.998.
  - Inflation: Mean 6.253; St. Dev. 2.071.
  - VIX: Mean 33.623; St. Dev. 9.082.
  - EONIA: Mean 2.389; St. Dev. 1.542.
- Bank moments (Table 1):
  - Bank size (log assets): Mean 23.56; St. Dev. 1.077.
  - Bank capital (Tier 1 % assets): Mean 7.472; St. Dev. 3.272.
  - Bank liquidity: Mean 2.584; St. Dev. 1.847.
  - Bank ROA: Mean 0.992; St. Dev. 1.836.
  - Bank NPL: Mean 3.263; St. Dev. 4.500.
  - Bank risk profile (RWA/assets): Mean 65.106; St. Dev. 10.38.
  - Bank foreign funding (percent of assets): Mean 18.891; Median 15.252; St. Dev. 24.98.
  - Foreign bank dummy: Mean 0.812; St. Dev. 0.391.

_Italic source: wp1813 (IMF working paper chapter content)._

### 1. Introduction  3

### 1    Introduction

### Motivation and research question
- Banking crises often follow periods of strong bank credit growth financed with foreign liquidity (Schularick and Taylor, 2012; Gourinchas and Obstfeld, 2012; Reinhart and Rogoff, 2009) and are accompanied by credit crunches with significant real effects (Laeven and Valencia, 2013; Bernanke, 1983).
- Macroprudential policy is defined as instruments that aim to ensure financial stability by limiting cyclical vulnerabilities and systemic risk over the credit cycle.
- Research gap: No systematic evidence exists on the ability of macroprudential policies to smooth local household credit cycles or to mitigate spillovers from the global financial cycle on the local bank credit cycle.
- Central question: How do macroprudential policies affect local bank credit to households and how do they interact with international factors such as FX lending, foreign bank funding, and global risk appetite?

### Laboratory and data
- Country and period: Romania, 2004-2012, quarterly frequency.
- Data source and coverage:
  - Comprehensive household credit register containing the universe of bank loans to individuals in Romania.
  - Close to 3,000,000 individual loans—residential mortgages and consumer loans—from 42 commercial banks.
  - Loan-level characteristics include loan volume and rate, loan-to-value ratio (LTV), debt-service-to-income ratio (DTI), etc.
  - Matched with high-frequency supervisory information on bank balance sheets.
- Romania features relevant for study:
  - Bank-dependent, small open emerging market economy in the European Union.
  - Large portion of the banking sector is foreign-owned.
  - Substantial foreign bank funding.
  - About one third of household loans are extended in foreign currencies (mainly EUR and CHF).

### Institutional and macroeconomic context (selected facts)
- GDP and credit booms and busts:
  - Average GDP growth of 7.3% between 2004 and 2008.
  - Total bank credit, including in foreign currencies, grew at an average of 23% in real terms during 2004-2008.
  - Credit-to-GDP ratio reached 40% of GDP in 2008 (more than tripled in four years).
  - Real GDP fell by 7.8% during 2009-2010 and averaged 1.5% in 2011-2012.
  - Post-crisis bank loan claims, at end-2012, were only four-fifths of their pre-crisis peak level.
- Banking sector structure:
  - Sample comprises 42 banks: 30 private commercial banks, 2 state-owned and development banks, and 10 branches of foreign banks.
  - The largest five banks account for almost 80% of total banking sector assets.
  - Average capital ratios (Tier 1 capital in percent of total assets): 7.4%.
  - Foreign bank assets account for 79.2% of total banking sector assets over 2004-2012 (54% in 2004 peaking at 89% at end of the boom).
  - Reliance on foreign funding at almost 20% of total assets.
- Household credit composition:
  - Household credit claims account for about half of total private credit.
  - Between 2005 and mid-2008 household debt increased at an average annual rate of 77%.
  - Ten percent of individual loans are residential mortgages and almost 90% are consumer loans (including mortgage-backed consumer loans).
  - Consumer loans account for 60% of total credit volume and mortgages for 40%.
  - Mortgages are mostly denominated in foreign currency: 81% of mortgages in EUR, 7% in CHF; the rest in USD, GBP, and YEN.
  - About 20% of consumer loans are extended in forex (mainly EUR).
  - In the credit register, 4% of loans originated during 2004-2012 were restructured or rescheduled and 9.7% were non-performing.

### Macroprudential policy measures and index construction
- Policies deployed by the National Bank of Romania (NBR) during 2004-2012 included:
  - Changes in minimum reserve requirements on local and foreign currency deposits (including foreign bank funding).
  - Limits on credit exposures in foreign currencies.
  - Changes in capital requirements and provisioning rules.
  - Time-varying ceilings on loan-to-value (LTV) and debt-service-to-income (DTI) ratios for household loans.
- Index construction:
  - Each introduction or change in a macroprudential instrument is coded as a tightening (+1) or loosening (−1).
  - Following Cerutti et al. (2017), the macroprudential index is defined as the running (cumulative) sum of these values so each instrument is reflected in the index for the duration it is in place.
  - Higher values of the index indicate a tightening of macroprudential conditions.
  - Two additional fine-grained indices are defined to group instruments:
    - Bank-side measures (e.g., limits on leverage and provisioning rules).
    - Borrower-side measures (e.g., LTV and DTI limits).

### Identification and empirical strategy (summary)
- Empirical challenges:
  - Macroprudential policy responds to and is correlated with local financial and economic conditions.
- Strategy to obtain identification:
  - Exploit cross-sectional differences in loan, borrower, and bank characteristics using microdata from the NBR credit register.
  - Emphasize double and triple interactions of the macroprudential index with:
    - Loan characteristics (FX vs. local currency).
    - Borrower characteristics (more vs. less risky, measured with DTI at loan origination).
    - Bank characteristics (more vs. less dependent on foreign funding).
  - Control for local and international macroeconomic and financial variables, including interactions.
  - Use high-frequency panel with bank×period and borrower county×period fixed effects to control for time-varying unobserved local macroeconomic shocks and bank characteristics.
  - Include loan-type×period fixed effects to control for differences between mortgages and consumer loans.
  - Analyze macroprudential policy interactions with global financial cycle proxies (VIX, Eurozone monetary policy).

### Main empirical findings
- Aggregate and composition effects:
  - Tighter macroprudential conditions are associated with a decline in the volume of household credit, especially for foreign (vs. local) currency loans.
  - Estimates are statistically and economically significant for all FX loans, with similar magnitudes across EUR and CHF loans.
  - A tightening of macroprudential policy by one standard deviation reduces average foreign currency loan volume by 12% more than it does local currency loan volume.
- Risk and substitution effects:
  - Mitigating effects of tighter macroprudential policy on FX credit are stronger for ex-ante riskier borrowers (measured with DTI at loan origination).
  - For borrowers with the same level of risk, a tightening of macroprudential policies reduces FX credit relative to local currency credit, implying banks substitute away from riskier lending.
- Bank heterogeneity:
  - Macroprudential policy is more effective in reducing loan growth in (riskier) foreign currencies by banks that are relatively more exposed to foreign funding.
- Interaction with global financial conditions:
  - Macroprudential policy more potently dampens lending in (riskier) foreign currencies when global risk appetite is high (proxied by low European VIX) and when foreign monetary policy rates are low.
  - Largest quantitative effects occur in tranquil periods of low VIX—suggesting greater effectiveness of macroprudential policies during economic booms compared to busts.
- Disaggregated instrument effects:
  - Borrower-side measures are generally more effective at dampening risky lending decisions than bank-side measures.
  - Bank-side measures are relatively more effective for banks with greater reliance on foreign funding.

### Contribution to literature
- Novelty:
  - First paper to employ a household credit register to study the effectiveness of macroprudential policies in mitigating risky household lending over a full economic cycle and to document the ability of macroprudential policies to dampen effects of the global financial cycle on local credit cycles.
- Relation to existing studies:
  - Complements cross-country macroprudential literature by exploiting detailed microdata to analyze heterogeneity across loans, borrowers, and banks.
  - Differs from Jiménez et al. (2017) who find bank-side measures stronger during crises; this paper finds borrower-side measures larger during tranquil times (low VIX).
  - Extends literature on cross-border spillovers by assessing whether local macroprudential policies mitigate transmission of global financial conditions to household credit.

*Source: wp1813 - 1. Introduction*

### 2.2    Macroprudential Policies

### 2.2    Macroprudential Policies

### Adoption and evolution of macroprudential measures (2004–2012)
- The NBR adopted a wide range of macroprudential measures during the 2004-2012 boom-bust cycle.
- Early objectives (2004-2006): limit impact of strong capital inflows on domestic credit and address risks of FX lending to unhedged borrowers.
- Key measures and chronology (high-level summary from text):
  - 2004-2006: Gradually raised reserve requirements on foreign currency deposits and cut those on local currency deposits.
  - 2005: Introduced an outright limit on foreign currency credit exposures to unhedged individuals and firms (in percent of banks’ common equity).
  - 2005 onward: Imposed limits on loan-to-value (LTV) ratios for mortgages and debt-service-to-income (DTI) ratios for all loans; DTI ceiling later reduced and applied to borrowers’ total debt.
  - 2007 (EU accession): Reversed or harmonized several measures with EU regulations — banks allowed to set LTV and DTI ceilings based on internal models (subject to central bank guidelines), foreign currency credit limits removed, minimum capital requirements reduced from 12% to 8%, and Basel II enforced.
  - August 2008: Required banks to consider interest and exchange rate risk in DTI limits (using internal risk models) and raised provisioning requirements for FX credits to unhedged borrowers.
  - 2009 (crisis): Reserve requirements for all bank deposits were drastically reduced; government subsidy program launched for first-time home buyers exempting new mortgages from LTV limits.
  - 2011: NBR set new currency-specific LTV and DTI ceilings.

### Macroprudential Policy Index (MPP)
- Construction:
  - Follows Cerutti et al. (2017) approach: index manually coded from list of NBR measures (see Table A1 for instrument list and coding).
  - Each instrument coded as +1 for a tightening and −1 for a loosening during the period it is in place (starting quarter of introduction until quarter removed if within sample).
  - Simultaneous introduction of two or three measures coded as +2 or +3.
  - MPP is the running (cumulative) sum of this variable starting in 2004:Q1.
- Summary statistics and properties:
  - Index range: between 0 and 12.
  - Mean: 5.943.
  - Standard deviation: 3.581.
  - Maximum attained in 2006:Q4.
  - Minimum attained in 2011:Q2-Q3.
- Sub-indices:
  - Bank-based MPP: captures quantitative restrictions on financial institutions’ balance sheets (e.g., reserve requirements on FX liabilities, ceilings on FX credit exposures to unhedged borrowers, provisioning and loan classification rules).
  - Borrower-based MPP: captures instruments aimed at reducing household indebtedness risk (e.g., time-varying LTV and DTI limits).

### Data and empirical strategy (overview)
- Datasets:
  - Loan-level dataset (household credit register) with loan originations for entire banking system.
  - Bank-level supervisory dataset with balance sheet information.
  - Merged on unique bank identifier; quarterly frequency over 2004:Q1-2012:Q4.
  - Loan and bank variables winsorized at the 1% level.
- Additional data:
  - Domestic and global macroeconomic and financial conditions from IMF’s International Financial Statistics.
- Empirical focus:
  - Estimate impact of macroprudential policies on household credit.
  - Use double and triple interaction terms with demanding fixed effects to identify causal effects and mitigate endogeneity concerns.
  - Main econometric estimator: OLS with standard errors clustered on bank.
  - Macroeconomic and bank-level variables in baseline specifications averaged over the past two quarters.

### Credit register (microdata)
- Coverage and thresholds:
  - Central Credit Register maintained by NBR; regulatory reports recorded monthly.
  - Includes loans larger than RON 20,000 (approximately USD 4,500).
- Loan-level information observed:
  - Issuing bank, loan terms (amount, currency, maturity; interest rate not available for origination until 2015), LTV and DTI ratios, borrower location (one of 41 counties), borrower age, and loan status over time (non-performing, rescheduled, restructured).
- Sample counts and amounts:
  - Total individual loans in final dataset over 2004-2012: 2,965,479.
  - Consumer loans: 2,654,962.
  - Mortgages: 310,517.
  - Foreign currency loans: 996,240 (868,745 in EUR, 118,007 in CHF, remainder OTHER).
  - Unique individual borrowers: 1,437,059 (282,364 mortgage borrowers).
  - Loans extended by 42 banks.
  - Average loan amounts:
    - Mortgages: approximately USD 44,000.
    - Consumer loans: approximately USD 11,000.
- Note on interest rates:
  - Actual interest rate at loan origination available in credit register starting in 2015; paper back-calculates interest from repayment schedules and found no systematic loan pricing impact of macroprudential policies.

### Macro variables used
- Two proxies for global conditions:
  - European VIX: Euro Stoxx 50 Volatility Index (lower VIX = lower uncertainty and higher investor risk appetite).
  - EONIA: Euro OverNight Index Average (quarterly average of monthly values) as proxy for Eurozone interbank rate.
- Domestic controls:
  - NBR policy rate (7-day repo rate).
  - Real GDP growth rate (seasonally-adjusted).
  - CPI inflation.
- In some specifications, include these macro variables and interactions to control for local cycle and mitigate confounding.

### Empirical specifications (key equations and identification)
- Level effect specification (equation 1):
  - Lijkt = αit + ηkt + β MPPt−z + Xt−z δ0 + Zit−z δ1 + Wjkt δ2 + εijkt
  - Lijkt is log(loan volume) by bank i to borrower j in county k at time t; z denotes lag.
  - Controls: Xt−z (policy rate, GDP growth, CPI inflation, VIX), Zit−z (bank size, capital and liquidity ratios, ROA, NPL, risk profile, share of foreign funding, foreign-owned indicator), Wjkt (borrower age, FX loan indicator, first-home mortgage program indicator).
  - Fixed effects: loan-type×period among others.
  - β expected to be negative (tighter MPP → lower loan volumes).
- Differential effect specification (equation 2):
  - Lijkt = αit + ηkt + β MPPt−z × RISK + Zit−z δ1 + Wjkt δ2 + εijkt
  - RISK can be loan-, borrower-, or bank-level (e.g., FX loan dummy, bank reliance on foreign funding, DTI).
  - Coefficient on MPP×FX captures differential effect on FX loans vs local currency loans.
- Spillover and triple interaction specification (equation 3):
  - Lijkt = αit + ηkt + β MPPt−z × SPILLOV ERt−z × RISK + Zit−z δ1 + Wjkt δ2 + εijkt
  - SPILLOV ER is the VIX or EONIA; RISK is an FX loan dummy.
  - Triple interaction tests whether MPP mitigates risky lending when external conditions are loose (VIX low) or foreign policy rates are low.

### Identification and robustness strategies
- Address endogeneity concerns (reverse causality and omitted variables) by:
  - Exploiting heterogeneity across loans, borrowers, and banks with double and triple interactions.
  - Including demanding fixed effects: bank×period and borrower county×period (quarterly or semester depending on variation).
  - Including interactions of MPP with GDP growth where GDP growth is a robust determinant of MPP (Table A3 evidence).
  - Averaging macroeconomic and bank-level variables over past two quarters (baseline).
  - Results robust but weaker with one-quarter lag; similar to baseline with two-quarter lag.

### Results — overview and FX loans
- Level effect (equation 1, column 1 summary):
  - Negative and statistically significant link between MPP and bank credit: a tighter macroprudential policy stance is associated with lower loan volumes.
  - Note: raw data correlation between MPP and bank credit is positive (MPP tightens when credit picks up), implying potential attenuation bias from reverse causality.
- Differential impact on currency composition (equation 2, columns 2-3 summary):
  - Focus on interaction MPP×FX (FX = foreign currency loan dummy).
  - Specification includes county×year-quarter fixed effects and loan-type×year-quarter fixed effects to compare loans of same type across banks in same county and quarter.
  - Includes interactions of GDP growth with the FX dummy to avoid confounding with local cycle.
  - Estimates indicate that a tightening of macroprudential policy by one standard deviation (SD) ... [text cut off in source at this point; full magnitude and statistical details reported in Table 2 in source].

*Source: wp1813 - 2.2    Macroprudential Policies (IMF working paper chapter content).*

### 3.581 units reduces average foreign currency loan volume by close to 18% more than it does for

### wp1813 - 3.581 units reduces average foreign currency loan volume by close to 18% more than it does for

### Differential effects on foreign-currency (FX) vs. local-currency (RON) loan volumes
- Using the estimates in column 1: 3.581×(−0.0500)=−0.1224 or 17.9%.
- Column 3: the coefficient on MPP×FX for the bank index is statistically insignificant.
- Column 4: borrower-oriented measures are more effective in reducing risky FX lending than bank-oriented measures.
- Table A5: similar coefficient magnitudes for EUR and CHF loans; larger magnitudes for other currency loans (USD, GBP, YEN), though other-currency FX loans make up less than 0.5% of the sample.
- Result: macroprudential policies, especially borrower-based measures, reduce FX loan growth uniformly across major currencies such as EUR and CHF.

### Borrower riskiness (DTI, FX)
- Measure of ex-ante borrower creditworthiness: borrower DTI.
- Column 1 (Table 3): coefficient on MPP×DTI is negative but statistically insignificant.
- Column 2: spline of MPP×DTI with FX and RON loan dummies:
  - MPP×DTI×FX is negative and statistically significant.
  - MPP×DTI×RON is positive and statistically significant.
- Columns 3-4: similar pattern for bank- and borrower-based MPP indices.
- Interpretation: tightening MPP shifts credit composition away from riskier FX lending toward less risky local-currency lending for borrowers with the same DTI.
- Quantified effect (using column 2 estimates): a tightening of MPP by one SD (3.581 units) reduces FX loan growth by 11.6% and increases RON loan growth by 6.3% on average, for a differential of 17.9 percentage points.
  - Calculations: 3.581×(−0.0323) = 11.6% for FX loans and 3.581×(−0.0177) = 6.3% for RON loans.
- Robustness: no effects found for mortgage lending using LTV; no effects using income as borrower risk measure; results robust to concern that high-debt borrowers sort into FX loans.

### Bank exposure to foreign funding
- Foreign funding measure: share of non-resident foreign currency deposits in banks’ total assets, constructed to match policy targets over subperiods (short-term FX non-resident deposits during 2004-2005 with maturity <1 year; all FX non-resident deposits during 2005Q1-2009Q1; short-term FX non-resident deposits during 2009Q2-2012Q4 with maturity <2 years).
- Bank funding composition facts: almost 90% of banks’ foreign funding comes from parent banks; about 5% from international development banks; during 2009-2016 about 43% of non-resident deposits were short term (<2 years); over the same period, >70% of non-resident deposits were in EUR and close to 20% in RON.
- Foreign banks rely more heavily on foreign funding: average ratio is 25% for foreign banks vs. 9% for domestic banks.
- Table 4: triple interaction MPP×FX×ForeignFunding is negative and statistically significant.
  - Column 2 coefficient on MPP×FX×ForeignFunding: −0.025.
  - Using column 2: (−0.025×34.64)−(−0.025×2.01) = 8.15% — the dampening effect of MPP on FX loan volume is larger by 8.2 percentage points for a bank at the 10th percentile (2.01%) vs. the 90th percentile (34.64%) of foreign funding exposure.
- Columns 3-4: negative and statistically significant for both bank- and borrower-based indices; economic magnitudes indicate bank-based measures are quantitatively more impactful.
  - Standardized coefficients on MPP×FX×ForeignFunding in columns 3-4 are −0.0012 and −0.0061, respectively, using SDs: bank-based MPP SD = 3.593, borrower-based MPP SD = 1.052, and SD of dependent variable log(credit volume).
  - Conclusion: bank-based macroprudential policies have a dampening effect on FX credit about 5 times larger than borrower-based policies, for a given level of foreign funding reliance.
- Novel finding: higher effectiveness of bank-based instruments to reduce spillovers from banks’ access to foreign funding on the local credit cycle.

### Spillovers from global uncertainty and risk appetite (VIX)
- Focus variable: European VIX as proxy for global uncertainty and investor risk appetite.
- Table 5, column 1: with MPP=0, a reduction of the VIX by one unit (≈ one tenth of a SD) raises household credit by 6.5%; this effect is undone by an increase in MPP by 3.7 units (one SD).
  - Calculation: 0.0655/0.0177 = 3.7 units; SD of MPP index is 3.581.
- Columns 2-5: allowing differential effects for FX vs. local-currency loans:
  - Positive and statistically significant coefficients for MPP×VIX by loan currency; coefficients not statistically different from one another at the 1% level.
  - Four-term spline (breaking currency-specific MPP effects into high vs. low VIX periods) shows:
    - MPP reduces FX loan growth relatively more than local-currency loan growth during periods of low VIX (high risk appetite).
    - MPP reduces FX loan growth more than loan growth in any currency during periods of high VIX (low risk appetite).
  - One-sided t-tests: coefficients on MPP×LowVIX×FX are greater (in absolute value) than those on remaining spline terms for all MPP indices considered.
- Table A7 (loan-type decomposition): when risk appetite is high (VIX low), elasticity of loan growth w.r.t. overall MPP index and individual indices is larger for consumer loans than for mortgages; triple interaction MPP×LowVIX×FX coefficient is larger than other terms.

### Spillovers from foreign monetary policy (EONIA)
- Proxy for regional monetary conditions: EONIA, given high eurorization of Romanian economy.
- Table 6: EONIA×FX interactions indicate periods of low Eurozone policy rates are associated with higher FX lending, evidencing cross-border spillovers.
- Triple interaction EONIA×MPP×FX is positive and statistically significant, indicating MPP reduces the spillover from foreign monetary conditions to FX household credit.
- Specification controls for EONIA×GDP growth×FX to rule out confounding from domestic macroeconomic conditions.
- Columns 2-3: stronger effects for borrower-based measures.

### Conclusions
- Macroprudential policies materially affect domestic bank credit to households and mitigate spillovers from the global financial cycle.
- Key empirical findings:
  - Stronger dampening effects on foreign- vs. local-currency loans.
  - Stronger dampening effects on riskier vs. less risky borrowers (measured by DTI).
  - Tighter MPP reduces household credit extended by banks more reliant on foreign funding, especially FX-denominated credit.
  - MPP is more effective at taming FX credit growth when global risk appetite is high (low VIX) and when foreign monetary policy rates are low (EONIA low).
- Policy implication: macroprudential instruments can dampen risky forms of lending and mitigate international spillovers on local household credit growth by shifting lending composition away from FX and toward local-currency loans.

*Source: wp1813 (excerpt).*

### References

### References

### Cited literature
- Bibliographic entries include works on macroprudential policy, capital flows, banking globalization, monetary policy transmission, credit booms, household debt, and systemic risk. Selected authors cited: Acharya, Aiyar, Akinci, Auer, Ayyagari, Baskaya, Bernanke, Bhutta, Borio, Bruno, Camors, Cerutti, Cetorelli, Claessens, Coimbra, Crowe, De Haas, Dell’Ariccia, Diamond, Dimova, Duenwald, Everaert, Forbes, Freixas, Ghosh, Giannetti, Gourinchas, IMF (2009, 2010, 2013, 2014), IMF-FSB-BIS (2016), Jiménez, Kashyap, Keys, Laeven, Lim, Mian, Miranda-Agrippino, Morais, Neagu, Ongena, Reinhart, Rey, Schularick, Skimmyhorn, Vandenbussche, Vandenbussche et al.

### Figures — key notes
- Figure 1: Household bank credit growth (real, year-on-year) and macroprudential policy (MPP) index during 2004-2012. MPP index constructed by coding tightenings (+1) and loosenings (−1) and defined as the cumulative sum; higher values indicate a tightening. Household credit is deflated by the CPI 2005=100. Data source: National Bank of Romania.
- Figure 2: Plots of total bank credit by type (mortgages, consumer loans) and currency (RON, EUR, CHF, and other currencies) during 2004-2012. Data source: National Bank of Romania.

### Table 1 — Descriptive statistics (sample period 2004-2012)
- Observations and key sample moments (Obs, Mean, Median, St. Dev.) for loan-level and aggregate variables:
  - Loan amount (in local currency: RON): Obs 2,965,479; Mean 68,173; Median 37,216; St. Dev. 208,357
  - Log(loan amount, in local currency: RON): Obs 2,965,479; Mean 9.80; Median 10.53; St. Dev. 2.78
  - Borrower age (years): Obs 2,965,479; Mean 39.17; Median 38.02; St. Dev. 10.01
  - Debt-service-to-income ratio (DTI): Obs 2,139,977; Mean 61.66; Median 42.65; St. Dev. 56.49 (DTI trimmed at a maximum value of 300%)
  - First-home mortgage dummy: Obs 2,965,479; Mean 0.026; Median 0.000; St. Dev. 0.160
- Loan type shares (dummies, Obs 2,965,479):
  - Foreign currency loan (FX): Mean 0.336; Median 0.000; St. Dev. 0.472
  - Foreign currency loan in EUR: Mean 0.293; CHF: Mean 0.039; OTHER: Mean 0.004
  - Local currency loan (RON): Mean 0.664
  - Mortgage (MGG): Mean 0.104; Consumer loan (CONS): Mean 0.896
  - Foreign currency mortgage (MGG-FX): Mean 0.093; Local currency mortgage (MGG-RON): Mean 0.011
  - Foreign currency consumer loan (CONS-FX): Mean 0.243; Local currency consumer loan (CONS-RON): Mean 0.653
- Macro variables (Obs 2,965,479):
  - Macroprudential policy index (MPP)—Overall: Mean 5.943; Median 7.000; St. Dev. 3.581
  - MPP subindex—Bank: Mean 2.333; Median 1.500; St. Dev. 3.593
  - MPP subindex—Borrower: Mean 2.25; Median 2.000; St. Dev. 1.052
  - Monetary policy rate (domestic): Mean 8.136; Median 7.500; St. Dev. 2.373
  - GDP growth: Mean 4.367; Median 6.340; St. Dev. 4.998
  - Inflation: Mean 6.253; Median 6.692; St. Dev. 2.071
  - VIX: Mean 33.623; Median 35.199; St. Dev. 9.082
  - EONIA: Mean 2.389; Median 2.634; St. Dev. 1.542
- Bank variables (Obs varies):
  - Bank size (Obs 2,943,757): Mean 23.56; Median 23.71; St. Dev. 1.077
  - Bank capital (Obs 2,777,235): Mean 7.472; Median 7.046; St. Dev. 3.272
  - Bank liquidity (Obs 2,943,757): Mean 2.584; Median 2.092; St. Dev. 1.847
  - Bank ROA (Obs 2,965,479): Mean 0.992; Median 1.118; St. Dev. 1.836
  - Bank NPL (Obs 2,965,479): Mean 3.263; Median 0.962; St. Dev. 4.500
  - Bank risk profile (RWA/assets) (Obs 2,777,234): Mean 65.106; Median 65.131; St. Dev. 10.38
  - Bank foreign funding (Obs 2,965,479): Mean 18.891; Median 15.252; St. Dev. 24.98
  - Foreign bank dummy (Obs 2,965,479): Mean 0.812; Median 1.000; St. Dev. 0.391
- Notes: Loan amount in RON; OTHER currencies include USD, GBP, JPY (account for less than 0.5% of all loans). Loan and bank variables winsorized at the 1%. See Table A2 for variable definitions and sources.

### Table 2 — Macroprudential Policies and Household Credit: FX Loans (OLS, 2004-2012)
- Dependent variable: log loan-volume Lijkt extended by bank i to borrower j in county k during year t (column 1) or quarter t (columns 2-4). Observations at bank-loan-quarter level.
- Key coefficients (standard errors in parentheses):
  - Macroprudential policy (column 1): -0.0531*** (0.018)
  - Macroprudential policy × FX (column 2): -0.0500*** (0.018)
  - FX loan coefficients across columns: 1.6617*** (0.109); 2.1067*** (0.241); 1.9483*** (0.211); 2.5370*** (0.250)
- Observations: 2,753,494 (col 1); 2,965,459 (cols 2-4). R-squared: 0.2190 (col 1); 0.2630 (cols 2-4).
- Other details: Controls include macro, bank, borrower age, loan dummies; GDP growth interactions included; fixed effects vary by specification. Standard errors clustered on bank. Significance: *** 1% level.

### Table 3 — Macroprudential Policies and Household Credit: Borrower Riskiness (DTI, FX)
- Dependent variable: log loan-volume at bank-loan-quarter level. DTI divided by 100.
- Key coefficients:
  - Macroprudential policy × DTI (col 1): -0.0002 (0.005)
  - Macroprudential policy × DTI × FX (cols 2-4): -0.0323*** (0.007); -0.0296*** (0.005); -0.1317*** (0.027)
  - Macroprudential policy × DTI × RON (cols 2-4): 0.0177*** (0.004); 0.0091* (0.005); 0.0447*** (0.011)
  - Borrower DTI: 0.6595*** (0.072); 0.6489*** (0.071); 0.6695*** (0.067); 0.6625*** (0.074)
  - FX loan coefficients: 1.3612*** (0.068); 1.5589*** (0.085); 1.5096*** (0.082); 1.5888*** (0.085)
- Observations: 2,139,941. R-squared range: 0.2710–0.2730.
- Notes: All macro and bank variables spanned by fixed effects; controls include borrower age and loan dummies. Standard errors clustered on bank.

### Table 4 — Macroprudential Policies and Household Credit: Bank Exposure to Foreign Funding
- Dependent variable: log loan-volume at bank-loan-quarter level.
- Key coefficients:
  - Macroprudential policy × FX × Foreign-funding (cols 1-4): 0.0023*** (0.001); -0.0025*** (0.001); -0.0035** (0.001); -0.0048** (0.002)
  - Macroprudential policy × FX (cols 1-4): -0.0166 (0.017); -0.0024 (0.038); 0.0232 (0.045); -0.4736*** (0.128)
  - FX loan (cols 1-4): 2.0756*** (0.237); 2.2313*** (0.361); 2.1754*** (0.391); 3.0888*** (0.368)
- Observations: 2,965,459. R-squared: 0.2630–0.2640.
- Notes: Macro and bank variables spanned by fixed effects including Macroprudential policy × Foreign-funding. Standard errors clustered on bank.

### Table 5 — Spillovers of Global Risk Appetite on Household Credit
- Dependent variable: log loan-volume at bank-loan-quarter level (year t). Controls include macro, bank, borrower age, loan dummies.
- Key coefficients:
  - Macroprudential policy (cols 1-2): -0.5885*** (0.119); -0.5981*** (0.119)
  - VIX (cols 1-5): -0.0655*** (0.020); -0.0667*** (0.020); -0.0100 (0.007); 0.0069 (0.009); -0.0243*** (0.008)
  - Macroprudential policy × VIX (col 2): 0.0177*** (0.004)
  - Macroprudential policy × VIX × FX (col 3): 0.0189*** (0.005)
  - Macroprudential policy × VIX × RON (col 3): 0.0174*** (0.005)
  - Macroprudential policy × Low VIX × FX (cols with this term): -0.2317*** (0.050); -0.2850*** (0.071); -0.8161*** (0.128)
  - Macroprudential policy × High VIX × FX: -0.0928** (0.036); 0.0273 (0.069); -0.6135*** (0.088)
  - FX loan (cols 1-5): 1.6474*** (0.105); 1.5234*** (0.201); 1.8998*** (0.225); 1.8813*** (0.200); 2.3223*** (0.235)
- Observations: 2,753,494 across columns. R-squared: 0.2280–0.2290.
- P-value t-tests reported as 0.000 for tests comparing MPP×Low VIX×FX greater than MPP×Low VIX×RON, and similar tests for High VIX contrasts.
- Notes: Fixed effects at Semester level for bank, county, and loan-type. Standard errors clustered on bank.

### Table 6 — Spillovers of Foreign Monetary Policy on Household Credit
- Dependent variable: log loan-volume at bank-loan-quarter level.
- Key coefficients:
  - EONIA × FX (cols 1-3): -0.3015*** (0.101); -0.0966 (0.111); -0.6035*** (0.109)
  - EONIA × Macroprudential policy × FX (cols 1-3): 0.0448*** (0.015); -0.0263 (0.030); 0.2411*** (0.045)
  - Macroprudential policy × FX (cols 1-3): -0.2791*** (0.051); -0.1139 (0.074); -1.3423*** (0.235)
  - VIX × Macroprudential policy × FX (cols 1-3): 0.0060*** (0.002); 0.0046** (0.002); 0.0284*** (0.009)
  - FX loan (cols 1-3): 2.6930*** (0.390); 1.6440*** (0.309); 3.8198*** (0.581)
- Observations: 2,965,459. R-squared: 0.2650 across columns.
- Notes: Macro and bank variables spanned by fixed effects; GDP growth interactions refer to EONIA × GDP growth × FX. Standard errors clustered on bank.

*Italic source: References section and associated figures/tables from wp1813 - References (PDF).*

### Appendix figures and tables

### Appendix figures and tables (wp1813)

### Household and Corporate Credit Growth (Figure A1)
- Plots household and corporate credit growth in real terms (year-on-year), credit deflated by the CPI (2005=100).
- Data source: National Bank of Romania.

### Macroprudential Policy Measures, 2004-2012 (Table A1 — selected entries and coding)
- Coding rule: +1 if tightening and −1 if loosening. Indices reported as Overall index, Bank index, Borrower index.
- Selected chronological measures and their index entries:
  - 2004Q1: consumer credit installments ceiling 30% of net incomes; mortgage installments ceiling 35% → Overall 1, Bank 0, Borrower 1
  - 2004Q1: consumer credit downpayment >= 25%; mortgage credit value <= 75% of property value → Overall 1, Bank 0, Borrower 1
  - 2004Q3: reserve requirement on foreign currency deposits from 25% to 30% (domestic stays 18%) → Overall 1, Bank 1, Borrower 0
  - 2005Q3: DTI installments sum not exceed 40% of net incomes → Overall 1, Bank 0, Borrower 1
  - 2005Q3: reserve requirements on domestic currency liabilities reduced from 18% to 16% → Overall -1, Bank -1, Borrower 0
  - 2006Q1: reserve requirements on foreign currency liabilities raised from 30% to 35% and later to 40% → Overall 2, Bank 2, Borrower 0
  - 2006Q2: reserve requirements increased from 16% to 20% → Overall 1, Bank 1, Borrower 0
  - 2007Q1: eligibility criteria for DTI defined by banks’ internal models → Overall -1, Bank 0, Borrower -1
  - 2007Q1: foreign currency credit exposure limits removed → Overall -1, Bank -1, Borrower 0
  - 2007Q1: minimum capital requirement lowered from 12% to 8% → Overall -1, Bank -1, Borrower 0
  - 2008Q1: higher provisioning rate for loans to unhedged foreign currency borrowers → Overall 1, Bank 1, Borrower 0
  - 2009Q2: launch of the “first home” mortgage subsidy government program → Overall -1, Bank -1, Borrower -1
  - 2009Q2: reserve requirements on foreign currency liabilities with residual maturity > 2 years reduced from 40% to 0% → Overall -1, Bank -1, Borrower 0
  - 2011Q4: introduce LTV ceilings by loan currency denomination and specific foreign currency shocks; LTV limits: 85% (local-currency mortgages), 80% (foreign currency loans to hedged borrowers), 75% (EURO loans to unhedged borrowers), 60% (other currency loans to unhedged borrowers); consumer credit in foreign currency value of purchased goods shall not exceed 133%; maturity of consumer credit max 5 years → Overall 3, Bank 0, Borrower 3
  - 2012Q4: extension of regulatory measures to non-financial companies unhedged to currency risk → Overall 1, Bank 1, Borrower 0

### Variable definitions and data sources (Table A2 — key variables)
- CREDIT REGISTER DATA
  - Loan amount (in local currency: RON): Loan amount granted to an individual, expressed in Romanian New Leu (RON). Source: NBR
  - Borrower age (years): age at time of loan granting. Source: NBR
  - Debt-to-income ratio (DTI): debt-service-to-income ratio at loan origination; available for mortgage and consumer loans. Sources: NBR and Ministry of Public Finances
  - First-home mortgage: dummy = 1 for loans under first-time home ownership government program. Source: NBR
- MACRO VARIABLES
  - Macroprudential policy (MPP)—Overall: running sum index coded from Table A1; tightening = +1, loosening = -1. Source: Authors’ calculations
  - Macroprudential policy index—Bank / —Borrower: as above but limited to bank-based or borrower-based instruments. Source: Authors’ calculations
  - Monetary policy rate: NBR’s 7-day repo rate. Source: NBR
  - GDP growth: Real (yoy) growth of seasonally-adjusted GDP. Source: International Financial Statistics
  - Inflation: year-on-year CPI. Source: International Financial Statistics
  - European VIX: Euro Stoxx 50 Volatility Index. Source: STOXX
  - EONIA: Euro OverNight Index Average. Source: Reserve Bank of New Zealand
- BANK VARIABLES (selected)
  - Size: logarithm of total assets. Source: NBR
  - Capital: Tier 1 capital in percent of total assets. Source: NBR
  - Liquidity: liquid assets divided by required liquid assets. Source: NBR
  - ROA: net income divided by total assets. Source: NBR
  - NPL: non-performing loans divided by gross loans. Source: NBR
  - Foreign funding: non-resident deposits (mostly EUR) scaled by total assets; defined by maturity rules across periods. Source: NBR
  - Foreign bank: dummy for majority foreign ownership. Source: NBR

### Local macro determinants of macroprudential policies (Table A3 — OLS, 2004-2012; dependent variable: Overall MPP index)
- Column (1): Monetary policy rate coefficient -0.0212 (std. err. 0.141)
- Column (2): GDP growth coefficient 0.3308*** (std. err. 0.103)
- Column (3): Monetary policy rate coefficient -0.1823 (std. err. 0.211)
- Column (4): GDP growth coefficient 0.4345*** (std. err. 0.122)
- Column (1)-(5) other entries:
  - Inflation in columns reported: 0.0135 (0.224) and -0.0308 (0.341)
  - VIX coefficients shown: -0.0329 (0.065) and 0.0500 (0.064)
- Observations: 36 in each column
- R-squared reported: 0.001, 0.233, 0.000, 0.008, 0.300
- Notes: quarterly data 2004-2012; robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.10.

### Macroprudential policies and household credit — FX loans (Table A4 — OLS, 2004-2012; dependent variable: log loan-volume Lijkt)
- Column (1) (Overall index):
  - Macroprudential policy: -0.0531*** (0.018)
  - Macroprudential policy×FX: -0.0317** (0.015)
  - FX loan: 1.6617*** (0.109)
  - Monetary policy rate: -0.1079*** (0.026)
  - Inflation: 0.0741*** (0.015)
  - Bank size: -0.5412** (0.233)
  - Bank capital: -0.0334** (0.016)
  - Bank liquidity: -0.0601** (0.027)
  - Bank NPL: -0.1939** (0.076)
  - Bank risk profile: -0.0162** (0.006)
  - Foreign bank: -0.1909** (0.079)
  - Borrower age: -0.0080*** (0.003)
- Column (2) (Overall index, alternate specification):
  - Macroprudential policy: -0.0388 (0.024)
  - Macroprudential policy×FX: -0.0317** (0.015) reported in col (1) vs col (2) shows variation
  - FX loan: 1.9455*** (0.205)
  - Bank size: -0.5190** (0.233)
  - Bank capital: -0.0309* (0.016)
  - Bank liquidity: -0.0621** (0.027)
  - Bank NPL: -0.1912** (0.078)
  - Bank risk profile: -0.0166** (0.006)
  - Foreign bank: -0.2016** (0.076)
  - Borrower age: -0.0081*** (0.003)
- Column (3) includes bank×year-quarter, county×year-quarter, loan-type×year-quarter FE:
  - Borrower age: -0.0074*** (0.003)
  - First-home mortgage coefficients shown: -0.0228 (0.143), -0.0839 (0.140), -0.2119 (0.173) across cols (1)-(3)
- Observations: 2,753,494 (cols 1-2), 2,965,459 (col 3)
- R-squared: 0.219 (cols 1-2), 0.263 (col 3)
- Notes: macro and bank variables averaged over last two quarters; standard errors clustered on bank.

### Macroprudential policies and household credit — FX loans by currency (Table A5)
- Interaction effects (Macroprudential policy×FX×Currency):
  - EUR:
    - Overall index: -0.0488** (0.018)
    - Bank index: -0.0301 (0.019)
    - Borrower index: -0.3552*** (0.068)
  - CHF:
    - Overall index: -0.0551 (0.033)
    - Bank index: -0.0416 (0.036)
    - Borrower index: -0.3191*** (0.097)
  - OTHER:
    - Overall index: -0.1415** (0.054)
    - Bank index: -0.0941** (0.046)
    - Borrower index: -0.7633*** (0.201)
- FX loan coefficients:
  - Overall index: 2.1041*** (0.241)
  - Bank index: 1.9479*** (0.211)
  - Borrower index: 2.5248*** (0.249)
- Observations: 2,965,459 in each column
- R-squared: 0.263 (cols 1-2), 0.264 (col 3)
- Notes: macro and bank variables spanned by fixed effects; controls include borrower age and first-home mortgage dummy.

### Macroprudential policies and household credit — Borrower riskiness by currency (Table A6)
- Interaction effects (Macroprudential policy×DTI×Currency):
  - EUR:
    - Overall index: -0.0340*** (0.007)
    - Bank index: -0.0312*** (0.006)
    - Borrower index: -0.1358*** (0.028)
  - CHF:
    - Overall index: -0.0239 (0.015)
    - Bank index: -0.0334* (0.017)
    - Borrower index: -0.0695 (0.041)
  - OTHER:
    - Overall index: -0.0536* (0.027)
    - Bank index: -0.0352* (0.019)
    - Borrower index: -0.2797** (0.110)
  - RON:
    - Overall index: 0.0175*** (0.004)
    - Bank index: 0.0089* (0.005)
    - Borrower index: 0.0451*** (0.011)
- Borrower DTI coefficient: 0.6504*** (0.071), 0.6700*** (0.067), 0.6622*** (0.074) across columns
- FX loan coefficients: 1.5616*** (0.085), 1.5118*** (0.083), 1.5907*** (0.085)
- Observations: 2,139,941 in each column
- R-squared: 0.273, 0.272, 0.273
- Notes: DTI ratio divided by 100. Standard errors clustered on bank.

### Spillovers of global risk appetite on household credit — By loan type (Table A7)
- Residential mortgages (columns 1–3) and Consumer loans (columns 4–6) reported separately.
- Key interaction coefficients (Macroprudential policy×Low VIX×FX):
  - Residential mortgages Overall index: -0.0314*** (0.010)
  - Residential mortgages Bank index: -0.0390** (0.016)
  - Residential mortgages Borrower index: -0.0784*** (0.024)
  - Consumer loans Overall index: -0.2564*** (0.053)
  - Consumer loans Bank index: -0.3118*** (0.075)
  - Consumer loans Borrower index: -0.8438*** (0.127)
- Macroprudential policy×Low VIX×RON:
  - Residential mortgages Overall: 0.0066 (0.023)
  - Consumer loans Overall: -0.1984*** (0.056)
- Macroprudential policy×High VIX×FX:
  - Residential mortgages Overall: -0.0053 (0.015)
  - Consumer loans Overall: -0.1109** (0.041)
- VIX coefficients:
  - Residential mortgages: -0.0035 (0.002), 0.0006 (0.002), -0.0065** (0.003) across columns
  - Consumer loans: -0.0106 (0.008), 0.0059 (0.010), -0.0263*** (0.009)
- FX loan coefficients:
  - Residential mortgages: 0.5604*** (0.190), 0.5657*** (0.155), 0.3428** (0.143)
  - Consumer loans: 2.0168*** (0.249), 2.0021*** (0.221), 2.2373*** (0.258)
- P-value t-tests for equality of coefficients (MPP×Low VIX×FX greater than …):
  - MPP×Low VIX×RON: 0.0000 (residential), 0.0000 (consumer), 0.0050 (borrower/residential), 0.0000 (others)
  - MPP×High VIX×FX: 0.0000 across reported tests
  - MPP×High VIX×RON: 0.0070, 0.0000, 0.0010, 0.0000, 0.0000, 0.000 reported as applicable
- Observations:
  - Residential mortgages: 277,590 (cols 1–3)
  - Consumer loans: 2,475,877 (cols 4–6)
- R-squared:
  - Residential mortgages: 0.197, 0.197, 0.198
  - Consumer loans: 0.185, 0.186, 0.186
- Notes: OLS 2004-2012; dependent variable log loan-volume Lijkt at bank-loan-quarter level. Controls include local monetary policy, GDP growth, inflation, bank variables (size, capital, liquidity, ROA, NPL, risk profile, share of foreign funding, foreign bank dummy), borrower age, and first-home mortgage dummy. GDP growth interactions refer to GDP growth×VIX×FX and GDP growth×VIX×RON. Standard errors clustered on bank. *** p<0.01, ** p<0.05, * p<0.10.

*Appendix figures and tables — wp1813.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp1813.pdf_