## wpiea2019066 - 1.5 ppts.

## Source details

**Canonical URL:** [wpiea2019066 - 1.5 ppts.](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019066.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019066.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019066.pdf.json)

---

### Effects on household credit and house prices
- Loan-targeted tools (LTV and DSTI limits; supply-loans such as limits on credit growth, loan loss provisions, loan restrictions) robustly reduce household credit growth across country groups.
- Supply-side loan-targeted measures have larger effects on household credit than demand-side measures:
  - A tightening event of loan supply tools is associated with an approximately 3 ppts decline in household credit growth after four quarters in both AEs and EMDEs.
- Unconditional household credit growth averages about 8 ppts per year (ALL); in EMDEs it averages 11.6 ppts, while in AEs household credit increases yearly by 5.6 percent.
- Demand-side tightening (DSTI and LTV) is robustly and negatively associated with household credit:
  - In EMDEs, one tightening (loosening) action moderates (raises) household credit by about 6 percent for these tools.
- Effects on house prices are mostly weaker:
  - Loan-targeted measures can mitigate house price growth, reducing house price growth by about one ppt for the entire sample and by 1.7 ppts for AEs.
  - Demand-targeted measures appear most potent for house prices.
  - Tax-related macroprudential policies (e.g., stamp duties) in AEs appear to dampen house price growth by a substantial 5 ppt, while having statistically insignificant impacts on household credit growth.

### Side-effects on private consumption and GDP
- Overall evidence of side effects on consumption is limited, but loan-targeted actions—especially those targeting the supply of bank loans—are negatively associated with private consumption growth:
  - A tightening of loan supply measures can reduce consumption growth by up to about 2 ppts when all loan supply measures are combined in an index.
  - Negative effects are stronger in AEs.
- Individual instrument effects on private consumption (cumulative after one year):
  - LTV limits: roughly 0.8 ppts reduction.
  - Loan restrictions: roughly 3.8 ppts reduction.
  - In AEs, loan loss provisions and loan restrictions show negative effects with magnitudes between 2.2 and 4.3 ppts.
  - In EMDEs, tax-related macroprudential policies reduce private consumption growth by 1.4 ppts.
- Real GDP growth effects are generally weak and not statistically significant:
  - Exception: limits on credit growth in AEs reduce real GDP growth by 0.8 ppts.

### Efficiency, nonlinearity, and policy interpretation
- Loan-targeted policy actions have significant effects on household credit growth; effects on house prices are somewhat weaker.
- Demand-side tools (DSTI and LTV limits) are relatively efficient: sizable effects on credit growth with more modest side-effects on consumption.
- Nonlinearity and diminishing returns:
  - AIPW estimation finds per-one-ppt LTV-tightening effects diminish as the size of tightening increases (policy leakage and regulatory arbitrage cited as explanations).
  - Estimated causal effects (AIPW, cumulative after four quarters for a one-ppt LTV tightening):
    - Tightening by Less than 10 ppts: -0.65***
    - Tightening by 10- 25 ppts: -0.36***
    - Average effect of any tightening (three-bucket AIPW): -0.31 ppts (compared to -0.16 ppts in the FE regression).
- Tradeoff:
  - Tightening LTV limits entails a tradeoff: relatively larger side effects on consumption when initial LTV limits are already tight.
- Policy recommendation:
  - Countries with tight LTV limits might be better off considering other macroprudential tools to complement the existing portfolio of tools, in line with international best practices (IMF-FSB-BIS, 2016).
- Broader conclusion:
  - A full welfare analysis is required for a comprehensive cost-benefit assessment.

### Quantifying one-percentage-point changes in LTV limits (AIPW / ATE results)
- Methodology:
  - Used an Augmented Inverse Propensity-Weighted (AIPW) estimator with an ordered logit for the propensity score (policy action buckets {-20, -10, 0, 10, 20}) and panel data for 58 countries over 2000 Q1 to 2016 Q4.
  - Estimated ATEs rescaled by the average ΔLTV for each bucket to obtain the effect of a one-percentage-point change in the average LTV limit.
- Estimated causal effects (AIPW):
  - For a tightening of less than 10 ppts: cumulative decline in real household credit growth after four quarters is estimated at 0.65 ppts per one-ppt LTV tightening.
  - For a tightening of more than or equal to 10 ppts and less than 25 ppts: estimated at -0.36 ppts per one-ppt tightening.
  - The average effect on household credit growth of any tightening measure is estimated at -0.31 ppts (compared to -0.16 ppts in the FE regression) based on AIPW with three buckets (tightening, loosening, control).
- Nonlinearity and diminishing returns:
  - Estimated effects per one-ppt tightening diminish as the size of the LTV adjustment increases.
  - Smaller effects for larger tightenings may reflect policy leakage and regulatory arbitrage (e.g., shift to non-bank credit or foreign funding not covered by domestic LTV limits).
- Caveats:
  - Magnitudes depend on bucket thresholds and sample size; large LTV changes are subject to limited observations and outlier influence.
  - Linear models would likely be misspecified given observed nonlinearity.

### Estimation approach, data, and robustness
- Data sources: The iMaPP database, Bloomberg, BIS, OECD, others (see Appendix IV), and the authors’ estimation.
- Estimation methods:
  - Augmented inverse propensity-score weighted (“AIPW”) estimation.
  - Fixed effects estimation with the timing assumption (“FE regression”).
  - Arellano-Bover-Blundell-Bond system GMM used as robustness check.
  - Quantile regressions and propensity-score-based methods used to address endogeneity and reverse causality.
- Sample and timing:
  - Quarterly panel data for about 58 countries over 2000 Q1 to 2016 Q4 (various tables reference periods up to 1991Q1 – 2016Q4 for robustness).
  - Confidence levels: *** p<0.01, ** p<0.05, * p<0.1. Standard errors clustered by country.
- Key robustness findings (selected magnitudes from Appendix VIII):
  - Appendix VIII Table 1 — Real Household Credit (yoy growth): MaPP (All, Sum): FE -0.842***; GMM -1.061***. MaPP (Loan, Sum): FE -1.883***; GMM -2.443***. ltv: FE -2.557**; GMM -3.658**. dsti: FE -2.648***; GMM -3.015**.
  - Appendix VIII Table 2 — Real Consumption (yoy growth): MaPP (Loan, Sum): FE -0.999***; GMM -0.990***. ltv: FE -0.778*; GMM -1.243**. loanr: FE -3.841***; GMM -4.346***.
  - Appendix VIII Table 3 — Real House Prices (yoy growth): MaPP (Loan, Sum): FE -1.066**; GMM -1.337**. ltv: FE -1.952***; GMM -2.127***. llp: FE -2.987**; GMM -3.733***.
  - Appendix VIII Table 4 — Real GDP (yoy growth): MaPP (All, Sum): coefficients small and statistically insignificant in reported columns; MaPP (Supply Capital, Sum): GMM shows -0.578*.
- Appendix VI and VIII subgroup findings:
  - Heterogeneity: Effects stronger in EMDEs, Asia and Pacific, countries with high indebtedness among low-income borrowers, and when credit gap is positive.
  - Sum coeff LTV level (Table 9, cumulative after four quarters): ALL: 0.15; AE: 0.07; EMDE: 0.37; ASIA: 0.30; EUROPE: 0.09; AMERICAS: 0.45.
  - For private consumption (Table 10, cumulative after four quarters): ALL: 0.08; AE: 0.04; EMDE: 0.11; AMERICAS: 0.16; Hi DTI Low-Income Borrowers: 0.38.
- Identification and data handling:
  - Timing assumption: LTV limits do not affect credit growth within the same quarter.
  - Observations with ΔLTV ≤ -25 ppts are excluded to mitigate outlier influence.
  - When creating the average LTV limit, values set to 100 (or 110 where pre-introduction regulatory limits exceeded 100) for pre-introduction periods to correctly capture tightenings when new loan categories are regulated.

### Policy-relevant takeaways
- LTV and DSTI tightenings effectively dampen household credit growth; the largest marginal per-ppt impacts occur for smaller tightenings (under 10 ppts).
- Side effects on consumption are present but modest in magnitude (around 0.1 ppts per one-ppt LTV tightening in many specifications); side effects rise when initial LTVs are already tight.
- Macroprudential policy design should account for nonlinearity, country-specific characteristics (exchange rate regime, capital openness, financial development, indebtedness distribution), and potential policy leakage to non-bank or foreign funding channels.
- AIPW provides stronger evidence of causal and nonlinear effects on household credit than timing-assumption FE regressions; nevertheless, a comprehensive welfare-cost benefit analysis remains necessary.

*Source: IMF working paper (wpiea2019066) content unit.*

### 1.5 ppts.

### wpiea2019066 - 1.5 ppts.

### Effects on household credit and house prices
- Loan-targeted tools (LTV and DSTI limits; supply-loans such as limits on credit growth, loan loss provisions, loan restrictions) robustly reduce household credit growth across country groups.
- Supply-side loan-targeted measures have larger effects on household credit than demand-side measures:
  - A tightening event of loan supply tools is associated with an approximately 3 ppts decline in household credit growth after four quarters in both AEs and EMDEs.
- Unconditional household credit growth averages about 8 ppts per year (ALL); in EMDEs it averages 11.6 ppts, while in AEs household credit increases yearly by 5.6 percent.
- Demand-side tightening (DSTI and LTV) is robustly and negatively associated with household credit:
  - In EMDEs, one tightening (loosening) action moderates (raises) household credit by about 6 percent for these tools.
- Effects on house prices are mostly weaker:
  - Loan-targeted measures can mitigate house price growth, reducing house price growth by about one ppt for the entire sample and by 1.7 ppts for AEs.
  - Demand-targeted measures appear most potent for house prices.
  - Tax-related macroprudential policies (e.g., stamp duties) in AEs appear to dampen house price growth by a substantial 5 ppt, while having statistically insignificant impacts on household credit growth.

### Side-effects on private consumption and GDP
- Overall evidence of side effects on consumption is limited, but loan-targeted actions—especially those targeting the supply of bank loans—are negatively associated with private consumption growth:
  - A tightening of loan supply measures can reduce consumption growth by up to about 2 ppts when all loan supply measures are combined in an index.
  - Negative effects are stronger in AEs.
- Individual instrument effects on private consumption (cumulative after one year):
  - LTV limits: roughly 0.8 ppts reduction.
  - Loan restrictions: roughly 3.8 ppts reduction.
  - In AEs, loan loss provisions and loan restrictions show negative effects with magnitudes between 2.2 and 4.3 ppts.
  - In EMDEs, tax-related macroprudential policies reduce private consumption growth by 1.4 ppts.
- Real GDP growth effects are generally weak and not statistically significant:
  - Exception: limits on credit growth in AEs reduce real GDP growth by 0.8 ppts.

### Efficiency and policy interpretation
- Loan-targeted policy actions have significant effects on household credit growth; effects on house prices are somewhat weaker.
- Demand-side tools (DSTI and LTV limits) are relatively efficient: sizable effects on credit growth with more modest side-effects on consumption.
- A full welfare analysis is required for a comprehensive cost-benefit assessment.

### Robustness and identification concerns
- Potential biases include reverse causality, limited usage of some tools (few countries), and dynamic panel bias from including lagged dependent variables.
- Robustness checks include system GMM panel estimates and quantile regressions (appendices); reverse causality is addressed with a propensity-score-based method.

### Quantifying one-percentage-point changes in LTV limits (AIPW / ATE results)
- Methodology:
  - Used an Augmented Inverse Propensity-Weighted (AIPW) estimator with an ordered logit for the propensity score (policy action buckets {-20, -10, 0, 10, 20}) and panel data for 58 countries over 2000 Q1 to 2016 Q4.
  - Estimated ATEs rescaled by the average ΔLTV for each bucket to obtain the effect of a one-percentage-point change in the average LTV limit.
- Estimated causal effects (AIPW):
  - For a tightening of less than 10 ppts: cumulative decline in real household credit growth after four quarters is estimated at 0.65 ppts per one-ppt LTV tightening.
  - For a tightening of more than or equal to 10 ppts and less than 25 ppts: estimated at -0.36 ppts per one-ppt tightening.
  - The average effect on household credit growth of any tightening measure is estimated at -0.31 ppts (compared to -0.16 ppts in the FE regression) based on AIPW with three buckets (tightening, loosening, control).
- Nonlinearity and diminishing returns:
  - Estimated effects per one-ppt tightening diminish as the size of the LTV adjustment increases.
  - Smaller effects for larger tightenings may reflect policy leakage and regulatory arbitrage (e.g., shift to non-bank credit or foreign funding not covered by domestic LTV limits).
- Caveats:
  - Magnitudes depend on bucket thresholds and sample size; large LTV changes are subject to limited observations and outlier influence.
  - Linear models would likely be misspecified given observed nonlinearity.

*Source: IMF working paper (wpiea2019066) content unit.*

### 1.  Real Household Credit Growth  2.  Real Consumption Growth

### 1.  Real Household Credit Growth  2.  Real Consumption Growth

### Estimation approach and data notes
- Sources: The iMaPP database, Bloomberg, BIS, OECD, others (see Appendix IV), and the authors’ estimation.
- Estimation methods:
  - Augmented inverse propensity-score weighted (“AIPW”) estimation.
  - Fixed effects estimation with the timing assumption (“FE regression”).
- FE regression uses interaction terms of ΔLTV with dummy variables for buckets: tightening by less than 10 ppts, and tightening by more than or equal to 10 ppts and less than 25 ppts.
- To mitigate influence of outliers, observations with ΔLTV less than or equal to -25 ppts are excluded.
- Confidence levels: *** p<0.01, ** p<0.05, * p<0.1. Standard errors clustered by country.

### Main findings — effects on household credit
- Nonlinearity:
  - The AIPW estimation finds strong and nonlinear effects on household credit; effects per one percentage point (ppt) LTV tightening are diminishing with the size of the LTV adjustment.
  - The FE regression based on the timing assumption also finds nonlinear effects on household credit, but estimates are smaller and not significant.
  - The timing-assumption FE regression result supports the notion of attenuation bias in typical regression estimates (Appendix III), and that AIPW better addresses endogeneity.
- Key quantitative estimates (cumulative effects after four quarters for a one-ppt LTV tightening):
  - AIPW (tightening by Less than 10 ppts): -0.65***
  - AIPW (tightening by 10- 25 ppts): -0.36***
  - FE regression (tightening by Less than 10 ppts): -0.43
  - FE regression (tightening by 10- 25 ppts): -0.10
- Summary interpretation:
  - The largest per-unit impact is for tightening of less than 10 ppts: "a one ppt LTV tightening cumulatively reduces household credit growth by up to 0.65 ppts after one year."
  - Diminishing marginal effect likely driven by policy leakage effects.

### Main findings — effects on private consumption
- Magnitude and pattern:
  - The estimated side effects on consumption are smaller and less robust than effects on credit.
  - AIPW estimates of the consumption growth decline are around 0.1 ppts and show no clear nonlinear pattern in aggregate.
- Key quantitative estimates (cumulative effects after four quarters for a one-ppt LTV tightening):
  - AIPW (tightening by Less than 10 ppts): -0.15*
  - AIPW (tightening by 10- 25 ppts): -0.11***
  - FE regression (tightening by Less than 10 ppts): 0.11
  - FE regression (tightening by 10- 25 ppts): 0.00
- Summary interpretation:
  - Side effects on consumption are statistically significant in some specifications but moderate in size (around 0.1 ppts).
  - Tightening by less than 10 ppts appears more efficient: larger effects on credit with smaller side effects on consumption.

### Do initial LTV levels matter?
- Thresholds for “tight” versus “loose” initial LTV levels:
  - Advanced economies (AEs): 100 percent (median LTV level).
  - Emerging market and developing economies (EMDEs): 90 percent (median LTV level).
- Estimation strategy:
  - Panel regression with the timing assumption to distinguish AEs and EMDEs.
  - AIPW used as robustness check without distinguishing AEs and EMDEs.
- Findings:
  - Consumption:
    - The effect of an additional tightening on consumption is larger when initial LTV limits are already tight.
    - In general, a tightening by one percentage point is only significantly associated with a decline in consumption if the LTV is already tight; result consistent across subsamples (Panel 2 of Figure 7).
  - Household credit:
    - Effects on household credit growth of a tightening are generally more pronounced when starting from a loose level.
    - In AEs, effects on household credit are only significant when the initial LTV cap is loose (greater or equal to 100 percent).
    - In EMDEs, effects are slightly stronger when initial LTV levels are high, though differences are not always statistically significant.
  - AIPW corroboration:
    - Excluding observations with initial LTV levels greater or equal to 100 percent, estimated effect on household credit becomes smaller, but side effect on consumption increases compared with full sample results (Appendix VI Table 1).
- Behavioral interpretation reconciling differential effects:
  - From a loose initial LTV: borrowers reduce loan size and meet higher downpayment from own funds — credit drops but consumption effect muted because purchases still proceed.
  - From an already tight initial LTV: further tightening leads borrowers to increase savings to meet new LTV limits rather than reduce borrowing further — larger decline in consumption.

### Policy implications and conclusions
- Tradeoff:
  - Tightening LTV limits entails a tradeoff: relatively larger side effects on consumption when initial LTV limits are already tight.
- Policy recommendation:
  - Countries with tight LTV limits might be better off considering other macroprudential tools to complement the existing portfolio of tools, in line with international best practices (IMF-FSB-BIS, 2016).
- Broader conclusions:
  - Importance of accounting for nonlinearity and reverse causality in estimating policy effects.
  - AIPW provides an approach that better addresses endogeneity and uncovers stronger nonlinear effects on household credit than timing-assumption FE regressions.
  - The new iMaPP database will be updated annually with the IMF’s Annual Survey on Macroprudential Policies, enabling further research, including quantification of effects of other macroprudential policies and cost-benefit analysis.

*Source: wpiea2019066 - 1.  Real Household Credit Growth  2.  Real Consumption Growth (IMF authors’ estimation and iMaPP database).*

### APPENDIX II. CONSTRUCTING THE AVERAGE LTV LIMIT

### APPENDIX II. CONSTRUCTING THE AVERAGE LTV LIMIT

### Adjustments when constructing the average LTV limit
- When a country introduces a LTV limit on a new category of loans, for the periods prior to the introduction the value is set at 100 for that category.  
  - Rationale: This treatment is necessary for the average LTV limit to correctly suggest a tightening when an LTV limit is introduced for an additional loan category (illustrated in Appendix II Table 1).
- When the introduced regulatory limits on LTV ratios are above 100 percent (example countries given: Netherlands and Slovakia), the value is set at 110 for the periods prior to the introduction.  
  - Note (footnote 28): "110 percent is an ad-hoc value, although average LTV ratio of all existing mortgage loans in Netherlands were about 110 percent when the LTV limits were introduced."
- When a country does not have any LTV limits at all, the value is set at 100, consistent with the logic above.

### Rationale for using the simple average of regulatory LTV limits
- The average LTV limit series summarizes regulatory developments reasonably well even when there are multiple LTV limits in a country.
- Using the tightest limit instead of the simple average could miss regulatory changes in other important loan categories if those categories are not the tightest.
- The simple average of regulatory LTV limits:
  - Reflects changes across all categories.
  - Empirically exhibits a picture consistent with the dummy-type index for the developments of LTV limits.

### Recording dates and data notes
- The average LTV limit and dummy-type policy action indicators are recorded based on their effective dates, because announcement dates are often not available.
- The text information in the iMaPP database contains announcement dates when they are available.

### Illustration note (Appendix II Table 1)
- Without the treatment of setting pre-introduction values to 100 (or 110 where applicable), the average LTV limit can mistakenly appear as a loosening when an LTV limit on an additional loan category is introduced; with the treatment the average LTV limit correctly indicates a tightening.

*Source: APPENDIX II. CONSTRUCTING THE AVERAGE LTV LIMIT (authors).*

### Appendix VI Table 1: AIPW Results – Effects of Initial LTV Levels

### Appendix VI Table 1: AIPW Results – Effects of Initial LTV Levels

### Data, sample, and estimation approach
- Sources: The iMaPP database, Bloomberg, BIS, OECD, others (see Appendix IV), and the authors’ estimation.
- Sample and period: Quarterly data from about 58 countries over the period 2000 Q1 to 2016 Q4 for panel regressions; period sample 1991Q1 – 2016Q4 referenced for robustness tables.
- Estimation methods:
  - AIPW estimation for groups (table reports group of tightening less than 10 ppts using a five-bucket model).
  - Panel regressions with fixed effects and lagged dependent variable (equation (A.VII.1)).
  - Robustness: Arellano-Bover-Blundell-Bond system GMM used as robustness check.
- Key identification assumption: Timing assumption — LTV limits do not affect credit growth within the same quarter.
- Exclusions/controls:
  - Observations with ΔLTV ≤ -25 ppts are excluded to mitigate influence of outliers.
  - Observations with “LTVt-1≥100” are not used for estimation.
  - Country and time fixed-effect dummies could not be used in some treatment groups; AE and EMDE dummies used to control some country fixed effects; quarter dummies and VIX control some time fixed effects.
  - Control vector X includes real GDP growth, domestic real interest rates, and other macroprudential policies.

### Main empirical findings — Effects on real household credit
- Focus metric: Sum of ΔLTV coefficients (cumulative effects over previous 4 lags).
- Significant aggregate and subgroup effects:
  - Effects of a one ppt tightening of LTV limit on real household credit vary from about 0.07 ppts to 0.37 ppts.
  - Effects are found significant for EMDEs, Asia and Pacific region, countries with high indebtedness level of low-income borrowers, and when the credit gap is positive in EMDEs.
- Persistence in credit growth (lagged credit growth coefficient):
  - Ranges from 0.82 for North and Latin American countries to 0.97 in countries with high indebtedness level of the low-income borrowers.
- Interest rate and business cycle controls:
  - Higher short-term interest rates negatively impact future household credit growth; interest rate effects slightly stronger in EMDEs than in AEs.
  - Past output growth (proxy for business cycle) is positively associated with credit growth.
- Dependent variable definition: Total private credit to household sector (in real terms), which includes both mortgage and consumer credit from all sources.

### Main empirical findings — Effects on real private consumption
- Significant effects on consumption in several samples:
  - Effects are significant when all countries are considered, for the EMDE group, for countries with highly indebted low-income borrowers, and when EMDEs experience a credit boom.
- Magnitude:
  - The magnitude of a one ppt tightening of LTV limit on real private consumption varies from about 0.08 ppts to 0.18 ppts.

### Robustness and additional checks
- Results are robust to:
  - Controlling for other macroprudential policies (Appendix VIII Tables 11-12).
  - Use of the Arellano-Bover-Blundell-Bond system GMM estimator (Appendix VIII Tables 13-14).
  - Subsample analyses by country groups (EMDE/AE), regions (Asia, Europe, Americas), exchange rate flexibility, capital openness, financial development, high debt-to-income among low-income borrowers, and conditioning on positive credit gap.
- Micro-level split for indebtedness: Countries split into high/low indebtedness of low-income borrowers if the average debt-to-income ratio of the bottom 40 percent households by income is above/below the median.

### Selected coefficient magnitudes from Appendix VIII robustness tables (fixed effects and GMM results)
- Appendix VIII Table 1 — Dependent Variable: Real Household Credit (yoy growth)
  - MaPP (All, Sum): FE -0.842***; GMM -1.061*** (standard errors reported in table).
  - MaPP (Loan, Sum): FE -1.883***; GMM -2.443***.
  - ltv: FE -2.557**; GMM -3.658** (in specified columns).
  - dsti: FE -2.648***; GMM -3.015**.
  - Observations: 4,492 (various column samples reported).
- Appendix VIII Table 2 — Dependent Variable: Real Consumption (yoy growth)
  - MaPP (Loan, Sum): FE -0.999***; GMM -0.990***.
  - MaPP (Supply Loans, Sum): FE -2.006**; GMM -1.950**.
  - ltv: FE -0.778*; GMM -1.243** (in specified columns).
  - loanr: FE -3.841***; GMM -4.346***.
  - Observations: 4,639 (various column samples reported).
- Appendix VIII Table 3 — Dependent Variable: Real House Prices (yoy growth)
  - MaPP (Loan, Sum): FE -1.066**; GMM -1.337**.
  - MaPP (Demand, Sum): FE -1.406**; GMM -1.774**.
  - ltv: FE -1.952***; GMM -2.127***.
  - llp: FE -2.987**; GMM -3.733***.
  - Observations: 4,111 (various column samples reported).
- Appendix VIII Table 4 — Dependent Variable: Real GDP (yoy growth)
  - MaPP (All, Sum): coefficients small and statistically insignificant in reported columns (e.g., 0.0466, 0.0108).
  - MaPP (Supply Capital, Sum): GMM column shows -0.578*.
  - Observations: 5,416 (various column samples reported).
- Appendix VIII Table 5 — Baseline: Dependent Variable: Real Household Credit (yoy growth)
  - MaPP (All, Sum): quantile and OLS results include -0.550*, -0.842***, -0.843*** across columns.
  - MaPP (Loan, Sum): reported values include -1.538, -1.883***, -1.458*** in different specifications.
  - ltv: reported values across specifications include -1.840*, -2.557**, -2.057*** and others depending on column.
  - dsti: reported values include -4.485**, -2.648***, -1.713*** in different columns.
  - Observations: 4,492 in several reported columns; number of groups and N(countries) vary by column.

### Interpretation and policy implications (from reported analysis)
- LTV tightenings have measurable contractionary effects on household credit and private consumption:
  - A one ppt tightening of the LTV limit reduces real household credit growth by between about 0.07 ppts and 0.37 ppts in the panel estimates.
  - A one ppt tightening reduces real private consumption growth by between about 0.08 ppts and 0.18 ppts in the panel estimates.
- Heterogeneity matters for policy effectiveness:
  - Stronger effects in EMDEs, Asia and Pacific, and in countries where low-income borrowers are more indebted or where credit gaps are positive.
  - Policy design should consider country-specific characteristics (exchange rate regime, capital openness, financial development, distribution of household indebtedness).
- Complementary macroprudential policies and interest rate environment influence outcomes:
  - Controlling for other macroprudential measures does not eliminate LTV effects.
  - Short-term interest rates amplify or dampen credit responses; interest-rate-sensitive contexts (EMDEs, Europe) show stronger relationships.

*Source: Appendix VI–VIII materials from the provided IMF working paper content unit.*

### Appendix VIII Table 6. Baseline: The Effects of Macroprudential Policies on

### Appendix VIII Table 6. Baseline: The Effects of Macroprudential Policies on Real Private Consumption Growth

### Estimation setup and sample
- Fixed effects (FE) estimation with timing assumption; country and time fixed effects; includes lagged dependent variable, real output growth (lag), and domestic interest rates (lag).
- Each macroprudential policy indicator is the average over the previous four quarters and considered individually.
- Quantile regressions refer to 10th and 90th percentile.
- Period sample: 1991Q1 – 2016Q4.
- Standard errors clustered at the country level. Confidence levels: *** p<0.01, ** p<0.05, * p<0.1.
- Dependent Variable: Real Consumption (yoy growth).

### Selected coefficient estimates (cumulative effects after four quarters / other controls)
- MaPP (All, Sum): -0.139; -0.150; -0.197; 0.0459; -0.170; -0.401; -0.666*; -0.197; -0.361 (standard errors in table).
- MaPP (Loan, Sum): -0.861; -0.999***; -0.983*; -0.516; -0.888**; -0.758***; -1.245***; -0.914; -0.958.
- MaPP (Demand, Sum): -0.838; -0.649*; -0.618*; -0.224; -0.527*; -0.767*; -1.744; -0.607; -0.468.
- MaPP (Supply All, Sum): 0.205; 0.0964; 0.502*; 0.557; 0.476; 0.0524; -0.461; -0.176; -0.209.
- MaPP (Supply Loans, Sum): -1.491; -2.006**; -2.016***; -1.319; -2.707**; -2.134; -1.769; -1.370; -1.054.
- MaPP (Supply General, Sum): 0.962; 0.359; 0.472; 2.002***; 0.998*; 1.136; -0.009390; 0.0276; -0.649.
- MaPP (Supply Capital, Sum): -0.504; -0.137; 0.685; -0.0986; -0.0225; -0.433; -1.266; -0.453; 0.207.
- Key controls (examples): dsti: -0.262; -1.054; -1.603; -0.500; -1.066; -0.589; -1.801; -1.569; -0.890. loanr: -3.727**; -3.841***; -3.956***; -4.249*; -4.267**; -2.505**; -1.318; -2.486; -2.520.
- Observations / Number of groups / R-squared (avg):
  - Observations: 4,639; 4,639; 4,639; 2,863; 2,863; 2,863; 2,631? (table shows several groupings; full listing in table).
  - Number of groups: 55; 55; 55; 31; 31; 31; 24; 24; 24.
  - R-squared (avg): 0.816; 0.788; 0.824.

---

### Appendix VIII Table 7. Baseline: The Effects of Macroprudential Policies on Real House Prices Growth

### Estimation setup and sample
- Same FE timing specification and sample: 1991Q1 – 2016Q4.
- Dependent Variable: Real House Prices (yoy growth).

### Selected coefficient estimates (cumulative effects after four quarters / other controls)
- MaPP (All, Sum): -0.713*; -0.0877; -0.234; -0.727; -0.898**; -1.129***; -0.5240; 0.0619; -0.0631.
- MaPP (Loan, Sum): -1.154; -1.066**; -0.509; -1.161; -1.667**; -1.374**; -1.042; -0.615; -0.566**.
- MaPP (Demand, Sum): -0.637; -1.406**; -0.768; -1.332***; -2.049**; -1.402; 0.00370; -0.213; -0.274.
- MaPP (Supply Loans, Sum): -2.511*; -1.191; -0.0899; -2.677; -1.970; -1.634*; -1.719***; -1.191; 1.010.
- MaPP (Supply All, Sum): -0.588; 0.253; -0.00772; -0.364; 0.0116; -1.010*; -0.717; 0.007890; 0.0208.
- Controls and notable coefficients:
  - ltv: -2.243; -1.952***; -0.491; -2.872***; -2.627***; -1.817; -1.480; -1.020; -1.487.
  - llp: -1.538*; -2.987**; -1.629; -0.392; -1.510; -2.124; -3.289**; -4.529**; -4.618***.
  - loanr: -4.538***; -1.385; 0.166; -7.895***; -4.164*; -1.756; -3.352**; -0.262; 2.963.
- Observations / Number of groups / R-squared (avg):
  - Observations: 4,111; 4,111; 4,111; 2,780; 2,780; 2,780; 1,331; 1,331; 1,331.
  - Number of groups: 55; 55; 55; 34; 34; 34; 21; 21; 21.
  - R-squared (avg): 0.823; 0.855; 0.800.

---

### Appendix VIII Table 8. Baseline: The Effects of Macroprudential Policies on Real GDP Growth

### Estimation setup and sample
- FE timing specification; period sample: 1991Q1 – 2016Q4.
- In addition to country and time fixed effects, includes lagged dependent variable and domestic interest rates (lag).
- Dependent Variable: Real GDP (yoy growth).

### Selected coefficient estimates (cumulative effects after four quarters / other controls)
- MaPP (All, Sum): 0.0651; 0.0466; -0.0158; 0.0289; -0.0414; -0.147**; -0.0670; 0.0427; -0.0179.
- MaPP (Loan, Sum): -0.0143; -0.0729; -0.132; -0.0580; -0.119; -0.255**; 0.0830; -0.0582; -0.147.
- MaPP (Demand, Sum): -0.0903; -0.112; -0.254; -0.0764; -0.131; -0.144; 0.0440; -0.138; -0.337***.
- MaPP (Supply Capital, Sum): -0.292; -0.118; -0.0669; 0.0937; 0.103; -0.439***; -0.826**; -0.360; -0.0104.
- Controls and notable coefficients:
  - dsti: -0.175; -0.191; 0.197; -0.0514; -0.207; 0.826**; 0.775; -0.134; -1.122***.
  - llp: -0.125; -0.482; -0.781***; -0.436; -0.614; -1.284; 0.604*; -0.482; -0.466.
  - capital: -0.155; 0.015; -0.00611; 0.110; 0.162; -0.258; -0.567**; -0.147; -0.0115.
- Observations / Number of groups / R-squared (avg):
  - Observations: 5,416; 5,416; 5,416; 3,125; 3,125; 3,125; 2,912; 2,912; 2,912.
  - Number of groups: 64; 64; 64; 34; 34; 34; 30; 30; 30.
  - R-squared (avg): 0.932; 0.943; 0.920.

---

### Appendix VIII Table 9. Baseline: The Effects of ΔLTV Limits on Real Household Credit Growth

### Estimation setup and sample
- Fixed effects estimation with timing assumption. Reports effects of a one-ppt LTV tightening; the row “Sum coeff LTV level” shows cumulative effects after four quarters.
- Period sample: 2000Q1 – 2016Q4.
- Subsamples/groups: ALL; AE; EMDE; ASIA; EUROPE; AMERICAS; Hi FX Regime Flexibility; Hi Capital Openness; Hi Financial Development; Hi DTI Low-Income Borrowers; Positive Credit Gap; Positive Credit Gap & EMDE only.
- Dependent Variable: Household credit (% change, yoy, real).
- Standard errors clustered at the country level. Confidence levels: *** p<0.01, ** p<0.05, * p<0.1.

### Key coefficients and diagnostics (selected)
- Autoregressive control: Household credit (lag=1q): 0.890***; 0.885***; 0.895***; 0.836***; 0.910***; 0.820***; 0.886***; 0.907***; 0.835***; 0.967***; 0.842***; 0.855***.
- Short-term interest rate (lag=1q): -0.0917***; -0.0992; -0.110***; -0.371; -0.101***; -0.0577*; -0.0902***; -0.0471***; -0.137; -0.101**; -0.148**; -0.145**.
- GDP growth (real, lag=1q): 0.153***; 0.147**; 0.181**; 0.0614; 0.0825; 0.339**; 0.146*; 0.0817; 0.150*; 0.0118; 0.144*; -0.0982.
- LTV level (change) lags (examples):
  - lag=1q: 0.0204; 0.0347; 0.0406; 0.0695*; -0.0342; 0.158**; 0.00833; 0.0292; 0.0592; 0.0464; -0.0529; -0.0570.
  - lag=2q: 0.0392; -0.0180; 0.133***; 0.0792**; 0.0453; 0.0346; 0.0456; 0.0148; 0.0330; 0.0720*; 0.0468; 0.108***.
  - lag=3q: 0.0522; 0.0398; 0.106**; 0.0881***; 0.0508; 0.153; 0.0649; 0.0319; 0.0419; 0.0786; 0.0682; 0.121*.
  - lag=4q: 0.0352; 0.0133; 0.0940***; 0.0631*; 0.0239; 0.109; 0.0537; 0.00418; 0.0343; 0.131**; 0.0569; 0.117***.
- Sum coeff LTV level (cumulative after four quarters) by group:
  - ALL: 0.15
  - AE: 0.07
  - EMDE: 0.37
  - ASIA: 0.30
  - EUROPE: 0.09
  - AMERICAS: 0.45
  - Hi FX Regime Flexibility: 0.17
  - Hi Capital Openness: 0.08
  - Hi Financial Development: 0.17
  - Hi DTI Low-Income Borrowers: 0.33
  - Positive Credit Gap: 0.12
  - Positive Credit Gap & EMDE only: 0.29
- F-test p-value for sum of LTV level = 0 (by group):
  - ALL: 0.19
  - AE: 0.58
  - EMDE: 0.01***
  - ASIA: 0.03**
  - EUROPE: 0.60
  - AMERICAS: 0.32
  - Hi FX Regime Flexibility: 0.28
  - Hi Capital Openness: 0.25
  - Hi Financial Development: 0.21
  - Hi DTI Low-Income Borrowers: 0.03**
  - Positive Credit Gap: 0.43
  - Positive Credit Gap & EMDE only: 0.07*
- Observations / R-squared / Country no (examples):
  - Observations: 3,236; 2,018; 1,262; 674; 1,883; 479; 1,799; 1,790; 1,302; 435; 1,513; 526.
  - R-squared: 0.918; 0.912; 0.934; 0.877; 0.945; 0.893; 0.907; 0.933; 0.854; 0.986; 0.902; 0.940.
  - Country no: 58; 32; 41; 33; 28; 31; 22; 75; 72; 4? (table lists many counts; full listing in table).

---

### Appendix VIII Table 10. Baseline: The Effects of ΔLTV Limits on Real Private Consumption Growth

### Estimation setup and sample
- Fixed effects estimation with timing assumption. Reports effects of a one-ppt LTV tightening; “Sum coeff LTV level” shows cumulative effects after four quarters.
- Period sample: 2000Q1 – 2016Q4.
- Same subgroup structure as Table 9.
- Dependent Variable: Private Consumption (% change, yoy, real).

### Key coefficients and diagnostics (selected)
- Autoregressive control: Private consumption (lag=1q): 0.830***; 0.842***; 0.827***; 0.687***; 0.846***; 0.822***; 0.839***; 0.843***; 0.758***; 0.833***; 0.822***; 0.785***.
- Short-term interest rate (lag=1q): -0.0856***; -0.0825**; -0.0843***; -0.211**; -0.0596***; -0.119; -0.0864***; -0.0714**; -0.0176; -0.0953; -0.112***; -0.133***.
- GDP growth (real, lag=1q): -0.0370; -0.0623; -0.0268; -0.0616; -0.0346; -0.0578; -0.108***; -0.0310; -0.0430; 0.00525; -0.141***; -0.158*.
- LTV level (change) lags (examples):
  - lag=1q: -0.000850; 0.0157; 0.002370; 0.0112; -0.0145; 0.0147; -0.001050; 0.0151; 0.0169; 0.0683; -0.0208; 0.0142.
  - lag=2q: 0.0219; 0.007900; 0.0311; -0.0158; 0.0104; 0.0181; 0.0188; 0.0346; -0.0179; 0.0403; 0.0520*; 0.0773*.
  - lag=3q: 0.0130; 0.0179; 0.004270; 0.008230; 0.0174; 0.158; 0.0311**; 0.006000; 0.0190; -0.002630; 0.003690; 0.00988.
  - lag=4q: 0.0422*; -0.005920; 0.0767**; 0.0396; 0.0398; -0.0294; 0.0259; 0.0312; -0.008450; 0.2710; 0.05000; 0.0766*.
- Sum coeff LTV level (cumulative after four quarters) by group:
  - ALL: 0.08
  - AE: 0.04
  - EMDE: 0.11
  - ASIA: 0.04
  - EUROPE: 0.05
  - AMERICAS: 0.16
  - Hi FX Regime Flexibility: 0.07
  - Hi Capital Openness: 0.09
  - Hi Financial Development: 0.01
  - Hi DTI Low-Income Borrowers: 0.38
  - Positive Credit Gap: 0.08
  - Positive Credit Gap & EMDE only: 0.18
- F-test p-value for sum of LTV level = 0 (by group):
  - ALL: 0.08*
  - AE: 0.48
  - EMDE: 0.09*
  - ASIA: 0.57
  - EUROPE: 0.43
  - AMERICAS: 0.40
  - Hi FX Regime Flexibility: 0.16
  - Hi Capital Openness: 0.27
  - Hi Financial Development: 0.89
  - Hi DTI Low-Income Borrowers: 0.17
  - Positive Credit Gap: 0.21
  - Positive Credit Gap & EMDE only: 0.06*
- Observations / R-squared / Country no (examples):
  - Observations: 3,187; 1,953; 1,360; 693; 1,817; 468; 1,854; 1,764; 1,197; 441; 1,400; 545.
  - R-squared: 0.824; 0.814; 0.841; 0.724; 0.837; 0.874; 0.851; 0.808; 0.778; 0.840; 0.833; 0.858.
  - Country no: 54; 31; 23; 11? (table lists many counts; full listing in table).

---

### Appendix VIII Table 11. Robustness (Control for other MaPPs): Effects of ΔLTV Limits on Real Household Credit Growth

### Estimation setup and sample
- Same as Table 9 but additionally controlling for other macroprudential policy (MPP) actions (lags).
- Dependent Variable: Household Credit (% change, yoy, real).
- Period sample: 2000Q1 – 2016Q4.

### Key coefficients and diagnostics (selected)
- Household credit (lag=1q): 0.898***; 0.910***; 0.894***; 0.841***; 0.911***; 0.820***; 0.886***; 0.906***; 0.900***; 0.965***; 0.867***; 0.855***.
- Short-term interest rate (lag=1q): -0.0857***; -0.101; -0.0995**; -0.413; -0.101***; -0.0525; -0.0859***; -0.0459***; -0.129*; -0.0993**; -0.145**; -0.141**.
- GDP growth (real, lag=1q): 0.147***; 0.112*; 0.201**; 0.0717; 0.0971; 0.345***; 0.156**; 0.0937; 0.105; 0.0144; 0.0747; -0.0909.
- LTV level (change) lags (examples):
  - lag=1q: 0.008080; 0.0337; 0.0194; 0.0629*; -0.0519; 0.163***; 0.0109; 0.0298; 0.0590; 0.0288; -0.0442; -0.0599.
  - lag=2q: 0.0187; -0.0365; 0.100**; 0.0614; 0.0207; 0.0250; 0.0404; 0.0129; 0.002850; 0.0735; 0.0516; 0.0992***.
  - lag=3q: 0.0310; 0.0284; 0.0795*; 0.0687**; 0.0257; 0.158; 0.0526; 0.0234; 0.0107; 0.0587; 0.0641; 0.107*.
  - lag=4q: 0.0215; 0.002610; 0.0797**; 0.0493; 0.0139; 0.0803; 0.0402; 0.00950; -0.000995; 0.132**; 0.0599*; 0.115***.
- Controls: Other MPPs (except LTV) lags (examples):
  - lag=1q: -0.0836; 0.0185; -0.261; -0.251; 0.0383; -0.0660; 0.1080; 0.169; -0.260**; -0.185; 0.296; 0.0323.
  - lag=2q: -0.380**; -0.194*; -0.480**; -0.408; -0.468*; -0.191; -0.164; -0.260; -0.480***; -0.0993; -0.0488; -0.192.
  - lag=3q: -0.233*; -0.153; -0.270; -0.302*; -0.276; 0.118; -0.111; 0.158; -0.430***; 0.0583; -0.0172; -0.206.
  - lag=4q: -0.307*; -0.449**; -0.263; -0.479; -0.188; -0.435; -0.354; -0.568; -0.521**; -0.160; 0.0676; -0.0219.
- Sum coeff LTV level (cumulative after four quarters) by group (robustness specification):
  - ALL: 0.08
  - AE: 0.03
  - EMDE: 0.28
  - ASIA: 0.24
  - EUROPE: 0.01
  - AMERICAS: 0.43
  - Hi FX Regime Flexibility: 0.14
  - Hi Capital Openness: 0.08
  - Hi Financial Development: 0.07
  - Hi DTI Low-Income Borrowers: 0.29
  - Positive Credit Gap: 0.13
  - Positive Credit Gap & EMDE only: 0.26
- F-test p-value for sum of LTV level = 0 (robustness specification):
  - ALL: 0.51
  - AE: 0.84
  - EMDE: 0.04**
  - ASIA: 0.05*
  - EUROPE: 0.96
  - AMERICAS: 0.33
  - Hi FX Regime Flexibility: 0.36
  - Hi Capital Openness: 0.30
  - Hi Financial Development: 0.60
  - Hi DTI Low-Income Borrowers: 0.03**
  - Positive Credit Gap: 0.38
  - Positive Credit Gap & EMDE only: 0.08*
- Observations / R-squared / Countries (examples):
  - Observations: 3,118; 1,900; 1,262; 612; 1,883; 479; 1,799; 1,728; 1,184; 435; 1,455; 526.
  - R-squared: 0.927; 0.938; 0.935; 0.882; 0.945; 0.893; 0.907; 0.934; 0.929; 0.986; 0.926; 0.940.
  - Countries: 56; 32; 24; 12; 32; 83? (table lists many counts; full listing in table).

*Source: The iMaPP database, Bloomberg, BIS, OECD, others (see Appendix IV), and the authors’ estimation.*

### Appendix VIII Table 12. Robustness (Control for Other MaPPs): The Effects of ΔLTV Limits on Real Private

### Appendix VIII Table 12. Robustness (Control for Other MaPPs): The Effects of ΔLTV Limits on Real Private Consumption Growth

### Estimation context
- Period sample: 2000Q1 – 2016Q4.
- Method: Fixed effects estimation with timing assumption. All specifications include country and time fixed effects.
- Dependent variable: Private Consumption (% change, yoy, real).
- Confidence levels: *** p<0.01, ** p<0.05, * p<0.1.
- Standard errors clustered at the country level (reported in parentheses).

### Key coefficient patterns (selected)
- Private consumption (lag=1q) coefficients (columns 1–12):  
  0.832***, 0.845***, 0.828***, 0.681***, 0.847***, 0.824***, 0.839***, 0.847***, 0.766***, 0.823***, 0.821***, 0.789*** (standard errors reported in table).
- Short-term interest rate (lag=1q) notable entries include: -0.0851***, -0.0852**, -0.0843***, -0.281***, -0.0594***, -0.127, -0.0852***, -0.0743**, -0.0305, -0.0937, -0.114***, -0.133***.
- LTV level (change) coefficients by lag (column 1, ALL):  
  - lag=1q: 0.000548 (0.0189)  
  - lag=2q: 0.0242 (0.0203)  
  - lag=3q: 0.0142 (0.0140)  
  - lag=4q: 0.0423* (0.0244)
- Other MPPs (except LTV) display varied coefficients across lags and groups; e.g., Other MPPs (lag=1q) column 2: 0.191* (0.106).

### Cumulative effects and significance (after four quarters)
- Sum coeff LTV level (columns 1–12):  
  0.08, 0.03, 0.13, 0.06, 0.05, 0.20, 0.07, 0.08, 0.00, 0.30, 0.12, 0.21
- F-test p-value for sum being zero (columns 1–12):  
  0.08*, 0.57, 0.08*, 0.47, 0.44, 0.32, 0.14, 0.341, 0.000, 0.28, 0.12, 0.04*
- Observations (columns 1–12):  
  3,124; 1,890; 1,360; 630; 1,817; 468; 1,854; 1,701; 1,134; 441; 1,377; 545
- R-squared (columns 1–12):  
  0.826; 0.818; 0.842; 0.738; 0.838; 0.876; 0.851; 0.813; 0.789; 0.842; 0.834; 0.859
- Countries (columns 1–12):  
  53; 30; 23; 10; 29; 8; 28; 17; 5; 2; 3; 7

---

### Appendix VIII Table 13. Robustness (system GMM): The Effects of ΔLTV Limits on Real Household Credit Growth

### Estimation context
- Period sample: 2000Q1 – 2016Q4.
- Method: Arellano-Bover-Blundell-Bond system GMM with timing assumption. All specifications include country and time fixed effects.
- Dependent variable: Household Credit (% change, yoy, real).
- Reported: AR(1) and AR(2) test p-values, Sum coeff LTV level, F-test p-value for sum being zero. Standard errors clustered at the country level.

### Key coefficient patterns (selected)
- Household credit (lag=1q) coefficients (columns 1–12):  
  0.912***, 0.912***, 0.907***, 0.809***, 0.929***, 0.922***, 0.908***, 0.925***, 0.857***, 0.958***, 0.893***, 0.917***.
- Short-term interest rate (lag=1q) examples: -0.0337**, -0.0753**, -0.0458**, 0.0112, -0.0587***, -0.0236, -0.0252*, -0.0266***, -0.00330, -0.0268, -0.0252, -0.0331.
- LTV level (change) coefficients by lag (column 1, ALL):  
  - lag=1q: 0.0418 (0.0616)  
  - lag=2q: 0.0321 (0.0412)  
  - lag=3q: 0.0472* (0.0287)  
  - lag=4q: 0.0535** (0.0250)

### Cumulative effects and tests
- Sum coeff LTV level (columns 1–12):  
  0.17, 0.10, 0.38, 0.23, 0.05, 0.13, 0.14, 0.12, 0.12, 0.23, 0.13, 0.22
- F-test p-value (columns 1–12):  
  0.15, 0.28, 0.01***, 0.07*, 0.72, 0.80, 0.25, 0.05**, 0.34, 0.11, 0.16, 0.02**
- AR(1) Test p-values (columns 1–12):  
  0.002, 0.064, 0.001, 0.044, 0.034, 0.064, 0.020, 0.113, 0.125, 0.024, 0.005, 0.003
- AR(2) Test p-values (columns 1–12):  
  0.557, 0.838, 0.729, 0.154, 0.0915, 0.812, 0.386, 0.117, 0.294, 0.205, 0.672, 0.482
- Observations (columns 1–12):  
  3,236; 2,018; 1,262; 674; 1,883; 479; 1,799; 1,790; 1,302; 435; 1,513; 526
- Number of countries (columns 1–12):  
  58; 34; 24; 13; 28; 31; 22; 7; 5; 72; 4;  (table reports: 58 34 24 13 28 31 22 7 5 72 4 — presented as in source)

---

### Appendix VIII Table 14. Robustness (System GMM): The Effects of ΔLTV Limits on Real Private Consumption growth

### Estimation context
- Period sample: 2000Q1 – 2016Q4.
- Method: Arellano-Bover-Blundell-Bond system GMM with timing assumption. All specifications include country and time fixed effects.
- Dependent variable: Private Consumption (% change, yoy, real).
- Reported: AR(1) and AR(2) test p-values, Sum coeff LTV level, F-test p-value for sum being zero. Standard errors clustered at the country level.

### Key coefficient patterns (selected)
- Private consumption (lag=1q) coefficients (columns 1–12):  
  0.861***, 0.856***, 0.876***, 0.747***, 0.856***, 0.934***, 0.884***, 0.855***, 0.815***, 0.853***, 0.910***, 0.908***.
- Short-term interest rate (lag=1q) examples: -0.0374**, -0.0480***, -0.0330**, -0.0747*, -0.0424***, -0.0165*, -0.0259**, -0.0400**, 0.0117, -0.0416, -0.0204**, -0.0311**.
- LTV level (change) coefficients by lag (column 1, ALL):  
  - lag=1q: 0.00349 (0.0329)  
  - lag=2q: 0.0170 (0.0188)  
  - lag=3q: 0.00651 (0.0139)  
  - lag=4q: 0.0348 (0.0231)

### Cumulative effects and tests
- Sum coeff LTV level (columns 1–12):  
  0.06, 0.00, 0.12, 0.04, 0.04, -0.06, 0.02, 0.09, -0.01, 0.34, 0.04, 0.15
- F-test p-value (columns 1–12):  
  0.25, 0.99, 0.05*, 0.56, 0.53, 0.76, 0.76, 0.21, 0.90, 0.07*, 0.63, 0.04**
- AR(1) Test (columns 1–12) reported as extremely small p-values for many columns (e.g., 4.68e-07, 7.61e-05, 0.000125, ...).
- AR(2) Test p-values (columns 1–12):  
  0.488, 0.953, 0.424, 0.218, 0.532, 0.194, 0.169, 0.0979, 0.356, 0.608, 0.548, 0.252
- Observations (columns 1–12):  
  3,187; 1,953; 1,360; 693; 1,817; 468; 1,854; 1,764; 1,197; 441; 1,400; 545
- Number of countries (columns 1–12):  
  54; 31; 23; 11; 29; 8; 30; 29; 7; 5; 12; 3

---

### Appendix VIII Table 15. Interactions: The Effects of ΔLTV Limits, Conditional on LTV Level

### Estimation context
- Period sample: 2000Q1 – 2016Q4.
- Method: Fixed effects estimation with timing assumption. Interaction terms with dummies for high and low initial LTV levels.
- Dependent variables: Household Credit and Private Consumption (% change, yoy, real).
- Standard errors clustered at the country level.

### Key coefficients and interactions (selected)
- Household credit (lag=1q) (columns 1–3): 0.889***, 0.884***, 0.890*** (standard errors: 0.0181, 0.0440, 0.0142).
- Private consumption (lag=1q) (columns 4–6): 0.830***, 0.842***, 0.805*** (standard errors: 0.0217, 0.0238, 0.0319).
- Short-term interest rate (lag=1q) examples: -0.0927***, -0.102, -0.106***, -0.0852***, -0.0818**, -0.0955***.
- ΔLTV level (lag = 1q) entries: 0.0204, 0.0295, 0.0120 (household credit columns) and 0.000974, 0.0142, -0.00167 (private consumption columns).
- LTV level (high) * ΔLTV level interactions include values such as -0.0563, 0.0435, -0.00549 (household credit) and -0.00914, -0.0199, 0.0265 (private consumption).

### Cumulative effects after four quarters (Sum coefficients)
- Sum coefficients (LTV level low): 0.12, 0.053, 0.196 (household credit columns 1–3); 0.082, 0.036, 0.133 (private consumption columns 4–6).
- Sum coefficients (LTV level high): 0.173, 0.196, 0.224 (household credit columns 1–3); 0.062, 0.028, 0.041 (private consumption columns 4–6).

### Sample and fit
- Observations (columns): 3,236; 2,018; 1,218; 3,187; 1,953; 1,234 (presented as in source).
- R-squared (columns): 0.918; 0.912; 0.925; 0.824; 0.815; 0.836

---

*Sources: The iMaPP database, Bloomberg, BIS, OECD, others (see Appendix IV), and the authors’ estimation.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019066.pdf_