## 2.1  Household survey data

## Source details

**Canonical URL:** [2.1  Household survey data](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023043-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023043-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023043-print-pdf.pdf.json)

---

### Data sources and sample construction
- Household data source: Household Finance and Consumption Survey (HFCS), coordinated by the ECB. Three survey waves: fieldwork for the first wave in 2010 and 2011; second wave between 2013 and the first half of 2015; third wave in 2017.
- Macroprudential policy data source: ECB MaPPED dataset for 28 European Union countries over the period 1995-2018; includes 11 categories and 53 subcategories of policy instruments, tracks dates of introduction, recalibration, and termination; direction interpreted as tightening, loosening, or ambiguous.
- Outcome of interest: newly issued mortgage loans to households. HFCS asks year mortgage was taken (or refinanced) and its amount, allowing construction of an annual chronology of new mortgage loans for each household.
- Sample construction trade-off: household characteristics available only for the survey year; analysis restricted to mortgage loans within three years of a survey wave and uses household characteristics as measured in the survey year throughout that period.
- Sample size: 4,582 observations (number of observations restricted to the regression sample).

### Sample restrictions and variable definitions
- Loans restricted to mortgages taken to purchase a household’s primary residence, using same property as collateral.
- Excluded: loans that use several properties as collateral; loans on second or third properties.
- LTV definition: ratio of mortgage loan amount to value of collateral property at the time the mortgage was taken.
- LTI definition: ratio of mortgage loan amount to the household’s current annual income.
- DSTI definition: ratio of current yearly payments on the mortgage to current annual income.
- MaPP measure used: number of tightening actions minus loosening actions for each country-year and for each of the 11 policy categories; larger number = tighter macroprudential policy.
- Ambiguous actions: about 19 percent of policy actions denoted as “ambiguous” in MaPPED, excluded from analysis.
- Note: no observation for the MaPP category “leverage ratio” in the sample.

### HFCS summary statistics (selected figures)
- Mean household Age: 41.49 (SD 0.25)
- No Economically Active in HH: 1.71 (SD 0.02)
- HH Gross Income: 51040.66 (SD 925.43) Euros
- Gender: Male 0.72
- Education: At Least some College 0.51
- Employment Status: Employed/Self-employed 0.93
- Initial Amount Borrowed: 110564.52 (SD 1703.79) Euros
- Loan-to-Value Ratio (LTV): 68.55 (SD 0.68) percent
- Length of Loan at Origination: 19.04 (SD 0.25) years
- Current Interest Rate: 2.78 (SD 0.04) percent
- Downpayment as % of House Value: 31.35 (SD 0.72) percent
- Rate Type: Adjustable 0.33
- Loan Type: Refinanced 0.27
- No Obs: 4582.00

### Income and wealth distribution in sample (shares)
- Gross income distribution (shares):
  - 0% - 10%: 0.02
  - 11% - 20%: 0.01
  - 21% - 30%: 0.04
  - 31% - 40%: 0.05
  - 41% - 50%: 0.07
  - 51% - 60%: 0.10
  - 61% - 70%: 0.14
  - 71% - 80%: 0.17
  - 81% - 90%: 0.22
  - 91% - 100%: 0.19
- Note: higher-income households are over-represented in HFCS; survey weights used to correct for oversampling.

### Average loan ratios by income quintile
- LTV, LTI, DSTI by income quintile:
  - 0%-20%: LTV 72.2, LTI 492.0, DSTI 27.9
  - 21%-40%: LTV 72.2, LTI 393.5, DSTI 23.4
  - 41%-60%: LTV 71.6, LTI 301.8, DSTI 18.6
  - 61%-80%: LTV 68.3, LTI 231.0, DSTI 14.9
  - 81%-100%: LTV 61.6, LTI 170.6, DSTI 11.7
- No Obs: LTV 4582.0, LTI 4480.0, DSTI 4148.0
- Pattern: lower-income households on average have higher LTV, LTI, and DSTI than higher-income households; all three ratios monotonically increase with households’ income (as reported in the source).

### MaPPED policy-action variation and focus
- Summary:
  - Two MaPP categories account for most non-zero action variation: levies/taxes on financial institutions (40% of non-zero actions) and minimum capital requirement (30% of non-zero actions).
  - Other categories show limited variability (about 10% loosening and tightening actions combined) or broader coverage not specific to residential lending.
- Authors focus main analysis on:
  - (i) levy/tax on financial institutions (targets lenders’ total assets);
  - (ii) minimum capital requirements (focuses on risk-weighted assets).
  - Rationale: different targets and channels; other categories lack variability or target broader exposures.

### Regression framework for loan amount (baseline)
- Baseline regression (Equation (1)):
  - Y_lict = α H_ict + β MaPP_ct × IncomeDecile_it + γ L_l + λ_ct + ε_lict
  - Indices: l = loan, i = household, c = country, t = year.
  - Outcome Y: amount of newly obtained mortgage credit deflated to 2015 euro (natural log of initial mortgage amount in baseline regressions).
  - H_ict: household controls — income decile within country, net wealth decile, household size measured by number of working household members, gender, age linear and quadratic, education, employment status.
  - L_l: loan characteristics — loan maturity; dummies for adjustable vs fixed rate, refinanced vs new loan, high LTV loans LTV>80.
  - λ_ct: country×year fixed effects (absorbs standalone MaPP term).
- Identification: variation across households within same country and same year.
- Inference: bootstrap standard errors with 1,000 replications using HFCS replicate weights (rescaling bootstrap of Rao and Wu (1988) as specified by Rao, Wu, and Yue (1992)).

### Main loan-size results (interaction coefficients)
- Levy/Tax X HH Income: -0.00181 (0.000719) — ∗∗ (p<0.05)
- Min. Cap. X HH Income: 0.00112 (0.000488) — ∗∗ (p<0.05)
- Observations: 4582
- Interpretation provided:
  - Levy/tax tightening: negative β implies higher-income households on average experience a larger reduction in mortgage loan size than lower-income households when levy/tax tightens.
    - Magnitude: an increase in household income by one decile is associated with a decrease in the mortgage loan amount by about 0.2 percent when levy/tax tightens. Increase in income from bottom decile to top decile implies decrease in average loan of about 2 percent or 2,300 euro.
  - Minimum capital requirement tightening: positive β implies lower-income households on average experience a larger reduction in mortgage loan size than higher-income households when minimum capital requirement tightens.
    - Magnitude: an increase in household income by one decile is associated with an increase in the mortgage loan amount by about 0.1 percent when minimum capital requirement tightens. Increase in income from bottom decile to top decile implies decrease in average loan of about 1 percent or 1,150 euro.
- Controls: loan amount increases with age (hump shape), income, net wealth, education; loan amount increases with maturity and for purchases vs refinancing; LTV>80 dummy positive and significant.

### Channel testing — Cost of borrowing (borrowing-cost / hedging channel)
- Two-step empirical test:
  1. Regression for Rate_lict on MaPP to assess association with loan rate at issuance (Equation (2)).
  2. Re-estimate baseline with down payment (% of collateral value) as dependent variable; test sign of MaPP × IncomeDecile.
- Table 7 results:
  - Column (1) Current rate: Levy/Tax = 0.140 (0.0569) — ∗∗ (p<0.05).
    - Interpretation: A one unit increase in the net tightening measure is associated with an increase the cost of borrowing by 14 basis points.
  - Column (2) Downpayment: Levy/Tax X HH Income = 0.0541 (0.0204) — ∗∗∗ (p<0.01).
    - Interpretation: Higher-income households on average experience a larger increase in their down payment than lower-income households when levy/tax tightens — consistent with a hedging channel.
- Minimum capital requirement:
  - Column (1) Current rate: Min. Cap. = 0.0124 (0.0538), not statistically significant.
    - Interpretation: No evidence that tightening minimum capital requirements raises overall costs of borrowing for households in this sample/time period.

### Channel testing — Risk-taking channel
- Hypothesis: Banks could respond to higher lending costs by increasing risk exposure (reallocating credit toward lower-income, higher-LTV borrowers).
- Test 1: Intensive margin (regression (3)), sample split 0-40th percentile vs 41-100th percentile.
  - Table 9:
    - For income>40th percentile: Levy/Tax = 1.667 (0.944) — ∗ (p<0.10). Medium-/high-income households tend to increase down payments during tightening.
    - For income≤40th percentile: Levy/Tax = -0.862 (1.464), not significant.
  - Conclusion: Results go against the risk-taking channel and support the borrowing-cost / hedging channel.
- Test 2: Extensive margin (LowIncomeHighLTV dummy):
  - Table 10: Levy/Tax = -0.0130 (0.0134), not significant.
  - Conclusion: No evidence banks increase the number of riskier loans to low-income, high-LTV households after tightening; further evidence against risk-shifting.

### Channel testing — Flight-to-quality (minimum capital requirement)
- Hypothesis: Tighter minimum capital requirements induce banks to reduce risky loan exposure, disproportionately affecting lower-income households with higher-LTV loans.
- Subsample regressions by LTV:
  - Table 11:
    - Column (1) Sample LTV≤100: Min. Cap. X HH Income = 0.00112 (0.000488) — ∗∗ (p<0.05).
    - Column (2) Sample LTV∈[50,100]: Min. Cap. X HH Income = 0.00137 (0.000475) — ∗∗∗ (p<0.01).
    - Column (3) Sample LTV<50: Min. Cap. X HH Income = 0.000670 (0.000540), not significant.
  - Interpretation: Interaction coefficient increases and becomes more significant in the high-LTV subsample, consistent with lenders contracting risky loans under higher capital requirements and low-income households experiencing larger reductions in mortgage loans in the high-risk segment.

### Differential effects on house prices
- Micro-level regressions (log value of house at purchase) show no robust evidence that levies/taxes or minimum capital requirements systematically affect micro-level purchase prices.
- Aggregate regressions by income group (regression (4)) — levies/taxes:
  - Tightening associated with households in bottom income bracket buying less expensive houses; result significant at 10 percent level.
  - Top 20th percentile appear to buy more expensive houses following tightening, but not statistically significant.
  - Interpretation: Lower-income households hedge higher borrowing costs by buying less expensive houses.
- Minimum capital requirements:
  - Signs suggest lower-income households buy less expensive houses and higher-income households (60th percentile and above) buy more expensive houses following tightening; results not significant at conventional levels but consistent with flight-to-quality.
- Overall: Evidence on house prices weaker than loan/down payment results but directionally consistent with identified channels.

### Robustness tests
- Interaction controls: adding interactions of MaPP with age, net wealth, gender, employment status, education — income interaction coefficients retain sign and significance.
- Monetary policy control:
  - Monetary policy shock measure: OIS yield with 3-month maturity (monetary event window).
  - Robustness table (Appendix Table19):
    - Levy/Tax X HH Income: -0.00137 (0.000784) — *
    - Min. Cap. X HH Income: 0.00132 (0.000679) — *
    - Monetary policy Shock X HH Income (column 1): 0.00009430 (0.000105), not significant
    - Monetary policy Shock X HH Income (column 2): 0.000143 (0.0000942), not significant
    - Observations: 4435 (both columns)
  - Interpretation: effects are robust and not driven by monetary policy; monetary policy interaction coefficients positive but not significant.
- Controlling for other MaPP policies:
  - Baseline augmented with interaction of income and other MaPP categories and with aggregate MaPP index — levy/tax and minimum capital requirement income interactions remain statistically significant with same signs.
- Selected robustness coefficient patterns for Levy/Tax X HH Income (various specifications):
  - Examples: -0.00161 (0.000757) ∗∗; -0.00181 (0.000719) ∗∗; -0.00185 (0.000711) ∗∗∗; repeated specifications show stability.
- Selected robustness coefficient patterns for Min. Cap. X HH Income (various specifications):
  - Examples: 0.00263 (0.000613) ∗∗∗; 0.00107 (0.000493) ∗∗; 0.00112 (0.000483) ∗∗; repeated specifications show stability.

### Main empirical findings
- Higher-income households on average experience a larger reduction in mortgage loan size than lower-income households when levy/tax on financial institutions tightens.
- Lower-income households on average experience a larger reduction in mortgage loan size than higher-income households when minimum capital requirements tighten.
- Evidence is consistent with a flight-to-quality effect on mortgage lending when minimum capital requirement tightens; evidence supports borrowing-cost / hedging channel for levies/taxes.
- Baseline results robust to controls for other MaPP categories and to inclusion of monetary policy shocks.

### Policy implications and suggested research directions
- Results illuminate the effectiveness of lender-based macroprudential tools and their uneven impacts across income groups.
- Equity implications of macroprudential regulation identified as "an important and fruitful venue for further research."

### Data cleaning and regression setup (procedural steps)
- HFCS preprocessing: merge imputations, reduce memory, consolidate waves into mortgage-level dataset; keep first mortgage on primary residence and up to 2 additional properties; define took_loan and loan_for variables; keep mortgages taken within 3 years of household survey.
- MaPP dataset: use implementation dates; count tightenings and loosenings by MaPP category to obtain net_tightening variables.
- Merge HFCS and MaPP on country-year; add macrovariables (GDP growth, CPI, unemployment rate) from WEO; add real house price growth from BIS; rebase CPI to 2015.
- Baseline regressions: deflate nominal variables; take natural log of initial mortgage amount; drop implausible age/education values; categorize age, education, employment; construct LTV, LTI, DSTI; winsorize specified variables; standardize income, wealth, financial variables; remove observations with LTV > 100 or DSTI > 80.

*Excerpt from IMF working paper (section 2.1 and associated tables) as provided in the supplied content.*

### 2.1  Household survey data

### 2.1  Household survey data

### Data sources and sample construction
- Household data source: Household Finance and Consumption Survey (HFCS), coordinated by the ECB. Three survey waves: fieldwork for the first wave in 2010 and 2011; second wave between 2013 and the first half of 2015; third wave in 2017.
- Macroprudential policy data source: ECB MaPPED dataset for 28 European Union countries over the period 1995-2018; includes 11 categories and 53 subcategories of policy instruments, tracks dates of introduction, recalibration, and termination; direction interpreted as tightening, loosening, or ambiguous.
- Outcome of interest: newly issued mortgage loans to households. HFCS asks year mortgage was taken (or refinanced) and its amount, allowing construction of an annual chronology of new mortgage loans for each household.
- Sample construction trade-off: household characteristics (e.g., household income, household size) available only for the survey year. Authors choose to restrict analysis to mortgage loans within three years of a survey wave (e.g., for a household surveyed in 2015, include loans during 2012-2015) and use household characteristics as measured in the survey year throughout that period.
- Sample size: 4,582 observations (number of observations restricted to the regression sample).

### Sample restrictions and definitions
- Loans restricted to mortgages taken to purchase a household’s primary residence, using same property as collateral.
- Excluded: loans that use several properties as collateral; loans on second or third properties.
- LTV definition: ratio of mortgage loan amount to value of collateral property at the time the mortgage was taken.
- LTI definition: ratio of mortgage loan amount to the household’s current annual income.
- DSTI definition: ratio of current yearly payments on the mortgage to current annual income.
- MaPP measure used in analysis: number of tightening actions minus loosening actions for each country-year and for each of the 11 policy categories; larger number = tighter macroprudential policy.
- Ambiguous actions: about 19 percent of policy actions denoted as “ambiguous” in MaPPED, excluded from analysis.
- Note: no observation for the MaPP category “leverage ratio” in the sample.

### HFCS summary statistics (selected figures from Table 1)
- Mean household Age: 41.49 (SD 0.25)
- No Economically Active in HH: 1.71 (SD 0.02)
- HH Gross Income: 51040.66 (SD 925.43) Euros
- Gender: Male 0.72
- Education: At Least some College 0.51
- Employment Status: Employed/Self-employed 0.93
- Initial Amount Borrowed: 110564.52 (SD 1703.79) Euros
- Loan-to-Value Ratio (LTV): 68.55 (SD 0.68) percent
- Length of Loan at Origination: 19.04 (SD 0.25) years
- Current Interest Rate: 2.78 (SD 0.04) percent
- Downpayment as % of House Value: 31.35 (SD 0.72) percent
- Rate Type: Adjustable 0.33
- Loan Type: Refinanced 0.27
- No Obs: 4582.00

### Income and wealth distribution in sample (Table 2, shares)
- Gross income distribution (shares):  
  - 0% - 10%: 0.02  
  - 11% - 20%: 0.01  
  - 21% - 30%: 0.04  
  - 31% - 40%: 0.05  
  - 41% - 50%: 0.07  
  - 51% - 60%: 0.10  
  - 61% - 70%: 0.14  
  - 71% - 80%: 0.17  
  - 81% - 90%: 0.22  
  - 91% - 100%: 0.19  
- Net wealth distribution (shares) reported in same table with similar cell structure.
- Note: higher-income households are over-represented in HFCS; survey weights used in analysis to correct for oversampling of high-income households.

### Average ratios by income quintile (Table 3)
- LTV, LTI, DSTI by income quintile:
  - 0%-20%: LTV 72.2, LTI 492.0, DSTI 27.9
  - 21%-40%: LTV 72.2, LTI 393.5, DSTI 23.4
  - 41%-60%: LTV 71.6, LTI 301.8, DSTI 18.6
  - 61%-80%: LTV 68.3, LTI 231.0, DSTI 14.9
  - 81%-100%: LTV 61.6, LTI 170.6, DSTI 11.7
- No Obs: LTV 4582.0, LTI 4480.0, DSTI 4148.0
- Pattern: lower-income households have on average higher LTV, LTI, and DSTI ratios than higher-income households; all three ratios monotonically increase with households’ income (as reported in the source).

### MaPPED policy-action variation and focus
- Table 4 and Table 5 summarize counts and shares of loosening (<0), neutral (0), and tightening (>0) actions for MaPP categories.
- Two categories exhibit most variation in non-zero actions: levies/taxes on financial institutions and minimum capital requirement — these two categories account for 40% and 30% percent of non-zero actions, respectively (as reported).
- Share table highlights limited variability in other categories (about 10% loosening and tightening actions combined).
- Authors focus main analysis on: (i) levy/tax on financial institutions (targets lenders’ total assets), and (ii) minimum capital requirements (focuses on risk-weighted assets). Rationale: different targets and channels; other categories suffer lack of variability or broader coverage (e.g., LTV/LTI limits not specific to residential loans and include commercial real estate and auto loans; soft vs. hard limits differ across countries).

### Regression framework for loan amount (equation summary)
- Baseline regression (Equation (1)):
  - Y_lict = α H_ict + β MaPP_ct × IncomeDecile_it + γ L_l + λ_ct + ε_lict
  - Indices: l = loan, i = household, c = country, t = year.
  - Outcome Y: amount of newly obtained mortgage credit deflated to 2015 euro.
  - H_ict: vector of household socioeconomic and demographic characteristics (income decile within country, net wealth decile, household size measured by number of working household members, gender, age linear and quadratic, education, employment status).
  - L_l: loan characteristics (loan maturity; dummies for adjustable vs fixed rate, refinanced vs new loan, high LTV loans LTV>80).
  - λ_ct: country×year fixed effects.
  - Identification from variations across households within same country and same year; standalone MaPP term absorbed by country×year fixed effects.
- Standard errors: bootstrap standard errors with 1,000 replications using HFCS replicate weights (rescaling bootstrap of Rao and Wu (1988) as specified by Rao, Wu, and Yue (1992)).

### Main loan-size results (Table 6)
- Coefficients on interaction term MaPP × HH Income (only main coefficients shown):
  - Levy/Tax X HH Income: -0.00181 (0.000719) — ∗∗ (p<0.05)
  - Min. Cap. X HH Income: 0.00112 (0.000488) — ∗∗ (p<0.05)
- Observations: 4582
- Interpretation provided in source:
  - Levy/tax on financial institutions tightening: negative β implies higher-income households on average experience a larger reduction in mortgage loan size than lower-income households when levy/tax tightens.
    - Magnitude interpretation: an increase in household income by one decile is associated with a decrease in the mortgage loan amount by about 0.2 percent when levy/tax tightens. Increase in income from bottom decile to top decile implies decrease in average loan of about 2 percent or 2,300 euro.
  - Minimum capital requirement tightening: positive β implies lower-income households on average experience a larger reduction in mortgage loan size than higher-income households when minimum capital requirement tightens.
    - Magnitude interpretation: an increase in household income by one decile is associated with an increase in the mortgage loan amount by about 0.1 percent when minimum capital requirement tightens. Increase in income from bottom decile to top decile implies decrease in average loan of about 1 percent or 1,150 euro.
- Controls and robustness (summary from source): full regression output in Appendix Table 12; controls have expected signs and majority statistically significant. Loan amount increases with age (hump shape), income, net wealth, education; loan amount increases with maturity and for purchases vs refinancing; LTV>80 dummy positive and significant.

### Channel testing (overview)
- Hypothesized channel for levy/tax on financial institutions: borrowing-cost channel. Tightening (higher levy/tax) makes large loan portfolios more costly for lenders; lenders may pass cost to borrowers via higher loan rates. Higher-income households with more liquidity can hedge higher borrowing costs by substituting larger loans with larger down payments.
- Two-step empirical test (as described):
  1. Estimate regression (Equation (2)) for Rate_lict = α H_ict + β MaPP_ct + γ L_l + δ GDP_ct + σ_c + τ_t + ε_lict, where Rate_lict is loan rate at issuance, GDP_ct is country real GDP growth, σ_c and τ_t are country and time fixed effects. Main variable: levy/tax on financial institutions to assess association with borrowing cost.
  2. Re-estimate baseline (Equation (1)) with down payment (percent of collateral value) as dependent variable and test sign of β on MaPP × IncomeDecile; β >0 would suggest higher-income households increase down payment relative to low-income households when levy/tax tightens.
- Results for these tests: Table 7 and full regression results referenced in Appendix (not included in provided excerpt).

*Italic source attribution: Excerpt from IMF working paper (section 2.1 and associated tables) as provided in the supplied content.*

### 13. The first column reports theβcoefficient from regression (2), and the second

### 13. The first column reports theβcoefficient from regression (2), and the second column in the table reports theβcoefficient on the interaction term on macroprudential policy and household income in (1).

### Cost of borrowing channel (Section 3.2.1)
- Table 7: Loan Rate & Down payment Regression
  - Column (1) Current rate: Levy/Tax = 0.140 (0.0569), significant at ∗∗ (p<0.05).
    - Interpretation: A one unit increase in the net tightening measure is associated with an increase the cost of borrowing by 14 basis points.
  - Column (2) Downpayment: Levy/Tax X HH Income = 0.0541 (0.0204), significant at ∗∗∗ (p<0.01).
    - Interpretation: Higher-income households on average experience a larger increase in their down payment than lower-income households when the regulation on levy/tax on financial institutions tightens — consistent with a hedging channel.
- Minimum capital requirements as macroprudential measure (Table 8)
  - Column (1) Current rate: Min. Cap. = 0.0124 (0.0538), not statistically significant.
    - Interpretation: No evidence that a tightening in minimum capital requirements raises the overall costs of borrowing for households in this sample/time period.

### Risk-taking channel (Section 3.2.2)
- Hypothesis: Banks could respond to higher lending costs by increasing risk exposure (reallocating credit toward lower-income, higher-LTV borrowers), which would produce differential effects driven by the lower end of the income distribution.
- Test 1: Intensive margin of credit reallocation (regression (3))
  - Sample split: 0-40th percentile (low-income) vs 41-100th percentile (medium-/high-income).
  - Table 9: Testing the Intensive Margin of Credit Re-allocation
    - For income>40th percentile (medium-/high-income): Levy/Tax = 1.667 (0.944), significant at ∗ (p<0.10).
      - Interpretation: Medium-/high-income households tend to increase down payments during a policy tightening.
    - For income≤40th percentile (low-income): Levy/Tax = -0.862 (1.464), not significant.
      - Interpretation: No statistically significant change in down payments for low-income households during tightening episodes.
  - Conclusion: Results go against the risk-taking channel and support the borrowing cost / hedging channel.
- Test 2: Extensive margin of credit reallocation
  - Dependent variable: LowIncomeHighLTV dummy (household in 0-40th percentile and LTV > 50).
  - Table 10: Testing the Extensive Margin of Credit Re-allocation
    - Levy/Tax = -0.0130 (0.0134), not significant.
      - Interpretation: Following a tightening in levies/taxes, the likelihood of banks lending to low-income households with high-LTV does not increase significantly.
  - Conclusion: No evidence that banks increase the number of riskier loans to low-income households after tightening; further evidence against the risk-taking shifting channel.

### Flight-to-quality channel (Section 3.2.3)
- Hypothesis: Tighter minimum capital requirements induce banks to reduce risky loan exposure, disproportionately affecting lower-income households who hold higher-risk loans (higher LTV).
- Test: Run regression (1) on subsamples split by LTV ratios (>50 and <50).
- Table 11: Loan Amounts: Splitting the Sample (New Loan (ln))
  - Column (1) Sample LTV≤100: Min. Cap. X HH Income = 0.00112 (0.000488), significant at ∗∗ (p<0.05).
  - Column (2) Sample LTV∈[50,100]: Min. Cap. X HH Income = 0.00137 (0.000475), significant at ∗∗∗ (p<0.01).
  - Column (3) Sample LTV<50: Min. Cap. X HH Income = 0.000670 (0.000540), not significant.
  - Interpretation: The interaction coefficient increases and becomes more significant in the high-LTV subsample, consistent with lenders contracting risky loans under higher capital requirements and low-income households experiencing a larger reduction in mortgage loans in the high-risk segment.

### Differential effects on house prices (Section 3.3)
- Micro-level regressions (log value of house at purchase) show no robust evidence that lender-based macroprudential tools (levies/taxes or minimum capital requirements) systematically affect the price of the property households purchase at the micro level.
- Aggregate-level regressions by income group (regression (4): HousePrice_jct = β_j MaPP_ct + γ GDP_ct + σ_c + τ_t + ε_ct)
  - Levies/taxes (left panel in Figure 1):
    - A tightening in levies/taxes is associated with households in the bottom income bracket buying less expensive houses; result significant at 10 percent level.
    - Households in the top 20th percentile appear to buy more expensive houses following a tightening, but results are not statistically significant.
    - Interpretation: Lower-income households hedge against higher borrowing costs from tighter levies/taxes by buying less expensive houses.
  - Minimum capital requirements (right panel in Figure 1):
    - Signs suggest lower-income households buy less expensive houses and higher-income households (60th percentile and above) buy more expensive houses following tightening; results are not significant at conventional levels but consistent with flight-to-quality channel.
- Overall: Evidence on house prices is weaker than loan/down payment results but consistent with lower-income households hedging via cheaper purchases under levy/tax tightening and suggestive (though weak) flight-to-quality effects for capital requirements.

### Robustness tests (Section 4)
- Controlling for interactions of macroprudential policy with other household characteristics:
  - Added interactions one at a time between macroprudential policy and age, net wealth, gender, employment status, education level.
  - Results (Table 17 and Table 18 in appendix): Coefficients of the interaction of macroprudential policy and income retain sign and statistical significance for both levies/taxes and minimum capital requirements.
  - Interpretation: Baseline differential effects by household income are robust to inclusion of other household characteristics.
- Controlling for potential differential effects arising from monetary policy:
  - Reran baseline regressions for loan amount including interaction between a monetary policy shock and household income.
  - Monetary policy shock measure: OIS yield with 3-month maturity (monetary event window), constructed by Altavilla et al. (2019).

*Source: wpiea2023043-print-pdf (chapter/section content provided).*

### Appendix Table19reports the results. The coefficients on the interaction

### Appendix Table19reports the results.

### Robustness: MaPP versus Monetary Policy
- Levy/Tax X HH Income: -0.00137
  - Standard error: (0.000784)
  - Significance: *
- Min. Cap. X HH Income: 0.00132
  - Standard error: (0.000679)
  - Significance: *
- Monetary policy Shock X HH Income (column 1): 0.00009430
  - Standard error: (0.000105)
  - Not significant
- Monetary policy Shock X HH Income (column 2): 0.000143
  - Standard error: (0.0000942)
  - Not significant
- Observations: 4435 (column 1) and 4435 (column 2)
- Country X Time FE: Yes (both columns)
- Interpretation from text: "The interaction of household income with monetary policy indicates that the effects of a monetary policy tightening are less effective for high-income households as the coefficient on the interaction term is positive, but they are not significant. Therefore our baseline results are robust and not driven by monetary policy."

### Robustness: Controlling for Other Macroprudential Policies
- Approach: baseline specification augmented one at a time with interaction of household income and other MaPP listed in Table4; also a regression controlling for interaction of household income and the aggregate MaPP index.
- Result summary: "Across all regressions, the coefficient on the interaction term between income and levies/taxes on financial institutions as well as minimum capital requirements remain statistically significant with the same signs as before. Other macroprudential policies may additionally yield differential effect as their interaction with household income is significant in some cases, but importantly our results on the levy/tax on financial institutions and minimum capital requirements are unaffected."

### Robustness Check IV: Differential Effects of Other MaPP (Levy/Tax) — Selected coefficients
- Levy/Tax X HH Income (various specifications): 
  - -0.00161 (0.000757) ∗∗
  - -0.00181 (0.000723) ∗∗
  - -0.00181 (0.000719) ∗∗
  - -0.00167 (0.000709) ∗∗
  - -0.00185 (0.000711) ∗∗∗
  - -0.00181 (0.000719) ∗∗
  - -0.00181 (0.000721) ∗∗
  - -0.00181 (0.000719) ∗∗
  - -0.00161 (0.000727) ∗∗
- Other MaPP interactions reported:
  - All Mapp X HH Income: -0.000298 (0.000284) [not significant]
  - Liqu. Requirements X HH Income: -0.00464 (0.00121) ∗∗∗
  - Risk Weights X HH Income: -0.000495 (0.000885) [not significant]
  - Cap. Buff. X HH Income: -0.00248 (0.000690) ∗∗∗
  - Lending Stds. Rest. X HH Income: -0.00177 (0.000877) ∗∗
  - Cred. Growth & Vol. X HH Income: 0.00117 (0.00453) [not significant]
  - expo. & concentration X HH Income: -0.0000281 (0.00146) [not significant]
  - Loan-Loss X HH Income: 0.00253 (0.00331) [not significant]
  - Min. Cap. X HH Income: 0.000942 (0.000508) ∗

- Sample size: Obs 4582 for all specifications
- Fixed effects: Country X Time FE: Yes

### Robustness Check V: Differential Effects of Other MaPP (Minimum Capital Requirements) — Selected coefficients
- Min. Cap. X HH Income (various specifications):
  - 0.00263 (0.000613) ∗∗∗
  - 0.00107 (0.000493) ∗∗
  - 0.000942 (0.000508) ∗
  - 0.00112 (0.000483) ∗∗
  - 0.00127 (0.000476) ∗∗∗
  - 0.00129 (0.000467) ∗∗∗
  - 0.00112 (0.000488) ∗∗
  - 0.00112 (0.000491) ∗∗
  - 0.00116 (0.000492) ∗∗
- Other MaPP interactions reported:
  - All Mapp X HH Income: -0.00136 (0.000333) ∗∗∗
  - Liqu. Requirements X HH Income: -0.00440 (0.00121) ∗∗∗
  - Levy/Tax X HH Income: -0.00161 (0.000727) ∗∗
  - Risk Weights X HH Income: -0.000497 (0.000918) [not significant]
  - Cap. Buff. X HH Income: -0.00285 (0.000645) ∗∗∗
  - Lending Stds. Rest. X HH Income: -0.00193 (0.000892) ∗∗
  - Cred. Growth & Vol. X HH Income: 0.000617 (0.00459) [not significant]
  - expo. & concentration X HH Income: -0.000114 (0.00148) [not significant]
  - Loan-Loss X HH Income: 0.00351 (0.00345) [not significant]
- Sample size: Obs 4582 for all specifications
- Fixed effects: Country X Time FE: Yes

### Main empirical findings (from conclusion and supported by tables)
- Higher-income households on average experience a larger reduction in mortgage loan size than lower-income households when regulation on levy/tax on financial institutions tightens.
- Lower-income households on average experience a larger reduction in mortgage loan size than higher-income households when regulatory minimum capital requirements tighten.
- Evidence is consistent with flight-to-quality on mortgage lending when the minimum capital requirement tightens.
- Baseline results are robust to controls for other MaPP categories and to inclusion of monetary policy shocks; monetary policy interaction coefficients are positive but not statistically significant.

### Policy implications and suggested research directions
- The results shed light on the overall effectiveness of macroprudential regulation and its potential uneven impact across income groups.
- Equity implications of macroprudential regulation are highlighted as "an important and fruitful venue for further research."

### Data cleaning and regression setup (key procedural steps)
- HFCS preprocessing: merge imputations, reduce memory, consolidate waves into mortgage-level dataset; keep first mortgage on primary residence and up to 2 additional properties; define took_loan and loan_for variables; keep mortgages taken within 3 years of household survey.
- MaPP dataset: use implementation dates; count tightenings and loosenings by MaPP category to obtain net_tightening variables.
- Merge HFCS and MaPP on country-year; add macrovariables (GDP growth, CPI, unemployment rate) from WEO; add real house price growth from BIS; rebase CPI to 2015.
- Baseline regressions: deflate nominal variables; take natural log of initial mortgage amount; drop implausible age/education values; categorize age, education, employment; construct LTV, LTI, DSTI; winsorize specified variables; standardize income, wealth, financial variables; remove observations with LTV > 100 or DSTI > 80.

*Differential Effects of Macroprudential Policy Working Paper No. WP/23/43*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023043-print-pdf.pdf_