## wpiea2023247-print-pdf

## Source details

**Canonical URL:** [wpiea2023247-print-pdf](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023247-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023247-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023247-print-pdf.pdf.json)

---

### Background and motivation
- During the short but deep COVID-19 recession, house prices rose at record levels in many countries due to demand and supply factors.
- Demand-side drivers:
  - Policy support measures that sustained income.
  - Financing conditions remaining favorable.
  - Many households searching for more space.
- Supply-side drivers:
  - Constraints on new construction.
  - Constraints on mobility, limiting available properties.
- Concerns:
  - Financial risks perceived as limited due to post-GFC policies, but housing affordability concerns were raised.
  - Affordability was already a policy concern prior to the pandemic as prices resumed upward trends after the correction/pause around the Global Financial Crisis (GFC).

### Limitations of existing affordability metrics
- Price-to-income ratio (PIR):
  - Available from BIS and OECD as an index comparing house prices to income per capita.
  - Downside: does not incorporate mortgage market or housing characteristics (e.g., mortgage interest rate changes affect monthly payments).
- OECD housing cost overburden rate:
  - Captures percentage of households spending over 40% of income on housing.
  - Does not consider whether a typical family that prefers ownership can qualify for a mortgage given prevailing prices and interest rates.
- BIS debt service ratio:
  - Measures burden of debt on household income but does not specifically capture housing affordability.
- Country-specific indices:
  - Incorporate mortgage market or housing characteristics but methodologies differ, creating comparability challenges.

### Objective and definition of the new HAI
- Aim: construct a housing affordability index (HAI) that:
  - Covers a large panel of countries over a long time period.
  - Incorporates mortgage and housing market aspects such as cost of financing and typical home size.
- Sample and period:
  - 40 countries for the period between 1970Q1 and 2021Q4.
- Affordability definition:
  - The ability of a median-income household to qualify for a mortgage loan needed to purchase an average-priced home within a country (median household focus).
- Methodological benchmark:
  - Mainly follow National Association of Realtors (NAR) methodology and extend to 39 additional countries.
  - Collected variables: (i) house price level (price per square meter), (ii) median household income, (iii) mortgage rates, (iv) typical loan-to-value (LTV) ratios, (v) average term-to-maturity of mortgage loans, (vi) average property size, (vii) average household size.
  - Methodological distinction: use average house price rather than median due to cross-country data availability.

### Data sources and construction approach
- House price level:
  - Use Bricongne et al. (2019) for average cost per square meter (in euros) for an unbalanced panel of 40 OECD countries over 1970–2017.
  - Extend by selecting 2015 as a base year, converting to local currency using 2015 exchange rate vis-à-vis the euro, and applying BIS quarterly nominal growth rates to generate nominal prices per square meter.
  - Coverage spans 1970Q1 to 2021Q4 for countries with maximum observations; minimum coverage at least 2010Q1 to 2021Q4 for Poland and Turkey.
- Mortgage lending rates:
  - Residential mortgage rates from multiple sources: ECB for 18 countries; Haver Analytics for the remainder (except Iceland).
  - Iceland: general interest rate from Central Bank of Iceland adjusted for inflation used as mortgage proxy.
  - When mortgage rate series are shorter than house price series, back-cast mortgage rates by regressing available mortgage rate on contemporaneous lending rate (long-term lending rate or policy rate) and using predicted values for missing periods.
  - Long-term borrowing rates have a median R-squared of 0.82 (Appendix note).
- LTV ratios and maturities:
  - LTV: data for 30 out of 40 countries from Hypostat 2020; Greece, Hungary, Norway from Cerutti, Dagher, and Dell'Ariccia (2017); remaining from national or international sources.
  - Maturities: data for 31 countries from Hypostat 2020; Cerutti, Dagher, and Dell'Ariccia (2017) for Greece, Hungary, Latvia, Malta; remaining from national or international sources.
- Property and household size:
  - Average property size: Eurostat for 29 countries (measured as of 2012); other countries from national statistics or literature.
  - Average household size: Eurostat for 33 countries, United Nations for 6 countries, PRB for Russia. UN data cover 2013–2016 and household size treated as relatively constant over time.
- Median household income:
  - Eurostat provides median household income for 32 countries; remaining 8 countries use national sources, CEIC, or OECD Income Distribution Database.
  - Exceptions where income is not reported as equalized: Canada, Hong Kong SAR, South Korea, United States.
  - For most economies, equalized income multiplied by average household size to recover household-level income.
  - Median household income mostly annual; quarterlies generated by linear interpolation.
  - Time series gaps extended using GDP per capita growth rates from IMF’s World Economic Outlook Database, assuming median household income grows with GDP per capita.

### Challenges, assumptions, and remedies
- Data challenges:
  - Lack of house price levels in local currency; limited centralized databases for mortgage rates, LTVs, and maturities.
- Remedies:
  - Use Bricongne et al. (2019) levels and extend with BIS index growth rates.
  - Draw variables from multiple sources; extend series using linear interpolation and back-casting.
  - Adopt proxies where necessary (e.g., Iceland mortgage proxy).
- Caveats:
  - Constructed series depend on imputed data and mismeasurement cannot be ruled out.
  - Choice of base year (2015) is somewhat arbitrary but indices robust to base year selection.
  - Assuming median household income grows with GDP per capita could overestimate affordability if income distribution becomes more skewed.

### Methodology overview and construction steps
- HAI combines: price of a typical house, income of a typical household, average mortgage rates, typical LTV ratio, typical mortgage maturity.
- Building blocks:
  1. Price of a typical house (HP)
     - HP = price per square meter × average size of a home.
  2. Monthly mortgage payment (PMT)
     - PMT = HP * LTV * (IR/12) * [1 − (1 + IR/12)^(−MA)]^(−1),
     - where IR denotes the annualized mortgage rate and MA denotes the maturity (in months).
     - Emphasis on typical LTVs rather than regulatory LTVs.
     - Observation: while a higher LTV can decrease affordability via larger servicing costs, a lower down payment requirement could enable more households to consider ownership.
  3. Necessary annual income to qualify (QINC)
     - QINC = PMT * 4 * 12,
     - Assumption: monthly mortgage payment should not exceed a quarter of the household’s monthly income.
     - This rule of thumb aligns with the OECD benchmark referenced: housing considered affordable if households do not spend more than 30% of their gross income on housing costs including insurance and taxes.
- Key numeric/procedural points:
  - Eurostat coverage: 32 countries; other sources: 8 countries.
  - Interpolation: annual-to-quarterly by linear interpolation.
  - Qualification rule: monthly mortgage payment ≤ a quarter of monthly household income.
  - Extensions of income series use GDP per capita growth rates from IMF WEO Database.

### Definition and interpretation of the HAI
- HAI formula:
  - HAI = (MEDINC / QINC) * 100, where MEDINC is the median household income.
- Interpretation:
  - HAI = 100: median-income household has exactly enough income to qualify for a mortgage on an average-priced home.
  - HAI > 100: median-income household has more than the qualifying income.
  - HAI < 100: median-income household does not have sufficient income to qualify.

### Validation and comparison with NAR index (United States)
- Methodology: follow NAR methodology; compare computed HAI to NAR-reported HAI for the United States.
- Key statistics (1970q1-2021q4):
  - Our index mean: 122.8
  - NAR index mean: 127.5
  - Our index standard deviation: 27.8
  - NAR index standard deviation: 30.1
  - Correlation between indices: 0.96
- Sources of slight differences:
  - NAR uses median house price; this study uses average house price per square meter.
  - Both use Census median household income annual data interpolated to quarterly and forecast/back-cast using nominal GDP growth rate.
  - NAR uses effective mortgage rate from Federal Housing Finance Board; this study draws mortgage rates from Haver (definitions vary).

### Sample composition, summary statistics, and dynamics
- Sample:
  - Total countries: 40
  - Advanced economies (AEs): 33
  - Emerging markets (EMs): 7
- Data coverage:
  - AEs: coverage tends to start in the 1970s.
  - EMs: coverage tends to start in the 2000s.
- Summary statistics (2001q1 to 2021q4):
  - Mean HAI in AEs: 118
  - Mean HAI in EMs: 85
  - Standard deviation in AEs: 48
  - Standard deviation in EMs: 53
- Time patterns:
  - 1970s–mid-1990s: median HAI below 100; even the 75th percentile below 100.
  - 2000s: cross-country median HAI exceeded 100.
  - Global Financial Crisis (GFC): affordability collapsed but recovered quickly.
  - Since 2010: median HAI remained comfortably above 100; 25th percentile hovered just below 100.
- Quantitative summary (Table 1: Summary statistics of HAI):
  - EM: Number of observations 440; 25th percentile 40; Median 71; 75th percentile 114; Mean 85; Standard deviation 53
  - AE (all): Number of observations 4,890; 25th percentile 66; Median 94; 75th percentile 130; Mean 100; Standard deviation 46
  - AE (after 2001): Number of observations 2,572; 25th percentile 82; Median 117; 75th percentile 153; Mean 118; Standard deviation 48
  - Total: Number of observations 5,530; 25th percentile 64; Median 93; 75th percentile 130; Mean 99; Standard deviation 47

### Drivers and decomposition
- Time-varying components: nominal mortgage rates, nominal median household income, nominal house prices.
- Observed long-run trends (GDP-weighted averages):
  - Nominal mortgage rates declined over the sample period.
  - Nominal household income increased consistently.
  - Nominal house prices increased consistently, with a notable correction during the GFC and subsequent recovery.
- Simple cross-country average growth analysis:
  - Mid-1970s and early 1980s HAI decline coincided with increases in house prices and borrowing rates not offset by household income growth.
  - Post-GFC improvement coincided with lower borrowing rates and falling house prices; household income moved little.
  - During the pandemic, house price growth drove the decline in affordability despite other components pointing to improved affordability.
- Decomposition for the United States (total differentials; Appendix B):
  - Changes in household income: roughly 40 percent of changes in affordability.
  - Changes in house prices: about a third.
  - Changes in mortgage rates: the remainder.
- Caveat: HAI captures immediate impact; it does not account for future interest rate adjustments or refinancing needs—relevant where adjustable-rate mortgages predominate.

### Cross-sectional patterns and caveats
- Cross-country variation:
  - Eastern European countries (e.g., the Baltics, the Czech Republic, Poland) have average HAI values touching or exceeding 150.
  - Economies with average HAI below 50 include Australia, Hong Kong SAR, Russia, and Turkey.
- Possible explanations:
  - High HAI in Eastern Europe may reflect high homeownership rates and relatively old (cheaper) housing stock.
  - Low HAI cases differ: Australia—long-run rapid house price increases; Turkey—lower mortgage market development and alternative purchase methods.
- Limitations:
  - HAI focuses narrowly on mortgage repayment affordability; does not capture outright ownership without mortgage.
  - Does not reflect sustainability of homeownership over loan lifetime given shocks.
  - Improvements driven by low interest rates can be ephemeral as rate resets occur.
  - HAI measures affordability conditional on access to a mortgage at prevalent market rate; does not measure informal financing.

### Regression analysis: government spending and housing affordability
- Research question: extent to which government spending on housing programs affects housing affordability.
- Data on government spending: Eurostat (limits regression sample to 21 AEs and 4 EMs out of 40).
- Types of spending:
  - Housing benefits and allowances (direct support).
  - Housing development spending (supply-side).
- Empirical approach:
  - Local projection framework (Jorda, 2005) with specification:
    - HAI_{i,t+h} = α_h + ∑_{k=1}^{2} γ_k^h HAI_{i,t−k} + ∑_{k=1}^{2} β_k^h Gov_{i,t−k} + ∑_{k=1}^{2} θ_k^h X_{i,t−k} + C_i^h + T_t^h + ε_{i,t}^h
    - HAI_{i,t+h} denotes the log of the housing affordability index for country i in year t+ h ∀ h∈[0,4].
    - Gov_{i,t−k} denotes the log of government expenditure on housing benefits or housing development as percent of GDP with lag k.
    - X_{i,t−k} denotes lagged controls.
    - Country and year fixed effects included.
  - Baseline controls: real GDP growth and lagged HAI. Results robust to additional controls.
- Results (Figure 9 summary):
  - Government spending on housing benefits:
    - Associated with an increase in housing affordability.
    - Effects significant at a one-year horizon following the increase in spending on benefits.
    - Effects remain positive thereafter but lose statistical significance.
  - Government spending on housing development:
    - Positive effect on affordability but with smaller magnitude and not statistically significant.
- Interpretation and cautions:
  - Housing benefits likely have a more direct and immediate effect by lowering borrowing costs for households.
  - Housing development spending operates indirectly via supply; impact depends on supply response and may be muted or delayed.
  - Public expenditure measures are crude; targeting may cause noise relative to average-household HAI.
  - Demand-side subsidies may raise house prices in the longer term, potentially negating affordability gains and increasing financial stability risks when subsidies cease.
  - Possible reverse causality (affordability influencing government spending) may understate estimated effects.

### Conclusion and policy implications
- Dataset and methodology:
  - Novel dataset on housing affordability for 40 countries spanning in some cases as early as the 1970s until 2021Q4.
  - Compiled mortgage rates, typical maturity and LTV ratios, average house size, and average household members for each country.
  - Distinct from simple price-to-income proxies by incorporating mortgage and housing market characteristics.
- Main findings:
  - Since the 1990s, affordability has improved, despite a setback during the GFC.
  - Across the cross-section, housing is more affordable in advanced than in emerging market economies.
  - Mortgage market development and modes of becoming owner-occupiers matter for affordability.
  - During periods of lower interest rates and favorable economic conditions, affordability improved in all countries.
  - Uncertainty remains about future affordability, particularly given the possibility of a "soft landing" where interest rates decline rapidly while incomes do not decrease significantly.
- Policy implications and research agenda:
  - Analysis is illustrative and identifies areas for future research:
    - More systematic cross-country analysis of HAI to understand structural determinants of housing tenure choices.
    - Analysis of policy impacts at more granular levels and over different horizons to assess costs and benefits, including intertemporal trade-offs.
    - Exploration of how housing affordability affects private consumption and housing investment.

*IMF WORKING PAPERS — Dataset on Housing Affordability, Introduction (content unit wpiea2023247-print-pdf).*

### Introduction ...........................................................................................................

### Introduction

### Background and motivation
- During the short but deep COVID-19 recession, house prices rose at record levels in many countries, reflecting a combination of demand and supply factors.
- Demand-side drivers highlighted:
  - Policy support measures that sustained income.
  - Financing conditions remaining favorable.
  - Many households searching for more space.
- Supply-side drivers highlighted:
  - Constraints on new construction.
  - Constraints on mobility, limiting available properties.
- Concerns:
  - Financial risks were perceived as limited due to post-GFC policies, but housing affordability concerns were raised.
  - Affordability was already a policy concern prior to the pandemic as prices resumed upward trends after the correction/pause around the Global Financial Crisis (GFC).

### Limitations of existing affordability metrics
- Commonly used proxy: price-to-income ratio (PIR)
  - Available from BIS and OECD as an index comparing house prices to income per capita.
  - Main downside: does not incorporate mortgage market or housing characteristics (e.g., mortgage interest rate changes affect monthly payments and therefore affordability even if PIR falls).
- OECD housing cost overburden rate
  - Captures percentage of households spending over 40% of income on housing (rent, mortgage payments, utilities, maintenance).
  - Does not consider whether a typical family that prefers ownership can qualify for a mortgage given prevailing prices and interest rates.
- BIS debt service ratio
  - Measures burden of debt on household income but does not specifically capture housing affordability.
- Country-specific indices (Australia, Canada, New Zealand, United Kingdom, United States)
  - Incorporate mortgage market or housing characteristics but methodologies differ across countries, creating comparability challenges for cross-country analysis.

### Objective and definition of the new HAI
- Aim: construct a housing affordability index (HAI) that:
  - Covers a large panel of countries over a long time period.
  - Incorporates mortgage and housing market aspects such as cost of financing and typical home size.
- Sample and period:
  - 40 countries for the period between 1970Q1 and 2021Q4.
- Affordability definition used:
  - The ability of a median-income household to qualify for a mortgage loan needed to purchase an average-priced home within a country.
  - Note: the index captures affordability for the median household, not the broader income distribution.
- Methodological benchmark:
  - Mainly follow National Association of Realtors (NAR) methodology (US) and extend to 39 additional countries.
  - Collected variables: (i) house price level (price per square meter), (ii) median household income, (iii) mortgage rates, (iv) typical loan-to-value (LTV) ratios, (v) average term-to-maturity of mortgage loans, (vi) average property size, (vii) average household size.
  - One methodological distinction: use average house price rather than median due to cross-country data availability.

### Data sources and construction approach
- Core house price level construction:
  - Use Bricongne et al. (2019) for average cost of housing per square meter (in euros) for an unbalanced panel of 40 OECD countries over 1970–2017.
  - Extend these series by:
    - Selecting 2015 as a base year and converting house price per square meter to local currency using the 2015 exchange rate vis-à-vis the euro.
    - Drawing house price indices from BIS, calculating quarterly nominal growth rates, and applying growth rates to the 2015 base year to generate average nominal house prices per square meter.
  - Resulting house price coverage spans 1970Q1 to 2021Q4 for countries with maximum observations and at least 2010Q1 to 2021Q4 for countries with minimum observations (Poland and Turkey).
- Mortgage lending rates:
  - Residential mortgage rates collected from multiple sources.
  - For 18 countries, data drawn from the European Central Bank (ECB).
  - For the remainder (except Iceland) rates drawn from Haver Analytics.
  - Iceland: use general interest rate from Central Bank of Iceland adjusted for inflation because mortgage rates for house purchases are not available in Haver and Icelandic mortgage rates are typically inflation indexed.
  - When mortgage rate series are shorter than house price series, back-cast mortgage rates using long-term rates (OECD) or policy rates (BIS or national sources):
    - Back-casting method: regress available mortgage rate on contemporaneous lending rate (long-term lending rate or policy rate) and use predicted values for missing periods.
    - Long-term borrowing rates have a median R-squared of 0.82 (see Appendix note).
- Loan-to-value (LTV) ratios:
  - Data for 30 out of 40 countries from Hypostat report dated 2020.
  - Greece, Hungary, Norway: LTV from Cerutti, Dagher, and Dell'Ariccia (2017).
  - Remaining countries: national sources (Australia, Canada, Sweden) or international reports (Finland, Italy, Japan, Slovenia).
- Typical mortgage loan maturity:
  - Data for 31 countries from Hypostat report 2020, supplemented by Cerutti, Dagher, and Dell'Ariccia (2017) for Greece, Hungary, Latvia, Malta.
  - Remaining countries: national sources (Australia, Canada, Romania) or international organizations (Japan, Italy).
- Average property size (square meters):
  - Eurostat provides data for 29 countries, measured as of 2012 (latest publication date).
  - Remaining countries: official national statistics offices or literature sources.
- Average household size:
  - Eurostat for 33 countries, United Nations for 6 countries, Population Reference Bureau (PRB) for Russia.
  - UN data cover 2013–2016; household size treated as relatively constant over time within a country.

### Challenges, assumptions, and remedies
- Main data challenges:
  - Lack of house price levels in local currency (many cross-country sources report indices, not levels).
  - Limited centralized databases for mortgage rates, LTV ratios, and average terms.
- Remedies employed:
  - Use Bricongne et al. (2019) levels and extend with BIS index growth rates to obtain levels in local currency.
  - Draw mortgage-market and housing variables from multiple sources including national statistical offices and international reports.
  - Extend and complement series using linear interpolation and back-casting.
  - Adopt proxies for missing variables (e.g., Iceland mortgage proxy; back-casted mortgage rates via regression on long-term or policy rates).
- Caveats:
  - Constructed series depend on imputed data and mismeasurement cannot be ruled out.
  - Choice of base year (2015) is somewhat arbitrary but indices are robust to base year selection.

### Validation exercises and empirical strategy
- Performance assessments performed:
  - For the United States: compare constructed index to NAR index.
  - Across countries over time: analyze summary statistics and drivers of HAI dynamics.
  - Policy linkage: regression analysis examining the link between government spending on housing and affordability.
- Analytical steps summarized in the paper:
  - Stocktaking of affordability indices in selected countries.
  - Detailed data description and methodology to construct HAI.
  - Documentation of empirical properties of HAI over time and across countries.
  - Regression analysis on government spending and housing affordability.
  - Conclusion summarizing key findings.

*IMF WORKING PAPERS — Dataset on Housing Affordability, Introduction (content unit wpiea2023247-print-pdf).*

### 0.75 in the sample of countries.

### wpiea2023247-print-pdf - 0.75 in the sample of countries.

### Data sources and coverage
- Median household income: data come mainly from surveys conducted by Eurostat (32 countries).
- Remaining 8 countries: data come either from national sources, CEIC, or OECD Income Distribution Database.
- Exceptions where median household income is not reported as equalized: Canada, Hong Kong SAR, South Korea, and the United States.
- For most economies (apart from Canada, Hong Kong SAR, South Korea, and the United States) median household income is reported in terms of equalized household income; the dataset multiplies the income data by the average household size to recover household-level income.
- Median household income is in most cases available only at annual frequency; quarterly series are generated by linearly interpolating the annual data.
- Time series gaps are extended backwards and forwards using GDP per capita growth rates from the IMF’s World Economic Outlook Database, assuming median household income has the same growth rate as GDP per capita within a country.
- Caveat: assuming median household income grows with GDP per capita could lead to overestimation of affordability, particularly if income distribution becomes more skewed and upper deciles enjoy larger gains.

### Methodology overview
- The housing affordability index (HAI) combines information on:
  - price of a typical house,
  - income of a typical household,
  - average mortgage rates,
  - typical LTV ratio,
  - typical maturity of a mortgage.
- The computation follows discrete building blocks calculated for each country.

### Construction steps (building blocks)
- 1. Price of a typical house (HP)
  - HP = price per square meter × average size of a home.
- 2. Monthly mortgage payment (PMT)
  - PMT is computed for principal and interest on a loan amount HP * LTV using the formula:
    - 푃푀푇 = 퐻푃 ∗ 퐿푇푉 ∗ (퐼푅/12) ∗ [1 − (1 + 퐼푅/12)^(−푀퐴)]^(−1),
    - where IR denotes the annualized mortgage rate and MA denotes the maturity (in months).
  - Note: The methodology follows the NAR's approach; HAI measures affordability based on the capacity of households to qualify for a mortgage at a typical LTV ratio.
  - Emphasis is placed on typical LTVs rather than regulatory LTVs to highlight long-term structural features of mortgage markets rather than cyclical fluctuations.
  - Observation: while a higher LTV can decrease affordability due to increased costs of servicing larger loans, a lower down payment requirement could potentially enable more households to consider property ownership.
- 3. Necessary annual income to qualify (QINC)
  - QINC = PMT ∗ 4 ∗ 12,
  - Assumption: for a household to qualify for a mortgage loan, the monthly mortgage payment should not exceed a quarter of the household’s monthly income.
  - This rule of thumb aligns with the standard often used in the international debate on housing affordability: housing is considered affordable if households do not spend more than 30% of their gross income on housing costs including insurance and taxes (OECD).

### Key numeric and procedural points
- Eurostat coverage: 32 countries.
- Other sources coverage: 8 countries.
- Interpolation: annual-to-quarterly by linear interpolation.
- Qualification rule: monthly mortgage payment ≤ a quarter of monthly household income.
- OECD affordability benchmark referenced: 30% of gross income on housing costs.
- Extensions of income series use GDP per capita growth rates from the IMF’s World Economic Outlook Database.

*IMF WORKING PAPERS Dataset on Housing Affordability (excerpt).*

### 4. Then, the housing affordability index (HAI) is obtained as:

### 4. Then, the housing affordability index (HAI) is obtained as:

### Definition of the HAI
- 퐻퐴퐼=(푀퐸퐷퐼푁퐶/푄퐼푁퐶)∗100, where MEDINC is the median household income.
- Interpretation:
  - HAI = 100: a median-income household has exactly enough income to qualify for a mortgage loan on an average-priced home.
  - HAI > 100: a median-income household has more than the qualifying income to apply for a mortgage on an average-priced home.
  - HAI < 100: a median-income household does not have sufficient income to qualify for a mortgage on an average-priced home.
- The index measures the degree to which a typical family can afford monthly mortgage payments for a home valued at the average market price within a country.

### Validation: Comparison with NAR index (United States)
- Methodology: follow NAR methodology; compare computed HAI to NAR-reported HAI for the United States.
- Key statistics (1970q1-2021q4):
  - Our index mean: 122.8
  - NAR index mean: 127.5
  - Our index standard deviation: 27.8
  - NAR index standard deviation: 30.1
  - Correlation between indices: 0.96
- Sources of slight differences:
  - NAR uses median house price; this study uses average house price per square meter due to cross-country data constraints.
  - Both use Census median household income annual data interpolated to quarterly and forecast/back-cast using nominal GDP growth rate for comparability.
  - NAR uses effective mortgage rate from the Federal Housing Finance Board; this study draws mortgage rates from Haver (national sources; definitions vary by country).

### Housing affordability dynamics over time (sample overview)
- Sample composition:
  - Total countries in sample: 40
  - Advanced economies (AEs): 33
  - Emerging markets (EMs): 7
- Data coverage:
  - AEs: coverage tends to start in the 1970s.
  - EMs: coverage tends to start in the 2000s.
- Summary statistics (2001q1 to 2021q4, for comparability):
  - Mean HAI in AEs: 118
  - Mean HAI in EMs: 85
  - Standard deviation in AEs: 48
  - Standard deviation in EMs: 53
- Cross-country trends:
  - 1970s–mid-1990s: median HAI below 100 (median-income households struggled to obtain mortgage for an average-priced house); even the 75th percentile was below 100.
  - 2000s: cross-country median HAI exceeded 100 (improved affordability).
  - Global Financial Crisis (GFC): affordability collapsed but recovered quickly.
  - Since 2010: median HAI remained comfortably above 100; 25th percentile hovered just below 100.
- Note: country sample size varies over time; mean calculated as the average weighted by GDP.

### Drivers of HAI over time (time-varying components)
- Time-varying components: nominal mortgage rates, nominal median household income, nominal house prices.
- Observed long-run trends (GDP-weighted averages):
  - Nominal mortgage rates declined over the sample period.
  - Nominal household income increased consistently.
  - Nominal house prices increased consistently, with a notable correction during the GFC and subsequent recovery.
- Simple cross-country average growth analysis:
  - Mid-1970s and early 1980s decline in HAI coincided with increases in house prices and borrowing rates not offset by household income growth.
  - Post-GFC improvement in affordability coincided with lower borrowing rates and falling house prices; household income moved little.
  - During the pandemic, house price growth drove the decline in affordability despite other components pointing to improved affordability.
- Decomposition for the United States (total differentials; Appendix B):
  - Rough contributions through the sample period:
    - Changes in household income: roughly 40 percent of changes in affordability.
    - Changes in house prices: about a third.
    - Changes in mortgage rates: the remainder.
  - Caveat: HAI captures immediate impact; it does not account for future interest rate adjustments or refinancing needs—particularly relevant where adjustable-rate mortgages predominate.

### Cross-sectional patterns and caveats
- Cross-country variation:
  - Eastern European countries (e.g., the Baltics, the Czech Republic, Poland) have average HAI values touching or exceeding 150.
  - Economies with average HAI below 50 include Australia, Hong Kong SAR, Russia, and Turkey.
- Possible explanations:
  - High HAI in Eastern Europe may reflect high homeownership rates and relatively old (cheaper) housing stock.
  - Low HAI cases differ: e.g., Australia—long-run rapid house price increases; Turkey—lower mortgage market development and alternative home-purchase methods (savings, informal borrowing).
- Limitations of HAI:
  - Focuses narrowly on mortgage repayment affordability; does not capture outright ownership without mortgage.
  - May not reflect sustainability of homeownership over loan lifetime given shocks to interest rates and household income.
  - Improvements driven by low interest rates can be ephemeral as rate resets occur.
  - HAI measures affordability conditional on access to a mortgage at prevalent market rate; does not measure other forms of access or informal financing.

### Regression analysis: government programs and housing affordability
- Research question: extent to which government spending on housing programs affects housing affordability.
- Data on government spending: Eurostat (general government expenditure by economic function). Note: Eurostat coverage limits regression sample to 21 AEs and 4 EMs out of 40.
- Types of government spending considered:
  - Housing benefits and allowances (direct support to households).
  - Housing development spending (supply-side spending to increase housing stock).
- Empirical approach:
  - Local projection framework (Jorda, 2005) with specification:
    - 퐻퐴퐼푖,푡+ℎ = αℎ + ∑_{k=1}^{2} γ_{k}^{h} 퐻퐴퐼푖,푡−k + ∑_{k=1}^{2} β_{k}^{h} Gov_{i,t−k} + ∑_{k=1}^{2} θ_{k}^{h} X_{i,t−k} + C_{i}^{h} + T_{t}^{h} + ε_{i,t}^{h}
    - 퐻퐴퐼푖,푡+ℎ denotes the log of the housing affordability index for country i in year t+ h ∀ h∈[0,4].
    - Gov_{i,t−k} denotes the log of government expenditure on housing benefits or housing development as percent of GDP with lag k.
    - X_{i,t−k} denotes lagged controls.
    - Country and year fixed effects included.
  - Baseline controls: real GDP growth and lagged HAI. Results robust to additional controls (population growth, 10-year government bond yield, stock market index growth, demographic variables).
- Regression results (Figure 9 summary):
  - Government spending on housing benefits:
    - Associated with an increase in housing affordability.
    - Effects significant at a one-year horizon following the increase in spending on benefits.
    - Effects remain positive thereafter but lose statistical significance.
  - Government spending on housing development:
    - Positive effect on affordability but with smaller magnitude and not statistically significant.
- Interpretation:
  - Housing benefits likely have a more direct and immediate effect by lowering borrowing costs for households.
  - Housing development spending operates indirectly via supply; its impact depends on the extent of supply response relative to demand and may be more muted or delayed.
- Limitations and cautions:
  - Public expenditure measures are crude; targeting (by income group or housing segment) may cause noise relative to average-household HAI.
  - Potential trade-offs:
    - Demand-side subsidies (housing benefits) may raise house prices in the longer term, potentially negating affordability gains and increasing financial stability risks if households become overburdened when subsidies cease.
    - Supply-side spending (housing development) could offer more sustainable affordability improvements but may yield smaller immediate effects.
  - Possible reverse causality (affordability influencing government spending) may understate estimated effects.

*Source: wpiea2023247-print-pdf - 4.*

### Conclusion

### Conclusion

### Dataset and methodology
- Novel dataset on housing affordability for a sample of 40 countries spanning in some cases as early as the 1970s until 2021Q4.
- To construct the index, compiled information on mortgage rates, typical maturity and LTV ratios, average size of a house, and average number of household members for each of the 40 countries in the sample.
- Distinct from measures that proxy affordability as the ratio of house prices and households’ income: this measure considers characteristics of the mortgage and housing markets.

### Main findings
- Since the 1990s, affordability of housing has improved, despite a setback suffered during the GFC.
- Across the cross-section, housing is more affordable in advanced than in emerging market economies.
- An important dimension is the development of mortgage markets and the mode by which households become owner-occupiers (with or without a mortgage).
- During periods of lower interest rates and favorable economic conditions, affordability improved in all countries.
- Uncertainty remains about future affordability, particularly given the possibility of a "soft landing"—a scenario where interest rates decline rapidly while incomes do not experience significant decreases. The outcome of this situation will play a crucial role in determining future affordability.

### Quantitative summary (Table 1: Summary statistics of HAI)
- EM: Number of observations 440; 25th percentile 40; Median 71; 75th percentile 114; Mean 85; Standard deviation 53
- AE (all): Number of observations 4,890; 25th percentile 66; Median 94; 75th percentile 130; Mean 100; Standard deviation 46
- AE (after 2001): Number of observations 2,572; 25th percentile 82; Median 117; 75th percentile 153; Mean 118; Standard deviation 48
- Total: Number of observations 5,530; 25th percentile 64; Median 93; 75th percentile 130; Mean 99; Standard deviation 47

### Regression analysis on fiscal spending and housing affordability
- Regression analysis indicates government spending on housing could have an impact on housing affordability.
- Only the impact of spending on housing benefits appears to be significant.
- Government spending on housing development may yield a positive effect, but the impact is insignificant for the sample of countries and for the time-period considered.

### Policy implications and research agenda
- The presented analysis is largely illustrative and identifies areas where future research can build on these findings.
- Suggested directions for future research:
  - More systematic analysis of cross-country differences in the HAI to shed light on structural determinants of housing tenure choices.
  - Analysis of the impact on affordability of different government policies at a more granular level and over different horizons to help policymakers assess relative costs and benefits, including intertemporal trade-offs in achieving affordability goals.
  - Exploration of how housing affordability can help understand private consumption and housing investment.

*IMF Working Papers — Dataset on Housing Affordability*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023247-print-pdf.pdf_