## Executive Summary (wpiea2026124-source-pdf)

## Source details

**Canonical URL:** [Executive Summary (wpiea2026124-source-pdf)](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026124-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026124-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026124-source-pdf.pdf.json)

---

### Overview
- Housing affordability is a prominent social, political, and economic issue across the EU for the period 2010–2024; public concern rose sharply between 2015 and 2025.
- Numerous national initiatives and an EU-level plan are in place; the paper focuses on consequences rather than drivers.
- Analysis period: 2010–2024 (core empirical estimates use annual Eurostat data for EU27 over 2010–2023).

### Main channels through which housing affordability affects outcomes
- Loss of opportunities: poverty and labor market consequences
  - Job or training offers may be declined because housing near workplaces is unaffordable, producing higher poverty, lower labor force participation, reduced mobility, and spatial misallocation.
- Adequacy of housing: overcrowding and severe housing deprivation
  - Higher housing costs worsen overcrowding and severe housing deprivation, with negative health implications.
- Reduction in essential non-housing consumption: fertility and health consequences
  - Low-income households cut non-housing spending (health, education, nutrition, fertility) when housing costs rise.

### Econometric approach and core findings
- Methodology: Double Machine Learning with Instrumental Variables (DML‑IV) applied to country‑year panel (EU27).
- Broad empirical findings:
  - Large impacts on housing adequacy, poverty, and health.
  - Somewhat smaller impacts on fertility and labor force participation.
  - Small impacts on average age of leaving parental home and age of first childbirth.
  - All three mechanisms (housing adequacy deterioration; lost opportunities; reduced non‑housing consumption) are empirically active in the EU.

### Heterogeneity and time variation
- Cross‑country heterogeneity: impacts differ across Member States reflecting policies and structural features.
- Time variation: impacts increased pre‑pandemic but declined afterwards; remote work may have contributed to the reversal.

### Policy implications
- Housing policy complements social policies; addressing affordability supports poverty reduction, labor market participation, health, and demographic goals.
- Rapid social policy responses can mitigate short‑term harms while supply‑side measures (increasing housing supply) take effect.

*Source: wpiea2026124-source-pdf - Executive Summary*

---

### 1. Housing: a growing concern (key facts and trends)

### Public perceptions and survey evidence
- Perception changes (2015–2025):
  - Share of Europeans perceiving housing as one of two main personal issues increased by 80 percent between 2015 and 2025.
  - Share perceiving housing as one of two main national issues increased by 63 percent.
- Country concentration of concern increases: particularly strong in Ireland, Luxembourg, The Netherlands, and Spain; notably concentrated 2022–2024 before stabilizing.
- Selected survey results:
  - OECD Risks That Matter (17 EU countries): 45 percent concerned about housing in next year or two; 52 percent concerned over next 10 years.
  - In 2022, over 59 percent of respondents aged 18 to 29 expressed concerns about housing in next year or two—12 percentage points higher than respondents aged 30 to 49 and 23 percentage points higher than respondents aged 50 to 64.
  - Eurofound: share likely to have to leave home in next 2–3 months because they cannot afford it reached 7.6 percent in 2025 (up 43 percent since 2021).
  - Gallup World Poll (OECD): satisfaction with housing affordability: 50 percent in 2017, 49 percent in 2020, 40 percent in 2023, 43 percent in 2024.
  - France (September 2025): 57 percent cite inflation and purchasing power as main concern; 78 percent say housing weighs most on purchasing power; 87 percent say housing price rose most in past 10–15 years.

### Affordability and drivers (highlights)
- Pandemic period:
  - Real house price growth (deflated by HICP) accelerated and outpaced income growth, reducing purchase affordability.
- Post-pandemic:
  - Prices and rents grew less than income, but energy price increases drove housing costs up.
  - Share unable to keep home adequately warm rose from 6.9 percent in 2019 and 2021 to 10.6 percent in 2023; energy prices stabilized in 2023 and 2025 and inability-to-keep-warm declined starting 2024.

### Macro and market indicators (selected statistics)
- Over 2011–25, housing costs have risen at broadly the same rate as income.
- House-price-to-income ratio: peaked in 2022 at 15 percent higher than in 2015, then declined rapidly; by 2024 it was below its long‑term average.
- Rent share in disposable income for renters: peaked at 24.5 in 2013–15, declined to 22.5 percent in 2025.
- Housing-cost burden and overburden: lower than early 2010s; post‑pandemic rebound limited and short‑lived; both ratios declined in 2024–25.
- Housing-related arrears: lower than early 2010s but higher than 2019; decline in 2024–25 limited.

### Homeownership, borrowing capacity, and financialization
- EU27 homeownership rate: 71 percent in 2012, 70 percent in 2020, declined to 68.5 percent in 2025.
- Borrowing capacity influenced by interest rates and macro‑prudential borrower‑based measures; young adversely affected by lower incomes and fixed‑term contracts.
- Non‑resident investor share in EU housing market rose from 10.4% in 2010 to 14.1% in 2020 (EC, 2025a), coinciding with reduced home ownership in some regions.

### Housing adequacy measures
- Overcrowding and severe housing deprivation monitored; adequacy generally improved over time but young experienced erosion post‑pandemic.
- Overcrowding and severe deprivation series reported for 2015–2023 (note: no data for severe deprivation in 2021 and 2022 in figures).

### Key empirical takeaways (concise)
- Public concern rose sharply 2015–2025, especially 2022–2024; affordability dominated protests and surveys.
- Aggregate indicators show no clear sustained deterioration in affordability across EU over past 10–15 years; some improvements relative to early 2010s.
- Short‑lived post‑pandemic spikes attributable in part to energy shocks; homeownership declined to 68.5 percent in 2025.

*Source: wpiea2026124-source-pdf - 1. Housing: a growing concern...*

---

### 3. Reasons for the gap between perceptions and data; conceptual framework

### Why aggregate data may understate problems
- Household formation responses (young staying with parents or sharing) can raise household income and mechanically reduce measured burden.
- Data omissions:
  - Principal mortgage payments excluded from housing cost measures (Annex I).
  - HICP excludes owner‑occupiers’ housing costs (OOH); rents component covers existing rents not new rents.
- Jurisdiction‑specific indices that include OOH can reverse purchasing‑power conclusions (example: Catalonia adjustments noted in source).

### Concentration of affordability problems
- Affordability concentrated among the young, lower‑income households, urban populations, and tenants.
- Social housing share: estimated 6–7 percent by EC, 8 percent by OECD; large cross‑country variation.
- France examples:
  - 25 percent of population aged 18–34 report inadequate size vs 15 percent for population above 35 (Odoxa, 2026).
  - 69 percent of young report multiple housing quality issues — 14 percentage points more than population aged 35+ (Odoxa, 2026; Eurofound, 2023).
  - Over 80 percent of new hiring are fixed‑term or temporary (France), impeding access to housing.
  - Tenants at reduced price (or free): 11.1 percent in 2010 → 9.5 percent in 2019 → 10.7 percent in 2025.
  - Survey: in France, 30 percent of young (18–34) report that due to “housing problem” they could not accept a job (Odoxa, 2026).

### Non‑financial dimensions and volatility
- Perceptions reflect availability, adequacy, loan access, wealth inequalities, and volatility (large consequences from cost increases).

### Conceptual framework — three interacting mechanisms
1. Housing costs → housing adequacy deterioration.
2. Housing costs → lost opportunities and reduced mobility.
3. Housing costs → cuts in non‑housing consumption.

### Forms of inadequacy and consequences
- Homelessness: rising in the EU; decline in social housing since 2010 may have contributed.
- Inadequate location, size (overcrowding), and quality (severe housing deprivation) affect health, well‑being, productivity, child development, and intergenerational inequality.
- New‑construction responses (smaller dwellings) can entrench adequacy problems.

*Source: wpiea2026124-source-pdf - 3. Possible reasons for the gap between perceptions and data*

---

### Econometric approach and estimation method (DML‑IV summary)

### Rationale and challenges
- Two main complications: many potential confounders and endogeneity (housing affordability co‑moves with income, credit, policy).
- Standard 2SLS limitations: strong linearity, small conditioning sets, finite‑sample bias, overfitting risk when expanded.

### Adopted estimator: Double Machine Learning with Instrumental Variables (DML‑IV)
- Key features:
  - Neyman‑orthogonal scores to reduce sensitivity to nuisance estimation errors.
  - Cross‑fitting to prevent overfitting.
- Advantages:
  - Allows rich macroeconomic and institutional controls, nonlinearities, heterogeneous effects, and panel settings with limited time periods.
- Target parameter: average marginal effect of change in affordability on socioeconomic outcomes, conditional on covariates.
- Data transformation: all variables expressed as log deviations from contemporaneous EU average (difference between natural logarithm of country value and logarithm of EU27 average in that year).

### Instruments used and motivation
- Instruments:
  - Lagged construction activity: building permits in square meters per thousand inhabitants (lag 1 year).
  - Construction costs: annual growth in construction producer prices for new residential buildings.
- Economic logic:
  - Lagged permits capture effective stock additions with delay; construction costs constrain supply and affect prices.
- Validation:
  - First‑stage relevance and exogeneity tests adapted for DML‑IV; cluster‑robust bootstrap inference.
- Key caveat:
  - Exclusion restriction untestable; rich controls, reverse‑causality tests, and overidentification tests provide supporting evidence.

### Estimation and inference
- Four‑step estimation: data assembly → nuisance estimation (out‑of‑sample at country level) → causal effect estimation using instrument‑driven variation → robustness and validity checks.
- Inference: wild cluster bootstrap at country level to account for limited time span and within‑country serial correlation.
- Data sources: annual Eurostat indicators for EU27, 2010–2023.

*Source: wpiea2026124-source-pdf - 1. Econometric approach and estimation method*

---

### 3. Results (core quantitative findings and heterogeneity)

### Overview of housing affordability measures
- Two indicators: housing cost share in disposable income and housing cost overburden rate.
- Housing cost share in disposable income preferred: captures intensity across distribution and yields larger, more consistent coefficients.

### A. Impact on housing adequacy
- Statistical significance:
  - Estimates using housing cost share significant at 10 percent or less; overcrowding and severe housing deprivation often significant at 1 percent.
- Magnitudes (preserve exact reported elasticities and illustrative numbers):
  - 1 percent increase in housing cost share in disposable income (relative to EU average) increases overcrowding rate by 2.34 percent.
    - Illustration: if a country’s share of housing cost in disposable income is 0.2 ppt higher than EU average of 20.8 percent, its overcrowding rate would be 0.4 ppt higher than EU average of 17.4 percent.
  - 1 percent increase in housing cost overburden rate increases overcrowding rate by 1.41 percent for entire population.
    - Illustration: if a country’s housing cost overburden rate is 0.1 ppt higher than EU average of 9.9 percent, its severe housing overcrowding rate would be 0.25 ppts higher than EU average of 17.4 percent.
  - 1 percent increase in housing cost share in disposable income increases severe housing deprivation rate by 1.43 percent.
    - Illustration: if housing cost share is 0.2 ppt higher than EU average, severe housing deprivation would be less than 0.1 ppt higher than EU average of 5.0 percent.
- Young and leaving parental home:
  - 1 percent increase in housing cost share increases average age of leaving parents by 0.08 percent (average age virtually unchanged at 26.5 years); result barely significant at 10 percent.
- Temporal and regional patterns:
  - Impact on adequacy rose pre‑pandemic and declined post‑pandemic; regionally always negative in Central, Eastern, Southern Europe, and Ireland; mixed in Western Europe and Nordics depending on indicator.
- Role of social transfers:
  - Social spending 1 percent higher than EU level reduces:
    - Impact on overcrowding from 2.34 percent to about 2.11 percent.
    - Impact on severe housing deprivation from 1.43 percent to about 1.10 percent.
    - Impact of overburden rate on overcrowding from 1.41 percent to about 1.29 percent.
  - Social transfers reduce impact on average age of leaving parents by about one quarter.

### B. Impact on labor force participation (LFP)
- Estimated effects:
  - 1 percent increase in housing cost share in disposable income reduces total LFP by 0.53 percent.
  - 1 percent increase in housing cost share reduces female LFP by 0.59 percent.
    - Illustration: if country’s housing cost share is 0.2 ppt higher than EU average of 20.8 percent, total LFP would be 0.4 ppt lower than EU average (71.7 percent vs 72.1 percent); female LFP would be 0.4 ppt lower than EU average (66.2 percent vs 66.6 percent).
- Temporal/regional variation:
  - Negative impact on LFP declined post‑pandemic (marginally for total, larger for female LFP).
- Childcare and social benefits:
  - Lower enrollment in childcare below age 3 marginally increases negative female LFP impact (from -0.59 to about -0.60).
  - Social benefits mitigate LFP impact modestly.

### C. Impact on poverty (AROP)
- Estimated effects:
  - 1 percent increase in housing cost share in disposable income increases at‑risk‑of‑poverty rate (AROP) by 0.92 percent for entire population.
  - 1 percent increase in housing cost overburden rate increases AROP by 0.29 percent for entire population.
  - 1 percent increase in housing cost overburden rate increases AROP by 0.65 percent for the young (aged 20 to 29).
    - Illustrations:
      - If housing cost share is 0.2 ppt above EU average of 20.8 percent, AROP would be 0.2 ppt higher than EU average of 16.8 percent.
      - If housing cost overburden rate is 0.1 ppt above EU average of 9.9 percent, AROP for entire population would be 0.05 ppt higher than EU average of 16.8 percent.
      - If overburden rate of the young is 0.1 ppt above EU average of 12.6 percent, AROP of the young would be 0.1 ppt higher than EU average of 19.4 percent.
- Temporal/cross‑country:
  - Increase in housing-cost‑share affects all countries; effects broadly stable though some lower impact post‑pandemic.
  - Social transfers marginally reduce impact; larger offset for overburden rate especially for the young.

### D. Impact on fertility
- Measurement note: indicators aggregate tenants and homeowners; do not capture direct cost of buying a house.
- Estimated effects:
  - 1 percent increase in housing cost share increases average age at first birth by 0.21 percent.
    - Illustration: if country’s housing cost share is 0.2 ppt higher than EU average (21.0 percent vs 20.8 percent), women would have first child less than 1 month later (29.4 vs 29.3 years).
  - 1 percent increase in housing cost share lowers fertility rate by 0.56 percent.
    - Illustration: 0.2 ppt higher housing cost share → fertility rate 0.1 lower than EU average of 1.51.
  - Housing cost overburden has limited and less significant impact on fertility (example: overburden 10.0 percent vs 9.9 percent → fertility 1.509 vs EU average 1.507).
- Regional/time heterogeneity:
  - Negative fertility impact driven by South and South‑Eastern Europe and Ireland; positive in Western Europe and Nordics.
  - Negative pre‑pandemic trend declined post‑pandemic (remote work cited as possible factor).
- Role of social spending and childcare:
  - Social spending reduces impact on fertility by about 11%:
    - Social spending 1 percent above EU average reduces impact of overburden on fertility by 0.01 percentage point (from -0.08 to about -0.07).
    - Reduces impact of housing cost share on fertility from -0.56 to about -0.50.
    - Reduces impact on age of first childbirth from 0.21 to about 0.20 (about 8 percent reduction).
  - Childcare reduces negative impact of housing affordability on women’s LFP but does not increase fertility in estimates.

### Impact on health
- Core estimate:
  - 1 percent increase in housing cost share in disposable income relative to EU average increases share of population perceiving their health as bad or very bad by 1 percent.
    - Illustration: if housing cost share is 0.2 ppt higher than EU average of 20.8 percent, share perceiving health as bad or very bad would be 0.1 ppt higher than EU average of 9.2 percent.
- Geographic/time patterns:
  - Impact stronger in Southern and Eastern parts of EU and Ireland.
  - Impact increased pre‑pandemic and declined afterwards.
- Mechanisms: stress, material hardship, reduced healthcare and nutrition, poor housing quality; intergenerational health and developmental consequences.

### Summary magnitudes and relative importance
- Particularly large effects: housing adequacy, poverty, and health.
- Somewhat smaller effects: fertility and labor force participation.
- Small effects: average age of leaving parental home and age at first childbearing (in absolute terms).

*Source: IMF Working Paper — “The (Many) Consequences of the Housing Affordability Problem in the EU”, 3. Results*

---

### Conclusions and policy implications

### Main conclusions
- Empirical evidence supports three principal channels: deterioration in housing adequacy, loss of opportunities, and reductions in non‑housing consumption.
- Consequences are statistically significant for housing adequacy, poverty, labor force participation, fertility, and health.
- Pre‑pandemic worsening of impacts reversed after the pandemic; remote work likely a contributing factor.

### Policy recommendations and complementarities
- Increase housing supply (e.g., European Affordable Housing Plan) to dampen housing costs and long‑run negative consequences.
- Coordinate social spending with housing policies:
  - Social transfers can mitigate immediate impacts on overcrowding, severe deprivation, poverty, and fertility.
  - Childcare availability reduces negative female LFP impacts.
- Complement long‑term structural solutions with faster‑delivering measures given public perceptions of an affordability emergency.

*Source: wpiea2026124-source-pdf — Conclusions and Annex I definitions*

---

*Source: wpiea2026124-source-pdf - The (Many) Consequences of the Housing Affordability Problem in the EU (Working Paper No. WP/2026/124).*

### Executive Summary ......................................................................................................

### Executive Summary

### Overview
- Housing affordability has become a prominent social, political, and economic issue throughout the European Union (EU), with protests and surveys indicating widespread concerns that have become a major topic in several local elections.
- Numerous national policy initiatives have been launched and are now complemented by an EU-level plan.
- This paper focuses on the consequences of housing affordability rather than its drivers, covering the period 2010-2024 for the EU.

### Main channels through which housing affordability affects outcomes
- Loss of opportunities: poverty and labor market consequences
  - Individuals may turn down job offers or learning/training opportunities due to inability to find or afford housing near prospective workplaces.
  - Expected outcomes include increases in the overall poverty rate, lower labor force participation, reduced labor mobility, and spatial misallocation of labor.
- Adequacy of housing: overcrowding and severe housing deprivation
  - When households cannot afford adequate housing, they may live in inadequate accommodation in terms of size, quality, and/or location.
  - An increase in housing costs is expected to worsen housing overcrowding rate and severe housing deprivation, with negative implications notably for health.
- Reduction in essential non-housing consumption: fertility and health consequences
  - Low-income households may cut non-housing consumption when housing costs are high, affecting health, education, nutrition, and fertility.

### Econometric approach and core findings
- Methodology
  - Estimation uses a double machine learning framework with instrumental variables.
- Broad findings
  - The impact of an increase in housing cost burden is:
    - Large on housing adequacy, poverty, and health.
    - Somewhat smaller on fertility and labor force participation.
    - Small on the average age at which children leave the parental house and on age at which women have their first child.
  - All three identified mechanisms (loss of opportunities, housing inadequacy, reduced non-housing consumption) are empirically active in the EU.

### Heterogeneity and time variation
- Country and time differences
  - The impact of housing cost differs across countries, reflecting differences in policies and structural features.
  - The impact changed over time: for all indicators, the impact increased pre-pandemic but declined afterwards.
  - Further research is needed to identify the reasons for this reversal; the development of remote work may have played a role.

### Policy implications
- Housing affordability problems exacerbate social and economic issues that other public policies aim to address, implying synergies between housing policies and broader social policies.
- Addressing housing affordability can therefore support goals related to poverty reduction, labor market participation, health, and demographic outcomes.

*Source: wpiea2026124-source-pdf - Executive Summary*

### 1. Housing: a growing concern...

### 1. Housing: a growing concern...

### Public perceptions and survey evidence
- The share of Europeans that perceive housing as one of the two main issues they face increased by 80 percent between 2015 and 2025; the share perceiving housing as one of the two main issues their country faces increased by 63 percent.
- The increase in concern was particularly strong in Ireland, Luxembourg, The Netherlands, and Spain, and was notably concentrated from 2022 to 2024 before stabilizing.
- Eurobarometer indicators:
  - Personal issue and national issue series tracked for Autumn 2015 through Autumn 2025 show rising shares (figures present detailed time series).
- Table of major housing-related protests (2021–2025) shows that when protests occurred the main grievance was usually housing affordability (examples include The Netherlands 2021–2024, France 2023, Portugal 2024, Ireland 2024, Spain 2024–2025).
- OECD Risk That Matters survey (17 EU countries surveyed):
  - 45 percent are concerned about finding or maintaining adequate housing in the next year or two.
  - 52 percent are concerned about finding or maintaining adequate housing in the next 10 years.
  - In 2022, over 59 percent of respondents aged 18 to 29 expressed concerns about not being able to find or maintain adequate housing in the next year or two—12 percentage points higher than respondents aged 30 to 49 and 23 percentage points higher than respondents aged 50 to 64.
- Eurofound survey:
  - The share of the population declaring it is likely they will have to leave their home in the next 2 to 3 months because they cannot afford it increased by 43 percent since 2021, reaching 7.6 percent in 2025.
- Gallup World Poll (OECD):
  - Satisfaction with housing affordability: 50 percent in 2017, 49 percent in 2020, 40 percent in 2023, and 43 percent in 2024.
- France (September 2025 survey):
  - 57 percent of respondents pointed to inflation and purchasing power as their main concern.
  - 78 percent highlighted that housing weighs most on their purchasing power because 87 percent said housing was the consumption item whose price increased the most in the past 10 to 15 years.

### Affordability and its drivers
- When housing concerns trigger protests, affordability is predominantly emphasized.
- The recent increase in public concerns coincides with a sharp post-pandemic increase in housing costs.
- During the COVID-19 pandemic:
  - Real house price growth (deflated by the HICP) accelerated markedly and outpaced income growth, making home purchase less affordable.
- After the pandemic:
  - Housing prices and rents grew less than income, but housing costs shot up driven by energy prices.
  - The share of the population unable to keep their home adequately warm rose from 6.9 percent in 2019 and 2021 to 10.6 percent in 2023; this level was noted as never reached since 2012 (Eurostat, 2025).
  - Energy prices stabilized in 2023 and 2025; the share unable to keep homes adequately warm declined starting in 2024.
  - Income caught up with housing costs; most surveys indicate concerns stopped increasing after the spike.

### Macro and market trends in housing costs and affordability measures
- Over the period 2011–25, housing costs have risen at broadly the same rate as income (Figure 5).
- House-price-to-income ratio:
  - Exhibited volatility over past two decades with no sustained deterioration overall.
  - The ratio decreased after the burst of the real estate bubble, increased starting in 2015 to peak in 2022 (15 percent higher than in 2015), and declined rapidly thereafter; by 2024 it was below its long-term average.
  - The magnitude of country-level changes in price-to-income ratios is only weakly correlated with changes in the perception of housing as an issue.
- Rent-to-income:
  - Harmonized rent-to-income ratio not available; for renters, the share of rent in disposable income declined from a peak of 24.5 in 2013-15 to 22.5 percent in 2025.
  - The increase during the pandemic was limited and quickly reversed.
- Housing-cost burden indicators:
  - The share of housing cost in disposable income and the share of population facing housing cost overburden are much lower than in the early 2010s.
  - The post-pandemic rebound was limited and short-lived; both ratios declined in 2024-25.
- Housing-related arrears:
  - Lower than in the early 2010s but remain higher than in 2019; the decline in 2024-25 was limited.

### Homeownership, borrowing capacity, and financialization
- Homeownership rate:
  - Varies significantly across EU members; generally larger in newer Member States.
  - Declined in most countries since 2010.
  - EU27 homeownership rate: 71 percent in 2012 (reported), 70 percent in 2020, and declined to 68.5 percent in 2025.
  - The faster decline in 2020s coincides with increased population concerns, but country-level data suggest no strong link between ownership decline and perception changes.
- Factors affecting capacity to purchase:
  - Borrowing capacity depends on more than price-to-income; access to credit matters.
  - From the Global Financial Crisis to the pandemic, low interest rates made borrowing easier, while macro-prudential borrower-based measures reduced access to home loans for parts of the population.
  - The young were particularly affected due to lower incomes and higher prevalence of fixed-term employment contracts reducing income predictability.
- Financialization:
  - Institutional investors (pension funds, insurance companies) increased investments in real estate during periods of abundant liquidity and low interest rates.
  - ECB data noted non-resident investors' share in the EU housing market increased from 10.4% in 2010 to 14.1% in 2020, coinciding with a reduction in home ownership (EC, 2025a).
  - Increased financialization may have diminished housing available for individual buyers, contributed to higher prices and rents in some regions, and delayed transition to first-time ownership in stressed areas.

### Housing adequacy, overcrowding, and deprivation
- Measures of housing adequacy:
  - Overcrowding rate captures smaller dwelling / sharing choices.
  - Severe housing deprivation captures overcrowding plus poor dwelling quality (leaking roof, no bath/shower and no indoor toilet, or dwelling considered too dark).
- EU-level trends:
  - Housing adequacy has generally improved over time, though improvements for the young were eroded post-pandemic.
  - Positive but weak link between change in housing adequacy and change in perception of housing as an issue across EU members.
- Specific statistics:
  - Overcrowding and severe housing deprivation time series presented for 2015–2023 in figures (no data for 2021 and 2022 for severe housing deprivation).
  - Housing-related arrears and arrears on mortgage or rent, utility bills or hire purchase tracked from 2010 through 2025 (figures provided).

### Key empirical takeaways
- Public concern about housing rose sharply between 2015 and 2025, especially 2022–2024, with affordability the dominant grievance in protests and surveys.
- Multiple surveys (Eurobarometer, OECD Risk That Matters, Eurofound, Gallup, national surveys) consistently show increased concern post-pandemic and heightened vulnerability among younger cohorts.
- Broad aggregate indicators (price-to-income, housing-cost burden, arrears, overcrowding, severe housing deprivation) do not show a clear, sustained deterioration in housing affordability across the EU over the past 10–15 years; some indicators improved relative to the early 2010s.
- Short-lived post-pandemic spikes in housing costs were driven in part by energy price shocks; these spikes contributed to temporary increases in housing insecurity (inability to keep home adequately warm) and heightened public concern.
- Homeownership declined to 68.5 percent in 2025 at the EU27 level, influenced by complex interactions among housing prices, borrowing capacity, macro-prudential measures, and increased institutional investor activity.

*Source: wpiea2026124-source-pdf - 1. Housing: a growing concern...*

### 3. Possible reasons for the gap between perceptions and data

### 3. Possible reasons for the gap between perceptions and data

### Why aggregate data may understate housing affordability problems
- Aggregate, household-level measures can bias downward the measured housing cost burden because individuals respond to unaffordability by household formation changes:
  - Young workers who cannot afford a dwelling may stay longer with parents or share housing, increasing household income and reducing measured burden.
- Housing cost measures are incomplete:
  - Data include rents and interest payments on home loans but not principal payments (considered investment/financial transaction, not consumption; Annex I).
  - The HICP covers consumption and includes rents but excludes owner-occupiers’ housing costs (OOH).
  - The rents component of the HICP covers all existing rents rather than new rents, which can dampen measured rents inflation relative to current market dynamics and contribute to the data–perception gap.
- Jurisdictional adjustments that include OOH can change purchasing-power comparisons:
  - The Catalan government’s adjusted consumer price index that includes the cost of buying a house found that average gross salary increased more than inflation during 2014 and 2023, but inflation adjusted to incorporate the OOH increased more than the average gross salary — implying purchasing power gain versus purchasing power loss depending on the price indicator used.

### Affordability is concentrated in specific groups and countries
- Housing affordability is not uniformly distributed; problems are concentrated among:
  - The young, lower income households, urban populations, and tenants.
- EU averages mask cross-country heterogeneity; affordability has evolved differently across EU countries (Figures 11–17 referenced in source).
- Group- and country-specific statistics from the source:
  - Social housing estimated at 6-7 percent of the housing stock by the EC and at 8 percent by the OECD; shares differ massively across EU countries.
  - In France, 25 percent of the population aged 18 to 34 reports living in housing inadequate in size compared to 15 percent for the population above 35 (Odoxa, 2026).
  - In France, 69 percent of the young report living in housing that is too small, too noisy, too damp, facing heating issues, or too dark — 14 ppts more than for the population aged 35 or more (Odoxa, 2026; Eurofound, 2023).
  - In France, more than 80 percent of new hiring are fixed-term contracts or temporary assignments, which can be an obstacle to access to housing.
  - The share of the population that are tenants at reduced price (or free) declined from 11.1 percent in 2010 to 9.5 percent in 2019; it then increased to reach 10.7 percent in 2025.
  - Survey evidence on mobility and opportunities:
    - In France, 30 percent of the young (aged 18-34) report that due to “housing problem” they could not accept a job (Odoxa, 2026).

### Non-financial dimensions and volatility contributing to concerns
- Perceptions of a housing affordability crisis often reflect concerns beyond direct financial affordability:
  - Availability and adequacy of housing.
  - Access to housing loans and homeownership prospects.
  - Housing and wealth inequalities.
  - Volatility in housing costs and the significant consequences of cost increases.
- The rest of the paper documents numerous and significant consequences of higher housing costs across these dimensions.

### Conceptual framework: three mechanisms through which affordability problems have consequences
- The paper investigates three interacting mechanisms:
  1. Housing costs affect housing adequacy.
  2. Housing costs result in lost opportunities and reduced mobility.
  3. Housing costs force low-income households to cut non-housing consumption.
- These mechanisms can interact (example: inadequate housing increases health hazards and reduced resources for healthcare).

### 1. Housing affordability problems reduce housing adequacy
- Forms of inadequacy:
  - Homelessness:
    - Homelessness has been on the rise in the EU (EC, 2025b and c; Eurofound, 2023; European Economic and Social Committee, 2024).
    - The European Affordability Plan (December 2025) recognizes homelessness is “closely linked to unaffordability of housing” and includes actions such as the EU Anti-Poverty Strategy (to be adopted in 2026) and measures to increase social housing availability.
    - Decline in social housing stock since 2010 (partly due to lower investment and sales to tenants) may have contributed to increased homelessness.
  - Inadequate location:
    - Affordable housing may be unavailable near work or education, increasing commute cost and time and reducing access to infrastructure or adequate security; evidence shows impacts on health, well-being, happiness, and child development.
  - Inadequate housing size (overcrowding):
    - Overcrowding used as proxy for inadequate size; low-income households and the young are most affected (EC, 2025b).
    - Anecdotal/statistic: In France, 25 percent of 18–34 vs 15 percent of 35+ report inadequate size (Odoxa, 2026).
  - Inadequate quality:
    - Measured via severe housing deprivation rate (EU Social Scoreboard); affects rural population and the young more severely.
    - In France, share suffering severe housing deprivation has increased in recent years; 69 percent of the young report multiple housing quality issues (Odoxa, 2026).
- Consequences of inadequate housing:
  - Directly reduces material well-being.
  - Contributes to sense of demotion, economic and financial insecurity, subjective poverty.
  - Negatively affects physical and mental health, productivity, labor participation (absenteeism), income, and creates fiscal costs.
  - Particularly damaging for children’s development and can reinforce intergenerational inequality.
  - New-construction responses to affordability (e.g., declining apartment sizes) can entrench adequacy problems.

### 2. Lost opportunities: impact on poverty, labor force participation, and spatial allocation
- Mechanisms and evidence:
  - High housing costs can prevent relocation to jobs or training or make longer commutes unaffordable, limiting employment and education participation.
  - Survey evidence: In France, 30 percent of young (aged 18–34) report not accepting a job due to “housing problem” (Odoxa, 2026).
  - Access constraints: landlords increasingly require income stability; high share of fixed-term hiring (more than 80 percent in France) impedes access for new labor-market entrants.
- Economic impacts:
  - Turning down jobs or training reduces current and future income and increases poverty risk; “poverty is [like homelessness] closely linked to unaffordability of housing.”
  - Inability to accept jobs can result in unemployment or exit from labor force, lowering productivity and potential growth.
  - Housing affordability can entrench poverty and inequality across generations.
- Spatial misallocation:
  - High housing costs push low-income households out of high-cost metropolitan areas, contributing to shortages of essential-service workers in those areas.
  - Housing costs can prevent workers from moving to high-cost areas, contributing to regional differences in unemployment and hampering productivity growth (documented in United States, Spain, Sweden, and other studies).
  - Swedish Productivity Commission highlighted housing and rental markets as barriers to labor mobility and linked to slowdown in productivity.

### 3. Housing affordability problems constrain non-housing consumption (fertility and health examples)
- General effect:
  - For budget-constrained households with little savings, higher housing costs reduce resources for essential spending: food, healthcare, education, energy.
- Fertility:
  - Survey evidence:
    - In 2021, 26 percent of Poles indicated decision to have a child conditional on housing stability (Eurofound, 2023).
    - In 2024, 87 percent of respondents from 17 EU countries identified housing conditions as barriers to childbearing (OECD, 2025) — simple average of responses from the 17 countries surveyed.
    - In 2025, 82 percent of French aged 18 to 34 reported that, due to housing problem, they had to constrain other types of spending and saving; 29 percent had given up on medical care and 21 percent said they gave up on the idea of having a child (or one more child) due to housing (Odoxa, 2026).
  - Demographic trends:
    - Fertility rate declined from 1.57 in 2010 to 1.34 in 2024.
    - Mean age of women at birth of first child increased from 28.8 in 2013 to 29.8 in 2024.
    - Number of births in the EU was 22.8 percent lower in 2024 than in 2010.
  - Empirical literature highlights multiple pathways:
    - Rising house prices decrease fertility among tenants but raise it for homeowners due to wealth effects (observed in United States, Canada, England, Australia).
    - Access to housing can affect age at first birth and overall fertility (e.g., Brazil, Netherlands, panel of 39 low-fertility countries).
    - Home loans and lending growth influence fertility (e.g., lending growth facilitates transition to parenthood; low interest rates post-GFC increased births among households with adjustable-rate loans in the UK and US).
- Health:
  - Inadequate housing affects physical health via exposure to toxins, poor insulation/dampness, poor living environment, higher injury risk, and resulting illnesses.
  - Housing costs can reduce capacity to afford medical visits, preventive care, healthcare and prescriptions, and can lead to insufficient diet.
  - Mental health impacts:
    - Financial strain from housing costs induces stress; housing affordability affects mental health over and above general financial hardship.
    - Housing quality, neighborhood, and insecurity (including relocations, homelessness, school changes) are associated with depressive symptoms, reduced happiness, and linked to increases in suicide.
  - Intergenerational effects:
    - Parental stress affects children’s stress, health, and development.

### Next steps in the paper
- The paper proceeds to estimate the consequences of housing affordability in the EU over the past 15 years using 15 different estimates and discusses econometric approach, estimation method, data, and validity of estimates.

*Source: IMF Working Paper — 3. Possible reasons for the gap between perceptions and data*

### 1. Econometric approach and estimation method

### 1. Econometric approach and estimation method

### Rationale for methodology
- Estimating effects of housing affordability requires avoiding restrictive parametric assumptions unsuited to complex economic relationships and rich covariate information.
- Two main econometric complications:
  - Considerable number of potentially relevant confounders.
  - Endogeneity concerns: housing affordability co-moves with income, credit conditions, and policy interventions and is influenced by broad structural and cyclical factors.
- Limitations of standard 2SLS in this context:
  - Relies on strong linearity assumptions, parsimonious controls, and prespecified functional form.
  - May perform poorly with rich covariate sets, nonlinear interactions, and heterogeneous treatment effects.
  - Risk of finite-sample bias, reduced efficiency, and omitted variable bias when conditioning set is too small; overfitting and weakened first-stage relationships when expanded.

### Adopted estimator: Double Machine Learning with Instrumental Variables (DML-IV)
- DML-IV is designed for settings with endogenous treatments and high-dimensional controls (Chernozhukov and others (2018) referenced in source).
- Key features that address machine-learning challenges:
  - Neyman‑orthogonal scores: render causal estimate first‑order insensitive to errors in nuisance estimation.
  - Cross‑fitting: prevents overfitting by estimating nuisance components on samples separate from those used for causal estimation.
- DML-IV advantages in this study:
  - Allows inclusion of a rich set of macroeconomic and institutional controls without ad hoc variable selection.
  - Accommodates nonlinearities and interactions present during periods with significant shocks.
  - Provides a natural framework for exploring treatment effect heterogeneity, important given EU cross‑country differences.
  - Suited to panels with a limited number of time periods by exploiting cross‑sectional variation and flexibly controlling for confounders rather than relying on long time series.

### Causal interpretation and instrumental-variable assumptions
- DML‑IV retains standard IV identification assumptions:
  - Relevance and conditional exogeneity of instruments.
  - Instruments must be sufficiently strong and plausibly exogenous to isolate variation in housing costs orthogonal to unobserved determinants of outcomes.
- DML‑IV incorporates instruments in a high‑dimensional environment, allowing rich controls while maintaining inferential validity.
- Ensuring instrument relevance, validity, and sufficient variation remains a prerequisite for correct application.
- The approach orthogonalizes treatment and outcome equations with respect to controls and exploits plausibly exogenous variation in housing affordability to isolate causal responses, conditional on underlying IV assumptions.

### Estimation steps and validation (overview)
- Four-step estimation and validation method (Figure 13 in source):
  1. Start from a rich country-year panel combining social outcomes, housing affordability indicators, macroeconomic controls, and predetermined supply-side instruments.
  2. Estimate nuisance functions: flexibly relate outcomes and housing affordability to observed controls (including country and year effects) using out-of-sample estimation at the country level to absorb common trends, persistent country characteristics, and nonlinear macro relationships.
  3. Causal effect estimation: isolate component of housing affordability associated with predetermined supply-side instruments and relate cleaned outcomes to instrument-driven housing variation.
  4. Robustness and validity checks: instrument relevance, over-identification tests, reverse-causality tests, leave-one-country-out sensitivity, cluster-robust bootstrap inference.
- Technical details and robustness/validity tests are presented in Annex III in the source.

### Instruments used
- Instruments for housing affordability:
  - Lagged construction activity: building permits in square meters per thousand inhabitants taken with a lag of 1 year.
  - Construction costs: annual growth in the construction producer prices for new residential buildings.
- Economic motivation:
  - Higher past construction activity increases effective housing stock with a lag.
  - Increased past construction costs constrain supply and are quickly reflected in housing prices and affordability.
- Empirical validation:
  - Strength and validity of the first-stage relationship supported by additional tests adapted for the DML‑IV framework.
  - Validation includes identification tests for instrument relevance and exogeneity, reverse causality checks, and cluster‑robust bootstrap inference to account for small time series and cluster structure.

### Addressing concerns about instruments and controls
- Potential concern: construction activity and costs may proxy for broader macroeconomic conditions (interest rates, wages, aggregate demand) that directly affect outcomes.
- Responses implemented:
  - Include a rich set of macroeconomic controls, country fixed effects, and time fixed effects (details in Annex IV).
  - Interpret remaining variation in construction costs and lagged permits as reflecting housing supply dynamics driven by regulatory constraints, planning lags, and input cost shocks.
  - Acknowledge the exclusion restriction cannot be tested directly and remains an identifying assumption.
- Combination of rich controls, absence of reverse causality, and supportive overidentification tests provides a consistent picture in favor of the proposed identification strategy, while noting residual correlation with broader forces cannot be entirely ruled out.

### Target parameter and outcomes
- Parameter of interest: the average marginal effect of a change in affordability on socioeconomic outcomes, conditional on observed covariates.
- Endogenous regressor: housing affordability.
- Outcomes used include several social and economic indicators such as poverty, fertility, labor force participation, etc. (full list of specifications in Annex IV).
- DML‑IV orthogonalizes estimating equations so the parameter of interest is locally insensitive to errors in nuisance estimation.

### Estimation inference
- In lieu of standard inference methods, apply cluster‑robust bootstrap to account for the relatively small time series and cluster structure of the data.
- Robustness and validity tests and further technical details are provided in Annex III (per source).

### Data sources and transformation
- Data:
  - Annual data from Eurostat covering the EU27 over the period 2010–2023.
  - Eurostat provides harmonized indicators enabling inclusion of rich controls for macroeconomic conditions, labor market dynamics, and social protection systems.
- Variable transformation:
  - All variables are expressed as log deviations from the contemporaneous EU average.
  - For each country and year, variables are transformed into the difference between the natural logarithm of the country‑specific value and the logarithm of the EU27 average in that year.
- Purpose and implications of transformation:
  - Removes common EU‑wide trends (e.g., driven by monetary conditions or synchronized business cycle fluctuations) and centers analysis on country‑specific deviations.
  - Advantages:
    - Mitigates non‑stationarity and scale differences across countries.
    - Improves numerical stability and interpretability.
    - Allows coefficients to be read as elasticities with respect to relative positions within the EU distribution.
    - Abstracts from aggregate EU‑level shocks, strengthening identification of effects driven by cross‑country heterogeneity.
  - Limitation:
    - Developments common across all EU countries (such as EU‑wide policy shifts, global financial shocks, or synchronized macroeconomic trends) are differenced out and thus not captured in estimated effects.

*Source: IMF Working Paper chapter "1. Econometric approach and estimation method" (annual Eurostat data for EU27, 2010–2023). *

### 3. Results

### 3. Results

### Overview of housing affordability measures
- Two housing costs indicators are used: housing cost share in disposable income and housing cost overburden rate.
- Housing cost share in disposable income often gives larger and more consistently significant coefficients because it captures intensity of housing costs across the entire distribution, while the housing cost overburden rate identifies only the share of households with severe affordability distress.
- The impact of a given housing cost burden depends on household income position: for a high-income household spending 40 percent or more of disposable income on housing (threshold for housing cost overburden) may not be a burden, while for a low-income household spending 10 percent may be problematic.
- The housing cost share in disposable income is the preferred measure to assess consequences of housing affordability.

### A. Impact on housing adequacy
- General:
  - Affordability pressures primarily translate into deteriorating living standards rather than immediate demographic shifts.
  - Estimates using housing cost share in disposable income have the expected sign and are significant at 10 percent or less.
  - Housing adequacy (overcrowding rate and severe housing deprivation) is a core adjustment margin; estimates are most often significant at 1 percent.
  - Housing costs also deteriorate health status, reduce labor force participation, and increase at-risk-of-poverty rate.
  - Immediate impact on fertility and age of first childbearing is comparatively less pronounced but highly statistically significant.
- Magnitude (Table 2 interpretations preserved):
  - 1 percent increase in housing cost share in disposable income (relative to the EU average) increases the overcrowding rate by 2.34 percent.
    - Example: If a country’s share of housing cost in disposable income is 0.2 ppt higher than EU average of 20.8 percent, its overcrowding rate would be 0.4 ppt higher than EU average of 17.4 percent.
  - 1 percent increase in the housing cost overburden rate increases the overcrowding rate by 1.41 percent for the entire population.
    - Example: If a country’s housing cost overburden rate is 0.1 ppt higher than EU average of 9.9 percent, its severe housing overcrowding rate would be 0.25 ppts higher than EU average of 17.4 percent.
  - 1 percent increase in housing cost share in disposable income (relative to the EU average) increases the severe housing deprivation rate by 1.43 percent.
    - Example: If a country’s share of housing cost in disposable income is 0.2 ppt higher than EU average, its severe housing deprivation rate would be less than 0.1 ppt higher than EU average of 5.0 percent.
- Young population and leaving parental home:
  - 1 percent increase in housing cost share in disposable income increases the average age of leaving parents by 0.08 percent, leaving the average age virtually unchanged at 26.5 years.
  - This result is (barely) significant at 10 percent; no robust and significant impact of housing overburden rate on age of leaving parental house is found.
  - Thus the strong impact on the young is not primarily due to delayed leaving of parental home.
- Temporal and regional variation:
  - The impact on housing adequacy was increasing before the pandemic but declined afterwards; the change is sizable for housing cost in disposable income but more limited for housing cost overburden rate.
  - Possible drivers: decline in stock of social housing since 2010 may have increased pre-pandemic impact; post-pandemic developments like remote work may have partially offset impacts.
  - Impact differs across EU Member States:
    - Always negative in Central, Eastern Europe, Southern Europe, and Ireland.
    - In Western Europe and Nordic countries, the impact depends on the indicator: an increase in overburden rate always increases overcrowding; impact of housing cost share in disposable income varies by country.
- Role of social transfers:
  - Having social spending 1 percent higher than EU level reduces:
    - Impact of housing cost share in disposable income on overcrowding from 2.34 percent to about 2.11 percent.
    - Impact on severe housing deprivation from 1.43 percent to about 1.10 percent.
    - Impact of housing overburden rate on overcrowding from 1.41 percent to about 1.29 percent.
  - Social transfers also reduce impact of housing cost in disposable income on average age of leaving parents by about a quarter.

### B. Impact on labor force participation
- Net effect expectation is ambiguous (incentive to increase labor supply vs. mobility constraints reducing opportunities).
- Empirical estimates (Table 3 interpretations preserved):
  - 1 percent increase in housing cost share in disposable income reduces labor force participation (LFP) rate by 0.53 percent for the total population.
  - 1 percent increase in housing cost share in disposable income reduces female LFP by 0.59 percent.
    - Example: If a country’s share of housing cost in disposable income is 0.2 ppt higher than the EU average of 20.8 percent, its LFP would be 0.4 ppt lower than EU average (71.7 percent compared to 72.1 percent); female LFP would be 0.4 ppt lower than EU average (66.2 percent compared to 66.6 percent).
- Temporal and regional variation:
  - Negative impact on LFP is declining post-pandemic; decline is marginal for total population but larger for female population.
  - While negative impact on total LFP is experienced across all EU countries, it is not uniform for female LFP.
  - Differences may reflect legislations, practices related to female labor, and access to childcare.
- Childcare and social benefits:
  - Higher enrollment in childcare is associated with higher female LFP.
  - Lower enrollment in childcare of children below 3 increases the negative impact of housing cost on female LFP marginally (from -0.59 to about -0.60).
  - Social benefits mitigate the LFP impact of housing affordability, though the impact is relatively small for both total and female LFP.

### C. Impact on poverty
- Mechanisms: reduced LFP, turning down job offers, and limited access to training/education due to housing costs can increase present and future income poverty.
- Empirical estimates (Table 4 interpretations preserved):
  - 1 percent increase in housing cost share in disposable income increases at-risk-of-poverty rate (AROP) by 0.92 percent for the entire population.
  - 1 percent increase in housing cost overburden rate increases AROP by 0.29 percent for the entire population.
  - 1 percent increase in housing cost overburden rate increases AROP by 0.65 percent for the young (aged 20 to 29).
    - Example: If a country’s share of housing cost in disposable income is 0.2 ppt above the EU average of 20.8 percent, the AROP would be 0.2 ppt higher than EU average of 16.8 percent.
    - Example: If the housing cost overburden rate is 0.1 ppt above EU average of 9.9 percent, the AROP for the entire population would be 0.05 ppt higher than EU average of 16.8 percent.
    - If the housing cost overburden rate of the young is 0.1 ppt above EU average of 12.6 percent, the AROP of the young would be 0.1 ppt higher than EU average of 19.4 percent.
- Cross-country and temporal patterns:
  - Increase in housing costs in disposable income affects all countries; impact has been broadly stable over time though signs of lower impact post-pandemic exist.
  - Impact of housing cost overburden on poverty is muted in some countries, possibly reflecting policy differences (notably social transfers).
- Social transfers:
  - Social transfers marginally reduce the impact of housing cost in disposable income on AROP; offsetting impact is larger for the housing cost overburden rate, especially for the young.
- Note on AROP definition (from source):
  - The at-risk-of-poverty rate is the proportion of persons with an equivalized disposable income below that 60 percent of the national median equivalized disposable income after social transfers.

### D. Impact on fertility
- Measurement note:
  - Housing cost indicators aggregate costs for both tenants and homeowners; they account for variations in homeownership rates, mortgage market differences, real estate taxation, and energy price increases following the war in Ukraine. They do not capture direct changes in the cost of buying a house.
- Empirical estimates (Table 5 interpretations preserved):
  - 1 percent increase in share of housing cost in disposable income increases the average age at which women have their first child by 0.21 percent.
    - Example: If a country’s share of housing cost in disposable income is 0.2 ppt higher than EU average (21.0 percent instead of 20.8 percent), women would have their first child on average less than 1 months later than the EU average (at 29.4 years instead of 29.3).
  - 1 percent increase in share of housing cost in disposable income (relative to the EU average) lowers the fertility rate by 0.56 percent.
    - Example: If a country’s share of housing cost in disposable income is 0.2 ppt higher than EU average, the fertility rate would be 0.1 lower than the EU average of 1.51.
  - Housing cost overburden impact on fertility is even more limited and less statistically significant:
    - If housing cost overburden rate is 10.0 percent instead of EU average of 9.9 percent, the fertility rate would be 1.509 compared to an EU average of 1.507.
- Channels and magnitude:
  - Delaying childbearing contributes to reduction in total fertility rate.
  - The impact of housing costs on average age of leaving parental house is very small, so delay in leaving parental home is not a meaningful channel for fertility effects.
- Temporal and regional variation:
  - Negative impact on fertility differs across the EU and has changed over time:
    - Negative impact driven by South and South-Eastern Europe and Ireland.
    - Impact is positive in Western Europe and the Nordics.
  - Impact of housing costs on fertility and age of first childbirth was increasingly negative before the pandemic but has declined since the pandemic.
  - Possible post-pandemic drivers: development of work from home reduced demographic impact by allowing moves to cheaper housing areas or reducing commuting time and childcare burdens, which can foster fertility.
- Role of social spending and childcare:
  - Social spending reduces the impact of housing costs on fertility by about 11 percent:
    - When social spending exceeds EU average by 1 percent, it reduces the impact of housing overburden on fertility by 0.01 percentage point (from -0.08 to about -0.07).
    - It reduces the negative impact of higher housing costs as a share of income from -0.56 to about -0.50.
    - Impact on age of first childbirth is reduced from 0.21 to about 0.20 (about an 8 percent reduction).
  - Childcare reduces the impact of housing affordability issues on women’s labor force participation but does not increase fertility in the estimates.

*Source: IMF Working Paper — “The (Many) Consequences of the Housing Affordability Problem in the EU”, 3. Results (authors’ calculations).*

### 0.32 children per women (about 14 percent) higher than for couples where neither does.

### wpiea2026124-source-pdf - 0.32 children per women (about 14 percent) higher than for couples where neither does.

### Impact on health
- Core empirical finding:
  - A 1 percent increase in the share of housing cost in disposable income relative to the EU average increases by 1 percent the share of population perceiving their health as bad or very bad.
- Numerical illustration from the text:
  - If in a country, the share of housing cost in disposable income is 0.2 ppt higher than the EU average of 20.8 percent, the share of population perceiving their health as bad or very bad would be 0.1 ppt higher than the EU average of 9.2 percent.
- Geographic and temporal patterns:
  - The negative impact of housing cost on health is generalized across the EU but is stronger in Southern and Eastern part as well as Ireland.
  - The impact of housing on health increased before the pandemic but declined afterwards. (Figure 26: No data for 2013.)
- Mechanisms and broader consequences:
  - Housing affordability affects health via stress and material hardship, which harms adults and children, reducing cognitive and socioemotional development, educational achievement, well-being, prospects, and productivity (Hallaert and others, 2023, Newman and Holupka, 2025).
  - Housing affordability problems increase health inequality and reduce population happiness and sense of control (ESS, 2025).
- Selected reported regression table items (Table 6):
  - Health status (Bad or very bad) — Housing cost in disposable income: 1.00*
  - Social benefits excluding pensions: -0.03
  - Value of the explained variable (average): 9.2

### Conclusions (summary of main findings and implications)
- Context and recent drivers:
  - Reports of housing affordability problems in Europe are ubiquitous; housing affordability is high on the European agenda.
  - The inflation shock (notably energy prices increase) triggered by the war in Ukraine increased housing costs faster than income, real price of houses accelerated, and the increase in interest rates reduced home-buying power of households.
  - Despite perceptions, when measured by housing costs in a long-term perspective, housing affordability remains relatively low overall, though another increase in energy prices could rapidly raise housing costs and concerns.
- Three identified pathways through which housing cost burden has consequences:
  - A deterioration in housing adequacy.
  - A loss of opportunities.
  - A reduction in non-housing spending.
- Empirical results on affected outcomes:
  - Housing cost burden affects, in a statistically significant way, housing adequacy, poverty, labor force participation, fertility, and health.
  - Impact magnitudes (relative scale from text):
    - Particularly large effects on housing adequacy, poverty, and health.
    - Somewhat smaller effects on fertility and labor force participation.
    - Small effects on average age at which children leave parental house or on delaying entry into parenthood.
- Temporal trend:
  - Consequences of high housing costs were increasingly severe before the pandemic, but the trend was reversed afterwards; remote work / work from home likely contributed to this reversal.
- Policy implications and recommendations:
  - Complementarities exist between housing policy and other social policies; increasing housing supply (as in the European Affordable Housing Plan) would help alleviate pressure on housing prices and rents, dampening housing costs and negative consequences.
  - Coordinating social spending with housing policies is important because increasing housing supply will take time, while social policies can have an immediate impact.
  - Given widespread perceptions of a housing affordability emergency, it may be worth complementing long-term structural solutions with measures that can deliver faster results.

### Annex I. Definition and Description of Housing Costs
- 1. Housing costs — definition and measurement:
  - Housing cost is defined as the monthly expenses associated with the right to live in a dwelling. This includes the cost of utilities (water, electricity, gas, and heating), structural insurance, mandatory services and charges (e.g., sewage and refuse removal), regular maintenance and repairs, and taxes.
  - For homeowners:
    - Housing cost calculation includes mortgage interest payments net of any tax relief, and gross of housing benefits (i.e., housing benefits should not be subtracted from the total housing cost).
    - It does not include payment of the principal of the mortgage loan (considered a saving/investment rather than a consumption item). This omission leads to disposable income remaining after paying housing costs being overestimated and distorts comparisons between renters and homeowners.
  - For tenants:
    - Housing cost includes rental payments gross of housing benefits (i.e., housing benefits should not be subtracted from the total housing cost).
- 2. Housing cost overburden rate — definition and limitations:
  - Definition:
    - The housing cost overburden rate is the percentage of the population living in a household where the total housing costs ('net' of housing allowances) represent more than 40 percent of disposable income ('net' of housing allowances).
  - Uses:
    - This indicator is used in the Social Scoreboard of the European Pillar of Social Rights and is available by age group.
  - Advantages:
    - It changes with both variations in housing cost and income and allows assessment for young people (age-disaggregated data).
  - Main limitations (as stated in the source):
    - The 40 percent threshold is somewhat arbitrary (other countries such as the United States consider a threshold of 30 percent).
    - The implications of spending more than 40 percent of income on housing are highly dependent on a household’s position in the income distribution; high-income households may spend 40 percent without feeling burdened, while low-income households may find 10 percent problematic.
    - Individuals facing affordability problems may share accommodation or have adult children stay longer with parents; increased household income in such cases can reduce the measured overburden and lead to underestimation of affordability problems.
- 3. Housing cost in disposable household income:
  - Definition:
    - The housing cost in disposable income is the financial burden of housing costs relative to the total income available to the household.
  - Comparison with overburden rate:
    - Unlike the housing cost overburden rate, it is not dependent on a specific threshold and measures the burden for the entire population.
  - Shared characteristics with the overburden rate:
    - Covers both tenants and homeowners; changes due to variation in housing cost or decline in income; calculated at the household level and may underestimate problems if individuals share housing or remain in parental homes.

*Source: IMF Working Papers — The (Many) Consequences of the Housing Affordability Problem in the EU (content from wpiea2026124-source-pdf).*

### Annex II. Disconnect Between Perception of a

### Annex II. Disconnect Between Perception of a Housing Affordability Problems and Various Measures of Housing Affordability

### Overview
- The Annex examines correlations between changes in various housing affordability measures and changes in the perception of housing as a concern (Eurobarometer) across EU27 Member States.
- To limit cultural expectation effects, the analysis focuses on changes in perceptions rather than their levels.

### Housing prices
- Changes in housing prices are not correlated with changes in perception of housing as an issue for either period considered (2015-24 or 2019-24).
- The change in the price-to-income ratio is correlated with the change in perception of housing as a problem, whether considering:
  - the change since 2015 (when the ratio was at its lowest in 2 decades) or post pandemic, and
  - perception of housing as a personal issue or as a national issue.
- Correlation strength (selected R² values reported in figures):
  - Change in housing price index (2015-24) vs Change in perceived housing as a personal issue (2015-25): R² = 0.0328
  - Change in housing price index (2015-24) vs Change in perceived housing as a national issue (2015-25): R² = 8E-05
  - Change in housing price index (2019-24) vs Change in perceived housing as a national issue (2019-25): R² = 0.0027
  - Change in housing price index (2019-24) vs Change in perceived housing as a personal issue (2019-25): R² = 0.0103
- Heterogeneity by homeownership rates:
  - In Newer Member States (highest homeownership rates in EU27), the correlation between change in perception and change in the price-to-income ratio is negative or non-existent depending on the period.
  - For the rest of the EU members, the correlation is positive.
- Correlation strength for price-to-income ratio vs perceptions (selected R² values):
  - Change price-to-income ratio (2015-24) vs Change in perceived housing as a personal issue (2015-25): R² = 0.0715
  - Change price-to-income ratio (2015-24) vs Change in perceived housing as a national issue (2015-25): R² = 0.2498
  - Change price-to-income ratio (2019-24) vs Change in perceived housing as a personal issue (2019-25): R² = 0.0797
  - Change price-to-income ratio (2019-24) vs Change in perceived housing as a national issue (2019-25): R² = 0.1989
  - EU15: Change price-to-income ratio (2015-24) vs Change in perceived housing as a personal issue (2015-25): R² = 0.1624
  - Newer Member States: Change price-to-income ratio (2019-24) vs Change in perceived housing as a personal issue (2019-25): R² = 0.0001
  - Newer Member States: Change price-to-income ratio (2015-24) vs Change in perceived housing as a personal issue (2015-25): R² = 0.0392
- Homeownership rate changes:
  - Change in the homeownership rate is not associated with changes in perceptions over the past decade; the link is weak and disappears for the most recent period when homeownership rate declined the most.
  - Footnote: "29 The positively link is due to Newer Member States. There is no association for the rest of the EU27."
  - Correlation strength examples:
    - Change Ownership Rate (2015-24) vs Change in perceived housing as a personal issue (2015-25): R² = 0.0518
    - Change Ownership Rate (2015-24) vs Change in perceived housing as a national issue (2015-25): R² = 0.0433
    - Change Ownership Rate (2019-24) vs Change in perceived housing as a personal issue (2019-25): R² = 0.0542
    - Change Ownership Rate (2019-24) vs Change in perceived housing as a national issue (2019-25): R² = 0.0293

### Housing costs
- Changes in housing affordability measures (housing-cost-to-disposable-income, housing-cost overburden, housing-related arrears) are not positively correlated with changes in perceptions of housing as an issue (personally or nationally) for any period considered.
- Correlation strength (selected R² values reported in figures):
  - Change in share of housing cost in disposable income (2015-24) vs Change in perceived housing as a personal issue (2015-25): R² = 0.035
  - Change in share of housing cost in disposable income (2015-24) vs Change in perceived housing as a national issue (2015-25): R² = 0.0073
  - Change in share of housing cost in disposable income (2019-24) vs Change in perceived housing as a personal issue (2019-25): R² = 0.1277
  - Change in share of housing cost in disposable income (2019-24) vs Change in perceived housing as a national issue (2019-25): R² = 0.0467
  - Change Housing Cost Overburden Rate (2015-24) vs Change in perceived housing as a personal issue (2015-25): R² = 0.0109
  - Change Housing Cost Overburden Rate (2015-24) vs Change in perceived housing as a national issue (2015-25): R² = 0.0018
  - Change Housing Cost Overburden Rate (2019-24) vs Change in perceived housing as a personal issue (2019-25): R² = 0.0393
  - Change Housing Cost Overburden Rate (2019-24) vs Change in perceived housing as a national issue (2019-25): R² = 0.008
- Definition included in figures:
  - Housing Cost Overburden Rate: "Percentage of the population living in a household where total housing costs (net of housing allowances) represent more than 40 percent of the total disposable household income (net of housing allowances)."
- Housing-related arrears:
  - Change in arrears and change in perceived housing as an issue show very low correlation in presented figures (selected R² values near zero).
  - Examples:
    - Evolution of Arrears and Perception of Housing as a National Issue Since 2015: R² = 0.0041
    - Evolution of Arrears and Perception of Housing as a Personal Issue Since 2015: R² = 0.0062

### Housing adequacy
- Changes in housing adequacy ratios are positively associated with changes in perceptions of housing as a problem, but the link is weak and often driven by outliers.
- Measures and selected R² values:
  - Change in arrears on mortgage or rent, utility bills or hire purchase (2019-24) vs Change in perceived housing as a personal issue (2019-25): R² = 0.011
  - Change in perceived health status as "Bad or very bad" (2015-24) vs Change in perceived housing as a national issue (2019-25): R² = 0.0021
  - Change in overcrowding (2015-24) vs Change in perceived housing as a personal issue (2015-25): R² = 0.0788
  - Change in overcrowding (2015-24) vs Change in perceived housing as a national issue (2015-25): R² = 0.023
  - Change in overcrowding (2019-24) vs Change in perceived housing as a personal issue (2019-25): R² = 0.0388
  - Change in overcrowding (2019-24) vs Change in perceived housing as a national issue (2019-25): R² = 0.0177
  - Change in severe housing deprivation (2015-23) vs Change in perceived housing as a personal issue (2015-25): R² = 0.1367
    - Note: "2/ 2020 for Ireland due to data availability."
  - Change in severe housing deprivation (2015-23) vs Change in perceived housing as a national issue (2015-25): R² = 0.13
  - Change in severe housing deprivation (2019-23) vs Change in perceived housing as a personal issue (2019-25): R² = 0.0337
  - Change in severe housing deprivation (2019-23) vs Change in perceived housing as a national issue (2019-25): R² = 0.054

### Key takeaways
- Overall correlations between objective housing affordability measures and changes in public perceptions of housing as a problem are generally weak across EU27.
- The strongest (though still modest) associations observed are for certain price-to-income ratio changes (notably for national-level perceptions) and for severe housing deprivation changes (somewhat stronger R² values, e.g., R² = 0.1367).
- Homeownership status and sample heterogeneity (EU15 vs Newer Member States) matter: Newer Member States often drive different correlation signs/strengths compared with EU15.
- Many relationships are sensitive to the period considered (2015-24 vs 2019-24) and are often influenced by outliers.

*Source: Eurostat, Eurobarometer, and Authors' calculations.*

### Annex III. The Double Machine Learning with

### Annex III. The Double Machine Learning with Instrumental Variables, Instrument Validity, and Diagnostic Checks

### Structural model and setup
- Structural model:
  - Y = θ(X) T + g(X, W) + ε
  - Treatment assignment: T = h(X, W, Z) + ν
- Variables:
  - X: low-dimensional covariates entering the parameter of interest
  - W: potentially high-dimensional vector of controls
  - Z: instrument affecting T but influencing Y only through T
  - θ(X): possibly heterogeneous causal effect of the treatment
- Motivation:
  - Standard linear IV estimators (e.g., 2SLS) impose restrictive functional forms and require strong dimension reduction.
  - DML-IV combines machine learning with orthogonalized moment conditions to address high dimensionality and nonlinearities.

### Identification assumptions
- Relevance:
  - E[T | X, W, Z] ≠ E[T | X, W]
  - The instrument Z induces variation in the treatment after conditioning on observables.
- Exogeneity (exclusion restriction):
  - E[ε | X, W, Z] = E[ε | X, W]
  - The instrument affects the outcome only through the treatment.
- Overlap (common support):
  - The conditional distribution of Z given (X, W) exhibits sufficient variation to identify the effect of T.
- Note:
  - These assumptions are identical in spirit to those required for classical IV estimation; DML-IV does not weaken identification requirements but improves robustness to high dimensionality and functional form misspecification.

### Nuisance function estimation and orthogonalization
- Estimated conditional expectations (treated as nuisance parameters and estimated with flexible machine learning):
  - Outcome regression: mY(X, W) = E[Y | X, W]
  - Treatment regression without instrument: mT(X, W) = E[T | X, W]
  - Treatment regression with instrument: mT(X, W, Z) = E[T | X, W, Z]
- Residualized (orthogonalized) variables using estimated nuisance functions:
  - Ỹ = Y − m̂Y(X, W)
  - T̃ = m̂T(X, W, Z) − m̂T(X, W)
- Interpretation:
  - Ỹ removes variation in Y explained by observed covariates.
  - T̃ isolates the component of T driven exclusively by the instrument, net of controls.
  - Orthogonalization eliminates regularization bias from machine learning estimation of nuisance functions.
- Final orthogonalized equation:
  - Ỹ = θ(X) T̃ + u, where u is orthogonal to T̃.

### Estimation of treatment effects
- Practical approximation:
  - θ(X) often approximated by a linear function of X to allow observed heterogeneity while keeping interpretability.
  - Final-stage estimation carried out using elastic net on residualized variables.
- Treatment effect summaries:
  - Average treatment effect (ATE): ATE = E_X(θ̂(X_it))
  - Conditional average treatment effect for i-th country and t-th year: CATE_it = θ̂(X_it)
  - Approach: estimate a global causal model with countries and periods entering as fixed effects and through features; obtain ATE and CATE by averaging predicted values from the globally estimated causal effect.

### Sample splitting, cross-fitting, and asymptotics
- Sample splitting and cross-fitting:
  - Sample divided into folds.
  - Nuisance functions estimated on one subsample and evaluated on another to avoid overfitting and ensure valid inference.
  - Cross-fitting improves efficiency and stability while preserving orthogonality for asymptotic normality.
- Asymptotic properties (under regularity conditions):
  - √n-consistent (estimation error shrinks at the rate of 1/√n as sample size n increases)
  - Asymptotically normal
  - Robust to high-dimensional controls and flexible functional forms

### Advantages and limitations relative to 2SLS
- Advantages:
  - Functional form flexibility: nonlinear relationships allowed
  - High-dimensional controls: large covariate sets can be included without ad hoc selection
  - Heterogeneous treatment effects: effects may vary with observed characteristics
  - Bias reduction: orthogonalization mitigates overfitting and regularization bias
- Limitations:
  - Higher computational complexity
  - More demanding implementation
  - Less transparent mechanics compared to linear IV models
- Interpretation:
  - DML-IV preserves the causal interpretation of IVs; estimated parameter captures causal effect of treatment variation induced by the instrument, conditional on observed characteristics, analogous to a local average treatment effect under standard IV assumptions.

### Cluster bootstrap inference (wild cluster bootstrap algorithm)
- Context:
  - Panel data with a short time span; clustering at the country level (27 cross-sectional units, time dimension of around 10 years).
  - Wild cluster bootstrap accounts for within-cluster dependence and finite-sample uncertainty.
- Implemented algorithm steps:
  1. Cross-fitting of the nuisance functions mY(X, W), mT(X, W) and mT(X, W, Z) with grouped cross-validation to account for data clusters created by the presence of data of different countries.
  2. Calculation of the orthogonalized outcome and instrument-induced treatment variation Ỹ and T̃.
  3. Estimation of the final-stage regression Ỹ = θ(X) T̃ + u.
  4. Calculation of the ATE as the sample average of the predicted treatment effect.
  5. Wild cluster bootstrap using Rademacher weights. For each bootstrap replication:
     - Generate cluster-level Rademacher weights {w_g} ∈ {−1, +1}
     - Multiply the residuals û_i by the corresponding cluster weight w_g(i)
     - Construct pseudo-outcomes Y_i* = Ŷ_i + w_g(i) û_i
     - Re-estimate the final-stage regression using Y* and compute the bootstrap ATE τ̂_b*
  6. Inference and confidence intervals:
     - Bootstrap standard error = standard deviation of {τ̂_b*}_b=1^B
     - Construct percentile-based confidence intervals
     - Two-sided p-value obtained by comparing absolute deviation of bootstrap draws from the original estimate
- Rationale for clustering:
  - Clustering at the country level allows for arbitrary intra-country correlation over time, important when the number of time periods is small and standard asymptotic approximations may be unreliable.

### Instrument validity and diagnostic checks
- Dimension assessed: relevance, exogeneity, and absence of reverse causality.
- Relevance diagnostics:
  - Measured by out-of-sample partial R²: relative reduction in mean squared error when instrument set is added to the treatment model conditional on covariates.
  - Findings:
    - All partial R² values are above the 0.02 threshold.
    - The majority of partial R² values are above 0.05.
    - Instruments provide meaningful variation in housing costs beyond observed controls.
  - Additional tests:
    - Cluster-robust Wald test of joint significance in auxiliary regression of residualized treatment on residualized instruments yields p-values well below conventional thresholds.
    - Time-preserving permutation test of instrument relevance implemented to account for short time dimension and potential serial correlation.
    - Specifications show relevance at 10 percent significance level or better.
    - Exceptions: specifications with at-risk-of-poverty for age 20-29 and severe housing deprivation are somewhat borderline, but very close to the 10 percent threshold.
- Exogeneity diagnostics:
  - Moment-based overidentification (J) test adapted to DML-IV using orthogonalized residuals from final-stage regression with fully out-of-fold predictions.
  - Test whether residuals are uncorrelated with residualized instruments.
  - Because of a relatively small number of clusters, p-values obtained using a wild cluster bootstrap.
  - Findings:
    - Exogeneity cannot be rejected at the 10 percent significance level.
- Causality / reverse causality diagnostics:
  - Auxiliary tests assess whether the outcome predicts the treatment (reverse causality).
  - For each specification, p-values reported for:
    - (i) Granger-style test including one lag of the outcome
    - (ii) Granger-style test including two lags of the outcome
    - (iii) Joint Wald test of the null that all included outcome lags have zero coefficients in the treatment equation
  - Tests conducted conditional on full set of controls, fixed effects, and instruments used in DML-IV estimation.
  - Findings:
    - Across specifications, p-values are generally large, implying that lagged outcomes do not significantly predict the treatment and providing no evidence of reverse causality.
    - The horizontal line in figures indicates the 10 percent significance threshold.

*Source: Annex III. The Double Machine Learning with Instrumental Variables, Instrument Validity, and Diagnostic Checks (wpiea2026124-source-pdf).*

### References

### wpiea2026124-source-pdf - References

### Fertility, Demography, and Housing Prices
- Aksoy, Cevat Giray (2016). Short-Term Effects of House Prices on Birth Rates, London, EBRD, Working Paper 192, 45 p.
- _______________, Jose Maria Barrero, Nicholas Bloom, Katelyn Cranney, Steven J. Davis, Mathias Dolls, and Pablo Zarate (2026). Work from Home and Fertility, Paris and London: CEPR, Discussion Paper DP21250, 59 p.
- Atalay, Kadir, Ang Li, and Stephen Whelan (2021). Housing Wealth, Fertility Intentions and Fertility, Journal of Housing Economics, vol. 54, 8 p.
- Becker, Gary S. (1960), An Economic Analysis of Fertility, Chapter 7 in Demographic and Economic Change in Developed Countries, New York: Columbia University Press and NBER, pp. 209–240.
- Biljanovska, Nina, Chenxu Fu, and Deniz Igan (2023).  Housing Affordability: A New Dataset, Washington, DC: International Monetary Fund, Working Paper WP/23/247, 33 p.
- Brauner-Otto, Sarah H. (2023) Housing and Fertility: a Macro-Level, Multi-Country Investigation, 1993-2017, Housing Studies, vol. 38(4), pp. 569-596.
- Clark, Jeremy and Ana Ferrer (2019). The Effect of House Prices on Fertility: Evidence from Canada, Economics, vol. 13(2019-38), 32 p.
- Cumming, Fergus and Lisa Dettling (2024). Monetary Policy and Birth Rates: The Effect of Mortgage Rate Pass-Through on Fertility, Review of Economic Studies, vol. 91(1), pp. 229–258.
- Dettling, Lisa J. and Melissa S. Kearney (2014). House Prices and Birth Rates: the Impact of the Real Estate Market on the Decision to Have a Baby, Journal of Public Economics, vol. 110, pp. 82–100.
- Doepke, Matthias, Anne Hannusch, Fabian Kindermann, and Michèle Tertilt (2022). The Economics of Fertility: A New Era, Cambridge, MA: NBER, Working Paper 29948, 130 p.
- Kearney, Melissa Schettini and Phillip B. Levine (2025, revised in 2026), Why is Fertility so Low in High Income Countries?  Cambridge, MA: NBER, Working Paper 33989 (revised January 2026) 50 p.
- Lovenheim, Michael F. and Kevin J. Mumford (2013). Do Family Wealth Shocks Affect Fertility Choices? Evidence from the Housing Market, The Review of Economics and Statistics, vol. 95, pp. 464-475
- Malmberg, Bo and Gebrenegus Ghilagaber. Influences of Lending Growth and House Prices on Fertility 1992-2020: Evidence from a Panel of OECD Countries, Stockholm: Stockholm University, mimeo, 2026, 28 p.
- van Doornik, Bernardus, Dimas Fazio, Tarun Ramadorai, and Janis Skrastins (2024). Housing And Fertility, Brasilia: Banco Central Do Brasil, Working Paper 612, 74 p.
- van Wijk, Daniël and Peteke Feijten (2025). Rising House Prices, Falling Fertility? How Rising House Prices Widen Fertility Differences between Tenure Groups, European Journal of Population, vol. 41, 31 p.
- Skirbekk, Vegard (2022). Decline and Prosper! Changing Global Birth Rates and The Advantages of Fewer Children, Cham: Palgrave Macmillan, 396 p.

### Housing Affordability, Mental Health, and Wellbeing
- Amerio, Andrea, Andrea Brambilla, Alessandro Morganti, Andrea Aguglia, Davide Bianchi, Francesca Santi, Luigi Costantini, Anna Odone, Alessandra Costanza, Carlo Signorelli, Gianluca Serafini, Mario Amore. and Stefano Capolongo (2020). COVID-19 Lockdown: Housing Built Environment’s Effects on Mental Health, International Journal of Environmental Research and Public Health, vol. 17(16), 10 p.
- Arundel, Rowan, Ang Li, Emma Baker, and Rebecca Bentley (2024). Housing Unaffordability and Mental Health: Dynamics across Age and Tenure, International Journal of Housing Policy, vol. 24(1), pp. 44-74.
- Bentley, Rebecca, Emma Baker, Kate Mason, S. V. Subramanian, and Anne M Kavanagh (2011). Association between Housing Affordability and Mental Health: a Longitudinal Analysis of a Nationally Representative Household Survey in Australia, American Journal of Epidemiology, vol. 174(7), pp. 753–60.
- Botha, Ferdi, Rebecca Bentley, Ang Li, and Ilan Wiesel (2024). Housing Affordability Stress and Mental Health: The Role of Financial Wellbeing, Australian Economic Papers, vol. 63(3), pp. 473-492.
- Chung, Roger Yat-Nork, Gary Ka-Ki Chung, David Gordon, Jonathan Ka-Long Mak, Ling-Fei Zhang, Dicken Chan, Francisco Tsz Tsun Lai, Hung Wong, and Samuel Yeung-Shan Wong (2020). Housing Affordability Effects on Physical and Mental Health: Household Survey in a Population with the World’s Greatest Housing Affordability Stress, Journal of Epidemiology & Community Health, vol. 74(2), pp. 164-172.
- Fowler, Katherine, Matthew Gladden, Kevin Vagi, Jamar Barnes, and Leroy Frazier (2015). Increase in Suicides Associated with Home eviction and Foreclosure During the US Housing Crisis: Findings from 16 National Violent Death Reporting System States, 2005-2010, American Journal of Public Health, vol. 105 (2), February, pp. 311-316.
- Kirkpatrick, Sharon I and Valerie Tarasuk (2007). Adequacy of Food Spending is Related to Housing Expenditures among Lower-Income Canadian Households, Public Health Nutrition, vol. 10(12), pp. 1464-1473.
- _____________ (2011). Housing Circumstances are Associated with Household Food Access among Low-Income Urban Families, Journal of Urban Health, vol. 88(2), pp. 284-96.
- Newman, Sandra and Scott Holupka (2014). Housing Affordability and Investment in Children, Journal of Housing Economics, vol. 24, pp. 89-110.
- _____________ (2025). Assisted housing and healthy child development, Journal of Housing Economics, vol. 70, 16 p.
- Pollack, Craig Evan, Beth Ann Griffin, and Julia Lynch (2010). Housing affordability and health among homeowners and renters. American Journal of Public Health, vol. 39(6), pp. 515–521.
- Rautio Nina, Svetlana Filatova, Heli Lehtiniemi, Jouko Miettunen (2017). Living Environment and its Relationship to Depressive Mood: A Systematic Review, International Journal of Social Psychiatry, Vol. 64(1), pp. 92-103.
- Singh, Ankur, Zoe Aitken, Emma Baker, and Rebecca Bentley (2020). Do Financial Hardship and Social Support Mediate the Effect of Unaffordable Housing on Mental Health? Social Psychiatry and Psychiatric Epidemiology, vol. 55, 705–713.
- Taylor, Mark P., David J. Pevalin, and Jennifer Todd (2007). The Psychological Costs of Unsustainable Housing Commitments, Psychological Medicine, vol. 37(7), pp. 1027-1036.
- Otodom (2021), Szczęśliwy dom: Badanie dobrostanu Polaków (Happy Home: A Study of the Well-being of Poles), Poznań, 57 p.
- Rhone, Kailyn (2024). The Key to Affordable Living Is Moving In With Your Sibling, Wall Street Journal, December 7.
- Burns, Sarah (2018). Single People ‘Discriminated Against’ on Social Housing, Irish Times, 23 May 2018.

### Housing Markets, Supply, Mobility, and Macroeconomic Implications
- Anthony, Jerry (2023). Housing Affordability and Economic Growth, Housing Policy Debate, vol. 33(5), pp. 1167-1186.
- Gabriel, Stuart and Gary Painter. Why Affordability Matters, Regional Science and Urban Economics, vol. 80, January 2020, 6 p.
- Glaeser, Edward and Joseph Gyourko (2018). The Economic Implications of Housing Supply, Journal of Economic Perspectives, vol. 32(1), pp. 3-30.
- Haffner, Marietta E. A. and Kath Hulse (2021). A fresh look at contemporary perspectives on urban housing affordability, International Journal of Urban Sciences, vol. 25 Sup. 1, Special issue “The Global Crisis in Housing,” pp. 59-79.
- Hsieh, Chag-Ti and Enrico Moretti (2019). Housing Constraints and Spatial Misallocation, American Economic Journal: Macroeconomics, vol. 11(2), pp. 1-39.
- Eliasson. Kent and Olle Westerlund (2024). Housing Markets and Geographical Labour Mobility to High-Productivity Regions: The case of Stockholm, European Urban and Regional Studies, vol. 31(3), pp. 259-280.
- Eiglsperger, Martin, Rodolfo Arioli, Bernhard Goldhammer, Eduardo Gonçalves, and Omiros Kouvavas (2022). Owner-Occupied Housing and Inflation Measurement, ECB Economic Bulletin, Issue 1/2022, pp. 74-92.
- Hallaert, Jean-Jacques, Iglika Vassileva, and Tingyun Chen (2023). Rising Child Poverty in Europe: Mitigating the Scarring from the COVID-19 Pandemic, Washington, DC: International Monetary Fund, Working Paper WP/23/134, 59 p.
- Hick, Rod, Marco Pomati, and Mark Stephens (2022). Housing and Poverty in Europe: Examining the Interconnections in the Face of Rising House Prices, Cardiff: Cardiff University, 37 p.
- ____________________________________ (2025). Housing Affordability and Poverty in Europe: On the Deteriorating Position of Market Renters, Journal of Social Policy, vol. 54(4), pp. 1072-1095.
- Kouvavas, Omiros and Desislava Rusinova (2024). How Big Is the Household Housing Burden? Evidence from the ECB Consumer Expectation Survey, ECB Economic Bulletin, Issue 3/2024, pp. 47-52.
- Nguyen, Ha, Ashwini Arulrajhan, Carlo Pizzinelli, and Ippei Shibata (2026). The Impact of House Prices on Internal Migration: The case of Spain, Washington, DC: International Monetary Fund, Working Paper WP/26/65, 38 p.
- Hsieh, Chag-Ti and Enrico Moretti (2019). Housing Constraints and Spatial Misallocation, American Economic Journal: Macroeconomics, vol. 11(2), pp. 1-39.
- Peverini, Marco (2023). Promoting Rental Housing Affordability in European Cities - Learning from the cases of Milan and Vienna, Springer, SpringerBriefs in Applied Sciences and Technology, 142 p.
- Eliasson. Kent and Olle Westerlund (2024). Housing Markets and Geographical Labour Mobility to High-Productivity Regions: The case of Stockholm, European Urban and Regional Studies, vol. 31(3), pp. 259-280.

### Policy, Data Sources, and Institutional Reports
- European Commission (2018). The Development of Childcare Facilities for Young Children with a View to Increase Female Labour Participation, Strike a Work-Life Balance for Working Parents and Bring about Sustainable and Inclusive Growth in Europe (the “Barcelona objectives”), Brussels: EC, Report from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, Brussels: European Commission, 24 p.
- __________________ (2025a). Housing in the European Union: Marker Development, Underlying Drivers, and Policies, Luxembourg: Publications Office of the European Union, European Economy Discussion Paper 228, 82 p.
- __________________ (2025b). The European Affordable Housing Plan. Strasbourg: EC, Communication from the Commission to the European Parliament, The Council, The European Economic and Social Committee and the Committee of Regions, COM(2025) 1025 final, 21 p.
- __________________ (2025c). Understanding the housing crisis. Strasbourg: EC, Commission Staff Working Document Accompanying the Communication from the Commission to the European Parliament, The Council, The European Economic and Social Committee and the Committee of Regions – European Affordable Housing Plan. SWD(2025) 1053 final, Part 1/2, 176 p.
- European Economic and Social Committee (2024). Conclusions of the EESC Conference on the Housing Crisis in Europe – the way forward? Brussels: EESEC, 5 p.
- Eurofound (2023). Unaffordable and Inadequate Housing in Europe, Luxembourg: Publications Office of the European Union, 76 p.
- ________ (2025). Housing affordability: Approaches to Measurement and Key Data Insights – Background Paper, Dublin: Eurofound, 7 p.
- Eurostat (2023). Owner-Occupied Housing and the Harmonised Index of Consumer Prices, Luxembourg: Publications Office of the European Union, 47 p.
- ______ (2025). Housing in Europe - 2025 Edition, Luxembourg: Eurostat, interactive publication.
- OECD (2023). Main Findings from the 2022 OECD Risks that Matter Survey, Paris: OECD Publishing, 66 p.
- _____ (2024). Affordable Housing Database - indicator PH4.2. Social rental housing stock, Paris: OECD, OECD Affordable Housing Database, last updated November 11, 2025, 6 p.
- _____ (2025). More Effective Social Protection for Stronger Economic Growth: Main Findings from the 2024 OECD Risks that Matter Survey, Paris: OECD Publishing, 48 p.
- World Health Organization (2018). WHO Housing and Health Guidelines, Geneva: WHO, 149 p.
- International Monetary Fund (2025). Sweden: 2025 Article IV Consultation-Press Release; Staff Report; and Statement by the Executive Director for Sweden, Washington D.C.: IMF, Country Report No. 25/79, 76 p.
- Vigers, Benedict and John Reimnitz (2025). Housing Affordability Crisis Hits Wealthy Economies, Gallup World Poll.
- Odoxa (2025). Baromètre du logement et de l’immobilier – Vague 1, Une étude menée par Odoxa pour Nexity et BFM Business, September, 18 p.
- ______ (2026). Les jeunes et l’accès au logement (Young people and access to housing), Une étude menée par Odoxa pour Nexity, 32 p.
- CE Noticias Financieras (2026a). The Rise in Housing Costs Wipes out the Improvement in Purchasing Power over the Last Decade, 23 February.
- ___________________ (2026b). Small Houses are Making a Comeback in Spain: Developers are Sacrificing Square Footage to Adjust Prices, 27 February.
- ___________________ (2026c). The Construction of Affordable Housing Plummeted by 23% in 2025 despite the Housing Affordability Crisis, 27 March.
- Chocron, Véronique (2026). La « galère » des locataires sans CDI (The ordeal of tenants without permanent contracts), Le Monde, 3 February 2026.

### Methods, Measurement, and Machine Learning
- Chernozhukov Victor, Denis Chetverikov, Mert Demirer, Esther Duflo, Christian Hansen, Whitney Newey, James Robins (2018). Double/debiased machine learning for treatment and structural parameters, The Econometrics Journal, vol. 21(1), pp. C1–C68.
- Eiglsperger, Martin, Rodolfo Arioli, Bernhard Goldhammer, Eduardo Gonçalves, and Omiros Kouvavas (2022). Owner-Occupied Housing and Inflation Measurement, ECB Economic Bulletin, Issue 1/2022, pp. 74-92.
- Kouvavas, Omiros and Desislava Rusinova (2024). How Big Is the Household Housing Burden? Evidence from the ECB Consumer Expectation Survey, ECB Economic Bulletin, Issue 3/2024, pp. 47-52.
- Biljanovska, Nina, Chenxu Fu, and Deniz Igan (2023).  Housing Affordability: A New Dataset, Washington, DC: International Monetary Fund, Working Paper WP/23/247, 33 p.
- Nguyen, Ha, Ashwini Arulrajhan, Carlo Pizzinelli, and Ippei Shibata (2026). The Impact of House Prices on Internal Migration: The case of Spain, Washington, DC: International Monetary Fund, Working Paper WP/26/65, 38 p.

*The (Many) Consequences of the Housing Affordability Problem in the EU: A Machine-Learning Empirical Analysis Working Paper No. WP/2026/124*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026124-source-pdf.pdf_