## 1.    Introduction

## Source details

**Canonical URL:** [1.    Introduction](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025024-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025024-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025024-print-pdf.pdf.json)

---

### Overview and motivation
- Global context: economy emerged from the steepest and most coordinated monetary tightening cycle in decades after a long low-rate period (policy rates close to zero between 2009and2020).
- Mortgage-market stylus: FRMs rose relative to ARMs during the low-rate decade; ARMs picked up again during the post-pandemic tightening cycle.
- Core questions:
  - What is the link between monetary policy and the uptake of FRMs relative to ARMs?
  - To what extent does mortgage composition matter for future monetary transmission?

### Key findings (path-dependency and state-dependency)
- Monetary policy affects the composition of new mortgages (path-dependency):
  - Loosening episodes can significantly increase the share of new fixed-rate mortgages in new flows; tightening episodes make adjustable-rate mortgages relatively more prevalent.
  - Quantitative estimate: a rise (fall) of 100bpp in policy rates increases (decreases) the proportion of ARMs in the flow of new mortgages by 10percentage points, on average (about two-thirds of the within-country standard deviation in ARM flows in the sample).
- Mortgage composition affects monetary transmission (state-dependency):
  - Fixed-rate mortgages weaken transmission strength and increase lags.
  - Quantitative estimate: six quarters after a 100bpp policy rate increase (decrease), real private consumption drops (increases) 5percentage points more in a world with only ARMs relative to one with only FRMs, controlling for the level of mortgage debt.

### Data constructed and novelty
- Novel dataset features:
  - Flow data: share of fixed- and adjustable-rate mortgages in new originations and respective interest rates for 27AEs and EMEs.
  - Stock data: composition of the mortgage stock over time for 35AEs and EMEs — first publicly available outstanding-loan fixation composition based on national central bank data (authors' knowledge).
  - Country-specific monetary policy shocks based on surprises vs analyst forecasts (up to the day prior to each announcement), cleaned of information effects.
- Harmonization choices:
  - FRM defined as mortgages whose rate does not reset for at least 12months; ARMs reset within 12months or less. Applied to flows (status at origination) and stocks (residual maturity).
  - 12months cutoff likely underestimates ARMs (mortgages resetting within 2–5 years may be classified as FRMs) but ensures within-country and cross-country comparability.
- Sample features:
  - Unbalanced panel, quarterly frequency, about 15years on average.
  - Flow composition data: 27 economies. Stock composition data: 35 economies.
  - Stylized trend: by 2020, almost half of mortgages outstanding across 35 economies were FRMs, up from 40percent in 2011. ARMs in new originations fell after 2009 and rose after 2020.

### Interpretative mechanism and literature link
- Mechanism:
  - Monetary policy changes alter borrowing costs and household budget constraints, shifting demand toward the mortgage type relatively cheaper or less binding on budgets.
  - FRM rates are generally less responsive to policy than ARM rates, but both move with policy.
  - Higher rates: ARMs (typically cheaper at origination) become more appealing to constrained borrowers; lower rates: FRMs become more attractive despite higher initial costs.
- Consistency with literature:
  - Aligns with findings that higher FRM shares limit monetary transmission (Calza et al.,2013; Corsetti et al.,2022; Pica,2021) and household-level cash-flow channel evidence.

### Methodology overview
- Identification of monetary policy shocks:
  - Shocks = actual policy-rate announcement − average Bloomberg analyst forecasts up to the day prior.
  - Orthogonalization (Bauer and Swanson logic): purge Bloomberg surprises of information effects by regressing on recent RGDP and inflation surprises and national stock-market changes; residuals are shocks aggregated to quarterly frequency.
  - Regression includes: two prior quarterly RGDP surprises, six prior monthly inflation surprises, and change in national stock-price index (day prior vs 180 days earlier).
  - Resulting shocks correlate 0.58 with high-frequency shocks available for five major economies.
- Empirical frameworks:
  - Path effects: IV-local projections (IV-LP) instrumenting Deltarate_{c,t} with country-specific monetary policy shocks; outcomes include ARM share in new originations, ARM/FRM rates, government yields; controls and lags included; horizons h = 0,...,8 quarters.
  - State-dependent effects: composition of ex-ante mortgage stock used in state-dependent IV-LP to estimate how fixation shares alter transmission to real private consumption and other macro outcomes.

### Dataset limitations and caveats
- 12months cutoff may understate ARMs in jurisdictions using 3–5 year fixation periods.
- Flows vs stocks differ conceptually/empirically; flows are volatile and changes in flows can take years to affect stocks.
- Unbalanced coverage across countries and time; Appendix tables detail coverage, definitions, and sources.

### Appendix A.1 — IV-LP with state-dependent interaction (model summary)
- Core regression (symbols preserved):
  - y_{c,t+h} − y_{c,t−1} = α_h + β^h_1 Deltarate_{c,t} × ARMdebt_{c,t−1}^V + β^h_2 Deltarate_{c,t}^V + β^h_3 ARMdebt_{c,t−1} + ∑_{l=0}^3 β^h_l X_{c,t−l} + ∑_{l=1}^4 ρ^h_l ∆y_{c,t−l} + θ^h_t + γ^h_c + ε^h_{c,t+h}
  - Dependent variables: cumulative changes in real GDP, real private consumption, durables consumption, and real house prices.
- Parameter of interest:
  - β_1 measures the differential effect of a 100bpp change in policy rates for each percentage point increase in ARMdebt measured one quarter before the intervention.

### Definition and rescaling of ARMdebt
- ARMdebt = share of adjustable-rate mortgages in the stock of debt, rescaled as a proportion of a country’s household debt-to-GDP ratio.
- Rationale: importance of ARMs depends on both mortgage prevalence and how much households borrow.
- Note: if ARM share is expressed as share of total mortgages, ARMdebt may understate ARMs in the debt-to-GDP ratio since mortgage debt is a subset of household debt.

### Estimation details and inference
- Preferred specification: combines share of ARMs out of mortgage stock and the level of debt-to-GDP.
- Alternatives: separate inclusion of variables and horse-race model.
- Standard errors: Driscoll-Kraay SE with 3 lags; robustness checked with quarter-level clustered SEs.
- Inference robust to alternative SE choices.

### Regression evidence and correlations (selected results)
- Simple correlations (Figure 2 described):
  - ARM share in new flows: weak correlation with policy rates; strong correlation with average FRM interest rates and FRM-ARM spread; non-linear with ARM rates.
- Table 1 (selected coefficients and diagnostics; heteroskedasticity-robust SEs clustered at quarter level):
  - Column (1): Spread coefficient = 8.191*** (SE 0.559).
  - Column (2): Risk premia coefficient = 6.213*** (SE 0.472).
  - Column (3): Spread coefficient = 9.141*** (SE 1.067); Risk premia coefficient = -0.979 (SE 1.124).
  - Column (4): Spread coefficient = 8.752*** (SE 1.210); Risk premia coefficient = -3.237*** (SE 1.125); FRM rates coefficient = 3.149*** (SE 0.851).
  - Observations: 1,354 in each column.
  - R-squared: 0.907, 0.902, 0.907, 0.908 for columns (1)-(4).
  - Time FE: Yes; Country FE: Yes; Controls: Yes.
- Interpretation:
  - Current FRM-ARM spread positively correlated with ARM share in new originations.
  - Risk premia positive in isolation but lose significance and turn negative when controlling for current spreads.
  - Higher contemporaneous FRM borrowing costs correlate with higher ARM share in new originations.
  - Behavioral implication: short-run interest-cost minimization dominates mortgage choice; results in this subsection are associative, not causal.

### Path dependency: rates, spreads, and flow responses (impulse-response evidence)
- Mortgage-rate reactions to policy:
  - FRM rates change by 0.5pp for each percentage point rise (fall) in policy rates.
  - ARM rates respond almost one-for-one with policy rates (~1.0pp per 1 percentage point policy move).
  - FRM responses over time resemble long-term government bond responses; ARM responses resemble short-term government bond yields.
  - All rates revert to initial level after about 4-6quarters.
- Spread consequence:
  - FRM-ARM spread declines (increases) by about 0.5pp for each percentage point increase (decline) in policy rates.
  - Sample average FRM-ARM spread at any point ≈ 1pp; hence after a 1pp tightening FRMs remain 0.5pp more expensive than ARMs on average.
- Flow consequence:
  - Despite spreads declining, ARM share in new originations increases (declines) following tightening (loosening).
  - Quantified: a 1 percentage point increase (decline) in policy rates increases (lowers) relative prevalence of ARMs in new originations by 10percentage point after one year, on average (≈ two thirds of within-country SD).

### Impacts on mortgage-stock composition (stock responses)
- A 100bpp increase (decline) in policy rates increases (reduces) the share of ARMs in stock by about 4percentage points within 12quarters, with effects persisting up to 20quarters.
- This corresponds to about 50percent of a within-country standard deviation in the sample.

### State dependency: transmission conditional on ARM stock
- ARM stock (as proportion of household debt-to-GDP) conditions marginal effects of policy on aggregates:
  - For real private consumption and output:
    - The same 100bpp tightening (loosening) leads to a decline (increase) in real private consumption about 0.05percentage points larger for each additional percentage point increase in the share of ARMs as a proportion of GDP. Differential effects noticeable after six quarters.
    - Differential effects in durables consumption noticeable after two quarters.
    - Differential response in real GDP about 0.02percent (half the consumption differential).
- Mechanisms:
  - Cash-flow channel: ARM resets directly affect disposable income and consumption.
  - Debt level channel: household debt-to-GDP amplifies sensitivity via wealth/collateral through house prices.
- Horse-race evidence:
  - Household-debt-to-GDP ratios matter more for durables consumption transmission.
  - ARM share in stock matters more for GDP and both measures of private consumption, consistent with dominant cash-flow channel when ARM prevalence is higher.

### Quantitative magnitudes and illustrative calculations
- β_1 interpretation: differential effect per percentage point of ARMdebt for a 100bpp policy change.
- Exact illustrative statement:
  - For a given level of household debt, a 100basis points increase (decrease) in policy rates induces a consumption response that is 5 percentage points stronger in a world with 100percent ARMs relative to one with only FRMs (back-of-the-envelope).
- Timing and asymmetry:
  - Differential effects manifest with lags (consumption after six quarters; durables earlier).
  - Effects larger during tightening cycles when refinancing incentives are limited.

### Asymmetry and refinancing channel (Appendix B.1 summary)
- Conceptual point:
  - FRMs with free prepayment can create asymmetry: FRMs impair transmission more during tightening than loosening because borrowers are less likely to refinance in tightenings.
- Empirical approach:
  - Asymmetric local projections include interactions of |MP_{c,t}| with Tightening_{c,t} and Loosening_{c,t}, interacted with ARM_{c,t-1}.
- Country subsamples:
  - “Free prepayment” countries include United States, Denmark, Mexico, Italy, Japan, Canada.
- Main empirical findings:
  - Free prepayment subsample: tightening transmission is significantly weaker the higher the FRM share in stock; no analogous differential during loosening.
  - Penalty subsample: asymmetry much less pronounced; coefficients smaller and often not statistically different from zero.

### Estimation diagnostics and robustness (selected appendix statistics)
- First-stage IV results (coefficient on MP shocks; Driscoll-Kraay SE with 3 lags in parentheses):
  - (ARMflow) 1.088*** (0.189)
  - (FRM rates) 1.117*** (0.170)
  - (ARM rates) 1.117*** (0.169)
  - (FRM-ARM spread) 1.080*** (0.182)
  - (10y govt yield) 1.091*** (0.160)
  - (1y govt yield) 1.080*** (0.175)
- First-stage observations:
  - ARMflow: 1,079; FRM rates: 1,346; ARM rates: 1,344; FRM-ARM spread: 1,344; 10y govt yield: 1,121; 1y govt yield: 1,283.
- First-stage F-tests / Kleibergen-Paap rk Wald F statistics:
  - F-test values reported: ARMflow 33.16; FRM rates 43.36; ARM rates 43.44; FRM-ARM spread 35.34; 10y govt yield 46.70; 1y govt yield 37.93.
  - Kleibergen-Paap rk Wald F statistic ≥ 34 for main IVs; ≥ 24 for ARM stock dynamics robustness.
- Impulse response inference:
  - Shaded 90percent confidence intervals based on Driscoll-Kraay SE with 3 lags reported in figures.
- Sample sizes:
  - Panel for flow correlations: 26 economies.
  - Panel for flow regressions and flow analysis: 27 economies.
  - Table 1 observations: 1,354.

### Policy-relevant implications and conclusions
- Two central conclusions:
  1. Monetary policy affects relative take-up of FRMs vs ARMs: tightening raises ARM take-up in new flows; loosening raises FRM take-up.
  2. Ex-ante composition of outstanding mortgage stock (ARM share) significantly determines monetary policy transmission strength and timing — monetary policy is state-dependent.
- Combined implication:
  - Prolonged loosening phases induce households to take cheap FRMs, which over time reduces monetary authorities’ ability to stabilize the economy ex-post because a larger FRM stock weakens interest-rate pass-through.
- Asymmetric transmission:
  - FRMs introduce both path and state dependence and can more strongly impair transmission during tightening episodes, especially where free FRM prepayment exists.

*Source: wpiea2025024-print-pdf - 1. Introduction; Appendix A.1; Appendix B excerpts from the provided IMF working paper PDF content.*

### 1.    Introduction

### wpiea2025024-print-pdf - 1.    Introduction

### Overview and motivation
- The global economy recently emerged from the steepest and most coordinated monetary tightening cycle in decades, reversing a long period of historically low interest rates during which many central banks kept policy rates close to zero between 2009and2020.
- During the low-rate decade, the prevalence of fixed-rate mortgages (FRMs) rose relative to adjustable-rate mortgages (ARMs) across many countries; ARMs picked up again during the post-pandemic tightening cycle.
- Key questions: What is the link between monetary policy and the uptake of FRMs relative to ARMs? To what extent does mortgage composition matter for future monetary transmission?

### Key findings
- Monetary policy affects the composition of new mortgages (path-dependency):
  - Loosening episodes can significantly increase the share of new fixed-rate mortgages in new flows; tightening episodes make adjustable-rate mortgages relatively more prevalent.
  - Quantitatively: a rise (fall) of 100bpp in policy rates increases (decreases) the proportion of ARMs in the flow of new mortgages by 10percentage points, on average (about two-thirds of the within-country standard deviation in ARM flows in the sample).
- Mortgage composition affects monetary transmission (state-dependency):
  - Fixed-rate mortgages weaken transmission strength and increase lags.
  - Quantitatively: six quarters after a 100bpp policy rate increase (decrease), real private consumption drops (increases) 5percentage points more in a world with only ARMs relative to one with only FRMs, controlling for the level of mortgage debt.

### Data constructed and its novelty
- A novel dataset compiles information on mortgage-market structure for an unbalanced panel of countries at a quarterly frequency for up to two decades.
- Coverage and components:
  - Flow data: share of fixed- and adjustable-rate mortgages in new originations and respective interest rates for 27advanced (AEs) and emerging market economies (EMEs).
  - Stock data: composition of the mortgage stock over time for 35AEs and EMEs — presented as, to the authors' knowledge, the first publicly available source of outstanding-loan fixation composition based on national central bank data.
  - Country-specific monetary policy shocks based on surprises vis-a-vis analyst forecasts (up to the day prior to each announcement), cleaned of information effects, are included to identify causal effects.
- Data measurement and harmonization choices:
  - FRM definition harmonized to include mortgages whose rate does not reset for at least 12months; ARMs reset within 12months or less.
  - This classification applies to both flows (status at origination) and stocks (residual maturity).
  - The chosen 12months cutoff likely underestimates the share of ARMs (mortgages resetting within 2–5 years from origination may be classified as FRMs), but provides consistent within-country and cross-country comparability.
- Sample characteristics:
  - Unbalanced panel covering a wide set of AEs and EMEs, observed at quarterly frequency for about 15years on average.
  - Flow composition data harmonized for 27economies; stock composition data for 35economies.
  - Noted stylized trends: by 2020, almost half of all mortgages outstanding across 35advanced and emerging-market economies were FRMs, up from 40percent in 2011 (data availability precludes reporting the 2009 figure). The share of ARMs in new originations fell steadily after 2009 and rose again after 2020 in the post-pandemic hiking cycle.

### Interpretative mechanism and relation to existing literature
- Mechanism: monetary policy changes alter borrowing costs and household budget constraints, shifting demand toward the mortgage type that is relatively cheaper or less binding on the household budget.
  - FRM rates are generally less responsive to policy than ARM rates, but both rise (fall) with monetary tightening (loosening).
  - As interest rates increase, ARMs — typically cheaper at origination — become more appealing to constrained borrowers; when rates decline and budget constraints ease, FRMs become more attractive despite typically higher initial costs.
- Consistency with related findings:
  - Results are consistent with studies showing higher FRM shares limit monetary transmission (Calza et al.,2013; Corsetti et al.,2022; Pica,2021) and with household-level cash-flow channel evidence in country-specific studies.

### Methodology overview
- Identification of exogenous monetary policy shocks:
  - Monetary policy surprises are defined as the difference between actual policy-rate announcements and the average forecasts submitted to Bloomberg by professional analysts up to the day prior to each announcement.
  - A two-step orthogonalization follows Bauer and Swanson (BS23) logic: Bloomberg policy-rate surprises are purged of information effects by regressing them on recent economic surprises and national stock-market changes to obtain shocks orthogonal to recent RGDP and inflation surprises and stock-market movements.
  - The regression for each country c and announcement a includes two prior quarterly RGDP surprises, six prior monthly inflation surprises, and the change in the national stock-price index on the day prior to the announcement relative to its value 180 days earlier. Residuals from this regression constitute the monetary policy shocks; shocks are aggregated to quarterly frequency.
  - Resulting shocks correlate 0.58 with high-frequency shocks available for five major economies (Appendix Figure A5).
- Empirical framework:
  - Path effects (monetary policy → mortgage flows): instrumental-variable local projections (IV-LP) following Jordà et al. (2015). The key endogenous regressor, change in policy rate (Deltarate_c,t), is instrumented with country-specific monetary policy shocks. Outcomes include the share of ARMs in new originations, interest rates on ARMs and FRMs, and government yields. Controls include contemporaneous and three lags of log changes in real GDP, CPI, nominal house prices, real private consumption, household debt, and four lags of changes in the dependent variable; horizons h = 0,...,8 quarters.
  - State-dependent effects (mortgage stock composition → transmission to aggregates): use the composition of the ex-ante mortgage stock (from the novel stock data) in a state-dependent IV-LP framework to estimate how mortgage fixation shares alter the strength and timing of policy transmission to real private consumption and other macro outcomes.

### Dataset limitations and definitional caveats
- The 12months cutoff for FRM vs ARM classification is shorter than fixation periods used in some jurisdictions (e.g., 3 or 5 years) and therefore likely underestimates ARMs in the sample.
- Flows and stocks differ conceptually and empirically because mortgages are long-dated, flows can be volatile, and definitions differ (origination vs residual maturity), so changes in flows can take years to affect stocks.
- Data coverage varies across countries and time (unbalanced panel); Appendix tables detail country coverage, definitions, and sources.

*Source: wpiea2025024-print-pdf - 1.    Introduction*

### Appendix A.1for more details.

### Appendix A.1for more details.

### Methodology: IV-LP with state-dependent interaction
- Estimation model: an IV-local-projections (IV-LP) specification augmenting equation (2) with an interaction term to capture state-dependent (differential) effects of policy-rate changes depending on the relative prevalence of ARMs in the outstanding debt-to-GDP stock.
- Core regression form (symbols preserved as in source):
  - y_{c,t+h} − y_{c,t−1} = α_h + β^h_1 Deltarate_{c,t} × ARMdebt_{c,t−1}^V + β^h_2 Deltarate_{c,t}^V + β^h_3 ARMdebt_{c,t−1} + ∑_{l=0}^3 β^h_l X_{c,t−l} + ∑_{l=1}^4 ρ^h_l ∆y_{c,t−l} + θ^h_t + γ^h_c + ε^h_{c,t+h}
  - Dependent variables y_{c,t+h} − y_{c,t−1}: cumulative changes in real GDP, real private consumption, durables consumption, and real house prices.
- Parameter of interest:
  - β_1 measures the differential effect of a 100bpp change in policy rates for each percentage point increase in ARMdebt measured one quarter before the monetary policy intervention.

### Definition and rescaling of ARMdebt
- ARMdebt is defined as the share of adjustable-rate mortgages in the stock of debt, rescaled to be expressed as a proportion of a country’s household debt-to-GDP ratio.
- Rationale for rescaling:
  - The importance of ARMs for monetary transmission depends on both how much households borrow and how many households have a mortgage.
  - If only a small percentage of the population has a mortgage, monetary transmission through household cash-flow channels could be muted even if the entire mortgage stock were adjustable-rate.
  - Conversely, when mortgage debt is prevalent, small increases in the share of ARMs in the stock may have important implications for monetary policy transmission.
- Note: In countries where the ARM share is expressed as share of total mortgages, ARMdebt may understate the share of adjustable-rate mortgages in the debt-to-GDP ratio, since mortgage debt is always a subset of household debt.

### Taking stock of the literature: mortgage choice drivers
- Literature findings summarized:
  - Time-series and cross-sectional variation in the relative prevalence of ARMs is positively correlated with the spread between FRM and ARM rates (Badarinza et al., 2018).
  - ARM prevalence is negatively correlated with expectations of future rates or inflation (Andersen et al., 2023; Botsch and Malmendier, 2023; Koijen et al., 2009).
  - Short-run cost minimization is often dominant in mortgage choice decisions.
- Dataset contribution:
  - Expands the cross-sectional dimension relative to existing studies; panel composed of 27 advanced and emerging market economies for which time-varying information on the composition of new mortgage flows was collected.
- Simple correlations (Figure 2, described):
  - The share of adjustable-rate mortgages in new mortgage flows correlates weakly with policy rates, strongly with average interest rates on new fixed-rate mortgages, strongly with the FRM-ARM spread, and non-linearly with ARM rates.
  - Figure construction notes: each dot represents an average across 50 data points around the interval shown; country fixed-effects included; panel includes 26 advanced and emerging market economies for within-country plots.

### Regression evidence: Table 1 (interest spreads, risk premia, and ARM flows)
- Regression setup: tests how flows of new ARMs relate to current FRM-ARM spread and a measure of expected future ARM rates (risk premia), with extensions controlling for contemporaneous level of borrowing costs.
- Definitions:
  - Spread = (FRM rates - ARM rates).
  - Risk premia = difference between current FRM rates and the avg ARM rate over the previous year.
  - FRM rate = average mortgage rate on new FRM loans at any given point in time.
  - Controls include (log) nominal house prices, CPI, real GDP, real private consumption, household debt levels; country and quarter fixed effects included.
  - Heteroskedasticity-robust SEs are clustered at the quarter level.
- Selected coefficient estimates and diagnostics (as reported):
  - Column (1): Spread coefficient = 8.191*** (SE 0.559).
  - Column (2): Risk premia coefficient = 6.213*** (SE 0.472).
  - Column (3): Spread coefficient = 9.141*** (SE 1.067); Risk premia coefficient = -0.979 (SE 1.124).
  - Column (4): Spread coefficient = 8.752*** (SE 1.210); Risk premia coefficient = -3.237*** (SE 1.125); FRM rates coefficient = 3.149*** (SE 0.851).
  - Observations: 1,354 in each column.
  - R-squared: 0.907, 0.902, 0.907, 0.908 for columns (1)-(4) respectively.
  - Time FE: Yes; Country FE: Yes; Controls: Yes.
- Interpretation:
  - Consistent with Badarinza et al. (2018): current FRM-ARM spread is positively correlated with the share of ARMs in new originations.
  - Risk premia are positively correlated with ARM flows in isolation but lose significance and turn negative when controlling for current spreads.
  - When controlling for contemporaneous borrowing costs (FRM rate), higher borrowing costs are significantly correlated with a higher share of ARMs in new mortgage originations, beyond the effects of spreads and risk premia.
  - Overall suggestion: households’ mortgage choice behavior aligns with short-run interest-cost minimization, and the level of borrowing costs plays a key role; results are associations and not interpreted causally in this subsection.

### Path dependency: monetary policy effects on rates, spreads, and mortgage choice
- Impulse-response evidence (Figure 3, described):
  - Mortgage rates react almost instantaneously to a change in monetary policy rates.
  - FRM rates change by 0.5pp for each percentage point rise (fall) in policy rates.
  - ARM rates respond more strongly, increasing almost one-on-one with the change in policy rates.
  - FRM responses over time resemble long-term government bond responses; ARM responses resemble short-term government bond yields.
  - All rates revert back to initial level after about 4-6quarters.
- Consequence for spreads:
  - Because ARM rates respond more than FRM rates, the FRM-ARM spread declines (increases) by about 0.5pp for each percentage point increase (decline) in policy rates.
- Effect on mortgage flows:
  - Despite spreads declining following monetary tightening, the share of ARM flows increases (declines) significantly following monetary tightening (loosening).
  - Quantified effect: a one percentage point increase (decline) in policy rates increases (lowers) the relative prevalence of ARMs in new originations by 10percentage point after one year, on average.
  - Magnitude context: this 10percentage point change corresponds to about two thirds of a within-country standard deviation in ARM flows in this sample.
- Proposed mechanism:
  - Budget constraints alongside short-run cost minimization:
    - Tightening raises borrowing costs on all mortgage products; even if spreads decline, FRM nominal rates rise and FRMs are generally more expensive due to the insurance premium charged by banks.
    - Higher absolute FRM costs can push households toward ARMs because fixed-rate affordability is reduced and underwriting constraints (e.g., debt-to-income thresholds) may bind.
  - Empirical note: Bosshardt et al., 2024 documents many potential borrowers pushed above underwriting debt-to-income thresholds as interest rates rose, generating a substantial decline in home purchasing activity in 2022 and 2023, especially among low-income groups (contextual citation from source).

### Key statistics and temporal dynamics (summary bullets)
- β_1 interprets differential effect per percentage point of ARMdebt for a 100bpp policy change.
- Driscoll-Kraay standard errors with 3 lags used for primary inference.
- Sample sizes and panels:
  - Panel for flows correlations: 26 advanced and emerging market economies (Figure 2 descriptions).
  - Panel for Table 1 regressions and flow analysis: 27 advanced and emerging market economies (data collection for new mortgage originations).
- Regression outcomes (Table 1):
  - Observations: 1,354.
  - R-squared range: 0.902–0.908 across model specifications.
  - Key estimated coefficients: Spread ~ 8.191 to 9.141 (significant), Risk premia effect changes sign when controlling for spread and FRM rates, FRM rate coefficient = 3.149*** in extended specification.
- Policy-rate shocks to mortgage composition:
  - FRM rates: ~0.5pp change per 1 percentage point policy-rate move.
  - ARM rates: ~1.0pp change per 1 percentage point policy-rate move.
  - FRM-ARM spread: ~−0.5pp per 1 percentage point policy-rate increase.
  - ARM share in new originations: ~+10percentage point after one year per 1 percentage point policy-rate increase.

*Source: Appendix A.1 (excerpts) from the provided IMF working paper PDF content.*

### Appendix figure B1provides a reality check on the model and on monetary policy shocks, plotting

### Appendix figure B1provides a reality check on the model and on monetary policy shocks, plotting

### Monetary policy and mortgage choice (flow responses)
- Monetary tightening increases the share of new adjustable-rate mortgages (ARMs); loosening increases the relative prevalence of fixed-rate mortgages (FRMs) in new flows.
- Responses in GDP, house prices, household debt levels and real private consumption all have the expected signs and significance.
- FRM-ARM spread dynamics:
  - The FRM-ARM spread declines on average by about0.5pp on average following a1pp increase in policy rates (Figure3).
  - The average spread between FRMs and ARMs in the sample at any given point in time is about1pp (Figure A2).
  - This implies FRMs would remain0.5pp more expensive than ARMs following a1pp tightening episode.
- Behavior and affordability channel:
  - Borrowers facing binding budget constraints tend to choose the cheaper mortgage product after tightening episodes.
  - As borrowing costs decline following monetary loosening, affordability constraints ease and FRMs become relatively more attractive despite rising FRM-ARM spreads.
  - Evidence is consistent with a general preference for FRMs when budget constraints are less binding, possibly because FRMs shield households from interest rate risk.

### Impacts on mortgage-stock composition (stock responses)
- A100bpp increase (decline) in policy rates increases (reduces) the share of ARMs in stock by about4percentage points within12quarters, with the effect persisting for up to20quarters (Appendix Figure B2).
- This effect corresponds to about50percent of a within-country standard deviation in this sample.

### State dependency: monetary policy transmission and ARM stock
- The stock (ex-ante share) of outstanding ARMs, expressed as a proportion of household debt-to-GDP ratios, conditions the marginal effect of policy rate changes on macro aggregates.
- For consumption and output:
  - The same100bpp tightening (loosening) of policy rates leads to a decline (increase) in real private consumption which is about0.05percentage points larger for each additional percentage point increase in the share of ARMs as a proportion of GDP. These differential effects are noticeable after six quarters.
  - Differential effects in durables consumption are noticeable after two quarters.
  - The differential response in real GDP is about half of this magnitude (0.02percent).
- Interpretation of mechanisms:
  - Two mechanisms potentially operate jointly:
    1. Cash flow channel: mortgage rate resets on ARMs directly affect disposable income and consumption.
    2. Debt level channel: household debt-to-GDP ratios affect sensitivity to interest rates and amplify wealth/collateral channels through house-price effects.
  - Horse-race evidence:
    - Household-debt-to-GDP ratios matter more for transmission to durables consumption.
    - The share of ARMs in stock matters more for transmission to GDP and both measures of private consumption, consistent with a dominant cash-flow channel when ARM prevalence is higher.

### Quantitative magnitudes and illustrative calculation
- For a given level of household debt, the same100bpp policy rate increase will be0.05pp more effective in reducing real private consumption6quarters ahead for each percentage point increase in the share of ARMs in stock.
- These estimates correspond to exactly one standard deviation in cumulative real private consumption changes at6quarters in the sample.
- Back-of-the-envelope: for a given level of household debt-to-GDP, a100basis points increase (decrease) in policy rates induces a consumption response that is5 percentage points stronger in a world with100percent ARMs relative to one with only FRMs.
- Effects are likely more pronounced during tightening cycles when refinancing incentives are limited; FRMs can impair transmission more during tightenings, especially with free prepayment options available.

### Key conclusions and policy-relevant implications
- Two central findings:
  1. Monetary policy affects the relative take-up of FRMs versus ARMs: tightening raises ARM take-up in new flows; loosening raises FRM take-up.
  2. The ex-ante composition of the outstanding mortgage stock (ARM share) is a significant determinant of the strength of monetary policy transmission to the real economy — monetary policy is state-dependent.
- Combined implication:
  - Prolonged loosening phases induce households to take on cheap FRMs, which over time reduces the monetary authority’s ability to stabilize the economy ex-post because a larger FRM stock weakens interest-rate pass-through.
- Asymmetric transmission:
  - FRMs introduce both path and state dependence in monetary policy decisions and can more strongly impair transmission during tightening episodes.

*Source: wpiea2025024-print-pdf - Appendix figure B1provides a reality check on the model and on monetary policy shocks, plotting*

### References

### wpiea2025024-print-pdf - References (Appendices A–B)

### Data coverage and sources (Appendix A)
- Country grouping:
  - Advanced Economies and Emerging Markets listed in Appendix Table A1 (countries enumerated in source).
- Core data series and sources (Appendix Table A2):
  - Policy Interest Rate: Bloomberg
  - Monetary Policy Shocks: Bloomberg
  - Residential House Price: Bank of International Settlements (BIS)
  - Household Credit to GDP: BIS
  - Share of Fixed Rate Mortgages: Flow and Stock: European Central Bank (ECB), National Central Banks
  - Interest Rates on new mortgages (FRMs and ARMs originations): Haver Analytics, ECB, National Central Banks
  - Yields on government bonds: Haver Analytics, OECD, Bloomberg
  - GDP, Headline CPI, GDP per Capita, Private Consumption, Durables Consumption: World Economic Outlook; Haver Analytics for some series
- Coverage windows and definitions:
  - Flow of ARMs (Appendix Table A3) and stock of ARMs (Appendix Table A4) report country-specific coverage windows (examples: United States 1999:Q2–2023:Q2 for flow; United States 2013:Q1–2023:Q1 for stock).
  - Notes specify that where “duration of contract” is used, FRMs are defined as loans not floating at a given quarter, irrespectively of fixation period; rate resets of floating loans may be subject to contracted limits and gradual phasing-in.
- Cross-country housing finance characteristics (Appendix Table A5):
  - Data collected for 2022:Q4 unless otherwise specified.
  - Table reports country-level metrics including Share of Fixed Rate Mortgages (Flow), Share of Fixed Rate Mortgages (Stock), Household Debt to GDP (Ratio), Typical Fixation Period (Years), Availability of Free FRM Prepayment.
  - Footnotes: e.g., Chile — “In Chile all mortgages are inflation-indexed and are thus classified as floating.”; 2022:Q1 and 2021:Q4 timestamps noted for specific entries.

### Construction of monetary policy shocks (Appendix A.1)
- Orthogonalization approach to account for “the central bank reacting to information” channel (Bauer and Swanson, 2023):
  - Regressors used for orthogonalization at announcement-level:
    - Two prior GDP surprises: difference between actual release and mean of analyst forecasts from Bloomberg; different Bloomberg indexes used to maximize coverage; backward window up to 11 months; for most countries the two prior GDP announcements occur within 6–7 months before announcements.
    - Six lags of inflation surprises within a backward-looking window of 11 months.
    - Stock price changes: change in domestic stock market price index on the day prior to announcement relative to its value 180 days earlier; series constructed from largest ETFs listed on each country’s exchange.
  - After orthogonalization, announcement-level shocks are aggregated to quarterly frequency.
  - For pegged economies, shocks for major currencies are used (see Appendix Table A6 on Eurozone and Euro Area pegs).
- Validation and comparability:
  - Figure A5 compares shocks from literature vs estimated shocks: unbalanced panel covering 1999:1 to 2023:1; correlation coefficient for the whole sample is 0.56.
  - One outlying data point for GBR (−0.83, −0.41) was excluded from the figure to improve visibility.

### Baseline first-stage IV results (Appendix Table A7)
- First-stage coefficient on MP shocks (dependent variable: change in policy rates at time t) reported for six IV-LP models (one per dependent variable):
  - Coefficient estimates on MP shocks:
    - (ARMflow) 1.088*** (0.189)
    - (FRM rates) 1.117*** (0.170)
    - (ARM rates) 1.117*** (0.169)
    - (FRM-ARM spread) 1.080*** (0.182)
    - (10y govt yield) 1.091*** (0.160)
    - (1y govt yield) 1.080*** (0.175)
  - Observations:
    - ARMflow: 1,079
    - FRM rates: 1,346
    - ARM rates: 1,344
    - FRM-ARM spread: 1,344
    - 10y govt yield: 1,121
    - 1y govt yield: 1,283
  - F-test values:
    - ARMflow: 33.16
    - FRM rates: 43.36
    - ARM rates: 43.44
    - FRM-ARM spread: 35.34
    - 10y govt yield: 46.70
    - 1y govt yield: 37.93
  - Estimation details:
    - Time fixed effects: Yes
    - Country fixed effects: Yes
    - Controls included: (log) nominal house prices, CPI, real GDP, real private consumption, household debt levels
    - Standard errors: Driscoll-Kraay SE with 3 lags (reported in parentheses)

### Additional results and robustness (Appendix B)
- Response to a 100 bpp change in interest rates (Figure B1):
  - Impulse responses plotted for: real house prices; real private consumption; real durables consumption; real GDP.
  - Country and quarter FE included.
  - Shaded areas represent 90percent confidence intervals based on Driscoll-Kraay SE with 3 lags.
  - Kleibergen-Paap rk Wald F statistic in first stage is equal or greater than 34.
- ARM stock dynamics (Figure B2):
  - Response of ARM stock to a 100 bpp change in interest rates shown over extended projection horizon up to 20 quarters.
  - Controls include four lags of change in ARM stock and in real house prices; real private consumption; real GDP; real durables consumption.
  - Country and quarter FE included.
  - Shaded areas represent 90percent confidence intervals based on Driscoll-Kraay SE with 3 lags.
  - Kleibergen-Paap rk Wald F statistic in first stage is equal or greater than 24.

### Asymmetry and the refinancing channel of monetary policy (Appendix B.1)
- Conceptual insight:
  - FRMs with free mortgage prepayment can create asymmetry: FRMs are likely to impair transmission more during tightening cycles than during loosening cycles because borrowers are much less likely to refinance in tightening cycles (Berger et al., 2021; Eichenbaum et al., 2022).
- Empirical specification:
  - Asymmetric local projections modify equation (2) to include interaction terms of |MP_{c,t}| with Tightening_{c,t} and Loosening_{c,t}, interacted with ARM_{c,t-1}, and control variables and lags as specified in the source. Projection horizon h = 0, . . . , 8.
- Hypotheses tested:
  - H1: FRMs impair transmission during a tightening impulse more than during a loosening one.
  - H2: Asymmetry is more pronounced in countries where FRM prepayment is free (or cheap).
- Country classification on FRM prepayment (Appendix Table A5):
  - Countries where FRM prepayment is free include: United States, Denmark, Mexico, Italy, Japan, Canada (notes in source cite Zanforlin and Espinosa-Vega, 2008; Jappelli and Scognamiglio, 2018; Kishimoto and Kim, 2014).
  - For other countries, prepayment is possible but not free; penalties vary across jurisdictions.
- Main empirical findings:
  - Where FRM prepayment is free (“Free prepayment” subsample):
    - Transmission of a tightening impulse is significantly weaker the higher the share of FRMs in the stock (Figure B3).
    - No analogous differential effect is observed during loosening episodes — implying loosening transmission is less affected by FRM share, consistent with refinancing in loosening phases.
  - Where FRM prepayment is costly (“Penalty” subsample):
    - Degree of asymmetry is much less pronounced.
    - Over time, ARMs (FRMs) help (impair) transmission in a similar way in both tightening and loosening phases; coefficients are smaller and not statistically different from zero.
- Visualization details:
  - Figure B3 plots differential responses of real private consumption to tightening shocks (red) and loosening shocks (blue) depending on ex-ante share of ARM as a percentage of outstanding loan stock.
  - Country and quarter FE included; shaded areas represent 90percent confidence intervals based on Driscoll-Kraay SE with 3 lags.

*Source: Appendix material and References from the provided IMF content unit.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025024-print-pdf.pdf_