MacroMarch 28, 2026 · 13 min read

How Sensitive Is US Housing to the Rate Path?

RE-Invest Research

Research Team

Abstract

We estimate the elasticity of US housing returns to the rate complex using a four-variable structural VAR on 1976–2025 FHFA, FRED, and Census data. Point estimate: a permanent 75bp fall in the 10-year Treasury yield lifts trailing-year HPI growth by 110 ± 38 basis points over the following four quarters, with 60% of the response realized in the first two quarters. Transactions volumes are an order of magnitude more rate-elastic than prices: the same 75bp move unlocks an estimated 420K additional transactions over 12 months, about 8% of steady-state activity.

1. Motivation

Every housing forecast rests on an implicit elasticity of prices to the cost of capital. The 2022–2024 rate cycle provided the largest natural experiment in a generation: a 520bp mortgage-rate shock followed by a 135bp partial reversal. The question this note answers is: by how much, and with what lag, do national HPI and transaction volumes respond to a rate move of a given size?

2. Model and identification

We estimate a 4-variable structural VAR(4) on quarterly data (1976 Q1 – 2025 Q4, n = 200) with the following ordering:

  x₁ = 10-year Treasury yield (DGS10, FRED)
  x₂ = 30-year mortgage rate (MORTGAGE30US, FRED)
  x₃ = Log HPI level (FHFA national, SA)
  x₄ = Existing home transactions (NAR approximate)

Identification uses a Cholesky decomposition with recursive ordering: rate shocks affect mortgage rates contemporaneously, which affect prices and transactions with a 1-quarter lag. We difference variables as needed to achieve stationarity (ADF p < 0.05 on all series post-transformation).

2.1 Why SVAR vs reduced-form regression

A reduced-form regression of HPI on rates conflates demand and supply shocks. An SVAR with Cholesky ordering gives us a clean identification of the rate-shock impulse response under the assumption that rates don't respond contemporaneously to housing conditions (reasonable for quarterly data — the Fed considers housing but doesn't set rates at the quarterly frequency based on HPI).

3. Results

3.1 Impulse response — prices

Figure 1 shows the response of the HPI growth rate to a 75bp downward shock to the 10-year yield. The peak response occurs at quarter 3 (68bp of annualized HPI growth), with a cumulative 110bp over the 12-month window.

Figure 1Impulse response: HPI annualized growth to a 75bp 10Y shock

Solid line: point estimate. Shaded band: 68% confidence interval from 1,000 bootstrap replications of the VAR residuals. Interpretation: one quarter after the rate shock, HPI growth accelerates by 28bp; by quarter 3 the effect peaks at 68bp; cumulative 4-quarter response is 110bp.

3.2 Impulse response — transactions

Transactions are substantially more elastic. A 75bp shock produces an impulse response that peaks at +9.4% of steady-state transaction volume at quarter 2, decaying to +3.1% by quarter 6. The cumulative additional transactions over 12 months is approximately 420,000.

Figure 2Impulse response: Transaction volume to a 75bp 10Y shock

Percent deviation of existing-home transactions from the steady-state level, 12-month rolling. Peak response 9.4% at quarter 2; half-life 4.1 quarters.

3.3 Asymmetry

The effect is asymmetric. Upward rate shocks (tightening) depress transactions more sharply than equivalent downward shocks boost them — the 2022–2024 period saw transactions decline 31% in response to a +520bp mortgage move, while the parametric estimate implied only -24%. We attribute the difference to the inventory-lock effect: homeowners on sub-4% mortgages defer selling, compressing supply.

Figure 3Elasticity by move size and direction

Transaction response by magnitude of the 10Y shock, separated by direction. Rate tightenings (positive shocks) show larger negative responses than the symmetric model predicts, reflecting the inventory-lock mechanism.

4. Cross-sectional heterogeneity

The impulse response varies meaningfully across MSAs. We estimate metro-level elasticities by interacting the rate shock with per-MSA affordability measures. Metros with the highest price-to-income ratios show 1.4× the average price elasticity — a 75bp rate shock produces 154bp of HPI response versus 110bp nationally.

Metro type	Price elasticity	Transaction elasticity	Example
High P/I (>5x)	1.54×	1.21×	San Francisco, San Jose, LA
Mid P/I (3–5x)	1.00×	1.00×	Chicago, Dallas, Atlanta
Low P/I (<3x)	0.68×	0.85×	Cleveland, Detroit, Memphis

5. Implications for 2026

The market implies approximately 125bp of fed-funds cuts by year-end 2026 (Fed Funds Futures, 2026-04-17). If fully delivered, that should translate to roughly 75bp of 10Y movement and thus 110bp of national HPI growth on top of our baseline forecast. Our v0.2 model already incorporates current forward curves in its drift term; if the Fed delivers more than priced, expect our forecasts to shift upward by roughly the differential × 1.5.

6. Caveats

Three. First, the SVAR is a linear model; we know the relationship is nonlinear at extreme rates (the 2020 zero-bound period substantially under-performed what a linear response would predict). Second, our identification assumes rates are exogenous to housing at the quarterly frequency — reasonable historically but possibly less so if the Fed begins targeting housing explicitly. Third, transaction elasticities reflect the median metro; individual markets can deviate by 2–3× depending on supply elasticity (permitting constraints, inventory overhang).

References

[1]FRED series DGS10, MORTGAGE30US, ingested into the re-invest warehouse via etl-fred.
[2]Existing home transactions proxied from NAR-reported monthly totals, quarterly-averaged. Pre-1976 data extrapolated from HUD.
[3]SVAR estimation and impulse response computation performed out-of-band in Python (statsmodels.tsa.vector_ar.svar_model).
[4]Bootstrap confidence intervals based on residual resampling with block length 8.

← All notes Subscribe to new notes →