Phoenix, Austin, Tampa: the sunbelt correction, mapped
In six large sunbelt metros — Phoenix, Austin, Tampa, Jacksonville, Raleigh, and Boise — home prices have fallen 1.8% to 6.4% over the last year. That’s given back somewhere between 12% and 27% of the price spike those same metros saw between 2020 and 2022. Our model is bearish on all six for the next 12 months. This piece answers two questions: how much further does each one fall, and is the correction cyclical (it reverses in a year or two) or structural (it doesn’t)?
1. The setup
Between Q2 2020 and Q2 2022, the sunbelt outran the national HPI index by 14.2pp on average, driven by net in-migration, tech-hire clustering, and an exceptional supply of new construction that nonetheless could not keep up with demand. Starting Q3 2023, the dynamic inverted: the same six metros underperformed the national index by 6.1pp through Q4 2025. This note treats that reversal as a natural experiment.
2. Identifying the cohort
We define the "sunbelt correction cohort" as metros satisfying all three of:
- 2020–2022 HPI gain > 35%
- 2023–2025 HPI change < 0%
- Active-inventory growth (Zillow ZHVI + Realtor) > 30% YoY in 2025
Six metros meet all three conditions at the 2026 Q1 cutoff:
| Metro | 2020–22 gain | 2023–25 change | Inventory YoY |
|---|---|---|---|
| Phoenix-Mesa-Chandler | +47% | -6.4% | +52% |
| Austin-Round Rock | +42% | -5.8% | +68% |
| Tampa-St. Petersburg | +45% | -4.9% | +47% |
| Jacksonville | +38% | -3.2% | +44% |
| Raleigh-Cary | +36% | -2.6% | +38% |
| Boise City | +53% | -4.1% | +61% |
3. Decomposition
We decompose the correction into three drivers: (i) price-to-rent reversion, (ii) inventory overhang, and (iii) demand pullback (in-migration rollover). We estimate each driver's contribution by regressing quarterly HPI change on the relevant proxy and summing contributions.
Price-to-rent reversion is the dominant term, contributing -2.8pp of the -4.7pp cohort-average correction. These metros saw P/rent ratios that hit 2x their 2015–2019 means; returning to even 1.3x the pre-pandemic baseline requires a further -1.1pp of price adjustment (or a +13% rent gain, which the ZORI data shows is not happening).
4. Hypothesis test: cyclical vs structural
The question for investors is whether the correction reverts (cyclical) or stalls (structural). We build a two-hypothesis test using the v0.2 model and two counterfactuals:
H1: Cyclical (mean reversion)
If the correction is cyclical, 2026 should see prices continue to fall for another 2–4 quarters until P/rent ratios converge with pre-pandemic norms, then stabilize. Prediction: 2026 HPI change between -1% and -4% for the cohort.
H2: Structural (re-rating)
If the correction is structural, it reflects a permanent re-rating of sunbelt real estate as the cost of capital rises and remote work loses marginal effect. Prediction: prices stabilize at current levels or modestly rise.
5. Implications for portfolio allocation
For long-only equity holders in sunbelt real estate: consider rebalancing exposure toward metros that still pass our top-quartile forecast screen (Dallas-Fort Worth, Houston, Charlotte) and away from the cohort identified above. For fixed-income holders of CMBS and SFR securitizations with exposure to the cohort: we expect modestly higher default rates in the 2026–2027 window for loans originated at 2021–2022 peak valuations. Our forward default rate estimate for sunbelt SFR is 1.6%, versus 0.9% for national.
6. Limits and caveats
Three caveats. First, the cohort is small (n=6) and the panel regression uses only ~120 quarterly observations — the point estimates are noisy at the individual-metro level. Second, we do not attempt to time the bottom; our forecast is cumulative, not path-dependent. Third, local policy changes (property tax caps, zoning reform, corporate relocations) can individually offset the aggregate drift — model signals should be cross-checked against metro-level news before acting.
References
- [1]FHFA HPI metro files, hpi_at_metro.csv, release 2026-03-25.
- [2]Zillow Observed Rent Index (ZORI), metro-level. Inventory figures from Zillow For-Sale Inventory.
- [3]Counterfactual estimated by re-fitting v0.2 with the mean-reversion term's weight set to 0 and re-normalizing remaining weights to sum to 1.