The ALEX Market Trend Model
One model, everywhere: the same index carries every comparable sale in an ALEX valuation to today, draws the trend on every market statistics page, and supplies every price outlook.
Section 1 · Table 1 · Today’s reading · all five counties
| County | 12-month change | Index peak | Since peak | 6-month forecast | 6-month range | 12-month forecast | 12-month range |
|---|---|---|---|---|---|---|---|
| Travis | −2.1% | Apr 2022 | −26.9% | −1.3% | −5.9% to +3.4% | −2.5% | −9.8% to +4.9% |
| Williamson | −3.7% | May 2022 | −27.6% | −1.4% | −6.3% to +3.6% | −2.6% | −10.3% to +5.2% |
| Hays | −4.7% | May 2022 | −26.9% | −1.1% | −7.1% to +5.3% | −2.2% | −12.1% to +8.0% |
| Bastrop | −4.6% | Apr 2022 | −21.9% | −1.5% | −11.1% to +9.2% | −2.8% | −18.5% to +14.6% |
| Caldwell | +9.7% | Aug 2022 | −17.5% | −1.4% | −5.6% to +2.9% | −2.5% | −9.2% to +4.1% |
All figures are read from ledger row e79670256f6f (origin September 2026, issued October 8, 2026). The 12-month change is the index from September 2025 to September 2026. Since peak is September 2026 against the county’s highest index month. Forecasts are the median; ranges run from the 10th to the 90th percentile. Caldwell carries the metro forecast.
Section 2 · Figure 1 · Williamson County index and forecast
The crimson line is measured. The gold line and band are the forecast from September 2026. The band is narrower than the real uncertainty (Section 8).
Section 3 · In plain terms
What the trend chart shows
The red line is what actually sold. Each point is the middle sale price per square foot for one month, after seller-paid closing costs and repair credits. A month needs at least 25 sales to be drawn.
The gold line is the market trend. It comes from homes that sold twice. Comparing the same house at two dates removes the question of which homes happened to sell. A month full of large new homes can lift the red line. It does not lift the gold line.
The dashed line is the forecast at 6 and 12 months. The shaded band is the model’s range.
Why it replaced the straight line
The old line was fitted to the same 36 monthly medians it was drawn over. A few unusual months could tilt it, and it knew nothing beyond those 36 points. The new line is estimated from about 289,000 repeat sales across the five-county Austin area, back to 1996.
Which area it describes
The trend and forecast are for the county. They are drawn at the level of the area shown, so the line sits on that area’s own prices. A subdivision or ZIP can move differently. ZIP-level indices are too noisy to draw: ZIP 78633 moved 8.6% in a single month on one.
Section 4 · Table 2 · Data · which repeat sales are kept
ABOR MLS closed residential sales (PropertyType = Residential, StandardStatus = Closed, ClosePrice > $10,000). The price is the net price: ClosePrice less seller-paid buyer closing costs and repair credits. Two sales of one property form a pair.
| Rule | A pair is kept only if |
|---|---|
| Holding period | At least 180 days between the two sales |
| Same house | Living area changed by no more than 10% |
| Plausible return | Absolute annualised log return of at most 0.5 |
| New construction | Not a builder’s first sale resold within 365 days |
| Duplicates | Not the same property and close date, nor the same price within 31 days |
| Geography | Second sale in the five-county MSA: Travis, Williamson, Hays, Bastrop, Caldwell |
| September 2026 vintage | 289,319 clean pairs · May 1996 (Caldwell June 1997) to September 2026 |
Section 5 · The repeat-sales index
Case–Shiller three-stage weighted repeat sales (WRS), monthly, in logs. For pair i with sales in months s < t and log price change yi:
| Stage | Equation |
|---|---|
| 1 · Bailey–Muth–Nourse OLS | yi = It − Is + ei |
| 2 · Interval variance | êi² = a + b·hi + ui, hi = t − s in months |
| 3 · Weighted least squares | weights 1/σ²(hi) from stage 2, σ² floored at 0.1·mean(ê²); Cov(Î) = s²(X′WX)−1, s² = Σw r² / (n − k) |
| County shrinkage | Icounty,shrunk(m) = w·Icounty(m) + (1 − w)·Imetro(m), w = nm / (nm + 20) |
Base month January 2000 (I = 0). Only months connected in the pair graph are estimated; nothing is interpolated. A metro index is fitted on all pairs, and each county with at least 400 pairs gets its own (all five qualify). nm counts pair endpoints in month m. The constant 20 is the production rule and was not tuned. Each index is a real-time vintage: only pairs whose second sale closed by the end of origin month T, frozen in that month’s ledger row.
Section 6 · Placing the index on a market page
The index measures change, not level. On a page with monthly medians pm (net $/SF, months with at least 25 sales) the drawn line is
L(m) = exp( I(m) + c ), c = mean over drawn months of [ log pm − I(m) ]
Least squares on log levels with the slope fixed at one. One constant is estimated from the area’s own sales; the shape is the county’s. The 12-month change is exp(I(T) − I(T−12)) − 1. Forecast points are L(T+h) = exp( I(T) + c + qτ(h) ), τ ∈ {0.10, 0.50, 0.90}.
Section 7 · Table 3 · The v2 market forecast
For each unit (five counties and metro), origin T and horizon h ∈ {6, 12, 24, 36} months, the forecast of I(T+h) − I(T) is an equal-weight quantile average (Vincentisation) of three components at 99 levels. The weights are 1/3 each and are not estimated.
| Component | Point forecast |
|---|---|
| rw | No change: 0. |
| fhfa | Four-quarter log change of the FHFA purchase-only HPI, Austin MSA, NSA, as known at T (each quarter usable 3 months after quarter end), × h/12. Same for every county. |
| bvar_H60 | Four-variable monthly BVAR per unit, 12 lags: index log growth (vintage T), 30-year mortgage rate (MORTGAGE30US), payroll growth (12-month log change of AUST448NA), log months of supply. Minnesota prior, Litterman ridge form, λ1 = 0.2, λ2 = 0.5, λ3 = 1. Observation s weighted 0.5(T−s)/60 (60-month half-life). Posterior-mean coefficients iterated forward; growth summed over h. |
| Ranges (method B) | Qτ = point + σ(T,h)·√h·zτ |
zτ is a weighted quantile of two pooled sets of standardised errors: the component’s own past errors (origins s ≤ T−h across the five counties, expanding from January 2004, used once 24 exist) and FHFA all-transactions index shocks (ATNHPIUS12420Q, ALFRED vintages as of T). Each set is weighted by its number of non-overlapping h-month windows; FHFA carries about 0.76–0.92. σ blends a 24-month-half-life EWMA standard deviation of monthly index growth with its full-history standard deviation, weight 0.5h/24.
Section 8 · Table 4 · Evidence and calibration
| Replay period, metro | CRPS 12 months, v2 | CRPS 12 months, no change | 80% range hit rate, 6 months | 80% range hit rate, 12 months |
|---|---|---|---|---|
| 2005–2015 | 0.0382 | 0.0519 | 65.8% | 68.5% |
| 2016–2022 | 0.0735 | 0.0912 | 61.9% | 63.1% |
CRPS (continuous ranked probability score; lower is better) favours v2 over no change in both periods. The market quantiles are not calibrated: adaptive quantile tracking in V2_SPEC applies to property-level forecasts only. Replay results are exploratory, scored against the final index, which still revises. On the 2023–2026 holdout, the BVAR component against no change at 12 months had a bootstrap p-value of 0.498: no significant gain. ALEX’s forecasting studies found that no market model reliably beat “no change” at 12 months. The forecast is shown because it is the official, recorded model. It is not a promise.
Pre-registered prospective test. One ledger row per month, issued on the 15th; market truth frozen from the index vintage published at T + h + 3 months; formal readout after 60 matured origins. As of October 22, 2026, the newest row is origin September 2026; none has matured.
Section 9 · Provenance and limitations
A record, not a revision. Each month’s model is written once to a hash-chained ledger and never edited. Pages read the newest row and check the chain first. If the chain fails, or the county is outside the five, the page draws the measured medians only.
| Limitation | What it means |
|---|---|
| County level | A county market forecast, not a forecast for one subdivision or one home. |
| Interval width | The ranges are too narrow by the replay evidence in Section 8. |
| Index revisions | The newest months rest on fewer pairs; late-reported sales revise them. |
| Shared inputs | Payroll and FHFA are metro-level series, so every county shares them. |
Section 10 · References
- Bailey, M. J., Muth, R. F. & Nourse, H. O. (1963). A regression method for real estate price index construction. Journal of the American Statistical Association, 58, 933–942.
- Case, K. E. & Shiller, R. J. (1987). Prices of single-family homes since 1970. New England Economic Review, Sept/Oct, 45–56.
- Litterman, R. B. (1986). Forecasting with Bayesian vector autoregressions: five years of experience. Journal of Business & Economic Statistics, 4, 25–38.
- Lichtendahl, K. C., Grushka-Cockayne, Y. & Winkler, R. L. (2013). Is it better to average probabilities or quantiles? Management Science, 59, 1594–1611.
- Smith, J. & Wallis, K. F. (2009). A simple explanation of the forecast combination puzzle. Oxford Bulletin of Economics and Statistics, 71, 331–355.
- Timmermann, A. (2006). Forecast combinations. Handbook of Economic Forecasting, 1, 135–196.
- Gneiting, T. & Raftery, A. E. (2007). Strictly proper scoring rules, prediction, and estimation. Journal of the American Statistical Association, 102, 359–378.
Methodology, not a valuation. This document describes how ALEX builds its market trend and forecast. Every figure in Tables 1–4 and Figure 1 is read from the official v2 ledger row named above or from the v2 replay record; the forecast is a county-level model output with the limitations stated in Sections 8 and 9, not an appraisal and not a promise of future prices. Market statistics pages: alex-companies.com/market-data/. ALEX Companies · Licensed by the Texas Real Estate Commission · Broker: Alexander Tiffany Southwest, LLC, #565194-BB.