How accurate is this home price forecast?

Across the 51 state-level regions with enough history to test (the 50 states plus the District of Columbia), 51 of 51 actual prices landed inside the model's ±1σ band in the one-year hindcast, and the median absolute error was 1.5%. The hindcast refits the model on data through May 31, 2025, hides the following twelve months, then compares the projection to what actually printed. A one-year test is a sanity check on the shape of the model, not proof that the five-year path is right — there is not enough monthly history here to test a five-year horizon.

Which state is forecast to grow fastest through 2031?

Connecticut carries the fastest projected path: the median line runs from $498,000 as of May 31, 2026 to $661,099 in May 2031, a +32.8% total change, with a year-5 band of $386k to $1.13M. Read the band, not the midpoint. Texas sits at the other end at +12.4% on the median path.

How is the forecast calculated?

Base growth is half the state's trailing Redfin CAGR and half the FHFA 1991–2024 long-run CAGR for that state, an empirical-Bayes shrinkage that keeps a single hot or cold stretch from running the projection. forecast.js asks for a 60-month CAGR first, but the trend carries 60 monthly observations, which reach only 59 months back from the last print, so that lookback returns null on every state file here and the blend falls back to the 36-month CAGR (May 31, 2023 to May 31, 2026), i.e. 3 years, not five. Volatility is half the observed monthly-return σ and half the FHFA σ. The band is applied as exp(±σ×T) in log space, so its half-width grows linearly with the horizon rather than with √T — it is a true ±1σ interval (about 68%) at year 1 and roughly ±2.2σ (about 97%) by year 5. The national long-run anchor is 4.40% a year. Full detail is in the method block on this page, and the code is src/lib/forecast.js.

Is this forecast free?

Yes. No signup, no email wall, no subscription. The same state rows behind this table are served as JSON at /api/states/index.json, and reuse terms are at /fair-use/.

Home Price Forecast by State

A five-year median-price projection for every state, each one shown with its confidence band and its own one-year hindcast error, so you can see how the model did before you decide what to make of it. The projection centers on a blend of each state's trailing Redfin growth and its FHFA long-run rate — it is not a prediction that prices will do any particular thing. Market figures are as of May 31, 2026 (Redfin month-end); the pipeline last ran July 30, 2026. Free, no signup, no paywall. Same numbers available as JSON at /api/states/index.json.

Fastest projected 5-yr path Connecticut +32.8% $498,000 → $661,099 · band $386k–$1.13M

Slowest projected 5-yr path Texas +12.4% $356,100 → $400,205 · band $299k–$535k

1-year hindcast 51 of 51 inside the ±1σ band median absolute error 1.5% · fit through May 31, 2025, next 12 months held out

Five-year home price forecast by state

All 51 state files carry a 60-month Redfin trend, so every one produces a projection. Sorted by total change on the median path from May 31, 2026 to May 2031. The band column is the honest part, and it is wider than "±1σ" suggests: forecast.js widens the band linearly with the horizon, so the year-5 band spans ±2.2σ (about 97% coverage under a lognormal assumption), and in high-volatility states it is very wide. A midpoint quoted without its band is not a forecast.

#	State	Median price (May 2026)	Year 1 median (May 2027)	Year 3 median (May 2029)	Year 5 median (May 2031)	5-yr change (median path)	Year-5 band (±2.2σ)	Blended annual growth	1-yr hindcast error
1	Connecticut	$498,000	$527,032	$590,271	$661,099	+32.8%	$386k – $1.13M	5.83%	+1.3%
2	New Jersey	$579,900	$611,508	$679,985	$756,131	+30.4%	$526k – $1.09M	5.45%	-1.9%
3	New Hampshire	$537,900	$567,065	$630,226	$700,421	+30.2%	$470k – $1.04M	5.42%	-0.6%
4	Wisconsin	$361,600	$380,586	$421,600	$467,034	+29.2%	$319k – $683k	5.25%	+1.3%
5	New York	$620,500	$651,764	$719,098	$793,387	+27.9%	$521k – $1.21M	5.04%	+1.4%
6	California	$887,400	$931,487	$1,026,340	$1,130,852	+27.4%	$704k – $1.82M	4.97%	+0.1%
7	Maine	$439,200	$461,001	$507,902	$559,575	+27.4%	$337k – $928k	4.96%	-2.4%
8	Illinois	$337,900	$354,631	$390,620	$430,261	+27.3%	$284k – $651k	4.95%	+1.8%
9	Pennsylvania	$330,200	$346,508	$381,580	$420,202	+27.3%	$271k – $651k	4.94%	+1.2%
10	Rhode Island	$536,900	$562,786	$618,364	$679,429	+26.5%	$422k – $1.09M	4.82%	-2.1%
11	West Virginia	$265,200	$277,963	$305,363	$335,463	+26.5%	$229k – $492k	4.81%	+4.5%
12	Vermont	$448,400	$469,899	$516,038	$566,707	+26.4%	$296k – $1.09M	4.79%	-3.7%
13	Delaware	$384,500	$402,778	$441,982	$485,001	+26.1%	$308k – $763k	4.75%	+1.6%
14	North Dakota	$311,200	$325,890	$357,382	$391,918	+25.9%	$204k – $754k	4.72%	+0.5%
15	Massachusetts	$688,100	$719,223	$785,757	$858,445	+24.8%	$547k – $1.35M	4.52%	-3.6%
16	Missouri	$297,500	$310,942	$339,675	$371,063	+24.7%	$251k – $549k	4.52%	+0.3%
17	New Mexico	$395,500	$413,345	$451,488	$493,151	+24.7%	$194k – $1.25M	4.51%	+7.4%
18	Ohio	$282,600	$295,200	$322,109	$351,471	+24.4%	$236k – $524k	4.46%	+1.4%
19	Nevada	$481,200	$502,387	$547,601	$596,884	+24.0%	$393k – $907k	4.40%	+0.7%
20	Michigan	$297,900	$310,916	$338,680	$368,923	+23.8%	$229k – $595k	4.37%	+0.6%
21	Alaska	$427,100	$445,260	$483,929	$525,956	+23.1%	$349k – $793k	4.25%	+0.4%
22	Kansas	$316,300	$329,626	$357,985	$388,785	+22.9%	$246k – $615k	4.21%	-0.9%
23	Virginia	$499,300	$520,259	$564,854	$613,270	+22.8%	$399k – $943k	4.20%	+0.2%
24	Maryland	$477,300	$497,273	$539,761	$585,879	+22.7%	$393k – $873k	4.18%	+2.0%
25	District of Columbia	$740,000	$770,382	$834,939	$904,906	+22.3%	$417k – $1.96M	4.11%	+1.6%
26	Tennessee	$413,200	$429,927	$465,438	$503,883	+21.9%	$358k – $708k	4.05%	-0.1%
27	Kentucky	$284,400	$295,662	$319,541	$345,349	+21.4%	$244k – $488k	3.96%	-1.8%
28	Utah	$560,200	$582,064	$628,385	$678,392	+21.1%	$478k – $963k	3.90%	-3.0%
29	Iowa	$258,700	$268,773	$290,111	$313,143	+21.0%	$225k – $435k	3.89%	-1.7%
30	Idaho	$503,400	$522,797	$563,861	$608,151	+20.8%	$395k – $937k	3.85%	-0.6%
31	Alabama	$312,600	$324,256	$348,887	$375,390	+20.1%	$268k – $526k	3.73%	+4.2%
32	Colorado	$617,000	$639,908	$688,306	$740,365	+20.0%	$499k – $1.10M	3.71%	-2.0%
33	Indiana	$287,300	$297,883	$320,234	$344,262	+19.8%	$238k – $498k	3.68%	+0.5%
34	Wyoming	$464,500	$481,541	$517,520	$556,188	+19.7%	$142k – $2.18M	3.67%	+17.3%
35	Arkansas	$275,500	$285,576	$306,848	$329,704	+19.7%	$229k – $475k	3.66%	-1.1%
36	South Dakota	$346,600	$359,180	$385,725	$414,232	+19.5%	$266k – $645k	3.63%	+0.2%
37	Washington	$651,800	$674,798	$723,258	$775,198	+18.9%	$510k – $1.18M	3.53%	-2.5%
38	Minnesota	$372,300	$385,404	$413,012	$442,598	+18.9%	$313k – $626k	3.52%	-1.5%
39	Nebraska	$319,100	$330,306	$353,913	$379,207	+18.8%	$266k – $542k	3.51%	-2.1%
40	Mississippi	$284,300	$294,029	$314,499	$336,393	+18.3%	$220k – $515k	3.42%	+2.4%
41	South Carolina	$394,000	$407,341	$435,392	$465,376	+18.1%	$331k – $653k	3.39%	-2.2%
42	Montana	$528,600	$546,204	$583,189	$622,680	+17.8%	$377k – $1.03M	3.33%	-2.6%
43	Hawaii	$741,300	$765,455	$816,152	$870,206	+17.4%	$518k – $1.46M	3.26%	-3.6%
44	North Carolina	$397,600	$410,389	$437,213	$465,791	+17.2%	$361k – $601k	3.22%	-1.1%
45	Oregon	$525,500	$541,920	$576,316	$612,894	+16.6%	$436k – $862k	3.12%	-2.5%
46	Florida	$421,500	$434,620	$462,099	$491,315	+16.6%	$358k – $675k	3.11%	-0.1%
47	Oklahoma	$264,600	$272,827	$290,058	$308,376	+16.5%	$216k – $441k	3.11%	-1.6%
48	Louisiana	$269,000	$277,268	$294,574	$312,960	+16.3%	$222k – $441k	3.07%	+0.9%
49	Georgia	$389,000	$400,782	$425,426	$451,587	+16.1%	$326k – $626k	3.03%	-1.1%
50	Arizona	$453,800	$467,437	$495,953	$526,209	+16.0%	$372k – $744k	3.01%	-1.4%
51	Texas	$356,100	$364,514	$381,943	$400,205	+12.4%	$299k – $535k	2.36%	+0.8%

Housing market crash risk by state

The same 0–100 composite the dashboard uses, with all five inputs shown rather than rolled into a single opaque number. Each sub-score is normalized 0–100 where 100 is maximum downside pressure, then weighted 30/20/20/15/15. 0 states currently reads elevated (70+) and 14 read moderate (50–69). A dash means that input is not published for that region and the composite silently substitutes a neutral 50 for it: District of Columbia. Treat those scores as partial.

#	State	Crash-risk score (0–100)	Reading	Price-to-income (30%)	Price YoY (20%)	Inventory surplus (20%)	DOM spike (15%)	Permits (15%)
1	Colorado	63	Moderate downside risk	91	44	85	0	67
2	Washington	61	Moderate downside risk	96	51	100	0	15
3	Tennessee	60	Moderate downside risk	76	35	87	20	67
4	Florida	58	Moderate downside risk	68	37	63	19	100
5	Hawaii	58	Moderate downside risk	100	51	68	22	7
6	California	55	Moderate downside risk	100	33	66	1	35
7	Massachusetts	55	Moderate downside risk	94	46	88	2	0
8	Utah	55	Moderate downside risk	75	53	87	0	31
9	Montana	54	Moderate downside risk	100	42	73	0	4
10	Arizona	53	Moderate downside risk	69	48	70	28	27
11	North Carolina	51	Moderate downside risk	63	39	94	21	14
12	Oregon	51	Moderate downside risk	87	50	74	0	0
13	South Carolina	51	Moderate downside risk	69	44	77	19	21
14	New Mexico	50	Moderate downside risk	82	0	53	9	92
15	Idaho	49	Low downside risk	93	44	27	0	47
16	Vermont	49	Low downside risk	64	42	85	31	0
17	New York	48	Low downside risk	100	25	37	39	0
18	Arkansas	46	Low downside risk	34	35	73	21	71
19	Nevada	46	Low downside risk	82	34	60	19	0
20	Rhode Island	46	Low downside risk	78	34	42	3	49
21	Kentucky	44	Low downside risk	30	39	73	22	63
22	Maine	43	Low downside risk	75	35	66	0	0
23	Texas	43	Low downside risk	33	40	75	7	63
24	New Hampshire	42	Low downside risk	61	27	83	14	0
25	District of Columbia	41	Low downside risk	—	23	69	0	—
26	Mississippi	40	Low downside risk	48	21	61	18	41
27	Maryland	39	Low downside risk	34	23	82	13	40
28	New Jersey	38	Low downside risk	64	32	40	3	29
29	Virginia	37	Low downside risk	57	30	71	0	0
30	Delaware	36	Low downside risk	33	21	75	21	24
31	Ohio	36	Low downside risk	16	23	59	9	90
32	Wyoming	36	Low downside risk	77	0	35	19	21
33	Georgia	35	Low downside risk	49	42	55	4	0
34	Illinois	34	Low downside risk	18	21	39	10	100
35	Indiana	33	Low downside risk	17	29	50	0	79
36	Alabama	32	Low downside risk	44	14	70	15	0
37	Missouri	31	Low downside risk	23	27	65	0	37
38	Louisiana	29	Minimal downside risk	28	34	41	0	37
39	Michigan	29	Minimal downside risk	20	27	54	0	49
40	Minnesota	29	Minimal downside risk	21	44	68	0	0
41	Oklahoma	29	Minimal downside risk	19	40	65	5	10
42	West Virginia	29	Minimal downside risk	31	11	47	6	47
43	Iowa	28	Minimal downside risk	1	39	41	3	73
44	North Dakota	28	Minimal downside risk	17	20	28	4	84
45	Alaska	27	Minimal downside risk	37	29	22	0	37
46	Kansas	27	Minimal downside risk	24	37	33	0	37
47	Wisconsin	27	Minimal downside risk	37	17	53	9	0
48	South Dakota	26	Minimal downside risk	37	28	46	0	0
49	Pennsylvania	25	Minimal downside risk	24	23	46	5	22
50	Connecticut	23	Minimal downside risk	52	14	25	0	0
51	Nebraska	22	Minimal downside risk	22	36	16	10	25

Method, limits, and common questions

Base growth = 0.5 × the state's trailing CAGR from the Redfin median-sale-price trend, + 0.5 × the FHFA long-run state CAGR. That is empirical-Bayes shrinkage toward a prior: it keeps one hot stretch or one bad year from carrying a five-year projection. Across the 51 regions the blended rate runs from 2.36% to 5.83% a year.
The trailing leg is 36 months, not 60. forecast.js asks for a 60-month CAGR first, but the trend carries 60 monthly observations, which reach only 59 months back from the last print, so that lookback returns null on every state file here and the blend falls back to the 36-month CAGR (May 31, 2023 to May 31, 2026), i.e. 3 years, not five. The variable it feeds is still named trailingCAGR5y in src/lib/forecast.js, which is misleading in the code and would be misleading here if we repeated it. Practical effect: the observed half of the blend is a 3-year rate, so it reacts faster to the recent market than a true five-year rate would.
Confidence band. σ is 0.5 × the observed monthly-log-return volatility of the state's own trend + 0.5 × the FHFA long-run σ for that state. Read the implementation before you read the band: projectPrices() computes σ_T = σ×√T and then applies it as exp(±σ_T×√T), so the half-width in log space is σ×T. It grows linearly with the horizon instead of with √T, which means the band is a true ±1σ interval (about 68% under a lognormal assumption) only at year 1, and by year 5 it spans ±2.2σ, about 97%. That is wider than the √T scaling a random walk implies. It is not a floor and not a ceiling.
Horizon is hard-capped at five years in the library itself. Anything longer would be extrapolation past what 60 months of monthly data can support.
The FHFA anchor table is external and hard-coded. LONGRUN_HPI_CAGR_FHFA_1991_2024 and HPI_ANNUAL_VOL_FHFA_1991_2024 in src/lib/forecast.js are literal tables attributed to the FHFA All-Transactions House Price Index, 1991Q1–2024Q4. They were not re-derived by this pipeline — the data this site collects only carries FHFA HPI back to 2016 nationally. Audit them against the source: fhfa.gov/data/hpi/datasets. The national anchor is 4.40%/yr.
Hindcast = refit the whole model on the trend minus its last 12 months (cutoff May 31, 2025), project one year forward, then compare to what actually printed. Error is (actual − projected) ÷ projected, so a positive number means the market came in above the projection. Result: 51 of 51 inside the ±1σ band, median absolute error 1.5%. A one-year test cannot validate a five-year horizon; it validates the shape.
Nominal, not inflation-adjusted. Every dollar figure here is nominal. The median state sits at +21.9% over five years on the median path, and a nominal gain of that size is a different thing from a real one.
Crash-risk weights are 30% price-to-income, 20% price YoY, 20% inventory surplus vs long-run average, 15% days-on-market spike, 15% building-permit YoY. Those weights live as a module-private constant in src/lib/crash-risk.js, so they are restated here by hand. Every source behind the factor choices, 6 of them, is listed on the crash-risk methodology page. It is a relative-ranking composite, not a forecast, and it contains no macro inputs — Fed policy, employment, and recession risk are all outside it.
Missing inputs are not penalized, they are averaged. Each normalizer returns a neutral 50 when its input is null, so a region missing an input drifts toward the middle instead of dropping out. That currently affects District of Columbia, shown with a dash in the affected columns.
Dates. Market figures are as of May 31, 2026, the Redfin month-end in the data, identical across all 51 state files. July 30, 2026 is when the pipeline last ran, which is a different thing and is not the date of the data.
How accurate is this home price forecast? Across the 51 state-level regions with enough history to test (the 50 states plus the District of Columbia), 51 of 51 actual prices landed inside the model's ±1σ band in the one-year hindcast, and the median absolute error was 1.5%. The hindcast refits the model on data through May 31, 2025, hides the following twelve months, then compares the projection to what actually printed. A one-year test is a sanity check on the shape of the model, not proof that the five-year path is right — there is not enough monthly history here to test a five-year horizon.
Which state is forecast to grow fastest through 2031? Connecticut carries the fastest projected path: the median line runs from $498,000 as of May 31, 2026 to $661,099 in May 2031, a +32.8% total change, with a year-5 band of $386k to $1.13M. Read the band, not the midpoint. Texas sits at the other end at +12.4% on the median path.
How is the forecast calculated? Base growth is half the state's trailing Redfin CAGR and half the FHFA 1991–2024 long-run CAGR for that state, an empirical-Bayes shrinkage that keeps a single hot or cold stretch from running the projection. forecast.js asks for a 60-month CAGR first, but the trend carries 60 monthly observations, which reach only 59 months back from the last print, so that lookback returns null on every state file here and the blend falls back to the 36-month CAGR (May 31, 2023 to May 31, 2026), i.e. 3 years, not five. Volatility is half the observed monthly-return σ and half the FHFA σ. The band is applied as exp(±σ×T) in log space, so its half-width grows linearly with the horizon rather than with √T — it is a true ±1σ interval (about 68%) at year 1 and roughly ±2.2σ (about 97%) by year 5. The national long-run anchor is 4.40% a year. Full detail is in the method block on this page, and the code is src/lib/forecast.js.
Is this forecast free? Yes. No signup, no email wall, no subscription. The same state rows behind this table are served as JSON at /api/states/index.json, and reuse terms are at /fair-use/. If you want to see how this compares to a paid subscription product, we wrote that up at /compare-vs-reventure/.
Related: per-state charts and the interactive version of this projection live on each state page. See also market movers, the permit supply pipeline, and county distress watch.

Data as of May 31, 2026 (Redfin month-end); pipeline last run July 30, 2026. Sources: Redfin, FRED (St. Louis Fed), FHFA, U.S. Census Bureau, plus the FHFA HPI datasets for the long-run CAGR and volatility anchors. Forecast code: src/lib/forecast.js. Crash-risk code: src/lib/crash-risk.js. Full source list at /sources/. This is published information, not investment or lending advice.