King County actual vs. rent-predicted homelessness rate

What this dataset measures

Large urban CoCs whose boundaries can be rebuilt from whole Census counties or places, so each homeless count can be matched to the population and rents of the same area. Built to test the common explanation that high housing costs account for Seattle/King County's homelessness.

Geography

United States, Washington, King County, WA, Seattle, WA

Unit of observation

Continuum of Care (CoC) by PIT count year

Update cadence

Annually, after HUD releases each year's PIT counts by CoC

Curator

FactHub

Curated data tables

King County actual vs. rent-predicted homelessness rate

For each year, comparison sample and measure: King County's rate, the rate predicted from rent by a fit that leaves King out, and the 95% prediction interval for the difference.

Unit of observation: Year x comparison sample x measure

king_rent_residuals.csv text/csv · 2.8 KB (2,883 bytes)

All 35 rows 13 columns Download the file

All 35 rows

year	sample	measure	n CoCs	king_median_rent US dollars per month, nominal	actual_rate per 10,000 residents	predicted_rate per 10,000 residents	residual per 10,000 residents	pi95_low per 10,000 residents	pi95_high per 10,000 residents	slope_per_100 per 10,000 residents per $100	slope_se per 10,000 residents per $100	r2
2015	all_full_count	total	66	1354	47.81	29.61	18.2	-26.33	62.73	1.48	0.9	0.04
2016	all_full_count	total	54	1418	49.91	29.5	20.41	-26.65	67.48	1.69	1.15	0.04
2017	all_full_count	total	66	1555	53.2	29.85	23.35	-19.64	66.35	1.42	0.75	0.053
2018	all_full_count	total	55	1674	54.24	31.18	23.06	-18.35	64.47	1.91	0.92	0.074
2019	all_full_count	total	66	1736	49.71	33.51	16.2	-25.07	57.47	1.86	0.66	0.111
2022	all_full_count	total	63	1917	58.97	33.1	25.87	-13.35	65.09	1.81	0.59	0.134
2024	all_full_count	total	59	2110	72.08	38.05	34.03	-22.33	90.39	1.68	0.87	0.062
2015	major_city	total	42	1354	47.81	45.88	1.93	-45.25	49.1	5.03	1.4	0.244
2016	major_city	total	34	1418	49.91	58.02	-8.11	-57.39	41.17	7.85	1.98	0.329
2017	major_city	total	42	1555	53.2	46.97	6.23	-38.39	50.86	4.25	1.1	0.273
2018	major_city	total	34	1674	54.24	61.74	-7.5	-50.63	35.62	6.61	1.47	0.388
2019	major_city	total	42	1736	49.71	52.34	-2.63	-43.38	38.12	4.49	0.89	0.387
2022	major_city	total	39	1917	58.97	49.59	9.38	-27.93	46.68	4.07	0.76	0.439
2024	major_city	total	37	2110	72.08	57.84	14.24	-47.74	76.21	4.53	1.37	0.237
2015	excl_nyc	total	65	1354	47.81	28.14	19.67	-22.64	61.98	1.23	0.86	0.032
2016	excl_nyc	total	53	1418	49.91	27.07	22.84	-21.75	67.42	1.25	1.1	0.025
2017	excl_nyc	total	65	1555	53.2	28.38	24.82	-15.6	65.24	1.27	0.71	0.048
2018	excl_nyc	total	54	1674	54.24	27.97	26.27	-10.89	63.44	1.48	0.84	0.057
2019	excl_nyc	total	65	1736	49.71	31.91	17.8	-20.33	55.94	1.71	0.61	0.112
2022	excl_nyc	total	62	1917	58.97	32.12	26.85	-10.93	64.64	1.73	0.57	0.135
2024	excl_nyc	total	58	2110	72.08	35.06	37.02	-7.03	81.07	1.49	0.68	0.079
2015	all_full_count	unsheltered	66	1354	17.96	9.88	8.08	-8.94	25.09	1.14	0.34	0.147
2016	all_full_count	unsheltered	54	1418	20.95	7.1	13.85	1.18	26.51	0.53	0.31	0.054
2017	all_full_count	unsheltered	66	1555	25.06	11.81	13.25	-3.93	30.43	1.16	0.3	0.188
2018	all_full_count	unsheltered	55	1674	28.3	9.12	19.18	5.91	32.44	0.67	0.3	0.089
2019	all_full_count	unsheltered	66	1736	23.21	15.69	7.52	-13.01	28.04	1.52	0.33	0.253
2022	all_full_count	unsheltered	63	1917	33.9	16.42	17.48	-7.38	42.33	1.53	0.37	0.216
2024	all_full_count	unsheltered	59	2110	41.92	15.96	25.96	0.7	51.22	1.21	0.39	0.144
2015	major_city	unsheltered	42	1354	17.96	15.45	2.51	-15.76	20.79	2.4	0.54	0.328
2016	major_city	unsheltered	34	1418	20.95	10.05	10.9	-3.28	25.08	1.12	0.57	0.108
2017	major_city	unsheltered	42	1555	25.06	18.8	6.26	-11.04	23.55	2.35	0.43	0.432
2018	major_city	unsheltered	34	1674	28.3	13	15.3	-0.08	30.67	1.22	0.52	0.145
2019	major_city	unsheltered	42	1736	23.21	24.04	-0.83	-21.56	19.9	2.72	0.45	0.473
2022	major_city	unsheltered	39	1917	33.9	24.66	9.24	-17.59	36.08	2.61	0.54	0.383
2024	major_city	unsheltered	37	2110	41.92	22.27	19.65	-7.35	46.66	2.14	0.6	0.268

13 field definitions

Field	Type	Meaning	Units	Value type
year	integer	PIT year (full-count years only)	—	observed
sample	string	all_full_count, major_city (HUD 2025 Major City CoCs) or excl_nyc	—	derived
measure	string	total or unsheltered rate	—	derived
n	integer	CoCs in the fit (King excluded)	CoCs	derived
king_median_rent	number	King County median gross rent that year	US dollars per month, nominal	observed
actual_rate	number	King County rate	per 10,000 residents	derived
predicted_rate	number	Rate predicted for King's rent	per 10,000 residents	estimated
residual	number	actual_rate minus predicted_rate	per 10,000 residents	estimated
pi95_low	number	Lower bound of the 95% prediction interval for the residual	per 10,000 residents	estimated
pi95_high	number	Upper bound of the 95% prediction interval	per 10,000 residents	estimated
slope_per_100	number	Fitted change in rate per $100 of median rent	per 10,000 residents per $100	estimated
slope_se	number	Standard error of the slope	per 10,000 residents per $100	estimated
r2	number	R-squared of the fit	—	estimated

Supporting files

Source snapshots, manifests, quality reports, documentation, and other materials used to produce or review this release.

• build_and_analyze.js Transformation manifest · text/javascript · 11.8 KB (12,057 bytes)

  Node.js script that builds coc-panel from the HUD PIT sheets and ACS JSON and fits the per-year regressions.

• xlsb_to_csv.ps1 Transformation manifest · text/x-powershell · 977 bytes

  PowerShell script that exports each sheet of the HUD .xlsb workbook to CSV through Excel with macros disabled.

• build_rate_rent.js Transformation manifest · text/javascript · 8.2 KB (8,372 bytes)

  Node.js script that computes king-residuals and the saved analyses (leave-one-out fits, prediction intervals, rent-growth correlations).

• peers.js Transformation manifest · text/javascript · 10.2 KB (10,432 bytes)

  Node.js script that builds peer-changes (z-scored distances, 2019-to-latest changes, rent growth).

• findings.md Documentation · text/markdown · 14 KB (14,292 bytes)

  Method notes, results tables and caveats for the panel and regressions.

• findings-peers.md Documentation · text/markdown · 16.5 KB (16,944 bytes)

  Method notes, results and caveats for the peer comparison, including peers' count-method changes.

Sources

1. HUD, 2007-2025 Point-in-Time Counts by CoC (AHAR Part 1 data) Homeless counts (total, sheltered, unsheltered) and count type for each CoC and year. Downloaded from the Internet Archive snapshot because huduser.gov blocks scripted requests; contentDigest is the SHA-256 of that downloaded workbook (8,859,426 bytes).

   Retrieved October 10, 2026. U.S. government work (public domain). Public domain; not rehosted here. Retrieve from HUD or the archived copy.

   Archived copy captured May 30, 2026

2. U.S. Census Bureau, American Community Survey 1-year estimates, tables B01003, B25064, B25004, B25003 (2015-2019, 2021-2024) Population, median gross rent and rental vacancy for the counties and places that make up each CoC.

   Retrieved October 10, 2026. U.S. government work (public domain). Public domain; values used are in coc-panel.

   Archived copy captured May 21, 2026

3. KCRHA, 2024 Point-in-Time Count FAQ v2 Documents King County's respondent-driven sampling (RDS) method from 2022 and the 2024 revision to 16,868.

   Retrieved October 10, 2026. Public agency document. Linked, not rehosted.

   Archived copy captured October 10, 2026

4. Almquist ZW et al., "Innovating a community-driven enumeration and needs assessment of people experiencing homelessness," American Journal of Epidemiology 194(6), 2024 Only quantitative comparison of King County's 2022 RDS estimate with a forecast of the old visual-count method; basis for the method-adjustment sensitivity range.

   Retrieved October 10, 2026. Open-access journal article. Linked, not rehosted.

   Archived copy captured October 10, 2026

Limitations and reproducibility

Reproducible from public sources: Download the HUD workbook and ACS tables listed in sources, export the workbook sheets with xlsb-export-script, then run build-panel-script, peers-script and residuals-script (Node.js). Outputs match the tables here.

• PIT counts are single January nights; methods vary by CoC and change over time.
• King County changed its unsheltered count method in 2022 (respondent-driven sampling). The estimated effect ranges from 0 to about 3,600 people in 2024.
• Rates per 10,000 are FactHub calculations. HUD does not publish CoC-level rates.
• CoC boundaries are rebuilt from whole Census geographies with small gaps (see coc-geography).
• Multi-county median rent is a renter-weighted average of county medians, not a true median.
• Rent alone explains 2-14% of the variation in total rates across all 67 CoCs (24-45% among Major City CoCs).
• All results are associations, not causal estimates.