# King County's pre-2020 peers: how did their homelessness change from 2019 to 2024?

Part of the FactHub Seattle/King County homelessness fact check (see `../findings.md`, `../../research/count-method.md`, `../../research/comparisons.md`). Built 2026-10-10. Saved from the research agent's final report, which the agent itself was blocked from writing.

**Question:** which large CoCs looked most like King County (WA-500) before 2020, and did their homelessness change differently from King's through 2024?

**Short answer: most of King's closest peers also got worse.**
- Of the 10 nearest peers, 9 had a higher total rate per 10k at their latest full count than in 2019, and 8 had a higher unsheltered rate.
- As counted, King's rise (+45% total, +81% unsheltered) was larger than the peer median (+26% / +29%).
- With the central method adjustment (+36% / +60%), King still rose more than 7 of the 10. With the high-end adjustment (+14% / +14%) it would sit roughly mid-pack.
- No close peer clearly improved. The ones that look best from a 2019 base improved on one measure only, mostly from 2019 spikes, and most of them changed their count method.
- King's rent grew more slowly than almost every other CoC's, so "it was just rents" does not explain why King rose more than its peers.

## Files
- `peers.csv`: all 67 CoCs with baseline features, distance to King (two versions), a simple-screen flag, 2019 and end values, changes, and rent growth.
- `king_adjusted.json`: King's change as counted and with each method adjustment, plus the per-year sample median rents.
- `scripts/peers.js`: builds both files. Run `node peers/scripts/peers.js .` from `data/`.

## Method
1. **Baseline features, 2015-2019.**
   - **Rates:** total rate and unsheltered rate per 10k, averaged over each CoC's full-count years (3-5 years). Sheltered-only years carry the unsheltered count forward, so they are dropped.
   - **Relative rent:** each year's ACS median gross rent divided by the median across the 67 CoCs that year, then averaged. This removes inflation and the national rent trend.
   - **Rent growth:** 2019 rent / 2015 rent − 1.
   - **Population (optional):** log of 2019 population.
2. **Distance:** each feature is z-scored across the 67 CoCs. Distance is the Euclidean distance to King, with all features weighted equally.
3. **Simple screen:** 2019 median rent within ±15% of King's $1,736 (that is, $1,476-$1,996), and 2019 total rate within ±40% of King's 49.7 (29.8-69.6).
4. **Change:** from 2019 (a full count for all 67 CoCs) to 2024. Where 2024 was sheltered-only (Portland, San Jose, Riverside), the endpoint is 2025.
5. **King adjustment** (`count-method.md` §6): subtract the RDS method premium from King's 2024 unsheltered count, and therefore from the total. Central estimate 1,100 people; range 0 to 3,600.

## Peer list
King's baseline: total 51.0, unsheltered 23.1, relative rent 1.38, rent growth 2015-19 +28%.

| Rank | CoC | Distance | Rank incl. pop | Base total | Base unsh | Rel. rent | Rent growth 15-19 |
|---|---|---|---|---|---|---|---|
| 1 | CA-606 Long Beach | 1.01 | 8 | 43.4 | 28.4 | 1.15 | +28% |
| 2 | CA-502 Oakland/Alameda | 1.04 | 1 | 35.5 | 25.2 | 1.56 | +31% |
| 3 | OR-501 Portland/Multnomah | 1.05 | 4 | 49.7 | 23.2 | 1.07 | +27% |
| 4 | CA-601 San Diego | 1.64 | 2 | 26.0 | 14.5 | 1.42 | +23% |
| 5 | CA-503 Sacramento | 1.74 | 3 | 25.8 | 14.9 | 1.08 | +28% |
| 6 | CA-600 Los Angeles | 1.80 | 14 | 52.4 | 38.7 | 1.26 | +23% |
| 7 | GA-500 Atlanta | 1.81 | 16 | 75.5 | 16.6 | 1.01 | +28% |
| 8 | CA-505 Contra Costa | 2.10 | 6 | 17.3 | 11.4 | 1.57 | +26% |
| 9 | CA-500 San Jose/Santa Clara | 2.12 | 5 | 40.9 | 31.1 | 1.95 | +25% |
| 10 | CA-511 Stockton/San Joaquin | 2.21 | 12 | 26.2 | 11.8 | 1.03 | +22% |

- **Effect of adding population:** Long Beach, LA, Atlanta and Stockton drop out, and Las Vegas, Riverside and Pinellas come in. The core of the list (Oakland, San Diego, Sacramento, Portland, San Jose, Contra Costa) stays the same.
- **Simple screen:** only Oakland, LA and Honolulu pass. Honolulu (HI-501) ranks 24th by distance and is added below as a robustness peer.
- **Geographic concentration:** 8 of the 10 peers are on the West Coast, and 7 are in California. This is not a national comparison.

## Results: 2019 to the latest full count (rates per 10k)

| CoC | 2019 total | End total | Change | 2019 unsh | End unsh | Change | Rent growth 19-24 |
|---|---|---|---|---|---|---|---|
| **King, as counted** | 49.7 | 72.1 | **+45%** | 23.2 | 41.9 | **+81%** | +22% |
| King, central adj. (−1,100) | 49.7 | 67.4 | +36% | 23.2 | 37.2 | +60% | |
| King, high adj. (−3,600) | 49.7 | 56.7 | +14% | 23.2 | 26.5 | +14% | |
| Long Beach | 40.9 | 74.9 | +83% | 27.6 | 54.4 | +98% | +34% |
| Oakland/Alameda | 48.0 | 57.3 | +19% | 37.8 | 38.5 | +2% | +17% |
| Portland (2025) | 49.4 | 132.3 | +168% | 25.1 | 86.8 | +247% | +26% |
| San Diego | 24.3 | 32.1 | +32% | 13.4 | 18.5 | +38% | +33% |
| Sacramento | 35.8 | 41.1 | +15% | 25.1 | 24.5 | −3% | +37% |
| Los Angeles | 60.9 | 79.3 | +30% | 46.0 | 55.1 | +20% | +27% |
| Atlanta | 63.5 | 55.1 | −13% | 14.2 | 20.0 | +41% | +40% |
| Contra Costa | 19.9 | 24.3 | +22% | 14.1 | 16.7 | +19% | +20% |
| San Jose (2025) | 50.4 | 55.6 | +10% | 41.1 | 38.8 | −6% | +19% |
| Stockton | 34.5 | 58.0 | +68% | 20.4 | 42.5 | +108% | +42% |
| *Honolulu (screen)* | 45.3 | 45.0 | −1% | 24.7 | 27.7 | +12% | +13% |
| **Top-10 median** | | | **+26%** | | | **+29%** | |
| All 66 other CoCs, median | | | +20% | | | +29% | +34.5% |

**How King compares**
- **Broad increase.** 2019-2024 was a period of broad increase: 76% of the 66 other large CoCs rose on each measure.
- **King rose more than most peers, as counted.** Only Long Beach, Portland and Stockton rose more. In per-10k terms, King's unsheltered increase (+18.7) ranks 4th of 67 and its total increase (+22.4) ranks 7th.
- **The method adjustment matters.** At the central estimate King is still above the peer median. At the high end it is below.

**Did any peer improve?**
- **No peer improved on both measures.**
  - Atlanta's total fell, but its unsheltered count rose 41%.
  - Sacramento and San Jose cut unsheltered rates slightly while their totals rose.
  - Oakland was flat on unsheltered.
  - Honolulu was flat on total.
- **The 2019 base flatters three of the improvers.** Oakland (5,629 people in 2017 to 8,022 in 2019), Sacramento (3,665 to 5,561) and San Jose (7,394 to 9,706) all spiked into 2019.
  - **Against the 2015-19 average:** their totals are up 61%, 59% and 36%, and their unsheltered counts up 53%, 64% and 25%.
  - **King on the same basis:** total +41%, unsheltered +82%. So King's total rise is mid-pack, and its unsheltered rise is still near the top, behind Portland (+274%), Stockton (+260%) and Long Beach (+92%).
  - **King's own 2019 was low:** it was the only decline of the visual-count era.
- **Trajectories after 2019 differ.**
  - Sacramento rose to 9,278 in 2022 and fell to 6,615 in 2024.
  - Oakland went from 9,747 to 9,450 over the same years.
  - San Jose was roughly flat.
  - King rose at every count: 11,199 (2019), 13,368 (2022), 16,868 (2024).

## Rent-growth check ("it was just rents")
- **No relationship in the full sample.** Across the 66 other CoCs, nominal rent growth from 2019 to 2024 is uncorrelated with the % change in total rate (r = −0.04) and in unsheltered rate (r = −0.01). Against the per-10k change in total rate, r = −0.14.
- **Improvers had faster rent growth, not slower.** The 16 CoCs whose total rate fell had median rent growth of +38.5%. CoCs whose rate rose had +33.9%.
- **King's rent grew slowly.** King's rent rose +21.5% ($1,736 to $2,110), 60th of 67. Its rent relative to the sample median fell from 1.41 to 1.20. If rent growth drove the change, King should have outperformed its peers.
- **Within the top 10 the relationship is weak** (r = 0.03 for total, 0.21 for unsheltered).
  - The slow-rent-growth peers, Oakland (+17%) and San Jose (+19%), did relatively well. But King's rent growth was similar and it did not.
  - Sacramento (+37%) and Atlanta (+40%) did relatively well despite fast rent growth.
- **Limits:**
  - ACS median gross rent lags asking rents.
  - Rent *level* is still the strongest cross-sectional correlate in `../findings.md`.
  - This check tests only whether *changes* in rent explain *changes* in counts.

## What the "improvers" changed, and whether their count method changed
Sources were found by web research subagents. Most PDFs could not be opened, so items marked [U] rest on search snippets or press reports.

**CA-500 San Jose/Santa Clara (unsheltered −6%, 2019 to 2025)**
- **System changes:**
  - Aug 2020: the 2020-25 Community Plan aimed to house 20,000, cut inflow 30% and double temporary beds [U]. https://destinationhomesv.org/2020/08/13/santa-clara-county-continuum-of-care-releases-new-roadmap-for-ending-homelessness/
  - 2021: temporary beds rose from 1,882 to 2,336, and inflow fell from 4,778 to 3,403 [U], per the same source.
  - Measure A (2016, $950M bond) has funded 5,135 homes [U]. https://sanjosespotlight.com/santa-clara-county-bond-brings-thousands-of-affordable-homes/
  - 2020: Measure E transfer tax.
  - 2023-26, under Mayor Mahan: interim housing sites went from 7 to 23. Shelter beds per 100 homeless people went from 29.0 to 40.6. https://www.kqed.org/news/12072999/tiny-homes-big-ambitions-matt-mahans-run-for-governor-spotlights-his-shelter-strategy
  - 20,830 people housed since 2020 [U]. https://destinationhomesv.org/documents/2026/02/2025-year-end-progress-report.pdf/
- **Method: a major break in 2025.**
  - The vendor changed from ASR to Simtech.
  - The full-coverage street count was replaced by counts of likely areas plus a random sample, extrapolated to the rest.
  - The county said 2025 is not comparable with earlier counts. https://sanjosespotlight.com/santa-clara-county-switches-how-homeless-people-are-counted/
  - **The 2019-2025 change cannot be credited to policy.**

**CA-503 Sacramento (unsheltered −3%, 2019 to 2024)**
- **System changes:**
  - Aug 2021: city siting plan for 20 shelter, safe-ground and safe-parking sites [U]. https://sacramentocityexpress.com/2021/08/20/city-outlines-initial-steps-for-implementing-its-comprehensive-siting-plan-to-address-homelessness/
  - Nov-Dec 2022: Measure O, a camping ban that applies once shelter has been offered [U]. https://www.cityofsacramento.gov/city-manager/projects-and-programs/measureo
  - Dec 6, 2022: a 5-year city-county partnership agreement.
    - Joint encampment teams target 20 encampments a month.
    - Shelter beds go into one shared coordinated access system, which handled 918 referrals in its first 6 months [U].
    - https://capradio.org/articles/2022/12/01/city-and-county-of-sacramento-reach-landmark-deal-on-addressing-homelessness
  - Dec 2022: Coordinated Access System launched with 2-1-1 [U]. https://sacramentocityexpress.com/2024/06/03/thousands-helped-into-shelter-and-housing-through-sacramentos-coordinated-access-entry-system/
- **Method: partial changes.**
  - 2019 to 2022: the same Sac State team, but with expanded mapping. The researchers credited part of the 2022 rise to "better counting techniques." https://csus.edu/news/newsroom/stories/2022/7/homeless-count-2022.html
  - 2024: the vendor changed from Sac State to Simtech. Sacramento Steps Forward says the method was the same. https://calmatters.org/housing/homelessness/2024/09/pit-count-analysis-2024/
  - The 2022 peak and the 2024 drop (−29% total, −41% unsheltered) are both partly method artifacts. The 2019-to-2024 comparison is the least affected, but it is not clean.

**CA-502 Oakland/Alameda (unsheltered +2%)**
- **System changes:**
  - Jan 2020: the county Office of Homeless Care and Coordination was created [U].
  - Sep 2020 onward: Homekey hotel conversions (county $14.5M for a 104-unit Comfort Inn; Oakland $20M plus about $17.5M) [U]. https://covid-19.acgov.org/covid19-assets/docs/press/press-release-2020.09.29.pdf
  - Community cabins and a Jul 2021 RV safe-parking site [U].
  - 2022: the Home Together 2026 plan added shallow subsidies and improved coordinated entry, and moved more than 1,700 people into housing in its first year [U]. https://health.alamedacountyca.gov/home-together-2026/
  - Sep 2023: county homelessness emergency declared [U].
- **Method: changed.**
  - 2019: ASR counted every census tract on foot.
  - 2022: known locations plus a random sample, counted mostly from cars, with new multipliers [U]. https://oaklandnorth.net/2022/03/29/oakland-alameda-county-homeless-count/
  - 2024: geographic sampling, with a probable vendor change to Simtech [U].
  - **Treat the comparison with caution.**

**GA-500 Atlanta (total −13%, unsheltered +41%)**
- **System changes:**
  - 2017 onward: HomeFirst ($25M city plus $25M private) shifted funding from shelter toward permanent housing [U].
  - 2020-22: Emergency Housing Vouchers and ARPA money expanded rapid rehousing. The count fell 38% to 2,017 in 2022. https://www.wabe.org/city-of-atlanta-reports-nearly-40-drop-in-homeless/
  - Aug 2022: LIFT 2.0, with a goal of housing 1,500 people.
    - It paid landlords holding fees to reserve units in bulk.
    - About 180 coordinated-entry assessors.
    - Encampment closures in 2023.
    - https://endhomelessness.org/resources/case-studies/housing-focused-responses-to-unsheltered-homelessness/atlanta/
  - Apr 2024: The Melody, 40 container units [U].
- **Method:** no change found.
- **Why the total fell:** the sheltered count fell from 2,498 to 1,827, likely because shelter and transitional-housing capacity shrank as funding moved to permanent housing. Meanwhile the unsheltered count rose from 719 to 1,040. **This is not an across-the-board improvement.**

**HI-501 Honolulu (total −1%, unsheltered +12%)**
- **System changes:**
  - Oct 2021: Honolulu CORE launched, and encampment sweeps continued [U].
  - 2022: Act 252 created a statewide homelessness office [U].
  - Jan 2023 onward: Governor Green's emergency proclamations for kauhale (tiny-home villages) [U].
  - 2024: 3 kauhale opened on Oahu [U]. https://governor.hawaii.gov/wp-content/uploads/2024/05/240515-Kauhale-Slides.pdf
- **Method: changed.**
  - 2020: the count moved from several days to a single day [U].
  - 2022: the count was held in March, and COVID had cut shelter capacity.
  - 2024: more volunteers took part. The governor called comparison "very difficult." https://www.hawaiipublicradio.org/local-news/2024-05-16/new-report-shows-increase-homelessness-oahu

**Pattern across the improvers**
- Four of the five changed their count method or vendor, and three of those changed vendor (to Simtech).
- The common policy themes were:
  - shelter or interim-housing expansion (San Jose, Sacramento);
  - joint city-county encampment teams with coordinated access (Sacramento);
  - landlord holding fees and rapid rehousing (Atlanta);
  - Homekey hotel acquisitions (Oakland).
- Seattle/King did many of the same things, so the comparison cannot isolate any one of them.

## Caveats
- **PIT noise.** These are single-night counts. Swings of 20-50% can come from weather, volunteer turnout, sweeps before the count, or shelter openings. Small CoCs swing on a few hundred people.
- **Method changes:**
  - King switched to RDS (central premium about 1,100 people; range 0-3,600).
  - Portland added by-name-list data in 2025.
  - San Jose (2025), Sacramento and Oakland changed vendor or sampling.
  - Honolulu went to a single-day count.
- **Endpoint mismatch.** Portland, San Jose and Riverside end in 2025. The 2025 rows use 2024 ACS population.
- **Baseline choice.** Results change depending on whether the baseline is 2019 or the 2015-19 average.
- **Small n.** Ten peers, mostly in California. The peer set depends on which features are used and how they are weighted.
- **Geography.** City-only CoCs (Atlanta, Long Beach) have small denominators.
- **No causal identification.** Policies, shocks (COVID, fentanyl, the end of eviction relief, migrant arrivals) and count methods all changed at the same time.

## What this does and doesn't show
**It does show:**
- King's closest pre-2020 peers mostly saw rising rates from 2019 to 2024, so a rise in King is not unusual on its own.
- As counted, King rose more than most of its peers, especially on unsheltered. That remains true at the central method adjustment but not at the high end.
- Rent growth since 2019 does not explain the differences, and King's rent grew unusually slowly.

**It does not show:**
- That any peer's policy caused a better outcome. The improvements were modest, came on one measure each, and mostly coincided with method changes or 2019 peaks.
- How much of King's excess rise is real rather than a method effect. The 0-3,600 range covers most of the gap.
- Anything about 2026 (King: 18,365) or about pre-2015 trends.