# Shelter and housing bed capacity (HUD Housing Inventory Count) across large US Continuums of Care
Research dataset for a FactHub fact check on Seattle/King County (WA-500) homelessness. Built 2026-10-10. It uses the same 67 CoCs as `../coc_homelessness_rent_panel.csv` (see `../findings.md`). Saved from the research agent's report, which the agent itself was blocked from writing.

## Files
- **`hic_panel.csv`:** 67 CoCs x 2012-2025 (938 rows). HIC beds by type, joined to HUD PIT counts and ACS population.
- **`hic_results.json`:** every regression (slope, SE, intercept, R², RMSE) and the permanent-bed table.
- **`scripts/build_hic.js`:** builds the panel. Run `node scripts/build_hic.js . ..` from this folder.
- **`scripts/analyze_hic.js`:** prints the tables below. Run `node scripts/analyze_hic.js .`
- **`scripts/csvlib.js`:** CSV parsing. **`scripts/headers.js`** and **`scripts/show_coc.js`** are inspection helpers.
- **`raw/`:** unmodified downloads. `raw/hic2025_csv/` and `raw/hic2024ed_csv/` hold each sheet exported to CSV with `../scripts/xlsb_to_csv.ps1` (Excel, macros force-disabled).

### Columns
- **`es_beds` … `oph_beds`:** HUD's "Total Year-Round Beds" for Emergency Shelter, Transitional Housing, Safe Haven, Rapid Re-Housing, Permanent Supportive Housing and Other Permanent Housing.
  - 2014-2016 RRH is HUD's "RRH & DEM" column, which adds the RRH demonstration projects. For WA-500, DEM is 0.
  - The 2012 sheet has no RRH or OPH column, and the 2013 sheet has no OPH column. Those cells are blank, and `perm_beds` for those years counts only what was reported.
- **`es_seasonal_beds`, `es_overflow_beds`:** HUD's seasonal and overflow ES beds. They are not included in `es_beds` or `temp_beds`.
- **`temp_beds`** = ES + TH + SH. **`perm_beds`** = RRH + PSH + OPH.
- **`pit_count_type`, `pit_total`, `pit_sheltered`, `pit_unsheltered`:** Count Types, Overall, Sheltered Total and Unsheltered Homeless from HUD's 2007-2025 PIT workbook (the same file as `../findings.md`).
- **`population`:** taken from the existing panel (ACS 1-year B01003). It is blank for 2012-2014 and 2020. 2025 uses 2024 ACS.
- **`temp_beds_per_homeless`** = temp_beds / pit_total.
- **`unsheltered_share`** = pit_unsheltered / pit_total. It is blank in sheltered-only count years, because the unsheltered number there is carried forward (or 0 in 2021).
- **`temp_beds_per_10k`, `perm_beds_per_10k`:** beds per 10,000 residents.
- **Also included:** `short_name` and `hud_category_2025` (copied from the panel).

## Method
1. **Download.** HUD's "2007-2025 HIC Counts by CoC" (.xlsx), one sheet per year. The HIC is taken on the same January night as the PIT, and the CoC codes match the PIT workbook.
   - The 2007-2024 edition was also downloaded. WA-500's 2019, 2023 and 2024 values are identical in both editions.
   - The 2025 edition's "Revisions" sheet is empty.
2. **Columns.** Columns are found by header name, not position, because the layout changes in 2013, 2014 and 2017.
   - MO-604a is mapped to MO-604, as in the existing panel.
   - All 67 CoCs are present in every year from 2012 to 2025.
3. **Fits.** Cross-CoC fits are ordinary least squares within one year.
   - They use only CoCs with a full (sheltered and unsheltered) PIT count that year: n = 67 in 2019 and 60 in 2024. Portland, San Jose and Austin did sheltered-only counts in 2024.
   - Each fit is run on three samples: all CoCs, all except NYC, and HUD 2025 "Major City CoC" only.

## Sources
- **HUD, 2007-2025 HIC Counts by CoC.**
  - Live URL: https://www.huduser.gov/portal/sites/default/files/xls/2007-2025-HIC-Counts-by-CoC.xlsx. It returns an HTTP 202 bot challenge to scripts.
  - Copy used: https://web.archive.org/web/20260530191310/https://www.huduser.gov/portal/sites/default/files/xls/2007-2025-HIC-Counts-by-CoC.xlsx, downloaded via the `id_` raw-file form. SHA-256 d4dd3a9dc84384976fa2efcde2d89081a4deeea57be43c73eb587c77ed527bc1.
- **HUD, 2007-2024 HIC Counts by CoC** (used only for the cross-check): https://web.archive.org/web/20250211110540/https://www.huduser.gov/portal/sites/default/files/xls/2007-2024-HIC-Counts-by-CoC.xlsx. SHA-256 4f881539918d89cfe83e7eedf736522fc8ff847fc54c32cea907d1b8560c3a24.
- **HUD, 2007-2025 PIT Counts by CoC:** https://web.archive.org/web/20260530191417/https://www.huduser.gov/portal/sites/default/files/xls/2007-2025-PIT-Counts-by-CoC.xlsb
- **Population:** ACS 1-year B01003, via `../coc_homelessness_rent_panel.csv`.

## Results
### King County (WA-500): year-round beds by type vs. PIT

| Year | ES | TH | SH | RRH | PSH | OPH | ES seasonal | ES overflow | Temp (ES+TH+SH) | Perm (RRH+PSH+OPH) | PIT total | Sheltered | Unsheltered | Count | Temp beds / homeless | Sheltered / temp beds |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2012 | 2,590 | 4,102 | 45 | – | 4,225 | – | 326 | 184 | 6,737 | 4,225 | 8,899 | 6,281 | 2,618 | full | 0.76 | 0.93 |
| 2013 | 2,645 | 4,111 | 45 | 47 | 4,365 | – | 431 | 133 | 6,801 | 4,412 | 9,106 | 6,370 | 2,736 | full | 0.75 | 0.94 |
| 2014 | 2,754 | 3,865 | 45 | 78 | 4,687 | 0 | 315 | 207 | 6,664 | 4,765 | 8,949 | 6,213 | 2,736* | sheltered-only | 0.75 | 0.93 |
| 2015 | 2,828 | 3,579 | 45 | 1,326 | 4,675 | 1,420 | 370 | 407 | 6,452 | 7,421 | 10,122 | 6,319 | 3,803 | full | 0.64 | 0.98 |
| 2016 | 3,083 | 3,358 | 45 | 2,291 | 4,848 | 1,411 | 420 | 188 | 6,486 | 8,550 | 10,730 | 6,225 | 4,505 | full | 0.60 | 0.96 |
| 2017 | 3,498 | 2,902 | 45 | 781 | 5,469 | 2,086 | 390 | 126 | 6,445 | 8,336 | 11,643 | 6,158 | 5,485 | full | 0.55 | 0.96 |
| 2018 | 3,842 | 2,578 | 50 | 1,068 | 5,767 | 1,794 | 389 | 14 | 6,470 | 8,629 | 12,112 | 5,792 | 6,320 | full | 0.53 | 0.90 |
| 2019 | 4,540 | 2,129 | 45 | 1,102 | 6,066 | 1,652 | 265 | 6 | 6,714 | 8,820 | 11,199 | 5,971 | 5,228 | full | 0.60 | 0.89 |
| 2020 | 4,857 | 2,131 | 84 | 964 | 6,249 | 1,731 | 171 | 32 | 7,072 | 8,944 | 11,751 | 6,173 | 5,578 | full | 0.60 | 0.87 |
| 2021 | 4,070 | 1,884 | 84 | 1,207 | 6,266 | 1,594 | 30 | 242 | 6,038 | 9,067 | 5,183 | 5,183 | 0 | sheltered-only | n/a | 0.86 |
| 2022 | 5,376 | 1,966 | 76 | 1,463 | 6,785 | 2,848 | 0 | 46 | 7,418 | 11,096 | 13,368 | 5,683 | 7,685 | full | 0.56 | 0.77 |
| 2023 | 5,288 | 1,890 | 76 | 1,247 | 7,251 | 4,057 | 42 | 3 | 7,254 | 12,555 | 14,149 | 6,464 | 7,685* | sheltered-only | 0.51 | 0.89 |
| 2024 | 5,173 | 1,742 | 76 | 1,990 | 7,768 | 1,505 | 786 | 188 | 6,991 | 11,263 | 16,868 | 7,058 | 9,810 | full | 0.41 | 1.01 |
| 2025 | 5,958 | 1,712 | 76 | 2,388 | 8,034 | 1,757 | 44 | 83 | 7,746 | 12,179 | 16,936 | 7,126 | 9,810* | sheltered-only | 0.46 | 0.92 |

\* Unsheltered is carried forward from the previous full count. In 2021 the PIT total is sheltered only, so the ratio is not comparable.

**What the King County series shows (2015 → 2024):**
- **Emergency shelter grew.** ES went from 2,828 to 5,173 (+83%). From 2012 to 2025 it went from 2,590 to 5,958.
- **Transitional housing fell.** TH went from 3,579 to 1,742 (−51%); from 4,102 in 2012 to 1,712 in 2025.
- **Total temporary beds were roughly flat.** ES + TH + SH stayed between 6,038 and 7,746 from 2012 to 2025. Per 10,000 residents it was 30.5 in 2015 and 29.9 in 2024.
- **Permanent beds grew.**
  - PSH rose from 4,675 to 7,768 (+66%), and RRH + PSH + OPH from 7,421 to 11,263 (+52%).
  - Permanent beds' share of all year-round capacity (temporary + permanent) rose from 53% to 62%.
  - OPH spiked in 2022-2023 (2,848 and 4,057) and fell back to 1,505 in 2024. The timing matches when Emergency Housing Vouchers were being leased (see `../ehv/findings-ehv.md`), but that link is an untested hypothesis.
- **The homeless count grew faster than temporary beds.** The PIT total went from 10,122 to 16,868. Temporary beds per homeless person fell from 0.64 to 0.41, and the unsheltered share rose from 37.6% to 58.2%. Counts from 2022 on use respondent-driven sampling (RDS); see `../findings.md`.
- **Occupancy.** PIT sheltered ÷ temporary beds (a rough occupancy proxy) ran 0.77-1.01.

**On the claim that "capacity shifted from emergency shelter to PSH":**
- In HUD's HIC for WA-500, ES beds rose in most years. They dipped in 2021 (pandemic de-intensification) and in 2023-2024, and reached a series high in 2025.
- The temporary-bed type that shrank was transitional housing.
- Permanent beds (mainly PSH) grew much faster than temporary beds. So the data show a relative shift toward permanent housing, not a shift away from emergency shelter.
- Whether PSH growth displaced additional shelter is a counterfactual this data can't test.

**Reporting notes:**
- **Tiny-home villages.** The HIC-by-CoC file has no project detail inside ES, so it can't confirm that tiny-home villages entered ES around 2022. ES did rise 1,306 beds from 2021 to 2022, but the 2021 base was pandemic-depressed.
- **Safe Haven in 2025.** The 2007-2025 workbook still reports SH separately in 2025: WA-500 has 76 beds, and 83 CoCs nationally have 2,309 (2,252 in 2024). A merger of SH into ES is not reflected in this file. Either way, SH is negligible for King.

### Cross-CoC: unsheltered share vs. temporary beds (full-count CoCs, OLS)

| Year | Predictor | All: n, slope (SE), R² | Excluding NYC: R² | Major City only: n, R² |
|---|---|---|---|---|
| 2019 | Temp beds per homeless person | 67, −8.9 pts per 0.1 (0.5), 0.84 | 0.84 | 43, 0.88 |
| 2019 | Temp beds per 10k residents | 67, −4.6 pts per 10 (1.6), 0.12 | 0.10 | 43, 0.17 |
| 2024 | Temp beds per homeless person | 60, −8.2 pts per 0.1 (0.7), 0.69 | 0.70 | 38, 0.67 |
| 2024 | Temp beds per 10k residents | 60, −2.5 pts per 10 (1.2), 0.07 | 0.04 | 38, 0.13 |

**2019** (ranks out of 67; "Predicted" and "Residual" come from the beds-per-homeless fit):

| CoC | Temp beds / homeless (rank) | Temp beds / 10k (rank) | Unsheltered share (rank, 1 = highest) | Predicted | Residual (pts) |
|---|---|---|---|---|---|
| King County | 0.60 (44) | 29.8 (10) | 46.7% (24) | 42.8% | +3.9 |
| NYC | 1.00 (5) | 94.0 (1) | 4.6% (62) | 7.5% | −2.9 |
| Boston | 0.98 (7) | 88.4 (2) | 1.9% (66) | 8.8% | −6.9 |
| Houston | 0.70 (38) | 4.5 (62) | 41.0% (27) | 34.2% | +6.8 |
| San Diego | 0.49 (51) | 11.8 (32) | 55.2% (18) | 53.0% | +2.2 |
| Denver | 0.91 (15) | 16.1 (22) | 16.4% (47) | 15.7% | +0.7 |
| Dallas | 0.71 (36) | 8.8 (50) | 32.0% (32) | 32.9% | −0.9 |
| Los Angeles | 0.26 (63) | 15.7 (23) | 75.5% (4) | 73.2% | +2.3 |
| San Francisco | 0.42 (55) | 38.6 (5) | 64.5% (12) | 58.4% | +6.1 |
| Portland | 0.55 (47) | 27.4 (11) | 50.7% (20) | 46.9% | +3.8 |

Medians across the 67: 0.77 temp beds per homeless person, 11.2 per 10k, 30.2% unsheltered.

**2024** (ranks out of 60):

| CoC | Temp beds / homeless (rank) | Temp beds / 10k (rank) | Unsheltered share (rank) | Predicted | Residual (pts) |
|---|---|---|---|---|---|
| King County | 0.41 (53) | 29.9 (10) | 58.2% (13) | 59.4% | −1.2 |
| NYC | 0.80 (25) | 132.5 (1) | 3.1% (59) | 27.8% | −24.7 |
| Boston | 1.18 (1) | 103.0 (2) | 2.3% (60) | −2.8%† | +5.1 |
| Houston | 0.92 (11) | 4.5 (59) | 33.8% (31) | 18.5% | +15.3 |
| San Diego | 0.54 (43) | 17.4 (27) | 57.6% (14) | 49.1% | +8.5 |
| Denver | 0.59 (39) | 25.5 (14) | 20.4% (38) | 44.8% | −24.4 |
| Dallas | 0.90 (15) | 8.6 (52) | 29.2% (33) | 19.8% | +9.4 |
| Los Angeles | 0.37 (55) | 29.6 (11) | 69.5% (5) | 62.7% | +6.8 |
| San Francisco | 0.53 (46) | 52.8 (5) | 52.3% (19) | 50.4% | +1.9 |
| Portland | 0.52 | 48.0 | not in fit (sheltered-only count) | – | – |

† A linear extrapolation below 0%.

Medians across the 60: 0.75 temp beds per homeless person, 15.0 per 10k, 34.2% unsheltered.

**How to read this:**
- **The high R² for beds per homeless person is largely mechanical.** Sheltered people occupy temporary beds, so beds ÷ homeless ≈ sheltered share ÷ occupancy, and unsheltered share = 1 − sheltered share.
- **Per resident, temporary beds explain only 4-17%** of the variation in unsheltered share.
- **King** has high temporary capacity per resident: 10th, about 2x the median. It has low capacity relative to its count: 44th of 67 in 2019 and 53rd of 60 in 2024. It sits near the beds-per-homeless line in both years.
- **NYC and Denver in 2024** are about 25 points below the line. Their sheltered counts exceed or approach their year-round HIC beds, consistent with right-to-shelter and migrant sheltering outside the year-round inventory.
- **Houston and Dallas** sit above the line.

### Permanent beds (RRH + PSH + OPH) per 10k residents, 2015 vs 2024
Ranks are out of 67; rank 1 = most beds or largest increase.

| CoC | 2015 (rank) | 2024 (rank) | Change (rank) | PSH only, 2015 → 2024 | Temp beds/10k, 2015 → 2024 |
|---|---|---|---|---|---|
| King County | 35.1 (14) | 48.1 (14) | +13.1 (28) | 22.1 → 33.2 | 30.5 → 29.9 |
| NYC | 34.2 (15) | 67.1 (9) | +32.9 (8) | 31.3 → 46.9 | 87.8 → 132.5 |
| Boston | 98.7 (2) | 120.6 (3) | +21.9 (13) | 76.1 → 49.8 | 95.1 → 103.0 |
| Houston | 7.6 (63) | 16.2 (54) | +8.6 (33) | 6.5 → 10.0 | 5.3 → 4.5 |
| San Diego | 11.1 (48) | 35.3 (26) | +24.3 (12) | 9.7 → 22.1 | 14.0 → 17.4 |
| Denver | 8.5 (61) | 32.8 (28) | +24.4 (11) | 6.9 → 14.7 | 15.7 → 25.5 |
| Dallas | 9.3 (55) | 13.4 (60) | +4.1 (50) | 9.0 → 6.4 | 10.4 → 8.6 |
| Los Angeles | 20.8 (24) | 41.2 (18) | +20.4 (16) | 17.0 → 29.4 | 15.2 → 29.6 |
| San Francisco | 90.2 (3) | 198.0 (2) | +107.7 (2) | 81.5 → 69.6 | 29.2 → 52.8 |
| Portland | 57.2 (7) | 117.0 (4) | +59.8 (3) | 46.3 → 75.3 | 21.2 → 48.0 |

Medians across the 67 CoCs:
- Permanent beds per 10k: 16.9 in 2015, 27.9 in 2024. The median change was +8.2.
- Temporary beds per 10k: 14.5 in 2015, 15.0 in 2024.

**How the named CoCs compare:**
- King has had about 2x the median permanent capacity per resident in both years.
- King's increase was above the median but smaller than SF, Portland, NYC, Denver, San Diego, Boston or LA.
- King, Houston and Dallas are the named CoCs whose temporary beds per resident fell.

## Caveats
- **RRH "beds" depend on funding.** They are rental-assistance slots in private-market units, not buildings, so they swing with grants (King: 2,291 in 2016, 781 in 2017).
- **Vouchers.** PSH and OPH include scattered-site, tenant-based vouchers (e.g. HUD-VASH, mainstream vouchers, and possibly EHV), not only dedicated buildings.
- **Permanent beds serve people the PIT doesn't count.** People in RRH, PSH or OPH are not counted as homeless, so permanent beds are not shelter capacity for the PIT population.
- **Inventory, not utilization.** The HIC lists beds that exist on count night, not whether they were staffed or full.
  - Sheltered ÷ temporary beds is only a proxy for occupancy.
  - The sheltered count can include hotel and voucher stays that are outside the year-round HIC, as seen in NYC and Denver.
- **Reporting changes.**
  - RRH starts in 2013, OPH in 2014, and 2014-2016 RRH includes demonstration projects.
  - Project-type classification shifts over time (TH vs ES, PSH vs OPH). For example, Boston's PSH fell while its OPH rose.
  - Year-to-year jumps, like King's OPH in 2022-2024, may reflect reporting rather than real capacity.
  - Seasonal and overflow beds are excluded from all totals.
- **PIT method changes.** King switched to RDS in 2022, and sheltered-only years carry the unsheltered count forward. Every ratio built on the PIT inherits this.
- **Population and boundaries** follow the existing panel. ACS population is blank for 2012-2014 and 2020.
- **Associations only.** The cross-sectional fits are within single years, and the beds-per-homeless fit is near-tautological.

## What this does and does not show
**It does show:**
- King's ES beds about doubled from 2012 to 2025 while TH fell by more than half. Total temporary beds stayed around 6,000-7,700.
- King's permanent inventory, mostly PSH, grew 52% from 2015 to 2024 (PSH alone +66%).
- King's homeless count outgrew its temporary beds. It now ranks near the bottom on temporary beds per homeless person and near the top on temporary beds per resident.
- Across CoCs, the unsheltered share tracks beds relative to the count very closely (partly by construction), but tracks beds per resident only weakly.

**It does not show:**
- A shift *from* emergency shelter to PSH. ES grew; the decline was in TH.
- Whether more shelter would have lowered unsheltered homelessness in King, or whether PSH spending crowded out shelter.
- Whether beds were open and used. It also doesn't show which ES projects are tiny-home villages, or why King's OPH spiked in 2022-2023.
- Any causal effect of the bed mix.