GAO, Homelessness: Better HUD Oversight of Data Collection (GAO-20-433, 2020)

Summary

Homelessness: Better HUD Oversight of Data Collection Could Improve Estimates of Homeless Population (GAO-20-433, July 2020) is a U.S. Government Accountability Office report to the Chairwoman of the House Financial Services Committee. Its main purpose is oversight — to assess how the Department of Housing and Urban Development collects and quality-controls the annual point-in-time (PIT) counts of people experiencing homelessness. But in the course of that review GAO built its own econometric model of what community-level factors move homelessness, and it is that model — specifically its $100-rent / ~9-percent finding — that has become load-bearing on this wiki's Homelessness is a housing-cost problem page. It supplies the within-market, over-time half of the case, complementing the cross-market evidence of Colburn & Aldern.

Key Findings

• The headline panel estimate. Tracking the same Continuum-of-Care (CoC) communities over 2012–2018 in a weighted linear fixed-effects model, GAO found that "a $100 increase in median rental price was associated with about a 9 percent increase in the estimated homelessness rate. For instance, in a CoC with a homelessness rate of 16 individuals per 10,000, a $100 increase in household median rent would have an associated increase to about 17.4 individuals per 10,000."[1]
• Rent was the robust variable. The model included housing, demographic and economic characteristics; the relationship between median rent and homelessness "remained positive and significant" across sensitivity analyses — GAO removed localities with unusual year-to-year count changes and controlled for weather and for the counting methodology used.[1] Housing characteristics in the model were median rent (in 2018 dollars), the share of rental units vacant, and the share of units that were renter-occupied — the same rent-and-vacancy family of predictors Colburn & Aldern and Quigley et al. identify.[1]
• The design is a within-community fixed-effects panel. The estimating equation models the natural log of the homelessness rate per 10,000 for each CoC and year on lagged housing, demographic and economic factors with year and community fixed effects, so the identifying variation is change within a community over time, not differences between communities.[1]
• GAO states its own limits. The report interprets the results "with some degree of caution": "our regression models may be subject to omitted variable bias—it is unlikely that we have been able to quantify and include all factors relevant to homelessness."[1] And the whole report is, at bottom, a critique of the PIT data: it documents that HUD's one-night counts vary in method across communities and are prone to undercount — the very data-quality problems the housing-cost literature inherits.

Relation to the Georgist Case

GAO is not a Georgist source and does not discuss land or land value taxation. Its value to this wiki is evidentiary: it is the cleanest panel demonstration that when rents rise within a community, its measured homelessness rises — the temporal complement to the cross-sectional geography Colburn & Aldern document. Together they support the two-level reading on the outcome page (housing-market conditions set how many people a market pushes into homelessness). The step from "housing costs" to "land costs" is not GAO's; it runs through work like Knoll, Schularick & Steger showing land prices, not construction costs, drive long-run housing costs.

Nuances and Limits

• Association, not experiment. The fixed-effects panel controls for stable community differences but is still observational; GAO flags omitted-variable bias explicitly.
• The count data are the object of the critique. Because the report's main charge is that PIT counts are inconsistent and undercount, the $100-rent estimate is built on the same imperfect measure it is warning about — a caveat the outcome page states in its own voice.
• A within-market, not a policy, result. The estimate says rising rents track rising homelessness; it does not by itself establish that a given rent-lowering policy (LVT, upzoning, subsidy) would cut homelessness one-for-one — that is a separate, more contested question.

Bears On

• Outcome: Homelessness is a housing-cost problem — GAO supplies the within-market panel half of the case.
• Research: Homelessness Is a Housing Problem (Colburn & Aldern) — the cross-market complement.
• Research: Inflection points in community-level homeless rates (Glynn, Byrne & Culhane) — a nonlinear reading of the same rent–homelessness relationship.

See Also

• Outcome: Homelessness is a housing-cost problem
• Research: Homelessness Is a Housing Problem (Colburn & Aldern)
• Research: Inflection points in community-level homeless rates (Glynn, Byrne & Culhane)
• Research: Do Local Economic Conditions Affect Homelessness? (Hanratty)
• Narrative: The Housing Crisis Is a Land Crisis

Sources

1. U.S. Government Accountability Office, Homelessness: Better HUD Oversight of Data Collection Could Improve Estimates of Homeless Population, GAO-20-433, July 2020 — used for every figure and quotation on this page. The $100-rent/~9-percent estimate and the 16→17.4-per-10,000 illustration (p. 29), the sensitivity-analysis and robustness language (pp. 29–30), the omitted-variable-bias caveat, and the fixed-effects model specification (app. II, pp. 51ff.) were verified verbatim against the full report PDF this session (fetched via an Internet Archive mirror of gao.gov, which returned HTTP 403 to this wiki's direct egress). gao.gov/products/gao-20-433 · Full PDF (gao.gov)