Incentives Shape Outcomes: A Comprehensive System-Level Analysis of Housing, Homelessness, Workforce, and Public Policy in America
Di Tran University — The College of Humanization
Publication-ready research article (WordPress format)
Executive Summary
- Homelessness is not a single problem; it is a multi-agency system outcome. In FY 2024, targeted federal homelessness funding spanned HUD, VA, HHS, FEMA, ED, DOJ, DOL, USDA, and more—totaling $9.852B in enacted funding across major programs, each with distinct goals, eligibility rules, and performance measures. (USICH, 2024) [1]
- America’s most-cited homelessness number is a “snapshot,” not a census. The HUD Point-in-Time (PIT) count occurs on one night (typically in the last 10 days of January) and measures sheltered and unsheltered homelessness as defined for PIT methodology. (HUD, 2024) [2]
- In January 2024, HUD estimated 771,480 people experiencing homelessness on a single night, including 497,256 sheltered and 274,224 unsheltered—the highest total since consistent reporting began. (HUD, 2024) [3]
- Definitions drive who “counts,” who qualifies, and what gets funded. HUD’s program definitions differ from the education definition under McKinney‑Vento, which explicitly includes many “doubled-up” situations. This creates legitimate, recurring confusion across sectors—and measurable gaps between need and eligibility. (NCHE, n.d.; eCFR, 24 CFR 576.2) [4]
- Housing cost pressure is a repeatable upstream driver. In 2023, 49.7% of renter households (over 21 million) spent more than 30% of income on housing—meeting the standard definition of “cost burden.” (U.S. Census Bureau, 2024) [5]
- Research converges on “threshold effects”: once rent burdens cross a tipping point, homelessness can rise sharply. A prominent “inflection point” analysis finds the expected homelessness rate increases sharply once median rents exceed ~32% of median income. (Glynn/Byrne/Culhane, 2018) [6]
- The system often measures activity more easily than outcomes. “Beds funded,” “bed nights,” “encampment contacts,” and “case management sessions” are straightforward to count; “housing stability 12–24 months later,” “income growth,” and “returns to homelessness” are harder—yet more aligned with public goals. (HUD, 2024; HMIS Data Standards) [7]
- Humanization requires a broader success definition than shelter alone. Evidence suggests many people experiencing homelessness have some formal employment yet extremely low incomes—indicating that housing policy and workforce policy cannot be cleanly separated if the goal is durable stability. (Meyer et al., 2024) [8]
Pull quote: To understand a system, follow the incentives. To change a system, align them.
Introduction
Homelessness in America is frequently debated as a moral issue, a housing supply issue, a behavioral health issue, a public safety issue, or a governance issue. Empirically, it is all of these—and therefore it is best understood as a system outcome.
HUD’s PIT count—often treated in public discourse as the homelessness number—counts people in shelters, transitional housing, safe havens, and those sleeping in places not meant for human habitation on a single night. (HUD, 2024) [9] That number is indispensable as an indicator. But it is not sufficient as a theory of causation, a performance measure, or a governance model.
This article offers a system-level framework connecting:
- funding streams →
- performance metrics →
- incentives →
- organizational behavior →
- and population outcomes.
The goal is not to assign blame to any political party, administration, city, provider, or profession. The goal is more practical: to explain why sincere efforts can coexist with worsening outcomes—and how incentive-aligned redesign can improve results without demonizing stakeholders.
The Incentive Principle
Why incentives outrank intentions in complex public systems
In economics and organizational theory, the principal–agent problem describes how the people who fund or oversee work (principals) rely on other parties (agents) to deliver outcomes—yet agents often have different constraints, information, and motivations. (Jensen & Meckling, 1976; Holmström, 1979) [10]
This dynamic is not a character flaw. It is structural:
- Funders often cannot directly observe “true outcomes” in real time.
- So they fund what they can measure (outputs, compliance, units delivered).
- Agents rationally orient their operations around what keeps funding stable.
This is how well-meaning systems drift into metric-driven behavior—sometimes called “goal displacement” in public administration and management literature.
Pull quote: When money follows metrics—and metrics lag reality—systems can optimize the measurable while missing the meaningful.
The “system beats person” corollary
Quality improvement literature popularized a core insight: outcomes are often produced by system design, not by individual effort alone. (Deming Institute, 2015) [11] This matters because homelessness response is frequently framed as provider virtue versus provider failure—or individual responsibility versus societal responsibility. A system lens reduces moral heat and increases diagnostic precision: what is the system rewarding, tolerating, or unintentionally producing?
Defining and Measuring Homelessness
Defining homelessness
HUD program definitions (housing policy lens)
Under federal regulation, the ESG “homeless” definition includes (among other categories) people lacking a fixed, regular, adequate nighttime residence, including those sleeping in places not meant for human habitation or staying in shelters, transitional housing, or certain hotel/motel arrangements paid for by government or charities. (eCFR, 24 CFR 576.2) [12]
HUD’s PIT methodology further clarifies that sheltered homelessness includes people in emergency shelters, transitional housing, or safe havens on the night of the count, and it excludes people in rapid rehousing, permanent supportive housing, or other permanent housing programs (even if those programs serve formerly homeless households). (HUD, 2024) [13]
Education definition (human services lens)
The McKinney‑Vento education definition is broader in some respects, explicitly including children and youth sharing housing due to loss of housing or economic hardship (“doubled up”), among other situations. (NCHE, n.d.) [14]
Implication: A person may be considered homeless by an educational institution and not qualify under a HUD program pathway, or vice versa. This friction does not only affect individuals—it shapes data and funding eligibility across agencies.
Measuring homelessness
The PIT count: essential, but structurally limited
HUD describes the PIT as a one-night snapshot, and acknowledges that unsheltered counts are difficult and may undercount the true population due to visibility and local capacity constraints. (HUD, 2024) [9]
HUD also notes that the PIT does not include people temporarily staying with family or friends (“doubled-up” / “couch surfing”), even if unstable. (HUD, 2024) [15]
HMIS: operational measurement of service systems
A Homeless Management Information System (HMIS) is a local information system designated by a Continuum of Care to record and analyze client, service, and housing data for people experiencing homelessness or at risk. (HUD HMIS Data Standards, FY2024) [16]
HMIS enables performance measurement—but performance depends on what is defined, required, and rewarded.
Key metrics table: what the system can measure vs what society needs
| Metric category | Examples commonly tracked | Primary data source(s) | What it tells you | Core limitation |
| Snapshot prevalence | PIT total; sheltered vs unsheltered | PIT / AHAR | Scale at a point in time | Under-count risk; not a flow measure; excludes doubled-up (HUD, 2024) [17] |
| System throughput | Entries; exits; bed nights; projects served | HMIS | Volume of service activity | High activity ≠ reduced homelessness (HMIS Data Standards, FY2024) [16] |
| Outcome durability | 12/24-month housing stability; returns | HMIS + follow-up | Whether housing “sticks” | Requires longer tracking; data linkage challenges |
| Upstream pressure | rent burden; evictions; overcrowding | Census/ACS; local courts | Risk environment | Not controlled by homelessness agencies (Census, 2024; HUD, 2019) [18] |
| Human flourishing | earnings; job stability; health utilization | Admin linkage / studies | Independence and health | Often outside homelessness program scope (Meyer et al., 2024) [8] |
Pull quote: If we only measure what is easy, we will reward what is measurable—not what is meaningful.
Historical Evolution of Housing and Homelessness Policy
A full history is beyond the scope of a single article, but several structural shifts help explain why homelessness has remained difficult to reduce even amid significant spending.
Postwar housing expansion and long-run distributional effects
Post‑World War II homeownership growth was shaped by multiple forces, including federal policy structures and postwar economic conditions; research highlights the role of GI Bill-era human capital investment and broader suburban development dynamics. (Harvard JCHS, 2022) [19]
Deinstitutionalization and the “system transfer” problem
Deinstitutionalization of psychiatric hospitals is often cited in public discourse as a homelessness driver. Serious scholarship emphasizes that the consequential issue is not deinstitutionalization alone, but the downstream system capacity (community care, housing supports, and crisis systems). (Yohanna, 2013) [20]
The modern federal homelessness response architecture
The McKinney‑Vento framework established a durable policy backbone for homelessness response, including today’s education provisions, and homelessness assistance architecture has evolved through regulation and appropriations over time. (NCHE, n.d.; eCFR, 24 CFR 576.2) [21]
System Structure, Funding Flows, and Incentive Design
System structure analysis: an interlocking governance chain
The homelessness/housing ecosystem typically involves:
- Federal agencies and appropriations (statutory eligibility, program rules, major funding vehicles)
- State and local governments (matching, subawards, planning, regulatory environment, landlord-tenant policy)
- Continuums of Care (CoCs) (local planning and coordination structures)
- Nonprofit and public providers (frontline program delivery)
- Data systems (HMIS, PIT, administrative records)
The CoC concept itself is defined as a group of representatives across nonprofit providers, victim service providers, governments, businesses, public housing agencies, school districts, health systems, universities, law enforcement, and people with lived experience—organized to plan and deliver a system of outreach, shelter, rapid rehousing, permanent housing, and prevention strategies. (eCFR, 24 CFR 576.2) [12]
Funding & incentive design: why “what gets funded” matters
Targeted federal homelessness funding is multi-agency
A single year of enacted federal funding includes large investments in:
- HUD Homeless Assistance Grants
- VA housing and service programs (e.g., HUD‑VASH, SSVF)
- HHS health and behavioral health programs
- Education funding for homeless children and youth
- Emergency food/shelter programs, and more
(USICH, 2024) [1]
Table: Targeted federal homelessness funding (FY 2024 enacted, selected programs)
| Agency | Program | FY 2024 enacted ($M) |
| U.S. Department of Housing and Urban Development[22] | Homeless Assistance Grants | 4,051 |
| U.S. Department of Veterans Affairs[23] | HUD‑VASH | 1,046 |
| U.S. Department of Health and Human Services[24] | HRSA Health Care for the Homeless | 627 |
| Federal Emergency Management Agency[25] | Emergency Food and Shelter Program | 117 |
| U.S. Department of Education[26] | Education for Homeless Children and Youth | 129 |
| U.S. Department of Labor[27] | Homeless Veterans’ Reintegration Program | 66 |
| U.S. Department of Justice[28] | Transitional Housing Assistance Grants (sexual assault victims) | 50 |
| Total (all listed in USICH table) | 9,852 |
(USICH, 2024) [1]
Interpretation: When multiple agencies fund overlapping populations with different “success” metrics, the frontline system experiences incentive cross‑pressure: providers must optimize for multiple scorecards.
HUD’s Continuum of Care competitive funding
HUD announced “more than $3.5 billion” in competitive funding through the Continuum of Care program—described as the largest source of federal grant funding for homeless services and housing programs—and noted that nearly 400 CoC communities apply and HUD funds ~7,000 projects annually through the program. (HUD, 2024) [29]
State & local variation: the same nation, different systems
HUD reports a national homelessness rate of 23 people per 10,000 in 2024, with higher rates in some states (e.g., 48 per 10,000 in California and 81 per 10,000 in New York). (HUD, 2024) [9]
But the variation is not only in totals; it is also in sheltered vs unsheltered composition, which reflects climate, shelter capacity, policy choices, and local housing markets. (HUD, 2024) [9]
Table: State variation snapshot (PIT 2024, selected states)
| State | Total experiencing homelessness | % unsheltered | Homelessness per 10,000 people | Notable composition |
| California[30] | 187,084 | 66.3% | 48 | High unsheltered; high chronic count (HUD, 2024) [31] |
| New York[32] | 158,019 | 3.6% | 81 | Predominantly sheltered (HUD, 2024) [33] |
| Florida[34] | 31,362 | 53.8% | 14 | Majority unsheltered (HUD, 2024) [35] |
| Kentucky[36] | 5,231 | 32.8% | 12 | Majority sheltered (HUD, 2024) [37] |
| Hawaii[38] | 11,637 | 34.7% | 81 | High rate; disaster displacement context noted (HUD, 2024) [39] |
The nonprofit role: implementers inside incentive constraints
In practice, nonprofit providers are often the operational backbone for outreach, shelters, rapid rehousing, supportive housing, and services. HUD explicitly notes that CoC‑funded projects are operated by nonprofit providers (alongside states, tribes, and local governments). (HUD, 2024) [29]
This matters because nonprofit behavior is frequently evaluated morally (“good” or “bad”) rather than structurally (“what did the funding rules reward or punish?”). A system audit treats nonprofits as responders to incentive environments, not as the origin of the incentive environment.
Incentive Map: from funding to outcomes
Below is the core causal chain needed for a system audit. The mechanism is conceptually simple—even if implementation is complex.
flowchart LR
A[Funding rules & appropriations] –> B[Metrics & compliance requirements]
B –> C[Incentives for agencies, CoCs, providers]
C –> D[Operational behavior\n(shelter rules, service mix, targeting)]
D –> E[Outcomes\n(unsheltered rates, stability, returns, earnings)]
E –> B
(Feedback loop note: outcomes often reshape future metrics and funding narratives.)
From Activity to Outcomes: Activity vs. Outcome, Perverse Incentives, and Advanced Analysis
Activity vs Outcome: the current trap
A system can become “busy” without becoming “effective” if:
- funding requires high-volume reporting,
- compliance consumes staff capacity,
- and outcomes are measured with long time lags.
HUD’s own data architecture distinguishes between people currently homeless in PIT categories and people served in permanent housing programs who are not counted as homeless in PIT. (HUD, 2024) [13] This distinction is correct for measurement—but can confuse public accountability when the public expects homelessness to fall as permanent housing capacity rises.
Pull quote: A system can increase services and still lose ground if inflows rise faster than outflows.
Perverse incentives: not conspiracies—misalignments
Perverse incentives are predictable when:
- Metrics are proxies (e.g., “contacts made” instead of “people stabilized”).
- Funding is tied to process more than outcome because outcomes are hard to measure.
- Agencies optimize locally (their own program success) rather than globally (system success).
A concrete example: HUD’s PIT and Housing Inventory Count (HIC) data matter in competitive processes. HUD notes that PIT/HIC data play “a critical role” and that the CoC program competition scoring historically incentivized annual counts. (HUD, 2023 Notice) [40]
That incentive is not inherently bad—better data is good. But whenever data influences resources, the system must protect measurement integrity and interpret changes carefully (especially when methodology improves or shelter capacity changes).
Advanced analysis: inflows, outflows, and “pressure”
A practical advanced frame is stock-and-flow:
- Stock: number of people experiencing homelessness at a point in time
- Inflow: entries into homelessness (economic shocks, eviction, domestic violence, discharge from institutions, disaster)
- Outflow: exits to stable housing (RRH → permanent housing, PSH, reunification when safe, etc.)
- Returns: people who re-enter homelessness after exit
A PIT count primarily observes the stock. HMIS can observe flows—depending on coverage and data quality. (HMIS Data Standards, FY2024) [16]
Upstream pressure strongly affects inflow. The U.S. Census Bureau reports that nearly half of renter households are cost-burdened, reflecting widespread affordability strain. (U.S. Census Bureau, 2024) [5]
HUD research also finds housing market factors—high median rents, overcrowding, and evictions—are strong predictors of homelessness rates in high-cost markets, while higher housing density is “protective.” (Nisar et al., 2019) [41]
And “inflection point” work finds homelessness rates rise sharply once rent-to-income crosses a threshold range. (Glynn/Byrne/Culhane, 2018) [6]
System Redesign Principles, Workforce Integration, Implementation, Citizen Accountability, and Humanization
This final section consolidates the remaining outline elements into a single, coherent redesign agenda—because in real governance, these levers must move together.
System redesign principles
A system-level redesign that aims for durable reductions in homelessness typically requires:
Principle: Align metrics with outcomes.
If the public goal is fewer people living unsheltered and more people stably housed, then success measures must emphasize:
- housing stability at 12/24 months
- returns to homelessness
- inflow reduction (prevention)
- and (where appropriate) earned income growth and disability stabilization.
Principle: Risk-adjust outcomes.
Communities face different housing markets, disaster risks, migration dynamics, and shelter infrastructures. HUD’s own reporting emphasizes distinct drivers across states (e.g., eviction backlogs and rent pressures in New York, disaster displacement in Hawaii). (HUD, 2024) [42]
Principle: Reduce administrative load where it does not improve outcomes.
HUD’s move to a two‑year CoC NOFO cycle was explicitly framed as reducing administrative burden so staff can focus more on core duties. (HUD, 2024) [29]
Workforce integration: an essential missing layer
Housing stability and workforce stability are often treated as separate service silos. Yet empirical evidence suggests many people experiencing homelessness have some formal employment—with very low earnings.
A large administrative-data study finds that about 52% of sheltered homeless individuals and 40% of unsheltered homeless individuals had formal employment, but with low median earnings (about $8,300 among workers), implying sporadic/low-wage work. (Meyer et al., 2024) [8]
The same research finds extremely low incomes overall and shows that safety net transfers play a crucial role in material survival. (Meyer et al., 2024) [8]
Humanization implication: A humane system cannot define “success” as shelter utilization alone. If dignity includes stability, agency, and participation, then workforce pathways (or appropriate disability supports) must be treated as core—not peripheral.
Pull quote: Housing without income is fragile. Income without housing is unstable. The system must stop forcing people to choose.
Integrated system model: housing + health + work + accountability
An integrated model does not mean “one agency does everything.” It means:
- A shared outcome definition (stability, not throughput).
- A shared minimum dataset (privacy-protecting, but linkable).
- Coordinated funding that avoids punishing cross‑boundary collaboration.
The practical target is alignment across the chain:
- Funding: supports interventions proven to reduce homelessness duration and increase stability (e.g., supportive housing approaches).
- Metrics: prioritize durable exits rather than raw service volume.
- Incentives: reward prevention and stability, not churn.
Evidence syntheses find Housing First models can reduce homelessness and increase housing stability; economic reviews report that benefits can exceed costs in U.S. studies, while noting limitations and context dependence. (Community Guide economic review, 2022) [43]
Implementation & scaling: what changes first
A feasible scaling sequence often looks like:
Step: Publish the incentive map locally.
Make it explicit: Which funding streams exist? What metrics do they require? What behaviors do they implicitly reward?
Step: Convert outcome goals into contracting language.
If contracts pay primarily for “units delivered,” providers will rationally deliver units. If contracts pay for “stable exits,” systems begin to orient toward stability.
Step: Strengthen measurement integrity.
Because PIT counts can undercount unsheltered people and vary with methodology, systems should interpret changes cautiously and invest in consistent methods. (HUD, 2024; HUD Notice, 2023) [2]
Citizen accountability
A democracy cannot outsource accountability solely to agencies and nonprofits—especially when the system spans multiple agencies and billions in funding. (USICH, 2024) [1]
Citizen accountability increases when questions shift from abstract debate (“who is to blame?”) to system clarity (“what is being funded and rewarded?”).
Citizen Framework: five questions that audit any homelessness or housing system
Use these five questions to evaluate budgets, public dashboards, provider contracts, and political promises:
- What is funded?
- What is measured?
- What is rewarded (explicitly or implicitly)?
- What behavior is increasing as a result?
- What outcome is produced—and for whom?
These questions are nonpartisan by design: they can be applied to any administration, any city, any ideology, and any provider network.
Real-world application: a short example
Consider a city that reports:
- rising “people served”
- rising shelter utilization
- rising outreach contacts
- and rising unsheltered visibility
A system audit might find:
- upstream rent burden increased (raising inflow) (Census, 2024) [5]
- money was tied to shelter operations and throughput metrics
- outcome tracking for 12–24 month stability was weak
- workforce linkages were optional and underfunded
The conclusion would not be “providers failed.” The conclusion would be: incentives were aligned to manage the crisis, not to reduce the inflow or stabilize exits. Redesign then targets the incentive chain.
Synthesis
Homelessness is where multiple American systems collide:
- housing affordability stress
- health and behavioral health capacity gaps
- labor market precarity
- disaster displacement
- institutional discharge pipelines
- and governance fragmentation across agencies
HUD’s own reporting notes that PIT counts are snapshots, that definitions exclude doubled-up situations, and that methodology varies; meanwhile, national rental cost burden is historically high and research links housing costs to homelessness rates. (HUD, 2024; U.S. Census Bureau, 2024; Nisar et al., 2019; Glynn/Byrne/Culhane, 2018) [44]
If the nation wants different outcomes, system incentives—not merely slogans—must change.
Humanization: what the system ultimately owes the person
A humanized homelessness policy does not romanticize suffering or reduce people to metrics. It asserts a grounded ethic:
- People deserve safety now (crisis response).
- People deserve stability soon (durable housing outcomes).
- People deserve dignity always (agency, respect, and pathways to independence or appropriate long-term support).
Evidence that many people experiencing homelessness have some formal work but extremely low incomes suggests a moral and practical truth: the “housing problem” and the “workforce problem” are entangled in the lived experience of poverty and instability. (Meyer et al., 2024) [8]
A system that humanizes does not only ask: “How do we shelter people tonight?”
It also asks: “How do we design a society where fewer people enter homelessness—and where exits lead to a real life?”
Di Tran University Disclaimer (exact text)
This publication is provided by Di Tran University — The College of Humanization for educational and informational purposes only. It does not constitute legal advice, medical advice, financial advice, or official public policy guidance. The analysis is nonpartisan and is based on publicly available sources cited inline; data definitions and methodologies vary by agency and program and may change over time. Any examples are illustrative unless explicitly stated otherwise. Views expressed are those of the author(s) and do not necessarily reflect the views of any government agency or funding entity.
[2] [3] [9] [13] [15] [17] [22] [23] [28] [30] [31] [33] [34] [35] [37] [39] [42] [44] The 2024 Annual Homelessness Assessment Report (AHAR to Congress) Part 1: Point-In-Time Estimates of Homelessness, December 2024
[4] [14] [25] McKinney-Vento Definition – National Center for Homeless Education
[5] [18] Nearly Half of Renter Households Are Cost-Burdened
[6] wp.zillowstatic.com
[7] [16] FY 2024 HMIS Data Standards Manual
[10] [32] [38] Theory of the Firm: Managerial Behavior, Agency Costs …
[11] A Bad System Will Beat a Good Person Every Time – The W. Edwards Deming Institute
[12] eCFR :: 24 CFR 576.2 — Definitions.
[19] The Homeownership Rate and Housing Finance Policy – Part 2
[20] Deinstitutionalization of People with Mental Illness: Causes …
[21] McKinney-Vento – National Center for Homeless Education
[29] HUD Archives: HUD Announces Over $3.5 Billion to Help People Experiencing Homelessness
[40] Final 2024 HIC and PIT Data Collection Notice Final (10.31.23) Final
[41] Market Predictors of Homelessness
[43] Permanent Supportive Housing With Housing First: Findings From a Community Guide Systematic Economic Review