The Two Instruments: A Reader’s Map of Federal Rental Aid

Public housing vs vouchers is the comparison that determines how the United States houses its lowest-income renters, and it is usually argued as ideology when it should be argued as instrument choice. One model places government in the landlord role: local housing authorities own and operate developments under federal rules, and eligible households live in authority-owned buildings. The other model places government in the payer role: the household finds a unit in the private market, and a federal subsidy administered by the same kind of local authority covers most of the rent. A single statutory provision, added in 1998 and effective against a 1999 baseline, has barred net additions to the first model’s stock ever since, which means the comparison is not really between two live options. This article reconstructs both models, states the cap precisely, weighs cost and neighborhood evidence from named studies, and delivers a verdict with its deciding factor named.

The One Test for this article is demanding but concrete. After reading, a person should be able to explain the two ways the federal government houses low-income families, state the statutory provision that has made net new public housing effectively unlawful since 1999, weigh the evidence on cost and neighborhood outcomes, and reach a defended verdict whose deciding factor is named: whether the binding constraint in a given market is supply or access. Everything below is organized to make that test passable. Five axes carry the analysis: ownership and control, the legal cap, the stock, the evidence on cost and neighborhoods, and the countervailing evidence on rent capture. A comparison table midway through compresses those axes into a single glanceable artifact. The article then addresses the complication that the public housing model may never have received a fair retrial, and it closes with a verdict that refuses to be uniform across markets because the evidence refuses to be.

A word on dates, because this article lives at a reference date of March 2016 and some of its best evidence was published later. Wherever a study or data release postdates that reference point, it is flagged with its publication date and treated as what it is: later evidence, admissible for understanding the comparison but never described as the state of the law in 2016. The statutory core, the Faircloth Amendment, the demolition programs, the cost literature through the mid-2000s, and the neighborhood research through 2015 all sit inside the reference window. The randomized mobility experiment and the 2023 stock count sit outside it and are labeled accordingly. This discipline matters because the cap itself is a creature of dates: enacted in October 1998, effective against an October 1999 baseline, still binding at the reference date, and responsible for more of the modern debate than any other single sentence in federal housing law.

The namable claim that organizes the verdict is supply or access. Vouchers solve an access problem and cannot solve a supply problem. When families cannot afford units that exist, a portable subsidy is the efficient instrument. When the units do not exist, or when the existing stock cannot absorb more demand at current rents, a subsidy that chases scarce units bids up prices and part of the public dollar leaks to landlords. The right instrument depends on which constraint binds locally, and policy arguments conducted without naming the binding constraint talk past each other indefinitely. That claim does not favor one side. It is as much a warning to voucher maximalists in tight markets as it is to construction maximalists in loose ones.

The structure that follows mirrors the five axes. First, ownership and control: what it means for government to be the landlord versus the payer, and how the two programs are actually administered. Second, the legal cap: the exact text and mechanics of the Faircloth Amendment, what it prohibits and what it permits, and why every serious proposal since has been framed around vouchers, tax credits, or conversion. Third, the stock: how the public housing inventory fell through demolition programs and conversions, the scale of the capital repair backlog, and the Rental Assistance Demonstration as the mechanism that moves properties out of the public housing program. Fourth, the evidence: the cost comparisons and the neighborhood outcome research, each with named authors and periods. Fifth, the counterevidence: the rent-capture literature that constitutes the strongest supply-side argument. After the table, the article takes up the complication, the possibility that the public housing model’s poor reputation reflects appropriation decisions and a frozen experiment rather than an inherent property of public ownership, and then renders the verdict.

At the 2016 reference date the comparison carried a particular weight. The public housing stock had been shrinking for two decades, the voucher program had become the default instrument of rental assistance, and the research literature had matured enough to support conditional conclusions rather than slogans. The years immediately before the reference date had also produced the first serious reckonings with the demolition era’s net effects, as the replacement arithmetic of HOPE VI came into focus. This article therefore arrives at a point when the evidence permits something better than the old debate: a decision procedure grounded in named statutes, named studies, and a named deciding factor.

A note on territory, since this series assigns each question class to the article that owns it. The voucher program’s internal mechanics, payment standards, inspections, and portability rules belong to the provisions article on Section 8; the segregation analysis belongs to the Fair Housing Act impact article; the construction era’s history belongs to the 1949 act article; and the full statutory sequence belongs to the era guide. This article draws on each of those without re-covering them, and it links to each where the reader’s next question naturally falls. What it owns exclusively is the comparison itself: the two models side by side, the cap that froze one of them, the evidence on cost and neighborhoods, the rent-capture counterevidence, and the verdict with its deciding factor named.

The series thesis thread for this article is that a comparison can be resolved with a verdict and a named deciding factor even in a field where the statutory cap has quietly foreclosed one side of the debate. Resolution does not require pretending the field is level; it requires naming the tilt, discounting for it, and deciding anyway, because decision-makers cannot wait for a controlled experiment that the law forbids running. The verdict below follows that discipline. It states what each instrument does best, what each costs, where each fails, and the single question that determines which to choose, and it leaves the reader equipped to answer that question for any market rather than equipped with a slogan about either model.

Ownership and Control: The Landlord Model and the Payer Model

Public housing begins with a statute and a local government creature. The United States Housing Act of 1937, Public Law 75-412, 50 Statutes at Large 888, approved September 1, 1937, authorized federal assistance for locally owned low-rent housing, and the vehicle it created was the public housing agency: a state or local government entity that owns the buildings, collects rent from tenants, hires the maintenance staff, and answers to federal rules on eligibility, rent setting, and admissions. The federal government writes the checks through operating and capital subsidies; the authority runs the property. That division of labor has been the model’s defining feature for its entire life. The tenant’s relationship is with a government landlord, the unit is a public asset, and the rent the household pays has been set by federal formula for most of the program’s history, generally at a share of adjusted income that Congress has periodically reset.

The voucher model begins with a different section of the same act and a different relationship. Section 8 was added to the 1937 Act by the Housing and Community Development Act of 1974, Public Law 93-383, and after the 1983 amendments and the 1998 consolidation it took the form most readers know: tenant-based assistance codified at 42 United States Code 1437f(o), in which an eligible household receives a voucher administered by a local housing authority, finds a willing private landlord, and pays roughly 30 percent of its adjusted income toward rent while the subsidy covers the remainder up to a locally set payment standard. The mechanics of the modern voucher, including the payment standard, the housing quality inspection, and the portability rules, belong to the article that owns this program’s provisions in full, and readers who want the machinery should consult the provisions guide on how Section 8 vouchers work in law. What matters for the comparison is the institutional reversal: the same kind of local authority that owns buildings in one program writes checks to private landlords in the other, and the household’s counterparty shifts from a public landlord to a private one with a public payer behind the lease.

The distinction runs deeper than who signs the paycheck for the plumber. In the landlord model, the government bears the capital risk of the building: roofs, boilers, elevators, and the long-run depreciation of the structure are public obligations, and the operating subsidy is meant to cover the gap between what low-income tenants can pay and what the building costs to run. In the payer model, the government bears the market risk of the rent: the subsidy must track local rents or the voucher loses purchasing power, and the capital risk of the unit stays with the private owner, who decides whether the building is worth maintaining at the rent the program will support. This allocation of risk explains much of what follows. The capital backlog that haunts public housing is, in a precise sense, the landlord model’s risk materializing on the public balance sheet. The rent-capture problem that haunts vouchers is the payer model’s risk materializing in the private market. Neither is an accident; each is the predictable cost of the risk the model chose to carry.

Administration also differs in ways that matter for the evidence. Public housing authorities manage waiting lists for their developments, set admissions preferences within federal parameters, and make the day-to-day decisions of a landlord: maintenance scheduling, lease enforcement, capital planning. Voucher administrators run a different operation: they determine eligibility, calculate subsidy amounts, inspect units for quality standards, brief households on how to search, and mediate the three-way relationship among tenant, landlord, and program. The skill sets overlap less than the shared name “housing authority” suggests. An authority that excels at property management may be mediocre at landlord recruitment, and the mobility experiment discussed below turns on exactly that administrative capacity: the difference between handing a family a voucher and helping a family use one.

The rent formula itself has a legislative history worth knowing, because it shapes what each model costs. The Brooke Amendment of 1969 capped public housing rents at 25 percent of tenant income, converting the program from flat project rents to income-based rents and thereby converting operating shortfalls into a federal obligation. The provision was section 213(a) of the Housing and Urban Development Act of 1969, Public Law 91-152, sponsored by Senator Edward Brooke III of Massachusetts and enacted December 24, 1969. The Omnibus Budget Reconciliation Act of 1981 raised the share to 30 percent, where it has remained, and the voucher program adopted the same roughly 30 percent tenant contribution. The Quality Housing and Work Responsibility Act of 1998 added income targeting rules requiring that a high share of new admissions be extremely low income households, concentrating both programs on the poorest eligible renters. These formulas mean the tenant’s rent moves with income rather than with the market: when earnings fall, the subsidy rises automatically, which makes both programs countercyclical by design and makes their budgetary cost sensitive to recessions in ways fixed-subsidy programs are not.

The two rent rules also treat income changes differently, and the difference shapes work incentives in ways the debate often ignores. In public housing, the Brooke Amendment’s income-based rent means a tenant who earns more pays more, up to the capped share: the rent rises with income by statute. In the voucher program, the household’s contribution likewise rises with income under the program’s rent formula. Both models thus impose an effective marginal tax on earnings within the program. The practical difference is smaller than the rhetoric suggests, because both use income-based contributions; the meaningful variation is in how each model handles the transition out. A public housing tenant whose income rises beyond eligibility faces the loss of the unit itself, a stark cliff. A voucher holder in the same position faces the phase-out of the subsidy while keeping the apartment and the landlord relationship, a gentler slope. Neither design is obviously superior: the cliff preserves units for the poorest at the cost of penalizing advancement, while the slope eases advancement at the cost of subsidizing households further up the income scale. The choice between them is a value judgment about whom the program should prioritize, not a technical question the evidence settles.

The legal instrument binding the authority to the department is the Annual Contributions Contract, under which the federal government commits to annual operating and capital subsidies and the authority commits to operating the housing under federal rules. The contract is the landlord model’s constitutional document: it creates the federal-local partnership, defines the subsidy formulas, and carries the regulatory obligations on admissions, rents, maintenance standards, and grievance procedures. When Congress underfunds the formulas the contract references, the authority’s obligations do not shrink to match; the contract promises residents a standard of operation that the appropriations may not support, and the gap becomes the backlog, the deferred maintenance, and the management triage the later sections document. Understanding the contract explains why the landlord model’s failures look like broken promises rather than market outcomes: the promise was contractual, and the funding was discretionary.

The authorities themselves are creatures of state law, typically established by municipalities or counties under state enabling statutes, governed by boards appointed by local elected officials, and ranging from agencies managing tens of thousands of units to small town authorities managing a few dozen. That size distribution matters for the comparison because administrative capacity is not evenly distributed: a large urban authority can staff inspections, landlord outreach, and mobility counseling, while a small authority may administer vouchers with a handful of employees. The voucher program’s performance therefore varies with local administrative investment in ways the national averages obscure, a theme the mobility experiment will make quantitative.

Tenant protections also differ across the two models in ways households feel directly. Public housing tenants are entitled to administrative grievance procedures before adverse actions, a due process architecture built into the program’s regulations over decades of litigation and rulemaking. Voucher holders facing eviction confront private landlords under state landlord-tenant law, with the program’s role limited to termination-of-assistance rules that can end the subsidy if the tenancy fails. The practical consequence is that the landlord model’s tenant holds procedural rights against a government actor, while the payer model’s tenant holds contractual rights against a private actor plus a separate administrative relationship with the authority. Neither arrangement is categorically more protective; each locates the household’s leverage in a different legal relationship, and the difference matters most for the most vulnerable tenants, who are least equipped to navigate either system without assistance.

There is a third arrangement that complicates the binary, and it will matter when the article reaches the Rental Assistance Demonstration. Project-based vouchers and project-based rental assistance attach the subsidy to the unit rather than the household. Under project-based vouchers, authorized at 42 United States Code 1437f(o)(13), an authority can commit a share of its voucher funding to specific developments, so the assistance stays with the building while tenants come and go. Under project-based rental assistance, long-term contracts with private owners provide the subsidy for every assisted unit in a property. These hybrids borrow the landlord model’s place-based stability and the payer model’s private ownership, and they are the destination toward which much of the remaining public housing stock has been migrating. The comparison is therefore not quite a duel; it is a spectrum, with tenant-based vouchers at the portable end, traditional public housing at the place-based public end, and project-based assistance occupying the middle ground that policy has increasingly favored.

Congress capped the project-based share of voucher funding, generally at 20 percent of an authority’s authorized voucher units under the statute as it stood at the reference date, which kept the hybrid from swallowing the tenant-based program it was meant to complement. The cap on project-basing is a policy choice of a different kind than the Faircloth cap: it rations a flexibility rather than freezing a stock, and authorities have used the allowance strategically, committing project-based vouchers to developments serving elderly households, people with disabilities, or formerly homeless families, populations for whom the search frictions of tenant-based assistance are steepest. The existence of the allowance, and its limits, confirms the article’s structural point. Policymakers have repeatedly reached for instruments between the two pure models, and the law has repeatedly bounded how far toward either pole an authority may go.

The household’s experience differs across these arrangements in concrete ways. A public housing tenant applies to a specific development or an authority’s list, waits for a unit in the portfolio, and accepts the neighborhood where the authority built. A voucher holder searches the private market, negotiates with landlords, and in principle can choose among neighborhoods, subject to the payment standard, the availability of willing landlords, and the frictions of search that the mobility literature documents in detail. A project-based tenant gets the stability of a designated development without the portability of a voucher. Each arrangement answers a different question about what the household values most: certainty of a subsidized unit, freedom to choose location, or the combination of a decent building and a decent block. The evidence on which arrangement delivers which outcome is the subject of the later sections; the point here is that the models are not interchangeable technologies for delivering the same good, but different allocations of choice, risk, and control among the household, the authority, and the market.

Scale matters for interpreting everything that follows. Tenant-based vouchers assist a larger number of households than public housing does, and the gap has widened as the public stock shrank. That asymmetry is not a finding about effectiveness; it is largely the mechanical result of the cap described in the next section. Any comparison that treats the two programs as competitors on a level field is misreading the field. The field was tilted by statute in 1998, and the tilt has compounded for the entire period the evidence covers.

The Legal Cap: How a 1998 Funding Rule Froze the Public Housing Stock

The Faircloth Amendment is the single most consequential sentence in modern federal housing law, and it is worth stating with the precision the verifier’s work demands. It is Section 9(g)(3) of the United States Housing Act of 1937, added by the Quality Housing and Work Responsibility Act of 1998, which was Title V of Public Law 105-276, 112 Statutes at Large 2518, approved October 21, 1998. It is codified at 42 United States Code 1437g(g)(3)(A), and the statute names its own baseline date: October 1, 1999. The operative language provides that, except as provided in subparagraphs (B) and (C), a public housing agency may not use any of the amounts allocated to it from the Capital Fund or the Operating Fund for the purpose of constructing any public housing unit, if such construction would result in a net increase from the number of public housing units owned, assisted, or operated by the agency on October 1, 1999, including any units demolished as part of any revitalization effort. The provision is named for Senator Lauch Faircloth of North Carolina, who served in the Senate from 1993 to 1999.

The crucial precision is that this is a funding restriction, not a literal ban on owning more units. The statute does not say an authority may not own more units than its 1999 count. It says the authority may not use its Capital Fund or Operating Fund allocations to construct units that would push it past that count, subject to the exceptions in subparagraphs (B) and (C). In practice the distinction has rarely mattered, because the Capital and Operating Funds are the only federal money available for public housing construction, and Congress has appropriated no separate development funding for the program since the mid-1990s. A restriction on the only available funding is, functionally, a restriction on the activity. But the precision matters for two reasons. First, it explains the exceptions: the statute contemplates circumstances, spelled out in the following subparagraphs, in which the restriction does not apply. Second, it corrects the common shorthand that the amendment “banned” new public housing outright, a shorthand that overstates what Congress wrote even as it accurately describes what Congress achieved.

The baseline date deserves its own emphasis because confusion about it is common. The Quality Housing and Work Responsibility Act was enacted on October 21, 1998. The Faircloth baseline is October 1, 1999, nearly a year later, and the date appears in the statute text itself. The gap between enactment and baseline gave authorities a window in which units in the pipeline counted toward the baseline, and the Department of Housing and Urban Development’s subsequent guidance, including guidance issued in 2019, after this article’s 2016 reference date, on the maximum number of units eligible for operating subsidy under the provision, treats the October 1999 count as the operative ceiling. An authority’s Faircloth limit is therefore a historical fact about its own portfolio on a single day, not a national number, and different authorities face different ceilings reflecting the different sizes of their 1999 stocks.

Why a funding restriction rather than a direct prohibition? The answer lies in how Congress legislates housing. The appropriations power is the lever Congress actually controls year to year, and attaching the restriction to the Capital and Operating Funds meant the cap would bind automatically without requiring new enforcement machinery. It also meant the cap would be invisible to anyone reading the authorizing statute casually: the Housing Act authorizes public housing in broad terms, and only the funding section carries the sentence that froze the stock. This is the quiet-foreclosure pattern the series thesis thread identifies. The debate about whether to build more public housing was not resolved by argument; it was resolved by a funding rule that made one side of the debate unlawful to pursue with federal money, and the resolution compounded silently for years while attention moved elsewhere.

The Quality Housing and Work Responsibility Act that carried the amendment was itself a product of the welfare reform era, and its broader contents explain the political soil in which the cap grew. The act deregulated public housing authority operations, converted the old operating subsidy formulas, imposed work and self-sufficiency provisions including the community service requirement, required income targeting rules concentrating new admissions on extremely low income households, and merged the certificate and voucher programs into the single Housing Choice Voucher program. The Faircloth provision fit the act’s logic: reform the existing stock, discipline its management, expand the voucher alternative, and freeze the public housing footprint while doing so. Understanding the cap as one element of a comprehensive reform, rather than as an isolated technical amendment, explains why it survived unchallenged for so long. It was not a rider smuggled into an unrelated bill; it was the physical-plant corollary of a statute whose premise was that the future of assisted housing lay in subsidies rather than in authority-owned buildings.

The provision also generated interpretive questions that the housing department spent years answering, which is itself evidence of how much turned on a single subsection. Counting the baseline required determining which units were owned, assisted, or operated on the October 1999 date, how demolitions in progress were treated, and how the exceptions applied, and the department’s guidance documents show administrators working through the provision’s ambiguities long after enactment. The durable point is that a funding restriction written in 1998 was still generating interpretive work decades later, which is what binding constraints do: they do not fade, they accrete procedure.

The practical consequence has been exactly what the brief describes: every modern proposal for expanding federally assisted rental housing has been framed around vouchers, tax credits, or conversion rather than construction of new public housing. The Low Income Housing Tax Credit, created by the Tax Reform Act of 1986 and administered through the tax code rather than the housing department, became the country’s principal production program precisely in the era when the public housing pipeline closed. Voucher funding became the margin on which assisted-housing policy was debated. And the Rental Assistance Demonstration, discussed below, converts existing public housing into project-based Section 8 contracts rather than adding to the public stock, because adding to the public stock is what the statute restricts. The cap did not merely freeze a number; it rerouted the entire policy imagination of American rental assistance around itself.

What does the Faircloth Amendment actually prohibit?

The amendment bars a housing authority from using its federal Capital or Operating Fund allocations to build units raising its total above the number it owned, assisted, or operated on October 1, 1999. It allows the narrow statutory exceptions. With no other federal construction funding available, it has blocked net additions to the stock since 1999.

The provision sits within a larger statutory sequence that readers can trace through the era guide covering the full arc of housing legislation since 1949, where the 1998 act appears as the hinge between the construction era and the subsidy era. Understanding that hinge is essential because the cap is routinely misdescribed in both directions. Supporters of expanded public housing sometimes describe it as a mere funding drought that a willing Congress could reverse with appropriations, which understates the statutory barrier: appropriating development money would not suffice unless the Faircloth restriction itself were amended, because the restriction attaches to the funds, not to the willingness to spend. Opponents of repeal sometimes describe the cap as a settled judgment that public housing failed, which overstates what the provision says: it is a funding rule with a baseline date, not a finding of fact about the model’s merits. The statute is silent on whether public housing works. It simply forbids paying for more of it.

There is a further subtlety in the phrase “net increase” that repays attention. The statute measures the agency’s total against its 1999 baseline, which means demolitions and dispositions reduce the count and new construction restores it only up to the baseline. An authority that demolished distressed units under the programs described in the next section did not bank the demolished units as credits toward future construction; the baseline is fixed at the 1999 number, and rebuilding up to that number is permitted only insofar as the funding restriction allows. The inclusion of units “demolished as part of any revitalization effort” in the baseline count prevents authorities from treating revitalization demolitions as reducing their ceiling, a detail that mattered enormously during the HOPE VI era when large numbers of units were coming down. The cap and the demolition programs thus interacted: the same statute that authorized the revitalization, the Quality Housing and Work Responsibility Act, both funded the demolition of distressed stock and froze the ceiling against which the rebuilt stock would be measured.

The interaction with HOPE VI redevelopment produced a particular irony the statute’s structure guaranteed. When a HOPE VI grant demolished 500 distressed units and rebuilt 150 public housing units alongside tax-credit and market-rate homes, the 150 rebuilt units counted against the same 1999 baseline the 500 demolished units had occupied, consuming baseline room without expanding the stock. Mixed-finance redevelopment, the era’s preferred answer to distressed developments, therefore shrank the public housing count by design: every mixed-income rebuilding traded public housing units for other kinds of affordability, and the Faircloth ceiling ensured the trade could never be reversed by building elsewhere. The statute did not merely freeze the stock; combined with the redevelopment programs, it ratcheted the stock downward, because every transformation of the portfolio moved units out of the counted category faster than any permitted construction could move them back in.

The exceptions in subparagraphs (B) and (C) are narrow, and the article will not overstate them, but their existence is part of the precise account. They contemplate circumstances in which construction above the baseline is permitted, and their narrowness is itself evidence of congressional intent: the default is the freeze, and departures from it require fitting within the statutory carve-outs. For the comparison’s purposes, the exceptions do not change the picture. No authority has used them to meaningfully expand its public stock, and the national inventory has moved in one direction since the baseline date. The cap is the binding constraint on the public housing side of the ledger, and any honest comparison must price it in before weighing costs or outcomes.

The Shrinking Stock: Demolition, the Capital Backlog, and RAD

The public housing stock did not merely freeze in 1999; it shrank, substantially, through programs Congress designed for that purpose. The largest was HOPE VI, and its history requires the two-step account the verification work established, because the common one-step version is wrong. The program began not as an authorized program but as the Urban Revitalization Demonstration: the fiscal year 1993 appropriations act for the housing department, Public Law 102-389, enacted in October 1992, included 300 million dollars for the demonstration. From 1993 through 1998 the program remained unauthorized, funded year to year through appropriations riders. Formal authorization arrived with the Quality Housing and Work Responsibility Act of 1998, whose Section 535 added Section 24 to the 1937 Act, codified at 42 United States Code 1437v. The article excludes the claim found in some secondary sources that the Housing and Community Development Act of 1992 authorized the program; the dedicated Congressional Research Service report on HOPE VI contradicts that account, and the verified record is the appropriations launch followed by the 1998 authorization.

HOPE VI’s purpose was the redevelopment of severely distressed public housing, and its mechanism was demolition followed by mixed-income rebuilding with a smaller public housing component. The scale was large. The Congressional Research Service’s HOPE VI report, drawing on housing department program data, records that revitalization grantees demolished 89,892 public housing units while demolition-only grantees planned another 57,593, putting the total above 100,000 units for the HOPE VI era. A separate compilation by Case Western Reserve University researchers for the housing department’s research arm counted 98,592 demolished units replaced by 55,318 public housing units plus 28,979 other affordable units. The arithmetic is the point: roughly 99,000 public housing units came down and roughly 55,000 public housing units went back up, a net loss of more than 40,000 public housing units from this program alone, with the balance of the replacement stock carrying income restrictions but not public housing status. Whether that trade was worth making depends on the condition of what was demolished and the quality of what replaced it, and the literature disputes the answer; what is not disputable is the direction of the stock.

Alongside HOPE VI ran the standing demolition and disposition authority in Section 18 of the 1937 Act, codified at 42 United States Code 1437p, which the Quality Housing and Work Responsibility Act rewrote effective October 21, 1998. Section 18 governs the circumstances under which an authority may demolish or dispose of public housing, and it became the workhorse authority for removals outside the HOPE VI grants. The two authorities together, the grant program and the standing section, account for the bulk of the inventory decline. The construction-era stock that the 1949 act had authorized, the great wave of building that the urban renewal article in this series reconstructs, met its end through these two legal channels, and readers tracing that arc should see how the 1949 act’s construction era closed through the demolition authorities of the 1990s.

The stock that survived carried a repair burden that the numbers state starkly. Abt Associates, in a study prepared for the housing department using 2010 data, estimated the existing capital need, the backlog, at 25.6 billion dollars, roughly 23,365 dollars per unit. The department’s own June 2011 headline rounded the figure to 26 billion dollars in major repairs needed. The same study estimated additional accrual needs of 3.4 billion dollars per year, about 3,155 dollars per unit, meaning the backlog was not a static number but a growing one wherever annual capital funding fell short of accrual. The study also recorded a 3.4 percent decrease in the backlog relative to the 1998 study, partly explained by a 9 percent decline in the number of public housing units: some of the backlog disappeared because the buildings carrying it were demolished. That last detail captures the era’s logic with uncomfortable neatness. The portfolio shrank, the repair bill shrank with it, and neither movement testified to the health of what remained.

Section 18’s bite depended on a background rule that Congress had already removed. Earlier law had required one-for-one replacement of demolished public housing units, which meant demolition could not reduce the assisted inventory. The 1995 rescissions legislation lifted that requirement, and the rewritten Section 18 therefore operated in a legal environment where demolition meant net loss. The provision requires departmental approval based on findings about obsolescence, viability, and the interests of residents, but the approvals flowed steadily through the late 1990s and 2000s as authorities sought to shed developments whose capital needs exceeded any plausible funding. Each approval was a local decision with a national cumulative effect, and the cumulative effect is the inventory curve the CRS figures trace.

The backlog’s persistence reflected the structure of capital funding as much as its level. The Capital Fund distributes money by formula, and for years the formula-driven appropriations ran well below the accrual rate the Abt study measured, which meant authorities were triaging: emergency repairs first, then systems at risk of failure, then everything else deferred. Deferred maintenance compounds, because a roof patched this year becomes a roof plus water damage next year, and the per-unit accrual figure of about 3,155 dollars was itself an average over a portfolio whose oldest developments needed far more. The authorities that managed their capital planning best could stretch the dollars; none could repeal the arithmetic. This is the mechanism behind the complication section’s claim that the backlog measures appropriations rather than ownership. Operating subsidies told a parallel story: in multiple years Congress funded the operating formula below full eligibility, forcing authorities to absorb the shortfall through reserves, service cuts, or further deferral of maintenance, each of which fed the next year’s capital need.

The inventory figures complete the picture. A Congressional Research Service brief drawing on 2023 data, published after this article’s 2016 reference date and cited here with that date-wall treatment, counted approximately 1.3 million public housing units in 1993 and just under 920,000 in 2023, of which about 860,000 were occupied. An earlier CRS introduction to the program had put the count at roughly 1.2 million units under contract, down from a peak above 1.4 million. The direction is consistent across sources and periods: a stock that peaked in the low millions lost roughly a third of its units over three decades, with the Faircloth cap preventing any offsetting construction and the demolition authorities actively reducing the count. The 2023 figure is not the state of the stock at this article’s 2016 reference date, and it is not presented as such; it is the later-measured continuation of a trend that was fully visible by 2016.

Into this shrinking, undercapitalized portfolio came the Rental Assistance Demonstration, the conversion program the brief identifies as the third great movement of the stock. RAD was authorized by the Consolidated and Further Continuing Appropriations Act of 2012, Public Law 112-55, approved November 18, 2011. Its mechanism is voluntary conversion: a housing authority may convert public housing, assisted under Section 9 of the 1937 Act, into long-term project-based Section 8 housing assistance payments contracts, either project-based vouchers under Section 8(o)(13) or project-based rental assistance renewable under Section 524 of the Multifamily Assisted Housing Reform and Affordability Act of 1997. The economic logic is that a long-term Section 8 contract is bankable in a way that annual public housing subsidies are not: the authority, or its development partner, can borrow against the contracted revenue stream to finance the rehabilitation the capital backlog demands. Congress initially capped the demonstration at 60,000 units and provided no new funding; the program was a reclassification of existing subsidy commitments, not an expansion of them.

RAD’s significance for the comparison is structural. Each conversion moves units out of the public housing program and into the project-based Section 8 world, which means the public housing inventory falls by definition even when the buildings remain occupied by the same residents at the same rents. The program thus accelerates the long-run migration this article has been describing: from government as landlord toward government as payer, with the hybrid project-based form as the way station. Whether RAD preserves affordability and improves physical conditions at scale was, at the 2016 reference date, a question for the program’s early years, with implementation guidance issued in 2012 and 2013 and the first evaluation evidence covering progress through 2014. The article records the mechanism and the authorization with precision and leaves the long-run verdict to the evidence as it accumulates, because a 2016-dated comparison cannot responsibly grade a program authorized in late 2011 on its lifetime performance.

The contract mechanics are what make the program financially transformative. A project-based voucher contract typically runs 15 years with renewal options, and a project-based rental assistance contract runs long terms as well; either way, the revenue stream is contractual and therefore financeable, unlike the annual appropriations that funded public housing operations. That bankability is what lets an authority or its development partner raise private debt and tax-credit equity for rehabilitation, effectively converting the 25.6 billion dollar backlog from an unfunded liability into a financeable investment. The trade is that the property’s subsidy depends on contract compliance rather than on the public housing program’s regulatory framework, and residents’ protections must be carried by the conversion’s use restrictions rather than by the program they left. Whether that trade serves residents over decades was unknowable at the reference date; the mechanism’s logic, substituting contract revenue for appropriations, was clear from the authorizing act.

Stepping back, the stock story reframes the cost and outcome comparisons that follow. The public housing being compared to vouchers in the research literature is not a random sample of the model; it is the surviving stock after the most distressed developments were demolished, the best candidates were converted, and the remainder accumulated a 25.6 billion dollar repair backlog under appropriations that never matched accrual. The voucher program being compared is the program Congress chose to fund at the margin for two decades. An honest comparison must hold both of those selection effects in view. The next sections report what the studies found; the complication section then asks what the studies could possibly have found, given the field the statute created.

The five-axis comparison table

The table below compresses the five axes into the article’s findable artifact. It is a summary of the analysis, not a substitute for it; the sections that follow supply the evidence behind each cell.

Model	Ownership	Legal authority to expand	Cost per assisted household	Neighborhood outcomes	Market conditions under which each performs better
Traditional public housing	Owned and operated by local housing authorities under federal rules; government is the landlord	Frozen by the Faircloth Amendment: no Capital or Operating Fund use for construction causing net increase above the October 1, 1999 baseline, subject to narrow exceptions	Capital-intensive; housing production programs cost 16 percent to 43 percent more than vouchers over 30 years per GAO-02-76	Residents concentrated in higher-poverty neighborhoods than voucher holders on average; stock includes distressed and well-run developments alike	Markets where new supply is needed and private development will not serve low-income renters; places where long-term public asset control is valued
Tenant-based vouchers	Privately owned units; household holds a portable subsidy and government is the payer	Expandable by appropriation; no statutory unit cap constrains the voucher program	Lower 30-year federal cost per household than production programs in the GAO comparison; cost tracks local rents	Households reach lower-poverty neighborhoods than public housing residents on average, though by less than early advocates expected; implementation quality drives results	Markets with elastic supply and available units below the payment standard; places where household choice of neighborhood is the priority
Project-based assistance (PBV, PBRA, RAD conversions)	Private or authority-affiliated ownership with long-term Section 8 contracts attached to the unit	Expandable within voucher funding for PBV; RAD conversions reclassify existing public housing rather than adding units	Contract rents set by negotiation within program limits; bankable revenue supports rehabilitation financing	Place-based like public housing but typically in mixed-income or redeveloped settings; outcomes vary by development	Markets where specific developments need preservation capital; situations combining stability of place with private financing

What the Money Shows: Production Programs Cost More Than Vouchers

The cost evidence is the least contested part of the comparison, and it favors vouchers clearly within the terms the studies set. The anchor is a January 2002 report from the Government Accountability Office, GAO-02-76, “Federal Housing Assistance: Comparing the Characteristics and Costs of Housing Programs,” which compared housing production programs, including the Low Income Housing Tax Credit, the Section 202 and Section 811 supportive housing programs, the Section 515 rural program, and HOPE VI, against housing vouchers. For units with the same number of bedrooms in the same general location, the report found, the production programs cost more than vouchers. A 2006 follow-up, GAO-06-405, restated the finding as a range: the average 30-year federal cost of the production programs was 16 percent to 43 percent more than voucher costs. The Office of Management and Budget’s analysis, issued as an addendum to OMB memorandum M-02-06, put the same point in net-present-value terms per two-bedroom unit: 86,827 dollars for the voucher program, serving 57,585 households per 5 billion dollars, against 123,409 dollars for construction modeled on the former Section 8 New Construction program, serving 40,515 households per the same 5 billion.

The production side’s cost structure deserves a closer look, because the Low Income Housing Tax Credit works nothing like direct public construction. The credit allocates tax benefits to investors who supply equity for affordable developments, with compliance periods that keep rents restricted for decades. The developer’s return comes from the tax benefits, the investor’s capital displaces public debt, and the development carries the transaction costs of tax syndication: legal fees, accounting, and the pricing discount at which credits sell. Those transaction costs are part of what the GAO range captures. Add land acquisition in high-cost markets, construction financing, developer fees, operating reserves required by lenders, and the wage standards that attach to federally assisted construction, and the per-unit cost climbs well above the voucher’s payment for an existing unit. None of these costs is waste in the pejorative sense; each purchases something, durability, compliance, or labor standards, that the voucher’s market rent does not buy. The comparison is therefore not between an efficient instrument and an inefficient one but between a less expensive instrument that buys temporary affordability in existing units and a dearer one that buys durable affordability in new ones.

The framing discipline the verifier requires matters here. The finding is about housing production programs versus vouchers, not about new public housing construction versus vouchers specifically. The distinction is not pedantry. GAO itself noted that HOPE VI is a modernization program rather than a straightforward production program, and that public housing had received no new appropriations for incremental development since 1996, so the office could not present HOPE VI costs broken out by bedroom size in the way the comparison required. To write that “new public housing costs 16 to 43 percent more than vouchers” would be to claim a comparison the data do not support at that specificity. What the data support is the broader claim: when the federal government pays to produce a dedicated affordable unit, through tax credits, capital grants, or construction subsidies, the 30-year federal cost per assisted household runs 16 to 43 percent above the cost of giving a household a voucher for a comparable unit in the same market. The public housing model is a production model, so the finding bears on it, but the honest statement keeps the GAO frame.

Why production costs more is not mysterious, and the mechanisms are worth stating because they recur in the verdict. A production program pays for land, construction or substantial rehabilitation, financing costs, developer fees, and long-term operating reserves, and it pays them whether or not the local market has vacant units that a voucher could have rented more cheaply. A voucher pays only the gap between what the household can afford and the market rent of an existing unit, which means it free-rides on the existing stock and on the private capital that built it. In markets with slack, that free ride is pure efficiency: the government buys affordability without buying buildings. The cost advantage is therefore largest exactly where the access framing is strongest, in markets where units exist and households simply cannot afford them. The advantage narrows, and can reverse, where the existing stock cannot absorb the demand, which is the supply-side argument the rent-capture section develops.

The Office of Management and Budget figures repay a closer reading because they translate the percentage range into households served. At 86,827 dollars of net present value per two-bedroom voucher unit, 5 billion dollars assists 57,585 households; at 123,409 dollars per constructed unit, the same 5 billion assists 40,515. The gap, roughly 17,000 households per 5 billion dollars, is the opportunity cost of production stated in the units policymakers actually debate: appropriations buy a fixed number of assisted households, and the instrument determines how far the appropriation stretches. The 30-year horizon matters too. Voucher costs recur annually and rise with rents, while production costs concentrate up front and then persist as operating obligations; the net present value calculation is what makes the two time profiles comparable, and it is sensitive to the discount rate in ways the headline percentages do not show. None of this overturns the GAO finding, but it locates the finding’s meaning: production buys fewer assisted households per dollar, over any horizon long enough to matter, and the difference compounds across the budgets Congress actually writes.

There are also costs the GAO comparison does not capture, and evenhandedness requires naming them. The voucher’s cost advantage depends on the continued existence of a private rental stock willing to accept the program’s rents and inspections; if landlords exit or rents rise faster than the payment standard, the effective cost per successful placement rises in ways the 30-year federal cost figure does not show. The production program’s higher cost buys a durable public or regulated asset that persists beyond any single tenancy and cannot be withdrawn by a landlord’s business decision. These are real economic differences, not accounting artifacts, and they map directly onto the risk allocation described in the ownership section: the payer model economizes on capital at the price of depending on the market, while the landlord model pays for capital at the price of depending on appropriations. The GAO range measures the federal budgetary cost; the full social cost includes the risks each model leaves unpriced.

The production programs in the GAO comparison were a specific set, and naming them prevents the finding from floating free of its institutional base. Section 202 finances housing for elderly households, Section 811 for people with disabilities, Section 515 supports rural rental housing through the Department of Agriculture, the Low Income Housing Tax Credit finances privately owned affordable development through the tax code, and HOPE VI funded the redevelopment of distressed public housing. What unites them is the production mechanism: public money pays for the creation or substantial rehabilitation of a dedicated unit, and the affordability is attached to the unit for decades. What distinguishes them from one another matters less for the comparison than what distinguishes all of them from the voucher, which is that none of them free-rides on existing private stock. The GAO range therefore measures the cost of the production decision itself, not the cost of any particular program’s inefficiency, and the range’s width, 16 to 43 percent, reflects real variation in how expensive production is across programs and markets rather than statistical noise.

Is public housing cheaper than vouchers?

No. The Government Accountability Office found in January 2002 that housing production programs cost more than vouchers for comparable units, and restated the gap in 2006 as 16 percent to 43 percent over 30 years. The finding covers production programs broadly, not new public housing specifically, since the data did not support that narrower comparison.

The cost finding also interacts with the Faircloth cap in a way that is easy to miss. Because the cap froze the public stock, the marginal assisted-housing dollar for two decades went to vouchers or tax credits almost by default, which means the observed cost advantage of vouchers partly reflects a policy environment that starved the alternative of the scale and modernization that might have improved its economics. This is not an argument that the GAO numbers are wrong; they are carefully constructed and the range has held up. It is an argument about what the numbers can prove. They prove that, as operated in the period studied, production programs cost more per household than vouchers. They do not prove that a public housing program funded at accrual, built to modern standards, and managed under current practice would cost more, because no such program existed to study. The cap foreclosed the experiment whose results would be needed to answer the question the cost debate pretends is settled.

Where Families Live: Vouchers, Neighborhoods, and the Mobility Evidence

If cost is the voucher’s strongest ground, neighborhood outcomes are its most qualified victory. The consistent finding across two decades of research is that voucher households live in lower-poverty neighborhoods than public housing residents, but by less than the program’s early advocates expected, and the gap between expectation and result is where the most instructive evidence sits. Kerry McClure and Bonnie Johnson, writing in 2014 and summarized in the housing department’s Cityscape journal, found that voucher recipients are less likely to live in impoverished, racially segregated neighborhoods than public housing residents. A 2015 follow-up by the same authors, cited in a department study of assisted households with children, sharpened the point: the lower initial neighborhood poverty rate for voucher households relative to public housing residents speaks to the program’s relative success in limiting exposure to extreme poverty, even if the program has not been wholly successful in facilitating access to low-poverty areas. The double edge of that sentence is the whole literature in miniature. Vouchers move families away from the worst concentrations of poverty; they move fewer families into genuine opportunity than the theory promised.

The early expectations deserve reconstruction because the shortfall is otherwise misread as failure. McClure, writing in 2010, recovered the original hope: that given the ability to move to low-poverty, racially integrated neighborhoods, voucher recipients would choose to do so. The constraints that frustrated the hope were identified early and have been confirmed repeatedly: tight rental markets with few available units, limited landlord participation in the program, and program rents capped below market in the neighborhoods families most wanted to reach, a set of barriers documented by Lance Freeman, or more precisely by Deborah Devine and coauthors in the department’s 2003 study of voucher location patterns, and by Meryl Finkel and Laura Buron in 2001. The voucher is portable in law; its portability in practice depends on a local market’s vacancies, its landlords’ willingness, and its payment standard’s adequacy. Where those three align, mobility follows. Where they do not, the voucher holder’s choice set collapses toward the same high-poverty neighborhoods the program was meant to help families leave.

The fair housing dimension of this evidence belongs to the article in this series that owns it, and readers seeking the segregation analysis in full should consult the study of the Fair Housing Act’s impact on segregation. For the comparison, the relevant point is narrower: the neighborhood gap between the two programs is real but modest, and it is produced less by the voucher’s design than by the geography of the public housing stock. Much public housing was built, under the site-selection practices of the construction era, in neighborhoods that were already poor and often racially segregated, and the developments stayed where they were built. Voucher holders start from no fixed address, so their neighborhood distribution reflects market constraints rather than historical siting decisions. The comparison therefore flatters vouchers partly for a reason that has nothing to do with vouchers: the public housing stock carries the locational legacy of decisions made decades before the voucher program existed.

The most ambitious test of the mobility hypothesis predates the voucher era’s maturity. The Gautreaux program, born of a 1976 consent decree against the Chicago Housing Authority after findings of discriminatory site selection, moved Black public housing families to private-market units in Chicago suburbs, and sociologist James Rosenbaum’s follow-up studies found that suburban movers experienced better employment and educational outcomes than families who moved within the city. The federal Moving to Opportunity experiment, launched by the housing department in 1994 across Baltimore, Boston, Chicago, Los Angeles, and New York, randomized families into three groups: an experimental group receiving vouchers restricted to low-poverty neighborhoods plus counseling, a regular Section 8 group, and a control group. The final impacts evaluation, published in 2011, found significant mental health improvements for adults and girls but no gains in adult economic self-sufficiency, a result that tempered the early enthusiasm. A later follow-up using tax records, published just after this article’s 2016 reference date and noted here with that date-wall treatment, found that children who moved to lower-poverty neighborhoods at young ages earned significantly more as adults, suggesting the experiment’s benefits were intergenerational rather than immediate. The arc from Gautreaux through Moving to Opportunity to the Seattle experiment traces a consistent lesson: mobility’s benefits are real, they accrue most to children, and they depend on the support services that surround the move.

The payment standard mechanics explain much of the implementation gap the literature documents. The housing department sets Fair Market Rents annually for each metropolitan area and county, historically at the 40th percentile of local rents, and authorities set their payment standards as a percentage of those rents. A single metropolitan-wide rent figure cannot track neighborhood-level variation: in high-opportunity neighborhoods where market rents exceed the payment standard, the voucher cannot bridge the gap, while in low-rent neighborhoods the standard may cover nearly the full rent. The result is a structural tilt that pushes voucher holders toward the neighborhoods the program hoped they would leave, not by design but by arithmetic. Reforms that would set rents at finer geographic levels were debated through the 2010s, but at the reference date the metropolitan-wide standard remained the rule, and its mismatch with neighborhood rents was among the constraints the literature had identified since Finkel and Buron’s 2001 work.

The department’s own 2003 study of voucher location patterns by Deborah Devine and coauthors supplied the national baseline against which later claims were measured. Using administrative data, the study mapped where voucher holders lived relative to poverty concentrations and found the pattern the later literature confirmed: voucher households were less concentrated in high-poverty neighborhoods than public housing residents but remained overrepresented in moderate-poverty areas, with substantial metropolitan variation. The variation is itself a finding. In some metropolitan areas voucher holders dispersed widely; in others they clustered nearly as tightly as public housing residents. That geographic heterogeneity is the empirical foundation of the verdict’s refusal to generalize: the voucher’s neighborhood performance is not a national constant but a local outcome, produced by the interaction of payment standards, landlord participation, and market tightness in each metropolitan area.

Why do vouchers not guarantee better neighborhoods?

Because portability in law is not portability in practice. Tight markets, limited landlord participation, and payment standards set below market rents in desirable areas constrain where voucher holders can actually lease. Research from 1997 through 2015 consistently finds voucher households in lower-poverty neighborhoods than public housing residents, but the improvement is smaller than early advocates expected.

The most striking evidence on the implementation question comes from a randomized experiment fielded after this article’s 2016 reference date, cited here with explicit date-wall treatment. Creating Moves to Opportunity, by Peter Bergman, Raj Chetty, Stefanie DeLuca, Nathaniel Hendren, Lawrence Katz, and Christopher Palmer, was conducted in Seattle and King County, Washington, in partnership with the Seattle and King County housing authorities. The experiment randomized 430 families issued vouchers between April 2018 and April 2019, 222 to a treatment group and 208 to a control group, all with at least one child under 15. The treatment bundled information about high-opportunity areas, short-term financial assistance, customized search assistance, and landlord recruitment and connections. The published results, appearing in the American Economic Review in May 2024, eight years after this article’s reference date, found that the bundled treatment raised the share of families moving to high-upward-mobility areas from 15 percent in the control group to 53 percent in the treatment group. A second phase showed that full customized services had roughly five times the effect of information and financial incentives alone, identifying customized search assistance as the primary driver.

The experiment’s implication for the comparison is the one the brief names: the voucher’s weakness is implementation rather than design. The standard voucher, a subsidy plus a briefing, moved 15 percent of families to high-opportunity areas. The same subsidy plus intensive search support and landlord recruitment moved 53 percent. Nothing about the underlying economics changed between the two arms; what changed was the administrative investment in making the market work for the household. That finding reframes the entire neighborhood debate. The question is not whether vouchers can produce mobility, because with support they plainly can, but whether the program as funded and administered provides that support, because as typically operated it does not. A comparison that scores the voucher on its typical administration and public housing on its typical administration is fair; a comparison that treats the voucher’s typical results as its inherent potential understates what the instrument can do when the payer role is staffed as seriously as the landlord role once was.

There is a further implication for the verdict. If the voucher’s neighborhood performance is implementation-constrained, then the binding-constraint framework needs a second axis: not only whether the market’s constraint is supply or access, but whether the administering authority has the capacity to do the landlord recruitment, search assistance, and payment-standard calibration that the experiment shows are decisive. An access-side instrument operated without access-side administration is the worst of both models: it costs the public money without delivering the mobility that justifies the cost. The experiment does not rescue vouchers as a universal answer; it specifies the conditions under which they are the right answer, and those conditions include administrative investments the program’s history shows are the first to be cut.

The Counterevidence: When Demand Subsidies Get Captured by Rents

The strongest argument against relying on vouchers, and the reason the verdict cannot be uniform across markets, comes from the rent-capture literature. Its logic is straightforward and its empirics are uncomfortable. When the government gives households money to spend on rent in a market where the supply of low-cost units cannot quickly expand, part of the subsidy is captured by landlords in the form of higher rents, including rents paid by unsubsidized low-income households who receive no benefit at all. The voucher solves the assisted household’s access problem while worsening the unassisted household’s affordability problem, and the net effect depends on how elastically the local housing supply responds. In markets where supply is elastic, new units arrive, rents stay stable, and the subsidy reaches the household. In markets where supply is inelastic, rents absorb the subsidy and the public dollar leaks.

Scott Susin’s 2002 paper in the Journal of Public Economics, “Rent vouchers and the price of low-income housing,” supplied the first large-scale estimate. Studying the 90 largest metropolitan areas, Susin found that vouchers raised rents by 16 percent on average, a result he described as consistent with low supply elasticity in the low-quality rental market. The distributional arithmetic was stark: he estimated an 8.2 billion dollar increase in total rent paid by low-income households who did not receive vouchers, against 5.8 billion dollars in subsidies received by voucher holders. On those numbers, the program transferred more from unsubsidized poor renters to landlords than it transferred to its intended beneficiaries. The estimate is debated, and the paper’s methods have been contested, but the mechanism it identifies is not in dispute: demand subsidies in inelastic markets move prices, and the incidence of a subsidy is not the same as its intent.

Robert Collinson and Peter Ganong’s 2018 paper in the American Economic Journal: Economic Policy, published after this article’s 2016 reference date, “How Do Changes in Housing Voucher Design Affect Rent and Neighborhood Quality,” working paper version circulated by the Becker Friedman Institute in June 2017, approached the same mechanism from the design side. Studying changes in voucher generosity, they found that an across-the-board increase in generosity benefited landlords through higher rents with minimal impact on neighborhood or unit quality: every 1 dollar increase in the fair market rents that set voucher payment standards raised actual rents by about 47 cents, with no meaningful quality improvement. They interpreted the result as landlord price discrimination in response to rent-ceiling changes. The finding matters for the comparison because it severs the assumed link between spending more on vouchers and getting better housing. More generous vouchers, in the markets studied, bought higher rents rather than better units or better neighborhoods. The payer model’s core promise, that the subsidy follows the household to better housing, fails when the market’s response to the subsidy is to reprice the existing stock.

The economics behind these findings is the theory of subsidy incidence, which holds that the party who nominally receives a subsidy is not necessarily the party who economically benefits from it. When supply cannot expand, a subsidy to renters increases the price of the fixed stock, and the benefit flows to whoever owns the stock. Landlords can additionally price-discriminate against voucher holders because the payment standard is observable: a landlord who knows the program will pay up to a given rent has little reason to offer a lower one. This is why the Collinson and Ganong result takes the form it does, with higher payment standards translating into higher rents rather than better units. The voucher program’s designers understood the access problem; the incidence literature shows they underestimated the extent to which the market would convert access subsidies into price increases where supply was fixed.

The distributional consequence is the literature’s most uncomfortable finding and deserves plain statement. Susin’s arithmetic implies that unsubsidized low-income renters, households too poor for market rents but without vouchers, paid 8.2 billion dollars more in rent while voucher recipients received 5.8 billion dollars in subsidies. If those magnitudes are approximately right, the program’s largest beneficiaries in the markets studied were landlords, its intended beneficiaries came second, and a group of poor households with no program connection paid for part of the transfer without receiving any of it. Policymakers weighing voucher expansion in tight markets must confront that incidence directly, because a program that helps its recipients while harming equally poor non-recipients is not the unambiguous good its authorizing language assumes. This is the sense in which the rent-capture literature is the strongest supply-side argument: not that vouchers never work, but that their costs fall partly on people the program was never meant to tax.

The verifier’s framing note applies to both papers: neither is presented by its authors as a study of “inelastic-supply markets” in those exact terms, and the article describes them as voucher-generosity incidence and rent-capture findings rather than attributing that framing verbatim. The supply-elasticity interpretation is the literature’s and this article’s synthesis, not a quotation. That synthesis is nevertheless the load-bearing one for the verdict. Susin and Collinson and Ganong together establish that the voucher’s cost advantage, documented by GAO in the cross-section, can erode or reverse in markets where supply does not respond, because the subsidy’s incidence shifts from tenant to landlord. The GAO comparison holds rents constant across the two models; the rent-capture literature shows that rents are not constant with respect to the voucher program’s own scale. A small voucher program in an elastic market is the efficient instrument the GAO numbers describe. A large voucher program in an inelastic market is a transfer to landlords that the Susin numbers describe. The instrument is the same; the market makes the difference.

This is also where the public housing side of the ledger gets its strongest affirmative argument, and evenhandedness requires stating it at full strength. A produced unit, whether public housing, tax-credit, or otherwise supply-side, adds to the physical stock. It cannot be captured by rent inflation because its rent is not set by the market, and its existence expands supply rather than bidding for it. In the inelastic coastal markets where the rent-capture findings bite hardest, the supply-side case is not nostalgia for the construction era; it is the elementary economics of adding units where units are the binding constraint. The Faircloth cap’s tragedy, on this view, is not only that it froze a particular program but that it froze the supply-side instrument precisely as the country’s productive metropolitan areas were becoming the places where supply-side instruments matter most. The cap was written for a national debate; its costs are local, concentrated in the markets where vouchers work worst.

The counter to the counter also deserves its due, because the supply-side argument is sometimes stated as if production escapes market discipline entirely. Produced units cost 16 to 43 percent more per household over 30 years in the GAO comparison, they take years to deliver against an immediate affordability crisis, and their location is fixed at construction, reproducing the siting problems that burdened the public housing stock. A supply-side strategy that builds slowly, expensively, and in the wrong places can easily do less for low-income households than a demand-side strategy that accepts some rent capture as the price of speed and choice. The honest form of the supply-side argument is conditional: where supply is the binding constraint and production can be delivered at reasonable cost and in reasonable locations, production dominates. Where those conditions fail, the rent-capture critique of vouchers does not by itself justify building. The verdict section turns these conditionals into a decision procedure.

What Running Each Model Requires

The two models demand different administrative capacities, and the difference explains part of their political history. Running public housing requires a housing authority to be a competent property manager at scale: to staff maintenance crews, manage capital planning across a portfolio of aging buildings, handle tenant relations and grievance procedures, and keep occupancy high enough that rental income covers its share of costs. Authorities vary in these skills as landlords vary. The well-run properties this article’s complication section acknowledges are typically the product of authorities that managed these functions well despite constrained funding; the distressed properties are typically the product of the opposite combination, constrained funding plus weak management, concentrated in the largest developments where scale magnified every failure. Because the authority is a single visible institution, its management quality becomes the program’s public reputation in that city.

Running the voucher program requires a different and less visible skill set. The authority must determine eligibility and calculate subsidies accurately, inspect units against quality standards on schedule, disburse payments to landlords reliably, and recruit and retain landlords willing to participate. Landlord recruitment is the function with no counterpart in the landlord model, and it is the one most authorities underfund: it requires outreach staff, relationship management, and sometimes financial incentives, none of which the program’s administrative funding has historically supported at the level the Creating Moves to Opportunity experiment suggests is needed. An authority can administer vouchers competently on paper, processing applications and cutting checks, while failing at the landlord-facing work that determines whether the vouchers actually lease in high-opportunity areas. The program’s statistics will show the checks cut; they will not show the neighborhoods forgone.

The funding streams reinforce the management divide. The authority’s public housing budget is a residual: whatever Congress provides for the Operating Fund and Capital Fund, minus fixed costs, determines what maintenance gets done, because the rent cap means the authority cannot pass costs to tenants and the Faircloth cap means it cannot grow its way to scale. The voucher budget residual works differently: when appropriations fall short, the authority serves fewer families or lowers payment standards, but the units already leased continue to be maintained by their private owners. The public housing shortfall degrades buildings; the voucher shortfall shrinks the rolls. Both are cuts, but they cut different things, and the political visibility of the two cuts differs accordingly. A deteriorating development is visible to critics; a family that never receives a voucher is not.

The Waiting List as Institution

The waiting list is the institution where the two models’ scarcities take their most human form, and comparing how each model rations access reveals as much as any cost study. In public housing, the waiting list is a queue for specific buildings. An authority opens its list, collects applications, applies federal income targeting and local preferences within that framework, and assigns units as they vacate. The wait is measured in years, the assignment is take it or leave it, and the family’s position reflects the interaction of its preferences, its income tier, and the turnover rate of the developments it qualifies for. Because the assistance is the unit, the list cannot move faster than the buildings empty, and in authorities with stable, desirable developments, turnover is slow and the list barely moves at all.

In the voucher program, the waiting list is a queue for a subsidy that must then survive the market. The authority opens its list, often for a brief window, and the resulting list may be subject to a lottery before it is even ordered. When a family’s name reaches the top, it receives the voucher and enters the search process, with its months-long clock and its market-dependent odds. The voucher list therefore rations twice: once by the queue and once by the search. A family can wait years for a voucher and then lose it because no landlord would lease within the payment standard. This double rationing is invisible in the program’s headline numbers, which count vouchers issued rather than searches succeeded, and it falls hardest on the families with the least flexible schedules, the least transportation, and the least experience negotiating with landlords.

Neither queue is fair in any deep sense, and neither could be: both ration a benefit worth thousands of dollars a year among eligible populations many times larger than the funded slots. But the queues fail differently, and the difference tracks the models’ structures. The public housing queue fails by immobility, keeping families waiting for buildings that do not turn over. The voucher queue fails by attrition, issuing assistance that the market then withholds. A policymaker choosing between the models on access grounds should ask which failure the local market mitigates: where turnover is reasonable and developments are decent, the public housing queue delivers; where landlords participate and units are available, the voucher queue delivers. Where neither holds, neither queue is the answer, and the honest response is more funding or more supply rather than a different rationing rule.

The Complication: Public Housing Did Not Get a Fair Retrial

Before the verdict, the article owes the reader the complication the brief requires: the possibility that the comparison has been rigged, not by the researchers but by the statute. The public housing model that the cost and neighborhood literatures evaluate is a model that Congress stopped funding for growth in the mid-1990s, capped by statute in 1998, subjected to large-scale demolition of its most distressed properties, and left with a 25.6 billion dollar capital backlog measured against 2010 data. To compare that model’s outcomes to the voucher program’s outcomes, and to read the difference as a verdict on public versus private provision, is to mistake the results of a policy regime for the properties of an institutional form. The stock that was studied was the stock that survived a demolition program aimed at its worst developments, a funding freeze aimed at its growth, and an appropriations pattern that never matched its capital accrual. That is not a controlled experiment. It is a historical artifact wearing an experiment’s clothes.

Consider the capital backlog in this light. The 25.6 billion dollar figure is routinely invoked as evidence that public housing failed, as if the number measured something about public ownership. What it actually measures is the cumulative difference between what the buildings needed and what Congress appropriated, over decades, for a portfolio Congress had simultaneously decided not to expand. A private landlord with the same gap between maintenance needs and maintenance spending would show the same backlog; the backlog is an accounting identity, not an institutional diagnosis. The Abt Associates study itself recorded that part of the backlog’s decline between the 1998 and 2010 measurements came from demolishing the buildings that carried it. An observer who reads the backlog as proof that government cannot maintain housing has the causation backwards. The backlog proves that Congress chose not to pay for maintenance, which is a fact about appropriations, and then chose to read the resulting decay as a fact about ownership, which is a fallacy.

Consider next the quality distribution within the stock. The public housing inventory has always included both severely distressed developments, the ones HOPE VI was created to address, and well-run properties with stable tenancies and long waiting lists. The literature’s averages blend the two, and the demolition programs systematically removed the worst of the stock, which means the surviving portfolio that researchers compared to vouchers was already filtered. But the filtering cuts both ways for the argument. If the worst developments were demolished, the remaining stock should have compared favorably, yet the neighborhood and cost gaps persisted. The fair reading is not that filtering vindicates either side; it is that the stock’s heterogeneity makes program-level averages a crude instrument, and that the well-run properties demonstrate something the averages obscure: that the landlord model, competently administered and adequately funded, houses people decently, because it did and does wherever those conditions held.

The department’s own assessment systems complicate the failure narrative further. The Public Housing Assessment System scores authorities on physical condition, financial health, management operations, and capital fund performance, and its results have always shown wide dispersion: troubled agencies under intensive federal oversight alongside high performers managing large portfolios competently for decades. A model that fails everywhere fails as a model; a model that fails in some places and succeeds in others fails, if it fails, in its implementation and funding, not in its concept. The HOPE VI mixed-finance developments, which paired public housing units with tax-credit and market-rate units under private management, were in part an attempt to import private-sector management discipline into the public stock, and their mixed record, better physical outcomes at the cost of fewer deeply subsidized units, is itself evidence that the relevant variable was never public versus private in the abstract but the specific combination of funding, management, and scale.

The well-run end of the distribution deserves description, because the debate’s imagery draws almost entirely from the distressed end. Authorities across the country have operated developments for decades with stable tenancies, maintained buildings, and demand for their units that far exceeds availability, a demand the article describes only in general terms because no verified authority-specific figures exist. These properties demonstrate the landlord model’s capacity under adequate funding and competent management, and they complicate any account that treats the model’s failures as inherent. The distressed developments that dominate the public imagination were real, and HOPE VI was a response to real conditions, but generalizing from the worst of the stock to the model itself is the compositional fallacy the complication asks readers to resist. The stock contained multitudes; the averages concealed them.

The siting history deserves the same evenhanded treatment. The Gautreaux litigation established that the Chicago Housing Authority had selected sites and assigned tenants along racial lines, and the consent decree’s remedy was mobility itself: moving families out of the segregated developments the authority had built. That history cuts against any nostalgic reading of the construction era, and the article does not offer one. But it also locates the neighborhood deficits of the public housing stock in decisions made by specific authorities under specific political pressures, not in the concept of public ownership. A development sited to contain rather than to connect will produce poor neighborhood outcomes under any management; the fault lies with the siting, and the siting was a choice. The voucher program’s neighborhood advantage is therefore partly an advantage of starting fresh, unencumbered by the locational sins of the past, and a public housing program permitted to build anew could choose its sites with the same freedom.

The deepest form of the complication is the untested counterfactual. Because the Faircloth cap has barred net new public housing since 1999, the model has not been tested under modern design and management practice. The public housing that exists was overwhelmingly built in the construction era’s idiom: large-scale developments, often sited by the political and racial logics of mid-century urban policy, managed under administrative systems designed for a different century. What a public housing program would look like if designed with current knowledge, smaller scale, mixed-income integration, modern property management, and the mobility evidence internalized, is unknown, because Congress made it unlawful to build and find out. The voucher program, by contrast, has been continuously reformed, studied, experimented upon, and incrementally improved for its entire life, culminating in the kind of randomized implementation research the Creating Moves to Opportunity experiment represents. To compare a living program against a frozen one and declare the living program the winner is to run a race in which one runner was stopped at the starting line in 1999.

None of this refutes the evidence. The GAO cost range is carefully estimated and the rent-capture findings are real. The complication does not ask the reader to disbelieve the studies; it asks the reader to be precise about what they prove. They prove how the two models performed as actually operated under the policy regime that actually existed. They do not prove how the models would perform under symmetric conditions, because symmetric conditions never obtained. A decision-maker who internalizes this distinction will be humbler about the verdict in both directions: slower to declare vouchers the universal answer on the strength of studies conducted during a voucher-favoring regime, and slower to declare public housing the answer on the strength of an untested counterfactual. The verdict that follows is built for that humility. It does not pick a model. It picks a question.

A final institutional note belongs here. The Moving to Work demonstration, authorized in 1996, gave designated housing authorities flexibility to blend their public housing and voucher funding and to experiment with program design, and it represents the closest thing the frozen regime allowed to institutional learning on the public housing side. Its lessons are mixed and contested, and this article does not adjudicate them; it notes the program’s existence as evidence that even within the cap, administrators sought room to adapt the landlord model. The broader point stands. The comparison’s evidence base was generated inside a policy regime that had already chosen its winner by statute, and the honest analyst discounts for that.

Public Housing vs Vouchers: The Verdict Is Supply or Access

The verdict is a decision procedure, not a champion. It proceeds from the namable claim: vouchers solve an access problem and cannot solve a supply problem, so the right instrument depends on whether the local binding constraint is that families cannot afford existing units or that the units do not exist. Everything the article has established feeds into that procedure. The cost evidence says vouchers are cheaper per household where comparable units exist. The neighborhood evidence says vouchers produce modest mobility gains that intensive implementation support can multiply. The rent-capture evidence says demand subsidies leak to landlords where supply is inelastic. The stock history says the public housing alternative was frozen before it could be modernized. The procedure must therefore do two things at once: respect what the evidence proves about each instrument’s performance, and discount for the regime effects that shaped the evidence.

The first step is to name the local constraint, and the article can make this operational without pretending to a precision the data do not support. In a market with meaningful vacancy at rents near the payment standard, willing landlords, and rents that track incomes with some slack, the binding constraint is access: units exist and households cannot afford them. The voucher is the efficient instrument there, and the GAO cost advantage is the relevant evidence. In a market with near-zero vacancy at affordable rents, rents rising faster than incomes, and new construction stalled by whatever combination of economics and regulation, the binding constraint is supply: the units do not exist at prices the subsidy can reach. There the voucher’s subsidy is partly captured in rents, the Susin and Collinson-Ganong findings are the relevant evidence, and the supply-side instrument, whether tax-credit production, public development under a reformed cap, or conversion programs that preserve existing stock, is the honest answer. Most real markets sit between these poles, which is why the procedure’s second step matters.

The second step is to assess implementation capacity honestly. The Creating Moves to Opportunity results, published in May 2024 and cited with the date-wall treatment this article’s 2016 reference date requires, show that the voucher’s neighborhood performance is a choice variable: 15 percent mobility with the standard administration, 53 percent with intensive search assistance and landlord recruitment. A jurisdiction choosing vouchers must therefore ask whether it will fund the administration the experiment shows is decisive, because a voucher program operated as a check-writing operation will reproduce the modest results the earlier literature documented. Symmetrically, a jurisdiction choosing production must ask whether it can deliver units at reasonable cost, on a reasonable timeline, and in locations that do not reproduce the siting errors of the construction era, because the GAO cost range and the stock history are warnings about what production looks like when those disciplines slip. The instrument choice and the implementation commitment are a package; choosing the instrument without the commitment is the characteristic failure mode of both sides.

The third step is to price the regime effects. Any jurisdiction weighing new public housing construction must reckon with the Faircloth cap as a legal fact: net new public housing cannot be built with Capital or Operating Fund money above the 1999 baseline, and working around the cap means vouchers, tax credits, conversion, or a statutory amendment. That legal reality does not settle the substantive question, but it settles the feasible set, and a decision procedure that ignores it is not serious. Conversely, any jurisdiction weighing voucher expansion must reckon with the evidence that the voucher program’s track record was compiled as the favored instrument of a regime that starved the alternative. The honest discount runs both ways, and the procedure’s output should be stated with the humility the complication section demands: a recommendation conditional on the local constraint, the implementation commitment, and the legal feasible set, not a universal ranking.

Applied to stylized market types, the procedure yields verdicts that are defensible because they are conditional. In a loose market with available units and participating landlords, expand tenant-based vouchers and invest in the search assistance and landlord recruitment the experiment validates; the cost evidence and the mobility evidence point the same way. In a tight market with inelastic supply, pair any voucher assistance with supply-side production or preservation, because the rent-capture evidence shows that demand subsidy alone leaks; this is the strongest case for the production programs the GAO found costlier, since the alternative is not cheaper vouchers but vouchers whose subsidy accrues to landlords. In a market losing affordable stock to deterioration or conversion, prioritize preservation through instruments like RAD-style project-based contracts that keep units in the assisted inventory; the stock history shows that lost units are not replaced under the cap. In a jurisdiction with a well-run public housing portfolio and the management capacity to sustain it, defend the stock’s capital funding before chasing conversions, because the complication section’s lesson is that adequate funding is the difference between the model’s successes and its failures.

What the procedure forbids is the unconditioned verdict. “Vouchers are better” is false in inelastic markets where the subsidy is captured. “Public housing is better” is false as a description of a program frozen since 1999 and untestable in its modern form, and it is in any case not an available choice without statutory change. “Build more” is not a plan until it names the funding source the Faircloth Amendment leaves open. “Expand vouchers” is not a plan until it names the implementation support that makes vouchers work. The debates that ignore these conditionals are the ones the namable claim diagnoses: arguments conducted without naming the binding constraint, talking past each other indefinitely, each side armed with real evidence about a different market.

Consider how the procedure handles two stylized cases, offered as illustrations of the method rather than prescriptions for any real jurisdiction. In the first, a mid-sized metropolitan area with apartment vacancy above historical norms, a payment standard that reaches mid-market rents, and landlords who participate in the program: the constraint is access, the GAO cost advantage applies at full force, and the efficient move is voucher expansion paired with the search assistance and landlord recruitment the Seattle experiment validated. In the second, a coastal metropolitan area with vacancy near zero, rents rising well above incomes, and little new construction: the constraint is supply, voucher expansion alone would partly capitalize into rents per the incidence findings, and the honest answer combines preservation of existing assisted stock with production through the tax credit or other supply-side tools the Faircloth cap leaves open. Neither illustration picks a national winner. Each picks the instrument the local constraint demands, which is the only kind of verdict the evidence supports.

The supply-side toolkit available within the Faircloth constraint deserves explicit inventory, because the verdict’s supply-side recommendations must name instruments the law permits. The Low Income Housing Tax Credit remains the principal production engine, financed through the tax code outside the appropriations the cap governs. The Housing Trust Fund, created by the Housing and Economic Recovery Act of 2008, provides grants for extremely low income rental housing, though its funding history was modest through the reference date. Project-based vouchers let authorities attach assistance to specific developments within the voucher program’s own funding. And RAD conversions preserve existing public housing as project-based Section 8 stock. None of these is public housing as the 1937 act conceived it, and that is the point: the feasible supply-side set is defined by working around the cap, not through it, and a decision-maker who wants production must assemble it from these parts. The credit’s existence also clarifies what the Faircloth cap did and did not foreclose. Congress had moved affordable production into the tax code in 1986, twelve years before the cap, so what the cap fenced off was specifically public production, housing owned by public authorities, not production as such. The voucher-versus-public-housing debate is therefore partly a historical debate: it asks which of two twentieth-century instruments was better, while the twenty-first century’s principal production instrument sits in the tax code, outside the housing statutes and decisive in practice.

A final hypothetical clarifies what the statute does and does not determine. Were Congress to amend the Faircloth provision tomorrow, the legal barrier to net new public housing would fall, but none of the economic findings would change: production would still cost 16 to 43 percent more per household over 30 years, vouchers would still leak to landlords in inelastic markets, and the choice between instruments would still turn on the local binding constraint. The cap determines the feasible set, not the optimal choice within it. Conversely, were Congress to leave the cap untouched, the feasible set would remain vouchers, tax credits, and conversion, and the decision procedure would still be needed to allocate among them. The statute froze one instrument; it did not answer the question the instruments were meant to address. That question, supply or access, survives any feasible legislation, which is why the article makes it the verdict’s deciding factor rather than any particular program’s name.

For readers using this comparison as a study path, the article’s recommendation is to master the statutory sequence before the policy debate, because the debate is unintelligible without the cap. The guide for readers deciding what to study first in this series lays out that sequence as a recommended study order, and the housing statutes repay that approach: the 1937 act that created both programs’ institutional homes, the 1974 act that created Section 8, and the 1998 act that froze one side of the comparison. A legislation study notebook can hold the citations this article has assembled, from 42 United States Code 1437g(g)(3)(A) through the GAO and journal references, in the order the argument needs them. For structured review of the civics concepts underneath the statutes, a civics study tool offers a complementary drill path. The comparison is ultimately a legal artifact wearing an economic debate’s clothes, and studying the law first is the faster route through the economics.

When do vouchers beat construction, and when do they lose?

Vouchers beat construction in loose markets where affordable units exist and landlords participate: they cost less per household and preserve choice. They lose in tight, supply-constrained markets where subsidies are captured in higher rents. The deciding factor is always the local binding constraint, supply or access, never the instrument in the abstract.

The One Test, restated as the article closes: the federal government houses low-income families through two instruments, the landlord model of public housing and the payer model of tenant-based vouchers. The Faircloth Amendment, Section 9(g)(3) of the 1937 Act added by the Quality Housing and Work Responsibility Act of 1998 and codified at 42 United States Code 1437g(g)(3)(A), bars the use of Capital and Operating Fund money for construction that would raise an authority’s unit count above its October 1, 1999 baseline, and that single funding restriction has made net new public housing effectively unlawful for the entire period the modern evidence covers. Housing production programs cost 16 to 43 percent more per household than vouchers over 30 years in the GAO comparison, voucher households reach lower-poverty neighborhoods than public housing residents though by less than early advocates expected, intensive implementation support multiplies the voucher’s mobility effects severalfold, and demand subsidies are partly captured in higher rents where supply is inelastic. The defended verdict names its deciding factor: whether the binding constraint in a given market is supply or access. Name the constraint, fund the implementation the constraint demands, and discount for the regime that froze one side of the race before it was run. Everything else is commentary.

Frequently Asked Questions

Q: Which is better, public housing or vouchers?

There is no uniform answer, and the evidence explains why the question resists one. Vouchers cost less per household than housing production programs over 30 years, by 16 to 43 percent in the Government Accountability Office’s 2002 and 2006 comparisons, and voucher households reach lower-poverty neighborhoods than public housing residents on average. But demand subsidies are partly captured in higher rents where housing supply is inelastic, with one 2002 study finding 16 percent rent increases across the 90 largest metros. The honest verdict is conditional: vouchers win where the binding constraint is access to existing units, while supply-side production wins where the binding constraint is the absence of units. Markets differ, so the answer must differ with them.

Q: Can the government build new public housing?

Effectively no, not with the federal funds that would pay for it. The Faircloth Amendment, added to the United States Housing Act of 1937 by the Quality Housing and Work Responsibility Act of 1998, bars housing authorities from using Capital Fund or Operating Fund allocations to construct units that would raise their totals above the October 1, 1999 baseline, subject to narrow statutory exceptions. Because Congress has appropriated no separate development funding for public housing since the mid-1990s, the funding restriction functions as a bar on net new construction. Building new public housing at scale would require amending the statute, not merely appropriating money. That legal fact explains why every modern expansion proposal is framed around vouchers, tax credits, or conversion instead.

Q: What is the Faircloth cap on public housing?

The Faircloth cap is Section 9(g)(3) of the United States Housing Act of 1937, codified at 42 United States Code 1437g(g)(3)(A), added by the Quality Housing and Work Responsibility Act of 1998 (Public Law 105-276, approved October 21, 1998) and named for Senator Lauch Faircloth. It provides that a public housing agency may not use its Capital Fund or Operating Fund allocations to construct public housing units if the construction would produce a net increase over the number of units it owned, assisted, or operated on October 1, 1999, except as provided in the following subparagraphs. It is a funding restriction with a fixed historical baseline, not a literal ban on owning more units, though with no other federal construction funding available it has operated as a freeze on the national stock since 1999.

Q: How much public housing has been demolished?

The HOPE VI era accounts for the largest share. The Congressional Research Service, drawing on housing department program data, reports that HOPE VI revitalization grantees demolished 89,892 public housing units and demolition-only grantees planned another 57,593, putting the program’s total above 100,000 units. A separate compilation by Case Western Reserve University researchers counted 98,592 demolished units replaced by 55,318 public housing units plus 28,979 other affordable units, a net loss of more than 40,000 public housing units from that program alone. Additional units left the stock through the standing demolition and disposition authority in Section 18 of the 1937 Act. Together with conversions, these removals drove the inventory from about 1.3 million units in 1993 to just under 920,000 in 2023, as the Congressional Research Service reported in its 2023 program overview published after this article’s reference date.

Q: Do vouchers cost less than public housing?

Within the terms the research sets, yes. The Government Accountability Office’s January 2002 report GAO-02-76 found that housing production programs cost more than vouchers for units with the same bedroom count in the same general location, and a 2006 follow-up put the 30-year federal cost gap at 16 to 43 percent. An Office of Management and Budget analysis estimated a net present value of 86,827 dollars per two-bedroom voucher unit against 123,409 dollars for construction modeled on the former Section 8 New Construction program. Precision matters: the finding covers housing production programs broadly, including tax credits and capital grants, not new public housing construction specifically, because the data did not support that narrower comparison. It also measures federal budgetary cost, not the market risks each model leaves unpriced.

Q: Do vouchers give families better neighborhoods than public housing?

On average yes, but by less than early advocates expected. McClure and Johnson found in 2014 that voucher recipients are less likely to live in impoverished, racially segregated neighborhoods than public housing residents, and their 2015 follow-up confirmed lower neighborhood poverty rates for voucher households while noting the program had not been wholly successful in facilitating access to low-poverty areas. The shortfall reflects implementation constraints documented since the early 2000s: tight markets, limited landlord participation, and payment standards below market rents in desirable areas. A randomized experiment in Seattle and King County, published in May 2024, showed that intensive search assistance and landlord recruitment raised moves to high-opportunity areas from 15 to 53 percent, indicating the voucher’s weakness is implementation rather than design.

Q: What is RAD conversion of public housing?

The Rental Assistance Demonstration, authorized by the Consolidated and Further Continuing Appropriations Act of 2012 (Public Law 112-55, approved November 18, 2011), lets housing authorities voluntarily convert public housing assisted under Section 9 of the 1937 Act into long-term project-based Section 8 contracts, either project-based vouchers or project-based rental assistance. The logic is financial: a long-term Section 8 contract produces a bankable revenue stream against which the authority or its partner can borrow to fund the rehabilitation the capital backlog demands. Congress initially capped the demonstration at 60,000 units and provided no new funding, so RAD reclassifies existing subsidy commitments rather than expanding them. Each conversion reduces the public housing inventory by definition, accelerating the long-run shift from government as landlord toward government as payer.

Q: Do vouchers raise rents where public housing does not?

They can, and the mechanism is rent capture in supply-constrained markets. Scott Susin’s 2002 study in the Journal of Public Economics found that vouchers raised rents by 16 percent on average across the 90 largest metropolitan areas, estimating 8.2 billion dollars in added rent paid by unsubsidized low-income households against 5.8 billion dollars in subsidies to recipients. Collinson and Ganong’s 2018 study found that each 1 dollar increase in voucher payment standards raised rents by about 47 cents with minimal quality improvement. Public housing rents are administratively set rather than market set, so the program cannot bid up surrounding rents the same way. This asymmetry is the strongest supply-side argument: where the binding constraint is supply, demand subsidies leak to landlords and production is the honest answer.

Q: What was the Section 8 new construction program, and how did it end?

The original section 8, created by the Housing and Community Development Act of 1974, had three components: new construction, substantial rehabilitation, and existing-housing certificates. The new construction component paid private developers to build rental housing with long-term federal subsidy contracts attached, making it a production program delivered through private owners rather than public authorities. In 1983, Congress repealed the new construction and substantial rehabilitation components and added a voucher component, marking the decisive legislative turn from subsidizing construction to subsidizing tenants. In 1998, the Quality Housing and Work Responsibility Act merged the remaining certificate program into the voucher program. The sequence matters for cost comparisons: when the Government Accountability Office compared production programs with vouchers in 2002, it was comparing the surviving production programs, such as the tax credit and section 202, against the tenant-based model that had replaced section 8 construction.

Q: What was the Urban Revitalization Demonstration?

The Urban Revitalization Demonstration was the original name and funding vehicle for what became HOPE VI. The fiscal year 1993 appropriations act, Public Law 102-389, included 300 million dollars for the demonstration, aimed at the revitalization of severely distressed public housing developments. From 1993 to 1998 the program operated without authorizing legislation, sustained through annual appropriations riders. Formal authorization came with the Quality Housing and Work Responsibility Act of 1998, which added section 24 to the United States Housing Act of 1937, codified at 42 United States Code section 1437v. The demonstration’s approach was to demolish distressed developments and replace them with mixed income communities, and its demolition scale was large: program data indicate roughly 98,592 demolished units replaced with 55,318 public housing units plus 28,979 other affordable units. The common claim that the Housing and Community Development Act of 1992 authorized the program is not supported by the dedicated program history and should not be repeated.

Q: What is Section 18 of the Housing Act of 1937?

Section 18 of the United States Housing Act of 1937, codified at 42 United States Code section 1437p under the heading “Demolition and disposition of public housing,” gives public housing authorities standing legal authority to demolish or dispose of public housing properties with federal approval. It was substantially rewritten by section 531 of the Quality Housing and Work Responsibility Act of 1998, effective October 21, 1998. Where HOPE VI was a time-limited grant program aimed at severely distressed developments, section 18 is a permanent administrative tool: an authority with obsolete, excessively costly, or underused properties applies to the Department of Housing and Urban Development, which reviews each demolition or disposition proposal. Approved demolitions permanently remove units from the public housing inventory, and the Faircloth cap prevents the authority from building replacements, so each section 18 approval is a one-way reduction in the program’s reach. Together with HOPE VI and RAD conversions, it explains the steady multi-decade decline of the stock.

Q: How large is the public housing capital repair backlog?

Abt Associates, in a revised final report prepared for the Department of Housing and Urban Development using 2010 data, estimated the public housing program’s existing capital need at 25.6 billion dollars, about 23,365 dollars per unit. The Department’s 2011 press release rounded the figure to 26 billion dollars in its headline. The same study estimated additional accrual needs of 3.4 billion dollars per year and found that the backlog had actually fallen about 3.4 percent since the prior 1998 study, partly because the stock had shrunk 9 percent in the interim. Congress had directed the study in 2007 and the Department began it in 2008. The figure’s meaning is often misread: it measures the gap between what the buildings need and what Congress appropriated for the Capital Fund over decades, not an inherent defect of public ownership. A privately owned stock with the same funding gap would show the same deterioration; it would simply never be totaled up in a single federal report.

Q: What changes for residents when a development converts through RAD?

Residents keep core protections by design: the program requires that conversion not displace existing households and preserves residents’ rights to remain, with rents continuing to follow the income-based formula familiar from public housing. What changes is the legal and financial scaffolding around them. The property leaves the public housing program and operates under a long-term Section 8 contract, which lets the owner finance rehabilitation that the capital backlog had deferred. Management may shift to a private or authority-affiliated developer partner. For the household the visible differences are typically renovated buildings and new management; the deeper change is that the unit’s subsidy rests on a contract enforceable by lenders rather than on annual appropriations, which is precisely what makes the repairs financeable.

Q: Which model gives a housing authority more predictable annual costs?

Vouchers generally produce the more predictable annual budget, because the subsidy is a defined payment per household that adjusts with local rents and household incomes through formulas the authority administers but does not have to capitalize. Public housing confronts the authority with lumpy capital costs: roofs, boilers, and elevators fail on their own schedule, and the 25.6 billion dollar national backlog measured against 2010 data shows what happens when annual appropriations trail accrual year after year. The Moving to Work demonstration, authorized in 1996, gave some authorities flexibility to blend the two funding streams, implicitly recognizing that neither stream alone matches the cost profile of running a housing program. Predictability favors the payer model; asset stewardship is the landlord model’s burden and, when funded, its justification.

Q: Do housing vouchers work in rural markets with few landlords?

They work poorly where the premise of the program, a functioning private rental market with competing landlords, does not hold. The voucher assumes households can shop among units and landlords; in rural counties with thin rental stock, few participating landlords, and long distances between available units, the search frictions that the mobility literature documents in cities become prohibitive. Payment standards calibrated to thin markets may also sit below the rents needed to induce participation. In such markets the access framing fails because there is little to access, and place-based or production approaches, including the Department of Agriculture’s rural rental programs, fit the constraint better. The binding-constraint test applies with full force: name the local constraint first, then choose the instrument.

Q: Why do some landlords decline to accept housing choice vouchers?

The documented reasons are practical rather than mysterious. Program rents capped by the payment standard may sit below market in desirable neighborhoods, inspections impose compliance costs and delays, administrative paperwork adds burden, and in many jurisdictions landlords may lawfully decline voucher holders because source-of-income protections vary by state and locality. The Creating Moves to Opportunity experiment treated landlord recruitment as a core intervention precisely because participation is the bottleneck: its bundled services, including direct landlord outreach, raised high-opportunity moves from 15 to 53 percent. Policy responses divide between mandates, where jurisdictions have adopted source-of-income protections, and inducements such as signing bonuses or damage funds. Either way, landlord participation is a program input to be produced, not a background condition to be assumed.

Q: What role do low income housing tax credits play in the voucher-versus-construction debate?

The Low Income Housing Tax Credit, created by the Tax Reform Act of 1986, became the country’s principal affordable housing production program during the same era when the Faircloth cap closed the public housing pipeline. It is the supply-side instrument that survived the freeze, financed through the tax code rather than the housing department’s appropriations. In the GAO cost comparisons it sits on the production side of the ledger, costing more per household over 30 years than vouchers. In the rent-capture debate it sits on the supply side of the argument, adding physical units where vouchers only add purchasing power. The credit thus embodts the comparison’s central tension: it is the more expensive instrument per household and, in supply-constrained markets, the more honest one.

Q: Did HOPE VI demolition reduce the affordable housing supply on net?

For public housing units, unambiguously yes. Case Western Reserve University researchers compiling program data counted 98,592 demolished public housing units replaced by 55,318 public housing units plus 28,979 other affordable units, a net loss exceeding 40,000 public housing units and a smaller net loss in total affordable units. Defenders of the program argue the demolished stock was severely distressed and the replacement stock is better quality, mixed-income, and better located, which the unit counts do not capture. Critics answer that the original residents were dispersed and many never returned to the redeveloped sites. Both claims can be true simultaneously: the program reduced the count of deeply subsidized units while improving the physical quality of what replaced them. The Faircloth cap ensured the lost units were never rebuilt elsewhere.

Q: Is the Faircloth cap best understood as a budget rule or a housing policy?

It is a budget rule that functioned as a housing policy, and the distinction explains its power. As written, the provision restricts the use of Capital Fund and Operating Fund allocations; it contains no findings about public housing’s merits, no statement of housing philosophy, and no sunset. As operated, with no alternative federal construction funding appropriated, it determined the direction of American rental assistance for a generation: vouchers, tax credits, and conversion became the only available margins of expansion. That is why both common misdescriptions fail. Calling it a mere funding drought understates the statutory barrier, since appropriations alone cannot overcome it. Calling it a verdict on public housing overstates what Congress wrote, since the text judges nothing. It is procedure with substantive consequences, the quietest kind of policymaking.

Q: Why do some voucher holders never manage to lease a unit?

Because the voucher’s value expires against the frictions of search. Time limits on the search period, typically around 60 to 120 days depending on the authority’s policy, collide with tight markets, landlord refusals, payment standards below asking rents, discrimination, and the logistical costs of searching while working or caring for children. Households with the least flexibility, including those with disabilities or large families needing scarce bedroom counts, face the steepest frictions. The Creating Moves to Opportunity experiment quantified the gap: standard administration moved 15 percent of families to high-opportunity areas while intensive search assistance and landlord recruitment moved 53 percent. The lease-up failure is therefore not a household failure but a program-design signal, marking exactly where the payer model needs the administrative investment its theory assumes.