Scaling That Sticks
What housing policy teaches us about what scale actually requires.
Scalable solutions are the right goal. The ambition to move beyond one-off pilots, to reach more people, replicate what works, and build systems that keep delivering — that ambition is correct and worth pushing hard.
The harder question is what scale actually demands in practice. Most programs treat the solution as the thing that scales: the housing scheme, the subsidy formula, the technology platform. The discipline of solving is what actually scales — the ability to identify the binding constraint in a specific place, choose instruments that fit, build the institutions needed to deliver, and adapt as the system reacts. Get that distinction wrong and you get programs that look scaled but don’t deliver.
Affordable housing makes a useful test case precisely because the world is full of apparently scaled models. Singapore has its Housing and Development Board. Vienna has social housing. Japan deregulated zoning nationally. Chile has incremental housing. Brazil has Minha Casa, Minha Vida. Germany has one of the world’s most extensive rental markets. South Africa has mass subsidized housing. Each contains real lessons. None is a template.
Lant Pritchett and Justin Sandefur (American Economic Review, 2015) give us the analytical framework for why. Their work on external validity shows that imperfect local evidence often outperforms rigorous evidence from another context when the real question is what will work here. Scaling isn’t about transplanting a proven solution. It’s about transferring the discipline of diagnosing constraints, fitting instruments, and adapting when the system pushes back. That’s where most scaling efforts break down — not in the ambition, but in the assumptions about what travels.
One hidden assumption runs through almost every affordable housing case study: it’s about ownership. Across sub-Saharan Africa, 40 to 70% of urban households rent, according to UN-Habitat’s 2016 World Cities Report. The scalable principle for a renter in Mathare is categorically different from the scalable principle for a first-time buyer in Mumbai. Mortgage-based solutions require income documentation and repayment predictability that the bottom two urban income quintiles in most developing cities cannot provide, regardless of subsidy level. That segmentation matters more than almost any other variable in housing diagnostics, and most comparative literature skips past it.
The Invariant Principle
People need secure, well-located, serviced housing at a total cost they can sustain, in a tenure form that matches their income pattern, and the housing system must keep producing that access over time. The instruments that deliver this vary completely by context.
“Affordable housing” is one of the most abused phrases in public policy. A unit can be cheap to build but expensive to live in. A house can be subsidized but badly located. A mortgage can be affordable on paper but brittle when household income fluctuates. A mass housing program can deliver millions of units and still fail if it strands people far from jobs, schools, and transport. The standard question is not which country has the best model. It is what each model is actually solving, under what institutional conditions, and what part of that lesson can travel.
The Cases
Singapore’s public housing model houses close to 80% of the resident population, according to Singapore’s Department of Statistics (77% as of 2023). The surface read is: create a powerful housing agency and build subsidized apartments. Singapore’s actual lesson is more demanding. It combined land control, long-term planning, infrastructure coordination, housing finance, construction capacity, estate management, and political legitimacy inside one capable state apparatus. Countries that copy the product without the system get the shell without the engine.
Japan is the case the affordable housing literature most consistently ignores. Tokyo’s national zoning framework, reformed in the 1990s, delegates substantially less discretionary blocking power to local governments than equivalent systems in the US, UK, or Australia. The result: Tokyo’s capital region (the four prefectures of Tokyo, Kanagawa, Saitama, and Chiba) built nearly 294,000 housing units in 2023, according to Japan’s Ministry of Land, Infrastructure, Transport and Tourism — nearly three times California’s 109,000 units in the same year. The mechanism is direct. When statutory floor-area ratios are capped at 1.33, as Mumbai’s was for decades, the land cost per housing unit runs four to six times higher than in a city permitting ratios of 5 to 8. Alain Bertaud documents this in “Order Without Design” (MIT Press, 2018): regulatory supply constraints, not the absence of public programs, drive unaffordability in most cities. Zoning codes inherited from colonial administrations, designed for low-density European settlement, produce artificial land scarcity in urban centers and force expansion at the periphery. The instrument for fixing this varies — national override powers, inclusionary zoning, infrastructure grants tied to housing approvals — but the diagnosis is consistent.
Vienna teaches a different lesson: preserve affordability as permanent urban infrastructure. About 60% of Vienna’s residents live in subsidized or cost-limited housing (City of Vienna). Many countries treat affordable housing as a short-term supply response — build units, allocate beneficiaries, close the project. If public investment creates affordability today but the city loses it tomorrow through privatization or poor maintenance, the system has produced temporary access, not scale. France’s SRU law makes the same point through a different instrument, requiring municipalities to maintain social housing at 20 to 25% of main residences. Municipalities can exclude affordable housing through zoning and delay regardless of what national governments spend. Obligations on the actors who control land-use outcomes are a precondition for spatial justice, not an optional add-on.
Germany has the lowest homeownership rate in the EU, at about 46.5% in 2022 (Eurostat). Its rental model combines a large private landlord sector, long-term tenancies, and housing allowances for low-income renters. Historically, major German cities remained significantly more affordable than Paris or London. The Mietpreisbremse (rent brake), operative since 2015, is empirically contested: several studies suggest it reduced rental supply in Berlin and may have worsened the 2018 to 2023 price surge. The lesson survives the complication. Well-regulated rental markets with tenant protections and investment frameworks for small-scale private landlords can scale access for irregular-income households that ownership programs structurally exclude. For those households, housing microfinance — step-up lending products and group-guarantee schemes used by BancoSol in Bolivia and housing lenders across East Africa — extends the instrument set further than either mortgage subsidies or rental regulation alone.
Chile’s incremental housing experiments, especially Elemental’s Quinta Monroy, became famous for challenging the idea that social housing must deliver a complete unit. The lesson that circulated was “build half a house.” Elemental’s actual argument is harder: spend public money on the parts households cannot easily provide themselves — location, land regularity, structure, services, and a framework for future expansion. Households are co-producers of housing. Policy should invest in the platform that makes co-production safe, legal, and wealth-building.
Brazil’s Minha Casa, Minha Vida contracted 2.1 million homes and delivered more than 1.76 million between 2023 and 2025 (Brazilian government). That’s scale by any ordinary measure. But mass housing programs everywhere face the same temptation: land is cheaper at the edge, delivery is easier on greenfield sites, and success can be counted in keys handed over. Housing is a place in a city, and the scalable principle is measuring affordability as the household’s total life cost — including transport and access to opportunity — not just the unit price.
South Africa’s post-apartheid program delivered millions of subsidized houses. It also shows why scale alone isn’t sufficient. Subsidized houses built where land was cheap rather than where opportunity was accessible reproduced spatial exclusion while improving shelter. The state delivered a unit. It didn’t deliver urban citizenship. In countries with histories of segregation or unequal land markets, housing policy cannot be separated from urban structure. That has to be a design principle, not a retrofit.
Colombia’s Mi Casa Ya illustrates fiscal fragility. Demand-side subsidies depend on fiscal continuity, mortgage conditions, developer supply, and macroeconomic stability. Official notices have indicated suspension of new applications. In semi-dollarized credit environments, rising mortgage rates can eliminate the affordable segment of the market entirely, because the subsidy value is denominated in local currency while the real cost of credit moves with the dollar. A housing subsidy is a fiscal promise embedded in a financial system. It scales only when the funding, credit, supply, and targeting systems are stable enough to carry it.
What Scaling Actually Requires
Six diagnostic questions, in order. Together they form a practical toolkit for achieving scale that delivers rather than just scale that counts.
What’s the invariant principle? It is not “build lots of cheap houses”. Secure, serviced, well-located housing access at a sustainable cost, in a tenure form that matches the household’s income pattern, produced by a system that keeps supplying it.
What’s the binding local constraint? It differs dramatically: public land, trunk infrastructure, municipal zoning and regulatory supply restrictions (probably the most common bottleneck, per Bertaud), mortgage market depth, rental informality, developer finance, construction productivity, or insecure tenure. Diagnosing the constraint correctly is more than half the work.
What income segment are you designing for? Ownership programs with mortgage components are structurally unreachable for irregular-income households. Programs for the bottom two quintiles need to start from rental access, eviction protection, incremental improvement, and housing microfinance — not a subsidized mortgage they can’t sustain.
What capability must be built? A program that delivers units but leaves behind no stronger institutions has a limited shelf life. The country should emerge better able to plan land, target subsidies, regulate rental markets, finance infrastructure, protect tenants, and coordinate metropolitan growth.
What breaks at scale? Most programs perform better when small, supervised, and staffed by exceptional teams. Scale exposes the weak joints: procurement, land acquisition, beneficiary selection, construction quality, fiscal stress, and local political resistance.
Who is the Searcher? This is the question the framework-building instinct avoids. Bill Easterly’s distinction between Planners and Searchers cuts directly here. Planners design programs from above and measure success in inputs delivered. Searchers find out what works through trial, error, and direct feedback from the people they serve. The five questions above are useful, but they’re structured as if someone benevolent and capable will ask them honestly. The harder question is what institutional arrangement makes it in someone’s interest to ask them and act on the answers — and what happens to the person who detects failure and says so. In most government housing programs, that person faces institutional consequences rather than rewards. Designing the Searcher means designing protection for that person, not just a feedback mechanism. Singapore worked because a small technocratic elite was directly accountable to its population’s housing outcomes. South Africa’s spatial exclusion persisted partly because the institutional reward was units delivered, not access gained.
The Closing Argument
Pritchett and Sandefur’s insight isn’t anti-ambition. It’s anti-false-confidence. The development world has learned to be skeptical of weak causal claims inside a study. It’s slower to be skeptical of the much larger claim that a proven intervention can travel, scale, survive politics, and transform systems.
Affordable housing makes the issue concrete. Achieving real scale means producing secure, serviced, well-located affordability — in ownership or rental form matching the income pattern of actual households — within the political, fiscal, and institutional realities of specific places. The six questions above are a practical toolkit for getting there.
That toolkit is less glamorous than a ready-made solution. It’s also more likely to produce scale that actually sticks.
A thoughtful reflection on growth that is sustainable, meaningful, and rooted in long-term impact rather than speed alone.