This week I’m at YIMBYTown, and thought I would create this blog post for other attendees (and those who wish they could be there) to talk about how we evaluate housing grants on the Abundance and Growth Fund. If you know a law worth changing, a law that needs enforcing, or a permitting bottleneck keeping viable projects from getting built, please have a read below and then reach out.
Our current housing model can justify roughly $470 in grant funding per expected additional home credited to us in six supply-constrained metropolitan areas.1 A $1 million grant therefore needs to account for roughly 2,100 such homes, after allowing for uncertainty, what would happen anyway, and the contributions of other advocates and funders.
That’s a demanding benchmark–equivalent to a social benefit cost ratio of 1000x on our grantmaking–but, as I explain below, this bar exists because it reflects the next best opportunity of our grant funding on global growth, public health, & poverty relief. And note that I am laying out our internal process for better transparency and understanding: These are not minimum analysis requirements for you as a grantee, and you can start with a short inquiry through our application page. While more grantee-side data and credible analysis makes a decision easier and faster, the valuation process is internal to CG and you do not need a perfect economic model to apply.
Why our bar is so high
Coefficient Giving’s Global Health & Wellbeing work compares important, neglected, and tractable opportunities across causes by their expected contribution to human welfare, using the conventional tools of welfare economics. Housing competes for funding with CG’s opportunities to backfill cheap, lifesaving aid, including HIV treatment disrupted by foreign-aid cuts. Those are real uses of the money Coefficient Giving could otherwise spend a grant on, instead of a housing grant.2
To compare health and income gains, we use welfare-equivalent dollars, or CG$. One CG$ represents the subjective benefit of giving $1 to someone earning $50,000 a year; a dollar received by someone poorer gets more weight. (In our logarithmic social welfare function, a dollar is worth 10x more to someone consuming $10,000 annually than a dollar to someone consuming $100,000 annually).
Our current bar is roughly CG$ 1,000 of expected value per grant dollar. We call that ratio the grant’s social return on investment, or SROI. This corresponds to a social benefit cost ratio in traditional welfare analysis, but is calculated from the perspective of CG as a grantmaker rather than the federal government.
Rich-country housing can clear a very high bar when a modest advocacy budget changes the rules for thousands of high-value homes in the highest-demand locations. (Ideal grants combine a high value per home permitted, many homes affected, and a modest cost to change their prospects of success.)
It’s worth pointing out that while other funders generally don’t use our bar, a housing grant returning, say, 100x–900x under this framework can still be an extraordinary opportunity for a donor focused on domestic US policy. It can feel silly to issue a grant rejection saying, “I’m sorry: your idea only generates 500x more social value than it costs in expectation”. One non-monetary contribution we like to offer grantees is, when helpful & appropriate, to share our findings with other funders or with an organization’s existing sponsors, on grant proposals that generate excellent social returns without quite being able to crowd out a malaria net or AIDS drugs in the aftermath of USAID’s collapse. If you are a co-funder interested in strong domestic housing grant proposals, please reach out as we would be happy to share promising opportunities (with applicants’ permission).3
A familiar framework with different parameters
Doing housing policy philanthropy with this quantitative frame can seem unusual. However, if you have worked with the federal government’s Circular A-4 guidelines for social benefit cost analysis, much of this should sound familiar: define the counterfactual, estimate a policy’s benefits and costs over time, account for uncertainty, and test which assumptions drive the result in a sensitivity analysis. Our framework differs in its parameters, including how we value health, gains to people at different incomes, and how we treat effects across countries. Unlike a federal regulatory impact analysis—which compares a policy’s total social benefits and costs—we divide the welfare gain attributable to our grant by the amount of the grant itself.
If you perform your own conventional Circular A-4-style analysis on your work, and find a 1000:1 social benefit-cost ratio, submit it. We would still need to translate the assumptions into our framework and work out what our funding would change, but a conventional 1000x BCR calculated using standard federal methods would likely be a very strong opportunity.4
I also want to reassure skeptics of overconfident expertise in matters of housing regulation generally about our quantitative approach. No wonk or expert, no matter how smart or how well-informed, can settle inherently subjective and individual questions about how and where to live.5
Where housing proposals fit
Additional housing production in superstar cities is our core surrogate endpoint: an outcome we can front-load a ton of research to estimate once, and then reuse to predict welfare gains for particular grants without redoing custom calculations for every grant. More homes in the highest-value places let people reach better jobs and improve affordability, with benefits extending beyond the people who move into the new buildings.
Since 2015, our housing valuation has focused on the San Francisco, Los Angeles, Seattle, New York, Boston and Washington, DC, metropolitan areas. We are working to extend it beyond those six. The value of an additional home built varies with local conditions; we should generally expect less value where the Glaeser Wedge (the gap between housing prices and the cost of supplying housing) is smaller, and where patenting rates and global innovation spillovers are lower. But those gaps are widening beyond the classic superstar cities.6 That leaves room for state and regional reforms where action may be more tractable before shortages become as severe.
For a national proposal, we look at where the effects would occur, because different locations have extremely different social marginal values of an incremental home. Superstar housing production is not the only outcome we can value, however. Reducing the real cost of construction can save resources on homes that would have been built anyway. Other welfare-relevant consequences are fair game if you can explain the causal mechanism and why the grant could meet our SROI bar. Our estimates would need to include adverse effects and public resource costs, distinguish real resource savings from transfers, and avoid counting a benefit twice. The broad target is what Chris Elmendorf & I call the “three legs” of the reform stool:
Zoning
Permitting (permit entitlement, utility connections, certificates of occupancy, environmental review, design review, historic district review, & any other form of discretionary review)
Feasibility (fees, exactions, taxes, community benefit agreements, etc)
Building code reform (this can fit into the third leg insofar as lower construction costs boost feasibility of zoned & permitted projects, or you can call it a new “fourth leg” if you want)
You are a plausible fit if your work could materially change welfare-relevant outcomes and additional funding would let you do something valuable that would otherwise go undone. That might mean passing a reform, enforcing it, changing a building code, or producing research that changes a specific and consequential decision. Explain how your work would get more homes built in the highest-value areas or reduce construction costs over wide areas, and what additional funding would let you do.
Fieldbuilding related to research to inform cost-benefit analysis for building code reform and code reform advocacy is of special interest, as it’s the most neglected frontier of YIMBY work in the US right now.
From zoned capacity to homes
A statewide upzoning can put 100,000 units on paper without producing anything like 100,000 apartments. A narrower reform might still ultimately accomplish more social good if it makes construction feasible or gets homes built years earlier, or if it builds institutional and legislative momentum towards a wider reform in the future. To be sure, many different kinds of arbitrary growth control regulations are socially costly, and in principle all bad land use regulations should eventually be reformed, but to do the most good with limited funds we have to rank-order different relief pathways by their tractability and importance.
A theoretically ideal grant application would offer a parcel-level production forecast: identify potential redevelopment “soft sites,” test whether redevelopment pays, and estimate when construction is likely. The completed project’s value, less construction, demolition, financing, fees and a competitive normal developer return, must leave enough to justify replacing the existing use.7 In practice, however, this level of analysis is costly & difficult to achieve inside CG and virtually unheard-of among our housing grantees’ application materials.
Often we have much less information. In past back-of-the-envelope calculations, we have started with zoned capacity and an explicit “guesstimate” of how much will be built. Where possible we give that estimate a time horizon and explain it using comparable reforms, local development rates, and the permitting, design-review, fee and exaction environment. Again: there’s no one universal take-up rate for zoned capacity, and you don’t need a parcel-level feasibility study to apply.8
An example from City of Yes
Our $2 million general-support grant to Open New York helped fund its housing advocacy, including work on City of Yes. The grant supported a broader agenda, but the reform makes a useful public illustration.9
NYC’s Department of City Planning did much of the capacity-to-production work through its Reasonable Worst-Case Development Scenario. Its analysis combined site and zoning characteristics with neighborhood market conditions and likely buildout. For the adopted reform, it projected 82,100 additional homes by 2039 under its high-market scenario.10
Next we assess how much will materialize, how advocacy changed the outcome, and what our funding contributed. Suppose those adjustments leave either 5% or 10% of the city’s projection credited to ONY’s effort and the grant’s share of ONY’s budget:

At today’s $470 per metropolitan superstar home benchmark, the $2 million face amount breaks even at about 4,255 credited homes in expectation, or 5.2% of that forecast.11 This simplified illustration assigns the full general-support grant to one outcome and one grantee, but the full valuation also accounts for construction timing, benefit duration, funding costs and the grant’s other effects on other policies advocated by Open NY.12 With a plausible 5% credit attribution to ONY for the social outcomes of City of Yes alone, ONY easily cleared even this very high 1000x SROI bar.
Using upper and lower bounds to decide what needs more work
We can often learn enough without estimating every input precisely. If a defensible pessimistic estimate of expected return already comfortably clears a social benefit cost ratio of 1,000x generally (or, specifically, $470 per superstar house in our current methodology) we can stop refining the arithmetic and focus on the remaining questions about the grant. If a proposal falls short even with an optimistic estimate, then more decimal places won’t help it get over the line. For any selected grant, we ideally still want to develop a rough best guess of the expected return when practical, both to compare opportunities and to learn from our forecasts. But for grants below a certain threshold, attempting to refine a perfect SROI calibration is an inefficient use of everyone’s time.
One more note: none of this makes us a dumb spreadsheet algorithm. Judgment enters through the evidence, the counterfactual, the team’s ability to execute, the size of the award, and consequences that are hard to model. Writing down our estimates makes it possible to challenge our assumptions and lets us check our forecasts against what happens. We’re not fearful of “outcome-hacking” by applicants through this methodological transparency, either. If you can credibly present a reform forecast that is part of a plausible causal chain to the outcome of a large number of feasible homes in the most valuable above-cost metros, that’s a wonderful thing to submit and very hard to fake.
Why this helps advocates too
It is easy to get lost in the politics, communications and strategy around a symbolic victory. For instance, while still in my Niskanen role, my predecessors at CG asked me how many units each section of the federal ROAD package might possibly unlock. It was a clarifying moment: I had been so focused on the politics and comms, on the need for every Senate Banking member to have a messagable win of some kind in the bill, that I had momentarily forgotten the goal of reform is to unlock housing. Only a half dozen or so of ROAD’s several dozen titles have any plausible claim to unlocking any meaningful number of new homes. The politics and the comms value of the small or orthogonal bill titles were still essential to passing the substantive parts of the overall bill, but it’s important to (thoughtfully) remember what is being done for symbolism or comms and what is actually substantive.
Much of this is already contained in any advocacy strategy. A report assumes someone will read it and change a decision. An organizing campaign assumes a constituency can change the politics. But writing down the probabilities and magnitudes lets us interrogate assumptions that otherwise remain unspoken. It also helps us choose between a large but less tractable reform and a smaller one we can plausibly increase the odds of winning.
Ultimately, causality can still look like a shotgun blast of joint effort where it’s hard or costly to discern which individual contributor is uniquely responsible for each aspect of an outcome. Research, organizing, legislative work and litigation often reinforce one another, for instance, or several approaches may also be worth trying when we do not know which will work.
For a major reform vertical like national single stair reform, we can be more confident in a combined strategy than in identifying the singular decisive memo, meeting or lawsuit. In those cases, we can evaluate several grants under one umbrella: their cost and the expected effect of the package. In that umbrella package, we still judge what additional funding changes, which grants improve the strategy, at what size, and whether the necessary pieces can be funded and executed.
What to send
If you have a promising idea you think we should look at, use the proposal form linked from our application page. A short inquiry should explain:
Policy and place. Name the rule, the jurisdictions affected, and the path to implementation.
Expected benefit. If possible, estimate additional homes, earlier completion, cost reductions or another welfare gain, with a time horizon.
The baseline. Describe what would happen without grant support, who else is working on the issue absent our grant, and the main obstacles.
Your contribution and budget. Explain your role, your partners and the relevant policy ecosystem, other funding, and (if possible) what you would cut with half the award or add with twice as much.
Give a best guess and plausible ranges, with sources and assumptions. If missing evidence is the reason for a research proposal, explain which decision the research would inform. We receive more inquiries than we can investigate and follow up where there appears to be a fit.
And if you’re at YIMBYtown, come find me. Tell me what you want to change, why it would matter, and what funding would let you do. Then send the inquiry so we can follow up after the conference.
Coefficient Giving’s target metros were chosen years ago, when my Open Philanthropy predecessors became the first contemporary YIMBY movement funders in 2015, but they coincide with six of the seven highest-cost metros in Duranton & Puga’s 2023 paper.
For instance, GiveWell’s July 2026 announcement describes Coefficient Giving’s $1 billion commitment and funding gaps caused by aid cuts, including HIV treatment. Coefficient Giving’s Health Aid Transition Fund announcement describes bridge financing and support for health systems, including countries in sub-Saharan Africa.
One interesting open question: Our screening framework of “importance, neglectedness, and tractability” is in at-least-superficial tension with our social return maximizing goals. Should our program officers hand off our very best grantees to other funders, since the most extraordinary grants in our portfolio should be easy to persuade other funders to take on? Perhaps our maximum counterfactual impact would actually be to give all our best grant ideas to other foundations, and retain only the riskiest marginal grants above our bar subject to the constraint that they’re not so obviously good enough to hand off to co-funders.
N.b. all of the above is just a bonus if you do have the information available. You do not need to learn full social benefit cost analysis or even perfectly forecast your housing outcomes before telling us about an unusually valuable opportunity. This is just to explain how to think about how to make an informed guess about whether you’d be a good fit for an AGF housing grant (though of course if you have even some introductory part of a credible analysis of your proposal ready, that certainly doesn’t hurt your chances).
Do people want to live in big cities or small cities, car-oriented suburbs or transit-oriented towers? The post-1916 attempt to settle questions like that by political means from on high by so-called “experts”, through ever more stringent growth control regulations–rather than allowing free people to express their preferences in open and competitive land & structure markets how they want to live–is a key driver of America’s urban housing shortage. Our quantification approach seeks to identify the most cost-effective ways to accommodate latent demand for locations and lifestyles that regulation currently most strictly binds, but it is downstream of observed revealed preferences rather than a prescriptive claim about how Americans ought to live. We can infer from current and shadow prices in land, labor, and structure markets in high-demand metros that millions more Americans would choose to live in the highest-demand locations with good jobs, good infrastructure, & good weather if land use regulations allowed it.
Interregional mobility connects housing markets in “spatial equilibrium”. New homes in productive, desirable places can improve affordability elsewhere as movers vacate homes in other regions, which means the welfare effects are not confined to the city issuing the permits. Extending the model requires estimating the marginal social value of housing in each place, rather than applying the six-metro benchmark nationwide.
For a detailed redevelopment forecast, one would estimate each site’s probability of building within the horizon times its net added homes, separately with and without reform, then sum the difference. A rough capacity conversion approximates that calculation; state whether capacity is gross or net and avoid subtracting replacements or baseline production twice. Vincent Rollet’s “Zoning and the Dynamics of Urban Redevelopment” explains why existing buildings, local prices and the option to wait affect take-up. Implementation is part of our forecasting effort as well. A change to the International Building Code (a model code) needs adoption at the state and sometimes local level; or a state zoning preemption law may need enforcement before cities issue permits. A good grant application will explain where your organization can change the sequence. Timing matters too: a home built five years earlier supplies five additional home-years before the baseline catches up, and we put nonzero weight on nonsubsidy reforms that accelerate production even when homes would eventually have been built in some long-run steady state.
The data is public: $1,966,904 in general support to ONY in 2023 and $33,096 in 2024, totaling $2 million. Coefficient Giving’s 2024 grantee review describes ONY’s City of Yes advocacy.
DCP’s Final Scope of Work (April 26, 2024), pp. 88–89, describes market-based buildout rates checked against observed production. Revised Technical Memorandum 002 (December 3, 2024), pp. 34 and 49, gives 82,100 additional homes for the amended plan’s high-market scenario through 2039.
Again, these 5% and 10% shares are hypothetical, not our original credit estimates for the actual grant, and the resulting home credits are illustrative projections.
Our published methodology allows estimated credit shares as a simple “equity share” of a reform’s average total social value, or as a more complicated direct comparison of counterfactually modeled outcomes with and without the particular grant.


My idea is less of a legal or regulatory reform and more one of partnership opportunity. I manage housing and community development programs for the National Association of Realtors and sit on the advocacy team in DC. Part of our programs include modest grants available to state and local Realtor associations on topics related to housing opportunity, fair housing, smart growth, placemaking and rural outreach. These grants are intended to support partnership and collaboration with local housing and community development nonprofits and government agencies. Please reach out if you want more information or a connection to your local Realtor group to start conversations about how this funding can help spur local impact. lbenner@nar.realtor