Housing Action Coalition

What do they do? Housing Action Coalition works with housing practitioners and public decision makers on housing production, project support and implementation education. More

Why this approach interests us

• Helping a viable housing project move from permission to completion can address a bottleneck in increasing supply.
• Additional homes can benefit both new residents and other households affected by crowding, instability and rent pressure.
• The Developer Pathway Program offers a concrete example of technical assistance and education rather than only a broad policy ambition.

Our main reservations

• We cannot reliably isolate the housing an additional donation would cause beyond developers, public agencies and already-funded work.
• The charitable and advocacy entities have different donation routes, and our health calculation captures only part of the potential policy benefit.

What do you get for your dollar? Our central partial-health scenario is about $23.31 million per 10 Bay Area QALYs. The possible reach of housing policy makes this an interesting mechanism, but the current evidence does not establish a highly cost-effective health donation. More

What do they do?

The unit is a legal recipient, not a brand

The official donation form names both EINs and distinguishes educational support from advocacy-arm memberships. For this assessment, “ordinary gift” means an unrestricted gift to the c3 after the recipient is made explicit. It does not mean any payment made through the shared brand is interchangeable. Nor does it mean the recipient has agreed to the particular marginal allocation in our model. We have not submitted a donation or verified a restricted grant arrangement.

The distinction matters because the public work page mixes education, project support, policy activity and candidate endorsements. Our review describes this broader organizational setting without constructing an electoral persuasion model. The c3 may contribute useful research, education or technical analysis to an ecosystem in which other entities make decisions. That possibility is different from attributing the affiliate's entire policy success to a c3 gift. A proper grant assessment would identify who employs or contracts each contributor and how shared costs are allocated.

The latest c3 return identifies the c4 as a related organization. Its related-organization transaction schedule and revenue statement do not provide a clean reconciled account of all inter-entity flows. Consequently, we have not added both expense totals and called that a consolidated charity budget. The fact that two entities cooperate is not itself evidence that adding their reported costs is wrong; the problem is that eliminations and shared-service relationships have not been established. Our c3 boundary is explicit so that this uncertainty cannot disappear inside a broad HAC label.

Technical assistance is one pathway

The Developer Pathway Program describes project-readiness, financing and entitlement assistance over one year. Eligibility covers the nine Bay Area counties and Sacramento County, with a focus on eligible emerging or scaling developers. Applications are currently closed. The $500 standard participant fee, $250 eligible-alumni fee and advertised $12,000 service value mean different things: none is an audited marginal delivery cost. We retain the prior model's $15,000 central advisory-year allowance as an analyst planning input, not a quoted price.

The 2026 cohort page lists 14 named participants. Several biographies describe small infill work; one describes involvement in a 92-unit affordable development. These are not 14 completed program projects, and a participant's prior developments are not outputs of this program. The SF-based participant's low-density focus is particularly relevant when interpreting the favorable 100-home project assumption. We do not infer project location from professional residence or multiply historical career units by program membership.

Our modeled technical route is therefore a hypothetical expansion: additional advice reaches a real project, improves its probability or timing of completion compared with available alternatives, and eventually changes residents' housing. The regional Bay share and SF share remain judgments because the public cohort does not provide a project-by-project location and completion ledger. A named cohort makes follow-up feasible; it does not complete that follow-up for us.

Project support is broader than the new program

HAC's project-support description separates review, endorsement and subsequent support. Expert feedback could reveal a design or approval problem. An endorsement may help a project navigate an uncertain process. But neither service is equivalent to construction. Our model does not count all endorsed homes, and it does not separately add a second general project-support benefit for the same homes already credited through technical assistance or policy implementation.

There are several reasons that successful-looking support may have little marginal effect. A developer may already have experienced advisers. The project may satisfy an approval process that would conclude similarly without HAC. Financing may remain infeasible after approval. A faster project may displace construction from another site rather than increase total housing. Conversely, a modest intervention can be important if a correctable bottleneck would otherwise cause a long delay. We require a counterfactual for that specific bottleneck rather than assigning a universal effectiveness rate to endorsements.

The policy portfolio is real, but its pathways differ

The current policy page spans process reform, building and zoning codes, finance and fees, and preservation/protection/homeownership. These are plausible routes to housing outcomes, not interchangeable units of impact. A disclosure rule may affect transparency; a fee change may affect project feasibility; a code revision may affect layout and safety; preservation may prevent loss without creating a new home. Our numerical housing-production channel cannot be casually reused for every item on that list.

The 2025 impact report describes Bay work alongside Southern California expansion, supported in part by Arnold Ventures. It discusses rezoning, fees, ADUs, single-stair reform and AB 610, and acknowledges that laws have not always unlocked expected production. Its revenue table totals $2,781,459 across c3/c4 and event categories. This is useful evidence of a wider organization and already-funded work; it is not a Bay-only spending total, a causal evaluation or a current grant gap.

Our model gives the broad implementation-education channel a larger marginal allocation than the advisory program. That is a deliberate attempt to assess a systems organization as a systems organization. It still quantifies only one source-anchored housing scale rather than inventing an inventory of successful future laws. The remaining portfolio is evaluated qualitatively and remains a potential source of both additional benefits and risks.

Spending breakdown

Latest three available original returns

The official financial page links original 2022, 2023 and 2024 c3 returns. These are calendar-year filings. They should not be confused with a 2025 impact report, the date a PDF was uploaded, or the date we retrieved it. We read the original financial and program sections; the latest functional-expense, revenue and related-organization tables were also visually checked.

HAC — gross accounting shares

Calendar years; denominator includes netted event expense. Functional subtotal is not another category.

Shares are calculated as category expense divided by the stated annual denominator. Rounding may make displayed shares differ slightly from 100%.

Gross accounting expense adds direct event costs excluded from Part IX to avoid making fundraising events costless. It is not a measure of full societal resources or the cost of a particular additional activity. We do not subtract event sponsorships from the expense boundary or call the event's accounting loss the cost of producing housing. Revenue, costs and social benefits answer different questions.

The latest return's program category combines educational work rather than providing separate audited totals for developer assistance, policy implementation and general education. The three-way allocation in the model is not derived from the program percentage. We use the complete gift in the numerator, so allocating a smaller fraction to quantified work does not remove management or fundraising costs from the hypothetical donor budget.

What changed, and what cannot be inferred?

Reported revenue was $1,079,291 in 2022, $1,325,239 in 2023 and $616,190 in 2024. Year-end net assets were $54,953, $401,350 and negative $89,280. These figures indicate considerable annual variation. They do not establish the September 2026 cash position or whether a new gift would maintain, expand or substitute for work already financed. A deficit can reflect timing, use of prior restricted support, organizational transition or financial stress. We do not select among those explanations without evidence.

The 2024 balance sheet reports $70,441 in cash, $73,502 in assets and $162,782 in liabilities, including $92,824 in deferred revenue and $69,254 in unsecured loans. Deferred revenue is not automatically an overdue bill; restrictions and contractual obligations matter. Conversely, cash is not automatically free for a new donor's preferred program. The publicly available return does not provide a current cash-flow forecast or a prospective marginal spending plan.

There are two filing reconciliation cautions. Part VIII reports $199,932 in related-organization contributions, while Schedule R's transaction table reports no listed transactions or amounts. Also, Part VIII's event gross income is $1,500 higher than the corresponding Schedule G figure, although direct event costs agree at $132,144. These observations are not allegations of misconduct. They identify information that a consolidated financial analysis would need clarified, and explain why we retain the original line values rather than silently repairing the return.

The affiliate and newer revenue are not a substitute for a budget

The c4's original 2024 filing reports $355,195 in functional expense, $382,616 in revenue and $85,849 in net assets. We record these separately as context. They are not added to the c3 gift cost or used to claim c3 control over affiliate work. The combined public brand's cost structure needs inter-entity reconciliation before an ordinary payment to an unspecified HAC recipient could have a single financial interpretation.

The newer impact report's revenue categories include $163,324 c3 membership, $277,830 c3 grants/donations, $348,879 c4 membership, $1,626,830 c4 grants/donations, and two event-sponsorship categories. Most of the named grant/donation revenue is therefore c4, but revenue shares are not spending shares or health-attribution shares. We found no matched 2025 whole-expense or 2026 topic-level budget in the reviewed materials. A claim that current c3 operations cost the combined $2.78 million would be unsupported.

Real resources outside the charity

Homes require land, construction, utilities, public administration and often subsidized financing. Some of those costs would be incurred elsewhere or later in the counterfactual; some are genuine additions. A development loan is a financing mechanism, not automatically a social resource cost equal to its principal. An impact-fee reduction is a transfer unless it changes real service provision or incentives. We disclose the missing net resource account rather than pretending all complements are either free or equal to gross capital spending.

The main result is donor-budget effectiveness for a hypothetical ordinary gift. A separate sensitivity adds $1 million of net additional real resources once. That is an illustration of sensitivity to the boundary, not an estimate of actual construction cost. Donors comparing HAC with direct service should not interpret this result as a fully harmonized societal cost-effectiveness ratio.

Monitoring and information sharing

What the public record can monitor

HAC publishes activity descriptions, a named developer cohort, project-support processes and an impact report. The c3's 2023 return reports 14 endorsed projects and 61 member events. Those are potentially useful activity denominators, but neither measures additional completed housing. The 2024 return instead describes educational outreach and policy-related discussion without a comparable project-completion denominator. We cannot construct a year-on-year causal production series from these differently defined narratives.

For developer assistance, the pivotal unit is a project with a baseline status and a delivery record. A credible monitor would record entry date, site, proposed net units, assistance received, permit stages, financing, construction start, completion and occupancy. It would retain failed and withdrawn projects, not just successful graduates. The public cohort page does not provide that chain. Accordingly, the cost per advisory year, completion difference and health exposure remain judgments.

For policy education, an event or report is an output. The next observation is whether a public agency, practitioner or project actually uses it. A further observation is a changed decision or process. Only after that can we ask about completed housing relative to alternatives. Counting all attendees as influenced decision makers, all decisions as implemented, or all approved homes as additional would compress several uncertain links into a misleadingly concrete number.

Existing laws are not a fresh donor opportunity

The SF Planning implementation page records the Family Zoning Plan's December 2025 adoption and January 12, 2026 effective date. A September gift cannot cause that historical passage. It might help practitioners understand new rules, document obstacles, or improve implementation, but the relevant comparator already includes the law, public staff and funded private work. We therefore use only a small residual fraction of the source-scale housing opportunity, then a separate gift contribution within that residual.

The model does not count every future home made legally possible by the plan. Legal capacity is not a forecast of completed housing. Even a production forecast is not a forecast of donor impact. A building would have to be completed earlier or become genuinely additional because of the extra charitable work, and the health exposure must end when the comparator catches up. These requirements explain why the policy channel's source-scale figure is reduced before health conversion.

AB 610: a material correction to the claimed mechanism

HAC's impact report describes a three-year prohibition on new constraints after Housing Element certification. The enacted Chapter 494 text, published October 13, 2025, instead expressly says the disclosure provision does not prohibit adopting, amending or increasing a constraint. It requires disclosure of anticipated constraints under specified conditions. We rely on the enacted text for the legal mechanism, not the stronger provider description.

Disclosure can still matter. It can make a constraint visible earlier, improve analysis, support public scrutiny or help practitioners plan. But those effects require actors to use the information, and they do not mechanically prevent an unfavorable decision. We assign no additional numerical AB 610 benefit. In particular, we do not count a moratorium as an enforceable implementation result or add a separate AB 610 housing cohort to the Family Zoning scale already used.

This discrepancy is a useful qualitative test of evidence quality. The report supplies a lead to a real law and identifiable collaborators; the law supplies the operative boundary. It does not justify dismissing every provider statement, nor does it justify assuming the provider's account of all other bills is exact. For future updates, adopted text, administrative guidance and project-level implementation evidence should control over sponsorship summaries.

Public technical assistance and safety work already exist

The ABAG AB 130/SB 131 resource page is indexed as providing implementation resources, including an urban-infill CEQA eligibility checklist. Full-page access failed during this review, so we treat the indexed official description as a limited baseline observation, not a complete audit of that service. The Governor's June 30, 2025 signing notice independently establishes that major reforms were already enacted. HAC would need to fill a residual gap rather than receive credit for replacing an imagined absence of public work.

A San José Fire Department October 28, 2025 memorandum describes collaboration on single-exit-stair recommendations involving designers, researchers, testing laboratories and fire agencies. It is evidence of an existing technical process, not a completed local single-stair reform or a HAC-caused safety result. We do not infer the eventual 2026 outcome from the older memorandum. A credible building-code health model would need design-specific fire, evacuation and accessibility risks alongside any added housing.

Causal evidence and transfer

From supply to affordability

The Asquith, Mast and Reed working-paper summary reports lower nearby rents around new large apartment buildings relative to comparison locations, with increased in-migration from low-income areas. It supports a supply mechanism rather than the assumption that all market-rate construction necessarily raises nearby rents. The study is not randomized assignment of HAC support, and the large-building context is not automatically transferable to a small infill advisory cohort.

This distinction matters in both directions. Refusing any citywide benefit because new tenants may be relatively affluent would miss a plausible market pathway. Applying the study's rent change to every Bay renter would greatly overstate its scope. Our spillover term is a small per-net-home equivalent outside the directly affected residents. It is explicitly judgmental and has a zero-spillover sensitivity. It is not the study's rent coefficient relabeled as health.

The SF Controller's October 29, 2025 analysis forecasts 8,041–13,615 net additional homes over 2026–2045 under two scenarios for a pre-final rezoning proposal. Those totals already subtract additional demolitions. The model also projects citywide price changes, but does not quantify all tenant harms. Later amendments and actual market conditions may change the forecast. We use the low/high net-home totals only as scale anchors, not as current observed units or HAC output.

The controller's analysis makes the citywide pathway more concrete: a sufficiently large change in housing supply could matter beyond individual buildings. It does not identify which marginal implementation action is best, the share caused by the c3, or a health coefficient. Our residual and contribution assumptions supply those missing links transparently. They are not made empirical merely because their product multiplies an official forecast.

From affordability to health

The Bentley et al. Australian longitudinal study uses fixed-effects comparisons among 10,047 adults. Entering unaffordable housing was associated with a 1.19-point lower SF-36 mental-component score in lower-to-moderate-income households, with a 95% confidence interval from −1.97 to −0.41. The higher-income group did not show the same association. These are mental-component points, not preference-weighted utility units. We do not convert them mechanically into QALYs.

The result supports attention to who experiences a meaningful affordability change. A new occupant who would otherwise occupy a similarly healthy home may gain convenience or consumption without the health improvement assumed for someone leaving unstable or crowded conditions. Our direct-health parameter therefore covers only a fraction of a person-equivalent per additional home, not every occupant. This is a deliberately explicit transfer judgment, not a prevalence estimate from the Australian study.

The prior research also considered Denary et al., a longitudinal housing-assistance cohort. Its cross-sectional and within-person findings should not be treated as equivalent causal estimates. The within-person comparison did not establish the same clear benefit. This source is inherited from the immediately preceding research bridge; the PubMed page yielded no text on the present reopening. We retain it as contrary/uncertain evidence rather than claim a fresh full-paper review or infer proof of no effect.

No retrieved study measures HAC-attributable utility, and none fixes the .02 direct utility or .001 central spillover utility in our model. Those coefficients are priors. The finite calendar horizons are also priors, not empirical follow-up. This limits confidence more than arithmetic precision can repair. A more compelling quantitative report would connect a defined housing change to a comparable population with observed health outcomes and an appropriate counterfactual.

Why not use supportive-housing effects?

An intervention providing stable housing and intensive services to someone experiencing homelessness differs from enabling market housing production. It changes severity, eligibility, accompanying support and alternatives. Copying a supportive-housing utility or mortality effect into every newly built home would import benefits from a service HAC does not deliver. We do not do that. Similarly, the value of a housing voucher or emergency rent payment cannot be applied to a permitting reform without reconstructing the transmission pathway.

There may be indirect effects on homelessness through the housing market. There may also be health gains from less crowding, improved accessibility or proximity to services. These are reasons to investigate, not reasons to count all such benefits simultaneously. Without a matched cohort and deduplicated exposure, adding separate mental-health, homelessness, air-quality and access benefits can describe the same improvement several times.

Displacement, safety and other signed effects

Net new homes already subtract lost units, but that does not automatically compensate displaced residents for worse conditions or disrupted care. A demolition can create more units while harming particular households. Construction can add noise, air pollution and injury exposure. Fee reductions may affect public infrastructure or service capacity. Building-code changes can create safety tradeoffs even if average housing affordability improves. Our signed stress tests acknowledge these possibilities without alleging measured HAC-caused harms.

The central utility is interpreted as a net annual health change for its counted populations. The shared-harm scenario makes that net effect negative rather than subtracting the same harm twice. A separate donor-specific harm term represents a different pathway that can persist even when implementation has no positive effect. It is a stress test, not a documented injury estimate. Unquantified effects remain listed so readers do not mistake the numerical sum for the whole social value of housing policy.

Decision and revision record

This V2 replaces a program-only framing with a whole-c3-gift assessment that includes a source-anchored citywide implementation pathway. It preserves the prior conditional model for comparison, the first new outputs, the unfavorable central estimate and explicit null/harm cases. It does not establish a verified donor offer or a complete organization expected return. Our current disposition is exploratory research; not a priority health-effectiveness recommendation on the public evidence.

The most important new findings are the three-year c3 cost reconciliation, the distinction between newer mixed-entity revenue and spending, the enacted AB 610 correction, current funded/public baselines and a clearer finite health model. The next revision should be driven by evidence about marginal c3 work, attributable implementation, completion and resident outcomes. It should not be driven by a desire to match another organization's numerical rank.

Sources were retrieved September 11, 2026 unless explicitly marked inherited. Filing years are 2022–2024; the impact report concerns 2025 and does not establish 2026 finances. Original-source links appear beside relevant claims. The companion source and finance ledgers preserve access limitations and exact boundaries. No outreach was performed, so we make no claim about organizational responsiveness, confidential budgets or willingness to accept a specific grant.

The model, assumptions and saved results are available for independent review. Scenario assumptions were specified before the first calculation and were not subsequently adjusted to improve the result.

Qualitative assessment

What is reassuring?

There are publicly accessible original tax filings, a current policy agenda, a named cohort and a described project-support process. These make it possible to identify specific legal and accounting boundaries. The organization also describes limitations in translating laws into production, rather than presenting legislative activity as the entire solution. We regard inspectability as useful, while distinguishing it from demonstrated effectiveness or a validated monitoring system.

The public materials identify collaborators and funded work. That weakens a sole-attribution story but improves our ability to ask the right counterfactual questions. A systems organization should not be penalized merely for working in a coalition; cooperation may be essential. What cannot follow is full credit to each collaborator for the same outcome. Our contribution parameters and overlap rules exist precisely because a coalition can be valuable without each member being independently necessary.

What is most concerning?

The strongest substantive reservation is the absence of an observed marginal implementation-to-occupancy chain attributable to the charitable entity. This affects both modeled pathways. The second is that the health coefficients remain weakly matched to the actual residents and housing changes. The third is legal/financial ambiguity across the brand, made concrete by the AB 610 description and related-organization reconciliation issue. None is repaired by the breadth of the mission or the precision of the calculator.

The favorable case depends on a large advisory project, cheaper full delivery, stronger funding response, much more residual policy leverage, larger health-affected populations and longer policy benefit. These assumptions can be coherent together, but their joint plausibility is not established. The result should prompt a search for evidence of that world, not be presented as the charity's likely cost-effectiveness. A selective presentation of the favorable case would be materially misleading.

What remains outside the quantified model?

We leave unquantified preservation of existing housing, accessibility improvements, specific code-reform safety effects, financing-system changes, better public information, organizational resilience and some distributional consequences. We also exclude non-Bay health, electoral outcomes, direct monetization of rent savings, developer profits and a generic value for economic growth. Some omitted effects may be important; omission is not a claim of zero value.

The model is not uniformly conservative. Some missing harms would reduce benefit, some missing benefits would increase it, and several central assumptions could be optimistic. The correct description is incomplete and uncertain. Calling it a guaranteed lower bound would imply that every omitted effect is favorable and every modeled coefficient is defensible, neither of which has been shown.

What do you get for your dollar?

What the model asks

The new model asks what a hypothetical additional $100,000 ordinary c3 gift might accomplish. It allocates $30,000 to technical assistance, $50,000 to implementation education and $20,000 to general education/infrastructure. These are analyst marginal allocations, not observed spending shares. The entire $100,000 is charged. There is no second overhead markup and no deduction for participant payments or membership revenue.

The general-work share receives no separate quantified benefit. Some enabling infrastructure is implicit in the ability to deliver the two modeled pathways; we do not add another generic organizational multiplier. Unmodeled preservation, code, financing and educational effects could change the result in either direction. Thus “whole gift” describes the cost and portfolio boundary, not complete measurement of all benefits.

Technical pathway

Advisory Bay homes equal the allocated budget divided by full advisory-year cost, multiplied by conditional net project homes, completion-probability improvement, funding additionality, Bay share and a cross-pathway non-overlap factor. In the central case:

($30,000 / $15,000) × 20 × .10 × .50 × .75 × .80 = 1.2 expected net Bay homes.

The .10 is a difference in completion probability conditional on actual additional assistance, not the probability that any supported project ever completes. The .50 asks whether the gift creates assistance beyond existing underwriting and substitutes. These are different counterfactuals. We do not add another generic success discount. The .80 removes possible overlap with homes already credited through the policy channel; it is not a second haircut for project failure.

The budget is a fully loaded allowance inherited from the earlier conditional case, including failed effort. The new ordinary-gift scenario does not assert that two places are available. The .75 Bay share and .25 SF share within Bay are judgments; they are not inferred from headquarters. The favorable 100-home project assumption remains an explicit, unverified stress case.

Policy and citywide pathway

Policy Bay homes equal a source-scale SF net-home forecast multiplied by a residual implementation fraction and a gift contribution within that residual. In the central case:

8,041 × .02 × .02 = 3.2164 expected net SF homes.

The first .02 means that only a small portion of the already-law-enabled scale is still alterable through relevant residual implementation. The second .02 is the joint contribution of the allocated $50,000 c3 work: actual funding response, eligible technical/educational delivery, other actors, implementation and dilution across HAC's own past/future funded cohorts. The allocation is already embedded in the meaning of this contribution; multiplying by .50 again would double-discount it. Conversely, applying this .02 to an entire statewide housing shortage would inflate the scale beyond the source.

The policy term is not an estimate that HAC produced 3.2164 homes last year. It is an explicit scenario for future implementation at one source-anchored scale. No additional AB 610, CEQA, SPUR or Onyx home total is stacked onto it. A multi-organization donor portfolio would need to reconcile shared projects before summing these separate organizations' estimates.

For the reported SF subtotal, the model assigns 100% of policy direct and spillover health to SF beneficiaries, while assigning 25% of advisory Bay health to SF. The policy allocation is a simplifying beneficiary-geography prior, not a conclusion established merely by locating the source homes in SF. Migration and housing-market spillovers can cross city boundaries; a revised beneficiary ledger could redistribute the SF/rest-of-Bay subtotals without changing the currently assumed total Bay health. No coefficient is changed by this disclosure.

Health and time

For each net home, annual direct health is .1 affected person-equivalents multiplied by .02 net utility, or .002 QALYs per year. Annual spillover is .5 additional person-equivalents outside the direct group multiplied by .001 utility, or .0005 QALYs. The spillover term is an equivalent population summary, not a count of identified renters. Repeated people across housing chains must not be counted repeatedly for the same health period.

Advisory health starts in year three and lasts six calendar years, giving 5.03149369 discounted years at 3%. Policy health starts in year ten and lasts five, giving 3.45859642 discounted years. These finite windows are deliberately different from the twenty-year source forecast and from building life. They approximate the timing of the marginal cohort, not every forecast home's occupancy. If work merely accelerates construction, the duration must stop at counterfactual catch-up even if the building remains useful for decades.

The executable integral is exp(−ln(1+r)×start) × [1−exp(−ln(1+r)×duration)] / ln(1+r), with the zero-rate limit equal to duration. Discounted years are applied once. No benefits occur before occupancy, no infinite housing lifetime is included, and no mortality survival model is implied by this utility-only housing bridge.

First central output, preserved

The advisory pathway yields .0150944811 Bay QALYs; policy yields .0278105738. Their sum is .0429050549 Bay QALYs, including .0315841941 SF QALYs and .0113208608 elsewhere in the Bay. Dividing the complete $100,000 donor budget by this gain gives $23,307,277.040449742 per 10 Bay QALYs. These first outputs are preserved; no coefficient was changed to obtain a better rank.

There is no calibrated probability distribution. The favorable outcome does not receive an invented weight merely to produce an expected-value number. Tail concentration and “remove favorable world” weighted diagnostics are therefore not applicable; removing the favorable scenario leaves the central result unchanged. This is less informative than a defensible probability model, but more honest than unsupported precision.

Thresholds and prior-model comparison

A $100,000 gift must produce one QALY to reach $1 million per 10 QALYs, or ten QALYs to reach $100,000. The central scenario is about 23.3 times short of the first threshold. At central health per policy home, closing that gap would require roughly 114 additional policy-attributed homes while retaining the advisory result, far above the central 3.2164. This is a diagnostic, not a target used to tune the model.

The previous accepted Developer Pathway model estimated $15 million per 10 SF QALYs for one hypothetical $15,000 advisory year. It remains unchanged in the companion prior-model file. The new $23.31 million result answers a different question: an ordinary whole c3 gift, Bay-wide advisory allocation, overlapping pathways removed, explicit calendar timing, a policy channel and a fully costed unquantified share. The change is not evidence that HAC's measured performance worsened between reviews.

The older favorable $80,000 advisory-only result and new approximately $85,460 portfolio result are numerically close by coincidence of assumptions, not independent validation. Both rely on unverified strong output and health transfer. No broad historical advocacy model is added back to the new estimate. The report preserves unfavorable conclusions instead of choosing the scope with the lowest price.

Model input appendix

These are the locked numerical inputs, not measured HAC effects. The sole external numerical scale is the Controller forecast; even that is a historical forecast, not a current donor output. Null and harm scenarios override selected inputs as described in the executable model. All costs are USD.

The advisory and implementation channels cover mutually adjusted housing cohorts. The .80 advisory non-overlap factor removes potential duplicate homes, while the spillover equivalent excludes residents already counted in direct health. Neither adjustment removes additional costs. These two overlap issues are distinct from funding additionality and completion probability.

Policy-scale households are not multiplied by every year of the twenty-year forecast. Each scenario instead assigns one finite marginal completion cohort and one explicit health window; future information could justify a distributed completion schedule rather than this simplified timing. The present schedule is a modeling judgment, not a reproduction of the Controller’s annual housing trajectory.

Funding and previous grants

What could an ordinary gift change?

A charitable gift could support staff continuity, additional technical work, better implementation materials or organizational infrastructure. Those possibilities are not mutually exclusive, and unrestricted funding need not purchase a discrete visible slot to be useful. The model is allowed to explore such a response. But the public record does not establish which response would occur, how much current work is already funded, or how other donors would react.

The strongest immediate funding evidence is negative or ambiguous: the advisory pilot has an identified supporter and closed applications; the c3's historical deficit is not a current cash forecast; the newer report mixes entity revenues and identifies support for expansion outside the Bay. None proves there is no funding room. Together they rule out describing our advisory allowance as an available additional place or the combined revenue total as an unmet charitable budget.

A donor need not require a provider to promise a precise QALY return. More useful questions are operational. Which c3 work would otherwise be delayed, reduced or cancelled? What staff time or contractor capacity does the next tranche support? Which projects or public processes will use it? What financing and advice are already available? How will the organization distinguish changed timing from projects that would have completed anyway? These questions are closer to the uncertain causal inputs than a generic statement that every donation helps.

What information would change the decision?

The highest-value document would be a current c3 budget and marginal work plan reconciled to the c4 and restricted grants. It should identify available unrestricted cash, committed costs, proposed expansion and already-funded deliverables. A second useful document would be a project/implementation ledger with baseline stages, interventions, dates, outcomes and failures. Independent agency confirmation or contemporaneous project records would strengthen it more than another retrospective endorsement count.

For developer assistance, the key test is whether advice changes a bottleneck that matters for occupancy, rather than simply improving presentation. For policy education, it is whether a particular residual administrative or implementation choice changes because of the work. For the health bridge, the key test is whether the resulting housing materially improves conditions for residents compared with their alternatives. These need not all be known precisely before giving, but a strong health recommendation needs more evidence than is presently public.

Reasonable alternatives

A donor whose objective is measurable near-term health could favor a more direct access, treatment or equipment opportunity with a clearer delivery denominator and a verified marginal constraint. That is a comparative judgment about evidence and the central scenario, not a claim that housing systems never have high returns. A donor specifically seeking housing-policy change might accept greater uncertainty, especially if they can evaluate the legal recipient and work plan directly.

Supporting a public technical-assistance effort, another housing organization or an actual project are also counterfactuals, not automatically inferior options. Existing public guidance may already address the relevant need; a different nonprofit may have a more targeted implementation role; direct capital may address financing but require far more resources. We have not priced an alternative grant. The appropriate comparison depends on the bottleneck, not on whose website describes the broadest ambition.