The Housing Credit Expansion

Raise each state's allocation of the tax credit that finances subsidised housing by twelve percent permanently, and halve the bond threshold a project must meet to qualify for the smaller credit.

AI evaluation · not yet reviewed by a human

This evaluation was produced and sourced by an AI model; a human review is still pending. Figures and conclusions may still change. The review log is at the foot of the page.How review works →

Almost every subsidised apartment built in the United States since 1986 has been financed by a tax credit that states allocate to developers, who sell it to investors and use the money to build. The tax law of 2025 enlarged it in two ways from January 2026: each state's per-head allocation of the larger credit rose by twelve percent permanently, and the share of a project that must be financed with tax-exempt bonds to qualify for the smaller one fell from half to a quarter, which roughly doubles the number of projects a given volume of bonds can support. Industry estimates put the combined effect at up to 1.2 million additional homes over a decade. Rents in the resulting flats are capped for at least thirty years and tenants must earn below sixty percent of the local median. This evaluation looks thirty years ahead, because that is how long the rent caps last.

Balance

Better for the future · 0.73 previous scale

Balance on the previous scale. The Bilanz 2.0 simulation is not yet available for this evaluation. The category comes from the share of the debate on the pro side (r).

For 28 · 73 % Against 11 · 27 %

Size class: large Scale of this evaluation: Normalised Impact — unitless, calibrated to this topic. For comparison: one point here is worth roughly 2 billion euro per year. The result turns on two things and neither is the size of the credit. The first is the horizon: the credit is claimed over ten years and the rent cap lasts at least thirty, so a thirty-year view counts three decades of cheaper rent against one decade of forgone tax, and a ten-year view does not. The second is that what this measure is worth is a movement of money rather than a stock of housing. Since the sixteenth of September a home built for the open market is priced at what it costs to provide, so the home itself counts for nothing; what counts is that these flats are let below the market, and the gap between the capped rent and the market one is money that stays with the tenant instead of reaching the owner, who is paid for it through the tax that is not collected. That is why the displacement argument, which used to be the largest single figure against the measure at 3.1 billion euro a year, now carries nothing: a home that is not built does not consume what it would have cost. How much of this construction would have happened anyway still matters for what the programme achieves and it no longer moves the result, which now rests on the rent gap, the weight of a euro in these households and the length of the caps [6][11]. How we score →

Arguments for

Arguments against

6 arguments evaluated · Scoring v1.3 Δ absolute +17

Arguments — For

3 arguments

Rent that stays below the market

25of 100

A flat financed this way must be let to households earning below sixty percent of the local median, at a capped rent, for at least thirty years. The gap between that rent and the market one is what the tenant keeps.

Value 5 · Household budgetsImpact 7.8Plausibility 6.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is money that stays in a household because its rent is capped, priced at the middle of the scale like all money. The federal budget forgoes the tax that pays for it, which is the argument standing opposite, and the difference between the two is what a euro is worth in a household spending half its income on rent against what it is worth in the treasury. Half of American renter households now spend more than thirty percent of income on housing and a quarter spend more than half, which is why the weight here is the highest this site uses for household money. Nothing is counted for the security of a thirty-year rent cap as distinct from its level, which is real and is not separately measured. Nothing is counted for the investors who buy the credit, whose margin appears as a cost on the other side. The value is the middle of the scale, and the weight rather than the class is what makes this the largest argument here.

The two changes together are estimated to finance up to 1.2 million additional homes over a decade; 80,000 a year is used, in a range from 40,000 to 120,000, which is the conservative end of that estimate. Rents in these flats run about 5,000 euro a year below the market rent for an equivalent one, in a range from 2,500 to 9,000, with the gap widest in the coastal markets where the shortage is sharpest. Because the rent cap lasts at least thirty years and this evaluation looks thirty years ahead, an average of 15.5 annual cohorts are standing at any moment inside that horizon, which is 1.24 million homes. That gives 6.2 billion euro of rent not paid a year, at a weight of 2.5: the department's tenant data put more than two fifths of these households at or below thirty percent of the local median income and nearly four fifths at or below half of it ^[9], which places most of them in the bottom fifth of the national distribution and nearly all of them in the bottom two fifths; the weight of the bottom fifth is used, which is at the generous end. What is counted is the gap between the market rent and the capped one — money that stays in the household and does not reach the owner, who is compensated by the credit standing opposite — and not the worth of the flat itself, which is what it costs to provide and carries nothing. The horizon is what makes this figure large: on a ten-year view the same programme would show 440,000 homes rather than 1.24 million. The Impact is by far the largest here and it is a function of how long the rent caps last as much as of how many flats are built.

▸ Show calculation ▾ Hide calculation

	Additional homes financed a year Setting, range 40,000 to 120,000: the conservative end of the industry estimate [1]	up to 1.2 million over a decade	80,000 homes a year
×	Cohorts standing at any moment inside a thirty-year horizon the rent cap lasts at least thirty years, so cohorts accumulate across the whole horizon	15.5	1,240,000 homes
×	Rent below the market for an equivalent flat Setting, range 2,500 to 9,000 euro: widest in the coastal markets where the shortage is sharpest [7]	5,000 euro a year	6.2 billion euro a year
×	Weight of a euro in these households more than two fifths of tenant households sit at or below thirty percent of the local median income and nearly four fifths at or below half of it: mostly the bottom fifth of the national distribution, nearly all the bottom two [9]	2.5	15.5 billion euro a year
÷	Normalised Impact scale of this evaluation	2 billion euro a point	7.75

Score 7.75 Impact × 5 Value × 6.5 Plausibility ÷ 10 = 25 of 100

The mechanism is contractual rather than behavioural: a project that takes the credit signs a thirty-year use agreement capping rents, and the Treasury recaptures the credit if it does not. The counterfactual is the tenants in those households renting on the open market. What is estimated rather than fixed is the number of homes, which depends on how many developers use the enlarged allocation and the halved bond threshold, and the estimate used comes from the industry that would build them rather than from an independent source — a reason to sit at the conservative end of it, which is what is done. The confounder that matters is the rent gap itself: in weak markets a capped rent is close to the market rent and the tenant gains little, which is exactly where allocations are easiest to use, so the average gap may be smaller than assumed. That is named and unresolved. Reverse causation does not arise. The Plausibility is above the middle: the rent cap is a contract and the number of flats comes from the industry that would build them.

evidence basis: Precedent · P ceiling 7 identification: Definitional · no rung ceiling

Counterfactual: the same households renting on the open market. Design: definitional for the rent cap — a thirty-year use agreement with recapture if it is breached; the number of homes is a projection from the industry rather than a causal claim. Confounder: in weak markets the capped rent is close to the market rent, and those are the markets where allocations are easiest to use, so the average gap may be smaller than assumed; named and unresolved. Direction: not applicable. Ceiling: precedent 8.5, held at 7.0 because the unit count comes from the industry that would build the units.

Neighbourhoods that improve

2.2of 100

Where subsidised housing is built in a low-income neighbourhood, house prices rise and crime falls; where it is built in a high-income one, prices fall. Most of it is built in the first kind.

Value 6 · Housing supplyImpact 0.6Plausibility 6

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is the effect on everyone else in the streets around a new building: property values, reported crime, and whether a block that was losing people starts gaining them. This site places that with the working order of economic systems and the supply of housing rather than with life and health, because what is measured is a neighbourhood rather than a person. What is priced is the net of the two directions the research finds — an improvement in poor neighbourhoods and a small loss in rich ones — and the net is positive because most of this housing is built in the first kind. Nothing is counted for the tenants themselves, whose gain is the argument above. The value sits in the middle-upper part of the scale, and what is priced is the net of an improvement in poor neighbourhoods and a small loss in rich ones.

Diamond and McQuade compare house prices and crime on the streets around a new development, before and after it opens, against streets a little further away, and find that developments in neighbourhoods below about 26,000 dollars of median income raise surrounding house values by about six percent and lower violent crime, while developments in higher-income neighbourhoods lower values by about two and a half percent ^[3]. Their estimate of the net welfare effect is positive and concentrated in poor areas. Applied to the 1.24 million homes standing inside the horizon, a net neighbourhood benefit of 1,000 euro a year per home is used, in a range from nothing to 4,000 — well below what their capitalisation estimates would give, because a change in house prices is partly a transfer between owners and buyers rather than a gain to anybody. A neighbourhood that improves is a real good and carries no income weight; who lives there is in the value class. That is 1.24 billion euro a year. The Impact is a twelfth of the rent argument and it is the only effect here that reaches people who do not live in the buildings.

▸ Show calculation ▾ Hide calculation

	Homes standing inside the horizon [1]	from the argument above	1,240,000 homes
×	Net neighbourhood benefit per home a year Setting, range 0 to 4,000 euro: well below the capitalisation estimates, because a change in house prices is partly a transfer between owners and buyers [3]	1,000 euro	1.24 billion euro a year
÷	Normalised Impact scale of this evaluation	2 billion euro a point	0.62

Score 0.62 Impact × 6 Value × 6 Plausibility ÷ 10 = 2.2 of 100

The counterfactual is the same streets without the development, which the study constructs by comparing houses close to a new building with houses a little further away in the same neighbourhood, before and after it opens — a difference-in-differences design built around the timing and location of each project rather than a lottery. That design is quasi-experimental and it is the strongest evidence in this evaluation. The confounder that matters is that developers choose sites, and may choose blocks that were about to change anyway; the paper answers it by using the further ring of the same neighbourhood as the comparison, so that a trend common to the area drops out. Reverse causation is the same worry in another form and gets the same answer. The transfer is the weak point: the study covers developments built before this expansion, in a period of lower construction costs and different neighbourhood dynamics, and the conversion from capitalised house prices to an annual welfare figure is a construction of this evaluation rather than of the paper. The Plausibility is at the middle to upper part: the effect is identified around the opening of each development and the annual figure put on it is not the paper's.

evidence basis: Study · P ceiling 7 identification: Quasi-experimental · rung ceiling 8

Counterfactual: the same streets without the development. Design: quasi-experimental — difference-in-differences around the opening of each development, near ring against a further ring of the same neighbourhood (Diamond and McQuade) [3]; not a discontinuity in allocation scores. Confounder: developers choose sites and may choose blocks about to change anyway; addressed by using the further ring of the same neighbourhood as the comparison. Direction: reverse causation is the same worry and gets the same answer. Ceiling: quasi-experimental 8.0 less one point, because the conversion from capitalised house prices to an annual welfare figure is this evaluation's and not the paper's.

A flat you are not evicted from

1.1of 100

A capped rent in a building that must stay affordable for thirty years is the opposite of the private low-rent market, where a household spending half its income on rent moves under threat every few years.

Value 9.5 · HealthImpact 0.3Plausibility 4.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is health: the illness, sleep and mental health effects of housing instability, which are documented in a large literature on eviction and on severe rent burden. It sits just below the top of the scale, in the class this site uses for life and health, because almost none of what is counted is death. It is separate from the money above in the ordinary way: a household that would have paid the higher rent and stayed put gains from that argument alone, and one that would have been forced out gains from this one as well. Nothing is counted for children's schooling, which follows from stability and would enlarge this. Nothing is counted for the households who do not get one of these flats. The value sits just below the maximum: the stream is health, reached through not having to move under threat.

Across the 1.24 million homes standing inside the horizon, the health gain from a secure capped rent against the private low-rent market is put at 0.01 quality-adjusted years per household a year, in a range from 0.002 to 0.04 — about four days of good health, which is a low reading of a literature that finds substantial effects of eviction on mental and physical health. That gives 12,400 quality-adjusted years a year, or 496 million euro. The figure is deliberately small for two reasons: some of these households would have been stably housed anyway, and the counterfactual for the rest is not homelessness but a worse and less secure flat. What is not counted, and would enlarge this considerably, is the effect on the children in these households, for whom the eviction literature finds the largest and longest-lasting consequences. The Impact is a thirtieth of the rent argument, which is what happens when a real effect is spread thinly across a very large number of households.

▸ Show calculation ▾ Hide calculation

	Homes standing inside the horizon [1]	from the first argument	1,240,000 homes
×	Quality-adjusted years gained per household a year Setting, range 0.002 to 0.04: about four days of good health, a low reading because the counterfactual is a worse flat rather than homelessness [8]	0.01	12,400 quality-adjusted years
×	Value of the years the value of a healthy life year used across this site	40,000 euro each	496 million euro a year
÷	Normalised Impact scale of this evaluation	2 billion euro a point	0.25

Score 0.25 Impact × 9.5 Value × 4.5 Plausibility ÷ 10 = 1.1 of 100

The counterfactual is the same households in the private low-rent market. The chain is named — capped rent, no rent-driven move, fewer disrupted schooling and health episodes — and its first two links follow from the use agreement. The health literature on eviction and severe rent burden is large and consistent in direction, but it studies people who were evicted against people who were not, without exogenous variation in who gets evicted, which leaves the obvious confounder standing: households that lose a flat are in difficulty for reasons that also damage health. That is named and unresolved, and it is why the quality-adjusted figure used here is a small fraction of what those studies would imply. Reverse causation is live in the same way and is answered by the same choice. Nothing has measured the health of subsidised tenants against comparable private ones, so what places this number is a complete chain rather than a finding. The Plausibility is below the middle: the direction is well supported, the population studied is not this one, and the size is set low for that reason.

evidence basis: Mechanism · P ceiling 6 identification: Mechanistic · rung ceiling 6 band: Chain closed, unevidenced · P 4–5

Counterfactual: the same households in the private low-rent market. Design: mechanistic — the eviction and rent-burden literature is consistent in direction but compares evicted against non-evicted households without exogenous variation; nothing measures subsidised tenants against comparable private ones. Confounder: households that lose a flat are in difficulty for reasons that independently damage health; named, unresolved, and the reason the figure is a small fraction of what those studies imply. Direction: reverse causation is live and answered by the same choice. Ceiling: mechanistic 6.0 binds. Band: chain closed but unevidenced — links named, confounder answered by choosing a low figure, only the measurement missing.

Nothing measured argues against the claim; what is missing is any comparison of the health of subsidised tenants against comparable households in the private low-rent market. Read back: about half the time, a capped secure rent is worth roughly the four days of good health a year assumed here.

Open: Tenant rosters for these buildings and health records are both administrative and both dated. Following households allocated a flat against those on the same waiting list who were not would give the first clean estimate and could carry this above 6.

Arguments — Against

3 arguments

Tax that is not collected

9.7of 100

The credit is forgone revenue rather than an appropriation, which is why it never appears in a spending debate. Once the expansion is mature it runs to about six and a half billion euro a year on top of what the programme already costs.

Value 5 · Public financesImpact 3.2Plausibility 6

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is federal revenue that does not arrive, priced at the middle of the scale exactly as an appropriation would be; a credit and a cheque are the same money. The whole cost of the expansion stands here, and what it buys stands opposite as rent, neighbourhoods and health rather than as money. Nothing is added for the fact that the cost is invisible in the budget process, which is a real feature of the instrument and not a separate harm. Nothing is deducted for the tax the construction activity generates, which is small and would properly belong with the activity. The value is the middle of the scale, because a credit and a cheque are the same money.

The credits attached to one home come to roughly 110,000 dollars over the ten years in which they are claimed, or about 94,800 euro, at 9,480 euro a year — a setting in a range from 80,000 to 150,000 dollars. The order of magnitude can be checked: the programme as it stood cost about 14.4 billion dollars a year in forgone tax ^[10], and about 3.9 million homes have been placed in service since 1987, roughly 100,000 a year, so some 1 million homes are inside their ten-year claiming period at any moment, which is about 14,000 dollars a home a year across the existing mix; the additional homes here are mostly bond-financed and carry the smaller credit, which is why a lower figure is used. The credits are counted at their nominal value, not discounted, because this site discounts nothing on either side. Applied to 80,000 additional homes a year, and with an average of 8.5 annual cohorts claiming at any moment inside the thirty-year horizon, that is 6.45 billion euro a year of revenue forgone. The asymmetry with the argument opposite is not an accounting error but the point of the instrument: the credit is claimed for ten years and the rent cap lasts thirty, so the benefit accumulates for two decades after the cost has stopped. On a ten-year view instead, the cost would be 4.17 billion euro a year against 5.5 billion of rent relief rather than 15.5, and the measure would come out close to level. The Impact is the largest cost here, and the gap between it and the rent relief opposite is a consequence of the horizon rather than of the size of the subsidy.

▸ Show calculation ▾ Hide calculation

	Credits attached to one home over the ten years they are claimed Setting, range 80,000 to 150,000 dollars: 14.4 billion dollars a year of forgone tax across roughly 1 million homes inside their claiming period is about 14,000 dollars a home a year for the existing mix; the additional homes are mostly bond-financed and carry the smaller credit [6][10]	about 110,000 dollars at 1.16 to the euro, nominal and undiscounted	94,828 euro a home
÷	Claimed over ten years	÷ 10	9,483 euro a home a year
×	Additional homes a year [1]	80,000	759 million euro a cohort-year
×	Cohorts claiming at any moment inside a thirty-year horizon the credit runs for ten years per project, so claiming cohorts overlap for most of the horizon	8.5	6.45 billion euro a year
÷	Normalised Impact scale of this evaluation	2 billion euro a point	3.23

Score 3.23 Impact × 5 Value × 6 Plausibility ÷ 10 = 9.7 of 100

The credit amounts are statutory and the allocation formula is arithmetic: a per-head figure raised by twelve percent, multiplied by state populations. The counterfactual is the previous allocation and the previous fifty percent bond threshold. What is estimated is the number of projects, which is the same figure used opposite and carries the same uncertainty, and the average credit per home, which varies with construction cost and with whether a project uses the larger or the smaller credit. The confounder that would raise the cost is the halved bond threshold, whose effect on volume is genuinely unknown — it is a change in a qualification test rather than in an allocation, so nothing caps how many projects use it beyond the separate bond volume limit. That is named and unresolved. Reverse causation does not arise. The Plausibility is a little above the middle: the credit is statutory and how many projects claim it is the same open question as on the other side.

evidence basis: Projection · P ceiling 6 identification: Mechanistic · rung ceiling 6

Counterfactual: the previous allocation and the previous fifty percent bond threshold. Design: mechanistic — the allocation formula is statutory arithmetic, but how many projects claim the credit, especially under the halved bond test, carries the quantity. Confounder: the bond threshold is a qualification test rather than an allocation, so nothing caps volume beyond the separate bond limit; named and unresolved. Direction: no reverse causation. Ceiling: projection 6.0 binds and mechanistic gives the same.

The cost of the machinery

0.9of 100

A developer does not receive the credit; they sell it to an investor at a discount and pay lawyers, accountants and syndicators to arrange the sale. That part of the subsidy never becomes housing.

Value 5 · Enforcement costImpact 0.3Plausibility 5.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is real resources consumed in converting a tax credit into a building: legal and accounting work on the syndication, the investor's due diligence, the annual compliance monitoring for thirty years, and the state agency's allocation process. It is priced at the middle of the scale like other money spent on running a rule. The discount at which the credit is sold is not counted here, because that is a transfer to the investor rather than a resource consumed, and this site does not book a transfer as a loss without also booking the receipt. Nothing is counted for the developers' own overhead, which a direct subsidy would also require. The value is the middle of the scale, and only the resources actually consumed are counted, not the discount at which the credit changes hands.

The audit and accounting literature on this programme puts the transaction and compliance overhead at something like a tenth of the credit value once syndication, legal work, investor due diligence and thirty years of monitoring are counted, in a range from five to twenty percent. Applied to 6.45 billion euro of credits a year, that is 645 million euro a year of work that produces no housing. The comparison that makes the point is a rent voucher, which delivers the same subsidy to a household with an administrative cost in the low single digits of percent; the credit's overhead is the price of using the tax system to build rather than to pay. What is not counted is the investor's discount, typically ten to fifteen cents on the dollar of credit, which is a transfer rather than a resource loss and would roughly double this figure if it were treated as one. The Impact is a tenth of the public cost, which is the ordinary overhead of delivering a subsidy through the tax code rather than through a payment.

▸ Show calculation ▾ Hide calculation

	Credits claimed a year [6]	from the cost argument	6.45 billion euro a year
×	Share consumed in syndication, legal work and monitoring Setting, range 5 to 20 percent: small rural projects carry proportionally much higher overhead than large urban ones	10 %	645 million euro a year
÷	Normalised Impact scale of this evaluation	2 billion euro a point	0.32

Score 0.32 Impact × 5 Value × 5.5 Plausibility ÷ 10 = 0.9 of 100

The counterfactual is the same subsidy delivered directly. The chain has no behavioural link: a credit that is syndicated incurs syndication costs. What is estimated is the percentage, which comes from audit work and industry reporting rather than from a systematic survey, and which varies with the size of the deal — small rural projects carry proportionally much higher overhead than large urban ones. The confounder that would lower the figure is that some of this work would be required under any subsidy that has to be allocated and monitored, so the comparison with a voucher overstates the excess. That is named and unresolved. Reverse causation does not arise. The Plausibility is a little above the middle: the overhead is documented in audit work and its share varies with deal size in ways the single figure used here does not capture.

evidence basis: Precedent · P ceiling 6 identification: Mechanistic · rung ceiling 6

Counterfactual: the same subsidy delivered directly, as a payment rather than a credit. Design: mechanistic — a syndicated credit incurs syndication costs by construction, and the percentage comes from audit work and industry reporting rather than a systematic survey. Confounder: some of this work would be required under any allocated and monitored subsidy, so the comparison with a voucher overstates the excess; named and unresolved. Direction: no reverse causation. Ceiling: mechanistic 6.0 binds below the precedent ceiling of 8.5; the overhead share varies with deal size and a single figure is used.

Homes that would have been built anyway

0of 100

Subsidised construction competes for the same land, contractors and materials as unsubsidised construction, and a substantial share of what the credit finances displaces homes somebody would have built without it. It no longer carries a weight of its own: a market home is worth what it costs, so the home that is not built takes its cost with it.

Value 5 · Household budgetsImpact 0Plausibility 6

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is homes that do not exist because subsidised ones took their place, and the households who would have lived in them. Since the sixteenth of September a home sold or let at the market price is priced at what it costs to provide, on the same floor space on both sides, which is the price the people who live in it are willing to pay. A displaced home therefore takes its own cost with it: the land, the materials, the capital and the upkeep are not consumed, and the loss and the saving are the same number. What remains of the objection is distributional rather than material — who ends up in the housing that does get built — and that stands opposite as the rent the subsidised tenants do not pay. Nothing is counted for the builders, who build either way. The value question does not arise, because the argument carries no net quantity.

Estimates of how far subsidised construction displaces unsubsidised construction run from about a fifth to almost all of it, with the higher figures in markets where land and labour are the binding constraint rather than finance ^[4]^[5]. Half would be 40,000 homes a year that would have been built anyway and now are not, which across an average of 15.5 cohorts inside the horizon is 620,000 homes. Until September that was valued at 5,000 euro a year each and stood at 3.1 billion euro, the largest single figure on this side. It does not stand any more, and the reason is a change of convention rather than a new finding: a home built for the open market is worth what it costs to build and keep, so the home that is not built saves exactly what it would have been worth. The Accountability Office put the median development cost of a subsidised unit at about 165,000 dollars for the projects it examined ^[6], and at the 6.8 percent a thirty-year loan cost in September 2026 ^[11] that is well over 12,000 euro a year before upkeep — money that is not spent when the home is not built. The displacement is still real and it still matters for what the programme achieves, but what it moves is who lives in the homes rather than how many euro of housing exist, and the tenants' side of that already stands opposite. The Impact is nil by the site's accounting rule, and the argument is kept visible rather than deleted because the displacement itself is not in doubt.

▸ Show calculation ▾ Hide calculation

	Additional homes financed a year [1]	from the first argument	80,000 homes a year
×	Share that displaces homes somebody would have built anyway Setting, range 20 to 90 percent: the higher figures apply where land and labour bind rather than finance [4]	50 %	40,000 homes a year
×	Cohorts standing at any moment inside the horizon	15.5	620,000 homes
×	What each home would have been worth to the people living in it median development cost of a unit at the mortgage rate of September 2026, plus upkeep [6][11]	about 12,000 euro a year, the market price of the access	7.4 billion euro a year
−	What the same homes would have cost to build and keep a home for the open market is worth what it costs: what is not built is not paid for either	the same number, on the same floor space	0 billion euro a year
÷	Normalised Impact scale of this evaluation	2 billion euro a point	0

Score 0 Impact × 5 Value × 6 Plausibility ÷ 10 = 0 of 100

The counterfactual is the same metropolitan markets without the subsidised units, which is what the crowd-out literature constructs from variation in the volume of credits across places and years. Those designs are quasi-experimental in intent — instrumenting subsidised volume with allocation formulas and qualified census tract boundaries — and their results disagree sharply, which is the honest state of the question rather than a defect of any one paper. The confounder that separates them is what constrains construction in a given market: where the binding constraint is finance, a subsidy adds homes, and where it is land and permission, it mostly reallocates them. That is unresolved and it is why the range used here is a factor of four and a half wide. Reverse causation is present in the raw correlation, since credits are allocated to places with visible need, and the instruments are what address it. Because the finding is genuinely contested rather than merely uncertain, the plausibility sits with the argument opposite rather than above it. The Plausibility is a little above the middle: displacement is agreed to happen and its size is one of the more disputed numbers in housing economics.

Excluded (axis = 0): Excluded from the scoring rather than scored small. A home built for the open market is worth what it costs to build and keep — this site prices the two at the same level and the same floor space since the sixteenth of September — so a home that is displaced takes its cost with it: the land, the materials and the capital are not consumed either. What displacement moves is who lives in the housing that gets built, not how much housing exists in value terms, and that difference already stands opposite as the rent these tenants do not pay.

evidence basis: Converging studies · P ceiling 6 identification: Quasi-experimental · rung ceiling 8

Counterfactual: the same metropolitan markets without the subsidised units. Design: quasi-experimental — subsidised volume instrumented with allocation formulas and qualified census tract boundaries (Eriksen and Rosenthal; Baum-Snow and Marion) [4][5]. Confounder: what constrains construction in a given market — finance, in which case a subsidy adds homes, or land and permission, in which case it reallocates them; unresolved, and the reason the range spans a factor of four and a half. Direction: reverse causation is present in the raw correlation, since credits go to places with visible need, and the instruments address it. Ceiling: set at 6.0, below the quasi-experimental ceiling, because the estimates disagree sharply with one another.

Summary

This comes out clearly positive, and the reason is the calendar rather than the size of the subsidy: the credit is claimed for ten years and the rent cap it buys lasts at least thirty, so three decades of cheaper rent are set against one decade of forgone tax. About 80,000 additional homes a year, let at roughly 5,000 euro below the market rent to households earning under sixty percent of the local median, is what the money buys, and it is that gap — money that stays in the household rather than reaching the owner — that the measure is worth, not the flats themselves, which are worth what they cost to build. The objection that a large part of this construction would have happened anyway is real and is no longer scored: a home that is not built does not consume the land, materials and capital it would have taken, so what displacement moves is who lives in the housing rather than how much of it exists. What remains against the measure is its price — about six and a half billion euro a year of tax that is not collected once the expansion is mature — and the instrument, since about a tenth of the subsidy is consumed converting a tax credit into a building, which a direct payment would not require.

Outlook — effect over time

Better for the future · 0.73 previous scale

Sources

Northmarq: 2025 affordable housing tax changes: understanding the housing credit, bonds and opportunity zones. northmarq.com
Nixon Peabody: Low-income housing and community development tax credits in the 2025 tax law. nixonpeabody.com
Diamond and McQuade, Journal of Political Economy: Who Wants Affordable Housing in Their Backyard? An Equilibrium Analysis of Low-Income Property Development. journals.uchicago.edu
Eriksen and Rosenthal, Journal of Public Economics: Crowd out effects of place-based subsidized rental housing. sciencedirect.com
Baum-Snow and Marion, Journal of Public Economics: The effects of low income housing tax credit developments on neighborhoods. sciencedirect.com
U.S. Government Accountability Office: Low-income housing tax credit: development costs and oversight. gao.gov
Harvard Joint Center for Housing Studies: The State of the Nation's Housing. jchs.harvard.edu
National Low Income Housing Coalition: The Gap: a shortage of affordable homes. nlihc.org
U.S. Department of Housing and Urban Development, HUD User: Understanding Whom the LIHTC Serves: Data on Tenants in LIHTC Units as of December 31, 2017. huduser.gov
Congressional Research Service: An Introduction to the Low-Income Housing Tax Credit (RS22389, updated July 2025). congress.gov
Freddie Mac: Primary Mortgage Market Survey — 30-year fixed-rate mortgage averaged 6.76 percent, 10 September 2026. freddiemac.com

Last reviewed by Claude Opus 5 · September 16, 2026 · 3× AI, not yet reviewed by a human

September 16, 2026AI reviewClaude Opus 5record updated
Buchungsregel wohnbau-kosten/1.0 an den US-Setzungen geprüft: die 5.000 € je Wohnung sind Überschuss über die Bereitstellung (GAO-Median 165.000 $ je Einheit, 6,76 % Zins) — Kostenseite in con-2 und Maßstab beziffert, Ankerfrage ausgewiesen; Zahlen unverändert, r 0,65.
September 14, 2026AI reviewClaude Opus 5re-scored
Gesamtprüfung 08.09.: Diamond/McQuade als DiD korrigiert, w auf realen Gütern entfernt (pro-2 i 1,49→1,24, con-2 4,65→3,1), Summary bei 90 % Verdrängung richtiggestellt, HUD-/CRS-Belege ergänzt; Konto-Umbau offen; r 0,62→0,65.
September 6, 2026AI reviewClaude Opus 5First evaluation
First evaluation over a thirty-year horizon, because the rent cap outlasts the credit; displacement of unsubsidised construction set at half and decides the result.

Evaluations are produced with AI support and reviewed on a schedule for new developments; human passes are marked separately.How we review →