4% LIHTC with Tax-Exempt Bonds: Deal Structure, the 25% Test, and Modeling

Key Takeaways

How 4% Bond Deals Work

The 4% Low-Income Housing Tax Credit paired with tax-exempt bonds is a common financing structure for large-scale affordable housing production in the United States. Unlike the competitively allocated 9% credit, 4% LIHTC credits do not require a competitive credit allocation when a building meets the bond financing test under IRC Section 42(h)(4). That makes them the faster, more scalable path for developers who can assemble the rest of the capital stack.

In a bond deal, a state or local government entity, typically the state housing finance agency (HFA) or a local housing authority, issues tax-exempt private activity bonds on behalf of a developer. The bond proceeds finance the project's construction or acquisition. When the bonds finance enough of the building's aggregate basis (at least 50%, or at least 25% with the 5% post-2025 condition described below), the building's credit is not subject to the state housing credit ceiling and covers 100% of its eligible basis. The developer sells those credits to an investor for equity.

The critical difference from 9% deals is the capital stack. Because 4% credits are worth about 30% of qualified basis in present value (40% in undiscounted annual credits over ten years), they produce roughly 37% of eligible basis in equity at $0.92 pricing (0.04 × 10 × $0.92), versus roughly 83% for a 9% credit at the same pricing (0.09 × 10 × $0.92). The remaining gap must be filled by the bonds themselves (which become permanent debt, or are replaced by permanent debt at conversion), soft debt from state and local programs, deferred developer fee, HOME funds, and other subordinate sources. How these pieces interact, and how the bond tranche drives the whole structure, is the underwriting problem.

The Financed-By Test (50% or 25%)

The financed-by test, at IRC §42(h)(4)(B), is the gateway to 4% credits outside the state housing credit ceiling. If the building passes, its credit on 100% of eligible basis is not subject to the ceiling. If it fails, the 100%-of-building exemption is lost. Credit attributable to bond-financed eligible basis still falls outside the ceiling under §42(h)(4)(A), and any other credit needs a ceiling allocation. A building with tax-exempt bonds outstanding after placement in service is generally treated as federally subsidized (§42(i)(2)(A)), so the 9% minimum rate in §42(b)(2) is available only if the taxpayer elects to exclude the bond proceeds from eligible basis under §42(i)(2)(B).

Practitioners structure deals so the test is met. The example in Treas. Reg. §1.42-1T(f)(1)(iii) (written for the former 70% threshold) shows credit attributable to bond-financed eligible basis still falling outside the ceiling below the threshold. Graff and Langan (Journal of Affordable Housing, vol. 32, no. 3, 2024) note that the remaining basis could in principle receive an agency allocation, but report never having seen that done under the 50% test. Treat the test as binary for underwriting purposes. The same legislation raised the state housing credit ceiling by 12% beginning with calendar year 2026 (§42(h)(3)(I)), which bears on any 4% versus 9% comparison.

The formula

The test compares the bond proceeds used to finance a building's basis to the aggregate basis of the building and the land on which it is located:

Tax-exempt bond proceeds financing building basis ÷ Aggregate basis ≥ 50%, or ≥ 25% with the 5% new-bond condition

The 25% alternative applies to buildings placed in service in taxable years beginning after December 31, 2025, subject to the 5% new-bond condition described below. Before OBBBA, the threshold was 50%. The addition is a major structural change to 4% LIHTC deal structuring. The test is applied building by building.

What counts as aggregate basis

Aggregate basis is the entire depreciable basis of the buildings plus the cost of the land they sit on. In practice it is close to total development cost, but not identical. Items excluded include syndication and partnership costs, certain reserves not tied to the depreciable property, working capital, and other non-basis financing costs. Bond counsel will opine on the specific inclusions and exclusions for each project. In this article's example, aggregate basis is $47M of $50M TDC (94%). Do not use TDC as a shortcut in the test itself.

What counts as bond-financed

The numerator is the portion of bond proceeds actually used to fund aggregate basis costs, not the total face amount of bonds outstanding. Bonds sitting in a project account that are never spent on basis do not count. Bonds spent on non-basis uses (issuance costs, permanent loan fees, non-basis reserves, working capital) do not count either, and issuance costs above 2% of proceeds separately threaten the bonds' qualified status under §147(g). Do not use the outstanding bond amount as a proxy: bond counsel will look at proceeds used to finance basis when opining on the test.

The test also has a timing dimension, and the statute does not state a measurement date. IRS rulings, as summarized by Graff and Langan, treat the test as met at placement in service or, per PLR 201049018, at any point between placement in service and the end of the first year of the credit period. These rulings concern the 50% test under the pre-2026 statute; they are private letter rulings, which only the requesting taxpayer may rely on, and no authority loaded for this article applies them to the 25% alternative. Confirm timing with bond counsel. The bonds (or the bond loan) should stay outstanding until the test is satisfied for every building: redeeming or repaying after placement in service and after the test is met does not in itself cause failure (PLR 200147010, PLR 201049018), but repaying earlier can. Do not partially repay a multi-building bond loan before all buildings are placed in service.

Why the 25% change matters

The 25% alternative has three real consequences for 4% deal structuring:

1. Volume cap relief on cap-constrained deals. States allocate private activity bond volume cap from a finite annual pool. Where the 50% test forced projects to consume cap up to their debt capacity ceiling, the 25% alternative lets projects size bonds to NOI capacity instead of the test minimum. On low-NOI deals in weak markets that were cap-consumers, the alternative can meaningfully lower cap consumption per project. On high-NOI deals in strong markets, cap consumption barely changes when NOI-supportable debt already exceeded the old 50% requirement.
2. More flexible capital stacks. Developers can now blend a smaller tax-exempt bond tranche with taxable permanent debt (Freddie Mac conventional, agency, or bank). The tradeoff is that taxable debt sits at a higher rate than tax-exempt, so the blended cost of capital rises. In the typical yield curve environment the substitution effect offsets a meaningful share of the volume cap benefit. Do not model the savings from moving to 25% without pricing the taxable replacement debt.
3. Higher fixed costs per bond dollar on smaller issuances. Counsel, trustee, underwriter, and rating agency fees do not scale down in proportion to bond size, so a smaller bond tranche pushes the fixed-cost-per-dollar-of-proceeds up, not down. This makes very small deals harder to justify economically, not easier. Very small deals still struggle to spread the fixed cost.

The 5% New-Bond Rule

OBBBA does not simply drop the threshold. It adds a 25% alternative with a condition specific to the transition. Under §42(h)(4)(B), as amended by Pub. L. 119-21 section 70422(b), a building clears the test if either:

1. Bonds described in §42(h)(4)(A) finance 50% or more of the aggregate basis of the building and the land on which it is located; OR
2. Those bonds finance 25% or more of that aggregate basis, AND one or more of the bonds are part of an issue with an issue date after December 31, 2025 and provide the financing for not less than 5% of that aggregate basis (the new-bond condition).

The amendment applies to buildings placed in service in taxable years beginning after December 31, 2025. Pre-2026 issues count toward the 25% but cannot supply the 5%, and the 5% is measured by issue date, not draw date. The 50% prong remains available: a building whose bonds were all issued before 2026 can still qualify if those bonds finance at least 50% of aggregate basis, while a building with, for example, 40% of its basis financed by pre-2026 bonds and no post-2025 issue fails both prongs.

This is a modeling trap on any deal structuring across the 2025-2026 transition. If a developer sizes the 2026 tranche at exactly 5% of estimated aggregate basis, and construction cost inflation grows actual basis by 1% between application and cost certification, the tranche falls to 4.95% of actual basis and no longer satisfies the 5% requirement. The remedy is a supplemental issuance, which carries its own transaction costs. As an illustrative cushion, size the post-2025 tranche at 10% of estimated aggregate basis, with combined issuance at 30% or more, to leave room for cost drift.

State Agency Implementation

The federal test sets the statutory minimum for the §42(h)(4) exemption. It does not set how much private activity bond volume cap a state will allocate to a project. Each state housing finance agency or issuer applies its own allocation policy, and the number can be framed as a share of aggregate basis, as supportable permanent debt, or as a greater-of rule combining the two. Whether it operates as a floor, a ceiling, or a sizing rule differs by state, so a pro forma needs both the federal test and the state's current QAP and PAB allocation policy.

Two implications for modeling. First, check the direction of the state's number. A minimum the project must reach calls for additional soft sources if NOI does not support the bond amount, while a maximum the project may receive means any bond sizing above it must be funded elsewhere. Second, the state number interacts with NOI capacity. Where NOI-supportable debt exceeds both the federal and state numbers, DSCR binds. Where the state number exceeds NOI-supportable debt, the deal needs additional soft sources or taxable debt to close the gap between what NOI supports and what the state policy calls for.

Bond Volume Cap and Recycling

Private activity bonds draw from a finite state allocation governed by IRC Section 146. Each state receives an annual volume cap based on population, adjusted annually by the IRS for inflation. For calendar year 2026, the cap is the greater of $135 per resident or $397,625,000 (per Rev. Proc. 2025-32, IRB 2025-45). The IRS releases the annual inflation adjustment in the fall of the prior year; check for the current-year procedure before locking assumptions.

The state allocates its cap across all private activity bond uses: multifamily housing, single-family mortgage revenue bonds, manufacturing facilities, airports, and other permitted purposes. Housing competes with other uses for its share. In states where demand for bond cap exceeds supply, volume cap is the binding constraint on 4% production.

How volume cap works

When a state HFA or local issuer authorizes bonds for a LIHTC project, the bond amount counts against the state's annual volume cap. An issuing authority may elect to carry forward unused volume cap for a specified purpose, and the carryforward can be used for bonds issued in the next three calendar years (§146(f)). Each state sets its own process for allocating housing bond cap, and the agency that administers it may differ from the one that runs the QAP for 9% credits.

Bond recycling

Bond recycling is a technique that lets issuers stretch tax-exempt financing capacity further. Under §146(i)(6), if a loan financed by an issue 95% or more of the net proceeds of which finance §142(d) residential rental projects is repaid, and during the 6-month period beginning on the repayment date that repayment is used to provide a new loan for any such project, bonds issued to refinance the issue are treated as a refunding issue, to the extent they do not exceed the principal amount of the bonds refunded. A refunding issue does not draw new volume cap. In a short-term structure, bonds are issued to finance construction; after the financed-by test is met and the loan is repaid with permanent loan proceeds, the repayment can be recycled into another qualified residential rental project.

What recycling does not do is, by itself, put the second project outside the housing credit ceiling. This is the point most easily missed in 4% deal structuring.

The reason is the statutory plumbing. §42(h)(4)(A) removes credit from the ceiling for eligible basis financed by a tax-exempt obligation only if "(i) such obligation is taken into account under section 146, and (ii) principal payments on such financing are applied within a reasonable period to redeem obligations the proceeds of which were used to provide such financing or such financing is refunded as described in section 146(i)(6)." Clause (i) turns on volume cap. Under §146(i)(1), a bond issued to refund another bond is not a private activity bond to the extent it does not exceed the outstanding amount of the refunded bond, so a refunding issue does not draw new volume cap. Clause (ii) ties the exemption to the redemption or refunding of the original financing, which preserves the first project's treatment when its loan is repaid and the repayment is recycled. The second project's exemption still has to come from bonds that are taken into account under §146 and satisfy the financed-by test, which means its own new bond allocation.

The practical value of recycling in a 4% pipeline is the cost of capital. Recycled bonds lower the blended cost of capital by extending tax-exempt interest rates to deals that would otherwise use taxable debt in the same tranche. The size of the saving depends on the spread between tax-exempt and taxable debt, which moves with the yield curve, so use a live quote rather than a fixed spread. Recycling increases NOI-supportable debt on constrained deals and improves cash flow on well-covered ones. What it does not do is multiply the state's effective 4% credit output. That multiplier comes from the 25% alternative itself (each new deal can consume less new cap), not from recycling.

Constraints on recycling under §146(i)(6): the repayment must be used for a new loan to a §142(d) qualified residential rental project within the 6-month period beginning on the repayment date; the refunding issue must be issued not later than 4 years after the original issue date; its latest maturity date may not be later than 34 years after the date the refunded bond was issued; it must be approved under §147(f) before issuance; and the rule applies to only one refunding of the original issue. The repayment date and the bond redemption date can differ. Check the issuer's recycling guidelines for the procedural rules for pairing repaid proceeds with pipeline projects.

Short-Term vs Permanent Bonds

The choice of bond structure affects the capital stack, the financing timeline, and the deal's risk profile. There are two primary approaches.

Short-term construction bonds

Short-term bonds, illustratively 24 to 36 months, are issued to finance construction and count toward the financed-by test. After placed-in-service, and once the test is met for every building, the bonds are redeemed with permanent loan proceeds (a Freddie Mac Tax-Exempt Loan, FHA 221(d)(4) financing, or a conventional bank loan). This structure:

• Enables recycling of the bond proceeds into another qualified project
• Separates the bond (tax-exempt) and permanent debt (either tax-exempt or taxable) markets
• Adds conversion risk: the developer must secure permanent financing before bond maturity
• Requires a construction lender willing to fund alongside the bond issuance

Long-term permanent bonds

Permanent bonds serve as both the construction and permanent financing vehicle. The bonds are issued at closing, draw during construction, and convert to a permanent fixed-rate obligation. Freddie Mac's Optigo tax-exempt loans are one example of a vehicle for permanent tax-exempt bond financing.

• Simpler structure with no conversion event
• Tax-exempt rates are generally lower than comparable taxable rates, though the size of the spread varies with yield curve shape
• Bonds remain outstanding, so cap is not recycled
• Suits developers who can secure Freddie Mac or FHA terms

Sizing the Bond Tranche

Bond sizing in a 4% deal is driven by three constraints that must be satisfied simultaneously:

1. The federal financed-by test. Bonds must finance at least 50% of aggregate basis, or at least 25% with at least 5% of aggregate basis financed by bonds from an issue dated after 2025. This is the statutory test.
2. The state allocation policy. The state's PAB allocation policy sets how much volume cap the project can receive, framed as a share of basis, as supportable debt, or as a greater-of rule. Read whether it operates as a floor or a ceiling.
3. Debt service coverage. The bond amount (if permanent) or the replacement permanent debt (if short-term) must be supportable by the project's net operating income at the lender's required DSCR (the examples in this article use an illustrative 1.20x). This is the maximum.

Where the 50% test required bond amounts above what restricted rents could support, developers had to bridge the mismatch with short-term structures or unusual draw schedules. The 25% alternative eases that tension. When NOI capacity exceeds both the federal and state numbers, DSCR governs sizing and the other tests clear as a byproduct.

The sizing calculation

For a permanent bond deal, the sizing follows this sequence:

Table 1. Bond sizing sequence for a permanent bond deal, with illustrative inputs: NOI of $1,310,000, a 6.0% debt constant (a 5.0% rate on a 35-year amortization is about 6.06%), a 1.20x DSCR, and an illustrative 28% to 30% state sizing rule. Here NOI capacity governs and both the federal test and the state rule clear as a byproduct. Where a state rule exceeds NOI-supportable debt, the gap is closed with soft sources or taxable debt, not a larger bond.

When NOI-supportable debt exceeds the state number, size to NOI. When the state number exceeds NOI-supportable debt, either add soft sources (state HFA soft loan, HOME, deferred fee) to bridge the operating shortfall, fund the excess with taxable debt where the state's policy allows, or reduce the aggregate basis by trimming scope. Do not solve the gap by oversizing the bond above NOI capacity, because you cannot service debt from rents you are not allowed to charge.

Key variables in bond sizing

• Bond interest rate. Tax-exempt permanent rates are set by the lender or issuer based on term, DSCR, LTV, and prevailing benchmark spreads. Use an illustrative 4.5% to 5.5% range for permanent tax-exempt bonds, with short-term construction pricing higher, and replace it with a live quote before locking underwriting assumptions.
• DSCR requirement. A DSCR of 1.15x to 1.25x is a common underwriting assumption, and the examples in this article use an illustrative 1.20x. Higher DSCRs reduce supportable debt.
• Amortization period. Illustratively 35 to 40 years for new construction and 30 to 35 years for rehab. Longer amortization increases supportable loan amount.
• Net operating income. NOI in a LIHTC project is constrained by AMI-based rent limits and operating expense levels. Rents cannot be raised to market. There is limited upside to NOI above initial underwriting.

The 42(m) Determinations

Every bond-financed 4% deal needs two things under §42(m). First, §42(m)(1)(D) conditions §42(h)(4) on the project satisfying the requirements for allocation of a housing credit dollar amount under the qualified allocation plan applicable to the area, which the housing credit agency administers. Second, the governmental unit that issued the bonds (or on whose behalf they were issued) makes a determination, under rules similar to §42(m)(2)(A) and (B), that the credit amount does not exceed what is necessary for the financial feasibility of the project and its viability as a qualified low-income housing project throughout the credit period (§42(m)(2)(D)). Without both, §42(h)(4) does not apply, and the credits count against the state housing credit ceiling.

The feasibility determination is a subsidy-layering review. The reviewer considers the sources and uses of funds and the total financing planned for the project (bonds, soft debt, HOME, deferred developer fee, tax credit equity, GP equity), any proceeds or receipts expected from tax benefits, the percentage of the credit used for project costs other than the cost of intermediaries, and the reasonableness of developmental and operational costs (§42(m)(2)(B)). If equity from tax credits plus other sources exceeds what the project reasonably needs, the credit amount is reduced.

Three practical consequences for modeling:

• The final credit amount is set through the 42(m) review, not by the developer's spreadsheet
• Late-stage changes to costs or other sources can change the outcome at the next review point
• Size the credit request with a small cushion, not to the exact number that closes the gap in the model. The determination limits the credit to the amount necessary for feasibility and viability, not above it

Common Mistakes

The errors that derail 4% bond deal underwriting most frequently:

• Using the old 50% threshold as the only test. Any pro forma built before mid-2025 likely uses the 50% test. For buildings placed in service in taxable years beginning after December 31, 2025, the 25% alternative is available when a post-2025 issue finances at least 5% of aggregate basis. Sizing to 50% when 25% would pass can oversize the bond tranche by up to 2x, distorts the capital stack, and inflates debt service.
• Sizing to the federal test and ignoring the state allocation policy. The federal test is the statutory minimum for the exemption. The state's PAB allocation policy determines how much volume cap the project receives, and it can work as a floor, a ceiling, or a greater-of rule. Model to both the federal test and the state's current policy.
• Confusing aggregate basis with eligible basis, or with TDC. The financed-by test uses aggregate basis (land plus depreciable property, less certain non-basis items). The credit calculation uses eligible basis (excludes land, certain reserves, syndication costs). TDC is broader than either. Using the wrong denominator can overstate the ratio and produce a false pass.
• Ignoring the 2% cap on bond issuance costs. Under IRC §147(g), a private activity bond is not a qualified bond if the issuance costs financed by the issue (counsel, underwriter, trustee, rating agency) exceed 2% of the proceeds of the issue. On a $15M issuance the cap is $300K. Pay issuance costs above 2% from other sources. Related rule: 95% of net proceeds must finance qualified project costs (§142(a)); non-qualified uses (working capital, unrelated fees) must stay under 5%.
• Assuming recycled bonds put the next project outside the credit ceiling. Recycled proceeds provide low-cost financing to the next project, but a refunding issue does not draw new volume cap, and §42(h)(4) turns on bonds taken into account under §146. The next project still needs its own new bond allocation sized to meet the financed-by test.
• Sizing bonds to the test minimum without checking debt capacity. Clearing the federal test and the state policy is necessary but not sufficient. The bond amount must also be supportable by NOI at the required DSCR. A project that clears the threshold but can only support $8M in permanent debt on a $13M bond faces a conversion gap at placed-in-service.
• Forgetting the timing dimension. The statute does not state a measurement date. IRS rulings (private letter rulings concerning the 50% test under the pre-2026 statute) treat the test as met at placement in service or, per PLR 201049018, at any point between placement in service and the end of the first year of the credit period. Confirm timing for the 25% alternative with bond counsel. The bonds should stay outstanding until the test is satisfied for every building. If you plan to recycle or redeem, set the redemption after placed-in-service and after the test is met, and do not partially repay a multi-building bond loan before all buildings are placed in service. Model the draw schedule to confirm the test is met at the right moment.
• Double-counting the bonds as a permanent source. In a recycled bond structure, the bonds are redeemed after placed-in-service. They are not a permanent source of capital. Permanent sources and uses must show the replacement debt (Freddie Mac, FHA, bank loan). The bonds appear only in the construction-period sources and uses.
• Modeling the tax-exempt rate advantage as a fixed spread. The spread between tax-exempt and comparable taxable debt varies with the shape of the yield curve and with the specific benchmark. In flatter-curve environments the advantage compresses. Do not lock a fixed spread into a long-horizon model; use a live benchmark.

How to Model It

A 4% bond deal pro forma is more complex than a 9% model because it must integrate the bond structure, the financed-by test, and the construction-to-permanent conversion. Here is what each tab should contain.

Financed-By Test tab

This is the tab reviewers look at first. Build it as a standalone calculation:

• Numerator: Bond proceeds expended on aggregate basis as of the test date (placed in service, or the end of the first credit year if relying on PLR 201049018). If the full amount is issued at closing, par less issuance costs and non-basis uses is the maximum numerator; count only the portion actually expended on basis by the test date.
• Denominator: Aggregate basis of the building and the land on which it is located, building by building. Include land, all hard costs, all soft costs that are basis-eligible, all basis-eligible financing costs. Exclude syndication costs, working capital, and non-basis reserves.
• Result: Numerator divided by denominator. Must be at least 50%, or at least 25% with the 5% post-2025 condition met, and must also be compared with the state's allocation policy. Show the result as a percentage with two decimal places. Include 50%-pass, 25%-pass, 5%-new-bond-pass, and state-policy flags.
• Cushion: Target a cushion of 2 to 3 percentage points (an illustrative assumption) above the binding threshold to provide margin for cost increases during construction. A result that drops below the threshold after a change order can force a supplemental issuance or credit reduction.

Bond Draw Schedule tab

For recycled bond structures, model the monthly draw and repayment schedule:

• Month-by-month bond draws tied to the construction budget
• Bond proceeds expended on basis as of the placed-in-service test date (and the end of the first credit year, if relying on PLR 201049018); this is what feeds the financed-by test numerator
• Bond interest accrual (construction-period interest, typically funded from bond proceeds or a separate interest reserve)
• Redemption date and amount, set after placed-in-service and after the test is met for every building
• Permanent loan funding date and amount
• Transition tranche split (if applicable). For deals structuring across 2025-2026, model the pre-2026 and post-2025 bond tranches separately so you can confirm the 5% post-2025 requirement is met even after basis inflation.

Sources and Uses: Construction Period vs Permanent

Unlike a 9% deal where sources and uses are typically shown as a single permanent statement, a 4% bond deal needs two versions:

Table 2. Construction vs permanent sources. In a recycled bond deal, the tax-exempt bonds appear as a construction source and are replaced by permanent debt. The remaining sources carry through to the permanent statement.

Credit Calculation tab

The credit calculation for a 4% deal follows the same formula as any LIHTC deal, but with one important distinction: the credit rate is 4% (floored), not 9%. The calculation: eligible basis × basis boost (if QCT/DDA) × applicable fraction = qualified basis × 4% = annual credit × 10 years = total credits × credit pricing= tax credit equity.

Link the tax credit equity output directly to the sources and uses. When credit pricing changes, the equity changes, and the gap (filled by deferred developer fee, additional soft debt, or sponsor equity) should adjust automatically. Also account for deferred developer fee limits: check the state QAP for the cap on deferred fee as a share of the total fee and for the required repayment period.

Operating Pro Forma and Debt Service

The operating pro forma must demonstrate that the project's NOI supports the permanent debt service at the required DSCR. Rents are capped by AMI limits. Do not model rent growth above the HUD-published maximum rents for the applicable income tiers (typically 50% and 60% AMI). Operating expenses should be benchmarked against comparable LIHTC properties in the same market, not market-rate comparables, which carry different expense profiles.

A good 4% bond model passes a simple test. Change the bond interest rate by 50 bps and see whether the supportable debt amount, the sources and uses gap, and the DSCR all update automatically. If any cell requires a manual override, the model is not properly linked.

Related Articles

This article is part of the LIHTC underwriting series. Each article covers a specific aspect of tax credit deal structuring:

• LIHTC 101: 4% vs 9%
• 9% LIHTC Competitive Allocation
• Investor Returns
• The LIHTC Capital Stack
• Year 15 Exit Strategies
• Rent and Income Limits