This paper explains, in plain terms, how Lintel produces forward projections of HUD's annual income limits and the LIHTC rent ceilings that follow from them. It is written for developers, lenders, underwriters, asset managers, and anyone whose work depends on these numbers — not just to describe what we do, but to give you enough detail to defend the projection in front of a credit committee or investment partner.
1. Summary
Every year, the U.S. Department of Housing and Urban Development (HUD) publishes income limits that determine who qualifies for federally subsidized housing — and, through Section 42 of the tax code, the maximum rents that LIHTC (Low-Income Housing Tax Credit) properties can charge. The published limits arrive in April. Underwriting decisions happen all year.
Lintel projects, today, what HUD will publish next April. We use the same Census Bureau survey data HUD will use, apply the same Congressional Budget Office wage forecast HUD will apply, and run the calculation HUD documents in its annual methodology notice. The result is a defensible forward estimate that lets you model a deal closing six months from now without waiting for the official release.
Two things distinguish a Lintel report from a manual spreadsheet approximation. First, every projection ships with a 95% confidence interval derived from the published sampling error in the underlying Census data and historical variance in CBO forecasts. Second, every projection includes a multi-year cap-overhang trajectory showing how rents in cap-bound markets will likely evolve over the next 5–10 years as deferred upside releases.
2. Why projecting income limits matters
A LIHTC deal stands or falls on three numbers: the income limit, the rent ceiling derived from it, and the tax credit equity raised against the projected revenue. Get the income limit wrong and the rent ceiling is wrong; get the rent ceiling wrong and the equity raise is wrong; get the equity raise wrong and the project doesn't pencil.
Historically, underwriters have approached this projection in one of three ways: wait for the official April release (conservative but impractical — deals close on quarterly cycles, not HUD's annual one), build a spreadsheet model (accurate if maintained by a specialist, but brittle when HUD adjusts methodology), or buy a published projection (reliable but historically expensive at $500+ per county per year).
Lintel productizes the projection: every U.S. county, every fiscal year, accessible at $149 per report or $299 per month for unlimited access. The underlying calculation is exactly what HUD does — there is no proprietary model. The product is the engineering: keeping the calculation faithful to HUD's evolving methodology, refreshing the inputs as new data publishes, and packaging the result so an underwriter can use it without reverse-engineering a methodology document.
3. The data sources
Three external sources feed every projection. None of them are proprietary — they are all freely-available federal government publications. What Lintel does is integrate them faithfully and refresh them as new vintages release.
American Community Survey (ACS)
The U.S. Census Bureau publishes annual estimates of median family income for every county and metropolitan area. ACS comes in two vintages: a 1-year estimate based on the most recent year's survey responses (more current but with wider sampling error in small areas), and a 5-year estimate that pools five consecutive years (more precise but lags reality). HUD prefers the 1-year when it's statistically reliable for the area, and falls back to the 5-year otherwise. Lintel applies the same fallback rule.
Fair Market Rents (FMR)
HUD also publishes its own annual Fair Market Rent dataset — the 40th-percentile gross rent for standard-quality units in each FMR area, by bedroom size. FMRs matter to income limits because of a HUD rule called the high housing cost adjustment: in expensive markets where the standard 50%-of-median-income limit would be too low for a family to actually afford an apartment at the FMR, HUD raises the income limit to ensure affordability.
CBO Economic Projections
ACS data lags reality by a year and a half — the 2024 1-year ACS doesn't publish until September 2025. To produce an income limit for fiscal year 2026 (which HUD publishes in April 2026), HUD has to project the 2024 median income forward to a 2026 figure. They do this using the Congressional Budget Office's forecast of national per-capita wages and salaries, published in CBO's biannual Budget and Economic Outlook (typically released in January/February and August). Lintel uses the most recent CBO release available.
4. The calculation, at a glance
Once inputs are sourced, the calculation is mechanical — seven sequential steps that take you from a raw ACS estimate to a final 4-person Very Low Income Limit. The diagram below shows the full flow. The two amber-highlighted steps — high housing cost adjustment and the year-over-year cap — are the adjustments most likely to materially change the result.
The detailed sections below walk through each step in plain language.
Step 1 — Pick the right ACS estimate
For each area, take the most recent 1-year ACS median family income if it's statistically reliable; otherwise fall back to the 5-year ACS. HUD's reliability test: the published margin of error must be less than 50% of the estimate, and the estimate must rest on at least 100 observations. If neither vintage qualifies, HUD averages the three most recent reliable 5-year estimates. If that still fails, the area gets the state non-metropolitan median.
Step 2 — Inflate to the fiscal year
Multiply the ACS estimate by the CBO trend factor. The trend factor is defined as CBO-projected per-capita wages and salaries in the fiscal year, divided by per-capita wages and salaries in the ACS survey year. For fiscal year 2026, HUD used $48,641 / $46,125 = 1.0546. For fiscal year 2027, the analogous factor from the CBO's February 2026 projection is $50,374 / $47,206 = 1.0671. The result is rounded to the nearest $100. This is the projected Area Median Family Income (MFI) for the fiscal year.
Step 3 — Compute the preliminary Very Low Income Limit
Multiply the projected MFI by 50%. The result is the preliminary 4-person Very Low Income Limit (VLIL), before any adjustments. The VLIL is the centerpiece of the entire calculation: most LIHTC rents are derived directly from it.
Step 4 — Apply the high housing cost adjustment
In areas where rents are expensive relative to income, the preliminary VLIL is too low — a family at that income level couldn't actually afford a typical 2-bedroom apartment in the area. HUD's correction: if 35% of the preliminary VLIL is less than 85% of the annualized 2-bedroom FMR, raise the VLIL to the income at which 35% of household income equals 85% of the annualized FMR. Rounded to the nearest $50. This adjustment fires in roughly one-third of all U.S. counties and is the single largest driver of variation between markets.
Step 5 — Apply the low housing cost adjustment
In areas where rents are unusually cheap, HUD applies a ceiling instead. It rarely binds — only 1-3% of FMR areas — but exists to prevent runaway VLILs in markets where the formula would otherwise produce outliers.
Step 6 — Apply the state non-metropolitan floor
After housing-cost adjustments, the VLIL is checked against 50% of the state's non-metropolitan median family income. If the adjusted VLIL would fall below this floor, it is raised to it. For Florida in fiscal year 2026, that floor is $40,400.
Step 7 — Apply the year-over-year cap and floor
Beginning in fiscal year 2024, HUD added a hard cap on year-over-year changes: increases are limited to the greater of 5% or twice the national MFI change, with an absolute maximum of 10% per year. Decreases are limited to 5%. When the cap binds, the VLIL is set to prior year × 1.10 and rounded down to the nearest $50 — directional rounding, not nearest. When the floor binds, the VLIL is set to prior year × 0.95 and rounded up. The remaining uncapped value carries forward as deferred upside.
Note on the prior-year value. The previous-year VLIL the cap is measured against is not estimated or derived — it is HUD's officially-published number from the prior fiscal year, retrieved from HUD's Section 8 Income Limits API and stored in our database when each new fiscal year's limits release. When an area has no prior published value (newly-defined areas, or data gaps), the cap is skipped entirely rather than applied to a guess.
From VLILs to LIHTC rents
With the income limit schedule in hand, the LIHTC rent ceiling is a matter of Section 42(g)(2) of the Internal Revenue Code: maximum monthly rent equals one-twelfth of 30% of the income basis appropriate for the unit's bedroom size. The basis comes from HUD's bedroom-to-household-size convention: studios use the 1-person VLIL, 2-bedrooms use the 3-person VLIL, and so on. Tiers above 50% AMI scale proportionally — a 60% AMI rent is 120% of the 50% AMI rent.
5. The two differentiators
Lintel reports include two analytical features that aren't available from any other source. Both fall out naturally once the engine is built carefully; both genuinely improve underwriting decisions.
95% confidence intervals
Every projected VLIL ships with a 95% confidence interval. The interval combines two sources of uncertainty:
- ACS sampling error. Census publishes a 90% margin of error alongside every income estimate. Lintel converts this to a 95% standard error and propagates it through the calculation.
- CBO forecast variance. CBO publishes historical accuracy data for its short-term wage projections; the typical 1-year forecast standard error is approximately 1.5%. We add this in quadrature to the ACS error.
Where HUD's year-over-year cap binds, we truncate the confidence interval at the cap value — the cap is not a probabilistic constraint, it is a hard ceiling. This produces an asymmetric interval that accurately reflects the structural fact that the upper bound cannot exceed the cap, while the lower bound is governed only by sampling error.
The practical use: when the interval is tight, you can trust the point estimate as published. When it's wide (smaller counties with high ACS sampling error), you have a quantitative basis for either choosing a more conservative number or noting the uncertainty to your credit committee.
Multi-year cap-overhang trajectory
When the year-over-year cap binds, the "blocked" growth doesn't disappear — it carries forward. An area where the unconstrained formula wants +20% growth, but gets only +10% under the cap, has +9% of deferred upside baked into its FY+1 anchor; that area will almost certainly hit the cap again the following year as the overhang continues to release.
Lintel surfaces this dynamic explicitly. Every report includes a forward trajectory projecting the next 5 to 10 years of VLILs under three assumptions: Static (0% — what's already locked in), Custom (the user sets the growth assumption), and CBO trend (the actual year-by-year CBO per-capita wage growth forecast through 2036, applied separately for each projected fiscal year).
The trajectory distinguishes two market types. Pent-up markets (where the cap is binding today) climb at the cap rate for several years until the overhang unwinds. Headroom markets (where the formula is below the cap) follow whichever growth rate the user assumes. For a deal closing in 2027 that will operate through 2032, this distinction can mean substantial differences in projected revenue.
6. Worked examples
Two examples — one simple, one with the full set of adjustments — showing the calculation end-to-end. Both reproduce HUD's published FY2026 results exactly.
Gainesville, Florida — a non-adjusted area
Gainesville's 2024 1-year ACS median family income is $94,013, with a margin of error well within HUD's reliability bounds. The 2-bedroom FMR is $1,493 per month.
- Trended MFI:
$94,013 × 1.0546 = $99,141, rounded to $99,100. - Preliminary 4-person VLIL:
$99,100 × 50% = $49,550. - High housing cost test: 35% of $49,550 is $17,342; 85% of annualized FMR is $15,229. Since $17,342 > $15,229, the adjustment does not trigger. The preliminary VLIL stands.
- State non-metro floor (50% of FL non-metro median): $40,400. The $49,550 VLIL is above the floor.
- Year-over-year cap: prior year VLIL was $48,200; cap allows up to $53,020. The $49,550 is within the cap. No adjustment.
Final 4-person VLIL: $49,550. Matches HUD's published FY2026 figure exactly.
Miami-Dade, Florida — high-cost area with the cap binding
Miami-Dade's 2024 1-year ACS median family income is $85,113. The 2-bedroom 40th-percentile FMR is $2,436 per month — substantially higher than Gainesville, reflecting Miami's rental market.
- Trended MFI:
$85,113 × 1.0546 = $89,756, rounded to $89,800. - Preliminary 4-person VLIL:
$89,800 × 50% = $44,900. - High housing cost test: 35% of $44,900 is $15,715; 85% of the annualized FMR is $24,847. Since $15,715 < $24,847, the adjustment fires. The new VLIL is
$24,847 / 0.35 = $70,992, rounded to $71,000. This jump — from $44,900 driven by income, to $71,000 driven by rent — is the high housing cost adjustment in action. - State non-metro floor: $40,400. The $71,000 is well above.
- Year-over-year cap: prior year VLIL was $61,950; cap allows up to
$61,950 × 1.10 = $68,145. The formula wants $71,000 but the cap binds, rounding down to $68,100. The remaining $2,900 of upside ($71,000 minus $68,100) carries forward as deferred overhang.
Final 4-person VLIL: $68,100. This matches HUD's published FY2026 figure exactly, including the directional rounding of the cap.
7. Limits and honest caveats
It is a projection, not a forecast. HUD's actual published limits can differ from the projection for at least three reasons: (1) the 1-year ACS used in the projection is the most recent available, but if the next vintage publishes between projection and HUD's release, HUD may use it; (2) HUD occasionally changes methodology between fiscal years; (3) the CBO forecast itself revises. We validate every release against HUD's published anchors — the engine reproduces HUD's FY2026 calculations to the dollar for both non-adjusted areas (Gainesville: $49,550) and high-cost cap-bound areas (Miami-Dade: $68,100). A nationwide backtest is forthcoming.
The confidence interval covers measurable uncertainty, not methodology change. Sampling and forecast variance are quantifiable. HUD changing how it calculates the cap, as it did in FY2024, is not. The CI describes the noise around a given methodology; it doesn't protect against the methodology itself moving.
The multi-year trajectory is conditional. The out-year caps in our forward projection assume HUD continues to apply the 10% absolute ceiling. If HUD adjusts the cap structure — which it has done before — the trajectory after the change point becomes a provisional estimate. We flag the trajectory as provisional in the product itself for this reason.
This is not investment advice. Lintel reports are engineering: a faithful reproduction of HUD's public methodology applied to public data, with appropriate uncertainty quantification. Whether to base a deal on these projections is a decision for you and your credit committee. We recommend, as a matter of professional practice, using the lower bound of the confidence interval when sensitizing a deal's financial model — particularly when the projection is for a fiscal year more than 12 months out.
8. Sources and references
Every Lintel projection can be reproduced from public data using only the documents below. Nothing in our calculation is proprietary; the product is the engineering, not the formula.
- HUD Notice PDR-2026-01 — Annual Income Limits methodology.
- HUD Income Limits Methodology Document, FY2026 — step-by-step reference.
- U.S. Census Bureau ACS, table B19113 — Median Family Income.
- HUD Fair Market Rents FY2026 — 40th-percentile rents by area.
- HUD Section 8 Income Limits API — official prior-year VLILs.
- CBO Budget and Economic Outlook, Feb 2026 — publication 61882.
- Internal Revenue Code § 42(g)(2) — LIHTC maximum rent statute.
Questions, corrections, or methodology suggestions? Email hello@lintelusa.com. We treat methodology questions as first-class — if you spot a mistake in this paper, we want to hear about it.
See also: the methodology overview (the ten-step scannable summary) and the About page for context on why this exists.