Cost, construction type and the labor package — California

Four passes, and the binding number is the second one

Cost estimation here is not one event. It is four escalating passes, each with a different owner, precision band and timing.

The structural problem is the ordering. The number that matters legally — the one in the CTCAC application — is produced at roughly the second level of precision and then frozen. Construction start is commonly 12–24 months later or more. CTCAC re-tests at placed-in-service against the revised limits for the placed-in-service year (4 CCR §10327(c)(5), §10325(d)), which gives relief on the limit side and nothing at all on the cost side. The developer carries the escalation risk in between.

Who is actually in the room: developer project manager, architect, GC preconstruction estimator, LIHTC consultant, and — if any labor-standards trigger is present — a labor compliance consultant or third-party monitor. The prevailing wage determination itself is nobody's job until it becomes everybody's problem.

The threshold basis limit is the spine of the phase

Understand what these are. CTCAC derives them from a rolling five-year window of awarded-project cost data — 2021 through 2025 for the 2026 limits, adding the newest award year and dropping the oldest each year — inflated forward with the R.S. Means Historical Cost Index and City Cost Indexes. The system dates to 2008 and now works at the region level rather than pure county level, which is why Riverside, San Bernardino and Imperial share identical figures and why 26 mostly rural counties all show $560,000 at 2BR. CTCAC's 2026 memo notes R.S. Means showed relative stability through 2025, producing only a slight increase in some regions. These are regulatory caps, not cost estimates. Treating them as a cost benchmark is a classic and expensive error.

The boost menu at 4 CCR §10327(c)(5) is where unit mix, construction type and the labor decision converge.

Subparagraph (F) grants +10% for a High or Highest Resource location on the CTCAC/HCD Opportunity Area Map — but only where the county's unadjusted 9% 2BR limit is $500,000 or less. The lowest 2BR limit across Riverside, San Bernardino, Imperial, San Diego, Orange and Los Angeles is $531,200, so in 2026 that bump is unavailable in all six. The gate can flip either direction next January. Note separately that applicants may elect the Opportunity Map designation in effect when initial site control was obtained, up to seven years prior.

One genuinely unresolved question sits underneath all of this, and the sources do not settle it. §10327(c)(5) says "a twenty percent (20%) increase to limits," "a ten percent (10%) increase," and so on, and there is no aggregate cap outside the internal 20% cap within subparagraph (B). What the regulation does not say is whether stacked increases are additive on the base (20 + 10 + 10 = 130%) or compounding (1.20 × 1.10 × 1.10 = 145.2%). That is a material swing in the denominator of a non-discretionary kill switch. The research corpus flags this as its highest-priority unknown and states plainly that it must not be implemented from the regulation prose. Resolve it by reading the adjusted-limit formula cells in the published CTCAC application workbook, which is a free .xlsx with live formulas.

The high-cost test, and how to read the headroom

4 CCR §10325(d), with a parallel provision at §10317(i)(5), sets the high-cost test. Breaching the application-stage gate is not a warning — it is the end of the application cycle, plus the carrying cost of holding site control until the next round.

Four mechanics are worth getting exactly right.

A caution on the citation itself. The domain brief anchors the test at §10325(d) with a parallel at §10317(i)(5); the corpus synthesis logs a third rendering, §10323(i)(5), and treats the section number as unsettled pending a grep of the March 2026 regulations PDF. Verify the number against the current PDF before putting it in anything a regulator will read.

Whatever the section number, the useful output is not a percentage. Express remaining headroom in dollars, because that is the number that tells a developer how much cost risk the deal can absorb.

The 4-to-5 story triple cliff

This is the highest-value, least-modeled interaction in the domain, and it is a triple cliff because three independent rule systems break at the same line.

CTCAC partially compensates through the +10% Type III and +15% Type I basis limit bumps. Use those as the defensible proxy for the construction-type premium — they are the regulator's own published estimate. The widely repeated rule that Type V to Type III adds 12–22% per square foot and Type III to Type I adds a further 8–18% appears only in trade blogs and marketing content and traces to no primary or peer-reviewed source. Do not underwrite to it.

The delta that would let you price the cliff properly does not yet exist in the corpus. The residential side is verified below. The matching general/commercial determination has not been pulled, so no verified residential-to-commercial delta exists yet — pull it craft by craft for the same counties and effective period.

Watch the expiration date on whatever determination you are using. An expired determination is a live compliance risk, not a stale reference.

The labor package attaches to the capital stack, not the site

LIHTC alone triggers nothing — it is a tax credit, not a direct federal subsidy. Almost anything else in the stack does. A site screener structurally cannot answer whether a deal is prevailing wage; a feasibility model that knows the intended sources can.

California: Labor Code §1720(a)(1) defines public works as construction, alteration, demolition, installation or repair done under contract and "paid for in whole or in part out of public funds." §1720(b) defines public funds very broadly, and three of its six categories catch affordable deals constantly.

California courts construe those exemptions strictly. Reported case law holds that §1720(c)(6)(E) says nothing about a project partly financed by below-market public loans in combination with other financing — mixing an exempt source with a non-exempt one defeats the exemption. Prevailing wage applies to public works over $1,000, with relief up to $25,000 for construction and $15,000 for alteration or repair where the awarding body has an approved labor compliance program. Every real project is over the line. Surface the triggers and the citations; the binding answer comes from a DIR public works coverage determination or an opinion of counsel, never from a model.

Federal Davis-Bacon triggers are portable to every state and worth modeling as a separate layer.

Entitlement pathways now carry their own labor package.

Skilled and trained workforce under Pub. Contract Code §2601 requires that all workers in apprenticeable building trades be skilled journeypersons or registered apprentices, with 60% of journeypersons being graduates of an approved apprenticeship program for work performed on or after January 1, 2020 — the current, stable figure. §2602 requires monthly compliance reporting to the awarding body.

One live ambiguity to flag rather than resolve: Labor Code §1725.5 (initial nonrefundable registration fee $400, Director may set fees up to $800) was stated to remain in effect only until July 1, 2026 and to be repealed as of that date, with a successor structure at §1725.6 that AB 2011's §65912.130 already cites. The source research could not cleanly determine from secondary sources which section is currently operative or what the current fee is. Read the current statute and the DIR registration page before quoting a number.

What the benchmarks support, and what they do not

Escalation is the gap between when the application budget is locked and when construction starts.

Per the San Diego Housing Commission / BAE Affordable Housing Cost Study of April 17, 2025, drawing on CTCAC application data. The CCCI is published free by California DGS; the underlying ENR Building Cost Index is paywalled.

By construction type, from the same SDHC/BAE study:

The prevailing wage premium is the best-supported number in the phase. Terner Center, August 2, 2024 (n=859 California LIHTC awards, 2020 to 2023, covering 75,426 units, all figures in 2024 dollars): the raw uncontrolled gap is $428,000 per unit on prevailing wage applications against $258,000 on non-PW, and the authors explicitly say that overstates the effect.

The regional coefficients from the same model reshape how to think about geography in Southern California.

Terner's own stated limitations matter. The data are developer estimates at application, not final costs. Applications were inconsistent and required manual repair. The model is explicitly not causal. And it covers construction costs only — Site Work, Structures, General Requirements, Contractor Overhead, Contractor Profit, Prevailing Wages, General Liability Insurance and Third-party Construction Management — not land, architecture, legal, impact fees or construction interest.

73% of non-PW projects still used at least one other public funding source, which means most non-PW deals are one policy change away from being PW deals.

The structural caps, modular, and the order to run this in

Three caps bind the budget independently of the basis limit.

On modular: CTCAC's regulations contain no modular-specific basis limit adjustment. Modular projects compete under the same threshold basis limits and the same 14% builder OH&P cap. The question that actually determines whether modular saves money on a prevailing wage deal — whether factory labor is covered by California prevailing wage — is unresolved in the sources. The general principle in DIR coverage determinations and case law is that off-site work performed at a site established for the project can be covered while work at a permanent, general-purpose commercial facility usually is not, but no controlling DIR determination specific to modular multifamily housing was located. No credible California-specific modular cost-savings study was found either; the frequently cited "20% cheaper, 50% faster" claims trace to manufacturer materials. This is a question for counsel, not for a model.

The practical order:

Those first four decisions are jointly determined and each one moves the limit. That makes this a small optimization problem rather than a lookup, which is worth knowing before anyone starts iterating by hand.