brain/
conceptstock-market

Legislative divergence base rate: ~11% of laws land far from their forecast

Notes

Legislative divergence base rate: ~11% of laws land far from their forecast

One-line summary: Brian Potter scored 239 federal laws and found ~89% land roughly where expected — but ~11% diverge substantially, and the divergences are asymmetric: a throwaway provision becomes NEPA or the 401(k), while a headline reform (credit-rating agencies) does nothing at all. The discipline this imposes on policy-driven theses is a base rate, not a bias.

The insight

Policy-driven chains in this project usually reason forward from a law's stated purpose to its market effect. brian-potter's scoring exercise supplies the base rate that reasoning needs, and it cuts against the instinct in both directions.

The distribution is mostly boring. 95 of 239 laws (40%) scored −1 to +1 — they did what was expected. Only ~11% scored ±5 or more. "Most federal laws, it seems, do more or less what they're expected to do." Anyone whose policy thesis depends on a law having a surprising second-order effect is betting on a ~1-in-9 tail.

But the tail is fat on the upside and it is where the alpha lives. The distribution has a positive skew, and the two canonical cases are directly load-bearing for this project's book:

  • NEPA — the environmental-impact-statement provision was added last minute, "was not covered in any major media publication", and in Congress received "neither debate, nor opposition, nor affirmative endorsement." It became the primary tool for stalling infrastructure and seeded CEQA and imitators worldwide. This is the permitting friction that now gates datacenter, grid, and fab construction — i.e. a forcing function under ferc-large-load-to-dc-gridscale-construction and aging-grid-replacement-to-td-capex-supercycle originated in a provision nobody read.
  • The 401(k) — "originally considered an insignificant provision of the 1978 Revenue Act," scored as having "a 'negligible effect upon budget receipts'." It is now the fifth-biggest individual tax break and the primary US retirement vehicle — the statutory origin of exactly the defined-contribution flow trump-accounts-to-sp500-structural-inflows and passive-flows-db-to-dc trade. Potter quotes Daniel Halperin, then a senior Treasury official: "There was absolutely no discussion in '78 that if you do this, the world is going to change."
  • TFTEA (2015), scored +7 — sold as customs modernization; one provision raised the de minimis exemption from $200 to $800 and "is credited as a driver of the explosion of Chinese imports from companies like Shein/Temu over the next several years, until it was reversed by the second Trump administration."

The converse is the falsifier, and Potter states it himself. Creating a legal capability does not mean anyone uses it. The 2006 Credit Rating Agency Reform Act "did nothing to disrupt the market share of the existing agency oligopoly" — S&P, Moody's and Fitch still hold 95%. So the base rate is asymmetric surprise, not reliable amplification: a small provision is a wide distribution, not a call option. That is the sentence that should discipline any "this obscure clause will be huge" thesis in this vault.

The mechanism of divergence is Potter's most transferable contribution: laws are like technologies. A law creates a capability; once the capability exists, actors use it in ways the authors never modelled. Divergence comes from (a) legal operators — courts interpreting broadly (NEPA's "detailed statement"), prosecutors using laws for unexpected ends (the DNA Fingerprinting Act → ICE), agencies declining to use them at all (zero prosecutions under the 2012 STOCK Act), or future Congresses expanding/gutting them; and (b) the world responding unexpectedly (FedEx and air-cargo deregulation; Shein and de minimis).

Two structural patterns worth carrying: larger laws (more pages) skew more positive — plausibly a bundling effect, since bundling many draws from a positively-skewed distribution raises the chance one lands big. And bills that create substantially new programs have much higher variance than appropriations bills, even at the same mean.

Evidence

  • The headline result — From 2026-07-16-feed-construction-physics-how-predictable-are-laws: "Overall, I found that federal laws mostly do what they're expected to do. But a substantial fraction of them — around 11% — diverge significantly, having either much smaller or much larger effects than originally predicted." Distribution: 95/239 (40%) scored −1 to +1; 49 (20%) scored −2 to −4; 68 (28%) scored +2 to +4; ~11% scored ±5 or more.
  • NEPA — From the same source, quoting Potter and Alec Stapp: "[The provision] was not covered in any major media publication. In Congress, it received 'neither debate, nor opposition, nor affirmative endorsement.'" And: "Not until several months after NEPA was passed did environmental groups realize what a potent weapon they'd been handed."
  • The 401(k) — From the same source, quoting Bloomberg: "The initial provision was estimated to have a 'negligible effect upon budget receipts.' Now, defined contribution plans are the fifth-biggest tax break for individuals, with an estimated revenue loss to the government of $61.4 billion in fiscal 2014." And Halperin: "There was absolutely no discussion in '78 that if you do this, the world is going to change."
  • The falsifier (capabilities go unused) — From the same source: "Changing the rules for designating a 'nationally recognized' credit rating agency, as the 2006 Credit Rating Agency Reform Act did, did nothing to disrupt the market share of the existing agency oligopoly." Also the 2012 STOCK Act: "As of 2025, there have been zero prosecutions under this law despite suggestive evidence that congressional insider trading does occur."
  • Laws that under-delivered — Alaska Natural Gas Transportation Act 1976 (−7): the pipeline "was never completed." California housing laws: "limited to no impact on the state's housing supply" per YIMBY Law, "a surprising source: A new report by YIMBY Law, a pro-development nonprofit that would very much like to see these laws work." Adjacent politics-thread datapoint: same-state statutory under-delivery as gavin-newsom / california-fiscal-spiral-2026 governance-quality input — not a budget number.
  • Operator-side political examples — From the same source: DNA Fingerprinting Act 2005 (in a VAWA-focused DOJ bill) "was later used by ICE to collect DNA from immigration detainees." STOCK Act 2012: "As of 2025, there have been zero prosecutions under this law despite suggestive evidence that congressional insider trading does occur." SOX obstruction clause "was later used in an unexpected way: criminally charging hundreds of January 6 defendants (though this was later struck down by the Supreme Court)."
  • The framing — From the same source: "A law will create or modify some capability... aimed at accomplishing some particular thing. But once that capability is out there in the world, people might take advantage of it in different ways, finding uses for it that the creators of that capability never expected."
  • Structural patterns — From the same source: "If we graph a law's divergence score against the number of pages in the law, we see a small positive correlation." And: "Appropriations bills... have a lower variance than bills that create substantially new programs. Both types of bills have the same average divergence score, but the probability of a large or small divergence is much greater with the latter."

Why it matters to politics

This is institutional/legislative methodology, in-scope for the politics thread: it is a base rate on how statutes actually behave, which is the right prior for every Trump-era EO, industrial-policy bill, and "this obscure clause will remake X" claim this thread otherwise reasons from intent. Potter's operator-side examples are specifically political:

  • Prosecutors using a law for unexpected ends. DNA Fingerprinting Act of 2005 (in a VAWA-focused DOJ authorization) "was later used by ICE to collect DNA from immigration detainees." SOX §1512(c) "obstruction of an official proceeding" used to charge January 6 defendants — later struck down by SCOTUS. STOCK Act 2012: "zero prosecutions" despite "suggestive evidence that congressional insider trading does occur."
  • Future Congresses expanding or gutting. Parole Reorganization Act 1976 made almost moot by Sentencing Reform Act 1984 abolishing federal parole.
  • California housing laws Potter cites (YIMBY Law: "limited to no impact on the state's housing supply") are a same-state under-delivery datapoint adjacent to california-fiscal-spiral-2026 / gavin-newsom — a pro-housing group's own report that the statutes did not do what authors and opponents both predicted.
  • NEPA as accidental constraint on state capacity / industrial policy, not just a datacenter-permitting ticker story.

Do not treat the ~11% as a calibrated probability (Potter: "preliminary"; AI-scored). Use it as an order-of-magnitude prior against both "the law will do exactly what the press release says" and "this throwaway provision is obviously the real story."

Implications

Contradictions / tensions

  • ⚠ The methodology is AI-scored and the author says so. Potter had Claude Opus 4.8 estimate expected effect, actual effect, and a divergence score for each law. He is explicit about the limits: "I expect there to be errors in various evaluations, and I would regard these results as preliminary." He also flags that the scoring rubric produced an artifact — a dip at ±1 — "another reason why we should treat these results with a grain of salt."
  • Judging "expected effect" is irreducibly fraught, and Potter concedes it: distinguishing aspirational statutory goals from real ones "requires a process of guessing and using context clues that is likely to be error-prone" (his Indoor Radon Abatement Act example).
  • Attribution across successor laws is a judgment call — much of a law's effect often arrives via later laws that build on it (his Psychotropic Substances Act → civil asset forfeiture example, scored +4 rather than higher because "the later bill did most of the work").
  • Survivorship in the examples. The memorable cases (NEPA, 401(k), TFTEA) are selected because they diverged. The 11% figure is the disciplined number; the anecdotes are not evidence of frequency.
  • This is one analyst's preliminary blog analysis, not peer-reviewed work. Treat the ~11% as an order-of-magnitude prior, not a calibrated probability.

Open questions

  • Does the ~11% divergence rate hold for the specific subclass this project cares about — laws that re-price an industry — or is it diluted by post-office renamings and appropriations?
  • Potter published the full prompt and evaluations. Is the underlying dataset queryable for the regulatory categories in this book (energy, healthcare, semis, trade)?

Evidence (added 2026-08-07) — the 21st Century ROAD to Housing Act, scored by the same author who set this base rate

From 2026-08-06-feed-construction-physics-21st-century-road-to-housing-act. brian-potter's section-by-section read of the newly-enacted Act is a clean application of this concept by the person whose "How Predictable Are Laws?" piece anchored it.

The Act has 59 individual sections, and Potter splits them sharply: a few target genuinely binding constraints — legalizing single-stair buildings, reducing the burden of environmental review, and improving incentives for lenders to offer small-dollar mortgages — while "a collection of other provisions… make various positive changes regarding housing (new resources, new or improved programs) but that probably won't affect housing supply at all."

His framing is the concept restated: "it's not always obvious how binding the particular constraints that the law targets are, and how much housing we can expect to be unlocked," and separately that "the actual effects of laws can diverge significantly from lawmakers' intentions."

The transferable lesson for this wiki's policy-driven chains: a bill's size (59 sections, "enormous housing bill") carries almost no information about its supply effect. The question is always whether the specific constraint relieved was the binding one. Same test this project applies to ferc-large-load-to-dc-gridscale-construction and state-datacenter-siting-moratorium-risk — and it is why the environmental-review provision is the one worth tracking here, since NEPA/SEQRA-style review is the constraint this vault has repeatedly found to be binding.

⚠ Explicitly incomplete: this is Part 1. Potter states the next post covers Title II, "a long section with lots of relevant housing supply provisions in it" — so the supply-relevant detail is deferred by the author. Do not treat this as the Act's full supply read. Bears on homebuilder-land-bank-mirage-to-margin-impairment and nvr-asset-light-exemption-homebuilder-pair-trade, neither of which is moved today.

Related

Update (2026-08-10) — two clean instances, one measured

From 2026-08-08-feed-construction-physics-reading-list-08082026 and 2026-08-10-autoresearch-datacenter-siting-gate-load-relocation.

1. A measured effect size for a non-housing statute acting as a housing constraint. A Max Tabarrok study of >1,200 species listings, 1980–2024, finds the Endangered Species Act reduces housing construction by ~10% near protected habitat and ~4% of total US housing stock nationwide — newly listed species measurably cut annual permit flows in treated areas. This is the rare case where "did the statute bind?" has a number attached, and the number comes from a law that is not a housing law at all. It sharpens this concept's core claim in both directions: a bill's stated subject carries as little information as its size, and the binding constraint is frequently a statute nobody classified as relevant. brian-potter's own 21st-Century ROAD to Housing Act analysis (2026-08-06) already flagged environmental-review burden as one of the few genuinely binding provisions in a 59-section bill — this quantifies why.

2. The data-center moratorium wave is a textbook count-vs-effect divergence. As of June 2026, 14 states were considering restrictions and bills were introduced in 11 — and the statewide legislative route is mostly failing: Virginia, Georgia HB 1059, South Dakota SB 232 and Wisconsin SB 1061 all failed to advance; Ohio's constitutional amendment to ban data centers over 25 MW failed to qualify for the 2026 ballot, gathering ~70,000 of the 413,488 required signatures. Anyone counting bills would score this as a rising trend.

Meanwhile the effective action ran through two channels that a bill count cannot see: executive orders (NY EO 62 on 2026-07-14; the Texas PUCT/ERCOT audit directive on 2026-08-03 — together gating the largest queue in the country) and county ordinances (Indiana alone: 11 ordinances, 17+ temporary moratoriums, 2 outright bans).

The transferable lesson, stated as a rule: when measuring policy risk, count the instruments that bind, not the bills introduced — and note that the three instruments (executive, legislative, local) have different success rates and different durabilities. Legislation is the most durable and the least likely to pass; executive action is the fastest and the most reversible; local action is the least legible and, on this evidence, the most numerous. See state-datacenter-siting-moratorium-risk.

New instance (2026-08-11) — a rule in force for eight months and still operationally inert

From 2026-08-11-autoresearch-macro-buckets-healthcare-financials-naic-summer-meeting.

The NAIC's "Discretion Amendment" — which lets the NAIC challenge credit ratings that differ by three or more notches, the instrument at the centre of naic-rating-scrutiny-to-pe-insurer-flywheel-derate — took effect 2026-01-01, but "systems required to operationalize the process are still being developed." Alongside it, the RBC Governance Task Force review is ongoing with no finalized modification, and there are no announced changes to RBC charges for CLOs or collateral loans.

This is a cleaner instance than most in this concept, because the usual divergence is bill introduced vs. bill passed. Here the rule passed and commenced and still does not bind — the gap is between legal effect and operational capacity. It suggests a third category for this concept alongside "introduced ≠ passed" and "passed ≠ enforced": commenced ≠ operable.

The practical read for the book: an announced-and-commenced regulatory instrument should still be priced on its operational date, not its effective date.

Update (2026-08-17) — a worked example: the 21st Century ROAD to Housing Act

Brian Potter's section-by-section read of Titles II and III (17 sections) is a clean instance of this concept's claim — a bill can target real constraints across seventeen sections and still move almost nothing, because the binding question is never "does the provision address a constraint" but "was that constraint binding, and is the incentive strong enough."

"The biggest potential impacts are concentrated in a small number of sections (the Innovation Fund, the RAD limit increase, the manufactured home chassis rule), but their actual impact will be a function of how much the constraint it targets was actually binding and how strong the incentive is, which in general is difficult to know."

"There are a bunch of sections that target real problems…but probably don't have enough juice to make much of a dent in them."

Per-section scale, which is the useful part — most provisions are an order of magnitude too small to move a national market: Innovation Fund $200M/yr (awards ≤$10M); RESIDE Act ~$100M; PRICE Act $225M / 17 projects in 2024; CDBG $3.3B/yr with ≤20% eligible for housing; NEPA delegation worth a 15–30% review acceleration; manufactured-home chassis rule a 5–10% cost reduction that Potter still expects little uptake from; FHA made three manufactured-home loans in 2021. The two with real scale are structural rather than fiscal: the RAD cap +100,000 units and §206's NEPA categorical exclusions (projects ≤15 units, office conversions, infill), which Potter calls "probably the most important section in the bill so far."

Potter offers no aggregate supply estimate and names no homebuilder or building-product company. That absence is the point for this project: a bill with a name, a number and seventeen provisions still yields no tradeable, which is exactly the base rate this concept exists to enforce against headline-driven legislative theses.

(All from 2026-08-14-feed-construction-physics-road-to-housing-act-part-ii; Part I ingested 2026-08-06.)

Referenced by
Mechanisms
FERC RM26-4-000 Section 403 ruling → large-load interconnection streamlined → AI DC power bottleneck cleared → PWR/GEV/VRT accelerated order intake
Concepts
California fiscal spiral (2026): "balanced budget" vs tax-base exodusPBM profit already left the spread — pass-through promises are cheapSole-source concentration in generic sterile injectablesState NADAC-plus-dispensing-fee floors are the binding-now PBM forcing functionState-level data-center siting moratorium risk
Questions
Does the 2026 drug-pricing policy stack compress PBM "spread" enough to re-rate the vertically-integrated PBM-insurers (CVS / CI / UNH)?Does the NAIC's new authority to challenge private letter ratings break the PE-insurer capital flywheel that funds Apollo's and KKR's AUM growth?
Entities
Brian PotterCheniere Energy