The Denominator Problem: What CUSMA Is Actually Worth  ·  Part 3 of 4

The Insurance Policy and Who Pays the Premium

Part 2 ended on a puzzle. In 2024, the realized cash value of CUSMA (USMCA in US usage), the duties actually avoided at the border, was a rounding error against a trillion dollars of trade. Yet three governments had spent two years renegotiating it, legislatures ratified it, and firms kept pouring capital into supply chains configured around it. Either everyone involved was confused, or the cash value was never the product.

It was never the product. This part uses one metaphor, the series’ second, and cashes it out immediately. CUSMA was an insurance policy: the premium was real money paid every year, the payout was zero in every normal year by design, and the insured peril was a reversion to tariff walls on a continent whose factories had been built assuming there were none. A policy that never pays out is not worthless. It is what allows a company to put a 30-year assembly plant in Ontario or Coahuila instead of Ohio. In literal terms: the agreement’s value was the reduction in the probability distribution of future trade costs, priced into every long-lived capital decision on the continent, and that value never shows up in customs receipts until the peril actually arrives.

The history matters here, because the near-zero 2024 spread is routinely misread as evidence the agreement never did anything. The original NAFTA closed a wedge that was large and asymmetric. Mexico’s average applied tariff was 10 to 12% in 1993, and that was after a decade of unilateral liberalization that had already cut the share of imports requiring government licenses from over a third in the mid-1980s to about 22% by the end of 1988 (CBO, The Effects of NAFTA on U.S.-Mexican Trade and GDP, 2003). The United States, by contrast, admitted 51.2% of Mexican imports duty-free in 1993 and averaged 2.07% overall. NAFTA phased the Mexican wall down to zero. By 2020, when CUSMA replaced it, the visible spread between agreement rates and default rates had collapsed to the 2 points Part 1 described, not because the agreement was idle but because its work was finished and embedded in the baseline. A successful trade agreement is supposed to look worthless on this measure. The spread is small because the spread was eliminated.

And capital did deploy on the policy, which is the claim’s testable half. The suppression effect had just been measured in the other direction: during the 2017 to 2019 renegotiation, the Bank of Canada marked Canadian business investment down 3% by 2021 on trade uncertainty alone (Bank of Canada, July 2019), and the USITC’s ratification-case model located CUSMA’s largest gains in the provisions that reduce policy uncertainty (USITC Publication 4889). Once the agreement was in force, the US direct investment position in Mexico grew 43.8% over 2020 to 2024, to $159.2 billion, and in Canada 25.3%, to $459.6 billion (BEA direct investment statistics). Mexico set FDI records in 2023 and 2024 ($36.1 billion, then $36.9 billion), pushed Monterrey industrial vacancy near 1%, and in 2024 broke its 2017 vehicle-production record and its 2018 vehicle-export record (CNIE report to Congress, April 2026). Canada drew C$85.7 billion of inbound FDI in 2024, the most since 2007 (Statistics Canada). Two complications belong beside that record, and both sharpen the argument. Composition: Mexico’s record FDI years were 74 to 78% reinvested earnings, with genuinely new investment falling to 13.4% of the total in 2023 and 8.6% in 2024, incumbents deepening rather than entrants arriving. And timing: the certainty was being discounted before 2025. Tesla announced a Monterrey gigafactory in March 2023 and paused it in July 2024, with Musk telling the earnings call that if heavy tariffs on Mexican-built vehicles were coming, “it doesn’t make sense to invest a lot in Mexico if that is going to be the case,” and the Bank of Canada’s November 2024 business survey found US trade policy uncertainty already holding back investment plans (Business Outlook Survey, Q4 2024). Capital deployed where the policy was believed, and stalled exactly where its credibility thinned. That is what buying insurance while doubting the insurer looks like in the data.

So the payout side was dormant. The premium side never was. Four groups paid it continuously.

Producers paid in compliance. Preference is not free; it is earned through rules of origin, including regional value content requirements (RVC: the minimum share of a good’s value that must originate within North America, which for passenger vehicles CUSMA raised to 75%). Meeting them means tracing content through suppliers, keeping certification records, and restructuring sourcing when a component falls short. The Federal Reserve’s study of CUSMA automotive trade put these compliance costs at 1.4 to 2.5% of import value on an ad valorem equivalent basis, and extrapolated a potential annual compliance burden for the manufacturing sector of $39 to $71 billion (FEDS Note, July 18, 2025). The proof that this premium is real money is behavioural: the same study documents importers who chose to pay the 2.5% MFN duty on passenger cars rather than incur the cost of proving origin. When a firm pays a tariff it could legally avoid, the avoidance cost has revealed itself as larger than the tariff. That is a market price for the premium, set by the people paying it.

Specific constituencies paid in concessions. Canadian dairy paid through tariff-rate quotas (TRQs: quantities that may enter at low or zero tariff, with prohibitive tariffs beyond the quota) that opened guaranteed shares of the Canadian market to US product, a concession extracted in the 2018 renegotiation and disputed under the agreement’s panels since. Mexican labour paid, or from another angle was paid, through the labour reform and the Rapid Response Labor Mechanism, which lets the US suspend preferences at individual facilities found denying collective bargaining rights; Mexico rewrote domestic labour law as a condition of keeping the policy in force. And all three governments paid in bound policy autonomy, the standing constraint on instruments they might otherwise have used. These premiums are structural rather than annual, but they were the price of the certainty product all the same.

Consumers paid nothing while the policy held, then paid the most when it lapsed. Through 2024, US consumers were net beneficiaries: duty-free North American goods at prices that embedded no tariff. From March 2025, the incidence flipped. The IEEPA tariffs collected on non-compliant goods, roughly $166 billion across all countries before the Supreme Court struck the authority in February 2026 (a figure from litigation-adjacent reporting), plus the Section 232 duties that remain, entered at the border and passed through in the usual way: to importer margins first, then prices. Whatever share landed on US buyers, the direction is not in dispute, and it makes US households the largest single class of premium-payers in the post-2025 configuration. The Tax Foundation’s forward-looking version of the same point: ending the CUSMA exemptions entirely would cost about $300 per US household in 2027 alone (Tax Foundation, June 23, 2026).

And the fourth premium line, the one added after 2022: defensive subsidies. When the US Inflation Reduction Act created the 45X production credit, it paid $35 per kilowatt-hour for battery cells and $10 for modules manufactured in the United States, not North America (IRS, advanced manufacturing production credit); the companion 30D consumer credit required North American final assembly plus content sourcing tests (IRS, clean vehicle credit). Treaty membership gave Canadian plants no claim on the production credit at all. To keep two anchor investments from migrating south, Canada matched it: up to $13.2 billion in production support plus a $700 million construction contribution for Volkswagen’s St. Thomas plant, and up to $15 billion for the Stellantis-LG plant in Windsor, with the Parliamentary Budget Officer costing the combined production subsidies at $28.2 billion and estimating government break-even in about 20 years (PBO, September 2023; PBO costing update, November 2023). The federal-Ontario announcement states the mirror explicitly: “should the incentives offered under the U.S. IRA be reduced or cancelled, so would the performance incentives under the agreement” (ISED, July 6, 2023). Canada’s contract is denominated in American legislation.

Status, mid-2026, because commitments are not disbursements: the Windsor plant is producing (modules from October 2024, cells from November 2025, with LG Energy Solution acquiring full ownership from Stellantis in February 2026), and federal payments reported to date are at least $530 million, nearly all of it construction support, with the first $40.4 million in production claims paid at contract rates that mirror 45X exactly (CBC News, November 26, 2025). St. Thomas is under construction with production targeted for 2027, so its production support has not begun to pay. The 2025 US budget law terminated the 30D consumer credit as of September 30, 2025, but left 45X battery rates intact, so the liability Canada mirrored still stands through the credit’s 2030 to 2032 phase-down (IRS OBBB guidance). These figures are presented as order-of-magnitude premiums, committed fiscal capacity in the tens of billions, not as computed values; that is the same measurement discipline that kept subsidies out of Part 2’s arithmetic.

The comparative point is what matters for the series. In a subsidy race, fiscal depth substitutes for tariff preference: the party that can write the largest cheques can relocate investment regardless of what any treaty says about tariffs. The three CUSMA parties hold very unequal amounts of it. The United States legislated production credits at continental scale; Canada matched two plants and strained to do it; Mexico, whose entire attraction proposition is cost, is least able to pay at all. A competition the treaty does not regulate is one the poorest member loses by default.

Which brings the incidence analysis to its structural conclusion, stated factually. The peril this insurance policy covers is US protectionist action: that is not a characterization but the observed 2025 to 2026 fact pattern, in which every payout-triggering event was a US statute. And the premiums, wherever they start, largely accrue to the United States: compliance spending is directed at proving content shares defined to favour North American and increasingly US production; tariff collections on the non-compliant flow to the US Treasury; and the defensive subsidies exist to counter a US fiscal instrument, meaning their function is to raise the price Canada pays to keep what it already had. The insured party pays the premium to the party that controls the peril. That is the structure. It does not require an adjective, and the next part is about what it implies for anyone allocating capital against a treaty commitment.

Next in the series: Part 4, Repricing the Treaty, on the annual-review cycle, the instrument churn of 2026, and what the last 18 months established about what treaty text can and cannot secure.

Running Glossary  ·  Terms Introduced in Part 3

Regional value content (RVC). The minimum share of a good’s value that must originate in North America to qualify under CUSMA rules of origin. Passenger vehicles: 75% under CUSMA, up from 62.5% under NAFTA.

Tariff-rate quota (TRQ). A two-tier tariff: a set quantity enters at a low or zero rate, and imports beyond it face a much higher rate. The mechanism governing US access to the Canadian dairy market.

45X / 30D. Two Inflation Reduction Act credits. 45X pays US-based manufacturers per unit of battery cell and module production ($35/kWh cells, $10/kWh modules). 30D was the up-to-$7,500 consumer EV credit conditioned on North American assembly and content; terminated for vehicles acquired after September 30, 2025.

Rapid Response Labor Mechanism (RRLM). CUSMA enforcement tool allowing suspension of preferential treatment for goods from a specific facility found denying workers’ collective bargaining rights. Applies principally to Mexico.

(Parts 1-2 defined: preference margin, utilization rate, MFN, rules of origin, TiVA, Section 122, IEEPA, Section 232, effective tariff rate.)

A note on framing: Fully Briefed synthesizes publicly available government source material and translates it into financial terms. This is education, not legal, customs, or tax advice. The arithmetic in this series states its assumptions and sources so you can substitute your own inputs; for decisions that turn on your own tariff classification, origin documentation, or contract exposure, work with your customs broker and professional advisors.

Trevor Ryhorchuk, CPA, CIA, PMP

Fully Briefed — Canadian Trade Intelligence
tradeintel.fullybriefed.ca