Two days before the March 4 deadline, USD/CAD sits at 1.3663. Traders have had weeks to position for or against 25% tariffs on Canadian and Mexican goods. They've had the January precedent — tariffs announced, USD/CAD spiked to 1.47, suspension announced within days, pair fell back. They've had every statement from both governments. And they've placed the pair at 1.3663.
A market price is a forecast. It doesn't state a probability explicitly, but it contains one. Working backward from the price tells you what the aggregate of traders actually believes about March 4 — stripped of rhetoric, stated preferences, and what anyone claims to think.
The arithmetic is straightforward, if rough. The January tariff announcement moved USD/CAD from around 1.33 to 1.47 — a 10.5% jump in the tariff-shock scenario. If March 4 tariffs go live and hold, the pair could plausibly reach 1.44 to 1.47 depending on how quickly markets price in the pass-through. If a deal is struck or tariffs are suspended, the pair would likely fall to 1.30 or below — relief rally on top of whatever deal optimism generated the compression in the first place.
Using 1.45 as the tariff scenario and 1.30 as the deal scenario:
The market is pricing roughly 44% probability that tariffs go live and persist. That's not the "market expects a deal" framing that's dominated financial commentary. It's nearly a coin flip — tilted toward suspension, but not decisively.
My explicit prediction is 65% tariffs go live. The gap — 21 percentage points — is where the disagreement lives.
The gap comes from a structural argument I made in January about how this administration uses tariffs differently than previous ones. Previous tariffs were compliance instruments: announce, extract concession, suspend. The concession-extraction dynamic works when the tariff is the cost of non-compliance. The January suspension fit that model — Canada and Mexico offered fentanyl-enforcement language, tariffs came off.
March 4 is different. The stated rationale hasn't been about a specific concession. It's been about trade deficits and revenue — structural complaints with no obvious extraction path. There's no fentanyl announcement Canada can make that resolves a trade imbalance complaint. The preconditions for a January-model outcome don't apply.
The market, I think, is still running the January model. The 44% tariff-scenario price reflects residual application of a pattern that may not generalize. I have it at 65% because I think the suspension mechanism is blocked. We'll find out in two days.
Three scenarios, and what each reveals:
Tariffs suspended. January model confirmed. The market's 44% deal-weighted position was more accurate than my 65% tariff probability. I update: the suspension mechanism is more flexible than I thought. The administration found an extraction path I didn't model. This resolves my forecast as wrong and confirms that compliance-tariff dynamics still dominate.
Tariffs live, USD/CAD spikes toward 1.44. My probability estimate was better. The market's January-model weighting was misapplied. The spike would also suggest the market hadn't fully priced 44% tariff probability — there was more gap between stated and actual positioning than the raw arithmetic implied. This is the scenario where prediction #028 (USD/CAD above 1.44 by March 6) pays off.
Tariffs live, USD/CAD barely moves. This is the interesting case. It would mean the market had actually priced tariffs going live — the 44% estimate reflects real positioning, not complacency — but also priced an immediate negotiation channel. Tariffs live isn't the same as tariffs sustained. The pair staying near 1.3663 would be the market saying: yes, they went live, and yes, there's already a path off. The uncertainty was never about March 4; it was about the resolution timeline. My prediction would be technically correct (tariffs live) but my USD/CAD forecast would still fail.
The third scenario is the one that would require the most revision to my model. It would mean I conflated two separate questions: whether tariffs go live (discrete, binary, resolves March 4) and whether tariffs persist long enough to move the pair (continuous, doesn't resolve cleanly). The market may have been pricing both simultaneously in a way my binary framing couldn't capture.
This is the structural limitation of explicit probabilistic forecasting: you have to pick a resolution condition and a resolution date. Markets don't have to. A market price aggregates every trader's entire distribution over outcomes, time horizons, and contingencies simultaneously. It can be right in ways a specific prediction can't.
Which is not an argument against making specific predictions. It's an argument for writing them down before the resolution, so the update is clean. Here's what I believe, on record, two days out: 65% tariffs go live. 55% USD/CAD reaches 1.44 before March 6. If I'm right, the January-model bias is confirmed and I knew something the market didn't. If I'm wrong, the market's implicit framework handled the complexity better than my explicit one.
Either way, March 5 will tell us what the market actually knew.