Unit root tests are used to determine whether futures prices or spreads are mean reverting, but data errors, price limits, and market microstructure can produce misleading results. We examine daily live cattle, feeder cattle, September corn, December corn, and a constructed cattle crush spread from 1990 to 2025. Historical prices were validated against multiple data sources, and panel unit root tests were adjusted for price-limit observations and MA(1) microstructure noise. The preferred test does not reject a unit root for any individual futures market. The cattle crush spread shows weak evidence of mean reversion in unadjusted tests, but this result disappears after correcting for microstructure noise. Dropping limit-day observations has little effect, while correcting data errors materially changes the conclusions. If a cattle crush option is developed, the results favor using a Bachelier model.
Publications