We examine the performance of the threshold cointegration approach, specifically Band-TVECM, to price transmission analysis in an explicit context where trade decisions are made based on expectation of final prices, because trade takes time. We find that, following a standard inference strategy, a large portion of three-regime cases are not identified as such. Results show that transfer costs are systematically underestimated, particularly in three-regime models. The speed of price transmission is also biased in three-regime models. Furthermore, inferences about occurrence of trade are poor, with estimated models suggesting far lower market integration than is true in the data generating process .