In this study, we develop a new approach to investigate spatial market integration. In particular, it is a Markov-Switching autoregressive (MSAR) model with time-varying state transition probabilities. Studying market integration is an effective way to test whether the law of one price holds across geographically separated markets, in other words, to test whether these markets perform efficiently or not. In this model, we assume that the parameters depend on a state variable which describes two unobservable states of markets – non-arbitrage and arbitrage – and is governed by a time-varying transition probability matrix. The main advantage of this model is that it allows transition probabilities to be time-varying. The probability of being in one state at time t depends on the previous state and the previous levels of market prices. An EM (Expectation-Maximization) algorithm is applied in the estimation of this model. For the empirical application, we examine market integration among four regional corn (Statesville, Candor, Cofield, Roaring River) and three regional soybean markets (Fayetteville, Cofield, and Creswell) in North Carolina. The prices of these markets are quoted daily from 3/1/2005 to 6/30/2010. Six pairwise spatial price relationships for the corn markets, and three pairwise spatial price relationships for the soybean markets are examined. Our results demonstrate that significant regime switching relationships characterize these markets. This has important implications for more conventional models of spatial price relationships and market integration. Our results are consistent with efficient arbitrage subject to transactions costs.