This paper explores the theoretical and empirical issues in unit root non-stationarity for multiproduct acreage response systems. The policy implication of a unit root in a multivariate autoregressive time series like a system of acreage responses is very disturbing; it implies that even a one-shot government policy that influences acreage will have a permanent response associated with it. A wheat/barley acreage response system for the Prairie province region of Canada is estimated and the existence of a unit root cannot ~e rejected empirically. The response of the stationary system of acreage responses to a policy shock is shown to be radically different from the response of a non-stationary system of acreage responses even though the empirical estimates of the reduced form parameters are only marginally different between the two models.