This study uses U.S.-level time-series data (1935-2015) to test the present value model of farmland prices. Following Engle and Granger (1987) we test each individual time-series for non-stationarity. We subject real farmland prices per acre, real returns per acre, and real interest rates (log-log form) to a set of unit root tests designed to test for (weak) stationarity. We find evidence from some of these tests to support stationarity (KPSS test) and also evidence to support non-stationarity. We also find that structural breaks in the data series may at least partly explain why we cannot reject the null that farmland prices and returns/acre are cointegrated. The observed breaks may be due to changes in the required risk premium on farmland investments, such as during the US farm financial crisis of the early 1980s. Also, the standard unit root and co-integration tests may not be powerful enough to detect co-integration.