Present value models of agricultural land prices in England and Wales
This study employs recently developed techniques in time series econometrics to estimate linear models of equilibrium price determination in a competitive market for durable assets. Motivating this study is the unstructured approach employed in previous land price research, where the theoretical model of agent behaviour is invariably mis-specified or left undeveloped and the empirical model prone to the problems of spurious regression. The joint issues of theoretical and statistical congruence play important roles here. Specifically, a theoretical model is developed in which market participants are assumed to price land using present value methods. At the market level this yields a reduced form expression of equilibrium price determination which can be estimated empirically using aggregate data for England and Wales. The concepts of error correction and cointegration are then investigated and applied to the land price model. A unique long run relationship is identified between real agricultural land prices, inflation and real agricultural rents. Taking account of inflation-hedging as a motivation for acquiring farmland, land prices are shown to be principally determined by the returns to land, as embodied by market rents. The empirical model is also congruent with theoretical predictions regarding the unit elasticity between asset prices and returns. The error correction representation of the cointegrating set indicates that the short run response of land prices to rent and inflation is larger than the long run response. Consequently, land prices initially overshoot their equilibrium values following changes in rents or inflation. The period of adjustment to long run equilibrium lasts around three or fours years. The long run real rate of discount on agricultural land is estimated at 3.6% confirming the widely held belief that real rates of return on farmland are low. Present value models incorporating naive, adaptive and rational expectations are also estimated and the adaptive model is favoured by the data.