Farmland Prices in New Zealand and Finland Myyrä, S. 1 and F.G. Scrimgeour 2 1 MTT Agrifoor Research Finland, Economic Research, Email: sami.myyra@mtt.fi 2 University of Waikato, New Zealand, Email: scrim@waikato.ac.nz Keywords: Farmland prices, GDP, aggregated agricultural income, convergence EXTENDED ABSTRACT Farmland prices are a critical determinant of farm profitability. Yet around the world analyses of farm land prices rarely produces models which are both contextually and empirically robust. This paper examines the movement of land prices in both Denmark and New Zealand during the period 1981 to 2005. Typically, farm prices are modelled in terms of their fundamental economic values as reflected by discounted future earnings (net-present value type of models (NPV)). Hedonic modelling approaches are helpful in relating earnings potential to land characteristics but still leave significant variation unexplained. This paper takes an alternative approach which is, in part, motivated by the observation that significant areas of land in Finland are owned by non-farmers and this is also true in New Zealand, though to a lesser extent. Hence the paper explores the impact of the wider economic performance on stochastic trend in land values by examining the convergence between Gross Domestic Product (GDP), aggregated agricultural income (AAI) and farmland prices (FP). A number of regional policy issues are also related to heavily subsidised agriculture in Finland. Our approach is to convert data series into a stationary form and in this way ensure the necessary conditions for analysis. Results from unit root testing provide evidence that logs of average farmland price (FP), GDP and aggregated agricultural income (AAI) series are first difference stationary I(1) series. In the case of the New Zealand data, the ADF was capable of confirming the result. For the Finnish data, the Hadri test was done in level as well as in first differences. On first differences, the null of joint stationarity is accepted. This suggests that the Finnish data series of GDP, aggregated agricultural income and farmland prices are also first difference stationary. However the Hadri test gives mixed results of stationarity of New Zealand data at levels, suggesting that these series are joint stationary I(0). This result is not supported by other panel data joint stationary tests. Unit root tests clearly suggest that modelling must be done on first differences. Results do not support prediction of fast structural change in factors affecting Finnish farmland prices. However, they support the view that agricultural factors do not have a large effect on agricultural land prices in Finland. This might partially explain why hedonic land price models fail to give reasonable explanation power over variations in farmland prices. Models where value of the land is given by the capitalized value of current and expected future streams of net income from agriculture fail because, at least in the Finnish case, they capture only a limited part of stochastic variation in land prices. However, the situation is completely different within the New Zealand data. Here the results give a reasonable base for models relying on the present value approach, because a reasonable part of the stochastic variation in land prices is explained by agricultural factors affecting agricultural income. Results also support the idea that agriculture’s influence on agricultural land prices is weakening. However, changes are not as dramatic as could be expected by agricultural product price statistics. These results show the importance of analysts extending their analyses to make better use of indicators of changes in the wider economy when seeking to explain fluctuation in agricultural land prices and in looking to anticipate future changes. 1061