Segmenting agricultural land market according to development potential: a latent class approach

Abstract

Not all farmlands are purchased for farming. Where development pressures are strong and urban boundaries still fluid, some farmlands are purchased for non-agricultural purposes. However, since the future development use is not evident or pre-determined at the time of transaction, the farmland market may appear to operate as one albeit with latent segments. Analyses of land price determinants should involve some measures to ascertain the cause and the degree of functional segmentation in the market, so that the shadow prices of different land attributes can be differentiated by market segments. Using an extensive dataset of over 2,000 Malaysian farmland sales, our Latent Class Analysis confirms that there are two underlying distinct distributions and that within each distribution, relationships between variables display considerable local independence. Strength of potential drivers of farmland price is proven to differ according to segments. In addition, we are able to show that the segment classification results based on the parcel’s ‘developability’ was fairly accurate when compared to the classification given by official land valuation documents. This exercise proves that unobserved segmentation can be predicted with a reasonable degree of accuracy simply by letting the data ‘speak for itself’. In terms of agricultural support funding, the segmentation may allow for the country’s better targeting of recipients and refinement of farm support programs.