The influence of urban sprawl on farmland prices in Belgium (refereed paper)

Since over half a century, Europe has been undergoing periurbanization; this phenomenon is similar to suburbanisation or 'urban sprawl' in the U.S. Hence, for each plot of land, there is competition for land use (agriculture, urbanisation). We here study the effect of urban sprawl on the price of farmland in Belgium. Using a set of very traditional urban economics variables, we show that the determinants of residential prices also explain the spatial variation of farmland. Indeed, the prospect of conversion from agriculture to residential land uses are high in Belgium, which is a densely inhabited country characterized by a tight urban network and where, moreover, land zoning is permissive. Therefore, urban sprawl largely dominates residential choice since over 40 years leading to fragmented landscapes where agriculture coexists with housing. We here use the Capozza & Hesley (1989) model that was recently developed by Plantinga & Miller (2001) and Cavailhés & Wavresky (2003). This model aims at analyzing urban growth from a microeconomic point of view, where agricultural land is converted in residential plot. Until conversion, the price of farmland is equal to the capitalization of the agricultural land rent; after conversion it is equal to the capitalization of the residential rent. It basically depends upon the commuting distance and the anticipated date of conversion (other determinants are also discussed: Ricardian land fertility, number of inhabitants of the township, income, border effects leading to discontinuities, etc.) A hedonic log-linear equation is used where spatial autocorrelation is controlled by Anselinsí method (1988) and the multicollinearity (if any) by partial least squares (PLS). The same econometric equation is estimated for both developable and agricultural land price. Data were aggregated at the level of the 589 municipalities and made available for an 11 years period (1995-2005). Data are mainly provided by the Belgian National Institute of Statistics (SPF …conomie - Direction Générale Statistiques): for all communes we know the price, surface and the number of transactions per year, for both developable land (but not yet developed) and farmland (meadow and arable land). Distance between the centroid of a commune and the centroid of the closest hub-city is the shortest road distance computed on the real road network (see Vandenbulcke et al., 2007, for more details). Econometric findings show that R2 values are slightly better for developable land (R2 = 0,92) than for farmland (R2 = 0,86), but that the partial R2 value associated to urban influence variables is better for farmland (0,85) than for developable land (0,38)3. The explanation of this paradox can be theoretical (combined effect of distance and anticipated land conversion date), as well as empirical (due to data imperfections: data are aggregated by municipalities). Results also reveal that agricultural prices decrease with distance from job centres and increase with the demographic size of the commune, with population rate of increase, households' income as well as with the contiguity to the coastline. The same variables enter the regression estimated with residential land price. Figure 1 (provisional results) shows the relationship between the slope of land price according to distance to urban centres (CBD) in the 25 urban areas, for farmland as well as developable land. Explaining farmland slope (dependent variable, y-axis) by developable land slope (explanatory variable, x-axis) indeed leads to a high determination coefficient (R2 = 0,62).