Spatial patterns of land-use and neighbourhood diversity: a multilevel analysis of residential land prices in Luxembourg

There is increasing interest in understanding how the spatial organisation of land uses within a neighbourhood impact on the perceived quality of a residential place. Not only this diversity may lead to higher attractiveness but also is usually seen as a key aspect of sustainable urban growth (mixed use development). Geoghegan et al. (1997) assume that increasing land use diversity might affect property values in two ways: negatively as they introduce higher chances of negative visual and noise externalities, but in the meantime positively as diversity may implicitly mean the proximity to important local urban amenities. Our paper aims at revealing the role of land-use diversity in determining the price of residential land, hence the attractiveness of a location beyond the structural qualities of houses.
We conduct a hedonic price analysis of all residential land transactions across the Grand Duchy of Luxembourg between 2007 and 2011. Land-use diversity is measured by the Shannon diversity index. In addition to these unconventional geographical measures, we rely also on a multilevel modelling approach, which is still quite rare in land and housing hedonic literature (notable exceptions are Orford (2000), and Chasco and Le Gallo (2012)). We believe the multi-level approach is needed here to account for the nested nature of the data and to relax the assumption of a unitary equilibrium land market. As an alternative to the single-level model, the multilevel model accounts for the hierarchical structure of the spatial units, by modelling the variability at each of the considered levels and allowing individual observations within a particular spatial unit to be more similar than a random sample (Jones, 1991). Orford (2000) in particular emphasises the capacity of multi-level models to deal with spatial segmentations of the market and spatial dependence effects.
In our paper we first identify the variability of transaction price at the different levels compared to the overall mean. Second, we check for spatial variations in the valuation of land-use diversity and parcel size in the residential land market. Third we test for remaining spatial effects via a cross-regressive multilevel model, as suggested by Chasco and Le Gallo (2012).
Our results confirm the usefulness of the multilevel model approach and a negative valuation of close land-use diversity, whereas it is considered a positive externality in walking distance. Further, random coefficients for the Shannon indices confirm spatial variations in the valuation of land-use diversity. Via the cross-regressive multilevel model we test for remaining spatial effects and conclude that our model cleared up the entire spatial dependence in the land price data, conversely to the case in Chasco and Le Gallo (2012) and thus rather supporting suggestions from Orford (2000).