The impact of earthquake risk on the spatial pattern of house prices: A multi-level analysis

There is considerable evidence, exemplified by the work of Onder et al (2004) that explores the effects of the 1999 Marmara event, that earthquake risk has distorted spatial and temporal house price patterns in the Istanbul metropolitan area. Last yearís earthquake activity in Eastern Turkey, although not sufficiently proximous to be directly problematic in our study area, reinforced the need for policy makers, real estate agents, valuers, mortgage lenders and housebuyers within the wider geographic region to be able to accurately measure the impact of perceived earthquake risk on property values. Although typically researchers have used either hedonic or willingness to pay methods to generate such measures, both of these approaches have been subject to extensive critiques. The purpose of this paper is to consider and illustrate the potential utility of an alternative approach: the multi-level method. Multi-level models tend to be employed where the observations being examined are clustered and correlated and where the underlying causal processes operate at multiple spatial scales. The model allows us to disentangle the impact of different property attributes on values at different spatial scales. This means that, when compared with hedonics, the estimated impact of the earthquake risk measures will be quantitatively different (arguably more accurate; Costello et al, 2011) and more spatially refined. The paper applies the multi-level method to the analysis of house prices from more than 2000 dwellings in Istanbul during 2006/7. The paper concludes that the estimates are robust and offer a richer picture of spatial differences than traditional hedonic estimates.