Exploring changes in risk perception through house price differentials following a natural disaster: a case study of the Christchurch earthquakes of 2010 and 2011

<p>This study examines the impacts of the devastating 2010-2011 Christchurch earthquakes on the local housing market. Specifically, the focus of this study is the pre-existing and new land hazard zones and their impact on consumers pricing of residential real estate.</p> <p>This case study is unique because it provides a quasi-natural experiment and a robust means of detecting changing perceptions of risk through house price differentials.</p> <p>A mixed methods approach has been applied to systematically assess the impacts of the earthquakes and new land hazard zones. The first method is a Difference in Difference (DID) Ordinary Least Squares (OLS) regression model using cross-sectional sales transaction data spanning the period February 2007 to October 2012. Spatial data are added to this dataset which includes the new liquefaction Technical Category (TC) zones and their pre-earthquake equivalent zones. The first DID regression estimate uses a sample of 4,898 sales transactions, where the land hazard zones remained consistent before and after the earthquake events. From the analysis of results, it is evident that consumers ignored or downplayed the land hazard risk zones prior to the earthquakes since there is no consistent association between house prices and pre-earthquake land hazard risk zones. However, this changed abruptly after the February 2011 earthquake which led to house price premiums in the order of 15.1-18.8% for safer land areas compared to high risk (TC3) zoned land, ceteris paribus. The second estimation using a larger sample of 20,377 sales transactions yielded inconsistent results, with respect to land combinations that changed. This inconsistency is likely to do with the staged approach to remapping of liquefaction after February 2011. Another important finding is that the traditional predictors of price such as land area and floor area had diminished relevance after February 2011.</p> <p>A cluster analysis was employed on the smaller, consistently zoned dataset. This method showed that the mean price effects from February 2011 provide support for the findings from the hedonic OLS modelling.</p> <p>A web-based survey was also conducted to examine behavioural aspects of risk perception. The survey was designed to trace the associations between risk perception and demographic variables such as gender, age, education level and religion. Only the association between gender and the perception of risk associated with widespread flooding was found. The survey response rate was low and even though 4,100 fliers were distributed, the results are treated with caution.</p> <p>An analysis of census data was employed to examine the change in population within statistical Area Units (AU) between 2006 and 2013 Censuses. Another OLS regression is used to explain changes in population using demographic data and TC zones as independent variables. As with the survey, no demographic variables helped explain population changes. Only the red land zone explained population changes; a 1% increase in the proportion of red zone land within an AU explained a loss of 27.06 people on average, ceteris paribus.</p>