I conduct a wide review of the literature studying the amenity value of social housing, and find little consensus between most studies. Based on this analysis I note three crucial areas for improvement: (1) Post trends in many, and maybe even most studies, have not been correctly controlled for - creating a situation where findings may be driven by neighbourhood shocks rather than legitimate treatment effects. (2) There are few studies in the literature evaluating social housing outside the United States. (3) The modelling methodology used has not changed significantly since the 1990s, with slow adoption of spatial models and alternative satellite data sources.

My study compiles a new comprehensive dataset of social housing in Christchurch (New Zealand), and uses this to estimate treatment effects across several different models. Post trends associated with social housing appear to be negative in my study area, thus incorporating post trends tends to increase treatment estimates. I also propose a novel hedonic model which incorporates machine learning principles and global satellite data. My estimates suggest that after three years, social housing tends to increase nearby surrounding house prices by between 9.1% and 14.7%. I conclude with a series of recommendations and lessons from this analysis to inform future work.