Gentrification is an ambiguous term, which roughly speaking means the replacement of poor residents in a community by the rich, and a related change in the character of the community and its amenities. There are two broad mechanisms for gentrification – displacement and succession. Displacement is where the influx of rich residents actually increases the likelihood that poor residents move away (e.g. due to higher housing costs). Succession implies that rich households simply move in after poor residents that moved away for other reasons.
The distinction is important because displacement implies gentrification may be harmful whereas succession implies that it is a more benign process. My latest CEP Urban and Spatial Programme discussion paper is the first study to provide empirical evidence that gentrification involves displacement of poor residents. While it’s true that several studies look at the question already, none of them find any evidence of displacement. Instead, these studies suggest that gentrification occurs through succession.
Displacement studies usually combine two types of data. Firstly, studies use data on the proportion of higher socioeconomic class households living in a neighbourhood (e.g. based on a Census). Neighbourhoods are then characterised as gentrifying or not according to whether there was a large increase in the share of high socioeconomic class residents over say ten years. Secondly, studies use data from longitudinal household surveys. Such datasets allow researchers to track individual households across all the different neighbourhoods they live in over the years. The usual approach is to link these data together in order to examine whether living in a gentrifying neighbourhood means households are more likely to move away. Previous studies find that poor households living in a neighbourhood characterised as gentrifying are no more likely to move away than poor households living in non-gentrifying neighbourhoods. This is interpreted as evidence that gentrification occurs through succession rather than displacement.
In my paper, however, I argue that previous estimates may be biased by the fact that different types of household (with different natural mobility rates) tend to live in different types of neighbourhood. This well documented phenomenon is called ‘sorting’ and means that previous studies might miss actual displacement. My approach makes use of year-to-year variation in winter temperatures in Great Britain. I argue that if displacement does happen, then it will be more pronounced in years with colder winters. The reason is that households will be less able to withstand rising rents resulting from gentrification if budgets are already stretched by higher fuel bills. This novel approach reveals a ‘causal’ effect because the type of household living in gentrifying neighbourhoods does not differ in cold years.
I use data from the UK Census to compute a measure of gentrification for every neighbourhood in Great Britain over two periods: the 1990s and the 2000s. Neighbourhoods are defined as gentrifying if there is an above-average increase in the share of residents with a university degree. Figure 1 illustrates my gentrification measure for London neighbourhoods in the 1990s (TTWA is the London Travel to Work Area, MSOAs are small census areas). Gentrification in the region, according to this definition, is evidently concentrated towards inner London but there are pockets elsewhere. I use this gentrification measure to estimate displacement effects for a sample of low-income private renter households from the British Household Panel Survey (BHPS). The BHPS is a survey of households that has been following a large sample of households since the 1990s, and so allows me to track which households move and when.
Figure 1: Gentrification index for London in the 1990s
I find that that gentrification does displace low-income households. In fact, my estimates show that you need to have a household income of more than 1.5 times the average for the city and year to have no chance of being displaced. My findings also indicate that displacement may be avoided if gentrification occurs slowly enough. Figure 2 illustrates the size of displacement effect (left axis) relative to the speed of gentrification (bottom axis). The figure shows that there are no significant displacement effects resulting from small increases (or decreases) in neighbourhood degree share, i.e. a slow pace of gentrification. Households only start to be displaced when the degree share increases by 10 percentage points more than average (which equates to 0.1 on the bottom axis). These findings suggest a need to rethink gentrification and its consequences.
Figure 2: Displacement effects at different levels of gentrification
A lot of place-based policies aim to encourage ‘mixed communities’ on the grounds of it being beneficial for existing low-income residents. While the evidence on whether mixed communities help is inconclusive, my findings suggest that such policies may end up displacing original residents altogether. If policymakers wish to improve outcomes for low-income private renters, it may be more effective to target housing assistance to households living in already gentrifying neighbourhoods.