Major cities around the world are gentrifying as high-income newcomers drive up housing prices and displace lower-income incumbent residents. Increasing the housing supply should mitigate rent increases, but new buildings could also stimulate demand for nearby housing by improving neighborhood quality. The net impact depends on how the relative sizes of these supply and demand effects play out over space.

This paper identifies the causal impact of new construction on nearby rents, displacement, and gentrification in San Francisco by exploiting random variation in the location of new construction induced by serious building fires. I combine parcel-level data on fires and new construction with an original dataset of historic Craigslist rents and a panel of individual migration histories to test the impact of proximity to new construction.

I find that rents fall by 2% for parcels within 100m of new construction. Renters’ risk of displacement to a lower-income neighborhood falls by 17%. Both effects decay linearly to zero within 1.5km. Next, I identify a hyperlocal demand effect, with building renovations and business turnover spiking within 100m. Gentrification follows the pattern of this demand effect: parcels within 100m of new construction are 2.5 percentage points (29.5%) more likely to experience a net increase in richer residents. Affordable housing and endogenously located construction do not affect rents, displacement, or gentrification.

These findings suggest that increasing the supply of market rate housing has beneficial spillover effects for incumbent renters, reducing rents and displacement pressure while improving neighborhood quality.

[Keywords: displacement, gentrification, housing supply, spatial econometrics]

[“Amazingly, it still takes a long time to build on these burned lots—nearly 5 years to get a permit approved and 7.2 years before completion! Nevertheless, burned lots are much more likely to be redeveloped than similar unburned lots. The bottom line is that burned lots are a good as-if experiment for what would happen if a random set of lots were developed.”]

See Also: