“Urban renewal” has been a dominant theme of the past 60 years, and not only in the US. The basic idea is that housing values depend on the value of nearby houses, in that poorly-kept houses lower nearby house values, and well-groomed lawns do the opposite. That is, the quality of housing stock has a large externality component. Of course, the magnitude of this effect is very difficult to measure. If I renovate my home, and you – the analyst – see house prices increase in my neighborhood, there is a ton of endogeneity; perhaps I renovated because there was some secular gentrifying trend within my neighborhood already present.
The authors of this recent JPE use a unique dataset to get around this problem. In the 1990s, the city of Richmond, Virginia gave grant money to CDCs to renovate housing stock in four poor neighborhoods in that city; millions of dollars in each neighborhood were given. Of note, a fifth neighborhood was considered and left out of the program because it was in the same city council district as a neighborhood already chosen. Using over 100,000 residential house sales, the authors strip housing prices down to land value using a semiparametric technique, then investigate what happens to land prices in the city after the program ends. The result is striking: home prices very close the the center of CDC work increases 6-10% faster per year than elsewhere in the city, the effect was noticeable at a distance of half a mile. Discounting future tax revenue, the housing externalities in run-down neighborhoods near the city center were so strong as to almost pay for the renovation themselves.
(I particularly like this paper because I spent a good chunk of time when I was at the Richmond Fed working with this data. The takeaway for my own future work is that nonparametric estimation, when done smartly, can be used even with very large datasets, and that urban data (land value, crime rates, etc.) are far too “lumpy” to be captured with the simple econometric techniques often deployed in urban econ. A paper with one of the authors of this JPE that was published in a Fed journal a few years ago shows this quite plainly when it comes to land values in a metro area.
http://126.96.36.199/research/economists/bios/pdfs/sarte_housing_externalities_revisedpaper.pdf (like to NBER WP version. Final version published in JPE 118.3)