It’s surprising that they took measurements at 2000 and 2020 and did not mention the Great Recession at all, which severely reduced ability of households to get a mortgage and disrupted both demand and supply. I don’t think their attempt to control for the “kinked” supply curve by dropping the bottom 25% of metros by income growth is sufficient to eliminate metros that were devastated by the Great Recession.

Also, I am not sure whether comparing to median value of constraint index (e.g. Wharton residential land use regulation index) is a particularly good way to categorizing metros into constrained and unconstrained, since it’s hard to know which specific constraints of the “regulatory hydra” are important for future development (https://twitter.com/TheJakeSchmidt/status/157220787234744729...). One metro may need to reduce minimum lot sizes to allow the next increment of development, whereas another metro may need to allow lot mergers and relax density limits instead.

So I think this paper is basically concluding that land use constraints are harder to measure than income growth.