In Homelessness is a Housing Problem, Prof. Gregg Colburn and data scientist Clayton Page Aldern seek to answer the question: why is homelessness much more common in some cities than in others?
They find that only two factors are significant: 1) overall rents and 2) rental vacancy rates. Where housing is scarce and rents are high, lots of people are homeless. Where rents are lower, fewer people are homeless, even in very poor places. (In fact, high city incomes correlate positively with homelessness, because more and better jobs lead to higher demand for housing).
By contrast, many other factors that one might think are related to homelessness in fact are not correlated on a citywide basis. For example, since homeless people are generally poor, one might think that places with high poverty rates or high unemployment rates have lots of homelessness. The authors show that this is not the case. Where most people are poor, there is less demand for housing, which translates into lower rents and less homelessness.
One might also think that places with warm weather have lots of homelessness, because homeless people might be attracted to them. But high-rent cold cities like Boston have above-average levels of homelessness, while cheaper warm-weather cities like Orlando and Charlotte do not. However, homeless people are more likely to have temporary shelter in cold cities than in expensive warm-weather cities like San Diego- either because city governments are less motivated to build homeless shelters when no one is at risk of freezing to death, or because the homeless themselves are less eager to use shelters. I suspect that if the authors focused only on highly visible unsheltered homelessness, they might have found a stronger correlation with weather).
It might be argued that shelters themselves (or other social services) attract the destitute. However, the authors find that “a region’s proportion of in-migrants with incomes below the federal poverty line … is entirely unrelated, statistically speaker, to per capita rates of homelessness.”
What about drug use and mental illness? The authors were unable to unearth any city-level data on the frequency of either- but state-level data do not show a strong correlation between the amount of drug use or mental illness and homelessness rates. I didn’t find this surprising, because even if half of homeless people are mentally ill and/or addicted, even some people in these categories might be functional enough that they could find cheap housing if such housing existed. (Having said that, it seems to me possible that unsheltered homeless people might be more severely impaired- so I wonder if the authors would find similar results if they focused on the number of unsheltered homeless).
The last chapter, on policy responses, mentions in passing that reducing zoning regulations might increase housing supply. However, the authors are more focused on infusing money into local governments to support lower-income housing. As a result, their treatment of land use regulation is a bit shallow; they criticize single-family zoning, but housing supply is constrained by a much wider variety of regulations. For example, even in areas zoned for multifamily housing, government limits housing supply by limiting the number of units that can be built on a parcel and requiring land to be used for parking that could otherwise be used for more housing.