Fundamental Flaws with "Homelessness is a Housing Problem"

Let's dissect their arguments

Since its publication in 2022, the book “Homelessness is a Housing Problem” has become mainstream dogma. The authors list individual factors commonly blamed for homelessness (e.g., mental illness, drug abuse, poverty) and show that these are not correlated with area homelessness rates. They also look at area characteristics (e.g., weather, welfare benefits, political leaning) and purport to show that these are also poorly correlated with local homelessness rates. They found that housing prices were positively correlated with homelessness rates and concluded, therefore, that housing costs are the main cause of homelessness. Noahpinion has a good guest blog detailing its main arguments.

There are several problems with their analysis, which I’ll detail below.

The homeless are mobile

The authors look at possible causes of homelessness in individuals, calculate whether the homelessness rate is correlated with the prevalence of that possible cause in a given area, then claim that the lack of a correlation rules out the possible cause.

But individual risk factors increase the likelihood of an individual becoming homeless. It does not follow that areas with greater proportions of at-risk individuals will have higher rates of homelessness. For that to be true, we need an added assumption: that the homeless do not move. If the homeless are mobile, there is no reason to expect a correlation.

The authors themselves admit “Public and charitable supports – from employment centers to food banks and emergency shelters–are more frequently provided in cities, for example, than in suburban settings. Accordingly, people seeking these services are more likely to travel to cities, sometimes permanently.”

There is ample evidence for the mobility of homeless populations. Many jurisdictions offer free bus tickets to the homeless. Rawson-Neal Psychiatric Hospital in Las Vegas discharged over 1500 of its patients with one-way bus tickets out of state to destinations like San Francisco. At most 27% of San Francisco homeless rented or owned in the city just prior to becoming homeless, indicating that up to 73% moved from elsewhere. 35% of the homeless in Los Angeles lost their housing elsewhere, as did 35% of Austin’s homeless population. Given that the homeless do move to different cities, the dismissal of individual factors (e.g. mental illness, poverty, drug abuse) is invalid.

Poorly measured policies

To discount the idea that attractive policies may draw homeless people from out of town, the authors compare different locales based on a ratio of Temporary Assistance for Needy Families (TANF) benefits to median rent on a two-bedroom apartment.

The chosen metric does a poor job of evaluating the attractiveness of local homeless benefits given that:

1. TANF applies only to families with children, which represent just 30% of all homelessness

2. TANF benefits are set at the state level, not city or county

3. Area median rent is irrelevant since the homeless do not pay rent

A serious attempt to analyze the relationship between public benefits and homelessness rates would look instead at a city’s annual homelessness budget per homeless person, which would give an idea of the resources available. Or, it could compare the dollar amount in cash and food benefits each homeless person receives on average. Both of these alternatives take into account local variance in benefits, whereas the authors’ chosen metric does not.

The authors also examine low-income migration and find that it’s negatively correlated with local homelessness rates. But this is also irrelevant: having low-income does not qualify an individual for homeless services or cash benefits, so this will not tell us whether the homeless are attracted by these policies! Why not examine the migration patterns of homeless people to be more exact?

Policies, not political parties

“If progressive policies were the main driver of homelessness, wouldn’t progressive cities have similarly high levels of homelessness?” is too simplistic. A thorough assessment requires an examination of actual policies, not just party preference. It’s not enough to say “this city voted for Joe Biden” or “the mayor is a democrat” as was done in the book.

An honest assessment of policy would — at minimum — compare the amount of free cash offered, the laxness of local laws and law enforcement regarding public drug use, theft, peddling of stolen goods and sidewalk camping. But none of these policies were examined or even mentioned.

One particularly glaring omission is the Ninth Circuit’s Martin v. Boise ruling, which effectively prohibits enforcement against public camping if a city doesn’t first offer a shelter bed. The authors address the ruling in just a few lines, but only to moralize that we must find a solution to homelessness other than criminalization. They failed to acknowledge that as of 2022, seven of the nine states in the Ninth Circuit are also in the top ten states with the highest homelessness rates:

Given that the Ninth Circuit contains <20% of all states and but comprises 70% of the states with the highest homelessness rates, it seems disingenuous to deny how policy can affect homelessness.

Data interpretation issues

The book’s thesis is that high housing costs are the primary driver of homelessness. Their logic is that high housing costs mean fewer people can afford housing — thus, they live on the streets instead.

But high housing costs are only part of the equation. Housing affordability is more relevant in whether someone becomes homeless. Affordability depends on both housing costs and income. If housing affordability is supposed to be the underlying cause of homelessness (and not merely correlated), then some of the data should have given the authors immediate pause. For example, they found :

• an inverse relationship between poverty and homelessness rates

• an inverse relationship between unemployment and homelessness rates

• weather explains 60% of the variation in unsheltered homelessness in cities — compare with median rent and rent burden (which explain 55% and 13% of the variation in general homelessness, respectively)

• absolute rent accounts for more than 4x the variance in homelessness that rent burden does

This last point is important. Is a person more likely to become homeless simply because the absolute rent is higher? Consider someone who pays $2000/month in rent but makes $20,000/month. Is that person more likely to end up on the streets than someone who pays $1500/month in rent but only earns $3000/month? Which person is at higher risk of becoming homeless due to an unexpected emergency expense or job loss? The authors hand-wave this away, claiming “Minimum wage labor, public assistance, and support from family and friends can be enough to help you get by.” I found this entirely unconvincing.

Taken together, these relationships point to a hidden variable that’s correlated with high absolute rent, low unemployment and low poverty rates.

Questionable assumptions => questionable conclusions

Is housing price correlated with homelessness rate? Sure. But so is rent control. Homelessness is directly correlated with median income. And inversely correlated with unemployment. Yet, I doubt anyone would argue that we should eliminate rent control, lower incomes and raise the unemployment rate to solve homelessness. The authors would undoubtedly claim these correlations do not indicate causation, or that there must be a hidden variable. Then why are they so eager to conclude that housing price is the primary driver of homelessness? It’s the same faulty logic.

The obvious next step in their analysis would be examining changes in housing prices over time and observing whether these have the predicted effect on area homelessness rates. Without such analysis, they cannot claim causality. And there’s ample evidence from their own book to call such a claim into question.

Why else might there be a correlation between housing cost and homelessness? High housing costs can only exist where local incomes are also high. In large cities and counties, high incomes mean the local government has money, and is likely spending generously on “the homeless problem” — making the city ever more attractive to homeless individuals.

To test my hypothesis, we need only examine cities with high housing costs that do not provide generous benefits and services to the homeless. These cities are conveniently excluded from the book’s analysis, which covers Continuums of Care that represent 30 of the largest cities and counties in the country. Consider small wealthy enclaves with the most expensive housing in the country but few homeless services. For example, in 2022, the homelessness rate in Newport Beach was 0.11% — same as Missouri, which is considered to be low.

As I’ve mentioned in a previous post, if affordable housing were the main goal of homeless individuals, the percent of homeless people who arrive in San Francisco after becoming homeless elsewhere would be zero. Not 30%, not 30 people. Zero. That someone would become homeless then choose to move to or remain in any of the world’s most expensive real estate markets (e.g., New York City, San Francisco, Los Angeles) means whatever caused that individual to become homeless will not be solved with cheaper housing because being housed is not their primary goal. If it were, they would eagerly move to where housing is already cheap.

Does housing cost really drive homelessness? If so, why would someone serious about policy solutions only suggest creating more affordable housing in expensive markets? Cheaper housing is available throughout the country, today. Just one bus ride away, yet the authors never once suggest relocation as a solution.

Perhaps even they don’t truly believe their thesis.