A new study has run the numbers, and has concluded that social welfare is optimized by putting affordable housing in very poor neighborhoods, rather than wealthier (and especially whiter) ones.

Authored by Rebecca Diamond and Timothy McQuade of the Stanford School of Business, the study really has two major conclusions. First, building affordable housing in very low-income neighborhoods creates major benefits for the surrounding area. Second, there are major social costs to placing affordable housing developments in higher-income neighborhoods, though they calculate that these costs are outweighed by the income benefits to the affordable housing residents.

While there are some valuable findings here, as you might imagine, we have a few issues.

For one, the kind of affordable housing the study looks at is targeted to people with incomes that are 50 percent or more higher than is typical in low-income neighborhoods. The paper looks at developments funded by the Low Income Housing Tax Credit, or LIHTC. LIHTC buildings generally target households making 60 percent of Area Median Income; the study’s authors estimate that in their sample, that averages about $40,000. By contrast, median income in the low-income neighborhoods they find benefit the most from new LIHTC buildings top out at about $26,000. And since $26,000 is the high end, most of these neighborhoods are actually poorer than that.

In other words, they’re talking about neighborhoods so poor that building low-income housing increases the community’s average income. That makes their finding that LIHTC projects increase housing prices by about 6 percent within 0.1 miles (yes, you read that right—more on the geographic range of the effects later) much less surprising. It also means that the finding is less about “affordable housing” per se and more specific to LIHTC, or other subsidies with similar income targets. It’s questionable whether the same results would hold for other kinds of subsidized housing targeted to lower-income households.

Second, it’s worth taking a second to underline exactly what the “costs” of LIHTC buildings are to higher-income neighborhoods. Diamond and McQuade find that LIHTC buildings don’t increase crime. And yet they also find that the average homeowner in such a neighborhood would pay nearly $4,000 to avoid having to live within 0.1 miles of a LIHTC building. But note that we said “homeowner”: renters appear to have no such preference. Even more curiously, this aversion to low-income housing only appears in higher-income neighborhoods with low Black and Latino populations.

What would create such a pattern? The authors have an idea. “If local residents have preferences over the demographics of their neighbors,” they write, “new in-migrants could make the neighborhood more or less desirable.” This may be the world’s politest way of saying “mostly white homeowners appear to be discriminating against Blacks, Latinos, and/or poor people.”

Now, that’s not necessarily a surprise: it confirms many years of research about how racism and the perception of the presence of lower-income people affect housing markets. But it raises a question that anyone in housing policy or urban planning needs to be able to answer: are preferences of advantaged groups for segregation—segregation that we know is harmful for lower-income people and people of color—just another legitimate interest that we need to weigh against the interests others might have in desegregation? As it happens, the authors estimated the gains of integration, and found that they outweighed the costs. But there’s no reason the numbers had to work out that way. If the model’s results had shown that the benefits of segregation to mostly white, mostly higher-income homeowners were greater than the costs to disproportionately Black and Latino lower-income households, would that mean they would have come out in favor of segregation?

Finally, the way the authors do try to quantify the benefits of integration is extremely limited. Their estimates are based on Raj Chetty et al’s findings about the increase in average lifetime earnings for low-income households in higher-income neighborhoods. Of course, we’re big fans of Chetty’s work, and we’ve cited it ourselves extensively. But it’s a huge mistake not to include other potential benefits in a cost-benefit analysis. Other studies, for example, have shown major improvements in mental health; you might also expect better educational outcomes, which arguably have value beyond simply their contribution to future income. There’s also the reduced likelihood of crime victimization; potentially shorter commutes; and so on. None of these are weighed in when the authors conclude that the benefits of building LIHTC in high-income areas are exceeded by the benefits of building in the very low-income neighborhoods we talked about earlier.

So what should we take away from all of this?

  1. Building LIHTC units in very poor neighborhoods may, in fact, be a kind of place-based development strategy with some payoffs, as reflected by rising home prices. But it’s not clear how far it goes as a broad strategy for revitalizing these neighborhoods. For one thing, the strongest gains are in a very small area—just 0.1 miles from the project—with quickly declining improvements beyond that. Moreover, just as their estimates of the benefits of integration are limited, so are their estimates of the costs of segregation. While they do find that LIHTC projects help lower crime, it’s not clear whether there are improvements on other indicators that residents are likely to care about beyond home prices: schools, local retail options, and so on.
  2. Are all preferences made equal? We can use home prices to quantify the preferences of homeowners—but that doesn’t mean we should weigh every kind of preference in the same way. It turns out that for many people, the presence of people of color or lower-income people is enough to cause them to value their homes less. Evaluating policy options always involves value judgments, and econometric models—while often helpful—are not a substitute.
  3. When we do use econometric models, we need to be aware of what’s being left out. This, in fact, is part of our value judgments, whether we’re aware of it or not. Do we value better mental health for low-income people? Do we value education beyond its income effects? Do we value giving people the option to live somewhere they otherwise couldn’t? If none of those are in the model, then we are effectively answering “no.”
  4. Leaving integration up to local governments is unlikely to be successful. This is a point we’ve made before, most recently, perhaps, in citing research by Michael Lens and Paavo Monkkonen that showed that metropolitan areas with more local power in development decisions are more segregated than ones where states play a bigger role. This study underlines that while many people in all sorts of neighborhoods value diversity and integration, some do not, and they are willing to pay thousands of dollars to avoid having low-income neighbors. (This may also be about the kind of “prisoner’s dilemma” of who might get “stuck with more than their fair share” of low-income housing. In fact, research suggests that when integration is widespread, the dynamics of neighborhood change are altered in ways that reduce the incentive for self-segregation fo the advantaged.) When those preferences are combined with hyper-local power over what kinds of housing gets built where, it’s inevitable that many jurisdictions and neighborhoods will create regulatory barriers to low-income people living in their communities: in other words, exclusionary zoning.