Run a Google search for “gentrification” and you’ll get thousands of news items and scholarly articles on how urban revitalization risks pushing low-income communities out of cities. Despite all that research, Ken Steif, director of the Urban Spatial Analytics grad program at the University of Pennsylvania, says it’s remarkable how cities are still struggling to stop displacement when investment starts to rise.
On Jan. 31, he debuted a model that may help. It provides policymakers with a rough glimpse into the future by relying heavily on U.S. Census data to predict gentrification. The initial framework, which was rolled out through his consulting firm Urban Spatial, was applied to 29 legacy cities mostly clustered around the Northeast, but also included urban areas like Chicago, Birmingham, New Orleans and St. Louis.
Along with researchers Alan Mallach, Michael Fichman and Simon Kassel, Steif blended datasets of median housing prices for 3,991 census tracts in these cities during 1990, 2000 and 2010. They also recorded the average housing prices for groups of census tracts surrounding each census tract, and looked at other variables like changes in resident income, to measure the incline of housing values in a given region.
If they could plug data from 1990 and 2000 into this model and get a prediction for 2010 that lined up with what housing prices actually looked like in 2010, the experiment would’ve proven successful. And that’s sort of what happened. The margins of error in Cincinnati, Baltimore and Pittsburgh, for example, rested between 14 percent and 12 percent. At the lowest end of the spectrum, South Bend, Indiana, and Erie, Pennsylvania, were both under 5 percent.
What researchers did find notable was that there weren’t any distinct patterns of error — for example, larger cities on average didn’t show different proportions of error than smaller cities. “The model is not biased toward smaller cities or larger ones or those with booming economies,” note the authors, concluding that “this is evidence that our final model is generalizable to a variety of urban contexts.”
It’s an exciting addition to the debate on how to slow gentrification’s negative influence, with an eye towards the power it could bring to community development financial institutions, nonprofits or policymakers who want to channel city growth into a more equitable, accessible future. “If developers can commonly access data like this, if these sorts of analytics can be democratized, it puts us all on the same page,” says Steif. To him, that could directly translate to outlining a game plan for transit improvements based on neighborhood fluctuations among low-income populations dependent on the bus. Or, organizers using the model to help frame their community benefit agreements with major developers looking to move into a neighborhood.
But it’s not the first time researchers have attempted to peer into future growth patterns. In a 2016 article published at Cityscape, the U.S. Department of Housing and Urban Development’s research journal, authors Karen Chapple and Miriam Zuk from the University of California, Berkeley point out that policymakers have been consulting “early warning systems” to scope out neighborhood shifts since at least the 1980s.
Some of these systems were extra diligent in their approach, compiling data on everything from building permits to condominium conversions to mortgage lending characteristics within the region in question. That type of thoroughness can burden a prediction model, though, if its source information isn’t easily collectable on a year-to-year basis, and you’re a community group with barely enough resources to research local poverty in its present form.
Then there’s the issue of whether or not that data communicates a clear message to politicians or nonprofits who aren’t versed in statistics or the sciences. “I can develop stat models left and right, but if nobody can [read] them, it’s pointless,” says Steif, suggesting why past models rarely led to successful preventive action on the ground.
Efficiency also has its risks. James Jennings, a researcher at Tufts University who’s investigated gentrification trends in Boston for decades, says on top of relying on official sources — and presenting their implications in a way that’s understandable by all tiers of the community — researchers need to speak with the people in the neighborhoods they’re forecasting and collect their observations as another form of data.
“A lot of the literature on this completely leaves out the voices of people feeling the angst of gentrification — the ones actually being displaced,” says Jennings. “It’s those local voices that help to highlight nuances and subtleties and little pieces of information that give us a better sense of how our city enables change.”
That piece of research by Chapple and Zuk buttresses Jennings’ observation. In the Cityscape article, they point out how an empty lot, or crippled building, can appear to outside observation as a sign of disinvestment. Sometimes it’s just the case that the landowner is sitting on that property and waiting for the value to increase before rehabbing it — the type of information you could get from residents next door if the data isn’t easily available from the city planning department.
The goal for Steif now is to partner up with funders or other think tanks to reformat this model to additional sets of data culled from the city level. He says there’s at least one local government that’s already reached out to him to talk about adopting his model, which he describes as “using data to make improvements in the lives of people.”
“That is the overall goal,” he adds.
The Equity Factor is made possible with the support of the Surdna Foundation.