The challenges of mapping gentrification
Cities are looking to each other for ways to identify harmful neighborhood change. But a new study cautions against a “one-size-fits-all” approach.
For all the attention paid to gentrification by urbanists — and increasingly by popular media — there’s very little consensus around the term. There’s no clear way of defining or measuring it, and there’s even wide disagreement about whether it’s good or bad. Despite conventional wisdom that gentrification is always harmful, a good deal of research finds that it doesn’t cause widespread displacement and that it can even benefit existing communities.
There’s at least one area of general agreement: cities should develop policies that mitigate any negative impacts of neighborhood change (such as evictions, displacement, or rent-burdened households) while encouraging widely beneficial forms of population growth, economic opportunity, and local investment. But to apply those policies effectively, cities need to map where harmful change is happening — and according to new research, that’s yet another case of dramatic divergence.
The study — published last week in the journal Urban Studies — comes courtesy of an MIT research team led by Benjamin Preis. The team gathered four recent maps of gentrification and displacement in major U.S. cities: Los Angeles, Philadelphia, Portland, and Seattle. Critically, each map had been developed based on very different measures of gentrification (more on that below), reflecting the general lack of consensus in this area.
Then Preis and company applied those four underlying map methods to the same city (Boston). The result was striking. There was very little overlap in terms of gentrification areas, with only seven common census tracts (out of 180 tracts in all of Boston) marked as gentrifying or “at risk” of gentrifying across the four map methods. There was also a very wide range of map coverage: the most conservative map method identified 25 at-risk tracts, while the most lenient identified 119.
The authors caution that cities could reach “very different conclusions about the location and severity of gentrification based on the method they choose.” They add:
From different variable choices to varying risk thresholds, the assumptions embedded within the methods have significant effects on what tracts are identified as vulnerable and, in turn, where city policy responses would be targeted if these methods were used.
Let’s take a closer look into why that matters.
Cities copy other cities
The four underlying maps chosen by Preis and team shared some important characteristics. The biggest was that all four had been developed by (or on behalf of) local government. That official status makes them arguably more likely to directly inform policy interventions than, say, maps produced by academics or community organizations.
Another important factor: as recent precedents in the gentrification mapping space, these maps (and the methods underlying them) would seem to be good candidates for other cities to copy. Indeed, at the time of the MIT study, Denver had directly adopted Portland’s mapping approach, while Boston had planned to adopt Seattle’s method.
That’s not too surprising. Cities share insights and innovations all the time, from ride-hail regulations to bike-share systems to High Line knockoffs. But given the global concern around gentrification, and the high likelihood that yet more cities will be looking for ways of mapping the trend, it’s essential to know whether one method is replicable in another place.
In other words, if different mapping methods lead to wildly different conclusions about the geography of gentrification — as this study suggests they do — then any new map based on those methods might not be reliable. That could lead to inefficient or ineffective policy interventions, channeling limited resources to the wrong places.
At the core of this map gap is the selection of factors used to measure harmful neighborhood change. In the case of the four maps reviewed for this study, these factors included measures of non-white, college-educated, and rent-burdened populations; statistics around average household income, rental cost, and housing prices; and the ratio of low-to-high income households.
But the specific factors varied widely by map method, based largely on what each effort determined to be a clear cause or sign of gentrification. As the researchers note:
The variables included in or excluded from the different methods relate at the most fundamental level to their authors’ decisions about the most salient causes and indicators of gentrification, highlighting the degree to which they understand gentrification to be driven by private investment, rent gaps, state-led public investment or changing consumer preferences towards city living, to give a few examples.
I’ve shared a bit more on each method in the bullets below, but it’s worth noting that far more detail is available in the full paper. (It’s also worth noting that the study itself didn’t aim to evaluate the validity of each method.)
- Los Angeles. The city’s “innovation team” developed an Index of Displacement Pressure that mapped neighborhoods by two scores: the first used six factors to identify neighborhood change, and the second used seven different factors to identify displacement pressures. Key factors included rental and housing costs, race, availability of affordable housing, and proximity to public transit.
- Philadelphia. The Philadelphia Federal Reserve produced a study of neighborhood gentrification that included a section meant to guide practitioners and city officials. The approach focused strictly on the rent gap, excluding factors such as race as well as public investments such as transit, and instead focusing on changes in household income, education attainment, and housing (or rental) costs.
- Portland. The city commissioned a Gentrification and Displacement Study developed by Portland State housing scholar Lisa Bates. The approach focused on housing markets and private investment (factors included housing price, proximity to affluent areas, and household income, but not public transit), laddering up to three key dimensions: vulnerability to displacement, demographic change, and housing market appreciation.
- Seattle. The city’s 2035 Comprehensive Plan mapped potential gentrification and displacement based on 14 variables, mapping a total “Displacement Risk” score. The approach focused on rent price and development potential, with key variables including measurements of income, race, and rent-burden, as well as a very wide range of neighborhood amenities, such as transit, businesses, and job proximity.
Different measures, different maps
If the different mapping approaches were equally reliable, then the four Boston maps should have identified more or less the same neighborhoods as being at risk of (or already experiencing) gentrification. Suffice it to say, the four maps were not more or less the same, with the four methods producing what the researchers call “very different maps of gentrification-related displacement risk.” That much is clear from maps of the four methods shown side by side:
The mapping approaches used by Portland and Philadelphia resulted in the most conservative indications of Boston gentrification, with many areas labeled “not at risk” or “not gentrifiable.” By contrast, the Seattle mapping approach was driven by less restrictive factors, and thus its Boston map has zero “no risk” areas and few “low risk” ones. Los Angeles’s approach results in a map somewhere in between, although statistically closer to the Seattle map in nature.
Dig deeper and the map gap gets even wider. For the Portland, L.A., and Philadelphia approaches, at least 40 percent of their at-risk tracts are not considered to be at-risk by a different method. As mentioned earlier, the range of “at risk” Boston census tracts was huge, from 25 (Philadelphia method) to 119 (Seattle method).
Philly and Seattle represent the most glaring departures. The Philly approach (which didn’t focus on race) was unique in excluding most of Dorchester, a heavily African American area, while the Seattle approach (which included public amenities) was unique in including most of Allston and Brighton, where areas with major plans for rezoning and transit improvement. Philly’s at-risk tracts covered 16 percent of the total population and 14 percent of the non-white population, whereas Seattle covered 73 percent total and 80 percent non-white.
There was one area where the models agreed: all had a disproportionately greater share of evictions in at-risk areas than total households in these areas. But even that differed dramatically in degree, if not kind. The Philly map covered just 16 percent of evictions; the Seattle map covered a whopping 88 percent; Portland and L.A. covered 32 and 43 percent, respectively.
As mentioned, there were only seven Boston tracts that all four methods agreed were at high risk of gentrification or displacement. To be sure, these areas included many neighborhoods that locals usually consider “at risk”: northern Dorchester, northern Roxbury, Jamaica Plain, Chinatown, and East Boston. But the researchers caution against the idea that combining all four methods will produce a “true” picture of at-risk neighborhoods, given the vast range of inputs and outputs.
They conclude:
[T]his research illuminates for advocates and policymakers that there is no “one-size-fits-all” method for mapping gentrification and displacement risk; practitioners must ensure that the methodology they use fits the temporal, spatial and socioeconomic context of the city at hand.
Moving forward
The research shows that different ways of measuring gentrification can lead to dramatically different ways of mapping it, and thus different answers to the seemingly basic questions: is it happening, and if so, where?
It is certainly not news to find that different data inputs create different analysis outputs. But again, the fact that other cities will no doubt adopt existing approaches creates significant policy implications for choosing one method over another — or, for that matter, choosing to develop a new one entirely. It’s quite possible that the root causes of neighborhood change are so place-based, driven by unique historical factors and social values, that no one city can adopt another’s mapping method.
The study authors seem to support this hyper-localized approach:
To the extent that the causes of gentrification are informed by local historical and contextual factors, there is reason to question the wholesale adoption of models developed within academia or the adoption of a model used in one city for another city, without attempting to account for the particularities of gentrification experienced in each individual city, or at least consciously choosing between different methodologies and the theories that they operationalise.
That said, there does seem to be an opportunity for better data to improve the measurements themselves, even as applied to local context. For example, the four maps explored here didn’t include eviction data, nor do they seem to have considered data on people leaving the neighborhood (often called out-migration). More importantly, they don’t seem to have compared out-migration across neighborhoods, which can show whether displacement is the result of acute neighborhood change or whether people are broadly leaving a city for other reasons.
And none of these measures address the troubling trend of “displacement by decline” — the fact that concentrated poverty and lack of investment in a neighborhood can be just as harmful, if not more, than gentrification.
It doesn’t help that, even if you define gentrification carefully, it remains very difficult to measure. As urban scholar Lance Freeman told me a few years back, displacement is — by definition — the absence of someone, which means it’s hard to detect this change (and why it occurred) unless you follow specific populations very closely over a long period of time. As such retrospective analysis improves, cities should be able to develop stronger models for predicting harmful change.
That’s a long view, but there’s still hope for right now. We can make progress on certain pressures that exist no matter how you map gentrification, such as reforming zoning rules that restrict dense new development, or creating more housing options for all incomes. It now costs $750,000 to produce a single unit of affordable housing in San Francisco; lowering the cost of housing production is possible in the near future— and something technology is particularly well suited to address.
There may not be a single roadmap for positive neighborhood change, but there is a way forward.
Follow Sidewalk Labs with our weekly newsletter and our podcast, “City of the Future.”
