Does gentrification push people out? The answer depends on how it is measured

Gentrification is one of those hot-button topics where people can look at the same city and see opposite truths. Some see disinvested neighborhoods finally getting access to grocery stores, improvements in public transit and safer neighborhoods while others see decreased affordability, the loss of cultural identity and the exit of longtime residents – often people of color – who are no longer able to afford to stay. Central to this disagreement is the question of whether the displacement of minority residents is inherently an inevitable outcome of gentrification.

One of the main reasons this disagreement persists is because gentrification hasn’t historically been defined in one consistent way. Different researchers set different eligibility rules (i.e., which places could gentrify), different thresholds (i.e., how much change counts as gentrification) and different indicators (i.e., income, housing costs, college degrees or property values). Depending on the indicators used, the map continually changes — along with the conclusions about the causes of gentrification and its relationship to displacement. 

In response to this, NCRC and our partners at the University of Michigan recently published an article in the Journal of Urban Affairs that assesses several widely used methods for measuring gentrification. Instead of picking one definition and defending it, the central analysis sought to figure out which definitions best track the kinds of changes most people associate with gentrification. We utilized an approach borrowed from data science and medical testing protocols called the Receiver Operating Characteristic (ROC) analysis, which scores how well each definition lines up with real-world changes without getting stuck on any one single threshold.

Our starting point is basic: gentrification isn’t just about buildings and prices. It’s about who has the means and the power to live in any given place. To that end, we looked at indicators of economic change (income, property values, rents, education) and indicators of social change (who’s moving in, who’s moving out and whether higher-status jobs are concentrating).

We then took each gentrification method’s analysis map and compared it with three independent signals of neighborhood transformation:

• Occupational change: The number of residents in professional/managerial/technical jobs (a marker of class shift);
• Racial demographics: Change in the share of non-Hispanic, White residents (a common pattern of racial turnover in US cities) and;
• Housing market or rental market pressure: Growth in home purchase mortgages (a sign of reinvestment and demand).

If a gentrification method’s analysis map lines up with large shifts in these indicators, it is considered as a valid mechanism for measuring gentrification. If not, it isn’t considered valid.

We focused on Washington, DC during the 1990–2020 period for the larger analysis because it’s one of the most studied – and most debated – cases of gentrification in the country. The District has seen major public investment, shifting housing markets and dramatic demographic changes across the past three decades, making it a strong test bed.

The map of Washington, DC below shows five methods of assessing whether or not neighborhoods gentrified between 1990 and 2020, with different results. Depending on which method is used, the estimates of displacement varied widely:

Figure 1 – Different ways of assessing gentrification in Washington DC by decade – (1990s to 2010s).

When we ran the accuracy checks in assessing the true gentrification rates for Washington, DC from the 1990 to 2010 period, five approaches – including one used by NCRC – were consistently matched to real-world patterns of class change, racial turnover and reinvestment better than other approaches. Freeman’s method (2004) is one of the original ways of using census data to assess whether neighborhood gentrification occurred while using the most indicators related to income, home value, housing age and education in its method. The other approaches follow Freeman’s procedures more closely, but use different indicators and different geographic boundaries (i.e., downtown areas, metropolitan areas, etc.) in their methods. McKinnish (2010) and Ellen & O’Regan’s (2011) methods focus primarily on changes in median income. Ding, et al. (2016) added more indicators in a study focused on Philadelphia. NCRC’s (Mitchell, et al., 2019) method largely followed Ding’s method with a particular focus on central city neighborhoods and was tested at the national level. 

Using these stronger-performing methods, we estimate that about 60,000 Black residents moved out of neighborhoods with indications of gentrification between 1990 and 2020. However, that does not mean that every move was forced. People move for a variety of reasons. But, it does mean that the loss of Black residents was concentrated in gentrifying areas, while White, and in many areas Hispanic, populations grew in those same areas. In short: where these methods say gentrification occurred, racialized population loss also followed. The chart below shows how the changes in residents vary depending on which method of assessing gentrification was used compared to the US Census’ estimates of changes in the District’s overall population:

Table 1 – Different methods of assessing gentrification and displacement.

The approach uses the Receiver Operating Characteristic curve analysis model to evaluate competing definitions of gentrification by testing which ones most closely track observable social and market changes. Greater detail on how to produce the ROC evaluations can be found in this article. Its main conclusion is that while most of the methods performed adequately, the NCRC and Ding methods have a slightly greater validity in this Washington, DC-based comparison. By directly comparing methods against real-world outcomes, the overall approach stays objective, which keeps the test transparent, reusable and shows which methods track the changes that communities are actually experiencing. This approach provides several tangible benefits:

• Improves the public debate. We can move past the “it depends on the method” approach to determining gentrification towards one that is best matched to observing societal and market changes.
• Supports targeted solutions. Cities can focus on tenant protections, right-to-return policies, long-term affordability covenants, community land trusts and anti-displacement funds where validated measures show pressure is highest.
• Scales well to other cities. The same accuracy check can be used for cities with differing contexts, such as Los Angeles, Atlanta or Seattle, to create local definitions and target interventions where they’ll matter most.

Ultimately, the causes and impacts of gentrification are both economic and social. When multiple definitions are tested against real indicators of class, race and community investment, the strongest methods showed that DC’s gentrifying neighborhoods were the main sites of large Black population loss – about 60,000 people over three decades. That makes the stakes more clear, and it gives cities a practical way to aim policy where it counts. When gentrification is measured accurately, patterns of racialized population loss become harder to dismiss and easier to address with targeted policy.

Bruce C. Mitchell is the Principal Researcher with NCRC’s Research team.

Photo credit: Haikal Omar via Pexels.