Gentrification, the process by which communities are displaced economically, culturally, or physically from neighborhoods they have long called home, is a pressing threat in many cities. First documented by Ruth Glass in mid-twentieth century London, the process takes many forms and has been linked to many causes, including parks and green infrastructure. My research in gentrification, then, began as an outgrowth of my interest in green space in cities, particularly urban agriculture. However, the work has become a research topic unto itself, including two funded projects. 

The first project, detailed below, sought to develop new methods for high resolution modeling of gentrification across jurisdictional boundaries. This work was funded by Poverty Solutions at the University of Michigan. I am also Co-PI of a Sea Grant funded project, set to begin in late 2023, focused on understanding gentrification in post-industrial waterfront communities in Michigan. This work emerged from conversations with stakeholders and non-profits in impacts communities who report that recent research documenting the economic benefits of post-industrial investment has disguised inequities and gentrification. For more information about this ongoing project or about the project below, feel free to reach out

High-resolution assessment of gentrification across cities

Gentrification is notoriously difficult to measure, particularly across jurisdictions (Loukaitou-Sideris et al., 2019). Gentrification is a complex process which includes physical (e.g., building rehabilitation) and economic (e.g., business types/ownership) transitions as well as social (wealth, education) and cultural (community organizations) transitions. Traditional measures of gentrification are either qualitative or quantitative (Preis et al., 2020). Qualitative measures typically provide a robust look at economic and cultural transitions in particular neighborhoods, while quantitative measures typically offer a clearer look at economic and social trends across entire cities (bunten et al., 2023; Preis et al., 2020). While a few isolated projects have sought to integrate these into mixed methods indicators (Pettit et al., 2019), scholars continue to search for indicators that bridge the strengths of qualitative and quantitative assessment - a way to map gentrification at a large scale that offers an early warning system as well as a way to measure the effects over time. 

I was awarded a Poverty Solutions fellowship in the summer of 2023 to pursue improved solutions for modeling gentrification across jurisdictions, particularly in the context of amenity-driven gentrification, which we have worked on previously. For this work, I explored the utility of nationwide point data for households and businesses as a way to create very high resolution information on gentrification processes, outcomes, and futures. I built on the work of bunten et al. (2023) who documented that measuring the gap between income in a neighborhood and home value in a neighborhood can foreshadow transitions. Using household-level data provided by InfoGroup, a marketing firm, I assessed the physical and social transitions associated with gentrification with income and home value data, including the development of an early warning system for gentrification. This novel early warning system for gentrification can be calculated at the household level for any city in the United States. It can then be combined with business data, also available at the level of individual structures, and census data to consider the predictiveness of our early warning system and the implications for future planning. For the purposes of this project, we focus on coastal Michigan as a case study in early warning system development, work which sets the stage for broader analysis of tourism-driven gentrification on the “pleasant peninsula.” 

Developing the early warning system

Gentrification scholars have long theorized that the primary driver behind the phenomenon is the so-called rent gap (Smith, 1979). Although the exact definition and operationalization of a rent gap has long been debated (Hammel, 1999), the general principle is that a rent gap occurs when there is some distance between the potential and realized value of a parcel. While some scholars argue that land use and land rent theory are incompatible, others point out that a rent gap can appear even under the complexities of mixed land use. Building on bunten et al. (2023), this work assumes that a rent gap occurs and can be measured anytime the expectations for a plot of land differ from the realized value extracted. Specifically, we explore the bounds of bunten’s expectations-based measure which argues that an emerging gap between local home values and local incomes demonstrates the high potential for near-term investment and neighborhood change. 

While bunten et al. model and test this hypothesis in a number of large urban centers, we have extended this model in two key ways. First, we operationalized the model at the household level, allowing extreme flexibility in the scope and scale of assessment. Second, we applied this model to large towns and small cities in Michigan, exploring the potential of this expectations-based metric to offer early warnings of forms of gentrification that occur outside large urban agglomerations (Cocola-Gant, 2018)

To do this, we employed three key datasets: 1. a household-level dataset provided by InfoGroup, 2. business data provided by InfoGroup, and 3. US census data. Data cleaning and processing occurred in R; code are available as an R markdown document upon request and data can be shared with University of Michigan affiliates under the library license. The scope of analysis was coastal urban areas in Michigan, following the Census Bureau’s definition of urban as at least 2,000 housing units or at least 5,000 people (2020 Census Urban Areas FAQs, 2022). Within each of those coastal urban areas, we calculated household income and home value percentiles before assessing the gap between these. In our base case (based on the work of bunten et al.), a percentile gap higher than 25% marked a house as susceptible to gentrification, and neighborhoods where more than ⅓ of homes met this criterion were considered to be in the early stages of gentrification. Sensitivity analysis of both of these thresholds is discussed below. We also test how much lag appears between the appearance of the rent gap and other documented changes in neighborhoods. 

Using the expectations measure to predict future changes

To determine which percentile gap and time delay were most predictive of community change, we searched for trends in the correlation between rent gap variables and known indicators of gentrification. Specifically, we considered seven different values for the gap size, seven values for the percent of the neighborhood reaching a gap, and four values for the number of years after a gap was identified. 

Across the state of Michigan, the utility of our measure of gentrification seems to peak when we focus on a gap of 30% between the income and home value. Future work should explore why the predictivity of the measure actually decreases as the rent gap gets higher. 

When rent gap is held steady at 30%, the percentage of households which need to meet this threshold is optimized at 20%. While this is relatively low, it may indicate that the rate of investment can increase dramatically after a few houses are determined to be good purchases. 

From here, we can zero in on the most appropriate lag time to look for changes after a gap has been identified. According to our findings, 7 years is the appropriate amount of time to wait. 

Now that we have identified the appropriate thresholds, the question remains - does this measure actually detect significant differences between gentrifying and stable neighborhoods? In short, yes. All variables of interest differ significantly across gentrifying and stable neighborhoods in the directions we would expect. Purchasing power, the number of coffee shops, and the percent of young adults all grow in gentrifying neighborhoods, while the average length of residency declines. 

Gentrification in coastal Michigan today and tomorrow

Given the thresholds we have determined, we can map gentrification across coastal Michigan. The map on the following page shows a subset of Michigan coastal cities, showing expected gentrification in 2020. 

Figure 4. Map of gentrification in Michigan coastal cities. Red indicates gentrification was expected in the census block group around 2020. A.) Detroit and St. Clair suburbs, B.) Muskegon and Grand Haven, C.) Traverse City, D.) Houghton, E.) Port Huron, F.) Alpena

Future work

This project has explored the potential for mapping gentrification at extremely high resolution across multiple municipalities. Results-to-date suggest that this is possible and that there is great potential for insights across scales. Two areas of future work will be the focus of upcoming projects: 


2020 Census Urban Areas FAQs. (2022). US Census Bureau.

bunten,  devin michelle, Preis, B., & Aron-Dine, S. (2023). Re-measuring gentrification. Urban Studies, 00420980231173846.

Cocola-Gant, A. (2018). Tourism gentrification. Handbook of Gentrification Studies, 281–293.

Hammel, D. J. (1999). Re-establishing the Rent Gap: An Alternative View of Capitalised Land Rent. Urban Studies, 36(8), 1283–1293.

Loukaitou-Sideris, A., Gonzalez, S., & Ong, P. (2019). Triangulating Neighborhood Knowledge to Understand Neighborhood Change: Methods to Study Gentrification. Journal of Planning Education and Research, 39(2), 227–242.

Pettit, K. L. S., Cohen, M., & Levy, D. K. (2019, April 19). Turning the Corner Project Overview: Monitoring Neighborhood Change to Prevent Displacement in Five Cities. Urban Institute.

Preis, B., Janakiraman, A., Bob, A., & Steil, J. (2020). Mapping gentrification and displacement pressure: An exploration of four distinct methodologies. Urban Studies, 0042098020903011.

Smith, N. (1979). Toward a Theory of Gentrification A Back to the City Movement by Capital, not People. Journal of the American Planning Association, 45(4), 538–548.