Dr. Tim Thomas is a professional researcher at the University of California, Berkeley, where he served as research director of the Urban Displacement Project. He also founded and serves as the director of the Eviction Research Network. He recently served as the chief research and data officer at the King County Regional Homeless Authority, where he led efforts to use data-driven strategies to address homelessness. Thomas's research focuses on neighborhood change, housing disparity, and urban displacement. His works informed tenant protection laws across the country, as well as the eviction moratorium during the COVID pandemic.
Episode Description
It can take decades for an urban community to develop, but a fraction of that time for it to be torn apart. Each year, urban residents are forced out of their homes and out of their communities to make way for gentrification or other kinds of urban development. When these individuals are forced to leave their homes, they're also forced to leave behind their histories and cultures. These changes often rob communities of support systems, which is a focus of this episode of Stats and Stories with guest Dr. Tim Thomas.
Transcript
Rosemary Pennington
It can take decades for an urban community to develop, but a fraction of that time for it to be torn apart. Each year, urban residents are forced out of their homes and out of their communities to make way for gentrification or other kinds of urban development. When these individuals are forced to leave their homes, they're also forced to leave behind their histories and cultures. This urban displacement often robs them of support systems as well.
It's the focus of this episode of Stats + Stories, where we explore the statistics behind the stories and the stories behind the statistics.
I'm Rosemary Pennington. Stats + Stories is a production of the American Statistical Association in partnership with Miami University's Departments of Statistics and Media, Journalism, and Film.
Joining me is regular panelist John Bailer, emeritus professor of statistics at Miami University.
Our guest today is Tim Thomas. Thomas is a professional researcher at the University of California, Berkeley, where he served as research director of the Urban Displacement Project. He also founded and serves as the director of the Eviction Research Network.
He recently served as the chief research and data officer at the King County Regional Homeless Authority, where he led efforts to use data-driven strategies to address homelessness.
Thomas's research focuses on neighborhood change, housing disparity, and urban displacement. His work has informed tenant protection laws across the country, as well as the eviction moratorium during the COVID-19 pandemic.
Tim, thank you so much for joining us today.
Tim Thomas
Thank you so much for having me. It's a pleasure.
Rosemary Pennington
What is urban displacement?
Tim Thomas
You know, it's a really good question that we don't talk a lot about. We hear this term displacement, we hear terms like gentrification, but I think it's really nice and easy to situate this into three buckets, right?
There's exclusive neighborhoods, the really expensive neighborhoods we can't really get into. These are spaces that wealthier folks move into, and individuals who are higher income tend to be more mobile. They can move to spaces they want to. They can take advantage of certain areas that have really nice amenities or good networks. They can move downtown or something like that.
So those are areas that are kind of blocked off for most of society.
Then there's the term gentrification, right? That's where a neighborhood that traditionally was lower income, potentially segregated, has already switched over. That term was coined in the 1960s by someone named Ruth Glass, but it really took hold in the '90s, particularly in America, where there was urban revitalization, people moving back to the urban core, and you started to see disparities in terms of who was able to take advantage of gentrified areas.
Gentrification still goes on. There's constant development. Think about the new high-rise buildings going up across entire parts of America. This is not secluded to San Francisco, Seattle, or New York. Gentrification is every city, right? Those are spaces that have seen a shift and are kind of starting to lean toward becoming exclusive, but that's not necessarily always the case.
Displacement, on the other hand, is a term that we've talked about at UC Berkeley, particularly at the Urban Displacement Project. It kind of precedes gentrification.
There are low-income neighborhoods that are ripe for investment, but they also tend to start out lower income and segregated. Not all low-income areas are going to gentrify, right?
If you think about an urban core, an area that has shifted and where rents have gone up, think about a block or a neighborhood. Back in the day, a coffee shop kind of predicted where the next gentrifying space was. Now, I think it's Trader Joe's and other commercial spaces that trigger this sense of, Oh wow, the neighborhood has changed.
When I think about displacement, I look for neighborhoods nearby that have lower rents and tend to be less, quote unquote, desirable. I want to be careful with that term because who deems something desirable is also a pretty interesting question.
But displacement tends to happen in spaces that are close to gentrifying areas.
I think about displacement in two terms: hard displacement and soft displacement.
Soft displacement is when somebody making about 80% of the area median income has a hard time paying the rent and thinks, Well, I don't want to move, but I definitely don't want to be paying these prices.r
So, I call that soft displacement. That's middle-income folks in America tending to move to other areas. Those people, when they move to certain spaces, may start to gentrify another area that's more affordable.
What has particularly happened for Black and Brown communities is that displacement has occurred in formerly redlined, segregated spaces and pushed people farther away from the urban core, farther away from their networks, and farther away from their jobs. People have to commute longer distances to get to work, so it causes a lot of problems across the board.
Granted, they had a quote unquote choice. That's the key word for soft displacement. It doesn't necessarily mean they wanted to do that.
Hard displacement is another topic that I don't think we talk about a lot. That's when you're forced to move. That's where evictions happen, and by and large, I think this is the population most vulnerable to homelessness. They have nowhere else to go.
When we've done research on where people can go, I coined this term, highways of migration.
If you look at a map—any city—and think about where it's affordable for somebody making 80% of the area median income, or 50% of the area median income... I think 50% is kind of the rough cutoff where most people experiencing hard displacement tend to be.
There's only a handful of places they can afford to move to. Maybe they were paying $1,000 or $800 in rent. What other places in the city have apartments available at those prices?
At the end of the day, we start to see that only certain neighborhoods are available for that population to move into.
This is creating a big problem because we had this issue called segregation that created certain spaces where certain people could move. I think it creates huge problems for both the histories and the futures that households face.
John Bailer
You know, as you were talking about this, these are concepts that seem fairly accessible. As you're describing them, these categories are meaningful and definable, but I just think, boy, it's got to be tough to measure this and to think about the data sources.
Can you talk a little bit about how you operationalize measuring ideas of displacement or gentrification, and what kind of data sources you access to do so?
Tim Thomas
Yeah, absolutely. That's a great question.
Categorically, it's nice. We started out with thresholds of medians, where, if a neighborhood or a census tract—which tends to be a rough boundary for a neighborhood—was experiencing more or less mobility, per se, from the U.S. Census.
But that's highly flawed, right? We're throwing out a lot of useful information by looking at just medians, just the halfway threshold there. And census data itself is rough data.
You look at the American Community Survey. Right now, 2023 is available, but that spans across five years, so already, when I'm looking at census data, I'm looking at the past by about three to seven years.
Luckily—not luckily—but displacement is a long game, and it happens over a long period of time.
But really, what we need, and scholars have pointed out this problem, is individual-level data that tracks individual moves over time.
We also don't have information about why people are moving, so we can only make assumptions based on certain information.
We developed this tool called the Housing Precarity Risk Model, which is available on our website, evictionresearch.net.
What it does is use a machine learning model with a Bayesian generalizer. In other words, it's a really fancy model that can take hundreds of variables at a time.
It uses residential data. There's data collected on us all the time based on credit scores, social media, voting records, and even those crappy coupons we get in the mail. People put that data together, and we can obtain yearly addresses.
We basically calculate a net migration rate of people by certain income levels inside a census tract.
When we see more people at, say, 50% of area median income leaving than coming in—because neighborhoods change all the time and people are constantly moving in and out—when you see more exits than entrances, we call that a pseudo form of loss or displacement.
Then we calculate and account for the fact that there's a lot of data and a lot of error because some of that data isn't very good.
Using the U.S. Census, we validate whether we have a strong enough sample of people from that dataset. Then we come up with a best fit and determine which variables predict this net migration rate for different groups.
We then apply those predictions to census tracts across the country to see which areas and demographics are experiencing more out-migration than in-migration.
That's how we identify places where we might be seeing a risk of displacement.
Rosemary Pennington
I led some students to DC a few years ago, and we spent most of the spring semester studying together. As part of that trip, we had a tour of the U Street neighborhood of DC with some housing advocates and activists who were talking about how that place has been gentrified and what the residents have been doing to try to resist gentrification.
I was just reading earlier about attempts to gentrify Anacostia, which was a part of DC where, when people were moving into the center of DC, that pushed African Americans across the river into Anacostia, and now that area is being gentrified as well.
Both U Street and Anacostia are Black parts of DC, Black neighborhoods, and I wonder, as I was listening to you talk, what can residents in these communities do to resist these attempts at displacement?
Tim Thomas
That’s a question I get all the time, and it's difficult.
I think one of the best ways is, first, to be aware of what's going on. Be aware of your contributions and what you're doing. Supporting local businesses and those kinds of things are very important, but also advocating for policy.
It's not just a one-person or one-sided issue or way to approach things.
Back in 2019, when I was involved in helping pass tenant protection legislation in Washington state, I figured out what I call the three-legged stool of progressive policy.
What that is, is we need advocates to share their stories, right? Honestly, they have the best research questions. I can come up with research questions all day.
The second part is data analysts like me who are willing to put together the data. The data is the truth.
And then third, you need policymakers who are willing to champion new bills and new policies.
If you find yourself fitting within one of those three legs, it's important to speak up or stand up.
Otherwise, it's really important to recognize what privileges you might have. Martin Luther King Jr. talked about allies during the Civil Rights Movement, White allies in particular. They can enter and exit a movement at any time simply because of their skin tone.
I'm a White person, and there are definitely privileges that I have.
I think it's important to recognize the impact we can have within the White community, but also to recognize the respect we need to maintain. If we're going to be allies in this, it's a serious commitment, and we need to be aware that what we do may have positive or negative effects.
It's just the data. It's just the research.
What happens in gentrifying neighborhoods, when you see more White individuals moving in, is that you tend to see the neighborhood flipping.
There's a classic four-stage model. This was coined in the '70s, so the language is a little rough, but basically it said, watch where the lesbians moved to, right?
The idea was that individuals who have been marginalized are often willing to move into other marginalized neighborhoods. They start to use quote unquote sweat equity to build up certain areas because they want single-family homes.
Then you start seeing the second stage, which is young families wanting to move into a certain space.
At that moment, you have diversity, but what's really interesting is that the desires of the community and the voting position start to shift a little bit.
So when we're in these spaces, when we're in these communities, be careful about what our wants and needs are versus what the cultural community's needs are.
The other stages are risky investment and then high-gain capital investment. That's where it's kind of gone to the system at that point.
But when we are in those spaces, I think it's important for us to recognize and honor those cultural spaces.
Rosemary Pennington
You're listening to Stats + Stories, and we're talking with Tim Thomas, director of the Eviction Research Network, about urban displacement.
John Bailer
Tim, I'm curious about the types of variables that may be influencing displacement. You have this precarity model, and there are individual factors at the level of a household, or maybe an individual within a household. There might be neighborhood-level factors, but then beyond that, there are structural, historical factors.
You mentioned redlining earlier as defining certain communities. There are other kinds of pressures I could well imagine are coming into play, and I was hoping you could give some illustrations of important features operating at these different levels.
Tim Thomas
Yeah, that's a great question.
Our dependent variable is this quote unquote net migration rate, right? It's the inflow and outflow of low-income households. Are we seeing more or less movement? That's our dependent variable.
The controls that we throw in there, we draw from history because this is where the beauty of data science and computing power comes in.
We use this Bayesian Additive Regression Trees model, this machine learning model that takes up a lot of resources. We started with over 600 features—600 variables—and then whittled those down so we could have more precision based on those features.
I highly encourage people to look up BART models. They're really powerful and fun.
Structurally, we're limited because the individual-level data is on the dependent side—that's the person moving. We count how many of those people are moving, and we use the census tract as the unit of analysis. We have to confine it within a certain bucket.
So think about all the variables that can go in there. We have race, racial composition, income, and housing structure.
The EPA has datasets on vulnerability scores. There are index scores and all sorts of different measures that go in there. Then you have larger countywide factors.
It's so important not to rely on aggregated large data because you lose the detail of what's going on inside a certain space. You may have a county that looks perfectly fine, but when you look closer at the census tract level, you see certain areas experiencing higher or lower levels of one thing or another.
We added a dummy variable to the census tract for voting patterns in that area because whether people voted more Democratic versus Republican kind of determined the types of policies that were being adopted or that people were ready for.
We also accounted for changes between the 2020 election and the preceding election.
We looked at those sociopolitical dynamics, and we also looked at infrastructure. How many affordable housing units were there? How many HUD vouchers were available?
All of those things add to the potential for someone to be able to stay in a neighborhood versus not.
The top predictors are somewhat what you would expect. More segregated Black neighborhoods tended to see a lot more eviction risk, whereas displacement was happening more in areas where there was a greater influx of people moving in.
So we're able to look at those details at a very high level.
Voting really mattered a lot too, which was really interesting. And, to your earlier question about what we can do: make sure you vote.
Rosemary Pennington
How many Americans are going through this each year? And do you have a sense of whether this is a problem that is increasing, or any sense of the trajectory of it?
Tim Thomas
When we created the Housing Precarity Risk Model, the first shock was how prevalent this was across the country.
By and large, no one is untouched. Every city is affected. Even small cities and rural areas in Texas are impacted. It kind of blew me away.
We had to limit this model to some degree because we wanted people to take it and use it appropriately. We limited it to census tracts where there were more than 50 people per square kilometer, and even with that limitation, even trying to make it a little more conservative, there's just so much precarity going on across the entire country.
You definitely can see a lot of precarity in the big hitters—San Francisco, Seattle, New York, Philadelphia, and all these different areas.
But what was really interesting, and what kind of shocked me when we built the model, was that, remember earlier when I was talking about hard and soft displacement? They're not one and one. They don't necessarily overlap all the time.
Displacement tends to happen to households that make 50% to 80% of the area median income, which is a HUD cutoff for qualifying for certain subsidies. People making less than 50% tended to see more eviction risk—hard displacement.
If you looked at San Francisco, what was really wild about the model was that it picked up the impacts of local policies in that area.
If you look at the displacement map of San Francisco, for example, you see a lot of displacement, which makes a lot of sense. It's a very volatile market, with a lot of movement, but they also have a lot of tenant protections against eviction, and you don't see as much eviction risk there.
The displacement risk is huge, while the eviction risk is larger in certain spaces.
What's going on right now is that we don't know much about post-pandemic America. The data is catching up, and that's very problematic right now.
I'm still giving it about two more years before I start analyzing quote unquote post-pandemic America because of the delay of the ACS and the five-year ACS.
But what we see is that, if we look at raw counts of evictions, most states are evicting more people than they ever have in recorded history.
Washington State, California, the Bay Area, Philadelphia, and many other places are really impacted.
What happened was a fundamental shift during the pandemic. People who would have been in that quote unquote soft displacement area, or who were potentially just hanging on financially, were hit when the service sector and many jobs shut down. It really pushed a lot of people into the precarity population.
What we don't really pay attention to is the longevity of this.
Scientists, and even the public listening to this, know that it takes a while for certain things to gestate and mature.
What happened during the pandemic was one of the biggest impacts. I call it the silent depression of sorts because it had a massive impact on people, and it's going to continue to have an impact.
On top of that, we saw a massive, rapid increase in rents.
What was really wild was that, when we look at rents, very urban spaces like San Francisco actually saw a drop in average monthly rent. That's because people were leaving these urban spaces as we went into sheltering in place and all of those other changes.
However, every county around there—Contra Costa County, Marin County—saw nothing but increases, rapid increases, statistically significant increases in rent that they never would have seen had they not experienced migration from those urban areas into more rural counties.
That's causing a huge problem.
Now we also have the commodification of housing, and we're moving into a subscription model of housing, so the concept of the American Dream has fallen apart as well.
We're in a perfect storm right now. The data is not mature enough to fully show it, but I hate to say this: I think we're in the worst period in America's history when it comes to secure housing.
John Bailer
That's a statement that's really hard to follow up with a question on. That is dramatic.
I'm thinking about when you were talking about the vulnerability index, and you talked about HUD vouchers, these housing vouchers that have existed. There was information that was being centrally provided, and there were support systems that were being centrally supported.
There have been reductions in the type of data that's being collected in recent times. Even with vouchers—which, I can't remember the actual statistics, but less than 30% of the need is met by vouchers—even if vouchers are secured, people still have to find housing providers who will accept them.
So there are fewer vouchers, and there are even fewer places that will take them.
What has to happen to make things better?
Tim Thomas
Honestly, we need the data. We need people championing and creating this data.
One data point, for example, that isn't easily accessible is eviction data.
There have been efforts across the country to collect and show what's going on, but every court system is different, every jurisdiction is different, and some of the data is buried within PDF images of court records.
We can't really get a lot of information out of that unless we can digitize it.
That's what we did in Washington state, which was exactly the problem. All records are kind of held at the county clerk level, so you need to access that clerk's system and download the PDFs.
We created this NLP pipeline, basically a natural language processing pipeline, to extract the address of the individual. That allowed us, using the last name and the address, to estimate the race of the individual.
We tested that against real data, and we were within tewo to five percentage points.
I always tell my students all statistics are wrong, but some are useful, and this is one of the more useful statistical models.
We were able to show that disparity, which changed the minds of state legislators on advancing the pay-or-vacate period from three days to 14 days.
Can you imagine being evicted and having to come up with payment in three days? Even two weeks is rough.
What’s striking is how much of the history we’ve studied—in college and throughout our lives—continues to resurface. I often find myself returning to these foundational concepts: segregation, redlining, and related issues.
I think eviction and displacement, and earlier I talked about the highways of migration, are recreating many of the concepts we've talked about with segregation.
It's kind of the new redlining. It's kind of the new set of discriminatory problems that we're facing.
Rosemary Pennington:
That’s all the time we have for this episode of Stats and Stories. Tim, thank you so much for joining us today.
Tim Thomas
Thank you for having me.
Rosemary Pennington:
Stats and Stories is a partnership between the American Statistical Association and Miami University's Departments of Statistics and Media, Journalism, and Film.
You can follow us on Spotify, Apple Podcasts, or wherever you get your podcasts. To share your thoughts, email statsstories@amstat.orgor visit statsandstories.net.
Be sure to join us for future episodes of Stats and Stories, where we explore the statistics behind the stories—and the stories behind the statistics.