The Eviction Lab has created an expansive, nationwide database on evictions, and they’ve just launched an interactive map that the the New York Times featured. James Minton, Pat Sier, Noele Lusano, and Lane Olson of the Eviction Lab web team tell us how they used our maps and data visualization tools to bring the data to life.
What’s the story behind Eviction Lab?
The Eviction Lab grew out of Professor Matthew Desmond’s work studying housing, poverty, and eviction in 2008, living and working alongside poor tenants and their landlords in Milwaukee, WI. He realized the need to collect national data on eviction to address fundamental questions about residential instability, forced moves, and poverty in America. Until now, publicly available eviction data has been limited.
Professor Desmond and the research team at Princeton spent the past few years acquiring, cleaning, and validating tens of millions of eviction records. This is the first nationwide database on evictions at this level of detail, and they created an interactive site for people to explore and use this data.
What do you hope an interactive map like this will accomplish?
At a high level, we’re hoping to change the narrative about residential instability in America by illustrating the prevalence of evictions nationwide and providing policymakers, researchers, journalists, and citizens with accurate information. Beyond that, we aim to depict “hotspots” where the problem is concentrated and allow for easy comparisons between areas, from states down to census block groups.
One fundamental question driving this work is, how is this problem spatially (as well as temporally) distributed and concentrated? A map is the most useful visualization we could employ to depict that. Allowing people to search for their locations helps drive home the idea that this is a national issue, even if it is concentrated in certain parts of the country.
Why did you use our tools to build the map?
Your platform was an easy choice for us after discovering vector tiles and the Mapbox GL library. It would have been difficult to manage an interactive visualization with as much data as we have without using vector tiles and your web API, Mapbox GL JS. We’re expecting a lot of traffic, so we also needed a reliable solution.
What were some of the technical problems you had to solve?
Balancing accuracy of feature shapes, amount of data loaded at once, and showing layers at low zooms was a challenge for us. It quickly became apparent that displaying eviction data all the way down to the census block group level would require huge, unmanageable GeoJSON files. Luckily, we came across your open-source command line tool, Tippecanoe, which allowed us to automate data injection and optimize our geometries. Compressing the data as well as geography into vector tiles with Tippecanoe meant we could load it all from pre-built static files rather than needing to build an API.
How about challenges related to geocoding and search?
We had planned to use Mapzen for geocoding because we had more experience with it, but when Mapzen closed, we found moving to Mapbox geocoding to be relatively painless. The two APIs are pretty similar; the only pain point was that your geocoder doesn’t return counties, but we just added our own small subset of data for counties in case someone searches by county.
How did you approach the design and map styling?
The first major challenge from a design perspective — since we were tasked with displaying not just eviction rates, but also a variety of census data — was finding a way to elegantly display two variables at once atop an already-complex visualization. We had great help early on from our technical consultant, Chris Groskopf, in developing a bivariate mapping strategy using the standard choropleths for census data, plus “danger color” red bubbles for evictions. Since that strategy was adopted, a major focus of our work has been keeping all these elements on the map — choropleths, bubbles, place names, boundaries, streets — in balance, so the user can find their desired location, as well as the data they’re seeking.
We also wondered how to depict null data to give users an honest picture, rather than showing nothing, which could be misinterpreted as no evictions. We decided on a few visual indicators that would be explained in the map legend. For example, we use white bubbles with a grey outline instead of red to indicate missing data.
You also used our tools to control the presentation of map data at each zoom.
We needed to figure out how to show just enough geographic detail and associated labels so users can orient themselves and browse through unfamiliar geography. We opted to confine the most detailed geographic features (major roads and highways, towns, and villages) to the highest levels of zoom, maintaining only larger place names throughout most other levels of zoom. We also added a hover tooltip at high zoom levels that tells you where you are to help improve user navigation.
This is a really inspiring project and now an impressive map. What’s next for your team?
One feature we definitely plan to implement is a Top Evictors ranking tool that lists the landlords who evict the most in a given area, with the aim of redressing the information asymmetry between eviction defendants and plaintiffs. We came together as a web development team for the first time on this project, and we’ve garnered a good deal of hard-won expertise, especially with regard to the map tool. We’d love to devote that expertise to future mapping and data visualization projects. We’re interested in digging more into Expressions in GL JS, as well as some of the other features like heatmaps and extrusions.
The entire evictions data set is publicly available at evictionlab.org, view the full map. Use our data viz tools to create interactive maps and tell complex stories with your data. Learn more about how our Community team is supporting positive impact projects across the world.
Mapping nationwide evictions: How we built it was originally published in Points of interest on Medium, where people are continuing the conversation by highlighting and responding to this story.