After The Marshall Project and the New York Times’s The Upshot published an investigation debunking the often-repeated idea that immigrants increase crime in the U.S., many readers asked: What about undocumented immigrants?
We knew we wanted to try to answer this question. The problem was that very little data exists about undocumented immigrants. So when the Pew Research Center released new undocumented population estimates across the country, for the first time it was possible to compare population changes to changes in crime in the last decade, and show that undocumented immigrants, too, do not increase crime.
Advocates for immigration reform have used this work to rebut misleading narratives on immigration at White House news conferences where ICE and other law enforcement officials were speaking. The Washington Post published an opinion piece from their editorial board about our findings. In January, the report was used to motivate a bill passed by the Washington senate to curb discrimination against undocumented immigrants.
This report was shared and viewed widely, with hundreds of thousands of views on the Marshall Project website and our partner New York Times page. It was posted to social media by the thousands, including by many respected journalists, politicians, organizations and leaders such as Peter Baker, Glenn Thrush, Nicholas Kristof, Sam Vinograd, Sahil Kapur, Rep. Nydia Velazquez, the ACLU, the U.S. House Committee on Homeland Security, the Urban Institute, the Sentencing Project, Prison Legal News, and others. Flagg was interviewed on CNN on Michael Smerconish’s morning show, and on the radio on SiriusXM and NPR.
Dozens of other news organizations picked up or otherwise covered the analysis, including NPR, Politico, The Washington Post, The Atlantic, Rolling Stone, The Boston Globe, The Trace, the Southern Poverty Law Center, Documented NY, AZ Central, Daily Kos, Yahoo News, Splinter News, ThinkProgress, The Huffington Post, New York Magazine, Mother Jones, AM New York, and a range of Spanish and international outlets such as Univision, Diario De Noticias, Proceso Digital, Al Dia, The Brazilian Times, La Prensa, Radio Bilingüe, Gestión, El Sol de Mexico, El Diario NY and others.
Some localized the data we provided on request to produce stories focused on their own areas, including Patch.com’s series of local articles covering Chicago, Philadelphia, Boston, Charlotte, Providence, Nashua and Lubbock.
To investigate the potential relationship between undocumented immigrants and crime, The Marshall Project downloaded and merged all types of violent and property crime data published by the FBI Uniform Crime Reporting program for the same areas and time period covered by the Pew undocumented population estimates that had just been released. Historical changes in the legal definition of rape and inconsistencies in how motor vehicle theft is recorded in different areas meant both of these types of crime had to be removed from analysis, so we used raw numbers of reported crimes and populations to produce amended rates for the rest of the categories.
After calculating 3-year averages and changes in crime rates, The Marshall Project fit regressions to model the relationship between changes in an area’s undocumented population and changes in violent crime, property crime, and their components of aggravated assault, robbery, murder, burglary and larceny. None of the models found evidence of a connection. This analysis was done in R.
We demonstrated our analysis visually, allowing viewers to see and understand the data directly for themselves. We used Illustrator and D3.js for this design and web development work.
What was the hardest part of this project?
Our main challenge in terms of data analysis for this project was the scarcity of available data about undocumented populations. Thanks to the work Pew does, we were able to get estimates of these populations. But the estimates Pew publishes as part of their standard work are generally raw numbers of undocumented immigrants, and our analysis required percent change over time – nontrivial to derive due to the error margins in the estimate formulas. So Pew researchers generously worked with us to get the estimates in the form we needed for a robust analysis.
The other time-consuming part of this analysis was tracking crime rates by metropolitan area, the geographies of which change over time. Sometimes a metro would grow to encapsulate new area, sometimes it would divide into multiple smaller regions. Consulting technical documentation for the roughly 180 areas in our study by hand, we determined for each area when a changed geography could still be an appropriate match to the original, when smaller areas would need to be combined for a proper match, or when no accurate match was possible.
In editorial terms, our biggest challenge was to cut through the vast amount of misinformation and flawed data reporting on the subject of immigrants, and the fear that comes with such misinformation. Inspired by the specific needs our readers expressed, we did everything we could think of to encourage their trust by making our analysis process transparent and understandable, including presenting the information visually in a way that people could browse the data and make up their own minds.
What can others learn from this project?
One thing we learned from the process of reporting this story was the value of listening to the questions and needs of our readers. By paying attention to readers’ voices, we were able to identify an opportunity to provide clarity on a question that was important to them, and we are so grateful to readers for giving us that opportunity.