Research
Gun violence prevention Education that's equitable
I came to the topics of gun violence and educational inequality through different paths, but they are directly linked. Research has shown (opens in a new tab) that a homicide near a child's home can lower their reading and vocabulary scores in the days that follow. And more schooling reduces criminal offending (opens in a new tab).
Gun violence · Where I started
Does a gun make your home safer?
I was first drawn into this research by a simple question. Decades of research show that a gun in the home increases the risk of suicide, homicide (opens in a new tab), and intimate partner violence (opens in a new tab) for the people who live there. But most Americans believe the opposite (opens in a new tab). Why do so many people see a gun at home as protective?
That question turned into my honors thesis (Amherst, May 2025), which I wrote as four papers on national survey and public-health data. It closes with a proposal for a public dashboard to help people without a formal statistical background better understand the empirical research around gun violence. Two of the papers are now under review.
My main theory, after doing all of this research, is stories. Why would someone own a gun for protection? It's pretty easy to conjure up a story: a bad guy breaks into the house, and I want to protect myself and my family.
But let me tell you a different story. A bad guy breaks into the house because he wants to steal something from me. All he cares about is walking away with that thing, but he's armed, "just in case." I raise a gun. He sees it. He shoots first.
If I hadn't had a gun on hand, I might well have walked out alive. But that's not the story we conjure up when we think about someone breaking into our home.
Gun violence · At the UChicago Crime Lab
What causes homicide, and how do we stop it?
Here's one more story, and this time, take a guess first. When most of us worry about homicide, we think about being killed by a stranger, often while they're in the process of committing another crime (e.g., someone robs you and kills you in the process, a mass shooting, etc.).
For every 10 people who are killed by a stranger in the process of committing another crime, how many people do you think are killed by someone they know in an argument?
The correct answer is 75.
10 Killed by a stranger during another crime
75 Killed by someone they know, in an argument
Source
Supplementary Homicide Reports Multiply Imputed Database, 1976–2023 (opens in a new tab) (James Alan Fox, Northeastern University), 2024 data provided by Fox; author's calculation for 2024. "Stranger, during another crime": a stranger, in a felony circumstance (robbery, burglary, a drug offense, etc.). "Someone they know, in an argument": any known relationship, in an argument or brawl.
So the thing Americans think they're protecting against by owning a gun—a stranger coming in and killing them—is actually far less common than being killed by someone they know. Jens Ludwig's NBER working paper, "Crime in Covid Times (opens in a new tab)," takes on the question of what drives homicide in America. I was the research assistant on it, and I'm working on a related essay to be published soon.
This is another case where the stories we tell ourselves don't match reality. And it helps explain why having a gun in the home increases the risk of homicide. As Ludwig argues, homicide in America mostly happens when there's an argument and a gun on hand. More guns on hand means that when arguments escalate in the heat of the moment, they're more likely to turn deadly.
Education · At the UChicago Education Lab
Which students does a program help most?
A randomized controlled trial (RCT) usually reports one number: on average, did the program help, hurt, or make no difference (the average treatment effect)? But if you're a principal deciding who gets a tutor, you need to know which students it would help the most, and the average can hide that. So we try to predict how much an individual student will benefit (or not) from a given program, such as tutoring. Statisticians call these estimates heterogeneous treatment effects (HTEs).
With other members of the Data Science team, I've worked on model selection algorithms, built and implemented novel HTE algorithms, and even built an AI chatbot that helps principals with a limited budget schedule tutoring sessions, so the tutoring goes to the students who will benefit the most.
Under review
"Perceptions of Safety: A Quarter-Century of Change in Attitudes Toward Community Firearms."
How do Americans' feelings about more guns in their communities compare in 1999 and 2024?
"Demographic Drivers of the Gender Gap in Surveys of Household Firearm Ownership."
Why do married men and women answer survey questions about household gun ownership differently?
Working papers & essays in progress
"Gun Ownership and Voting in Presidential Elections: A Response to Joslyn."
Does knowing someone owns a gun help predict their presidential vote?
"U.S. Ecological Analysis of Firearm Ownership and Suicide Rates."
How does household gun ownership relate to suicide across U.S. states?
"Protest or Pandemic: An Alternative Theory for Why U.S. Homicide Rose in May of 2020."
What drove the 2020 surge in U.S. homicides?
"Rubin meets Rashomon: Mitigating model multiplicity in heterogeneous treatment effect estimation" (working title).
"Pivoting the paradigm: the role of spreadsheets in K-12 data science."
"Factors associated with Study Analgesic Medication Persistence for post-surgical pain in the Opioid Analgesic Reduction Study (OARS)."
P. B. Matheson, O. Tirschwell, and 11 co-authors.
What predicts whether patients keep taking their study pain medication?
Thesis
"A Fatal Fallacy: Understanding and Bridging the Gap Between Empirical Research and Public Opinion on Safety Implications of a Firearm in the Home."
Why do so many people see a gun at home as protective?
Research assistance
Jens Ludwig, "Crime in Covid Times."
Talks
- Education "Ask the Data: How Conversational Analytics May Change Educational Research, Evaluation, and Planning." Co-presenting with Joel Rabin (Inform 2 Inspire). We built an AI chatbot for California's literacy initiatives that lets non-technical users explore public California literacy data by asking simple questions in English. California Educational Research Association annual conference, Rancho Mirage, CA. Session details (opens in a new tab)
- Gun violence "Changing Minds, Saving Lives." Three Minute Thesis, Amherst College. People's Choice Award. Watch the talk (opens in a new tab)
- Gun violence Demographic Drivers of the Gender Gap in Surveys of Household Firearm Ownership, at Wesleyan's Center for the Study of Guns & Society undergraduate research conference.
Questions about the research?
Email me, especially if you're working on similar questions. I'd love to compare notes.