My entire education was subsidized by taxpayers. From kindergarten through high school I attended public schools in Ottawa. My undergraduate degree is from the University of Toronto, which receives about a quarter of its funding from a government, while my graduate degrees are from the University of Michigan, which receives a collapsing but not ignorable proportion of its funding from a government. I think all researchers have a responsibility to make their research not just available to the public, but also understandable to people at large, and this is especially true for those of us who try to research questions of real public interest. But I also feel an obligation to the Ontarians and Michiganders who picked up much of the bill for me to learn how to do any of this stuff in the first place.
I see my work as falling into about four groupings, and I'll give some take-home messages from my research in each of those clusters. I will try to keep it short. Please note that this is not an essay on the most important things to know about an area, or things that I think are deeply true. I am only telling you what we found in our studies. This document would be a good piece to read at my funeral, but a very bad meta-analysis.
I wrote my PhD dissertation about simulating elections in the computer, and the main application was to simulate how many seats different political parties would have obtained in a specific election, if the election had been conducted under a different electoral system (that is, if the country switched from a voting rule like single-vote plurality to some kind of proportional representation). The headline was that these kinds of simulations are possible, they seem accurate, and they can tell us interesting things about what could happen if we changed how votes are cast and counted. It also seemed that, when political parties take positions on which electoral systems they support, they might not always be acting in their own self-interest; I found several examples of parties supporting systems that, I think, would probably hurt their electoral performance. The Rackham newsletter ran a nice piece that explains this work well.
More recently, in the paper "Votes can be confidently bought in some ranked ballot elections, and what to do about it", we present a hypothetical vote-buying scheme that could be used to purchase votes in elections that use ranked ballots. We don't believe that anyone has ever used this scheme, but we explain how it could be done, and we discuss the risk factors that could make a ranked ballot election more susceptible to this kind of illicit activity.
The administration of American elections is the main concern of the MIT Election Data and Science Lab (MEDSL), so naturally, when I joined the lab, it became one of my main concerns as well. We've written a few pieces in this area.
The paper "American election results at the precinct level" presents the results of three national elections' worth of gathering precinct-level results of American elections. MEDSL builds and hosts one of the main repositories of these results, which are the most granular level of election data reporting that come anywhere close to national coverage. They take an enormous amount of time and resources to collect, clean, and standardize, since each state reports them individually, and often in radically different ways. Our paper talks about the more than 40 million rows of candidate-precinct-mode level results that we had collected at that time, and works through why you should believe that they're basically accurate.
In the more recent paper "Trust in the Count: Improving Voter Confidence with Post-Election Audits", we ran a survey experiment to check whether or not election audits that find a similar result to the original vote count actually do increase voters' trust in elections, which everyone assumes they do, but which had been tested very few times before. Beyond adding one more data point to suggest that they do, we actually designed the experiment in a way that allowed us to test which attributes of audits are most important for bolstering confidence; so, is it who does the audit, or how many votes they re-count, or something else? Of the attributes we tested, we found that the most important one was how transparently results are reported. This is unfortunate, since in some of my ongoing work I've learned that the results of American election audits are usually not available to the public without substantial effort.
Two other papers in this group, the ones about election officials, are companion pieces. For years there has been extremely good reason to worry that election officials are severely mistreated by the public, often accused with absolutely no evidence --- and sometimes viciously, with credible threats of violence --- of engaging in illicit activity to subvert the elections that, in truth, they work hard to conduct with few resources and under great pressure. To objectively verify that this is happening, and to gauge its extent, we collected over a million tweets from state and local election officials, and we checked the words used in reply to them, as well as the tone of those replies. We found that the way people address election officials on social media has indeed grown steadily more harsh over the last decade. One of my coauthors wrote a great rundown of the results here.
I used to study comparative political science, where we're admittely still trying to wrap our minds around what democracy is and what it does. Two of my papers deal with these questions. In the chapter that appears in Why Democracies Develop and Decline, which is a book that applies Varietes of Democracy (V-Dem) data to important political science questions, we used V-Dem data to re-test several of the most prominent theories about how democracy relates to various other types of major political institutions. If you're interested in the connection between democracy and other institutions like state capacity and party systems, this was our take on the classics. Probably my favourite part of this chapter is that we introduced a new way of testing hypotheses for many different definitions of democracy all at once, resulting in some really fun pseudo-"caterpillar plots". Michael Coppedge spoke about the whole book on the Democracy Paradox podcast, and it's a nice overview of all the findings together.
In "An unexpected consensus among diverse ways to measure democracy", we showed a (to me, very surprising) result in how democracy is measured. A classic question is whether democracy is a binary attribute (you're either a democracy or you're not), or whether it's continuous (one country can be, say, 30% democratic, while another is 80.3% democratic). Sometimes you want to take one of these many-valued measures and split it into a binary measure; for example, people sometimes take every country in some many-valued measure that's at least 50% democratic and call it a democracy, whereas if you're below 50%, you're an autocracy. A natural question is: why 50%? We were curious what happens if you take a many-valued measure and split it up in whichever way matches one of the major binary measures as closely as possible. The answer was: you get exactly the same number, over and over and over, to such an extent that I kept thinking there was a bug in my program. This paper grapples with what that regularity means for the study of democracy.
In 2020, I spent about 650 hours on a volunteer project: writing a Wikipedia page about a political scientist every single day of the year. I undertook this project specifically to reduce the severe biases in which political scientists (and more broadly, which scientists, and indeed which people) have a page devoted to them on Wikipedia. I focused on writing pages about women, Americans of colour, and people outside the United States or other over-represented countries. The paper writes up both the bias that existed, how I went about the project, and also how very small a dent all that effort made in the bias that still exists. I also wrote about this project in The Washington Post.
The Clock Game paper in a recreational math magazine gives a thorough solution to a game that my friend and I used to play when we were kids, where you take a sequence of numbers that you see somewhere in the world (like reading the time off a clock), and try to make them into a true mathematical statement by inserting operations in between each of the digits. This one's just for fun.