Click to view my Research Statement as a pdf.
Click to view my Research Statement as text.
I am an applied economist. I use modern econometric methods to answer empirical, policy-relevant, thought-provoking questions. My research to date studies two broad questions:
- How does marriage policy affect coupling?
- How can we reduce racial disparities in policing outcomes?
Economics of the Household
Recent legal, social, and demographic changes make marriage and cohabitation increasingly distinct and give same-sex couples the right to marry. These changes provide policy variation that lends itself to answering causal questions about marriage policy. Along with the popularity of marriage, the importance of child-rearing, and household decision-making, these changes also affirm the importance of continued research on the topic of marriage.
I leverage a change in the federal government's definition of spouse to estimate how spousal visa policy affects couple formation and marriage rates in my job market paper, “Spousal Visas and Couple Formation: Evidence from the End of the Defense Of Marriage Act.” By recognizing same-sex spouses, the federal government gives same-sex couples access to spousal visas for the first time and legalizes same-sex marriage for non-permanent residents. In response, the marriage and coupling rates for same-sex couples with a citizen and non-citizen partner increase dramatically.
My job market paper is the first to study an extensive margin change in access to spousal visas. I use a difference-in-differences-in-differences design to estimate the treatment-on-the-treated effect of the policy on coupling and marriage rates. Same-sex couples with a citizen partner and a non-citizen partner are the treatment group. The triple difference design removes selection bias due to coupling trends in other same-sex or mixed-citizenship couples. I also implement this with a Poisson count model. This has two advantages. First, the log-linear relationship permits interpreting the estimates as effects on rates in addition to counts. Second, unlike other log-linear models, the Poisson model preserves zeros in the outcome variable, so the sample maintains representativeness.
I find that access to spousal visas causes an increase in coupling rates by 36% and marriage rates by 78%. Hence, spousal visa policy substantially benefits non-permanent residents with citizen partners. Back-of-the-envelope calculations suggest that millions of people directly have their current partners thanks to spousal visa policy.
Legalizing same-sex marriage provides fertile ground for promising empirical research on the effects of marriage. I plan to leverage this variation in future research. In my current early-stage work, I ask if access to the legal marriage contract affects assortative mating or the surplus generated by matches. If so, marriage law favors the creation of some couples over others. I answer this question using variation in state-level same-sex marriage legalization. I calibrate a model (Ciscato, Galichon, Goussé JPE 2020) that quantifies the relative extent of assortative mating and matching surplus across marriage markets. State, year, and couple type (same-sex, different-sex) define the marriage markets. I then estimate how same-sex marriage legalization affects these quantities, using a staggered difference-in-differences design.
My research adds to the contemporary marriage literature that recognizes the distinction between cohabitation and the legal marriage contract. Underappreciating this distinction can confuse analyses of marriage. My research also reframes the experiences of the LGBTQ+ population. In addition to understanding this understudied population, we can learn about significant policies that impact everyone.
Economics of Crime and Policing
Policing is a substantial expense for municipalities, and there is growing concern regarding the outcomes of policing. Heated public debate over crime and policing demonstrates the urgent need for research on these topics.
My co-author, Romaine Campbell, and I have detailed administrative data from a large urban police department in the United States. We exploit these data to gain policy-relevant insight into police officer behavior.
In our working paper, “Officer Language and Suspect Race: A Text Analysis of Police Reports,” we construct an officer-level measure of text-based racial slant. We then leverage the random assignment of officers to 911 call dispatches to estimate the effect of police officer racial slant on arrest probability. Preliminary results suggest that officers exhibiting racial slant make disproportionately more arrests in white neighborhoods.
To create the measure of racial slant, I use Natural Language Processing: machine learning methods that allow me to use text as data. I use adjectives and adverbs from thousands of police reports to predict the observed suspect race with an elastic net logistic regression. The elastic net optimally combines \(\ell_1\) (lasso) and \(\ell_2\) (ridge) penalties on the word coefficients, setting many to zero. Therefore, the machine learning model implicitly pinpoints words that encode suspect race and neutral words. We interpret the closeness of the predicted and true suspect race as a measure of a police report's ability to encode race implicitly.
In work-in-progress, we use these administrative data to estimate the effect of body-worn cameras on dispatch outcomes. The police department we work with was not an early adopter of body-worn cameras and was institutionally unwilling to adopt them with a randomized control trial. Therefore, it may reflect an average police department better than one willing to implement a randomized control trial. We compare dispatches made in broad daylight to those made at night before and after the rollout of body-worn cameras, using a difference-in-differences design. Preliminary results suggest that requiring officers to wear cameras changes their propensity to initiate interactions with civilians.
Our research contributes to a growing literature on policing. We hope to provide much-needed evidence on effective policing to help policy-makers find appropriate solutions to pressing problems.
I plan to further develop my text analysis skills in future research projects. These methods are increasingly popular in social science and develop our econometric toolkit to tap into the world of text. I am also grateful for the new difference-in-differences literature clarifying how to implement staggered treatment designs. These designs fit many meaningful contexts. I plan to continue to employ these and other methods to estimate credible causal effects of meaningful policies.
Job Market Paper
Abstract: Policy can impact partner choice and match quality. Spousal visa policy permits non-residents to marry citizens, arguably an assimilation capstone. I ask how this policy affects couple rates, marriage rates, and assortative mating by citizenship and birth country. Without immigration policy variation, I identify the effect of spousal visa access by exploiting a change in the federal government’s definition of spouse. When the Supreme Court ended the Defense of Marriage Act in United States v. Windsor, same-sex couples gained access to spousal visas for the first time. I estimate the effect of this policy change for mixed-citizenship same-sex couples, accounting for aggregate changes in other same-sex and mixed-citizenship couples, using a triple difference design. Spousal visa access causes a 36% increase in coupling rates and a 72% increase in marriage rates. Transfer benefits, health insurance, roommates, moving, fraud, or state-level heterogeneity do not explain the results. Informal calculations suggest that 1.5 million people currently have partners, directly thanks to spousal visa policy.
“Officer Language and Suspect Race: A Text Analysis of Police Reports” (with Romaine Campbell)
(due to data restrictions this paper is available upon request only)
Abstract: We ask if police officers use adjectives and adverbs that systematically differ by suspect race (race-predictive language), if race-predictive language use relates to other officer characteristics, and if officers with more race-predictive language have different 911 call dispatch outcomes in Black compared with White neighborhoods. We leverage a novel data set containing police report text from a single large urban police department. We identify race-predictive language using an elastic net with word counts, then use predicted race to construct an officer-level measure of race-predictive language. By exploiting the conditionally random assignment of officers to 911 dispatches, we test whether officers with greater race-predictive language affect suspect outcomes. We find evidence that officers use different adjectives and adverbs in reports for Black versus White suspects and that such language correlates positively with officer inexperience, the number of complaints, the number of instances of use-of-force, and the racial gap in suspect search rates. We do not find a relationship between officer race and race-predictive language. Finally, we find evidence that officers with greater race-predictive language are more likely to assist and less likely to arrest in Black relative to White neighborhoods.
Works in Progress
“Marriage Legalization, Assortative Mating, and Match Surplus”
Abstract: Marriage is a social phenomenon. Marriage is also a legal contract. Does access to the legal marriage contract affect assortative mating or the surplus generated by matches? If so, marriage policy favors the creation of some couples over others. I answer this question using variation in state law. Previously states barred same-sex couples from the legal marriage contract. I calibrate a model (Ciscato, Gousse, Galichon JPE 2020) that quantifies the relative extent of assortative mating and total matching surplus across marriage markets defined by state, year, and couple type: same-sex and different-sex. I then estimate how same-sex marriage legalization affects these quantities, using a staggered diff-in-diff design.
“Body-Worn Cameras and Police Stops” (with Romaine Campbell)
Abstract: We study the effects of body-worn cameras on officer interactions in a police department that was not an early adopter of body-worn cameras and was institutionally unwilling to adopt them with a randomized control trial (RCT). We hypothesize that this sizeable urban police department may better reflect an average department than early adopters or departments willing to implement an RCT. We compare dispatches made during daylight hours to those made at night before and after the rollout of body-worn cameras. We find that officers use less force during self-dispatched calls and make fewer self-dispatches after the rollout. We do not find changes in officer use of force for 911 call dispatches.
“Risk Aversion, Offsetting Community Effects, and COVID-19: Evidence from an Indoor Political Rally” Journal of Risk and Uncertainty 63, 133–167 (2021) (with Dhaval M. Dave, Andrew I. Friedson, Kyutaro Matsuzawa, Drew McNichols, Joseph J. Sabia)
Abstract: The Centers for Disease Control and Prevention (CDC) deem large indoor gatherings without social distancing the “highest risk” activity for COVID-19 contagion. On June 20, 2020, President Donald J. Trump held his first mass campaign rally following the US coronavirus outbreak at the indoor Bank of Oklahoma arena. In the weeks following the event, numerous high-profile national news outlets reported that the Trump rally was “more than likely” the cause of a coronavirus surge in Tulsa County based on time series data. This study is the first to rigorously explore the impacts of this event on social distancing and COVID-19 spread. First, using data from SafeGraph Inc, we show that while non-resident visits to census block groups hosting the Trump event grew by approximately 25 percent, there was no decline in net stay-at-home behavior in Tulsa County, reflecting important offsetting behavioral effects. Then, using data on COVID-19 cases from the CDC and a synthetic control design, we find little evidence that COVID-19 grew more rapidly in Tulsa County, its border counties, or in the state of Oklahoma than each’s estimated counterfactual during the five-week post-treatment period we observe. Difference-in-differences estimates further provide no evidence that COVID-19 rates grew faster in counties that drew relatively larger shares of residents to the event. We conclude that offsetting risk-related behavioral responses to the rally—including voluntary closures of restaurants and bars in downtown Tulsa, increases in stay-at-home behavior, displacement of usual activities of weekend inflows, and smaller-than-expected crowd attendance—may be important mechanisms.