To Serve and Collect: The Fiscal and Racial Determinants of Law Enforcement

Latest twitter thread (via https://threadreaderapp.com/thread/1039876308526874626.html)

New (accepted!) paper and thread: “To Serve and Collect: The Fiscal and Racial Determinants of Law Enforcement” by myself, Thomas Stratmann, and @ATabarrok, forthcoming in the Journal of Legal Studies 1/17
papers.ssrn.com/sol3/papers.cf…

Punchline: when local governments are running budget deficits, black and Hispanic arrest rates increase, while white arrests remain (mostly) unchanged, *but only if* local police are legally able to retain forfeiture revenues in their budget 2/17

To better understand our results, I’d like to walk through two important *correlations* first, and then our discuss our strategy for identifying a *causal* relationship between fiscal incentives and arrests 3/17

Correlations: 1) Local government fine and forfeiture (F&F) revenue increases more with drug arrests than other types of arrests (not surprising, drug exchanges require cash), 2) F&F revenue increases faster with black and Hispanic drug arrests than white drug arrests🤔. 4/17

The positive correlations between drug arrests and fine and forfeiture revenues, especially for black and Hispanic drug arrests, show that police departments have the motive and opportunity for revenue driven law enforcement 5/17

What about means? There are only so many murders/robbery arrests to make. Drug arrests, however, are more of a police choice variable, able to be ramped up almost at will. Thus, in addition to motive & opportunity, police also have the means for revenue driven enforcement 6/19

Potential explanations: different types of drug crime, racial bias in arrest category/fines/bail, rates of guilty pleas, sentence bargaining, charge reductions, reliance on public defenders, and the economic & racial determinants of successfully challenging at trial 7/17

See nber.org/papers/w24579 by @amandayagan, Freedman, & Owens, or scholar.harvard.edu/files/cyang/fi… by Arnold, Dobbie, and Yang, for related work. Lots more cited in the paper! 8/17

Regardless of the source(s) of these differentials, they point towards important differences in police incentives. To tell a causal story, however, we need variation in these incentives. Our ID strategy leverages laws that allow police to retain $$ from seized property 9/17

The catch: seizure laws are not randomly assigned. States with larger black populations are more likely to allow their police to retain seizure revenue🤔. 10/17

Solution: budget deficits. Specifically, we include the effect of seizure laws and local budget deficits in our regressions, and then identify off of the interaction of the two i.e. the marginal impact of additional budget deficit in states where seizure revenue is retained 11/17

We find in states where seizure revenues are retained i) black and Hispanic arrests for drugs, DUI, and prostitution arrests are all increasing with local govt deficits. ii) white arrests for drugs and DUI are *entirely insensitive* 12/17

Side note, iii) black, Hispanic, & white prostitution arrests are all increasing w/ deficits (where seizure $$ are retained) at a similar rate. Prostitution arrests are *far* less numerous than drug or DUI arrests, but it’s still interesting (is this surprising @causalinf?) 13/17

Similarly, we find that the rate of property seizures from blacks and Hispanics are both increasing with deficits in states where police can retain revenues 14/17

While white arrest rates are largely unchanged, we do observe increases in the rate of seizures from whites (the magnitudes are 50% smaller and much noisier (p<0.10)). Still, this suggests that white arrestees are not immune from these police incentives 15/17

Limitations:1) our data and ID strategy are coarse. 2) we’re only able to make causal inferences at the margin. 3) We use deficits to *identify* the problem, but the problem of revenue driven law enforcement probably exists wherever police can retain revenues 16/17

Our results raise questions about the wisdom of non-revenue neutral law enforcement. It’s hard to imagine “optimal deterrence” or “justice”, when each potential arrestee is measured not just by the weight of their transgression, but by how much $ can be extracted from them 17/17

[Mike Editorializing 1] There are *so many* issues at the intersection of race and law enforcement, but how often do we come across one for which there is a direct and immediately accessible policy fix? 1/2

[Mike Editorializing 2] In a perfect world, law enforcement would be revenue-neutral, but in this case there is a 2nd-best solution on the sidewalk: F&F $$ should never be retained by the police. Put it in general/state budget. Fungible, sure, but the flypaper effect is real 2/2

Politics, Unemployment, and Immigration Enforcement

The Clemson School of Business and Behavioral Sciences put out a nice press release about my 2014 paper with Thomas Stratmann looking at the political determinants of immigration audits. The paper is a year old, but the subject has been made relevant recently thanks to a particularly awful presidential candidate whom I will not name. Suffice it to say, there is no shortage of ways for politicians to blame immigrants for our ills. And, as our paper seeks to demonstrate, immigration enforcement presents  another opportunity for enforcement agent discretion to be internalized by government principals and transformed into political capital.

2014 Center For Advanced Modeling Graduate Workshop

BRINGING TOGETHER THE NEXT GENERATION OF COMPUTATIONAL SOCIAL SCIENTISTS
JOHNS HOPKINS UNIVERSITY, BALTIMORE, MD, JULY 10-12, 2014

Over the last 20 years, agent-based modeling has progressed from an avant-garde invasion into the social sciences to a widespread methodology used to make contributions of remarkable interdisciplinary range. Young modelers can often find themselves on a methodological island within their departments, confronted by the institutional barriers limiting their interactions with the methodologically like-minded in other departments and disciplines. The Center for Advanced Modeling (CAM) Graduate Student Workshop is an opportunity for students to present work, at a variety of stages, to an audience of their intellectually and methodologically diverse peers, as well as senior faculty from a range of fields. The goal of the workshop in to create a setting in which students can present their work to faculty and other students, share ideas, and begin building the interdisciplinary network of colleagues and co-authors necessary for success.

To this end, we invite working papers, dissertation chapters, and recently published work that leverages agent/individual –based computational modeling from a variety of disciplines including, but not limited to:

  • Economics, Political Science, Sociology, Demography, and History
  • Ecology, Evolutionary Biology, Epidemiology, and Geography
  • Systems, Industrial, or Biomedical Engineering
  • Computer Science, Applied Mathematics, and Network Science

We will be limiting the scale of the workshop in order to allow for an informal and lively discussion of the papers and ideas. Participation will be free of charge, and lodging and meals will be provided for all participants, but participants will need to cover their own travel expenses. Presentation slots are for students only, but there will be lodging available for a handful of senior faculty as well. Potential participants are asked to send an extended abstract (500 -1000 words) or a full paper with a short abstract tocenterforadvancedmodeling@gmail.com before May 10, 2014. Notifications about paper acceptance will be sent out before May 15, 2014.

The work shop agenda will include 5 2-hour sessions over 2 days, each featuring 2 graduate student presentations, as well as:

  1. A presentation by CAM director Joshua Epstein of his new book Agent_Zero: Toward Neurocognitive Foundations for Generative Social Science, Princeton University Press. An NIH Director’s Pioneer Award (DP1) funded both his book and this Workshop, which will produce new ideas and collaborations continuing this line of research.
  2. A Saturday morning round table discussion on “Building a research agenda and academic career as an agent-based modeler” lead by Michael Makowsky (Johns Hopkins University)