Jennifer Doleac on Crime

Episode Summary

Executive Summary: Russ Roberts and economist Jennifer Doliak examine how technology is reshaping criminal justice, focusing on surveillance tools, risk assessment, body cameras, DNA databases, and re-entry policy. The conversation highlights a recurring theme: many well-intended interventions have mixed or null effects, can create unintended consequences, and should be judged by causal evidence rather than intuition alone.

Main Topics: Technology in criminal justice (Priority: 5/5): Doliak explains how cameras, GPS monitors, DNA databases, and algorithmic tools are increasingly used in policing, supervision, and sentencing, while warning that their effects can differ from expectations. AI, risk assessment, and sentencing (Priority: 5/5): The discussion covers algorithmic risk scores, their potential to reduce bias compared with human judgment, and the danger that they can embed existing disparities under a false veneer of scientific objectivity. Body-worn cameras and police accountability (Priority: 5/5): They review evidence on police body cameras, including randomized trials showing little average effect on officer behavior, while noting that policymakers may still value them for transparency, rare-event documentation, and public trust. DNA databases and deterrence (Priority: 4/5): Doliak summarizes research showing that expanding DNA databases sharply lowers recidivism by increasing expected detection, while also raising privacy and civil-liberties concerns. Incarceration, rehabilitation, and prison quality (Priority: 4/5): The hosts discuss how prison affects later employment and recidivism, emphasizing that effects depend heavily on context, sentence margin, and prison conditions; some evidence suggests marginal incarceration in the U.S. can harm outcomes. Re-entry policy and 'ban the box' (Priority: 4/5): Doliak explains that removing criminal-history questions from job applications can backfire by prompting employer statistical discrimination, reducing employment for young Black men without college degrees. Economics, causality, and unintended consequences (Priority: 5/5): The conversation closes with a defense of economics’ focus on incentives, trade-offs, and causal identification, while acknowledging the profession’s tendency to overvalue contrarian findings and underestimate unmeasured costs.

Key Arguments: Technology in criminal justice should be evaluated empirically because new tools do not automatically improve outcomes and may create financial, privacy, or fairness costs. Algorithmic risk assessments may improve over human hunches by reducing arbitrary bias, but they can also reproduce existing bias and gain unwarranted legitimacy from appearing scientific. Body-worn cameras are appealing as transparency tools, but randomized evidence often shows little average impact on complaints, use of force, or other daily outcomes; their value may lie in rare-event documentation and accountability. DNA database expansions appear to deter reoffending substantially by raising the probability of detection and arrest, producing large reductions in recidivism. The main concern with DNA databases is not the data string itself but the privacy and surveillance implications, including possible expansion into familial search and private genealogy databases. Incarceration on the margin often harms employment and other life outcomes in the U.S., implying that current incarceration rates may be too high. Prison effects vary by institution and country; Norway is highlighted as a case where prison can function more like rehabilitation and job training than punishment. Ban-the-box policies can backfire because employers, unable to ask directly about records, infer criminal history using race and other proxies, reducing hiring of Black men. Well-meaning wraparound re-entry programs may fail or even worsen outcomes if they reduce agency, take up time, or are not well targeted. Economists contribute by forcing explicit trade-off analysis and causal evaluation, even though they may be too attracted to surprising negative unintended consequences.

Data Points: Date of episode: December 13, 2018 - Opening of the EconTalk episode Employment effect of ban the box: 5% decline - Young Black men without a college degree saw employment fall in the years after ban-the-box policies were adopted Recidivism reduction from DNA databases: about 40% - Doliak cites Danish evidence that recidivism falls dramatically when offenders are added to DNA databases Body-worn camera RCT result: no effect - The Washington, D.C. randomized trial found no meaningful change in officer behavior across many outcomes Research frontier: current U.S. data limitations on criminal-record linkage - Doliak notes the lack of census or survey questions on criminal records makes U.S. causal research difficult More police and crime: more police equals less crime - She notes this is a settled finding in economics based on causal studies Prison context: Norway prison leads to better outcomes - Used as an example of how prison quality and services affect post-release life

Pivotal Quotes: "We don't have to come up with a tool that's perfect. We just have to come up with a tool that's better than the status quo." — Jennifer Doliak: On algorithmic risk assessment and comparing new technology to human judgment "We should assume everything we try will fail." — Jennifer Doliak: On policy experimentation, evaluation, and avoiding premature attachment to programs "It's not worth it, it is still not worth it. If we're saving X dollars every year due to reduced crime and you look at that number and you say, I don't know, it's still not worth it." — Jennifer Doliak: On using evidence to benchmark privacy costs against crime reductions from DNA databases

Implications: Listeners should expect criminal-justice technologies to have mixed effects and demand rigorous evaluation. Policymakers should design interventions with explicit goals, monitor unintended consequences, and be willing to abandon programs that fail—even when they are popular or well intended.