1 / 23

Benefit-Cost Analysis for Crime Policy

Benefit-Cost Analysis for Crime Policy. Roseanna Ander Executive Director, University of Chicago Crime Lab & Jens Ludwig McCormick Foundation Professor, University of Chicago. Why benefit-cost analysis (BCA)?. Crime control is costly…

memanuel
Télécharger la présentation

Benefit-Cost Analysis for Crime Policy

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Benefit-Cost Analysis for Crime Policy Roseanna Ander Executive Director, University of Chicago Crime Lab & Jens Ludwig McCormick Foundation Professor, University of Chicago

  2. Why benefit-cost analysis (BCA)? • Crime control is costly… • With >2 million people behind bars, U.S. has highest incarceration rate in world by far • 2006 expenditures on criminal justice = $214 billion • Ignores all the (surely massive) non-monetary social costs of crime control, such as impacts on distressed communities • With massive budget deficits at every level of government, growing interest in scaling back

  3. Why benefit-cost analysis (BCA)? • But crime itself is also costly… • $1 to $2 trillion per year • Disproportionately affects most disadvantaged among us (ex: leading cause of death, blacks 15-24) • How much crime control should we do? • This is one important use of BCA for crime policy: • Helps us weigh benefits of scaling back criminal justice system, vs. costs of having more crime • Good BCA counts non-monetary impacts as well

  4. Second important use of BCA for crime policy • Figure out how to get maximum amount of crime control for given costs of crime control • (ex) 2006 we spent $98.9b on police, $46.9b on judicial / legal, $68.7b on corrections • Is this remotely close to optimal mix? Could we get more crime control w/ same or lower cost?

  5. An example from the MIT Poverty Lab (2008) • What are the costs of obtaining an additional child-year of education in developing countries? • Conditional cash transfer program • Hiring extra teachers, class size 80→40 • De-worming treatment

  6. An example from the MIT Poverty Lab (2008) • What are the costs of obtaining an additional child-year of education in developing countries? • Conditional cash transfer program ($6,000) • Hiring extra teachers, class size 80→40 • De-worming treatment

  7. An example from the MIT Poverty Lab (2008) • What are the costs of obtaining an additional child-year of education in developing countries? • Conditional cash transfer program ($6,000) • Hiring extra teachers, class size 80→40 ($200) • De-worming treatment

  8. An example from the MIT Poverty Lab (2008) • What are the costs of obtaining an additional child-year of education in developing countries? • Conditional cash transfer program ($6,000) • Hiring extra teachers, class size 80→40 ($200) • De-worming treatment ($3.25)

  9. An example from the MIT Poverty Lab (2008) • What are the costs of obtaining an additional child-year of education in developing countries? • Conditional cash transfer program ($6,000) • Hiring extra teachers, class size 80→40 ($200) • De-worming treatment ($3.25) • Lessons: • Each intervention ‘works,’ but social good from given $ can be increased thousand-fold by allocating resources to most efficient uses • In resource-limited world, crucial need to know which interventions are most cost-effective

  10. How to obtain benefit-cost estimates? • Two steps: • Measure the causal impact of a policy intervention (measured in natural units like arrests, etc.) • Convert impacts in natural units to common dollar metric, to facilitate comparison to costs • Each of these steps is much more challenging than they might initially appear…

  11. How to measure the causal impacts of policies? • Selection bias concern (e.g., effects of Boy Scouts) • Randomized clinical trials solve this for sure: • Treatment and control groups randomly assigned • B/c of random assignment, any difference in average outcomes for two groups unambiguously due to the “treatment” (program or policy) • Are clinical trials only way to measure impacts? • “Randomistas” (e.g., Larry Sherman) vs. “Research pluralists” (e.g., Rob Sampson)

  12. Why this matters hugely for crime policy • No guarantee bias need be small • (Ex) Hormone Replacement Therapy (HRT) • For many years, encouraged for post-menopausal survivors of breast cancer • Based on observational (epidemiological) studies that controlled for rich set of covariates • OK if we have good theory, but I think our theories quite bad (so this design is like “regression-adjust and pray”) • Finally a randomized trial of HRT was carried out…

  13. Why this matters hugely for crime policy • No guarantee bias need be small • (Ex) Hormone Replacement Therapy (HRT) • For many years, encouraged for post-menopausal survivors of breast cancer • Based on observational (epidemiological) studies that controlled for rich set of covariates • OK if we have good theory, but I think our theories quite bad (so this design is like “regression-adjust and pray”) • Finally a randomized trial of HRT was carried out… • Had to be stopped early (HRT increased risk of cancer 3x) • BCA of biased impact estimate leads to bad decisions

  14. Middle ground? • Not “RCT or bust,” but rather, aspiration for as much “design-based” research as possible • Institutional knowledge about how / why some groups but not others receive “treatment” • Plausible that assignments unrelated to potential outcomes (conditional on observable variables) • Feasible “natural experiment” crime examples: • Lotteries • Discretion creating nearly random assignment • Assignment through prioritization • Policy changes over time

  15. Lotteries (Already common in education, housing) • Often excess demand for a government service • Allocate limited resources among eligibles by lottery • Note randomization not inconsistent with prioritization, b/c there is eligibility screen (key is, do # eligibles > # slots?) • Compare eligible persons who did vs. did not win lottery (groups similar aside from lottery outcome) • Example from University of Chicago Crime Lab: • Cook County Juvenile Temporary Detention Center, convert 500 bed facility to run like 10 50-bed facilities • Half of residential units currently therapeutic • Randomize youth for whom JTDC no strong priors

  16. Discretion • Institutional structure of criminal justice system creates quasi-lottery • Decision-makers possess discretion, and vary in their propensity to assign “treatments” • Defendants randomly assigned to decision-makers • Examples from Chicago: • Judge payola scandal, so now randomization to judges • Charles Loeffler (2010), judges vary in prison sentence lengths (judge as instrumental variable for estimating effects of prison time on labor market outcomes) • Judges also vary in assigning alternatives to detention

  17. Assignment through prioritization • Often occurs when reducing decision-maker discretion is a goal • Keys are ranking of people / places, and fixed thresholds in treatment decision • People or places prioritized for some intervention on basis of some scalar score (like predicted risk) • Those just above threshold receive one treatment, those just below receive a different treatment • Person just above & below threshold nearly identical, difference in ranking in narrow range close to random • Regression discontinuity – compare above / below threshold

  18. (Ex) Chen and Shapiro American Law & Economics Review, 2007

  19. Policy changes over time(Ex) Ludwig & Cook, 2000 JAMA

  20. Second crucial step for BCA • Convert impacts in natural units (arrests, dropouts, etc.) to dollar metric • Without this, no way to compare benefits to costs (which are already in dollars) • Dollar metric does not imply that only tangible “economic” costs matter! • Indeed, intangible costs drive social costs of crime • (Ex) Cook and Ludwig 2000, social costs of gun violence each year ~$100b • Medical costs & lost earnings tiny share of that

  21. 3 methods for measuring intangible costs • Jury trial awards • Not a representative sample of crimes • “What would it take to make victim whole?” not the right question for crime policy • Labor & housing market risk / price tradeoffs • Omitted variables concerns • Plus wrong target (miss altruistic concern for others) • Contingent valuation (CV) – hypothetical market Qs • People have incentives to think about this already (labor & housing markets), but hypothetical behavior is hypothetical • High priority: Learn more about CV in crime application (compare to environmental area)

  22. Why BCA is so important • Each extra $1 spent on police, ~$4 to $8 worth of ↓ crime (B/C ratios of 4:1 to 8:1) • From study of COPS program, Evans & Owens (2006) • BCA elaboration from Donohue & Ludwig (2007) • Now consider secondary prevention experiment Crime Lab carried out in Chicago • Social-cognitive skill program for at-risk middle school and 9th grade boys in Chicago Public Schools • $ value of crime impacts imply B/C ratio of 1.4:1 to 9:1, but imprecisely estimated, not significant • Considering impact on high school graduation as well suggests B/C ratio of 8:1 (upper end of police range)

  23. Conclusions • Benefit-cost analysis only makes sense if we are sure we have good estimate for policy impact • Since intangible costs drive social costs of crime, crucial that we learn more about how to measure those (and intangible costs of crime control also) • Without better crime policy analysis (causal inference and benefit-cost analysis) we are blinding policymakers to the difficult tradeoffs they face in this area

More Related