Police, prisons, and punishment: the empirical evidence on crime deterrence
Citation: Jonathan Klick, Alexander Tabarrok (2012) Police, prisons, and punishment: the empirical evidence on crime deterrence. Handbook on the Economics of Crime (RSS)
Internet Archive Scholar (search for fulltext): Police, prisons, and punishment: the empirical evidence on crime deterrence
US 2007, spending on police and prisons about $175 billion. US 1999 cost of crime may exceed $1 trillion.
New data sets and econometric analysis have provided new insight into variation of crime.
Authors review recent attempts to measure deterrence effect of police and prisons.
Becker (1968) stated policymakers have two tools: probability and severity of punishment. More police can increase probability, including by lowering cost of victim/bystander contacting police.
Correlation between police and crime -- do police cause crime? More likely, crime causes hiring police, and more police reduce crime. How to disentangle multiple chains of causality?
Useful quasi-natural experiment would increase or decrease police randomly in large number of jurisdictions for reasons not related to crime.
Levitt (1997) found police increase during mayoral and gubernatorial election years, and analyzed data from 59 US cities 1970-72, but McCrary (2002) found errors in Levitt's work, leaving the question open. Levitt (2002) then used police and firefighter hiring to indicate when former not a response to crime, and found more police reduce crime.
Others looked at effect of locally increased police in response to terror or terror alert, finding more police reduce crime, but reliant on data from single shock and/or single jurisdiction, and possibly displacing crime.
Another approach is to look at funding shocks.
Others have tried to look at police effort, rather than just staffing levels, by comparing crime after police have won or lost a salary arbitration.
Overall it seems more police largely reduce opportunistic crimes, in the short term, but long term not studied empirically, and property and violent crime ought be related.
Author claims given above studies, an elasticity of -0.35 for police with respect to crime is not unreasonable. What policy implication would 10% more police causing a 3.5% reduction in crime have? Assigning costs to crime, and average salary of $54k to police, number of police should be approximately doubled in US.
Econometric research on prisons in crime in 2 categories: number/duration of prisoners, conditions of prisons.
First has to disentangle deterrence and incapacitation.
Levitt (1996) looked at shock of prisoner release due to overcrowding lawsuits, found release of single prisoner causes 15 crimes, but it is questionable how much of a random shock lawsuits are. Another approach is to look at sentencing changes that only impact some prisoners. Another is to look at recidivism of individual prisoners following sentencing rule change, finding deterrence effects for 3-strikes laws.
Effect of prison conditions more ambiguous.