Less Prison More Policing Less Crime How Criminology

Less Prison + More Policing = Less Crime How Criminology Can Save the States from Bankruptcy Lawrence W. Sherman Jerry Lee Centers of Criminology Universities of Cambridge and Pennsylvania April 21, 2010

Summary If criminal justice policy were based only on current evidence of its cost-effectiveness, we would have • • Less money spent on prison More money spent on police Fewer Serious Crimes, Less Total Harm Fewer states threatened with bankruptcy

Cost-Effectiveness of Prison? 1. General Deterrence? Unknown 2. Specific Deterrence? Unknown (Weak evidence of No Effect, or Net Increase) 3. Incapacitation: Wasted on Many. • Most people want “Willie Hortons” in prison • Relatively few people in prison are “Willie Hortons” • Many “Willie Hortons” get probation, parole • Only Virginia has adopted risk-based sentencing • Even Virginia fails to use seriousness as key factor

Cost-Effectiveness of Police? • General Deterrence: Hot Spots Patrols (displacement debate) Traffic and minor crime enforcement Possibly total N of police • Incapacitation Serious Repeat Offenders • Situational Crime Prevention: POP • Specific Deterrence: More diversion, less court

Less Crime? Lock Up The Worst, Manage the Rest • Lock Up The Worst: Less Serious Crime • Manage the Rest: More Surveillance by Police with probation & parole • Criminal Events vs. People Who Commit Crime 1. Individuals Events in Hot Spots 2. Police presence deters events 3. POP alters causal structure of events • Crime Victims Prefer Diversion, Also Reducing Crime

Criminologists at Work Daniel Nagin Joan Petersilia Frank Cullen Anthony Braga Richard Berk David Weisburd Kate Bowers Chris Koper

Outline 1. Premises 2. Key Concepts --“Push-button” Policy Systems -- Crime Harm Index (CHI) -- Risk-Based Policies --A General Theory of Crime, Prisons & Police 3. Evidence on Prisons 4. Evidence on Policing 5. Who Can Push the Buttons? How?

1. Premises: What Criminal Justice Needs People Need Criminal Justice to 1. Protect them from loss of liberty to be secure 2. Hold offenders accountable for their crimes 3. Cost taxpayers as little as possible This Requires a System of Criminal Justice that a. Produces consequences for every decision that are b. Predictable, based on good evidence, and c. Chosen democratically, by managing the system

2. Key Concepts A. “Push-Button” Policy Systems B. Crime Harm Index: CHI C. Risk-Based Policies D. General Theory of Crime, Prisons & Police

2. a. “Push-Button” Policy Systems

Contrast Before Keynes • • Case Studies Trends Histories Explanations Micro-level studies Little intervention No “buttons” to push Economy beyond control After Keynes • Government has key buttons • Pushing can help a lot • Which ones to push? • When to push them? • What effect predicted? • How do buttons affect each other?

Push-Button Criminology • Case-by-case crime policy now “Laissez-Faire” • Interventionist policy: more control of trends • Key buttons to push: prisons, police, probation & parole • Pushing wrong buttons, case-by-case: research evidence shows --Too much prison for too little benefit --Not enough optimizing of police • Pushing right buttons requires system policy • Guided by a general theory

A General Theory 1. Macro-economics: --societies as the unit of analysis --not individuals --not transactions 2. Rates across societies: --Independent variables (causes) --Dependent variables (effects) Money Employment

Neo-Keynesian Economics Causes (Independent) • • • Money supply Interest Rates Tariffs Taxation Deficits Effects (Dependent) • Employment • GDP • Gross National Happiness (Bhutan): --crime --disease --mortality --inequality

Macro-Criminology Causes (Independent) Effects (Dependent) • Prison • Crime • Police • Detection Rates • Probation/Parole • Crime Harm Index (CHI)

2. b. Crime Harm Index ( CHI )? • • Not UCR Index—Challenge to BJS! (US Senate) Like Consumer Price Index (CPI) CHI Elements weighted according to formula Sum of each (Crime X Cost = Cost per event) Divided by Population (CHI person) Murder = $5 million, Burglary = $1, 000 British Crime Survey uses some costs of crime Could also be based on public opinion data Limited to reliably counted categories

Beyond Neo-Classical Deterrence Assumptions that • All crimes cause equal harm • All offenders commit crimes of equal harm • All prisons prevent equal harm per inmate • All police deter equal numbers of crimes • General, not specific, deterrence rules most • Punishment never escalates crime, just deters

BAD “Push-Button Criminology”

Total Incarceration Rate

Non-Police Data: Different Trend (BJS—NCVS)

Police Count Crime Count

Not Complex Enough for CHI • Prisoners vary highly in risk of “Willie Horton” crimes • Incapacitation effects on Crime Harm Index vary by risk level of people imprisoned • Imprisonment effects on CHI vary by age, first offense, etc. • Police impact on CHI varies highly by what they do • Especially vary in allocating time by risk level to 1. places 2. offenders 3. victims • Theories of CHI Must Focus on Risk of Serious Harm

2. c. Risk-Based Policies • Police • Prosecutors • Sentencing; Prisons • Parole and Probation

Peaks and Valleys of Crime (Distribution of Violent Offenses in Tokyo)

A “Hockey Stick” Cumulative Curve: The “Power Few”

Normal Distribution

The Willie Horton Problem: Over-Prediction of High Risk Extreme cases put low-risk people in prison • “Just in case” • No Prediction More False Positives -------------- • Response Assumes Normal Curve • Increases dosage for all, not few • Rising Prison Rates • Not needed if responding to the “power few” •

Risk-Based Policy: Foundation of Cost-Effectiveness • • Pew Trust Report RAND 1982 Report on Selective Incapacitation Idea rejected by 1986 NAS Report on Error False positives too high for values But prison rate has tripled False positives are embedded in sentencing Actuarial Risk could get them out, not put them in

Since 1986: New Generation of Risk Forecasting • Based only on Prior Charges, Residence, Age, Sex—no PSR • Nothing qualitative • More like a short-term weather forecast • Based on huge samples • E. g. , 30, 000 in Philadelphia • Journal Royal Statistical Society, Series A, 2009 Richard Berk

Clinical vs. Statistical Forecasts • Since Meehl 1954 Clinical Versus Statistical Prediction • Statistical Prediction beats Clinical • Most contests before statistical learning • Cheap supercomputers • “Data mining” that assays every possible pattern of precursors • Identifies best prediction combinations

2 -Dimensional Risk—or More? Low Seriousness Low High 1 2 High 3 Frequency* *Not probability (prevalence) 4

Geoffrey Barnes (2007) 2 -Year Berk Forecast Test, Philadelphia APPD Cases High Risk (2%) Neither High nor Low Risk (38%) Low Risk (60%)

High Risk 2% vs. Bottom 60% Two Years From Forecast Date Charges for Any Offence 8 X more Charges Serious Offence 10 X more Charges Murder or Attempt 75 X more

Group Average Number of Charges for ANY Offense Within Two Years of Probation Start High Neither Low

Average Number of Charges for SERIOUS Crimes Within Two Years of Probation Start High Neither Low

Average Charges for MURDER or Attempted Murder Within Two Years of Probation Start High Neither Low

Total Charges: High Risk 2% Over 2 Years For every 100 Offenders: --37 charges of murder or attempted murder --318 charges of serious crime (murder, rape, robbery) --1800 charges for all crimes

2. d. A General Theory of Crime, Prisons & Police 1. The higher the proportion of inmates who are highfrequency and high-harm, the more cost-effective prisons will be in lowering CHI. 2. The higher the proportion of police time spent on high-risk places, victims and offenders, the more cost-effective police will be in lowering CHI. 3. The more low-risk offenders police hold accountable for their crimes without using prison, the more cost-effective police at reducing CHI.

1. Prisons The higher the proportion of inmates who are high-frequency and high-harm, the more cost-effective prisons will be in lowering CHI. NB: Prisons would have to have many fewer inmates to become more cost-effective

2. Police Time The higher the proportion of police time on high -risk places, victims and offenders, the more cost-effective police will be in lowering CHI. NB: a. Just adding police is not predicted to cause less crime or harm b. Reductions in crime harm could come with even fewer police, depending on their focus

3. Police Dispositions The more low-risk offenders police hold accountable for their crimes without using prison, the more cost-effective police will be at lowering CHI. NB: If police stop using arrest and prosecution where tested alternatives are available, they will be more cost-effective without spending more money.

My Title is Not Quite Right Less prison, more police = less crime? Not necessarily Easier to say it that way But risk-based policy is crucial Theory of cost-effective CHI effects depends on it Thus, conditional on optimizing risk-based allocations, • More prison could, with More Police = Less Crime • But More Prison would not Cost Effectiveness • • •

Taking Cost Into Account • Prison rate hits diminishing returns quickly • Very few offenders per 100 seem worth locking up • Look at the evidence: prisons and police

3. Evidence on Prisons If we use risk analysis, • Who Goes to Prison? • Who Doesn’t? Preliminary analysis in Philadelphia suggests • half of all sentenced to prison are low-risk, yet • half of all high-risk offenders not given prison

Nagin and Cullen, 2009 Crime and Justice Vol. 38 Daniel Nagin Frank Cullen “Special” Effects of Imprisonment on the Imprisoned: • 5 Experimental studies of custody more crime • 11 Matching studies juvenile custody more crime • Propensity Score Matching mixed effects

4. Evidence on Police: Clear effects, More on Costs Needed a. Places 1. Hot Spot Patrols 2. Problem-Oriented Policing 3. Displacement b. Offenders High-frequency, high-harm c. Dispositions Youth offenders Domestic Violence Restorative Justice

Campbell Collaboration • • • Systematic Reviews Many with meta-analyses Peer-Reviewed NIJ support (UK, Holland, Canada, Sweden) Chinese and English versions Norwegian Government website: www. campbellcollaboration. org

a. Policing Places 1. Anthony Braga review: Increasing police in high risk places reduces crime & disorder there 2. Koper & Mayo-Wilson: policing gun crime places less gun crime 3. Weisburd, et al: Problem-Oriented Policing POP less crime & disorder 4. Displacement: Weisburd, Bowers, others

b. Policing High-Risk Offenders Two Randomized Controlled Trials: Police Foundation (Washington) RAND (Albuquerque, NM)

c. Detection, Dispositions, Accountability Campbell Reviews: • Juvenile Prosecution: Petrosino et al 2010 • Domestic Violence: NIJ experiments • Restorative Justice: UK, Australia, US • Shapland, et al 2008 Cost effectiveness on RJ in UK = 8 to 1 return

Grounded Theory of Crime, Police and Prisons • Police and prisons may cause crime as well as prevent it • Effects of police and prisons on crime depend on who is treated how • In general, prison does more harm than policing • In general, police do more good than prisons • Precise decisions needed to reduce harm, increase prevention.

5. Who Can Push the Buttons? How? 1. Governors and State Legislatures --Sentencing Guidelines --Parole Violations (less stick? ) --Bigger carrot: State funding more local police with prison savings 2. County Prosecutors & Judges --Diversion from prosecution , Based on risk --Incentives of more police in their county 3. Local Police Chiefs --Hot spot patrols --Problem-Oriented Policing (POP) --Restorative solutions 4. Federal Leadership: --showing the way --funding more police, as incentives to states

Bankrupting States NY Times: Schwarzenegger Seeks Shift From Prisons to Schools (by Jennifer Steinhauer Jan 6, 2010 ) • State Constitutional Amendment • Referendum • No more money to prisons than to U of Calif. • But how to do it? • Privatization? Probably not.

HOW? Political Evidence • Reducing prison by early release may backfire • IF Money appears to be only motive • No analysis of public safety • No claim that less prison could less crime

Front-End Beats Back-End Arrest Prosecution Sentencing Guidelines Risk Analysis at each step—before prison. “These people were not sent to prison on purpose!” • Why? Because it would have wasted your money. They weren’t dangerous enough. • • •

UK Early Release • • 14 Days before sentence to end 2007 -2010 Many crimes committed 2 weeks after release But same would be true after full term No evidence that net crime went up But no randomized trial to test that theory Result: Labour Prime Minister blamed

The Politics of Early Release

Emergency Release--US • NY Times: (Monica Davey, March 4, 2010) “Safety Is Issue as Budget Cuts Free Prisoners” • Oregon suspends program—radio ads • Illinois program a “big mistake. ” • Colorado—$14 million savings lost • Michigan—big drop, big challenges

Carrots and Sticks • Cost-saving is a stick, at front end or back • More police is a carrot • Cohen’s national data show people prefer to spend more money on police than prisons • Use some savings on prison budgets to hire more police

A Federal Carrot? • Plan A: Federal program could match local funds invested in policing out of less prison for low-risk offenders. • Plan B: Federal program could match local funds invested in policing high-harm targets • Plan C: Both A and B • Federal buttons to lower Crime Harm Index

Federal Leadership: Research and Development 1. More support for automated risk forecasting 2. More Randomized trials, COST-EFFECTIVENESS, • • • Risk-Based Policing Risk-Based Charging Risk-Based Sentencing Risk-Based Parole Release Risk-Based Parole Revocation

Beyond a Crisis Bankruptcy is a bad reason to do this. But it is a good opportunity. If the result is less crime, cost issues may fade The Obama administration has made history in health care • Can we now make history in criminal justice as well? • • My answer: Yes we can.

THANK YOU Lawrence W. Sherman Jerry Lee Centers of Criminology Universities of Cambridge and Pennsylvania April 21, 2010