Question for help
Chapter 12
Economics of Incarceration: Deterrence and Incapacitation
Introduction
For most of human history, corporal punishment, using force to cause pain, was used to enforce criminal law. This was used because it was an effective deterrent and it was low cost.
Incarceration has become the dominant sanction over the last 300 years
Incarceration influences crime through deterrence, incapacitation, and rehabilitation.
Introduction
Incapacitation – changes in crime caused from removing criminals from the general public
Deterrence – changes in offending due to disutility associated with incarceration
Rehabilitation – Providing services such as drug treatment, education, and recreation aiming to provide skills that will allow them to leave criminal behavior after serving their sentence.
Introduction
Deterrence and rehabilitation can work in ways that conflict since rehabilitation can make incarceration attractive
https://abcnews.go.com/Health/Wellness/nc-man-allegedly-robs-bank-health-care-jail/story?id=13887040
Testing the Deterrent Effect of Police on Crime
Increasing the number of police officers increases the probability of arrest and conviction. This not only has a deterrence effect on crime, but also an incapacitation effect since the incarcerated population is increased as well.
Testing the Deterrent Effect of Police on Crime
A naïve OLS regression finds positive relationship between the rate of police and crime rates. This seems to indicate that hiring more police will increase crime rates.
Testing for a causal relationship requires an exogenous shock.
Evans and Owens (2007) uses variation in COPS funding to estimate the effect of police on crime
Testing the Deterrent Effect of Police on Crime
COPS is a federal program that sends grants to local law enforcement agencies to fund additional police officers for 3 years.
The allocation of grants is not based on current crime rates.
Evans and Owens (2007) uses variation in the size of grants as an instrument for police officers
Their results estimate the elasticity of crime with respect to police force size to be -0.26 for property crime and -0.99 for violent crime
In other words, a 1% increase in number of officers per 10,000 decreases property crime by 0.26% per 10,000 people and violent crime by 0.99% per 10,000 people.
Testing the Deterrent Effect of Police on Crime
Mello (2019) performed similar analysis on the COPS program except around the years of the great recession.
Funding for COPS slowed in 2000 and ended in 2005. Once the recession hit, one billion dollars was allocated towards COPS to combat potential increases in crime from the recession.
Qualifications for COPS Grant
Applicants submitted statistical information on
Fiscal Health
Local Unemployment and Poverty
Local Crime Rates
Community Policing Strategies
Applicants were then given a score
Scores above a cutoff were given funding
Higher scoring cities were usually larger, poorer, and had higher crime rates
COPS Grant Distribution Rules
Required to allocate
at least 1.5% of funding to each state
at least 50% of total funding to jurisdiction with populations exceeding 150,000
These rules created variation in cutoff score by state and population of jurisdiction
This further strengthens the argument that these grants are a valid instrument since jurisdictions may have identical crime rates, but their eligibility will depend on their state and the population of the jurisdiction.
Testing the Deterrent Effect of Police on Crime
On July 18th, 1994, a terrorist attack destroyed the main Jewish center in Buenos Aires, Argentina. One week later the, the government began providing 24-hour police protection at 270 Jewish and Muslim institutions throughout the city.
This creates a natural experiment to understand the effects of reallocating police within a city.
Tella and Schargrodsky (2004) estimates that automobile theft decreased 75% in blocks with 24/7 police protection relative to areas that were one or two blocks away.
Testing the Deterrent Effect of DNA Databases on Crime
Beginning in 1988 states began passing legislation to create databases of criminal offenders’ DNA profiles.
The goal of these were to catch reoffenders more quickly and defer profiled offenders from committing more crime by raising the probability of getting caught/conviction.
In 1988, Colorado began taking DNA samples on some sex offenders
In 1989, Virginia began requiring DNA samples for sex and violent offenders
By 1999, every state had an established offender DNA database.
Testing the Deterrent Effect of DNA Databases on Crime
Using variation in the timing of the introduction these databases, Doleac (2017) estimates the effect DNA databases on crime rates and on the probability of a convict recidivating.
Interpretation of Results
These results imply that for violent criminals, person whose DNA is in the database is 3 percentage points less likely to recidivate within 3 years of release than a person whose DNA is not.
And 5 percentage points less likely that they will recidivate within 5 years.
The results showed weaker and less consistent results for violators of property crimes.
Estimating the Effect of DNA Database Size on Crime Rates
DNA databases are designed to affect crime by increasing the probability that a known offender gets caught if they reoffend
The size of the database should be highly correlated with the probability that a criminal is in the database.
Problem: database size as a treatment will have reverse causality bias since states with higher crime rates will have larger databases creating an upward bias on the effect of database size on crime rates
Instrument for Database Size
Uses predicted qualified convictions as an instrument for database size
Predicted qualified convictions is an estimated variable based on 1999 conviction rates and inmate rates
This instrument is highly correlated with database size, but uncorrelated with future crime rates through any other channel.
Example of the Instrument
Consider a state that adds convicted and incarcerated burglars to its database in 2002 and the state had 50 burglary convictions (per 100,000) and 200 incarcerated burglars (per 100,000) in 1999, then the predicted qualified convictions would be
0 in 2001
250 in 2002
300 in 2003
350 in 2004
IV Estimation Model
First Stage:
Interpretation of Results
The average database grew by 2,183 profiles (per 100,000 residents) over this time period, so this confidence interval implies that DNA databases reduced annual violent crime by 7%-45%. These results are statistically significant, but very imprecise.
Testing for Incapacitation Effects on Crime
What is the problem with an OLS regression of this form?
States with higher crime rates will have a higher proportion of the population in prison. This will bias the estimate in the positive direction.
Testing for Incapacitation Effects on Crime
Levitt (1996) prison overcrowding as a natural experiment to estimate the incapacitation effects.
From 1971 to 1993 the prison population quadrupled which led to prisoners being released if there was not sufficient space.
States continued to prosecute and sentence offenders, so there was no change in expected sanction/ deterrence effect, but occasionally groups of inmates who ordinarily would be incarcerated were released.
Testing for Incapacitation Effects on Crime
Levitt (1996) used prisoner releases as an instrument for the prison population.
He reported results for both OLS and IV estimation.
The elasticity of violent crime rate with respect to prison population was -0.1 for OLS and -0.4 for IV
These estimates suggest that a 10% increase in prison population lowers violent crime by 4%