Jacqueline Cohen

Characterizing Criminal Careers

Most knowledge about crime and criminals derives from cross-sectional analyses that link crime rates in a community with a communitys attributes. The criminal-career approach focuses on individual offenders and considers their crime-committing patterns as a longitudinal stochastic process. This approach, which invokes parameters characterizing participation rate, initiation rate, termination rate and the associated career length, and individual offending frequency, offers some important new insights. For example, annual offending frequency appears to be reasonably constant with age for those offenders who stay criminally active, termination rates are relatively low for active offenders in their 30s, and offending frequencies seem to be relatively insensitive to demographic attributes for active offenders. All these observations are opposite to those that would be derived from cross-sectional analysis.

Diffusion in Homicide: Exploring a General Method for Detecting Spatial Diffusion Processes

This article proposes a new method for examining dynamic changes in thespatial distribution of a phenomenon. Recently introduced exploratoryspatial data analysis (ESDA) techniques provide social scientists with anew set of tools for distinguishing between random and nonrandom spatialpatterns of events (Anselin, 1998). Existing ESDA measures, however, arestatic and do not permit comparisons of distributions of events in the samespace but across different time periods. One ESDA method—the Moranscatterplot—has special heuristic value because it visually displayslocal spatial relationships between each spatial unit and its neighbors. Weextend this static cross-sectional view of the spatial distribution ofevents to consider dynamic features of changes over time in spatialdependencies. The method distinguishes between contagious diffusion betweenadjoining units and hierarchical diffusion that spreads broadly throughcommonly shared influences. We apply the method to homicide data, lookingfor evidence of spatial diffusion of youth-gang homicides acrossneighborhoods in a city. Contagious diffusion between neighboring censustracts is evident only during the year of peak growth in total homicides,when high local rates of youth-gang homicides are followed by significantincreases in neighboring youth- nongang rates. This pattern is consistentwith a spread of homicides from gang youth to nongang youth. Otherwise, theincreases in both youth-gang and youth- nongang homicides generally occursimultaneously in nonneighboring tracts.

Deterrent Effects of the Police on Crime: A Replication and Theoretical Extension

This study replicates and then extends Wilson and Bolands (1978) theory of the deterrent effect of policing on crime rates in American cities by linking it to recent thinking on control of urban disorder and incivilities (Sherman, 1986; Skogan and Maxfield, 1981). The theory posits that police departments with a legalistic style tend to generate policies of proactive patrol (e.g., high traffic citation rate and frequent stops of suspicious or disorderly persons), which in turn may decrease crime rates either (1) indirectly, by increasing the probability of arrest, or (2) directly, by decreasing the crime rate through a deterrent effect regarding perceived threat of social control. We test both these propositions in an examination of robbery rates in 171 American cities in 1980. Overall, the major results suggest that proactive policing has direct inverse effects on aggregate robbery rates, independent of known determinants of crime (e.g., poverty, inequality, region, and family disruption). Moreover, when we demographically disaggregate the robbery rate the direct inverse effect of aggressive policing on robbery is largest for adult offenders and black offenders. We examine the reasons for these findings and discuss their theoretical and policy implications.

Specialization and Seriousness During Adult Criminal Careers

Crime-type switching between arrests is examined for tendencies by adult offenders to specialize in crime types or to escalate in seriousness as offending continues. The adult offenders examined display higher levels of specialization than have been previously reported for juveniles; among adult offenders, those who remain criminally active until older ages are also more specialized. Also, there is some evidence of trends toward a worsening of offending: for selected crime types, adult offending becomes more specialized and escalates in seriousness for white offenders. However, similar trends are not observed for black adult offenders.

Demographically disaggregated projections of prison populations

Abstract The need for improved long-run projections of prison populations has increased in recent years because of record-high numbers of inmates and severe overcrowding in state and federal prisons, and because of the growing importance of changing demographic factors in influencing corrections populations. A model is developed for projecting: general population demographics; demographic- and offense-specific arrest rates, imprisonment probabilities, and times served; and then the size and composition of prison populations. Model parameters are estimated for Pennsylvania and are shown to be sensitive to demographic factors, particularly age and race. Projections of future arrests, prison commitments, and prison population are developed for Pennsylvania using projections of demographic changes in the states population. Arrests are expected to peak in 1980, prison commitments are expected to peak in 1985, and prison populations are expected to peak in 1990, with the subsequent declines reflecting the maturation of the postwar baby boom children out of the highly crime-prone ages and, somewhat later, out of the highly prison-prone ages.

Predicting criminal recidivism: A comparison of neural network models with statistical methods

This article applies neural network and conventional statistical models to predicting criminal recidivism. While having promising properties for predicting recidivism, the network models do not exhibit any advantage over the other methods in an application on a well-known data set. Analysis suggests that currently available prediction variables have limited information content for discriminating recidivists, regardless of the models or methods used.

The Role of Drug Markets and Gangs in Local Homicide Rates

The presence of ongoing enterprises, sustained by a purpose and life beyond those of the current participants, distinguishes drug- and gang-related homicides. Without these enterprises that are sustained over time within distinctive spatial areas, violence would be reduced to individual acts precipitated by unique features of interactions between distinct individuals. Consistent with expectation, drug and gang homicides display substantial concentrations of violent incidents spatially and temporally within different neighborhoods in a city. The authors also find evidence for localized temporal dependencies in which a recent homicide alters the probability of another homicide in the same local area. Both drug and gang homicides display evidence of self-limiting suppression effects that inhibit extended periods of sustained high rates of these types of violence in the same area. Finally, there is evidence of cross-type diffusion from drug and gang homicides to subsequently higher rates of other gun homicides involving young offenders or victims.

The Influence of Capacity on Prison Population: A Critical Review of Some Recent Evidence

A recent study by Abt Associates (Abt/Carlson) purported to show that increments to prison capacity would lead to growth in prison population to fill that added capacity two years later. That finding has rapidly attained broad circulation and widespread acceptance. The original conclusion was based on a coefficient of 1.02 in a simple regression equation that represents change in prison population as a function of lagged changes in prison population and capacity. Reanalysis of the data shows that the original estimates resulted from a computation error; when that error is corrected the coefficient estimate is reduced to .264. Furthermore, two data points were particularly influential in the regression analysis, and omitting them results in a coefficient of .095 which is not statistic ally significant. Thus, the coefficient on which the original conclusion was based is eliminated in importance.

Hierarchical Bayesian Analysis of Arrest Rates

Abstract A Bayesian hierarchical model provides the basis for calibrating the crimes avoided by incarceration of individuals convicted of drug offenses compared to those convicted of nondrug offenses. Two methods for constructing reference priors for hierarchical models both lead to the same prior in the final model. We use Markov chain Monte Carlo methods to fit the model to data from a random sample of past arrest records of all felons convicted of drug trafficking, drug possession, robbery, or burglary in Los Angeles County in 1986 and 1990. The value of this formal analysis, as opposed to a simpler analysis that does not use the formal machinery of a Bayesian hierarchical model, is to provide interval estimates that account for the uncertainty due to the random effects.

THE LEGAL CONTROL OF DRUNKEN DRIVING: A COMMENT ON METHODOLOGICAL CONCERNS IN ASSESSING DETERRENT EFFECTIVENESS

Appearing in this volume are papers by Votey and Phillips, Ray, and Votey that raise basic questions about the effectiveness of efforts to reduce drunken driving through legal controls implemented in the criminal justice system. The primary issues of extent, source, and duration of any effect are raised by those authors. This paper discusses the concern for deterrence expressed by Phillips, et al., and particularly focuses on some methodological problems in demonstrating the deterrent effectiveness of criminal sanctions on drunken driving.