Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Summary THE CRIMINAL CAREER APPROACH Much research on crime has focused on aggregate crime rates, i.e., crimes per cap- ita in the general population. In contrast, the criminal career approach seeks to an- alyze the activity the careers-of the in- dividuals who commit criminal offenses. In this approach, aggregate rates are par- titioned into two primary components: the percentage of the population that commits crimes and the nature ant! extent of activity of those people who are ac- tively engaging in crime (i.e., active of- fenders). This partitioning is important because the two components can be in- fluenced by very different factors ant! call for quite different policy responses: the first participation is associated with ef- forts to prevent indivicluals from ever be- coming involved in crime; the second] frequency, seriousness, ant! career length is central to the decisions of the criminal justice system. Together, these four key dimensions characterize crimi- nal careers: 1 1. Participation the distinction be- tween those who engage in crime and those who do not; 2. Frequency the rate of criminal ac- tivity of those who are active; 3. Seriousness of offenses committed; 4. Career length the length of time an offender is active. Frequency, seriousness, and length of career may vary substantially across of- fenders. At one extreme are offenders whose careers consist of only one offense. At the other extreme are "career crimi- nals" also variously characterized as dangerous, habitual, or chronic offend- ers who commit serious offenses with high frequency over exten(led periods of time. Both of these extreme career types are of policy interest. Offenders with short careers who commit very few of- fenses are of interest for identifying exter- nal factors that foster early termination of careers and that can be affecter! by public policy. Such knowle(lge will be espe- cially useful in the design of crime con
2 trot policies, including prevention achieved through the imposition of sanc- tions for deterrent or rehabilitative pur- poses. Career criminals are of interest for formulating policies that can identify these offenders early in their careers and selectively target criminal justice re- sources at preventing their crimes, pri- marily through incarceration. Partitioning criminal activity enables research to focus more directly on the development and evolution of individu- als' offending behaviors over time and their relationship to aggregate crime rates. For example, a drop in the aggre- gate index crime rate may reflect a de- cline in participation, a decrease in fre- quency, a shift to less serious crimes, or a shortening of career lengths. If a causal factor strongly affected only one of these career dimensions, variation in the others could mask that relationship when crime is measured in terms of the aggregate rate. Since different factors are likely to affect the different dimensions of criminal careers, it is important to isolate those dimensions to assess each factor's respec- tive influence. Measurement of the dimensions of criminal careers is inherently difficult. It would be easy, of course, if one could find a large representative sample of the pop- ulation, each of whom would cooperate by maintaining and regularly submitting a daily log of any criminal activity. Since that option does not exist, criminology has developed two primary sources: ret- rospective self-reports from indivicluals about whether they have committed crimes and, if so, when, what kinds, and how often; and official arrest histories, which are viewed as a sample of crimes committed. Each of these approaches has its own flaws-nonresponse, distortion, and recall error in self-reports and varia- tions in police arrest decisions or offender skill that lead to biases in the sampling process that generates official records. But the findings derived from one ap CRIMINAL CAREERS AND CAREER CRIMINALS preach can ant] should be tested against the findings clerived from the other to develop more reliable measures of the underlying crime process. FINDINGS ON CRIMINAL CAREERS Some of the factors most strongly asso- ciated with aggregate crime and arrest rates are the demographic factors of age, race, and sex. By separately analyzing the links of these factors to participation and frequency, one can begin to see the im- portance of the criminal career approach. The data show that these demographic variables are strongly associated with par- ticipation and only weakly associated with individual frequency. That is, teen- agers, blacks, and mates are most heavily represented in aggregate measures of of- fending because larger percentages of those populations than of other popula- tions become involved in offending. But among all offenders, frequency of offend- ing is much less related to the demo- graphic variables. Because criminal jus- tice system clients have already passed through the participation "filter," sanc- tion decisions that invoke those demo- graphic variables are likely to be seri- ously in error from a crime control perspective. Thus, separating the partici- pation rate from the course of active of fenders' criminal careers enables one to isolate influences on the initiation of of- fending from influences on the frequency or duration of a criminal career. In prac- tical terms, prospective indicators of a high participation risk are useful in sug- gesting prevention strategies to reduce the fraction of offenders in the popula- tion. Participation in Offending Measures of Participation Participation can be measured either cumulatively (the fraction of a given group ever committing at least one crime
SUMMARY before attaining some age), or currently (the fraction committing at least one crime during a particular observation pe- riod). It is possible to tabulate participa- tion rates baser! on either official records of arrests or convictions or self-reports of offending. Scholars concemed with crime and delinquency have tended to study urban areas, where crime is most heavily con- centrated, and mates, who dominate ag- gregate arrest statistics. Therefore, knowI- ecige is most complete concerning participation by urban mates, and com- parative participation data are most readily available for the racial groups in urban areas. Participation in criminal be- havior, as measured by arrests, is higher than many people expect: 25~5 percent of urban males have been arrested by age 18 for a nontragic offense; 12-18 percent for an FBI index offense (murder, rape, robbery, aggravates! assault, burglary, lar- ceny, and auto theft); and ~8 percent for a crime involving personal injury. About half of those ever arrested during their lifetimes are first arrester! before they reach 18. The earliest criminal careers can begin before age 10; the rate of initiation rises to a maximum in the late teens and then drops off rapidly. At the peak age in the teenage years, the maximum probability of a first arrest for an index offense is about 6 percent for blacks and 2 percent for whites. Based on a number of studies, the black/white ratio of males ever ar- rested by age 18 is about 1.8 to 1 for nontraffic offenses. Other data suggest that the ratio is about 3 to 1 for FBI index offenses and at least 4 to 1 for offenses involving injury. The greatest demo- graphic differential is that attributable to sex: even for all nontragic offenses, male participation rates based on official rec- ords are 3 to 5 times those for females; the ratio is higher for more serious crimes. 3 Risk Factors in Criminal Participation Research suggests that adverse family influences and early antisocial behaviors are major risk factors associated with par- ticipation in delinquency and crime. The family influences most strongly associ- ated with subsequent participation in- clude inadequate parenting, as mani- fested by inconsistent or sporadically violent discipline, poor parent-child com- munication, and poor supervision; paren- tal delinquency or criminality; parental discord and family breakup; large family size, especially in poor families; and some indicators of Tow socioeconomic sta- tus Primarily for serious crimes). High rates of delinquency participation have been found among children exhibit- ing the following antisocial behaviors at an early age: covert antisocial behavior (e.g., lying, stealing); overt antisocial be- havior (e.g., aggression, fighting); poor school performance; abuse of hard drugs or alcohol; and association with delin- quent peers. Of course, all ofthese behav- iors occur frequently in many children who never begin criminal careers, but their chronic occurrence is correlated with a higher risk of participation in crim- inal activity. Although these risk factors have been suspected "causes of crime" for some time, recognition of their links to partici- pation could provide some basis for cle- signing strategies to prevent the initiation of criminal careers. However, several considerations hamper efforts to do so: · Difficulty in prospectively identify- ing offenders. Most youth who present any one of Me precursors do not initiate . . ~ serious criminal . careers. · Uncertainty about effective interven- tions. Attempts to remecliate precursor behaviors associated with participation may not influence participation itself;
4 most evaluations of preventive programs carried out thus far have design problems that preclude definitive conclusions about effectiveness. · Uncertainty about the optimal age for intervention. While the behavior of young children (e.g., ages 4~) is more responsive to interventions, it is more difficult to correctly identify high-risk children at those young ages. · Other concerns such as program cost, unintended] undesirable consequences, and legal and ethical limitations on per- missible interventions with youths who have not been charged with any offense. While a few interventions have demon- strated effectiveness in welI-designed experiments, they have involves! small samples. Larger, more heterogeneous samples in multiple settings shouIc! be used to further explore Head Start pre- schoo! enrichment programs and pro- grams to train parents and teachers in communication with and supervision of children and in behavioral methods for modifying chiTctren's antisocial behav iors. Individual Frequencies for Active Offenders Overall Rates Individual offending frequency (A) is measured by the number of crimes com- mitted per year by an active offender. Despite differences in observation tech- niques (self-reports or arrest recorcis), in the samples used (inmates, arresters, or general population samples), and in the jurisdictions examined (with their differ- ent population characteristics and crimi- nal justice system practices), there is con- siderable convergence in estimates of A for any given offense type: averages of 2 to 4 violent crimes per year for active violent offenders and 5 to 10 property CRIMINaL CAREERS AND CAREER CRIMINALS crimes per year for active property of- fenders. The average A of inmates is higher than that of offenders not incarcerated, which is to be expected since the greater activity of high-rate offenders increases their ex- posure to arrest and because selectivity in criminal justice decisions increases the risk of incarceration of those offenders if arrested. Estimates derived from self- reports by inmates in California and Michigan indicate that, before being in- carcerated, those who were active in rob- bery committed an average of 15 to 20 robberies per year and those active in burglary committed an average of 45 to 50 burglaries per year. Rates estimated for Texas inmates more closely resemble those of offenders in the community. Most significantly, A varies consider- ably across offenders so that the distribu- tion of A is highly skewe(l: the median offender commits only a handful of crimes per year, while a small percentage of offenders commit more than 100 crimes per year. This finding is obviously espe- cially important in cleveloping policies to reduce crime by concentrating on high- rate offenders, or "career criminals." Demographic Differences in A Sex, age, ant! race data are routinely collected on offenders, primarily for oper- ational identification purposes, and esti- mates of A are often contrasted across demographic subgroups defined in those terms. In contrast to the large demo- graphic differences observed in participa- tion rates, average values of A do not vary substantially with the demographic at- tributes of sex, age, or race. Most notably, black/white ratios for A range from about 1-1.3 to 1 for most crimes and are less than 2 to 1 even for robbery and other violent offenses that show very substantial race differences in aggregate population arrest rates. In other
SUMMARY words, the large race difference observed in aggregate crime or arrest rates is clue mostly to black/white differences in par . ~ · . - .1 .1 ~ in A. Preliminary data also suggest that male/ female frequency ratios are generally less than 2 to 1, even for the traclitionally mate offenses of assault and robbery. And, fi- nally, A is relatively insensitive to age. tlclpatlon rather than clluerences ~ a. . . . Other Factors Related to A Differences in A are associates! with a number of nonclemographic factors, in- cluding age of initiation of criminal ca- reers, drug use, unemployment, and prior criminal record. Although their theoreti- cal significance is sometimes ambiguous, these factors offer policy potential, espe- cially as a basis for iclentif~in~ hi~h-rate ~. ~ ~ ~ ~ is onenaers. Fencers who begin criminal careers at younger ages generally have higher values of A than offenders who begin at older ages. High frequencies are found among active offenders who are currently using drugs, especially multiple drugs, and among those who used drugs as juveniles. The values of A for drug users are at least twice those for other offenders and can be 6 times higher dur- ing periods of heavy drug use. The length of time spent unemployed is a significant factor associated with A: offenders who are unemployed] for substantial periods of time commit crimes at higher rates. Last, high frequencies in We past are a good indicator of high frequencies in the future for offenders who remain active. T~ · ~ - renas in Seriousness Most active offenders engage in a con- siclerable variety of crime types, with a somewhat greater tendency for offenders to repeat the same crime or to repeat within the group of property crimes or the group of violent crimes. For juveniles, 5 offense seriousness generally escalates over successive arrests, but it is difficult to determine how much of that escalation is due to differences among offenders (with more persistent offenders also commit- ting more serious crimes) or to a tendency of offenders generally to escalate the se- riousness of their crimes as their careers progress. For adult offenders who are arrested more than once, the evidence suggests no clear trends in seriousness. Career Length Since participation in criminal activity is more widespread among teenage males than among adult males and A is relatively stable over age for those offenders who remain active, many careers must be very short, en(ling abler only brief ventures into crime as teenagers. This conclusion is consistent with the decline with age observed in aggregate arrest measures. Research on duration of careers in fact suggests that adult careers average only about ~ years for offenders committing index offenses. However, this average hides major differences across offenders. Residual career length (the expected time still remaining in careers) is 5 years for 18-year-oIc3 inclex offenders, but it is 10 years for index offenders who are still active in their 30s, and it does not begin to clecline for active offenders until they are in their 40s. This finding that residual career length increases through the 20s to reach a maximum in the 30s-contradicts the wiclely hell! view that the Tow arrest rates in the 30s reflect high rates of career termination at those ages. These low ar- rest rates result from the high termination rates at the earlier ages, but those few persistent offenders who continue to be active into their 30s clisplay the lowest termination rates and the longest resiclual , v careers.
6 USE OF CRIMINAL CAREER INFORMATION IN CRIMINAL JUSTICE DECISION MAKING The findings from research on criminal careers can play an important role in the decisions of the criminal justice system. Those decisions take place at five points: arrest, pretrial release, prosecution, sen- tencing, and parole. At each point, deci- sion makers are motivated by different objectives such as maintaining public orcler, ensuring appearance at trial, attain- ing convictions, imposing deserved pun- ishment, and maintaining order in pris- ons but each also shares in the objective of crime control. To further this objective, decision makers may attempt to modify ers. criminal careers, either directly through behavioral or other individual therapies, or indirectly, by modifying substance abuse, employment prospects, or other characteristics associated with the fre- quency or residual duration of serious offending. Career Modification An earlier Panel on Research on Reha- bilitative Techniques reporter! in 1980 that no particular technique hac! been found to be broacITy effective in directly modifying criminal careers. For juvenile offenders, some community-based pro- grams result in recidivism rates that are no higher than those for juveniles who were incarcerated. A few recently studied programs with behavior modification ori- entations are reported to have favorably modified the careers of small samples of clelinquents during short (e.g., 1-year) fol- low-up periods; these warrant replica- tions with longer follow-up periods and larger samples. Among indirect ap- proaches to career modification, intensive drug abuse surveillance and treatment appear to reduce the frequency of serious offending by criminals ad(licte(1 to hard CRIMINAL CAREERS AND CAREER CRIMINALS drugs. Evaluations of interventions to up- grade employment skills or to assist the reentry of ax-inmates into the community have not documented success in modify- ing offenders' careers in the United States, although such programs have met with some success in the United King- dom. Given the Beard of interventions that have demonstrated general effectiveness in modifying careers, decision strategies focused on career criminals have recently become especially prominent, which has led to research on how criminal career information can be used to achieve greater crime control, especially through incapacitation of the most serious offend Incapacitation Under 1970 incarceration policies, in- capacitation was estimated to have re- duced the number of FBI index crimes by 10 to 20 percent. For robberies and bur- glaries, incapacitation is estimated to have reduced their number by 25~5 per- cent in 1973; in 1982, after the national inmate population had almost doubled, the incapacitative effect for these offenses is estimated to have increased to about 35~5 percent. For general increases in incarceration to reduce index crimes by an additional 10 to 20 percent from 1982 levels, inmate populations again would have to have more than doubled. Since the increments to crime control from incapacitation are modest, even with very large general increases in in- mate populations, and since considerable pressure on prison and jail resources ex- ists, interest has been stimulated in de- veloping policies that selectively target incapacitation on the most active offend- ers. Such policies have the potential of achieving the same, or even improved, crime reduction, with much smaller in- creases in inmate populations. Two forms
SUMMARY of selective incapacitation policies been considered: offencler-based have poli- cies, which extend the time served for those predicted] to be high-rate offenders among all those convicted of the same charge; and charge-based policies, which impose mandatory minimum terms on all offenders convicted of selected offense types in which high-rate offenders tend to engage. Of the selective incapacitation policies that have been proposed, the estimates! crime reduction effects through incapaci- tation are modest, ranging at most up to 10 percent. Comparer! with the 100 per- cent increase in inmate populations re- quirecl to achieve this reduction through general increases in incarceration, how- ever, selective incapacitation policies in- volve only 10 to 20 percent increases in total inmate populations. The achievable incapacitative effect will be Tower, how- ever, if the offenses of incarcerated of- fenders are continued by others or re- placec! by new recruits. The degree to which replacement occurs can be ex- pectecl to vary by crime type (e.g., pre- sumably higher for drug sales than for acts of personal violence), but there are no estimates of rates of replacement. Furthermore, although selective inca- pacitation policies can offer an attractive making. tradeoff between crime reduction and in- mate population increases, they raise a number of ethical ant! operational issues: · the inevitable errors associated with any classification rule that distinguishes between high-risk ant! Tow-risk offenders; · the proper balance in criminal justice decisions between cleserved punishment and public protection Concern for future crimes); · the degree to which the addec! crime control from selective incapacitation pol- icies warrants sentence disparity in of- fender-based policies or mandatory sen- tences in charge-based policies; and 7 · the possibility that some selection rules might disproportionately penalize offenders who are from minority groups. PREDICTION-BASED CLASSIFICATION RULES There have been a number of recent developments in systems for classifying offenders on the basis of assessments of their criminal careers. These systems in- volve two components: a statistically cle- rived prediction scale that relates of- fencler characteristics to the course of the criminal career and a rule for classifying offenders as high risk or Tow risk on the basis of their prediction scale scores. In assessing such classification rules, the pane! consiclered several factors: · the extent to which existing decision practices aIreacly focus on offenders whose criminal careers tend to be the most serious; · ethical limitations on the proper in- fluence of predictions in decision making and on the choice of predictor variables; · methodological issues in construct- ing and validating a prediction scale and classification rule; and · the record-keeping implications of a criminal career orientation to decision Existing Decision Practices Because criminal justice decision mak- ers do not routinely articulate their deci- sion processes, it is difficult to know the extent to which predictions about offend- ers are incorporated into their decisions. However, using statistical techniques, studies have tested whether decision out- comes are consistent with variables asso- ciated with the criminal career: the in- stant offense; the adult criminal history, measured by the record of previous ar- rests or convictions; the juvenile record, including police contacts and adjudica
8 lions; indicators of multiple illicit drug use; and employment history and current status. The panel found that the ~. . . . . ~ instant of- tense has the clom~nant ~ntiuence on de- cisions at all stages of the criminal justice process, from arrest through parole. The second most influential variable is the record of previous arrests or convictions. Use of the juvenile record is largely lim- ited to sentencing and parole because laws in most jurisdictions separate juve- nile records from adult records. Drug use is associated at the sentencing stage with higher probabilities for jail or prison sen- tences. Employment history and status is associated with decisions only at the stage of pretrial release. On the basis of criminal career re- search, current decision practices could be improved, in terms of crime control through incapacitation, if more weight were given to the juvenile record and to serious drug use. Ethical Limitations Opinions differ on the extent to which it is ethically proper for predictive consid- erations to influence the choice of crimi- nal justice sanctions. Strict adherents to a just-deserts philosophy exclude predic- tions altogether, arguing that the blame- worthiness of the instant offense is the only permissible basis for sanctions. A1- temative views are that while blamewor- thiness for the offense determines a pre- sumptive range of acceptable sanctions, the appropriate sanction in a particular case may vary within that range and could be based on prediction-based classifica- tions. The weight one considers acceptable for prediction-based classification rules depends on the gravity of the harm one is trying to prevent by means of classifica- tion and the accuracy ofthe classifications in assessing risk. In assessing whether a CRIMINAL CAREERS AND CAREER CRIMINALS potential predictor is ethically accept- able, its relationship to the blameworthi- ness ofthe offender and the empirical and logical relationship of the predictor to the behavior being predicted are commonly invoked criteria. In light of these considerations, vari- ables such as prior adult convictions are more widely accepted as predictors than behaviors or characteristics such as em- ployment status, over which the offender is presumed to have less control. Charac- teristics such as race, ethnicity, and reli- gion are especially unacceptable as can- didate predictors because they have no relationship to blameworthiness, they have no logical relationship to offending patterns, and their use affronts basic so- cial values. Methodological Issues Since offenders' behavior cannot be predicted perfectly, two types of classifi- cation error occur: false-positive classifi- cation of offenders as high risk when they are truly low risk, and false-negative clas- sification of offenders as low risk when they are truly high risk. For any predic- tion scale, the number of false-positive errors can be reduced by raising the cut point to classify fewer offenders as high risk at the cost of more false-negative errors. Any final choice of a prediction- based classification rule should estimate and weigh both types of errors, in devel- oping the final classification rule. Because the criminal justice process consists of a series of decisions that result in the release of some offenders while others are confined, classification rules are frequently distorted by selection bias when a rule developed for use at one stage- parole, for example-is applied at another, such as sentencing or assign- ment to a career criminal unit for prose- cution. Also, because the mix of offender characteristics at any given stage is likely
SUMMARY to stiffer across jurisdictions, classification rules constructed and validated in one jurisdiction should be revaTidatec] before use in another jurisdiction, even if they are to be applied at the same stage of the criminal justice system. Finally, even af- ter a rule is implemented, its accuracy may change over time because of changes in offenders' underlying behavior pat- terns, changes in the composition of the offender population, or changes in crimi- nal justice selection processes. Therefore, periodic revaTidations should be per- formect to be certain that the classification rule is still effective. Specific Classification Rules Several prediction-based rules for cIas- sifying offenders have attracted sufficient attention to warrant assessment by the panel. Two have been applied in parole decision making and invoke similar vari- ables: the Salient Factor Score of the United States Parole Commission and the Iowa Risk Assessment Instrument. Using a broad criterion variable with a large base rate (rearrest during follow-up peri- ods of 3 years or more), the rate of false positives Offenders cIassifiec3 as high risks who were not arrested cluring the follow-up period) is less than 30 percent in construction samples ant! less than 40 percent in validation samples. The accu- racy of these scales has generally been maintainer] in applications to new sam- ples of offenders ant! after various per- sonal attributes of questionable appropri- ateness have been eliminated from them. A third! scale has been proposed by INSLAW for use in selecting cases for high-priority federal prosecution. This scale is noteworthy for having focused on the continuous measure of time until re- cidivism rather than merely the dichoto- mous variable of the occurrence of recicl- ivism. The accuracy of this scale in its construction sample is comparable with 9 the parole scales, but it has not yet been tested in a validation sample. A fourth scale, developed from ciata collected in the Ranc! inmate survey, is unique in its use of Gentling frequency (A) as the cri- terion variable. However, when the scale was used to classify the offenders as high A, the false-positive rate was 60 percent for a sample of California inmates and 55 percent for the entire three-state sample. Future scale development shouts] be aimed at improving the classification of offenders in terms of their frequency of offending and their residual career length. While these career dimensions are not directly observable, their values can be inferred using statistical tech- niques that analyze time to recidivism. Record-Keeping Implications For criminal career knowledge to be helpful in decisions (whether formally or informally), the relevant information has to be accurate and available to decision makers on a timely basis. Much of this information is stored in the record sys- tems of adult ant! juvenile justice a~en- cies as well as social welfare agencies. While there is no universal agreement about what information should be permit- ted to influence criminal justice decisions (e.g., some people believe that only aclult convictions shouIc] be user! and neither arrests nor juvenile court actions), it is clear that full adult ant! juvenile arrest records do provide valuable information about criminal careers. The improvement of arrest and dispo- sition data shouIc! be a high-priority ob- iective of criminal justice systems. Arrest records should be macle more complete and accurate, and pertinent attributes of an offense should also be reported on the arrest history. Arrest records to be used in decisions shouts] be augmenter! with cor- responding dispositions. When an arrest is fount! to have been clearly unwar in, . ~ '
10 ranted, as in a case of mistaken identity, there should be a prominent notation and an opportunity for expungement. Adult justice system agencies should gain access to the juvenile record at the time of a person's first serious criminal involvement as an adult. In defining cri- teria for access, different jurisdictions might choose different thresholds of crime seriousness and different stages of criminal justice processing (e.g., arrest, indictment, or conviction). Requiring a high level of seriousness would reflect a clesire to avoid opening a juvenile record in connection with only a minor offense. Precluding access until after conviction would reflect a belief in the principle that a juvenile record should not influence decisions in the adult system that are made before guilt is established with le- gal certainty. Altematively, permitting prosecutors and judges to see a juvenile record when making charging, pretrial release, and plea bargaining decisions would allow them to become more fully informer! of the public safety risks associ- atec] with their decisions. Different juris- dictions will weigh these concerns dif- ferently and so may choose different thresholds and stages. One choice that seems reasonable would be the first adult arrest for a felony. A juvenile record would be appended to an aclult record only if the adult arrest leads to a convic- tion. The juvenile record, with appropri- ate safeguards against dissemination, wouIcl be retained in the juvenile repos- itory at least as Tong as it retains value for decision making in the adult criminal justice system. AN AGENDA FOR FUTURE RESEARCH The area of criminal career research is sufficiently promising from both knowI- edge and policy perspectives to warrant a significant effort with a long-term com CRIMINAL CAREERS AND CAREER CRIMINALS mitment for support. Key features of the effort should be improved measurement of the dimensions of criminal careers; measurement of their distributions, which are still known only imprecisely; measurement of the variation of the di- mensions over the course of a criminal career; and better identification of the factors that influence criminal careers. The most important criminal career dimension is individual frequency, par- ticularly of serious offenses. The two ap- proaches now used to measure frequen- cy self-reports and official arrest rec- ords-provide reasonable measures, but estimates would be improved apprecia- bly if the two methods were applied to the same samples. Particular attention should be directed at learning how the probability of arrest following a crime varies across subpopulations or crime types. More knowledge about the error structures of the two approaches, and their links to the underlying crime proc- ess, would increase the usefulness of the large number of official-record data sets that are available as primary record sources for estimating criminal career di- mensions. Research should focus espe- cially on the path of A over time as of- fenders age, variation in A with age for active offenders, the factors associated with intermittent spurts of high-rate and low-rate offending, and differences in A by crime type. While some correlates of career dimen- sions are reasonably well established (e.g., employment), it is still not clear whether the correlations reflect causal re- lationships that couIcl be manipulated to facilitate earlier termination. For exam- ple, it would be clesirable to follow up earlier experiments that provided postre- lease subsidy payments to parolees to recluce their economic needs with a pro- gram designed to avoid the work disin- centive effects observed in that experi- ment (perhaps by requiring the recipients
SUMMARY to perform community service). Of partic- ular policy significance is exploration of the factors most influential in terminating a criminal career. For example, little is known about the effect of time in prison on residual career length, especially about whether such time simply post- pones career termination. Questions of effects of particular treatments on career dimensions are generally best aciciressed through an experimental approach. Research activities that relate closely to incapacitation policy include the devel- opment of improved methods for cIassify- ing individuals in terms of their criminal career dimensions. Strong candidate cIas- sification variables include those that characterize prior juvenile and adult criminal activity the age of initiation, the frequency and total number of arrests and convictions, and the mix of crime types. It is important to assess the clegree to which an individuals record of juvenile and adult arrests provides significantly more predictive information than does the adult conviction record alone. Since active drug use is strongly associated with high offending frequencies, both the uses of this knowledge in decision making and the role of drug treatment in crime control should be further explored. It is also important to refine and test theoretical explanations of initiation of clelinquency or criminal activity and of continuation into serious adult careers. Better information is needed on particu- lar behaviors, life events, and early career patterns that can be linked to subsequent 11 career paths. One aspect of criminal ca- reers, especially among juveniles, is the role of others in influencing initiation and continuation of offending. It would be particularly importer to identify the cle- gree to which some iclentifiable individual offenders serve as recruiters to offending and to what extent behavior can be ex- plained as compliant responsiveness to recruitment efforts. Also with regard to initiation, it is important to explore more carefully the events that may influence the initiation of careers but that have not been sufficiently studied in a longitudinal context to establish causation. Such events include abuse at the hands of family mem- bers, parental alcohol or drug abuse, and divorce or other traumatic life events. Many of these issues are best ad- dressed through a longituclinal research clesign, especially in view of the inher- ently longitudinal and dynamic character of criminal careers. A major new Tongitu- dinal stucly that follows multiple cohorts of inclividuals, linking their criminal ca- reers to personal characteristics and to other life events, would be very valuable. Such a major project should be under- taken in a way that reflects multiple dis- ciplinary and theoretical perspectives rel- evant to unclerstanding criminal careers. The project should also try to incorporate opportunities for experimental interven- tions that can be tested on the population being studied longitudinally. Such a pro- gram shouIcl be a major national priority as the next phase of research on criminal careers.
