Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 319
JUVENILE CRIME APPENDIX B The Indeterminacy of Forecasts of Crime Rates and Juvenile Offenses Kenneth C. Land and Patricia L. McCall How much crime will there be in the United States in the next 5 or 10 years? Will crime rates go up or down or remain about the same? Since juvenile crime often is a leading edge of crime problems to come, how many juvenile offenses will there be? Will the number of juvenile serious violent offenders/homicide perpetrators increase? What will be the resulting demands on the juvenile and criminal justice systems? Over the past three decades, criminologists have made a number of attempts to address these and related questions. These usually have taken the form of efforts to explain past variations or to project future levels of crime by applying techniques of demographic and statistical analysis. These techniques typically consist of: the application of demographic age standardization methods to combine relatively accurate estimates of the age structure of the American population with age-specific arrest rates for various types of crimes and categories of the population to calculate expected numbers of criminal offenses or crime rates or the construction of an explanatory time-series regression or structural equation models to explain or predict variations in crime rates over time. Kenneth C. Land is John Franklin Crowell Professor of Sociology, Duke University. Patricia L. McCall is Associate Professor of Sociology, North Carolina State University.
OCR for page 320
JUVENILE CRIME Such analyses may be useful exercises with respect to explaining past experiences in the ups and downs of observed crime or the projection of recent trends in order to anticipate future problems and needs for levels of resources in the juvenile and criminal justice systems. Yet even a casual review of the various projections of crime rates or offenses that have been made over the years suggests that they contain large amounts of uncertainty. That is, the mere fact that a projection indicates that, say, juvenile homicide offenders may increase (or decrease) by some specific percentage over the next 5 or 10 years does not mean that the rates will, in fact, exhibit such an increase (or decrease). The purposes of this paper are twofold. First, we review a number of extant demographic projections of crime rates and offenses that have been made for the United States over the past few decades, with a special focus on projections of juvenile crime rates and offenses. We commence in the next section with a brief summary of demographic analyses of the crime wave in the 1960s based on the coming of age of the baby boomers. This is followed by a review of projections of downturns in crime rates in the 1980s based on the smaller “baby bust” birth cohorts. More recently, following the rise in delinquent and criminal offenses by adolescents and teenagers in the 1985-1993 period, criminologists have produced some scary projections, which we next describe, of increasing numbers of violent criminal offenses expected in the period 1995-2005, as the “echoboomers” enter their teenage years. It will be seen that one characteristic of most extant projections of juvenile and criminal offenses is that, until recently, they have produced only expected or average values of future levels of crime rates or offenses. But temporal variability of age-specific crime rates has been a key characteristic of offending patterns, especially for juveniles, in recent years. Yet most projections of criminal and juvenile offending rates and numbers of offenses disregard the uncertainty associated with such projections. To emphasize the significance of the uncertainty of projections of criminal and juvenile offenses, a second objective of the paper is to describe some exercises in the construction of plausible national projections of expected numbers of male juvenile homicide offenders—as well as upper and lower bounds for the expected numbers—for each year from 1998 to 2007. A final section contains a statement of the major conclusions from our review and analyses. THE BABY BOOMERS COME OF AGE IN THE 1960s AND 1970s One of the first attempts to examine the impact of a changing demographic age composition of the population on numbers of criminal offenses reported to the police was made during the 1960s—when the
OCR for page 321
JUVENILE CRIME United States was stunned by skyrocketing crime rates. On the heels of the relatively low levels of criminal and juvenile offending in the 1950s, scholars and politicians began searching for reasons behind the dramatic increase in crime rates during the 1960s. Criminologists were well aware of the fact that the Uniform Crime Reports (UCR) published annually by the Federal Bureau of Investigation (FBI), the primary measure of national crime levels and rates at that time, was at best a politically influenced undercount and a weak indicator of the extent of criminal activity in the United States (see, e.g., Biderman, 1966). This was an unsettling notion especially in light of the growing magnitude of crime. One of the outcomes of this crime wave was the development of the annual National Crime Victimization Survey (NCVS) beginning in 1973.1 This survey was introduced as a new tool to determine the extent of criminal activity in the United States by surveying individual households in the population regarding the victimization status of their members in the past year. Criminologists attempted to determine the impetus behind the crime surge of the 1960s. Philip Sagi and Charles Wellford, in their 1967 report to the President's Commission on Law Enforcement and the Administration of Justice, identified the centrality of shifts in the age composition of the population as an explanation. Using a variety of demographic techniques, they attempted to accurately estimate the extent to which the increasing crime rate was due to an increase in individuals' crime proneness versus the changing composition of the population with respect to age, race, and geographic location. In particular, Sagi and Wellford (1968) cited the contribution that the post-World War II baby boom generation was making to the crime wave in the 1960s.2 They argued that, during the early 1960s, individuals born in the early years of the baby boom hit peak criminal offending ages, i.e., their late teens and early 20s. Using techniques of demographic age standardization, Sagi and Wellford demonstrated that the population increase in these young ages between 1958 (the low point of national crime rates in the late 1950s) and 1964 (the most recent year's data available at the time of their study), in and of itself, accounted for 24 percent of the increase in 1 The NCVS was originally called the National Crime Survey. It was redesigned and renamed in 1992 (see Bureau of Justice Statistics, 1995). 2 Demographers have defined baby boomers in the United States as individuals born in the 18 high birth rate years from 1946 to 1964 (see, e.g., Crispell, 1993). Birth cohorts from these years are relatively large compared with those both earlier and later, and their movement through the age structure has been associated with various social movements and changes in social institutions. Sometimes “early boomers” born in 1946-1955 are further distinguished from “late boomers” born 1956-1964 (Gibson, 1993).
OCR for page 322
JUVENILE CRIME FBI Index (or UCR Part I) offenses. They further demonstrated that changes in population race, age, and place of residence in combination accounted for 46 percent of that increase (President's Commission on Law Enforcement and Administration of Justice, 1967:208; see also Sagi and Wellford, 1968). Wellford (1973) followed up this analysis by extending the time series to include annual index crime rates through 1969. He first computed age- and crime-specific arrest rates. Then Wellford computed the age-standardized total offense rate by adjusting for the underrepresentation of the total U.S. population in the UCR. He estimated that the percentage increases in person or violent (homicide, aggravated assault, and robbery) and property (burglary, larceny-theft, and motor vehicle theft) crimes were 148 and 92 percent (age-standardized crime rates) as opposed to 165 and 117 percent (as indicated by the crude crime rates reported in the UCR), respectively. Even though the increase in offending rates during the 1960s was not as large as the official crime rates would lead one to believe, the disconcerting news was that the rate of violent crimes rose more than that of property crimes among youth during this period. The other disturbing results of Wellford 's cohort analysis showed that, with one exception, each cohort born during the baby boom was exhibiting crime rates higher than the one before. The main point made in his research was that “minimally, age composition effects must be controlled in attempting to estimate crime increase” rather than relying solely on “rates reflecting only the changes in the size of the total population” (Wellford, 1973:63). Sagi and Wellford did not attempt to project the crime rates past the 1960s. But they noted that it is possible to forecast fairly accurately the size and age composition of the juvenile and adult populations for one or two decades into the future and, thereby, to project the extent of crime for that period. Beside noting the problems inherent in attempting to make these estimates, they also warned of the necessity to obtain arrest information for each sex and age group within each race and geographic location category in order to better account for the impact that the changing population composition has on crime trends and “to make much better judgments as to how much of any particular increase or decrease in crime rates was due to a change in the criminality of the persons involved” (President 's Commission on Law Enforcement and Administration of Justice, 1967:210). To a large extent, data readily available from the FBI today still possess the shortcomings identified by Sagi and Wellford over three decades ago.
OCR for page 323
JUVENILE CRIME DECREASING CRIME RATES FOR THE BABY BUSTERS IN THE 1980S During the years after Sagi and Wellford conducted their initial study, the United States saw an increasing crime trend through the 1970s and into the early 1980s, at which time the crime rates began falling. With an additional 10 to 15 years of crime data at their disposal, James A. Fox (1978), Lawrence E. Cohen and Kenneth C. Land (1987), and Darrell Steffensmeier and Miles Harer (1987) reexamined the impact of age composition on crime trends. The question posed by the latter two groups of investigators was whether the decline in crime rates in the 1980s was due to the baby boomers aging out of those crime prone ages—adolescence and young adulthood. Steffensmeier and Harer (1987) studied changes in crime rates among index crimes between 1980 and 1984. Echoing Sagi and Wellford's concern with the changes in crime trends reported in the UCR and the National Crime Survey, they noted that the official “crime figures are not age specific but are crude rates based on the U.S. population as a whole” (Steffensmeier and Harer, 1987:29). Using methods similar to Sagi and Wellford's, Steffensmeier and Harer applied the demographic technique of indirect standardization of crime rates on age-specific arrest rates adjusted for the proportions of the U.S. population not covered in the annual UCR series. By applying this age-adjustment method to data derived from the UCR and the National Crime Survey reports, they compared percentage changes between 1980 and 1984 in unadjusted crude rates (the traditional measure of change) with their adjusted percentage change that corrects for the changes in the age structure. They showed that the age composition accounted for approximately 30 to 70 percent of declines in property and robbery crime rates, since the baby boomers had aged past the property crime prone ages of adolescence and the early twenties. Violent crimes had not enjoyed such a large decline as the baby boomers had not quite reached the ages where the violent criminal offending tends to drop—that is, the late 20s and early 30s. Based on those findings, Steffensmeier and Harer (1987) used age-specific estimates of the U.S. population produced by the Bureau of the Census through the end of the 20th century to forecast reductions in the nation's crime rate from 1980 to 2000. The forecasts assumed that age-specific offending rates would remain constant into the future and thus were based solely on changes in age composition. Specifically, they noted that the proportion of young people (ages 15-24), those at high risk for property crime, was estimated to decline sharply into the early 1990s and the proportion of youth and young adults (ages 15-35), those at high risk for person crime, were expected to decline steadily into and even beyond
OCR for page 324
JUVENILE CRIME the year 2000. This was due to the arrival of the “baby busters” at these high crime-prone ages.3 The projections made by Steffensmeier and Harer (1987) showed a steadily declining violent crime rate until the year 2000 and a somewhat steeper declining property crime rate until the mid-1990s, when the rate would plateau and begin a slight increase in the late 1990s. More precisely, they forecasted that violent crime rates would fall about 13 percent compared with about 20 percent for property crime rates during the 1980 to 2000 period. The projections of crime rates in the 1980s and 1990s of Steffensmeier and Harer (1987) based on demographic standardization can be compared with those of two studies, Fox (1978) and Cohen and Land (1987), based on regression models of crime rate time series. Fox was the first researcher to publish forecasts of U.S. crime rates based on this type of analysis.4 Using national crime rate data for the years 1950 through 1974, he studied the impact the baby boomers had on the surge in crime rates during the 1960s and 1970s. This led to a conclusion similar to that of Steffensmeier and Harer (1987)—that crime rates would fall during the 1980s when the baby boomers matured out of the crime prone ages and were replaced by the baby busters. Fox constructed structural equation models that estimated not only the impact that race and age composition had on crime trends, but also that of socioeconomic characteristics of the population as well as police activities and expenditures. Based on his study, Fox concluded (1978:51): “The crime rate forecasts reveal a general reduction in upward trend during the 1980s and a trend increase during the 1990s. In fact, the violent crime rate . . . should decline in the 1980s before increasing once again in the 1990s.” These projections were based primarily on age-and race-specific population estimates and projections published by the Census Bureau for the last quarter of the century. Cohen and Land (1987) also conducted an analysis of crime trends in the United States through the mid-1980s based on a time-series regression analysis for a somewhat longer post-World War II period, 1946 through 1984, to determine the extent to which changes in the age structure influenced the crime trends. By relating their analysis to the question of the relationship between age and crime then debated by Hirschi and Gottfredson (1983) and Greenberg (1985), they attempted to answer whether the decline in crime rates beginning in the mid-1980s would continue to 3 Demographers generally refer to individuals in the United States born in the relatively low birth rate years from 1965 through 1976 that followed the baby boom years as baby busters (Crispell, 1993). Members of the baby buster birth cohorts also have been labeled in the popular press as “Generation X.” 4 See also Cohen et al. (1980) for a time-series regression analysis of U.S. property crime rates, 1947-1974, with projections to the mid-1980s.
OCR for page 325
JUVENILE CRIME decrease symmetrically (i.e., proportionally to the declining population in the high crime-prone ages) versus asymmetrically (whereby cohortspecific effects produce non decreasing crime propensities throughout the life courses of high crime-prone cohorts as argued by Greenberg , as well as suggested in the crime patterns displayed by the cohort analysis conducted by Wellford ). Cohen and Land (1987) focused specifically on homicide and motor vehicle theft rates and controlled for other social forces affecting crime rates: trends in business cycles as well as in criminal opportunity and the rate of imprisonment. They first identified the peak ages of offending for homicide and motor vehicle theft—15 to 29 and 15 to 24, respectively. By overlaying the trends in graphic form, Cohen and Land demonstrated that the homicide and motor vehicle trends mirrored the trends in age structure for these two youthful groups (see Figures B-1 and B-2, which reprint Figures 3 and 4 from their 1987 report). In their time-series regression analysis, they included the percentage of the population ages 15 to 29 as the age composition control in the homicide model and the percentage ages 15 to 24 in the motor vehicle theft model. In addition, they introduced measures for age-proneness shifts among the cohorts computed as FIGURE B-1 Annual estimates of vehicle theft rate and the percentage of the population ages 15 to 24, United States, 1946-1984, with projections of the latter to 2001. Reproduction of Figure 3 from Cohen and Land (1987). Reprinted with permission.
OCR for page 326
JUVENILE CRIME FIGURE B-2 Annual estimates of the murder rates and the percentage of the population ages 15 to 29, United States, 1946-1984, with projections of the latter to 2001. Reproduction of Figure 4 from Cohen and Land (1987). Reprinted with permission. the product of a dummy variable, equal to one for the years 1966-1984 and zero for 1947-1965, times the natural logarithm of the age structure index. This variable was “incorporated in order to test for the time series significance of changes in the levels of the age-specific crime rates in the later as compared to the earlier part of the sample period” (Cohen and Land, 1987:178). The results of their time-series analyses showed that the age structure variables in both homicide and motor vehicle theft models were significant, but that the age proneness shift measure was not (contrary to the finding of Wellford 's 1973 study). They argued that whatever cohort changes have occurred are not of sufficient magnitude net of those cohort differences transmitted through the other causal measures included in their models: unemployment rate, unemployment fluctuations, criminal opportunity, and imprisonment rate variables. They concluded from their analyses that the age structure-crime relationship, at least as evident in the homicide and motor vehicle theft rates series up to the mid-1980s, appeared to be symmetric. Cohen and Land (1987) compared their findings to those of Steffensmeier and Harer (1987), noting that the latter's use of age-
OCR for page 327
JUVENILE CRIME standardized arrest rates focused on the offender population and attributed all variation either to changes in the entire age composition of the population or to changes in offending rates. Cohen and Land argued that by using a single age composition index for each crime model, they concentrated exclusively on the relative frequency of adolescents and young adults in the population, which takes into account the pool of potential victims as well as offenders. Whereas Steffensmeier and Harer's techniques accounted for about two-thirds of the decline in motor vehicle theft and none of the decline in homicide, Cohen and Land's analysis accounted for about 26 percent of the year-to-year change in the vehicle theft series and about 58 percent of the change in the homicide series. Based on Census Bureau projections of the U.S. population age composition into the 21st century, Cohen and Land cautiously forecasted generally declining homicide and vehicle theft rates for the post-1985 period into the mid-1990s to be followed by increases into the next decade. Noting that increases already had occurred in the homicide and motor vehicle theft rates for 1985 and 1986 after they concluded their analysis of crime trends in the 1946-1984 period, Cohen and Land (1987) further conjectured that these increases could be explained by three possible scenarios. One was that the increases were due to a short-term illegal “drugs/crime bubble,” which their models were not designed to capture. This conjecture proved somewhat prophetic relative to recent explanations of the high levels of crime reported in the late 1980s and the early 1990s. SCARY PROJECTIONS OF INCREASING VIOLENT CRIME RATES FOR THE ECHO-BOOMERS IN THE 1990s Instead of the predicted drop in crime trends through the 1980s, the American public enjoyed only a five-year hiatus from the surging crime trends of the previous two decades. The increasing violent crime rates in 1985 and 1986 noted by Cohen and Land (1987) continued to climb until 1993. Violent crime rates particularly spiked for teenage males. Since males ages 15-19 in, say, 1990 were born in the 1971-1975 period, they were members of the baby buster birth cohorts who, according to both the demographic standardization and the time-series analyses cited above, were expected to have relatively low crime rates. Yet it became evident in the late 1980s and early 1990s that these members of the tail end of the baby buster cohorts were not behaving with respect to participation in index crimes, especially violent crimes, as had been expected. Responding to these increases, some criminologists projected that the age-specific violent crime trends of young offenders (ages 14-24) would continue to rise throughout the latter part of the 1990s and into the next
OCR for page 328
JUVENILE CRIME decade (Fox, 1996). This trend in violence among adolescents—particularly shocking to the public—raised serious concerns about the potential harm posed by these youths. In addition, some analysts argued that what had typically been only a threat to lower-class, inner-city dwellers, might become a reality for the rest of society. “Americans are sitting on a demographic crime bomb . . . . Despite the recent decline in murder rates, homicides committed by 14- to 17-year-olds between 1985 and 1993 increased by 165 percent (more for minority males). The next wave of homicidal and near-homicidal violence among urban youth is bound to reach adjacent neighborhoods, inner-ring suburbs, and even the rural heartland” (DiIulio, 1995a:15). Prominent criminologists and policy scientists such as John J. DiIulio (1997), James A. Fox (1996), and James Q. Wilson (1995) also warned that rising violent crime trends would only worsen as the echo boomers aged into their crime-prone years—a phenomenon that would begin during the first two decades of the 21st century.5 Census Bureau population projections supported this contention (U.S. Bureau of the Census, 1985; 1995; 1996). The numbers of teenage males in America were due to climb by 1 million from 1995 to 2000. Based on extant cohort studies that estimated that 6 percent of the youthful population become high rate, repeat offenders, Wilson estimated that there would be 30,000 more serious offenders on the streets by the turn of the century. “Get ready,” he warned (Wilson, 1995:507). Heeding Wilson's warning and expanding his depiction of this growing tide of youthful offenders, DiIulio coined the term “superpredators”— noting that today's offenders are worse than yesterday 's and that tomorrow's will be worse than today's. “According to Professor Wolfgang, . . . each male cohort has been about three times as violent as the one before it. We concur” (Bennett et al., 1996:29). DiIulio's (as well as Wilson's) characterization of the 21st century youthful offenders is nothing short of scary (Bennett et al., 1996:27): America is now home to thickening ranks of juvenile “super-predators”— radically impulsive, brutally remorseless youngsters, including ever more preteenage boys, who murder, assault, rape, rob, burglarize, deal deadly drugs, join gun-toting gangs, and create serious communal disorders. They do not fear the stigma of arrest, the pains of imprisonment, 5 Due to the large size of the baby boomer cohorts, even with lower birth rates than their parents, the number of children they bore is larger than the number of children in the baby buster years. Because these baby boomlet birth cohorts, born 1977-1995, thus reflect their parents' large cohorts, they often are referred to as echo boomers (Crispell, 1993). In the popular press, following the labeling of the cohorts who were born just before them as Generation X, the echo boomers have been dubbed “Generation Y.”
OCR for page 329
JUVENILE CRIME or the pangs of conscience. They perceive hardly any relationship between doing right (or wrong) now and being rewarded (or punished) for it later. To these mean-street youngsters, the words “right” and “wrong” have no fixed moral meaning. Touting the success of tougher law enforcement efforts against adult offenders especially during the war on drugs (DiIulio, 1995b) and demanding the incarceration of youthful offenders as a minimum requirement for curbing the tide of violence among youthful offenders, DiIulio (1997) stated that most juvenile criminals still received no punishment for their crimes and that prosecutors and judges are unduly burdened with caseloads, which leaves them impotent against this struggle to bring justice to and incarcerate juvenile offenders. The bottom line for DiIulio (DiIulio, 1995a:16) is that “we must remain deadly serious about targeting hardened adult and juvenile criminals for arrest, prosecution, and incarceration.”6 IS AN IMPENDING EXPLOSION IN YOUTH VIOLENCE REALISTIC? Besides the projected “baby boomerang” effect of the echo boomers on violent crime rates of the 21st century (Fox, 1996:1), there is little hard evidence that changes in other social and economic forces would exacerbate or relieve the forecasted explosion in youth violence. To be sure, the demographic force of increasing numbers of echo boomer adolescents and teenagers up to about the year 2010 is inexorable. Assuming further that age-specific delinquent and crime rates, especially violent crime rates, remain constant at the high levels experienced in the 1985-1993 period, it would appear inevitable that juvenile crimes would increase substantially, especially in the years 1996-2005. Assuming also that a constant proportion of birth cohorts become high rate, repeat offenders, Wilson 's conclusion that the numbers of such offenders in the youth population — and the associated numbers of offenses they commit—will increase dramatically and disturbingly during these years also appears indisputable. With the benefit of several additional years of data, however, it is clear that the age-specific delinquent and crime rates of adolescents and teenagers rose dramatically in the 1985 to 1993 years (relative to the rates 6 In spite of his call for enhanced law enforcement efforts, DiIulio (1997: A23) argues that the superpredators—“these more savage than salvageable young criminals could not and should not be punished into submission. Instead, the only responsible option is to try and save these typically abused, neglected, fatherless, Godless and impoverished children before it's too late, working mainly through the youth outreach efforts of local churches.”
OCR for page 338
JUVENILE CRIME statistical approaches with expert judgment. One variation of this approach consists of asking a group of interacting experts to give both a point estimate and a range for fertility, mortality, and migration (Lutz et al., 1999). Another variation applies a formal Bayesian statistics framework to the combination of expert judgments in demographic forecasting (Daponte et al., 1997). One advantage of this general approach is that the combination of subjective probability distributions of a number of experts to form one joint predictive probability distribution diminishes the danger of individual bias. This approach may be particularly useful for forecasting when structural changes or unanticipated events need to be factored into the forecasts. Its main drawback is the difficulty of eliciting the necessary input from experts. A careful and sophisticated application of the stochastic and combined stochastic-expert judgment approaches to the production of forecasts of crime levels and rates for a decade or two into the future clearly requires a large research project (or projects) and is beyond the scope of this paper. However, the application of the variants/scenarios forecasting recipe combined with a dash of expert judgment can be illustrated. For this, we focused on the construction of plausible national projections of expected numbers of male teenage (ages 14-17) homicide offenders—as well as upper and lower bounds for the expected numbers—for each year from 1998 to 2007.12 This is the age group and the crime that led to the scary projections of DiIulio, (1995b), Fox (1996), and Wilson (1995) reviewed above. It thus is an instructive exercise to examine the plausibility of the assumptions necessary to produce the high-level wave of teen-age homicide offenders cited by these analysts. Since the last year for which official estimates of homicide offending rates for this age group are available is 1997, we used 1997 as a jump-off year for the projection series and constructed high and low projection series annually for 10 years from 1998 to the year 2007. To generate the high and low projection series, we first accessed homicide offending rate time series for the 14-17 age group provided in 12 Only the results of our projection exercises for annual numbers of teenage homicide offenders are reported here. However, we also have produced high and low projection series of numbers of offenders for the years 1998 to 2007 for the following crimes and population-age groups: white male homicide offenders (ages 18-24, 25-59), black male homicide offenders (ages 18-24, 25-59), all violent offenders (ages 10-17, 18-24, and 25-59), all property offenders (ages, 10-17, 18-24, 25-59), white violent offenders (ages 0-17), black violent offenders (ages 0-17), white property offenders (ages 0-17), and black property offenders (ages 0-17). Generally, the results from the projection cones for these other crimes are comparable to those for teenage homicide as reported in Figures B-6 and B-7. They are available from the authors on request.
OCR for page 339
JUVENILE CRIME Fox (1997) for the years 1980 through 1997.13 These rates were estimated by Fox from FBI Supplementary Homicide Reports and include both known perpetrators as well as an estimated share of unidentified perpetrators computed by a statistical imputation procedure. Since almost all teenage homicide offenders are males, we focused on projections for males only. Furthermore, since the white male and black male rates are quite different, we constructed projections for these two groups separately. Second, we examined the ages 14-17 black male and white male homicide rates for the years 1980-1997 to determine the highest and lowest rates observed during this period. These are: Population Category Year and Low Homicide Rate Year and HighHomicide Rate Black males, 14-17 years old 1984 - 47.6 per 100,000 1993 - 244.1 per 100,000 White males, 14-17 years old 1984 - 9.4 per 100,000 1994 - 22.4 per 100,000 Consistent with previous characterizations of trends in crime over the past two decades reviewed above, it can be seen that 1984 was the low year for homicide rates for both race groups in the 1980-1997 period—just before the 1985-1993 upsurge in young male homicide rates. By comparison, the high years occurred in 1993 for black male teenagers and 1994 for white male teens. As a third step, we next conjectured about the highest and lowest bounds that these rates could plausibly attain in the 10 years 1998-2007 — given (1) the “observed” time series of homicide offending rates for the observation period 1980-1997 and (2) the high and low rates noted above. In doing so, we had the advantage, compared to Fox (1997), of information about the teenage homicide offending rates for 1997 as well as preliminary UCR data on aggregate homicide levels for 1998. These preliminary data indicate overall declines in homicide of about 7 percent from 1997 to 1998. These overall homicide trends have not yet been transformed into age-specific offending rates (as in Fox, 1997). However, we know from Fox's (1997) data that homicide declines for teenagers in the mid-1990s were on the order of 2 to 2.5 times larger than the declines in the overall homicide levels. 13 Fox actually provides estimates of homicide offending rates by age back to 1976. But, for consistency with high and low projection series, we generated for other crime categories, we used Fox's data series only back to 1980. We also used the update for 1997 of teenage homicide offending rates provided by the Federal Bureau of Investigation at the Internet address: http://www.fbi.gov/ucr/prelim98.pdf.
OCR for page 340
JUVENILE CRIME Assuming this pattern has continued, these overall homicide rate declines suggest that homicide offending rates for black and white males ages 14-17 continued to decline by 15 to 20 percent in 1998. Although informal observations suggest the declines in overall and teenage homicide offending have continued through mid-1999 (the date of construction of our projections), it is impossible to know how long this downward trend in teenage homicides will continue. But it is clear that the lower bound of a plausible projection cone for the annual numbers of homicides for these two teenage populations must accommodate the continuing rapid decline in 1998 and possibly for a few additional years into the future. Accordingly, we chose a lower bound to which the homicide offending rates could decline of 25 percent of the lowest rate observed during the 1980-1997 period. Furthermore, since the declines in the 1997 and 1998 offending rates have continued the rapid pace of the mid-1990s, we chose to allow the lower bound for the projection cones to decline linearly to this rate within five years from the jump-off year, i.e., from 1998 to 2002, and then remain fixed for the years 2003 to 2007. With respect to plausible upper bounds for the homicide offending rates, we conjectured that if teenage homicide offending rates were to reach 125 percent of the highest rates observed during the 1980-1997 period, the public outcry would be so strong that all sorts of societal homeostatic mechanisms—from even more active policing to more active involvement of school, religious, community, and civic organizations in juvenile crime prevention programs—would come into play to stabilize the rates and pressure them down again. And yet the possibility of a new wave of teenage homicide offending associated with the coming of age of the echo boomers—like that of the 1985-1993 period—should not entirely be ruled out of a projection cone designed to contain with a high probability the possible range of future teen homicide offending. Accordingly, we set the upper bound for our projection series to 125 percent of the highest rates reported above for each race group, 1980-1997. We also chose to allow the high bound for the projection cones to increase linearly to this level over a five-year period beginning in 1998 and then remain fixed for 2003 to 2007. The fourth step in the calculation of our projection cones consisted of multiplying the projected homicide offending rates for the entire 1998 through 2007 period by Census Bureau population race-specific projections for the 14-17 age group (U.S. Census Bureau, 1996). The results of the projected upper and lower bounds for the years 1998-2007 are displayed in Figures B-6 (black males) and B-7 (white males) together with the observed series (based on the rates provided in Fox, 1997) for the years 1980 to 1997.
OCR for page 341
JUVENILE CRIME FIGURE B-6 Black male homicide offenders, ages 14-17: Observed series 1980 to 1997 with projected upper and lower bounds to 2007. FIGURE B-7 White male homicide offenders, ages 14-17: Observed series 1980 to 1997 with projected upper and lower bounds to 2007.
OCR for page 342
JUVENILE CRIME Several observations can be made about these projection cones. First, it can be seen that the projection cones widen fairly rapidly from the jump-off year of 1997 to the year 2002. This is consistent with our decision to allow the projected lower and upper bounds of homicide offending rates for these two populations to reach the respective limits in five years. Second, at the same time, on the basis of the preliminary evidence regarding homicide trends from 1997 to 1998, and evidently from 1998 to 1999, cited above, the lower bounds of the projection cones decline just rapidly enough to envelop the expected numbers of teenage homicide offenders for these two years. Third, if indeed these declines are on the order of magnitude we expect, this will reduce the numbers of teenage black and white male homicide offenders to levels last seen in the middle 1980s. Fourth, it is evident that the effects of allowing the teenage homicide offending rates to grow to a maximum of 125 percent of the highest rates observed in the 1980-1997 period by the year 2002 are to produce upper bounds that increase by 2002 to about 3,800 for black males and about 1,800 for white males. Fifth, after 2002, the lower and upper bounds for both population groups continue to exhibit slow growth to 2007. Since the homicide offending rates are held constant for these years, these increases are due to continuing growth in the teenage populations at risk during this remaining five years of the forecast period. Since we approached this projections exercise primarily from a variants/scenarios perspective rather than as an attempt at the formal combination of stochastic forecasting with expert judgment, we have not sampled expert opinion about the probabilities that should be attached to the high and low bounds of our projection cones. However, based on the historical record of juvenile homicide offending rates, we believe they would contain future numbers of teenage homicide offenders with a high (.9 or .95) probability. Within the confines of the broad upper and lower bounds for the projection cones plotted in Figures B-6 and B-7, we also can describe the trajectories of expected values as well as the probability surfaces for various paths of juvenile homicide offenders across the years shown in the graphs. For us, these probability densities initially are highest along ridges—corresponding to the paths of expected values of the series (i.e., the paths of the annual numbers of offenders we consider most likely)— running close to the lower bounds of the graphs. This is necessary in order to accommodate what evidently are continuing declines in juvenile homicide offenses in 1998 and 1999 (despite highly visible and shocking mass shootings in public middle and high schools during these years). Nonetheless, because of the inherent unpredictability of the series plotted in the figures, we also allow for small but nonzero probability densities (corresponding to the possibility that they could occur) of numbers of
OCR for page 343
JUVENILE CRIME offenders in the middle and upper reaches of the projection cones for these two years. For the years 2000 through 2002, we then allow the ridges containing our most likely scenarios/expected values of juvenile homicide offenders to continue to decline, but at decelerating rates. Because of greater uncertainty with increasing years into the forecast period, however, we concentrate the probability densities somewhat less in this region of the projection cones. For the years 2003 to 2007, we then locate the probability ridges along slightly increasing lines toward the middle part of the projection cones, due to the larger numbers of echo-boomer juveniles at risk of homicide offending in these years. We also flatten the probability surface for our forecasts for these years even more—allowing for somewhat higher probabilities that there may be another upsurge in juvenile homicide offenders later in the 10-year forecast horizon. These exercises in the calculation of expected values, probability density surfaces, and high and low rate projection bounds for juvenile homicide offenders also can be used to assess the plausibility of the forecasts of homicide and other crimes by Fox (1996, 1997) and Steffensmeier and Harer (1999) summarized above in Figures B-3 and B-5. As noted earlier, the forecasts by these analysts were in the form of single expected values for each of a series of years into the first decade of the 21st century. In contrast, the forecast cones exhibited in Figures B-6 and B-7 are in the spirit of the conventional demographic high-medium-low projection scenarios or variants. As such, they provide lower and upper bounds within which the expected values of single-series forecasts should be contained. Recall that the forecasts of Steffensmeier and Harer (1999) did not focus on teenage homicides specifically but pertained to the general categories of person and property index crimes. Assume, however, that the slow increases in the person index crime rate that they expect over the years 1997 to 2010 (to a maximum increase of about 5 percent by 2010) also imply slow increases in teenage homicide offending rates. Then it clearly is the case that the Steffensmeier and Harer forecasts would fall well within the upper and lower projection series exhibited in Figures B-6 and B-7. In fact, this even would be true if teenage homicide rates over the projection period grow at twice the general rate of increase Steffensmeier and Harer expect for person crimes. A somewhat more direct comparison can be made between Fox's single-expected-value forecast series and the projection cones in Figures B-6 and B-7 by summing the bounds in these figures to compare with the non-race-specific forecasts of total teenage homicide offenders reproduced in Figures B-4 and B-5 above. Specifically, the upper bounds of our projection cones in 2005 sum to a total number of teenage homicide offenders of about 6,200, which is well below the approximately 8,500 upper bound
OCR for page 344
JUVENILE CRIME of Fox's (1996) forecast reproduced above in Figure B-4. Our upper bound for 2005 does, however, contain Fox's (1997) forecast for this year of approximately 4,000. But it also is the case that even the latter forecast requires a considerable growth in homicide offending rates for the two teenage groups in Figures B-6 and B-7. Put otherwise, Fox's (1997) forecast lies in the upper regions of the projection cones of Figures B-6 and B-7. Thus, his 1997 forecasts are not entirely implausible, but, in view of the apparent continuing declines in homicide in 1997 and 1998, perhaps not as plausible as forecasts that fall further within our upper and lower bounds projection series. CONCLUSIONS Criminologists have engaged in a number of attempts to forecast both numbers of criminal offenders and crime rates in the United States over the past three decades. In addition to their sheer intellectual interest, there are other reasons for an increasing interest in crime forecasts, such as the policy need to plan for resources for the juvenile and criminal justice systems. Our review of several existing contributions to the crime forecasting literature suggests, first of all, that these forecasts often contain continuity biases, i.e., are heavily influenced by recent trends in crime rates in the years just prior to the period for which the forecasts are made. Admittedly, forecasts of crime rates/offenses have various purposes, one of which could be the projection of recent trends into the future in order to draw out their implications (as in the case of the Fox, 1996, projections). However, to the extent that crime forecasts are meant to go beyond drawing out the implications of recent trends to represent likely paths that crime rates and offenses may take, they should attempt to minimize, or at least be cognizant of, the effects of continuity bias on the forecasts. A second characteristic of existing crime forecasts is that they typically produce only single-expected-value projections of juvenile or adult crime rates into the future and fail to recognize the uncertainty surrounding such forecasts. It is clear, however, that just because the projection of recent levels of crime rates or trends therein into the immediate future suggests that, say, juvenile crime will rise by a certain percentage does not mean that juvenile crime will in fact rise by that amount. In other words, there is a lot of indeterminancy or, in statistical terminology, uncertainty in crime forecasts. Future efforts in crime forecasting should recognize this and attempt to provide bounds on levels of uncertainty in the forecasts. We have illustrated some ways in which this can be done by adapting and applying the high-medium-low scenarios approach widely employed in demography to the projection of annual numbers of juvenile homicide
OCR for page 345
JUVENILE CRIME offenders for the years 1998 to 2007. Based on the high-low projection cones reported above, we concluded that scary forecasts of a new wave of juvenile homicide offenders in the first decade of the 21st century are relatively implausible. Rather, it is more likely that the numbers of juvenile male homicide offenders will continue to decline during the period 1998 to 2002 and then increase slightly thereafter to the year 2007. However, the possibility that members of the relatively large echo-boomer birth cohorts will develop—as they age into their teen and young adult years—a new fad or fashion related to dangerous and violent interpersonal activities (such as a new attachment to illegal drugs) and, accordingly, that the annual numbers of teenage homicide offenders will again increase in the 1998-2007 period cannot be entirely ruled out. Our exercise in forecasting juvenile homicide offenders also illustrates two additional implications of uncertainty in forecasts of crime rates and offenders. These are that the periods over which crime forecasts are made should be as short as possible and that the forecasts should be updated as often as possible (i.e., when new or updated data are available). As noted above, large-scale social systems have elements of complexity or nonlinear dynamics and chaos that militate against the accuracy of long-term forecasts. In practical terms, this means that forecasting cones (upper and lower bounds) for enveloping the ranges within which crime forecasts are likely to fall with a high probability will grow very rapidly from the base year into the future. For instance, the forecasting cones for juvenile homicide offenders developed herein lose their informative content very rapidly (i.e., the probability surfaces of the projections become less and less concentrated around the expected values). By the fifth year into the forecasting period, the probability density surfaces for these forecasts have diffused quite extensively. This corresponds to the fact that juvenile homicide offending rates can change very rapidly. To take this into account, the time periods of the forecasts should be relatively short and the forecasts should be revised when new information becomes available. For most police, court, and penal components of the juvenile and criminal justice systems, this is not particularly problematic, as forecasts typically are necessary only for one- or two-year government budgeting cycles. Only occasionally are projections more that five years into the future required for budgeting and/or planning purposes. In sum, future forecasts of crime rates/offenders should: guard against continuity biases or at least recognize their presence in projections the objective of which is to draw out implications of recent trends;
OCR for page 346
JUVENILE CRIME take into account uncertainty in the forecasts by developing upper and lower bounds within which paths of crime rates and offenses are expected to lie; shorten the forecast time period as much as the purpose for which the forecasts are produced will allow; and be updated as often as possible. The incorporation of these characteristics into crime forecasts should result in more realistic uses and assessments of the forecasts. REFERENCES Ahlburg, D.A., and K.C. Land 1992 Special issue: Population forecasting. International Journal of Forecasting 8: 289-542. Ahlburg, D.A., and W. Lutz 1999 Introduction: The need to rethink approaches to population forecasts . Pp. 1-14 in Frontiers of Population Forecasting, Supplement to Volume 24, 1998, W. Lutz, J. Vaupel, and D. Ahlburg, eds. Ahlo, J.M. 1990 Stochastic methods in population forecasting. International Journal of Forecasting 6: 521-530. Bennett, W.J., J.J. DiIulio, and J.P. Walters 1996 Body Count: Moral Poverty and How to Win America's War Against Crime and Drugs. New York: Simon and Schuster. Biderman, A.D. 1966 Social indicators and goals. Pp. 68-153 in Social Indicators, R.A. Bauer, ed. Cambridge, MA: MIT Press. Blumstein, A. 1995 Youth violence, guns and the illicit-drug industry. The Journal of Criminal Law and Criminology 86(1): 10-36. Bureau of Justice Statistics 1995 National Crime Victimization Survey Redesign. Fact Sheet. 1998 Serious violent crime levels continued to decline in 1997. Four Measures of Serious Violent Crime. [Online.] Available: http://www.ojp.usdoj.gov/bjs/glance.cv2.htm Cohen, L.E., M. Felson, and K.C. Land 1980 Property crime rates in the United States: A macrodynamic analysis, 1947-1974, with ex ante forecasts for the mid-1980s. American Journal of Sociology 86: 90-118. Cohen, L.E., and K.C. Land 1987 Age structure and crime: Symmetry versus asymmetry and the projection of crime rates through the 1990s. American Sociological Review 52 (170-183). Crispell, D. 1993 Where generations divide: A guide. American Demographics 15(May): 9-10. Daponte, B.O., J.B. Kadane, and L.J. Wolfson 1997 Bayesian demography: Projecting the Kurdish population, 1977-1990 . Journal of the American Statistical Association 92: 1256-1267. DiIulio, J.J., Jr. 1995a Arresting ideas: Tougher law enforcement is driving down urban crime . Policy Review 74: 12-16. 1995b Why violent crime rates have dropped. The Wall Street Journal (September 6): A19.
OCR for page 347
JUVENILE CRIME 1997 Jail alone won't stop juvenile super-predators. The Wall Street Journal (June 11): A23. Fox, J.A. 1978 Forecasting Crime Data: An Econometric Analysis. Lexington, MA: Lexington Books. 1996 Trends in Juvenile Violence: A Report to the United States Attorney General on Current and Future Rates of Juvenile Offending, March. Washington, DC: Bureau of Justice Statistics. 1997 Update on Trends in Juvenile Violence: A Report to the United States Attorney General on Current and Future Rates of Juvenile Offending . Washington, DC: Bureau of Justice Statistics. Gibson, C. 1993 The four baby booms. American Demographics 15(November): 36-40. Greenberg, D. 1985 Age, crime and social explanation. American Journal of Sociology 91: 1-21. Hirschi, T., and M. Gottfredson 1983 Age and the explanation of crime. American Journal of Sociology 89: 552-584. Land, K.C. 1986 Methods for national population forecasts: A review. Journal of the American Statistical Association 81: 888-901. Land, K.C., and S.H. Schneider 1987 Forecasting in the social and natural sciences: An overview and analysis of isomorphisms. Pp. 7-31 in Forecasting in the Social and Natural Sciences, K.C. Land and S.H. Schneider, eds. Boston: D. Reidel Publishing. Lee, R.E. 1993 Modeling and forecasting the times series of U.S. fertility: Age patterns, range, and ultimate level. International Journal of Forecasting 9: 187-202. 1999 Probabilistic approaches to population forecasting. Pp. 156-191 in Frontiers of Population Forecasting, Supplement to Volume 24,1998, W. Lutz, J. Vaupel, and D. Ahlburg, eds. Lee, R.E., and L. Carter 1992 Modeling and forecasting the time series of U.S. mortality. Journal of the American Statistical Association 87: 659-671. Lee, R.E., and S. Tuljapurkar 1994 Stochastic population projections for the United States: Beyond high, medium, and low. Journal of the American Statistical Association 89: 1175-1189. Lutz, W., W. Sanderson, and S. Scherbov 1999 Expert-based population projections. Pp. 139-155 in Frontiers of Population Forecasting, Supplement to Volume 24, 1998, W. Lutz, J. Vaupel, and D. Ahlburg, eds. President's Commission on Law Enforcement and Administration of Justice 1967 Task Force Report: Crime and its Impact: An Assessment. Washington, DC: U.S. Government Printing Office. Sagi, P.C., and C.F. Wellford 1968 Age composition and patterns of change in criminal statistics. The Journal of Criminal Law, Criminology and Police Science 59(1): 29-36. Steffensmeier, D., and M.D. Harer 1987 Is the crime rate really falling? Journal of Research in Crime and Delinquency 24: 23-48. 1999 Making sense of recent U.S. crime trends, 1980-96/8: Age-composition effects and other explanations. Journal of Research in Crime and Delinquency 36 (3).
OCR for page 348
JUVENILE CRIME Stoto, M.A. 1983 The accuracy of population projections. Journal of the American Statistical Association 78: 13-20. U.S. Census Bureau 1985 Estimates of the population of the United States and components of change by age, sex, and race: 1980 to 1984. Current Population Reports, Series P-25, No. 965. Washington, DC: U.S. Government Printing Office. 1995 National and state population estimates: 1990 to 1994. Current Population Reports, Series P-25, No. 1127. Washington, DC: U.S. Government Printing Office. 1996 Population Projections of the United States by Age, Sex, Race, and Hispanic Origin: 1995 to 2050. Current Population Reports, Series P-25, No. 1130. Washington, DC: U.S. Government Printing Office. Wellford, C. 1973 Age composition and the increase in recorded crime. Criminology 11(1): 61-70. Wilson, J.Q. 1995 Crime and public policy. Crime, J.Q. Wilson and J. Petersilia, eds. San Francisco, CA: Institute for Contemporary Studies Press.
Representative terms from entire chapter: