You are on page 1of 32

POLITICAL ECONOMY WORKING PAPERS

Getting tough on juvenile crime: An analysis of costs and benefits Simon M. Fass and Chung-Ron Pi
Political Economy Working Paper 04/01 OCTOBER 2001

SCHOOL OF SOCIAL SCIENCES THE UNIVERSITY OF TEXAS AT DALLAS RICHARDSON, TEXAS 75083-0688
AN EQUAL OPPORTUNITY/ AFFIRMATIVE ACTION UNIVERSITY

GETTING TOUGH ON JUVENILE CRIME: AN ANALYSIS OF COSTS AND BENEFITS

Simon M. Fass Associate Professor of Political Economy, School of Social Sciences University of Texas at Dallas

Chung-Ron Pi Senior Analyst, Dallas County Juvenile Department

Abstract: Recent decades have seen juvenile justice shift focus from the child and treatment to the offense and accountability. This change, manifest in juvenile codes that support punishment and doctrines that include transfers to adult criminal court, has had significant caseload and fiscal impact. However, scarcity of pertinent research and of analysis of costs and benefits leaves unclear whether the new, get tough approach achieves greater delinquency reduction than its predecessor. Combined with a quasi-experimental empirical simulation of the effects of punitive sanctions, a cost-benefit analysis of alternative dispositions in Dallas County, Texas, suggests that harsher sentence can indeed prevent some offenses. The value of this gain, however, is much less than what it costs to produce. As a result, by consuming public resources that might otherwise be invested in more productive purposes within or outside the justice system, the policy of toughness visits substantial opportunity costs on communities that embrace it.

This paper is based on findings of a research project sponsored by the Office of Juvenile Justice and Delinquency Prevention (OJJDP), U.S. Department of Justice (grant number 98-JN-FX-0001). The opinions expressed herein are those of the authors. They do not necessarily reflect views of the OJJDP.

GETTING TOUGH ON JUVENILE CRIME: AN ANALYSIS OF COSTS AND BENEFITS During the last thirty years the juvenile justice system's traditional focus on the child and on applying appropriate methods of diagnosis and treatment has been supplanted by an approach that focuses on the offense and on ways of holding the child accountable for it (Feld, 1991, 1999a). This transformation reveals itself in a number of ways. Important among them are: juvenile codes that explicitly support the aims of punishment, accountability, and protection of public safety; courts that operate under offense-based doctrines that include determinate sentencing, extended jurisdiction statutes, and mandatory minimum sentences; and relaxation of obstacles to transferring juvenile offenders to adult criminal court (Bishop et. al., 1998; Feld, 1990, 1999b; Torbet et. al., 1996). One factor contributing to this transformation has been the inability of researchers to show evidence that rehabilitative strategies curb delinquency (Bishop et. al., 1998). Although an early claim by Martinson (1974) that nothing works proved an overstatement, subsequent investigations, such as Lipsey (1992), showed that unambiguously effective programs were, and remain, exceedingly rare. Another factor is public perception of increasing rates of juvenile crime, especially of the violent sort. Highly-publicized media accounts of violence, exaggerated talk about an explosion in juvenile crime rates during the 1980s and early 1990s, hyperbolic predictions by Blumstein (1995), DiIulio (1995), and Snyder and Sickmund (1995) of an imminent demographic "crime bomb" that might accelerate these rates, and other like claims fueled growing public apprehension and frustration (Smith et. al., 1999). This climate contributed to charges that juvenile justice authorities, rather than addressing public safety concerns, coddled offenders by continuing to operate within the treatmentrehabilitation framework (Butterfield, 1997). It also contributed to incessant political calls to do something to assure public safety and, ultimately, to broadened electorate and legislative support for "get tough" policies. An extreme example in the 1995 Texas juvenile code is determinant sentences of up to forty years for certain felonies (Bishop et. al., 1998; Fritsch and Hemmens, 1996). 1 The impact of this shift is evident in national-level data on juvenile court practices. Between 1988 and 1997 there was an estimated 48% increase in the total number of delinquencies (Puzzanchera et. al., 2000). At the same time, there was a disproportionate 75% increase in the number of petitioned delinquency cases, a 67% increase in the number of adjudicated cases resulting in formal probation, and a 56% jump in the number of cases that resulted in out-of-home placement (i.e., incarceration). Impact is also evident at the state level. The 1995 Texas code, by exposing substantially higher numbers of young people to formal sentencing with progressively restrictive sanctions rather than diversion to informal intervention and counseling, has led to exploding caseloads, especially placement, and mounting fiscal pressure. This pressure, in turn, has driven local boards that administer the code to experiment with alternate programs they hope will achieve outcomes comparable to placement but at lower cost, such as various forms of intensive supervision probation, and short-term boot camp instead of long-term placement. Caseload and fiscal impacts of the new juvenile justice paradigm are clear to agencies that directly bear these burdens. Less clear is the answer to the question of whether the more punitive, more costly approach, through deterrence or rehabilitation, can achieve significantly greater reduction in juvenile crime than the less punitive, less costly approach that it replaces. Also unclear for those who subscribe to the view that the doctrine is less about deterrence and/or rehabilitation and more about assuring that offenders receive their just deserts, for its own sake or as means to improve electoral

standing, is whether the added suffering and/or votes are commensurate with the extra effort needed to elicit them. Two factors contribute to this lack of clarity. One is scarcity of pertinent empirical research (Bishop et. al., 1998). A recent review by Aos et. al. (2001), for example, identifies only a dozen rigorous studies that use experimental or quasi-experimental methods to compare outcomes between milder and harsher sentences on juveniles with similar offense histories and demographic characteristics. 2 The other factor is that cost-benefit analysis, a method of social inquiry expressly designed to help answer questions about the relationship between gains and the effort required to produce them, is infrequent in juvenile justice research (DiIulio, 1996; Welsh and Farrington, 2000a, 2000b). 3 Addressing these factors directly, we present a case study analysis that combines a quasiexperimental empirical simulation of the effects of more punitive dispositions with dollar value estimates of associated benefits and costs to clarify answers for one jurisdiction: Dallas County, Texas. Specifically, we look at net benefits associated with sentencing juveniles, whose offense histories qualify them for stronger treatment, to: probation rather than deferred prosecution; intensive supervision probation rather than regular probation; local residential placement under county custody rather than intensive supervision; and placement under state custody by the Texas Youth Commission (TYC) rather than local placement. After briefly reviewing shortcomings and strengths of the cost-benefit procedure, we lay out the data and methods used to estimate: differences in numbers of re-offenses and re-dispositions that follow initial shifts to more restrictive sanctions; and dollar values for unit costs and benefits. Combining these estimates, we then present our findings for each disposition pair in terms of net benefit and in terms of cost-effectiveness. These suggest that tangible costs incurred by the justice system are not matched by tangible cost-saving benefits. Together, system benefits from reduction in redispositions, and victim and other benefits from reduction in re-offenses, are not large enough to offset initial investment in harsher sanction. This observation does not necessarily imply that getting tough on juveniles is wrong. For one thing, cost-benefit analysis does not have a special capacity to make such judgement. Also, certain intangible benefits that we ignore, retribution for its own sake and vote enhancement, for example, may compensate for the shortfall. However, our finding does suggest that by consuming tangible public resources that might otherwise be invested in more productive purposes within or outside the justice system, toughness visits substantial opportunity costs upon communities that embrace it. Overview of Cost-Benefit Analysis Cost-benefit analysis, at root, is a procedure for systematically displaying and comparing favorable and unfavorable consequences of alternative courses of action. By identifying and, where appropriate, quantifying and valuing the positive and negative consequences of policy, it can help researchers, administrators, officials, voters, and interested others better understand some of the more important tradeoffs, distributional consequences (i.e., winners and losers), and other implications of their preferences (Arrow et. al., 1996; Campen, 1986; Schmid, 1989). At the same time, cost-benefit analysis conveys the idea, implicitly if not explicitly, that it is useful to treat maximization of societal benefits or, conversely, minimization of societal costs, as a policy objective no less important than retribution, rehabilitation, incapacitation, deterrence, restoration, compensation, or other justice goals.

Unfortunately, as can happen when there is failure to distinguish between method and the way it is applied, the promise that cost-benefit analysis holds for juvenile justice is sometimes undermined by exaggeration of its usefulness. Some zealous advocates view the procedure as a logical, rigorous technique that supplies definitive answers to policy questions. Adopt the action that maximizes benefits while minimizing costs, they say, and all will be well. 4 But cost-benefit analysis cannot produce unequivocal answers. Especially in domains such as juvenile justice where it is hard to quantify, monetize, or even identify many important benefits and costs (e.g., pain and suffering of victims and offenders), the procedure yields findings that are too uncertain to serve as exclusive policy guides. One source of uncertainty, common to all forms of social analysis, is the already-noted scarcity of suitable data on which to base robust estimates of policy effects. Most studies in the justice field, as a result, invent hypothetical scenarios and then suggest what cost-benefit relationships might look like under them. 5 Another source, as DiIulio and Piehl (1991) note, is imprecision in assigning money values. There is wide variation in dollar estimates for tangible costs and benefits, i.e., items that have market prices, and intangible costs and benefits, i.e., items that dont have these prices. Especially awkward, because estimates can vary by a factor of two thousand or more, are dollar values of such things as lives saved or lost, pain and suffering incurred or avoided, and quality of life heightened or diminished. 6 A third source of uncertainty is analytical bias. Cost-benefit analysis is a subjective operation that always adopts a particular point of view: that of society as a whole viewed through the lens of economic theory, which after Gittinger (1982) we call economic analysis; or those of individuals and organizations within it, financial analysis. 7 For any perspective, analysts have wide latitude to decide which costs and which benefits they include, to determine which of these costs and benefits are assigned monetary values, and to choose among several methods of estimating the values. In their turn, decision makers and other consumers of studies have scope to argue that findings of costbenefit analysis at odds with their opinions are caused by incorrect dollar estimates, inappropriate inclusion or exclusion of certain costs and benefits, or some other source of prejudice. Taken together, these uncertainties make it hard for cost-benefit analysis to provide unambiguous answers about whether benefits are sufficient in relation to costs. Difficulty, however, should in no way suggest that the method is useless - among other sensible reasons because the alternative to ambiguous answers is, simply, no answers. Absent answers, deliberations on the merits of policy are condemned to continue to turn mainly on preferences shaped by congeniality, ideology, and political expediency of the moment; quality of rhetoric brought to bear on arguments; and other enigmatic influences. In other words, although cost-benefit analysis yields uncertain answers, incorporating them into the decision making process helps to narrow the much greater, incessant uncertainty that confounds juvenile justice. Even if the method cant supply the answers, it can help a lot, as the following tries to demonstrate. Patterns of Re-offense and Re-disposition Most young people referred to the Dallas County Juvenile Department are subject to two informal and five formal dispositions that impose increasingly tighter limits on freedom of action. 8 Under the states juvenile code, the department applies milder sanctions on children having few and/or relatively minor infractions and progressively harsher sanctions on those with many repeated and/or serious infractions. 9 Each step in the disposition sequence demands greater expenditure by the

county and state. Main benefits expected from these extra outlays, depending on point of view, are additional gains in child maturation (manifesting themselves, for instance, as immediate or longerterm improvements in employment, earnings, health, civic participation, school attendance, academic grades, etc.), greater offender suffering, more offenses prevented, or a combination of these. We do not have reliable measures for behavioral progress or for distress. Our analysis, necessarily partial, therefore focuses on a relatively narrow question: What are the likely effects, first in terms of differences in numbers of re-offenses and associated re-dispositions and then in terms of differences in money values, of expending extra resources on harsher sanction? Specifically we ask: What would happen if juvenile offenders get regular probation rather than deferred prosecution, intensive supervision probation instead of regular probation, local placement rather than intensive supervision probation, and TYC rather than local placement? 10 Data to answer the question about numbers come from the records of 13,144 individuals referred to the department during 1994-97. They committed 58,650 juvenile and adult infractions and received 17,124 dispositions between their first referral and the end of 1999 (Table 1). 11 Each of our observations is an initial disposition event that ranges from the mildest sanction, deferred prosecution, to the harshest, TYC placement. 12 Although data allow tracking for up to five years of followup, 1994-99, sample size in the fifth year is too small to produce meaningful results. Accordingly, our analysis tracks four years (though most tables show three because there are no statistically significant differences in year four).
Table 1: Sample Size, by disposition and length of followup followup from start of disposition disposition deferred prosecution probation intensive supervision probation local placement TYC placement all dispositions 1 year 7,124 5,117 1,317 2,331 1,235 17,124 2 years 6,801 4,799 1,188 2,165 1,136 16,089 3 years 5,068 3,263 649 1,533 842 11,355 4 years 3,242 1,862 255 944 548 6,851

To estimate effects of substituting one disposition for another, we apply to these data a negative binomial regression procedure that controls for the influences on re-offense and re-disposition patterns of race and ethnicity, sex, age at first referral and at disposition, number and type of prior referrals, and number and type of prior adjudications.13 The exercise yields an estimate of the probable range of differences in numbers of re-offenses and re-dispositions for each disposition pair and followup year.14 Table 2 shows the results for re-offenses.15 It suggests that changing the sentence of 100 juveniles from deferred prosecution to probation would produce an estimated increase of 20.9 technical violation and status re-offenses in the first year, with a 95% probability that the true figure lies between 17.1 and 24.9, and 1.5 such re-offenses in the second year (in a range of 0.0 to 3.7). There are no differences in years three (or four). These increases stem from the combination of scrutiny that

offenders receive from probation officers and imposition of court orders that they must follow. Juveniles do not necessarily behave differently under probation than under deferred prosecution. It is just that their misbehavior is spotted. However, as a result of being picked up for technical violations and status re-offenses in the first two years, the number of misdemeanors declines by 5.4 offenses in the first year (in range of 2.2 to 8.5), and 3.6 in the second (0.1 to 6.9). For felonies, significant in the third year only, the model predicts a mid-range increase of 3.9 offenses.
Table 2: Estimated Difference in Number of Re-offenses Resulting from Change in Initial Disposition, 1 per 100 initial juveniles, by type of offense and year of followup technical, status offenses 2 misdemeanors 2 felonies 2 lower midupper lower midupper lower midinitial disposition year 3 bound range bound bound range bound bound range 1 17.1 20.9 24.9 (8.5) (5.4) (2.2) probation 2 0.0 1.5 3.7 (6.9) (3.6) (0.1) rather than 3 0.3 3.9 deferred prosecution Total 1 intensive supervision rather than probation 2 3 Total 1 local placement rather than intensive supervision 2 3 Total 1 TYC rather than local placement 2 3 (39.0) (40.8) (10.1) (0.2) (31.8) (37.5) (7.8) (0.1) (24.6) (34.1) (5.4) 0.0 (20.0) (10.5) (0.6) 3.9 (17.9) (7.1) 14.4 (2.0) (3.6) 25.2 14.1 0.0 5.3 (39.0) 11.7 (31.8) 18.5 (24.6) 0.0 (21.8) 0.0 (16.4) 0.0 (11.1) 0.0 (20.2) 0.0 0.0 (20.2) (7.4) 0.0 (15.2) 6.5 6.7 (2.0) (3.8) 0.0 (10.2) 13.2 14.7 17.7 (0.1) 17.1 5.3 22.4 11.7 28.6 18.5 (15.4) (9.0) (2.3) 0.3 3.9

upper bound

Total 2 (mid-range) 15.5 (2.1)

7.7 7.7

3.9 17.3 11.7 0.0 0.0 11.7 (63.4) 6.5 21.1 (35.8) (44.9) (7.8) (10.6) (63.3)

Total (51.1) (45.4) (39.5) (27.1) (14.1) (0.6) (7.4) (3.8) (0.1) Notes: 1. All figures shown are statistically significant at the 5% probability level. 2. Numbers in parentheses are negative, indicating fewer re-offenses compared to the milder sanction. 3. Year 4 is omitted from the table because differences in that year are not statistically significant.

Additional monitoring imposed by substituting intensive supervision for regular probation yields a single result, a mid-range rise of 11.7 technical violation and status re-offenses in the first year. Broader impacts appear only with the offense suppression effects of incarceration, as when local placement replaces intensive supervision. Here there are large mid-range declines in technical violations (31.8), misdemeanors (16.4), and felonies (15.2) in the first year. These are followed by increases: 6.5 felonies in year two, 14.4 misdemeanors and 6.7 felonies in year three. Similarly, when TYC substitutes for local placement, significant drops in numbers of violations show up in all years, misdemeanors during the first and third years, and felonies during the first year. The overall effect is summarized in the right-most column of Table 2, which shows mid-range values for the net number of all re-offenses. Probation in lieu of deferred prosecution engenders a net rise of 15.5 re-offenses in year one, a drop of 2.1 in year two, and a rise of 3.9 in year three, or a total net increase of 17.3 offenses for the whole period. Local placement in lieu of intensive supervision, likewise, produces a total net decline of 35.8 offenses, TYC in lieu of local placement a decline of 63.3.

These differences result from and run together with differences in re-dispositions, which we show for mid-range values in Table 3. Numbers are necessarily less in this case because many re-offenses are not serious enough to warrant prosecution and because re-dispositions often follow a string of infractions rather than every one. 16Our re-disposition models thus consist of five equations. Two are for first and second year re-dispositions in the juvenile system, with a third equation capturing a time-lag effect in our data of large numbers of first-year offenses adjudicated during the second year. We do not apply these equations to juvenile data for the third and fourth years because results are statistically insignificant. For these years we apply two equations to the sub-sample of adult data. Further, because the re-disposition models use only cases that have at least one re-offense during the followup period, resulting in a smaller number of valid cases, we impose another restriction. Instead of predicting re-dispositions for each offense class, feasible only for first and second year juvenile data, our estimates for years three and four predict re-dispositions for all offenses. Reflecting the juvenile codes emphasis on accountability, the table underscores that most re-dispositions are severe. In the first and second years, roughly 60% of all re-dispositions for the probation cohort are local and TYC placements. Proportions rise progressively toward 100% for the other cohorts. For adults, likewise, jail or prison is the dominant sentence in 70% or more of cases.
Table 3: Estimated Difference in Number of Re-dispositions Resulting from Change in Initial Disposition 1 per 100 initial juveniles, by type of offense and year of followup difference in number of re-dispositions 2 juvenile system initial disposition year3 probation 2.08 (0.03) intensive local TYC supervision placement placement 2.08 4.58 1.66 (0.03) 2.05 1.33 (0.04) 4.54 4.24 (0.04) 0.21 1.62 1.73 0.21 0.36 0.36 0.08 0.08 0.11 0.11 adult system probation county jail state jail prison 10.41 (0.14) 0.76 11.03 7.77 Total

probation 1 rather than deferred prosecution 2 3

Total 2.05 intensive supervision 1 0.47 rather than 2 probation 3 Total 0.47 local placement 1 (1.92) rather than 2 0.06 intensive supervision 3 Total (1.86) TYC rather than local placement 1 2 3 (0.51)

1.33 (0.77) 0.06 (0.71)

4.24 (13.82) 0.19 (13.63)

1.73 (22.26) 0.70

0.00

0.00

0.00

0.00

7.77 (38.77) 1.02

0.78 (21.56) (5.13) (0.51) (0.35) 0.78

1.84 1.84

0.57 0.57

0.90 0.90

4.09 (33.66) (5.64) (0.51)

(1.51)

(0.55) (0.81)

(3.22) (9.37)

Total (0.51) 0.00 0.00 (5.64) (0.35) (1.51) (0.55) (0.81) Notes: 1. Re-disposition figures are the mid-range of estimated probability distributions. 2. Numbers in parentheses are negative, indicating fewer re-dispositions compared to the milder sanction. 3. Year 4 is omitted from the table because differences are not statistically significant.

This severity of sanction, combined with differences in the sequence of re-dispositions that follow different initial dispositions, helps to explain the re-offense patterns described earlier. Probation, for example, because it involves closer supervision, detects more re-offenses than deferred prosecution.

This accounts for the increase in numbers of technical and status offenses in the first year, and in the second year to a lesser extent, when probation substitutes for deferred prosecution. When juveniles on probation re-offend again, most are placed in local or TYC facilities for extended periods, constraining their ability to re-offend during these periods. However, when juveniles on deferred prosecution re-offend, most receive probation. Although this probation increases detection of further re-offenses, the number of individuals on probation that start their trajectory with deferred prosecution is much less than the number initially sentenced to probation. The number of detected reoffenses is therefore small. As a result, the offense detection effect is overshadowed by the offense suppression effect of placing re-offenders that start their trajectory with probation (i.e., 100 individuals initially sentenced to it). This accounts for the drop in misdemeanors in the first and second years when probation substitutes for deferred prosecution. Conversely, when the offense suppression effect is removed as juveniles reach adulthood, mainly in the third year, there is an increase in felonies. With some variation in details, similar combinations of offense suppression and offense detection effects also explain differences in re-offense patterns for the other disposition pairs. Unit Values for Justice System Costs and Benefits Movements of individuals from shifts to more punitive sanction through subsequent re-offenses and re-dispositions are accompanied by streams of costs and benefits. The initial change in sentencing, which we imagine to apply to 100 juveniles for each disposition pair, incurs an initial incremental cost equal to the difference in average justice system costs between the harsher and milder sanction. Subsequent increases in re-dispositions then visit additional, marginal costs on the system, while declines provide marginal cost-saving benefits. In parallel, as we discuss in the next section, subsequent increases in re-offenses impose added costs on victims, their families, and various service providers (e.g., fire, ambulance, medical, etc.), while decreases offer loss-prevention gains. Average, incremental, and marginal unit costs that we use to calculate net justice system benefits are summarized in Table 4. To estimate total average costs we followed an activity-based costing procedure described in Wayson and Funke (1989), adopting a few shortcuts due to time and resource limitations. The first step in this procedure traced the processes through which individuals move for each type of disposition. The next step identified all component activities that take place under each disposition, and all resources consumed in producing each activity: direct and indirect, labor and capital. Table A-2, in the appendix, details the component activities and their unit costs. Establishing the level of resources consumed required repeat interviews with staff of the Dallas Police Department and the countys juvenile probation unit, public defenders office, district attorneys office, and other pertinent individuals to derive estimates of personnel time allocated to the different activities (e.g., hours or days per week or month per client). Multiplying staff time by wages and salaries yielded estimates of the direct human resources cost of each activity for each disposition. In some instances, such as court processing, it was necessary to observe activities directly. For others, such as police overhead, we examined budget and expenditure reports in detail. As standard procedure, however, we relied on a combination of these approaches. That is, working from the specific to the general, we aggregated detailed activity data at the individual staff level to build a picture of the whole. At the same time, working from the general to the specific, we took line item expenditures and workload measures at the aggregate organizational level and broke them down to get figures for individual activities

Though there are differences between cases that follow different disposition tracks, especially between deferred prosecution and the others, after police activities are done most go through a standard process that starts at intake assessment and ends at final disposition. Main differences between dispositions tend to involve ancillary activities that are part of one process but not another, such as matching the child with an appropriate placement facility. There are also variations in intensity or duration for the same activities, such as length of stay in detention pending court appearance, duration of supervision in the community, or time spent in out-of-home placement. Another difference is type of court hearing, e.g., plea and disposition, trial before court, trial before jury, etc., but we found no systematic variation across dispositions on this score. Large cost differentials between dispositions result mainly from what happens after a disposition decision, i.e., short-term community supervision at one end versus long-term placement in an out-of-home treatment facility at the other.
Table 4: Justice System Unit Costs, by type of disposition ($2000) cost per case deferred probation prosecution $420 $610 $90 $120 $1,250 $610 $1,200 $540 $2,200 $1,070 $1,050 $7,070 $6,020 $1,240 $2,120 probation

Juvenile System police (referral, arrest, and/or investigation) detention screening, case review Detention intake, assessment court activities 1 supervision (probation and/or parole) residential facility 2 services (community and/or aftercare) Total average cost Incremental cost for initial shift in disposition Marginal cost per re-disposition 4 - economic - financial Adult System

intensive supervision $750 $120 $2,170 $610 $1,200 $3,670 $1,580 $10,100 $3,030 $1,770 $3,070 county jail

local placement $700 $120 $6,670 $610 $1,200 $25,850 $600 $35,750 $25,650 $6,260 $27,250 state jail

TYC placement $1,060 $120 $6,500 $610 $1,200 $840 $31,510 $41,850 $6,100 $7,320 $33,320

prison and parole Total average cost 5 $5,630 $3,400 $21,060 $58,460 Marginal cost per re-disposition - economic and financial 4 $985 $595 $3,685 $10,230 Note: 1. Includes court, district attorney, liaison, defense counsel and public defender. Values are the same for all dispositions. 2. Includes facility matching and case preparation for local placement. 3. Incremental cost is the additional cost of shifting to the harsher disposition, i.e., the difference in average costs. 4. Marginal cost is the cost associated with one additional re-disposition. 5. Includes police and court costs.

To arrive at final estimates for each activity and disposition, we multiplied the direct human resources cost by three indirect cost factors derived separately for departmental overhead (e.g., personnel benefits, support staffing, equipment and vehicle usage, etc.), overhead of the countys central administration, and capital (e.g., buildings, equipment, etc.).17 This calculation did not include the residential facility cost of TYC placement. Here we used figures from Fabelo (1999) and later updates from the Texas Criminal Justice Policy Council. Fabelo and the council also supplied averages for adult system costs: probation, $3,070; county jail, $840; state jail, $18,500; and prison followed by parole, $55,900 - to which we added our own estimates of adult arrest and related police costs ($560) and court costs ($2,000).

Incremental, or additional costs of shifting from milder to harsher sanction are then the differences in adjacent averages. For example, substituting TYC for local placement requires an incremental cost of $6,100 per juvenile (i.e., $35,750 minus $41,850). Averages are appropriate for establishing these incremental costs when we imagine disposition shifts that involve 100 juveniles at a time. 18 For marginal costs (or benefits) associated with change in numbers of subsequent re-dispositions, values that are very hard to estimate accurately, we adopted a low-high estimate range: 10% and 25% of average costs. Because analysis results were almost the same when using 10% or 25%, we show marginal values in Table 4 at the center of the range, 17.5%, to simplify presentation. Marginal values for economic cost-benefit analysis, however, are not the same as for financial analysis. This is because Dallas County and the TYC use private contractors to perform many disposition-related activities. In economic analysis, where the presumed point of view is the community as a whole seen through the lens of economic theory, the marginal cost to society of adding one or a few more juveniles to a public or private counseling service or to a public or private placement facility operating at less than full capacity is small. The fee that a private contractor charges to provide for the juvenile is usually based on the average cost of supplying the service, not the actual marginal cost. As consequence, economic cost-benefit analysis does not equate marginal values with the fee. Rather, marginal economic costs for the juvenile system in Table 4 are set equal to 17.5% of average costs. In financial cost-benefit analysis the point of view is that of actors within society. In this perspective, an agency wanting to estimate how much more it would have to spend later on, or how much costsaving it might expect as a result of shifting juveniles to harsher disposition, must equate marginal cost with the contractors fee, i.e., with average cost. 19 Because the 17.5% applies only to activities carried out internally while costs for contractor services within each disposition are set equal to their average values, marginal financial costs for the juvenile system are substantially higher than economic costs. For the adult system, however, marginal economic and financial costs are both set at 17.5% of average costs because private contractor involvement, mainly for county jail, is modest. Given the small number of cases sent to county jail, application of a separate (and higher) estimate for the marginal financial cost of this sentence would have negligible impact on analysis results. Unit Values for Costs and Benefits to Victims and Others Re-offenses that precede re-dispositions engender tangible and intangible costs to victims, offenders, and others in the community, the prevention of which register as benefits. Miller et. al. (1996) provide a long list of such costs and benefits. However, we choose a narrower set of items that, in addition to being easier to monetize, are seen as especially important in the eyes of victims and most others: property losses from damage, theft, fraud, etc.; outlays for emergency responses by fire, ambulance, and police services; medical expenses to treat injuries; social services (mainly for child victims); mental health outlays to redress psychological harm; and foregone output due to death, injury, court appearances, or other events causing loss of time that might otherwise be engaged in paid or unpaid productive activity. Also incorporated into some of our calculations are losses and gains in quality of life associated with pain and suffering of victims and their families (but not of offenders and their kin) stemming from death or injury. Inclusion of this intangible item in cost-benefit analysis is controversial, among

other reasons because values for it are arbitrary (e.g., they range from $7,500 to more than $15 million per life) and because strategic selection of a particular value allows cost-benefit analysis to reach any conclusion.20 We nevertheless include this item to see how it affects results in financial analysis. The lower portion of Table 5 summarizes our schedule of estimated average economic and financial costs for each offense class under differing assumptions. The figures vary considerably. Total economic costs associated with a felony, for instance, range from $10,290 when assuming low values for services and output, to $19,940 when assuming high values.21 Likewise, total financial costs stretch from $12,270 with low estimates for services and output to $22,370 with high estimates or, including quality of life at $34,550 per felony, $56,920.
Table 5: Unit Costs to Victims and Others, by type of offense ($2000) cost per offense status offense, delinquency technical violation misdemeanor felony property loss, damage - excluding value of stolen items 1 $0 $32 $365 - including value of stolen items $0 $126 $2,795 2 police, fire, ambulance $0 $1 $78 medical services $0 $40 $454 social, victim services - low estimate $0 $1 $5 - high estimate $0 $6 $53 mental health services - low estimate $0 $2 $27 - high estimate $1 $17 $267 output, earnings - low estimate $0 $50 $9,360 - high estimate $0 $100 $18,720 quality of life $0 $1,157 $34,550 Total for economic analysis3 - low estimate for services, output $0 $130 $10,290 - high estimate for services, output $1 $200 $19,940 Total for financial analysis4 - low estimate for services, output $0 $220 $12,720 - high estimate for services, output $1 $290 $22,370 - high estimate plus quality of life $0 $1,450 $56,920 Notes: 1. Includes value of damage caused during robbery and vehicle theft. 2. Covers non-disposition activities at time of offense, e.g., traffic control, backup police, fire fighting, etc . 3. Excludes value of stolen items. 4. Includes value of stolen itemsThe first of several steps to arrive at these values was dis-aggregation cost item assumption

of the 58,650 juvenile and adult infractions in our sample into their constituent elements. 22 Second was to distinguish between offenses that incur victim and other non-justice system losses and those that do not. Relying on judgement of Juvenile Department staff, this exercise identified only one status or technical violation, intoxicated driving, that engendered losses. There were 53 such infractions during the reference period, or about 0.6 % of all violations. For misdemeanors, we estimated that 14,630 offenses, or 59 % of the total, incurred losses. The share for felonies was 86 %. Third was to assign baseline dollar values for victim and other losses to each discrete charge that incurs them. Estimates of property loss from theft and/or material damage drew from several sources. For damage incurred during the course of infractions, such as household items broken during assaults, we used data from the 1996-97 National Crime Victimization Survey (NCVS). For theft and vandalism we mainly relied on Dallas County offense records. These assign a dollar range for specific infractions, such as theft of more than $20 but less than $200. 23 Where records did not give values, in most instances of burglary and robbery for example, we again relied on NCVS data. Hall

(1997) provided values for arson-related losses, Maguire and Pastore (1999) for burglary of vehicles, and Miller et. al. (1996) for remaining items. Values for fire, ambulance, and police services not already included in disposition costs are based on analysis of detailed expenditure data for the City of Dallas. Police and fire department data are organized in a way that allows straightforward calculation of basic response costs (including or excluding transport to hospital), false alarms, and so on, and of a factor to account for central administrative overhead and capital. Miller et. al. (1996) is our source for baseline medical service costs, and for valuations of social and victim services, mental health services, and lost output and earnings. 24 We also use Miller et. als value for quality of life, i.e., $1.9 million for death, lesser sums for injuries. In step four, after adjusting figures to year 2000 dollar terms, we derived weighted average values for each category of loss by offense class and sub-class, with variations for the different assumptions noted above. Table 5 summarizes the result, detailed in appendix Table A-3. These tables indicate that the average value of property loss and damage from a misdemeanor is $126 when stolen items are included, i.e., viewed from victims financial perspective. When the items are excluded, i.e., viewed in the economic perspective as a transfer within society, the value is $32. For felonies, respective amounts are $2,795 and $365. 25 The tables also show two values for social and victim services, mental health services, and lost output, one labeled high estimate and the other labeled low. High values for social and victim, and mental health services are figures that we borrow from Miller et. al. (1996). However, Miller et. al. seem to exaggerate the frequency at which victims avail themselves of these services. But we have no better estimates. Accordingly, to see how lower figures affect analysis, we take low estimates as 10% of the high ones. Likewise, the high value for lost output borrows from Miller et. al. Low estimates are half these values. We incorporate a low estimate here to account for the observation that victims of juvenile crime, by and large, have characteristics similar to those of offenders, notably lower than average earnings potential and shorter than average life expectancies. Analysis Findings Tables 6 through 9 present the results of our analysis. In different ways, they show the net benefit over four years - positive when benefits exceed costs and negative when costs exceed benefits associated with shifting 100 juveniles to the more restrictive sanction in each of the four disposition pairs. 26 Our first finding is that harsher sanctions do not produce positive net benefits for the justice system (Table 6). Whether looked at from an economic or from a financial perspective, cost-saving benefits, if any, from reduction in numbers of re-dispositions during the followup period are not sufficient to fully offset the initial incremental cost of tougher sentencing. From a financial perspective, county, state, municipal, and federal agencies expend a combined net of $801,000 (plus or minus $35,500) over four years when probation replaces deferred prosecution. Likewise, net midrange costs for intensive supervision instead of probation, local placement instead of intensive supervision, and TYC instead of local placement are, respectively, $486,000, $1,519,000, and $395,000.
Table 6 : Net Benefit to Justice System of More Restrictive Sanction, per 100 initial juveniles, $2000 1 initial disposition initial incremental cost economic net benefit financial net benefit

10

probation rather than deferred prosecution

($602,000)

($654,000) $10,500

($801,000) $35,500

intensive supervision rather than probation ($303,000) local placement rather than intensive supervision TYC rather than local placement

($346,000) $16,000

($486,000) $72,500

($2,565,000)

($2,368,000) $53,500

($1,519,000) $133,000

($610,000)

($548,000) $15,500

($395,000) $40,500

Note: 1. Values in parentheses are negative net benefits, i.e., net costs.

Financial outlays even exceed initial incremental costs in the first two disposition pairs, probation instead of deferred prosecution and intensive supervision instead of probation. This is because these sentences, as we discussed in relation to Table 3, are followed by increases in re-dispositions. The inverse applies to the other two pairs because they are followed by fewer re-dispositions. That is, although local placement instead of intensive supervision requires an initial incremental investment of $2,565,000, cost-saving benefits stemming from fewer re-dispositions (i.e., $1,046,000 at the midrange), lower the aggregate financial burden to $1,519,000. Similarly, cost-saving benefits of TYC instead of local placement reduce net fiscal outlays from $610,000 at the start to $395,000 at the end of four years. Our second finding is that tougher sanction does not always produce positive net benefits for victims and others and, when it does, gains are substantial only after including a value for quality of life.(Table 7). Specifically, probation in lieu of deferred prosecution, rather than reducing them, increases costs to victims and others. The reason, noted in our earlier discussion of re-offenses (see Table 2), is that this shift, though it suppresses misdemeanors during the first two years, is followed in the third year by increased numbers of felonies imposing high victim costs that wipe out prior gains. The result is a net (mid-range) loss over the period of $76,580, or $188,600 with quality of life. This is a lose-loseoutcome, with public agencies and victims both bearing higher costs. Less perverse, intensive supervision in lieu of regular probation yields no victim benefits warranting mention because this action produces only increases in technical violations and status offenses, infractions with negligible impact. Although the justice system bears a higher cost, the benefit to victims and others is effectively zero.
Table 7: Net Benefit to Victims and Others of More Restrictive Sanction, per 100 initial juveniles, $2000 1 category 2 property police, fire medical social, mental output loss, ambulance services victim health damage services services services

initial disposition

probation rather than ($8,780) ($270) ($1,260) ($140) ($820) deferred prosecution intensive supervision $0 $0 $0 $0 ($12) rather than probation local placement rather $8,620 $230 $1,470 $180 $870 than intensive supervision $88,800 $235,300 TYC placement rather $12,280 $310 $2,250 $280 $1,290 $72,440 $146,490 than local placement $34,500 $114,500 Notes: 1. Financial perspective, assuming high dollar value for services and output. 2. Values are mid-ranges of probability distributions. Numbers in parentheses are negative net benefits, i.e., net costs.

all categories 3 excluding quality of life ($76,600) ($65,330) ($112,020) $29,500 ($12) $0 $0 $1 $66,600 $55,220 $105,190 $114,000 quality of life

including quality of life ($188,600) $105,000 ($12) $1 $171,800 $354,000

11

3. Due to rounding, these totals may not sum exactly.

In contrast, local placement instead of intensive supervision generates positive net benefits of $66,600 at the mid-range, or $171,800 with quality of life. Here, big gains associated with firstyear declines in misdemeanors are cut back somewhat by large increases in numbers of misdemeanors and felonies in subsequent years. Also, because the probability range for our estimate of total re-offenses is quite wide - running from -17.9 to +14.1 misdemeanors and from -20.2 to 17.7 felonies (see Table 2) - the range for victim net benefits is also wide: $114,000, or $354,000 with quality of life. These figures, bigger than their respective mid-range values (i.e., $66,600 and $171,800), imply that this disposition shift also carries with it a chance of producing negative net benefits for victims and others. So the only disposition substitution that produces unambiguous positive net benefits for victims and others is TYC for local placement, with net gains at the midrange value of $88,800, or $235,300 with quality of life. Combining justice system and victim net benefits, our third finding is that total net benefits are negative for all disposition pairs (Table 8). Whether viewed from an economic or financial perspective, the total dollar value of benefits from reduced numbers of re-dispositions and reoffenses, even when adopting a high estimate for services and output and incorporating quality of life, is unlikely to come close to the total dollar value of costs. The possible exception, seen from a financial perspective that incorporates $1.9 million as the value of human life, is TYC in lieu of local placement. The estimated probability range here is a negative net benefit lying between -$318,000 and -$8,000 (i.e., -$163,000 $155,000). The upper bound of the range is still a net cost of $8,000, or $80 per juvenile, but it is close enough to the break-even point to imply that benefits have a chance of equaling or exceeding costs when the number of serious offenses and associated redispositions prevented is large in relation to the incremental investment needed to produce these effects. Only substitution of TYC for local placement exhibits this capacity in our model, and only when quality of life is assigned a value in excess of $1.9 million. However, if one rejects the premise, as does the economic perspective, that gains or losses in quality of life are benefits or costs to society, or if one views the putting of a dollar value on life as lacking analytic meaning or utility (e.g., because it is so arbitrary), then substitution of TYC for local placement looks no better in cost-benefit terms than the other disposition pairs.
Table 8: Total Net Benefit of More Restrictive Sanction, per 100 initial juveniles, $2000 1 initial disposition total net benefit 2 assuming:: low value for high value for services and output services and output Economic Net Benefit probation rather than deferred prosecution ($691,000) $29,500 ($726,000) $50,500 intensive supervision rather than probation local placement rather than intensive supervision TYC rather than local placement Financial Net Benefit probation rather than deferred prosecution intensive supervision rather than probation ($346,000) $16,000 ($2,337,000) $110,000 ($509,000) $31,500 ($346,000) $16,000 ($2,308,000) $178,500 ($473,000) $53,000

high value for services and output plus quality of life

not applicable not applicable not applicable not applicable

($845,000) $47,500 ($486,000) $72,500

($879,000) $65,000 ($486,000) $72,500

($994.000) $140,500 ($486,000) $72,500

12

local placement rather than intensive supervision TYC rather than local placement

($1,479,000) $188,000 ($345,000) $57,500

($1,448,000) $247,000 ($308,000) $75,000

($1,370,000) $487,000 ($163,000) $155,000

Notes: 1. Numbers in parentheses are negative net benefits, i.e., costs. 2. Total net benefit combines net benefit to justice system with net benefit to victims and others.

13

These findings are highly sensitive, among lots of other things, to our choice of non-system cost and benefit items to include in analysis, and to our dollar valuations of them. Weve not examined gains and losses to offenders and their families, nor the value of retribution (i.e., punishment for its own sake), nor other items. It is often impossible to quantify such things or to derive reasonable dollar estimates for them. Had we used $7,500 as the value of a human life rather than $1.9 million, our findings would remain intact. Adoption of $15 million might change everything. It is helpful in this context to also examine the situations cost-effectiveness, i.e., net justice system dollars required to avoid one delinquency, and leave decision makers and others free to judge whether the outlay is worth what it yields. Our finding here, consistent with the previous, is that except for TYC in lieu of local placement, harsher sanction is a fairly expensive way to prevent serious offenses (Table 9). Substitution of probation for deferred prosecution, for instance, leads to a mid-range decrease of 9.0 misdemeanors, an increase of 3.9 felonies, and thus a net reduction of 5.1 delinquencies over the followup period. This means that the justice system expends between $150,000 and $164,000 (i.e.,-$157,000 $7,000) in financial terms to prevent one delinquency.
Table 9: Effect of More Restrictive Sanction on Justice System Cost Effectiveness ($2000) 1 justice system net benefit per 100 initial juveniles difference in number of delinquent re-offenses 2 misdemeanors felonies total net justice system outlay per delinquency prevented

initial disposition Economic Cost-Effectiveness probation rather than deferred prosecution intensive supervision rather than probation local placement rather than intensive supervision TYC rather than local placement Financial Cost-Effectiveness probation rather than deferred prosecution intensive supervision rather than probation local placement rather than intensive supervision TYC rather than local placement

($654,000) $10,500 ($346,000) $16,000 ($2,368,000) $53,500 ($548,000) $15,500

(9.0) 0.0 (2.0) (14.1)

3.9 0.0 (2.0) (3.8)

(5.1) 0.0 (4.0) (17.9)

($128,000) $2,100 not applicable ($592,000) $13,400 ($31,000) $900

($801,000) $35,500 ($486,000) $72,500 ($1,519,000) $133,000 ($395,000) $40,500

(9.0) 0.0 (2.0) (14.1)

3.9 0.0 (2.0) (3.8)

(5.1) 0.0 (4.0) (17.9)

($157,000) $7,000 not applicable ($380,000) $33,300 ($22,000) $2,300

Note: 1. Numbers in parentheses indicate decrease in number of re-offenses.

At the very least, this outlay is superior to that of intensive supervision in lieu of probation, where a mid-range cost of $486,000 produces no delinquency reductions (because this shift only affects technical offenses and status violations). It is also superior to local placement in lieu of intensive supervision which, as a result of being unable to match high incremental expenditure with high prevention, reduces serious offenses at a net financial cost of $380,000 per occurrence. In contrast, and precisely because the incremental system cost of making the shift is small while the delinquency prevention effect, 17.9 serious offenses, is large, TYC instead of local placement reduces these infractions at a net cost of $22,000 per occurrence. These indicators suggest that decision makers and others who favor harsher sanction to protect society from serious offenses or simply to punish bad behavior place a high implicit value on delinquent acts and/or their effects. If they prefer probation to deferred prosecution, for example,
14

then they imply that they believe the cost to the community of one delinquency has a mid-range value of at least $157,000. If worth less, then beyond serving as means to augment constituency support in the political arena, there is no obvious purpose in their willingness to spend that many tax dollars to prevent it. Likewise, a delinquency must be perceived to cost the community at least $380,000 if the way to avoid it is by substituting local placement for intensive supervision, $22,000 if TYC is preferred to local placement. Conclusion Our findings, only as accurate and as useful as our data and methods allow, are specific to Dallas County. They cannot be readily generalized to jurisdictions with very different justice policy regimes, system cost structures, demographic characteristics, and economic milieus. Notwithstanding this caution, we are confident in suggesting that communities which pursue policies of more restrictive sanction, if curious about relationships between costs and benefits, are likely to discover that material gains to victims and others, though perhaps substantial, do not make up for the added spending needed to produce the gains. Even if these policies reduce offenses, including serious ones, the composition of infractions that fall under the rubric of delinquency is such that the average value of prevented losses to victims and others is modest. At the outside, we reckon it to be about $57,000 per felony. At the same time, the instrument used to obtain this benefit is extraordinarily blunt, our model suggesting, for example, that it requires local placement of 100 juveniles to generate a net mid-range reduction of two felonies over four years. The arithmetic is then unambiguous: a $2,565,000 incremental investment in incarceration prevents two felonies that, had they happened, would have cost victims and others $114,000. Though it may point out that this relationship represents an inefficient use of resources, costbenefit analysis does not and cannot say that spending $2,565,000 to get a return of $114,000 is bad. The procedure is not certain enough, not complete enough, to provide such a definitive assertion. In addition to costs and benefits to offenders, the value of retribution, and other things, weve not accounted for the political pros and cons of advocating harsher sanction. Dollars are important. But in the political arena of decision making, where judges and other key actors are elected, so too are votes, building of constituencies, and the like. In this arena decision makers may have sound reason to act, or at least to appear to act tough. Nevertheless, our cost-benefit analysis suggests that while there may be tangible and ethereal gains, say, from actions that prevent two offenses, these actions impose substantial opportunity costs because equal or greater tangible gains also flow from not preventing them and allowing victims and others to suffer the consequences. Not spending $2,565,000 on incarceration is $2,565,000 that can be expended on valued public goods and services within the justice system and outside, or returned to taxpayers. This sum is consequential for a jurisdiction with two million people, as $360 million would be for the entire country if the county figure is extrapolated to it. So if real or imagined political necessity seems to insist upon getting tough, cost-benefit analysis informs decision makers, voters, scholars, and any others who are of mind to listen that this policy exacts a heavy communal cost. More generally, it reminds that retribution, rehabilitation, incapacitation, deterrence, restoration, compensation, and other goals are achieved at a price. Justice policy, accordingly, is ill-served when the list of goals ignores or downplays the import of maximizing benefits and minimizing costs for the whole community.

15

References Acton, J. (1973), Evaluating Public Programs to Save Lives: the Case of Heart Attacks, Report R950-RC. Santa Monica, CA: Rand Corporation. Aos, S., Phipps, P., Barnoski, R., and Lieb, R. (2001), The Comparative Costs and Benefits of Programs to Reduce Crime, Olympia, WA: Washington State Institute for Public Policy. May. Arrow, K.J., Cropper, M.L., Eads, G.C., Hahn, R.W., Lave, L.B., Noll, R.G., Portney; P.R., Russell, M., Schmalensee, R., Smith, V.K., Stavins, R.N. (1996), "Is There a Role for Benefit-Cost Analysis in Environmental, Health and Safety Regulation?" Science, v272, n5259 , p. 221(2), April. Baron, D. (1992), The Analysis of Count Data: Overdispersion and Autocorrelation. Sociological Methodology, vol. 22, pp. 179-220. Bartko, J.J. (1961), The negative binomial distribution: a review of properties and applications, Virginia Journal of Science, vol. 12, pp. 18-37. Barton, W. H. and Butts, J. A. (1990), "Viable Options: Intensive Supervision Programs for Juvenile Delinquents," Crime and Delinquency 36( 2): 238- 256. Bishop, D.M., Lanza-Kaduce, L., and Frazier, C.E. (1998), Juvenile Justice Under Attack: an Analysis of The Causes and Impact of Recent Reforms, University of Florida Journal of Law and Public Policy, 10, pp 129-160 Blumstein, A., (1995), "Violence by Young People: Why the Deadly Nexus?" National Institute of Justice Journal, 229 (August), pp. 2-9 Bourque, B.B., Cronin, R.C., Pearson, F.R., Felker, D.B., Han, M. and Hill, S.M. (1996), Boot Camps for Juvenile Offenders: An Implementation Evaluation of Three Demonstration Programs, Washington, DC: U.S. Department of Justice. pp.103-104 Butterfield, F.(1997), "Juvenile Courts in Chaos: Critics Propose Their Demise," New York Times, July 21, p.1. Cameron, A.C., and Trivedi, P.K. (1986), Econometric Models Based on Count Data: Comparisons and Applications of Some Estimators and Tests, Journal of Applied Econometrics, vol. 1, pp. 29-53. Campen, J.T. (1986), Benefit, Cost and Beyond, New York: Ballinger Colgan, C.S. (1998), Cost/Benefit Analysis, Maine Correctional System, State of Maine Correctional Facilities Capital Plan, Cost/Benefit Analysis, Final Report. Davidson, W. S. and Redner, R. (1988), "The Prevention of Juvenile Delinquency: Diversion from the Juvenile Justice System," in Price, R.H., Cowen, E.L., Lorion, R.P., and Ramos-McKay, J. (eds.), 14 Ounces of Prevention: A Casebook for Practitioners, Washington, D.C.: American Psychological Association. pp. 123- 137.

16

DiIulio, J.J. and Piehl, A.M. (1991), Does prison-pay? Brookings Review 4: 28-35. DiIulio, J.J. (1996), Help Wanted: Economists, Crime and Public Policy, Journal of Economic Perspectives, vol. 10 (1). Winter. pp.3-24 DiIulio, J.J. (1995), "Arresting Ideas," Policy Review, 74 Dunford, F.W., Osgood, D.W, and Weichselbaum, H F. (1982). National Evaluation of Diversion Projects - Final Report, U.S. Department of Justice, National Institute for Juvenile Justice and Delinquency Prevention, Washington D.C. Elrod, H.P. and Kevin, M. (1992), "Second Wave Evaluation of a Multi-Faceted Intervention for Juvenile Court Probationers," International Journal of Offender Therapy and Comparative Criminology 36 (3): 247- 262. Empey, L.T. and Erickson, M.L. (1972), The Provo Experiment: Evaluating Community Control of Delinquency. Lexington, MA: Lexington. Fabelo, T. (1999), "Oranges to Oranges: Comparing The Operational Costs of Juvenile and Adult Correctional Programs in Texas," Criminal Policy Justice Council, Austin, Texas. January Fagan, J. and Reinarman, C. (1991), "The Social Context of Intensive Supervision: Organizational and Ecological Influences on Community Treatment," in Armstrong, T.L. (ed), Intensive Interventions with High Risk Youth, New York: Willow Tree Press. pp 341- 394 Feld, B.C. (1999a), A Funny Thing Happened on the Way to the Centenary: Social Structure, Race and the Transformation of the Juvenile Court, Punishment and Society, 1, (2), pp. 187-214. Feld, B.C. (1999b), The Transformation of the Juvenile CourtPart II: Race and the `Crack Down' on Youth Crime, Minnesota Law Review, 84, (2), pp. 327-395. Feld, B.C. (1991), The Transformation of the Juvenile Court, Minnesota Law Review, 75, (3), pp. 691-725. Feld, B.C. (1990), The Punitive Juvenile Court and the Quality of Procedural Justice: Disjunctions Between Rhetoric and Reality, Crime and Delinquency, 36, (4), pp. 443-466.. Frank, R.H. (2000), Why is Cost-Benefit Analysis So Controversial? Journal of Legal Studies, vol. XXIX, June. pp. 913-930 Fritsch, E.J. and Hemmens, C. (1996), An Assessment of Legislative Approaches to the Problem of Serious Juvenile Crime: a Case Study of Texas 1973-1995, American Journal of Criminal Law (23) Spring. pp. 563-608 Gardner, W., Mulvey, E., and Shaw, E. (1995), Regression Analyses of Counts and Rates: Poisson, Overdispersed Poisson, and Negative Binomial Models. Psychological Bulletin, vol. 118, no. 3, pp. 392-404. Gittinger, J.P. (1982), University Press Economic Analysis of Agricultural Projects, Baltimore: Johns Hopkins

17

Gray, C.M., Conover, C.J., and Hennessey, T.M. (1978), Cost Effectiveness of Residential Community Corrections: An Analytical Prototype, Evaluation Quarterly, 2(3):375-400. Greenwood, P.W., Model, K.E., Rydell, C.P., and Chiesa, J. (1996a), Diverting Children From a Life of Crime: Measuring Costs and Benefits, Santa Monica CA: Rand Corporation. Greenwood, M. and Yule, G.U. (1920), An Inquiry into the Nature of Frequency Distributions Representative of Multiple Happenings with Particular Reference to the Occurrence of Multiple Attacks of Disease or of Repeated Accidents, Journal of the Royal Statistical Society, Series A, vol. 83, pp.255-279. Hall, J.R. (1997), U.S. Arson Trends and Patterns - 1996, Fire Analysis and Research Division, National Fire Protection Association. Hausman, J., Hall, B.H., and Grilliches, Z. (1984), Econometric Models for Count Data with an Application to the Patents-R&D Relationship, Econometrica, vol. 52, pp.909-938. Hser, Y. and Anglin, M.D. (1991), Cost-Effectiveness of Drug Abuse Treatment: Relevant Issues and Alternative Longitudinal Modeling Approaches, in Cartwright, W.S. and Kaple, J.M. (eds.), Economic Costs, Cost-Effectiveness, Financing, and Community-Based Drug Treatment, Research Monograph 113, U.S. Department of Health and Human Services, Rockville, MD: pp 67-93. Hubbard, R.L. and. French, M.T. (1991), New Perspectives on the Benefit-Cost and CostEffectiveness of Drug Abuse Treatment, in Cartwright, W.S. and Kaple, J.M. (eds.), Economic Costs, Cost-Effectiveness, Financing, and Community-Based Drug Treatment, Research Monograph 113, U.S. Department of Health and Human Services, Rockville, MD: pp. 94-113. Johnson, N., Kotz, S., and Kemp, A. (1992), Univariate Discrete Distributions. John Wiley: New York. Jones-Lee, M.W. (1976), The Value of Life: an Economic Analysis. Chicago, IL: University of Chicago Press. Karoly, L.A., Greenwood, P.W., Everingham, S.S., Houb, J., Kilburn, M.R., Rydell, C.P., Sanders, M., and Chiesa, J. (1998), Investing In Our Children, Santa Monica CA: Rand Corporation. Land, K., McCall, P., and Nagin, D. (1996), A Comparison of Poisson, Negative Binomial, and Semi-parametric Mixed Poisson Regression Models, Sociological Methods and Research, vol. 24, no. 4, pp. 387-442, May. Land, K.C., McCall, P.L., and Williams, J.R. (1992), "Intensive Supervision of Status Offenders: Evidence on Continuity of Treatment Effects for Juveniles and a 'Hawthorne Effect' for Counselors," in McCord, J. and Tremblay, R.E. (eds), Preventing Antisocial Behavior, Interventions from Birth through Adolescence, New York: Guilford Press. Ch. 15. Lanoie, P., Carmen, P., and Latour, R. (1995), The Value of a Statistical Life: A Comparison of Two Approaches, Journal of Risk and Uncertainty, 10:235.257. Lerman, P. (1975), Community Treatment and Social Control. Chicago: University of Chicago Press.

18

Levitt, S.D. and Venkatesh, S.A.(1998), An Economic Analysis of a Drug-Selling Gangs Finances, National Bureau of Economic Research Working Paper Series, Working Paper 6592, June. Lipsey, M.W., (1984), "Is Delinquency Prevention A Cost-Effective Strategy: A California Perspective," Journal of Research in Crime and Delinquency, vol. 21, no. 4, 279-302, November. Lipsey, M.W. (1992), "Juvenile Delinquency Treatment: A Meta-Analytic Inquiry into the Variability of Effects," Cook, T.D., Cooper, H., Cordray, D.S., Hartmann, H., Hedges, L.V., Light, R.J., Louis, T.A., and Mosteller, F. (eds.) Meta-Analysis for Explanation, New York: Russell Sage Foundation, pp 83-127. Maguire, K. and Pastore, A.L. eds. (1999), Sourcebook of Criminal Justice Statistics [Online]. available: http://www.albany.edu/sourcebook Martinson, R. (1974), What Works? - Questions and Answers About Prison Reform, The Public Interest 35 (Spring): 2254. Miller,T.R., Cohen, M.A., and Wiersema, B. (1996), Victim Costs and Consequences: A New Look, National Institute of Justice, Research Report 155282, Washington D.C. February. National Council on Crime and Delinquency. (1987), The Impact of Juvenile Court Intervention, San Francisco: NCCD. Puzzanchera, C., Stahl, A.L., Finnegan, T.A., Snyder, H.N., Poole, R.S., and Tierney, N. (2000), Juvenile Court Statistics 1997. Washington, DC: Office of Juvenile Justice and Delinquency Prevention. Quay, H.C. and Love, C.T. (1977), "The Effect of a Juvenile Diversion Program on Re-arrests, Criminal Justice and Behavior, 4: 377- 396. Rasmussen, D. and Yu, Y. (1996), An Evaluation of Juvenile Justice Innovations in Duval County, Florida, Florida State University (mimeo). Roberts, A.R. and Camasso, M.J. (1991), Juvenile Offender Treatment Programs and Cost-Benefit Analysis, Juvenile and Family Court Journal, pp37-47. Schmid, A.A. (1989), Benefit-Cost Analysis: A Political Economy Approach, Boulder:Westview Press. Severy, L.J. and Whitaker, J.M. (1982), "Juvenile Diversion: An Experimental Analysis of Effectiveness," Evaluation Review, 6( 6): 753- 774. Smith, S.K., Steadman, G.W., Minton, T.D., and Townsend, M. (1999), Criminal Victimization and Perceptions of Community Safety in 12 Cities, 1998, NCJ Report Number173940. Washington, D.C.: U.S. Department of Justice. May. Snyder, H.N. and Sickmund, M. (1995), Juvenile Offenders and Victims: a Focus on Violence, NCJ Report Number 153570, Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

19

Torbet, P.M., Gable, R., Hurst, H., Montgomery, I., Szymanski, L., and Thomas, D. (1996), State Responses to Serious and Violent Juvenile Crime,. Wayson, B.L., and Funke, G.S. (1989), What Price Justice?: A Handbook for the Analysis of Criminal Justice Costs, NCJ Report Number 106777, Washington, DC: National Institute of Justice, U.S. Department of Justice. Welsh, B.C. and Farrington, D.P. (2000a), Correctional Intervention Programs and Cost-Benefit Analysis, Criminal Justice and Behavior, vol. 27, no.1, February. pp. 115-133 Welsh, B.C. and Farrington, D.P. (2000b), Monetary Costs and Benefits of Crime Prevention Programs, in Tonry, M. (ed.), Crime and Justice: A Review of Research, vol. 27. Chicago: University of Chicago Press. pp. 305-361 Winkelmann, R.and Zimmerman, K. (1995). Recent Developments in Count Data Modeling, Journal of Economic Surveys, vol. 9, pp. 1-23. Zimring, F.E. and Hawkins, G. (1995), Incapacitation, New York: Oxford University Press.

20

Appendix
Table A-1: Independent Variables Used to Estimate Re-offenses and Re-dispositions variable name description RACE_W caucasian = 1, african-american is the reference group RACE_H hispanic = 1, african-american is the reference group RACE_O other ethnic group = 1, african -american is the reference group FEMALE female = 1, male = 0 AGEF1314 age at first referral 13-14 = 1, ages under 12 is the reference group AGEF15 age at first referral 15 = 1, ages under 12 is the reference group AGEF16M age at first referral 16 and over = 1, ages under 12 is the reference group AGED1314 age at disposition 13-14 = 1, ages under 12 is the reference group AGED15 age at disposition 15 = 1, ages under 12 is the reference group AGED16M age at disposition 16 and over = 1, ages under 12 is the reference group NUMFEL1 one prior felony referral = 1, no prior felony referral is the reference group NUMFEL2 two prior felony referrals = 1, no prior felony referral is the reference group NUMFEL3M three prior felony referrals = 1, no prior felony referral is the reference group NUMREF2 two total prior referrals = 1, one prior referral is the reference group NUMREF3 three total prior referrals = 1, one prior referral is the reference group NUMREF4M four and more total prior referrals = 1, one prior referral is the reference group DRUG_DUM any prior drug referral = 1, otherwise = 0 PERS_DUM any prior assault referral = 1, otherwise = 0 STAT_DUM any prior status referral = 1, otherwise = 0 PROP1 one prior property referral = 1, none is the reference group PROP2 two prior property referrals = 1, none is the reference group PROP3M three prior property referrals = 1, none is the reference group TVIOL1 one prior technical violation = 1, none is the reference group TVIOL2M two or more prior technical violations = 1, none is the reference group PRADJ_DU any prior adjudication = 1, otherwise = 0 PRPLM_DU any prior residential placement = 1, otherwise = 0 mean 0.2850 0.2913 0.0227 0.1820 0.4362 0.2078 0.1616 0.2661 0.2572 0.4187 0.3273 0.1462 0.1517 0.1880 0.1156 0.3179 0.2037 0.3392 0.1679 0.3918 0.1581 0.1766 0.0865 0.0540 0.2001 0.0601 standard dev. 0.4514 0.4544 0.1488 0.3859 0.4959 0.4057 0.3681 0.4419 0.4371 0.4934 0.4692 0.3533 0.3588 0.3907 0.3198 0.4657 0.4028 0.4734 0.3738 0.4882 0.3649 0.3813 0.2811 0.2259 0.4001 0.2378

21

Table A-2: Estimated Juvenile Justice System Unit Costs, by type of disposition and activity direct cost indirect disposition and associated activity unit cost per unit subtotal Deferred Prosecution police case $260 $260 $140 detention screening case $20 $20 $30 case review case $20 $20 $20 six-month supervision (180 days) day $1 $180 $320 Total $480 $510 Probation police case $380 $380 $200 detention screening case $50 $50 $70 detention (6.93 days) day $80 $560 $620 intake assessment case $240 $240 $340 district attorney case $80 $80 $120 district court case $200 $200 $280 court liaison case $80 $80 $110 defense counsel, public defender case $260 $260 $0 supervision (365 days) day $2 $870 $1,210 community services (44% of cases for 80 days) day $10 $420 $590 Total $3,140 $3,540 Intensive Supervision Probation police case $460 $460 $240 detention screening case $50 $50 $70 detention (12.6 days) day $80 $970 $1,080 intake assessment case $240 $240 $340 district attorney case $80 $80 $120 district court case $200 $200 $280 court liaison case $80 $80 $110 defense counsel, public defender case $260 $260 $0 ISP supervision (182.5 days) day $6 $1,010 $1,410 probation supervision (182.5 days) day $2 $430 $600 community services (65% of cases for 80 days) day $10 $620 $870 Total $4,400 $5,120 Local Placement police case $430 $430 $230 detention screening case $50 $50 $70 detention (41 days) day $70 $2,980 $3,320 intake assessment case $240 $240 $340 district attorney case $80 $80 $120 district court case $200 $200 $280 court liaison case $80 $80 $110 defense counsel, public defender case $260 $260 $0 placement matching and case preparation case $160 $160 $220 contract, department facility (270 days) day $90 $24,010 $0 aftercare services (72 days) day $3 $240 $330 Total $28,730 $5,020 TYC Placement police case $660 $660 $340 detention screening case $50 $50 $70 detention (39.6 days) day $70 $2,900 $3,230 intake assessment case $240 $240 $340 district attorney case $80 $80 $120 district court case $200 $200 $280 court liaison case $80 $80 $110 defense counsel, public defender case $260 $260 $0 TYC facility (270 days) day $110 $29,730 $0 TYC parole (90 days) day $9 $800 $0 Total $35,000 $4,490 Note: Columns and rows may not sum exactly due to rounding.

total cost $1997 $390 $50 $40 $510 $990 $580 $110 $1,180 $580 $200 $480 $190 $260 $2,080 $1,010 $6,670 $700 $110 $2,040 $580 $200 $480 $190 $260 $2,430 $1,040 $1,490 $9,520 $660 $110 $6,290 $580 $200 $480 $190 $260 $370 $24,010 $570 $33,720 $1,000 $110 $6,130 $580 $200 $480 $190 $260 $29,730 $800 $39,480

$2000 $420 $50 $40 $540 $1,050 $610 $120 $1,250 $610 $210 $510 $200 $280 $2,200 $1,070 $7,060 $750 $120 $2,170 $610 $210 $510 $200 $280 $2,570 $1,100 $1,580 $10,100 $700 $120 $6,670 $610 $210 $510 $200 $280 $400 $25,450 $600 $35,750 $1,060 $120 $6,500 $610 $210 $510 $200 $280 $31,510 $840 $41,850

22

Table A-3: Estimated Unit Costs to Victims and Others, by offense class and sub-class ($2000)

property loss, damage offense class and sub-class including theft ALL STATUS/TECHNICAL VIOLATIONS $0 MISDEMEANORS assault, injury assault, no injury criminal mischief and other damage possession/delivery of drugs, other items theft, burglary, forgery, etc. escape/evasion criminal trespass false alarms obstruction disruption administrative other misdemeanors ALL MISDEMEANORS (weighted average) excluding theft 1 $0

police, fire,medical ambulance 2 services

social, victim services high low $0

mental health services high $1 low $0

output

quality life low $0 $0

of

high $0

$0

$0

$0

$19 $14 $335 $0 $302 $0 $36 $0 $0 $0 $0 $1 $126

$19 $14 $335 $0 $12 $0 $36 $0 $0 $0 $0 $1 $32

$0 $0 $0 $0 $0 $0 $0 $60 $0 $0 $0 $0 $1

$366 $0 $0 $0 $0 $0 $0 $0 $0 $0 $0 $13 $40

$54 $10 $0 $0 $0 $0 $0 $0 $0 $0 $0 $0 $6

$5 $1 $0 $0 $0 $0 $0 $0 $0 $0 $0 $0 $1

$114 $77 $7 $0 $6 $0 $0 $0 $0 $4 $0 $44 $17

$11 $8 $1 $0 $1 $0 $0 $0 $0 $0 $0 $4 $2

$897 $83 $0 $0 $3 $0 $0 $0 $0 $0 $0 $0 $100

$449 $41 $0 $0 $2 $0 $0 $0 $0 $0 $0 $0 $50

$10,200 $1,734 $0 $0 $72 $0 $0 $0 $0 $0 $0 $0 $1,157

FELONIES assault, injury $109 $109 $67 $1,405 $257 $26 assault, no injury $20 $20 $0 $0 $475 $48 child neglect $0 $0 $122 $4 $991 $99 criminal mischief and other damage $7,900 $7,900 $342 $0 $0 $0 possession/delivery of drugs, other items $0 $0 $0 $0 $0 $0 theft, burglary, forgery, etc. $3,900 $165 $61 $0 $5 $0 escape/evasion $6 $6 $119 $172 $0 $0 false alarms $0 $0 $1,043 $0 $0 $0 obstruction/disruption/admin. offence $0 $0 $0 $0 $0 $0 homicide $234 $234 $1,264 $20,062 $0 $0 other felonies $0 $0 $0 $0 $0 $0 ALL FELONIES (weighted average) $2,795 $365 $78 $454 $53 $5 Notes: 1. Includes value of damage caused during robbery and vehicle theft. 2. Non-disposition activities at time of offense, e.g., traffic control, backup police, fire fighting, etc.

$903 $1,577 $1,074 $29 $0 $11 $0 $77 $0 $5,665 $0 $267

$91 $158 $107 $3 $0 $1 $0 $8 $0 $566 $0 $27

$3,390 $298 $30 $3 $0 $51 $0 $0 $0 $1,180,140 $0 $18,718

$1,695 $149 $14 $2 $0 $26 $0 $0 $0 $590,070 $0 $9,360

$38,577 $6,084 $8,058 $167 $0 $394 $0 $0 $0 $1,938,000 $0 $34,556

23

The 1995 Texas code requires local juvenile justice boards to adopt a seven-step policy of progressive sanctions. Each step defines sanction options available to judges and probation officials for different classifications of delinquents. These classifications are based on type of offense; past criminal or delinquent behavior; effectiveness of previous interventions; and an assessment of special treatment, counseling, or training needs. Boards may deviate from these guidelines when they deem other actions more appropriate. All boards exercise this option, most using it for the bulk of cases.
2

Five studies compare diversion to regular juvenile court processing, i.e., Davidson and Redner (1988), Dunford et. al. (1982), National Council on Crime and Delinquency (1987), Quay and Love (1977), and Severy and Whitaker (1982). Five compare intensive supervision probation to regular probation, i.e., Elrod and Minor (1992), Empey and Erickson (1972), Fagan and Reinarman (1991), Land et. al. (1992), and National Council on Crime and Delinquency (1987). Two compare intensive supervision probation to incarceration, i.e., Barton and Butts (1990) and Lerman (1975).

Noteworthy applications of cost-benefit analysis include Greenwood et. al. (1996a) and Karoly et. al. (1998), who look at longer-term impacts of early childhood intervention programs on delinquency, and Lipsey (1984), who looks at delinquency prevention. Hser and Anglin (1991) and Hubbard and French (1991) examine drug treatment, while Colgan (1998), Gray et. al. (1978), Rasmussen and Yu (1996), and Roberts and Cammasso (1991) explore sentencing options. In the most exhaustive use, Aos et. al. (2001) apply the method to fourteen juvenile programs, including: multi-systemic and functional family therapy; aggression replacement training; multidimensional treatment foster care; diversion; intensive supervision probation/parole; sex offender treatment; and boot camps.
4

The position of these advocates stirs opposition which, at the other extreme, rejects all possibility of the method serving constructive purpose (see Frank, 2000). Because it implies rejection of the utility of weighing pros and cons, a characteristic human behavior, such extreme opposition is a bit over the top.
5

Aos et. al. (2001), for example, rely on findings from what they consider rigorous evaluations in different states to estimate effects for Washington state. Because the evaluations involve small samples and short followup periods, e.g., less than 260 individuals and two years in diversion studies from Michigan, Aos et. al. use Washington recidivism data to extend the analytical time horizon to seven years. Though the work comes closer to reality than most, it remains an analysis of hypothetical scenarios. This is the unavoidable consequence of data scarcity.
The literature reports estimated costs for juvenile probation (in $2000 terms) that range from $2.20 per day in Bourque et. al. (1996) to $5.90 in Aos et. al. (2001). Residential placement estimates run from $33 in Roberts and Camasso (1991) to $115 in Fabelo (1999). More extreme are values for human life. These range from $7,500 per life in Acton (1973) and Levitt and Venkatesh (1998) to more than $15 million in Jones-Lee (1976) and Lanoie et. al. (1995).
7 6

Economic cost-benefit analysis, among other strictures imposed by adherence to theory, insists that costs and benefits incurred and received by all affected societal actors be counted. Thus material and psychological losses resulting from theft are balanced by material and psychological gains to the wrongdoer and others who may benefit from use of the stolen items. If net gains to

victims. This may explain why few of the cost-benefit studies noted earlier, even those by economists, are of this type. They are financial analyses, adopting perspectives that are mainly shaped by personal, cultural, occupational, political, bureaucratic, and other considerations.
8

The two informal dispositions are supervisory caution, where a child is counseled and then released, and deferred prosecution, where the child is sent home with a warning to stay out of trouble. The mildest formal disposition is probation, under which a probation officer visits the child once every month or two for a period set by the court, usually a year. Next is intensive supervision probation, involving more frequent, weekly monitoring of behavior for an initial period of six months followed, if the child does not re-offend, by six months of regular probation. Third is local placement in a residential facility operated directly by the county or by a private contractor. This incarceration can be in a secure facility from which getting away is hard or an open facility from which leaving is easy. Placement averages 270 days, followed by 70 days of counseling, education, and other aftercare services. Fourth is placement in a secure state facility, operated directly by the TYC or for it by private contractors, also for an average of 270 days, followed by 90 days of parole. Last is determinant sentencing, the most severe juvenile disposition. There is also a very rare disposition: transfer to the adult criminal court system. Juveniles sentenced to TYC placement during 1994-99, for example, had a mean of 2.4 years of prior contact with the justice system, during which they accumulated an average of 7.3 offenses, largely felonies, for assaults and property crimes. Those given deferred prosecution had a mean of 0.5 years of contact and an accumulation of 1.47 offenses, mainly misdemeanors.
10 9

We compare effects between adjacent rather than distant dispositions for two reasons. First, juvenile characteristics are similar between proximate dispositions. It is hard to predict what would happen if placement substitutes for deferred prosecution because children receiving such dispositions are very different from each other, especially with respect to their offense histories and probable reoffense and re-adjudication trajectories. Of those who got deferred prosecution during 1994-97, for example, 18% were re-adjudicated during the following 48 months, and a bit more than 7% were placed. But among juveniles who received local placement, more than half were re-adjudicated and 41% were placed again. Differences are much smaller between neighboring dispositions. The second reason is that courts exercise decisions on the margin between dispositions. Judges do not choose harsher treatment unless milder ones prove ineffective. To be useful, analysis should compare realistic alternatives.
11

These data come from three sources. First is the Dallas County Juvenile Departments database on offenders arrested and referred between the ages of 10 and 17 years. It supplies all information on re-offenses and re-dispositions while individuals are under jurisdiction of the juvenile court. Second is the Texas Department of Public Safety (DPS), providing data on arrests after individuals reach adulthood. Third, necessary because DPS records dont give reliable information on adult dispositions, is the countys criminal court database. Here we drew a random sub-sample of 1,774 cases from the juvenile pool who received probation, intensive supervision, local and TYC placement during 1994-96. Most reached adulthood by 1999. After matching them with DPS offense records, we extracted from court data the associated dispositions and length of sentence.
12

By initial we mean that each observation represents the first instance that a juvenile

statistical models, such as ordinary least squares regression, which assume a continuous distribution for the endogenous variable (Johnson et. al., 1992; Cameron and Trivedi, 1986). A common discrete distribution is the Poisson count (Hausman et. al., 1984; Bartko, 1961; Greenwood and Yule, 1920). But this model is limited by its distribution assumptions, which lead to biased as well as inefficient estimates when applied to over-dispersed data, i.e., where the dependent variable has a variance that exceeds its expected value (Land, et. al.,1996; Johnson, et. al, 1992; Winkelmann et. al.,1995). The negative binomial regression model, commonly used to address over-dispersed data, allows for a less restrictive variance and, as a result, yields less biased estimates (Baron, 1992; Gardner, et. al, 1995). We specify two sets of dependent variable in our equations. For differences in re-offenses, dependent variables are number of technical violations and status offenses, number of misdemeanors, and number of felonies committed during each year. For re-dispositions that result from the offenses, the dependent variable is total number of re-dispositions each year. Applying the negative binomial regression procedure, we model expected number of re-offenses and re-dispositions (yi), with coyi ~ Poisson(*i) variates xi, as: *i = exp( xi + i) where and ei ~ gamma(1/,1/) Here, ei represents the unknown heterogeneity across observations. It follows a gamma distribution with mean of 1 and variance .. The set of co-variates xi in the models include the basic demographic and offense history variables shown in appendix Table A-1. By controlling statistically for the influence of other variables presumed to affect rates of re-offense and re-disposition, the model allows some isolation of the effect of sentencing. At the same time, model coefficients, or parameters, allow generation of re-offense and re-disposition predictions for different cohorts. That is, for re-offenses, the model simulates what a typical individuals re-offense count would be if he or she received the harsher disposition. For re-dispositions, it simulates the probability of being re-adjudicated for additional offenses of one type or another under different initial dispositions. With this statistical procedure, we generate predictions of differences in numbers of re-offenses between disposition pairs through manipulation of "dummy" variables that represent each disposition cohort. For example, in simulating probation versus deferred prosecution, the model specification is:
14

*i = exp(1i xi + 2i D i + i)
Here, Di is the dummy variable for type of disposition, set to 1 if the child received probation or to 0 if the child received deferred prosecution. Other independent control variables are represented by xi. The coefficient for the dummy variable, 2i , is estimated from a pooled sample, i.e., containing individuals in both deferred prosecution and probation cohorts. This coefficient represents the estimated difference in number of re-offenses for the probation cohort relative to the deferred prosecution cohort. Given the parameter estimates for x and D, and the fact that D is set to 0 for deferred prosecution, the predicted number of actual re-offenses for individuals who received deferred prosecution is then:

deferred prosecution cases with the same offender characteristics. These new re-offense predictions, however, are what the model predicts would result from changing the disposition from deferred prosecution to probation. The difference between the two predictions, one based on the actual disposition dummy value (i.e., D = 0) and the other on the modified dummy value (i.e., D = 1) while holding other offender characteristics constant (i.e., as represented by xi), yields the predicted difference in number of re-offenses that our simulation model says would result from shifting to the harsher sentence.
15

Table 2 shows only the statistically significant results. The model predicts differences for every disposition pair, but many of them are not significant at a 95% probability level, i.e., where the probability that an observed difference stems from chance is more than 5% or, conversely, where the probability that the difference is not zero is less than 95%. The determining factor is the statistical significance of the disposition dummy variable.
16

We used a procedure similar to that for offenses, with two modifications. One is that there is no manipulation of disposition dummy variables. Instead, we derive them from estimated coefficients of re-offense variables in the re-disposition models. That is, rather than using actual numbers of redispositions for re-offenders, we apply statistical models to generate expected numbers of for a given number of predicted re-offenses. We do this because, given that the first step in our procedure is simulation of differences in numbers of re-offenses, subsequent steps involving re-dispositions must maintain the same underlying assumptions. Thus in our re-disposition model we try to account for relationships that are likely to exist between a juveniles initial disposition, his or her offense and other characteristics, re-offense pattern, and the probability of being re-adjudicated in each year of the followup period. Setting number of re-dispositions as the dependent variable and re-offense counts and other characteristics as independent variables, we estimate coefficients for re-offenses. These represent the estimated number of re-dispositions for each re-offense, which we use to generate re-disposition predictions for each cohort. In the second modification, whereas re-offense prediction models involve separate estimates for each of three offense classes for every year of followup, we estimate type of re-disposition by applying the countys actual disposition distribution pattern to total re-disposition counts.
17

Overhead includes salary and wage benefits, salaries and benefits of support and supervisory staff, supplies, and materials. Central administration overhead represents the proportion of supporting activities provided by other county agencies that contribute to the operation of direct units. Major items here are building maintenance and utilities, central administrative personnel, and budget control and management. For capital equipment and facilities, our analysis of county and police expenditures indicated that depreciation of structures over forty years, and vehicles and equipment over three years, yields an annualized expenditure equal to less than 3% of annual operating budgets. To be safe, we adopted a 5% cost factor for these items.
18

Dallas County averages 1,780 deferred prosecution cases per year, 1,280 probation, 330 intensive supervision, 580 local placement, and 310 TYC placement. With the spirit of realism that directed us to compare adjacent rather than distant dispositions also telling us to take a realistic view of the magnitude of feasible shifts between dispositions in any one period, we think that a scale of 100 juveniles is reasonable. It is small relative to the frequency of deferred prosecution and probation

community services are provided by contractors. Departmental policy, logically, is to first fill its own facilities and services and, in the opposite direction, to first drain contractor facilities and services. On the margin, therefore, movement of juveniles into or out of placement, and into or out of community services associated with this and other dispositions, are shifts toward and away from contractors. The situation is the same for TYC placement. In both instances the marginal cost of services and of placement (i.e., excluding disposition-related activities done in house) is equal to the contractor fee.
20

Several methods are used to derive estimates for the value of human life and, less often, of physical and mental anguish associated with injury. Common among them is the revealed preference approach, which assumes that individuals, implicitly or explicitly by the choices they make in the market, express their beliefs as to what their own lives are worth. Also common is the contingent valuation, or survey approach, which presumes that one can arrive at a value for life by asking people what they think it is rather than trying to deduce it from their market behavior. Less common is the jury award method, used by Miller et. al. (1996), which presumes that juries are representative samples of the communities from which they are drawn, and that the awards they make represent the values that these communities attach to the pain and suffering involved in different kinds of cases. This method is rarer than the others, among other reasons because with published jury awards that reach beyond $200,000,000 per case, most analysts share Zimring and Hawkins (1995) view that it yields estimates that ...are opportunistic, arbitrary, inconsistent, and too high (p. 138).
21

Values in Table 5 would be somewhat lower with use of marginal rather than average costs for applicable items, i.e., police, medical, social, and mental health services. However, when values range by multiples of more than five for misdemeanors to more than seven for felonies, this adjustment would make only negligible difference in results. That is why we do not use marginal pricing here.
22

The 58,650 infractions comprise: 9,710 status violations distributed across 15 discrete charges; 24,790 misdemeanors distributed across 12 sub-classes and 134 charges; and 24,150 felonies distributed across 11 sub-classes and 178 discrete charges. Sub-class refers to general subcategories of each offense class. For example, misdemeanor sub-classes include: assault with injury, assault with no injury, criminal mischief and other damage, possession or delivery of drugs or other illegal items, etc. Discrete charge refers to specific offenses within each sub-class. Under criminal mischief and other damage, for instance, charges include: criminal mischief valued at less than $20, criminal mischief valued at between $30 and $50, reckless damage or destruction, desecration of a venerated object, failure to give notice upon striking an unattended vehicle, graffiti, etc.
23

To pinpoint a value to represent each range, we used 85% of the mid-point of the bracket deriving this proportion from analysis of relationships between means and mid-points for comparable ranges in the NCVS.
24

Our estimate of the present value of lifetime output and earnings losses resulting from premature death, following methods described by Hartunian et. al. (1981) and Rice et. al. (1989), yielded a figure of $880,000, about the same as would obtain after converting the Hartunian and Rice estimates into year 2000 dollar terms. But because Miller et. al. account for certain details that

26

The tables are based on application of system unit costs to predicted differences in numbers of re-dispositions, and unit costs for victims and others to differences in numbers of re-offenses. Given the range of uncertainty in our predictions of re-offenses and re-dispositions, there is necessarily a corresponding range of uncertainty for estimates of net benefit. At the same time, there is variation in the assumptions that underlie our dollar value estimates (i.e., economic versus financial perspective, high estimate for services and output versus low, inclusion versus exclusion of quality of life). This means that there is a specific range of probable values for the net benefit for each combination of assumptions. To address uncertainty, we adopted a Monte Carlo simulation procedure to generate probable ranges for net benefits under each set of assumptions. The procedure randomly sampled predicted differences in re-offenses and re-dispositions from their respective 95% probability distributions, i.e., between their lower and upper bounds. Every run of the simulation selected one re-offense and one re-disposition figure for each disposition pair and year of followup. At the same time, it multiplied each figure by the relevant unit cost for the system and for each of five sets of assumptions about unit costs to victims and others (i.e., two economic net benefits, one assuming a low and the other a high estimate of service and output costs; and three financial net benefits, one assuming a low estimate of service and output costs, the second a high estimate, and the third the same high estimate plus a value for quality of life), all discounted at an annual rate of 5% to bring them to year 2000, present value terms. Summed over four years, the procedure yielded one net benefit estimate for the justice system and, after adding victim and other losses and gains, two total economic net benefit estimates and three total financial net benefit estimates. Repetition of this procedure 1000 times for every disposition pair produced 1000 estimates for each of the six net benefits, distributed as 95% probability distributions between lower and upper bounds on each side of a mid-range dollar value.

You might also like