GENERAL RISK AND SERIOUS ADOLESCENT OFFENDING
Intervening with adolescent offenders to prevent continued offending would be a relatively straightforward task if one could identify those who would be chronic, serious, and/or violent offenders early in their offending careers and correct the factors that were most influential in producing this pattern of behavior. As noted in Chapter 1, however, this amounts to predicting and intervening to stop a relatively rare event; serious, violent, chronic adolescent offenders are a small proportion of the general adolescent offending population. This group is both proportionately and numerically quite small, and when the focus is restricted to the most serious delinquent offenders, for example, the chronically violent offender, it is exceedingly small (Snyder, 1998). In addition, the markers that differentiate this group cleanly at the start of their offending careers are rather limited in their predictive power.1
The power of a risk marker to predict future arrest or the impact of an intervention to reduce the likelihood of future arrest is often depicted in terms of an “effect size.” An effect size is a metric that can be compared across multiple studies; it indicates how much impact a particular risk variable or intervention has on whether an individual is arrested. It is useful for comparing results across studies because, unlike indicators of statistical significance, it is less affected by the size of the samples examined. In the studies of interventions considered later in the chapter, the effect size indicates the average observed difference in arrest rate between a treated group and a comparison group. If a study indicates that a treated group has an arrest rate of 25 percent and the comparison group has an arrest rate of 35 percent, that intervention has an effect size of .10, a 10 percent lower rate of rearrest. Effect sizes across multiple studies are examined using a technique called meta-analysis, which uses regression approaches to identify aspects of programs that are related to larger or smaller effect sizes among the pool of studies examined.
1 The term “risk marker” is used throughout this section. This is in keeping with the distinction made by Kraemer and colleagues (1997), in which a marker has a documented association with a later outcome, and a factor has substantiation that the observed association with the later outcome is causal (i.e., changing the risk factor has been shown to reduce the likelihood of the outcome). Overwhelmingly, the research on risk for future delinquency has demonstrated the presence of risk markers, with much less evidence that these risk indicators are risk factors related to later delinquency. The literature uses these terms loosely and interchangeably. The wording used here is believed to be reflective of the general state of the literature, and further specific distinctions would be distracting.