Our previous work in this area (Gottfredson et al., 2002; Wilson et al., 2001) summarized earlier focused on the effects across different treatment modalities, such as instructionally based self-control or so- cial competency programs using cognitive–behavioral methods and the general focus of the programs (envi- ronmental vs. individual change approaches). Below we focus on the relationship between other aspects of a program’s characteristics and program effects: pro- gram duration and who delivers the intervention.
The length of the evaluated interventions varied
greatly, ranging from less than 1 month to 84 months,
not counting summers. Although the modal program
length was 9 months, or a full school year, half of
the programs were implemented over a period of less
than 41/ months. A random effects inverse variance 2
weighted regression analysis found a small positive,
statistically nonsignificant relationship between the length of the intervention (measured in months) and the observed effect size (unstandardized B = 0.03, p = .06, using logged length of program). The scatter plot in Fig. 1 shows the relationship between length of the program (on a logged scale) and effect size. As can be seen, the positive relationship is driven in large part by a single outlier.6 The relationship remains non- significant without this outlier. Thus, the evidence sug- gests that program length is not predictive of student outcomes.
Duration of prevention services is also related to the type of prevention service provided. As in other similar analyses, the number of studies on which modality-specific analyses are based is small, but examining this relationship separately for the five modalities with 10 or more effect sizes we also found no relationship between program duration and effect size with the exception of a miscellaneous category of “other instructional programs.” The diversity of pro- grams in this residual category prevents any mean- ingful conclusion regarding duration of instructional