SECTION 4 Implications for Research and Policy
Summary of Findings and Recommendations for Future Research
An overview of our findings provides strong support that violent offending has distinct etiological features above and beyond those it shares with nonviolent offending. Even employing gross comparisons, made here relying in large part on studies never designed to test our hypothesis, we see patterns that direct our attention to “good prospects” for identifying people at greatest risk of committing violent offending. We begin with a set of general recommendations for future research, and then we turn to each of the constructs covered in the book in more detail.
Methods for Testing the Differential Etiology of Violence
We started the book with a discussion about the overlap between violent and nonviolent offending (in particular the finding that violent offenders have much in common with chronic offenders). Though we have highlighted studies where techniques are employed to disentangle the two, a great deal remains to be learned. First, data sets that include indicators of both violent and nonviolent offending could easily be applied to the cause, and many authors with such data sets have not yet exploited them for this purpose. Second, using statistical modeling, similar to that used by Savage et al. (2014), would help authors report findings that allow the reader to learn about this distinction. Modeling to test the differential etiology of violence may include comparison of coefficients across models of violent and nonviolent offending using appropriate formulae, employing a control for frequency of nonviolent offending in models of violent offending, and using binary dependent variables which compare violent to nonviolent-only offenders or violent to frequent, nonviolent-only offenders.