The Effects of Sanction Intensity on Criminal Conduct - JDAI Helpdesk

Types <strong>of</strong> Outcomes
Eligible studies measure recidivism in terms <strong>of</strong> new arrests and/or convictions.
Technical violations <strong>of</strong> probation, such as absconding or failing a drug test, are also
included as a separate outcome measure. While technical violations do not inevitably
result in a recorded arrest or charge for a new <strong>of</strong>fense, they represent a failure to comply
with probation conditions that could be affected by the intensity <strong>of</strong> supervision.
<strong>The</strong> use <strong>of</strong> technical violations as an outcome measure comes with the caveat that
increased supervision intensity could increase the likelihood <strong>of</strong> a violation being detected
through increased surveillance, rather than simply a failure to comply. This caveat
applies to new criminal cases too, but to a lesser extent. New crimes are more likely to
be detected by the police than by probation <strong>of</strong>ficers, so future arrests are less likely to be
affected by the <strong>of</strong>fender’s probation status. This also makes arrest a preferable outcome
measure to charges or convictions that come further along the criminal justice process
and may be more affected by disclosure <strong>of</strong> prior sentences. Of course, police <strong>of</strong>ficers in
smaller beat areas probably know the repeat <strong>of</strong>fenders too and will adjust their discretion
to arrest accordingly. All recidivism measures suffer from inherent limitations.
Offending measured by self-report is not excluded, but most ISP studies use
<strong>of</strong>ficial records. This is a limitation <strong>of</strong> our research: it is well-known that <strong>of</strong>ficial records
can underestimate the prevalence <strong>of</strong> re<strong>of</strong>fending, and there may be confounding between
the treatment and response that could be partly overcome by using self-reports.
However, these data were simply not available to the extent needed to conduct a
meaningful analysis.
14