A last limitation is that our data are based on surveys, and so
causal inferences are limited. We did try to handle this by measuring
the mediator at Time 2 in Study 2 and through a complex
within- and between-individuals design in Study 3. Ideally, we
would have liked to have a bigger sample in Study 3 to increase
statistical power, but we made a classical trade-off between the
costs of data collection and the likelihood of finding the effects
(Snijders & Bosker, 1999). Still, recent research suggests that the
results of Study 3 were unlikely biased by the sample size, particularly
because bootstrapping was also used (Maas & Hox,
2005). Further, we note from social psychology experimental
results that victimization causes self-regulation impairment
(Baumeister et al., 2005), which causes harmful behaviors (Stucke
& Baumeister, 2006). Their evidence, the preliminary evidence we
cited earlier, and our results suggest that our proposed causal chain
is plausible.