Miller is correct that causality can be inferred only from randomized experimental design. Statistical controls in quasi-experimental design only rule out the alternative causal explanations specifically controlled for in the analyses. Accordingly, while we presented our results as being consistent or inconsistent with causal hypotheses, there is no instance in our report in which we explicitly claim that spanking causes children's aggression.
Miller is also correct that 3 points of measurement permit one to rule out more alternative explanations than 2 permit. Nevertheless, our study and the study by Straus et al are the first longitudinal investigations of physical punishment by parents to rule out the autoregressive effect of child misbehavior (ie, the stability of children's aggression over time) and thus, the child effect on parent discipline. If representative randomized studies or 3-wave longitudinal studies with appropriate controls existed, we certainly would rely more heavily on them. In their absence, studies like ours with an intermediate level of causal confidence are important to bridge the gap until more methodologically stringent research has been conducted. While "measurement at only two points in time is a less-than-ideal design...two points are certainly much better than one."1
Lindner Gunnoe M, Mariner CL. In Reply. Arch Pediatr Adolesc Med. 1998;152(3):309. doi:
Customize your JAMA Network experience by selecting one or more topics from the list below.
Create a personal account or sign in to: