Ekland,
This is a great policy question, with likely far reaching implications. What is the best evidence that you know about showing that incarceration adversely affects families of the incarcerated? What are the limitations of those studies? Do you think there are other plausible explanations for the associations, e.g., confounding or selection processes?
Your pre-post matching idea is roughly the approach of the best evaluations for this type of intervention (when it is very difficult to randomize) but you are exactly right about the challenges. The synthetic control design might be useful (see comments to someone else re synthetic controls - like K Rudolph's paper that we read for seminar).
Maria