What I see the author doing as I read his work is finding a lot of ways to use different definitions of terms to determine which offenders count as "drug offenders". His assumptions are notably different from the assumptions of some other researchers. It seems possible to me, at least on first skimming of these articles, that a lot of the differences in the data that Pfaff uses can be accounted for by these alternative definitions of terms.

For instance, I grant that a "violent offender" who is in jail on a principal charge of shooting a police officer while in the process of being arrested on a drug charge or during a drug raid was indeed violent, and is rightly viewed differently than a "non-violent drug offender".

But we still have to acknowledge that if there were no War On Drugs, then there wouldn't be as many of those drug raids and drug arrests to begin with. The circumstances for many of those violent crimes simply would not occur.

So, the process by which you sort the offenders into batches seems to have a strong impact on the conclusions you make.