Commentary on Coulson & Caulkins (2012): Optimizing drug scheduling


Coulson & Caulkins [1] critically examine the past performance of administrative drug-scheduling processes in dealing with drugs that have emerged since 1971. The authors consider whether the existing drug-scheduling process slants toward over-scheduling, incurring costs including premature cessation of research on compounds that might turn out to have medical benefits. Evaluating dozens of decisions in the United States against the standards set by the international conventions and national laws, and relying on published sources for evidence of error, they find only two possible administrative ‘Type I’ errors (cases of overscheduling): the placements of the possibly non-addictive Propiram and the potentially useful 3,4-methylenedioxymethamphetamine (MDMA) in Schedule I. They identify four possible ‘Type II’ errors (cases where harmful drugs remained unscheduled). The authors also find that legislative action is responsible for one possible ‘Type I’ error: the placement of γ-hydroxybutyric acid (GHB) in Schedule I despite potential medical benefits. Those results suggest a tolerably accurate administrative process, with no clear bias toward overscheduling.

This undoubtedly useful analysis does not answer all of the questions that might be asked.

As the authors note, it might be objected that counting decisions, giving each molecule equal weight, might not be appropriate. If, for example, it were to turn out that MDMA has value in the treatment of post-traumatic stress disorder [2], and that hasty scheduling in 1985 had the effect of delaying research into those benefits by a quarter of a century, that one error might outweigh, in terms of importance, scores of correct decisions to schedule drugs that never would have found substantial user populations.

The authors chose to treat as fixed the scheduling provisions of the Controlled Substances Act (21 USC 812), which controls the scheduling process in the United States, but that law embodies an important logical lacuna. To fit the criteria for Schedule I, a drug must have ‘high potential for abuse’ and must not have any ‘currently accepted medical use’. Schedules II–V are for drugs that have recognized medical use, and reflect decreasing levels of abuse potential. There is thus no category for a drug of modest abuse potential but no currently recognized therapeutic value, nor for a drug with medical potential that waits to be determined. The proposal, mentioned by the authors, to make MDMA a Schedule III drug to reduce its commercial sale but facilitate medical research would not have been in accord with the statute, at least as currently interpreted. The United Kingdom, by contrast, recognizes three different classes of banned drugs, according to their associated levels of risk.

Taking the current legal framework as given limits the analysis in other ways. As the authors note, the conventions and the Controlled Substances Act recognize only a single class of benefit from the use of any drug: the treatment of disease. The mere pleasure-giving quality of a drug does not count, or rather counts as a negative, as widespread voluntary non-medical use is treated as evidence of ‘abuse potential’. (As one administrator involved in the process summarized the matter: ‘If it's fun, it's Schedule One’.) Nor does the possibility that, under some circumstances, a drug can facilitate experiences resembling mystical epiphanies [3], with possibly lasting and beneficial personal changes as a result [4].

Relatedly, the possible conflicts between the drug laws and religious liberty and the recognized rights of indigenous peoples to pursue their traditional practices have no weight in the scheduling decision.

As a purely logical matter, the fact that a given drug may induce abuse or dependency in some fraction of those who use it does not imply that all use of that drug is harmful. Indeed, for almost all drugs (the exception being nicotine in cigarette form) the majority of users never meet clinical criteria for abuse or dependency, and there is no good reason to think in general that use, without abuse, is harmful. The ordinary canons of economic analysis treat providing consumers with what they want as a benefit, and in the absence of clinical evidence of abuse or dependency there is no good reason to treat drug use differently.

That being the case, a broader analysis of the scheduling issue might take a risk–benefit perspective, weighing the avoided costs of abuse against the foregone benefits of non-abusive consumption. An alternative, and necessarily more complex, analysis might use dynamic programming to consider scheduling not as a yes/no decision but as a series of choices over time: at each moment, the agency may choose either to schedule or to wait. Waiting allows for an eventual decision to be made with the benefit of additional information, but incurs some risk of allowing the growth of what may be dangerous practices around the substance in question. A dynamic-programming perspective would shift the discussion from one of accuracy to one of timing: of a set of decisions, how many were made too early and how many too late (including the possibility that the substance in question should not be scheduled at all)?

Even, then, accepting the Coulson & Caulkins verdict of ‘not guilty’ with respect to the charge of administrative hyperactivity, one might still regard the existing scheduling regime as a whole as in need of substantial reform.

Declaration of interests



Steven Davenport provided invaluable research assistance.