We thank the discussants for their comments, particularly David Nutt for his highly constructive suggestions that build upon, rather than attack, our critique [1] of the Nutt, King & Phillips [2] analysis (hereafter NKP).

The commentaries make minimal challenge to technical aspects of our analysis. Most focus instead on the failings of the current classification system, which we do not and will not defend. Fischer & Kendall [3], Room [4], Rossow [5] and Nutt himself [6] all argue that the NKP approach [2] yields results that sensibly challenge a scheduling system that is driven by political considerations. Ergo, a severe critique of NKP [2] is not helpful. We disagree. It is true that a principal justification for assessing the harms of specific drugs is to inform public policy, of which scheduling is one important element. However, scientific soundness is the criterion for judging whether the assessment tool is a good one; it is not simply whether it would give a better policy outcome by some independent but unarticulated criterion.

Obot [7] usefully distinguishes between the general desire to bring order to a domain and univariate rankings as just one tool for doing that. In that framing, our point is that univariate harm rankings are not the right tool for informing decisions about whether a substance should or should not be scheduled. They do not even have the right unit of analysis for that question, which should be the change in harm associated with a change in policy, not the level of harm under one (the current) policy.

Another argument the commentaries raise is, to caricature, that the public expects single-dimensional ratings and, by implication, is unable to work with more complex systems; yet if the public has trouble grasping multi-dimensional scales, that should be seen as a hurdle to overcome, not a restraint that needs to be accepted.

Furthermore, the decisions are not made by direct democracy, and the political actors and processes that make scheduling decisions do not need to produce scalar ratings of harm. The Food and Drug Administration and similar organizations can decide which food additives to ban without publishing univariate danger scales.

We grant Rossow's point [5] that aggregate measures reflecting the relative scale of different problems can be useful, a purpose served by, for example, cost of illness studies; yet policy choices, including scheduling decisions, must take into account multiple factors, as Fischer & Kendall note [3]. Something like the matrix that we suggested will be necessary, implicitly if not explicitly. (Fischer & Kendall [3] and Room [4] note that others have made similar arguments; we were not claiming originality and thank them for providing readers with cites to that important earlier work.)

Our discussants run the risk of trivializing the problem by suggesting that health harms are everything. Consider the idea of prohibiting cigarettes. The gains in reduced mortality and morbidity would swamp the health harms associated with resulting black markets; yet even public health hawks are wary of criminal prohibition of tobacco production. This suggests that decisions about legal status reflect the weighing of many considerations. That the current tobacco regulations weight corporate financial interests so heavily, as opposed to public health and other legitimate policy goals, is a serious issue, but nevertheless the multi-dimensionality of the decision and the goals cannot be ignored.

Room's insight [4] that psychopharmacologists do take pleasure into account but negatively is fascinating. It raises both conceptual and scaling issues. From some public health perspectives it is a problem when dangerous activities are fun, whether the activity is smoking crack, getting drunk, driving fast or motorcycling without a helmet; whereas for the economist, the fun per se is a good thing. Standard economic analysis becomes muddier when the activity itself alters the individual's tastes and preferences regarding further participation in that activity—as with drug dependence and rational addiction models. However, we stick to our claim that both NKP [2] and the current system depart from the canons of standard policy analysis because when comparing different substances, they give no positive weight to the pleasures of any substances not currently legal and unscheduled.

Declarations of interest