Use and cumulation of evidence from modelling studies to inform policy on food taxes and subsidies: biting off more than we can chew?

BACKGROUND Food tax-subsidy policies are proposed to hold promise for helping to produce healthier patterns of food purchasing and consumption at population level. Evidence for their effects derives largely from simulation studies that explore the potential effects of untried policies using a mathematical modelling framework. This paper provides a critique first of the nature of the evidence derived from such simulation studies, and second of the challenges of cumulating that evidence to inform public health policy. DISCUSSION Effects estimated by simulation studies of food taxes and subsidies can be expected to diverge in potentially important ways from those that would accrue in practice because these models are simplified, typically static, representations of complex adaptive systems. The level of confidence that can be placed in modelled estimates of effects is correspondingly low, and the level of associated uncertainty is high. Moreover, evidence from food tax-subsidy simulation studies cannot meaningfully be cumulated using currently available quantitative evidence synthesis methods, to reduce uncertainty about effects. Simulation studies are critical for the initial phases of an incremental research process, for drawing together diverse evidence and exploring potential longer-term effects. While simulation studies of food taxes and subsidies provide a valuable and necessary input to the formulation of public health policy in this area, they are unlikely to be sufficient, and policy makers should not place excessive reliance on evidence from such studies, either singly or cumulatively. To reflect known and unknown limitations of the models, results of such studies should be interpreted cautiously as tentative projections. Modelling studies should increasingly be integrated with more empirical studies of the effects of food tax and subsidy policies in practice.