In many high-income countries, the rise in obesity and incidence of chronic diseases has lead to calls for the imposition of targeted food taxes and subsidies in order to improve the quality of diets, lower energy intakes, and reduce the burden of diet-related diseases. Finland has followed the trend with the introduction on 1st January 2011 of a small excise tax on confectionaries, ice-cream, and soft drinks, but the debate about the desirability of fiscal instruments to promote public health is ongoing. Important but unresolved questions surround the form of the tax (food-based or nutrient-based? Excise or ad-valorem?), its coverage in terms of product groups or nutrients, whether it should be combined with subsidies to promote consumption of healthy foods, the costs that it imposes on the food industry, and its effectiveness in improving diets and nutritional health. We sought to contribute to that debate by reviewing the theoretical basis for differentiated food taxes (DFTs), and the recent but quickly expanding empirical literature on the subject. At a conceptual level, DFTs can be justified by the large external costs that are caused by unhealthy diets and related diseases, although it is over-consumption of unhealthy foods, as opposed to moderate consumption, that should ideally be taxed - unfortunately a practical impossibility. More convincingly, there is evidence that consumers do not act rationally when choosing foods and diets, and in that context taxes can be used to correct behavioural failures, such as a lack of self-control. It is important to acknowledge, however, that a tax imposes costs on private individuals and the public sector, and therefore should only be considered socially desirable if its benefits outweigh its costs. This is an empirical question, which relies crucially on the effectiveness of DFTs in driving food choices towards healthier options. In Finland, that question has only been addressed in two recent studies that unfortunately reach different conclusions, with only Kotakorpi et al. (2011) supporting the idea that moderate DFTs would induce sizeable improvements in nutritional health. The other study (Irz, 2010) suggests that energy intake and obesity would hardly be affected by a tax targeting foods high in fat and sugar, while the change in diet quality would be ambiguous. Although unsatisfactory, the lack of certainty regarding the effect of DFTs in Finland is rather typical of the situation in other countries, which points to the inherent limitations of the econometric techniques that have been applied to date to analyse the subject. The main problem lies with the difficulty of estimating own-price and cross-price elasticities accurately at a level of product disaggregation that makes sense from a nutritional - rather than economic - point of view. Neither of the two Finnish studies seems entirely satisfactory at that level, and it is therefore difficult to argue that, in the Finnish context, the introduction of DFTs is truly evidence-based. In spite of the absence of unambiguous conclusions in the literature, a few important messages emerge from the review. First, the effect of DFTs on food choices, dietary intakes, and health are complex because taxing a particular food has implications for the entire diet due to the complex and largely unknown relationships of substitutability and complementarity that foods entertain with one another. This is a point of practical importance, as some studies have shown that, due to unintended effects, well-meaning taxes would most likely increase the risk of diet-related diseases rather than decrease it. The complexity of the food choices that have been put to light by empirical research stands in sharp contrast with the way DFTs are typically communicated to the public by labeling the targeted foods as 'junk' and relying on the intuitive appeal of the law of demand. Second, moderate taxes applied to a broad range of foods deemed unhealthy do not typically generate large behavioural responses and improvements in diet quality. The Finnish study by Kotakorpi et al. (2011) stands as an exception but the fact that their results rely largely on a single parameter, namely the own-price elasticity of demand for 'sweets and sugar', which is extremely high by international standards, should invite caution when drawing the policy implications of that study. Third, achieving an unambiguous improvement in diet quality and effective targeting of the policy often requires the combination of several taxes and subsidies. The design of an efficient policy therefore requires the careful comparison of different tax and subsidy schemes in terms of their effects on diet quality and health. Once in place, any policy should be regularly monitored and evaluated, with possible changes to tax rates and range of products or nutrients subject to the tax. An additional advantage of combining a subsidy to a tax would be to alleviate the suspicion that DFTs are introduced as a way of raising government revenue rather than improving public health. Fourth, an efficient incentive scheme should align its instruments as closely as possible to its public health objective. For instance, if the health policy seeks to reduce consumption of particular energy-dense but nutrient-poor foods (i.e., "empty calories" such as sugar-sweetened beverages), an excise tax which is simple and easy to administer is likely to be more efficient than a nutrient tax. On the other hand, if the objective is to reduce intakes of a given nutrient, taxing foods differentially according to their nutrient content, and hence making it possible for consumers to substitute low-content foods for high-content ones within product groups, is likely to be more efficient than a tax applied uniformly to broad food groups. A nutrient tax presents the additional advantage of creating an incentive for product reformulation towards healthier options by food manufacturers. Ultimately, however, policies should be compared in terms of the costs required to achieve the same improvement in health outcomes. Fifth, it should not be forgotten that fiscal instruments affect supply as well demand of foods. Rather than assuming a 100% pass-through rate from suppliers to consumers, the transmission of a tax along the food chain needs to be investigated empirically as well. More specifically about the Finnish policy context, several avenues should be explored in order to inform the ongoing debate: o In the search for a cost-effective and efficient policy, and noting the limited range of policy instruments that have been studied in Finland to date, more work is needed to identify the incentive schemes capable of delivering large improvement s in public health at minimum cost. With respect to the ongoing policy debate, we note that none of the existing studies is capable of capturing the defining feature of a nutrient tax (e.g., sugar tax), which is to link the tax rate of each food to its nutrient content. The comparison of food-based and nutrient-based taxes therefore remains unexplored, and future studies should also include additional food groups (e.g., dairy products) and/or nutrients (e.g., saturated fat, fibers) in the analysis. o Because modern consumers are faced with a wide choice of foods and, even within narrow product groups, the nutritional content of those foods varies greatly, quality issues reflecting within-group substitutions need to be better taken into account when assessing the effects of taxes and subsidies. o The experience of other countries suggests that, ultimately, the policy debate is unlikely to be resolved by econometric studies alone. Real-life experiments and randomized controlled trials provide an alternative and potentially fruitful area of research, but are costly and difficult to organize. In the near future, the evaluation of the current confectionary tax in Finland should be set as a priority, and the policy discussion surrounding the saturated fat tax introduced in October 2011 in Denmark should be followed closely. o The impact of the proposed policies on the food industry, which to date has been largely neglected, should be investigated. This is necessary even to understand the public health impact of the policies, given that the transmission of any tax or subsidy in a concentrated sector such as the food industry is unlikely to be perfect. Further, the costs imposed on industry should be taken into account in any impact assessment seeking to establish whether the proposed policy is socially desirable in the sense that its benefits outweigh its costs.