Table 3 Examples of simulation-based methods used for distal outcomes or comparing sweetened beverage tax designs

Demand systems models such as Almost Ideal Demand System (AIDS); Quadratic Almost Ideal Demand System (QUAIDS), Rotterdam model, Exact Affine Stone Index model (EASI)

Accounts for simultaneous substitutions and complementarity across beverage & food categories to obtain own- and cross-price elasticities of demand as well as income or expenditure elasticities

If sample large enough, able to disaggregate by income or other characteristic of interest to get subpopulation specific elasticities.

Elasticity estimates may be sensitive to how model used, beverage/food categories included in the system and sample sizes.

Beverage/food categories that are not often purchased require more complex econometric models.

Demand models are usually static and does not account for habit formation.

Connecticut & Massachusetts, USA: Uses QUAIDS to estimate SSB demand and tax among food assistance participants [74].

Chile: Uses LAIDS to derive beverage and food elasticities to infer what a tax might mean in terms of changing consumption [75].

Demand- and supply-estimations for differentiated product markets using the Berry, Levinsohn and Pakes (BLP) model

Able to account for beverage market structure and shares of firms (manufacturers and/or retailers) and household socio-demographic characteristics to assess how much of the tax is passed onto consumers (strategic pricing) and hence what the resultant consumption changes accounting for substitutions, and new market share distribution might be.

If a tax is in place, possible to validate the structural demand and supply models.

Requires detailed data on product attributes (e.g., brand, tax status) purchased by households.

Determination of inside- vs outside-option may limit interpretation as the model assumes that the price of the outside option is unchanged.

Only provides estimates of short-term supply-side response to a tax as other strategies like changing portfolio mix and reformulations may follow.

Demand models are usually static and does not account for habit formation.

France: Comparing firms’ strategic price responses to an ad valorem vs excise taxes on sweetened beverages [73].

Mexico: Comparing changes in volume of and sugar from taxed vs untaxed beverages purchased as well as tax revenues generated under SSB taxes based on sugar-density vs volume [76].

Population-based microsimulation models (PSM) of which extended cost-effectiveness analyses (ECEA) are a subset

Uses existing distribution of population characteristics collated from various data sources to construct a hypothetical population. Various policies or interventions and empirically informed effect sizes between dietary intake changes and health outcomes are applied as parameters to compare how outcomes would vary across these policies or interventions vis-à-vis the status-quo.

Assumes that diseases are independent of each other. Frequently due to data limitations, only key diet-disease relationships are included in models

Unable to account for industry responses such as reformulations or changes in marketing.

Need to define time horizon given population cohort and assume discount rates.

Validation of assumption and methods needed but often difficult.

USA: Applies the CHOICES model to estimate cost-effectiveness of a 1 cent/ounce tax on SSBs [70].

Australia: Multi-state lifetable model of a hypothetical 20% SSB tax on the monetized productivity of adults 20y or older [77].

South Africa: Estimates changes in Type 2 Diabetes-related deaths for different income groups and the resultant burden to individual and public payers due to a 20% SSB tax [78].

Computable general equilibrium (CGE) models of which input-output models or social accounting matrix are a part of.

Able to assess macroeconomic/economy-wide implications (employment, sector-specific productivity, trade, gross domestic product) using representative agents (consumers, producers, government) and accounts for import-exports for country

Assumes that demand elasticities are fixed and independent of policy

Requires additional parameters from demand systems model estimates, market share changes, PSM and cost-effectiveness estimates and thus only possible later in the lifecycle of the evaluation.

Guatemala: Considers the whole value chain, from the production of sugar to the different productive sectors that use sugar and the final consumer to evaluate the overall effects various SSB tax policies [79].