Optimal Taxation and Food Policy: Impacts of Food Taxes on Nutrient Intakes New Directions in Welfare – OECD, Paris – July 2011 Thomas Allen (University of Perpignan, CIHEAM/IAMM-MOISA and INRA-ALISS) Olivier Allais (INRA-ALISS) Véronique Nichèle (INRA-ALISS) Martine Padilla (CIHEAM/IAMM-MOISA) Outline of the presentation • Background • Research objectives • Methodology • Results • Discussion Background • Increase in the prevalence of obesity and overweight in France since 1990 (Obépi, 2009); • Higher risk of illnesses for which nutrition is an essential determinant, among the low-income groups (InVS, 2006) • Nutrient-rich food are associated with higher diet costs and energy-dense food with lower costs (Darmon et al., 2007); • Public Health authorities’ questioning and academic discussion on the prospect of potential « fat taxes ». Research Question How best to design a fiscal policy improving households’ allocation of goods in terms of nutrient adequacy to recommendations? Objective Identify the optimal price conditions improving households’ diet quality. Review of the Litterature • Food consumption economics : Estimation of a food demand system to capture price elasticities (Deaton et al.) • Health studies : Definition of the public health question and tools of analysis (Drewnowski et al.) • Public economics: Modelisation of the optimal taxation conditions (Ramsey, Murty et al.) Optimal taxation model Ramsey's model (1927) Taxes' objective: Raise funds. Planner's ojective: Maximise social welfare under the constraint that tax revenue covers a given level of public expenditure. Max .V Vp,I s.c. Rti xi i Optimal taxation model Inverse elasticity rule • Ramsey rule: The reduction in demand for each good, caused by the tax system, should be proportional for each good. • Inverse elasticity rule: Optimal tax rates on each good should be inversely proportional to the good’s own–price elasticity of demand. tk 1 p e k kk Optimal taxation model Application to a nutritional policy objective • Taxes' objective: Transforming consumption behaviours. • Planner's objectif: Maximise social welfare under the constraint that the overall diet quality of consumers' food basket reach a minimum level in terms of nutrient adequacy to recommendations. Max .V Vp,I s.c. Rti xi i obj quali quali ix i i Optimal taxation model A nutritional quality/price ratio • Optimal financing criteria : The optimal tax rates, for each good, are decreasing functions of their own-price elasticity of demand. • Optimal adequation criteria: The optimal tax rates, for each good, are decreasing functions of their « nutritional quality/ price » ratio. 1 t quali 2 1 k k p 2 e 2 p k kk k Optimal taxation model System of simultaneous equations • The maximization program results in a system of equations where each optimal price variation, tk, : * * t f quali , e , p , x , t , k Where quali, p and x are vectors of the diet quality indicators, initial prices and quantities associated with each good and e the own and cross price elasticities. • Solving this sytem requires to estimate a complete food demand system. Methodology – Demand model A conditionally linear system • Selection of the Almost Ideal Demand System model (Deaton and Muellbauer, 1980): n w ln p (ln M ln P ) i i ij j i i j 1 • Iterated Least Square Estimator (Blundell and Robin, 1999). Methodology – Pseudo-Panel Data • A panel of scanner data: - 156 periods: 1996-2007 - 8 cohorts: Date of birth/Social status - 27 food groups • Group agregation: Homogenous categories in terms of nutritional content (fruits/vegetables fresh/processed, snacks/already prepared meals, vegetable/animal fat, salty/sugary fat). • Price construction: 24 clusters of price according to Localisation/Social status. Methodology - Nutrient adequacy indicators • MAR: • LIM: • SAIN: 1nnutri ij MAR 100 i n j 1RNI j Add _ suga Sat _ fat Na 1 i i i LIM 10 i 3 3153 2250 1n nutri ij 100 nj 1RNI j SAIN 100 i ENER i Nutrient adequacy indicators MAR - Mean adequacy ratio The MAR for a 100g of food i: n n Nutri 1 ij 1 MAR NAR i ij n RNI n j 1 j 1 j The MAR for a food basket: np p 1 x MAR P i NAR ij x i MAR i n j 1 i 1 i 1 Nutrient adequacy indicators LIM – Score des composés à limiter The LIM for a 100g o food i: n Na Sat _ fat Adde _ sug 1 1 ij ij ij LIM LA i ij 3 3153 22 50 n j 1 The LIM for a food basket: np p 1 x LIM P i LAR ij x i LIM i n j 1 i 1 i 1 Nutrient adequacy indicators SAIN – Score d’adéquation individuel aux recommandations nutritionnelles The SAIN for a 100g of food i: n ij 1 NAR SAIN i nj i 1ENER The SAIN for a food basket: p n p ij NAR ENER i 1x SAIN P i x i SAIN i n ENER P ENER P j 1 i 1 i 1 Results – Price elasticty of demand Uncompensated own-price elasticities • Statistically significant. • Negatives. • Low and inelastic. • Within usual range. Simulations – Optimal taxation MAR • Goods to tax: Fish, meat, poultry, deli meat, snacks, sugar, animal fat, beverages • Goods to subsidize: Fruits and vegetables, yoghurt, milk, cereals and starches, potatoes, vegetable fat and salty snacks Simulations – Optimal taxation LIM • Goods to tax: Fruits and soft drinks, deli meat, snacks, mixed dishes, dairy products, cereals and starches, vegetable and animal fat, sweets and salty snacks. • Goods to subsidize: Fish, meat, poultry, vegetables, potatoes, water coffee and tea and alcoholic beverages. Simulations – Optimal taxation SAIN • Improvements once calorie intakes are taken into consideration: • Mixed dishes are to be taxed; water to be subsidized. • Meat are more heavily taxed; fruits and vegetables more heavily subsidized. Fiscal incidence Welfare losses homogeneously spread over all income groups. Conclusion Results and policy implications • Theoretical result: A « diet quality/price » ratio and an augmented inverse elasticity rule; • Empirical results: Mixed evidence supporting food taxation: - Low price elasticities and high tax rates; Weak convergence on food groups to tax/ subsidize accross nutrient adequacy indicators. Appendices Methodology – Optimal taxation (2) Use of the Lagragian Method to obtain a system of n+2 linear and non-linear equations and n+2 unknowns. ( 0 ) ( 0 ) ( 0 ) x x x 1 2 0 ) i i k ( k , quali . e t . e t . e x 0 i i , k i i , k i k , i k ( 0 ) ( 0 ) ( 0 ) 2 p 2 p p i i i k k i ( 0 ) x j ( 0 ) quali . t . e obj .% . quali . x 0 j i j , i( j j 0 ) p j i j i ( 0 ) x j ( 0 ) t . x t . t . e R j j j i j , i ( 0 ) p j j i i with Vpc,I I Methodology – Optimal taxation (3) • Using the Lagrangian method: x x 1 2 k , x MAR . e t . e 2 2 p p * k i ( 0 ) i i i , k ( 0 ) k with ( 0 ) i i i , k ( 0 ) k i Vpc,I I • And assuming a differentiable demand function: ( 0 ) x ( 0 ) k * k , x x e k , i . t i k k ( 0 ) p i i Methodology – Optimal taxation (4) • Increasing the MAR objective until the other constraints collapse is equivalent to: • Maximisation Program: Max . MAR MAR i x i p c , I i s.c. tx(p,I) Cost i i i c Methodology – Optimal taxation (5) • Maximisation Program: pc,I Max .V V s.c. MAR x ( p ,I ) Object . ii c i tx(p,I) Cost i i c i (0 ) k , tk 40 %. p k