The impact of government advertising on demand for fruit and vegetables Rachel Griffith Lars Nesheim Martin O’Connell Institute for Fiscal Studies and UCL © Institute for Fiscal Studies Summary of paper • Aim is to evaluate the impact of a government advertising campaign which sought to increase fruit and vegetable consumption – Known as the 5 A DAY campaign • We want to know – Did the policy increase fruit and vegetable consumption on average? – Which groups were most effected? • Taking into account the pricing response of imperfectly competitive firms © Institute for Fiscal Studies Contribution • There is a great deal of work assessing past government advertising campaigns (e.g. anti-smoking, drunk driving and drugs) • A recent paper (Capacci and Mazzochi (2011)) evaluates the UK 5 A DAY programme • They estimate impact on demand controlling for potentially confounding factors (i.e. prices) • Show that as fruit and vegetable prices have been rising, controlling for them results in larger consumption response • They treat prices as exogenous, but price may change in response to the government advertising • We specify a structural model that allows us to estimate the impact of government advertising campaign on prices © Institute for Fiscal Studies Impact of advertising on demand and prices • Two predominant views of firm advertising – Informative – A complementary good in the utility function • Impact on price elasticity and equilibrium pricing theoretically ambiguous – e.g. persuasive advertising often said to make demand less elastic, but Becker and Murphy (1993) show that if it raises the willingness to pay of marginal consumers, it can raise the price elasticity • Government advertising may be informative or complementary – Impact on demand curve may lead imperfectly competitive firms to change their price – Which could have a perverse impact for some consumers (e.g. those with demand relatively insensitive to advertising, but sensitive to price) © Institute for Fiscal Studies Modelling fruit and vegetable demand • We treat retailers as setting the final price of fruit and vegetable products • There are four main retailers (and an additional ‘other retailer’ category), each selling 100s of different fruit and vegetable products • Therefore we must aggregate over the products • For now we define 5 products (each available at all retailers) – Fresh fruit – Vegetables – Fruit juice – Tinned fruit – Salad © Institute for Fiscal Studies Demand model • We model consumers as: 1. Deciding upon the number of fruit and vegetable consumption occasions in a given interval of time – This determines the total quantity of all fruit and vegetables demanded 2. For each consumption occasion, choosing a store-product combination, which involves – Deciding on which retailer to shop in – And deciding on which product to buy: © Institute for Fiscal Studies Demand model • We model the choice over consumption occasions as a Poisson process – Consumer i chooses the intensity where to is the expected per consumption occasion utility • We model store-product choice using a mixed logit – Consumer chooses the store-product j that gives the highest utility where are product characteristics and characteristics – and are store are individual specific taste parameters • Model is in the spirit of Dubé (2004) and Hendel (1999), but closer to Burda et al (2011) who show model is consistent with utility maximising behaviour © Institute for Fiscal Studies Purchase data • Use data on food and drink products purchased and brought into the home by 100,000 households over 2003-07, including – Product (barcode) bought – Store shopped in – Quantity and expenditure • Each time a household visits the supermarket, we know how much fruit and vegetables they purchase and exactly what products they buy © Institute for Fiscal Studies Evolution of mean number of portions per person per day • Vertical line denotes beginning of government TV campaign • Target is 5 portions per person per day Deseasonalised portions 2.80 2.75 2.70 2.65 2.60 2.55 2.50 2.45 2.40 2.35 2.30 © Institute for Fiscal Studies Advertising data • Contains details of – All expenditure on TV, radio, and press advertising by firms and government in the food market over 2001-06 • Evolution of total private fruit and vegetable advertising expenditure: Total monthly expenditure (£m) 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 © Institute for Fiscal Studies Impact of advertising • Demand impact – model captures: – Impact of firm ‘fruit and vegetable’ advertising on overall fruit and vegetable demand – Impact of government advertising on overall demand – Impact of store advertising on store choice – Impact of product-specific advertising on product choice • Pricing impact: – The store-product elasticities plus an assumption about the form of retailers’ strategic interactions, allows us to estimate supply side parameters – And therefore help us identify the impact of government advertising on pricing © Institute for Fiscal Studies Evidence of the impact of advertising on total fruit and vegetable consumption • We model portions of fruit and vegetables per household member per day as Poisson process • Control for: – Household characteristics (class, composition, shopping frequency, total annual food expenditure) – Store characteristics – Price of products faced by consumer in the week and store in which they shopped – Time effects • And estimate the impact of government 5 A DAY and firm fruit and vegetable advertising on number of portions – Allowing for intercept and slope effect – And for heterogeneous treatment effect © Institute for Fiscal Studies Heterogeneity is important e.g. intercept effect of government advertising Household type Socioeconomi c status With kids Skilled 0.059 0.012 Semi-skilled 0.078 0.005 Unskilled 0.110 0.009 Skilled 0.027 0.012 Semi-skilled 0.025 0.006 Unskilled 0.040 0.005 Skilled 0.048 0.009 Semi-skilled 0.038 0.004 Unskilled 0.059 0.007 Pensioner Other © Institute for Fiscal Studies Coefficient Standard error Government advertising associated with a 4% increase in demand CDF of increase in daily per capita portions due to government advertising © Institute for Fiscal Studies Increased private advertising associated with a further 2% increase in demand CDF of increase in daily per capita portions due to higher level of private advertising © Institute for Fiscal Studies Evidence of the impact of advertising on storeproduct choice • We estimate a mixed logit – Random coefficients on store dummies means model nests nested logit where consumer chooses store and then product • Control for: – Store characteristics – Product fixed effects – Seasonality – Weekly store-product specific prices • Allow store level advertising to affect store choice and product level advertising to effect product choice – Letting each type of advertising effect intercept and slope of demand curve © Institute for Fiscal Studies Advertising rotates the demand curve Mean Variable Coefficient Standard deviation Standard error Coefficient Standard error Store advertising -0.061 0.015 0.002 0.0005 Product advertising -0.274 0.048 0.021 0.014 Price -0.344 0.025 0.278 0.082 Store advertising*Price 0.037 0.008 - - Product advertising*Price 0.212 0.030 - - © Institute for Fiscal Studies Illustration: the impact of one particular private advertising campaign on demand • In Summer 2005 the retailer Sainsbury’s rebranded its fruit and vegetable range • Accompanying rebranding was a large advertising campaign 1.8 Total monthly expenditure (£m) 1.6 1.4 1.2 1.0 Fruit&veg advertising 0.8 Other advertising 0.6 0.4 0.2 © Institute for Fiscal Studies Oct-06 Jul-06 Apr-06 Jan-06 Oct-05 Jul-05 Apr-05 Jan-05 Oct-04 Jul-04 Apr-04 Jan-04 Oct-03 Jul-03 Apr-03 Jan-03 0.0 Impact on demand for Sainsbury’s vegetables in the absence of advertising campaign 2.5 Price (£) 2.0 1.5 True demand Demand in absence of advertising 1.0 0.5 0.0 4.8% 5.0% 5.2% 5.4% Market share © Institute for Fiscal Studies 5.6% 5.8% Summary • Aim is evaluate a government advertising campaign accounting for the pricing response of oligopolistic firms – Potentially important as any price increases generated as a result of the campaign may diminish its effectiveness – And may lead the campaign to have perverse effects for some groups of consumers • We are currently in the process of estimating the structural demand model, which we can use to estimate how prices and demand responded to the campaign • Although we control for private advertising, we currently treat it as exogenous • In future work we would like to endogenise this © Institute for Fiscal Studies