Skip to main content

Empirical evidence on the formation of habits, self-control and non-separabilities in food choices

Periodic Reporting for period 3 - FOODHABITS (Empirical evidence on the formation of habits, self-control and non-separabilities in food choices)

Reporting period: 2019-09-01 to 2020-12-31

Governments around the world regard obesity as a major policy concern. Our research builds the evidence base to help design effective policies that address this issue. Central to our research are two key considerations. First, many policies are targeted at junk food markets, which tend to be dominated by a small number of firms that heavily advertise their products. This means that accounting for how firms will respond to policies is crucial to understand what the overall impact of a policy will be. Second, the costs of poor nutrition vary across people. Policies will also vary in how effective they are at changing people's behaviour; thus the overall effectiveness of a policy will depend on how well it targets those for whom the costs are highest.
Restrictions to advertising of junk foods is one policy option. In Dubois, Griffith and O'Connell (2018) we study the impacts of banning advertising in the UK market for potato chips. We show that the effects of advertising on product level demands vary across products and consumers. Advertising of one brand may steal market share from some rival brands, while boosting demand of others; advertising also acts to change consumers’ willingness to pay for more healthy products. We show that banning advertising, leads to a reduction in the quantity of potato chips sold. However, one effect of advertising is to lower consumer sensitivity to price, reducing the slope of market demands. Therefore, the ban acts to make the market more competitive and firms respond by lowering prices. Lower prices lead to an offsetting increase in demand, meaning that in equilibrium the advertising ban lowers the quantity of potato chips sold by around 10%.

Another popular policy is corrective taxation to reduce excess consumption. In Griffith, O’Connell, Smith (2019) we show how the optimal design of alcohol taxes depends on the correlation of consumers’ product level demands with the social costs that arise from their excess alcohol consumption. We study the empirical importance of this for the UK alcohol market, and show that the government can implement better targeted policy by setting higher tax rates on products disproportionately purchased by heavy drinkers.

In Dubois, Griffith and O'Connell (2020) we study whether taxes on sugary drinks are effective at lowering sugar consumption for those individuals that policy has targeted, and for whom the consequences of high intake are thought to be most severe, namely young people and those who consume a lot of sugar. We study consumption on-the-go – which accounts for around half of sugar from soft drinks – using novel data on British individuals, including teenagers and young adults. Much of the existing literature focuses only on purchases brought into the home, with children and young adults typically not identified as a distinct group of decision makers within household level data. We estimate consumer choice in the non-alcoholic drinks market and simulate the introduction of a soda tax, accounting for how firms’ pass-through the tax to prices. We show that soda taxes do not do a good job at targeting individuals from households with high total dietary sugar, however they are relatively effective at targeting young consumers and those from low income households (see figure attached in image section).

The main justification for policy to reduce obesity is that individuals fail to completely account for costs resulting from their food choices. To design policy well it is important to know the size of these costs and how they vary across individuals and within individual over time. A prominent example of such costs are self-control problems that lead individuals to make choices that they later regret. Our research has provided some of the first estimates of self-control problems using large-scale observational data on food purchases (Cherchye et al, 2017). We find that younger and lower income individuals suffer from larger self-control problems than older and higher income individuals.

One way for government to improve diet among low-income households is to target benefits on the purchase of healthy food. In Griffith, von Hinke and Smith (2018) we study the UK Healthy Start scheme: a large, nationally-implemented scheme that distributes vouchers for specific healthy foods (fruit, vegetables and milk) to low-income households with young children. Standard economic theory predicts that the effect of such vouchers will be greatest for those who would, in the absence of the vouchers, spend less than their value on healthy foods, and will be equivalent to cash for those who would otherwise spend at least the value of the vouchers on healthy foods. We use a discontinuity in eligibility to show that mean monthly expenditure on fresh fruit and vegetable of eligible households increased by 15% compared to pre-reform levels, and the effect of the vouchers is larger than an equivalent-value cash benefit.
In Dubois, Griffith and O'Connell (2020) we specify consumer preferences as individual level parameters that we estimate. This departs from the standard approach to modelling consumer preference heterogeneity in discrete choice models, where preferences are treated as random effects drawn from a known distribution. The main advantage of our approach is that we do not need to make assumptions restricting or ruling out correlation in consumer level preferences with consumer attributes (including purchase behaviour for other goods). We are therefore able to directly relate individual level predictions of the impact of the tax to consumer characteristics in a flexible way. This means that we can assess precisely which individuals respond to the tax and on whom the economic burden of the tax falls most heavily.

In Griffith, Nesheim and O'Connell (2018) we relax the assumption typically imposed in discrete choice demand models that utility is linear in consumption of the outside good.
This imposes independence between the total resources a consumer allocates to a set of separable goods and demand for those goods (i.e. it imposes no “income effects”) and places restrictions on the curvature of individual and market demand, and hence on predictions of pass-through of cost shocks and taxes to consumer prices. Nevertheless, it is commonly believed that for small budget share product categories the restriction that the marginal utility of expenditure is constant is a reasonable approximation. We show that flexibly modelling the impact of total expenditure on demand in a discrete choice demand model can be important, in particular for evaluating the distributional implications of a policy reform.
In Cherchye et al (2017) we use insights from the literature on household economics, and in particular the collective approach to choice behaviour, and apply them to individual self-control problems. We develop an empirically tractable framework that allows us to account for the heterogeneous food choice behaviour of individuals and that provides a structural interpretation of the within-person time-series fluctuations in diet quality. The framework interprets this within person variation in choice behaviour through the lens of multi-selves models influential in the theoretical literature on behavioural economics.
Decile of distribution of share of calories from added sugar