Cognitive Biases and Behavioural Segmentation in Food Demand

Lead Research Organisation: University of Reading
Department Name: Sch of Agriculture Policy and Dev

Abstract

The study will consist of three parts.

The first part of the study will segment consumers based on their preferences and consumption patterns using economic models of demand to facilitate better targeting of dietary interventions. Previous models of food demand have provided the average response of the population as a whole to changes in economic and demographic circumstances. Our work will extend previous models to recognise that food preferences differ between different individuals and, therefore, the notion of an average response to changing socio-economic circumstances may provide only limited insights into consumer response to dietary interventions. Our modelling approach will enable us to identify sub-groups within the population within which food choices and food preferences are similar. This part of the study will use data from the UK Government's Living Costs and Food Survey which covers a nationally representative sample of nearly 7000 UK households, stratified by Government Office Region and other socio-economic variables.

The second part of the study will attempt to explain these preferences and consumption patterns based on factors influencing behaviours derived from social psychology and behavioural economics. It will explore how susceptible consumers within particular preference and food consumption clusters are to the range of "biases" influencing food choice (e.g., tolerance of risk, conformity to social norms, discount rates, impulsivity etc). This part of the study will use a questionnaire-based survey of 1000 UK households to explore the range of social and cognitive factors that explain the susceptibility to biases influencing food choice. It will provide insights into the design of dietary interventions relevant for different consumer segments. The study will facilitate the development of a suite of interventions that are appropriate for different consumer preferences and behavioural patterns. The study can be used to develop diagnostic or screening tools to identify the most appropriate interventions for different segments of consumers.

The third part of the study will use these insights to develop experimental manipulations aimed at changing the behaviour of some of the key sub-groups based on their susceptibility to behavioural biases and associated food preference patterns. This will help in assessing the potential efficacy of dietary interventions based on behavioural economics insights.

The project will assemble a unique dataset which brings together food purchase behaviour and the cognitive biases and social factors influencing food choice which will be very valuable for future interdisciplinary research on food choice behaviour.

Planned Impact

This research aims to support the promotion of healthier eating strategies. The outputs from this research will be relevant for all key actors in the implementation of healthy eating measures, but will be primarily focused on policy making bodies and agencies responsible for designing and managing dietary interventions and information campaigns. The other potential users of the research include the food industry, consumer organisations (including non-governmental organisations working for the improvement of diet and nutrition or prevention of specific diseases), and the media and food research networks.

For policy makers and agencies responsible for implementing dietary interventions, the impacts will flow from the better design and targeting of dietary interventions made possible by this research. These impacts are expected to be realised through facilitation of:
(1) Development of a novel approach to segmentation of consumers based on their preferences.
(2) Development of screening/diagnostic tools for identification of preferences of individual consumers and their responsiveness to different dietary interventions.
(3) Development of dietary interventions that explicitly take into account consumer preferences and susceptibility to cognitive biases and
(4) Improved targeting of dietary interventions to consumer segments based on the potential responsiveness of different preference segments to different interventions.

The research can contribute to development of dietary interventions and healthy eating messages that are tailored to the attitudes, norms and response characteristics of household groups that make the least healthy dietary choices. Both government agencies like the Department of Health/Food Standards Agency and consumer organisations can use the insights gained from this research to develop campaigns aimed at inducing behavioural change. The identification of household clusters based on distinctive preference patterns and susceptibility to cognitive biases will be a useful input for the food industry in the design of industry level initiatives for promoting healthy eating, dissemination of nutritional information and the development of marketing strategies for nutritionally enhanced or healthier products. The improved design and targeting of dietary interventions can be implemented only over the medium term. As the effectiveness of interventions improves, better adherence to healthy eating norms, improved dietary outcomes and associated gains in health can be expected to be realised over the longer term.

Publications

10 25 50
 
Description Significant new knowledge generated
The first phase of the study was to understand how food consumption is driven by consumers' underlying dietary preferences which differ between households. These differences in dietary preferences are often ignored in the design and evaluation of policy measures and there is a focus on average responses of the population to changes in the economic and social environment. While some measures have used observable characteristics such as income or age to group households, they are not realistic as they assume that households with the same observable characteristics have the same preferences and respond to policies in the same way. In this research, a mixture model was applied to the Living Cost and Food Survey 2013 (n=5,692) to group households according to their similarities in dietary preferences for five types of food: dairy & eggs; meat & fish; fruits & vegetables; and drinks. The analysis (see Figure 1) identified five distinct segments of households. The model ensured that each segment had similar underlying dietary preferences and we were able to show that each segment had a variety of dietary patterns, socio-demographic characteristics and markers of dietary health. This implies that a range of measures targeted at unhealthy eaters are likely to be needed to address the problems of unhealthy eating.
The results show people's discounting behaviour (the extent to which immediate rewards are favoured over larger but delayed rewards) and the degree to which people can override quick 'gut feeling' answers show the most interesting patterns across the segments and in relation to key variables. Discounting shows a positive relationship with the probabilities that individuals will end up in each of the five segments, with a positive relationship with Segments 1 & 2, no relationship with Segment 3, and a negative relationship Segments 4 & 5. Discounting also shows a relationship with the food that people buy, with people who were more prone to discounting less likely to buy unprocessed foods (raw ingredients) and more likely to buy ready meals than those who were less prone to discounting. Performance on the Cognitive Reflection Test had the strongest relationship with BMI independently of a participants' segment, with those better able to suppress quick "gut feeling" answers having lower BMIs. Other biases that showed significant relationships with BMI include sunk costs, discounting and risk framing, with a stronger propensity to show the bias being linked to higher BMIs.
To complete the study, two interventions, tailored around the key biases identified in phase 2, were carried out to see if people's food choices can be differentially influenced.
Study 1 looked at whether cuing the immediate rewards of a food product (i.e. taste and luxury) increase selection of that item, particularly for people who are high temporal discounters. A UK-wide sample of 500 participants took part in a simulated online grocery shopping task, where the same healthy food items were presented in either in terms of their healthiness, how they tasted or how luxurious they were. We aimed to see whether changing the description of the items would influence how often they chosen by participants, and if immediate benefits would be more influential for those high in discounting. Participants also completed measures of discounting, the Cognitive Reflection Test, and a set of demographic questions (incl. BMI).
Study 2 followed on from Study 1 but accounted for the differences uncovered in preference for raw ingredients vs. ready meals. 500 new participants were asked to imagine they were shopping for dinner and asked to select from a range of items in an online store environment. Those items could be either: 1) raw ingredients (which can be combined to create a range of meals); 2) pre-prepped ingredients (i.e. easy cooking); or 3) ready meals (no cooking, just heating). Promotional messaging interventions (again of taste or luxury) were applied to the raw ingredients only to see if they incentivised increased raw ingredient purchasing, especially for high discounters.
The results of both of these studies will be analysed to see whether there is a difference in uptake of the healthier options based on both the manipulation (whether immediate or delayed rewards are emphasised) and on participants' susceptibility to key biases. The intention is that the key findings will be submitted for publication within the next year.
New or improved research methods or skills developed
A novel econometric model was developed which can segment consumers based on their food preferences. It is a mixture distribution of Almost Ideal Demand Systems where the number of components of the mixture distribution gives the number of consumer segments. The model is implemented using a Bayesian methodology and incorporates a Tobit mechanism to deal with censoring.
Important new research resources identified
A survey was conducted over 750 households. The main food purchaser completed a 2 week food diary recording all food purchases. This food diary is compatible with the UK expenditure and food survey. In addition to the diary, the respondent completed a questionnaire designed to elicit their susceptibility to a range of cognitive biases. The resulting data set is unique in recording both food purchases and psychometric variables.
Important new research questions opened up
Whilst we successfully identified segments and a number of cognitive biases in our data, we were unable to establish a particularly strong link between the two. We hypothesise that one reason for this is the aggregation of individuals within the household. This raises important questions about the allocation of food within the household and in particular how the preferences and personalities of the household members interact to determine the overall diet. A research proposal aimed at investigating how aspects of this affect childhood obesity is to be submitted imminently.
Significant negative results or research paths closed off
None
Particularly noteworthy new research networks, collaborations or partnerships, or combinations of these
The project team collaborated with sensory scientists during the project. A PhD student worked in parallel with the Psychology team to analyse how sensory perceptions might be correlated with membership of the 5 segments identified in the econometric analysis.

Increased research capability generated from training delivered in specialist skills
Two post-doctoral researchers were employed on the research project and extended their skill set in survey methods, statistical analysis, project management and presentation.
The interdisciplinary collaboration between psychologists, economists and food scientists at the University of Reading has deepened as a result of the work in this project. This is leading to new research projects and is being further extended to create a significant interdisciplinary capability for analysing diet and health policy at Reading.
Exploitation Route The findings have the following implications for others: For policy makers, for the first time we have demonstrated that groups of consumers can be shown to have differing food preferences and associated cognitive biases. This implies that a more nuanced approach to policy design is required. For researchers: we have identified the fact that a more detailed understanding of how food choices are determined within a household is important.
Sectors Agriculture, Food and Drink

 
Description Advisor Economics Working Group FSA
Geographic Reach National 
Policy Influence Type Participation in a advisory committee
Impact Raising the profile of economics research for the Food Standards Agency's work
 
Description Advisor Food Economy Task Force
Geographic Reach National 
Policy Influence Type Participation in a advisory committee
Impact Help to inform education policies to better meet industry needs
 
Description Advisor Food and You (FSA)
Geographic Reach National 
Policy Influence Type Participation in a advisory committee
Impact Gave insights to inform policy making on food safety in the home
 
Description SARIC outline panel member
Geographic Reach National 
Policy Influence Type Membership of a guideline committee
Impact Identify grants with potential to deliver impact in industry
 
Description Collaboration with David Just at Cornell University 
Organisation Cornell University
Country United States 
Sector Academic/University 
PI Contribution As part of our ESRC funded project, we have support to fund David's travel to the UK. We are discussing ways in which we might take the collaboration further.
Collaborator Contribution David sits on the advisory board for the poject and has given seminars and a lecture in Reading.
Impact No outputs yet
Start Year 2014
 
Description HealthWatch Debate 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact Debate motion: This house believes sugar is harmful so ALL sugary foods should be taxed, not just soft drinks.
Spoke on the evidence against the motion.
Year(s) Of Engagement Activity 2016
 
Description Keynote address at Scottish Graduate Programme in Economics (SGPE) annual conference 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Postgraduate students
Results and Impact Invited to give a keynote address (Cake or Death? Tax or Stealth? How far can economics help?) to postgraduate students registered on the SGPE MSc programmes in economics and SGPE PhD students from 8 Scottish departments in economics.
Year(s) Of Engagement Activity 2016
 
Description Presentation at AAEA Conference (San Francisco) 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Other audiences
Results and Impact Presentation given on 'applying behavioural and experimental economics to food and agri-environmental issues'.
Year(s) Of Engagement Activity 2015
 
Description Presentation at Keele World Affairs 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Public/other audiences
Results and Impact Gave a talk titled 'Challenges in Global Food Security' to an audience of around 300 members of Keele World Affairs. This prompted debate and questions on the subject area.
Year(s) Of Engagement Activity 2016
 
Description Presentation at the FSA on ESRC funded cognitive biases project 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Policymakers/politicians
Results and Impact Requested to present on ESRC-funded study (cognitive biases and behavioural segmentation of food demand) to a general FSA audience.
Year(s) Of Engagement Activity 2015
 
Description Talk at British Library 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? Yes
Geographic Reach Regional
Primary Audience Public/other audiences
Results and Impact Talk stimulated a lively audience discussion.

A number of participants indicated that their views have changed.
Year(s) Of Engagement Activity 2014
URL https://www.youtube.com/watch?v=KgYIiMM5Z_w&feature=youtu.be
 
Description Talk at British Nutrition Foundation Event 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? Yes
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact Talk stimulated discussion and was reported in the BMJ.

A number of contacts were made and continue to e active.
Year(s) Of Engagement Activity 2014
URL http://www.nutrition.org.uk/bnfevents/pastevents/foodtaxes