Quantifying and understanding inconsistency in the generation of consumer food waste is particularly important, to target intervention strategies where they are most efficient. This report analyses two of the largest available datasets to define consumer behavioural typologies and develop a systems map to illustrate potential links between consumer behaviour and the creation/reduction of food waste.

The reduction of food waste has become one of 17 established United Nations Sustainable Development Goals. There is now an international target of halving per capita food waste at the retail and consumer level and reducing food losses along production and supply chains by 2030. Understanding consumer behaviour in relation to food waste generation is therefore critical in developing and targeting cost-effective interventions to achieve food waste reductions.

Quantifying and understanding inconsistency in the generation of consumer food waste is particularly important as identification of consumer waste typologies allows targeting of intervention strategies where they are most efficient.

The current literature identifies large numbers of characteristics which identify typologies, but many of these relationships are likely to be spurious, as a result of multiple testing, small sample sizes, selective reporting and lack of pre-specification of hypotheses limit the proportion of true to untrue relationships under investigation.

We addressed these problems by analysing the two largest available datasets using a range of analytical techniques that minimise type I (false positive) errors. Use of data from WRAP had the further advantage of allowing comparison of self-reported waste with measured waste; whilst Euro-barometer data facilitated generalisation across Europe.

We analysed both data sets using multiple regression models. Rather than specifying a single model, we accepted that different model structures could lead to different results and used model averaging to account for this uncertainty therefore reducing the false-positive error rate. We also utilised machine learning methodologies (random forests and Bayesian Networks) to corroborate the findings.

Analyses consistently indicated that consumers (households) were variable in their food waste behaviour, reinforcing the importance of identifying typologies. They also consistently identified household composition as a key typology or determinant of food waste, with large households generating more waste than small households. Households should therefore be considered as an important unit of analysis in further work, although this does not preclude further exploration of individual actors as well.

Analyses of Euro-barometer consistently identified Country as an important predictor with some indication that countries with grocery spending per capita in excess of €3000 had higher food waste but no apparent relationship with GDP (these relationships will be explored further in future work to ascertain how robust they are). Demographics, specifically age, education and occupation were important predictors in some analyses but not others, and are therefore tentative candidate variables for consumer waste typologies.

In addition to identifying typologies, we were also concerned with generating systems map to allow holistic modelling of waste as an emergent property of a complex system. Identifying commonalities and developing systems models of consumer behaviour is a pre-requisite for developing accurate predictive models to inform policy.

We used machine learnt Bayesian Networks to develop systems maps of the consumer food waste nexus. Different linkages were emphasised to different degrees in models based on different data but one important commonality emerged: Consumer behaviour before shopping, in the retail environment, and in the home predicted food waste. Modelling of consumer behaviour should therefore not be restricted to a single environment. The clustering of behaviours and drivers within each environment requires further unpacking to ascertain which behaviours are key in which environments.

The strength of evidence underpinning the generation of typologies and systems map is currently low. We did not explicitly evaluate the evidence base for publication or selective reporting biases as analyses were based on raw data, but we highlight the pervasive nature of these biases. Further data collection can ameliorate these problems, particularly in relation to precision and inconsistency; but the uncertainty inherent in information on consumer behaviour requires appropriate propagation in probabilistic models, to inform coherent decision-making.

Citation: 

Matthew Grainger, Gavin Stewart, 2016 "Consumers behavioural economic interrelationships and typologies, Project report",  H2020 REFRESH, Newcastle University, Newcastle-Upon-Tyne, UK.

Language: 

  • English

Publishing date: 

18/10/2016

Language: 

Citation: 

Matthew Grainger, Gavin Stewart, 2016 "Consumers behavioural economic interrelationships and typologies, Project report",  H2020 REFRESH, Newcastle University, Newcastle-Upon-Tyne, UK.

ISBN: 

978-94-6257-989-7