Model selection and averaging in the assessment of the drivers of household food waste to reduce the probability of false positives

Abstract

Food waste from households contributes the greatest proportion to total food waste in developed countries. Therefore, food waste reduction requires an understanding of the socio-economic (contextual and behavioural) factors that lead to its generation within the household. Addressing such a complex subject calls for sound methodological approaches that until now have been conditioned by the large number of factors involved in waste generation, by the lack of a recognised definition, and by limited available data. This work contributes to food waste generation literature by using one of the largest available datasets that includes data on the objective amount of avoidable household food waste, along with information on a series of socio-economic factors. In order to address one aspect of the complexity of the problem, machine learning algorithms (random forests and boruta) for variable selection integrated with linear modelling, model selection and averaging are implemented. Model selection addresses model structural uncertainty, which is not routinely considered in assessments of food waste in literature. The main drivers of food waste in the home selected in the most parsimonious models include household size, the presence of fussy eaters, employment status, home ownership status, and the local authority. Results, regardless of which variable set the models are run on, point toward large households as being a key target element for food waste reduction interventions.

Exploring the Drivers behind Self-Reported and Measured Food Wastage

“…This database included statistics regarding objective amount of avoidable household food waste (the terms 'avoidable food waste', 'food waste', and 'food wastage' are being used alternately throughout the paper to describe edible, avoidable food discarded by households) (kg per household; measured through a physical waste survey), sociodemographic factors (gender, household age structure, household size, household composition, home ownership status, type of residence, local authority, presence of children, employment status, and socioeconomic status), and behavioral variables (meal preplanning, pre-shopping cupboard checking, preparing a shopping list, using leftovers, food storage practices, type and frequency of shopping trips, and presence of fussy eaters). Household size, the presence of fussy eaters, employment status, home ownership status, and the local authority were all identified as main drivers of food waste [15]. Grainger et al [15] found no significant effect of any of the behavioral factors on measured amount of food waste.…”

“…Household size, the presence of fussy eaters, employment status, home ownership status, and the local authority were all identified as main drivers of food waste [15]. Grainger et al [15] found no significant effect of any of the behavioral factors on measured amount of food waste.…”

“…A major barrier toward a comprehensive understanding of food waste drivers is the fact that most studies rely on self-reports, which are difficult to verify [15,16]. More specifically, the relationship between different behaviors, attitudes, and food waste quantities has been almost exclusively explored by questionnaires [8], of which validity has been recently questioned [8,[17][18][19][20].…”

“…In the first study, Grainger et al [15] analyzed the Waste and Resources Action Program database [45]. This database included statistics regarding objective amount of avoidable household food waste (the terms 'avoidable food waste', 'food waste', and 'food wastage' are being used alternately throughout the paper to describe edible, avoidable food discarded by households) (kg per household; measured through a physical waste survey), sociodemographic factors (gender, household age structure, household size, household composition, home ownership status, type of residence, local authority, presence of children, employment status, and socioeconomic status), and behavioral variables (meal preplanning, pre-shopping cupboard checking, preparing a shopping list, using leftovers, food storage practices, type and frequency of shopping trips, and presence of fussy eaters).…”

“…To the best of our knowledge, this is the first study to measure and compare drivers of food waste using both self-reports and measurements of actual households' food wastage. As suggested by Grainger et al [15], the use of measured amounts of avoidable food waste as a dependent variable has the potential of reducing the 'social desirability bias'. The comparison between the two measurements provides us a unique opportunity to explore the relations of factors that were suggested as food waste drivers with both self-reported and measured food wastage.…”