(AGENPARL) – ven 07 giugno 2024 Journal of Cleaner Production 452 (2024) 142086
Contents lists available at ScienceDirect
Journal of Cleaner Production
journal homepage: http://www.elsevier.com/locate/jclepro
Household food waste in five territories in Europe and Northern Africa:
Evaluation of differences and similarities as implication for actions.
` Di Veroli a, *, Umberto Scognamiglio a, Irene Baiamonte a,
Benedetta Peronti a, Jacopo Niccolo
Lilliana Stefanovic, Susanne Gjedsted Bügel c, Lea Ellen Matthiessen c, Youssef Aboussaleh d,
´ralska-Walczak e, Laura Rossi a
Chaimae Belfakira d, Dominika Srednicka-Tober
, Rita Go
Council for Agricultural Research and Economics – Research Centre for Food and Nutrition (CREA – Food and Nutrition), Via Ardeatina 546 – 00178 Rome, Italy
Section of Organic Food Quality Faculty Organic Agricultural Sciences University of Kassel, Germany
Department of Nutrition, Exercise, and Sports, University of Copenhagen, Frederiksberg, Denmark
Ibn Tofail University, Faculty of Sciences, Department of Life Sciences, Kenitra, Morocco
Department of Functional and Organic Food, Institute of Human Nutrition Sciences, Warsaw University of Life Sciences, Nowoursynowska 159c, 02-776 Warsaw,
Poland
A R T I C L E I N F O
A B S T R A C T
Handling Editor: Tomas B. Ramos
Reduction of food waste is an important element of the sustainable transformation of food systems. This study
focused on food waste quantification, its causes, and perception in 5 territories: North Hessia (Germany), Cilento
Bio-District (Italy), Kenitra (Morocco), Warsaw (Poland), Copenhagen (Denmark) with the main objective of
assessing whether different cultures affected the levels and the profiles of household food waste. A validated
questionnaire was used to assess the quantities and typologies of food waste (completely unused, partially used,
meal leftovers, leftovers after storing). In addition, the reasons for food waste and how food waste was perceived
were investigated. In a sample of 2154 respondents, the level of still edible food that was wasted amounted to
399 g per family per week, equivalent to 153 g per capita. Kenitra showed the highest amount of FW per
household (539 g), but the lowest amount of food waste per capita (125 g). Citizens of rural communities, e.g.,
Cilento Bio-District (136 g), North Hessia Federal State (132 g), and Kenitra (125 g), had more effective food
waste prevention practices than citizens of urban areas, e.g., Copenhagen (201 g) and Warsaw (179 g). Family
size was identified as a significant factor in FW generation, with households having 5 or more members showing
lower FW per capita (85 g) than single-member families (309 g). The study underscores the need for tailored
strategies to reduce FW considering the above-reported territorial differences.
Keywords:
Food waste
Household
Europe
Northern Africa
Territorial comparisons
SysOrg project
1. Introduction
The agri-food system has both economic and environmental impacts,
consuming resources such as water, soil, and fuels (Mekonnen and
Hoekstra, 2011); each stage of the supply chain determines a loss of
resources (De Corato, 2020). The environmental impact of food pro­
duction is exacerbated by food loss and waste (FLW), which causes more
energy to be consumed for disposal and increased production demand
(Kummu et al., 2012; Scherhaufer et al., 2018). For these reasons, one of
the steps to mitigate the environmental footprint of the food system is to
reduce FLW throughout the supply chain as foreseen by target 12.3 of
the United Nations Sustainable Development Goals (United Nation,
2015).
The Environment Program of the United Nations (UNEP, 2021)
estimated that 17% of the food produced in the world is thrown away; in
2019 this percentage accounted for about 931 million tons of Food
Waste (FW), of which 61% from the domestic sector, 26% from food
services and 13% from retail services. These estimates are indicative of
the magnitude of the FW phenomenon and led to the consideration that
if the global amount of FW would correspond to a country, it would be
the third largest greenhouse gas emitter, after China and the United
States (UNEP, 2021). At the European level, EUROSTAT FW measure­
ments related to 2020 showed that almost 59 million tonnes of fresh
food mass were thrown away, corresponding to 10% of the food prod­
ucts available in Europe (Eurostat, 2022).
According to a study carried out by the European Joint Research
* Corresponding author.
https://doi.org/10.1016/j.jclepro.2024.142086
Received 27 November 2023; Received in revised form 27 February 2024; Accepted 1 April 2024
Available online 10 April 2024
0959-6526/© 2024 Elsevier Ltd. All rights reserved.
B. Peronti et al.
Journal of Cleaner Production 452 (2024) 142086
Centre, the reduction of FW would improve the productivity of the agrifood system, bringing benefits in economic terms. In addition, FW
reduction would have an important and positive impact on the mitiga­
tion of greenhouse gas emissions with an estimated reduction of up to 18
million tons of carbon dioxide equivalent (European Commission. Joint
Research Centre, 2023).
Reduction of FW is an essential part of the transformation of the food
systems that need to be reshaped to have less impact on the environment
(Kennedy et al., 2021). The increased sustainability of food production
and consumption was the inspiration principle of the project “Organic
agro-food systems as models for sustainable food systems in Europe and
Northern Africa” – SysOrg (SysOrg project, 2021). The SysOrg project
focuses on territorial food systems investigating the role of different
dietary approaches and models, reduction of FW, and enhanced organic
food and farming as sustainability elements. The main aim of the project
is to identify the critical points within the food system in five selected
territories to establish possible common interventions to improve the
local food system also with the reduction and prevention of FW. Hence,
the SysOrg project mapped and analyzed FW in five territories: North
Hessia Federal State (Germany), Cilento Bio-District (Italy), Kenitra
Province (Morocco), Warsaw Municipality (Poland), Copenhagen Mu­
nicipality (Denmark) having specific characteristics and different levels
of implementation of FW policies.
Against this background, the main objectives of this study were to
measure the quantity, frequency, and typology of household FW in the
five territories to evaluate differences and similarities. The data
collected had the scope to design recommendations for policy actions
aimed at FW prevention and reduction. Specific purposes of the work
were the evaluation of the effect of sociodemographic characteristics on
household FW, the assessment of consumers’ attitudes toward FW, and
its relationship with the quantity of waste generated by the families.
prioritized (Vereinte Nationen, 2014). Therefore, the theoretical hy­
pothesis underlying this work was that food habits, different local cul­
tures, settings, and the level of implementation of FW policies would
impact the quantity and typology of food thrown away in the house­
holds. A comprehensive assessment of FW in the five territories was
carried out including, besides the quantitative dimension, the potential
and hypothetical causes of FW.
This study intended to answer the following research questions: i)
what is the actual level of waste in the five territories covered by the
SysOrg project? ii) which families’ characteristics impact FW produc­
tion? iii) what are the reasons, motivations, or barriers for waste
reduction and prevention? iv) it is possible to characterize the patterns
of FW in the five territories?
3. Materials and methods
3.1. The survey methodology and the questionnaire
The present study is a cross-sectional assessment conducted by
administrating a questionnaire to adult (>18 years old) residents in the
above-mentioned five territories. The compilation of the questionnaire
was voluntary and anonymous, and the participants were informed
about the objectives of the study and the intention to publish the results.
Data were collected following the European Commission General Data
Protection Regulation (679/2016) and the study was conducted ac­
cording to the guidelines of the Declaration of Helsinki (World Medical
Association, 2018).
The data collection was carried out from January to June 2022. A
convenience sample size of a maximum of 500 completed questionnaires
per territory was fixed. Respondents were recruited using the “river”
sampling methodology (Lehdonvirta et al., 2021), a sampling procedure
that was slightly different in the five territories characterized by cultural
peculiarities and variable geographical extension (SysOrg project,
2021). In North Hessia, Kenitra, Warsaw, and Copenhagen respondents
were recruited via social media channels and questionnaires were
completed online using the Lime Survey© data collection tool. In
Kenitra, in consideration of the low coverage of internet access (Orga­
nisation for Economic Co-operation and Development and United Na­
tions, 2001), the online data collection was coupled with direct
face-to-face interviews of randomly selected people in public places.
The limited geographical extension and density of the population of the
Cilento Bio-District (Cilento Bio-District, 2023) would not permit the
recruitment of respondents via social media. Hence the administration
of the questionnaire was carried out with the assistance of a specialized
research agency, Format Research© S.r.l., Italy. The random selection of
respondents was carried out using the municipalities’ personal data lists
of the Cilento Bio-District residents. The interviews were administered
through the Cati system (Computer Assisted Telephone Interview) or Cawi
system (Computer Assisted Web Interview). A validated questionnaire
(Grant et al., 2023; Scalvedi and Rossi, 2021; Van Herpen et al., 2019)
aimed to quantify still edible food that was wasted and to evaluate the
perception of FW by consumers was used. The final questionnaire and
the modalities of translation into the languages of the five territories
were reported in Table A1 and Fig. A1 (Appendix A).
2. Theory
The cross-territory analysis of FW, as conceptualized in this study,
would offer insights for developing models and inputs aimed at reaching
the 12.3 target of the Sustainable Development Goals aimed at sub­
stantially reducing waste generation through prevention, reduction,
recycling, and reuse (United Nation, 2015). The value of this research is
related to the fact that the analysis of FW status and causes in the five
territories in five different countries, geographically distributed over
Northern, Eastern, Central, and Southern Europe as well as North Africa,
contributes to a broad transnational and multi-actor discussion on the
FW reduction and prevention measures, with the possibility of trans­
ferring the obtained results to other regions as starting or accelerator
points for FW reduction. Critical points (barriers and levers) for FW
prevention and reduction could be identified and proposed on the basis
of the study results. The selected territories (two urban areas and three
rural settings) were mapped as far as concerning the programs to fight
FW present on the ground and in general regarding their approach to
food system sustainability. The analyzed areas have in common initia­
tives to significantly contribute to FW reduction even at different
implementation levels. The two urban territories diverged in terms of
policy actions with the Municipality of Copenhagen having several
ongoing campaigns supporting the reduction of food waste at the public
and private level in the framework of an advanced stage of promotion of
sustainability of food choices (City of Copenhagen, 2024). On the other
hand, Warsaw is at an early stage of FW management mainly focused on
food sharing and donation (Foodsharing Polska, 2024; Foodsi, 2024) in
the framework of a dynamic increase of Warsaw consumers’ ecological
awareness. The three rural areas studied in this research included the
North Hessia Federal State and the Cilento Bio-District, both advanced in
terms of territorial protection and food waste (Pugliese and Antonelli,
2015; Schmidt et al., 2019) and the Kenitra Province in which FW policy
actions are new and still not embedded in the local food system, with
recommendations of low environmental impact of food choices not
3.2. Data analysis
After the data cleaning procedure (Table B1 – Appendix B) that im­
plies the elimination of 2033 units, the final sample consisted of 2154
respondents. Also, the absolute values of income levels in the five ter­
ritories were different; hence the income variable was categorized into 7
levels from the lowest to the highest.
A descriptive analysis was performed using means and frequencies
related to FW.
ANOVA was carried out to investigate per capita FW among the
sociodemographic, attitude constructs, and to assess the difference be
B. Peronti et al.
Journal of Cleaner Production 452 (2024) 142086
tween territories of the most wasted food categories. To understand the
strength of the relationship R2 was calculated according to the following
formula:
(yi ? ?
y i )2
y i ? y)2 ESS
R2 = 1 ?
= 1 ? ?i=1
= ?i=1
y)2 TSS
i=1 (yi ?
i=1 (yi ?
A model-based clustering approach (Seri, 2023) was employed to
identify respondents’ waste profiles. This clustering approach considers
the data as arising from a mixture of distinct probability distributions,
each usually corresponding to a distinct cluster. In the current work, a
mixture of multivariate normal distributions was used. This type of
approach allows for an assessment of uncertainty about the assignment
of units to clusters. Each data point is probabilistically assigned to
different clusters, allowing for fuzzy and non-rigid classification in
waste profiles (Henning et al., 2020; McNicholas, 2016). For this anal­
ysis, carried out with the mclust package of R software (Scrucca et al.,
2023), the most wasted food groups (in quantity and frequencies) were
included as the variables with the highest level of variance. Based on the
BIC (Bayesian Information Criterion) (Schwarz, 1978) and ICL (Inte­
grated Complete-data Likelihood) (Biernacki et al., 2000), a 6-compo­
nent model with EEV (Equal volume, Equal shape, Variable
orientation) parametrization was chosen (Celeux and Govaert, 1995).
Finally, considering the uncertainty of assignment to the clusters, a lo­
gistics model was implemented, using the cluster as the response vari­
able and the territory as the explanatory variable; each unit was
weighted with the probability of belonging to the cluster to which it was
assigned.
The statistical analysis was performed using R Software, version
Where: TSS = Total Sum of Squares; RSS = Residual Sum of Square; ESS =
Explained Sum of Squares.
A linear model was applied to evaluate differences in per capita FW
among sociodemographic groups and attitudes and to analyze variations
of the most wasted food categories between territories. A dichotomic
logistic model was used to study the probability of not wasting among
sociodemographic variables and attitudes. The model was also used to
evaluate the probability of waste in the different waste typologies and
for each of the most wasted food categories. The logistic model was
defined as follows:
pr{?}
logit(pr{?}) = log( Odd(pr{?}) ) = log
= ?o + (?x)
1 ? pr{?}
Where ? = {not waste} = {Y = no}
For each logistic model with factorial explanatory variables (variable
X, sociodemographic variables), the probability of not wasting was
calculated for each category a of each variable x using the following
formula:
Pr{? | x = a} =
4. Results and discussion
Odd(? | x = a)
exp {?0 + ?x=a x}
1 + Odd(? | x = a) 1 + exp {?0 + ?x=a x}
For each logistic model with continuous explanatory variables
(variable X, the attitudes), the odds ratio (OR) was calculated to measure
the association between the attitudes and the probability of not wasting
(Pr{? }):
/1 ? pr{? | X = x + 1} ?
? = exp { ? }
OR = ?
pr{? | X = x}/1 ? pr{? | X = x}
The characteristics of the sample are reported in Table C1 (Appendix
4.1. Household FW: quantities, frequency, and waste typologies
Fig. 1 (Panel A) shows the quantitative evaluation of FW in the five
territories. Considering only the families that wasted, a mean of 399 g of
FW per household per week, equivalent to 153 g per capita per week,
was found. The lowest levels of household waste were found in North
Hessia (274 g/week), while the highest were found in Kenitra (539 g/
week). Family size influenced the total amount of FW, hence normal­
izing the household FW for the number of family members, the Kenitra
Province showed the lowest level of FW (125 g/capita/week). This is
related to the fact that Kenitra had the highest percentage of large
households (60% with five or more members, data not shown) while in
the other territories, this percentage is about 10%. Copenhagen (201 g)
In the above-described model, the OR represents the effect of a unit
increase in the explanatory variable on the probability of not wasting.
Specifically, if the other variables remain constant, a unit increase from
X = x to X = x + 1 results in a change in logit equal to:
?pr{? | X = x + 1}/1 ? pr{? | X = x + 1} ?
? = [ ?o + (?(x + 1)) ? [ ?o + (?x) = ?o ? ?o + ?(x + 1)
logit( pr{? | X = x + 1} ) ? logit( pr{? | X = x} ) = log?
pr{? | X = x}/
1 ? pr{? | X = x}
? (?x) = ?
and Warsaw (180 g) Municipalities resulted in the highest level of waste
expressed as quantity per capita per week. In the present study, the
measurement of food waste was performed using questionnaires
applying the recall method, a self-reported data collection relying on
respondents’ memory. Food waste quantification with the questionnaire
is an indirect measurement that has been reported to underestimate the
amount of food waste (Elimelech et al., 2019). The different methods for
measuring food waste (e.g., diaries, waste composition analysis, ques­
tionnaires, etc.) have advantages and disadvantages, but any of them is
superior to the others (Van Herpen et al., 2016, 2019). For example,
waste composition analysis that provides detailed and accurate insights
into the level of food waste is an impractical approach for use across
An OR = 1 indicates that the attitude does not impact the odds of not
wasting; an OR > 1 indicates direct proportionality between the score
(Likert scale) of the behavior variable and the probability of not wasting;
an OR 0) (Panel A), frequency of
families that wasted (Panel B), and waste typologies (Panel C) in the five territories.
large samples of households (Lebersorger and Schneider, 2011). In this
study, the selection of the methodology for food waste assessment was
carried out by combining the performance efficiency, the acceptance by
respondents, the reliability of the data collected, and, the applicability in
large-scale surveys that were possible with questionnaire and not with
other approaches for costs and practical reasons (Grant et al., 2023).
As reported in Panel B of Fig. 1, approximately 63% of households
reported that they throw away at least one food product belonging to
one of the food groups. The highest frequency of FW was observed in
Copenhagen Municipality (82%) and the lowest was found in Cilento
Bio-District (34%). The households in Cilento Bio-District and Warsaw
had the most similar behavior in terms of waste typologies and Copen­
hagen showed a waste typologies profile similar to the mean of the
whole sample (Panel C – Fig. 1). The information about FW typology is
important to be taken into consideration when implementing/shaping
policies. When food is wasted as partly used or unused food as in Cilento
Bio-District (partly, 38%; unused, 20%) and Warsaw (partly, 38%; un­
used, 22%), preventive actions should focus on the purchase habits and
pantry or fridge organization. On the other hand, when food is mainly
wasted as meal leftovers or stored leftovers as in Kenitra (meal leftover,
46%; stored leftover, 19%) and North Hessia (meal leftovers, 23%;
stored leftover 33%), the policy actions should focus on the kitchen’s
abilities and the capacity to evaluate the food quantities to cook.
The five territories showed differences in terms of typologies of
generated food waste (Table 1), with Kenitra Province characterized by
the higher frequency of generating each waste typology when compared
to other territories. FW as completely unused had the lowest probability
of occurring in the Cilento Bio–District (15%), while FW as partly used
had the highest probability of occurring in Copenhagen and Warsaw
Municipalities (both 52%). The probability of throwing away meal
leftovers was significantly higher in Kenitra (62%) than in other terri­
tories, while waste leftovers after storage had a lower probability of
occurrence in the Cilento Bio-District (9%) than in other territories.
4.2. Household FW and sociodemographic variables: quantities and the
probability of not wasting
The investigation of the relationship between sociodemographic
variables and the quantity of per capita FW as well as the probability of
not wasting permitted to answer the research question related to the
identification of families’ characteristics that impact FW generation. As
reported in Table 2, the number of family members is the characteristic
most strongly linked with the quantity of waste produced in the
household (the family size explained almost 13% of the variability of
waste) while age and level of education were not associated with FW.
Belonging to different territories has a small but significant effect on the
quantity of FW (the variable territory explained more than 2% of the
variance of waste) as household income which explains slightly more
than 1% of the variance of waste.
The linear models (Table 2) confirmed these results with the family
B. Peronti et al.
Journal of Cleaner Production 452 (2024) 142086
larger families (29% for four-member families and 27% for 5 or more
members). Hence, larger households are more likely to waste but waste
less in terms of per capita quantities. This data could be interpreted
considering that larger families while being more efficient in food uti­
lization (Parizeau et al., 2015), were potentially at higher risk of
generating waste in consideration of their large food needs. In addition,
in larger households, there may be a tendency to over-purchase leading
to potential waste when products are unused or expire before they can
be consumed (Babbitt et al., 2021). Territorial belongings significantly
influence the probability of not wasting which was higher in Cilento
Bio-District (66%), followed by North Hessia (41%), and lowest in the
other three territories with the minimum in Copenhagen Municipality
(18%). The probability of not wasting decreased as the income
increased, being 47% and 36% for the two lowest income levels up to
28% and 23% for the highest income levels. According to the available
literature, the effects of income on FW level were not univocal, with
some studies demonstrating that the lowest was the income the highest
was the FW (Stancu et al., 2016), others reporting the opposite (Stefan
et al., 2013; Szab´
o-B´
odi et al., 2018), and others not showing any rela­
tion between FW and income (Koivupuro et al., 2012; Qi and Roe,
2016). As commonly reported (Grant et al., 2023; Grasso et al., 2019),
older people (50% for 55–64 years and 70% for >65 years) have a higher
probability of not wasting than the youngest (about 27 % for 18–34
years and 35–44 years).
Table 1
Analysis of typology of waste among 5 territories. Generalized Linear Models
(Poisson model and dichotomic logistic model); significant p-value 10 resulted in the construct “I usually
leave food in the fridge for too long because I don’t know how to cook it”
(? = 13), “it is difficult to prepare a meal with the food I usually have at
home” (? = 13), and “I rather waste leftovers from meals to avoid
spoilage” (? = 10). Other constructs such as “I do not have enough ca­
pacity in my kitchen to store food leftovers”, “I like to prepare meals of
fresh food instead of leftovers for tasty reasons” and “I avoid storing food
leftovers because it ends up as waste anyway in a while” resulted with a
significative ? lower than 10.
In Table 3 the results of the dichotomous logistic models for the study
of the probability of not wasting (ORs) were reported. The construct
“consequence of waste for future generations” showed a significant result
with an OR less than 1 (OR = 0.92), which means that as the Likert scale
size being the variable with the strongest association with per capita
waste meaning that at the increase of the number of household mem­
bers, a corresponding decrease of per capita waste was observed. This
finding was also reported in other studies that stressed the importance of
the packaging size, often not designed to meet the consumer needs of
smaller households, as one of the reasons for the high FW in these
families (Williams et al., 2020). The models demonstrated a specific
effect of the territories on FW with the polarization of North Hessia,
Cilento Bio-District, and Kenitra wasting about 130 g/capita/week, and
Warsaw and Copenhagen Municipalities wasting approximately 190
g/capita/week. This finding could be explained by considering that
consumers in rural areas had more efficient still edible FW management
than urban areas (Grant et al., 2023; Secondi et al., 2015).
According to Table 2, the probability of not wasting decreases pro­
gressively as the number of household members increases, with smaller
families having a higher probability of not wasting (41% for singlemember families and 47% for two-member families) compared to
B. Peronti et al.
Journal of Cleaner Production 452 (2024) 142086
Table 2
Relationship between the quantity of food waste and sociodemographic variables. ANOVA and Generalized Linear Models (linear model and dichotomic logistic
model); significant p-value