Appetite 165 (2021) 105283

Contents lists available at ScienceDirect

Appetite
journal homepage: www.elsevier.com/locate/appet

Explaining inequalities in fruit and vegetable intake in Europe: The role of

capabilities, opportunities and motivations
Daniela Craveiro a, b, *, Sibila Marques a, Iva Zvěřinová c, Vojtěch Máca c, Milan Ščasný c,
Aline Chiabai d, Cristina Suarez e, Pablo Martinez-Juarez e, f, Silvestre García de Jalón d, g,
Sonia Quiroga e, Timothy Taylor h
a
Instituto Universitário de Lisboa (ISCTE-IUL), CIS-IUL, Lisbon, Portugal
b
Lisbon School of Economics and Management, CSG/SOCIUS-ISEG, Universidade de Lisboa, Lisbon, Portugal
c
Environment Centre, Charles University, Prague, Czech Republic
d
Basque Centre for Climate Change, BC3, Leioa, Spain
e
Department of Economics, Universidad de Alcalá, Alcalá, Spain
f
Department of Communication and Media Studies, University Carlos III of Madrid, Madrid, Spain
g
Department of Agricultural Economics, Statistics and Business Management, Universidad Politécnica de Madrid, Madrid, Spain
h
European Centre for Environment and Human Health, University of Exeter Medical School, Exeter, United Kingdom

A R T I C L E I N F O A B S T R A C T

Keywords: People who do not eat enough fruit and vegetables (F&V) have incremental health risks. Most Europeans do not
Fruits and vegetables comply with health recommendations relating to F&V consumption and this is especially true for those with
Inequality lower-level education, which reinforces structural inequalities in health and wellbeing among Europeans. This
Education
study investigated the role of key behavioural triggers – capabilities, opportunities and motivation (in the COM-B
Health
Diet
model) – as pathways for educational differentials in F&V intake in Europe. A cross-sectional survey-based study
was conducted in five European countries differing widely in their consumption habits, wealth, and climatic
conditions. A structural equation model was designed to study how capabilities (diet perceived knowledge,
health purchase criteria), opportunities (financial availability, social norms), and motivations (health value,
habits strength) affect educational inequalities in the intake of F&V (5 portions a day) as mediators. Multi-group
comparisons assessed country differences. People with higher levels of education were more likely to eat the
recommended diet, i.e., at least 5 portions of F&V a day. Countries in the sample vary significantly in the per­
centage of people complying with the recommendation, but not significantly in terms of relative education
differentials. The educational gap in the intake of F&V is mainly explained by education differentials in financial
availability, diet knowledge, and habits in inserting F&V in main meals. Policies targeting dietary inequalities
should address behavioural triggers affecting dietary intake, for example by subsidising F&V, developing tar­
geted dietary awareness campaigns, or by intervening in mass catering contexts to facilitate the implementation
of healthy habits.

1. Introduction F&V, when complemented with a diet low in fat, salt and sugar, may also
contribute to a reduction in the risk of obesity, which is an additional
The consumption of sufficient quantities of fruit and vegetables risk factor for NCDs.
(F&V) is an important component of a healthy diet, recognised by ex­ The World Health Organization (WHO) attributed 3.9 million deaths
perts, health professionals and citizens across Europe (e.g., Ridder, worldwide to insufficient F&V intake in 2017 (WHO, 2019). Despite this
Kroese, Evers, Adriaanse, & Gillebaart, 2017). The connection between evidence, people in countries all over Europe do not on average eat the
improved F&V intake and a number of health outcomes is well known, recommended amount of F&V that would ensure a health protection
such as the reduction of non-communicable diseases (NCDs) including effect. According to the European Health Interview Survey (EHIS;
cardiovascular diseases and cancer (WHO, 2019). The consumption of 2013–2015), only 14.1% of European adults consume 5 portions of F&V

* Corresponding author.
E-mail address: daniela.craveiro@iscte-iul.pt (D. Craveiro).

https://doi.org/10.1016/j.appet.2021.105283
Received 5 March 2020; Received in revised form 20 April 2021; Accepted 22 April 2021
Available online 12 May 2021
0195-6663/© 2021 Elsevier Ltd. All rights reserved.
D. Craveiro et al. Appetite 165 (2021) 105283

per day (400 g assuming 80 g per portion), as recommended by the WHO consumption of F&V vary greatly across EU-28 countries. Relying on
and FAO (2005), which forms the basis of most national guidelines in EU data from the European Health Interview Survey (EHIS; 2014–2015), we
Member States (EUROSTAT, 2018). The percentage of the actual con­ can observe that among the countries with higher consumption profile –
sumption varies across countries and varies consistently across educa­ United Kingdom, Ireland, Denmark, Netherlands and Portugal – 20–30%
tional and socioeconomic groups. of the population eat at least five portions a day. This figure is less than
Several reviews of the literature have reported the inverse associa­ 8% among countries with the lowest consumption, such as Romania,
tion between socioeconomic status and fruit and vegetable intake. Less Bulgaria, Croatia, Greece, Austria, or Slovenia (EUROSTAT, 2018).
educated, lower income or lower occupational status individuals tend to Surprisingly, the countries with higher percentages of people reporting
have a lower consumption of F&V and less healthy diets (e.g., Darmon & eating at least 5 portions of F&V a day are those with the largest
Drewnowski, 2008; Giskes, Avendaňo, Brug, & Kunst, 2010; Kamphuis educational gaps (EUROSTAT, 2018).
et al., 2006; Ridder et al., 2017; Roos, Johansson, Kasmel, Klumbiené, & Cultural patterns and the availability of F&V at the national level are
Prättälä, 2001). The trend is reported in almost all EU-28 countries, with some of the possible explanations for this complex pattern (e.g. Hall
only a few exceptions (EUROSTAT, 2018). et al., 2009; Naska et al., 2000). Cross-country differences in dietary
Behavioural disparities in health such as dietary choices underline inequalities suggest different pathways between socioeconomic status
different exposures and vulnerabilities to health risks between different and food consumption.
social groups contributing to these health inequalities. This can be Many interventions to promote F&V intake have been proposed, but
framed under the Theory of Fundamental Causes, according to which if the mechanisms by which socioeconomic factors influence dietary
social conditions are the main determinants of health-relevant actions in patterns are not understood there is a risk that actions may accentuate
keeping with the access to resources that increase the scope for engaging these differences and reinforce the disadvantage of the most economi­
in health-enhancing or health-protective behaviours (Phelan, Link, & cally vulnerable groups (McGill et al., 2015; Rekhy & McConchie, 2014).
Tehranifar, 2010). We identified from the literature some of these re­ It is therefore important to explore the pathways and understand the
sources by consulting previous studies on explanatory variables for in­ factors that account for socioeconomic differences in dietary behaviours
equalities in F&V intake. in order to be able to reduce those inequalities or minimize the degree to
A lack of access to F&V, be it in terms of financial or physical bar­ which they influence health.
riers, has been advanced as an explanation for inequalities in diets. Even Taking into consideration the multiplicity of factors discussed in the
though healthy diets can be achieved on a low budget in some settings literature, approaches require broad models on human behaviour for a
(e.g. Tharrey, Perignon, Dubois, Gaigi, & Darmon, 2019), research comprehensive view on how these dimensions may operate. The COM-B
findings based on normalized international data suggest that, on model (Capability, Opportunity, Motivation and Behaviour) is particu­
average, following a healthier diet pattern with higher intakes of F&V larly tailored for such an exercise as a comprehensive and coherent
(such as the Mediterranean-type diet) is more expensive than opting for understanding of behavioural determinants (Michie, Van Stralen, &
a less healthy diet pattern (Rao, Afshin, Singh, & Mozaffarian, 2013). West, 2011). This model affirms how behaviour (B) requires the inter­
There is also evidence that residents from lower-income neighbourhoods action of three distinct factors: capability (C), the physical and psy­
have less access to healthier options and higher exposure to unhealthy chological capacities to engage in behaviours; opportunity (O), that
food outlets (e.g. Andreyeva, Long, & Brownell, 2010; Black, Moon, & covers the possibilities prevailing in the physical and social context; and
Baird, 2014; Giskes et al., 2010). Perceptions of availability and motivation (M), or the internal processes that trigger behaviours (Michie
affordability (e.g. Ball et al., 2006), food expenditure (Pechey & Mon­ et al., 2011).
sivais, 2016) and diet costs (Aggarwal, Monsivais, Cook, & Drewnowski, Previous studies that address indirect effects of socioeconomic var­
2011) have been identified as relevant mediators for socioeconomic iables on the intake of F&V tend to focus on specific dimensions, instead
differences in fruit and vegetable consumption. of exploring the interplay of multiple pathways. They test mediations
A range of other factors may also influence dietary inequalities. For using coefficient comparisons, considered to be a weak analytical
example, in a multilevel study with a sample of Australian women, strategy for establishing mediation (Claassen, Klein, Bratanova, Claes, &
socio-economic differences in F&V intake were shown to be partially Corneille, 2019). Finally, these studies do not assess country level
mediated by perceived social support and nutritional knowledge (Ball, specificities nor do they employ a common framework that would allow
Crawford, & Mishra, 2006). These factors have also explained educa­ for mapping the diverse relevant theoretical pathways.
tional differences in diet quality among first-time mothers in the same In this paper, we attempt to address all these gaps. This paper ex­
country (McLeod, Campbell, & Hesketh, 2011). Attention to health amines the associations between education and F&V intake, drawing on
when buying food, and organic food consumption (i.e. attitudes about a survey implemented in five European countries. We aim to explore the
healthy eating) partially explained educational differences in fruit and multiple drivers of inequalities, drawing on the COM-B framework to
vegetable intake in a regional subsample of an epidemiological study in explore the factors that mediate the impact of education on intake using
France (Lê et al., 2013). The indirect effect of education on F&V con­ path analysis, probing the statistical significance of each indirect path.
sumption was also found to be mediated through control beliefs (self-­ We also explore the potential that the mediation effects differ between
efficacy and health locus of control), in a sample of 45-year-old women countries.
residing in the city of Bergen, Norway (Leganger & Kraft, 2003). Other Relying on recent literature reviews on determinants of F&V intake
psychosocial factors, such as perceived health status, perceived life and healthy eating, we select the key attributes involved in capability,
control and social cohesion, were found to be contributing to education opportunity and motivation determinants for diet change, and intake of
differences in vegetable consumption in a sample of residents of the city F&V in particular. Concerning the capability to eat healthily or to eat
of Utrecht in the Netherlands (Mulder, De Bruin, Schreurs, Van Ameij­ enough F&V, research highlights the role of cognitive factors, such as
den, & Van Woerkum, 2011). knowledge and beliefs, and the volitional factors related to self-
Research has reported important differences across countries and regulatory skills (e.g., Godinho, Carvalho, & Lima, 2014). To address
regions in Europe (Boylan et al., 2010; Hall, Moore, Harper, & Lynch, this dimension, we include indicators related to health food purchase
2009; Naska et al., 2000; Pechey et al., 2013; Roos et al., 2001). Roos criteria and self-assessments of diet healthiness.
and collaborators concluded, based on a meta-analysis, that those in In terms of opportunities, material and social environments emerge
southern European countries tend to eat more F&V than those in as important dimensions to behavioural changes, shaping the opportu­
northern and central European countries and that there are lower so­ nities for healthier diets (Bowen, Barrington, & Beresford, 2015; Giskes
cioeconomic disparities in consumption levels (Roos et al., 2001). et al., 2010; Story, Kaphingst, Robinson-O’Brien, & Glanz, 2008). Ma­
More recent data, however, present a less coherent pattern. The daily terial environments concern the multiple settings where people eat or

2
D. Craveiro et al. Appetite 165 (2021) 105283

acquire food; financial and physical accessibility to healthy food rank as Committee from Rsu (Riga, Latvia, 1/21.12.2017), University of Exeter
key features in these micro-contexts (e.g., Black et al., 2014; Giskes Medical School (Exeter, United Kingdom, RG/CB/CA249); Instituto
et al., 2010). Social environments relate to the interactions among Universitário de Lisboa (Lisbon, Portugal, 24/2017); University of
people that influence behaviour through social factors and social cues, Alcalá (Alcalá, Spain, CEI/HU/2018/02); Charles University Environ­
such as social norms (Ajzen, 2011). To address the opportunities ment Center (Czech Republic, Prague, 4/2017). The dataset analysed in
dimension, we include indicators related to financial availability (ma­ this paper includes 7582 valid responses. Table 1 presents an overview
terial opportunity) and social norms about healthy eating (social of the sample in the different countries.
opportunity).
Regarding motivation (M), the authors refer to the importance of
reflexive motivation processes and automatic motivation processes 2.2. Measures
(Michie et al., 2011). Reflexive motivation involves what people
consciously want, decide and/or plan to do. This dimension has been The INHERIT household survey was developed to examine attitudes,
explored extensively in the literature on diet change within the scope of preferences and behaviours related to consuming, moving and living
the Theory of Planned Behaviour, anchored in the notion of intentions (Zvěřinová et al., 2018). A set of variables was selected to test the
(e.g., Canova and Manganelli, 2016; Kothe et al., 2012). In a broader relevance of capabilities, opportunities and motivations to explain dif­
way, reflexive motivation regarding eating healthy can also be ferences in F&V intake across individuals with differing levels of edu­
addressed through peoples’ personal values, that is, on the end states or cation. The indicators were selected based on criteria of theoretical
behaviours perceived as good and desirable (e.g. Verplanken & Holland, soundness and empirical validation (internal consistency of scales,
2002; Verplanken & Roy, 2016). As regards automatic motivation pro­ regression assumption studies).
cesses, research has demonstrated the empirical value of habits for
predicting eating behaviours (e.g., van’t Riet, Sijtsema, Dagevos, & de 2.2.1. Capabilities
Bruijn, 2011). This duality of motivational triggers for eating healthily is People are able to increase their fruit intake or opt for healthier diets
included in our study with the model considering the relative impor­ if they have the skills and knowledge to do so (Rimal, Moon, Balasu­
tance of health as a personal value (reflexive) and the strength of habits bramanian, & Miljkovic, 2011). In our survey, two indicators are
(automatic) of eating F&V at main meals. considered to address these dimensions: (i) concern to own health as an
Our main research hypothesis is that capabilities, opportunities and important criteria to buy food (health purchase criteria); and (ii) the
motivations to eat healthy food partially explain educational differen­ self-perceived healthiness of the personal diet (perceived diet knowledge),
tials in F&V intake. assessed on a 7 points Likert scale (“How healthy or unhealthy do you
The socioeconomic position encapsulates the relative position of the think your current food consumption is on a scale from 1, very un­
individual on a continuum of variables that describe the key structural healthy, to 7, very healthy?“).
domains of social stratification, such as education, income, occupation,
and wealth (Krieger, Williams, & Moss, 1997; Lahema, 2010). In this 2.2.2. Opportunities
study, we focus on different socioeconomic segments based on different Material and social environments shape opportunities to follow a
levels of educational attainment. This choice stems from two key rea­ healthy diet with enough fruits and vegetables (Bowen et al., 2015;
sons: relevance and convenience. Education is one of the stronger pre­ Giskes et al., 2010; Story et al., 2008). Generally, more educated people
dictors of health behaviours, such as diet or exercise (e.g., Brunello, Fort, tend to have the financial resources to live and work in settings with
Schneeweis, & Winter-Ebmer, 2016; Park, Cho, & Moore, 2018). Addi­ better access to healthier options (e.g., Black et al., 2014; Giskes et al.,
tionally, education level has clearly been identified as one of the most 2010). Therefore, in our model we use financial availability to measure
relevant mechanisms in the social stratification of health inequality in material environments, in terms of household income.1
Europe, determining occupational class, employment prospects, income, The description of social environment applies a social norms measure.
and wealth (Phelan et al., 2010). The use of education as a proxy for The survey included a three-item scale following the standard wording
socioeconomic position is also convenient given the relative ease of data on normative beliefs recommended by Ajzen (1985, pp. 11–39). The
collection and comparison across different international countries due to scale is designed to capture contextual understandings on what people
established protocols (e.g., the International Standard Classification of believe they are expected to do.2 We consider this as a measure of social
Education). norms – as opposed to descriptive norms, thus, as how people believe
The contribution of COM-B factors to education related inequalities others behave (Ham, Jeger, & Ivkovic, 2015). As preliminary analyses
in diets are examined in a unique data sample gathered by our own (in terms of item correlation and internal consistency) did not validate
survey carried out in five European countries that differ significantly in the combination of the three items into a single score (Supplementary
current lifestyles and diets as well as their respective educational material, Table A3), we selected the indicator with the highest face
endowments. validity (“People who are important to me would disapprove/approve of
my eating of a healthy diet most of the time”).
2. Material and methods
2.2.3. Motivation
The motivation to eat is influenced by reflexive and automatic
2.1. Data
motivational processes (e.g. Wood and Rünger, 2016).
For the reflexive motivational trigger, we applied the role of personal
Data come from a survey conducted in five countries – the Czech
values. It is plausible to assume that people who value their personal
Republic, Latvia, Portugal, Spain, and in the United Kingdom –
compiling different political and socio-economic contexts and ensuring a
broad European perspective on different subjects (Zvěřinová, Ščasný, & 1
Midpoints from 12 income brackets shown in the questionnaire divided by
Máca, 2018). This survey was conducted in 2019 as part of the Horizon
the square root of the size of the household, converted to Purchasing Power
2020 funded INHERIT project (inherit.eu).
Standards and then square rooted (with this last mathematical calculation
The country subsamples were selected using quota sampling from performed to account for the non-linear association between education and
online access panels and are representative of national populations aged financial availability).
18–65 years in terms of gender, age, region, and education (EUROSTAT, 2
In the Theory of Planned Behaviour, “subjective norms” are measured by the
2017). All participants gave their informed consent. The study obtained strength of “normative beliefs” weighted by the strength of individual moti­
ethics approval in each country where data were collected: ethical vation to comply with them (Ajzen, 1985).

3
D. Craveiro et al. Appetite 165 (2021) 105283

Table 1
Sample description by country.
CZ ES LV PT UK Total

N % N % N % N % N % N %

Age group
18–34 years old 587 31.5 504 31.6 328 34.3 495 40.6 591 30.4 2505 33.0
35–49 years old 709 38.0 654 41.0 352 36.8 455 37.3 659 33.9 2829 37.3
50–65 years old 570 30.5 437 27.4 276 28.9 270 22.1 695 35.7 2248 29.6
Gender
Female 947 50.8 805 50.5 515 53.9 593 48.6 1073 55.2 3933 51.9
Male 919 49.2 790 49.5 439 45.9 626 51.3 870 44.7 3644 48.1
Other 0 0.0 0 0.0 2 0.2 1 0.1 2 0.1 5 0.1
Marital status
No partner 584 31.3 571 35.8 318 33.3 495 40.6 836 43.0 2804 37.0
With partner 1282 68.7 1024 64.2 638 66.7 725 59.4 1109 57.0 4778 63.0
Education attainment
Primary/lower secondary 799 42.8 387 24.3 94 9.8 349 28.6 474 24.4 2103 27.7
Secondary education 692 37.1 539 33.8 498 52.1 479 39.3 685 35.2 2893 38.2
Tertiary 375 20.1 669 41.9 364 38.1 392 32.1 786 40.4 2586 34.1
Total 1866 100.0 1595 100.0 956 100.0 1220 100.0 1945 100.0 7582 100.0

Notes. Czech Republic (CZ), Spain (ES), Latvia (LV), Portugal (PT), United Kingdom (UK). Frequency (N) and percentage (%) by sample.

health tend to have a greater intention to opt for healthier diet options,
such as eating F&V in main meals. An indicator of Schwartz scale (e.g.,
Schwartz, 1992) concerning the importance attributed to health as a
personal value was included in the model (health value, evaluated on a 8-
point Likert scale: from − 1, opposed to my values, to 7, of supreme
importance).3 To minimize the interference of differences in rating
styles the value score was centred by the individuals’ average score on
the full scale.
The motivation to eat F&V is also influenced by automatic triggers,
such as habits. Habits are generated by contextualized-learned behav­
iours; when a habit is established, situational cues are able to trigger the
behaviour, without any conscious decision to do so (e.g. Gardner, 2015).
The strength of the habit of eating F&V is measured with an adapted
short version of the Self-Report Habit Index (Gardner, Abraham, Lally, &
de Bruijn, 2012), considering the consumption of F&V and in main Fig. 1. Representation of indirect and indirect effects of education on fruit and
meals on weekdays. The habit strength score results from the average of vegetable intake.
six items, assessed on an agreement scale to three sentences concerning
the lunch and dinner situation (“Eating fruit or vegetables at lunch/­ F&V were estimated using the conversion key made available by Cleg­
dinner time on weekdays is something that …. I do without thinking; … horn et al. (2016). After summing the variables, responses were coded
is natural for me to do; … I do automatically”). The scale demonstrated a into a dichotomous variable showing compliance or not with the stan­
good level of internal consistency (Cronbach’s Alpha = 0.914; Supple­ dard of 5 portions or more a day (0 = less than 5 portions a day; 1 = 5 or
mentary material, Table A2). more portions a day).

2.2.4. Education
2.3. Analyses
The educational inequalities in F&V intake are explored in this
article to contribute to the understanding of pathways to decrease health
A structural equation model was designed to study the contribution
inequalities in Europe. Based on the International Standard Classifica­
of capabilities (diet perceived knowledge, health purchase criteria),
tion of Education (ISCE), country specific variables for highest level of
opportunities (financial availability, social norms), and motivations
educational attainment were categorized in three levels: up to low sec­
(health value, habits strength) to educational inequalities in the intake
ondary education (primary/low secondary education), secondary edu­
F&V (5 portions a day) as mediators. These hypotheses are theoretically
cation, and tertiary education.
grounded (COM-B model) and were specified before data treatment.
The mediated effects of education on fruit and vegetable intake
2.2.5. F&V intake
through differences in capabilities, opportunities and motivations (in­
A Self-Reported short-form Food Frequency Questionnaire was
direct paths) are estimated based on the direct effects of educational
included in the survey to collect information on diets (Cleghorn et al.,
variables on each mediator variable (Paths a) and each mediator and the
2016 adapted by Zvěřinová et al., 2018). Respondents were asked to
dependent variables (Paths b), as presented in Fig. 1.
indicate how often they consume F&V separately on a scale of 9 fre­
Regression coefficients and overall fit of the model were estimated
quency categories (see Figure A1 in Appendix). The daily portions of
using the Lavaan package in R (Rosseel et al., 2019), based on robust
estimations. The model included regression paths between education
variables (introduced in the model as two dummy variables) and each
3
Single item usage of the health item receives support from preliminary mediator, and from each mediator variable to F&V intake, with a total of
psychometric assessments of the scale that failed to frame the health item 7 regressions and 12 bivariate correlations, with pair-wise comparisons
clearly in one specific factor (basic human values), being relatively less among the scale level mediators, to take into account all relevant effects.
compliant than other items with the structure of Schwartz’ human values scale All regression paths include a set of controls (gender, age group, marital
(Schwartz, 1992). status, country). Due to the existence of exogenous dichotomous

4
D. Craveiro et al. Appetite 165 (2021) 105283

variables, model parameters were estimated using diagonally weighted Estimated coefficients of the indirect paths are presented in Table 4.
least squares, but the full weight matrix is used to compute robust All estimated indirect effects are considered statistically significant (p <
standard errors and mean- and variance-adjusted test statistics. The .001). Thanks to the standardization process, the coefficients are com­
estimation of the indirect effects was computed based on MacKinnon parable and provide information on which pathways present higher
and Dwyer (1993), adapting the product of coefficients approach to correlations with the consumption of F&V. According to these calcula­
models with dichotomous outcomes. tions, the education-related advantages in F&V intake in the five Euro­
Multigroup comparisons were conducted to assess whether the pean countries are mostly related to financial availability. Higher
regression pathways under study significantly vary across country educated people may have more income available to make healthier diet
samples. The multigroup modelling relies on a version of the model choices than less educated people, making it easier to access and opt for
without the country control variables and it is based on the comparison higher intakes of F&V.
of the overall fit between “free” (all parameters are allowed to differ The second and third strongest indirect effects concerns habits and
between groups) and “constrained” models (regression parameters are diet knowledge, respectively. Those with higher levels of education are
fixed to be equal across groups). more likely to include F&V in main meals as a habit. They also have
Preliminary analysis included the study of bi-variate associations increased understanding of healthier diets.
between educational attainment and the variables in the model Using health as a criteria when purchasing food, valuing health as a
(Table 2), the overall significance of the regressions that compose the personal value and perceiving social support when opting for healthier
model (Table 3), and the study of the key linear regression models based diets also significantly contribute to explaining educational inequalities
on residuals analysis (Supplementary material, Table A3, Figure A2). in diets (p < .05).
A set of multigroup comparisons by country were conducted
3. Results (Table 5). Comparing the overall fit between constrained and free
models, we find a substantial increase in the overall model fit, sug­
The results show that, drawing on the full sample, the association gesting differences across country groups in the regression coefficients
between educational attainment and F&V intake is similar to that in (Δχ2(172) = 187, p < .01). The systematic comparisons between free and
previous studies - higher educated groups are shown to be more likely to partially constrained models suggest that the relationships in different
eat the recommended amount of F&V (χ2(2) = 184.654, p < .01). countries may be considered broadly consistent, though a few factors
Overall, 22% of the pooled sample said that they ate at least 5 portions a may be more or less responsible for preference heterogeneity in con­
day; 17 and 18% Czech and Spanish respondents, respectively, 21% sumption of F&V across the countries.
Latvian and Portuguese respondents, and 32% of British respondents. Country samples differ only in the effect of education variables on
However, the share of respondents who eat the recommended portions the health value. Regression coefficients by country were consulted to
of F&V (or more) vary greatly across education levels. This percentage is understand this result (details in the Supplementary material, Table A5).
lower among the least educated (11–15% and 23% in UK) and higher Concerning the health value score, the correlation between education
among the most educated respondents (Tertiary education, 25–29% and variables and health value scores is found to be positive in the Czech
41% in UK). While countries vary in the percentage of people that say Republic (Secondary education: B = 0.28, SE = 0.06, p < .01; Tertiary
that they eat more than 5 portions of F&V a day (from 17% in Czech education: B = 0.44, SE = 0.08, p < .01), Latvia (Secondary education: B
Republic to 32% in the United Kingdom), the educational differentials = 0.38, SE = 0.14, p < .01; Tertiary education: B = 0.54, SE = 0.14, p <
are consistent across samples (Fig. 2). .01) and Portugal (Secondary education: B = 0.09, SE = 0.07, p > .05;
All indicators selected to describe capabilities, opportunities and Tertiary education: B = 0.17, SE = 0.08, p < .05). The same was not true
motivations to eat healthier increase significantly with the levels of in the UK and Spain, where health concerns appear to be equal across
education, with the exception of the health value score variable (p > educational levels, whilst being more socially stratified (and relevant to
.05) (Table 2). inequality) across Latvia and the Czech Republic.
The path analysis was modelled to test the predicted mediation ef­ Countries do not vary significantly in the regression coefficients
fects between education and F&V intake (details in the Supplementary concerning the direct impact of education on diets. Education in­
material). The overall significance of each regression path was tested equalities in F&V intake do not significantly vary across countries
previously, assessing fitness differences between the estimated models (Δχ2(8) = 6.564, p = . 5843). Countries do not differ either in the in­
and respective intercept-only model. fluence of capability, motivation, and opportunity variables in the
No severe infringements of the regression assumptions were chances of eating 5 portions a day (Table 5.).
observed (Supplementary material, Table A3; Figure A2). All estimated
models provide a better fit than the intercept-only models. 4. Discussion
Fitness measures indicated a good overall fit to the data (CFI =
0.989, TLI = 0.921; RSMEA = 0.042, IC90%[0.031–0.053], p = .882; The results give further evidence of low intake of F&V in Europe. The
SRMR = 0.14). The model with mediation paths provides a better fit majority of people (78%) from five European countries stated that they
than the model with no mediation paths (Δχ 2(12) = 918.12, p < .001).4 ate less than the recommended 5 portions of F&V a day. Additionally,
The estimated coefficients of the direct paths are presented in Table 3. the intake of F&V significantly varies according to education level.
Higher educational attainment levels (secondary and tertiary education Tertiary and secondary educated people are more likely to eat the rec­
in relation to lower secondary) are positively correlated to capabilities ommended 5 portions in the overall sample and in each country sample.
(Diet knowledge, Health purchase criteria), opportunities (Financial These results are similar to those found in the literature (e.g. Darmon &
availability, Social norms), and motivations (Habits strength, Health Drewnowski, 2008; Giskes et al., 2010; Kamphuis et al., 2006; Ridder
value score) to eat healthy food, controlling for country, gender, age et al., 2017; Roos et al., 2001), signalling relevant health risks among the
group and marital status (p < .001). Capabilities, opportunities and general population, and especially among those less educated.
motivation variables are shown to be positively correlated with the By exploring the multiple theoretical pathways that may potentially
likelihood of eating at least 5 portions of F&V a day (p < .001). be involved in the explanation of educational inequalities in diet, it is
possible to highlight key factors related to this behavioural pattern. The
most relevant pathways identified in our study concern the three trig­
4
Model omitting paths a (education to mediation variables) and paths b gers for behaviour change as proposed in the COM-B model (Michie
(mediation variables to dependent variable): CFI = 722, TLI = 0.611, RSMEA = et al., 2011): opportunity (financial availability), motivation (habits)
0.092, IC90%[0.087-0.097], p = .000, SRMR = 0.014). and capability (healthy diet knowledge). This theoretical divide appears

5
D. Craveiro et al. Appetite 165 (2021) 105283

Table 2
Capabilities, opportunities and motivation variables by education level.
Lower secondary Secondary Tertiary Total

N % N % N % N %

Capability Purchase criteria No 1492 70.9 1905 65.8 1572 60.8 4969 65.5
Yes 611 29.1 988 34.2 1014 39.2 2613 34.5
M SD M SD M SD M SD
a
Diet knowledge 4.33 1.24 4.46 a,b 1.17 4.72 a,b
1.15 4.52 1.19
c
Opportunity Financial availability 39.6 10.5 42.3c,d 11.4 47.2 c,d
12.4 43.2 11.9
e
Social norms 5.70 1.36 5.75 f 1.31 5.87 e,f
1.20 5.78 1.29
g
Motivations Habit strength 4.49 1.70 4.61 g,h 1.63 4.82 g,h
1.59 4.65 1.64
Health value score 1.17 1.26 1.20 1.19 1.21 1.16 1.19 1.20

Notes. Frequency (N) and relative percentage (%). Mean (M) and standard deviation (SD). Health purchase criteria and Education (χ2(2) = 53.181, p < .01). Pair-wise
comparisons (Bonferroni correction) a, b, c, d, e, f, g, h: p < .05.

Table 3
Estimated regression coefficients included in the path-model.
Path Independent variable Dependent variable Coefficient Standard
Error

Direct path Secondary education 5 portions a day 0.138*** 0.043

Tertiary education 0.340*** 0.045
Paths a Secondary education Purchase criteria 0.175*** 0.039
Tertiary education 0.279*** 0.040
Secondary education Diet knowledge 0.141*** 0.033
Tertiary education 0.351*** 0.035
Secondary education Financial availability 3.648*** 0.324
Tertiary education 8.039*** 0.326
Secondary education Social norms 0.123*** 0.036
Tertiary education 0.206*** 0.039
Secondary education Habit strength 0.190*** 0.046
Tertiary education 0.357*** 0.048
Secondary education Health value score 0.157*** 0.033
Tertiary education 0.248*** 0.035
Paths b Capability Purchase criteria 5 portions a day 0.104*** 0.022
Diet knowledge 5 portions a day 0.179*** 0.015
Opportunity Financial availability 5 portions a day 0.005*** 0.001
Social norms 5 portions a day 0.042*** 0.013
Motivation Habit strength 5 portions a day 0.189*** 0.010
Health value score 5 portions a day 0.045*** 0.014

Notes. All regressions paths include a set of 8 control variables: gender (male as reference category), age group dummies (50+ years old as reference category), marital
status (no partner as reference category), and country dummies (PT as reference category). ***p < .001.

to be relevant not only for understanding behaviour change patterns but exposure to food insecurity risk. Thus, subsidies for F&V could be more
also understanding behavioural inequalities. helpful for less educated and low-income people. There is a good evi­
The Theory of Fundamental Causes puts forward a key theoretical dence from plenty of other studies that healthy food subsidies are
framework for health inequalities (Phelan et al., 2010). Within this effective in triggering consumption change (for systematic reviews see
scope, the concept of flexible resources is crucial. This idea encapsulates Andreyeva et al., 2010; Thow, Downs, & Jan 2014) and may improve the
how privileged socioeconomic positions correlate with access to multi­
ple resources that facilitate healthier options (Phelan et al., 2010). This
foresees the relative importance of particular resources as depending on
the specific historical and geographical context even though these ad­
vantages persist over time (Phelan et al., 2010). Our approach allows for
the comparison of the relative relevance of such resources (or pathways)
and provides some clues on the priorities for intervention among Eu­
ropean adult populations.
The results suggest that to be more efficient, the design of in­
terventions to decrease diet-related social inequalities and consequent
increased health risks should address the different behavioural triggers.
The educational inequalities in F&V intake in the five European
countries are mostly related to financial situation. As expected, more
highly educated groups have more disposable income than less educated
people, which partially explains a higher intake of F&V. In the health
inequalities literature, the differentials between social groups in terms of
material resources have been identified as relevant pathways for in­
equalities for decades (e.g., Phelan et al., 2010). People with lower ed­ Fig. 2. Percentage of people that declare eating at least 5 portions of fruit and
ucation levels get less qualified jobs, ensuring lower incomes, and vegetables a day by country and education.
therefore restrictions in the food expenditure budget, or a higher Notes. Czech Republic (CZ), Spain (ES), Latvia (LV), Portugal (PT), United
Kingdom (UK).

6
D. Craveiro et al. Appetite 165 (2021) 105283

Table 4
Estimated indirect effects of education on fruits and vegetables intake (5 portions a day) through each mediator in the path-model.
Education variables Mediators Coefficient Sobel(Z)

Secondary education Capability Purchase criteria .004 3.254**

Tertiary education .006 5.049***
Secondary education Diet knowledge .014 3.509***
Tertiary education .033 8.297***
Secondary education Opportunity Financial availability .051 10.920***
Tertiary education .056 22.046 ***
Secondary education Social norms .003 2.175*
Tertiary education .005 3.467***
Secondary education Motivation Habit strength .027 3.850***
Tertiary education .049 4.452***
Secondary education Health value score .004 3.105***
Tertiary education .006 4.705***

Notes. All regressions paths include a set of 8 control variables: gender (male as reference category), age group variables (50+ years old as reference category), marital
status (no partner as reference category), and country dummies (PT as reference category). For education variables Primary/Lower secondary education is the
reference category. *p < .05, **p < .01, ***p < .001. Standardized indirect effects were calculated as a product of coefficients (Path a) and (Path b), made comparable
following MacKinnon and Dwyer (1993).

Table 5
Multigroup comparison (by country sample).
χ2 DF Δχ2 DF (Δχ2) p

Unconstrained model 59.970 15

Constrained models:
All regression coefficients equal by country 692.69 187 333.68*** 172 0.000
Education regression coefficients on intake equal by country 62.856 23 6.5638 8 0.5843
Education regression coefficients on Purchase criteria equal by country 72.728 23 6.0852 8 0.6377
Education regression coefficients on Diet knowledge equal by country 86.900 23 9.9424 8 0.2691
Education regression coefficients on Social norms equal by country 83.185 23 8.9313 8 0.3481
Education regression coefficients on Habit strength equal by country 70.074 23 8.1198 8 0.4219
Education regression coefficients on Health value score equal by country 91.373 23 16.323* 8 0.0380
Education regression coefficients on Financial availability equal by country 84.607 23 13.717 8 0.0895
Purchase criteria regression coefficients on intake equal by country 63.793 19 3.410 4 0.492
Diet knowledge regression coefficients on intake equal by country 62.987 19 3.056 4 0.548
Social norms regression coefficients on intake equal by country 63.063 19 2.902 4 0.574
Habit strength regression coefficients on intake equal by country 65.933 19 5.379 4 0.251
Health value score regression coefficients on intake equal by country 72.112 19 7.497 4 0.112
Financial availability regression coefficients on intake equal by country 61.323 19 2.8507 4 0.583

*p < .05, **p < .01, ***p < .001.

diets of people of lower socioeconomic status (McGill et al., 2015). order to establish the inclusion of F&V as a default choice (Gardner,
Nevertheless, we would stress that our model does not predict linear 2015). This resonates with the “nudge” approach that intervenes in the
correlations between either education and income or between income contextual features surrounding habits in order to steer human behav­
and F&V consumption. In our sample, the increase in income translates iour towards better options (e.g. Thaler & Sunstein, 2008; Evans &
into greater F&V consumption growth all the while the level of income Stanovich, 2013; Vecchio & Cavallo, 2019). Our data suggest that the
increases. The efficacy of healthy food subsidies may be constrained automatic route in the decision-making process in meals contributes to
(counterintuitively) among more deprived people. the disadvantage of people with lower educational resources. Therefore,
Studies on consumer stratification and profiling have signalled such interventions where people make decisions related to diets, may prior­
a complex interplay. Low F&V intake profiles are socially heterogeneous itize targeting lower socioeconomic class neighbourhoods. These can
(e.g., Bertail & Caillavet, 2008; Raaijmakers, Sijtsema, Labrie, & Snoek, include, for example, interventions in school canteens, such as including
2018). In addition, studies have also reported profiles of lower-income soup and fruit in all lunch menus; presenting fruit bowls in waiting
consumers with medium and high F&V consumption (e.g. Raaijmakers rooms, meeting rooms, or lunchrooms in schools, social and municipal
et al., 2018) or with profiles of lower-income consumers who are services; or even visually promote the consumption of F&V where
particularly insensitive to economic incentives (e.g. Bertail & Caillavet, people eat or buy food. Fostering schemes for bundle buy of healthy
2008). Lower financial availability may be more relevant for F&V access foods with vouchers (Vecchio & Cavallo, 2019), or regular subscriptions
in some social groups than others. Most studies (ours included) do not of F&V baskets (Bell et al., 2019), for example, may also ensure that
consider, for example, the role of subsistence farming or family’s people buy enough fresh F&V to have in their household and so rely less
budgeting strategies in the access to and domestic availability of fresh on intentional decision making.
food among lower income groups. This may confound the results while Additionally, interventions should also target the dietary knowledge
also providing clues for intervention (e.g. Tharrey et al., 2019). There­ of people. In our study, the assessment of the healthiness of personal diet
fore, this requires the consideration of multiple strategies to ensure served as an indicator of healthy diet knowledge. Less educated people
positive results across the income spectrum. self-assessed as being less knowledgeable as to how to eat healthily
Habit strength is the second most influential factor that explains which partially explained their lower fruit and vegetable intake. This
educational inequalities in F&V intake. More highly educated people are therefore leads to the recommendation for the tailoring of information
better able to combine the consumption of F&V in a main meal. To campaigns to populations with lower educational resources.
promote better diet habits among the less educated it is important to F&V intake promotion campaigns should make clear the advantages
intervene in the contexts where people consume their main meals in of this behaviour and possible action plans based on realistic scenarios

7
D. Craveiro et al. Appetite 165 (2021) 105283

for the target populations (for example, disclosing strategies on how to Additionally, even though the survey samples are designed to collect
obtain healthy diets on limited budgets; portraying people eating F&V in country representative samples of adult populations, the survey samples
workplace settings linked to lower qualified jobs or on typical did not include people without Internet access with data collection
commuting routes). through web-based questionnaires administered by online access panels.
Promotional campaigns should be long-term, and require (1) inter Although the Internet coverage is high in all these counties (The internet
sectorial collaboration (industry, retail, NGOs, state, city, etc.), (2) ac­ is available in 95% of British, 86% of Czech and Spanish, 82% of Latvian,
tion across the multiple dimensions of behaviour change, (3) clear 79% of Portuguese households EUROSTAT, 2019), the results are to be
messages, (4) interactive approaches; and (5) cultural targeting (Rekhy understood with caution and making generalizations about the national
& McConchie, 2014). The accumulated evidence on campaign efficacy differences should be done only with reservations.
should guide and direct the development of these initiatives (e.g., Kothe,
Mullan, & Butow, 2012; Rekhy & McConchie, 2014; Ungar, Sieverding, 5. Conclusion
& Stadnitski, 2013).
Educational differences in F&V intakes emerge as very consistent The study examined the educational inequalities in F&V intake in
across the countries selected in this study. Despite the national vari­ order to contribute to the understanding of pathways to decrease health
ability in the percentages of people that eat at least five portions of F&V inequalities in Europe. Building on the COM-B framework (Michie et al.,
a day, the educational gap in diets are of similar magnitude across 2011), the association between education and F&V intake is shown to be
samples. These findings do not align with the consulted literature (e.g. related to capability, motivation, and opportunity differentials. The
EUROSTAT, 2018), possibly due to methodological differences in the most educated people have an increased likelihood of consuming at least
assessment of F&V intakes and the social gradient in such intakes. 5 portions of F&V a day. The educational gap is mainly explained by
Overall, our data collected in 2019 suggest a higher consumption of factors such as financial availability, diet knowledge and habits in
F&V than the 2014 EHIS. In this EHIS, information on F&V intake stems inserting F&V into main meals.
from the frequency of consumption, and correspondingly asking re­ Public health policies need to pay special attention to support less
spondents only to specify the number of portions when a person educated people to increase their F&V intake. This can be achieved by
declared eating more than once a day. The data sets are therefore not targeting key triggers for behaviour change, by combining subsidies for
directly comparable. Additionally, in our study, education differentials F&V, targeting awareness campaigns and nudge approaches to enable
are assessed while maintaining gender, age groups, and marital status as less educated people to change their diets.
constants. These dimensions influence dietary intake and are added to
the regression models as controls to return a better understanding of Ethical statement
educational inequalities independently of other important factors. One
part of the national differences in education differentials may arise from The study resorts to survey data collected by online panel sampling.
compositional differences in the populations. Future research should All participants gave their informed consent.
address this hypothesis with multilevel models in order to identify and The study obtained ethics approval in each country where data was
explain country effects on diets. collected: ethical Committee from Rsu (Riga, Latvia, 1/21.12.2017),
As in any such case, there are limitations in the approaches deployed University of Exeter Medical School (Exeter, United Kingdom, RG/CB/
in this paper. First, we apply a self-reported short form of food frequency CA249); Instituto Universitário de Lisboa (Lisbon, Portugal, 24/2017);
questionnaire (FFQ) to estimate F&V intake. The short form of FFQ we University of Alcalá (Alcalá, Spain, CEI/HU/2018/02); Charles Uni­
adapted was validated against an extensive FFQ for the UK (Cleghorn versity Environment Center (Czech Republic, Prague, 4/2017).
et al., 2016). The diet quality of participants is therefore roughly
assessed and not based on energy or nutrition intake. Research suggests Author contributions
that the estimate of energy and nutrient intake based on self-reported
FFQ experiences bias from multiple factors (Watanabe et al., 2019) – Conceptualization: D.C. & S.M.; Data curation: All authors under I.Z.,
including education, our key variable. Future research should assess V.M., & M. Š. supervision; Formal analysis: D.C. & S.M; Methodology: D.
whether key explanatory dimensions contributing to education differ­ C. & S.M; Supervision: S.M.; Writing - original draft: DC; Writing - re­
entials in F&V intakes differ when diets are assessed with other methods. view & editing: D.C., S.M., I.Z., V.M., M. Š., A.C., C⋅S., P.M.J., S.G.J., S.Q.
The operationalization of the dimensions of capability, opportunities and T.T.
and motivation to eat healthily was also constrained by data availability.
The study relied on analysis of a multi-thematic survey that did not Acknowledgements
include detailed measures for all the concepts involved even though the
questionnaire was heavily reliant on indicators tested by previous This work was supported by the European Union’s Horizon 2020
research and subject to empirical validation by a pilot study. Most research and innovation, by funding of the INHERIT project (www.
measures relied on single items, challenging content validity (poten­ inherit.eu), coordinated by EuroHealthNet (Grant Agreement No.
tially not capturing the construct) and reliability assessments (no in­ 667364). The funding sponsors had no role in the design of the study; in
ternal consistency measures) (e.g., Fisher, Matthews, & Gibbons, 2016). the collection, analyses, or interpretation of data; in the writing of the
This is particularly challenging in the measurement of abstract con­ manuscript, and in the decision to publish the results. All authors had
structs that potentially hold multiple meanings (e.g. Fisher et al., 2016) full access to the data reported in the manuscript. Declarations of in­
– for example, our measure of health score value relies on a single item terest: none.
even though we may assume health means different things to different
people (e.g., Raaijmakers et al., 2018) or our measure of social norms
Appendix A. Supplementary data
that strives to capture global beliefs about the people that are most
relevant to them (e.g. Ajzen, 2011; Ham et al., 2015).
Supplementary data to this article can be found online at https://doi.
The broadness of our scope may compromise the depth of the anal­
org/10.1016/j.appet.2021.105283.
ysis but nevertheless it can be used to inform future research by high­
lighting the main dimensions contributing to the educational
References
differentials and is worthy of further exploration. For example, this
approach may support future comparative analysis of the different Aggarwal, A., Monsivais, P., Cook, A. J., & Drewnowski, A. (2011). Does diet cost
behaviour models that the COM-B model aims to compile. mediate the relation between socioeconomic position and diet quality? European

8
D. Craveiro et al. Appetite 165 (2021) 105283

Journal of Clinical Nutrition, 65(9), 1059–1066. https://doi.org/10.1038/ consumption among adults: A systematic review. British Journal of Nutrition, 96,
ejcn.2011.72 620–635. https://doi.org/10.1079/bjn20061896, 2006.
Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In J. Kuhl, & Kothe, E. J., Mullan, B., & Butow, P. (2012). Promoting fruit and vegetable consumption.
J. Beckman (Eds.), Action-control: Springer series in social psychology. Berlin, Testing an intervention based on the theory of planned behaviour. Appetite, 58(3),
Heidelberg: Springer. https://doi.org/10.1007/978-3-642-69746-3_2. 997–1004. https://doi.org/10.1016/j.appet.2012.02.012
Ajzen, I. (2011). The theory of planned behaviour: Reactions and reflections. Psychology Krieger, N., Williams, D. R., & Moss, N. E. (1997). Measuring social class in US public
and Health, 26(9), 1113–1127. https://doi.org/10.1080/08870446.2011.613995 health research: Concepts, methodologies, and guidelines. Annual Review of Public
Andreyeva, T., Long, M. W., & Brownell, K. D. (2010). The impact of food prices on Health, 18(1), 341–378. https://doi.org/10.1177/1359105315576606
consumption: A systematic review of research on the price elasticity of demand for Lahema, E. (2010). In W. C. Cockerham (Ed.), The new blackwell Companion to medical
food. American Journal of Public Health, 100(2), 216–222. sociology (71-96)Health and social stratification. West Sussex, UK: Blackwell
Ball, K., Crawford, D., & Mishra, G. (2006). Socio-economic inequalities in women’s fruit Publishing Ltd.
and vegetable intakes: A multilevel study of individual, social and environmental Lê, J., Dallongeville, J., Wagner, A., Arveiler, D., Haas, B., Cottel, D., et al. (2013).
mediators. Public Health Nutrition, 9(5), 623–630. https://doi.org/10.1079/ Attitudes toward healthy eating: A mediator of the educational level-diet
PHN2005897 relationship. European Journal of Clinical Nutrition, 67(8), 808–814. https://doi.org/
Bell, R., Khan, M., Romeo-Velilla, M., Stegeman, I., Godfrey, A., Taylor, T., et al. (2019). 10.1038/ejcn.2013.110
Ten lessons for good practice for the INHERIT triple win: Health, equity, and Leganger, A., & Kraft, P. (2003). Control constructs: Do they mediate the relation
environmental sustainability. International Journal of Environmental Research and between educational attainment and health behaviour? Journal of Health Psychology,
Public Health, 16(22), 4546. https://doi.org/10.3390/ijerph16224546 8(3), 361–372. https://doi.org/10.1177/13591053030083006
Bertail, P., & Caillavet, F. (2008). Fruit and vegetable consumption patterns: A MacKinnon, D. P., & Dwyer, J. H. (1993). Estimating mediated effects in prevention
segmentation approach. American Journal of Agricultural Economics, 90(3), 827–842. studies. Evaluation Review, 17(2), 144–158. https://doi.org/10.1177/
https://doi.org/10.1111/j.1467-8276.2008.01138.x 0193841X9301700202
Black, C., Moon, G., & Baird, J. (2014). Dietary inequalities: What is the evidence for the McGill, R., Anwar, E., Orton, L., Lloyd-Williams, F., O’Flaherty, M., Taylor-Robinson, D.,
effect of the neighbourhood food environment? Health & Place, 27, 229–242. et al. (2015). Are interventions to promote healthy eating equally effective for all?
https://doi.org/10.1016/j.healthplace.2013.09.015 Systematic review of socioeconomic inequalities in impact. BMC Public Health, 15
Bowen, D. J., Barrington, W. E., & Beresford, S. A. A. (2015). Identifying the effects of (457), 1–15. https://doi.org/10.1186/s12889-015-1781-7
environmental and policy change interventions on healthy eating. Annual Review of McLeod, E. R., Campbell, K. J., & Hesketh, K. D. (2011). Nutrition knowledge: A mediator
Public Health, 18(36), 289–306. https://doi.org/10.1146/annurev-publhealth- between socioeconomic position and diet quality in Australian first-time mothers.
032013-182516 Journal of the Academy of Nutrition and Dietetics, 111(5), 696–704. https://doi.org/
Boylan, S., Lallukka, T., Lahelma, E., Pikhart, H., Malyutina, S., Pajak, A., et al. (2010). 10.1016/j.jada.2011.02.011
Socio-economic circumstances and food habits in Eastern, Central and Western Michie, S., Van Stralen, M. M., & West, R. (2011). The behaviour change wheel: A new
European populations. Public Health Nutrition, 14(4), 1–10. https://doi.org/10.1017/ method for characterising and designing behaviour change interventions.
S1368980010002570 Implementation Science, 6(1), 42. https://doi.org/10.1186/1748-5908-6-42
Brunello, G., Fort, M., Schneeweis, N., & Winter-Ebmer, R. (2016). The causal effect of Mulder, B. C., De Bruin, M., Schreurs, H., Van Ameijden, E. J., & Van Woerkum, C. M.
education on health: What is the role of health behaviors? Health Economics (Vol. 25, (2011). Stressors and resources mediate the association of socioeconomic position
(3), 314–336. https://doi.org/10.1002/hec.3141 with health behaviours. BMC Public Health, 11(798), 1–10. https://doi.org/10.1186/
Canova, L., & Manganelli, A. (2016). Fruit and vegetable consumption as snacks among 1471-2458-11-798
young people. The role of descriptive norm and habit in the Theory of Planned Naska, A., Vasdeki, V.G. S., Trichopoulou, A., Friel, S., LeonhaÈuser, I.U., Moreiras, O., et
Behavior. TPM: Testing, Psychometr., Methodol., 23(1), 83–97. https://doi.org/ al Fruit and vegetable availability among ten European countries: How does it
10.4473/TPM23.1.6 compare with the ‘five-a-day’ recommendation? British Journal of Nutrition, 84(4),
Claassen, M. A., Klein, O., Bratanova, B., Claes, N., & Corneille, O. (2019). A systematic 549-556.
review of psychosocial explanations for the relationship between socioeconomic Park, C. L., Cho, D., & Moore, P. J. (2018). How does education lead to healthier
status and body mass index. Appetite, 132(June), 208–221. https://doi.org/10.1016/ behaviours? Testing the mediational roles of perceived control, health literacy and
j.appet.2018.07.017, 2018. social support. Psychology and Health, 33(11), 1416–1429. https://doi.org/10.1080/
Cleghorn, C. L., Harrison, R. A., Ransley, J. K., Wilkinson, S., Thomas, J., & Cade, J. E. 08870446.2018.1510932
(2016). Can a dietary quality score derived from a short-form FFQ assess dietary Pechey, R., Jebb, S. A., Kelly, M. P., Almiron-Roig, E., Conde, S., Nakamura, R., et al.
quality in UK adult population surveys? Public Health Nutrition, 19(16), 2915–2923, (2013). Socioeconomic differences in purchases of more vs. less healthy foods and
0.1017/S1368980016001099. beverages: Analysis of over 25,000 British households in 2010. Social Science &
Darmon, N., & Drewnowski, A. (2008). Perspective Does social class predict diet quality ? Medicine, 92(100), 22–26. https://doi.org/10.1016/j.socscimed.2013.05.012
American Journal of Clinical Nutrition, 87, 1101–1117. https://doi.org/10.1093/ajcn/ Pechey, R., & Monsivais, P. (2016). Socioeconomic inequalities in the healthiness of food
87.5.1107 choices: Exploring the contributions of food expenditures. Preventive Medicine, 88,
Eurostat. (2017). Population on 1 January by age group and sex. Available at: https://ec. 203–209. https://doi.org/10.1016/j.ypmed.2016.04.012
europa.eu/eurostat/web/population-demography-migration-projections/data/data Phelan, J. C., Link, B. G., & Tehranifar, P. (2010). Social conditions as fundamental
base. causes of health inequalities: Theory, evidence, and policy implications. Journal of
Eurostat. (2018). Fruit and vegetable consumption statistics. Available at: https://ec.eur Health and Social Behavior, 51(1_suppl), S28–S40. https://doi.org/10.1177/
opa.eu/eurostat/statistics-explained/index.php/Fruit_and_vegetable_consumption_st 0022146510383498
atistics. Raaijmakers, I., Sijtsema, S., Labrie, C., & Snoek, H. (2018). Consumer segmentation
Eurostat. (2019). Level of internet access - households [TIN00134]. Available at: https based on health-related motive orientations and fruit and vegetable consumption.
://ec.europa.eu/eurostat/databrowser/view/tin00134/default/table?lang=en. British Food Journal, 120(8), 1749–1763. https://doi.org/10.1108/BFJ-02-2018-
Evans, J. S. B., & Stanovich, K. E. (2013). Dual-process theories of higher cognition: 0098
Advancing the debate. Perspectives on Psychological Science, 8(3), 223–241. https:// Rao, M., Afshin, A., Singh, G., & Mozaffarian, D. (2013). Do healthier foods and diet
doi.org/10.1177/1745691612460685 patterns cost more than less healthy options? A systematic review and meta-analysis.
Fisher, G. G., Matthews, R. A., & Gibbons, A. M. (2016). Developing and investigating the BMJ Open, 3(12), 1–16. https://doi.org/10.1136/bmjopen-2013-004277
use of single-item measures in organizational research. Journal of Occupational Health Rekhy, R., & McConchie, R. (2014). Promoting consumption of fruit and vegetables for
Psychology, 21(1), 3–23. https://doi.org/10.1037/a0039139 better health. Have campaigns delivered on the goals. Appetite, 79, 113–123. https://
Gardner, B. (2015). A review and analysis of the use of “habit” in understanding, doi.org/10.1016/j.appet.2014.04.012
predicting and influencing health-related behaviour. Health Psychology Review, 9(3), Ridder, D., Kroese, F., Evers, C., Adriaanse, M., & Gillebaart, M. (2017). Healthy diet:
277–295. https://doi.org/10.1080/17437199.2013.876238 Health impact, prevalence, correlates, and interventions. Psychology and Health, 32
Gardner, B., Abraham, C., Lally, P., & de Bruijn, G. J. (2012). Towards parsimony in habit (8), 907–941.
measurement: Testing the convergent and predictive validity of an automaticity van’t Riet, J., Sijtsema, S. J., Dagevos, H., & de Bruijn, G. J. (2011). The importance of
subscale of the Self-Report Habit Index. International Journal of Behavioral Nutrition habits in eating behaviour. An overview and recommendations for future research.
and Physical Activity, 9(102), 1–12. https://doi.org/10.1186/1479-5868-9-102 Appetite, 57(3), 585–596. https://doi.org/10.1016/j.appet.2011.07.010
Giskes, K., Avendaňo, M., Brug, J., & Kunst, A. E. (2010). A systematic review of studies Rimal, A., Moon, W., Balasubramanian, S. K., & Miljkovic, D. (2011). Self-efficacy as a
on socioeconomic inequalities in dietary intakes associated with weight gain and mediator of the relationship between dietary knowledge and behavior. Journal of
overweight/obesity conducted among European adults. Obesity Reviews, 11(6), Food Distribution Research, 42(3), 1–14.
413–429. https://doi.org/10.1111/j.1467-789X.2009.00658.x Roos, G., Johansson, L., Kasmel, A., Klumbiené, J., & Prättälä, R. (2001). Disparities in
Godinho, C. A., Carvalho, J., & Lima, M. L. (2014). Promoting healthy eating: A brief vegetable and fruit consumption: European cases from the north to the south. Public
review of research on predictors and interventions. Transcultural, 4(1), 39–53. Health Nutrition, 4(1), 35–43. https://doi.org/10.1079/PHN200048
Hall, J. N., Moore, S., Harper, S. B., & Lynch, J. W. (2009). Global variability in fruit and Rosseel, Y., Jorgensen, T. D., Oberski, D., Byrnes, J., Vanbrabant, L., Savalei, V., et al.
vegetable consumption. American Journal of Preventive Medicine, 36(5), 402–409. (2019). Lavaan: Latent variable analysis. Available at: https://cran.r-project.org/we
https://doi.org/10.1016/j.amepre.2009.01.029. e5. b/packages/lavaan/index.html.
Ham, M., Jeger, M., & Ivković, A. F. (2015). The role of subjective norms in forming the Story, M., Kaphingst, K. M., Robinson-O’Brien, R., & Glanz, K. (2008). Creating healthy
intention to purchase green food. Economic Research-Ekonomska Istrazivanja, 28(1), food and eating environments: Policy and environmental approaches. Annual Review
738–748. https://doi.org/10.1080/1331677X.2015.1083875 of Public Health, 29, 253–272. https://doi.org/10.1146/annurev.
Kamphuis, C. B. M., Giskes, K., de Bruijn, G. J., Wendel-Vos, W., Brug, J., & van publhealth.29.020907.090926
Lenthe, F. J. (2006). Environmental determinants of fruit and vegetable Thaler, R. H., & Sunstein, C. R. (2008). Nudge: Improving decisions about health, wealth,
and happiness. Yale University Press.

9
D. Craveiro et al. Appetite 165 (2021) 105283

Tharrey, M., Perignon, M., Dubois, C., Gaigi, H., & Darmon, N. (2019). Opticourses: Environmental Psychology, 45, 127–134. https://doi.org/10.1016/j.
Promoting healthier food purchases at no additional cost in low-income households. jenvp.2015.11.008
The European Journal of Public Health, 29(Supplement_4). https://doi.org/10.1093/ Watanabe, D., Nanri, H., Sagayama, H., Yoshida, T., Itoi, A., Yamaguchi, M., et al.
eurpub/ckz186.464, 2019. (2019). Estimation of energy intake by a food frequency questionnaire: Calibration
Thow, A. M., Downs, S., & Jan, S. (2014). A systematic review of the effectiveness of food and validation with the doubly labeled water method in Japanese older people.
taxes and subsidies to improve diets: Understanding the recent evidence. Nutrition Nutrients, 11(7), 1–15. https://doi.org/10.3390/nu11071546, 1546.
Reviews, 72(9), 551–565. https://doi.org/10.1111/nure.12123 Wood, W., & Rünger, D. (2016). Psychology of habit. Annual Review of Psychology, 67,
Ungar, N., Sieverding, M., & Stadnitski, T. (2013). Increasing fruit and vegetable intake. 289–314. https://doi.org/10.1146/annurev-psych-122414-033417
“Five a day” versus “just one more. Appetite, 65, 200–204. https://doi.org/10.1016/ World Health Organization (WHO). (2019). Increasing fruit and vegetable consumption to
j.appet.2013.02.007 reduce the risk of noncommunicable diseases. e-Library of Evidence for Nutrition
Vecchio, R., & Cavallo, C. (2019). Increasing healthy food choices through nudges: A Actions (eLENA). Available at: https://www.who.int/elena/titles/fruit_vegetables_
systematic review. Food Quality and Preference, 78(103714), 1–12. https://doi.org/ ncds/en/.
10.1016/j.foodqual.2019.05.014 World Health Organization (WHO) and Food and Agriculture Organization of the United
Verplanken, B., & Holland, R. W. (2002). Motivated decision making: Effects of Nations (FAO). (2005). Fruit and vegetables for health. Report of a Joint FAO/WHO
activation and self-centrality of values on choices and behavior. Journal of Personality Workshop, 1-3 September, 2004, Kobe, Japan. Available at: http://www.fao.
and Social Psychology, 82(3), 434–447. org/3/a-y5861e.pdf.
Verplanken, B., & Roy, D. (2016). Empowering interventions to promote sustainable Zvěřinová, I., Ščasný, M., & Máca, V. (2018). Barriers and potential for adopting healthier,
lifestyles: Testing the habit discontinuity hypothesis in a field experiment. Journal of more equitable and environmentally friendly solutions identified in a five-country survey.
Available online: . (Accessed 1 January 2019).

10