Exploring everyday life dynamics in meat reduction - A cluster analysis of flexitarians in Denmark

Exploring everyday life dynamics in meat reduction – a cluster analysis of flexitarians in Denmark. Flexitarians are attracting increasing attention in the research on meat reduction. But there has been limited focus on comprehensive understandings of a broader range of dynamics that can work as barriers and facilitators for meat reduction. In this article, we use social practice theory (SPT) as a comprehensive approach to barriers and facilitators in meat reduction in everyday life. We present an analysis of data from a representative Danish cross-sectional survey. We show, first, that Danish flexitarians can be divided into four distinct clusters (what we will refer to as classes) in accordance with combinations of everyday facilitators and barriers. Second, we show that the prevalence of these classes varies considerably depending on how long people have been flexitarians. We argue that the patterns in this variation indicate that over time people transition to other classes where barriers to plant-rich eating become less significant, and routinization emerges in different ways. Finally, third, we show that flexitarians do report eating less meat than consumers who label themselves as eating meat with no restrictions. But we also highlight that the difference is relatively modest. Indeed, meat intake is still quite common even in classes where routinization is highest. Throughout the paper, we discuss similarities and differences between the SPT framework and another recent framework, the COM-B model, that also provides a comprehensive approach to the understanding of behavioural change.


Introduction
Meat consumption, and especially the consumption of beef, is known to be a major contributor to the high levels of greenhouse gas emissions from food production (Gerber et al., 2013).For this reason, meat reduction and increased consumption of plant-based foods are considered to be changes that allow us to adopt a more climate-friendly diet (IPCC, August 2019).
In recent years, the question of how we are to encourage ordinary consumers to change their dietary patterns to incorporate more plantbased ways of cooking and eating has attracted considerable attention, across the globe, from policy makers, public bodies dealing with food and the environment, and food market actors (de Boer & Aiking, 2017; Fuentes & Fuentes, 2017;Gerbers-Leenes, 2017;Lang & Heasman, 2015, p. 2;Soule & Sekhon, 2019;Whitley et al., 2018).
It remains far from clear, however, how such dietary transition can be implemented in large population segments.Although a large number of studies on meat reduction focus on single factors such as information (e.g.Vainio, 2019), attitudes (e.g.Slade, 2018), psychological barriers (e.g.Collier et al., 2021, p. 167) and price reactions (e.g.Malek et al., 2019), there is not much consensus on the combinations of single factors into more comprehensive understandings of facilitators and barriers to meat reduction.Vegetarian and vegan (as well as pescatarian) diets clearly result in the lowest greenhouse gas emissions (Willett et al., 2019).But vegetarians and vegans so far seem to make up a relatively small share of consumers (Lund et al., 2021;Paslakis et al., 2020), and few meat eaters appear ready to abandon meat altogether (Hielkema & Lund, 2021;Malek and Umberger, 2021).Thus, in order to shed light upon facilitators and barriers to meat reduction in large population segments, it makes less sense to take a starting point in the otherwise large number of studies on vegetarians and vegans (e.g.Markowski & Roxburgh, 2019;Rosenfeld & Burrow, 2017;Twine, 2018).
So, growing attention is now being given to consumers who are willing to take a step in the desired direction and eat less meat.Indeed, globally a significant proportion (though not a majority) of consumers are currently planning to reduce, or are reducing, their meat intake (Dagevos, 2021).In line with Dagevos (2021), we shall refer to these consumers as flexitarians, although it should be noted that the way flexitarians are defined and measured varies considerably in the studies published to date (see Dagevos, 2021).Probably, the reason many consumers are ready to become flexitarians is that reducing meat intake is easier than adopting a vegetarian and vegan diet, since meat can still be included in dishes and meals.On the other hand, the non-binding character of the flexitarian diet leaves it an open question how much flexitarians are reducing their meat intake in practice (see Dagevos & Voordouw, 2013;Verain et al., 2015).In a recent literature overview, Dagevos (2021) demonstrated that most flexitarians are not breaking significantly with their meat-eating behaviour.In light of the many meat substitute products currently entering the market, and the innovations in cuisine in which meat is becoming less central, it might be expected that recent studies of flexitarians would identify more striking breaks in meat consumption.However, this is not the case: the data collected in 2019 and later suggest that most flexitarians still eat many meat-based meals (Malek and Umberger, 2021;Hielkema & Lund, 2021).Further, it has been found that in the Netherlands average meat reduction did not change dramatically, among flexitarians, in 2019 as compared with 2011, and in fact it increased among the most engaged flexitarians (Verain et al., 2022.On the other hand, the aforementioned studies show, after all, that flexitarians are eating less meat than regular meat eaters (Malek and Umberger, 2021;Hielkema & Lund, 2021;Verain et al., 2022.Thus, we do need to understand more fully what works as barriers and facilitators in meat reduction. In this article, we contribute to the recent literature on flexitarians by exploring how everyday life dynamics support and hamper Danish flexitarians' progress towards reduced meat consumption.The data we draw on are from a questionnaire-based survey of a representative sample of Danish consumers.The situation in Denmark is especially interesting because the country has a high level of meat consumption per capita relative to the global level and in comparison with many other European countries (Ritchie & Roser, 2017).Furthermore, Denmark is a large producer and exporter of animal products, in particular pork, and in Denmark, the EU funding for advancing sales of agricultural products is dominantly used for advancing animal products (Klimarådet, 2021).At the same time, Danish consumers are front-runners regarding other environmentally friendly kinds of food consumption, such as consumption of organic foodstuff, where Denmark is second-highest in EU regarding per capita spending (www.statista, 2023).
First, we place our approach and study in relation to the existing research literature on meat reduction, outline shortly social practice theories (SPT) and the COM-B model, and explain why our approach is inspired by SPT, and which everyday barriers and facilitators are included in our empirical study.Second, materials and methods involved in our survey study of Danish food consumers and the barriers and facilitators for meat reduction are described.Third, we outline the results of the study.Finally, we provide a conclusion and discussion.

Barriers and facilitators in meat reduction
The fast growing research literature on barriers and facilitators in meat reduction tends to be dominated by studies where a limited set of factors are taken into consideration.These factors focus on intention, and different kinds of motives such as attitudes, values, psychological barriers, information, reactions to prices, and meat attachment.This tendency to focus on single factors or a combination of single factors can be seen across the whole literature on barriers and facilitators in meat reduction, i.e. in studies investigating meat reducing consumers more generally (e.g.Becker & Lawrence, 2021, p. 164;Cheah et al., 2020;Cliceri et al., 2018;Çoker and van der Linden, 2020;Graca et al., 2015;Kwasny et al., 2022, p. 168;Lentz et al., 2018;Malek et al., 2019;Sanchez-Sabate et al., 2019;Slade, 2018;Vainio, 2019;Wolstenholme et al., 2021, p. 166;Zur & Klöckner, 2014), and studies where flexitarians are examined (Collier et al., 2021, p. 167;Lacroix & Gifford, 2019;Malek & Umberger, 2021;Randers et al., 2022;Randers & Thøgersen, 2023, p. 180;Verain & Dagevos, 2022;Verain et al., 2022).All of the studies just laid out clearly offer important research contributions.However, they do not necessarily attempt to address broader dynamics in everyday life.

Comprehensive understandings of barriers and facilitators in meat reduction
Rather than only studying single factors or combinations of single factors such as attitudinal and other cognitive factors, we also need perspectives that cover a broader range of everyday life activities and relations that may impact consumers' ability to embark upon and persist with a changed behaviour, meat reduction.Two conceptual frameworks, recently highlighted as ways of studying transition to plant-based diets, provide such more comprehensive research perspectives.The two approaches are social practice theory and the COM-B model, respectively.
Emanating from sociological theory, social practice theory (SPT) focuses on the social character of meal provisioning and eating, and on the everyday life dynamics that shape food consumption such as routines, competences and social conventions (Halkier & Holm, 2021).Social practice theory assumes that eating is a composite practice embedded in everyday life (Warde, 2005;, 2016), and that changes in dietary patterns happen gradually and in a piecemeal fashion (e.g.Delormier et al., 2009).For this reason, its advocates argue, it is the everyday practices of routines and competencies regarding cooking and procurement, food cravings and taste priorities, and how these practices change over time, that need to be studied (Plessz & Étile, 2019).Studies are now appearing that draw upon the social practice framework to understand flexitarians' behavioureither directly (Daly, 2020;Khara et al., 2021;Mylan, 2018;White et al., 2022, p. 175), or indirectly by pointing to everyday conditions and dynamics that hinder or promote the adoption of a plant-based diet (North et al., 2021, p. 167;Pohjolainen et al., 2015).
Others have suggested that the COM-B model is an approach which enables a more comprehensive understanding of barriers and facilitators of meat reduction (Graca et al., 2019b).The COM-B model conceptualises the following overarching components in behavioural change: capabilities, opportunities, and motivations (Michie et al., 2011).Hence, the barriers and facilitators related to motives, attitudes and intentions, that are important in a majority of the existing single factor studies on meat reduction, are included in the COM-B model, but supplemented with other broader dynamics in everyday life food consumption.As Graca and colleagues point out, the consumer's understanding of difficulties and opportunities related to capabilities and opportunities in everyday life has so far received limited attention (2019b).Fortunately, studies are now emerging that also give insights into these latter two components of flexitarian practice (Berg et al., 2022;Daly, 2020;Grassian, 2020;Kemper, 2020, p. 150;Kemper and White, 2021;Mylan 2018;Sijtsema et al., 2021).
There are commonalities between SPT and the COM-B model, (White et al., 2022, p. 175), for example that both approaches include a broad range of everyday life aspects, such as routines and habitual behaviour.There are also specific overlaps in categories, such as 'competences' from practice theories that mirrors parts of 'capability' from COM-B.We discuss this further in the next subsection ("Everyday barriers and facilitators included in this study").However, there are important differences.Often, practical research using the COM-B model does not take all components into consideration at the same time, even though they should be considered conjointly in order to understand and promote behaviour change (Graca et al., 2019).Further, the model does not specifically aim to understand how behavioural change is embedded in everyday life.
In this article, we use social practice theory as our approach to enable a more comprehensive understanding of barriers and facilitators of meat reduction in everyday food consumption.Due to the clear conceptual links between SPT and COM-B, SPT-based research can feed into the B. Halkier and T.B. Lund understanding of how the COM-B components operate together.While barriers and facilitators in everyday life are (explicit or implicit) cornerstones in SPT studies (see e.g.White et al., 2022, p. 175), the tripartite focus on barriers, facilitators and behavioural change from the COM-B model have not attracted particular attention.We believe that SPT-inspired research can be improved by adding domain-relevant behavioural outcome variables to the examination of the social practices.
Hence, in the following, we outline the practice theoretically inspired themes of barriers and facilitators that was included in the survey study on Danish flexitarians reported in this article.We carefully discuss how these categories fit in with the COM-B components and further highlight how findings from other studies on meat reduction inspired the questionnaire construction.

Everyday barriers and facilitators included in this study
There are several versions of social practice theory (SPT).The version drawing upon Shove et al. (2012) with focus on meanings, competences and materials has been used in several recent qualitative studies of meat reduction (Khara et al., 2021;White et al., 2022, p. 175), and this version tends to focus more on the elements organizing practices.Another version of SPT, drawing upon Warde (2005), tends to focus more on the performing (or 'practicing' of) practices, where the activities are held together by understandings, procedures and engagements.We are drawing upon the Warde (2005;2016) version which tends to focus on the practicing of food activities and not solely on the organization of food practices.We believe that this version of SPT is well matched with our use of questionnaire data since questions that focus on performing or behavior can be formulated in the quantitative format.
The specific themes of barriers and facilitators we aimed to capture are specified below.First, the activities whereby food practices are performed (Warde, 2005:134) are characterised by routines, or habits (Castelo et al., 2021, p. 157;Warde, 2016), meaning that people predominantly provide for and eat what they are used to.This dynamic of routines seems somewhat parallel to parts of the COM-B component called opportunity, because this covers processes that make meat reduction …"affordable, appropriate and easy to perform the behaviour" (Graca et al.,19b: 381).At the same time, habitual behaviour is explicitly a part of the COM-B component called motivations, because motivations cover "… automated processes" (ibid).This nicely illustrates how the different components in COM-B are intertwined, which is parallel to the overlap of the different characteristics of food activities highlighted by SPT.This habitual character of how people procure food, and then cook and eat it can work both as a barrier to meat reduction and as a facilitator once meat-reducing routines have been established.A Finnish quantitative study of plant-based eating (Pohjolainen et al., 2015) has demonstrated how food routines work as a barrier.Further, habits are mentioned as a barrier to reduction of meat consumption in a systematic literature review on consumption-side interventions in the context of meat consumption across disciplines (Kwasny et al., 2022, p. 168).It is mostly studies of vegetarians and vegans that show how routines can work as facilitators (e.g.Twine, 2018).
Second, food activities are guided by understandings (Warde, 2005:134), which seems parallel to parts of the COM-B component called motivation, because it covers "… inner reflective and automated processes that drive the behaviour" (Graca et al., 2019b: 381).We focus in particular on understandings situated close to practical everyday life (Daly, 2020;Mylan, 2018), such as those relating to taste, satiety and price.In two studies where flexitarians are clustered into subgroups (Malek & Umberger, 2021;Verain et al., 2022) motives were also included that appear to be close to everyday life food practices (such as appreciation and taste), parallel to our theme of understandings, as well as more attitudinal motives (such as health and environment).In the current literature, taste is mostly shown to work as a barrier to meat-reduction.Examples of this include a qualitative Australian study of meat consumption across different dietary groups (North et al., 2021, p. 167) a quantitative Portuguese study (Graca et al., 2015), and a quantitative Dutch study, using the COM-B model (Berg et al., 2022).
Third, food activities follow a variety of procedures (Warde, 2005:134), and some of these are closely linked with practical competencies and procurement (Janhonen et al., 2018;Plessz & Étile, 2019).The theme of procedures seems parallel to parts of the capability component in the COM-B model, since this covers "… knowledge, cognitive and interpersonal skills" (Graca et al., 2019b: 381).The literature suggests that competencies can work as barriers as well as facilitators.Thus a New Zealand qualitative study of meat-reduction at different life-stages finds the latter effect (Kemper, 2020, p. 150), but other studies (e.g.Khara et al., 2021;Pohjolainen et al., 2015) have identified lack of competencies to cook plant-based meals primarily as a barrier.How much competencies can be used and developed in practice also depend on procurement, i.e. the everyday organisation of food selection and delivery/purchase (Castelo et al., 2021, p. 157;Kwasny et al., 2022, p. 168).
Fourth, food activities take place in social relations and are thereby embedded in social conventions, similar to engagements (Warde, 2005:134), governing the appropriate ways to cook and eat (Halkier, 2021;Murcott, 2019).The theme of social relations and conventions is less easily dovetailed directly with the COM-B model, but perhaps it can be seen as part of the component opportunity, because Graca et al., 2019b describes opportunity as also covering social influences (381).The literature on meat-reduction also demonstrates the importance of social relations and norms.The British qualitative study of veganism mentioned above (Twine, 2018) underlines the importance of the social normalisation of plant-based eating.The same goes for some of the studies of meat reduction and of flexitarians (Cheah et al., 2020, p. 149;Ginn & Lickel, 2022;Verain et al., 2022).Here interpersonal relations and social conventions work as a facilitator.
The specific operationalisations of the four everyday themes as barriers and facilitators are shown in Table 2 in the article section "Materials and methods".
We aim to address the following research questions: (1) What everyday barriers to, and facilitators of, meat reduction are salient among flexitarians, and are there flexitarian subgroups that differ in respect of these everyday barriers and facilitators.Having identified some subgroups, we go on to examine (2) whether and how they develop over time, in tandem with duration of adherence to the diet, and (3) whether the target behaviour of eating less meat is achieved.
Our focus on flexitarian subgroups follows recommendations from two recent reviews in which segmentation was mentioned as an important research avenue (Graca et al., 2019b;Dagevos, 2021) and two recent empirical studies which has segmented flexitarians in subgroups (Malek & Umberger, 2021;Verain et al., 2022).It is likely that there will be substantial differences between consumersboth in general and, perhaps more particularly, during the conversion phase, where practises are being adapted to accommodate a new behaviour (in this case, less meat eating).Indeed, we find it likely that the flexitarian subgroups will co-develop with time elapsed (Research Question 2).It is also important to understand whether particular subgroups with distinctive complexes of practice switch more easily to their target behaviour (cf.Research Question 3).

Data
The study reported here used data from a cross-sectional, representative questionnaire survey of a sample of Danish consumers.The questionnaire was developed by the authors of this paper.After this a survey bureau was contracted to collect data through computer assisted telephone interviewing.The data were collected in June and July 2020.Participants were invited using random digit dialling (RDD) to both B. Halkier and T.B. Lund landline and cell-based telephones.Among the 10,859 individuals reached, 72% (7859) either refused to participate or failed to call back or answer the phone at a later (arranged) time, while 3000 (28%) completed the interview.A further 10 251 telephone numbers were not reached after five call attempts, giving an overall response rate of 14% (calculated using the RR3 formula (AAPOR, 2015)).
We compared the sociodemographic details of the sample (n = 3000) with census data for the adult Danish population (Table 1).
The sample deviates very modestly from the Danish adult population as regards gender and geographical location.On most accounts, the age distribution of the sample reflects the Danish population quite well, although participants above 70 years are overrepresented and young Danes (18-29 years) slightly underrepresented.Turning to education, the sample overrepresents Danes with medium-length and high levels of education, and underrepresents those with compulsory school and vocational education.This type of educational unrepresentativeness is common in surveys.Still, the sample in fact does include 353 participants with compulsory education, meaning that they are quite strongly represented in the survey.Given the sociodemographic mismatches, we applied weightings in all the descriptive analyses (with the following weight adjustments: gender, age, region, and highest completed education) to ensure the reported proportions were representative of the Danish population.

The questionnaire and measures
The questionnaire prompted for: sociodemographic details of the participant, what types of food the participant eats in general, and what types of food were eaten by the participant at the evening meal consumed the day before the interview.Following that, the questionnaire prompted about meat eating practices.Depending on the response to this, relevant follow-up questions were presented, including questions about why meat reduction was not considered, and (for those who had reduced their meat intake), what made it easy or difficult to eat plantbased foods.
The following measures were employed in the analysis.
Current meat diet group: Current meat diet groups were defined by responses to questions about which of four described diets characterized the respondent best: 1.I eat meat (including fish and poultry), 2. I eat little meat, 3. I eat vegetarian foods, and 4. I eat vegan foods.Participants giving response 1 were then asked whether they "had begun eating less meat the past year" (response options: 1 = yes and 2 = no).These two questions return five current meat diet groups: no reduction in meat intake, vegetarian diet, vegan diet, recent meat reducer, and eats small amounts of meat.In this study we treat participants in the last two groups as flexitarians, so henceforth when we refer to flexitarians it covers both groups.
Barriers and facilitators: To examine barriers to, and facilitators of, reduced meat consumption among flexitarians we developed two questions with multiple response options.In Table 2, we outline the responses offered to the respondents and divide them into the four themes presented earlier in the subsection entitled "Everyday barriers and facilitators included in this study".
Duration of flexitarian diet: Flexitarians were asked how long they had followed their diet (response options ranged from "six months" to "my whole life").We combined responses to this question with data on the respondents who had reported that they had begun eating less meat over the past year.This gave the following duration groups: 1-12 months, 1-2 years, 2-5 years, 5-10 years, more than 10 years, "whole life".
Consumption of meat (target behaviour): Three measures of the target behaviour, i.e. meat consumption, were developed.The first was based on a food frequency question, "How often  ) and fish, along with vegetables and other foods.From this question we developed two measures: whether the respondent had eaten "meat" for dinner yesterday, and whether he or she had eaten "red meat" for dinner yesterday.

Data analysis
Descriptive statistics were used to report the prevalence of meat eaters, flexitarians and vegans/vegetarians, and also to provide an overview of the barriers and facilitators affecting flexitarian consumers.
In addressing Research Question 1, latent class analysis (LCA) was employed to identify flexitarian subgroups with distinctive patterns of barriers and facilitators.Thus, in the rest of the article, we refer to the flexitarian subgroups as classes.LCA is a clustering technique that identifies possible latent classes through the analysis of interrelations among a number of manifest variables (or indicators) (McCutcheon, 1987).We inserted 20 response options from the measure of barriers and facilitators as binary indicators (0/1) in the LCA.We excluded two options requiring the respondent to be living with others, as these options were not relevant for all respondents.The response "none of these" was also omitted.We relied on statistical fit indices and substantial considerations when determining the number of latent classes for use in subsequent analyses.The fit indices reported in the study are the Akaike Information Criterion (AIC) (Akaike, 1987), the sample-size adjusted Bayesian Information Criterion (BIC) (Sclove, 1987), and entropy (Ramaswany et al., 1993).
Turning to Research Question 2, here we treated the variable indicating flexitarian classes as a dependent variable in a multinomial logistic regression, and inserted duration as an independent variable in order to study possible temporal trends between and within the classes.Sociodemographic factors (gender, age, education, household composition, employment status, and population density) were included in the regression as control variables.
Where Research Question 3 was concerned, we began by asking whether flexitarians generally consume less meat than non-flexitarian meat eaters.Here, we cross-tabulated these two groups with the measures of the meat consumption target behaviour, and then carried out X 2 tests of the associations.After that we examined whether the flexitarian classes and duration of a meat-reduced diet is associated with eating less meat using three logistic regression models in which each of the target behaviour variables were inserted as binary dependent variables.The following predictor variables were included in the regressions: flexitarian class, duration of flexitarian diet, and the sociodemographic controls mentioned above.

Danish consumers' meat-related diet, and food consumption characteristics
A majority of the respondents (approximately 54%) reported that they eat meat with no restrictions.Approximately 2.3% reported that they follow a vegetarian diet, and 0.8% a vegan diet.Between these poles were the flexitarian groups, with 24.7% reporting that they eat small amounts of meat and 18% reporting that they have begun eating less meat during the last year.

Barriers and facilitators among flexitarians
The shares of flexitarian consumers assenting to the facilitators we prompted about were higher than the shares of flexitarians assenting to the barriers (see the Total column in Table 3).Still, there are some quite widespread barriers, as more than half confirmed that routines ("I am used to eating meals with meat", 56%) and understandings in the form of preference for meat ("I like the taste of meat", 66%) makes it difficult to eat less meat.Also, approximately 60% agreed that it is difficult to eat less meat because "there are other things in everyday life that are more important than food".
To identify flexitarian classes with different patterns of barriers and facilitators, we ran LCA models with 1-6 classes (see fit statistics in Appendix 1).We found that the AIC and sample-size adjusted BIC (lower values reflect better-fitting latent class models) had not identified an optimal number at 6 classes.Nor did the Lo-Mendel-Rubin Likelihood Ratio test (in which a probability value > 0.05 indicates that the model with x+1 classes does not provide a significantly better fit to the data than a model with x number of classes) identify a statistically sufficient fit at 6 classes.A solution with four classes, however, returned a very high classification probability (entropy = 0.819), and the entropy values begin to decrease below 0.80 for models with 5 or more classes.Since the model solution with four classes also revealed strikingly different and sociologically relevant patterns of experienced barriers and facilitators, we chose to proceed with the four-class solution in the remaining analyses.
In Table 3, barriers and facilitators are reported for each class (see columns headed Class 1 to Class 4).We provide a conceptual overview of the different practices that characterise the four classes below.
Class 1 consumers (28% of the flexitarians; 12% of the Danish population) experience multiple barriers, although these do not include difficulties over the availability of meat replacements and plant-based products.Thus, 85% within this groups report that "It's easier for me to figure out how to make dishes with meat", and approximately four out of 10 find it difficult, time-consuming and expensive to make dishes without meat.Class 1 consumers are also less likely than Class 3 and Class 4 consumers to report that facilitators related to procedures and competencies make it easy for them to eat less meat, as only 20% of the respondents in these groups believe plant-based foods to be cheap, and 44% find it easy to make plant-based dishes.Understandings of taste and satiety, likewise, do not drive Class 1 respondents consistently, as only approximately 40% report that "plant-based food is tasty", and that "I feel full in a manner that is nicer for the body".
Class 2 consumers (17% of the flexitarians; 7% of the Danish population) differ markedly from Class 1, Class 3 and Class 4 consumers in that they tend to experience neither barriers nor facilitators to reduced meat consumption.The probable reason for this will become apparent below when we analyse the effect of duration of following a flexitarian diet.
Class 3 consumers (17% of the flexitarians; 7% of the Danish population) experience multiple facilitators of, but no barriers to, meat intake reduction.The large majority find it easy to make a dish without meat (83%), and that it is always possible to make a meal out of whatever vegetables are in the drawer (93%).Further, 65% of Class 3 consumers agree that "plant-based food is tasty", that "I feel full in a manner that is nicer for the body" (74%), and that they are "used to eating quite a few vegetable dishes" (87%).
Class 4 consumers (37% of the flexitarians; 16% of the Danish population) are like those in Class 3 in that they experience multiple facilitators.However, the class diverges significantly from Class 3 consumers as well, since a considerably higher share of Class 4 consumers see food as less important in daily life, are used to eating meat-based meals, and show an attachment to meat.Many assent to the statements that "There are other things in everyday life that are more important than food" (70%), that "I am used to eating meals with meat" (70%), and that "I like the taste of meat" (90%).

Flexitarian practices develop over time
We used multinomial logistic regression to examine the most B. Halkier and T.B. Lund important factors determining membership of the four flexitarian classes (see Appendix 2 for full details of the analysis).Among the sociodemographic control variables inserted into the regression, we found that gender was associated in a statistically significant way with membership of the flexitarian classes.However, the difference was relatively modest.Here, we will focus on what was clearly the most important factor explaining the flexitarian classes, namely: the duration over which less meat had been consumed.The pattern of this association is illustrated in Fig. 1 below.While we only have cross-sectional data available, and therefore cannot clarify how within-individual transitions between classes occur over time, the differences in class prevalence across the duration brackets do permit an interpretation of possible transition pathways to be made.We highlight these transitions wherever relevant below.
As can be seen from the figure, Class 1 consumers are found predominantly among flexitarians who have followed the diet for 1-12 months (36% share).The share of this class drops substantially (to around 10-15%) when the flexitarian diet is followed for more than 1 year.Class 4 consumers are noticeably prevalent in all duration categories.However, the share increases in the 1-2 years and 2-5 years brackets, and then declines in the 5-10 years and more than 10 years brackets.We recall that Class 4 consumers experience many facilitators but nonetheless report many barriers.They resemble Class 1 consumers as regards these motivational barriers, and it is therefore likely that Class a The shares of classes do not sum to 100 due to rounding error.b Question was not included in the latent class analysis, so the share is only reported in the Total column. 1 consumers who "survive" (i.e.do not relapse to former meat-eating patterns) for more than 12 months proceed to Class 4. In the same way, the reason that the share of Class 4 consumers decreases from the 5-10 years bracket and onwards could be that the barriers experienced diminish over time once these consumers become accustomed to the flexitarian practice, following which they transfer into either Class 2 or Class 3. Class 2 consumers are clearly most prevalent in the category of those who have been flexitarian for more than 10 years or their whole lives (almost 30%).They are less prevalent among those who have been flexitarian for less than 10 years.As Class 2 report experience neither of barriers nor of facilitators, we interpret the increased prevalence as time passes as a process of habituation to practices to the point where the cooking, procurement and eating activities involved in plant-based eating have become so routine and embedded in everyday life that no explicit challenges are experienced and there is no specific factor that makes eating less meat difficult.
Class 3 consumers experience no barriers and many facilitators.The share of this class is initially low (approximately 10% in the 1-12 months bracket), but it increases steadily over the first five years.At about 5 years, a plateau is reached, with a share of approximately 25% being sustained from that point onwards.

Flexitarian performance regarding meat consumption
In Table 4 (upper part), it can be seen that flexitarians in general also perform better, in practice, in reducing their intake of meat than nonflexitarians do.Thus, 27% of the flexitarians confirmed the statement "seldom or never eats red meat" as compared with 13% of the consumers reporting that they have not reduced meat intake.Also, 69% of the flexitarians had eaten meat (any kind) for dinner the day before the interview, while 80% of the consumers reporting that they have not reduced meat intake had done so.Finally, against the statement "Ate read meat for dinner yesterday" these groups scored 27% and 35%, respectively.
What explains the flexitarians' successful performance in the target behaviour?Do some flexitarian classes perform better than others?And how does durationtime elapsed since the flexitarian diet was begunaffect performance?We ran three multivariate logistic regressions in which the three meat consumption indicators were inserted as dependent variables.We provide full details of the three regression models in Appendix 3, Appendix 4 and Appendix 5, respectively.Here we report the main results of interest from this analysis in Table 4 (lower part).As specified in the "Materials and Methods" section, possible sociodemographic explanatory variables were included in the regressions models, but these were only treated as control variables.
We found substantial differences between the flexitarian classes in their frequency of eating red meat.Relative to the other classes, a noticeably smaller proportion of Class 1 consumers (17%) reported that they "seldom or never" ate red meat.This compares with Class 3 consumers (40%), Class 2 (28%) and Class 4 (29%).On the other hand, there were no significant differences between the classes when it came to having eaten some meat, or red meat specifically, for dinner the day before the interview.
Duration of following a flexitarian diet is also associated with two of the target behaviour indicators.A gradient trend is observable in Table 4 (lower part), in which a longer duration increases the likelihood of "seldom or never" eating red meat: 17% of consumers with 1-12 months of flexitarianism behind them report that they "seldom or never" eat red meat.For consumers with more than 10 years, the corresponding figure is 44%.These was an exception to this trend among consumers who had eaten less meat for their whole life: here 30% "seldom or never" ate red meat.A similar trend over time in the likelihood of having eaten meat for dinner the day before emerged: 75% of consumers with 1-12 months of the flexitarian diet had eaten meat for dinner yesterday, but that proportion decreases over time, dropping to 57% among consumers who had been flexitarian for more than 10 years.An unexpected exception to the trend here was that 72% of the respondents who had eaten small amounts of meat throughout their whole lives had eaten meat for dinner yesterday in this group.Finally, we identified no statistically significant differences between the duration categories as regards the probability of having eaten red meat for dinner the day before the interview.

Discussion
The contribution from our study to the existing research literature on meat reduction and flexitarians falls in one general and three more specific points.
The more general contribution is to use social practice theories (SPT) as a more comprehensive understanding of barriers and facilitators in the analysis of meat reduction among consumers.This is different from the dominant tendency in the existing literature which focus on single factors (e.g.Cheah et al., 2020, p. 149;Cliceri et al., 2018;Collier et al., 2021, p. 167;Randers & Thøgersen, 2023, p. 180;Vainio, 2019;Verain & Dagevos, 2022).Our SPT approach shares the comprehensiveness and several of the specific categories with the COM-B model (Graca et al., 2019b;Berg et al., 2022), but tend to see the barriers and facilitators as more embedded in everyday life than the COM-B model.
The first specific contribution of our paper, is the identification of four classes of flexitarians with distinct combinations of facilitators and barriers in everyday life.There are two prior studies where a clustering of flexitarians were also undertaken, which we find particularly relevant to consider.In a study from the Netherlands, Verain et al. (2022), using two comparable surveys from 2011 to 2019, identified three contrasting flexitarian groups: they labelled these unconscious, potential and conscious flexitarians.The input variables to the cluster analysis in the Dutch study involved a combination of single factors (such as personal norms, need for meat and ethical considerations) (Verain et al., 2022).In this sense, they have parallels with the barriers and facilitators

Table 4
Frequency of meat-eating (in general, and during yesterday's dinner) -for all meat-eating consumer groups, flexitarian classes and duration of following a flexitarian diet.belonging to the category "understandings" and to the category "social relations" that was employed in our identification of flexitarian classes.However, Verain et al., 2022 did not address the practical embeddedness in everyday life of flexitarian food activities.In a second, Australian study, Malek & Umberger (2021) identified five kinds of flexitarians: they labelled these heavy meat reducers, moderate meat reducers, and light meat reducers, dividing the last group into three subgroups.The input variables to the cluster analysis in this study were meat consumption frequency and willingness to reduce meat.In this way, the approach taken by Malek & Umberger (2021) also differs from ours, because in the former meat reduction activities related to, for example, routines, procedures, competencies and social relations were not included in the cluster analysis.At a general note then, neither Malek & Umberger, 2021 nor Verain et al., 2022 include facilitators and barriers embedded in everyday life in their clustering of flexitarian subgroups.In this sense, we believe that the flexitarian classes in our study come closer to consumer experiences of practical and social support and difficulties, irrespective of consumers' attitudes and intentions.
Our second specific contribution is that the flexitarian classes in our study differ in their meat consumption, but co-develop with time elapsed.Our results about the relationship between the flexitarian classes and the duration of following a flexitarian diet underpin the importance of temporality and routinization when changes in consumption are made in everyday life.The importance of this has been stressed by sociologists of consumption (Southerton, 2020;Warde, 2016), but does not seem to be part of the existing research in the field, even in the few studies comparing flexitarians over time (e.g.Verain et al., 2022).The COM-B model contributions to the field underscores routinization (Berg et al., 2022;Graca et al., 2019b), but does not explicitly suggest to study the temporal dimension.The same is the case for the qualitative studies of meat reduction, in the existing literature inspired by SPT (Khara et al., 2021;White et al., 2022, p. 175).We briefly present the main findings regarding this below, and discuss with other literature where relevant.We found that Class 1 consumers, who experience a lot of barriers, are more common when the flexitarian diet has been followed for a limited period of time (1-12 months).Barriers include some of the well-known ones from the existing literature, such as meat attachment (e.g.Berg et al., 2022;Graca et al., 2015).After this initial process, other classes with fewer (Class 4) or no (Class 2 and Class 3) barriers begin to predominate.While this is not a longitudinal study, and we therefore have not studied the same respondents over time, we think the association of class prevalence and duration indicates that flexitarians undergo a process of change in everyday life until many practices are routinized.In this connection, we note an interesting difference between Class 2 and Class 3 flexitarians, both of whom have few barriers: flexitarians in Class 2 had experienced no facilitators, but those in Class 3 had encountered many.This could be interpreted as a difference between routinized behaviour (not particularly difficult and requiring no explicit support) and identity-based consumption activities (no difficulties, but a need to "flag" supportive things in everyday life).Class 2 seems to dovetail with the COM-B model assumption on "automated processes" in the component called motivations (Graca et al., 2019b:381).In contrast, class 3, characterised by a recognition of how easy it is to eat less meat, appears to have a profile that reflects what Randers & Thøgersen (2023, p. 180) call a "flexitarian identity".On a general note, we believe that our results make clear that temporal change and routinization should be given further attention in future studies of flexitarians.This is also a clearly important perspective according to SPT scholars (Southerton, 2020;Warde, 2016), and habit formation scholars (Gardner et al., 2022).Preferably, changes in everyday practices and habits should be studied longitudinally, as this will strengthen our understanding of actual developmental routes, and also give a more accurate estimation of time that needs to elapse before substantial changes in flexitarian practices (both regarding barriers, facilitators, and the target behaviour of meat consumption) occur.
Our third and last specific contribution concerns actual meat reduction tendencies among the flexitarian classes where our social practice theory (SPT) approach was inspired by the COM-B model that emphasises the importance of the target behavior (e.g.Graca et al., 2019b;Grassian, 2020;Kemper & White, 2021).Our finding that many flexitarians appear to have routinized their practice is significant news, as it suggests that they are now less likely to relapse to increased meat consumption.Indeed, the more routinized flexitarians, i.e. especially Class 2 and Class 3, where there are few experienced barriers, reduced their frequency of eating meat the most.However, although we found that some flexitarian classes eat less meat, we stress that there is no dramatic decrease in any of the classes in comparison with non-flexitarians.There was, indeed, quite a high likelihood that the consumers who had followed a flexitarian diet for more than 10 years had eaten meat (unspecified or red) the day before we interviewed them.Also, Class 3 flexitarians, who might have been expected to have the highest success in reducing meat, as they experience practically no barriers at all and have multiple facilitators, regularly ate meat.Our results therefore suggest that there is a lot to overcome if we are to achieve a significantly greater reduction of meat intake even among the most engaged flexitarians.On this point, our study results are in line with findings from the Netherlands, where it would appear that the most engaged consumers are not dramatically reducing their meat intake over time (Verain et al., 2022).A possible reason for these discouraging results could be lagged adaptation where flexitarians have not yet picked up the many new meat replacement products that currently are entering the market, and are not yet aware of recent innovations in plant-based cuisines.In this view, it is just a question of time before they become widespread and people will adopt them into their everyday practices.In a more pessimistic interpretation, substantial reduction in meat intake will not come about by itself, even among flexitarians.We discuss this further below (see Study implications).
The study reported here has three limitations.First, the sample response rate of 14% was modest.Response rates are generally decreasing, and are sometimes very low nowadays.However, they are a poor indicator of nonresponse bias (Davern, 2013), and when population-adjusted weights are used (and we used them in all results relating to population prevalence) telephone surveys still provide accurate assessments of descriptive statistics (Pew Research Center, 2017).
Second, our measure of the duration of following a flexitarian diet is based on respondent self-reporting.It is notoriously difficult for respondents to recall the exact time at which changes occur earlier in their life.Therefore, there is an increased risk of memory bias for this measure.On the other hand, the response options offered included rather wide time intervals, which makes it easier for the respondents to retain relevant information from memory and match this with the response options.Also, we identified clear temporal differences in class prevalence, suggesting that the potential memory bias is limited.
Third, our measurement of meat-reduction relied on the selfreporting of the respondents, and only covered frequencies of meat intake, while quantities (in energy or kg meat) were not assessed.Although the studied flexitarians reported fewer meal episodes where meat is included they could be eating higher amounts of meat (higher kg volume) at the episodes where they do eat meat.While we believe that it is unlikely, in principle, the total kg intake of meat may then not be lower compared to the eating behavior before the consumers embarked on a flexitarian diet.
Meat consumption can be assessed in different ways (see Dagevos & Verbeke, 2022).We assessed the target behavior (meat eating) in two separate ways: through a food frequency question (how many times a day meat is eaten), and a question focussing on what was eaten at yesterday's dinner.There are pros and cons of both approaches.The first approach covers meat consumed the entire day, which is very desirable.But it also requires the respondents to provide an average account of the behavior.This may be difficult to assess accurately, and it may produce memory bias if respondents try to recall the behavior during e.g. the last month.Also, there may be between-individual variation in how B. Halkier and T.B. Lund respondents calculate this average assessment, and this may introduce differential response bias.The latter approach (yesterday's dinner) runs very little risk of memory and differential response bias, and it resembles the approach in 24-h dietary recall surveys used in dietary surveys.On the other hand, it does not cover the entire range of daily eating episodes where meat can be consumed.But, historically, meat has been a central part of a proper dinner in Denmark, and typically also the centrepiece of the dinner (Holm et al., 2015).Danes consume approximately one third of their daily energy intake during the dinner meal; this is a considerably higher rate than other episodes (Huseinovic et al., 2016).So the dinner meal is a culturally important meal insofar the reconfiguration of dishes (with and without meat) is concerned, and covers a large proportion of daily food intake.If real changes in the consumption of meat are taking place among flexitarians, we would expect to observe this in the dinner consumed.

Policy and market implications
More than half of Danes have not embarked on meat reduction.An important implication for policy makers and commercial actors is to raise awareness of the health and environmental benefits of eating less meat, and supporting the mainstreaming of meat reduction and replacement so that it becomes more normatively acceptable for a broader group of consumers (Stoll-Kleemann and Schmidt, 2017).Different tools exist to this effect.In a Danish context, the official dietary guidelines have recently added recommendations about limiting meat intake (Fødevarestyrelsen, 2021) which is an important step in the desired direction.Social marketing of plant-based foods is another important route (Stoll-Kleemann and Schmidt, 2017;Verfuerth et al., 2021).
However, changing people's intention and normative beliefs will not be enough.The results from this study together with the findings by Verain et al. (2022) are a cause of concern, as the reduction in meat intake appears to be modest even among the most conscious flexitarians (Verain et al., 2022), and among flexitarians where routinization in everyday life has emerged (this study).Communication campaigns targeting norms, priming, and salience (Byerly & et al, 2018) are less appropriate tools to target flexitarians, as they are already motivated to eat less meat.
Here it will be more important to develop learning tools focussing on the complexes and co-dependencies of food practices in everyday life (Castelo et al., 2021, p. 157) that will promote more plant-based eating, particularly for the two main reduction strategies that have been suggested in the literature (i.e. the 2R's of reduction and replacement, see Dagevos, 2021).Further, public and private food providers can play an important role in large-scale and continuous experiences with plant-based foods and dishes.For instance, trying out and getting acquainted with plant-based dishes in workplace and educational institutions can create positive spill-over to meals that are eaten at home (Verfuerth et al., 2021).

Conclusion
In this article, we contribute to the recent literature on flexitarians.In line with Hielkema & Lund, 2021 our results show that many Danish consumers are trying to reduce or have already reduced their meat intake.However, the combinations of facilitators and barriers in everyday life are differentially distributed on the four classes of Danish flexitarians that we have described and analysed.Furthermore, the four classes also differ according to the duration in time of their following a flexitarian diet, the pattern of which indicates that flexitarians transition to everyday practices where barriers are less strong and routinization emerges.Finally, our analysis showed that the Danish flexitarians in general report eating less meat than non-flexitarian Danish meat-eaters.
Here we also identify differences in meat eating behavior between the four classes of flexitarians.Surprisingly, meat intake is still quite common even in classes where routinization is highest.

Ethical statement
There have been no ethical issues involved in the empirical research that is the basis for this manuscript.

Declaration of competing interest
There are no conflict of interests related to this manuscript.

Table 1
Socio-demographic details of the study sample.
a The total in the study sample column does not sum to 100 because of rounding error.

Table 2
Overview of everyday practice themesdivided into facilitators and barriers.Halkier and T.B. Lunddo you eat red meat?", with response options ranging from "never/ seldom" to "more than one time per day".The second and third measures were based on a multiple response question prompting respondents about what they had eaten for dinner yesterday.Response options included multiple meat types (red meat, i.e. beef/veal/lamb, pork, poultry

Table 3
Results from latent class analysis of the barriers and facilitators experienced by flexitarians when reducing meat in everyday life (unweighted n = 1395) -reported as within column shares (in %).
b My cohabitant(s)/partner/kid(s) do not want to eat less meat

All meat-eating consumers (unweighted n¼2909)
a X 2 test of difference between flexitarians and consumers with no reduction in meat intake.bWald'sX 2 test of difference from multivariable logistic regressions.In addition to the latent class variable and duration of flexitarian diet, the following control variables were included in the regressions: gender, age, education, household composition, employment status, and population density,.B.Halkier and T.B. Lund