Abstract
Purpose
The purpose of this study is to explore how consumers apply clean-eating criteria to a range of food characteristics, and the extent to which individuals are consistent in how they apply clean-eating criteria across products. Further, this study investigates how the clean-eating approach relates to underlying food choice motives.
Design/methodology/approach
Data were collected in a consumer survey (n = 666) in Sweden, where participants were prompted about the importance of a set of intrinsic food attributes of the “free-from” and “added” types, for three different food product types (bread, processed meat, ready meals). Data were analyzed using latent class cluster analysis, to explore segments of consumers that place similar importance to the food characteristics and hold similar food choice motives.
Findings
Clean eating can be described by two distinctly different attainment strategies: avoiding undesirable characteristics or by simultaneously approaching desirable characteristics. Notably, individuals who apply clean-eating criteria in their food choices strive for healthy, natural and environmentally friendly food, but the clean-by-approach strategy implies a stronger focus on personal health in the form of weight control.
Originality/value
While claims and labels on food packages concerning clean eating are implemented by food manufacturers, it remains unregulated. This study provides information for future regulations on how consumers apply clean-eating criteria, and their motives thereof. Further, the results provide insights food manufacturers regarding motives for clean eating in different consumer segments.
Keywords
Citation
Edenbrandt, A.K. and Lagerkvist, C.-J. (2023), "Motives, propensities and consistencies among Swedish consumers in relation to the food choice concept of clean eating", British Food Journal, Vol. 125 No. 13, pp. 125-145. https://doi.org/10.1108/BFJ-03-2022-0217
Publisher
:Emerald Publishing Limited
Copyright © 2023, Anna Kristina Edenbrandt and Carl-Johan Lagerkvist
License
Published by Emerald Publishing Limited. This article is published under the Creative Commons Attribution (CC BY 4.0) license. Anyone may reproduce, distribute, translate and create derivative works of this article (for both commercial and non-commercial purposes), subject to full attribution to the original publication and authors. The full terms of this license may be seen at http://creativecommons.org/licences/by/4.0/ legalcode
Introduction
The dietary concept denoted as “clean eating” is a major trend in the food industry (Shelke, 2020), with 60.8 million #eatclean and 48.1 million #cleaneating posts on Instagram (as of 2022–10–22). With the increased interest and demand among consumers, clean eating is a concept that has gained much attention in industry and in food research, as well as in policy and regulation (Ingredion, 2014; Asioli et al., 2017; De Oliveira do Nascimento et al., 2018; European Commission, 2019; Martínez-Zamora et al., 2021; Roobab et al., 2021; Trujillo-Mayol et al., 2021). Importantly, the clean-eating concept is not clearly defined (Asioli et al., 2017), although it is often described by terms such as naturalness, “real food”, healthiness and absence of certain ingredients and additives that are perceived as harmful (Ambwani et al., 2020; Chen et al., 2022). Despite the increased interest and demand for clean labeling, there is not scientific evidence of such products holding higher levels of healthiness (Negowetti et al., 2022). Thus, the concept of clean eating and clean labeling of products may mislead consumers and be interpreted to hold qualities other than those that the consumers are seeking (Chen et al., 2022). Ultimately, unregulated clean labeling may even results in adverse health effects (Chen et al., 2022; Negowetti et al., 2022). This is a particular concern for vulnerable groups at risk of eating disorders (Allen et al., 2018; Ambwani, 2019).
Given the lack of clear definition and regulation of clean eating, the concept is open for interpretation among food manufacturers and consumers. Existing research suggest that an avoidance approach serve as the main behavioral motive for clean eating, meaning that the focus is to select food by some exclusion criteria, and in doing so screen for what the product needs to be “free from” such (Asioli et al., 2017). Other studies have documented avoidance of specific ingredients such as gluten, lactose, or palm-oil (Hartmann et al., 2018). However, a review by Asioli et al. (2017) also identified that clean-eating can relate to an approach-based search (i.e. seeking the presence of certain product quality attributes) such as organic, while other studies have focused on the degree of naturalness, degree of processing, freshness, number of ingredients and familiarity with ingredients as food choice aspects of clean eating (Aschemann-Wtizel and Peschel, 2019; Maruyama et al., 2021; Noguerol et al., 2021).
As evidenced in the review by Asioli et al., (2017), there is, to date, no study that characterize the concept of clean eating from a broader set of both presence and avoidance criteria, or which have investigated whether individuals in their role as consumers are consistent in applying the same search criteria for presence or absence of certain characteristics across food products. Therefore, the first objective of this study is to explore if there are underlying (i.e. latent) exclusive and exhaustive subgroups of individuals that adequately represent the heterogeneity in how they apply the presence and avoidance criteria. If so, what are these criteria and how large are these groups? A second, and related, objective is to investigate the extent to which individuals are consistent in applying the criteria across food products differing by their extent of processing.
Moreover, there is to our understanding no previous study investigating how the search behaviors related to avoidance and approach criteria match with consumers' underlying food choice motives. Investigation of the motivational basis for the clean-eating approach may gain insights from the Regulatory Focus Theory (Asioli et al., 2017), which proposes that there are two distinct motivational systems: promotion and prevention (Crowe and Higgins, 1997). An individual who is more prevention-oriented will then be expected to strive to attain a specific goal (such as healthiness) by avoiding unhealthy products or unhealthy characteristics in products. By contrast, individuals who are more promotion-oriented will strive to attain their goal by approaching healthy products or healthy characteristics in products (De Boer and Schösler, 2016). Therefore, a third objective of this study is to investigate how the search behaviors for presence or absence of certain characteristics across food products relate to underlying food choice motives. While reviews have identified healthiness as a key motivation for clean eating (Asioli et al., 2022; Chen et al., 2022), we investigate the heterogeneity in motives between the latent subgroups of individuals.
Currently, the concept of clean eating may disserve consumers, when used as a cue for healthiness, despite that there is not scientific support for this (Negowetti et al., 2022). It has been reported that clean-eating behavior, if persistent, comes in risk of contradicting public guidelines for a healthy diet with a long-term potential of bringing adverse medical health consequences (Nevin and Vartanian, 2017). It remains an open question how the concerns with using clean eating as a cue for healthiness should be addressed. The main contribution of this study to the literature on clean eating as a food choice behavior goes to the characterization of how individuals apply this concept across food products. Thus, this study does not discuss clean labeling per se, but rather contributes with understanding of what it is consumers seek in terms of presence and avoidance criteria, and what motivates their search behavior with respect to clean eating. Findings of a consistent search pattern across product categories would allow for more general actions from health professionals and/or from concerned public authorities, whereas more product specificity would warrant a more differentiated health information approach. Furthermore, the understanding of the motives underlying the proneness of clean eating and the further categorization of consumer segments by their proneness to search for presence or avoidance criteria may guide efforts to tailor information to better resonate with consumers.
Methods
Participants
Participants were recruited from an Internet panel administered by a research panel firm (userneeds.com) in September 2020 in Sweden. Quotas were used for gender and age to match the distribution of the Swedish population. This type of study and data collection does not require approval according to the Swedish Ethical Review Authority (Swedish Ethical Review Authority, 2021). Descriptive statistics for our sample (n = 666) is presented in Table 1, alongside Swedish population statistics for comparative purposes. The sample is also representative of the population with respect to household size, while individuals with university level education were overrepresented in the sample.
Products and product characteristics
Three products were included in the study, selected to represent foods of varying extent and type of processing, while also being expected to be purchased and consumed regularly by many consumers in Sweden. Based on these criteria, we included bread, cooked meat (represented by sausages) and ready-meals (represented by a broccoli pie with eggs). Each product was displayed using a generic image showing the product in a standard type of packaging but with no presentation of brand or other labels.
Procedure
Participants responded to an online survey developed in Qualtrics with three parts. The first part included self-reported personal characteristics (gender, age and educational level). The second part included Food Choice motives (see Section 2.4), while the third and final part asked the participants to rate the importance of the product-related clean-eating characteristics. Only respondents who confirmed that they had purchased the specific product at least once in the past month were qualified to respond to the product-specific parts of the questionnaire. This resulted in 53 observations being excluded after indicating that respondents had not purchased any of these products in the past month. After this exclusion, the number of participants varied by product with n = 626 for bread, n = 497 for cooked meat and n = 302 for ready-meals.
Food choice motives
In the second part of the survey, to gain insights into the motives for clean eating, respondents indicated the importance of the central concepts of clean eating: healthiness and naturalness. Specifically, based on the food choice questionnaire (Steptoe et al., 1995; Onwezen et al., 2019), we included motives related to health (general healthiness and healthiness in the form of weight control) and naturalness, and following the interconnectedness between naturalness and environmental sustainability (Rozin et al., 2004; Rozin, 2006), we also included environmental friendliness.
The importance of each of the motives was measured on scales from the Single Item Food Choice questionnaire (SI-FCQ) (Onwezen et al., 2019), which includes 11 items to measure 11 food choice motives. The questionnaire is developed based on the Food Choice questionnaire (Steptoe et al., 1995), which includes nine food choice motives (36 items). Respondents indicated their agreement with the statement “It is important to me that the food I eat on a typical day” on a seven-point scale ranging from “not at all important” (1) to “very important” (7).
Importance of the product-related clean-eating characteristics
In the third part of the survey, as applicable after the screening for product purchase, participants were prompted about the importance of a set of product-based intrinsic food attributes of the “free-from” and “added” types. These attributes were included in the study based on the existing literature (Bimbo et al., 2017; Hartmann et al., 2018; Aschemann-Wtizel and Peschel, 2019; Maruyama et al., 2021; Noguerol et al., 2021). The “free from” included eight characteristics: free from preservatives, colorants, palm oil, artificial sweeteners, added sugar, lactose and gluten, and finally, a more general cue for the free-from aspect by a short ingredient list. The free-from characteristics are avoidance-type qualities that consumers may use as cues to prevent harmful effects on their health. While there is no clear evidence that lactose-free and gluten-free food is healthier for non-intolerant individuals, some results suggest a perception that it is healthier and more natural to avoid lactose and gluten (Hartmann et al., 2018). Similarly, there is a perception that palm oil is unhealthy and unnatural (Hartmann et al., 2018). Furthermore, a second set of characteristics were included, in the “added” form: added vitamins, extra protein, or extra fiber. This reflects health attainment by enhancement, such that health is approached (Bimbo et al., 2017). While we sought to include the same attributes for all three products, certain adjustments were made, such that added fiber/wheat was excluded from the cooked meat category, and added vegetables was only included for the cooked meat.
For each of the three products, the importance of each food characteristic was measured with the statement: “How does the following characteristic impact the extent to which you want to purchase this [product type] or not?” The rate of agreement was indicated on a five-point scale (1 = very negative impact, 5 = very positive impact). The order of presentation of the clean-eating characteristics was randomized for each product to avoid ordering effects. The full list of food characteristics for each product is available in Table S1 in Supplementary Material.
Data analysis
Based on the indicated importance of the quality attributes for each product type, we first explored whether there were sub-groups of individuals who among themselves place similar importance to the different food attributes related to the clean-eating concept, while being distinctly different from other sub-groups. For this purpose, we took an exploratory approach to investigate the concept of clean eating as a product-based latent variable. The latent class cluster (LCC) model is appropriate for such analysis (Masyn, 2013; Nylund-Gibson and Choi, 2018). For the LCC-analysis we transformed the indicators (importance scale) into a binary form, where positive impact (4) and very positive impact (5) gave a value of one and zero otherwise. Binary indicators are most commonly used in LCC analysis (Nylund-Gibson and Choi, 2018), and transformation to binary indicators is suitable when categorical variables include low frequency response categories (Masyn, 2013). Based on this transformation, we estimated separate LCC models for each of the products included.
We further explored whether the importance of the clean-eating qualities hold across product types, such that there is consistency in individuals' sub-group belonging across product types. Based on the product-specific LCC models, we assigned individuals to the class with the highest posterior class membership probability.
Finally, we examine how the product-specific latent class membership is explained using the importance of the four food choice motives (healthiness, weight control, naturalness and environmental friendliness) as covariates. The inclusion of covariates as predictors of class membership followed the approach from Lanza et al. (2007).
The LCC modeling approach includes a measurement and a structural model, where the former describes the relation between product-based clean-eating attributes and a latent class variable. LCC analysis makes it possible to identify the most suitable number of classes in the measurement data and assign a probability of each variable to each class. It further enabled us to assign class membership probabilities for each individual to each class. The LCC-models were estimated in Latent Gold 5.1 (Vermunt and Magidson, 2013). A detailed description of the method and model selection is available in the Appendix. Moreover, appendix includes details on robustness checks with alternative model specifications.
Results
Inferring clean eating from product characteristics
For each of the products, we identified four distinct and non-trivial latent classes. While there are some differences in these classes between the products, the overall patters are similar. Based on their patterns, we labeled the classes as: Clean-by-avoidance, Clean-by-approach, Moderately engaged (in clean eating) and Unengaged (in clean eating). Figure 1 shows the share of individuals in each latent class for whom each characteristics have a positive to very positive importance in their product choice.
Both the clean-by-avoidance and the clean-by-approach classes were characterized by being positively influenced in their purchase decisions by avoiding preservatives, colorants, sweeteners, palm oil, chemical pesticides (organic) and additives in general (short ingredient list). However, differences emerged between the clean-eating classes in terms of product characteristics that are in the approach form (added vitamins, added fiber/wheat, added vegetables and high protein). Contrary to individuals in the clean-by-avoidance class, individuals in the clean-by-approach class were positively influenced by such characteristics in their willingness to purchase a product. Furthermore, the two attributes gluten-free and lactose-free positively impacted some individuals in the clean-by-approach class, while individuals in the clean-by-avoidance class did not assign importance to these qualities. Together, the clean-related classes constituted approximately 40% of the respondents in each product.
For the remaining classes, the moderately engaged class was characterized by intermediate levels of positive importance for most of the characteristics. The unengaged class consisted of individuals who did not find that most of the characteristics included had a positive impact on their willingness to purchase the products.
Is clean eating product-specific?
To explore whether the membership in a clean-eating class is consistent across products, respondents were assigned to the class with the highest membership probability in each of the product-specific models. Table 2 shows the correspondence of class membership for pairs of the products. Overall, there was movement between classes across products, which implied that many individuals were not in the same class for all products. However, some central patterns emerged. Importantly, individuals who were assigned to one of the clean-eating classes in one product model were often assigned to a clean-eating class in the other product models. For example, among individuals who were assigned to the clean-by-approach class in the bread model, 81% (6 + 75) were in a clean-eating class in the cooked meat model (Table 2a) and 87% (30 + 57) in the ready meal model (Table 2b). Furthermore, 80% (7 + 73) of the individuals in the clean-by-approach class in the ready meal model were in a clean-eating class in the cooked meat model (Table 2c). Similar patterns emerged for the clean-by-avoidance class membership in the bread model: 59% of the individuals were in a clean-eating class in the cooked meat model and 90% were in a clean-eating class in the ready meal model. Interestingly, while there was relatively high consistency in clean-eating class membership across products, the specific type of clean-eating class membership varies. For example, among the individuals in the clean-by-approach in the bread model, a large share (30%) was in the clean-by-avoidance in the ready meal model.
For the classes less engaged in clean eating, there was relatively high correspondence in class membership between the products: individuals who were in the unengaged class in one product model were typically also assigned to the moderately engaged or unengaged class in the other product models.
Some noteworthy deviations from these patterns are present. Overall, cooked meat stands out by displaying that many individuals are in different classes than for the other products. Specifically, many individuals who were in the clean-by-avoidance class when selecting bread were unengaged when selecting cooked meat (39%, Table 2a), and 45% of individuals who were clean-by-avoidance in the ready meal model were unengaged in the cooked meat model (Table 2c). More detailed results for the class constituency across products are available in Table S5 in Supplementary Materials.
Motivations for clean eating
Figure 2 shows the average importance scores for the motives for each class and product. In line with prior expectations, food choice motives related to naturalness and health were important among the clean-eating classes. Both clean-eating classes found health (general healthiness and weight control), naturalness and environmental friendliness to be more important food choice motives compared to the less engaged classes.
A noteworthy difference between the clean-by-approach and clean-by-avoidance classes is the importance of weight control, where the approach-oriented individuals found this more important (differences are statistically significant in the bread and ready meal models). Hence, both clean-eating classes strived for healthy, natural and environmental friendly food, but the clean-by-approach class was more focused on the personal health in the form of controlling their weight. This was echoed in the impact from individual food product characteristic; individuals in the clean-by-approach class valued personal health-promoting characteristics such as added vitamins and protein. Health was also important for the moderately engaged class, as reflected in the higher importance on fiber, protein and vitamins compared to the clean-by-avoidance class (Figure 1).
Discussion
Clean-eating is relatively stable across products-but there are two subgroups that apply different clean-eating criteria
A key contribution of this study is insights on how individuals apply clean eating across food product types. Three different product categories were included in this study, with varying degree of processing. We identified two subgroups of individuals that apply clean-eating criteria in their food choices. Together, the clean-eating classes constituted about 40% of the respondents in each product. In line with previous studies, avoidance of additives, pesticides, colorants and preservatives were important (Hartmann et al., 2018; Aschemann-Wtizel et al., 2019; Maruyama et al., 2021; Noguerol et al., 2021). Individuals in both subgroups found it important to avoid undesirable characteristics such as additives, preservatives, colorants and chemical pesticides (organic). However, contrary to the first subgroup, which achieved clean eating by avoidance, the second subgroup sought presence of added protein, fiber, vitamins, etc. This type of approach orientation rather fits with a related trend in the form of health promoting, or functional, food products, such as nutrition-enhanced products (Bimbo et al., 2017). Interestingly, although many consumers are consistent in that they apply clean-eating criteria across different product types, it is not uncommon for consumers to apply different clean-eating criteria for different types of products. For the selected product types included in this study, cooked meat (such as sausages) is different in terms of their clean-eating criteria. Further studies that include a wider range of products are needed in order to shed light on the reason for such differences.
Motivations for clean eating
The results of this study suggest that individual who are prone to search for clean-eating attributes are motivated to make food choices that are healthy, natural and environmentally friendly. While clean eating is centered around motives related to health and naturalness, we found that this applies to health in a broad sense, and only a segment of individuals are clean eaters in a strict sense, while another segment are more focused on personal healthiness. Thus, although both clean-eating groups strive for healthy, natural and environmentally friendly food, our results suggest that the clean-by-approach class have a strong focus on the personal health in the form of weight control. These findings imply that for a segment of consumers, applying clean-eating criteria as an avoidance strategy in their food choices does not contradict preferences for nutrition enhancements. Driven by a growing interest in clean eating among consumers, food manufacturers are adapting and reformulating their products towards ingredient lists that are more in line with the concept of clean eating (Noguerol et al., 2021). The findings in this study suggest that this appeals to a segment of consumers who are highly concerned with healthiness and naturalness, but a share of these consumers simultaneously value efforts to enrich products with healthy characteristics.
Limitations and future research
There are several opportunities for future research. First, a wider range of product types could be included, to systematically investigate differences between food categories. For example, future research may explore if clean-eating criteria are more or less important for products that are perceived as healthy (for example, müsli) than for products that are perceived as less healthy (e.g. chocolate). Second, while this study explores the importance of clean eating-related attributes among a sample of Swedish consumers, future studies could include respondents from other countries, to shed light on potential variations across cultures in the clean-eating criteria and motivations.
Furthermore, an interesting extension to this study would be to relate the clean-eating motives to consumers' perceptions about specific food product attributes and to examine how this is related to knowledge levels on nutrition and environmental impact. For example, how do consumers interpret the different “free-from” attributes, and does this interpretation vary with the nutritional literacy?
Implications for regulators
Clean eating is a food-choice approach based on the sought presence or absence of certain food characteristics (Ambwani et al., 2020). However, the clean-eating concept remains undefined, although it is typically described by absence of certain ingredients and additives, “pure” and “real” food, and naturalness (Asioli et al., 2017; Grant et al., 2019; Ambwani et al., 2020). Importantly, while there is interest and demand among for clean eating among consumers, there is no scientific support for higher healthiness in such products, and the clean-eating concept may disserve consumers seeking a healthy diet (Chen et al., 2022). Targeting the discrepancy between the qualities that the consumers may seek (e.g. healthiness) and the qualities that clean labeling implies can be targeted by different measures. ,
Clean labels are not well-defined or regulated, although many of the aspects of clean eating are covered by existing regulations on claims, particularly in the EU (Merten-Lentz, 2019; Mahy and Serve, 2020; Ghaderi, 2022; Negowetti et al., 2022). Importantly, regulating the use of clean labeling is associated with challenges (Negowetti et al., 2022), and it is not likely that misleading information on packages and websites can be eliminated (Ghaderi, 2022). Another venue for supporting consumers to identify products that are healthier is to increase nutrition literacy (Chen et al., 2022; Ghaderi, 2022). This study suggests that in the context of clean eating, there are large segments of consumers that do not value clean eating aspects, particularly the Moderately engaged and Unengaged latent classes. We find that the food characteristics that typically describe clean eating are mainly valued by consumer segments that find healthiness, naturalness and environmental friendliness important, and to a lesser degree weight loss, and this result applies across product categories. These findings could support the design and dissemination of educational efforts that aim to balance the present clean-eating claims and marketing with scientifically supported nutritional information.
Figures
Sociodemographic characteristics
| Variable | Sample | Population |
|---|---|---|
| Female (proportion) | 0.50 | |
| Age categories (proportions) | ||
| 18–34 | 0.28 | 0.28 |
| 35–49 | 0.28 | 0.24 |
| 50–64 | 0.22 | 0.23 |
| 65– | 0.23 | 0.25 |
| Household size | 2.3 | 2.2 |
| Education level (proportions) | ||
| Elementary | 0.07 | 0.17 |
| high school | 0.43 | 0.43 |
| university level or similar | 0.50 | 0.40 |
Note(s): Sample includes 666 individuals. Population statistics from Statistics Sweden (scb.se)
Correspondence in class membership between product types
| Cooked meat | ||||||
|---|---|---|---|---|---|---|
| Panel A | Clean-by-avoidance | Clean-by-approach | Moderately engaged | Unengaged | Total | |
| Bread | Clean-by-avoidance | 10% | 49% | 2% | 39% | 100% |
| Clean-by-approach | 6% | 75% | 0% | 19% | 100% | |
| Moderately engaged | 19% | 15% | 22% | 45% | 100% | |
| Unengaged | 3% | 3% | 84% | 11% | 100% | |
| Ready meal | ||||||
|---|---|---|---|---|---|---|
| Panel B | Clean-by-avoidance | Clean-by-approach | Moderately engaged | Unengaged | Total | |
| Bread | Clean-by-avoidance | 86% | 4% | 0% | 10% | 100% |
| Clean-by-approach | 30% | 57% | 11% | 2% | 100% | |
| Moderately engaged | 22% | 3% | 30% | 45% | 100% | |
| Unengaged | 2% | 0% | 16% | 82% | 100% | |
| Cooked meat | ||||||
|---|---|---|---|---|---|---|
| Panel C | Clean-by-avoidance | Clean-by-approach | Moderately engaged | Unengaged | Total | |
| Ready meal | Clean-by-avoidance | 5% | 48% | 3% | 45% | 100% |
| Clean-by-approach | 7% | 73% | 0% | 20% | 100% | |
| Moderately engaged | 16% | 18% | 32% | 34% | 100% | |
| Unengaged | 8% | 3% | 70% | 19% | 100% | |
Note(s): An example to assist the interpretation. Among individuals who were assigned to the clean-by-avoidance class in the Bread model (top row in Panel A), 10% were in a clean-by-avoidance in the cooked meat model, while 49% were in the clean-by-approach, 2% in the moderately engaged class, and 39% in the unengaged class
Categorization of food characteristics
| Bread | Cooked meat | Ready-meal | |
|---|---|---|---|
| Free-from characteristics | |||
| Lactose-free | ✓ | ✓ | ✓ |
| Gluten-free | ✓ | ✓ | ✓ |
| Free from palm oil | ✓ | ✓ | ✓ |
| Free from preservatives | ✓ | ✓ | ✓ |
| Free from colorants | ✓ | ✓ | ✓ |
| No added sugar | ✓ | ✓ | ✓ |
| Free from artificial sweeteners | ✓ | ✓ | ✓ |
| Short ingredient list | ✓ | ✓ | ✓ |
| Organic | ✓ | ✓ | ✓ |
| Added aspect | |||
| Extra protein | ✓ | ✓ | ✓ |
| Added vegetables | ✓ | ||
| Extra fiber/wheat | ✓ | ✓ | |
| Added vitamins | ✓ | ✓ | ✓ |
Note(s): In addition to the listed attributes, we asked respondents about labels and certificates. Domestically and locally produced (all products), and free-range eggs (ready meal)
Class enumeration
| # Classes | LL | BIC | AIC3 | SABIC | K | L2 | p-val | CE | LRT |
|---|---|---|---|---|---|---|---|---|---|
| Bread | |||||||||
| 1 | −4,582 | 9,242 | 9,201 | 9,204 | 12 | 2,739 | 0.000 | 0.00 | |
| 2 | −3,946 | 8,053 | 7,967 | 7,974 | 25 | 1,467 | 0.000 | 0.04 | 0.000 |
| 3 | −3,860 | 7,964 | 7,833 | 7,843 | 38 | 1,294 | 0.000 | 0.11 | 0.000 |
| 4 | −3,807 | 7,942 | 7,766 | 7,780 | 51 | 1,188 | 0.000 | 0.11 | 0.000 |
| 5 | −3,770 | 7,951 | 7,731 | 7,748 | 64 | 1,114 | 0.000 | 0.13 | 0.000 |
| 6 | −3,747 | 7,991 | 7,726 | 7,746 | 77 | 1,069 | 0.018 | 0.15 | 0.000 |
| 7 | −3,729 | 8,038 | 7,728 | 7,752 | 90 | 1,033 | 0.008 | 0.14 | 0.018 |
| 8 | −3,712 | 8,087 | 7,733 | 7,760 | 103 | 998 | 0.020 | 0.15 | 0.014 |
| 9 | −3,698 | 8,143 | 7,744 | 7,775 | 116 | 970 | 0.006 | 0.16 | 0.120 |
| 10 | −3,685 | 8,200 | 7,757 | 7,791 | 129 | 944 | 0.008 | 0.14 | 0.080 |
| 11 | −3,667 | 8,248 | 7,759 | 7,797 | 142 | 908 | 0.028 | 0.16 | 0.004 |
| 12 | −3,654 | 8,306 | 7,773 | 7,814 | 155 | 882 | 0.010 | 0.15 | 0.148 |
| Cooked meat | |||||||||
| 1 | −3,082 | 6,233 | 6,197 | 6,198 | 11 | 1742 | 0.000 | 0.00 | |
| 2 | −2,584 | 5,310 | 5,237 | 5,237 | 23 | 746 | 0.000 | 0.04 | 0.000 |
| 3 | −2,528 | 5,274 | 5,162 | 5,163 | 35 | 635 | 0.000 | 0.11 | 0.000 |
| 4 | −2,493 | 5,278 | 5,127 | 5,129 | 47 | 564 | 0.004 | 0.13 | 0.000 |
| 5 | −2,469 | 5,305 | 5,116 | 5,118 | 59 | 517 | 0.020 | 0.11 | 0.020 |
| 6 | −2,453 | 5,348 | 5,120 | 5,122 | 71 | 485 | 0.048 | 0.11 | 0.006 |
| 7 | −2,440 | 5,396 | 5,130 | 5,133 | 83 | 459 | 0.108 | 0.11 | 0.108 |
| 8 | −2,431 | 5,452 | 5,147 | 5,150 | 95 | 440 | 0.038 | 0.15 | 0.152 |
| 9 | −2,417 | 5,499 | 5,156 | 5,160 | 107 | 413 | 0.084 | 0.18 | 0.074 |
| 10 | −2,412 | 5,562 | 5,180 | 5,184 | 119 | 401 | 0.042 | 0.16 | 0.576 |
| 11 | −2,403 | 5,619 | 5,199 | 5,203 | 131 | 384 | 0.084 | 0.15 | 0.224 |
| 12 | −2,393 | 5,674 | 5,215 | 5,220 | 143 | 364 | 0.080 | 0.14 | 0.066 |
| Ready-meal | |||||||||
| 1 | −2052 | 4,168 | 4,138 | 4,133 | 11 | 1,553 | 0.000 | 0.00 | |
| 2 | −1,667 | 3,466 | 3,403 | 3,393 | 23 | 782 | 0.000 | 0.03 | 0.000 |
| 3 | −1,601 | 3,402 | 3,307 | 3,291 | 35 | 650 | 0.000 | 0.05 | 0.000 |
| 4 | −1,535 | 3,339 | 3,212 | 3,190 | 47 | 519 | 0.036 | 0.07 | 0.000 |
| 5 | −1,521 | 3,378 | 3,218 | 3,191 | 59 | 489 | 0.022 | 0.07 | 0.040 |
| 6 | −1,505 | 3,416 | 3,223 | 3,190 | 71 | 458 | 0.110 | 0.10 | 0.016 |
| 7 | −1,491 | 3,456 | 3,231 | 3,193 | 83 | 430 | 0.030 | 0.08 | 0.036 |
| 8 | −1,482 | 3,507 | 3,249 | 3,205 | 95 | 412 | 0.040 | 0.09 | 0.264 |
| 9 | −1,473 | 3,557 | 3,267 | 3,217 | 107 | 393 | 0.056 | 0.10 | 0.206 |
| 10 | −1,463 | 3,606 | 3,284 | 3,229 | 119 | 375 | 0.012 | 0.07 | 0.136 |
| 11 | −1,455 | 3,657 | 3,302 | 3,242 | 131 | 357 | 0.056 | 0.08 | 0.096 |
| 12 | −1,448 | 3,712 | 3,324 | 3,258 | 143 | 343 | 0.028 | 0.08 | 0.368 |
Note(s): LL = Log Likelihood, BIC=Bayesian information criterion, AIC3 = Akaike information criterion with 3 as penalty factor, SABIC = sample size adjusted BIC, L2 = likelihood ratio chi-square, CE = classification error, LRT = p-value from bootstrapped Likelihood ratio test
Within class average (Bread)
| Moderately engaged | Unengaged | Clean-by-avoidance | Clean-by-approach | |
|---|---|---|---|---|
| Class size | 34% | 28% | 19% | 19% |
| Indicators | ||||
| No preservatives | 0.56 | 0.03 | 0.98 | 0.92 |
| Added vitamin | 0.27 | 0.09 | 0.09 | 0.66 |
| Gluten-free | 0.06 | 0.04 | 0.01 | 0.25 |
| Lactose-free | 0.10 | 0.07 | 0.11 | 0.32 |
| Organic | 0.46 | 0.12 | 0.69 | 0.80 |
| High protein | 0.32 | 0.15 | 0.02 | 0.83 |
| No added sugar | 0.62 | 0.14 | 0.76 | 0.93 |
| Short ingredient list | 0.39 | 0.07 | 0.82 | 0.81 |
| No colorants | 0.50 | 0.06 | 0.98 | 0.99 |
| High share wheat | 0.79 | 0.26 | 0.67 | 0.90 |
| Free from palm oil | 0.43 | 0.23 | 0.98 | 0.95 |
| No sweeteners | 0.63 | 0.05 | 0.82 | 0.95 |
| Covariates | ||||
| Food motives | ||||
| Health | 5.33 | 4.58 | 5.57 | 5.92 |
| Naturalness | 4.92 | 4.11 | 5.98 | 5.72 |
| Weight | 4.42 | 3.53 | 4.02 | 4.88 |
| Environment | 4.58 | 3.78 | 5.30 | 5.60 |
Note(s): N = 626
Within class average (Cooked meat)
| Moderately engaged | Unengaged | Clean-by- approach | Clean-by-avoidance | |
|---|---|---|---|---|
| Cluster Size | 34% | 27% | 21% | 19% |
| Indicators | ||||
| No preservatives | 0.46 | 0.04 | 0.93 | 0.98 |
| Added vitamin | 0.18 | 0.01 | 0.45 | 0.05 |
| Gluten-free | 0.06 | 0.03 | 0.35 | 0.07 |
| Lactose-free | 0.08 | 0.06 | 0.33 | 0.04 |
| Organic | 0.49 | 0.19 | 0.91 | 0.67 |
| No added sugar | 0.48 | 0.01 | 0.95 | 0.87 |
| Short ingredient list | 0.44 | 0.08 | 0.90 | 0.76 |
| No colorants | 0.40 | 0.05 | 0.95 | 0.98 |
| No sweeteners | 0.51 | 0.02 | 0.98 | 0.87 |
| High in meat | 0.84 | 0.66 | 0.98 | 0.98 |
| Added vegetables | 0.14 | 0.00 | 0.59 | 0.02 |
| Covariates | ||||
| Food motives | ||||
| Health | 5.08 | 4.53 | 6.04 | 5.43 |
| Naturalness | 4.86 | 4.13 | 5.88 | 5.56 |
| Weight | 4.00 | 3.50 | 4.80 | 4.65 |
| Environment | 4.55 | 3.69 | 5.77 | 4.73 |
Note(s): N = 497
Within class average (Ready meal)
| Uninvolved | Clean-by-avoidance | Moderately engaged | Clean-by-approach | |
|---|---|---|---|---|
| Cluster Size | 43% | 27% | 17% | 13% |
| Indicators | ||||
| No preservatives | 0.10 | 0.93 | 0.32 | 0.97 |
| Added vitamin | 0.02 | 0.11 | 0.60 | 0.93 |
| Gluten-free | 0.04 | 0.07 | 0.11 | 0.65 |
| Lactose-free | 0.04 | 0.10 | 0.16 | 0.66 |
| Organic | 0.18 | 0.82 | 0.42 | 0.94 |
| High protein | 0.02 | 0.10 | 0.72 | 0.82 |
| No added sugar | 0.12 | 0.81 | 0.41 | 0.97 |
| Short ingredient list | 0.07 | 0.73 | 0.39 | 0.98 |
| No colorants | 0.09 | 0.99 | 0.47 | 0.93 |
| High in fiber | 0.12 | 0.46 | 0.76 | 0.84 |
| Free from palm oil | 0.22 | 0.90 | 0.46 | 0.99 |
| Covariates | ||||
| Food motives | ||||
| Health | 4.61 | 5.51 | 5.01 | 5.74 |
| Naturalness | 4.34 | 5.49 | 4.46 | 5.69 |
| Weight | 3.49 | 4.27 | 4.56 | 5.42 |
| Environment | 3.71 | 5.16 | 4.75 | 5.59 |
Note(s): N = 302
Parameters LCC model (Bread)
| Moderately engaged | Unengaged | Clean-by-avoidance | Clean-by-approach | p-val | R2 | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| Coef | z-val | Coef | z-val | Coef | z-val | Coef | z-val | |||
| Indicators | ||||||||||
| No preservatives | −0.48 | −1.55 | −4.23 | −8.47 | 3.04 | 4.67 | 1.67 | 4.25 | <0.001 | 0.55 |
| Added vitamin | 0.21 | 0.95 | −1.05 | −3.94 | −1.05 | −3.03 | 1.89 | 8.07 | <0.001 | 0.23 |
| Gluten-free | 0.16 | 0.33 | −0.36 | −0.71 | −1.58 | −1.39 | 1.78 | 4.07 | <0.001 | 0.09 |
| Lactose-free | −0.26 | −1.05 | −0.74 | −2.67 | −0.17 | −0.62 | 1.17 | 5.46 | <0.001 | 0.07 |
| Organic | −0.17 | −0.95 | −2.02 | −8.92 | 0.81 | 3.88 | 1.38 | 5.98 | <0.001 | 0.26 |
| High protein | 0.41 | 0.74 | −0.55 | −1.01 | −2.59 | −1.71 | 2.73 | 4.69 | <0.001 | 0.34 |
| No added sugar | −0.12 | −0.58 | −2.46 | −10.02 | 0.55 | 2.20 | 2.03 | 5.71 | <0.001 | 0.35 |
| Short ingredient list | −0.45 | −2.32 | −2.52 | −9.14 | 1.54 | 5.32 | 1.43 | 5.51 | <0.001 | 0.37 |
| No colorants | −1.57 | −2.04 | −4.32 | −5.36 | 2.21 | 1.98 | 3.68 | 1.76 | <0.001 | 0.57 |
| High share wheat | 0.53 | 2.44 | −1.85 | −9.78 | −0.12 | −0.56 | 1.44 | 4.96 | <0.001 | 0.27 |
| Free from palm oil | −1.60 | −4.07 | −2.51 | −6.51 | 2.48 | 2.58 | 1.62 | 2.58 | <0.001 | 0.40 |
| No sweeteners | 0.00 | 0.02 | −3.45 | −8.93 | 0.96 | 3.16 | 2.48 | 5.37 | <0.001 | 0.47 |
| Covariates | ||||||||||
| Intercept | 1.48 | 2.88 | 5.27 | 10.04 | −2.89 | −3.74 | −3.85 | −5.64 | <0.001 | |
| Food motives | ||||||||||
| Health | 0.12 | 1.21 | −0.11 | −1.20 | −0.25 | −2.26 | 0.24 | 2.12 | 0.030 | |
| Naturalness | −0.29 | −3.24 | −0.51 | −5.45 | 0.78 | 5.82 | 0.02 | 0.14 | <0.001 | |
| Weight | 0.12 | 2.10 | −0.17 | −2.76 | −0.12 | −1.97 | 0.18 | 2.82 | <0.001 | |
| Environment | −0.11 | −1.69 | −0.28 | −4.02 | 0.08 | 0.99 | 0.30 | 3.32 | <0.001 | |
Parameters LCC model (Cooked meat)
| Unengaged | Clean-by-avoidance | Clean-by- approach | Moderately engaged | p-val | R2 | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| Coef | z-val | Coef | z-val | Coef | z-val | Coef | z-val | |||
| Indicators | ||||||||||
| No preservatives | −0.87 | −1.68 | −3.96 | −5.66 | 1.81 | 2.92 | 3.03 | 2.38 | <0.001 | 0.55 |
| Added vitamin | 0.75 | 1.38 | −2.05 | −2.12 | 2.07 | 4.58 | −0.77 | −0.70 | <0.001 | 0.19 |
| Gluten-free | −0.49 | −1.06 | −1.01 | −1.89 | 1.70 | 5.41 | −0.21 | −0.40 | <0.001 | 0.14 |
| Lactose-free | −0.23 | −0.47 | −0.48 | −0.93 | 1.56 | 3.60 | −0.85 | −0.74 | <0.001 | 0.12 |
| Organic | −0.40 | −1.65 | −1.83 | −6.50 | 1.90 | 3.78 | 0.33 | 1.14 | <0.001 | 0.26 |
| No added sugar | −0.20 | −0.45 | −4.44 | −5.00 | 2.87 | 4.46 | 1.77 | 3.18 | <0.001 | 0.52 |
| Short ingredient list | −0.44 | −1.78 | −2.58 | −7.12 | 2.04 | 4.39 | 0.98 | 2.95 | <0.001 | 0.38 |
| No colorants | −1.31 | −2.17 | −3.83 | −5.74 | 1.98 | 2.67 | 3.17 | 1.97 | <0.001 | 0.57 |
| No sweeteners | −0.46 | −0.94 | −4.22 | −6.07 | 3.27 | 2.81 | 1.41 | 2.36 | <0.001 | 0.53 |
| High in meat | −0.84 | −1.82 | −1.80 | −4.60 | 1.25 | 1.41 | 1.39 | 1.32 | <0.001 | 0.12 |
| Added vegetables | 1.42 | 1.01 | −4.39 | −1.19 | 3.56 | 2.68 | −0.59 | −0.27 | <0.001 | 0.33 |
| Covariates | ||||||||||
| Intercept | 2.08 | 3.74 | 5.00 | 8.29 | −5.58 | −6.13 | −1.49 | −1.98 | <0.001 | |
| Food motives | ||||||||||
| Health | −0.05 | −0.52 | −0.16 | −1.52 | 0.37 | 2.45 | −0.15 | −1.02 | 0.097 | |
| Naturalness | −0.16 | −1.65 | −0.40 | −3.97 | 0.12 | 0.85 | 0.44 | 2.90 | <0.001 | |
| Weight | −0.07 | −1.05 | −0.19 | −2.76 | 0.06 | 0.64 | 0.21 | 1.81 | 0.022 | |
| Environment | −0.02 | −0.22 | −0.29 | −3.36 | 0.50 | 3.40 | −0.19 | −1.20 | <0.001 | |
Parameters LCC model (Ready meal)
| Uninvolved | Clean-by-avoidance | Moderately engaged | Clean-by-approach | p-val | R2 | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| Coef | z-val | Coef | z-val | Coef | z-val | Coef | z-val | |||
| Indicators | ||||||||||
| No preservatives | −3.05 | −5.17 | 1.85 | 2.53 | −1.57 | −2.66 | 2.76 | 1.71 | <0.001 | 0.62 |
| Added vitamin | −3.00 | −5.34 | −1.38 | −3.09 | 1.11 | 2.64 | 3.27 | 4.40 | <0.001 | 0.56 |
| Gluten-free | −1.45 | −3.27 | −0.78 | −1.82 | −0.24 | −0.53 | 2.46 | 6.56 | <0.001 | 0.34 |
| Lactose-free | −1.57 | −3.77 | −0.63 | −1.62 | −0.06 | −0.13 | 2.25 | 6.34 | <0.001 | 0.30 |
| Organic | −2.13 | −7.09 | 0.89 | 2.64 | −0.94 | −2.72 | 2.18 | 3.45 | <0.001 | 0.39 |
| High protein | −3.10 | −4.97 | −1.28 | −2.92 | 1.89 | 4.15 | 2.48 | 4.86 | <0.001 | 0.57 |
| No added sugar | −2.62 | −7.32 | 0.81 | 2.30 | −1.02 | −2.67 | 2.83 | 3.83 | <0.001 | 0.47 |
| Short ingredient list | −3.01 | −4.93 | 0.57 | 0.95 | −0.83 | −1.35 | 3.27 | 2.01 | <0.001 | 0.49 |
| No colorants | −3.45 | −5.79 | 3.17 | 2.03 | −1.22 | −2.00 | 1.50 | 1.96 | <0.001 | 0.65 |
| High in fiber | −2.17 | −7.30 | −0.31 | −1.18 | 1.00 | 2.78 | 1.47 | 3.64 | <0.001 | 0.34 |
| No palm oil | −2.54 | −4.38 | 0.88 | 1.37 | −1.47 | −2.34 | 3.12 | 1.88 | <0.001 | 0.43 |
| Covariates | ||||||||||
| Intercept | 4.69 | 7.25 | −1.32 | −1.97 | 0.78 | 1.02 | −4.15 | −3.90 | <0.001 | |
| Food motives | ||||||||||
| Health | −0.05 | −0.36 | 0.12 | 0.87 | −0.06 | −0.32 | −0.02 | −0.10 | 0.820 | |
| Naturalness | −0.05 | −0.42 | 0.31 | 2.34 | −0.45 | −3.06 | 0.18 | 1.02 | 0.008 | |
| Weight | −0.37 | −4.14 | −0.18 | −1.95 | 0.15 | 1.23 | 0.40 | 2.74 | <0.001 | |
| Environment | −0.44 | −4.54 | 0.05 | 0.48 | 0.19 | 1.43 | 0.20 | 1.35 | <0.001 | |
Correspondence of predicted membership between classes extracted from the food products. Sum of all percentages equals 100
| Cooked meat | |||||
|---|---|---|---|---|---|
| Bread | Clean-by-avoidance | Clean-by-approach | Moderately engaged | Unengaged | |
| Clean-by-avoidance | 2.2% | 11.1% | 0.4% | 8.9% | |
| Clean-by-approach | 0.9% | 10.4% | 0.0% | 2.6% | |
| Moderately engaged | 5.9% | 4.6% | 6.7% | 13.9% | |
| Unengaged | 0.9% | 0.9% | 27.2% | 3.5% | |
| Note(s): N = 460 | |||||
| Ready meal | |||||
|---|---|---|---|---|---|
| Bread | Clean-by-avoidance | Clean-by-approach | Moderately engaged | Unengaged | |
| Clean-by-avoidance | 15.1% | 0.7% | 0.0% | 1.7% | |
| Clean-by-approach | 5.5% | 10.7% | 2.1% | 0.3% | |
| Moderately engaged | 6.5% | 1.0% | 8.9% | 13.4% | |
| Unengaged | 0.7% | 0.0% | 5.5% | 27.8% | |
| Note(s): N = 291 | |||||
| Cooked meat | |||||
|---|---|---|---|---|---|
| Ready meal | Clean-by-avoidance | Clean-by-approach | Moderately involved | Unengaged | |
| Clean-by-avoidance | 1.2% | 12.8% | 0.8% | 11.9% | |
| Clean-by-approach | 0.8% | 9.1% | 0.0% | 2.5% | |
| Moderately engaged | 2.5% | 2.9% | 4.9% | 5.3% | |
| Unengaged | 3.7% | 1.2% | 31.7% | 8.6% | |
Note(s): N = 243
Paired comparisons for covariates (Bread)
| Moderately engaged (ME) | Unengaged (UE) | Clean-by-avoidance (C-av) | Clean-by-approach (C-ap) | |
|---|---|---|---|---|
| Health | C-Av | C-Ap | ME, C-Ap | UE, C-Av |
| Naturalness | C-Av | C-Av, C-Ap | C-Ap | UE, C-Av |
| Weight | UE, C-Av | ME, C-Ap | ME, C-Ap | UE, C-Av |
| Environment | C-Ap | C-Av, C-Ap | UE | UE |
Paired comparisons for covariates (Cooked meat)
| Moderately engaged (ME) | Unengaged (UE) | Clean-by-approach (C-ap) | Clean-by-avoidance (C-av) | |
|---|---|---|---|---|
| Health | C-Ap | C-Ap | ME, UE, C-Av | C-Ap |
| Naturalness | C-Av | C-Ap, C-Av | UE | UE |
| Weight | C-Av | UE | ||
| Environment | UE, C-Ap | ME, C-Ap | ME, UE, C-Av | C-Ap |
Paired comparisons for covariates (Ready meals)
| Uninvolved (UE) | Clean-by-avoidance (C-av) | Moderately engaged (ME) | Clean-by-approach (C-ap) | |
|---|---|---|---|---|
| Health | ||||
| Naturalness | C-Av, ME | UE, ME | UE, C-Av, C-Ap | ME |
| Weight | ME, C-Ap | C-Ap | UE | UE, C-Av |
| Environment | C-Av, ME, C-Ap | UE | UE | UE |
Class enumeration for models with all products included (bread, cooked meat, ready meal)
| # Classes | LL | BIC | AIC3 | SABIC | K | L2 | p-val | CE | Df | LRT |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | −4,840 | 9,865 | 9,781 | 9,757 | 34 | 7,207 | 0.024 | 0.00 | 201 | |
| 2 | −3,907 | 8,190 | 8,021 | 7,972 | 69 | 5,341 | 0.092 | 0.01 | 166 | <0.001 |
| 3 | −3,693 | 7,954 | 7,698 | 7,624 | 104 | 4,914 | 0.212 | 0.03 | 131 | <0.001 |
| 4 | −3,549 | 7,856 | 7,514 | 7,416 | 139 | 4,625 | 0.148 | 0.02 | 96 | <0.001 |
| 5 | −3,463 | 7,876 | 7,448 | 7,324 | 174 | 4,454 | 0.100 | 0.02 | 61 | <0.001 |
| 6 | −3,387 | 7,915 | 7,401 | 7,253 | 209 | 4,302 | 0.080 | 0.02 | 26 | <0.001 |
Note(s): LL = Log Likelihood, BIC=Bayesian information criterion, AIC3 = Akaike information criterion with 3 as penalty factor, SABIC = sample size adjusted BIC, L2 = likelihood ratio chi-square, CE = classification error, LRT = p-value from bootstrapped likelihood ratio test
Within class average (All products)
| Moderately engaged | Uninvolved | Clean-by-avoidance | Clean-by-approach | |
|---|---|---|---|---|
| Cluster Size | 34% | 27% | 20% | 19% |
| Bread | ||||
| No preservatives | 0.39 | 0.05 | 0.98 | 0.89 |
| Added vitamin | 0.26 | 0.04 | 0.09 | 0.76 |
| Gluten-free | 0.09 | 0.03 | 0.00 | 0.37 |
| Lactose-free | 0.11 | 0.07 | 0.08 | 0.49 |
| Organic | 0.35 | 0.11 | 0.59 | 0.84 |
| High protein | 0.31 | 0.08 | 0.20 | 0.78 |
| No added sugar | 0.50 | 0.15 | 0.78 | 0.96 |
| Short ingredient list | 0.31 | 0.04 | 0.78 | 0.76 |
| No colorants | 0.45 | 0.06 | 0.91 | 0.98 |
| High in fiber | 0.70 | 0.22 | 0.66 | 0.86 |
| Free from palm oil | 0.38 | 0.20 | 0.81 | 0.98 |
| No sweetener | 0.57 | 0.05 | 0.79 | 0.95 |
| Cooked meat | ||||
| No preservatives | 0.40 | 0.10 | 0.85 | 0.95 |
| Added vitamin | 0.20 | 0.00 | 0.00 | 0.61 |
| Gluten-free | 0.09 | 0.05 | 0.02 | 0.44 |
| Lactose-free | 0.08 | 0.09 | 0.02 | 0.46 |
| Organic | 0.45 | 0.20 | 0.78 | 0.87 |
| No added sugar | 0.45 | 0.05 | 0.81 | 0.91 |
| Short ingredient list | 0.39 | 0.03 | 0.77 | 0.87 |
| No colorants | 0.39 | 0.08 | 0.86 | 0.93 |
| High in meat | 0.82 | 0.74 | 0.96 | 0.98 |
| Added vegetables | 0.15 | 0.00 | 0.09 | 0.54 |
| No sweetener | 0.50 | 0.02 | 0.83 | 1.00 |
| Ready meal | ||||
| No preservatives | 0.29 | 0.06 | 0.91 | 0.95 |
| Added vitamin | 0.26 | 0.03 | 0.06 | 0.74 |
| Gluten-free | 0.06 | 0.05 | 0.04 | 0.50 |
| Lactose-free | 0.11 | 0.07 | 0.02 | 0.50 |
| Organic | 0.33 | 0.18 | 0.73 | 0.91 |
| High protein | 0.32 | 0.03 | 0.14 | 0.63 |
| No added sugar | 0.34 | 0.02 | 0.85 | 0.98 |
| Short ingredient list | 0.28 | 0.00 | 0.72 | 0.83 |
| No colorants | 0.36 | 0.05 | 0.95 | 0.93 |
| High in fiber | 0.49 | 0.03 | 0.39 | 0.89 |
| Free from palm-oil | 0.39 | 0.19 | 0.88 | 0.98 |
| Covariates | ||||
| Food motives | ||||
| Health | 4.61 | 5.51 | 5.01 | 5.74 |
| Naturalness | 4.34 | 5.49 | 4.46 | 5.69 |
| Weight | 3.49 | 4.27 | 4.56 | 5.42 |
| Environment | 3.71 | 5.16 | 4.75 | 5.59 |
Note(s): N = 235
When the LCC analysis is exploratory, as is the case in this study, the first step is to identify the model with the most appropriate number of classes. This should be guided by considering multiple measures of model fit and model diagnosis together with interpretability. Model fit can be evaluated by absolute model fit, where the likelihood ratio chi-square (L2) goodness of fit test, where the null hypothesis states that the model adequately fits the data. A significant p-value indicates a lack of adequate fit. As for relative model fit, the likelihood ratio test (LRT) tests whether the addition of one latent class improves model fit significantly. The null hypothesis is that there is no difference in model fit between the model compared to the model with one less class. Hence, an insignificant p-value suggests that the addition of a class does not significantly improve model fit. Moreover, information criteria measures can be compared between models, including BIC and AIC3, where the model with the lowest value has the best relative model fit. Finally, in classification diagnostics, the model precision of assigning individuals to the different latent classes can be evaluated. The classification error (CE) is based on estimated posterior class probabilities, and it measures the proportion of individuals that are estimated to be misclassified, wherefore values closer to zero is better (Vermunt and Magidson, 2005). Classification measures are not used for model selection, but rather indicate whether there are concerns with over-extraction of latent classes (Masyn, 2013).
For the Bread model, we found the four-class model to be the most suitable. The relative LRT test suggests that the eight-class model is the most suitable model, but this model includes classes that are similar and very small, and in such cases it is recommended to include fewer classes for the cause of interpretation. Overall, the information criteria suggest that the four-class model provides the best model fit. The BIC is lowest for four classes. The AIC3 and SABIC are lowest for six classes, but the improvement in the information criteria for these measures is relatively small from four classes.
For the Cooked Meat model, we also found that the model with four classes is most suitable. The BIC is smallest for the three-class model, for AIC3 and SABIC it is five classes, but the rate of improvement is relatively small from four classes. The chi-square test suggests that the seven-class model provides adequate model fit, while the LRT test suggests that the model with six classes has the best model fit. However, similar to the case of bread, this model provides several small classes that are similar.
Finally, for the ready meal-model, we found that the four-class model is most suitable. All three information criteria (BIC, AIC3, SABIC) are lowest for the four-class model. The chi-square test suggests that the six-class model has a good absolute model fit (p-value>0.05), while the LRT test suggests the seven-class model.
Information regarding the absolute fit and relative fit statistics and classification diagnosis are presented in the Supplementary Material (Table S2). A central assumption in LCC analysis is local independence, which implies that, conditional on the latent variable, the observed indicators should be independent. The local independence assumption is assessed by examining the bivariate residuals among all pairs of indicator. BVR values above 30 are considered severe violations (Asparouhov and Muthén, 2015). For the selected models, there are no severe violations.
We proceed by estimating LCC models where the covariates are included. This procedure, of first finding the model with the most appropriate number of classes prior to including covariates, is recommended (Masyn, 2013) since it provides more stable results and is less sensitive to possible misspecifications from the covariates. The full model results, with parameters and z-values, are presented in Table S3in Supplementary Material. When interpreting results of the LCC-models, it is intuitive to interpret the within-cluster distribution of indicators and covariates. These results are illustrated in Figure 1, while detailed results are available in Table S4 in Supplementary Material.
When exploring the extent to which individuals belong to the same type of class for each of the products, we assign each individual to the class with the highest probability, based on their responses to the questions. The correspondence of predicted membership between classes is presented in Table S5.
Finally, we explored the importance of food choice motives for the different classes, in each of the product models. Summary results are presented in Figure 2, while statistical tests for differences across classes, in the form of paired comparisons for the covariates, are presented in Table S6.
For robustness, we estimated models in which all indicators for all three product types are included in the same model. These models include only the individuals that responded to all three products, providing a smaller sample than the separate models. This model specification, where all product indicators are included in the same model further implies a large number of parameters relative to the separate models for each product types, and consequently less statistically significant parameters. Moreover, the large number of parameters implies that it is only possible to estimate models with six classes or fewer. Results are available in Table S7 (class enumeration) and Table S8 (parameter estimates for the selected model). Overall, the results are in line with the findings from the separate models: two separate clean eating classes are identified, where one has an avoidance focus and one has an approach focus.
References
Allen, M., Dickinson, K.M. and Prichard, I. (2018), “The dirt on clean eating: a cross sectional analysis of dietary intake, restrained eating and opinions about clean eating among women”, Nutrients. doi: 10.3390/nu10091266.
Ambwani, S., Shippe, M., Gao, Z. and Austin, S.B. (2019), “Is #cleaneating a healthy or harmful dietary strategy? Perceptions of clean eating and associations with disordered eating among young adults”, Journal of Eating Disorders, Vol. 7 No. 1, p. 17, doi: 10.1186/s40337-019-0246-2.
Ambwani, S., et al. (2020), “‘It's healthy because it's natural.’ Perceptions of ‘Clean’ eating among U.S. adoloscents and emerging adults”, Nutrients, Vol. 12 No. 1708.
Aschemann-Wtizel, J. and Peschel, A.O. (2019), “Consumer perception of plant-based proteins: the value of source transparency for alternative protein ingredients”, Food Hydrocolloids, Vol. 96, pp. 20-28.
Aschemann-Wtizel, J., Varela, P. and Peschel, A.O. (2019), “Consumers' categorization of food ingredients: do consumers perceive them as ‘clean label’ producers expect? An exploration with projective mapping”, Food Quality and Preference, Vol. 71 January 2018, pp. 117-128.
Asioli, D., Aschemann-Wtizel, J., Caputo, V., Vecchio, R., Annunziata, A., Næs, T. and Varela, P. (2017), “Making sense of the ‘clean label’ trends: a review of consumer food choice behavior and discussion of industry implications”, Food Research International, Vol. 99, pp. 58-71.
Asioli, D., Banovic, M., Barone, A.M., Grasso, S. and Nayga, R.M. Jr. (2022), “European consumers’ valuation for hybrid meat: does information matter?”, Applied Economic Perspectives and Policy, pp. 1-19.
Asparouhov, T. and Muthén, B. (2015), “Residual associations in latent class and latent transition analysis”, Structural Equation Modeling, Vol. 22 No. 2, pp. 169-177.
Bimbo, F., Bonanno, A., Nocella, G., Viscecchia, R., Nardone, G., De Devitiis, B. and Carlucci, D. (2017), “Consumers' acceptance and preferences for nutrition-modified and functional dairy products: a systematic review”, Appetite, Vol. 113, pp. 141-154, doi: 10.1016/j.appet.2017.02.031.
Chen, A., Kayrala, N., Trapeau, M., Aoun, M. and Bordenave, N. (2022), “The clean label trend: an ineffective heuristic that disserves both consumers and the food industry?”, Comprehensive Reviews in Food Science and Food Safety, Vol. 21, pp. 4921-4938.
Crowe, E. and Higgins, E.T. (1997), “Regulatory focus and strategic inclinations: promotion and prevention in decision-making”, Organizational Behavior and Human Decision Processes, Vol. 69 No. 2, pp. 117-132, doi: 10.1006/obhd.1996.2675.
De Boer, J. and Schösler, H. (2016), “Food and value motivation: linking consumer affinities to different types of food products”, Appetite, Vol. 103, pp. 95-104, doi: 10.1016/j.appet.2016.03.028.
De Oliveira do Nascimento, K., do Nascimento Dias Paes, S. and Maria Augusta, I. (2018), “A review ‘clean labeling’: applications of natural ingredients in bakery products”, Journal of Food and Nutrition Research, Vol. 6 No. 5, pp. 285-294, doi: 10.12691/jfnr-6-5-2.
European Commission (2019), “Special eurobarometer wave EB91.3”, available at: https://www.efsa.europa.eu/sites/default/files/corporate_publications/files/Eurobarometer2019_Food-safety-in-the-EU_Full-report.pdf
Ghaderi, A. (2022), “A European perspective on ‘clean eating’: commentary on Negowetti et al. (2021)”, International Journal of Eating Disorders, Wiley Online Library, Vol. 55 No. 1, pp. 49-51.
Grant, K., Gallardo, R.K. and McCluskey, J.J. (2019), “Measuring consumer preference for clean label in processed foods”, Agricultural and Applied Economics Association Annual Meeting, Atlanta. doi: 10.1017/CBO9781107415324.004.
Hartmann, C., Hieke, S., Taper, C. and Siegrist, M. (2018), “European consumer healthiness evaluation of ‘Free-from’ labelled food products”, Food Quality and Preference, Vol. 68, pp. 377-388.
Ingredion (2014), “The clean label guide to Europe”, available at: https://www.ingredion.es/content/ingredion/emea/Campaign/Clean-Label-Guide.html
Lanza, S.T., Collins, L.M., Lemmon, D.R. and Schafer, J.L. (2007), “Proc lca: a sas procedure for latent class analysis”, Structural Equation Modeling, Vol. 14 No. 4, pp. 671-694, doi: 10.1038/jid.2014.371.
Mahy, A. and Serve, A. (2020), “Belgian circular clarifies the use of ingredients in the ‘clean label’ trend”, European Food and Feed Law Review, Vol. 15 No. 1, pp. 59-61.
Martínez-Zamora, L., Peñalver, R., Ros, G. and Nieto, G. (2021), “Substitution of synthetic nitrates and antioxidants by spices, fruits and vegetables in clean label spanish chorizo”, Food Research International, Vol. 139, 109835, doi: 10.1016/J.FOODRES.2020.109835.
Masyn, K.E. (2013), “Latent Class analysis and finite mixture modeling”, in Little, T.D. (Ed.), The Oxford Handbook of Quantitative Methods, Oxford University Press, New York, Vol. 2, pp. 551-611, available at: http://books.google.com.au/books?id=FjNCnQEACAAJ
Maruyama, S., Streletskaya, N.A. and Lim, J. (2021), “Clean label: why this ingredient but not that one?”, Food Quality and Preference, Vol. 87 August 2020, doi: 10.1016/j.foodqual.2020.104062.
Merten-Lentz, K. (2019), “How the EU has left ‘clean labelling’ open to national initiatives”, National Law Review, Vol. IX No. 310, available at: https://www.natlawreview.com/article/how-eu-has-left-clean-labelling-open-to-national-initiatives
Negowetti, N., Ambwani, S., Karr, S., Rodgers, R.F. and Austin, S.B. (2022), “Digging up the dirt on ‘clean’ dietary labels: public health considerations and opportunities for increased Federal oversight”, International Journal of Eating Disorders, Vol. 55, pp. 39-48.
Nevin, S.M. and Vartanian, L.R. (2017), “The stigma of clean dieting and orthorexia nervosa”, Journal of Eating Disorders, Vol. 5 No. 1, p. 37, doi: 10.1186/s40337-017-0168-9.
Noguerol, A.T., Pagán, M.J., García-Segovia, P. and Varela, P. (2021), “Green or clean? Perception of clean label plant-based products by omnivorous, vegan, vegetarian and flexitarian consumers”, Food Research International, Elsevier, Vol. 149, 110652, doi: 10.1016/J.FOODRES.2021.110652.
Nylund-Gibson, K. and Choi, A.Y. (2018), “Ten frequently asked questions about latent class analysis”, Translational Issues in Psychological Science, Vol. 4 No. 4, pp. 440-461, doi: 10.1037/tps0000176.
Onwezen, M.C., Reinders, M.J., Verain, M.C.D. and Snoek, H.M. (2019), “The development of a single-item food choice questionnaire”, Food Quality and Preference, Vol. 71, pp. 34-45.
Roobab, U., Khan, A.W., Lorenzo, J.M., Arshad, R.N., Chen, B.R., Zeng, X.A., Bekhit, A.E.D., Suleman, R. and Aadil, R.M. (2021), “A systematic review of clean-label alternatives to synthetic additives in raw and processed meat with a special emphasis on high-pressure processing (2018-2021)”, Food Research International, Vol. 150, 110792, doi: 10.1016/J.FOODRES.2021.110792.
Rozin, P. (2006), “Naturalness judgments by lay Americans: process dominates content in judgments of food or water acceptability and naturalness”, Judgment and Decision Making, Vol. 1, pp. 91-97, available at: https://econpapers.repec.org/RePEc:jdm:journl:v:1:y:2006:i:p:91-97
Rozin, P., Spranca, M., Krieger, Z., Neuhaus, R., Surillo, D., Swerdlin, A. and Wood, K. (2004), “Preference for natural: instrumental and ideational/moral motivations, and the contrast between foods and medicines”, Appetite, Vol. 43, pp. 147-154, doi: 10.1016/j.appet.2004.03.005.
Shelke, K. (2020), “Clearing up clean label confusion”, Food Technology Magazine, available at: https://www.ift.org/news-and-publications/food-technology-magazine/issues/2020/february/features/clearing-up-clean-label-confusion
Swedish Ethical Review Authority (2021), “Etikprovningsmyndigheten”, available at: https://etikprovningsmyndigheten.se (Accessed 20 December 2021).
Steptoe, A., Pollard, T.M. and Wardle, J. (1995), “Development of a measure of the motives underlying the selection of food: the food choice questionnaire”, Appetite, Vol. 25 No. 3, pp. 267-284, doi: 10.1006/appe.1995.0061.
Trujillo-Mayol, I., Sobral, M.M.C., Viegas, O., Cunha, S.C., Alarcón-Enos, J., Pinho, O. and Ferreira, I.M.P.L.V.O. (2021), “Incorporation of avocado peel extract to reduce cooking-induced hazards in beef and soy burgers: a clean label ingredient”, Food Research International, Elsevier, Vol. 147, 110434, doi: 10.1016/J.FOODRES.2021.110434.
Vermunt, J.K. and Magidson, J. (2005), Latent GOLD 4.0 User's Guide, Statistical Innovations, Belmont, MA.
Vermunt, J.K. and Magidson, J. (2013), Technical Guide for Latent GOLD 5.0: Basic, Advanced, and Syntax, Statistical Innovations, Belmont, MA.
Acknowledgements
The study was funded by Formas (Swedish Research Council for Sustainable Development), Grant number 2018-01846.