EASY EATING: FOOD FOR THOUGHT
A Conjoint Experiment Examining Consumer Trade-Offs in
Contemporary Convenience Foods Across Different Consumer
Segments
EASY EATING: FOOD FOR THOUGHT
A Conjoint Experiment Examining Consumer Trade-Offs in Contemporary
Convenience Foods Across Different Consumer Segments
Pien van Vliet
January 11
th, 2021
Master thesis
MSc Marketing Management & Marketing Intelligence
Faculty of Economics and Business
University of Groningen
First Supervisor: M.T. van der Heide
Second Supervisor: dr. F. Eggers
Name: Pien van Vliet
Student Number: S3805514
Address: Coehoornsingel 67B, 9711BP, Groningen
Phone Number: +31651625306
Acknowledgements
First of all, I would like to thank my first supervisor Martine van der Heide and my second supervisor Felix Eggers for their valuable contributions, feedback and support during this intensive and informative thesis trajectory. In addition, I would also like to thank all the participants who made this research possible by participating in the experiment. Finally, I would like to thank my family and friends for their support and commitment over the past challenging months.
Thank you very much, Pien van Vliet
Abstract
In today’s western world the “need for speed” affects us in a variety of ways. As a result, the effective distribution of time is becoming increasingly challenging for a substantial part of the population. Partakers in the food chain have anticipated this development by introducing high-quality nutrition products, generally referred to as convenience foods or meal solutions, to accommodate consumers and their tight schedules. However, although convenience foods seem to be the perfect solution to accommodate today's busy consumer, due to a generally negative attitude towards the phenomenon it has not yet fully realized its potential.
The aim of this study was to shed a light on the relative importance of a variety of product features – nutrition claims, freshness guarantee claims, price in combination with food convenience – and the tradeoffs consumers make between them when deciding upon various meal solutions. The distribution of a choice-based conjoint analysis among 163 participants provided insights to address this question. This revealed that consumers in this study generally preferred products that carried a nutrition or freshness guarantee claim. In terms of pricing, the overall sample preferred the cheapest alternatives. As far as convenience is concerned, one cannot speak of an universal effect throughout the sample. Therefore, a segmentation analysis, in which four segments are distinguished, offers a more refined explanation of this effect for different types of consumers.
Table of Contents
1. Introduction ... 7 2. Literature review ... 11 2.1 Food convenience ... 11 2.1.1 Cooking skills ... 15 2.2 Nutrition claims ... 162.3 Freshness guarantee claims ... 18
2.4 Price ... 18 2.5 Consumer characteristics ... 19 2.5.1 Convenience orientation ... 19 2.5.2 Health consciousness ... 20 2.5.3 Demographics ... 21 2.6 Conceptual model ... 21 3. Methodology ... 22 3.1 Research methods ... 22
3.2 Product Attributes and Levels ... 23
3.2.1 Food convenience ... 23
3.2.2 Nutrition claims ... 25
3.2.3 Freshness guarantee claims ... 26
3.2.4 Price ... 27 3.2.5 “No-choice” option ... 28 3.3 Consumer characteristics ... 30 3.3.1 Cooking skills ... 30 3.3.2 Convenience orientation ... 30 3.3.3 Health consciousness ... 30 3.3.4 Demographics ... 31 4. Results ... 32 4.1 Data preparation ... 32 4.1.1 Scale reliability ... 32 4.2 Descriptive statistics ... 33 4.3 Hypothesis testing ... 35 4.3.1 Main effects ... 35
4.3.2 Moderation effect cooking skills ... 39
4.4 Segment identification ... 41
5. Discussion ... 45
5.1 Main findings ... 45
5.1.1 Effects of the product attributes ... 45
5.1.2 Moderating effect cooking skills ... 47
5.1.3 Segment identification ... 47
5.2 Theoretical contribution ... 49
5.3 Managerial implications ... 49
6. Conclusion ... 53
References ... 54
Appendix ... 59
Appendix A: Consumer characteristics scales ... 59
Appendix B: Effect cooking skills on remaining attributes ... 61
Appendix C: Comparison latent class solutions ... 62
Appendix D: Output latent class ... 63
1. Introduction
In today’s western world the “need for speed” affects us in a variety of ways. For a substantial part of the population for example, students or full-time employees, the effective distribution of time is becoming increasingly challenging (Costa, Dekker, Beumer, Rombouts & Jongen, 2001). One of the areas where this leaves its footprints is in the consumption of food. Cooking is often considered as a hassle: a time-consuming event in people's already crammed schedules. Only for special occasions people see the meal as an event for which they are willing to invest more time and effort in order to cook a sophisticated meal (Sloan, 1997). As a result, consumer demands in terms of food products and the consumption of it has changed (Hirogaki, 2013). Partakers in the food chain have anticipated to this by developing high-quality nutrition products to accommodate consumers and their tight schedules (Larson, 1998). This is the beginning of a movement within the food industry better known as "convenience foods" or meal solutions. Convenience foods can be characterized as “products of the food industries in which the degree of culinary preparation has been carried out to an advanced stage and which are purchased as labor-saving versions of less highly processed products” (Scholliers, 2015, p. 4). As of today, convenience foods comprise a turnover of 700 million Euro in the Netherlands, which represent a 40 percent increase as compared to 2015 (IRI, 2018).
In addition, it is worth investigating whether and how price influences these tradeoffs and consumer valuations. This topic has received hardly any attention in the extensive amount of available literature on consumers’ food choices.
While there is ample of quantitative literature available on consumer food choices (Hirogaki, 2013), these studies often omit the "convenience" component as influential factor. Whereas the concept of convenience foods is heavily researched in a qualitative context (Costa et al., 2001; Costa, Schoolmeester, Dekker & Jongen, 2007; Neumark-Sztainer, Story, Perry & Casey, 1999), very little quantitative research has been done to provide these findings with statistical evidence. Therefore, this paper contributes to the existing literature by bridging the gap between these streams of research. In order to do so, the following attributes are taken into account: food convenience (Costa et al., 2007; Candel, 2001), nutrition claims (Hirogaki, 2013), freshness guarantee claims (Realini & Marcos, 2014) and price (Hirogaki, 2013; Costa et al., 2007). Eventually the following research question has been formulated:
“What is the relative importance of the product attributes food convenience, nutrition claims, freshness guarantee claims and price when deciding upon various meal solutions?”
Complementary to this main research question, this study also investigates the moderating role of the cooking skills of the consumer. This variable is believed to affect the relationship between the aforementioned product attributes and consumers’ food choices (Baron & Kenny, 1986). The literature identifies the possession of cooking skills as a predictor of one's food choices (Short, 2003; Van der Horst, Brunner & Siegrist, 2011). In this context, consumers who lack cooking skills often attach great value to the ease with which a meal can be prepared and are therefore likely to opt for convenience foods rather than cooking from scratch (Peltner & Thiele, 2018). However, what if consumers are prompted to decide between different meal solutions that all provide convenience to some extent? Will people with the lowest skills automatically prefer the highest level of convenience, or are moderate or even minimum convenience products, such as pre-cut ingredients or meal kits, already satisfactory to fulfil consumers’ need for convenience? Unfortunately, this question remains unanswered in the available literature. Therefore, this yields the following question:
Finally, various other consumer characteristics are taken into account to further explore if the effects could be generalized across different types of consumers or whether effects differ across specific segments. Apart from one’s cooking skills, consumers' food choices may vary depending on the extent to which a consumer is convenience oriented. For example, Candel (2001) demonstrates that different degrees of convenience orientation cause varying preferences, possibly resulting in divergent food choices. Consumers’ health consciousness is included as another personality trait, which allows to assess the generalizability of the effects. Consumers with varying levels of health consciousness are likely to have widely divergent views on the role of nutrition (Lee, Conklin, Cranage & Lee, 2014). This in turn might alter the relative importance of certain attributes in consumer’s product valuation (Mai & Hoffmann, 2012). Finally, a number of demographic characteristics are considered, as the literature also designate this as an underlying factor influencing consumers’ food choices (Candel, 2001; Brunner et al., 2010; Neumark-Sztainer et al., 1999). Since all consumer characteristics are person-specific, including these variables is an appropriate way to examine the third, and last, research question:
“To what extent can the effects be generalized across different types of consumers?”
Since these research questions aims to shed a light on consumer tradeoffs and preferences in the food context, a choice-based conjoint analysis is selected. The article of Green and Rao (1971) introduced this research method to the marketing context and this research tool has acquired an indispensable position in marketing literature ever since. Moreover, the literature labels this type of research as an appropriate, possibly even the most appropriate, method to uncover consumer preferences (Eggers & Sattler, 2011). In conjoint analyses consumers are exposed to a bundle of different components into which a product can be decomposed, also referred to as attributes (Eggers & Sattler, 2011). In the food context one can think of attributes such as price, country of origin, volume and the use of certain labels (Hirogaki, 2013). Each attribute consists of different levels (i.e., multiple options within the attribute). By continuously requesting the consumer to select one option, while systematically alternating the compositions in the choice set in terms of levels, the relative importance of each attribute could be quantified (Green, Krieger & Wind, 2001).
gathers attributes from a variety of both quantitative and qualitative studies, which creates a unique composition of attributes. Merging these streams of literature into one choice-based conjoint experiment investigates the relative importance for the previously investigated attributes. In addition, it enriches the literature by providing the relative importance of the previously unquantified convenience concept. From both a theoretical and practical perspective, this contribution allows for an enhanced understanding of consumer motives and behavior. These insights are essential for practitioners, as they yield implications that managers can leverage in their efforts to optimize both store assortment and the design of their products. Subsequently, managers could align their marketing strategies accordingly.
2. Literature review
This section aims to provide a comprehensive overview with regard to the most important concepts of this research. Finally, all these findings will be aggregated and visualized in a conceptual model.
If the food consumption process that consumers go through was represented by means of a rough sketch, it would be displayed as two consecutive decisions that a consumer has to decide upon. First they need to decide “what food to consume”, i.e., the food choice, followed by the decision “how much to consume”, also referred to as the consumption volume (Rafacz, 2019). In this paper the focus will be on the first decision.
Similar to decisions rendered in any contexts, a consumer's food choice can be influenced by multiple variables (French, 2003). In order to decide which attributes to include, existing literature has been consulted. For this research the following attributes are taken into account: food convenience (Costa et al., 2007; Candel, 2001), nutrition claims (Hirogaki, 2013), freshness guarantee claims (Realini & Marcos, 2014) and price (Hirogaki, 2013; Costa et al., 2007). As this research aggregates findings from a variety of both qualitative and quantitative articles into a new and therefore unique composition of attributes, it contributes to the existing literature. Below, each of these attributes are specified based on existing literature and hypotheses are established accordingly.
2.1 Food convenience
associated with convenience foods that are frequently highlighted in scientific articles are "time saving" and “effort minimization” (Candel, 2001, De Boer, McCarthy, Cowan & Ryan, 2004). Authors whose definition is adopted regularly in existing literature are Traub and Odland (1979), who specify convenience foods as “any fully or partially prepared foods in which significant preparation time, culinary skills, or energy inputs have been transferred from the home kitchen to the food processor and distributor” (p. 3). Given its acclaim, this definition will also be retained in this study.
Table 1: Convenience food classification systems Classification according to ... Levels Description Havlicek et al. (1983) Processing Non-convenience Basic convenience Complex convenience Manufactured convenience
Fresh and home processed foods Single component processed items
High time saving & culinary skills built-in No home-prepared counterpart
Pearson et al. (1985)
Service & Processing 3 (Service) x 2 (Processing) No preparation + convenience No preparation + non-convenience Some preparation + convenience Some preparation + non-convenience Considerable preparation + convenience Considerable preparation + non-convenience
Not provided.
Costa et al. (2001)
Service Ready to eat
Ready to heat Ready to end-cook Ready to cook
No preparation required Requiring only mild heating
Require sufficient heating to finalize Require full cooking
Brunner et al. (2010)
Processing Salads
Single components
Moderately processed food items Highly processed food items
(fruit) Salads suitable for direct consumption Separate already prepared meal components (part of) Meals with most processing already done Complete ready/instant meals
Peltner & Thiele (2018)
Processing Raw food ingredients
Industrially processed foods Ready-to-eat foods
Fresh and home processed foods
Where several authors propose different classification systems in their articles (e.g. Brunner et al., 2010; Costa et al., 2001), these more detailed attempts boil down to approximately the same configuration. This is consistent with the article of Pepper (1980), in which he argues that convenience foods can be classified along two interconnected dimensions: either based on the level of pre-assembly (i.e., processing) or the remaining amount of preparation activities (i.e., service). For example, less pre-processing automatically implies a larger number of actions that the consumer has to perform independently and vice versa. In this study, the frequently used system of Costa et al. (2001) will be adhered, as this study yields the most comprehensive, yet unambiguous, classification system by reviewing and drawing on a combination of existing systems (Harrison, 1979; Havlicek et al. 1983; Pearson et al., 1985).
Although convenience foods seem to be the perfect solution to accommodate today's busy consumer, due to a generally negative attitude towards the phenomenon it has not yet fully realized its potential. Among other things, this has to do with the fact that a large proportion of consumers does not consider convenience foods to be a fully-fledged meal (De Boer et al., 2004; Brunner et al., 2010) and prefers to cook a homemade dinner instead (Moisio, Arnould & Price, 2004). This idea is supported by the so-called "IKEA effect", which implies that consumers increasingly appreciate products if they have contributed to its production themselves (Norton et al., 2012; Dohle, Rall & Siegrist, 2013). In addition, convenience foods are often perceived as unhealthy and unnatural (Gofton, 1995; Brunner et al., 2010), and inextricably linked to an unfavorable image in terms of taste and nutritional value (Van der Horst et al., 2011). Lastly, convenience food is often considered as less fresh, as it has undergone some form of pre-processing, which also has a negative impact on the perceived quality (Carrigan et al., 2006). Although these associations may be somewhat dated considering the expansion and diversification of the convenience offerings, research shows that this is still the prevailing thought among the majority of consumers today (Brunner et al., 2010; Jackson & Viehoff, 2016; Meah & Jackson, 2017).
will lead to an increasing appreciation of this meal (Dohle et al., 2013). Based on these statements, the following hypothesis can be formulated:
H1: An increase in convenience, operationalized as a higher degree of pre-processing, has a negative effect on a food product’s utility.
2.1.1 Cooking skills
Complementary to the examination of the above hypothesized convenience effect, this study also investigates a moderating effect. A moderating, so-called "third", variable is believed to influence a relationship between an independent and a dependent variable (Baron & Kenny, 1986). More specifically, a moderating variable could reinforce or mitigate the relationship between these variables. Within this study, the moderating effect of cooking skills on the relationship between convenience and product utility is investigated.
In recent years, literature from across the globe addresses the evolution of consumer cooking practices and related skills (Short, 2003; Caraher, Dixon, Lang & Carr-Hill, 1999; Engler-Stringer, 2010). Where the majority of articles devoted to this topic often describes the meaning of cooking skills ambiguous or simply exemplifies it by highlighting specific techniques (Short, 2003), in this study the definition of Begley and Gallegos (2010) is adhered. Here, the authors define cooking as “the conversion of raw material into cooked food with the use of fire or heat” (p. 27), whereas cooking skills concern “the efficiency of food preparation and/or confidence with cooking” (p. 27). The introduction of new food trends and changing consumer demands is accompanied by a shift of the corresponding cooking skills. In the food context, the trends and demand developed from “the almost exclusive use of raw ingredients to that of being heavily dependent on processed foods” (Barton, Wrieden & Anderson, 2011, p. 588). As these developments require different, often less sophisticated skills, many researchers identify this movement as the cause of consumers’ deteriorating cooking abilities (Van der Horst et al., 2010; Hartmann, Dohle & Siegrist, 2013). However, considering the heterogeneity amongst individuals, this may not automatically apply to everyone, i.e., (hobby) chefs.
consumers more inclined to opt for convenience food products instead of cooking a home-cooked meal from raw materials (Ternier, 2010; Van der Horst et al., 2010; Reicks, Trofholz, Stang & Laska, 2014). This effect is confirmed by Hartmann et al. (2013), who establish a negative correlation between cooking skills and the consumption of convenience foods. On the contrary, the article by Soliah & Walter (2006) shows that individuals are more willing to prepare fresh meals if they master sufficient food preparation knowledge and skills. The fact that individuals on both extremes, i.e., possessing either low or high cooking skills, exhibit reverse behaviors as a result of opposing preferences, would suggest a disordinal interaction
effect. In other words, this would imply that cooking skills affect consumers’ preferences in
terms of convenience in different ways, potentially altering the relation fundamentally.
As the convenience effect is expected to be highly dependent on one’s cooking skills, the interpretation of the previously hypothesized (generic) main effect is less meaningful. Hence, this yields the following, refined, hypothesis:
H2: The effect of convenience on a food product’s utility is moderated by an individual’s cooking skills. Specifically, the effect of convenience on a food product’s utility is positive when cooking skills are low, but turns negative as cooking skills increase
2.2 Nutrition claims
Another important trend in today’s society is that consumers are increasingly concerned about their well-being (IRI, 2018). In addition to the widely known importance of sufficient exercise activity, consumers acknowledge that nutrition occupies an indelible position on the road to a healthier lifestyle (Croll et al., 2001; Caraher & Lang, 1999). As a result, wholesomeness is a highly-valued attribute driving consumers food choices (Hirogaki, 2013).
comply with the criteria of the Nutrition Labeling and Education Act (Kozup, Creyer & Burton, 2003). Within this scope of claims, a distinction could be made between health- and nutrition claims (Williams, 2005). The article of Dean et al. (2012) outlines this distinction as follows: “Health claims promise a health-related benefit, whereas nutrition claims state the presence of a nutritionally beneficial component or composition leaving the consumer to make the connection between the component and health” (p. 129). Health claims, in most cases, appeal to a specific group of consumers as they “address the relationship between a specific nutrient and a disease or health condition” (Kozup et al., 2003, p. 20). Nutrition claims are more generic, e.g., “low in fat” or “sugar free” (Williams, 2005), and therefore appealing to a larger and more diverse audience (Dean et al., 2012). To prevent the experiment from excluding a large group of potential participants, this study therefore chooses to focus on nutrition claims.
Various empirical studies examined consumer perceptions of health related claims (Dean et al., 2012; Keller et al., 1997) and the positive effect of such claims on consumers' food choices (Hirogaki, 2013; Kozup et al., 2003). For example, it has been found that the presence of health related claims positively influenced consumers' judgement with regard to the healthiness of the product (Roe, Levy & Derby, 1999; Williams, 2005). Moreover, research has shown that consumers evaluate products displaying health related claims on its packaging more favorably than products for which these health claims are omitted (Kozup et al., 2003). Building on this, it is not entirely surprising that products emphasizing favorable nutrition claims result in enhanced consumers’ attitude towards the brand, their nutrition attitude and their intention to purchase (Roe et al., 1999; Garretson & Burton, 2000). Altogether, it implies that the presence of health related claims on food products induce a halo effect (Roe et al., 1999). This psychological phenomenon suggests that consumers’ positive associations regarding a particular attribute, in this case presented through the nutrition labels, triggers a more favorable evaluation of the overall product.
for labels, including the Nutri-Score, providing simplified information with regard to a product’s healthiness by aggregating information from different nutrition components and express these by means of one indicator (Hamlin & McNeill, 2016). Taking all aforementioned information into consideration, the following hypothesis can be formulated:
H3: Nutrition claims have a positive effect on a food product’s utility
2.3 Freshness guarantee claims
The food industry shows an upward trend in the demand for fresh, minimally processed, food products (Realini & Marcos, 2014), especially among females (Ternier, 2010; Larson, Perry, Story & Neumark-Sztainer, 2006). Consumers indicate that they consider freshness as an important attribute when making food choices (Hussin et al., 2010), particularly because many consumers consider freshness to be an important food-quality indicator (Brunsø, Fjord & Grunert, 2002; Dai & Davison, 2010). Literature demonstrates that, from a consumer perspective, the attractiveness of products that display freshness guarantee claims is significantly higher, compared to food products where such claims are omitted (Darby, Batte, Ernst, Roe, 2008). A frequently mentioned explanation for this effect is that fresh products make consumers feel more healthy (MarketingTribune, 2015). In addition, consumers generally value foods containing any kind of "positive" claims higher than foods without this claim (Darby, Batte, Ernst & Roe, 2006), and in this respect the freshness guarantee claim does not constitute an exception (Darby et al., 2008; MarketingTribune, 2015). Altogether, this leads to the following hypothesis:
H4: Presenting claims that highlight the food’s freshness has a positive effect on a food product’s utility
2.4 Price
In addition, the price charged for a product should be perceived by consumers as a fair price. Consumers can determine this price fairness by comparing the price paid with its reference price: the perceived average price that the market charges for the product in question (Petrick, 2005). When considering price as a driver for choices, ceteris paribus, consumers are expected to prefer the alternative with the lowest price (Darby et al., 2008).
When translating this attribute to the domain of the study, the food context, these findings replicate here. Several studies show that price is an important, if not the most important, attribute in explaining consumers’ food related behavior (Candel, 2001; Brunner et al., 2010; Eertmans, Victoir, Vansant & Van der Bergh, 2005; Avitia, Costa-Font, Gil & Lusk, 2015). While the price elasticity of food products may vary depending on the category (Andreyeva, Long & Brownell, 2010), the meta-analysis of Bijmolt, Van Heerde and Pieters (2005) produced an average price elasticity of -2.62. This implies that an increase in price yields a shrinking demand. This is supported by virtually all empirical findings in the nutrition domain, where researchers repeatedly establish significant negative effects in terms of price (Darby et al., 2008; Brunner et al., 2010; Hirogaki, 2013). Besides verifying these previous theoretical findings regarding price, this attribute has a large practical importance. The inclusion of the price attribute allows for a discrete measure to quantify consumers’ willingness-to-pay (Larson, 1998; Darby et al., 2008). Based on the literature presented above, the following effect can be hypothesized:
H5: The product having a low price (vs. regular or high) has a positive effect on a food product’s utility
2.5 Consumer characteristics
Lastly, there are multiple articles which suggest that consumer characteristics shape the choices individuals make. This could imply that, considering the heterogeneous nature of mankind, the effects as outlined above are not universal. Therefore, a bundle of characteristics is included to discover whether distinct consumer segments, as a result of varying preferences, could be identified.
2.5.1 Convenience orientation
convenience (Candel, 2001). A combination of improving prosperity and the ever-growing challenge to effectively distribute our time, has led to a substantial group of consumers that classify as “convenience-oriented” (Anderson Jr, 1971). This construct was originally introduced to the nutrition context in Candel's article (2001), in which he states that convenience orientation, abbreviated CONVOR, can be specified as “the degree to which a consumer is inclined to save time and energy as regards meal preparation” (p. 17). Therefore, convenience-oriented consumers are typified as those who tend to exhibit time and energy saving patterns when it comes to their consumption of food products (Anderson Jr, 1971). The extent to which this orientation is developed is person-specific and therefore may vary widely across individuals (Candel, 2001). Hence, including convenience orientation in the experiment is an appropriate instrument to determine the generalizability of the effects.
2.5.2 Health consciousness
The concept of health consciousness can be specified as “the motivational component that stimulates consumers to undertake health actions” (Mai & Hoffmann, 2012, p. 317). Health conscious consumers are those who deliberately engage in behaviors that positively contribute to their wellbeing, both physical and mental (Michaelidou & Hassan, 2008). In their pursuit to the best quality of life, these individuals are likely to be committed to adequate physical activity together with a balanced diet (Kraft & Goodell, 1993). Therefore, these individuals attach great
2.5.3 Demographics
Finally, differences in demographics may be an underlying factor influencing consumers' food choices. For instance, Candel’s article (2001) establishes that household size, presence of children and employment status affect consumers’ food choices. Furthermore, variables such as gender and age also yielded significant results when assessing their impact (Brunner et al., 2010; Neumark-Sztainer et al., 1999; Reed, McIlveen‐Farley & Strugnell, 2003). For this reason, a variety of demographic characteristics will be included in the study.
2.6 Conceptual model
The main concepts for this study have now been defined and explained. Subsequently, the associated expected relationships, derived from existing literature, have been hypothesized. These hypotheses can be aggregated and visualized in a conceptual framework (Figure 1).
3. Methodology
This section will elaborate on the methodology applied in this study, in order to investigate the research questions. First of all, the research methods will be explained and reasoned. Second, a brief explanation of the experimental context will be presented. Afterwards, this chapter will elaborate on the composition of the different product attributes and its operationalization, which results in the development of the experimental design. Moreover, the measurement scales, which are used for the purpose of measuring consumer characteristics, will be discussed. Finally, the data collection procedure of the entire online experiment will be addressed and the targeted participants will be specified.
3.1 Research methods
In order to investigate the formulated research questions and test the aforementioned hypotheses for significance, a deductive research approach is suitable. Deductive research focuses on formulating hypotheses based on existing literature and theories, followed by testing these hypotheses in order to strengthen or criticize existing literature (Bitektine, 2008). Within the context of this research, the aim is to quantify the relative importance of different independent variables (i.e., product attributes) and a moderating variable (i.e., cooking skills) in the consumer's choice process. The influence of these product attributes is investigated by considering their effect on the dependent variable (i.e., product utility). Subsequently, this study explores to what extent these effects can be generalized across different types of consumers.
consumer to select their preferred option, while systematically alternating the compositions in the choice set in terms of levels, the relative importance of each attribute could be quantified (Green et al., 2001).
Second, the experiment yields information about participant characteristics through multiple survey questions. This data collection method is intended to generate input for the second objective of the study: determine the generalizability of the effects that the aforementioned attributes have on the consumer’s’ preferences.
3.2 Product Attributes and Levels
With CBC analysis, participants are prompted to select their preferred option out of a systematically varying set of alternatives. In this context, in order to prevent the participant from being inundated due to increasing complexity of the available information, it is generally recommended to limit the number of attributes to a maximum of six (Eggers & Sattler, 2011). Within this experiment, each alternative consists of a bundle of four attributes: food convenience, nutrition claims, freshness guarantee claims and price. All these attributes can be present in an alternative to varying degrees. These varying degrees are also defined as the levels of an attribute. Thus, all the alternatives presented to the participant will contain each of the four attributes, but they will be heterogeneous in the levels for each of these attributes. Below, each attribute and associated levels are described individually.
3.2.1 Food convenience
For the food convenience levels the classification system of Costa et al. (2001) was applied, resulting in the levels: ready to heat, ready to end-cook and ready to cook. The operationalization of these levels is based on existing products of a large Dutch supermarket chain. In this respect, the ready to heat level, i.e., the high convenience level, included a pre-made bucket of soup which simply needs to be heated before consumption. For the ready to
end-cook level a similar soup was shown, only now the last steps in the preparation process
In order to develop suitable stimulus material and prevent the creation of confounds, a number of precautions have been put in place. First, there is the possibility that the choice of participants is influenced by the fact that they prefer one dish over another, e.g. one might prefer tomato soup over pumpkin soup. To avoid this taste issue it was decided to fix the dish and only vary the degree of convenience. In addition, parts of the packaging that might inadvertently influence the choice of the participant, such as labels, volume and/or calories, have been concealed with the help of photoshop. Lastly, participants might be influenced by the differences in volume that the distinct products contain. To prevent this from happening, the situation sketch provided within the experiment explicitly states that the volume of each alternative is 750 milliliter. Altogether, these precautions ensure that possible changes in consumer preferences can be attributed to the different convenience levels and that confounding variables are eliminated. After these image-adjustments, the three different types of products, each representing one of the convenience levels, are displayed to the participant using the images below (Figure 2).
Figure 2: Stimulus material convenience attribute
Table 2: Results convenience pre-test
Food convenience Answer Ready to heat Answer Ready to end-cook
Answer Ready to cook
Ready to heat 94% (N=49) 2% (N=1) 4% (N=2)
Ready to end-cook 6% (N=3) 75% (N=39) 19% (N=10)
Ready to cook 2% (N=1) 12% (N=6) 86% (N=45)
In order to prevent the creation of confounds a number of precautions have been put in place. First, there is the possibility that the choice of participants is influenced by the fact that they prefer one dish over another, e.g. one might prefer tomato soup over pumpkin soup. To avoid this taste issue it was decided to fix the dish and only vary the degree of convenience. In addition, parts of the packaging that might inadvertently influence the choice of the participant, such as labels, volume and/or calories, have been concealed with the help of photo shop. Lastly, participants might be influenced by the differences in volume that the actual products contain. To prevent this from happening, the situation sketch provided within the experiment explicitly states that the volume of each alternative is 750 milliliter. By doing so, the effect is only caused by the convenience and confounds induced by other factors can be prevented.
3.2.2 Nutrition claims
As highlighted in the literature review, to prevent the experiment from excluding a large group of potential participants, this study chooses to focus on nutrition claims. More specifically, considering the varying effectiveness across FOP labels, it has been decided to select an
interpretive summary indicator label as these generally yield the most promising results.
experiment as it is an universally applicable label and therefore relevant for all adopted products.
In order to determine a Nutri-Score that is plausible for every possible composition of the alternatives, the scores of several equivalent products were consulted. This revealed that for similar products in the Belgian market, where displaying a Nutri-Score is mandatory for all products, an A-score was assigned (Delhaize, n.d.). This Nutri-Score applies to both the ready to cook and the ready to end-cook product, as well as the ready to heat product.
In the experiment this attribute comprises two levels: either the alternative was equipped with a nutrition claim (present) or this nutrition claim was omitted (not present). The image below (Figure 3) is used in the experiment to present the Nutri-Score of a product. By incorporating this product attribute in the experiment, it can be investigated whether and to what extent the presence of this nutritional claim affects consumer preferences.
Figure 3: Stimulus material nutrition claim attribute
3.2.3 Freshness guarantee claims
Figure 4: Stimulus material nutrition claim attribute
3.2.4 Price
Regarding the price, the attribute is divided over three levels: low, moderate and high price. To ensure that the prices mimic a realistic situation, actual prices from a large Dutch supermarket chain have been consulted. For the low level the, the maintained prices for (tomato) soup meal kits have been reviewed. In addition, the moderate and high price levels are derived from the prices that are charged in the marketplace for the ready to end-cook and ready to heat alternatives. However, the actual prices for all these convenience products were relatively similar, resulting in rather subtle differences in terms of price. To put it into perspective: the price for the meal kit is €3.39, whereas the other alternatives were only marginally more expensive (€3.59 and €3.69 respectively). To properly determine the WTP of the consumer, a slightly wider price range is preferred. Hence, the decision has been made to set a price range which deviates from the actual prices by approximately 10 percent.
Table 3: Overview selected attributes and levels
Attribute Levels Description
Food convenience Ready to heat Pre-made tomato soup Ready to end-cook Mash tomato soup Ready to cook Meal kit tomato soup
Nutrition claim Present Nutri-Score A
Not present -
Freshness guarantee claim
Present “Harvested yesterday” logo
Not present -
Price €2.99 Price ready to heat (-10%)
€3.49 Price ready to end-cook
€3.99 Price ready to cook (+10%)
3.2.5 “No-choice” option
Figure 5: Choice set example
In sum, a total of 3 (food convenience) x 2 (nutrition claim) x 2 (freshness guarantee claim) x 3 (price) = 36 different product compositions could be generated. With the CBC design as presented above, featuring three alternatives per choice set, this would be equivalent to 7,140 potential choice sets. Given the repetitive nature of the task and the limited cognitive capacity of consumers, asking participants to indicate their preference for all these choice sets would be unrealistic. Hence, a fractional factorial design is maintained. In order to still guarantee the effectiveness of the design, a variety of choice design efficiency criteria (balance, orthogonality, minimal overlap and non-dominating choice sets) have been carefully considered.
!!
#!(!%#)!
=
)*!
3.3 Consumer characteristics
After completing the CBC analysis, participants continued the survey by answering a number of questions about themselves. In this part, questions related to the hypothesized moderator in this study, cooking skills, were addressed first. Subsequently, questions regarding convenience orientation and health consciousness were presented. The final part of the survey concerned questions related to demographic characteristics. This section will elaborate on the utilized scales to measure all consumer characteristics. For the sake of completeness, an overview of all scales mentioned below is presented in Appendix A.
3.3.1 Cooking skills
This competence is measured through a scale developed by Hartmann, Dohle and Siegrist (2013). As these authors build upon several previously initiated scales (Van der Horst et al., 2010; Brunner et al., 2010), this is considered to be an appropriate scale to measure this concept. Here, individuals are asked to assess their own cooking competence by means of seven statements relating to a diverse set of skills. Examples include “I am able to prepare a hot meal without a recipe” or “I consider my cooking skills as sufficient”. Participants are asked to use a 6-point scale (ranging from 1 = “Does not agree at all” to 6 = “Agree very much”) to indicate the extent to which they recognize themselves in the presented statements.
3.3.2 Convenience orientation
In order to determine the extent to which participants’ are oriented towards convenience, the CONVOR scale of Candel (2001) was adopted. The final version of this scale comprises six statements with regard to time and energy saving. Participants were asked to express the extent to which they thought the statements applied to them by means of a 7-point Likert scale (ranging from 1 = “completely disagree” to 7 = “completely agree”). One of the items included in this scale states the following: “The less physical energy I need to prepare a meal, the better”.
3.3.3 Health consciousness
typical” to 5 = “very typical”. Some examples of such statements are “I’m very involved with my health” and “I reflect about my health a lot”.
3.3.4 Demographics
4. Results
This chapter contains the results of the conducted experiment. The final survey was distributed through a variety of social media channels and generated a total of 163 responses over a period of 14 days. Prior to reviewing the results of the performed analyses, the first section briefly addresses the actions performed for the purpose of data preparation. Subsequently, in the second part the sample is characterized by means of descriptive statistics. In the third part of this chapter, the hypothesized effects are tested for acceptance using the obtained data. The last part contains insights gathered from the more exploratory part of the study, in which the generalizability of the results is examined by means of a latent class analysis. In this way, existing consumer heterogeneity can be highlighted, as this allows for the identification of distinct consumer segments in terms of their preferences.
4.1 Data preparation
4.1.1 Scale reliability
In the experiment a variety of multi-item questions were included in order to measure different consumer characteristics: cooking skills (7-items), convenience orientation (6-items) and health consciousness (9-items). Although the study only utilized validated scales when considering these constructs, for the ensuing analyses it is still important to verify whether the items for each consumer characteristic construct are internally consistent. The literature generally refers to a construct as being internally consistent whenever the corresponding Cronbach's Alpha exceeds the threshold value of 0.70 (Gliem & Gliem, 2003).
4.2 Descriptive statistics
For the purpose of data familiarization, generating descriptive statistics serves as an appropriate tool to obtain a first, basic impression of the sample. An overview of the demographic outcomes is presented in Table 4. From this information it can be deduced that the majority of the sample comprises relatively young (mean = 36.37), either working (50.31%) or studying (34.97%), female (58.28%) participants. This distribution can be explained by the fact that the experiment is mainly spread within private social networks.
Table 4: Demographic characteristics of the sample
Demographic Answer options Frequency Percentage
Gender Male Female Other 68 95 - 41.72% 58.28% - Household size 1 2 3 4 > 5 13 55 32 31 32 7.98% 33.74% 19.63% 19.02% 19.63% Presence of children Yes No 43 120 26.38% 73.62%
Education level No formal qualification
Secondary education
Intermediate vocational education College Undergraduate degree Graduate degree Doctoral degree Other 1 22 11 45 21 56 5 2 0.61% 13.50% 6.75% 27.61% 12.88% 34.35% 3.07% 1.23% Employment status
Employed full-time (36+ hours a week)
Employed part-time (less than 36 hours a week) Unemployed Student Retired Self-employed Unable to work 59 23 3 57 4 17 - 36.20% 14.11% 1.84% 34.97% 2.45% 10.43% -
To supplement these generic demographic variables, the study incorporated a number of other consumer characteristic questions specific to the research context: cooking skills, convenience
orientation and health consciousness. Relevant descriptive statistics related to these included
the participants the lowest possible score of 1 was registered. Second, regarding convenience
orientation, the majority of participants (74%) scored between two and four out of seven,
making the average equal to 3.37. This would suggest that overall the sample is relatively neutral with respect to their orientation towards convenience. Finally, in terms of health
consciousness this construct yielded approximately similar results, again suggesting a fairly
neutral sample.
Table 5: Characteristics of the sample cont.
Min. Max. Median Mean Std. Deviation
Age 19 70 32 36.37 14.68
Cooking skills 2 6 5 4.74 1.02
Convenience orientation 1 7 3 3.37 1.35
Health consciousness 2 5 4 3.77 0.68
Furthermore, understanding the interrelationships of these variables is important before proceeding to the estimation of the requested models. A correlation matrix incorporating these aforementioned characteristics is presented in Table 6. At first glance, this matrix presents no particularities as it reveals moderate, yet logically explainable correlations.
Table 6: Correlation matrix
Age Cooking skills Convenience
orientation Health consciousness
4.3 Hypothesis testing
To test the aforementioned hypotheses for acceptance a variety of multinomial logit models were estimated. In this process, the random utility theory constitutes the foundation for the analyses. Building on this theory, the assumption is that an individual (n) will eventually opt for the one product (i) that produces the highest overall utility (Kraus, Mitter, Eggers & Stieg, 2017). This overall utility consist of a systematic (V) and stochastic (ε) utility component:
(1)
In turn, the systematic utility component represents the aggregation of part-worth utilities (β), that an individual (n) attaches to the attributes levels (x) out of which that specific product composition (i) can be decomposed. This yields the following equation:
(2)
4.3.1 Main effects
The first model comprises all product attributes considered in this study – product convenience (C), nutrition claim (N), freshness claim (F) and price (P) – and their effects on the choice of the consumer. Building on the forgoing information, this model could be formally defined by means of the following expression:
(3)
When analyzing output that such model generates a fixed sequence of steps could be pursued to assess its overall quality: 1) the explained variance in the dependent variable captured by the overall model, 2) the overall significance of the estimated model, and 3) the significance of individual parameters.
Subsequently, the quality of this established model can be determined by comparing its performance to that of the NULL model. A likelihood-ratio test comparing both models shows that the more advanced model performs significantly better than the NULL model (p < 0.001). From this, the conclusion can be drawn that this model fits the data significantly better and hence is more adequate in predicting consumers’ choices.
Finally, one can look at the contribution of each individual parameter. Since effect coding has been applied, for the sake of completeness, the reference levels of each attribute must first be restored. The output listed in Table 7 reveals that, with the exception of the “Medium Price” level (p > 0.05), all inserted independent variables have a significant effect on the preference of the consumer. The attributes showed a low correlation and none of the cases produced a significant interaction effect. Hence, only statements concerning the main effects are deemed to be acceptable for the purpose of model interpretation.
First of all, consumers seem to prefer food products with “Low” convenience (β = 0.519, p < 0.001), over products that offer a “Moderate” (β = -0.236, p < 0.001) or even “High” degree of convenience (β = -0.284, p < 0.001). To determine the underlying significance between the levels, a similar model has been developed in which the levels of these variables have been dummy coded. In the case of dummy coding the interpretation of the established effects is relative to the reference level, which allows to determine underlying significance. The ensuing significant positive effect for “Low” convenience products proves that consumers favor “Low” convenience products over the reference level, i.e. “High” convenience (β = 0.8027, p < 0.001). However, no significant effect is found with regard to the “Moderate” convenience level (β = 0.0479, p = 0.525), indicating no unambiguous preference for this level as compared to “High” convenience. With this knowledge, the established corresponding hypothesis (H1) may only be partially accepted.
In addition, the presence of a nutrition claim also appears to have a significant effect on consumers’ choices, as consumers seem to prefer products that display a nutritional claim (β = 0.609, p < 0.001). Hence, the second hypothesis may also be accepted.
The last main effect, which relates to the price attribute, only produces significant results for the “Low” (β = 0.318, p < 0.001) and “High” (β = -0.306, p < 0.001) price levels. As with the convenience attribute, dummy coding provides clarity about the effects and its significance relative to the reference level, i.e. “High” price. A positive, significant effect for both the "Low" price (β = 0.624, p < 0.001) and "Moderate" price level (β = 0.295, p < 0.001) demonstrates that consumers are more likely to prefer cheaper alternatives and thus confirms the hypothesized negative effect of price (H5). The utility function of price, as well as those of all aforementioned attributes, are graphically represented in Figure 6.
Table 7: Results Model 1
Coefficients Estimate Std. Error Z-value Sig.
Food convenience Low 0.519 0.0357 14.5410 0.0000 *** Moderate -0.236 0.0414 -5.6958 0.0000 *** High -0.284 0.0421 -6.7458 0.0000 *** Nutrition claim Present 0.609 0.0594 10.2475 0.0000 *** Not present -0.609 0.0594 -10.2475 0.0000 *** Freshness claim Present 0.391 0.0583 6.6931 0.0000 *** Not present -0.391 0.0583 -6.6931 0.0000 *** Price Low 0.318 0.0367 8.6502 0.0000 *** Medium -0.012 0.0392 -0.2945 0.7684 High -0.306 0.0420 -7.2857 0.0000 *** None option -1.182 0.1239 -9.5353 0.0000 *** p < 0.001 ‘***’, p < 0.01 ‘**’, p < 0.05 ‘*’, p < 0.1 ‘.’
Based on these estimates, the relative importance of each of the attributes on consumers’ choices has been determined. By looking at Model 1, it can be concluded that consumers consider the appearance of a nutrition claim as the most important feature upon which they base their choices (36%), closely followed by both the freshness guarantee claim (23%) and the food convenience (23%) attribute. Interestingly, according to this calculation, price is relatively speaking the least important feature (18%) on which consumers base their choice. Figure 7 visually reproduces this relative importance distribution.
Figure 7: Relative importance attributes Model 1
4.3.2 Moderation effect cooking skills
In this phase of the analyses, the previously established model is supplemented with an interaction effect to test for the moderating role of cooking skills. As previously substantiated, the expectation is that people with a lower level of cooking skills are more likely to prefer simpler, more convenient alternatives. To verify this, Model 3 is extended by including an interaction term between a participant's cooking skills (CS) and the food convenience attribute (C). This has the following consequences with respect to the previously established formula:
(4)
Once again, a likelihood-ratio test could be performed to assess the quality of this model. Since it has already been established that the more parsimonious Model 2 functions significantly better than the NULL model, inserting this model in the likelihood-ratio test would yield more meaningful insights for the purpose of quality comparison. The results of this test demonstrate that the differences in performance of both models are highly significant in favor of the extended Model 3 (LL = -1740.7, p < 0.001). Statistics such as enhanced Pseudo R² (R² = 0.2297), improved adjusted R2 (Adj. R² = 0.2257) and a lower AIC (AIC = 3497.442) all confirm this finding. Building on this, one can safely infer that this model provides a more adequate fit. Table 8 summarizes the output produced by the model.
In other words, participants who possess a more advanced level of cooking skills prefer products with low convenience and vice versa. These findings confirm the expectation, hence hypothesis 2 may be accepted.
Table 8: Results Model 3
Coefficients Estimate Std. Error Z-value Sig.
Food convenience Low 0.529 0.0370 14.2868 0.0000 *** Moderate -0.202 0.0423 -4.7833 0.0000 *** High -0.327 0.0446 -7.3318 0.0000 *** Nutrition claim Present 0.653 0.0612 10.6743 0.0000 *** Not present -0.653 0.0612 -10.6743 0.0000 *** Freshness claim Present 0.398 0.0598 6.6604 0.0000 *** Not present -0.398 0.0598 -6.6604 0.0000 *** Price Price.value -0.670 0.0696 -9.6327 0.0000 *** Cooking skills int.
CS * FC Low 0.366 0.0373 9.8259 0.0000 ***
CS * FC Moderate -0.020 0.0409 -0.4895 0.6245
CS & FC High -0.346 0.0419 -8.2578 0.0000 ***
None option -3.449 0.2629 -13.1195 0.0000 ***
p < 0.001 ‘***’, p < 0.01 ‘**’, p < 0.05 ‘*’, p < 0.1 ‘.’
For the sake of completeness, an additional multinomial logit model was estimated to determine whether cooking skills, besides convenience, would also act as a moderating factor in the effects of the remaining attributes. The results of this model (Appendix B) show that with regard to both the nutrition and the fresh guarantee claim, no significant effect is reported. Interestingly, significant interaction estimates for the “Low” (β = -0.095, p = 0.011) as well as the “High” (β = 0.111, p = 0.010) price parameters reveal that cooking skills, in addition to convenience, also serves as a moderator in the effects of the price attribute. This manifests in a way that consumers with above-average cooking skills are less price-sensitive, whereas consumers who possess less sophisticated cooking skills firmly prefer the cheaper alternatives. A visual representation of this moderation is presented in Appendix B.
4.4 Segment identification
This last section of the results sheds light on the more exploratory part of the study in order to assess the generalizability of the aforementioned effects. Considering the heterogeneous nature of individuals, these effects are assumed to vary for different types of consumers. A latent class analysis, incorporating a variety of characteristics as class predictors, will explore whether distinct consumer segments could be identified based on their preferences.
For this purpose the initial Model 1 was utilized to explore, by systematically alternating between the number of classes, which model composition yielded the best model fit. An overview of the compared solutions is presented in Appendix C. After estimating several latent class solutions, the four segment solution is considered to be the most appropriate as this delivered the best results according to the consulted information criteria (AIC = 2161.29, BIC = 2435.55). The selected latent class model, henceforth referred to as Model 4, comprises the main effects of all adopted attributes and exploits the consumer characteristics cooking skills,
convenience orientation, health consciousness, age, gender, household size, presence of children and education level in order to predict the class membership. These added consumer
characteristics contribute to a more comprehensive understanding what 'type' of consumer resides in which class. The output of this model can be consulted in Appendix D.
3 can be classified as the relatively smaller segments with relatively smaller class sizes of 16.30 and 14.53 percent respectively. Since the model produces estimates per class, it is feasible to discover segment specific effects. A visual representation of the utility functions per segment is presented in Appendix E.
Class 1: The Easy Eaters
Consumers who can be classified as members of Class 1 consider convenience to be the most important driver in making their food choices. This type of consumer typically tends to favor convenient food alternatives, as they are likely to prefer “High” convenience (β = 3.0495, p < 0.001) over the “Moderate” (β = -0.9477, p = 0.042) or “Low” option (β = -2.102, p = 0.001). Consumers who identify with this segment tend to favor the alternative with the lowest price (β = 1.0143, p < 0.001). Since both the nutrition- (p = 0.2206) and the freshness guarantee claim (p = 0.1934) produced highly insignificant effects, no conclusions can be drawn with regard to these attributes. In terms of consumer characteristics, this segment primarily consist of males, who are part of a smaller or even single-person household. Lastly, this convenience seeking segment is characterized by having, relatively speaking, lower cooking skills.
Class 2: The Healthy Heroes
Class 3: The Multi Masters
In the case of Class 3, the latent class model yields significant effects for all parameters. For instance, this segment holds a strong preference for the “Moderate” convenience alternative (β = 2.0019, p < 0.001) and generally opts for products featuring the lowest price (β = 0.6973, p < 0.001). Similarly to the foregoing segment, this segment chooses products that preferably contain a nutrition- (β = 0.9727, p < 0.001) and/or freshness guarantee claim (β = 0.8873, p = 0.001). Although only a marginally significant effect, the results reveal that this cluster of participants possesses high cooking skills (β = 0.1799, p = 0.099). In terms of direction and magnitude, this segment replicates the Healthy Heroes in terms of health consciousness, which again indicates a highly health conscious group of consumers (β = 1.2102, p < 0.001). Lastly, with regard to gender, males are likely to belong to this segment (β = -1.0453, p = 0.004).
Class 4: The Culinary Cooks
As with the previous analysis, the above estimates permit to determine the relative importance of each attribute. Figure 9 visually reproduces this relative importance distribution per segment.
5. Discussion
In this last chapter, the findings of this study are critically reviewed by juxtaposing and comparing it with the existing literature. Subsequently, a number of managerial implications are formulated accordingly. Finally, the limitations of this study, which constitute the basis for the further research suggestions, are addressed.
Table 9: Hypothesis overview
Hypothesis Result
H1: An increase in convenience, operationalized as a higher degree of
pre-processing, has a negative effect on a food product’s utility.
Accepted *
H2: The effect of convenience on a food product’s utility is moderated
by an individual’s cooking skills. Specifically, the effect of convenience on a food product’s utility is positive when cooking skills are low, but turns negative as cooking skills increase
Accepted
H3: Nutrition claims attached on food products have a positive effect
on a food product’s utility
Accepted
H4: Presenting claims that highlight the food’s freshness have a
positive effect on a food product’s utility
Accepted
H5: The product having a low price (vs. regular or high) has a positive
effect on a food product’s utility
Accepted
* Although formally statistical evidence has been obtained to accept the negative main effect of convenience (H1), yet considering hypothesis two (H2) its interpretation is less meaningful: the effect of convenience appears different for people with different levels of cooking skills.
5.1 Main findings
5.1.1 Effects of the product attributes
Within this study, the aim was to provide an answer to a threefold research question. The initial research question intended to shed light on the relative importance of the included attributes –
food convenience, nutrition claims, freshness guarantee claims and price – in consumers’ food
a positive effect for both the nutrition and the freshness guarantee claim confirmed the previously documented evidence with respect to their effectiveness (Ikonen et al., 2020; Kozup et al., 2003; Roe et al., 1999; Darby et al., 2006; Darby et al., 2008). Consumers significantly preferred products that carry such claims over products where these have been omitted. Moreover, with regard to the price attribute, the established negative effect is consistent with findings from previous research, i.e. consumers prefer the cheaper alternatives (Hirogaki, 2013; Darby et al., 2008; Brunner et al., 2010).
For the previously unquantified convenience attribute, the question arose as to whether more convenience is invariably preferred or whether moderate or even minimum convenience products are already satisfactory to fulfil consumers’ need for convenience? In this respect, a negative main effect is observed, which implies that overall consumers significantly preferred low convenience over the moderate and high convenience alternatives. Initially this does not seem to coincide with consumers' contemporary need for convenience as described in the existing literature (Costa et al., 2007; Neumark-Sztainer et al., 1999; Candel, 2001). Although this finding does match the line of reasoning of the previously described "IKEA" effect (Norton et al., 2012; Dohle et al., 2013), as well as consumers’ prevailing perceptions on what to consider as a fully-fledged meal (De Boer et al., 2004; Brunner et al., 2010). Yet, it should be acknowledged that the interpretation of this (generic) main effect in isolation is less meaningful, as it is highly dependent on one's cooking skills and consequently adopts fundamentally different shapes.
As far as the distribution in relative importance is concerned, this breakdown is as follows with regard to the overall sample: nutrition claims (36%), freshness guarantee claims (23%), food
convenience (23%) and price (18%). Interestingly, the product price, which the literature
5.1.2 Moderating effect cooking skills
The subsequent part of the analysis examined whether the negative main effect established for convenience may differ between several types of consumers. In this study, the moderating role of cooking skills has been considered in order to do so. The literature claims that food developments in recent decades, including convenience foods, have led to deteriorating cooking skills among consumers (Van der Horst et al., 2010; Hartmann et al., 2013). This in turn affects the food choices that consumers make (Barton et al., 2011). Yet, given the heterogeneity among consumers, it is plausible to expect that this might not apply to every individual to the same extent (Ternier, 2010) causing product preferences to vary. Consistent with this expectation, the findings of this analysis indicate a disordinal interaction effect. This is why, as emphasized above, the interpretation of the (generic) main effect for convenience is less meaningful in isolation, considering the fact that this relation looks fundamentally different as a consequence of varying cooking skills.
For low cooking skills a positive convenience effect can be discerned, as consumers that typify as such attach the highest utility to the more convenient alternatives. The reverse is found in the case of high cooking skills, as here consumers show a gradually more negative utility as alternatives offer more convenience. In this respect, the results suggest that the more advanced cooking skills an individual possesses, the lower their preference for convenience and vice versa. This is in line with the previously identified negative correlation of Hartmann, Dohle and Siegrist (2013) between cooking skills and the consumption of convenience foods.
5.1.3 Segment identification
Considering the heterogeneous nature of individuals, the preconceived effects are assumed to vary for different types of consumers. Hence, the last part of the analyses concerned a more exploratory research approach. By means of a latent class analysis, incorporating a variety of characteristics as class predictors, the study examined whether distinct consumer segments could be identified based on their preferences.
The results demonstrate that four segments can be distinguished within the sample of this study:
The Easy Eaters, The Healthy Heroes, The Multi Masters and The Culinary Cooks. With regard
