TO WASTE OR NOT TO WASTE
The influence of marketing activities and consumer characteristics on the households’ virtue food disposal behavior
By
RICK G. WORMGOOR
TO WASTE OR NOT TO WASTE
The influence of marketing activities and consumer characteristics on the households’ virtue food disposal behavior
Master Thesis
Msc Marketing, specialization Marketing Management and Marketing Intelligence University of Groningen, Faculty of Economics and Business
By RICK G. WORMGOOR Student number: 2503107 Koningsstraat 2A-2 8911 KV Leeuwarden Phone: +31 (0)610 440 382 E-mail: R.G.Wormgoor@student.rug.nl
Completion date: 12th January 2015
- 1 -
ABSTRACT
By using a longitudinal study approach, this study investigates the influence of marketing activities as well as consumer characteristics on the quantity of avoidable food waste. More precisely, this study is the first of its kind researching the effect of buying promoted products, the individuals’ health motivation as well as individuals’ shopping habits on the share of waste within the virtue category. With a food waste diary approach, the quantity and composition of waste for 23 food categories is measured and analyzed. Results show that the amount of products bought on promotion as well as the individuals shopping habits do not affect the share of virtue waste. However, two states of health motivation significantly influence the share of waste within the virtue category; namely amotivation with a negative and controlled motivation with a positive influence.
- 2 -
MANAGEMENT SUMMARY
The issue of food waste is getting a more serious problem all around the world. As stated by the Food and Agriculture Organization of the United Nations more than 30% of the worldwide produced food for consumers, which are 1.3 billion tons of eatable food, is lost or wasted (FAO 2013). Breaking down the focus on Europe, the disposed food adds up to over 90 million tons each year (European Commission 2014a). This waste of food has negative however, avertible consequences for the environment as well as the economy. Literature often suggests that the consumer is mainly responsible for the high amount of waste since, in categories like dairy products, they are responsible for up to 65% of the produced waste (Gustavsson et al. 2011).
Though, very little information is provided regarding influencing factors of consumers’ disposal behavior. By visualizing the enormous economic and environmental effects of food wasting, extensive marketing research is needed to gain more insights into consumers’ food wasting behavior. Literature research indicates the influence of promotional activities, health motivation as well as shopping habits on the quantity of waste produced by a household. This research delivers new and more detailed insights regarding influencing factors on the amount of wasted food by focusing not merely on the total quantity of avoidable food waste but concentrating on the share of waste within the virtue food category.
Consequently, the problem statement of this research is defined as the following: ‘To what extent do consumer characteristics and marketing influence the consumers’ share of wasted virtue food?’ In this study, marketing influence is defined as the frequency of buying promoted products while consumer characteristics is subdivided into the health motivation (motivation to eat a healthy diet) and the shopping habits (frequency of main- and top-up shopping).
- 3 -
Countering former expectations, the total amount of products bought on promotion over all categories (virtue, vice and neither) as well as separated by categories did not show a significant effect on the share of wasted virtue products. The same held true for the frequency of main shopping and top-up shopping defined as the shopping habit of the participants. By focusing on the individuals’ health motivation, significant effects were found for two of the three states; namely amotivation and controlled motivation. While the effect of controlled motivation showed the anticipated positive effect on the share of virtue waste, indicating a higher share of waste for persons with a higher level of controlled motivation, the influence of amotivation differed from the expectations. Households where the person mainly responsible for shopping and cooking had a rather amotivated state influences the share of virtue waste negatively; resulting in a lower share of virtue waste for respondents with a higher level of amotivation. The third motivational state, so called autonomous motivation, had no significant effect on the share of virtue waste.
Since this study is the first one investigating influencing factors of food wasting on a more detailed level, it is a first step to fill the gap within academic literature and helps to further understand and explain influencing factors on households’ food wasting behavior. Furthermore, several implications can be derived from this study: First, public policy makers have to find a way to motivate people to eat healthy without enhancing the level of controlled motivation since this leads to a higher share for virtue waste. A possible solution is to trigger the consumers’ autonomous motivation which indicated to have no significant effect on the share of virtue waste. In addition, the advertising industry which often tries to prescribe the importance of conducting a healthy diet has to be aware of the influence of their advertising on the consumers’ level of controlled motivation and should try to reduce this effect to decrease the share of virtue waste.
Overall it can be conclude that all firms or persons motivating individuals to conduct a healthy diet or way of living have to know the influence of the motivational states on the food disposal behavior. Subsequently, they all are able to take an active role in increasing public health and at the same time lowering the share of virtue waste.
- 4 -
PREFACE
This report is my master thesis for the conclusion of my Master program at the economics and business at the Rijksuniversiteit Gronignen. I joined the Master theses group with the topic ‘sustainability’ however, since the first meeting my scope shifted to a subject where less research is done yet; food waste. Like many participants in this study I thought that I was not wasting much food – but I was proven wrong.
Herewith, I want to thank all people who helped me to finalize this project:
Firstly, I would like to thank my supervisor Jenny van Doorn, who provided guidance throughout the whole process of finalizing this thesis and who I could contact and ask for help at all times.
I also want to extend my gratitude to my master thesis group for their constructive feedback, ideas and fruitful discussions. I especially want to highlight Katharina Bergmann who reviewed this paper. In addition, I want to thank Patrik Nowak for also reviewing this research and for the motivating phone calls where we counted down the days until the final deadline.
Moreover, many people helped me by conducting at my 14-day food waste study. Thank you for all your feedback and for providing me with the data. Without you it would not have been possible to write this thesis.
Someone who must not be forgotten is the dog Paulchen and his ability to disturb several times a day. He often forced me to get some fresh air which (sometimes) cleaned my mind.
Last but not least, I want to thank Cornelia Fröhlich for supporting and motivating me during the whole writing period and who calmed me down when I (again) felt overchallanged by the work which was still to do and I nearly lost my ability of positive thinking. You are a huge support.
- 5 -
TABLE OF CONTENTS
1. Introduction ... 7
2. Literature review ... 12
2.1. Drivers of food waste ... 12
2.2. Promotions ... 13 2.3. Health motivation ... 15 2.4. Shopping habits ... 17 3. Hypotheses ... 18 3.1. Conceptual model ... 18 3.2. Promotional activities ... 20 3.3. Health motivation ... 22 3.4. Shopping habits ... 23 3.5. Control variables ... 24 4. Methodology ... 27
4.1. Data collection and participants ... 27
4.2. Descriptives of the collected data ... 33
-- 6 --
- 7 -
1.
INTRODUCTION
Food waste is getting a more serious problem all around the world. As stated by the Food and Agriculture Organization of the United Nations more than 30% of the worldwide produced food for consumers, which are 1.3 billion tons of eatable food, is lost or wasted (FAO 2013). Breaking down the focus on Europe, the disposed food adds up to over 90 million tons each year (European Commission 2014a). Throwing away eatable food has an impact on the environment as well as on the economy. In the UK 6.2 billion cubic meters water and 20 million tons of carbon dioxide are used to produce avoidable food waste each year (WRAP 2011). Furthermore, 60% of the wasted food was still eatable and has a total value of nearly €16 billion (WRAP 2013). Broadening the focus from the UK to industrialized countries in general, more than 40% of the food waste is caused on a retail and consumer level while the remaining percentages are divided in cultivation, postharvest and processing the food. Furthermore, consumers are responsible for up to 65% of the wastage in some categories e.g., dairy products (Gustavsson et al. 2011). As possible reasons the European Commission (2014b) state a lack of knowledge at the customer level (e.g., in form of the non-existence of shopping plans, the missing knowledge of handling leftovers and a general lack of awareness about food wasting) as well as on the retailer level by applying unfavorable marketing strategies such as ‘buy one get one free’ (bogof) promotions. The European Union (EU) has recognized the potential of enhancing resource efficiency and food security on a global level and therefore plans to halve the eatable food wasted until the year 2020 (European Commission 2011). Yet, much is known about the process of buying products as groceries since this is a well investigated and documented subject in academic literature (e.g., Brunsø and Grunert 1998; Rook and Fisher 1995; Stern 1962; Yim et al. 2014). In contrast, very little information is provided regarding the consumers’ disposal behavior (Cappellini 200λ; Cappellini, and Parsons 2013; Koivupuro et al. 2012). A possible explanation for the limited consideration of the issue ‘food wasting’ is, as described in marketing literature, that it is the last phase of the consumption chain and therefore might not require enhanced attention (Cappellini 2009).
- 8 -
intention as well as the anxieties of consumers and subsequently is able to derive opportunities to reduce the volume of wasted food.
To be able to investigate influencing factors on food wasting behavior, it is central to have a shared understanding of what food waste is. In academic literature several definitions exist (e.g., Gustavsson et al. 2011; Koivupuro et al. 2012; Lyndhurst, Cox, and Downing 2007; Parfitt, Barthel, and Macnaughton 2010 or Silvennoinen, Katajajuuri, and Hartikainen 2014). However, in this study food waste is defined as:
All kind of eatable material and food which could be used for human consumption and could be avoidable of being disposed by a household.
This is in line with the definition used in preceding researches e.g., by Silvennoinen, Katajajuuri, and Hartikainen (2014), Koivupuro et al. (2012) and Parfitt, Barthel, and Macnaughton (2010) and excludes the unavoidable wasted products like vegetable peel, bones or coffee grounds form the definition of wasted products. Furthermore, the terms food waste and food disposal are used interchangeably in this study.
- 9 -
assigned into vice nor virtue category (Hui, Bradlow, and Fader 2009; Van Doorn and Verhoef 2014).
Besides the effect of a different food wasting behavior across vice and virtue categories this distinction is additionally valuable, since choosing one product over the other is influenced by several factors e.g., promotional activities (e.g., Mishra and Mishra 2011; Wertenbroch 1998), the individuals motivation (e.g., Milkman, Rogers, and Bazerman 2009; Mishra and Mishra 2011) as well as their shopping behavior (e.g., Doron 2013; Wertenbroch 1998).
By focusing on promotional activities, anecdotal evidences illustrate that the amount of wasted food is positively influenced by promotional activities (e.g., Lyndhurst, Cox, and Downing 2007; Parfitt, Barthel, and Macnaughton 2010). It is not surprising that the proportion of retail categories sold on promotion differ between countries (IRI 2014). However, the overall proportion of volume on promotions for the virtue nourishment are the categories with the lowest promotion volume (IRI 2014). In contrast, relative vice food categories like confectionary have the highest promotion volume of 61.3% (IRI 2014) but are responsible for a rather limited amount of food disposal with 1.8% (WRAP 2013). A more detailed comparison of categories and countries is illustrated in appendix A. Consequently, virtue categories with a higher amount of food waste show a lower promotional density compared to the vice category with a higher number of promotional activities albeit a lower amount of food waste. Yet, merely few research studies investigate the effect of promotions on food disposal behavior (e.g., Koivupuro et al. 2012; Lyndhurst, Cox, and Downing 2007; Parfitt, Barthel, and Macnaughton 2010). However, their results are not consistent. Lyndhurst, Cox, and Downing (2007) indicate promotions and consequently buying too much food as one of the key factors of enhanced food waste. Contrary, Koivupuro and her colleagues (2012) encountered a higher amount of wasted food for participants who did not regularly buy promoted food. Therefore, buying more nourishment on promotion indicates a higher price sensitivity which results in lower food waste (Koivupuro et al. 2012).
- 10 -
behavior stated by Evans (2011; 2012) leads to purchasing more virtue products but consequently results in an increased amount of wasted food (Graham-Rowe, Jessop, and Sparks 2014).
Concentrating on these upcoming discrepancies, a possible explanation may be the underlying motivation of the individuals. The self-determination theory (SDT) distinguishes between behaviors based on a state of amotivation or extrinsic motivation (Gagné and Deci 2005). Consequently, based on these different states of motivation a behavior is more or less likely to be performed (Gagné and Deci 2005). Yet, the influence of health motivation on individuals’ food wasting behavior has not been researched.
Besides the already discussed factors promotion and motivation, the effect of individuals’ shopping behavior, e.g., frequency of shopping or usage of a shopping list etc., is a frequent subject of research (e.g., Brunsø and Grunert 1998; Evans 2011; Thomas and Garland 2004). Certainly, few researches investigate the resulting influence on food waste (e.g., Lyndhurst, Cox, and Downing 2007) however, non focuses on the differences between vice and virtue wasting behavior.
Summarizing, this study does not limit the focus merely on customer characteristics itself but also takes marketing actions, specified as promotional activities, on groceries on the retailers’ side and their influence on food wasting behavior into account. Therefore, this paper seeks to investigate the influence of (1) retailers promotional activities, (2) individuals’ health motivation as well as (3) individual shopping habits on the share of waste within the virtue category. Based on these first insights the problem statement is addressed as the following:
To what extent do consumer characteristics and marketing influence the consumers’ share of wasted virtue food products?
Consequently, to be able to resolve this issue several research questions are formulated: 1. How do promotional activities influence the food disposal behavior regarding the
percentage of disposed food in virtue category?
2. How does personal health motivation influence the food disposal regarding the percentage of disposed food in the virtue category?
- 11 -
To the best of the authors knowledge, this is the first study which not only examines the influence of promotional activities or individuals shopping habits on general food disposal behavior, but investigates additionally the influence of promotional activities as well as health motivation and shopping behavior on the trashing behavior across virtue food categories (see table 1 for a comparison).
Table 1
Selective comparison of existent literature on influence on food waste behavior
Author(s) Promotion Shopping
Behavior
Health
Motivation Dependent variable Study design
Evans (2011; 2012) x Amount of wasted
food
Ethnographic study design
Cappellini and
Parsons (2013) x Reusing left-overs
Ethnographic study design; Interviews Koivupuro et al. 2012 x x Amount of wasted food
Food waste diary study; Questionnaire
Lyndhurst, Cox, and
Downing (2007) x x
Amount of wasted
food Interviews
Watson and Meah
(2013) x Consumer anxieties about food Ethnographic study design; interviews; observations
This study x x x Share of wasted
virtue food
Food waste diary; Questionnaire
- 12 -
disposal behavior and therefore, provides deeper insights in the opportunity of retailers taking an active role in reducing the amount of wasted food. Finally, this study is meant to raise the awareness, that each individual has an opportunity to reduce food disposal.
To investigate the research questions, a food/shopping diary approach is conducted. The research lasts for two weeks and is comparable to the study of Koivupuro et al. (2012) or WRAP (2013). In total 41 households participated in the study.
The structure of this paper is as follows: First, the paper starts with a literature review including the conceptual model and hypotheses in chapter two. Next, the third chapter explains the applied methodologies and procedures. Thereafter, chapter four presents the results followed by the conclusion and recommendations in the fifth chapter.
2. LITERATURE REVIEW
This section elaborates and discusses the already existent literature regarding food waste in general and its influencing factors, namely promotional activities, health motivation and consumers shopping characteristics.
2.1. Drivers of food waste
- 13 -
waste based on three main feelings: firstly, wasting money, followed by throwing away good food, and thirdly the feeling of general guilt. At this moment, the majority of research concludes that customers are more focused and engaged on their own economic motives as a result of disposing food (e.g., Graham-Rowe, Jessop, and Sparks 2014).
Even though the consumer is responsible for 50% (Gustavsson et al. 2011), or as other research suggests nearly 60% (Griffin, Sobal, and Lyson 2009) of the disposed food, the easiest way to affect a change in their disposal behavior is to directly appeal the consumer. Obviously, this hypothesis is supported by the fact that households often purchase too much food and consequently surplus food has to be disposed (Evans 2012). Nonetheless, taking just the consumer into account of handling the problem of food waste does not seem to be appropriate (Evans 2011). The discrepancy between anxiety and the real behavior are explainable by the problems of everyday life, e.g., the social context like family relations, time or taste within households (Evans 2011; 2012).
Anecdotal evidence (e.g., Lyndhurst, Cox, and Downing 2007; Milkman, Rogers, and Bazerman 2008; Mishra and Mishra 2011) indicate that the amount of wasted food is non-exclusive due to the social context. Instead, several other factors compound the quantity of trashed food in a household. Among others, Parfitt, Barthel, and Macnaughton (2010) investigate that retail promotions are accountable for at least half the unplanned purchases made in the store due to impulsive buying. As a consequence of the enhanced impulsive buying behavior the probability of disposing more food enhances as well (Lyndhurst, Cox, and Downing 2007).
Besides the influence of retailers’ promotional activities also consumer characteristics as motivation to eat healthy and shopping behavior affects the quantity of food waste. These influencing factors will be elaborated separately in the following.
2.2. Promotions
Several studies investigate the influence of promotions on consumers’ buying behavior for vice and virtue products (e.g., Mishra and Mishra 2011; Parreño-Selva, Mas-Ruiz, and Ruiz-Conde 2014; Wertenbroch 1998) as well as the influence of promotional activities on food wasting behavior (e.g., Evans, Campbell, and Murcott 2013; Lyndhurst, Cox, and Downing 2007, Parfitt, Barthel, and Macnaughton 2010).
- 14 -
one free’). The research indicated that only direct price discounts seem to have an influence on self-reported buying behavior of customers. According to the research of Pauwels, Hanssens, and Siddarth (2002) promotional activities have a rather limited influence on sales. The data of their study indicate that more than 80% of the sales are not related to promotional activities (Pauwels, Hanssens, and Siddarth 2002). Furthermore, Pauwels and colleagues (2002) tested a long-term equivalent to Gupta’s (1λ88) 14/84/2 breakdown (focusing on purchase time acceleration, brand switching and stockpiling) caused by promotional activities. A variation to the study of Gupta (1988) is the distinction between perishable (yogurt) as well as storable (soup) products. For both categories the breakdown shifted to an enhanced effect of category incidence (66% for non-perishable products, 58% for non-perishable ones), brand switching occurs less often with a stronger effect for perishable products (39%) compared to storable products (11%). Increasing sales due to stockpiling are rather limited for perishable products (3%), however enhance for stackable products (23%).
Further research suggests different reasons for stockpiling for perishable and non-perishable product categories. Both categories illustrate an increased category expansion due to promotional activities (Ailawadi et al. 2007). The main reason for enlarging the category of perishable products, focusing on yogurt, is increased consumption and brand switching accounting for 90% of the expansion. Ailawadi et al. (2007) explain these findings due to the nature of the products since yogurt has a rather limited best before date. Compared to ketchup, a more non-perishable product, Ailawadi et al. (2007) indicate a smaller increase of consumption than for yogurt. Furthermore, the effect of preemptive switch as well as for loyalty acceleration is larger for stockable products compared to perishable ones.
Nijs et al. (2001) state that price promotions have a strong short-term effect on category demand, in other words on the expansion of the product category. The more often retailers use price promotions the sensitivity of consumers regarding these promotional actions increase. This category short-run enlarging effect is larger for perishable products compared to non-perishable ones. Nijs and his colleagues (2001) explain this by an increased usage of perishable products to prepare meals when these products are on promotion and therewith support the findings of Ailawadi and Neslin (1998).
- 15 -
expansion of the subcategory (Parreño-Selva, Mas-Ruiz, and Ruiz-Conde 2014) which supports the findings of a category expansion for perishable products as discussed above (Nijs et al. 2001). Focusing on promotional activities for vice products, research indicates a direct effect on the promoted product (Parreño-Selva, Mas-Ruiz, and Ruiz-Conde 2014). Therefore, a price discount leads to more sales of this particular product. Furthermore, a price-reduction approach is more preferred for this category since consumers have a feeling of guilt when consuming an increased amount of vice products. Nevertheless, the price discount is perceived as a valid reason to purchase a product, reduces the feeling of guilt and is a tradeoff for a self-controlled behavior like abandonment of the product (Mishra and Mishra 2011; Wertenbroch 1998). Consequently, price discounts help people to solve an inner conflict by saving money and therefore, only consuming a limited amount of vice food (Mishra and Mishra 2011; Wertenbroch 1998). This behavior is known as self-rationing proposing that consumers prevent themselves from overconsumption but permit a certain level of immediate pleasure created by the consumption of vice products (Wertenbroch 1998). This inner conflict is not existent for virtue products as consumers do not feel guilty for consuming healthy virtue food (Mishra and Mishra 2011).
Focusing on the influence of promotion on wasting food, Lyndhurst, Cox, and Downing (2007) indicate that promotional activities motivate consumers to purchase more food than needed as well as buying other unneeded products due to an enhanced probability of impulsive buying. Both is resulting in a higher chance that not all products can be consumed within the best-before date or before it gets moldy and thus needs to be trashed (Lyndhurst, Cox, and Downing 2007). Therefore, Lyndhurst, Cox, and Downing (2007) indicate that promotional activities are mainly responsible tempting consumers of buying too much food. These findings are supported by Parfitt, Barthel, and Macnaughton (2010) stating that half of the unintended, impulsive purchases are due to promotions. Subsequently, promotional activities are one of the main contributors to the quantity of wasted food.
2.3. Health motivation
- 16 -
However, as researched by Milkman, Rogers, and Bazerman (2009; 2008) people differ in their virtue and vice purchase behavior depending on the delay between purchase and consumption. With an increased distance between the planning and the actual behavior decisions are more influenced by the, as the authors named it, should self, which concentrates on the long term consequences of certain behavior. In comparison, decisions with a more immediate effect are influenced by the want self and try to maximize the short term pleasure. In other words, more immediate effects result in a higher amount of spending, more impulsive buying and purchase of an enhanced number of vice products and vice versa (Milkman, Rogers, and Bazerman 2009). As a consequence of the enhanced impulsive buying behavior the probability of disposing more food enhances (Lyndhurst, Cox, and Downing 2007).
Besides the distinction between the individuals should and want self’s and the arising tension between them, motivational discrepancies occur. Evans (2011; 2012) states a deviation between the willingness to eat healthy and the actual behavior in real life which is caused by the social context. By having an intention of eating healthy does not imply that this behavior is conducted. Still, consumers are improving their eating habits to a more healthy diet (Capacci and Mazzochi 2011). However, changing the diet to a healthier one is accompanied with buying more virtue products, with a more limited durability, to be able to serve proper meals (Evans 2011; 2012). Furthermore, these proper meals often need more time to be prepared which is a reason to break the intended behavior since habituated routines are often overruled by other situations of real life (Evans 2011; 2012). Consequently, buying more virtue products with a limited durability together with the fact that the intended behavior is not conducted (the discrepancy between willingness to eat healthy and the actual behavior) are factors which increases the quantity of disposes food (Evans 2011; 2012).
- 17 -
motivated goals increases the feeling of well-being (Ryan, Huta, and Deci 2008). Albeit, the influence of different motivational states on individuals’ food wasting behavior has not been researched yet.
2.4. Shopping habits
Besides the influence of the external factor promotional activities on the food wasting behavior as well as the motivation for eating healthier also individuals shopping behavior influences the amount of wasted food since consumers make about 70% of purchasing decisions in store (Vermeir and Van Kenhove 2005). Stilley, Inman, and Wakefield (2010) indicate that consumers have reserved a part of their mental budget for in-store decisions which are not planned beforehand. Consumers have a high probability to spend this money preferably on unhealthy and nonessential items (Stilley, Inman, and Wakefield 2010).
These in-store decisions are mainly based on impulsive buying decisions which have a higher probability to result in purchasing vice products (Wertenbroch 1998). These findings are in line with research of Milkman, Rogers, and Bazerman (2008) which indicates that actions with a more immediate outcome result in an increased level of impulsivity and purchasing a higher quantity of vice products. Subsequent research by Milkman, Rogers, and Bazerman (2009) stresses the differences between purchased amount of vice and virtue products depending on the time difference between ordering and delivering even more. Ordering groceries in an online store with a delivery delay between two and five days results in decreasing spending together with an increased percentage of virtue products and decreasing percentage of vice products with each additional day between ordering and delivering. Orderings one day before delivery show a higher amount of virtue products, however, this effect is not significant (Milkman, Rogers, and Bazerman 2009). In other words, more immediate effects result in a higher amount of spending, more impulsive buying and purchase of an enhanced number of vice products and vice versa (Milkman, Rogers, and Bazerman 2009). As a result, shopping with a higher frequency results in an increased impulsive buying and increased percentage of vice category shopping. Resulting from the higher level of impulsiveness, Lyndhurst, Cox, and Downing (2007) indicate as a consequence thereof a higher quantity of disposed food.
- 18 -
the highest quantity of disposed food. This finding is confirmed by Graham-Rowe, Jessop, and Sparks (2014) who explain that shopping less often results in a higher volume of stockpiling perishable products which have a higher probability to decompose and as a consequence get trashed.
An additional shopping behavior which influences food waste is preplanning the meals (e.g., Doron 2013; Lyndhurst, Cox, and Downing 2007; Quested et al. 2013) which is a prerequisite to prepare a shopping list. Quest and his colleagues (2013) investigate that the usage of a shopping list has a positive reducing influence on food waste. Continuative research by Doron (2013) supports this finding and implies that a lack of knowledge in meal planning is a major driver for food waste. In other words, being able to preplan meals and to prepare a shopping list reduces impulsive buying and consequently reduces food waste (Doron 2013). However, food waste is not reduced only by having a shopping list. In the study of Parfitt, Barthel, and Macnaughton (2010) more than 50% of the respondents admit not sticking to the shopping list always or sometimes.
3. HYPOTHESES
In the subsequent sections, the conceptual model is presented and hypotheses regarding the influence of promotional activities, health motivation and shopping habits on the share of wasted food within the virtue category are elaborated. Furthermore, the effect of sociodemographics as well as the effect of using shopping lists are discussed.
3.1. Conceptual model
- 19 -
Figure 1: Conceptual Model
- 20 - Table 2
Hypothesized signs of parameters
Variables Expected
sign Rationale Source
Promotional Activities
(vice category)
+
Promotional activities for vice, virtue and neither product categories positively affect food waste for virtue category. Stronger effect for virtue promotions.
e.g., Lyndhurst, Cox, and Downing (2007);
Mishra and Mishra (2011); Nijs et al. (2001)
(virtue category)
(neither category)
Health motivation
(amotivation) + no intention of eating healthy, higher
waste of virtue products e.g., Koestner et al. (2008);
Vansteenkiste, Lens, and Deci (2006);
Williams et al. (1996)
(autonomous) - lower probability of breaking intended
behavior, lower waste of virtue products
(controlled) + Higher probability of breaking intended
behavior, higher waste of virtue products
Shopping behavior
(list usage)
- Lower probability of impulse buying
behavior, lower waste of virtue products
e.g., Milkman, Rogers, and Bazerman (2009);
Stilley, Inman, and Wakefield (2010); Wertenbroch (1998) (higher frequency)
3.2. Promotional activities
- 21 -
instruments like price and promotional activities with which the retailers try to influence consumers’ behavior (Vermeir and Van Kenhove 2005) and mainly results in impulsive buying of products of the vice category (Milkman, Rogers, and Bazerman 2009).
Furthermore, Mishra and Mishra (2011) and Wertenbroch (1998) highlight the importance of promotions on the purchase quantity of vice and virtue products. However, the marketing instrument preferred by the consumer differs for both product categories. On the one hand, price discounts are more preferred for vice products since it reduces the feeling of guilt which arise by consuming vice products and therefore enhances the probability of impulsive buying of products within this category (Mishra and Mishra 2011; Wertenbroch 1998). On the other hand, bonus packs are the preferred promotion within products of the virtue category (Mishra and Mishra 2011). Therefore, independently of the specific preferred promotional activity within the categories, the quantity of the purchased products is enhanced throughout. Overall the influence of promotional activities on the quantity of waste are supported by Lyndhurst, Cox, and Downing (2007) and Evans, Campbell, and Murcott (2013).
Focusing on the influence of promotions on wasting effects within vice and virtue categories two assumptions have to be elaborated. On the one hand, Nijs and his colleagues (2001) assume that promotional activities increase the purchase of perishable products to prepare proper meals which would result in a higher quantity of wasting vice products. On the other hand, Evans (2011; 2012) argue that there is an intention to prepare healthy meals, however, due to the social context there is a discrepancy between buying promoted virtue products and using these ingredients to prepare a healthy meal. This discrepancy leads to an enhanced wasting effect within the virtue products caused by promotions for vice and virtue product categories.
- 22 -
H1 (a): Retailers’ promotional activities on virtue, vice and neither products positively
influence the share of disposed virtue food in a household.
H1 (b): The share of disposed food due to promotional activities is higher for virtue
products compared to vice products.
3.3. Health motivation
Issues regarding health, which includes the intention to eat healthy, are often described as processes with autonomous goals, since it directly effects and satisfies basic psychological needs (Vansteenkiste, Lens and Deci 2006). This is supported by Lindsay (2007), who reasons that eating healthy is a responsibility and moral obligation.
The self-determination theory reasons that different state of motivation influences individuals’ behavior to be effective or to behave in a healthy way (Vansteenkiste, Lens, and Deci 2006; Williams et al. 1996). According to the SDT individuals’ motivation is divided into a state of amotivation or extrinsic motivation. The later state is subdivided into controlled- or autonomous motivation. Stating that an individual has a responsibility or obligation to behave in a certain way are described as actions with a rather controlled motivation (Williams et al. 1996). Depending on the underlying motivation people are more or less likely to perform certain actions (Williams et al. 1996). Within the state amotivation people do not show any motivation to conduct a certain behavior (Deci and Ryan 2000; Ryan and Deci 2000) whereas controlled motivation is often based on external rewards like salary. Therefore controlled motivation is a contrast to autonomous motivation which is characterized by its internalization and integration into the consumers self, e.g., having internalized the objective to live healthy (Deci and Ryan 2000; Ryan and Deci 2000). Autonomous motivated decisions result in enhanced performance, higher self-esteem and increased excitement as well as creativity (Ryan and Deci 2000). In addition, people which base actions on autonomous motivation experience a higher degree of well-being and increased social function which results in an overall improved psychological health. Behaviors based on controlled motivation show no such relation (Ryan and Deci 2000; Ryan, Huta, and Deci 2008).
- 23 -
behavior in the future. Thus, people with an increased degree of controlled motivation of following a healthier lifestyle are more likely to break with their intention in maintaining a healthy behavior.
These findings are supported by Koestner et al. (2008) who state that autonomous motivation is connected to the pursuit of personal goals. Furthermore, the relation between autonomous motivation and fulfilling personal goals often comes together with the planning of implementations (Koestner et al. 2008). Where autonomous motivated participants are thinking about the quality of food and do eat healthier because they include (and plan for) healthy food in their eating habits, whereas more controlled motivated participants were less concerned about the quality but rather concerned about the quantity (Koestner et al. 2008).
These findings are translatable to this study, since the intention to eat healthier is motivated either autonomous or more controlled. As a consequence, behavior based on controlled motivation has a higher probability of falling back into consumption of vice products even though the intention was to consume more virtue products. Therefore, it is to expect that autonomous motivated people have a higher probability of using the virtue products they bought and as a consequence waste less compared to controlled motivated people who might waste more virtue products. In addition, amotivated participants have no intention to conduct a healthy behavior and therefore have an enhanced probability of wasting virtue nourishment. Based on these argumentations, the hypothesis is as follows:
H2: Share of waste within virtue category is influenced by an inverted U-shaped effect based
on degree of motivation to conduct a healthy behavior; resulting in a high share of waste for amotivated compared to lower share of waste for autonomous motivated and higher share of waste for controlled motivated individuals.
3.4. Shopping habits
- 24 -
shopping for groceries result in a lower amount of wasted food compared to a lower shopping frequency.
In addition, referring to the study of Milkman, Rogers and Bazerman (2009) one can conclude that the smaller the gap between planning a behavior and conducting it, the more vice products are purchased. Consequently, a higher frequency of purchasing nourishment results in buying more vice products. In other words, more immediate effects result in a higher amount of spending, more impulsive buying and purchase of an enhanced number of vice products and vice versa (Milkman, Rogers, and Bazerman 2009). Furthermore, research by Wertenbroch (1998) and Stilley, Inman, and Wakefield (2010) indicate a higher probability of purchasing vice products on impulse. Hence, by increasing the frequency of grocery shopping consumers face a trade-off between being able to better plan their shopping and the enhanced risk of impulse buying where vice products may replace virtue ones. Since previous research indicated a negative impact of the shopping frequency it can be argued that better planning outweigh the risk of impulsive buying. Therefore, it can be hypothesized:
H3: The frequency of shopping for groceries negatively influences the share of disposed
food in the virtue category.
3.5. Control variables
In this study age, gender, income, size of household and the usage of shopping lists are included as control variables. Research illustrate that household size, age, income, the gender of the main buyer of groceries as well as the usage of shopping lists influences the amount of food wasted in a household. The control variables will be discussed in the following. Even though, prior research focuses on the total amount of food, it is expected that the results are similar by investigating the effects on the share of waste for virtue products.
Household size
- 25 -
Macnaughton (2010) stating that the composition of the household has a significant influence on the food disposal behavior. Besides the number of persons in the household Parfitt (2010) also distinguishes between children who waste less food than adults. Though, households with children tend to waste more than households without children (Parfitt, Barthel, and Macnaughton 2010) which can be argued that children in general dislike more food compared to grownups (Lyndhurst, Cox, and Downing 2007). Therefore, the household size negatively influences the amount of disposed food within the virtue category.
Income
By focusing on income results of research are contradictory. According to Mitchell and Walsh (2004) a higher income leads to an increased impulsive buying behavior, since these individuals are able to compensate financial losses. As mentioned before, impulsive buying is often related to promotional activities and as a result ends in an increased food wasting behavior (Lyndhurst, Cox, and Downing 2007). This finding is supported by Parfitt, Barthel, and Macnaughton (2010) who state that lower income results in a lower amount of wasted food and vice versa. Though, other data does not find a correlation between income and food disposal behavior (Wenlock, Buss, and Derry 1980). Notwithstanding, due to the findings of Mitchell and Walsh (2004), Lyndhurst, Cox, and Downing (2007) and Parfitt, Barthel, and Macnaughton (2010) income is assumed to positively influence the amount of disposed food in households.
Age
Regarding the influence of age, Lyndhurst, Cox, and Downing (2007) claim that consumers between the age of 25 and 34 years waste the most food. As the people get older they seem to waste less food and from this it follows that the generation of 65 years and older waste the least amount of food (Lyndhurst, Cox, and Downing 2007). This results in a negatively influence of age on the amount of disposed food in households.
Gender
- 26 -
disposed food is increased compared to households where men or both are responsible for buying groceries (Koivupuro et al. 2012). Furthermore, research intents that men have a higher probability to purchase new clothes during the sale compared to women (Mitchell and Walsh 2004). However, since the fashion sector seems not be comparable to grocery shopping the authors state that the underlying general consumer traits do not change across categories. Therefore, it implies that these traits are stable across different contexts which also includes shopping for nourishment (Mitchell and Walsh 2004). In addition, Mitchell and Walsh (2004) indicate that males are more affected by promotional activities. On the other hand, Silvennoinen et al. (2014) argue that women have a deeper understanding of healthier living and therefore, purchase more virtue products. In addition, research suggests that women are more likely to buy too much food rather than not having enough later. Consequently, females have a positive influence on the amount of disposed food in households (Silvennoinen et al. 2014).
Based on the contrary findings of the influence of gender on food wasting behavior this paper follows the argumentation of Koivupuro et al. (2012) and Silvennoinen et al. (2014) who indicate that females are wasting more food, even though males are more affected by promotional activities (Mitchell and Walsh 2004). This results in an increased disposal behavior for women.
Usage of shopping lists
- 27 -
Now the theoretical framework is set, the next section briefly illustrate the applied methodology of this paper whereupon the results will be discussed.
4. METHODOLOGY
4.1. Data collection and participants
To be able to answer the research question the data collection was done by two different approaches: (1) A food waste diary approach and (2) an additional background questionnaire to obtain additional information about the participants. To minimize eventually occurring language barriers on the one hand and to extend the possible number of participants on the other, the food waste diary and the questionnaire were translated into three languages: English, German and Dutch. The data collection took place between October and December 2014 which enlarged the probability that most participants are not on vacations and stayed home during the research period since the study was carried out between the main summer and winter holidays.
After the data collection personal information like name and address are deleted permanently from the dataset and participants are only identifiable by their unique number. In the subsequent, all analysis are performed with an anonymous dataset.
Food waste diary
The food waste diary approach is similar to the method used by Koivupuro et al. (2012), Silvennoinen, Katajajuuri, and Hartikainen (2014) or in the ‘FoodBattle’ conducted by Bos-Browers et al. (2013) and is subdivided into four sections. The first two parts measure the frequency of dining in a restaurant / ordered Take-away food as well as the frequency of the individuals shopping habits (main shopping and top-up shopping). Furthermore, the amount and specification of purchased products on promotion (price discount and bogof) are collected. The last section measures the quantity of wasted products in the participants’ household.
- 28 -
categories and an additional others category to specify products which cannot be assigned to one of the labeled categories. According to Van Doorn and Verhoef (2014) these are the categories mainly purchased food categories which generate about 80% of all food purchases in the Netherlands. The amount of waste had to be weighed or estimated by the participants themselves each time they throw a product away. As a guidance some weight examples were provided similar to the study of Bos-Browers et al. (2013). Though, the focus is set on the waste of avoidable food waste which excludes the unavoidable wasted products like vegetable peel, bones or coffee grounds from the definition of wasted products. Besides the information about the quantity of waste, also data about the frequency of eating in a restaurant or purchasing Take-away meals, the rate of main shopping as well as top-up shopping behavior and finally the frequency and specification of purchasing promoted products (price discount or bogof) was collected. The participants were asked to record the mentioned factors for a sequential two weeks (14 days) period.
As stated by Silvennoinen, Katajajuuri, and Hartikainen (2014) there is no possible to “evaluate the accuracy and truthfulness of the diaries” (p.1061) since participants weighed the products without supervision. Furthermore, raising the awareness of food waste and weighing disposed food may already have a reducing influence on the produced food waste (Silvennoinen, Katajajuuri, and Hartikainen 2014).
Table 3
Subdivision of food categories a
Subdivision Food category
Virtue category Bread, Cereals, Dairy products, Eggs, Fruit, Juice, Soup, Take-away food,
Vegetables
Vice category Alcoholic drinks, Cheese, Chocolate, Cookies, Crisps, Nuts, Soft drinks,
Sweets
Neither category Chicken/Fish/Meat, Coffee/Tee, Dressings/Toppings, Meat products,
Rice/Pasta
a based on Hui, Bradlow, Fader (2009); Van Doorn and Verhoef (2014).
- 29 -
Verhoef (2014). In total ten products are categorized as relative virtue, eight as relative vice and five as neither the one nor the other. A detailed overview of the categories with all included products is illustrated in table 3.
Besides the measured total quantity of waste for a household and the quantity of waste for each category the households’ share of waste (SOW) for virtue, vice and neither products is calculated. The SOW for virtue products describes the quantity of wasted virtue products in household i during period t, relative to the total quantity of disposed products of household i in period t:
��� � �,� = � �� � ��� �,�� , (1)
where SOWVirtue,it describes the share of wasted virtue products of a household i in the time period t; WasteVirtue,it defines the quantity of wasted virtue food for a household i in period t. Wasteit is the
total amount of waste of a household i in time period t1.
After estimating the SOW for all three product categories One-Sample t-tests indicate significant differences (p < .001) between the wasted quantities in the categories. The average SOW for virtue products is with a percentage of 57.89% the highest, followed by 23.25% for the neither category and 8.86% for the vice category.
Questionnaire
Besides the food waste diary a second data collection method is conducted. To enhance the information about the participating households an additional background questionnaire (appendix C) had to be filled in. This questionnaire collected data about the households’ sociodemographics (e.g., gender, age, education, and household size), economic background (monthly net income) as well as data about the individual food shopping behavior (e.g., planning in advance, usage of shopping lists, and discipline sticking to list) and the main motive for wasting food. Furthermore, the individuals’ motivation as well as the perceived competence of the individual to maintain a healthy diet were collected. Additionally, the respondents could optional state their address if they want to receive a little compensation for their effort. An overview of the measured variables is provided in table 4.
1
- 30 - Table 4 Variables
Question Variables Example a Source
1-7, 20 Sociodemographics Age, Household size , Income
e.g., Bos-Browers et al. (2013); Lyndhurst, Cox, and Downing (2007); Silvennoinen, Katajajuuri, and Hartikainen (2014)
8-13 Shopping behavior Planning of shopping,
discipline in sticking to the list
e.g., Koivupuro et al. (2012; Lyndhurst, Cox, and Downing (2007)
14 Food waste Main reason for food waste
e.g., Bos-Browers et al. (2013); Lyndhurst, Cox, and Downing (2007); Silvennoinen, Katajajuuri, and Hartikainen (2014) 15.1 – 15.15 Motivation to maintain healthy diet
Reason for eating healthy diet is because I feel that I want to take responsibility for my own health.
Treatment Self-Regulation Questionnaire (TSRQ) e.g., Ryan and Connell (1989); Levesque et al. (2007)
16-19 Perceived competence in
maintaining healthy diet
I feel confident in my ability to maintain a healthy diet.
Perceived Competence Scale (PCS) by Williams et al. (1996); Williams, Freedman, and Deci (1998)
a The questionnaire with all measures variables and scales is attached in appendix C.
A summary of the reliability as well as descriptive statistics of the measured variables is illustrated in table 5. The quantity of wasted food in all categories, the products bought on promotion as well as dining in a restaurant or buying Take-away food is aggregated to a weekly bases to enhance the insights of the data. This is due to the fact that several food categories show a lack of data on daily level which means that no waste for these particular categories is existent.
- 31 - Table 5
Measures & Reliability
Variables Cronbach’s alpha Mean SD R with SOWvirtuea
Promotions
Price Discount n.a. .768 1.558 .014
Bogof n.a. .048 .217 -.021
Total Promotions n.a. .820 1.634 .011
Virtue Promotions n.a .410 .955 .003
Vice Promotions n.a .160 .429 -.097
Neither Promotions n.a .240 .677 .084
Shopping behavior (e.g., Koivupuro et al. 2012; Lyndhurst, Cox, and Downing 2007)
Main shopping n.a. 1.710 1.048 -.055
Top-up shopping n.a. 1.090 1.167 -.039
Planning in advance n.a. 2.800 0.999 -.071
Using shopping list n.a. 3.800 1.201 -.089
Discipline using list n.a. 2.390 .797 -.158
TSRQ (e.g., Ryan and Connell 1989; Levesque et al. (2007)
Autonomous motivation .790b 5.146 1.018 .186*
Controlled motivation .595c 2.395 .787 .231**
Amotivation .669 1.967 .965 -.081
PCS (e.g., Williams, Freedman, and Deci 1998)
Perceived competence .706 5.567 .992 -.099
a Correlation coefficient with share of waste in virtue category, where *** implies that R is significant at 1% level, **
significant at 5% level, * significant at 10% level. b Cronbach’s alpha when two items are deleted. c Cronbach’s alpha when one item is deleted.
- 32 -
amotivation. The autonomous state is described as “the most self-determined form of motivation” (SDT.org 2014). Thus, the TSRQ indicates the perceived autonomy of a participant (SDT.org 2014). In a validation research of Levesque et al. (2007) the validity of the TSRQ scale is supported since the internal consistency for the sub items was acceptable with most α > 0.73.
In this study, only the subscale for autonomous motivation exceed the critical value of 0.7 (Nunnally and Bernstein 1λλ4) with a value of α > 0.72. By deleting the first item of the TSRQ (“Because I feel that I want to take responsibility for my own health”) and the sixth item (“Because I have carefully thought about it and believe it is very important for many aspects of my life”) the Cronbach’s alpha for the autonomous motivation subscale increases from .715 to .7λ0. Furthermore, due to a deficient translation of the second item (“Because I would feel guilty or ashamed of myself if I did not eat a healthy diet”) eliminating this item improves the α of the subscale controlled motivation from .535 to α .5λ5. The Cronbach’s alpha of the subscale measuring amotivation could not be improved. Additionally, three principal component analysis with fixed numbers of extracted factors and Varimax rotation are executed to compare if the items loaded their respective construct. The results indicate the most appropriate solution for the model where all three items are deleted (see appendix D).
The perceived competence is measured with the PCS which contains of four items and has demonstrated its reliability with alpha values of higher than 0.9 (SDT.org 2014). Although the PCS indicate an α < 0.7 the value is not as high as stated in the literature but marginal exceed the critical Cronbach’s alpha value.
Hence, both scales, the TSRQ as well as the PCS, demonstrate a somewhat lower reliability than expected. An explanation for the low Cronbach’s alpha scores may be the small size of the sample with 41 respondents. Due to the fact, that the TSRQ and the PCS have proven their reliability in several other researches both scales are maintained within the study.
- 33 - 4.2. Descriptives of the collected data
Overall, households from the Netherlands, Germany, Switzerland and Austria participated at this research. In total, 44 households participated in this study and 41 households (respectively 76 adults and 17 children) provided all the necessary information and are therefore usable for further analysis.
The person mainly responsible for shopping grocery and preparing the meals were largely female (85.4%). Focusing on age of the respondents, 31.7% were younger than 31 years, 19.5% between 31 and 40 years and somewhat lower than 5% between 41 and 50 years. The largest age group (34.1%) was between 51 and 60 years old, whereas nearly 10% were older than 60 years. The average age was nearly 42 years. Regarding education, somewhat more than 46.3% stated a university degree as the highest completed education, followed by an obtained degree as doctor (17.1%) and completed middle school (14.6%). In addition, 22% of the sample were students, whereas more as double were employed (46.3%). In total, 73.2% of the sample household did not have any children living in the household while 12.2% had one child. Households with two children were represented by 14.6% of the sample. Consequently, the average household size including children was 2.28 persons a household. The main category of monthly net income of the sample was between 3000€ and 3λλλ€ with 31.7%. Moreover, 21.λ% had a monthly net income lower than 2000€; 1λ.5% between 2000€ and 2λλλ€ and 17.1% of more than 4000€. Focusing on dining in a restaurant or ordering meals the majority (63.4%) stated that they eat in a restaurant or order Take-away food once or twice a month. The second largest group (17.1%) specified eating in a restaurant or ordering Take-away three or four times a month. In contrast, 9.8% conduct this behavior at least 5 times a month. However, just as many respondents indicated never to eat in a restaurant or to order food.
- 34 -
somewhat lower with 30132 ml (SOW 46.50%) compared to the amount of wasted groceries with 43670 gram (SOW 53.50%).
Table 6
Overview disposed products a
Waste in Waste in units
(gram and ml) Mean Min Max SD Share of total Waste
b Virtue category 34704 423.22 0 1895 46.37 60.22% Vice category 5784 70.54 0 1250 18.69 9.16% Neither category 24314 296.51 0 7060 96.50 22.08% Total waste 64802 790.27 0 7590 118.38 100% Total liquids 30132 367.46 0 7210 97.56 46.50% Total groceries 34670 422.80 0 2235 49.98 53.50%
a Liquids (measured in ml) and groceries (measured in gram) are merged. The quantity has to be read as waste in
units.
b by summarizing the SOW of virtue, vice and neither the value differs from 100% since 5.7% of the respondents did
not waste any products.
Furthermore, in appendix E three correlations matrices of all variables are reported. Unsurprisingly, the share of waste within the virtue category correlates negative (p < 0.05) with the SOWVirtue. This correlation is even stronger (p < 0.01) for SOWNeither. In addition, the SOWVirtue
- 35 -
correlation (p < 0.10) implying that respondents who dine out more often have a higher share of waste for virtue food. Focusing on the shopping habits and sociodemographics variables and the quantity of Total Waste only Discipline of using a list shows a moderate positive (p < 0.05) correlation. Interestingly though, is the strong and negative correlation (p < 0.01) between Education and Age, implying that the younger the respondents the higher the finished education is which could be a consequence of a biased dataset as explained in the limitations.
4.3. Method
In this research the dependent variable was defined as the share of wasted food within the virtue category (SOWVirtue)2. Since numerous independent variables were measured at several points in time and the dependent variable is calculated based on ascertained data of different points in time, conducting an ordinary regression seems not appropriated. This is due to the fact, some variables might be state-invariant or time-invariant and thus do not vary over time (Baltagi 2005). An example for a time-invariant variable is education: once completed a certain education it is less likely that it will change much over time. Same accounts for state-invariant variables, where the state is constant: e.g., nationwide advertisement in the radio or TV (Baltagi 2005). As a consequence, using an ordinary regression does not control for individual heterogeneity and thus results are biased and misinterpreted based on the unobserved heterogeneity (Baltagi 2005). A second reason not to conduct an ordinary regression is the problem of facing multicollinearity issues by analyzing aggregated time series (Baltagi 2005).
To solve the issues mentioned above a method is needed which accounts for the individual heterogeneity of the sample. As stated by Baltagi (2005) panel analysis suggests this heterogeneity of the participants. Furthermore, using a panel data approach enhances the information gathered from the data and gives “more variability, less collinearity among the variables, more degrees of freedom and more efficiency” (Baltagi 2005, p.5)
Compared to a ‘normal’ dataset, the panel data approach has observation of its participants at discrete points in time (Sun and Zhao 2013) and can be described as a “pooling of observation on a cross-section of households […] over several time periods” (Baltagi 2005, p.15). Furthermore,
2 The effect of the independent variables have been estimated with three dependent variables (Total Waste, SOW Virtue
- 36 -
Baltagi (2005) highlights the difference between a regular regression and a panel regression based on the double subscripts on the variables within the panel regression e.g.:
yit= α+ X’it β + uit i=1,…,N; t=1,…,T , (2)
where i describes the household (cross-section dimension) while t illustrate the point in time; the so-called time-series dimension. Both is included for the dependent variable (y), the explanatory variables (X’) and the disturbance term (Batagi 2005). Often a one-way error component model for the disturbance term is included:
uit = i + vit , (3)
where i describes the unobservable individual-specific effect and therefore only varies over the
households and not over time. Additionally, vit accounts for the extant disturbance for households
as well as variations in time. Panel analysis methods are a widely used approach on several research areas, for example: economics (e.g., Bauer 2000), politics (e.g., Plümper, Troeger, and Manow 2005) and medicine (e.g., Abbring and Van den Berg 2004) or psychology (e.g., Begley and Czajka 1993).
Based on the assumptions regarding the error term different panel analysis methods are conducted. In general two different approaches are classified: (1) fixed effect (FE) models and (2) random effect (RE) models. While (1) FE models assumes the estimation of fixed parameter for i
and the remaining vit is independently and identically distributed (Batagi 2005). Furthermore, the
explanatory variables are assumed to be independent of vit for all households i and points in time t.
In addition, FE models are not able to estimate time-invariant variables (Batagi 2005) like gender, income or health motivation and are therefore omitted in the analysis. Therefore, a FE model is recommended for investigating the impact of time-invariant variables (Torres-Reyna 2007).
In contrast to the FE models, the (2) RE models assume a random distribution of i as well
as an independency of vit. Additionally, the explanatory variables are independent of uit (independent from i as well as vit) for all households i and points in time t (Batagi 2005). This
- 37 -
Moreover, an additional third method can be used to analyze panel data; the pooled OLS. This approach ignores the panel structure of the data; in other words it ignores state and time effects which is the reason why this model is not often applied in research (Batagi 2005).
To investigate which approach is most appropriate two test can be applied: First, the Hausman test indicates if an FE or RE model is more appropriated by estimating if i (the individual
invariant effects of the error term) correlates with the Xit. Therefore, H0 assumes that no correlation
is present (Torres-Reyna 2007). If the Hausman test shows significant results a FE model has to be applied; otherwise the RE model seems to be the more appropriate model.
However, the Hausman test only differentiates FE and RE approaches. Consequently a second test has to be conducted to obviate negligence of the simple OLS regression. Therefore, the Breusch-Pagan Lagrange multiplier (LM) specifies if an RE model or a simple OLS regression fits best to the data by analyzing the variance across entities. Here, H0 indicates no significant
differences between the households, consequently non-significance lead to a non-rejection of H0
and conducting a simple OLS regression is advised; otherwise the RE model is more appropriated (Torres-Reyna 2007).
Returning to the collected data of this study the Hausman test as well as the LM tests are conducted. First, the Hausman test shows insignificant results (p=.783; p>0.10) which indicates that an RE model compared to a FE model is more appropriated. However, as mentioned by Batagi (2005) just testing on fixed effects and random effects based on the Hausman test might not be enough. Therefore, a LM is applied which could confirm, with significant results, the findings of the Hausman test. Though, the LM shows insignificant results (p=1.000; p>0.10) which indicate that an RE model is not the most approppriated model and consequently a Pooled OLS estimation should be applied for the available data.
Based on these insights the impact of the independent variables on the dependent variable can be modeled. Therefore, the main focus is set on the Pooled OLS estimation. However, also the RE model is estimated simultaneously to be able to compare the results.
- 38 -
Yit = α0+ β1Totalpromotionit + β2Autonomousmotivationi +
β3Controlledmotivationi + β4Amotivationi + β5PerceivedCompetencei + β6Mainshoppingit +
β7Topupshoppingit + β8Planningshoppingi + β9Shoppinglisti + β10Disciplinelisti + β11Genderi + β12Agecati + β13Educationi + β14Occupationi +
β15Incomei + β16Housholdsizei + β17Dinnerouti +
β18Time + uit , (4)
uit = i + vit , (5)
where Yitdenotes SOWVirtue for a household i in time observation t; Totalpromotions describes the
amount of price and bogof promotions for all categories and varies like the Mainshopping (number of main shopping), Topupshopping (number of top-up shopping) for i in t. The Autonomousmotivation (degree of autonomous motivation) Controlledmotivation (degree of controlled motivation), Amotivation (degree of amotivation), PerceivedCompetence (degree of perceived compentence) as well as the Planningforshopping (planning for the shopping in advance), Shoppinglist (frequency of using a shopping list), Disciplinelist (the discipline of sticking to the list), the sociodemografic variables (Gender, Agecat, Education, Occupation, Income and Householdsize) and Dinnerout (the regular frequency of dining in a restaurant or ordered meals per month) are time-invariant and therefore only differ over i. Additionally, Time is included as a dummy variable. The error term uit denotes a random distribution of i and is independent of vit.
Moreover, the explanatory variables are independent of uit (independent from i as well as vit) for
all households i and points in time t.
