EAT MORE
-
MOVE MORE
The effect of real-time caloric feedback in form of physical
activity calorie equivalents (PACE) on consumer behavior.
EAT MORE - MOVE MORE
The effect of real-time caloric feedback in form of physical
activity calorie equivalents (PACE) on consumer behavior.
Master Thesis University of Groningen Faculty of Economics and Business
MSc Marketing Management
14th June 2016
First supervisor: Prof. Dr. Koert van Ittersum Second supervisor: Martine van der Heide
Abstract
While obesity rates are growing, policy makers are seeking for better approaches to encourage consumers to make healthier food choices. Building on previous in-store decision making and nutritional label theories, the present study introduces and evaluates a novel nutrition information format: real-time caloric feedback in form of physical activity calorie equivalents (PACE). Using an online shopping simulation, participants were randomly assigned to one of six conditions which differed in their caloric information schemes (n= 458). Unlike as expected, there were no significant differences between the real-time caloric feedback conditions and the commonly used Calorie label condition. However, it is noteworthy that the feedback/PACE condition led to the healthiest shopping basket in absolute numbers (2,009 cal). Moreover, this study contains key implications for policy makers by revealing that PACE label leads to significantly unhealthier shopping baskets (2,453 cal) than the currently used Calorie label (2,166 cal). In conclusion, results of this study can be seen as a major warning against the introduction of PACE labels. Furthermore, the results call for further research, especially in form of longitudinal studies, to test whether real-time caloric feedback in form of PACE truly has the power to decrease the amount of calories purchased in the supermarket setting.
Keywords: nutrition information, PACE, labeling, real-time caloric feedback, smart shopping carts, in-store shopping behavior, obesity
Preface
Healthy meals and sports on a regular basis are two main components of my daily life since early childhood. Unfortunately, this is not the case in all families and thus obesity rates are drastically rising. In my opinion, it is therefore crucial that more public intervention takes place in order to prevent people from becoming obese, and in a more effective way than it is done so far. With this thesis I want to contribute to the research on nutritional information by introducing new ways to make caloric information more salient to consumers and easier to interpret.
This thesis is the final component of my studies and reflects my whole journey as a student pretty well. It was a challenging and intense, but also a very exciting, pleasant and instructive time. It gave me the possibility to broaden my experience and knowledge in the field of Marketing, and also made me grow as a person.
Ending my studies at the University of Groningen, I would like to thank my supervisor Prof. Koert van Ittersum for his strong support, confidence and valuable feedback during our meetings. It has been a great pleasure to work with you! Moreover, I want to thank Martine for the smooth collaboration and helpful input regarding the survey as well as all my pre-testers for their constructive feedback on the questionnaire.
Finally, I want to thank all the amazing people I have met during this journey - in Bamberg, Lyon and Groningen - who made studying an unforgettable experience and stage of life. In particular, I want to thank Christoph for always listening and being there for me, even if far away. A very special thanks to my parents, not only for the values they taught me, but also for giving me the greatest support in everything I do. Thank you!
Groningen, 14th June 2016
Table of Content
List of Figures V
List of Tables V
1. Introduction 1
2. Literature Review 5
2.1 In-store decision making 5
2.1.1 Heuristics and Habits 6
2.1.2 Limited Self-Control Strength 6
2.1.3 Licensing Effects 7
2.2 Labeling Formats 8
2.2.1 Back-of-Package Labeling (Nutrition Fact Panels) 9
2.2.2 Front-of-Package Labeling 10
2.2.2.1 Nutrition specific FOP Labels 10
2.2.2.2 Summary FOP Labels 11
2.2.2.3 Physical Activity Calorie Equivalent (PACE) Labels 13
2.3 Real-time Caloric Feedback 17
2.4 Hypotheses 20
2.5 Conceptual Model 22
3. Methodology 23
3.1 Data Collection Method and Research Design 23
3.2 Acquisition of Participants 24
3.3 Pre-tests 24
3.4 Procedure of the Study 25
3.5 Measurements 27
3.5.1 Healthiness of the Shopping Basket 28
3.5.2 Nutrition Knowledge (Subjective) 28
3.5.4 Enjoyment of Running 29
3.5.5 PACE Check 29
3.5.6 Control Variables 30
4. Results 31
4.1 Descriptive Statistics and Preliminary Analysis 31
4.1.1 Data Editing and Cleaning 31
4.1.2 Sample Description 32
4.1.3 Random Assignment 32
4.1.4 Normality Test 34
4.2 Hypothesis Testing 35
4.2.1 The effects of real-time feedback on the healthiness of the shopping basket 35 4.2.2 The effect of real-time feedback in form of PACE on the basket's healthiness 38 4.2.3 The different impact of PACE on consumers that like/dislike running 39
4.3 Further Analysis 42
Discussion 43
5.1 General Discussion 43
5.2 Contribution to Marketing Theory 47
5.3 Contribution to Marketing Practice 48
5.4 Limitations and Future Research 49
References 52
Appendix I: Survey 57
Appendix II: SPSS Output 65
List of Figures
Figure 1: Conceptual Model 22
Figure 2: Example Product Category Milk (Condition 6) 26
Figure 3: Estimated Marginal Means of Total Calories Purchased for each Feedback Condition 36
Figure 4: Differences in Calories purchased within Feedback/No-Feedback Condition 38
Figure 5: Estimated Marginal Means of Total Calories Purchased for each PACE Condition 41
List of Tables
Table 1: 2x3 Research Design with the Factors Feedback Condition and Caloric Information Type 24
Table 2: Condition Distribution 33
Table 3: Results of ANOVA and Pearson Chi-squared Test 33
Table 4: Normality Tests for the Dependent Variable and Continuous Moderator 34
Table 5: Means and ANOVA Post Hoc Test, Pairwise Comparison for each Information Type 37
Table 6: ANOVA Post Hoc Test, Pairwise Comparison for each Feedback Condition 39
Table 7: Means and ANOVA Post Hoc Test, Pairwise Comparison for each PACE Condition 42
Table 8: Overview Hypotheses and Results of the Research 44
1. Introduction
"To curb the obesity epidemic and improve their health, many Americans must decrease the calories they consume and increase the calories they expend through physical activity."
- US Department of Agriculture -
The emergence of overweight and obesity is drastically increasing worldwide. In 2014, over 1.9 billion adults (39%) were overweight and amongst those more than 600 million people (13%) obese, tendency rising (WHO, 2015). The World Health Organization (WHO) defines overweight and obesity as an "abnormal or excessive fat accumulation that may impair health". The Body mass index (BMI) serves as main measurement to classify people as overweight and obese. According to the WHO (2015), adults having a BMI greater than or equal to 25 are overweight, 30 obesity, respectively. Researchers forecast a 130% increase in obesity prevalence by 2030 (Finkelstein et al., 2012). This rise in obesity is primarily the result of inadequate levels of physical activity and overconsumption of calories (RSPH, 2016; Variyam, 2005). 90% of the population is eating 60 to 100 calories too many per day and consequently gradually becoming fat (Chandon & Wansink, 2012). This trend has severe implications for individuals as well as society: Overweight and obesity causes serious chronic diseases such as heart disease, stroke, musculoskeletal disorders, some cancers and diabetes. At the present, obesity leads worldwide to more deaths than underweight, which makes it one of the main preventable causes of death (WHO, 2015). Moreover, obesity can trigger psychological depression (Kolotkin, Meter & Williams, 2001). By affecting millions of people, obesity contributes substantially to increasing health costs (Finkelstein, Ruhm & Kosa, 2005) and not least due to rising costs became a major public health concern.
To conquer what is often referred to as the obesity epidemic, policy makers are seeking for environmental and policy-based approaches. Since previous research states that consumers poorly judge the amount of calories, portion sizes and energy density (Elbel, 2011; Cohen, 2008), one of the most popular policy strategies is to provide consumers with nutritional information through
labels on packaged foods. Nutrition Fact Panels (NFP) were obligatorily introduced in the U.S. in the 1994 Nutrition Label and Education Act (NLEA). In Europe, the Nutrition Labeling Directive harmonized labeling in 1990. Even if initially non-obligatory, NFPs are currently used on 85% of food items sold and will become obligatory by December 2016 (Regulation (EU) No 1169/2011, 2011). NFPs require standardized information on key nutritions, serving size, percent daily values and calories of packaged foods (Hersey et al., 2013; Downs, Loewenstein & Wisdom, 2009). However, consumers have difficulties in interpreting these nutrition information tables (Jones & Richardson, 2007).
As a consequence, the traditional numerical NFP, usually found on the back of the package, has been supplemented by front-of-package (FOP) labels, which consumers view as a summary of a product's benefits and hazards (Roe, Levy & Derby, 1999). FOP labels facilitate the overall interpretation of a product's healthiness. They appear in various formats, amongst others as Guideline Daily Amounts (GDA), traffic lights and healthy logos. Despite the consumer's positive attitude towards these FOP labels (Aschemann-Witzel et al., 2013), previous studies reported mixed results on their effectiveness (Elbel, 2011). While some studies promote positive effects, others report that FOP labels even encourage overconsumption (Feunekes et al., 2008; Scott & Worsley, 1994). Especially low-fat nutrition labels and simple tick logos stimulate overconsumption by increasing the consumer's perceptions of appropriate serving sizes and lowering consumption guilt (Chandon & Wansink, 2012). But also reviews on more informational and numeric FOP labels such as GDA found only limited efficacy in reducing calorie intake (Hersey et al., 2013).
walk or miles to walk led to a selection of lower calorie meals compared to labels with traditional calorie information only or no nutritional information at all (e.g. Swartz et al., 2013; Dowray, Swartz, Braxton & Viera, 2013). Building on the positive effects of PACE menu labeling, this research aims to understand how exercise equivalents influence shopping behavior in the grocery store as a particularly important research setting for food decisions (Papies et al., 2014).
However, this study does not intend to introduce PACE as another alternative to already existing FOP labels, since the efficacy of labels is controversial as reported by previous literature. Instead, this research builds up on the idea of using smart shopping carts to provide consumers with real-time feedback. Smart carts are shopping carts which are supplemented by a scanner and an interactive screen. Initially, they were developed to track in-store spending behavior and to inform consumers about the current costs of their groceries while shopping by displaying the total price of their shopping basket in real-time. Moreover, smart carts could be used in order to offer customized and timely promotions, recommend complementary products, skip the check out line or share recipes and nutritional information (Van Ittersum, Wansink, Pennings & Sheehan, 2013).
The approach of this study is to use smart shopping carts to display the cumulated amount of calories being in a consumer's shopping basket in real-time. The purpose of this research is to examine the effect of real-time caloric feedback on consumer purchase behavior in a grocery store, resulting in the first research question:
(1) Does real-time caloric feedback positively influence the healthiness of a consumer's shopping basket?
Moreover, this study aims to investigate the impact of real-time feedback in form of exercise equivalents on consumer purchase decisions, leading to a second research question:
In addition, it will be examined whether PACE differently influence consumers depending on their attitude towards running, i.e. the exercise they would need to perform to expend the calories of a specified food item. Therefore, the last research question of this study is:
(3) Does feedback in form of PACE in minutes of running differently influence consumers depending on the extent to how much they like/dislike running?
This study adds to the existing body of literature on nutrition information and its effects on in-store decision making. By exposing consumers to caloric information in form of feedback, this research contributes to the small body of literature on real-time feedback and its influence on consumer purchase decisions. Moreover it adds to the range of possible services offered by smart carts and thus may promote their implementation in the retail environment. Next to evaluating the effectiveness of feedback in promoting healthier in-store food choices, this study attempts to detect which feedback format is the most effective one by analyzing how purchase behavior alters between different types of real-time feedback, i.e. calories only, PACE only, and a combination of both.
By introducing the novel nutrition information format PACE to the retail environment, consumers are not only confronted with numeric caloric data, but provided with tangible and easy interpretable caloric information. Hence, PACE is expected to better inform consumers about a foods calorie content and deter individuals from purchasing high-calorie products, resulting in an overall healthier shopping basket. This makes real-time feedback in form of PACE a promising and interesting tool for policy makers to curb the obesity epidemic and promote consumer well-being.
2. Literature Review
2.1 In-store decision making
2.1.1 Heuristics and Habits
In the supermarket environment, consumers are confronted with many distracting factors such as noise, time and an overwhelming number of products to choose from, resulting in limited opportunity for detailed information processing. Also, their motivation to engage in information processing is relatively low (Feunekes et al., 2008; Petty, Cacioppo and Schumann, 1983). In order to simplify the decision making process and reduce stress in the shopping situation, consumers try to avoid difficult trade-offs by prioritizing certain aspects and not integrating others (Connors et al., 2001). Instead of precisely seeking information and evaluating alternatives to arrive at better decisions, consumers only glance at nutritional information and process those information rather superficial (Scott & Worsley, 1994). Individuals make these fast and simple buying decisions within seconds at the point of purchase by primarily relying on heuristics and habits when shopping for groceries (Grunert, Wills & Fernándes-Celemín, 2010). It is well known that these habits, and therefore buying behavior, can be influenced by marketers, i.e. by nutritional health claims found on the front of the package (Roe et al., 1999).
2.1.2 Limited Self-Control Strength
foods and an overconsumption impulse for unhealthy foods. Thus, when consumer's self-control resources are fatigued or depleted and impulse takes over, consumers are more likely to choose alternatives with immediate benefits such as indulgent, hedonic foods, than persuading long-term health preservation goals.
2.1.3 Licensing Effects
Next to limited self-control and therefrom arising impulse buying, poor decision making may emerge from so called licensing (Khan & Dhar, 2006). When being faced with an opportunity to indulge, consumers activate their memory to ascertain whether enough progress towards a self-regulatory goal has been made in the past in order to license indulgence. Especially impulsive individuals tend to distort memories of past behavior in order to justify indulgence in the present (May & Irmak, 2014). In the domain of food decisions, licensing effects occur for example when the purchase of a virtuous product category, i.e. fruits or vegetables, increases the consumer's positive self-concept, which in turn increases the likelihood that the subsequent choice will be a more indulgent product. When looking at the healthiness of a shopping basket, it is assumed that at any moment during the shopping trip, the extent of the licensing effect is modeled by the current balance of virtue versus vice products. Thus it can be concluded that being able to point to virtuous past actions may license consumers to behave less virtuously in the future (Conway & Peetz, 2012).
All in all, preventing the rise in overweight and obesity is a complex task as it depends on the consumer's limited cognitive ability to make healthy food choices and a lack of self control as well as various factors outside the individual's control. An effective tool to help consumers in the long run to better control consumption and purchase decisions are public policy interventions in form of caloric information, giving consumers a higher perceived capability of making healthful choices (Aschemann-Witzel et al., 2013). According to the social cognitive theory, perceiving that one's capability is improved raises a person's perceived self-efficacy. In turn, higher levels of self-efficacy are crucial for favorable behavior change in eating healthy (Luszczynska, Tryburcy & Schwarzer, 2007). Therefore, providing caloric information is a promising strategy to promote a change in behavior towards healthier food decisions and thus curb obesity.
2.2 Labeling Formats
In order to make healthier food choices, consumers must be able to distinguish healthier products from less healthy ones. Therefore, previous research suggested to make the nutritional composition of a product transparent in the form of nutrition labels (e.g. Scott & Worsley, 1994). Nowadays, nutrition labels are in use around the world and serve as one of the main instruments for policy makers, non-governmental organizations and manufacturers to curb the obesity epidemic. Labels provide a basis for voluntary, informed and conscious consumer decision-making and hence promote healthier food choices (Capacci et al., 2012). Consumers generally have a positive attitude towards labeling. They find labels appealing, support labeling initiatives and appreciate that labels do not restrict them in their freedom to choose (Aschemann-Witzel et al., 2013).
information and make inferences about the food's healthiness, affecting the product's overall evaluation and eventually the purchase decision (Grunert et al., 2010).
2.2.1 Back-of-Package Labeling (Nutrition Fact Panels)
2.2.2 Front-of-Package Labeling
A research of Hersey et al. (2013) reviewed 111 articles on FOP nutrition labeling effects and came to the result that FOP labels can help consumers to make healthier food choices. Nevertheless, other studies reported mixed outcomes of different label formats (Elbel, 2011; Black & Rayner, 1992). One can distinguish between two general types of FOP systems that are currently in use: nutrient specific and summary systems. In the following, the most common nutrient specific and summary labels are described and their benefits as well as hazards discussed, resulting in the introduction of a novel labeling format, so called physical activity calorie equivalents (PACE).
2.2.2.1 Nutrition specific FOP Labels
The most popular under the nutrition specific FOP symbols are labels based on the percentage guideline daily amount concept (%GDA) and traffic light schemes (TL). Both typically comprise information on four key nutrients, i.e. total fat, saturated fat, sugar and salt, as well as energy density, i.e. calories (Grunert et al., 2010). %GDA schemes display nutrients per portion and include the amount in grams and as a percentage of a person's GDA for each nutrient. TL labeling schemes are color coded and more interpretive than GDA schemes. They usually display a ranking (e.g. low, medium, or high) of the key nutrients whereby the levels are assigned color codes of green, amber, and red (Hersey et al., 2013). These color codes intend to guide the attention of the consumer to the important nutritions and improve the accuracy of the healthiness ratings. Pairwise comparison tasks of GDA and TL schemes conducted in laboratory settings resulted in higher rates of correctly perceived healthiness of food items for TL than GDA-type formats (Jones & Richardson, 2007).
healthiness leading to poor food choices, i.e. products low in fat could well be high in sugar or salt (Feunekes et al., 2008; Black & Rayner, 1992). Moreover, Wansink (2003) states that the more difficult the FOP information is to quickly comprehend, the more likely it will be misinterpreted or even ignored.
Summing up, previous studies report contradicting results and a low impact of nutrition specific labeling on consumer decision-making. Many consumers do not understand what calories mean and represent, or lack numeracy skills to apply the fraction of one food item to the recommended amount of total calories consumed a day. Consumers simply might not know how to use caloric information, leading to the assumption that the available caloric information might not be sufficient to encourage and motivate consumers to change their behavior towards healthier food decisions. To simplify the interpretation of FOP labels, summary labels have been introduced.
2.2.2.2 Summary FOP Labels
Summary systems summarize the whole nutritional profile of a product and provide an easy interpretation of its healthiness by using an algorithm to come up with an overall nutrition score. Such labels consequently facilitate and improve consumer's decision-making with regard to healthy food choices (Feunekes et al., 2008). Summary labels can be binary such as the Choices Program logo, which displays a tick on a food package if the product meets specified nutrient criteria, or the Keyhole symbol, which is shown on ‘healthier’ food items. Other summary systems are the Guiding Stars system, which displays a ranking of zero to three star depending on the product's healthiness, or the NuVal system, which reports a product's nutrient score on a scale from 1 to 100 (Hersey et al., 2013).
are made within seconds. Next to the time, also the cognitive effort needed to process the information is reduced compared to more detailed labels (Scott & Worsley, 1994). Consequently, the cognitive resources are not depleted as fast and rational decisions can be made longer before impulse takes over. Plus, simpler labels do not require detailed nutritional knowledge which makes them also suitable for 'vulnerable' groups (Feunekes et al., 2008).
Another research by Sutherland, Kaley and Fischer (2010) examined the effectiveness of Guiding Stars as one of the simple labels and found significant changes in consumer behavior. Consumers purchased significantly more ready-to-eat cereals with stars, i.e. less added sugars and more dietary fiber and fewer no-star, i.e. high-sugar and low-fiber cereals. Therefore, Guiding Stars navigation programs can be considered as effective labeling scheme to guide consumers in their food choice by providing clear, concise, and simplified nutrition information.
Hersey et al. (2013) state as well that summary systems stimulate consumers to purchase healthier products. Moreover, he claims that nutrient-specific FOP labels that incorporate text and symbolic color to indicate nutrient levels facilitate the interpretation and thus motivate consumers more to select healthier products than labels that only emphasize numeric information, such as GDA schemes.
current nutrition labeling formats may confirm consumer's existing knowledge about less healthy products, and provide new information in that healthier products were initially seen as less healthy (Feunekes et al., 2008). This assumption is supported by a study of Scott and Worsley (1994), stating that FOP labels encourage overconsumption of particular products. Since existing FOP labels are not as effective as desired, or even have a reversed effect, better techniques are needed to nudge consumers into less calorie dense food decisions.
2.2.2.3 Physical Activity Calorie Equivalent (PACE) Labels
Literature exploring the potential impact of physical activity equivalents is limited, and existing research was mainly conducted in form of menu labeling:
One of the first studies on PACE labeling, a study conducted by Fitch et al. in 2009, quantitatively assessed the acceptability of and preference for the exercise equivalent label in fast food restaurants. 71% of participants preferred the usual calorie over the PACE format, some even cited the latter as demotivating. In addition to the general preference for calorie labels, they found that exercise equivalents had a more favorable impression among non-whites than whites, and among younger consumers. However, this study has noteworthy limitations since Fitch et al. (2009) tested the attitude towards exercise equivalents rather than their actual effect. Moreover, they presented PACE as an alternative instead of an addition to the standard calorie label, as well as most participants were not overweight nor obese, and non-white.
Bleich and Pollack (2010) included in their study sample white, black and hispanic U.S. american adults and reported equally divided preferences for calorie labels, PACE labels and percentage of total daily calories in fast food restaurants.
Another research by Bleich, Herring, Flagg and Gary-Webb (2012) examined the influence of different caloric information on the number of sugar sweetened beverages purchased amongst low-income African American adolescents. Participants purchased fewer high sugar content beverages when being exposed to a combination of calorie information and PACE in minutes of running. When providing participants with the absolute calorie count only and GDA scheme, no significant reduction in the amount of sweetened beverages purchased was found.
These findings are consistent with subsequent studies, which showed that PACE in form of menu labels led to less calorie dense meal choices:
calories were ordered in the calorie plus PACE conditions, i.e. 916 calories for PACE in minutes to walk, and 826 calories for PACE in miles to walk. Thus, the combination of calories and exercise equivalents motivated consumers most to order less calories, especially when PACE was indicated in miles to walk. Respondents with a normal BMI were stronger influenced by physical activity based labels than those who were overweight or obese. In total, 82% of the participants reported a preference for a combination of calories and exercise equivalents on menu labels. 45% preferred PACE in minutes, 37% PACE in miles to walk. The focus group even "overwhelmingly expressed their interest in seeing physical activity based labels on menus" (Dowray et al., 2013). All in all, this study reports a positive impact of PACE menu labels since they create a good understanding of caloric values and encourage consumers to order lower calorie food items. Whether these labels would be effective in real-life scenarios, however, remained to be tested according to the authors.
A real-world setting study of Platkin et al. (2014) analyzed the impact of PACE menu labeling by recruiting 62 females that ordered a meal within one of three menu formats: no calorie information, calorie information only, or calorie information plus PACE. Participant in the calorie only and calorie plus PACE condition ordered about 16% (206 cal) and 14% (162 cal) fewer calories from Lunch 1 to Lunch 2, whereas the no information group ordered barley fewer calories (2%, 25 cal). Interestingly, when only looking at unrestrained eaters, a larger decrease in calories ordered was observed when respondents were exposed to calorie plus exercise equivalent labels, compared to the calorie information only and no information condition. Therefore, the authors conclude that calorie menu labeling alone might be insufficient to reduce caloric intake and further research is needed in finding the most effective ways of presenting the menu labels for general public.
PACE label would apply to them. The emergence of this question is crucial and might be the reason why PACE labels are more effective in influencing consumer choices than currently available nutrition labels, since nutrition label efficacy highly depends on comprehension and relevance (Swartz et al., 2013). It might be exactly this shift towards personalized understanding and relevance that makes PACE labels more effective than common nutrition labels.
Viera and Antonelli (2015) conducted the first study that observed changes in parents' fast food meal choice for their children when being exposed to PACE labeling. 1000 parents were randomly assigned to one of four menu labeling formats (no labels, calories only, calories plus minutes, or calories plus miles needed to walk to burn the calories) and asked to place an order for their child. Parents assigned to the no label condition ordered an average of 1,294 calories, whereas those presented with calories only, calories plus minutes, or calories plus miles ordered 1,066 cal, 1,060 cal, and 1,099 cal respectively. In addition, parents were asked to rate the likelihood with which each label would make them encourage their child to exercise. Just 20% of parents would be “very likely” to encourage their children to exercise when only calories are displayed, versus 38% for calories plus PACE in minutes and 37% for calories plus PACE in miles to walk. These results led to the conclusion that PACE is a possible strategy to motivate parents to order fewer calories for their children as well as to encourage their kids to exercise.
Taking into account the outcomes of these studies, it can be concluded that PACE labeling might be a promising alternative to currently available labeling formats by providing individuals with tangible information and an immediate link between a food's calorie content and physical activity. Regarding the presentation of exercise equivalents, i.e. PACE in miles versus minutes to walk, findings of Dowray et al. (2013) suggest that consumers might be more responsive to PACE in form of miles to walk. However, contextualizing effort in terms of time may influence people's decisions more than effort in terms of miles. Moreover, although the miles label actually had the greatest influence on the calorie content of menu selections, more participants reported a preference for the PACE indication in minutes (Dowray et al., 2013). Therefore, this study uses PACE in minutes of running to burn the calories of a food item, since it is crucial for a label's effectiveness that consumers comprehend the label and have a positive attitude towards it.
Even if existing research indicates that labeling can influence consumer perceptions, preferences, prior expectations, and post-trial evaluations of a product (Bender & Derby, 1992), labels seem not to be the most effective way to inform consumers about a food's energy content after carefully analyzing their impact on consumer behavior in chapter 2.2.2.1 and 2.2.2.2. Therefore, in addition to existing studies on PACE labeling, further research on point-of-purchase interventions is needed in order to reveal a more effective way of presenting caloric information in form of PACE to the general public.
2.3 Real-time Caloric Feedback
A novel point-of-purchase intervention to present PACE to customers might be real-time feedback. Within the last years, the role of real-time feedback has become ever more important, since new technologies, such as smartphones, pedometers or smart-shopping carts, have revolutionized the presentation and processing of feedback by immediately providing feedback.
learning (Hutton, Mauser, Filiatrault & Ahtola, 1986). Feedback informs the recipient about a risk, i.e. threat in well-being, a problem or a solution, and intends to reinforce motivation for behavior change by creating a sense of caring, reaching more directly decisional consideration, increasing motivation, or providing social comparison and norms (DiClemente, Marinilli, Singh & Bellino, 2001). The effect of feedback depends on the timing, accuracy and completeness of the feedback message. Furthermore, feedback particularly reinforces motivation for behavior change when delivered in relation to progress towards goal attainment (Patterson, 2001; Greene & Rossi, 1998).
Bravata et al. (2007) evaluated in a systematic review of 26 studies the influence of pedometers as a form of feedback on physical activity among adults in the outpatient setting. Overall, pedometer users increased their physical activity by 26.9% over baseline. Another key factor that motivated individuals to increase their physical activity next to the feedback was having a step goal such as 10,000 steps per day or an alternative personalized step goal.
Conroy et al. (2011) tested in a Self-Monitoring And Recording using Technology (SMART) Trial how different forms of physical self-monitoring affect physical activity. 210 overweight adults were randomly exposed to paper recorded physical self-monitoring, a personal digital assistant with a self-monitoring software (PDA) or PDA with a daily tailored feedback message (PDA+FB). Participants in the PDA+FB condition received feedback on physical activity behavior in real-time, all other participants received delayed written feedback. The PDA+FB group was found to be most likely to perform physical activity self-monitoring and adhere to physical activity goals, which led in turn to higher physical activity levels than in the other two groups. Consequently, it can be stated that providing participants with incremental real-time feedback is more effective in achieving a goal than traditional forms of feedback. Presenting feedback in an immediate way helps consumer to adjust their behavior in time to make better progress toward a weekly goal (Conroy et al., 2011).
receive feedback on a daily basis, the daily feedback group achieved a significantly larger reduction in energy (-22.8% vs -14.0%) and saturated fat (-11.3% vs -0.5%) intake. Moreover, a trend toward a greater decrease in total fat intake was observed. Thus, it can be concluded that daily feedback messages which are tailored and delivered promptly may enhance motivation to reduce energy and fat intake.
Van Ittersum et al. (2013) went one step further by introducing the idea of providing feedback in real-time. They researched how real-time spending feedback displayed on smart shopping carts influences shopping behavior of budget and non-budget shoppers in the supermarket setting. Results showed that real-time spending feedback reduces spending uncertainty, and consequently impacts shopping experience positively for budget shoppers, which in turn increases their reportage intention. Effects are reversed for non-budget shoppers. Moreover, basket price salience goes up especially for non-budget shoppers, i.e. the level of attention shoppers pay to the total basket price during their shopping trip (Wathieu et al., 2004). Increased salience in price might lead to increased perceived importance of price and consequently impact subsequent spending decision.
Moreover, real-time feedback can be used in smart kitchens, in order to improve the healthiness of home cooked meals by providing calorie awareness of the ingredients used for the meal. Sensors track the number of calories in food ingredients, and immediately provide real-time feedback to the user by displaying the caloric values during cooking. At the same time, the recommended number of calories for a meal is indicated. The study by Chi, Chen, Chu and Lo (2008) showed that all participants reduced the calorie density of their meal from the pretest cooking phase (without caloric feedback) to the test cooking (with caloric feedback) by up to 25.9%.
where the purchase decision is made and in a visible and salient way (Nordfalt, 2009), this research analyzes the impact of real-time caloric feedback being displayed on smart-shopping carts while shopping. This study suggest that providing consumers with tailored real-time caloric feedback will improve consumer well-being. How? Reduced choice uncertainty, increased salience and a better overview of calories in the shopping basket lead to enhanced self-efficacy. In turn, the amount of calories purchased by an individual will decrease and thus the healthiness of the shopping basket increase. This belief is supported by Greene and Rossi (1998) who found that the effect of feedback on food choice improved behavior change by allowing a more salient and personalized intervention.
2.4 Hypotheses
The discussed literature is the foundation of the conceptual model of this research and thereof derived hypotheses. Summing up the findings of previous literature, it is crucial that nutritional information quickly captures the consumer's attention, is easily exposed, visible and salient to the shopper and directly provided at the point where purchase decisions are made (Aschemann-Witzel et al., 2013; Nordfalt, 2009). Real-time feedback displayed at smart shopping carts meets these requirements, and therefore might motivate consumers to make healthier food decisions. Consequently, the first hypothesis is as followed:
H1: The cumulated amount of calories per serving size purchased by an individual will decrease when presented with real-time caloric feedback during the shopping trip.
(Scott & Worsley, 1994), which is crucial in the supermarket environment. Under the assumption that information provided in the right format leads to desirable changes in behavior (Downs et al., 2009), the second hypothesis is the following:
H2: The effect of real-time caloric feedback on calories purchased will depend on the information format, where PACE information will significantly stronger reduce the amount of calories purchased than the calorie format.
Physical activity equivalents affect individuals differently: Studies by Fitch et al. (2009) and Van Kleef, Van Trijp, Paeps and Fernández-Celemín (2008) found that European consumers preferred other types of labels to exercise equivalents as they found the latter demotivating and patronizing. Other studies, however, reported consumer's preference for physical activity equivalent labels and even a motivation to exercise when being exposed to PACE (e.g. BMJ, 2016; Viera & Antonelli, 2015; Dowray et al., 2013). Existing research on PACE labels examined differences in the reduction of calories purchased depending on the consumer's BMI, health literacy, ethics and restraint versus unrestrained eaters. However, no study tested the influence of the consumer's attitude towards sports yet, i.e. attitude towards running, on purchase decision. Consumers that dislike running might consider running as punishment, and thus will be encouraged to purchase less calories, i.e. rather eat less calories, than running a large number of minutes to expend the consumed calories. Hence, the third hypothesis of this study is:
2.5 Conceptual Model
The variables of interest for this particular research are visualized in the conceptual model. Based on the three hypotheses presented, the model shows that the independent variable real-time caloric
feedback (H1) is hypothesized to have a positive impact on the dependent variable healthiness of the consumer's shopping basket, i.e. reduction in calories purchased. The moderator variables caloric information format (H2) and attitude towards running (H3) are anticipated to impact the
effect caused by the independent variable. The more appropriate the label format of calorie information is, the healthier the shopping basket will be, i.e. the less calories will be purchased, and the less the consumer likes running, the more will it impact the positive effect of real-time caloric feedback.
3. Methodology
The previous chapter reviewed the relevant literature for this study, followed by the conceptual model which summarizes the derived hypotheses. Chapter 3 describes the methodology of the research by specifying the research design, the acquisition of participants, the procedure of the questionnaire and the measurement of the different variables.
3.1 Data Collection Method and Research Design
This study aims to investigate the effectiveness of different types of real-time caloric feedback on the healthiness of a consumer's shopping basket. In order to test the hypotheses derived from the conceptual model, an empirical research was conducted in form of an online questionnaire. The Internet seems to be an appropriate survey method due to its easy and fast distribution, high reach and low costs. Moreover, the bias of social desirability is smaller, the researcher can easily program skip patterns and change the order of the questions (which is especially important for presenting the product categories, since different respondents should be exposed to different orders) as well as data is available in digital form, resulting in a more efficient processing of the results (Malhotra, 2010, p. 222). For the data collection of this study, the research platform Qualtrics was used.
Calories PACE Calories & PACE Caloric
Feedback
No Condition 1 Condition 2 Condition 3
Caloric
Feedback Yes Condition 4 Condition 5 Condition 6 Table 1: 2x3 Research Design with the Factors Feedback Condition and Caloric Information Type
3.2 Acquisition of Participants
Participants were acquired via Amazon Mechanical Turk (MTurk). MTurk is a crowdsourcing online marketplace that is used by individuals and businesses for tasks that require human intelligence. It is used for this study in order to get a representative dataset within shortest time since it has the advantage that a certain panel of people can be selected. In this case US citizens being the primary household shoppers were preselected. Participants received $2.25 in case of adequate completion of the survey. Furthermore, they had the chance of winning a prize package of $100 in value, consisting of cash and grocery products as incentive for realistic product choices. The questionnaire was activated 4th May 2016 and all data was gathered within one day.
The survey was accessible via computer, tablet and smartphone. The aim was to obtain a total of 450 participants, i.e. 75 individuals per condition. Respondents were randomly allocated to one of the six conditions.
3.3 Pre-tests
3.4 Procedure of the Study
In order to generate empirical data, an online survey including an online grocery shopping environment in the first section and self-report measures in the second part of the questionnaire was designed using Qualtrics software (2016 Qualtrics LLC). The structure of the survey is as follows. After a short introduction pointing out the framework of the survey and declaring that the study is conducted within the scope of the University of Groningen, participants had to declare consent to their participation as well as to confirm being the primary grocery shopper in their household as necessary precondition in order to participate in this study. As incentive, to motivate respondents to select options they would choose when they would be on a real shopping trip, all participants had a chance of winning a prize package of $100 in value, comprising the grocery products in their shopping basket, and additional cash.
Pictures, prices, serving sizes and calories for the products used in the survey were obtained from the online shop of the US retailer Walmart (Walmart.com). In case of inadequate information at Walmart.com, data was retrieved from the US online grocer FreshDirect (freshdirect.com). The PACE values were then calculated based on the energy expenditure of an adult with an average body weight of 160 lb, running at a rate of 10 minutes per mile as used in previous studies on PACE labeling (Dowray et al., 2013; Swartz et al., 2013). Thus, 12.8 kcal/min are burned according to an energy expenditure chart by Blair et al. (2001) that lists estimated calories burned by activity and body weight. In order to determine the number of minutes required to burn off calories of a food item, the calories per serving of the product were divided by the energy expenditure rate, i.e. 12.8 kcal/min. Sample calculations are included (Appendix B).
Moreover, in three of the six conditions participants were additionally presented with real-time caloric feedback on top of the page which was updated after each new product choice. In condition 4, the cumulated amount of calories was displayed, i.e. the amount of calories the person would consume by eating one serving of each product in her shopping basket at that point of time. Condition 5 provided the participant with the cumulated amount of PACE, i.e. the total amount of minutes needed to run to burn those calories and condition 6 featured a combination of both, i.e. the cumulated amount of calories and PACE based on the current shopping basket (see Figure 2).
Next to the four product choices, a no-choice option was included in every category to make the answer selection more realistic and more comparable to a real shopping trip. Respondents could only choose one item per category. Once a product was selected, the participant was transmitted to the next page to continue the shopping trip, i.e. choosing the next item. The order of the 25 categories as well as the order of the four products within one category was randomized, only the no-choice option was always displayed on the right and presented not very salient since participants should not been stimulated to select the no-choice option.
Before starting the virtual shopping trip, a brief listing of the elements shown to the consumer in the specific condition and, if applicable, a short explanation on PACE and real-time feedback was provided (Appendix C). The shopping trip was completed after respondents selected one item of each product category.
In the second part of the survey, participants were asked several questions about their shopping experience, their attitude towards health and physical activity, knowledge and use of nutritional information, and a few other questions, which however will not be used for the evaluation of this study but will serve for future research on grocery shopping behavior. The remainder of the questionnaire collected basic demographic information such as monthly household income and spending, gender, age, self-reported height and weight. Respondents were not able to move back during the whole survey, and all items were mandatory. In a last section, participants had the possibility to leave a comment, enter their Worker ID (enabling us to contact them in case of winning the prize), as well as they received a short debrief and finally the code needed to get paid for their participation before submitting the survey.
3.5 Measurements
3.5.1 Healthiness of the Shopping Basket
The dependent variable healthiness of the shopping basket is reflected by the amount of calories in the respondent's shopping basket and thus was measured by the cumulated calories selected by a respondent. The less calories the participant purchased, the healthier is his shopping basket; the more calories purchased, the unhealthier is the shopping basket respectively. In order to be able to state which condition leads to the healthiest shopping basket, the cumulated amount of calories per serving size per product in the shopping basket was recorded (but not shown to the respondent) after completing the virtual shopping trip, i.e. selecting one item of each product category. The mean scores for every condition was calculated, and hence conclusions drawn. The no-choice option was coded as zero calories (Appendix D). The condition with the lowest calorie score will be the most effective one in order to motivate consumers to make healthier product choices and the condition scoring highest in calorie-density of the shopping basket the least effective one, respectively. In addition, the survey collected variables that might explain variations in the calorie content of food choices among participants randomized to the same type of condition. These include according to previous literature nutritional knowledge, use and understanding of nutritional information, enjoyment of running and the respondent's BMI.
3.5.2 Nutrition Knowledge (Subjective)
confident about my ability to comprehend nutrition information on product labels. Moreover, at later moment of time, the actual understanding of PACE information was tested. The Nutrition Knowledge scale shows a strong internal consistency with a Cronbach’s Alpha of .886.
3.5.4 Enjoyment of Running
In order to make inferences between the continuous moderator Attitude towards Running and its influence on the effectiveness of PACE information on the healthiness of the shopping basket, the participant's enjoyment of running was measured. Therefore, the Physical Activity Enjoyment Scale (PACES) by Mullen et al. (2011) was used to measure Enjoyment of Running on a 8-item scale. This is a novel version of the original PACES which includes 18-items and is intended for a college-age population (Kendzierski & DeCarlo, 1991). Respondents were asked to rate "how you feel at the moment about the physical activity you have been doing" using a 7-point bipolar rating scale. Higher PACES scores reflect greater levels of enjoyment. For this study the introductory question was adapted to the exercise of running, i.e. "How do you feel about running?". The eight items measured on the seven-point bipolar rating scale remained the same as used by Mullen et al. (2011) and were displayed in a randomized order: I find it pleasurable; It's a lot of fun; It's very pleasant; It's very invigorating; It's very gratifying; It's very exhilarating; It's very stimulating; It's very refreshing. The eight items of the scale show a strong internal consistency with a Cronbach’s Alpha of .972.
3.5.5 PACE Check
of one serving of product A; A food item having a PACE value of 15 is healthier than product A; You would need to run 8 miles to burn the calories of one serving of product A. The statements were presented in randomized order with statement 1 being the correct answer choice.
3.5.6 Control Variables
4. Results
The present chapter provides a description and analysis of the data collected through the quantitative research explained in chapter 3. To run the analyses, the raw data was transported into SPPS Statistics. In a first step, relevant descriptive statistics are presented in order to provide insights into the respondents' demographics. Moreover, it is tested to what extent demographical differences occur within the six groups which resulted from the randomized allocation to the different conditions, followed by the test of normality and Levene’s test of homogeneity of variances. Subsequently, in the main part of this chapter, ANOVA analyses are performed in order to validate the hypotheses.
4.1 Descriptive Statistics and Preliminary Analysis
4.1.1 Data Editing and Cleaning
Furthermore, the amount of cumulated calories never exceeded the total amount of calories that could be purchased and hence no further respondents were deleted, resulting in a total of 446 valid respondents after editing and cleaning the data.
4.1.2 Sample Description
Overall, more women (n= 252) than men (n= 194) participated in the survey (56.5% females versus 43.5% males). This slight difference in female and male participants might be explained by the fact that traditionally women are the primary grocery shopper in a household. The respondents' age ranges from 20 to 75 years with an average age of 37 years (M= 37.29, SD= 11.38). All respondents had their geo-location within the boarders of the US. The amount of the participants' (as a household) monthly spending on groceries ranges from $50 to $1,600 with an average household spending of $407 (SD= 207.57).
4.1.3 Random Assignment
This study examined six different experimental conditions. According to Simmons, Nelson and Simonsohn (2013), at least 50 participants per condition are needed in order to receive reliable estimates, i.e. a total of at least 300 participants for this study. 446 valid responses are therefore a sufficient sample size to conduct reliable analyses. An overview of the randomized distribution of the 446 participants across the six different experimental conditions is presented in Table 2. The slightly uneven distribution of participants to the conditions may be explained by the deletion of participants during the data cleaning process. However, exactly half of the participants (n= 223) were presented with real-time caloric feedback, and the other 223 participant did not receive feedback during their online shopping trip.
ANOVA ANOVA ANOVA F Sig. Age . 789 . 558 Weight . 255 . 937 HH Spending 1.302 . 262 Pearson's χ2 Test Pearson's χ2 Test Pearson's χ2 Test Value Sig. Gender 4.425 . 490
gender was reviewed by Pearson’s Chi-squared test. The results of these tests depict that the null hypothesis (there is no significant difference) cannot be rejected (p > .05) for the variables age (p= . 558), weight (p= .937), household spending (p= .262) and gender (p= .490). This leads to the conclusion that the six experimental conditions did not significantly differ regarding the participants' demographics (see Table 3).
Condition N NoFeedback_Cal 76 NoFeedback_Pace 73 NoFeedback_Cal+Pace 74 Feedback_Cal 74 Feedback_Pace 77 Feedback_Cal+Pace 72 Total 446
Table 2: Condition Distribution Table 3: Results of ANOVA and Pearson Chi-squared Test
Moreover it was ensured that the subjective nutrition knowledge and confidence in using that knowledge did not significantly differ between groups (total Mean= 5.26, F= .228, p= .950).
4.1.4 Normality Test
To gain insights into the distribution of the data, normality tests were conducted for the dependent variable and the continuous moderator. As the independent variable and the moderator caloric information format are dummy coded, normality testing does not apply for these variables. To assess normality, the Kolmogorov-Smirnov test and the Shapiro-Wilk test were conducted. In order to extract valid conclusions from a regression analysis, the assumption of normality needs to be met (Field, 2009). Both tests show significant results (p< .05) for both variables, leading to the conclusion that they are significantly non-normal. Consequently, the null hypothesis of a normal distribution can be rejected. Table 4 depicts an detailed overview of the results.
However, when the sample size is large (> 200) and the amount of independent variables is small (< 5), these tests have limitations. Tests might turn out to be significant even if the scores are only slightly different from a normal distribution (Field, 2009). Therefore, skeweness and kurtiosis statistics are used to gain further insights into the distribution of the data. Whether the data is normally and symmetrical distributed is indicated by the skewness statistic, whereas information on how flat or peaked the data is, is obtained by the kurtosis statistic. In case of a perfect normal distribution, the values of both statistics equal zero. However, the data is still considered as normally distributed for skew values ranging between -0.5 and 0.5 and kurtosis values between -1.96 and 1.96 (Field, 2009). The results of the normality tests for this research depicts that all variables are situated within the acceptable scope for both statistics (see Table 4). Hence, the variables are normally distributed and the assumption of normal distribution is not violated.
Kolmogorov-Smirnov
Kolmogorov-Smirnov Shapiro WilkShapiro Wilk SkewnessSkewness KurtosisKurtosis Statistic Sig. Statistic Sig. Statistic Std. Error Statistic Std. Error
Total Calories . 076 . 000 . 976 . 000 -. 279 . 161 -. 757 . 231
Enjoyment Running . 082 . 000 . 943 . 000 . 118 . 116 -1.109 . 231
Lastly, Levene’s test of homogeneity of variances was applied (Appendix F), resulting in non-significant result (p= .107). Hence, the null hypothesis can not be rejected, meaning that the variance of the healthiness of a shopping basket is equal across the six groups, which is a desirable result and allows to perform further analyses.
4.2 Hypothesis Testing
In the following section, ANOVA analyses were performed to analyze the hypotheses established in the literature section. In order to analyze the main effects and interaction effect of real-time caloric feedback and caloric information format on the healthiness of the shopping basket, i.e. amount of calories purchased, a 2 (real-time caloric feedback: no vs. yes) x 3 (caloric information format: Calories vs. PACE vs. Calories+PACE) ANOVA on total calories purchased was performed. Subsequently, the continuous moderator was researched by running a second 2 (enjoyment of running: low vs. high) x 3 (PACE format: no PACE vs. PACE label vs. PACE feedback) ANOVA which analyzed how the effects of the PACE formats differed between people who like and dislike running. In order to run these analyses, the independent variable feedback (no= 0, yes= 1) and the moderators caloric information format (Calories= 1, PACE= 2, Calories+PACE= 3) and enjoyment of running (low= 1, high= 2) were dummy coded.
4.2.1 The effects of real-time feedback on the healthiness of the shopping basket
To answer H1, the first main effect of the ANOVA that depicts whether the healthiness of the shopping basket differs between the group being exposed to real-time caloric feedback and the no feedback group has been further analyzed. The ANOVA resulted in a significant difference between the no feedback and feedback condition, F(1, 440)= 17.284, p= .000. Also Figure 3 depicts that
feedback led in all three caloric information type conditions to a healthier shopping basket, however differences vary a lot.
Figure 3: Estimated Marginal Means of Total Calories Purchased for each Feedback Condition
To test the impact of real-time feedback in more detail, a specific contrast analysis was conducted. The specific contrast analysis (Appendix H) showed that the effect of feedback was only significant in the PACE condition, F(1, 440)= 12.307, p= .000, η2= .027 and the Calories+PACE condition,
F(1, 440)= 11.833, p= .001, η2= .026. In the PACE condition, participants presented with feedback
(M= 2,008.64, SD= 829.74.91) had a significantly healthier shopping basket than participants presented with PACE label only (M= 2,453.49, SD= 700.95, p= .000). The same holds for the Calories+PACE condition (MFeedback= 2151.56 (SD= 822.59), MnoFeedback= 2,593.58 (SD= 774.88), p= .001). However, unlike as expected, no significant results were found for differences between the feedback and no feedback condition for the information type Calories, F(1, 440)= .057 (<1), p= . 812, η2= .000; MFeedback= 2,135.52 (SD= 727.43), MnoFeedback= 2,165.75 (SD= 828.91), see Table 5.
exposed to feedback including the PACE format strongly reduced their scores of total calories purchased. However, these seemingly favorable results are driven by very perverse effects of PACE label without feedback on calories purchased. Thus, outcomes of an one-way ANOVA (Appendix I) comparing the effects of the Calories/no feedback condition (M= 2,165.75) and PACE/feedback condition (M= 2,008.64) on calories purchased were unlike as hypothesized non-significant, F(1, 440)= 1.373, p= .243.
The most important conclusion is that real-time caloric feedback is not significantly influential when comparing it to the typical, currently used caloric information in the supermarket setting, i.e. calorie labels, no matter which caloric information type (Calories, PACE or Calories+PACE) was used for presenting feedback. Hence, hypothesis 1, presenting the consumer with real-time caloric feedback reduces the amount of calories purchased, resulting in a healthier shopping basket, can not be confirmed.
Means of Total Calories Purchased Means of Total Calories Purchased Means of Total Calories Purchased Means of Total Calories Purchased Means of Total Calories Purchased
No Feedback
No Feedback FeedbackFeedback
Mean Std. Deviation Mean Std. Deviation
Calories 2,165.75 828.91 2,135.51 727.43 PACE 2,453.49 700.95 2,008.64 829.74 Calories + PACE 2,593.58 774.88 2,151.56 91.484 Univariate Test Univariate Test Univariate Test Univariate Test Univariate Test Univariate Test
Information Type df Mean Square F Sig. Partial Eta Squared
Calories 1 34,278.097 . 057 . 812 . 000
PACE 1 7,415,881.269 12.307 . 000 . 027
Calories+PACE 1 7,130,271.366 11.833 . 001 . 026
4.2.2 The effect of real-time feedback in form of PACE on the basket's healthiness
Even if real-time caloric feedback did not significantly impact the healthiness of the shopping basket in comparison to the Calories/no feedback condition, in absolute numbers feedback reduced the number of calories purchased in each information type condition, especially when delivered in form of PACE (see Figure 3). Hence, in a next step, the second hypothesis stating that real-time caloric feedback has a stronger impact on the healthiness of the shopping basket when presented in form of PACE was analyzed. As already stated in Chapter 4.2.1, both main effects as well as the interaction effect of the Two-way ANOVA appeared significant, meaning that the information type of presenting feedback has an impact on the dependent variable healthiness of the shopping basket. However, as depicted in Figure 4, differences between the caloric information formats on the dependent variable are pretty small within the feedback condition in comparison to differences within the no feedback condition.
Figure 4: Differences in Calories purchased within Feedback/No-Feedback Condition
In the no feedback condition, participants presented with Calories (M= 2,165.75, SD= 828.91) had a significantly healthier shopping basket than participants presented with PACE (M= 2,453.49, SD= 700.95, p= .024), and Calories+PACE (M= 2,593.58, SD= 774.88, p= .001) respectively. No significant results were found for differences between the PACE and Calories+PACE format in the no feedback condition, p= .097. In the feedback condition however, the caloric information effect was unlike as hypothesized not significant, F(2, 440)= 0.768 (<1), p= .465, η2= .003; MCalories=
2,135.52 (SD= 727.43), MPACE= 2,008.64 (SD= 829.74) and MCal+PACE= 2,151.56 (SD= 806.92) (see Table 5 and 6). Univariate Test Univariate Test Univariate Test Univariate Test Univariate Test Univariate Test
Condition df Mean Square F Sig. Partial Eta Squared
No Feedback 2 3,575,737.436 5.934 . 003 . 026
Feedback 2 462,640.443 . 768 . 465 . 003
Table 6: ANOVA Post Hoc Test, Pairwise Comparison for each Feedback Condition
Alternatively, the effect of the caloric information format was only significant within the no feedback condition, F(2, 440)= 5.934, p= .003, η2= .026 and not significant in the feedback
condition, F(2, 440) < 1, ns. Therefore, the second hypothesis, which claimed that the effectiveness of real-time caloric feedback on the healthiness of the shopping basket depends on the information format, whereby the PACE format will have a stronger positive impact than the Calorie format, can not be confirmed. Due to non-significant results in the feedback condition, H0 (= there are no differences between the caloric format groups) can not be rejected. However, when looking at the absolute amount of calories purchased, caloric feedback in form of PACE is the condition that led to the overall healthiest shopping basket (see Figure 3 and 4), even if at a non-significant level.
4.2.3 The different impact of PACE on consumers that like/dislike running
with PACE condition and enjoyment of running as independent variables. In a first step, the initial six were summarized into three conditions depending on the occurrence of PACE, resulting in a no PACE group (incl. initial condition 1 & 4), a PACE label group (incl. condition 2 &3) and a PACE feedback group (incl. condition 5 & 6). Moreover, the variable enjoyment of running was split into two groups based on its seven-point Likert scale (1= "I do not enjoy it at all", 7 = "I enjoy it a lot") leading to the groups low (<4) and high (>=4) levels of enjoyment. The dependent variable is, again, healthiness of the shopping basket, i.e. total calories purchased. The initial expectation was that PACE feedback will significantly reduce the calories purchased for consumers that do not like running (H3a), but will have no effect on consumers that do like running (H3b). The ANOVA (Appendix K) showed a significant result for the main effect PACE condition, F(1, 440)= 14.085,
p= .000. The least calories were purchased in the PACE feedback condition (M= 2,077.70), followed by the no PACE (only Calories) condition (M= 2,150.83) and most calories were purchased when PACE was displayed as label (M= 2,524.01). However, the other main effect enjoyment of running appeared not significant, F(2, 440)= .261, p= .609, neither did the interaction effect, F(2, 440)= 0.831, p= .436.
