The Hunger Games: does your sleepy brain trick you into buying
unhealthy foods?
A quantitative research about the effects of placebo sleep quality on making
food choices and susceptibility to price promotions, and the role of personal
value of sleep in this relationship
Myrthe Timmermans (10550615)
Master Thesis MSc Business Administration Specialization: Marketing Track
Faculty of Economics and Business University of Amsterdam
Academic year: 2016-2017 Supervisor: Andrea Weihrauch
Statement of originality
This document is written by Student Myrthe Timmermans who declares to take full
responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is
Table of contents Abstract 4 1. Introduction 5 2. Literature review 9 2.1 Sleep quality 9 2.2 Food choices 10
2.3 Measured sleep quality and food choices 11
2.4 The placebo effect 13
2.5 The personal value of sleep (WTP) 15
2.6 Price promotions 16
2.7 Research gap and question 17
2.8 Conceptual framework 18
2.9 Hypotheses 19
3. Data and Method 21
3.1 Setting 21 3.2 Subjects 21 3.3 Data collection 22 3.4 Pilot study 22 3.5 Procedure 23 3.6 Measures 23 3.7 Method 28 4. Results 29
4.1 Descriptive and frequencies statistics 29
4.2 Correlation matrix 31
4.3 Testing the hypotheses 33
4.3.1. H1: the effect of the sleep conditions on food choices 33 4.3.2. H2: moderation of willingness to pay for sleep quality 34 4.3.3. H3: effect of sleep conditions on switching behavior for price promotions 37
5. Discussion 39
5.1 Overall conclusion 39
5.2 Theoretical implications 41
5.3 Managerial implications 42
5.4 Limitations and directions for future research 43
6. References 45
Appendix 52
7.1 Tables 52
Abstract
An extensive amount of research has been conducted about the relationship between sleep and public health problems, showing that inadequate sleep leads to obesity. There are only few academics who focused on the relationship between sleep and food choices, which could be an important underlying mechanism for obesity. Therefore, this study aimed to examine the influence of placebo sleep quality on making (un)healthy food choices and the likelihood of switching to a different food product due to price promotions. Moreover, the role of personal value of sleep (willingness to pay) was researched in this relationship. Data was collected from 117 respondents who participated via an online between-subjects experiment.
Respondents were randomly assigned to condition one, where they were told that they had bad sleep quality the previous night, or condition two, where they were told that they had good sleep quality the previous night. The results showed that there were no significant effects of sleep quality on making (un)healthy food choices. Moreover, the analysis showed that there was not a significant effect of the sleep conditions on the likelihood of switching to another product due to price promotions. Lastly, the moderating effect of the willingness to pay for sleep was not demonstrated. These results indicate that it is difficult to use placebo sleep in communication. Therefore, companies can use sleep information in their
communication, but they should not assume that people are affected. Future research should research the placebo effect with high involvement categories and should use higher price promotion percentages.
1. Introduction
Nowadays, two parallel trends are visible: 1) individuals are not achieving adequate amounts of sleep (Demos et al., 2016) and 2) obesity is one of the biggest public health concerns (Garaulet et al., 2011). At first sight these two trends appear to work independently from each other, but research proves otherwise. Namely, extensive research has been done about this relationship, showing that lack of sleep leads to obesity (Chaput, Despres, Bouchard & Tremblay, 2011; Magee & Hale, 2012).
On average, we spend one third of our lives sleeping (Sejnowski & Destexhe, 2000). Sleep is essentially food for the brain and is important for both physical and mental health, as the body needs to restore itself and needs to process and store the information learned that day. As both sleep and adequate nutrition-intake are both basic needs for humans to survive, it grabbed the attention of researchers, resulting in a large amount of studies about the relationship between these two concepts. Previous studies already have looked extensively at the relation between sleep duration and obesity (Chaput et al., 2011; Magee & Hale, 2012), showing a wide range of underlying mechanisms for this relationship, such as increased ghrelin levels and decreased leptin levels (Chapman et al., 2013), increased calorie-intake and portion size (Schmid, Hallschmid, Jaunch-Chara, Born & Schultes, 2008), increased hunger levels (Hogenkamp et al., 2013) and higher levels of BMI, body fat and waist circumferences (Garaulet et al., 2011).
There are only few academics who focused on the relationship between sleep and food choices, which could be an important underlying mechanism for obesity as well. Sleep deprived adolescents are more likely to choose fast food consumption than vegetables and fruit (Kruger, Reither, Peppard, Krueger & Hale, 2014). Moreover, sleep deprived adults spend more money of their fixed budget on high caloric foods containing fat and sugar than
after a normal night of sleep (Chapman et al., 2013). These researchers are conducted in an experimental context and/or are based on self-reported data of the respondents.
The effect of real sleep deprivation thus has proven to have effects on body types and food choices. However, a lot of people still sleep well objectively, but they still might experience an effect between sleep and food choices. No study to date has looked at the effect placebo sleep quality can have on food choices. The placebo effect is ‘’any outcome that is not attributed to a specific treatment but rather to an individual’s mindset regarding the kind of treatment he or she is receiving’’ (Benson & Friedman, 1996, p. 194). It is difficult for people to really judge if their sleep quality is good or bad and they might be susceptible to external cues. For example, reading a news item that people sleep bad or a friend telling them that 5 hours of sleep is all you need to feel great. These external cues about sleep may lead to a subjective evaluation of sleep quality. Research of McFerran and Mukhopadhyay (2013) showed that people hold lay theories about phenomena in their environments. Lay theories are ‘’implicit assumptions that ordinary people hold about themselves and their world’’ (Karnani, McFerran & Mukhopadhyay, 2014, p. 10). Such lay theories are based on information of different sources, such as their cultural environment and everyday experiences. It was shown in their study that people’s lay beliefs about obesity impacted their food choices (McFerran & Mukhopadhyay, 2013), proving that these naïve beliefs guide actual behaviors. It is thus important to investigate if such external cues can influence changing the mindset of consumers. This research will thus look if the mindset of respondents can be manipulated, specifically regarding their perceived sleep quality, and look whether this has an effect on (un)healthy food choices. Moreover, reactions to price promotions in supermarkets could be an interesting way to see how “believing’’ to sleep bad versus good might affect consumers. Price promotions boost purchasing by reducing the price of products, giving customers a feeling of satisfaction of making a good deal and saving money (Hawkes, 2008).
Price changes can be used to deliver a message to the public which products they should buy, hereby promoting (un)healthy products (Andreyeva, Long & Brownell, 2010). Finding out the consequences of the placebo sleep quality on consumer behavior is an additional way to test the relation between placebo sleep quality and consumer evaluations, looking at the
susceptibility to switch their (un)healthy food choice to a price promoted (un)healthy product. Self-control is an important individual difference that can affect a wide array of decisions and behaviors. Haws, Davis and Dhlakia (2016) showed that healthy eating and responsible food spending are influenced by self-control. They showed that respondents with low self-control more easily ignore nutritional information about a product than people with high self-control; consequently, they would consume unhealthy food without thinking about the consequences.
In addition to testing if making consumers believe they slept bad (instead of measuring objective sleep deprivation in previous research) can affect their food choices, this thesis wants to examine if there could be differences between consumers on how strongly they react to such placebo effects. More precisely, the personal value of sleep could play an interesting part in this relationship. Imagine that people find sleep important for adequate decision-making and then telling them they slept badly. Consequently, they will think that they are not capable of making good decisions. Therefore, they will use the placebo sleep diagnosis for self-licensing, manipulating themselves into thinking they deserve unhealthy food. If the personal importance of sleep is low however, the placebo may have no effect. Thus, looking at the extent to which people find sleep important for decision making is an interesting way to see if the placebo will work in this research. To capture the personal valuation of sleep, willingness to pay (WTP) will be used. WTP is a respondent-reported outcome that uses a valuation method to measure preferences which provide the researcher with an absolute numeric value (Delfino, Holt, Taylor, Wittenberg & Qureshi, 2008). In previous researches, WTP for sleep has shown to have great differences under respondents (Miyakawa, Matsui &
Hiramatsu, 2010; Riethmuller, Muller, Knoblauch & Shoch, 2008), showing that attached importance to sleep thus varies. Combining the findings of previous researches, this thesis will seek to answer: ‘What is the effect of (placebo) sleep quality on healthy versus unhealthy
food choices as well as consumers’ susceptibility to price promotions of (un)healthy food items, and what is the role of the personal value of sleep in this relationship?’
This research has practical importance both for marketers and public policy makers and can be used as a tool to better present their products and information to consumers. If people believe that they slept good (even if they have not), and therefore make healthier choices, policy makers need to use this information for obesity-prevention and health-promotion programs. Moderating the mindsets via placebo information could be an easy way to enhance the health of consumers. However, this effect could be misused by companies: If people believe that they slept bad (even if they have not), and therefore make unhealthier choices, junk food companies can use this information in their marketing efforts to customers, thereby manipulating their mindset and subsequent buying behavior towards unhealthy food.
Consumers need to be aware about this so that they can protect themselves from this bias. Moreover, the effect of placebo sleep on price promotions can give some interesting insights on how to promote certain products to the public, regardless of the product type.
This thesis will be structured in five sections. The next chapter will explain relevant concepts used in the thesis and will shed some light on the status-quo about the literature of sleep quality and food choices, personal value of sleep (WTP) and price promotions. After that, the conceptual framework will be explained. Thirdly, the research design will be explained in detail and how the originating data is analyzed. The thesis will end with the main conclusions and limitations and a thorough explanation of the academic and managerial implications of this research.
2. Literature review
In this chapter, relevant concepts of the thesis are explained. Moreover, the existing literature about sleep quality, the personal value of sleep (WTP), food choices and price promotions is reviewed. After explaining the relevant literature, this section concludes with the conceptual framework, a research gap and the research question and hypotheses to fill the gap.
2.1 Sleep quality
Many healthy adults suffer from insufficient sleep. Previous studies have already looked extensively at the relation between sleep duration and obesity (Chaput et al., 2011; Magee & Hale, 2012), showing a wide range of explanations for this relationship. The medical
explanation is that sleep restriction impacts the hormonal regulation of food intake. For instance, ghrelin, an appetite increasing hormone, has shown to have higher levels after sleep deprivation, whereas leptin, an appetite decreasing hormone, has shown to have lower levels after sleep deprivation (Chaput et al., 2011). Previous studies about obesity point to a distinct relationship between sleep duration and body weight, suggesting a U-shaped association, showing that both shortened and extended sleep durations coincide with increases in body weight (Gangwisch, Malaspina, Boden-Albala, & Heymsfield, 2005).Studies with more simple explanations argue that short duration sleepers are more fatigue and therefore do less physical activity. Moreover, they simply have more opportunities and more time to eat. Sleep restriction has also proven to increase calorie-intake and portion size, showing that sleep deprived respondents take more snacks during the day and eat larger meals (Schmid et al., 2008). Moreover, sleep deprivation is also associated with hunger. For instance, Hogenkamp et al. (2013) showed that participants selected larger portions and reported higher hunger ratings after total sleep deprivation than after a normal night of sleep. Furthermore, Garaulet et al. (2011) showed that short duration sleepers had higher values of Body Mass Index, body
fat and waist circumferences than average and long duration sleepers. However, there is some inconsistency in these results, since a lot of studies only have proven the effect of sleep deprivation on weight gain. The meta-analysis of Magee and Hale (2011) showed that of the thirteen studies researched, four studies found an association only between short sleep duration and weight gain, four found an association between short sleep and weight gain and long sleep and weight gain, and the remaining five found no significant association between sleep duration and weight gain. These researches show there are many mechanisms in which short sleep duration ultimately has a negative effect on obesity. However, not all studies are consistent in showing the effect of sleep duration on energy balance, highlighting the importance for more research aimed at the mechanisms between sleep and obesity.
2.2 Food choices
While the above discussed researches looked at obesity and increased weight as an outcome of sleep deprivation, a relevant step in explaining potential consumer weight issues is to look at their choices which lead them to being overweight. Health consequences of food
consumption has been an important topic the last years. Prior research has shown that consumers show a natural tendency to consume both healthy and unhealthy foods (Talukdar & Lindsey, 2013). Nevertheless, there is an opposing trend visible: people tend to
overconsume unhealthy food and underconsume healthy food. This phenomenon can be explained by people’s intuitions when thinking about food. Raghunathan, Naylor and Hoyer (2006) found that there is systemic evidence that people think that unhealthy foods are tasty, and healthy food are assumed to be bland. Furthermore, Ostan, Poljsak, Simcic and Tijskens (2010) showed that although unhealthy, energy-dense food is bad for long-term health, people prefer it on the short term because healthy, non-energy-dense food, cannot give the same energy level on the short term. Therefore, people would prefer to eat unhealthy food when
choosing between healthy and unhealthy food. Another line of research about choosing healthy or unhealthy food originates from what is called the function of the consumption goal (Raghunathan et al., 2006). Research suggests that the hedonic and enjoyment goal is very important, and trade-offs must be made between the health benefits of a food offering versus the enjoyment of consuming something tasty. Therefore, when people have a salient hedonic goal in mind they would choose unhealthy food.
2.3 Measured sleep quality and food choices
There are relatively few studies that have focused on the effect of sleep quality on food choices, which could be an important underlying mechanism of the effect of sleep duration on obesity. Most of these studies used either self-reported sleep quality scales (e.g. Stanford Sleep Quality Scale or Stanford Sleepiness Scale) or simplified proxies such as quantity of sleep hours. Moreover, these studies have mainly focused on people who suffer from total sleep deprivation and people with short sleep duration (e.g. sleeping less than 6 hours per night). Kruger et al. (2014) showed that American adolescents who self-reported short sleep duration were less likely to choose vegetable and fruit and were more likely to choose fast food consumption. In a study with a food purchasing task, Chapman et al. (2013), studied two groups of normal-weight men. Compared to participants who had a normal night of sleep, participants who underwent a night of total sleep deprivation, spend more money of their fixed budget on high caloric foods containing fat and sugar. Furthermore, a recent study of Prather et al. (2016) about the relationship between sleep duration and the consumption of sugar-sweetened beverages (SSBs), showed that short duration sleepers had a 21% higher SSB consumption than long duration sleepers, and they chose fewer healthy drinks such as water, fruit juice and diet drinks.
Additionally, short sleep duration has proven to have a negative impact on impulsivity – defined as ‘’activing quick and without thorough consideration of consequences’’ (Demos et al., 2016, p. 214). Impulsivity has shown to be related to overeating and unsuccessful dieting because people prefer unhealthy food over healthy food (Meule, Lutz, Vogele & Kubler, 2014). Furthermore, in an experimental study of Benedict et al. (2012), they researched brain activity after normal sleep and total sleep deprivation using functional magnetic resonance imaging (fMRI), showing images of food to the respondents. The stimuli consisted of six low calorie foods and six high calorie foods. The results show that the sleep-deprived respondents gave 12% more yes responses to the question if they found the food more appetizing,
especially the high calorie foods. In addition to these results, another fMRI study performed by St-Onge, Wolfe, Sy, Shechter and Hirsch (2014) tested the neuronal responses to healthy and unhealthy food under sleep restricted respondents. Here the stimuli consisted of healthy food such as carrots and grapes, and unhealthy food such as pizza and candy. The results show that viewing unhealthy foods led to enhanced activation in the brain reward and food-sensitive areas compared to seeing healthy food. Moreover, the habitual sleep respondents did not show these specific activity patterns of unhealthy food.
In a research of Wells and Cruess (2006), students recorded subjective self-reported sleep quality, food consumption and food choices in a daily diary for four days. They then were given an assignment to sleep four hours or less for a night to use that as the partial sleep deprivation manipulation. The research showed that bad sleep causes respondents to be in a depressed, down mood, which may subsequently influence food choice; it can be used as an excuse to deserve unhealthy food. The respondents chose less healthy products and cared less about their weight. Moreover, another study using fMRI suggests that sleep deprivation is associated with an increased risk-taking for gains and a decreased loss-avoidance behavior (Phuong & Galván, 2016), suggesting that the desire to choose unhealthy food will be bigger.
Furthermore, Benedict et al. (2012) measured objective sleep quality in a sleep lab where half of the respondents underwent a night of total sleep deprivation and the other half had a normal night of sleep. The following fMRI study showed that sleep causes people to be more
sensitive to rewarding food stimuli, and thus unhealthy food.
2.4 The placebo effect
Although measured sleep quality is an important measurement to use in sleep studies, we know people are easily manipulated away from their beliefs. Instead of thus focusing on these validated measurements and methods, it would be interesting to hypothetically tell
respondents whether their perceived sleep quality is good or bad. If people think sleep is important for decision-making, they therefore presumably use their sleep quality as an indicator to base their decisions on. No study to date has researched the relationship between sleep quality and food choices using the placebo effect of telling people what their perceived sleep quality is. People find it difficult to judge for themselves if their sleep quality is good or bad and they might be susceptible to external cues. These cues ultimately may lead to a subjective evaluation of sleep quality. As mentioned previously, people hold lay theories about causes and consequences of many phenomena (McFerran & Mukhopadhyay, 2013). Lay theories are ‘’implicit assumptions that ordinary people hold about themselves and their world’’ (Karnani et al., 2014, p. 10). These lay theories can be grounded in all sorts of
information sources, such as everyday experiences and people in their environment. Although these lay beliefs may sometimes be in consensus with scientific research and in other cases do not, they have proven to exert an enormous influence on judgement and behaviors. Namely, in the research of McFerran and Mukhopadhyay (2013), it was shown that people’s lay beliefs about obesity impacted their food choices.
The placebo effect is another stream of research which proves how people can be
manipulated. The placebo effect is ‘’any outcome that is not attributed to a specific treatment but rather to an individual’s mindset regarding the kind of treatment he or she is receiving’’ (Benson & Friedman, 1996, p. 194). Two theories seem to underlie the placebo effect: expectancy theory and classical conditioning (Shiv, Carmon & Ariely, 2005). The former explains that beliefs about a product/procedure (which serves as the placebo) activates certain expectations that will occur, which in turn then influences the effectiveness of the stimuli. The latter, classical conditioning, views that the active substance/medication serves as the
unconditional stimuli and the methods to deliver the treatment is the conditioned stimuli (e.g. pills, drinks). Continuously pairing the unconditioned stimuli with the conditioned stimuli over time will eventually evoke that the pills, capsules and injections create a therapeutic effect as conditional response (Montgomery & Kirsch, 1997). The placebo effect is mostly used in pharmaceutical studies involving medicine, but can also easily be applied to aspects of everyday situations. For example, Draganich and Erdal (2014) have used placebo information in their sleep research. In their study, respondents themselves reported their sleep quality of the previous night. They then were randomly assigned to condition one, where they were told they had an above average sleep quality, or condition two, where they were told that they had a below average sleep quality. Participants then did the PASAT test (Paced Auditory Serial Addition Test), to assess the rate of information processing. Results show that when
participants were told they had a below average night of sleep, they performed worse on the PASAT test than participants who were told that they had an above average night of sleep. Their research thus proves that the perception of sleep quality can be influenced and can affect cognitive functioning of consumers. Moreover, other placebo related research has focused on experiments with false feedback, where bogus information is given to the
weight feedback. Respondents were weighted 2,5 kilograms heavier or lighter than their actual weight. Results show that respondents who were told they weighted less were not affected by the false weight feedback. However, respondents who were informed that they weighted more reported less positive moods and lower self-esteem. Furthermore, participants in this condition ate significantly more during a taste test than the other group. They let their worsening mood influence their dietary intake and that resulted in overindulgence of food.
2.5 The personal value of sleep (WTP)
No study to date has researched if the personal valuation of sleep can play a role in the relationship between sleep quality and food choices. Previous research has investigated the value of sleep in monetary terms, looking at the willingness to pay (WTP). WTP is a respondent-reported outcome that uses a valuation method to measure preferences which provide the researcher with an absolute numeric value. WTP is mostly used to derive the worth of a commodity, but a few health-related studies have used it to research what society wants to pay for health (Bobinac, Van Exel, Rutten & Brouwer, 2010). The outcome value is used as an indicator for the strength of preference. Bobinac et al. (2010) studied the individual willingness to pay for a quality-adjusted life-year (QALY), which compromises both length and quality of life. Participants expressed their WTP on a payment scale to use a painless medicine to avoid a decline in health for one year. The resulting WTP’s from the respondents differed from 0 to 2500 euro per month. In a sleep context, prior research has mainly focused on the willingness to pay for undisturbed sleep due to traffic noise and found that there were great differences between the willingness to pay values of the respondents (Miyakawa et al., 2010). Moreover, Riethmuller et al. (2008) found that noise-free sleep overall had a
remarkably high monetary value for the respondents, ranging from 8 – 24 dollars per night. Another line of research focuses on the willingness to pay for sleep for respondents suffering
from deceases. For instance, Delfino et al. (2008) looked at the willingness to pay for the ability to sleep under respondents suffering from Psoriasis, and showed that the willingness to pay ranged from 50 dollars to 5000 dollars a night. These studies show that people attach different monetary values to sleep. If people think sleep is important for adequate decision making, they most likely value sleep a lot.They will use their sleep quality to guide their behaviors. If the personal importance of sleep is low however, the placebo may have no effect, as consumers may use different parameters to base their behaviors on. Thus, looking at the extent to which people find sleep important for decision making is an interesting way to see if the placebo will work in this research. Value of sleep can thus be an interesting moderator in the relationship between sleep quality and food choices.
2.6 Price promotions
Besides willingness to pay for improving sleep quality, the price that people pay for products could have an important role as well for choosing (un)healthy food. Supermarkets have a lot of power and influence on what people buy, as they can determine what they sell.
Supermarkets have become important gatekeepers of the food supply, hereby being partially to blame for diet-related diseases and the associated risks, such as obesity, heart disease and cancers (Hawkes, 2008). They are the drivers in the diet switch of consumption of energy-dense, low-nutrition foods. However, they could use their power to prevent obesity, by increasing availability of healthy foods, such as fresh fruit, vegetables and lean meat, and reducing the availability or raising the prices of unhealthy foods (Glanz, Bader, & Iyer, 2012). In the research of Glanz et al. (2012), publications about food marketing and grocery stores were collected to make an extensive literature review. Their findings suggest ways to promote healthful eating by increasing affordability, price promotions and prominence of healthy products, hereby slowly de-marketing unhealthy food products. Price promotions are used in
stores to boost purchasing by reducing the price of products and stimulating impulsive buying behavior. Other research suggests that promotions on less-healthy products lure customers away from healthier, higher priced options (Nakamura et al., 2015). Moreover, there is a tendency for supermarkets to promote unhealthy food because the sales uplift is larger for less-healthy products than healthier categories. Looking if price promotions of food products (unhealthy and healthy foods) can cause consumers to switch their initial food choice can be an important tool for policy makers. Changes in prices alone may not be enough for public policy makers to increase the consumption of fruit and vegetables, but price changes may be a tool to deliver the message to the public by combining the price promotions with public education campaigns and regulations (Andreyeva et al., 2010). Price promotions may show the public what kind of products they need to buy for both public policy makers trying to decrease obesity but it is also a tool for marketers of junk food companies to increase their sales.
2.7 Research gap and question
To date, the causal mechanisms behind sleep quality and energy balance remain unclear (Kruger et al., 2014). Further research is needed to determine whether certain sleep durations impact food choices and nutrient intake (Grandner, Jackson, Gerstner & Knutson, 2013). Therefore, based on the above discussed literature, in this research sleep quality perceptions will be manipulated by using a placebo diagnosis and the effect thereof on a series of food choices will be tested. The placebo is then actually used as feedback on the subjective
measure of their subjective sleep data. This research has theoretical importance because it will contribute to the current consumer behavior research about mindset manipulation and the use of placebo information in communication efforts. Furthermore, this will be the first research looking at the moderating role of the personal value of sleep in the relation between sleep
quality and food choices. Moreover, this study will be the first to measure the relationship between sleep quality and food choices using placebo information. Additionally, researchers have disclosed that there should be a more complete understanding of the public health impact of food supply by changing the food prices (Andreyeva et al., 2010). Researching price
promotions of both healthy and unhealthy products has an impact to close this gap a bit further. The research elaborated in this paper aims to respond the following question: ‘What is
the effect of (placebo) sleep quality on healthy versus unhealthy food choices as well as consumers’ susceptibility to price promotions of (un)healthy food items, and what is the role of the personal value of sleep in this relationship?’
2.8 Conceptual framework
The main question is visualized in the following conceptual model (see figure 1), and can be used to understand the subject matter.
Figure 1. Conceptual model of the thesis. Placebo sleep quality Food choices Price promotions Personal valuation of sleep (WTP)
2.9 Hypotheses
This part of the chapter briefly introduces and states the hypotheses. More elaborate covering of the subjects can be found in the literature review above.
Placebo sleep quality and food choices
Using a cross-sectional three-year long health promotion project using surveys under overweight and non-overweight adolescents, Chen, Wang and Jeng (2006) showed that adequate, good sleep can be used as a motivation for health promoting behavior such as following a healthy diet, therefore choosing healthy food over unhealthy food. Giving participants a good sleep quality diagnosis may thus act as a positive stimulant to pursue healthy food choices, and thus a healthier lifestyle. Whereas, bad sleep causes the person to be in a depressed, down mood, which may influence food choices; it can be used as an excuse to deserve unhealthy food (Wells & Cruess, 2006). Respondents would care less about their weight and eating healthy foods. Moreover, they can manipulate themselves into thinking they deserve unhealthy foods as a tool to make the bad feeling go away. Eating unhealthy food gives us more enjoyment and thus can make us happier than healthier food when feeling down. Therefore, the following hypotheses can be compiled:
H1a: Participants in the good sleep condition will be more likely to choose healthier food options.
H1b: Participants in the bad sleep condition will be more likely to choose unhealthier food options.
Personal value of sleep (WTP)
If people think sleep is important for decision-making, they will attach a lot of value to it. People thus see sleep as an important indicator to base their decisions on. Participants in the
good sleep condition already obtain a positive feeling as they are told they have healthy sleep habits. They will let themselves think that healthy food is the way to go to maintain the good, healthy feeling. The more value is attached to sleep, the more people thus let their sleep quality guide their food choices. On the contrary, participants in the bad sleep condition hear that they slept badly; they will then think that they are not capable of making adequate
decisions. Therefore, they will use the placebo sleep diagnosis for self-licensing, manipulating themselves into thinking they deserve unhealthy food. Therefore, the following hypotheses can be compiled:
H2a: Personal value of sleep moderates the effect of the good sleep condition. The higher the WTP (valuation of sleep), the more positively the good sleep condition influences choosing healthier food choices.
H2b: Personal value of sleep moderates the effect of the bad sleep condition. The higher the WTP (valuation of sleep), the more positively the bad sleep condition influences choosing unhealthier food choices.
Placebo sleep quality and price promotions
Price promotions can lure consumers away from the (un)healthier choice as there are two extra advantages to pursue the price promotion: 1) saving money and 2) getting a good feeling by making a nice deal. Making a nice deal will strengthen the good feeling already obtained by having the good sleep diagnosis, whereas for the bad sleep diagnosis it will make up for the down, depressed mood.
H3: Participants in both sleep conditions (good and bad) will be more likely to choose food products that have price promotions, regardless of the product type (healthy or unhealthy).
3. Data and Method
The focus of this chapter is describing the research design. Moreover, this chapter describes the setting of the study and the subjects involved. Moreover, the used instruments, data collection and analysis procedures are discussed.
3.1 Setting
To answer the research question and hypotheses, a cross-sectional experimental research design is used. As for the respondents, the research is conducted at a broad age category, enabling the possibility to make comparisons between age groups. The minimum age for participating is 16 years old, as from that age children do not need parental approval to process online personal data (Ministerie van Justitie, 2001). Nonetheless, because everybody sleeps and eats, this research is relevant for everyone. Therefore, other than age, there were no further specific requirements for participating in the experiment. Specifically, a between-subjects with one factor (placebo sleep quality: good vs. bad) experimental design is used. This has the advantage that every score is independent of other scores, so crossover effects are prevented. An overview of the experimental conditions and variables is recapped in table 1.
Table 1. Overview of the conditions and variables of the thesis.
3.2 Subjects
In total, 130 people have participated in the experiment. A total of 117 respondents were taken into the final analyses phase. Not all the respondents filled in the survey completely. As I used forced response, incomplete surveys were thus unfinished surveys. Therefore, 13
Good quality sleep condition Bad quality sleep condition
- DV: food choices - DV: price promotions
- Moderator: personal valuation of sleep (WTP)
- Demographics & Controls
- DV: food choices - DV: price promotions
- Moderator: personal valuation of sleep (WTP)
- Demographics & Controls
respondents were dropped from the analyses. From the respondents, 41% is men and 76% is
women. The age of the respondents varies from 17 to 73 years, with an average age of 32 years (M = 32,24, SD = 15,53). Due to the random assignment to the conditions, 58 participants were assessed to condition one and 59 participants were assessed to condition two. A substantial number of participants implied to be restricting food intake to control weight (39,3%), indicating it is an important control variable to take into the analyses.
3.3 Data collection
The experiment is conducted as an online study. An online study is very cost-efficient way, as there are no costs for lab spaces, equipment and research conductors (Reips, 2000). Moreover, it is also time-efficient as it allows for simultaneous access by all the participants and is very easy to spread across networks. Furthermore, an online study allows for higher statistical power as a large data sample can be created (Reips, 2000). The survey was programmed in Qualtrics, the available survey program via the University of Amsterdam. In this program, it is possible to randomize the participants in the two conditions via the option ‘Randomizer’. The chosen approach for distribution of the questionnaire is the snowball method. The biggest advantage of the snowball sampling is that it helps the researcher find more participants that cannot be done via other sampling techniques (Gravetter & Forzano 2015). The questionnaire was spread out in the network of the researcher to family and friends via e-mail. Moreover, the questionnaire was also distributed on social networks such as Facebook and LinkedIn to reach the largest possible number of participants.
3.4 Pilot study
A small pilot study with five participants was conducted to test the configuration, introduction of the questionnaire, duration and the clarity of the questions. Their feedback was collected and implemented before starting with the main study.
3.5 Procedure
The respondents received an email in which the research is described and it is explained why they are asked to participate. Moreover, in the preface of the study, the process is explained and it is made clear that the respondent can stop at any time with the experiment when s/he does not want to participate anymore. The questionnaire begins with questions about the sleep habits of the respondents, measured by The Pittsburgh Sleep Quality Index (Buysse,
Reynolds, Monk, Berman & Kupfer, 1988). After that, a couple of demographic questions are asked. Afterwards, a false feedback diagnosis is shown (see below for detailed explanation), whereas half of the respondents sees the good sleep quality diagnosis and half sees the bad sleep quality diagnosis. Afterwards, a series of food choices is shown to the respondents. For these food choices, reactance is measured. Thereafter, with the same food products, price promotions of either the healthy or unhealthy food are shown, measuring their switching behavior to the other product. Next, the personal value of sleep is measured by asking the respondents their willingness-to-pay (WTP) for a medical pillow which increases their sleep quality. Next, the control variables are measured such as dieting status, allergy check and disliking of the used foods. Finally, respondents are asked to speculate about the purpose of the research, and are asked if they had any improvements for the questionnaire and want to join the lottery for the gift card. Lastly, the respondents are debriefed.
3.6 Measures
Independent variable: sleep quality and the fake sleep diagnosis. To measure sleep quality, a
lot of different scales exist, such as the Pittsburgh Sleep Quality Index (PSQI), Stanford Sleepiness Scale and the Epworth Sleepiness Scale. For this research, the PSQI will be used. The scale measures different components: sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication and daytime dysfunction
(Buysse et al., 1988). The purpose of asking these questions is that the placebo sleep quality diagnosis is then ‘calculated’ based on the self-reported data of the respondent, and therefore credible enough to avoid suspicion that the diagnosis is false feedback. To create the sleep diagnosis, a similar approach as the one of Draganich and Erdal (2014) will be used.
Respondents will be randomly assigned to either condition one, where they are informed that they have bad sleep quality, or condition two, where they are informed that they have good sleep quality. The respondents will have to wait a short while before the diagnosis is shown, so that it is more believable their score is being ‘calculated’. Moreover, the diagnosis will be illustrated with a score wheel in Qualtrics to make the diagnosis more credible. Furthermore,
this diagnosis should be told by someone who stands for authority, someone who knows what he is talking about to earn some credibility from the respondents. Research shows that people favor a physician in a professional attire with a white coat, which favorably influences trust, confidence and authority (Rehman, Nietert, Cope, Osborne & Kilpatrick, 2005). Therefore, an image of a doctor in a white coat will be also be used for the stimuli. The diagnosis will be accompanied with the following story: ‘’On average, we spend one third of our lives sleeping.
Sleep thus plays an essential role in maintaining good health and well-being throughout your life. Scientific research has shown that there is great correlation between actual sleep quality and self-reported data. Based on your reported sleeping habits and your demographics, a sleep quality score was calculated. According to our analysis, your sleep quality score of 76% is good (or: your sleep quality score of 34% is bad) compared to the average score of
participants of your age, gender, and body information in this study.’’
Moderator variable: the personal valuation of sleep (WTP). To measure the personal
valuation of sleep, willingness-to-pay (WTP) is used. WTP to measure sleep has already been used in a couple of sleep researches (Delfino et al., 2008; Riethmuller et al., 2008). For this
research, the method of Delfino et al. (2008) is used. To avoid getting answers all over the place, this question will be illustrated with a story about a pillow: ‘’Having the right pillow is
important for a good night’s sleep. Having the right cushioning for your upper body and positioning it correctly is one of the most important factors for sleeping success. Recently, Pillow Dreams launched a new pillow, called 'Sound Asleep', which can be used for different types of sleepers (back, front and side sleepers). This new pillow has scientifically shown to improve sleep quality significantly.’’ The WTP will then be assessed with the following
question: ‘’The cost of an average pillow with no sleep improving functions is 29,95 Euros.
Compared to an average pillow, how much would you be willing to pay for the Sound Asleep Pillow?’’ The survey question is asked as an open-ended bias to prevent respondents to
anchoring to ranges of bidding. Moreover, as a second question the following is asked: ‘‘At
the end of the survey, you will have a chance to participate in a lottery for a Bol.com gift card. If possible, would you switch the gift card to a Sound Asleep pillow?’’. This question
indicates whether respondents would attach more importance at improving their sleep quality or receiving a monetary gift.
Dependent variable: (un)healthy food choices. In the study of Chapman and Maclean (1993)
respondents were asked to give examples of unhealthy food and healthy foods. The most common examples of unhealthy food were junk food and candy and for healthy food fruit and vegetables were most cited. These respondents agreed that unhealthy food contains a lot of fat, calories and/or sugar, whereas healthy food represents high nutritional values such as vitamins and minerals. To avoid socially desirable responses in the direction of choosing the healthy option, the foods in this research are not too obvious the healthy or unhealthy option. Participants will have to choose between a series of two food options, involving a healthy and unhealthy option. To measure this variable, the reactance to the food will be measured by
asking the respondents: ‘’Given the choice between [..] and [..], which one would you pick?’’ The two products have the same price and the same content in the package so that these factors could be excluded. This construct is measured on a dichotomous scale (1 = healthy option, 2 = unhealthy option). Respondents will make a choice between the following food options: 1) smoothie (healthy) versus chocolate milk (unhealthy), 2) green olives (healthy) versus cocktail nuts (unhealthy) and 3) a nut bar (healthy) versus a chocolate bar (unhealthy).
Dependent variable: price promotions. After measuring the reactance to the food options, a
series of price promotions will be presented (showing the same food stimuli as used for the food choices). The participants who chose the healthy option will see the promotion of the unhealthy product and vice versa. For example, for someone who chose the smoothie, the following scenario is shown: ‘’You now find out that the chocolate milk is on promotion for
20%. How likely are you to switch your choice to the chocolate milk (instead of the
smoothie)?’’. This construct is measured on a 7-point Likert scale (1 = extremely unlikely, 7
= extremely likely). 20% and 15% are chosen as the discount amounts as these percentages are not too much of a reduction, but considerable enough to switch.
Age. The study of Talukdar and Lindsey (2013) showed that younger consumers are more
likely to do impulsive food purchases because the impulsive and risky behavior is sensitive during the adolescents’ period. Age will be asked with one open-ended question: ‘’What is
your age?’’
Gender. The study of Kruger et al. (2014) showed that shortened sleep duration is associated
with more consumption of fewer nutrient-dense food for men than for women. Fast food consumption differed significantly by gender, showing that males ate more than females. This
variable will be measured with one question: ‘‘What is your gender?’’
Height and weight. These demographics will be asked to make the sleep diagnosis more
credible, as the diagnosis will then be ‘calculated’ using the measurements of the participants. These constructs will be measured with the following questions: ‘’What is your height?'' and
‘’How much do you weigh?’’.
Dieting status. The study of Meule et al. (2014) used this as a control variable. It is likely that
respondents who are trying to lose weight would choose the healthy option, because it contains less calories. This variable will be measured with one question: ‘’Are you currently
restricting your food intake to control your weight (e.g. by eating less or avoiding certain foods)?’’
(Un)healthiness of products. For each of the used foods in the survey the following question
was asked: ‘‘In your opinion, how (un)healthy is this product?’’ This is an important manipulation check as people can differ in their opinions about what is healthy and what is not, and in the end they make choices based on their own beliefs.
Dislike of the used foods. If the participant does not like one of the food choices,
consequentially s/he will choose the other food option. The choice then will not be based on the sleep diagnosis. Therefore, this control variable should be used in the analyses. This construct is measured by asking: ‘’Do you dislike any of the foods used in this survey?’’
Food allergy and intolerance. Following the same logic as the control variable ‘dislike of the
the sleep diagnosis. Therefore, it is an important control variable to consider. This construct is measured by asking: ‘’Are you unable to eat any of the used foods in this survey due to food
allergy or intolerance (e.g. lactose, wheat/gluten, peanut?)’’
3.7 Method
All statistical analyses were conducted using SPSS version 23.0. To describe the demographic variables and the self-reported sleeping- and eating habits of the participants, frequency and descriptive analyses were used, consisting of means and standard deviations (SD). Moreover, a correlation matrix with all the main variables was performed.
For testing the hypotheses, to see if there is a difference between the scores of the good and bad sleep quality conditions on the health index scale, a one-way between subjects ANOVA analysis was conducted, as the independent variable is dichotomous and the dependent variable is continuous. Then, to look at the food choices separately, a series of logistic regressions were conducted. Logistic regressions allow for dichotomous dependent variables and can include covariates as well.
Furthermore, to measure the moderating role of the personal valuation of sleep (WTP), three analyses were conducted. Firstly, a two-way ANOVA analysis was performed, as both the independent (sleep condition) and moderator (willingness to pay, yes vs no) are dichotomous. Secondly, again a two-way ANOVA analysis was performed, as both the independent (sleep condition) and moderator (willingness to trade, yes vs no) are dichotomous, Lastly, to analyze if there are trends visible within the respondents who were willing to pay different amounts of money, a moderation analysis using PROCESS from Hayes was used. There is moderation when the direction or strength of the relationship between the independent (X) and dependent variable (Y) is influenced by the moderating variable (Z) (Baron & Kenny, 1986).
Moreover, to measure if there is a difference between the good and bad sleep quality
condition and their likelihood of switching to the price promotion product, one-way between subjects ANOVAs were conducted, as the independent variable is dichotomous and the dependent variable is continuous. Moreover, ANOVAs allows to research the effect of the control variables as well. Two separate analyses were performed for all the food choices: 1) the likelihood of switching from a healthy option to an unhealthy option due to a price
promotion and 2) the likelihood of switching from an unhealthy option to a healthy option due to a price promotion.
4. Results
In the following paragraphs the results of the analyses of the data in SPSS will be discussed in detail. Moreover, this chapter addresses all mentioned hypotheses, states the statistical tests that were used, and explains the results.
4.1 Descriptive and frequencies statistics
In this section, the commonly used variables are summarized and discussed. Moreover, the descriptive and frequency statistics will be examined in detail. As mentioned before, the respondents were assigned to condition one (bad sleep quality) or condition two (good sleep quality). In total, 117 respondents participated in the experiment, whereof 58 respondents were assigned to the bad sleep condition (49,6%) and 59 participants were assigned to the good sleep condition (50,4%). Due to the randomizer function in Qualtrics I thus managed to make the two groups of equal proportions. In the end of the survey, none of the participants revealed suspicion about the manipulation of the sleep quality score.
The dependent variable ‘food choices’ is a dichotomous variable. The results show that the choice between the smoothie and chocolate milk had an obvious favorite, showing that 87,2% of the respondents chose the smoothie over the chocolate milk (12,8%), thus choosing the healthier option. The choice between green olives and cocktail nuts shows a more equal distribution - 57,3% of the respondents chose the green olives and 42,7% chose the cocktail nuts. Here as well the healthier option is chosen a little more often. Moreover, the choice between the nut bar and chocolate bar displays the exact same numbers, showing that 57,3% of the respondents would choose the chocolate bar over the nut bar (42,7%). However, here the unhealthier option is chosen a bit more often. Moreover, on average, respondents scored in the middle of the scale which measured an aggregated total of all the food choices (M = 1,87, SD = 0,84). Furthermore, as in every food choice one product was conceptualized as being the healthier option, the results show that the products were interpreted as they were intended; respondents scored the nut bar higher on the healthiness scale (M = 4,31, SD = 1,31) than the chocolate bar (M = 2,18, SD = 1,06) (t (116) = -15,01, p = 0,000, CI [-2,41, -1,85]), the green olives (M = 4,95, SD = 1,17) higher than the cocktail nuts (M = 2,12, SD = 0,99) (t (116) = -23,16, p = 0,000, CI [-3,07, -2,59]) and the smoothie (M = 4,84, SD = 1,29) higher than the chocolate milk (M = 2,17, SD = 1,13) (t (116) = 19,73, p = 0,000, CI [2,40, 2,93]). The dependent variable ‘price promotions’ is a continuous variable (measured on a 7-point Likert scale). Respondents were initially exposed to two equally priced products (one healthy, the other one unhealthy), and once they made a choice they learned that the non-chosen option was on promotion and therefore became cheaper. They then evaluated their willingness to switch their initial preference due to the price promotion. The results of the analysis show that overall, the willingness to switch to the other food product (whether it was from the healthy to the unhealthy option, or the other way around) was low for all the used stimuli
(overall switching preference (M = 2,27, SD = 1,03), switch from unhealthy to healthy (M = 2,66, SD = 1,72), switch from healthy to unhealthy (M = 2,08, SD = 1,17)).
The mean of the moderator personal value of sleep (WTP) showed that 102 participants were willing to pay more for the sleep quality improving pillow than an average pillow, ranging from 30 to 120 euros (M = 50,51, SD = 17,80). Only 15 participants would thus not be willing to pay for the pillow. Thereafter, when asking if participants would be willing to trade the Bol.com gift card of 20 euros for the Sound Asleep pillow, 61 participants (52,1%) said yes and 56 participants (47,9%) said no.
4.2 Correlation matrix
Prior testing the hypotheses, a correlation analysis was conducted to check if there are any variables that correlate with one another (see table 2). Not all values could be calculated because the promotions of the food products are only seen by respondents who initially chose the other food option. So, SPSS treats the variable food choice as a constant variable and the Pearson correlation cannot be calculated. Therefore, in the correlation matrix, no correlations are measured between the food choices and their corresponding price promotions. The control variable age shows that it is significantly correlated with the food choice chocolate bar vs nut bar (r = .21, p < 0.05), as well as the price promotions of cocktail nuts (r = .25, p < 0.05) and green olives (r = .33, p < 0.05). Moreover, age correlates with the willingness to pay for the Sound Asleep pillow (r = .21, p < 0.05), indicating that the older the respondent, the more s/he is willing to pay more for the pillow. Moreover, as can be seen in the correlation matrix, a lot of the food choices and price promotions seem to correlate with the other food choices. Although the control variables diet and gender did not show significant correlations, previous researches disclosed that these are important confounds to consider (Kruger et al., 2014; Meule et al., 2014). Therefore, they still will be used in the hypotheses testing.
Table 2. Results of Pearson’s correlations including means and standard deviations (n = 117).
Note. * Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed). b. Cannot be computed because one of the
variables is constant.
Variables M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14
1. Gender .35 .48 _
2. Age 32.24 15.5 .21 _
3. Sleep quality condition .50 .50 -.17 -.03 _
4. Food choice smoothie/chocolate milk .87 .34 -.09 .60 .18 _
5. Food choice green olives/cocktail nuts .57 .50 -.08 .01 .08 .03 _
6. Food choice nut bar/chocolate bar .43 .50 -.01 .21* -.11 .13 .08 _
7. Promo chocolate milk 1.91 1.27 .00 -.07 -.03 .b -.04 .02 _
8. Promo smoothie 2.73 1.95 .35 -.41 -.03 .b -.56* -.24 .b _
9. Promo cocktail nuts 2.07 1.35 .32 -.25* .10 -.12 .b -.05 .37** .70 _
10. Promo green olives 2.28 1.81 .08 .33* .80 .06 .b .18 -.02 .33 .b _
11. Promo chocolate bar 2.46 1.48 -.08 .08 -.16 .14 -.32* .b .23 .94 .31 .24 _
12. Promo nut bar 2.78 1.77 .06 -.05 -.13 .24* -.03 .b .05 .57 .18 .33 .b _
13. WTP (paying) 50.50 17.81 .13 .21* -.14 -.02 -.11 -.09 .14 .18 -.01 .12 .04 -.21 _
14. WTP (trading) .52 .51 .13 -.08 -.10 -.06 .00 -.07 .03 -.18 -.01 -.01 -.02 -.01 .11 _
4.3 Testing the hypotheses
For all of the hypotheses, respondents were excluded from the analyses if they disliked or are allergic for one of the food products in the food choices. So, for example, testing effects on the food choice chocolate bar vs. nut bar, respondents indicating they are allergic for peanuts were excluded from the analyses. This is necessary to do, as they would automatically would choose the chocolate bar - the choice would be due to external factors rather than attributed to the effects of the sleep conditions.
4.3.1. H1: the effect of the sleep conditions on food choices
Participants were able to choose between a healthy and an unhealthy food option in three different categories (drinks and two snack categories). To check the effect of the sleep
conditions on the food choices, firstly a health score scale was made with all the food choices in all three categories together (ranging from 0 = only unhealthy choices, 3 = only healthy choices). A one-way between subjects ANOVA analysis was conducted to compare the effect of the bad sleep quality condition and the good sleep quality condition on (un)healthy food choices. The results show that there is not a significant effect of sleep quality on food choices at the p <.05 level for the two conditions (F (1, 65) = 0,26 p = 0,611). The control variables diet (F (1, 65) = 2,09, p = 0,15), gender (F (1, 75) = 2,06, p = 0,162) and age (F (35, 65) = 0,99, p = 0,500) were not significant as well. To look at the food choices separately, a series of logistic regressions were conducted (see table 3). Logistic regressions allow for
dichotomous variables and can include covariates as well. The logistic regression for food choice 1 (smoothie vs chocolate milk) shows that there is no significant association with sleep quality and the food choice (p = 0,661). Moreover, another logistic regression was conducted for food choice 2 (green olives vs cocktail nuts), again showing no significant association between sleep quality and food choice (p = 0,581).
Table 3. Summary of logistic regression analyses for variables predicting the individual food choices (n = 117).
Food choice 1 Food choice 2 Food choice 3
Variable B SE B P B SE B P B SE B P
Sleep conditions -1.38 0.65 0.661 0.21 0.39 0.581 -0.55 0.40 0.168
Age 0.02 0.02 0.336 0.01 0.01 0.725 0.03 0.01 0.017*
Gender -0.91 0.60 0.133 -0.36 0.41 0.383 -0.40 0.43 0.351
Diet 0.84 0.64 0.191 0.53 0.40 0.183 0.34 0.40 0.393
Note. * significant at p < 0,05 level.
Lastly, the logistic regression for food choice 3 (chocolate bar vs. nut bar) showed that
although sleep quality was not significant (p = 0,168), age did show a significant effect. A test of the full model against a constant only model was statistically significant, indicating that the predictors act as a set reliably distinguished between choosing the chocolate bar or the nut bar (χ2 = 5,704, p = 0,017 with df = 2). Nagelkerke’s R of 0.089 indicated a very weak
relationship between prediction and grouping. Prediction success overall was 60,7%. The Wald criterion demonstrated that only age made a significant contribution to prediction (p = 0,017). Sleep quality was not a significant predictor. The Exp(B) value indicates that when age is raised by one unit (one year) the odds ratio is 1 time as large and therefore older respondents are 1 more time more likely to choose the nut bar.
To conclude, the analyses showed that there were no significant effects of sleep quality on
food choices. Hypotheses 1A and 1B, which stated that participants in the good (bad) sleep condition will be more likely to choose healthier (unhealthier) food options are thus rejected.
4.3.2. H2: moderation of willingness to pay for sleep quality
To measure the effect of the sleep conditions on the food choices, again the computed health score was used (ranging from 0 = only unhealthy choices, 3 = only healthy choices). This
used in the experiment. Firstly, the analysis will turn to respondents who were willing to pay versus respondents who were not willing to pay for the sleep quality improving pillow.
Secondly, as a stricter method, respondents who were willing to trade the Bol.com gift card to the pillow versus respondents who were not willing to do this are compared. Finally, an analysis was conducted with only respondents who were willing to pay for the pillow, which as previously mentioned ranged from 30 to 120 euros.
Category I - Willing to pay vs. not willing to pay
Firstly, I looked at the differences between respondents who indicated if were willing to pay vs not willing to pay for the Sound Asleep pillow. To get answers on the moderation
hypothesis, firstly a two-way ANOVA analysis was performed, as both the independent (sleep condition, good vs bad) and moderator (willingness to pay, yes vs no) are dichotomous. Results show there was not a significant effect of sleep quality on food choices at the p <.05 level (F (3, 113) = 0,24, p = 0,629). Moreover, WTP (pay) (F (3, 113) = 0,35, p = 0,551) and the interaction term (F (1, 113) = 0,29, p = 0,549) were not significant. The control variables diet (F (1, 65) = 3,08, p = 0,081), gender (F (1, 75) = 2,74, p = 0,101) and age (F (35, 65) = 4,44, p = 0,501) were not significant as well. The models with the individual food choices are summarized in table 4, 5 and 6 in the appendix. Here as well no significant results were revealed, expect for the control variable age - which showed a significant effect on the third food choice (nut bar vs chocolate bar) (F (1, 110) = 5,62, p = 0,019).
Category II - Willing to trade vs not willing to trade
Secondly, the analysis will now look at if people were willing to trade the Bol.com gift card they could win for the Sound Asleep pillow. For this analysis, again a two-way ANOVA analysis was performed, as both the independent (sleep condition, good vs bad) and
moderator (willingness to trade, yes vs no) are dichotomous. The results again show that sleep quality is not significant (F (1, 110) = 2,21, p = 0,140), as well as WTP (trade) (F (1, 110) = 0,52, p = 0,471), or the interaction term (F (1, 110) = 0,49, p = 0,482). The control variable age does show a significant effect on the health score (F (1, 110) = 3,92, p = 0,050),
emphasizing the older the respondent, the more healthier food choices are made. The ANOVA results with the individual food choices are summarized in table 7, 8 and 9 in the appendix. In these analyses, no significant results were revealed.
Category III - Willing to pay range
As shown in the descriptive analysis, people differed in what they were willing to pay for the Sound Asleep pillow (M = 50,51, SD = 17,81) compared to an average pillow. To analyze if there are trends visible within this group, a moderation analysis using PROCESS was used. The results show that the regression coefficient for the interaction term (XM) is 0,00 and is not statistically different from zero (t (98) = -0,01, p = 0,998). There is thus no need to continue the analysis – as the interaction term is not significant, moderation thus not takes place. PROCESS analyses was also conducted for the individual food choices (summarized in tables 10, 11 and 12 in the appendix); again, no significant results were shown for food choice 1; the regression coefficient for the interaction term (XM) is -0,00 and is not statistically different from zero (t (98) = -0,01, p = 0,997), or for food choice 2; the regression coefficient for the interaction term (XM) is 0,01 and is not statistically different from zero (t (98) = 0,51,
p = 0,610), and also not for food choice 3; the regression coefficient for the interaction term
(XM) is -0,02 and is not statistically different from zero (t (98) = -0,94, p = 0,350).
To conclude, the analyses showed that there were no significant effects in the relationship between sleep quality, food choices, WTP (looking at paying (and how much) and trading). Hypotheses 2A and 2B, which stated that personal value of sleep moderates the effect of the
good (bad) sleep condition - the higher the WTP (valuation of sleep), the more positively the good (bad) sleep condition influences choosing healthier (unhealthier) food choices - is thus rejected.
4.3.3. H3: effect of sleep conditions on switching behavior for price promotions
To answer this hypothesis, one-way between subjects ANOVAs were conducted to compare the effect of the bad sleep quality condition and the good sleep quality condition for the probability of switching due to a price promotion. Two separate analyses were performed: 1) the likelihood of switching from a healthy option to an unhealthy option due to a price
promotion, 2) the likelihood of switching from an unhealthy option to a healthy option due to a price promotion. Analyses are reported per category.
Category 1 – Drinks The descriptive analysis showed that respondents in the bad sleep and good sleep condition were not likely to switch from the healthy product (smoothie) to the price promoted unhealthy product (chocolate milk) (Mbad = 1,94, SD = 1,29 and Mgood = 1,88, SD = 1,27). Moreover, sleep did not have a significant effect on switching from a smoothie to chocolate milk (F (1, 100) = 0,09, p = 0,764).
The other way around, switching from chocolate milk to a smoothie was also not significant (F (1, 13) = 0,04, p = 0,852). Respondents in the bad sleep condition and the good sleep condition did not score high on the scale which measured switching intention from the unhealthy product (chocolate milk) to the price promoted healthy product (smoothie) (Mbad= 2,75, SD = 2,06 and Mgood = 2,73, SD = 2,01).
Category 2 – Snack Choices I
