Consumers product preference and its effect on overweight

(1)

Consumers product preference and its effect on overweight

Participation in physical activity and its effect on the relationship between the

consumers product preference and obesity

Master Thesis

MSc Marketing Management

(2)

Participation in physical activity and its effect on the relationship between

the consumers product preference and obesity

by

JORICK NILS VAN DER GALIËN

University of Groningen

Faculty of Economics and Business

MSc Marketing March 2017 Leeuwerikweg 52 7971 DT Havelte (06)-21910607 j.n.van.der.galien@student.rug.nl Student number 3180611

First supervisor: Prof. Dr. Laurens M. Sloot

(3)

Table of content

Abstract ... 4 1. Introduction ... 4 2. Literature review ... 6 Obesity ... 6 Buying behavior ... 8 Processed food ... 9

Possible causes for the growing obesity rates ... 10

3. Conceptual model ... 12

4. Central research question and sub questions ... 13

5. Methodology ... 14 Sampling design ... 14 Materials ... 14 Procedure ... 17 6. Results ... 18 Descriptive statistics ... 18 Correlation analysis ... 21

Factor analysis and grouping of variables ... 23

Hypotheses testing ... 24

7. Conclusion ... 28

8. Discussion and limitations ... 30

Discussion ... 30

Limitations ... 31

9. Managerial implications and contribution to literature ... 33

Managerial implications ... 33

Contribution to literature ... 33

Bibliography ... 34

Appendix 1: Qualitative research: expert interview ... 37

Appendix 2: Quantitative research: survey ... 38

(4)

3

Consumers product preference and its effect on overweight | Jorick Nils van der Galiën

List of figures

Figure 2. 1: obesity and overweight trend over the years in The Netherlands ... 7

Figure 2. 2: obesity and overweight trend over the years in the United States of America ... 7

Figure 2. 3: obesity and overweight trend over the years in the United Kingdom ... 7

Figure 3. 1: Conceptual model ... 12

Figure 5. 1: product preference processed food ... 15

Figure 5. 2: product preference caloric rich food ... 16

Figure 5. 3: formula body mass index (BMI) ... 17

Figure 6. 1: distribution answers processed food question ... 20

Figure 6. 2: distribution answers caloric rich food ... 20

Figure 6. 3: distribution participants BMI in groups ... 21

Figure 6. 4: distribution participants BMI from low to high ... 21

List of tables

Table 4. 1: Hypotheses ... 13

Table 6. 1: Excluding participants ... 18

Table 6. 2: Descriptive Statistics ... 19

Table 6. 3: Correlation analysis processed food ... 22

Table 6. 4: Correlation analysis, processed food, convenience and health ... 22

Table 6. 5: correlation analysis caloric rich food ... 22

Table 6. 6: correlation analysis caloric rich food, convenience and health ... 23

Table 6. 7: results factor analysis and reliability analysis ... 23

Table 6. 8: correlation analysis new caloric rich food variable ... 24

Table 6. 9: correlation analysis new caloric rich food variable, convenience and health ... 24

Table 6. 10: results multiple linear regression hypothesis 1 ... 25

Table 6. 11: multiple linear regression hypothesis 2 ... 26

Table 6. 12: Moderation effect PA on PP and CRP on BMI ... 27

(5)

4

Participation in physical activity and its effect on the relationship between

the consumers product preference and obesity

Abstract

In the past decades a growing problem is recognized. It is the problem of overweight and obesity amongst populations in western countries like The Netherlands. People with overweight, or even obesity do have lower life satisfaction and it is related with several diseases: diabetes, high blood pressure, asthma and for example depression. Obesity is also related to lower life expectancies. There are no clear shopper segments that indicate which focus leads more to overweight and which focus leads to a healthy weight. This raises a question: Is it possible to relate preferences for processed and caloric rich food to overweight and are these variables moderated by physical activity?

1. Introduction

As already is known, obesity is a growing problem all over the world, especially in western countries. The goal of this research is to create insight about potential factors that (partially) cause the problem of obesity and the buying behavior of consumers. For future research, this report can be a starting point for others to come up with ideas and proposals to reduce the problems that obesity brings along. Past research show that there are several factors that can cause obesity or to reduce it. E.g. physical activity. This research is investigating the relation between consumers preference for processed products and caloric rich food and obesity. During this research I read a quote which made me start thinking. “Obesity affects every aspect

of a people’s lives, from health to relationships.” It made me realize that the consequences of

obesity do not stop at just being fat or just having overweight. Lee (2011) states that obese people often feel lonely, do have fewer friends and obese people often have a lower self-esteem. The chance to suffer on depression and experience serious health issues increase when being obese. Furthermore, obesity is very expensive (Lee, 2011). Therefore, it is not just the problems to the body that is making obesity such a hard problem to tackle but is also a problem in people’s mind, a mental problem.

(6)

5

Do people with obesity have a different focus on products when it comes to: processed versus non-processed food and the number of calories they contain compared to their normal-weight fellow man?

This paper tries to create insight about the assumption that obese people have a different product preference than people with normal weight. This product preference can be specified in this paper by the preference for processed products and caloric rich food. The moderator tries to explain if physical activity does have a significant effect on people’s product preference and overweight. Do people who do participate in physical activity have a different product preference compared to those who do not participate in physical activity?

(7)

6

2. Literature review

Obesity is a problem in many countries, specifically in western countries which is mentioned before. This problem has multiple causes and during this study food preference is researched. If the preference for certain types of food is positively related with overweight and obesity. During this study, two types of products are considered: processed or non-processed food and caloric rich food or caloric poor food. Besides these direct relationships between the types of food and obesity, also physical activity is considered as a moderator of these relationships.

Obesity

Obesity refers to an increased amounts of body fat (Sturm, 2002). Obesity is a problem which literally is increasing in size. Primarily because overweight and obesity is caused by a cluster of factors. The measure of obesity is called the Body Mass Index, also known as BMI. For adults counts, a BMI > 30 refers to obesity. The World Health Organization accepted the Body Mass Index, further called BMI, as an indicator for under- and overweight (James, 2008). The boundaries are < 18,5 for underweight and > 25 for overweight. Everything in-between 18,5 and 25 refers to a normal BMI. The BMI is calculated by the weight in kilograms divided by the participants height squared (Ogden, Carroll, Kit, & Flegal, 2013).

Overweight and obesity are accompanied by negative consequences. People with obesity are often struggling with issues related to their own self-respect and self-esteem. Besides, obese people do, in general, have less friends and are more likely to suffer from depression (Lee, 2011). These are the mental issues that can be caused by obesity but there are also some serious health concerns which need to be addressed. Health problems related to obesity are diabetes, high blood pressure, high cholesterol, asthma and arthritis (Sturm, 2002). Overweight and obesity are significantly associated with these kind of health concerns (Mokdad, et al., 2003). Finally, obesity is a very expensive problem. Obesity even outranks smoking and drinking in its effect on personal health and health care (Sturm, 2002). For example, in the United States of America, nearly one-fifth of the total expenses on health are spend on obesity. Obesity is associated with higher costs for the society. It increases health care costs and medication (Sturm, 2002). Several direct and indirect costs can be linked to obesity (Mokdad, et al., 2003). Some of these costs are lost future earnings and lower life expectancies (Lee, 2011). There is even a debate which states that obesity should be a disease of its own (Sturm, 2002).

(8)

7

Figure 2. 1: obesity and overweight trend over the years in The Netherlands (CBS, 2017)

As the CBS already shows figures that in 2016 more than 40% of the Dutch adults has overweight, the national health and nutrition examination survey shows that in 2014 37,7% of the American adults was obese as can be seen in figure 2.2.

Figure 2. 2: obesity and overweight trend over the years in the United States of America (NCHS, 2017)

In the USA the obesity rates for men and women are extremely high. A research by “the state of obesity” found that these figures for men are 35% and for women 40,4% (Obesity Rates & Trends Overview, 2017).

(9)

8

Finally, in England a lot of people are suffering with overweight or even obesity. Figure 2.3 shows a significant increase since 1993. In 2015 58% of women and 68% of men were overweight or worse. Adults being obese increased from 15% in 1993 to 27% in 2015 (Niblett, 2017). These figures show the urgency of the problem named obesity.

Buying behavior

Kotler and Armstrong (2010) stated that by doing and learning, people acquire believes and attitudes. These factors influence them in their buying behavior. While, Abideen and Saleem (2011) explain consumer buying behavior by the principle that it is based on the concept and idea a person simply decides, at the spot, to buy something. People are influenced, in their purchase decision-making, not only by the tangible product. They are influenced by something that is called the total product. This includes other features that are related to the purchase. These features are for example: services, warranties, packaging and pleasantries (Kotler, Atmospherics as a Marketing Tool, 1973).

However, the buying behavior of people is not the same for every product. It differs e.g. for an iPad, financial services or a bunch of bananas (Kotler & Armstrong, Principles of marketing, 2010). In the book, “principles of marketing”, Kotler and Armstrong use a model which explain consumer buying behavior by the level of involvement and the degree among differences in brands. There are four types of buying behavior explained in this model. Complex buying-, variety-seeking buying-, dissonance-reducing buying- and habitual buying behavior. Complex buying behavior means that people are high involved and there are significant differences in brands. Variety-seeking buying behavior means low involvement but significant differences in brands. Dissonance-reducing buying behavior is the other way around, high involvement but few differences in brands. Finally, habitual buying behavior has low involvement and few differences in brands (Kotler & Armstrong, Principles of marketing, 2010).

(10)

9

can have an influence on the buying behavior of consumers (Foxall & Greenley, 1999). Sense modalities and information can be part of these environmental variables. Sound, sight and touch can increase the consumer experience. The environmental psychology model proposes an environment that produces and emotional state in a consumer that can be linked to the three emotional variables above: pleasure, arousal and dominance, in a buying situation (Adelaar, Chang, Lancendorfer, Lee, & Morimoto, 2003).

Finally, advertisement can play a significant role in changing consumers buying behavior. That is also the major aim of advertising (Abideen & Saleem, 2011). Since ultra-processed food is heavily advertised in the past decades, it plays a significant role in the behavior of consumers to buy these products (Moodie, et al., 2013). To link buying behavior to this research two types of products are considered. The preference for processed products and the preference for caloric rich food related to obesity.

To create feeling with the subject researched, an expert interview is conducted. With a fellow student of mine we have been to Amsterdam to interview Simon Bunt. An expert when it comes to buying behavior of consumers and the reasons behind it. Simon is a developer at Questionmark and gave us a lot of information about product preferences of consumers. High caloric food is often associated with taste of the food (Davidson & Swithers, 2004) and mister Bunt mentioned that after price, palatability is the most important driver of consumers to buy/prefer products. Therefore, this study focusses on processed or non-processed and high- versus low caloric products. The entire summary of the expert interview is elaborated in Appendix 1.

Processed food

A study by Monteiro (2009) came up with three types of food. The first type is minimally processed food, fresh products. The second group is food with substances of whole foods like flour, sugar and oil. The ingredients are originally used for preparation and cooking of meals with fresh products, minimally processed. Finally, Monteiro (2009) stated that there is this group called ultra-processed food. It consists substances from the second group with no or minimally amounts of fresh products from the first group. Also, salt or other preservatives are added. Products within this group often contain cosmetic additives like flavors and colors in most cases. This group contains products like burgers, frozen pizza and ready-to-eat pasta dishes but also biscuits, cereal bars and sugared drinks (Moodie, et al., 2013). Ultra-processed products are often more energy dense, contain more free sugars and saturated fat than non-processed or minimally non-processed food. (Moubarac, et al., 2014) (Moodie, et al., 2013). This type of food is very profitable for the manufacturers due to the low production costs, the long shelf-life and the high retail value it contains (Stuckler, McKee, Ebrahim, & Basu, 2012). Negative relationships can be linked to processed food such as: a high consumption of highly processed food such as fast food can be linked to an increase in body-weight (Pereira, et al., 2005) (Moubarac, et al., 2014). The study by Canella et al. (2014) also showed that ultra-processed food can be associated with a high BMI. This indicates overweight over even worse, obesity (World Health Organization, 2013). Hypotheses 1 of this research is:

(11)

10

Moodie et al (2013) state that the consumption of ultra-processed food does not have to be a problem when the amounts are little. If other healthy sources of calories are consumed next to the small amounts of ultra-processed food, it is harmless (Moodie, et al., 2013). But the reasons why processed food or ultra-processed food can be related to an increase in body-weight are the typically the high proportions of ultra-processed products and the high amounts of fat, salt and sugar (Canella, et al., 2014) (Moubarac, et al., 2014) (Moodie, et al., 2013). Furthermore, ultra-processed food products are advertised and sold by large transnational corporations. They are durable, palatable and often ready-to-consume (Moodie, et al., 2013). These are advantages from ultra-processed food in comparison with the fresh and non-processed counterpart. Especially palatability is having a lot of influence since mister Bunt mentioned that is was the most important driver when it comes food preference behind price.

The study by Moubarac et al. (2014) also shows a heavy increase in monetary share of ultra-processed food in the total household food budget. The same results are found by the World Health Organization (2013). Moubarac et al. (2014) found that the monetary shares of non-processed food and culinary ingredient foods dropped. The substantial growth of ultra-processed food has contributed to an increase of obesity and multiple other diseases that are diet-related (Moodie, et al., 2013). In 2012 already 75% of the total food sales came from processed food.

The ultra-processed products, which do not need any preparation or cooking are harmful for human health. Ultra-processed food is also called ‘convenience food’ or ‘fast food’ because of this attribute. This attribute makes these kinds of products also very suitable for excess eating or even obesity (Monteiro, All the harmful effects of ultra-processed foods are not captured by nutrient profiling, 2009). Mister Bunt mentioned during the expert interview that convenience is indeed an important driver for people to consume.

The other independent variable in this research is the preference for caloric rich food. Food contains a certain number of calories, which differ across products. Ultra-processed is typically high in calories (Canella, et al., 2014). A disturbing fact is that several studies showed that obese people do have a higher preference for high-calorie food compared to their normal-weight fellow man (Mela, 2001) (Rissanen, et al., 2002). Another study by Burton, Smit and Lightowler (2007) showed that people with a higher BMI crave more for high-calorie food than people with a lower BMI. Does it stop by preference and craving for obese people? There are several studies which found that behavioral interventions can play a role in decreasing the preference for caloric rich food (Martin, et al., 2011). Besides, there is evidence that obese people show higher brain activity when watching pictures of caloric rich food (Murdaugh, Cox, Cook, & Weller, 2012). Therefore, it might be harder for obese people the lose weight and to keep losing weight (Murdaugh, Cox, Cook, & Weller, 2012). The second hypothesis is:

Hypothesis 2: The preference for caloric rich food is positively related to overweight.

Possible causes for the growing obesity rates

(12)

11

people can deal with themselves. So, it should not be too hard to fight obesity. Exercise more and eat less is. Practice shows that it is not that easy (Spear, et al., 2007).

People are more and more inclined to eat outside their own home or the buy ready-to-eat meals (Moodie, et al., 2013). This tendency can be the case because more and more women are working next to their husband and there is an increase in single households. Since the parents are more and more inclined to eat outside home or to buy ready-to-eat meals, the children of these parents do have less examples of cooking skills within their own houses. Of today’s young adults, 95% eat at a fast food restaurant or full-service restaurant once a week (Larson, Neumark-Sztainer, Laska, & Story, 2011). If they do not learn these cooking skills at home, they should be learned at school. However, also at school cooking skills are less part of the school curricula (James, 2008).

Physical activity and obesity are two terms which are connected and associated with each other for a long time (Ewing, Schmid, Killingsworth, Zlot, & Raudenbush, 2003). Usually the rule, if you participate enough in physical activity the chance for obesity declines applies. Participation in sport can be a foundation for adult health and a component for obesity prevention (Alfano, Klesges, Murray, Beech, & McClanahan, 2002). What will be investigated is the moderating effect of “physical activity” on the product preferences and obesity. Distinction will be made between people who do not participate in physical activity, people able to manage the recommended time of physical activity and a group which do participate in physical activity but do not participate enough to participate the recommended time (Ewing, Schmid, Killingsworth, Zlot, & Raudenbush, 2003). It is also known that when people lower their amount of time spend on physical activity the chance increases that these people gain weight compared to the people who keep being active for an equal amount of time (Reiner, Niermann, Jekauc, & Woll, 2013). Finally, the third hypothesis is divided into two sub questions:

Hypothesis 3A: The relationship between the tendency to buy processed products and overweight moderated by physical activity.

(13)

12

3. Conceptual model

During this research the dependent variable will be obesity and overweight. This variable will be separated in underweight, normal weight and overweight. This variable will be measured using the Body Mass Index (BMI).

The moderator of this research is participation in physical activity. Participation in physical activity will contain three different groups. The first group of participants is a group that does not participate in physical activity at all. Secondly, there is a group of participants that can manage the recommended time of participation in physical activity. Finally, there is a group that does participate in physical activity but does not participate enough to state that they manage the recommended time of participation.

The last variables in the research are the independent variables: the preference for processed food and the preference for caloric rich food. These variables are used to try to explain tendencies between the preferences for types of food and overweight. Furthermore, these tendencies are checked to be moderated by physical activity.

There are certain control variables that can be used to control the hypotheses. These variables are socio-demographic variables: gender, age, household size, education and income.

The entire conceptual model can be seen in Figure 3.1

(14)

13

4. Central research question and sub questions

During this research six hypotheses are used. These questions are already mentioned in the research parts above. To make a clear overview about what this research is about, the six questions are mentioned in table 4.1:

Number Hypotheses

1 The preference for processed products is positively related to overweight.

2 The preference for caloric rich food is positively related to overweight.

3A The relationship between the tendency to buy processed products and overweight is negatively moderated by physical activity.

3B The relationship between the tendency to buy high-caloric products and overweight is negatively moderated by physical activity.

Table 4. 1: Hypotheses

These questions will try to give insight about certain tendencies and relationships between different variables. First, the relationship between the preference for processed products and obesity will be investigated. Furthermore, the relationship between the preference for caloric rich food and obesity will be investigated. Finally, participation in physical activity and its moderating effect on the product preferences and obesity will be investigated as well. These answers on these hypotheses together give an answer to the central research question: Can

(15)

14

5. Methodology

Sampling design

Each research would be optimal if the entire population can be used. However, it is nearly impossible to realize this. This is because the population is almost finite (Etikan, 2016). Therefore, convenience sampling is used for this study. Convenience sampling is a type of nonprobability sampling. It is also called Accidental Sampling, the participants are often easy accessible, live nearby or are willing to participate. There a few advantages of convenience sampling like: it is affordable, easy and the participants are available (Etikan, 2016). There are also disadvantages for convenience sampling and the main disadvantage is that the participants often are biased. Finally, also outliers are a serious disadvantage of convenience sampling (Etikan, 2016).

The target group of the survey are Dutch adults (> 18 years). Therefore, all the survey questions were in Dutch to ensure valid responses and to enlarge the target audience. This is necessary since still not everyone does understand the English language completely and therefore might misinterpret the questions asked. Before the survey went online, several pre-tests were conducted. These pre-tests are conducted to minimize the chance that people misinterpret the question and therefore must be removed from the dataset. It is basically to eliminate problems beforehand, so they do not cause any trouble during the data collection. After the phase of pre-testing several questions are changed to make better to interpret. The survey is made in Qualtrics and provides a link which can be used for distribution. The distribution of the survey is done online via Facebook, email and WhatsApp. To enlarge my distribution area several other persons shared the link with colleagues, friends or co-students.

The study is designed with the online survey platform Qualtrics. The preference for processed products and the preference for caloric rich products are the independent variables, and obesity is the dependent variable. The moderator is physical activity. Furthermore, there are multiple control variables. These variables are socio-demographic variables, and these are age, gender, household size, education and income. Correlation analyses and multiple linear regressions are applied as well as a check for moderations. To get reliable data about the participants length and weight, questions about length and weight are asked at the very end of the survey. If the participant knows the purpose of the survey his or her answers can be biased.

To check for moderating effects, the PROCESS macro by Hayes will be conducted obesity as the dependent variable, processed- versus non-processed products & high- versus low-calorie products as the independent variable and physical activity is the moderator of the study.

Materials

Demographic question. First the participants were asked question related to their

(16)

15

far below average, below average, a little bit below average, average, a little above average, above average and far above average. This question is followed by education. What is you highest level of education: basisschool, MAVO, HAVO, VWO, MBO, HBO or Universiteit. These variables are used as a control variable during the study

Physical activity. To get a good insight about someone’s physical activity, several questions are

asked. One question is about their indication of the intensity of their job. Followed by a question about related to the number of hours a week the participant is active in sports. Finally, which percentage of the participants’ working day, they are sitting in a chair. These questions deliver a good picture about the physical activity of the participant and this is taken as a moderator during the study.

Product preference test (1). The participants were shown five sets of two comparable products

that differ on the level of food processing, e.g. canned green beans (processed) versus fresh green beans (non-processed), figure 5.1. These

five sets of two comparable products do represent product categories from a retail store and these five cover a great part of a retail store. The products needed to have a clear distinction between processed and non-processed products which can be seen by the participant. With this comparison there is always one processed option and one non-processed option. These products are chosen based on the clear difference between the level of processing. Furthermore, because in general the participant has huge familiarity with the product and therefore is able the answers the question correctly. Besides familiarity, it is expected that most of the participants do buy these products on a weekly basis. Recognition is easy for the processed and non-processed products to valid data. Moreover, different

product categories are chosen to create a clear picture about the entire retail store. The participant is asked which product they are more likely to buy. This is done on a seven-point bipolar scale with the fourth point as a neutral. The processed- and the non-processed product are randomly assigned on the right- or the left-hand side of the scale. Afterwards all individual scores will be summed up and divided by the seven products. The scores need to be adjusted before they are summed up to get a useful score. With the adjusted scores every processed product is on the right-hand side and the non-processed product is on the left-hand side. This means that the higher the score, the more a participant is likely to buy processed food. So, in this study the first-, second- and fourth comparison are adjusted.

(17)

16

Product preference test (2). Again, the participants were shown five sets of two comparable

products. This time they differ on the level caloric richness, e.g. Coca-Cola Regular (caloric rich) versus Coca-Cola Light (caloric poor), figure 5.2. There is one typically caloric rich product and one typically caloric poor product. The participant is asked which product they prefer more. These five products are chosen because they are regularly bought by consumers and there is a clear distinction between the product with a lot of calories and the product with fewer calories. This distinction must be clear to receive valid results. It is also expected that every participant buys these kinds of products regularly and that the participant is familiar with the product. The seven-point bipolar scale with the fourth point as a neutral is the same as the previous test and the high- versus low-calorie products are again randomly assigned on the right- or the left-hand side of the scale. Like the previous preference test, the adjusted scores are calculated. For caloric products the first-, fourth- and fifth question is adjusted to create a valid scored on this question.

Measurement constructs.

Measurement of health conscious: high- versus low-calorie products

To measure the health consciousness, a scale developed by Michaelidou and Hassan (2008) is used. The scale consists of six items which are measured on a seven-point Likert scale (from completely disagree to completely agree). In this study only five items are used since the sixed item caused misinterpretations. The five items used to measure the health consciousness of the participant are described in Appendix 3.

Measurement of convenience: processed- versus non-processed product

A scale developed by Scholderer and Grunert (2005) is used to measure the attitude of the participant towards convenience. This scale consist four items measured on a seven-point Likert scale (from completely disagree to completely agree). The items to measure the attitude of the participant towards convenience are described in Appendix 3.

Measurement of price conscious

To measure the impact of price consciousness towards buying products a scale developed by Donthu and Garcia (1999) is used. The scale consists of four items measured on a five-point Likert scale (from completely untrue to completely true). The items to measure price consciousness are described in Appendix 3.

Body Mass Index. Finally, the survey ends with questions regarding to the participants BMI.

These questions were open questions about the participants length, weight and how the participant compares their own weight with their own length. These final questions give an idea

(18)

17

about someone’s ideal BMI. To calculate the participants BMI, the weight of the participant is divided by the squared length.

Figure 5. 3: formula body mass index (BMI)

Procedure

(19)

18

6. Results

Descriptive statistics

The total amount of respondents is 262. During the analysis of the dataset 216 respondents are considered. This means that several respondents are excluded from the dataset. Excluding respondents from the dataset happened for various reasons. Some respondents did not fill their weight and therefore are excluded because the BMI could not be calculated. Others still misinterpreted the questions and therefore did not give reliable results. Two persons were 17 and therefore not within the target group. Others did not participate seriously based on the answers given at question 14 and 15 of the survey. Finally, someone became sick and the answers he or she gave could not be compared with a few months ago and is therefore excluded. The reasons why participants are excluded together with the corresponding number can be seen in table 6.1

Total amount of participants 262

Reason for exclusion Amount

Did not fill in their length or weight 6

Questions were misinterpreted 8

Younger than 18 2

No serious participation 29

Sick, the results cannot be compared with a few months ago

1

Number of participants considered 216

Table 6. 1: Excluding participants

The total amount of participants of which the results are used during the hypotheses testing is 216. General results from the 216 participants are shown in table 6.2. The table shows the general results of the questions about gender, age, household size, income, education and the most important one, their BMI.

Gender Male: Female: 41,7% 58.3% Age Range: 18-30 31-50 51-65 >65 M = 37.3, SD = 15.36 45,8% 27,3% 23,2% 3,7%

Household size Range:

(20)

19

Income Range:

Far below average Below average A bit below average Average

A bit above average Above average Far above average

M = 3.5, SD = 1.95 19,9% 20,4% 11,6% 12,0% 12,5% 19,4% 4,2% Education Range: LWO/VMBO MAVO HAVO VWO MBO HBO WO 1,4% 4,2% 11,1% 2,8% 23,6% 42,6% 14,3% BMI Range: Underweight Normal weight Over weight Obese M = 24.9, SD = 4.32 0,5% 59,2% 40,3% 10,6%

Table 6. 2: Descriptive Statistics

Based on the figures from the CBS (2018) the Dutch population is divided into 49,6% men and 50,4% woman. Secondly, the average age of the Dutch population in 2017 was 41,6 years based on figures of the CBS. The average age in the sample size is 37,3. This is a little below the average of the entire population. Furthermore, the CBS state that the average household size is 2,16 while the average household size of the sample size is 2,78. This is close to the average household size of the Dutch population. The distribution of the sample size is not too bad. The average income of the Dutch population is modal of course. The mean value based on the question about their monthly earnings do come close to modal. The mean value of the 216 participants is 3.52 on a scale from one to seven. According to overweight the CBS showed figures in 2016 that 49,2% of the Dutch adults, which is the target group of this study. The 40,3% of the participants with overweight within this study do come close to that number. According to these figures the sample size is a pretty good representation of the Dutch population if the minor differences are considered.

(21)

20

Figure 6. 1: distribution answers processed food question

The answers to this question are nearly equally distributed. Overall, for each product comparison, the non-processed option (point 1, 2 and 3) enjoys the highest preference. The average always enjoys the lowest preference but only the fourth point on the one-to-seven Bipolar scale is used for average. The preference for processed products (point 5, 6 and 7) is always preferred second.

The distribution for caloric rich products is not that equally divided as the distribution of the processed products. In three out of five times (Bread, Milk and Cheese), the caloric poor option enjoys the highest preference as can be seen in figure 6.2. The other two times (Cola and Chips), the caloric rich option enjoys the highest preference. For chips the average score (point four on the Bipolar scale) enjoys a higher preference than the caloric poor option. Even though for the average score only one option is considered and for the caloric poor score three options are considered.

Figure 6. 2: distribution answers caloric rich food 59,7% 11,6% 28,7% 57,9% 17,1%25,0% 40,7% 24,1%35,2% 64,4% 11,1%24,5% 46,3% 12,5% 41,2% 0,0% 10,0% 20,0% 30,0% 40,0% 50,0% 60,0% 70,0% 80,0% 90,0% 100,0% N o n -p ro ce ss e d A ve rag e Pro ce ss e d N o n -p ro ce ss e d A ve rag e Pro ce ss e d N o n -p ro ce ss e d A ve rag e Pro ce ss e d N o n -p ro ce ss e d A ve rag e Pro ce se e d N o n -p ro ce ss e d A ve rag e Pro ce ss e d

Beans Minched meat Tomato sauce Cheese Pineapple

P er ce n tag es

Distribution based on: non-processed, average and processed

Distribution preference for processed food

31,9%_23,6%44,5% 65,3% 4,6% 30,1% 56,0% 13,9%30,1% 39,3% 22,7%38,0%_16,2%25,5% 58,3% 0,0% 20,0% 40,0% 60,0% 80,0% 100,0% Calo ri c p p o r A ve rag e Calo ri c ri ch Calo ri c p o o r A ve rag e Calo ri c ri ch Calo ri c p o o r A ve rag e Cal o ri c ri ch Calo ri c p o o r A ve rag e Cal o ri c ri ch Calo ri c p o o r A ve rag e Cal o ri c ri ch

Cola Bread Milk Cheese Chips

P er ce n tag es

(22)

21

Finally, the distribution of the participants BMI can be seen in figure 6.3, below. Within the sample size only one person is underweight. Therefore, the low percentage of 0,5%. Nearly 60% of the participants has a normal weight. Which is a bit above the population figures. Over 40% was overweight and more than 10% of these participants are obese.

Figure 6. 3: distribution participants BMI in groups

The distribution of the BMI of the participants is given in figure 6.4. The scores gradually run up from 18,42 until 37,96. There is one participant with an extreme value of 50,08.

Figure 6. 4: distribution participants BMI from low to high

Correlation analysis

To find out if those questions that aim to answers the same hypothesis can be combined in one variable a correlation analysis is conducted.

First, the five sets of two comparable products that differ on the level of food processing aim to answers the hypothesis 1: The tendency to buy processed food is positively related to

overweight. The results of the correlation analysis are shown in table 6.3.

0,5% 59,2% 40,3% 10,6% 0,0% 20,0% 40,0% 60,0% 80,0% 100,0% Underweight < 18,5 Normal weight 18,5 - 25 Overweight > 25 Obese > 30 P er ce n tag es

Distribution based on: underweight, normal weight, overweight and obese

Distribution of the participants BMI

0 10 20 30 40 50 60 1 6 _{11 16 21 26 31} _{36 41 46 51 56 61 66 71 76 81 86 91 96} 101 106 111 116 121 126 131 136 141 146 151 156 161 166 171 176 181 186 191 196 201 206 211 216 B MI v lau e

The distribution of the BMI (from low to high)

BMI values from low to high

(23)

22

Correlation analysis based on the five sets of two comparable products that differ on the level of food processing Beans Minced meat Tomato sauce Cheese Pineapple Beans 1 0.386 (0.000) 0.190 (0.005) 0.273 (0.000) 0.331 (0.000) Minced meat 0.386 (0.000) 1 0.100 (0.142) 0.405 (0.000) 0.206 (0.000) Tomato sauce 0.190 (0.005) 0.100 (0.142) 1 0.156 (0.022) 0.355 (0.000) Cheese 0.273 (0.000) 0.405 (0.000) 0.156 (0.022) 1 0.184 (0.007) Pineapple 0.331 (0.000) 0.206 (0.000) 0.355 (0.000) 0.184 (0.007) 1

Table 6. 3: Correlation analysis processed food

The correlation analysis showed that each variable statistically significant correlates with the other variables. This means that a preference for processed products in one comparison often indicates a preference for the processed option on another comparison as well. However, there is one exception. Minced meat does not correlate with tomato sauce on a statistically significant level. All other variables correlate with each other on a significance level of P = <.05. To find support for this variable, the score on the preference for processed products is tested for correlation with convenience and health. The scale about convenience and health are about the level of convenient shopping and the engagement of the participant in their health. This correlation analysis showed the results that can be seen in table 6.4.

Correlation analysis based on the preference for processed food, convenient shopping and engagement in health

Preference Convenient Health Preference 1 0.241 (0.000) - 0.090 (0.186) Convenient 0.241 (0.000) 1 - 0.113 (0.097) Health - 0.090 (0.186) - 0.113 (0.097) 1

Table 6. 4: Correlation analysis, processed food, convenience and health

This correlation analysis showed that the preference for processed products is correlated with convenient shopping behavior. However, the preference for processed products is not related with the participants’ engagement in their health. Finally, the convenient shopping behavior and the participants’ engagement in health are not correlated as well.

Secondly, the five sets of two comparable products that differ on the level of caloric richness aim to answers the hypothesis 2: The tendency to buy caloric rich food is positively related to

overweight. The correlation analysis showed the following results, table 6.5.

Correlation analysis based on the five sets of two comparable products that differ on the level of caloric richness

Cola Bread Milk Cheese Chips

Cola 1 0.208 (0.002) 0.189 (0.005) 0.169 (0.013) 0.193 (0.004) Bread 0.208 (0.002) 1 0.344 (0.000) - 0.305 (0.000) - 0.253 (0.000) Milk 0.189 (0.005) 0.344 (0.000) 1 - 0.054(0.427) 0.001 (0.986) Cheese 0.169 (0.013) - 0.305 (0.000) - 0.054(0.427) 1 0.220 (0.001) Chips 0.193 (0.004) - 0.253 (0.000) 0.001 (0.986) 0.220 (0.001) 1

Table 6. 5: correlation analysis caloric rich food

(24)

23

Correlation analysis based on the preference for caloric rich food, convenient shopping and engagement in health

Table 6. 6: correlation analysis caloric rich food, convenience and health

The correlation analysis showed that the preference for caloric rich food is not correlated with either convenient shopping or engagement in the participants’ health. Also, convenient shopping and engagement in health are not correlated but this was shown in table 6.4 as well.

Factor analysis and grouping of variables

The class of procedures used to reduce data is called factor analysis (Malhorta, 2010). To group data, the data need to be checked on the level of appropriateness of the factor analysis using the Kayser-Meyer-Olkin (KMO) measure of sampling accuracy. Besides the Barlett’s test of sphericity need to be conducted. The score on the KMO measure needs to be between 0,5 and 1,0 while the Barlett’s test of sphericity needs to be significant.

After the KMO measure and the Barlett’s test of sphericity, the number of variables that can be grouped need to be determined. This can be done by looking at the eigenvalues of the different variables. The eigenvalues need to be > 1. Another option is to look at the percentage of variance explained by the variables. A single variable need to explain at least 5% or together the variables need to explain > 60%.

After determining the right number of variables that Cronbach’s alpha of the variables combined to be > 0.6. The Cronbach’s alpha ensures the internal consistency of the created variable. In table 6.7, several items are measured on the KMO, Barlett’s test of sphericity, eigenvalues, number of variables the explain > 5%, the total variance explained and the Cronbach’s Alpha. Kayser-Meyer-Olkin Barlett’s test of sphericity Extracting all values provided an eigenvalue of Number of variables that explain > 5% Total variance explained Cronbach’s Alpha Preference for processed food .67 <.001 2.048 All variables 41,0% .64 Preference for

caloric rich food with every variable (5) .51 <.001 1.628 All variables 32,6% .28 Computed preference for caloric rich variable (3) .58 <.001 1.501 All variables 50% .50 Health scale .82 <.001 3.038 All variables 60,8% .84 Convenience scale .70 <.001 1.935 All variables 48,4% .64

Table 6. 7: results factor analysis and reliability analysis

(25)

24

Correlation analysis based on the three sets of two comparable products that differ on the level of caloric richness

Cola Bread Milk

Cola 1 0.208 (0.002) 0.189 (0.005) Bread 0.208 (0.002) 1 0.344 (0.000) Milk 0.189 (0.005) 0.344 (0.000) 1

Table 6. 8: correlation analysis new caloric rich food variable

The three variables are all correlated on a significance level of P = > .05. Based on this new variable, another correlation analysis is conducted. The analysis showed that nothing changed. The preference for caloric rich food still is not correlated to either convenient shopping and engagement in the participants’ health. The results can be seen in table 6.9.

Correlation analysis based on the preference for caloric rich food, convenient shopping and engagement in health

Table 6. 9: correlation analysis new caloric rich food variable, convenience and health

The correlation analysis between the preference for caloric rich food and convenient shopping and the engagement in health is conducted to see if the variable of the preference for caloric rich food could be replaced due to the low level of reliability, Cronbach’s Alpha of .50.

Hypotheses testing

1) The tendency to buy processed food is positively related to overweight

To test the first hypotheses, the tendency to buy processed food is positively related to

overweight, I conducted a correlation analyses and a multiple linear regression to predict the

BMI of the participant based on two independent variables. Within the correlation analyses, the processed food score and the participants BMI are used as the variables. These variables are also used in the multiple linear regression. The score participants scored on processed food preference is used as an IV, as well as the score on convenience scale while the BMI of participants is used as the DV. There are also several control variables added to the multiple linear regression. These control variables are: gender, age and household size.

Based on the results of the correlation analysis, the preference for processed food is not related to the participants BMI, R = -.06, P = .34. The correlation analysis showed that the two variables are negatively correlated with an R = -.06. However, the correlation is not a statistically significant. This tells that a higher preference on processed products leads to a decrease in the participants BMI. This would support the exact opposite of the hypothesis. However, the results show that there is not enough evidence to state that it accounts for the entire Dutch population since it is an insignificant analysis.

(26)

25

control variables are included. These control variables are gender, age, household size, education and income. Of all the variables considered, only gender, age and income are statistically significant. The independent variables, preference for processed food has a P = .49. Household size and education are not statistically significant as well. The level of education is marginally significant with a P = .08. Age, gender and income are statistically significant. Age and income with both a P = .03 and age with a P = <.001. Gender and income are negatively related to the participants’ BMI. These results can be interpreted that woman do have a lower BMI than men and the more the participant earns the lower the BMI. Age is positively related to the participants’ BMI which means that the older a participant gets, the higher their BMI will be. The results of the multiple linear regression can be seen in table 6.10. The full elaboration of the regression analysis can be seen in Appendix 4.

Results of the regression processed, food score on BMI Processed

food

Gender Age Household

size Education Income Constant: 26.65 β -.15 -1.28 .12 -.15 -.35 -.40 P .49 .03 <.001 .48 .08 .03

Table 6. 10: results multiple linear regression hypothesis 1

Processed food is measured between 1 and 7 and the higher the score means that the participants preference is more towards processed food. The participants’ BMI decreases with .15 point for every point higher on the processed food score. This means that the higher the preference for processed food the lower the BMI. However, the results are insignificant and therefore it cannot be stated that these results that would account for the entire population. Secondly, the control variables household size and education do not have a significant effect on the preference for processed food and the participants BMI. However, gender, age and income do have a significant effect on the relation with a significant level of P = <.001 for age and P = .03 for gender and income. Based on these results it can be stated that the older a person gets, the BMI will increase, woman do have a higher BMI than men and the more a person earns the lower the BMI.

However, the tendency to buy processed or non-processed products does not have an influence on someone’s BMI. Therefore, the hypothesis cannot be accepted and is therefore rejected.

2) The tendency to buy high-calorie food is positively related to overweight

The second hypotheses, the tendency to buy high-calorie food is positively related to

overweight, is also tested via correlation analysis and a multiple linear regression. The variables

(27)

26

During the multiple linear regression analysis, the same control variables are used as in the previous hypothesis: gender, age, household size, education and income.

To start, the correlation analysis is conducted. Based on the results the preference for caloric rich food is not statistically significant related to the participants BMI, R = -.02, P = .77. The results state that the preference for caloric rich food is negatively related with the participants BMI. It tells that the higher the preference for caloric rich food, the lower the participants BMI. These results support the exact opposite of the hypothesis. Only the results are insignificant and therefore it cannot be stated that there is enough evidence to say that these results are generalizable for the population. To find significant results for the second hypothesis, the independent variable is tested for correlation with the convenient shopping scale and the participants’ engagement in their health. If there was a significant correlation, the independent variable could be replaced with an academic scale based on the participants’ engagement in their health. This is not the case and therefore, the score on the preference for caloric rich products is used as an independent variable.

Next to the correlation analysis, a multiple linear regression analysis is conducted. The preference for caloric rich food is used as the independent variable. The participants BMI is used as the dependent variable. Furthermore, age, gender, household, education and income are used as control variables. The analysis showed that the dependent variable (BMI) is for 17% explained by the independent variables since the adjusted R² = .17. The independent variable, preference for caloric rich food has a (M = 3.51, SD = 1.58) A significant regression equation was found F (6, 209) = 8.38, P = <.001. Like the previous hypothesis, the independent variable is again insignificant, P = 13. With household size as an exception, all the control variables are statistically significant. The results are showed in table 6.11. Gender and income are statistically significant with P = .03. Education is statistically significant with a P =.05. Finally, age is a significant predictor with a P = <.001. The full elaboration of the regression analysis can be seen in Appendix 4.

Results of the regression, caloric rich food score on BMI Caloric

rich food

Gender Age Household

size Education Income Constant: 27.11 β -.26 -1.24 .13 -.15 -.40 -.42 P .13 .03 <.001 .49 .05 .03

Table 6. 11: multiple linear regression hypothesis 2

(28)

27

with their weight. Older people are struggling harder to keep a low BMI, since the BMI increases in time someone gets older.

3) The buying behavior of people, regarding processed versus non-processed food and high- versus low-calorie food, will be negatively moderated by participation in physical activity,

when related to overweight.

To test whether physical activity moderates the direct effects in this study, the PROCESS macro by Hayes is conducted multiple times. In the first place, the PROCESS macro is conducted with the score on processed products as an IV and the BMI as a DV. Secondly the PROCESS macro is used with the score on caloric products as an IV and BMI as the DV.

Every time the PROCESS macro by Hayes is conducted, (model 1, 1000 bootstrap samples) are used. To check whether moderation has happened we look at the P-values. If the p-value < .05 there is a moderation effect. Besides the p-value, check whether the confidence interval does include zero. This means that the β-value is likely to be more than zero.

To test the negative moderating effect of the participation in physical activity on the relation between the preference for processed food and overweight, the PROCESS macro by Hayes is conducted. The preference for processed products is used as an independent variable with the BMI as the dependent variable. The participation in physical activity is used as a moderator. The results can be seen in table 6.12.

Moderation effect of physical activity (PA) on the preference for processed products (PP) and the preference for caloric rich products in relation to overweight (BMI)

β P 95% confidence interval

Processed product .138 .06 -.004 .280

Caloric rich food .063 .24 -.043 .170

Table 6. 12: Moderation effect PA on PP and CRP on BMI

This shows that physical activity does have a moderating effect on the preference for processed products and overweight. This effect is marginally significant. The confidence interval just includes zero. The lower level is at point -.043, just below zero. The hypothesis state that there is a negative moderating effect. However, the results show that there is a positive moderating effect. This means that physical activity moderates the relation between the preference for processed products and overweight. The results for hypothesis 3A are therefore not accepted although the test was nearly significant with a significance is level of .004 above the statistically significance level of .05. Also, the confidence interval barely includes zero. These are half of the results for the third hypotheses of this study: the buying behavior of people, regarding

processed versus non-processed food and high- versus low-calorie food, will be moderated by participation in physical activity, related to overweight.

(29)

28

overweight. The significance level is above .05 with a P = .24 and the confidence interval includes zero. The second part of hypothesis 3 (hypothesis 3B) can not be accepted and is therefore rejected.

Number Hypotheses Supported

1 The preference for processed products is positively related to overweight.

No

2 The preference for caloric rich food is positively related to overweight.

No

3A The relationship between the tendency to buy processed products and overweight is negatively moderated by physical activity. No

3B The relationship between the tendency to buy high-caloric products and overweight is negatively moderated by physical activity.

No

Table 6. 13: Overview hypotheses supported or not

(30)

29

7. Conclusion

The aim of the study is to give insight about the relations between the preference for processed products and the preference for caloric rich food on overweight. Besides these direct effects, the PROCESS macro by Hayes is conducted to test for moderation. Besides these variable, multiple control variables are used. These control variables are gender, age, household size, education and income. To gather data to give these insights, both qualitative research and quantitative research is conducted. The analyses can be used to create marketing advertisements, targeted at shopper segments that are more likely to have overweight.

The results of the analysis showed that both the preference for processed products and the preference for caloric rich products are insignificant predictors for the participants BMI. This corresponds to the correlation analyses. The correlation between the preference for processed products and the preference for caloric rich food with BMI were insignificant in both cases. However, physical activity scored close on a moderation effect between the preference for processed products and overweight. However, this moderation effect is in the opposite direction compared to the statement in the hypothesis. Physical activity does not have a significant moderating effect between the preference for caloric rich products and overweight. The results showed that none of the hypothesis can be supported by the data gathered.

The control variables were significant in most cases. Age turned out to be highly significant in both hypotheses. The same applies for the income- and gender variable. Education was statistically significant in the second regression where it was marginally significant in the first regression. Finally, household size was an insignificant variable in both regression. The results showed that woman do, in general, have lower BMI than men do, and older people often have a higher BMI. Another conclusion that can be drawn is that the more someone earns the lower the BMI. This also applies for education, the higher the level of education, the lower the BMI. Woman may be more concerned about their weight. Their concern about their weight might cause for a healthier lifestyle compared to men. This could also be an explanation why woman in general do have a lower BMI compared to men. The results also show that older people get, the higher the BMI. An explanation to this result can be that older people are less active in sports or other kinds of physical activity. Their change in lifestyle on these levels might results in a higher BMI. Especially if their eating behavior does not change. Finally, education and income do have a significant effect on overweight. I do not really have an explanation for education. People with a higher education make healthier choices because they might be better to oversee the consequences. Income might play a significant role because healthy food often is more expensive. People with lower incomes might be therefore more inclined to buy unhealthy food.

(31)

30

for future research. This approach might lead to a decrease in the level of overweight and obesity amongst the Dutch population.

8. Discussion and limitations

Discussion

First, the aim of this research is to find out if the preference for processed products and the preference for caloric rich products do have a significant effect on overweight. Secondly, this research aims to find significant results for a moderating effect between the two preferences and the participation in physical activities.

Some people do their groceries on a weekly basis. Others do them daily. This means that people must make decisions regarding products very often. Not only because they do their groceries so many times. Also, for the different products they buy each time. For each product they buy, they need to decide. Do I take the healthy kind and do I not? Do I take the processed kind because it is so convenient or do I not? Because price and palatability are the most important driver of someone’s buying behavior, it is hard for people to choose for the healthy and non-processed kind. The unhealthy product is often cheaper and perceived as more delicious, which makes it harder for people to choose for their health. This might also explain why income has a negative effect on a person’s BMI. People which do have more money to spend, are better able to buy the healthier problem.

This study indicates that it helps when people are more engaged with their health. Once someone is engaged with their health, they often how a lower BMI. This means that to decrease the problem of overweight and obesity, people need to get more and more engaged with their health. When people are more engaged with their health they probably make better choices regarding their product preference. When people start making better choices it might eventually lead to a decrease in overweight and obesity. These choices related to healthy products however this study could not support that preferring low-calorie products significantly leads to a lower BMI.

The study by Pereira, et al. (2005) stated that processed food can be linked to an increase in body-weight. These findings are supported by Mourabac, et al. (2014). This study does not find significant evidence to support these findings.

The study by Burton, Smit and Lightowler (2007) stated the people with a higher BMI crave more for high-calorie food. This would suggest that a preference for caloric rich products is positively related with overweight. Hence, this research did not find significant results to support this statement. Also studies by Mela (2001) and Rissanen, et al. (2002) have similar statements. Obese people do have a higher preference for caloric rich food compared to a normal weight person. The results in this study do not support these statements based on the its results.

(32)

31

findings by Ewing, et al (2003) maybe could have been supported if a direct relation was researched. Alfano, et al. (2002) and Reiner, et al (2013) have similar statements as Ewing, et al. therefore a direct relation between physical activity and overweight might be better to investigate.

Limitations and future research

Since every study has its flaws, also this study is not perfect. Therefore, there are some limitations that are worth to mention. The generalizability of this study is quite good. There is a lot of variance in income and education. Also, the participant’s BMI has a lot of variance which is the most important variable in the study. Since overweight is a major problem these days, it would be helpful if shopper segments can be created regarding to their product preferences. The study could not find clear distinctions between segments with different preferences and therefore different BMI’s. To get a better insight in these segments other products can be considered in future research or the list of products need to be extended. Within this survey there is only a fixed choice of products and this is different when a participant is shopping in real life. This is another reason to extend the survey with more and different products.

Corresponding to the first limitation, the answers given in the survey are based on self-reported behavior. Hence, the preferences for processed products or caloric rich products might be a bit different. To make themselves feel better, they might indicate that their preference is more inclined towards the healthier option. This might also be the case to give a better appearance. This should not be the case because of the anonymous character of the survey but participants might have forgotten that. Furthermore, the survey is about product preferences. It might be the case that it sticks to a preference. While doing their real groceries people might buy the healthy/unhealthy option when their preference is inclined with the other option.

Another limitation could be the length of the survey. This was not a very long survey and the questions might have directed participants in a certain way. Therefore, the given answers might maybe not be as valid. To avoid this problem in future research, the survey could be extended. With more questions that are about a different topic to distract the participant from the main goal.

Finally, the number of participants is not very large. The 216 participants considered are of course not representative for the entire Dutch population. A solution towards this problem might be the help of different researchers all over the country which are basically looking for the same answers. To create a survey with each other and distribute this survey with each other might give a better representation for the Dutch population and the quality of the survey might increase as well. Two persons know more than just one.

(33)

32

Another suggestion for future research is check if other variables might explain overweight. Or the differences in products preferences between men and women. This might explain why women, in general, do have lower BMI’s compared to men. The differences in eating behavior and physical activity between men and women might help to decrease the problem of overweight.