1
University of Amsterdam
Faculty of Economics and Business
MsC Business Studies – Marketing Track
The predicting role of cognitive age on
optimum stimulation level and
exploratory behavior
Master Thesis – 15
thof June 2014
Author: Jordi Caspers
Student number: 10282602
First supervisor: Drs. F. Slisser
2
Table of content
Abstract ... 4
Chapter 1: Introduction ... 5
Chapter 2: Literature Review ... 12
§ 2.1: Chronological Age ... 12
§ 2.2: Cognitive Age ... 13
§ 2.3: Consumer Behavior ... 18
§ 2.3.1: Optimum Stimulation Level ... 21
§ 2.3.2: Exploratory Behavior ... 24 § 2.4: Conceptual Model ... 27 Chapter 3: Methodology ... 28 § 3.1: Sample description ... 28 § 3.2: Procedure ... 28 § 3.3: Measurement scales ... 30 § 3.3.1: Dependent Variables ... 30 § 3.3.2: Independent Variable ... 31 § 3.4: Data analyses ... 32
§ 3.5: Strengths and limitations questionnaire method ... 32
Chapter 4: Results ... 35
§ 4.1: Preparing data for analysis ... 35
§ 4.2: Descriptive statistics ... 36
§ 4.3: Reliability and validity ... 38
§ 4.4: Hypotheses testing ... 39
§ 4.4.1: Hypotheses regarding demographics and cognitive age ... 39
3
§ 4.4.3: Hypothesis regarding cognitive age and exploratory behavior... 40
§ 4.5: Summary ... 43
§ 4.6: Final model ... 45
Chapter 5: Discussion ... 46
§ 5.1: Empirical findings and implications ... 46
§ 5.2: Limitations and further research ... 48
Appendix 1: Bibliography ... 51
Appendix 2: Questionnaires ... 57
4
Abstract
This study investigates the predicting role that cognitive age could have on optimum stimulation level and exploratory behavior; exploratory behavior is measured via the variables brand switching, innovativeness, repetitive behavior proneness and risk taking. This study focused on older consumers, because they generally perceive themselves to be at a different age from their
chronological age. While most studies have focused on the relationship between demographics and cognitive age, far fewer studies have focused on the relationship between cognitive age and
consumer behaviors. As the findings regarding the relationship between demographics and cognitive age are not consistent over time, this study also investigates the relationship between demographic variables such as chronological age, educational level, employment status and income.
Findings of this study are that chronological age has a significant positive relationship with cognitive age, whereas educational level, employment status and income do not have a significant relationship with cognitive age. Further, this study proved that there is a significant negative relationship between cognitive age and optimum stimulation level, indicating that if cognitive age goes up, optimum stimulation level goes down. The investigated relationship between cognitive age and general exploratory behavior is not found to be significant, but there is a significant negative relationship between cognitive age and brand switching, and a significant positive relationship between cognitive age and repetitive behavior proneness.
This study proved that cognitive age is a better predictor for behavior of older consumers than chronological age and calls for further research regarding this topic.
5
Chapter 1: Introduction
The world’s population is ageing at such a rapid pace that the number of older persons will exceed the number of children by 2047 (UN, 2007). At the moment this is already happening in many developed regions and the profundity of this demographic change will have an impact on a variety of subjects. This pervasive and truly global phenomenon can be seen as the most profound
demographic change in the history of humankind, and it is unlikely it will occur again in the future (Kohlbacher, Riley, & Hofmeister, 2011). To give a visual example of the expected growth of older people in the population, the following figures show the age distribution in the Netherlands in 2013 and 2030 (CBS, 2014), which states that in 2030 the older consumer segment grows by almost 50% in quantity.
Male x 1000 Female x 1000 Male x 1000 Female x 1000
Figure 1: Age distribution in the Netherlands, 2013 Figure 2: Age distribution in the Netherlands, 2030
Source: CBS (2014) Source: CBS (2014)
So it is clear that ageing populations will be an increasingly important market segment in the near future (Ahmad, 2002) and with this in mind, one would say that the elderly segment becomes a structural subject of research, as older consumers are known for their substantial levels of disposable
6 income and willingness to spend (Sawchuck, 1995). Indora (2009) claims that people over 50 years of age own about 80% of the total Dutch capital. Thereby, 42% of the 50-plussers indicate that they spend money more easily as they grow older. So with the ageing populations, older consumers will be responsible for greater shares of spending. In 2005 consumers over 55 were responsible for 28 per cent of all retail sales in the UK, and that share is predicted to increase to 31.6 per cent of all retail sales in 2015 (Verdict, 2006). Surprisingly, what is known about the behavior of this fast growing population segment lags far behind what is known about other important segments (Williams, Virpi, Wadleigh, & Chen, 2010; Yoon, et al., 2005). Despite the growing importance of the 50+ population and its perception as an attractive market segment, older consumers are still routinely neglected by many marketing and advertising practitioners (Niemelä-Nyrhinen, 2007; Simcock & Sudbury, 2006; Uncles & Lee, 2006) and companies do not have strategies in place to attract and retain older consumers (Ahmad, 2002).
The belief that older consumers are an important and potentially profitable segment and that research towards this group is needed, is gaining ground. Several researchers have addressed this topic and suggest further research intp understanding the elderly consumer and their behavior (Johnson, 1996; Kasper, Nelissen, & Groof, 2009; Lumpkin & Hite, 1988). Existing research on older consumers has primarily been conducted in the USA (Bartos, 1980; Moschis, 2003) and the
perception that older consumers represent an unattractive market segment is beginning to change (Myers & Lumbers, 2008) and researches towards older consumers and their behavior is in
advancing. A question which arises with this growing interest and researches towards elderly consumers is how older segments should be categorized (Kasper, Nelissen, & Groof, Op Weg naar een Nieuwe Typologie voor Ouderen, 2009). Several studies attempted to categorise older
consumers, based on affluence level (Bone, 1991), generations (Becker, 1992), variety seeking and autonomy (Gerritsen, 2005), household income (Sikkel & Keehnen, 2006), but the most common method used for segmenting older consumers is chronological age (Moschis, 1991). The reason why chronological age is commonly and frequently used as a variable in marketing and consumer
7 behavior research, is mostly because it is easily measurable and objective of nature (Barak and Schiffman, 1981; Settersten and Mayer, 1997). Unfortunately, there are some negative
consequences when using chronological age asa variable. The use of chronological age is
problematic for researchers interested in age-related research, particularly research that examines the attitudinal or behavioral patterns of the elderly. More precisely, chronological age does not lend itself well to functioning as a dependent variable; that is, it is exceedingly difficult to justify
employing almost any behavioral variable of interest to consumer researchers as a predictor of chronological age (Barak & Schiffman, 1981). Stated differently, the unique antecedent character of chronological age restricts its usefulness to being employed as a predictor variable and the overriding shortcoming of chronological age seems to be that it does not take into account the fact that people frequently perceive themselves to be at an age other than their birth age, and that this
self-perceived, or cognitive age, seems to influence consumer behaviors such as purchase behavior (Barak & Schiffman, 1981), fashion consumption (Lin & Xia, 2012) and shopping behavior (Jack & Powers, 2013). Kotler (1976) describes this phenomenom using an example of Ford cars. Ford designed a car to appeal to young people who wanted an inexpensive sporty automobile. Ford found that the car was being purchased by all age groups and realized that its target market was not chronologically young people, but people who perceived themselves to be young.
Chronological age may appear to be the easiest way of segmenting the mature market but is probably the least effective because it does not correlate well with behavior (Myers & Lumbers, 2008). The relevance and necessity of looking at cognitive age in the context of older consumers is to understand how changes due to ageing arise, which are better reflected by self-perceived, or
cognitive age, than chronological age (Moschis, 1994). Cognitive age can be defined as the age one perceives one’s self to be (Blau, 1956; Barak & Schiffman, 1981) and is often expressed in terms of how old one feels, thinks and looks, as well as by the things one does (Kastenbaum, Derbin, Sabatini, & Artt, 1972; Stephens, 1991). Most researches towards cognitive age focus on the relationship with demographic variables, whereas little or no research is done on the relationship or effect of cognitive
8 age on general consumer behaviors (Lin & Xia, 2012), this could be interesting as someone who did not feel old would preferably not be approached as an older person and would probably not be interested in products or services which are aimed at older consumers (Kasper, Nelissen, & Groof, 2009)
The American Marketing Association defines consumer behavior as the dynamic interaction of affect and cognition, behavior, and the environment by which human beings conduct the
exchange aspects of their lives (AMA, 2014). In other words, consumer behavior involves the thoughts and feelings that people experience and the actions they perform in consumption processes and it includes all the things in the environment that influence these thoughts, feelings, and actions (Peter & Olson, 2008). Since consumer behavior is such a broad topic, researches commonly focus on a single part of consumer behavior. This study will therefore focus on the link beween cognitive age and the important construct Optimum Stimulation Level (OSL). Previous research has shown that OSL explains a wide variety of consumer behaviors with strong exploratory components such as risk taking, innovativeness, brand switching and variety seeking (Steenkamp & Burgess, 2002). Therefore it can be very interesting to see if there is a relationship between cognitive age and OSL.
OSL is a model that characterizes an individual in terms of his general response to
environmental stimuli. Introduced in the psychology literature, OSL theory postulates that individual behavior is influenced by the intrinsically motivated desire to accomplish a specific level of
stimulation, the optimum stimulation level (Berlyne, Conflict, Arousal, and Curiosity, 1960). This level of optimum stimulation varies between individuals (Raju, 1980). When the stimulation derived from the environment is too low, individuals will attempt to increase stimulation and vice versa (Orth, Optimum Stimulation Level Theory and the Differential Impact of Olfactory Stimuli on Consumer Exploratory Tendencies, 2005). Psychological pleasantness is highest at the OSL, the level of
stimulation at which a person feels most comfortable (Benedict & Baumgartner, 1992) and behavior, aimed at modifying stimulation from the environment in the general direction towards the optimum
9 level of stimulation, has been termed “exploratory behavior”. Dominating the body of studies
towards exploratory behavior is Raju’s (1980) categorization of three general exploratory tendencies:
Risk taking describes exploratory behavior expressed through choices of innovative and unfamiliar
alternatives that are perceived as risky.
Variety seeking is expressed through an individual’s switching within familiar alternatives, including
brand switching, and an aversion to habitual behavior.
Curiosity-motivated behavior involves exploratory information seeking, interpersonal communication
and shopping. It can be inferred that cognitive age could be a mediator for OSL and exploratory behavior. Cognitive age influences one’s innovativeness (Blau, 1973) and those who perceive themselves to be younger are more likely to have had more education than those who perceive themselves as older (Rosow 1967, 1974; Peters 1971). According to Raju (1980), both education and innovativeness seem to influence OSL. Despite this promising prospects, no research has been conducted towards cognitive age and OSL yet. One would think it would be very interesting to see what the relationship is between cognitive age and both OSL and exploratory behavior. If there is a significant relationship between them, and cognitive age seems to predict certain exploratory
behaviors, this could be very helpful for marketing practitioners in order to be able to target the right people and it can help segmenting the market of older consumers.
Therefore, the purpose of this study is to investigate if cognitive age could act as predictor for OSL and exploratory behavior; this will be investigated using the following research question:
What is the relationship between cognitive age, Optimum Stimulation Level and exploratory behavior and to what extent can cognitive age act as predictor for Optimum Stimulation Level and exploratory behavior?
This research contributes to the existing body of knowledge in manifold ways. First of all, it helps to reveal and understand the relationship between cognitive age and environmental consumer
10 behavior. New insights about the relationship between cognitive age and OSL are going to extend existing theory, which also applies to the relationship between cognitive age and exploratory behavior. This research wil add information about the reasons why consumers, especially elderly consumers, behave in a certain way, and this will address an important gap in the literature (Kasper, Nelissen, & Groof, 2009). Furthermore, the study adds to the existing body of knowledge about consumer behavior, which can be of particular practical importance, as it provides a theoretical background for marketing practitioners. Finally, the work is intended to add to the existing
knowledge about cognitive age, and how it can be used for segmenting older consumers, underlining the importance and value of this concept in the current ageing world. As said before by Marthur & Moschis (2005) and Kasper, Nelissen, & Groof (2009), there is a need to replicate previous findings regarding cognitive age and research among the elderly should be directed towards cognitive age in order to find a better predictor for elderly consumer behavior.
In the next chapter, a more detailed overview of literature regarding cognitive age, optimum stimulation level and exploratory behavior will be presented and the conceptual model and
hypotheses will be stated in order to be able to answer the research question. Chapter 3 is about the methods and methodology used in this research, whereas chapter 4 is about data gathering and analysis. After completing the data analysis, the results are transformed into conclusions in chapter 5. Chapter 6 is about recommendations and implications about the outcomes of this research.
11
Chapter 6
Recommendations and Implications
Chapter 5
Conclusions
Chapter 4Results
Chapter 3Methodology
Chapter 2Literature Review
Chapter 1Introduction
12
Chapter 2: Literature Review
This literature review will start with a brief overview of chronological and cognitive age and their use in research in the past to get an idea about the popularity and limitations of chronological age, what has been studied and what the main discussions are about cognitive age and why cognitive age sometimes should be used as a substitute for chronological age. Also the first set of hypotheses about cognitive age in relation to demographics are set up, as one of the intentions of this research is to add to the existing knowledge about cognitive age, as several researchers conclude that it is necessary to find a better way to segment the older consumer market (Marthur & Moschis, 2005; Kasper, Nelissen, & Groof, 2009) .
After this introduction, in section 2.3 general consumer behavior and the importance of its possible relationship with cognitive age, and the literature of the two dependent variables of this research, optimum stimulation level and exploratory consumer behavior, will be discussed. Further the second set of hypotheses will be presented, they are meant to investigate the predicting role cognitive age has in respect to optimum stimulation level and exploratory behavior, it is intended to explain how cognitive age plays a role in exploratory consumer behavior, this has not been the subject of research yet, but research in this direction could help understanding the older consumer and his behavior (Trijp, 1997; Ahmad, 2002; Kasper, Nelissen, & Groof, 2009).
Finally, at the end of this literature review, the conceptual model of this research, which evolves out of the literature review, will be presented in section 2.4.
§ 2.1: Chronological Age
Chronological age is perhaps the most widely and frequently used measure in research. As a
demographic variable, it stands out from all other variables in terms of frequency of its use (Barak & Schiffman, 1981). The reason why chronological age is commonly and frequently used as variable in marketing and consumer behavior research, is mostly because it can be measured easily and is objective (Barak and Schiffman, 1981; Settersten and Mayer, 1997). Early on, studies identified
13 chronological age as either the number of years lived (Hendricks & Hendricks, 1936), or as the
distance from birth (Jarvik, 1975). As said, chronological age is widely used in all kinds of research in all kinds of research areas. In consumer behavior research, chronological age is often used in
descriptive consumer behavior studies, or to segment consumer markets (Barak & Schiffman, 1981). Despite its great popularity, there are some limitations to chronological age and its usefullness as an independent variable in research. Demographic variables are usually automatically selected without much thought being given to them. The use of chronological age is problematic for researchers interested in age-related research, especially research that focuses on behavior of the elderly. This is because chronological age does not lend itself well to functioning as an independent variable as the unique antecedent character of chronological age restricts its usefulness to being employed as a predictor variable (Barak & Schiffman, 1981).
As said in the introduction, the most important shortcoming, from a consumer behavior
perspective, is that chronological age does not consider the fact that people perceive themselves to be at a different age than their chronological age. In the mid fifties and early sixties, researchers started to think about these concerns and made a start with refining existing demographics and started to develop new ones (Barak & Schiffman, 1981), of which self-perceived age become an important subject of research in order to find a better indicator for behavior than chronological age.
§ 2.2: Cognitive Age
Three main developments of non-chronological age measures which are used as variables in social and psychological researches in the early 70s are personal age, other-perceived age and subjective age. Personal age is a selfreported age measure which consists of four age dimensions (feel-, look-, do- and interest-age) and is operationalized by Kastenbaum, Derbin, Sabatini and Artt in 1972. In addition to the four dimensions, respondents must also answer a large set of questions regarding their feelings about themselves and the comparison between the four dimensions and their
14 chronological age. Therefore, Barak & Schiffman conluded in 1981 that this measurement does not lend itself very well for research conducted by consumer and marketing researchers.
Other-perceived age is measured using the subjective evaluation of age status of an
individual as estimated by others. This way of measuring is largely based on perceived physical looks and the perceived social roles of the individual who has been observed (Lawrence, 1974) and is therefore not commonly used in consumer behavior literature.
Subjective age was one of the first concepts that measures an individual’s self-perception in terms of reference age groups, such as middle-aged, elderly, or old. Subjective age establishes how a person feels about such reference groups (Barak & Schiffman, 1981). The problem with the
measurement of subjective age is the way of rating (middle-aged, elderly, old). This kind of rating is different for every individual since every individual has his own feelings about these categorizations and thus there may be differences among individuals (Barak & Schiffman, 1981). The concept of subjective age was first introduced by Blau in 1956 and continued to be used by Peters (1971), Rosow (1967, 1974) and Ward (1977). Within the consumer behavior context, Roscoe, LeClaire and Schiffman suggested in 1977 that the age variable should be broadened, so that it reflects age related factors, including self-perceived age. In 1981, Barak and Schifmman expanded on this view and reviewed existing non-chronological measures of age, and proposed a new perceived age variable: cognitive age. Their self-perceived age measure is defined using the four dimensions as suggested by Karstenbaum, Derbin, Sabatini and Artt (1972). These dimensions capture aspects of age that are not adequately reflected in chronological age. The cognitive age measure lends itself well for consumer behavior studies that explore the scope and nature of cultural differences in the perception of age, and how these differences influence various aspects of consumer behavior (Barak & Schiffman, 1981).
Since then, much research has been conducted towards cognitive age and one of the important findings for this study is that cognitive age is particularly relevant for elderly consumer cohorts, as it
15 takes into account lifetime-related changes in perceptions and behavior that tend to occur in later life (Teller, Gittenberger, & Schnedlitz, 2013).
It therefore seems that cognitive age is perhaps a better way to predict behavior of (elderly) consumers than chronological age. This statement is supported by findings from studies that found that cognitive age can predict fashion consumption (Lin & Xia, 2012) , shopping behavior (Jack & Powers, 2013), and grocery store patronage (Teller, Gittenberger, & Schnedlitz, 2013). Moreover, cognitive age is perhaps a better way to segment the elderly market (Kasper, Nelissen, & Groof, 2009) and can act as a useful concept for advertising (Stephens, 1991). Because cognitive age, as an age self-image measure, focuses on identification with age-role reference groups and consists of four sub-dimensions (feel-age, look-age, do-age, and interest-age) it is more inclusive than chronological age (Barak & Schiffman, 1981; Barak & Gould, 1985). Furthermore, cognitive age is a more valid predictor of consumer behavior, as indicated in the sentences above, than chronological age, which reflects an ‘empty’ variable in the sense that chronological age itself cannot cause behavior and cognitive age is more sensitive to individual differences (Settersten & Mayer, 1997). Even though cognitive age, as a research variable, has become a subject of research since the 80s, from the mid-fifties on, research has moved towards self-perceived age and its influence on behavior. Blau (1956, 1973), Peters (1971) and Rosow (1967, 1974) found that the majority of elderly have a strong tendency to see themselves as considerably younger than their chronological age, a finding that is confirmed by more recent researchers (Mathur & Moschis, 2005; Sudbury & Simcock, 2009; Teller, Gittenberger, & Schnedlitz, 2013). Other researchers conclude that self-identification with a younger age group varies in terms of social class standing (Bengston, Kasschau and Ragan 1977; Peters 1971, Rosow 1967). Further, women are more sensitive to the negative stereotypes associated with "elderly" and "old," and tend to see their age differently than their male counterparts (Bengston, Kasschau and Ragan 1977; Peters 1971). Finally, research indicates that subjective age is related to subjective well-being (i.e., life satisfaction or morale) and self-confidence (Bengston, Kasschau and
16 Ragan 1977; Peters 1971), this is confirmed by Chua, Cote and Leong (1990) and Barak and Rahtz (1999).
Although these findings are consistent over time, other studies focusing on cognitive age have produced contradictory results. Barak and Stern (1986) reviewed several studies and found that some studies conclude that there is a significant positive relationship between variables as
retirement, widowhood and cognitive age, and a significant negative relationship between education, social class and cognitive age. Other studies did not find a relationship between these variables. Similarly, Henderson, Goldsmith, and Flynn (1995) examined the relationship between cognitive age and demographic variables such as gender, marital status, education (negatively tested), income (negatively tested), and race. However, these researchers did not find any significant relationships. Wilkes (1992) found that among women aged 60–69, marital status and income are negatively related with cognitive age, whereas work status was not a significant predictor of cognitive age.
Although there is a lot of research investigating cognitive age, the findings are not always consistent. Mathur and Moschis (2005) conclude that there is a need to replicate previous research to validate the findings. Kasper, Neelissen and De Groof (2009) came to the same conclusion; future research among the elderly should be directed towards cognitive age, in order to find a better predictor for elderly consumer behavior.
As said before, most early researches regarding cognitive age were focused on the relationship between demographics and cognitive age (Blau, 1973; Barak & Schiffman, 1981), whereas later researches from the 2000s and on, were more focused on cognitive age as a predictor for consumer behaviors such as fashion consumption (Lin & Xia, 2012) and shopping behavior (Jack & Powers, 2013). As the most conflicting results were found in the earlier researches, the first set of hypotheses will be stated in order to replicate previous findings regarding the predicting role demographics could have on cognitive age in order to verify earlier findings.
17 The influence of demographics on cognitive age is important to test, because they have intuitive appeal, are easily measured and are used in many social science theories (Henderson, Goldsmith, & Flynn, 1995).The most common demographic variable that is frequently used, is, as said before, chronological age. At first sight, it seems that chronological age is related to cognitive age. When one gets older, one would also feel older: maybe not as old as his or her chronological age, but it can be inferred that generally, someone of 60 feels himself older than 40 years ago when he was 20. This simple reasoning is confirmed by researches from Peters (1971), Ward (1977) and Bultena & Powers (1978), which conclude that there is a relationship between chronological age and self-perceived or cognitive age, although they also found that people tend to think of themselves as being at a different age, often younger, than their chronological age; this is confirmed by Markides & Boldt (1983) and Underhill & Caldwell (1983). This discrepancy can be explained by theories regarding societal stigmatization (Baum & Boxley, 1983). They stated that it could be that older people perceive themselves to be younger than they really are, because of the negative image associated with old age in society. So even though it could be that older people generally tend to feel younger than their chronological age, it can be inferred that chronological age can be used as a predictor for cognitive age; this will be tested through the following hypothesis:
H1a: There is a positive relationship between chronological age and cognitive age; if an individual is chronologically older, his cognitive age will also be older.
Socioeconomic demographics such as education level, employment status and income also seem to influence cognitive age. Researchers found that people higher on the socioeconomic ladder, who have a higher education, who are employed and have an substantial income above average, usually have a younger cognitive age (Rosow, 1967; Bultema & Powers, 1978). On the contrary, people of lower socioeconomic classes seem to have a higher cognitive age, because they accept the onset of old age earlier than people of higher socioeconomic classes (Markides & Boldt, 1983). An explanation
18 for this tendency could be that people of lower socioeconomic classes have lower life expectancies and more hardships and disabilities (Rosow, 1967). Further, in 1992, Wilkes found that people with a less than average income have an older cognitive age, also due to hardships and disabilities.
Unfortunately, these findings were not verified by researches from George, Mutran, & Pennybacker (1980) and Henderson, Goldsmith, & Flynn (1995), who did not find significant relationships between these variables. In order to verify and validate the previous findings described above, the following hypotheses were set up:
H1b: There is a negative relationship between educational level and cognitive age; the higher the educational level, the lower the cognitive age will be.
H1c: Employed individuals’ cognitive age is lower than the cognitive age of unemployed individuals.
H1d: There is a negative relationship between income and cognitive age; the higher the income, the lower the cognitive age will be.
§ 2.3: Consumer Behavior
As mentioned in the introduction, consumer behavior involves the thoughts and feelings of people’s experience and the actions they perform in consumption processes. It also includes all the things in the environment that influence these thoughts, feelings and actions (Peter & Olson, 2008). In general, there are two ways of thinking about consumer behavior, first you have the cognitive view which suggests a successful organization should aim to satisfy the needs of their consumers in order to make profit. To implement the cognitive view, organizations have to understand their customers and stay close to them to provide products and services that consumers will purchase and use appropriately (Peter & Olson, 2008). The cognitive view is about understanding the customer and
19 understanding his needs. To be able to do this, you have to understand the mental responses from consumers and have to consider affect, behavior, cognition and environment (Peter & Olson, 2008). According to the cognitive view, there are three approaches to studying consumer behavior; the interpretive approach, which has its base in cultural anthropology and seeks to develop a deep understanding of consumption and its meanings; the traditional approach, based on theories and methods from cognitive, behavioral and social psychology. This approach aims at developing theories and methods to explain consumer decision making and behavior. And, there is the marketing science approach, based on economics and statistics. This approach involves developing and testing
mathematical models to predict the impact of marketing strategies on consumer choice and behavior (Peter & Olson, 2008).
The model of consumer behavior through the eyes of the cognitive view is as follows:
Figure 3: Consumer Behavior according to the marketing concept (Peter & Olson, 2008)
On the other hand, there is the behavior modification perspective, which focuses on
environmental factors which influence behavior (Nord & Peter, 1980). All the processes which seem to occur within the person (needs, motives, attitudes etc.) are seen as problematic to predict and control and therefore speculation about them is shunned. This perspective thinks about the internal psychological processes as a ‘black box’ which is too difficult to understand. According to the behavior modification perspective, the model of consumer behavior is as follows:
20
Figure 4: The Behavior Modification Perspective (Nord & Peter, 1980)
This study follows the path of the behavior modification perspective. Trying to predict the way older consumers act and behave through using cognitive age as an predictor for this behavior is not about explaining the internal and mental processes that take place inside the consumer and understanding why a consumer behaves in a certain way, but about predicting this behavior.
As said, this study will investigate the relationship between cognitive age and OSL and
exploratory behavior. As cognitive age is measured through identification with role-reference groups, it is environmentally influenced. Therefore it is particularly interesting to investigate the relationship between cognitive age and the environmental part of consumer behavior. OSL is a model used to investigate the environmental influence on exploratory behavior, and so these two concepts will act as dependent variables in this study.
It can be inferred that cognitive age could be a predictor for OSL and exploratory behavior. Cognitive age influences one’s innovativeness (Blau, 1973) and those who perceive themselves to be younger are more likely to have had more education than those who perceive themselves as older (Rosow 1967, 1974; Peters 1971). According to Raju 1980) both education and innovativeness seem to influence OSL. Despite these promising prospects, no research has been conducted towards cognitive age and OSL yet. Therefore, the purpose of this research is to investigate if cognitive age could act as a predictor for OSL and if there is a direct relationship between cognitive age and exploratory behavior.
21
§ 2.3.1: Optimum Stimulation Level
Consumers seeking thrills, adventure, disinhibition, new experiences, fantasies, sensory stimulation, escape from boredom, and alternation among familiair things have been identified as engageing in exploratory consumer behaviors in order to raise their level of stimulation in life (Raju, 1980; Steenkamp & Baumgartner, 1992; Zuckermann, 1994). Research on this need for stimulation has shown that people tend to prefer intermediate levels of stimulation, referred to as the optimum stimulation level (OSL), and that there are reliable individual differences in the amount of stimulation considered optimal by a person (McReynolds, 1971). The psychological concept of OSL is a principle which connects three underlying psychological processes (curiosity, attribute-saturation and
boredom-solving behavior) with each other and relates these processes towards affect (Trijp, 1997). The theory of OSL states that the level of optimal stimulation at which an individual feels most comfortable and functions best, differs for each individual (Zuckermann, 1979). The difference between individuals is recognized by psychologists as a personal characteristic for which different measurement instruments have been developed (i.e. Mehrabian and Russell’s Arousal Seeking Scale (1973) and Zuckerman’s Sensation Seeking Scale (1979). OSL theory further states that people strive to reach the level of stimulation that corresponds with their optimal level of stimulation. To reach that optimal level, people exhibit stimulation-regulating behavior. When someone’s level of stimulation is too high, he or she will actively search for situations with low stimulation, in an attempt to restore the harmony. Someone with a level of stimulation that is too low, follows the same path in an attempt to restore the harmony and will look for situations which brings extra stimulation or arousal. The figure on the next page illustrates how arousal is related towards affect:
22
Figure 5: The relation between arousal potential as source for stimulation and affect (Trijp, 1997)
This inverted U-shaped model is the basic notion of OSL; the relationship between stimulation obtained from the environment or through internal means and a person’s affective reaction to stimulation (Benedict & Baumgartner, 1992). Psychological pleasantness is highest at the OSL, the level of stimulation at which a person feels most comfortable (Benedict & Baumgartner, 1992) and behavior, aimed at modifying stimulation from the environment in the general direction towards the optimum level of stimulation, has been termed “exploratory behavior”, this topic will be addressed in section 2.4. In his research Raju proposes a model which examines the relationships between OSL and selected personality traits, selected demographic variables and general exploratory tendencies in the consumer context. The following figure illustrates these relationships; the dotted arrow indicates that the exogenous variables are not directly part of psychological explanations of exploratory behavior:
23 Findings of his research were that younger, educated and employed people seem to have higher OSLs and further that those with high OSLs and low OSLs differ most with respect to risk taking and innovativeness and differ somewhat in brand switching and repetitive behavior proneness. These findings are consistent with more recent researches from Steenkamp and Baumgartner (1995, 2012) and Zuckerman (1994) who found that OSL decreases with age and increases with income and education and that OSL has a positive effect on the interest in exploratory products and innovativeness. Psychologists and consumer behavior researches have studied OSL extensively and the general conclusion is that OSL systematically influences consumer behaviors with strong exploratory tendencies. Raju (1980) states that the magnitude of OSL leads to attempts to adjust stimulation from the environment.
Raju (1980) also found that the exogenous variables he tested (age, employment status, education and income)do not all have a significant correlation with OSL. A more recent study from Mahatanankoon (2007) concludes that gender has no effect on OSL, but education and age does. So it seems that exogenous variables as a concept are not that good of a predictor for OSL. Cognitive age can be presumed as a predictor for OSL, because individuals who are cognitively young are less likely to be traditional and old fashioned and are more likely to have a high level of self-confidence (Stephens, 1991). This seems to correlate with OSL, as stated by Raju (1980); people with high OSL are more likely to explore new stimuli and manifest more risk taking in their behavior.
Therefore, individuals with a lower cognitive age might have a higher OSL. This statement will be tested through the following hypothesis:
H2: There is a negative relationship between cognitive age and OSL; the higher the cognitive age, the lower the OSL will be.
24
§ 2.3.2: Exploratory Behavior
Exploratory behavior is behavior with the sole function of changing the stimulus field (Berlyne, 1963). Exploratory consumer behavior tendencies have been categorized as curiosity-motivated behaviors, variety seeking, and risk taking (Raju, 1980). Curiosity-motivated behavior involves exploratory information seeking, interpersonal communication and shopping. Curiosity can be defined as the desire for knowledge for intrinsic reasons and a distinction can be made between specific and diversive curiosity-motivated behavior (Berlyne, 1960). Specific curiosity refers to behavior focused on exploration of a single stimulus in depth because it arouses a consumer’s curiosity. Diversive curiosity is different, it is a concept to seek stimulation in a variety of sources and occurs as a reaction to a state of boredom (Steenkamp & Baumgartner, 1992).
Variety seeking is expressed through an individual’s switching within familiar alternatives, including brand switching, and an aversion to habitual behavior (Orth, 2005). Risk taking describes exploratory behavior expressed through choices of innovative and unfamiliar alternatives that are perceived as risky. Risk is noted as a concern for human beings from the beginning of recorded history and most likely even before that (Grier, 1980), whereas risk taking is defined as ‘any consciously, non-consciously controlled behavior with a perceived uncertainty about its outcome, and/or about its possible benefits or costs for the physical, economic or psycho-social well-being of oneself or others’ (Trimpop, 1994). It can be concluded that every individual takes risks, some more than others. In previous researches, these concepts are all linked towards OSL, whereas a higher level of OSL means that an individual exhibits higher levels of risk taking, innovativeness, variety seeking and curiosity-motivated behavior.
In Raju’s (1980) model of OSL, the dotted arrow indicates that exogenous variables are not directly of influence on exploratory behavior, the exogenous variables age, employment status, education and income only seem to influence OSL, whereas OSL influences exploratory behavior. Part of this research is to test if cognitive age could be used as a predictor for both OSL and exploratory behavior; this is important in order to extend the existing literature regarding cognitive age and to
25 provide a theoretical rationale and empirical evidence for considering cognitive age as a substantial influencer and predictor of older consumers’ perceptions and their behavior related to brand switching, innovativeness and risk taking. This could help marketing practitioners in predicting the behavior of older consumers and in segmenting this increasing market.
Cognitive age influences one’s willingness to innovate (Blau, 1973), and one’s willingness to try new things; Individuals with a lower cognitive age are less tied to safe and well-established routines (Barak B. , 1998; Lin & Xia, 2012) and are more likely to seek information and display less cautiousness in buying (Stephens, 1991). These behaviors are all part of exploratory behavior and therefore it is hypothesized that cognitive age directly influences exploratory behavior through the following hypothesis:
H3: There is a negative relationship between cognitive age and exploratory behavior; the higher the cognitive age, the lower the exhibited exploratory behavior.
In this research, brand switching, innovativeness, repetitive behavior proneness and risk taking will be further scrutinized as dependent variables of exploratory behavior, influenced by cognitive age, because previous findings conclude that with these variables the difference between high and low OSL is the most profound. To test whether this also applies for cognitive age, the following
hypotheses are stated:
H3a: There is a negative relationship between cognitive age and brand switching; the higher the cognitive age, the lower the willingness to switch brands will be.
H3b: There is a negative relationship between cognitive age and innovativeness; the higher the cognitive age, the lower the willingness to innovate will be.
26
H3c: There is a positive relationship between cognitive age and repetitive behavior proneness, the higher the cognitive age, the higher the proneness to repeat behavior.
H3d: There is a negative relationship between cognitive age and risk taking; the higher the cognitive age, the lower the willingness to take risks will be.
In summary, this research tries to verify and validate previous findings regarding the predicting role demographics could have on cognitive age to add to existing knowledge of this increasingly
important concept in the current ageing world. Further, the relationship between cognitive age and OSL and exploratory behavior will be tested in order to add to the existing literature regarding cognitive age, which has, despite the promising aspects of these concepts, not been the subject of research before.
27
§ 2.4: Conceptual Model
Below is a schematic presentation of the conceptual model. The dark grey section indicates that its relationship will be tested with both of the lighter grey sections.
28
Chapter 3: Methodology
This methodology chapter is written in order to describe how the data were collected, analyses are made and conclusions are drawn based on the results obtained. This chapter elaborates on the method used to research the questions mentioned in the previous chapter and will be used to test the hypotheses, making it possible to draw a conclusion on how demographics, cognitive age, OSL and exploratory behavior are linked together. Two quantitative studies will be conducted; study 1 investigates the predicting role demographics has on cognitive age and study 2 investigate the influence of cognitive age towards OSL and exploratory behavior.
§ 3.1: Sample description
In order to be able to assess a medium effect (r = 0,30), using α = 5% and a power of 80%, 68 respondents are needed (Cohen, 1988). Taking into account the possibility of a low response rate of 30%, 230 questionnaires must be set out. But under the guise of the more the better, almost 330 questionnaires are sent to possible respondents.
Respondents were selected from the network of Woon- & Zorgcentrum Groot Hoogwaak, a residential home for elderly people, based in Noordwijk, Zuid Holland. Woon- & Zorgcentrum Groot Hoogwaak has over 600 clients of whom 170 are living at the residential home and around 300 employees with 80 employees are older than 50, meeting the requirement to be considered as a respondent. Together with the 80 volunteers of the organization aged over 50, this completes the sample of around 330 possible respondents. Woon- & Zorgcentrum Groot Hoogwaak is used as the source for respondents because I have access to the client and employee database and that makes it easy to obtain enough respondents.
§ 3.2: Procedure
Respondents were informed about the research with an invitation letter/email, which states that this research is done with approval of Woon- & Zorgcentrum Groot Hoogwaak, that the questionnaires
29 are anonymous, and answers will be treated confidentially. Further, an explanation of the
questionnaire was given and respondents were thanked in advance for their cooperation in this research. Respondents were not informed about the research topic, because this could have biassed their responses. Questionnaires to employees were sent via an email with a link to the questionnaire, questionnaires to clients were send per post, due to the fact that most clients did not have an email account or access to internet. If the questionnaire was send by post, a free returning envelope was been included.
The questionnaire consists of a total of 35 questions, divided over 4 categories;
demographics, cognitive age, OSL and exploratory behavior. In order to try to make the research as representative as possible, statistics from the Central Bureau for Statistics in The Netherlands are used to develop common answer scales for educational level, income and employment questions. When analyzing the data, it will be checked if the answers are consistent with the known statistics of the Dutch population. To make sure that instructions were clear, the questions were relevant to the intended respondents, respondents were willing to answer all the questions, and the questionnaire could be feasibly completed in the allotted time, there was a pilot test among several intended respondents and subject matter experts to find out if anything was missing from the questionnaire. Their feedback was been taken into account before the questionnaire was sent to all intended respondents.
The questionnaire is developed in English and translated into Dutch, because the intended respondents knowledge of the English language varies and confusion due to a lack of knowledge of the English language must be avoided. To make sure the questions remained the same after the translation, two different experts are asked to confirm the translations via back-translating the questions into English after the translation into Dutch.
30
§ 3.3: Measurement scales
§ 3.3.1: Dependent Variables
The dependent variables in this study are OSL and exploratory behavior.
Optimum Stimulation Level
As OSL has been a subject of research from the sixties, a number of different measurement scales has been developed over time. Three of the first major instruments to measure OSL are the Sensation Seeking scale from Zuckerman, Kolin, Price, and Zoob, developed in 1964, the Change Seeker Index from Garlington and Shimota, also developed in 1964 and ten years later the Arousal Seeking Tendency scale of Mehrabian and Russel (1974) made its entrance. More measurement methods of OSL have been developed over time, but these are the ones most used in consumer behavior researches (Raju, 1980). In 1992, Steenkamp and Baumgartner concluded that Garlington and Shimota’s (1964) 95-item Change Seeker Index (CSI) is a preferred instrument to measure OSL, but 95 items is a rather long measurement scale for research. Therefore, Steenkamp and
Baumgartner (1995) developed a new shortened 7-item form of the CSI and validated it cross-culturally in Belgium, the Netherlands, Great Britain, and the USA (Baumgartner & Steenkamp, 1998; Steenkamp & Baumgartner, 1995). Their findings indicated that, compared to the original scale, the shortened scale not only reduces the data collection burden for the respondent but has also
improved nomological validity and psychometric properties. CSI is typically rated on a 5-point Likert scale; however, in this study the same 7-point scale is used as for the measurement of exploratory behavior, to reduce potential respondent confusion.
It could be very possible and imaginable that Steenkamp and Baumgartner have used a Dutch questionnaire as part of their research was held in Belgium and The Netherlands , but despite some thorough research, it was not possible to retrieve the Dutch version.
31
Exploratory behavior
To measure general exploratory tendencies, the 39-item exploratory behavior scale from Raju (1980) will be used. This scale offers the best general representation of exploratory behavior in the
consumer context because it is not affected by social desirability considerations and it captures the essence of exploratory behavior (Raju, 1980). As this research only focuses on the tendencies brand switching, innovativeness, repetitive behavior proneness and risk taking, the questions regarding exploration through shopping, information seeking and interpersonal communication are taken out of the questionnaire. 24 items remained and are rated on a 7-point Likert scale.
§ 3.3.2: Independent Variable
The independent variable in this study is cognitive age and there are many different ways of measuring cognitive age. These methods fall into two major groups. The first, and oldest, is age identity (Cavan, Burgess, Havinghurst, & Goldhamer, 1949). Age identity concerns the age category (young, middle-aged, old) in which people perceive themselves to be, and is used extensively in gerontology studies. A second type of measure grew in response to the recognition that ageing is multidimensional (Birren, 1968), comprising biological, psychological, and sociological dimensions, none of which can be understood without reference to the others (Riley 1985). The cognitive age scale (Barak & Schiffman, 1981) is one such multidimensional scale, which incorporates the different dimensions of ageing. This typology has its roots in the consensus reached by philosophers
concerning the existential stances with regard to the human condition, which are knowing, feeling, and acting (Bengston, Reedy, & Gordon, 1985). The cognitive age scale has become the most popular measure of self-perceived age in marketing research pertaining to seniors. The superiority of the cognitive age scale over other available instruments is due to its ease of administration and understanding by respondents (Stephens, 1991), its validity (Auken & Barry, 1995), and its
32 in scale employment with samples of older people (Reinecke, 1993). For these reasons, the
multidimensional scale from Barak & Schiffman will be used to measure cognitive age.
This measurement method contains four questions about how old respondents think they look (biological), how old they feel (psychological and biological), and how old they rate their behavior and interests (social). For this scale, respondents are asked to answer in an age-decade scale from twenty till eighty. For respondents, the age-decade scale is easy to understand and use and for researchers easy to analyze and interpret (Stephens, 1991). Each set of four responses will be combined to yield a cognitive age score, the middle of the age-decade will be used for measurement, i.e. an answer in the age-decade scale 20s will be calculated as 25.
§ 3.4: Data analyses
Collected data were analyzed with SPSS. To avoid Type-I errors, an α of 5% was used. Likert scales are treated as interval variable, but were checked beforehand to see whether the data was normally distributed. H1a was tested with Pearson’s Correlation Coefficient, as H1a tests the
relationship between two ratio variables. H1b is tested with Spearman’s Correlation Coefficient, as it is intended to measure the relationship between multiple ordinal variables and a ratio variable, the same applies for H1d. H1c was measured with one-way Anova, because H1c tests the relationship between multiple nominal variables and a ratio variable. H2, H3, H3a, H3b and H3c were all
measured via Pearson’s Correlation Coefficient as they measure the relationship between an interval variable and ratio variable.
Before testing the hypotheses with the intended statistical tests, the mean and standard deviation were reported.
§ 3.5: Strengths and limitations questionnaire method
This research is cross-sectional, quantitative of nature and makes use of standardized,
33 or hardcopy is distributed. Advantages of this type of data-gathering are that respondents can answer the questionnaire at a self-selected time and at their own working speed; it is possible to send respondents a personal reminder when they forget to fill in the questionnaire, the absence of an interviewer- and response bias, as every single respondent is asked to respond to exactly the same set of questions (Saunders, Lewis, & Thornhill, 2009) and further is it possible to make sure the selected respondents are completing the questionnaire (Witmer, Colman, & Katzman, 1999).
A disadvantage of a quantitative approach is that, in comparison with a qualitative approach, results could be limited. Respondents are asked closed questions and the resulting data will be numerical descriptions. With a qualitative approach it is possible to explain questions and provide further explanations. Qualitative data will represent respondents needs, experiences and opinions and give respondents the opportunity to define and describe situations or events (Saunders, Lewis, & Thornhill, 2009). Another disadvantage of quantitative research via e-mail or post is bias caused by refusal (Saunders, Lewis, & Thornhill, 2009), when targeted respondents decide to ignore the mail. Despite these disadvantages, a quantitative approach is used because with a quantitative approach it is possible to reach a lot of respondents, which in this research contributes to the most integral and complete view of the topic and target groups. Another reason to choose for quantitative research is that most researches considering this research topic were also quantitative of nature.
Apart from some demographic questions, Likert-scales are used and questions will be measured as interval-scales. The most important point of critic about the use of Likert-scales is whether the ‘distance’ between ‘totally disagree’ and ‘disagree’ is equivalent to the distance
between ‘agree’ and ‘totally agree’. There is no consensus among researchers about this topic, but in marketing research, it is mostly accepted to use Likert-scales as interval-scales and therefore in the case of this study, Likert-scales will be measured as interval-scales.
The results of this research should be generalizable and transferable and the essence of this study must be useful in other studies concerning this topic. The questionnaire and interviews should
34 provide useful information on the research topic and should supply sufficient information for use in further research.
35
Chapter 4: Results
In this chapter, the results of the tested hypotheses will be shown. Before testing the hypotheses a description of what is done to prepare the obtained data for analysis will be given in § 4.1. § 4.2 explains more about the descriptive statistics of the dataset, whereas § 4.3 describes the reliability and validity of the data. In § 4.4 an overview of the outcomes of the tested hypotheses will be given and § 4.5 contains a summary of the results. In § 4.6 the final model will be exposed.
§ 4.1: Preparing data for analysis
Firstly, the completed questionnaires, hardcopy and digital, were transformed into a SPSS data file and a check was made of all questionnaires were useful for analysis. Six questionnaires were
incomplete (i.e. only the first page of the questionnaire was filled in); these were not included in the dataset. Thirteen questions of the questionnaire were recoded, as they were counter indicative. These questions were used in order to prevent the respondents from acquiescence bias, which happens if respondents tend to agree with everything that is asked to them (DJS Research). Also a hotdeck imputation was executed in order to replace missing values with a value of a similar donor that matches the donee in determined categories (Myers, 2011). Before executing the hotdeck imputation, frequencies were run to check whether missing data was below ten percent (Myers, 2011). This was the case for all the variables with missing data. The age variable had the most missing values (eight). These eight missing values for age were responsible for 5.5 percent of the total
response and justifies the use of hotdeck imputation.
To check which deck variable must be chosen (Myers, 2011), correlation tests were run to choose the variable which correlates most with the variable with the missing values. Values were rounded off at whole numbers. As a check, frequencies were run again to check whether all missing values were replaced, which was not the case.
Further, it was checked whether the obtained data was normally distributed. To see whether this was the case or not, the data was visually checked for normality. This was done by looking at
36 values that quantify aspects of a distribution (i.e. skewness and kurtosis) (Field, 2009), looking at the mean and standard deviatiosn and how these scores are related to the minimum and maximum values and by using P-P plots for looking at the cumulative probability of a variable against the cumulative probability of a particular distribution (Field, 2009). Visually checking for normality was chosen because checking for normality with the Shapiro-Wilk test or the Kolomogorov-Smirnov test in SPSS is not appropriate for larger samples; large samples will give rise to small standard errors and so when sample sizes are big, significant values arise even from small deviations from normality (Field, 2009). With large samples, it is more important to look at the shape of the distribution visually and to look at the value of the skewness and kurtosis statistics rather than calculate their significance (Field, 2009). All variables seem to be normally distributed, with no extreme skewness or kurtosis, and a standard deviation not too large in comparison with the mean, so the dataset was treated as if it was normally distributed.
After recoding variables, handling missing values and checking for normality, the last thing done before the data was ready for analysis, was to transform the separate questions of each variable (i.e.: the variable ‘cognitive age’ consists of four questions, whereas the variable ‘optimum stimulation level’ consists of seven questions) into one variable; this was done with the compute-variable function of SPSS. The new compute-variables are composed of the total mean of the questions regarding this variable, divided by the total number of questions (i.e. the separate means of the four questions about cognitive age were added up to create one coherent mean of cognitive age, and this mean was then divided by four). These newly created variables were visually checked for normality, and this was found to be the case.
§ 4.2: Descriptive statistics
330 questionnaires were send out, and in total 148 questionnaires were returned, resulting in a response rate of 44.8 percent. A total of 142 questionnaires were useful for analysis. Of these 142 respondents, 57 are male (40.1 percent), 85 respondents are female (59.9 percent). The average
37 chronological age of the respondents age is 69, with a standard deviation of 12. The youngest
respondent is 50 years old, whereas the oldest respondent is at the respectable age of 92 years. The average age of male respondents is significant slightly higher than the average age of female
respondents, respectively 72 years old versus 67 years old (t =2.27, df = 140, p = 0.025). The average cognitive age of the respondents is 61, with a standard deviation of 12. There is no significant difference between male and female respondents (t = 1.541, df = 140, p = .126). The difference between chronological age and cognitive age is significant (t = -8.418, df = 140, p < .001), indicating that older people tend to feel younger than their chronological age.
Seven respondents (4.9 percent) have primary education as their highest completed educational level, and 15 respondents (10.6 percent) have academic education as their highest educational level. Modus is at higher education, with 31 respondents (21.8 percent), closely followed by lower general secondary education, with 28 respondents (19,7 percent). Results of the Man-Whitney-U test show no significant difference between male and female respondents and educational level (z = -1.176, p = .24)
44 respondents (30.9 percent) have a job at this moment, whereas 23 respondents (16.2 percent) have a voluntary job. Five respondents (3.5 percent) indicate they do not have a job at the moment, and 70 respondents (49.3 percent) ae retired from working.
88 respondents (61.9 percent) indicate that they have an income below or around modal, which corresponds closely with the average in Noordwijk (CBS, 2010) and also with the average in The Netherlands (CBS, 2012). The incomes above modal also correspond with the average in The Netherlands; 25 respondents (17.6 percent) have an income one and a half times modal, whereas 19 respondents (13.4 percent) have an income about two times modal and 10 respondents (7.0 percent) have an income over two times modal. Results of the Man-Whitney-U test shows no significant difference between male and female respondents regarding income (z = 576, p = .449) and also no significant difference between educational level and income (z = -.535, p = .592).
38
§ 4.3: Reliability and validity
Before one could test for effect and relationships between different variables, these variables should be tested for reliability, to see if the measures produce consistent results when the same entities are measured under different conditions and validity and to check whether the measures indeed
measure what they intend to measure (Field, 2009). The scales used for this study were derived from different studies. The cognitive age scale was derived from Barak & Schiffman’s study (1980) towards non-chronological age variables, the OSL scale was derived from Steenkamp & Baumgartner’s study (1995) which was intended to find an easier way to measure OSL, and the exploratory behavior scale was derived from Raju’s study (1980) on the relationship between OSL and exploratory behavior. These scales are English by nature, so after consulting two experts, these scales were translated into Dutch, so they suited the respondents. This means that the newly Dutch worded items had to be checked for reliability. In advance, a choice was made for at least an Chronbach Alpha of .8 or higher, to indicate that these items are reliable (Field, 2009).
All cognitive age items showed excellent reliability, with an alpha of .937. Deleting any items won’t have increased Chronbach’s Alpha. The OSL items showed good reliability, with a Chronbach Alpha of .841, and also here deleting any items did not lead to a higher Chronbach Alpha. The exploratory behavior items did not scored so well on reliability at first sight, with a Chronbach Alpha of .781. After deleting item 26, Chronbach Alpha increased towards .826, which is considered to be good. If item 36 was also deleted, Chronbach’s Alpha would have increased towads .856, but the scale had already high internal consistency with Chronbach’s Alpha at .826 and the means of the different items did not differ that much from each other, so this item was kept.
All this together lead to the conclusion that the newly worded items have a relatively high internal consistency, because a reliability coefficient of .70 or higher is considered ‘acceptable’ in most social science studies (Lance, Butts, & Michels, 2006)
The construct validity is also investigated, with factor analysis used to do this. Factor analysis is used for identifying whether the correlations between a set of observed variables stem from their
39 relationship to one or more latent variables in the data, each of which take the form of a linear model (Field, 2009). The KMO-test was used, values above .5 will be seen as acceptable, but the closer to 1, the better (Kaiser, 1974; Hutcheson & Sofroniou, 1999). Outcome of the test was that all of the items scored above .5, at a significance level of < .001. Three items scored above .8, which are great values, seven items scored above .7, which is considered to be good. Fourteen items score above .6 and eleven items scored above .5, which are mediocre, but acceptable scores (Hutcheson & Sofroniou, 1999).
This analysis therefore indicated that the items are valid, and ‘measure what they have to measure’.
§ 4.4: Hypotheses testing
The scales and variables proved to be reliable, so it was legitimate to test the hypotheses. The first set of hypotheses were aimed at investigating the relationship between demographics such as age, gender, educational level, employment status and income and cognitive age, whereas the second hypothesis tested the relationship between cognitive age and OSL. The third set of hypotheses were set up in order to investigate the relationship between cognitive age and the exploratory behaviors brand switching, innovativeness, repetitive behavior proneness and risk taking. After testing the third set of hypotheses, a regression analysis was done in order to see how the exploratory behavior variables related to each other and cognitive age.
§ 4.4.1: Hypotheses regarding demographics and cognitive age
It was hypothesized that cognitive age is related to age, educational level, employment status and income. These sets of hypotheses were set up to verify earlier findings regarding these topics, which weren’t consistent over time. The tests used for analysis were Pearson’s correlation for the relation between chronological age and cognitive age, and Spearman’s correlation for the relation between
40 education and cognitive age, and income and cognitive age. A One-Way-Anova test was done for the relationship between employment status and cognitive age.
The results shown that chronological age and cognitive age are significantly related (r = .788, p < .001), but there was no significant relationship between education and cognitive age (r = -.054, p = .523) and between income and cognitive age (r = -.047, p = .578).
At first, it appears that there is a significant relationship between employment status and cognitive age (F = 36.15, df = 3, df2 = 138, p < .001), however, using chronological age as a covariate, this difference disappears (F = 1.211, df = 3, p = .308), and therefore there was also no significant relationship found between employment status and cognitive age. The only demographic variable that had a significant relation with cognitive age, was chronological age
§ 4.4.2: Hypothesis regarding cognitive age and OSL
The next step was to find out if there is a significant relation between cognitive age and OSL, which has not been the subject of research before and has some promising aspects. It was hypothesized that there is a negative relationship between cognitive age and OSL, indicating that if cognitive age goes up, OSL goes down. This hypothesis was tested with Spearman’s correlation.
The results showed that there is a significant relationship between cognitive age and OSL (r = -.383, p < .001). After this test, a partial correlation test was done to control the partial relation exploratory behavior could have on both these variables. Outcome of the test was that the relationship between cognitive age and OSL was a little weaker, but still significant (r = -.338, df = 139, p < .001)
§ 4.4.3: Hypothesis regarding cognitive age and exploratory behavior
As cognitive age relates to OSL substantially, it could be very well possible that cognitive age also influences exploratory behavior, as it has been proved by Raju (1980) that OSL can act as predictor for exploratory behavior. It was hypothesized that there is a negative relationship between cognitive
