A Value Inventory and Value Dimensions

ABSTRACT - Most contemporary research on values is based on the lists of Rokeach (1973). A new value inventory is constructed starting from qualitative interviews. The coded conversations yielded 1372 value descriptions, that were condensed to 160 representative values. From the database in which the condensation process was coded, a value thesaurus could be constructed. From the 160 values, six factors emerged: relations, social orientation, values from old, safety, family life and to be conformistic. These factors appeared to be stable. Some of the factors are similar to factors in other studies.


Joke Oppenhuisen and Dirk Sikkel (2001) ,"A Value Inventory and Value Dimensions", in E - European Advances in Consumer Research Volume 5, eds. Andrea Groeppel-Klien and Frank-Rudolf Esch, Provo, UT : Association for Consumer Research, Pages: 73-81.

European Advances in Consumer Research Volume 5, 2001      Pages 73-81


Joke Oppenhuisen, University of Amsterdam, The Netherlands

Dirk Sikkel, University of Tilburg, The Netherlands


Most contemporary research on values is based on the lists of Rokeach (1973). A new value inventory is constructed starting from qualitative interviews. The coded conversations yielded 1372 value descriptions, that were condensed to 160 representative values. From the database in which the condensation process was coded, a value thesaurus could be constructed. From the 160 values, six factors emerged: relations, social orientation, values from old, safety, family life and to be conformistic. These factors appeared to be stable. Some of the factors are similar to factors in other studies.


Since the ground breaking work of Rokeach (1973), values have become a common concept in the study of consumer behavior and an important tool in marketing research. Many marketing research agencies routinely include Rokeach’s 18 instrumental values and 18 end values into their questionnaires to measure life styles and preferences. After Rokeach, many approaches to apply values in consumer research have been developed. Well known examples in consumer research are LOV (Kahle, 1996), VALS (Mitchell, 1983) and VALS 2 (Winters, 1989). The value concept has been refined by Vinson, Scott & Lamont (1977) and Gutman (1982), Reynolds and Gutman (1984), where values are ordered in the form of chains.

A more general line of value researchis followed by Schwartz and Bilsky (1987, 1990) and Schwartz (1992, 1994). They hypothesized that values arise as a cognitive representation of general human needs that are general to all cultures:

- the needs of an individual as biological organism

- the needs of coded social interaction

- the needs of survival and well being of groups

Those needs give rise to eleven value domains into which all values can be grouped: power, achievement, hedonism, stimulation, self-direction, universalism, benevolence, tradition, conformity, security and spirituality. With the exception of spirituality, these domains are recovered as value clusters in a scaling analysis. Values within each of the clusters are positively correlated in the sense that when an individual gives one value within a cluster a high priority, he has a relatively high probability to give other values within the same cluster also high priority. The analysis is based on the Rokeach values, supplemented with 35 other values of which the origin is unclear; all values, can be attributed to the domains that were postulated by Schwartz and Bilsky.

Both lines of research have their value in our attempts to understand how values come into existence, how they can be compared across cultures and what role they play in consumer behavior. There is, however, one question that receives little attention in these research areas, namely: which values are there? This question was, of course, addressed by Rokeach as he provided the fundamental starting point from which the others carried on. The procedures he used, however, although quite laborious, seem to be undocumented and rather subjective. His 18 end values were based on

- research of the literature, both inside and outside the American society

- the personal end values of Rokeach himself

- the values of 30 psychology students

- values from a representative survey in the city of Lansing

This resulted in a list of several hundreds of values. This list was reduced on the basis of a number of considerations. The following values were deleted

- synonyms

- values that empirically appeared to be synonyms

- values that overlapped

- values that were too specific

- values that could not be end values

For the 18 instrumental values, similar procedures were used. Rokeach (1973, p. 30) contains the following comment

"As can be seen, the overall procedure employed in selecting the two lists is admittedly an intuitive one, and there is no reason to think that others, working independently would have come up with precisely the same lists of 18 terminal and 18 instrumental values. It would be interesting to see which values other might produce working independently and using the same criteria that have been described here."

This, however, never happened, at least not on the scale of the Rokeach project. Many researchers have used the Rokeach values or, like Schwartz, have taken it as a starting point. In almost all fundamental value research, the focus is on value structure or value dimensions, but not on procedures to construct a set of values that is in some sense complete.

In marketing practice, values are used for market segmentation and for brand positioning. Brand personalities mainly are made up of values, which are communicated through advertising and product design. The reason for starting a new value project was the feeling in advertising agencies that for communication purposes the Rokeach values were outdated. In the new project an attempt is made to elicit values in a way that

1. it satisfies scientific standards like verifiability, reproducibility and reliability.

2. it gives at least a rough idea about its completeness

3. it yields tools for practical application

The values are based on qualitative interviews with a representative sample in the Netherlands. Given the fact that the field of application is communication we have chosen to compare and scale the values on the basis of similarity of meaning (instead of correlation between priorities). This leads to a set of procedures that is entirely different from those used by Schwarz and Bilsky (1990).

Section 2 gives some general considerations about choices that had to be made at the start of the project. The actual value inventory is described in section 3. In section 4 the underlying value dimensions are discussed and the results are compared to those of others. Section 5 deals with validity and reliability issues. Section 6 concludes.


Value definitions

There are many value definitions. Lautmann (1971) found 178 definitions in 400 different references. According to Becker and Nauta (1983) the value definitions can be split into two different groups.

1. Values that are viewed from the needs of the individual; they are linked to motivation and satisfaction of needs.

2. Values that are viewed from the needs of society; they are linked to purposes, guidelines for behavior and evaluation of behavior.

For application to consumer behavior, the obvious approach is to take the individual level as starting point and to investigate what a consumer considers to be values. Schwartz and Bilsky (1987) conclude that the majority of value definitions have a number of common characteristics. In their view, values are

a) concepts or beliefs

b) about desirable end-states or behaviors

c) that transcend specific situations

d) guide selection of evaluation of behavior and events

e) are ordered by relative importance

The definition of Rokeach is in agreement with these elements:

a value is an enduring belief that a specific mode of conduct or end-state of existence is personally or socially preferable to an opposite or converse mode of conduct or end-state of existence

Rokeach explicitly states that each individual has a value system, which is an ordering of values by relative importance; this system serves to determine which value has priority in practical situations where there is a conflict between values, e.g. between honesty and responsibility or between friendship and patriotism. In the situation of the new inventory, thousands of sentences had to be examined with respect to the presence of values. This required a simple criterion to determine if a concept was a value. The following working definition, which is a simplified version of Rokeach’s definition, appeared to be convenient

a value is a state or a mode of conduct if people strive at it or consider it to be desirable

This definition covers only the elements a) and b) that were mentioned by Schwartz and Bilsky (1987). The requirement that values transcend specific situations was met by the design of the interviews (see section 3). The last two elements have more the character of properties and may be checked in subsequent research.

There is mixed evidence on the distinction between end-values and instrumental values. Heath and Fogel (1978) conducted an experiment in which they had their respondents classify Rokeach values into instrumental values and end-values. It appeared that there was a lot of disagreement among the respondents and between the classification of the respondents and the classification of Rokeach. Given the results of much laddering research it is obvious that at the individual level the distinction between instrumental values and end-values is valid, but that it is not possible to make fixed classification of values that holds for all individuals. Therefore the distinction between instrumental and end-values was dropped.

Cultural limitations

Values, and especially value descriptions, have a cultural component. Cultural differences may arise in

1. The value concept. A concept may exist in some cultures and not in other cultures. A striking example is the Rokeach end value 'salvation’; its Dutch translation, verlossing, has lost its religious meaning for the majority of the population. Many interpreted the word as 'delivery’ or 'stopping of pain’.

2. The perception of a concept as a value. A concept may exist in different cultures, but may not be seen as something to be strived at. A clean environment has not always been seen as something desirable.

3. The importance of a value. Dependent of economic conditions, a clean environment may or may not be seen as an issue of high priority.

4. The wording of a value. The word 'environment’ has only in the last decades obtained the meaning of 'ecological environment’.

All these considerations show that the usability of value lists have their limitations. The Rokeach value 'salvation’ has lost its meaning in Dutch society in 2000; similarly, values which are elicited from the Dutch population in 2000 can not (uncritically) be used in other societies. This does not necessarily hold true for underlying structures, which, according to Schwartz, may be universal but take their own shape in every different culture. For the value inventory it was decided to restrict it to the autochthonous population in the Netherlands.


The basic idea of the value inventory was to elicit values without using any preconceived notion or idea of how values can be ordered, clustered or arranged into dimensions. The values had to be based on empirical data; they had to be very or fairly general, thus satisfying requirement c) of Schwartz and Bilsky (1987). Given these requirements it was decided that the interviews had to cover general life domains. After experimentation with different interviewing techniques it appeared possible to elicit values from respondents in qualitative interviews, in which many aspects of life in general were discussed. Examples of such aspects are: children, work, friendship and politics. An exhaustive list is given in the appendix, table A1. The interviews were conducted in a style that resembles laddering. They had the following format: Each aspect of life was introduced by a photograph or a set of photographs, mainly taken from newspapers and magazines. After a few introductory questions on each subject, the respondent was prompted to explain why he had given his answer and, more in particular, why the values he mentioned were important. A significant difference from laddering interviews was that the values did not necessarily have to relate to the respondent, but could also be values of specific groups or other people in general.



The data were collected in 20 qualitative interviews, each of which lasted approximately three hours. The respondents were evenly spread with respect to sex, education and region. The conversations were taped and later transcribed. Every single sentence was scanned for values. The words of the respondents were taken literally and left as much as possible in their context. The result was that a large number (1372) of different value descriptions were obtained. These descriptions did not represent 1372 different values, but they did indicate 1372 different ways people talk about values. A problem with 1372 value descriptions is that their number is too large to do subsequent quantitative research. As Table 1 shows, there is however much redundancy in the meanings of the different value descriptions. From the 1372 descriptions, 160 values were selected by the researchers that were representatives for larger sets of value descriptions. The value descriptions in table 1 are connected to 'being boss’.

In selecting the 160 values, the following principles were used:

1. Synonyms; value descriptions that were synonymous were brought together

2. Ambiguity; value descriptions that had more than one meaning are excluded

3. Number of times a value is mentioned; frequently mentioned values are chosen

4. Dictionary; value descriptions which were correct according to the Dutch Van Dale dictionary were selected.

An English translation of the 160 values is given in the appendix (Table A2). Like in the selection process of Rokeach, this part of the process depended partly on the intuition of the researchers. The main difference is that the decisions of the researchers are documented in a data base and can easily be checked by others.

The attribution of value descriptions to sets is not exclusive, i.e. a value description can be a member of more that one set that is defined by a representative value.


Table A.2 shows that the number of representative value descriptions is still large, far too large to include them routinely into questionnaires. Moreover, the value descriptions may contain a hidden structure that may be of theoretical interest. It is, therefore, desirable to uncover such a presumed structure. Basically, there are two ways to obtain evidence about such a structure:

a. Collect data about the importance or priority of the values. This leads to a correlation matrix, where a pair of values that have simultaneously low and high importance has a positive correlation and a pair of values in which a high importance of one value predicts a low importance of the other value, has a negative correlation.

b. Collect data about the meaning of the value descriptions. This gives evidence about what consumers consider to be similar descriptions. Two values which may have simultaneously high and low importance still can be considered to be dissimilar.

Because the field of application, advertising and the construction of brand personality, mainly is concerned with meaning, it was decided to use the second method. The implication was that 160*159/2 = 12720 pairs of value descriptions had to be judged with respect to similarity of meaning. This was carried out by the CentERpanel, a panel of 2000 households who respond every week to a computerized questionnaire. The pairs of value descriptions were distributed over the respondents in such a way that every respondent judged approximately 60 pairs of values. In this way, for every distance between a pair of value descriptions, approximately 12 observations were obtained. This led to a distance matrix to which Principal Components Analysis (PCA) was applied. The factor solution was not rotated.

The PCA yielded six factors with a clear interpretation. For both directions the ten value descriptions with the most extreme factor scores are listed in Table 2.

All factors showed a contrast between value descriptions with a clearly different meaning. This meaning is printed bold in table 2. All factors have, surprisingly, a common element. All can be interpreted as the distinction between bond and freedom. The object of the bond is different for each factor. Factor 1 can be interpreted as the bond to relations and social life in contrast to the freedom of making a career (although it is the only factor where it could be argued that the object of the bond is at the opposite end: career, leaving no freedom for social relations). In factor 2, the object of the bond clearly is the fellow human being. Factor 3 shows a bond with rather traditional values like patriotism, pride, distinction and toughness. Factor 4 expresses a different type of bond; values like cleanliness, richness, safety and to make a living all point to personal security. Factor 5 expresses bond to family life. In factor 6, the object of the bond is the approval or admiration by others, with the extreme opposite 'to go your own way'. Given this interpretation, it is clear why the factors emerged this way: they all correspond to a very general aspect of life.




As it is claimed that the procedure that is used in this project is an improvement compared to Rokeach, the issue of validity and reliability is very relevant. The fact that in some sense few observations were used ('only' 20 qualitative interviews and 'only' 12 observations per distance between a pair of values), makes this issue even more pressing. In this section, some evidence with respect to validity and reliability is presented. We will discuss the stimuli that were used, the values extracted from the interviews and the stability of the factor solution.

The stimuli

The photographs that were used as stimuli were meant as an introduction to a subject. It was, of course, undesirable, that respondents reacted to specific details of the photographs. In the preparatory stage it appeared that some photographs caused such unwanted reactions. To test whether the particular choice of the photographs had any influence on the results, two different series were used. When the series are equivalent, there may be no difference in the average scores of the elicited values on the factors (a difference would indicate that a stimulus created a bias with respect to a factor); there also may be no difference in standard deviations of the factor scores (a smaller variance for a series would indicate that the issues that relate to the factor were not properly discussed in the interviews). In Table 3 the average factor scores and the standard deviations are listed; these scores are weighted by the number of occurrences of a value description.







The similarity of the figures for photo series a and b in Table 3 show no evidence that the specific characteristics of the photographs have led to any bias.

The number of qualitative interviews

A rule of thumb for qualitative interviewing is that the number of interviews is sufficient when no new information is encountered in new interviews. We can apply this rule both to the total number of value descriptions and to the 160 representative values. The rule can also be applied to the factor solutions.

Figure 1 shows the average number of value descriptions that were found in a given number of interviews. The average is taken over all possible combinations of interviews (so for 5 interviews, 20!/(5!xl5!) combinations were considered). The figure shows clearly that at 19 and 20 interviews the number of descriptions is still growing steadily. As a consequence, it has to be assumed that there exist many more value descriptions, and the list of 1372 descriptions is far from complete.

In Figure 2 it is shown that the average number of representative value descriptions practically stops growing after eleven interviews. At that point the average number has reached 148. Of course, the lack of growth at the 17 and more can be considered to be an artefact, since the total number of representative values was fixed to 160. Therefore other arguments have to be found to show that the 160 representative values are sufficient to reveal the underlying factor structure. This will happen in the next subsection.







Stability of the factor structure

The stability of the factor solution was tested by three split-half experiments

a. The judges of the distances between pairs of value descriptions were randomly split into two equal groups; the PCA's were replicated for each group

b. The qualitative inter-views were split into two groups; the PCA's were replicated for the value descriptions that came from each of the groups

c. The 160 representative value descriptions were randomly split into two groups of 80 descriptions; the PCA's were replicated for each of the groups

The results were evaluated on the basis of correlations between factor scores. Because of the possibility that a factor solution in a subsample was rotated with respect to another factor solution, also multiple regression was applied to see whether the factors spanned the same linear subspace.

In Tables 4a and 4b the results of experiment a are shown. The judges were split into two random groups. As a consequence, on the average each of the distances was judged six times. This led to two new distance matrices. On both of them PCA was applied. The first three factors appear to be stable. For the factors 4 through 6, however, the correlations, especially between subsamples 1 and 2 are low. This may be caused by the fact that the factors of the first subsample are rotated with respect to the factors of the second subsample. This can be verified by regression of the factors on the first subsample on those of the second subsample. The figures in Table 4b show that this is the case. The multiple correlation between the factor scores of factor 4 in the complete sample and the factor scores of factors 4 through 6 of subsample 1 is equal to .892. The multiple correlations between the factor scores in each of the subsamples are somewhat lower. This is, naturally, due to the fact that these subsamples are independent. Generally speaking, the correspondence between the different factor solutions is satisfactory and indicates that the distances between the value descriptions are judged by a sufficient number of judges.

In experiment b (Table 5), the interviews were split into two subsamples. The first subsample corresponded to photo series a, the second subsample to photo series b.

In each subsample, only those representative value descriptions were used that were elected in the corresponding interviews. The correlations now are calculated between the factor scores of the values that were in both samples that were compared. It is immediately obvious that all factor solutions are virtually identical, which could, of course, be expected as both subsamples have the majority of the representative value descriptions in common.



A much stricter test is experiment c, where it is examined how the factor solution depends on the particular choice of value descriptions. The representative value descriptions are randomly split into two groups of 80 descriptions. On both groups PCA is applied. In this case it is impossible to relate the two subsamples to each other as they have no value descriptions (and hence no factor scores) in common. It is, however, possible to relate the factor scores from the subsamples to the original factor scores, using multiple regression as in experiment a. The results are given in Table 6.

The correlations show how well the first 6 factors of the subsamples explain the corresponding factor scores of the complete sample of value descriptions. Again, the correlation coefficients are quite high, which indicates that the factor solution is stable. This result is very significant. Its implication is that, although we can not claim that the 160 representative values are in some sense complete, optimal or objective, their underlying semantical structure is stable and apparently does not depend very much on the particular choice of descriptions, as long as all value dimensions are sufficiently represented.


The study started at the same point where Rokeach started in the mid sixties: how to construct a list of values from scratch? It was decided to use only data from qualitative interviews. Given the results of the analysis:

- a set of 160 representative values could be constructed that was almost complete after 10 interviews,

- the underlying dimensions appeared to be stable and robust to the use of the representative values,

this decision seems to be justified. The underlying dimensions can be interpreted as the bond to six very abstract, but also very recognizable life domains: social life, the fellow human being, the past, personal security, family life and approval by others. These are domains in which, in the view of the Dutch respondent, values can be different for different people; apparently, such differences matter. A significant result also may be the value dimensions we did not find: politics and religion. These subjects were explicitly discussed in the interviews (see Table Al, topics 12, 13,14, 15, 19, 22) but did not lead to clearly contrasting values. Apparently, to the average citizen (and consumer) these subjects are not relevant in daily life.







Becker, JW. & Nauta, A.P.N. (1983). Enkele gegevens over de waarden in Nederland na 1945. 's Gravenhage: VUGA.

Gutman, J. (1982). A Means-End Chain Model based on Consumer Categorization Processes. Journal of Marketing 46, 60-72.

Hansen, F. & Kvwrk, A. (1998). Value dimensions - are they really that many? Copenhagen: Copenhagen Business School.

Heath, R.L. & Fogel, D.S. (1978). Terminal and Instrumental? An inquiry into Rokeach's Value Survey. Psychological Reports 42, 1141-1154.

Hofstede, G. (1991). Cultures and Organisations, Software of the Mind. London: McGraw Hill.

Kahle, L.R. (1996). Social Values and Consumer Behavior: Research from the List Of Values. Mahwah, NewJersey: Lawrence Erlbaum Associates.

Lautmann, R. (1971). Wert und Norm: Begriffsanalyse ffir die Soziologie. Opladen: Westdeutsche Verlag.

Mitchell, A. (1983). The nine American Lifestyles. Who we are and where we're going. New York: MacMillan Publishing Co. Inc.

Reynolds, T.J. & Gutman, J.(1984). Advertising is Image Management- Journal of Advertising Research 24, 27-34.

Rokeach, M. (1973). The Nature of Human Values. New York: The Free Press.

Schwartz, S.H.(1992). Universals in the Content and Structure of Values: Theoretical Advances and Empirical Tests in 20 Countries. Advances in Experimental Social Psychology 29, 1-65.

Schwartz, S.H. (1994). Are there Universal Aspects in the Structure and Content of Human Values? Journal of Social Issues 50, 1945.

Schwartz, S.H. & Bilsky, W. (1987). Toward a Universal Psychological Structure of Human Values. Journal of Personality and Social Psychology 53, 550-562.

Schwartz, S.H. & Bilsky, W. (1990). Toward a Theory of the Universal Content and Structure of Values: Extensions and Cross-Cultural Replications. Journal of Personality and Social Psychology 58, 878-891.

Vinson, D.E., Scott, J.E. & Lamont, L.M. (1977). The role of personal values in marketing and consumer behavior. Can personal values be used to assist marketeers on determining consumer choice behavior. Journal of Marketing 41, 44-50.

Vromen, M. (1995). Personal Drives Analysis. Amstelveen: FHV/ BBDO.

Vyncke, P. (1992). Imago management. Handboek voor Reclamestrategen. Gent: Mys & Brees.

Winters, W (1989). SRI Announces VALS 2. Marketing Research 2,67-69.



Joke Oppenhuisen, University of Amsterdam, The Netherlands
Dirk Sikkel, University of Tilburg, The Netherlands


E - European Advances in Consumer Research Volume 5 | 2001

Share Proceeding

Featured papers

See More


Major or Minor: When Foreign Language Increases Versus Decreases Cheating

Jia Gai, Erasmus University Rotterdam, The Netherlands
Stefano Puntoni, Erasmus University Rotterdam, The Netherlands

Read More


The Best of Both Worlds: Androgyny in Consumer Choice

Niusha Jones, University of North Texas
Blair Kidwell, University of North Texas

Read More


P2. The Upside of Myopic Loss Aversion

Daniel Wall, Carnegie Mellon University, USA
Gretchen Chapman, Carnegie Mellon University, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.