# Chapter 4: Measuring values

To go from theory to statistical tests to results can be a long and complicated process. The first step is to formulate specific questions that provide the information you want. The second step is to formulate a set of response alternatives that yield measures you can analyse. The third step is to collect the data. The fourth step is to prepare the variables for analysis, and the final step is to perform the analysis and interpret the results.

This chapter illustrates the first, second and fourth of these steps in greater detail. The first part discusses how we move from the conception of values to the set of questions with their response alternatives that measure values in the ESS survey. Then you will be guided through the process of preparing the original data for analysis.

Page 1

# Operationalization

Values are affect-laden beliefs that refer to a person’s desirable goals. Values guide the selection or evaluation of actions, policies, people and events. To measure peoples’ values, the ESS survey uses a modified version of a recently developed scale called the Portrait Values Questionnaire (PVQ). The PVQ includes short verbal portraits of different people. Each portrait describes a person’s goals, aspirations, or wishes that point implicitly to the importance of a single basic value. For example: “Thinking up new ideas and being creative is important to him. He likes to do things in his own original way,” describes a person for whom self-direction values are important. “It is important to him to be rich. He wants to have a lot of money and expensive things,” describes a person who cherishes power values. By describing each person in terms of what is important to him or her - the goals and wishes he or she pursues - the verbal portraits capture the person’s values without explicitly identifying values as the topic of investigation.1

For each portrait, respondents answer: “How much like you is this person?” The response alternatives are; “very much like me”, “like me”, “somewhat like me”, “a little like me”, “not like me”, and “not like me at all”. For each portrait, respondents choose their response by checking one of six boxes labelled with the response alternatives. Thus, respondents’ own values are inferred from their self-reported similarity to people who are described in terms of particular values. The similarity judgments are transformed into a 6-point numerical scale. Note that respondents are asked to compare the portrait to themselves rather than themselves to the portrait. Asking them to compare other to self directs attention only to the aspects of the other that are portrayed. Thus, the similarity judgment is also likely to focus on these value-relevant aspects. In contrast, asking respondents to compare self to other would focus attention on self and might cause respondents to think about the large range of self-characteristics accessible to them [Sru83] [Hol83] [Tve77]. Not finding these characteristics in the portrait, respondents might overlook the similarity of values.2 Because the PVQ uses concrete statements that are not cognitively complex, it is suitable for use with all segments of the population including those with little or no formal schooling.

The PVQ used in the ESS survey includes 21 value items. These items are combined into ten indexes, one for each of the ten basic values. There are three items to measure universalism, and two each for the other nine basic values. An item measures a basic value if the aims, goals, wishes, or efforts of the person described express or promote the central goal of the basic value or lead to its attainment. The items were constructed to cover the different conceptual components of each value.

Table 4.1 Operationalization of the ten basic values in the ESS survey
VALUE and central goal Items that measure each value with their ESS labels

### POWER

Social status and prestige, control or dominance over people and resources.
• Imprich: It is important to him to be rich. He wants to have a lot of money and expensive things.
• Iprspot: It is important to him to get respect from others. He wants people to do what he says.

### ACHIEVEMENT

Personal success through demonstrating competence according to social standards.
• Ipshabt: It is important to him to show his abilities. He wants people to admire what he does.
• Ipsuces: Being very successful is important to him. He hopes people will recognize his achievements.

### HEDONISM

Pleasure and sensuous gratification for oneself.
• Impfun: He seeks every chance he can to have fun. It is important to him to do things that give him pleasure.
• Ipgdtim: Having a good time is important to him. He likes to "spoil" himself.

### STIMULATION

Excitement, novelty, and challenge in life.
• Impdiff: He likes surprises and is always looking for new things to do. He thinks it is important to do lots of different things in life.
• Ipadvnt: He looks for adventures and likes to take risks. He wants to have an exciting life.

### SELF DIRECTION

Independent thought and action choosing, creating, exploring.
• Ipcrtiv: Thinking up new ideas and being creative is important to him. He likes to do things in his own original way.
• Impfree: It is important to him to make his own decisions about what he does. He likes to be free and not depend on others.

### UNIVERSALISM

Understanding, appreciation, tolerance and protection for the welfare of all people and for nature.
• Ipeqopt: He thinks it is important that every person in the world should be treated equally. He believes everyone should have equal opportunities in life.
• Ipudrst: It is important to him to listen to people who are different from him. Even when he disagrees with them, he still wants to understand them.
• Impenv: He strongly believes that people should care for nature. Looking after the environment is important to him.

### BENEVOLENCE

Preservation and enhancement of the welfare of people with whom one is in frequent personal contact.
• Iphlppl: It is very important to him to help the people around him. He wants to care for their well-being.
• Iplylfr: It is important to him to be loyal to his friends. He wants to devote himself to people close to him.

Respect, commitment and acceptance of the customs and ideas that one's culture or religion impose on the individual.
• Ipmodst: It is important to him to be humble and modest. He tries not to draw attention to himself.
• Imptrad: Tradition is important to him. He tries to follow the custom handed down by his religion or his family.

### CONFORMITY

Restraint of actions, inclinations, and impulses likely to upset or harm others and violate social expectations or norms.
• Ipbhprp: It is important to him always to behave properly. He wants to avoid doing anything people would say is wrong.
• Ipfrule: He believes that people should do what they are told. He thinks people should follow rules at all times, even when no-one is watching.

### SECURITY

Safety, harmony and stability of society, of relationships, and of self.
• Impsafe: It is important to him to live in secure surroundings. He avoids anything that might endanger his safety.
• Ipstrgv: It is important to him that the government ensures his safety against all threats. He wants the state to be strong so it can defend its citizens.

• [1] Separate male and female versions of the questionnaire were used in the first wave of the ESS. The difference between the two in English is in the prepositions: "he", "his", "him" in the male version, "she", "hers", "her" in the female version.
• [2] Greater accessibility of self-knowledge as compared with knowledge of others presumably characterises those from cultures where people have an independent construal of self (Markus & Kitayama, 1991). This is thought to include most of the population in European countries. It may not, however, characterise immigrant groups in European countries, or those from rural areas.
Page 2

# From Item to Index

In the preceding chapters, you have worked with ready-to-use indexes; now you will make them yourself. Instead of working online, you must download the data to your own computer. You are free to download the data in any of the available formats, but be aware that all syntax examples and explanations will be given with reference to SPSS.

### Introductory comments to those using SPSS

First, you are recommended to organize your work the following way:

• Make an extra copy of the downloaded file. If anything goes wrong with your file, you can start again with the backup.
• Store all your work in syntax files. You can generate syntax through the menu by clicking “PASTE” instead of “OK”. The great advantage with syntax is that it is easy and fast to document what you have done, to redo all your work, or to modify/correct errors. If something goes wrong with your data, a properly documented syntax can save you a lot of work.
• You should document your work by writing comments into your syntax. A comment must always start with “*” and end with a full stop “.”. You cannot write several sentences in one single comment; SPSS reads from the “*” to the first “.”.

Download the dataset in a format you are familiar with. You must decide if you would like to work with data from 2008 (ESS 4) or from 2002 (ESS 1). Please note that all examples and solutions will refer to 2002.

### First round of preparation, cleaning and recoding

The purpose of the Portrait Values Questionnaire is to assess each respondent’s value priorities. Sometimes this is not possible, either because of missing data or because respondents use only one or a few response categories. Missing data make it difficult to assess value priorities accurately because there may be no items or only one that measures how important a particular value is to a respondent. Respondents who use only one or two points on the response scale to answer almost all items have probably not tried to make discriminations among values, so their responses do not reflect their actual priorities. If our aim were to evaluate the quality of the Portrait Values Questionnaire, we would want to examine the amount of missing data and the proportion of respondents who do not discriminate among the values. However, if our aim is to investigate substantive questions such as the associations of values with one another or with behaviour, attitudes, and background variables, it is better to drop those who did not respond to many value items or who did not discriminate in their responses to the value items.

1. Compute a filter variable that identifies those who have more than 5 missing on the 21 value items, and those who have given the same answer to more than 16 of the 21 value items. Delete these respondents from the file.
SPSS

Open the file you downloaded in the first exercise. Paste the following syntax into a new syntax window (Open a new syntax window: File – New – Syntax.). Then you should run the syntax.

*This syntax eliminates 5822 respondents from the file.

COMPUTE drop=0.
COUNT
count1 = ipcrtiv to impfun (1)/count2 = ipcrtiv to impfun (2)/count3 = ipcrtiv to impfun(3)/
count4 = ipcrtiv to impfun (4)/count5 = ipcrtiv to impfun (5)/count6 = ipcrtiv to impfun (6)/
countmis = ipcrtiv to impfun (SYSMIS,7,8,9).
if (max(count1 to count6)>16 or countmis>5) drop=1.
value labels drop 0 'keep' 1 'drop'.
execute.
*Unselected cases are deleted from the file.
FILTER OFF.
USE ALL.
SELECT IF(drop = 0).
EXECUTE.
2. Frequency
Have a look at the frequencies for the variables "Ipmodst" and "Ipfrule". SPSS

You could either generate the syntax by clicking through the menu (Analyse – Descriptive statistics – Frequencies) and pasting it into a syntax window, or by writing the syntax into a syntax window.

FREQUENCIES
VARIABLES = ipmodst ipfrule
/ORDER= ANALYSIS.
EXECUTE.
3. Recode
The “Ipmodst” and “Ipfrule” variables are, like all of the value items, coded with 1 representing “very much like me” and 6 “not like me at all”. It is more intuitive that a large number represents “very much like me”, and a low number represents “not like me” (large number = much of something, low number = little of something). Reverse the coding of all the value items (old code 1 = new code 6, old code 2 = new code 5, etc.). SPSS

You are going to create 21 new variables with exactly the same information as the 21 original ones. The only difference is that the coding of the valid values is reversed. The proper command is recode. If you use the menu, you will have to repeat the procedure for each variable (Transform – Recode – Into different variables); by writing syntax you can do your work a lot faster. Each step is commented below.

RECODE: List the original variables, define old and new codes in () (please note that the missing codes 7, 8, 9 are not changed). INTO: List the new variables (please note that the order of the new variables must be identical to the order of the original variables. The new variable names start with the letter n, to indicate that they are new.). The command FORMATS is used to decide the format of a variable. F1.0 means that the variable is numeric, one character long and with zero decimals. Because the variables are placed after each other in the data file, we can use “to” (from the first to the last). The command VARIABLE LABELS adds labels to the newly created variables. The command VALUE LABELS adds labels to the newly created codes. Because the labels are identical, we only need to write them once. This gives the following syntax:

RECODE
ipcrtiv imprich ipeqopt ipshabt impsafe impdiff ipfrule ipudrst ipmodst ipgdtim impfree iphlppl ipsuces ipstrgv ipadvnt ipbhprp iprspot iplylfr impenv imptrad impfun (1=6) (2=5) (3=4) (4=3) (5=2) (6=1) (7=7) (8=8) (9=9)
INTO nipcrtiv nimprich nipeqopt nipshabt nimpsafe nimpdiff nipfrule nipudrst nipmodst nipgdtim nimpfree niphlppl nipsuces nipstrgv nipadvnt nipbhprp niprspot niplylfr nimpenv nimptrad nimpfun.
FORMATS nipcrtiv to nimpfun (f1.0).
VARIABLE LABELS
nipcrtiv 'Important to think new ideas and being creative'
nimprich 'Important to be rich, have money and expensive things'
nipeqopt 'Important that people are treated equally and have equal opportunities'
nipshabt 'Important to show abilities and be admired'
nimpsafe 'Important to live in secure and safe surroundings'
nimpdiff 'Important to try new and different things in life'
nipfrule 'Important to do what is told and follow rules'
nipudrst 'Important to understand different people'
nipmodst 'Important to be humble and modest, not draw attention'
nipgdtim 'Important to have a good time'
nimpfree 'Important to make own decisions and be free'
niphlppl 'Important to help people and care for others well-being'
nipsuces 'Important to be successful and that people recognize achievements'
nipstrgv 'Important that government is strong and ensures safety'
nipbhprp 'Important to behave properly'
niprspot 'Important to get respect from others'
niplylfr 'Important to be loyal to friends and devote to people close'
nimpenv 'Important to care for nature and environment'
nimpfun 'Important to seek fun and things that give'.
VALUE LABELS nipcrtiv to nimpfun 1 "Not like me at all" 2 "Not like me" 3 "A little like me" 4 "Somewhat like me" 5 "Like me" 6 "Very much like me" 7 "Refusal" 8 "Don't know" 9 "No answer".
Missing values nipcrtiv to nimpfun (7 to 9).
EXECUTE.

### Indexes

To compute a score for each of the ten values for an individual, use the items that index it from Table 4.2. The items are numbered from 1 to 21. Items that measure the same value were spread throughout the survey rather than clustered together.

Table 4.2. Value, item number and variable name
Values Items in index Variable name
Conformity 7,16 nipfrule, nipbhprp
Benevolence 12,18 niphlppl, niplylfr
Universalism 3,8,19 nipeqopt, nipudrst, nimpenv
Self-Direction 1,11 nipcrtiv, nimpfree
Hedonism 10,21 nipgdtim, nimpfun
Achievement 4,13 nipshabt, nipsuces
Power 2,17 nimprich, niprspot
Security 5,14 nimpsafe, nipstrgv

The score for each value is the mean of the raw ratings given to the items listed in table 4.2 for that value. You could compute these means in two different ways. You could either add the items together and divide the sum by the number of items, for example; Security = (nimpsafe + nipstrgv) / 2, or you could use a function called MEAN. Using the first alternative, a respondent must have answered both questions to get a valid value on the index; the latter alternative will still compute a mean even if only one question is answered. The first method uses all items, and is thus more accurate. The second method ensures that as many as possible of the respondents get a valid value on the index.

1. Create the indexes. You are recommended to use the MEAN function. Use the recoded variables. Later you are going to create another version of the indexes, so you must indicate either in the name or the label, that this is the first version. SPSS

If you choose to use the menu, select “Transform – Compute”.

Compute Apow = MEAN (nimprich, niprspot).
Compute Apow = MEAN (nimprich, niprspot).
Compute Aach = MEAN (nipshabt, nipsuces).
Compute Ahed = MEAN (nimpfun, nipgdtim).
Compute Asti = MEAN (nimpdiff, nipadvnt).
Compute Aself = MEAN (nipcrtiv, nimpfree).
Compute Auni = MEAN (nipeqopt, nipudrst, nimpenv).
Compute Aben = MEAN (niphlppl, niplylfr).
Compute Atra = MEAN (nipmodst, nimptrad).
Compute Acon = MEAN (nipbhprp, nipfrule).
Compute Asec = MEAN (nimpsafe, nipstrgv).
EXECUTE.
VARIABLE LABELS
APow 'Power – mean of raw rating'
Aach 'Achievement – mean of raw rating'
Ahed 'Hedonism – mean of raw rating'
Asti 'Stimulation – mean of raw rating'
Aself 'Self-Direction – mean of raw rating'
Auni 'Universalism – mean of raw rating'
ABen 'Benevolence – mean of raw rating'
Atra 'Tradition – mean of raw rating'
ACon 'Conformity – mean of raw rating'
ASec 'Security – mean of raw rating'.
EXECUTE.
Page 3

# Reliability

Indexes are built because they are expected to provide a more accurate measurement of the theoretical dimension than any single variable can do alone. Hence, the variables included in the index should all measure aspects of the same dimension. For example, if you were asked to create an index for body size, it would be sensible to include height and weight, but it would make no sense to include eye colour. One way of evaluating the extent to which an index measures one dimension that underlies all of its items is to perform a reliability analysis. Cronbach’s alpha is a measure of the internal consistency of the items in a scale, and it ranges from 0 to 1.1 An alpha greater than 0.7 is desirable for indexes that are used as a scale.

1. Find the alpha for each of the indexes. SPSS

Perform ten reliability analyses, one for each value. The example below gives the syntax for power. You could use either the original variables or the recoded variables (as in the example). If you prefer to use the menu: Analyse – Scale – Reliability Analysis.

RELIABILITY
/VARIABLES=nimprich niprspot
/FORMAT=NOLABELS
/SCALE(ALPHA)=ALL/MODEL=ALPHA.

You should not be surprised if the internal reliabilities of several PVQ indexes are relatively low. This reflects two facts. First, the items in the indexes were constructed and selected to cover the different conceptual components of each value, not to be nearly redundant measures of a narrowly defined concept. For example, the power value items tap both wealth and authority, and the universalism items tap understanding, concern for nature, and social concern. If items with more similar meanings were chosen for each index, alpha would be higher, but at the cost of poorer coverage of the breadth of meaning of each of the types of values. Second, each index includes only two or three items. With so few items it is virtually impossible to obtain high alphas unless the items are very similar to one another. Considering the small number of items used to measure each of the ten values and their necessary heterogeneity, even reliabilities of 0.4 are reasonable.

• [1] The explanation of Cronbach's alpha is rather complex, but it could be simplified to the following: The alpha depends upon the number of indicators in the index and the average intercorrelation among the items comprising the index. The more items there are in the index and the higher the average intercorrelation among them, the higher the alpha.
Page 4

# Correcting individueal differences in scale use

People differ in the way they use the response scale of the value items.1 Some individuals spread their responses across the whole scale; they say that some of the people described are a lot like them and that others are not like them. However, some individuals tend to concentrate their responses at one side of the scale (almost all of the people described are a lot like them) or the other side (almost all are not like them). If we ignore these response tendencies, interpreting the responses of these people as they appear, we would infer that all values are important to the first set of individuals and that no values are important to the second set. But that would be wrong.2

What really interests us is the relative importance of the ten values to a person, the person’s value priorities. This is because the way values affect attitudes, feelings and behaviour is through a trade-off or balancing among the different values that are simultaneously relevant to action (see discussion in chapter 3). Behaviours often have opposing implications for the relevant values (e.g., implications of rock climbing for stimulation vs. security). If we make predictions with the absolute importance of any single value for an individual or group, without considering the importance of this value relative to the individual’s or group’s other values, we will fail to take account of the fact that values function as a system [Sch96] [Sch05a] [Sch05b].

For instance, two people may both have scores of 3 for tradition values. One has lower scores (1 or 2) for all the other values, while the other has higher scores (4-6) for all the other values. Obviously, even though both people have the same score for tradition, tradition has higher relative priority for the first person than for the second. Despite their identical absolute scores for tradition, we would expect the first person to attend religious services more often. Thus, to measure value priorities, we need to correct for individual differences in scale use. The scale use correction converts absolute value scores into scores that indicate the relative importance of each value in the individual’s whole value system, i.e., the individual’s value priorities.

1. Compute each individual’s mean score on all 21 value-items. Call this variable MRAT. SPSS

*Use the recoded value items and the function MEAN.

COMPUTE MRAT = Mean (nipcrtiv, nimprich, nipeqopt, nipshabt, nimpsafe, nimpdiff, nipfrule, nipudrst, nipmodst, nipgdtim, nimpfree, niphlppl, nipsuces, nipstrgv, nipadvnt, nipbhprp, niprspot, niplylfr, nimpenv, nimptrad, nimpfun).
VARIABLE LABELS MRAT 'Mean score on all answered value items'.
execute.
2. Centre scores of each of the 10 values for an individual around that individual’s MRAT. Use the ten values computed above, and subtract MRAT from each value.3 Please indicate, either in the label or the name, that this is the centred value score. SPSS

Centre scores of each of the ten values for an individual around that individual’s MRAT. Use the indexes computed above and subtract MRAT.

Compute Cpow = Apow - MRAT.

Compute Cach = Aach - MRAT.
Compute Ched = Ahed - MRAT.
Compute Csti = Asti - MRAT.
Compute Cself = Aself - MRAT.
Compute Cuni = Auni - MRAT.
Compute Cben = Aben - MRAT.
Compute Ctra = Atra - MRAT.
Compute Ccon = Acon - MRAT.
Compute Csec = Asec - MRAT.
EXECUTE.
VARIABLE LABELS
CPow 'Power - Centred value score'
Cach 'Achievement - Centred value score'
Ched 'Hedonism - Centred value score'
Csti 'Stimulation - Centred value score'
Cself 'Self-Direction - Centred value score'
Cuni 'Universalism - Centred value score'
CBen 'Benevolence - Centred value score'
Ctra 'Tradition - Centred value score'
CCon 'Conformity - Centred value score'
CSec 'Security - Centred value score'.
EXECUTE.

These centred value scores should be used in the following types of analyses:

1. Correlation
2. Group mean comparisons, analysis of variance or of covariance (t- tests, ANOVA, MANOVA, ANCOVA, MANCOVA that use values as the dependent variables.
3. Regression:
1. If values are dependent variables
2. If values are predictor variables:
1. Enter up to 8 centred values as predictors in the regression.
1. If all 10 values are included, the regression coefficients for the values may be inaccurate and uninterpretable due to multicolinearity.
2. Choose the values to exclude as predictors a priori on theoretical grounds because they are irrelevant to the topic.
2. If you are interested only in the total variance accounted for by values and not in the regression coefficients, you may include all 10 values as predictors. The R2 is meaningful but, because the 10 values are exactly linearly dependent, the coefficients for each value are not precisely interpretable.
3. In publications, it is advisable to provide a table with the correlations between the centred values and the dependent variables in addition to any regression. These correlations will aid in understanding results and reduce confusion due either to multicolinearity or to intercorrelations among the values.

For multidimensional scaling, canonical, discriminant, confirmatory or exploratory factor analyses use the absolute scores for the 21 items or 10 value means.4 Note that exploratory factor analysis is not suited to reveal the circular structure of relations among the values.5

• [1] For a discussion of the general issue of scale use, see Saris (1988). Schwartz, et al. (1997) examine meanings of such scale use with values as an individual difference variable. Smith (2004) discusses correlates of scale use differences at the level of cultures.
• [2] This assumes that the set of ten individual level values is reasonably comprehensive of the major motivationally distinct values recognised across individuals and cultural groups. Otherwise, it might be that the values the first set of individuals considers unimportant and the values the second set considers important were just not included among the ten values. Empirical evidence supports the assumption that the list of ten values is, in fact, reasonably comprehensive (Schwartz, 1992, 1994).
• [3] When centring, do not divide by the individual's standard deviation across the 21 items. This is because individual differences in variances of value ratings are usually meaningful. Even if, on average, individuals attribute the same mean importance to the set of values, some individuals discriminate more sharply among their values and others discriminate less sharply. Standardising that makes everyone's variance the same (i.e., 1) would eliminate these real differences in the extent to which individuals discriminate among their values.
• [4] In these types of analysis, the exact linear dependence among items, created by centring, is problematic. Aspects of these types of analysis take care of the scale use problem in other ways beyond the scope of this presentation.
• [5] Factors obtained in an EFA with rotation will only partly overlap with the 10 values and exploit chance associations. The first unrotated factor represents scale use or acquiescence. It is not a substantive common factor. A crude representation of the circular structure of values can be obtained using EFA by plotting the value items in a two-dimensional space according to their loadings on factors 2 and 3 of the unrotated solution.
Page 5
• [Hol83] Holyoak, K. J., & Gordon, P.C. (1983). Social reference points. Journal of Personality and Social Psychology, 44, 881-887.
• [Sch05a] Schwartz, S. H. (2005a). Basic human values: Their content and structure across countries. In A. Tamayo & J. B. Porto (Eds.), Valores e comportamento nas organizacões [Values and behavior in organizations] pp. 21-55. Petrópolis, Brazil: Vozes.
• [Sch05b] Schwartz, S. H. (2005b). Robustness and fruitfulness of a theory of universals in individual human values. In A. Tamayo & J. B. Porto (Eds.), Valores e comportamento nas organizacões [Values and behavior in organizations] pp. 56-95. Petrópolis, Brazil: Vozes.
• [Sch96] Schwartz, S. H. (1996). Value priorities and behavior: Applying a theory of integrated value systems. In C. Seligman, J.M. Olson, & M.P. Zanna (Eds.), The Psychology of Values: The Ontario Symposium, Vol. 8 (pp. 1-24). Hillsdale, NJ: Erlbaum.
• [Sru83] Srull, T. K., & Gaelick, L. (1983). General principles and individual differences in the self as a habitual reference point: An examination of self-other judgments of similarity. Social Cognition, 2, 108-121.
• [Tve77] Tversky, A. (1977). Features of similarity. Psychological Review, 84, 327-352.