# 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

#### Footnotes

• [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.

#### References

• [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.