Basic Human Values 2 |
|
| Session Organisers |
Professor Eldad Davidov (University of Cologne, Cologne, Germany and University of Zurich, Zurich, Switzerland) Professor Peter Schmidt (University of Giessen, Giesen Germany) Professor Jan Ciecuich (Cardinal Stefan Wyszyéski University, Warsaw, Poland, and University of Zurich, Zurich, Switzerland) |
| Time | Wednesday 17th July, 11:00 - 12:30 |
| Room | D04 |
Values have held an important position in the social sciences since their inception. They have been used to explain the motivational bases of attitudes and behavior and to characterize differences between both individuals and societies. In 1992, Schwartz introduced a theory of ten basic human values, building on common elements in earlier approaches. The designers of the European Social Survey (ESS) chose this theory as the basis for developing a human values scale to include in the core of the survey. This theory has been extended to include 19 values (Schwartz et al., 2012) and a new scale, the PVQ-RR, has been developed to measure them.
In this session we welcome presentations on continuing work on basic human values as postulated by Schwartz, using the ESS and other data sources. Possible presentation topics may include (but are not limited to):
(1) the measurement of human values in various languages and cultures;
(2) values as predictors of attitudes, opinions and behaviour;
(3) values as consequences of various variables such as sociodemographic characteristics;
(4) value change and development among children, adolescents and adults, using various methods of data analysis for such data such, like latent growth curve modelling (LGM), mixture LGM, change scores, or autoregressive models, just to name a few;
(5) relations between different types of human values measurements (such as the PVQ-57, the PVQ-40 and the picture-based measures);
(6) multilevel and multigroup structural equation models using human values as individual and contextual predictors.
Both substantive and methodological papers using cross-sectional, cross-cultural or longitudinal datasets of basic human values are welcome.
Keywords: Basic human values; Portrait Value Questionnaire (PVQ); Value measurements; Value change
Professor Johann Bacher (University of Linz)
Dr Jacques De Wet (University of Cape Town) - Presenting Author
Ms Daniela Wetzelhütter (University for Applied Sciences Upper Austria )
Schwartz in his famous Theory of Basic Values follows Parsons and Rokeach in arguing that human values are trans-situational or context free. For any individual, the same personal value priorities exist across different life contexts such as the workplace, the school or the home. This assumption influenced the design of Schwartz’s Portrait Values Questionnaire (PVQ), which is widely used in the measurement of values. There is a tendency in the literature on values measurement to take this assumption for granted, but some cross-cultural research questions it. In our quest to improve validity and reliability in the measurement of values we used a quasi-experimental two-wave panel study to test Schwartz’s assumption, in the design of his PVQ, that values are trans-situational. Data was collected from Sociology classes at two universities: one in Austria (n=52) and the other in South Africa (n=61). In the first wave the respondents completed Schwartz’s context-free version of the PVQ, and thereafter they completed a second PVQ with their family/home context in mind. In the second wave, two weeks later, the respondents completed the PVQ with the university context in mind. We used various statistical methods in our analysis of the data including a modified Cronbach’s alpha, the Student t-test, Wilcoxon Signed-Rank test, Stouffer’s z-test and Multi-Dimensional Scaling. Our overall findings support a scenario where respondents have a universal set of values, but the way they prioritise their personal values is somewhat influenced by the value priorities associated with the life context they are thinking about.
Dr Maksim Rudnev (ISCTE-IUL) - Presenting Author
Ipsatization, or within-individual centering, is routinely applied to measures of basic values in order to clear out response styles and to modify absolute importance of values into implied preferences. However, it is rarely discussed what consequences it may bring and, otherwise, what would happen if one does not ipsatize value scores. We explicate the methodological and conceptual issues arising with ipsatization. We show that higher number of items, higher true variances, and lower indicator residuals are better replicated both with ipsatized and non-ipsatized scores. Expectedly, ipsatized scores are resistant to presence of the common factor, even when its variance is very high, whereas non-ipsatized scores deteriorate proportionally to a variance of the common factor, reaching zero when its variance is 10 times higher than variance of ‘substantive’ factors. The latter condition does not improve for non-ipsatized scores even when a sufficient number of indicators is available. In a simulation study manipulating correlations between factors, we found that ipsatized scores in general outperform non-ipsatized scores. Non-ipsatized scores show higher replication rates only when all the factors are positively interrelated, and especially so when they also correlate positively with the common factor. Severe bias arises when method factor is negatively related to substantive factors: in this case non-ipsatized scores show negative correlations with the true values, whereas ipsatized scores are robust to relations between factors unless there is a high positive collinearity between substantive factors. We list cases when ipsatization may increase bias, and conclude that ipsatization is in general preferable for computing unbiased value indices.
Professor Johann Bacher (University of Linz) - Presenting Author
Frequently, MDS is used to analyze Schwartz’s PVQ. A two-dimensional representation is expected. Stress coefficients and related measures are used in order to decide whether to accept the two-dimensions for further analysis. Thresholds for these coefficients are published in the literature. However, they provide only rough approximations because they ignore factors such as the number of objects (i.e. the items in the questionnaire), the dimensionality, the error in the data, the number of ties in the distance matrix and the number of missing data (Borg and Groenen 2005, p. 54-55).
Therefore, Borg and Groenen (2005) proposed applying simulation studies to compute precise threshold values. The procedure is similar to parallel analysis in factor analysis. A certain number of random data sets, e.g. 1000, is generated and MDS is applied to these random data sets. The result is an empirical random probability distribution with an expected mean and standard deviation. This distribution can be applied to test whether the empirical coefficient departs significantly from this random configuration or not.
The presentation describes the procedure and how it can be applied within IBM-SPSS. Results of simulation studies are reported, too.
Borg, I., Groenen, P.J.F. (2005) Modern Multidimensional Scaling: Theory and Applications. Springer, Heidelberg.