Fig 1.
Schwartz’s value circumplex, showing the relationship between human values.
Fig 2.
The distribution of each response style segment across (1) all rating categories and (2) all value segments. The size of each response style segment, as a percentage, is given in parentheses after its number in the vertical axis labels. The table on the left shows the response categories used by the respondents to indicate their answers. The darker the cell, the more the respective rating category was chosen. For instance, in response style segment 4 (RS4 in the text), category “2” was chosen for 52% of the items and category “3” for 32.2% of the items. The table on the right shows combinations of response style segments and value segments. It can be seen that response style segments and value segments are not independent; for instance, response styles segments 18 and 20, with many missing answers, are more prevalent in value segment VII.
Fig 3.
Barplots for the value segments I-VI.
Barplots for the first six value segments. The value items are grouped by value (e.g., the two benevolence items, BE1 and BE2, are next to each other), and ordered according to Schwartz’ theory. For ease of interpretation, the bars are coloured green for values considered more important than average and red for those considered less important than average. The horizontal axis represents the projection onto the first dimension of the LC-BML results; the more different from zero, the more/less important the value item is.
Fig 4.
Correspondence analysis for the segments and reported votes.
Correspondence analysis showing the graphical representation of the cross tabulation for the segments and reported votes over the years, using the adjusted poststratification weights applied to the posterior probabilities of Eq (5). The symbols for the segments are faded according to the explained inertia so that darker points fit better. No shading was done for the political parties. For clarity, each panel shows a different year, but the results are from the same single analysis. S2 Fig shows all years in a single plot. The first dimension can be labeled ‘openness-to-change’ and the second dimension ‘universalism’. The closer a party is to a segment, the more likely it is that this segment will vote for the party. For instance, dimension one distinguishes the traditional segment (III) from the other segments; Segment III is more likely to vote for SGP or CU and less likely to vote for GL. The plots also show that the positions of the political parties relative to the segments changes over time.
Table 1.
Analysis of variance for multinomial logit models estimating the parties voted for.
Table 2.
Estimated coefficients for the multinomial logit model fitted to voting behaviour.