The attitudinal space framework: Embracing the multidimensionality of attitudinal diversity

Abstract

Attitude polarization describes an increasing attitude difference between groups and is increasingly recognized as a multidimensional phenomenon. However, a unified framework to study polarization across multiple dimensions is lacking. We introduce the attitudinal space framework (ASF) to fully quantify attitudinal diversity. We highlight two key measures-attitudinal extremization and attitudinal dispersion-to quantify across- and within-group attitudinal patterns. First, we show that affective polarization in the US electorate is weaker than previously thought based on mean differences alone: in both Democrat and Republican partisans, attitudinal dispersion increased between 1988 and 2008. Second, we examined attitudes toward wolves in Germany. Despite attitude differences between regions with and without wolves, we did not find differences in attitudinal extremization or dispersion, suggesting only weak attitude polarization. These results illustrate how the ASF is applicable to a wide range of social systems and offers an important avenue to understanding societal transformations.

Keywords: Ecology; Nature conservation; Psychology; Social sciences.

Graphical abstract

Figure 1

Traditional vs. novel multidimensional approaches to attitudinal polarization To date, assessments of attitudinal polarization have mostly used unidimensional approaches to evaluate the increasing difference between attitude positions of social groups over time (i.e., attitude averages at T0 and T1, blue and orange tick marks on X axes, Row A), neglecting the extent of attitudinal diversity. In contrast, the multidimensional nature of attitudes is accounted for in the attitudinal space framework (Rows B & C); for simplicity, only 2 dimensions are displayed. Attitudinal extremization (i.e., the average distance to the joint centroid of the entire population) quantifies group differences across multiple attitudinal dimensions and thus, issue alignment (Row B). Attitudinal dispersion (i.e., average distance of each individual to the centroid of its respective social group, Row C) assesses within-group homogeneity. Taken together, the attitudinal space framework allows to differentiate between weak and strong polarization.

Figure 2

Increase in attitudinal dispersion in the American electorate The attitudinal space generated from the ANES dataset illustrates the position of partisans (strong Democrats and leaners in blue, strong Republicans and leaners in red) along two axes of feelings toward Republicans’ and Democrats’ party and candidates (Panel A; see Table S1 in supplemental information for details on questionnaire items d1-9 and r1-8), which explained 31% and 30% of the variance in the dataset, respectively. Traditional, one-dimensional approaches focusing on changes in mean attitude positions along Axis 1 (Panel B) and Axis 2 (Panel C) suggest an increase in affective polarization between the two groups from 1988 to 2008. The multidimensional attitudinal framework, however, reveals contrasting trends in different polarization measures, with a marked increase in attitudinal extremization (panel D) and in attitudinal dispersion (panel E). This together suggests an overall weaker phenomenon of affective polarization than previously assumed. All trends were significant (p-value < 0.05).

Figure 3

Weak polarization in attitudes toward wolves in Germany Attitudinal space based on survey respondents’ attitudes toward wolves in Germany, in regions without (green color) and with wolves (purple color) (Panel A; see Table S2 in supplemental information for details on questionnaire items a1-12). Attitudes significantly differed between regions with and without wolves (p-value < 0.05), with people in wolf regions expressing less positive appreciation (Panel B) and a greater desire to control wolf populations (Panel C). However, we found no significant differences in attitudinal extremization (Panel D) or dispersion (Panel E) between the two contexts, illustrating an overall weak polarization phenomenon. Numbers in brackets represent the percentage of variation explained by each attitudinal dimension.