Social Cohesion as a Real-life Phenomenon: Exploring the Validity of the Universalist and Particularist Perspectives

relatively enduring regimes of social cohesion. This is the case with social order,
social hierarchy and value diversity. On other components some groups show
relatively stable scores while others have more volatile patterns. This can be seen for
inequality (stable levels in the English-speaking and Continental European groups but
declining ones in the Nordics and in East Asia), social trust (declining in the English-
speaking group and stable in the other groups), passive participation (stable in the
English-speaking and continental group, but rising sharply in the Nordics), and ethnic
tolerance and political trust (declining sharply in East Asia and stable in the other
groups). Finally, there is one indicator (active participation) on which all four groups
show sharply rising levels over the 20 year period.

The second question stated above was explored with hierarchical cluster
analysis. In this analysis the forming of clusters of cases occurs in a series of stages.
At each stage the two cases (or clusters of cases) that are most similar on all the
variables are clustered. Thus, the further the analysis proceeds the fewer clusters
remain but the larger the differences between the clusters become and the more
heterogeneous the clusters become internally. In other words, in the initial stages it
produces many clusters which are relatively homogenous internally and in the final
stages just a few highly diversified clusters are left over (Cramer 2003). The best
visual representation of this process is a dendrogram, which displays the different
clusters as horizontal lines on the x-axis. The higher the value on this axis, the fewer
the number of clusters and thus the more internally diverse each cluster is. The
branching points on the axis (i.e. the moments when new clusters are formed) offer
good insight into the internal homogeneity of each cluster for a specific cluster
solution (by cluster solution we mean the number of clusters produced at a certain
stage).

Figure 4 shows three such dendrograms, one for each wave.¹ We used the
same countries and indicators for each of these analyses to ensure over time
comparability. Japan, South Korea were excluded because these countries had missing
data for passive and active participation in the first wave. The remaining European
and American countries are thus expected to fall in just three clusters - an English-
speaking one; a Scandinavian one and a Continental European. Focusing on the
dendrogram of the last wave, we can see that the three cluster solution (as shown by
the three horizontal lines) only partially corresponds to the predicted country
groupings. We can indeed discern a distinct Scandinavian cluster composed of
Denmark and Sweden although the Netherlands also forms part of this. Likewise we
see a large continental European continental but this group is heavily „polluted’ with
English-speaking countries (Britain, Ireland and Canada). The US forms a separate
cluster. In short, an English-speaking group of countries cannot at all be identified.
Thus, the geographic dispersion of English-speaking countries could be said to be
emblematic of their disparate social cohesion characters.

Figure 4 about here

A comparison of the three dendrograms over time confirms the relative cohesion of
the Scandinavian group and the absence of an English-speaking cluster. We can see

¹ We used standardized variables (Z-scores) for these cluster analyses as this ensures that each variable
is given equal weight in the analyses.

<<   10   11   12   13   14   15   16   >>

More intriguing information