ships. Table 14.4 presents the unrotated matrix for the principal components analysis of the correlation matrix, with only loadings above .45 reported. These four factors account for 71 percent of the total variance among the variables, with the first two factors alone accounting for 55 percent. It thus appears that variables 5.09 and 5.10 ("supranational integration" and "national integration") have little variance in common with the other eleven issues, which--while not demonstrating the unidimensionality expected for a single left-right or liberal- conservative interpretation--do break relatively cleanly into two dimensions that can be interpreted in analogous fashion.
The seven items that load high on factor I bespeak of leftism in the Marxist sense, while the four items that load high on factor II suggest an interpretation of leftism as classical liberalism--in the sense of opposition to authoritarian government. The seven variables linked with factor I were formed into a "Marxism" scale with a reliability of .90. The four linked with factor II were combined into a "liberalism" scale with a reliability of .81. These scales, which are virtually uncorrelated with each other (r = .08), seem to account for most of the cross-national variation among political parties on the issues for which they were scored, and they do so in a readily interpretable manner.
As is usually the case in scale construction, one can "purify" the scale by dropping items that tend to correlate the least with the others, and such is the case with the Marxism scale. The rotated factor matrix of the original thirteen variables, not reported here, indicates that the first four issues--the economic issues--tend to cluster together more closely than the other three ("secularization," "East West alignment," and "anticolonialism"). One can produce a four-item ''economic leftism" scale, with a reliability of .91, that deals only with the parties' positions on the role of the state in the economy--minus the extra ideological baggage contained in the three items eliminated. Obviously, the economic leftism scale, as a subset of the Marxism scale, correlates very highly with it (r = .94). Both scales are thought to be useful to retain- one for focus and the other for breadth.
Regarding the validity of these three scales measuring parties' issue orientations, a special avenue of investigation is present, the determination of predictive or criterion validity. This method holds that validity is established if the proposed measure conforms to some outside criterion the validity of which is either established or presumed. To demonstrate criterion validity, we need to obtain high correlations with another, presumably valid, rating of parties on the dimension.
We were able
to identify two extensive sets of ratings of party
ideologies from sources with offsetting ideological biases
of their own. One source is the U.S. State Department, which
for many years has classified parties as "Communist,"
"Non-Communist Left," "Center," and
"Conservative." The other source is a Soviet
publication that describes parties using a Marxist
vocabulary easily translatable into a "right-center-left"
classification. These "expert" ratings were
translated into scale scores (a four-point scale for the
United States and a three-point scale for the Soviets) that