12
4. VARIABLES AND PROCEDURES
Thirty-six parties were represented in the five Nordic parliaments in 1999 as well as being
sufficiently sampled the same year.
The five surveys' multichotomous party preference
variables were recoded into 36 dummy variables. Within each country, a score of one was
given the preference for a given party, whereas the remaining party preferences were given
the score of zero. These are the dependent variables of the analyses.
In each country the four independent variables are based on 19 items and measure the
political orientations of hierarchy (equality of procedure), egalitarianism (equality of
outcome), individualism (equality of opportunity), and fatalism (no equality possible). Within
each country, the orientations have been prepared for analysis in three steps. First, each of the
19 items was standardized into z-scores.
relationships between items and political orientation (Grendstad, 2001). For each respondent
in the survey, a new mean score across the standardized items was computed. A minimum of
three valid scores across the five items was required for inclusion in the analyses.
new dichotomized variables were constructed using the mean as the cutoff point.
Respondents with negative scores reject the political orientation in question and are given a
score of zero. Respondents with positive scores support the political orientation and are given
a score of one.
For the purpose of the QCA analyses, rejection of a political orientation was denoted
in lowercase while support was denoted in uppercase. Within each country a new variable
14
The response to party preference was obtained by asking 'Which party would you vote for if
there was a general election tomorrow?' The total sample across the five Nordic countries was
4,832 (Grendstad, Jensen, Kristinsson, Sjöberg, and Sundback, 1999).
15
Ie, each item received a mean of zero and a standard deviation of one. The main reason for
this standardisation is that the original items have different means which may stem from
different degrees of difficulty. For instance, if a fatalistic political orientation was expressed
too strongly, a majority of respondents may tend to disagree with it. Had the item been
formulated less strongly, one would obtain a more even distribution of responses. The
standardisation procedure permits one to balance the item and, in turn, to dichotomize it more
accurately. The "Don't know" option is set to missing.
16
One fatalistic item was removed prior to the analysis due to likely contamination: It seems
that whichever party you vote for things go on pretty much the same. The fatalist scales are
therefore based on a maximum of four items (see Grendstad, 2003b).
17
The cost of dichotomization increases when the mean is not used as the cutoff point
(Cohen, 1983).
Convention