16
software program. The OLS and GLS analyses are ideal given that the change dependent
variables are percentage change from one manifesto to a subsequent manifesto. Second, I run
White’s heteroskedasticity test and a test for multicollinearity on the OLS results. Third, I run a
Hausman test on the GLS random-effects and fixed-effects results to determine which is a more
appropriate model.
For all of the variables, the Hausman test specifies the fixed-effects model as the correct
choice. This test, the fixed-effects model, corrects for any cross-party differences. I am looking
at European political parties as an aggregate, whether or not European political parties as a whole
change their issue salience in response to disputes, and I do not want individual European
political parties to unduly bias the results. This fixed-effects choice allows more accurate
specifications of the processes determining whether party issue saliency is correlated with
disputes, and its results are considered in rejecting or supporting the hypotheses.
Results
The results of my GLS fixed-effects analysis are illustrated in Tables 1-12. Tables 4 and 5
illustrate the results of testing the first hypothesis with table 4 listing the results for the
per104change variable and table 5 the results for the per105change variable. Tables 7, 8, 9, and
10 illustrate the results of testing the second hypothesis with table 7 listing the results for the
per107change, table 8 the results for the per108change variable, table 9 the results for the
per109change variable, and table 10 the results for the per110change variable. Tables 1, 2, 3, 6,
11, and 12 illustrate the results of testing the third hypothesis with table 1 listing the results for
the per101change variable, table 2 the results for the per102change variable, table 3 the results
for the per103change variable, table 6 the results for the per106change variable, table 11 the
results for the per201change variable, and table 12 the results for the per202change variable.
Convention