Page 126 - Comparing Political Communication Theories, Cases, and Challenge

P1: JZZ/KAA P2: KaF
0521828317c05.xml CY425/Esser 0521828317 May 22, 2004 11:55

Werner Wirth and Steffen Kolb

in the context constructs. We cannot give any general suggestion as to
where an acceptable limit would be found, because this decision highly
depends on the research question.
(3) Test for construct equivalence.
By now, we have formed a hierarchical scheme of social phenomena
or scientific constructs, with a basic part and a context part. Start-
ing with the basic part,equivalence on the construct level has to be
discussed, tested for, and established. This can be done by pretesting,
focus-group discussions, or at least by analyzing strategies to control
possible sources of bias. For the less important context area,a discussion
of equivalence and a test for plausibility should be sufficient. The cur-
rent theoretical discussion in each country and the empirical “state of
the art” should be taken into account, in order to avoid construct bias in
the optimal way. Possibly some peripheral context constructs can be left
out completely if the expenses for the basic part have already been very
high.
(4) Test for item equivalence.
Whenever the measurement of the construct is to be undertaken et-
icly, the linguistic adaptation of the instrument to each culture must be
tested. Once again, tests for item equivalence can be applied either be-
fore the study (i.e., by pretests or by group discussions with linguists and
communication researchers on culture-specific connotations) or when
analyzing the data (i.e., by calculating the item-total-correlation for ev-
ery item used in an item battery). For content analyses, analogical tests
can be carried out depending on the scale of the variables.
(5) Test for method equivalence.
Usually,externalstatisticaldataonthestructureofacountry’spopulation
is easily available, so that sample bias can easily be avoided. The structure
of different media systems is not too difficult to analyze, either. Stacked
sampling before gathering the data or weighting procedures afterward
can help to establish equivalence on the sample level.
To test instrument equivalence, additional data on culture-specific
response sets (such as social desirability or acquiescence for surveys and
culture-specific coding for content analyses) are necessary. In the case
of surveys, different response rates, as well as the culture-specific ha-
bituation to different survey modes (i.e., mail, telephone, or personal
interviews, CATI, CAPI, etc.), should be taken into account.
Administration equivalence can be tested but can hardly be estab-
lished. For international studies, the variety of possible scientific and

106