The estimates of group-based voting behavior used in this chapter are
similar to the results from the previous chapter. We present them in this
appendix for the benefit of those readers who wish to follow the mod-
elling procedures more closely. In the first two columns of Table 7.A1,
we present the coefficients and standard errors for our preferred model
of group-specific political alignments in presidential elections from
1960 through 1992. We take the preferred model from the previous
chapter as our point of departure, and adopt Stanley and Niemi's
procedure for calculating these predicted probabilities. ²⁷ That is, we
calculate the expected probability for each individual respondent in
the pooled NES dataset; the predicted group-specific probability for a
given election is then the average of the predicted probabilities of all
(individual) group members in that year. Note also that the first step of
these calculations takes into account not only the main effects of year
and group membership, but also their interaction (if these are present in
the model) representing changes in that groups' political alignment. In
the remaining columns of Table 7.A1, we thus present the predicted
probabilities of favoring the Democratic candidate for each election.

Our interest in this chapter lies in using the predicted probabilities to
calculate the (margin-dependant) impact of social groups on party
coalitions, and we use the nine columns of predicted probabilities for
the subsequent calculations presented in the text of the chapter itself.

-198-