Segura and Fraga
13
Although the above described patterns are suggestive, we now set out to identify, on
multiple dimensions including but extending beyond race, which individual characteristics were
most associated with these two unusual patterns of vote choice.
Data and Analysis
To explore these questions, we employ a statewide exit poll of California voters,
conducted on October 7, 2003, by the Los Angeles Times. The sample frame was all California
voters. The total number of interviews was 5205 voters from 74 precincts. The resulting data
was then weighted to adjust for statewide sampling error, absentee voters, and declines-to-
participate.
9
Defectors—Recall Opponents who Did Not Support Bustamante
The first step in our analysis is to describe those voters—call them “defectors”—who
voted for the recall and then went on to vote for a gubernatorial candidate other than Lt. Gov.
Bustamante. We call these voters defectors because they exhibited a preference on the recall that
could broadly be understood to be “Democratic” in its partisan coloring, but went on to vote for a
non-Democrat in the election. The results of this analysis are presented in Table 4.
[Table 4 about here]
Table 4 reports one-way analyses-of-variance between the dependent variable—voting
for Cruz Bustamante—and an array of potentially important independent variables. This
comparison is made only among voters who voted against the recall. Each of the predictor
variables is self-explanatory, where the value one (1) indicates the presence of the characteristic
9
For more information, you can contact the Los Angeles Times Poll at 202 West 1
st
Street, Los Angeles, CA,
90012, or 213-237-2027. The data set is available from the Roper Center for Public Opinion Research at the
University of Connecticut, and is identified as study “USLAT2003-490.”
Convention