Segura and Fraga
19
recall supporters from the strategic among respondents who may have conceivably preferred a
Democratic governor, we re-estimate the model excluding Republicans and limiting the analysis
solely to Democrats and Independents who supported the recall—thereby lowering the N by 2/3.
The results and predicted probabilities are reported in the third and fourth columns of Table 7.
Latino Democrats and Independents are about 9% more likely to have engaged in this
strategic behavior than Democratic or Independent whites. Surprisingly, African-American
Democrats and Independents are almost 16% more likely to engage in this behavior than whites.
Democrats are about 5% more likely than independents to use this strategy, and Liberals about
8% more likely than conservative Democrats or Independents.
In general, we expect that most liberals, Democrats, African Americans and Latinos
would oppose the recall. But a share of them did not, and a curious share of those went on to
support Bustamante for governor. For our purposes, this is clear evidence of strategic behavior,
albeit a bad strategy with—from the point of view of these voters—disastrous results. In
summary, there is modest evidence for the contention that some small share of Liberal
Democrats, African Americans and Latinos voted to get rid of Gray Davis in hopes of placing
Cruz Bustamante in the governorship.
Discussion and Conclusion
Attributing motives on the basis of observed behavior is always a risky undertaking. The
unusual patterns of behavior in the 2003 recall election may well have been the product of an
uncertain and uninformed electorate voting for idiosyncratic reasons. The patterns in the data
suggest otherwise. That is, there are more than a few systematic elements to predicting which
voters cast ballots in the recall and replacement elections that appear, in partisan and ideological
terms, to be inconsistent.
Convention