18
conservative Democrat with a limited amount of political information (“3”), who does
not have a stated policy preference for limiting the political contributions of unions. This
is the archetype respondent for having confusion alter her vote. With these particular
characteristics she has 51% chance of voting ‘yes’ on this proposition. However, even
stacking the deck with this hypothetical respondent, making her confused does not alter
her ultimate choice as her likelihood of voting ‘no’ only increases only .16%.
The variable of interest, confusion, and the other control variables acted nearly
identically in Proposition 74 as they did in Proposition 75. Also, the first differences
results show no real effects changing the hypothetical respondent from “not confused” to
“confused.” This consistency across propositions indicates that this operationalization of
confusion clearly does not lead to more ‘no’ votes in initiative elections. Since the two
models (Proposition 74 and Proposition 75) are so similar, these results are very strong.
Confusion – The Four Different Types
If confusion is looked at in its entirety, there is little support for the conventional
wisdom that confusion leads to more ‘no’ votes in initiative elections. Breaking apart
confusion into its various categorizations presents some interesting relationships, and
suggest that the conventional wisdom may hold in certain circumstances. For both
Proposition 74 and 75, a logit regression was run separating out the four different types of
confusion. The results are presented in Table 5. The dependent variable, vote choice in
the respective propositions, and various controls remain unchanged for each proposition.
The four types of confusion, contradictory consensus cues, competing cues, bad cues and
absent cues are dichotomous variables coded “1” if the respondent is classified in that
manner or “0” otherwise.
Convention