19
Before turning to the estimation of the models, I will briefly discuss some of the
demographic characteristics of the respondents (see table 1 and figures 1-2). The vast majority
of the respondents identify with the Democratic party. Females make up 54% of the sample and
over 50% of the respondents are under the age of 40. The mode income category is $30,001-
$40,000. Individuals who only spoke English at home accounted for 48% of the sample, while
monolingual Spanish speakers accounted for 28% and the rest indicated that they were bilingual.
Two-thirds of the respondents were born in the United States.
[Insert Table 1]
[Insert Figure 1]
[Insert Figure 2]
Results
Latino Context-Free Model
For the estimation of all of the models I ran binary logistic regression. For the first set of
models (each set entails a baseline model and a model including the set of ethnic identification
measures), the dependent variable is a vote for the Democratic candidate.
12
Since I converted
nation of origin, partisanship, language, and gender into dummy variables I use non-Mexicans,
Independents, English speakers, and females as the baseline.
This first set of models is an analysis of Latino vote choice absent any Latino context,
where the choice is between a Republican and a Democratic candidate.
13
In other words, no
information on the candidate’s racial and or ethnic background is provided. I first run a baseline
12
I dropped those who indicated that they would vote for someone aside from Democrats and Republicans, given
that they made up only 3%.
13
I recognize that there could be some contextual influence if there is a Latino candidate running in one’s district.
However, given that the survey was taken in February of 2000 and the question made no reference to one’s own
district I define this as a Latino context-free case. Nevertheless, in the next stage of my work I will match the
respondent’s district with the candidates of that campaign.
Convention