married women being less likely to identify as Democrats (Plutzer and McBurnett 1991). The
variable married is dichotomous with 1 for women who are married and 0 for non-married
women.
Finally, the models include the demographic controls, age, education, income and U.S.-
born. Age is a continuous variable ranging from 18 years to 89 years. Education is a categorical
variable with 0 for “No schooling or Grade 1-8,” (1) for “High school incomplete,” (2) “High
school graduate,” (3) “business vocational,” (4) “some college,” (5) for “college graduate,” and
(6) for “post graduate/professional degree.” Income is a dichotomous variable, with 1 for
persons whose annual household incomes are above $50,000, and 0 for incomes below $50,000.
Finally, because Latinos have a large foreign-born segment, we include the variable U.S.-born, a
dichotomous variable with 1 for persons born in the United States and 0 for persons born outside
the U.S. In the case of Puerto Ricans, U.S.-born captures Puerto Ricans born in the continental
United States.
The multivariate estimations are presented in Tables 2, 3, and 4. Each table has
multivariate estimations for two questions separated by the national-origin groups. Hence in
each table there are six multivariate models. Logistic regressions are used when the dependent
variable is dichotomous (Questions 1 and 5), and ordered-logistic regressions are used and when
the dependent variables are coded on an ordinal scale (Questions 2, 3, 4, and 6). Since our
concern is with the sign and significance of the effect of the exogenous variables and not the
magnitude, we do not report the predicted probabilities.
(Table 2 about here)
The estimations for the first question show that even after the introduction of the selected
controls, the gender differences among Cubans and Puerto Ricans remain significant. Both
Cuban and Puerto Rican women were more likely to favor greater federal government
9
Convention