and civil unions. Further, I suggested an interaction effect between those in the ‘against
religion’ treatment group and frequent church attenders/born again subjects may further
increase opposition to marriage and unions.
Of the 259 subjects enrolled in the survey-experiment, twenty-four identified
themselves as ‘born again’. While this number is below the requisite thirty for statistical
analysis, I examined differences between those identifying themselves as born again and
those subjects who did not. One-way ANOVA results are presented in table 4 and
suggest that a subject’s born again status is highly significant both to responses regarding
same-sex marriage and those for civil unions. Figures 3 and 4 indicate that the
relationship is in the predicted direction, with born again respondents more exhibiting
higher mean opposition to civil unions and marriage.
Table 4 also shows that question order was significant in determining mean
opposition to same-sex marriage, and nearly significant when interacted with born again
status. While further analysis is necessary to completely understand this relationship,
these findings suggest that while question order may not have an effect at the aggregate
level, disaggregating the sample by religious preference may reveal effects.
Table 5 examines this supposition directly. It compares the between groups
effects of question order, frame, and religious preference on opposition to marriage and
civil unions. It shows that religious preference has a statistically significant effect on
opposition to both marriage and unions but that question order, even when interacted with
religious preference, is not. An important caveat here is that not all religious preference
categories have a large enough number of respondents to be evaluated seriously. The
groups which do have an N of at least 30 are: Protestant, Roman Catholic, and no
religion. Figure 5 is a graph of mean opposition to same-sex marriage by religious
19
Convention