Survey Question Order and the Political Preferences of College Students and their Parents

1
Elizabeth Gutting Survey Question Order and the
Political Party Preferences of College Students
and Their Parents, Daniel J. Willick and Richard
K. Ashley. Public Opinion Quarterly, Volume 35,
Issue 2 (Summer 1971), pages 189-199.
Survey Question Order and the Political
Preferences of College Students and their Parents
What could make up for the difference between the
2 surveys results? -The phone interviewees were
all from an intro sociology class, and were not
randomly selected while walking around the
campus. They all definitely had something in
common with each other (having taken
sociology). -Only 281 students were interviewed
on the phone, whereas 400 were approached on the
campus with a written survey. -The phone
interviewees did not have a sheet of paper to
look at and thus had less time to consider their
answers (they were answering on the spot).
Perhaps because they did not have the question in
front of them, they could not tell as easily as
the first survey group that the survey was most
likely being given to determine if their
political opinions aligned with their parents.
Therefore, they might have been less likely to
change their answers so that it did not seem like
they were merely following their parents views.
The data being examined is from a survey from
1968 which asked students their political party
preference and their parents preferences. 400
students were approached on the University of
California at Santa Barbara campus and asked to
fill out a quick survey. For 200 students, the
survey asked their political party preference
first. For the other half of the students, their
parents preference was asked first. The data
collected was examined to see if the question
order affected the students responses. A second
study was done to look at the question order.
However, this study was done over the phone and
the students who were called were not randomly
selected. Their names were gathered from a
sociology class questionnaire that was
distributed 3 months prior to the phone
interview. The last question on this
questionnaire asked the students to state their
political party preferencehowever, no data was
collected from the sociology survey, it was used
merely to gather names. The same procedure was
used with these students139 of the students
called were asked their parents political party
preference before their own, and 142 were asked
their political party preferences before their
parents.
This graph shows the results of the first survey.
I compared the values from the last 2 sections of
this graphthe students responses for the
surveys which asked their political party
preference first, and the students responses
when their parents preferences were asked first.
The graph shows that students were most likely
to say that they were independent when their
parents preference was asked first. But is the
difference between the students answers based on
the question order statistically significant? I
found the confidence interval in order to address
this question
95 C.I. (.130071, .319929) P-value 0.000 These
numbers indicate that the difference between the
students responses based on the question order
is statistically significant.
This graph shows the results of the second
survey, which was conducted over the phone. Once
again, I looked at the last 2 sections, which
display the students political party preference
responses. This graph, like the graph above,
seems to show that students were more likely to
reply that they were a member of the independent
party if they were asked what their parents
political party preference was before they were
asked their own party preference. Once again I
found the confidence interval to find if the
difference was statistically significant
95 C.I.(-.222955, .00601316) P-value.063 Becaus
e the p-value is .063 (and thus well above .05),
the difference between the students responses on
the phone based on question order is not
statistically significant.
What do Kenyon students report in 2003 for their
parents and their own political party
preferences? Because this survey is from 1968, I
decided to compare it to a survey of my own for
Kenyon students today. Instead of examining if
students said they were independent depending on
the order of the questions, I only looked for a
difference in opinion between students and
parents (because my survey size was so much
smaller). I used the same written format that the
surveyors used in 1968. I handed out 40 surveys
which asked for the students political party
preference first (out of 4 options democrat,
republican, independent, and other), and 40
surveys which asked for the students parents
political party preferences first (and then
theirs). I asked people in Peirce and in the
library to quickly fill out the survey. I found
that in the survey in which students were asked
their preference first, 50 answered that their
views were the same as their parents. The survey
which asked for parents political party
preference first reported that 80 of students
have the same view as their parents. 35 of
students from the first survey reported differing
views from both of their parents, and only .5
reported a difference from their parents in the
survey asking for their parents political views
first. I found the confidence intervals and
p-values to determine if the difference between
the 2 different question orders was significant.
The 95 C.I. for the students with opinions
differing from both parents(-.498431, -.101569)
with a p-value of .003. The 95 C.I. for students
with the same view as their parents(.137488,
.462512) with a p-value of 0.000. Thus, the
difference between the question order was shown
to be significant. Based on the p-values, it
could be said that more people answered with the
same political views as their parents when they
were asked their political party preference after
stating their parents. These results are the
opposite of the results of the survey from 1968,
which suggests that students were more likely to
respond with a political party different than
their parents (namely the independent party)
when they were asked their views last.
Lurking Variables There are many holes that must
be considered in my study, however. The studies
done in 1968 considered the opinions of many more
students (200 for each question order)80
students from Kenyon is only a small sample size
of the school. In addition to this, the study
looks at something a little different than what I
originally tested (the likelihood of a student
saying that they were independent depending on
the order of the survey questions). Nonetheless,
I wanted to include a more recent study that
applied to Kenyon and that might serve as a more
modern comparison. Is it possible that my survey
suggests that students of today are less likely
to rebel against their parents political party
preferences?