SURVEY RESEARCHagree/disagree question, respondents may respond
in the same way; however, we can capture both
aspects of the opinion by offering more choices
(strongly agree, agree) or with a contingency ques-
tion (agree/disagree and then how strongly do you
hold that opinion).Selective Refusals.In addition to the issue of sat-
isficing, by which respondents pick no or a neutral
response to avoid the effort of answering, some
respondents refuse to answer certain questions.
This often is the case involving a sensitive issue.
Respondents refuse rather than indicate a socially
inappropriate answer.
For example, in 1992 more than one-third of
Americans refused to answer a sensitive question
about racial integration. When many respondents
do not answer a question, the findings may be
misleading if the nonresponding people actually
hold an opinion. For example, if the respondents
who opposed racial integration answered “don’t
know,” the results appeared more favorable to inte-
gration than if all respondents had answered the
question. After adjusting for nonresponses, Berin-
sky (1999) found that the percentage of Americans
who favored racial integration dropped from 49.4
to 34.9 percent. He warned (p. 1225) that “the opin-
ions respondents express in the survey interview are
not necessarily identical to the opinions they con-
struct when coming to grips with a survey question.”
Agree/Disagree, Rankings or Ratings?Survey
researchers who measure values and attitudes have
debated two issues about responses offered.^31
Should a questionnaire item make a statement and
ask respondents whether they agree or disagree with
it, or should it offer respondents specific alterna-
tives? Should the questionnaire include a set of
items and ask respondents to rate them (e.g.,
approve, disapprove), or should it give them a list of
items and force them to rank order them (e.g., from
most favored to least favored)?
Offering respondents explicit alternatives is
best. For example, instead of asking, “Do you agree
or disagree with the statement, ‘Men are better
suited to run the nation?’” ask instead, “Do you
think men are better suited to run the nation, womenare better suited, or both are equally suited?” Less
well-educated respondents tend to agree with a state-
ment. Explicit forced-choice alternatives encourage
thought and avoid the response setbias—the ten-
dency of some respondents to agree.
Survey respondents asked about values often
show little differentiation and their responses pile
up at the extremes. One solution is to use a “rank-
then-rate” procedure. We first ask respondents to
rank values, most to least important. Next, we ask
them to assign each a rating. For example, respon-
dents rank values (e.g., world peace, personal
wealth, family security) in importance. Next they
assign a value, 1 to 10, from extremely important
to not important at all. A respondent may rank the
value of world peace ahead of personal wealth, but
when asked to rate the importance of world peace
or its personal significance, a respondent may give
world peace a 4 but personal wealth an 8.^32
Remember that we must present the alterna-
tives fairly and not offer a reason for respondents to
choose one alternative. For example, if you ask “Do
you support the law for energy conservation or
do you oppose it because the law would be difficult
to enforce?” instead of simply “Do you support or
oppose the law for energy conservation?” you
created a leading question against the energy con-
servation law. This is why we ask respondents to
choose among alternatives by ranking (e.g., please
give me you first choice, second choice, and third
choice) instead of rating items along an imaginary
continuum (e.g., which of these is best). Respon-
dents can rate several items equally high but place
them in a hierarchy if we ask them to rank the items
compared to one another.^33
Attaching numbers to a response scale can
assist respondents and give them a clue for under-
standing. Positive and negative numbers at the
extremes (e.g.,5 to 5) are best when we con-
ceptualize the variables as polar. It is best to use a
series of positive numbers (e.g., 0 to 10) if we con-
ceptualize the variable as a single continuum.
Again, how we do this tells us how we should organ-
ize the question and its answer choices.
Visual presentations, including the use of colors,
symbols, and pictures, can influence respondents’