For example, let’s say you were interested in the effect of political persuasions
on moral attitudes and behavior. Perhaps your hypothesis was that the more
conservative one is politically, the more conservative one might be morally. You’ve
operationalized your variables on political persuasion by assuming conservatives are
registered Republican and liberals are registered Democrat and that morally conser-
vative people will disapprove of divorce and be less likely to get a divorce. You decide
to test the hypotheses that because Republicans are less likely to approve of divorce
than Democrats are, then Republicans are less likely to get divorced (attitudes lead
to behavior).
You find that a reputable social scientific researcher had done a survey of a sam-
ple of Americans, but this researcher was interested only in gender and racial dif-
ferences in moral attitudes and behavior. It’s possible that the research contains other
background variables, such as age, political persuasion, educational background,
or occupation. Secondary analysis of the existing data will enable you to answer
your questions. In addition, you might be able to find data on statewide divorce
rates and statewide political attitudes; while these will not answer the question at
the more individual level, they can point to broad patterns about whether conser-
vatives are true to their beliefs and so less likely to divorce. (The answer is appar-
ently no; states that voted Republican in the last two presidential elections have
higher divorce rates than states that voted Democratic, with eleven “red states”
recording higher divorce rates than any “blue state”) (Crary, 1999; Dossier: Red
State Values, 2006).
Also, there may be different forms of data you can use. Sometimes, for example,
researchers will conduct an interviewand use only a numeric scale to register
responses. But then certain answers to certain questions might prompt the interviewer
to ask for more information. These responses may be written down as notes or sen-
tences on the initial interview forms. Going back to these forms might require you to
do content analysis of the narrative responses people gave to the questions.
For example, one of my students was perplexed by an apparent discrepancy in
the research on date and acquaintance rape. The National Crime Victimization Study
(NCVS) found that 25 percent of all college-age women had experiences that met the
TYPES OF SOCIOLOGICAL RESEARCH METHODS 123How to “Read” a Survey
- Four out of five doctors recommend Zytrolvan.
- Forty-three percent of Americans support the pres-
ident’s policy.
We hear statements like these all the time. But what
do they mean?
According to the American Association for Public Opinion
Research, an intelligent analysis of survey results requires that
you know some minimal information:- Who sponsored the survey, and who conducted it?
- What is the population being studied?
- What is the sample selection procedure?
- What is the size of the sample, and the completion rates?
- What is the wording of the questions?
- What are the method, location, and dates of data collection?
- How precise are the findings, including weighting or estimat-
ing procedures and sampling error? - Are some results based on parts of the sample rather than
the whole sample?
Unfortunately, very few of the survey results you hear about
in the mass media (or, for that matter, in many textbooks)
include all of the necessary information. Therefore you cannot
be sure of their accuracy. If the accuracy of the numbers is
important to you, look up the references. If there are no refer-
ences, start to worry.Sociologyand ourWorld