24.10.2014 Views

Asking Questions - The Definitive Guide To Questionnaire Design ...

Asking Questions - The Definitive Guide To Questionnaire Design ...

Asking Questions - The Definitive Guide To Questionnaire Design ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

90 ASKING QUESTIONS<br />

from 55 percent to 60 percent of voting-age citizens. In the end,<br />

though, this is only one attempt in a line of attempts to garner a<br />

more accurate response from the survey.<br />

Traffic Violations<br />

Four Examples of <strong>Questions</strong> on<br />

Socially Undesirable Behavior<br />

Traffic violations range in threat from relatively minor violations,<br />

such as a parking ticket, to much more serious violations, such as<br />

driving under the influence of alcohol. In a methodological study<br />

we conducted in Chicago, two separate strategies were used in an<br />

attempt to improve reporting. First, a series of questions was asked<br />

about various traffic violations so that respondents would not know<br />

that driving under the influence of alcohol was the topic of primary<br />

research interest. Second, randomized response procedures were<br />

used. Figure 3.6 shows how the interviewer presented these procedures.<br />

(<strong>The</strong>se procedures were also used on other topics in the questionnaire.)<br />

Using randomized response makes it possible to estimate<br />

the proportion of the population engaging in certain threatening<br />

behaviors, but it does not help determine the isolated behavior of a<br />

single, specific respondent. (<strong>The</strong> procedure is explained in greater<br />

detail later in the chapter.)<br />

Randomized response is one method for assuring respondents<br />

that the interviewer will not know what answer is given. It gives<br />

them a feeling of anonymity. Another method is to use selfadministered<br />

questionnaires. Although increasing the anonymity<br />

of response generally reduces reporting errors, such errors are<br />

not eliminated entirely. (See Figure 3.6.) For very threatening<br />

socially undesirable behavior such as drunk driving, 35 percent<br />

of a sample of respondents chosen from traffic court records<br />

denied being charged with drunken driving even with randomized<br />

response procedures. Still, this was a lower response error<br />

than the 50 percent underreporting that was found using standard<br />

procedures.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!