Description

The study the effect of context questions prior to target questions, researchers conducted a poll involving 1,054 subjects selected randomly from the Chicago phone directory. To include possibly unlisted phones, selected numbers were randomly altered in the last position. This data frame contains the responses to one of the questions asked concerning continuing U.S. aid to the Nicaraguan Contra rebels. Eight different versions of the interview were given, representing all possible combinations of three factors at each of two levels. The experimental factors were Context, Mode and Level.

Context refers to the type of context questions preceding the question about Nicaraguan aid. Some subjects received a context question about Vietnam, designed to elicit reticence about having the U.S. become involved in another foreign war in a third–world country. The other context question was about Cuba, designed to elicit anti–communist sentiments.

Mode refers to whether the target question immediately followed the context question or whether there were other questions scattered in between.

Level refers to two versions of the context question. In the "high" level the question was worded to elicit a higher level of agreement than in the "low" level wording.

Usage

Format

A data frame with 8 observations on the following 5 variables.

Context

Factor referring to the context of the question preceding the target question about U.S. aid to the Nicaraguan Contra rebels

Mode

Factor with levels "not" and "scattered", "scattered" is used if the target question was not asked directly after the context question

Level

Factor with levels "low" and "high", refers to the wording of the question

M

Number of people interviewed

Percent

Percentage in favour of Contra aid

Details

Increasingly, politicians look to public opinion surveys to shape their public stances. Does this represent the ultimate in democracy? Or are seemingly scientific polls being rigged by the manner of questioning? Psychologists believe that opinions—expressed as answers to questions—are usually generated at the time the question is asked. Answers are based on a quick sampling of relevant beliefs held by the subject, rather than a systematic canvas of all such beliefs. Furthermore, this sampling of beliefs tends to overrepresent whatever beliefs happen to be most accessible at the time the question is asked. This aspect of delivering opinions can be abused by the pollster. Here, for example, is one sequence of questions:

(1)

“Do you believe the Bill of Rights protects personal freedom?”

(2)

“Are you in favor of a ban on handguns?”

Here is another:

(1)

“Do you think something should be done to reduce violent crime?”

(2)

“Are you in favor of a ban on handguns?”

The proportion of yes answers to question 2 may be quite different depending on which question 1 is asked first.

Source

Ramsey, F.L. and Schafer, D.W. (2002). The Statistical Sleuth: A Course in Methods of Data Analysis (2nd ed), Duxbury.

References

Tourangeau, R., Rasinski, K.A., Bradburn, N. and D'Andrade, R. (1989). Belief Accessibility and Context Effects in Attitude Measurement, Journal of Experimental Social Psychology 25: 401–421.

Examples