Suchergebnisse

Benjamin Page's thoughtful critique of my book, The Illusion of Public Opinion, strives to reassure readers that all is well-despite the book's extensive documentation of measurement-error artifacts in numerous public opinion surveys. Page's own careful polling practices are not followed outside of elite academic survey centers. Moreover, even in such well-run surveys, the respondents are often ignorant of the issues being probed. The fact that nonrandom reasons of some sort must be determining on-the-spot survey responses may allow us to call the respondents 'rational' in a loose sense. But the responses still don't represent actual opinions about the specific policy issues being probed-let alone well-informed answers. Finally, while Page argues that measurement errors are most likely random, and should cancel out in the aggregate, the 'miracle of aggregation' doesn't occur when the form, wording, and context of questions varies from survey to survey. And even when wording and context are held constant, the respondents' understanding of the same words can vary dramatically or subtly over time, due to new political events and their changing interpretation. Adapted from the source document.

Critics of Converse's agenda-setting 1964 essay underexplored the seemingly technical issue of measurement error. Down this road not taken lie serious questions about the evidence for both of Converse's main theses. First, a thorough reexamination of the exact questions posed to a mass sample of the electorate and to an elite sample of congressional candidates suggests that the mass/elite difference in ideological constraint reported by Converse could be, in significant part, a measurement-error artifact caused by differences in question form. Furthermore, the vagueness and ambiguity of the questions not only calls into question the plausibility of Converse's main thesis-lack of ideological constraint among the public-but of his subsidiary thesis: that, measured over time, members of the public had such unstable political opinions that they might best be considered 'nonattitudes.'. Adapted from the source document.

SCHOLARS HAVE PREVIOUSLY DEBATED THE COMPARABILITY OF ALTERNATIVE FORMS OF THE QUESTION ABOUT PARTISANSHIP IN POLLS CONDUCTED BY GALLUP AND THE MICHIGAN SURVEY RESEARCH CENTER (SRC). IN THIS EXCHANGE OF VIEWS, GEORGE F. BISHOP, ALFRED J. TUCHFARBER, AND ANDREW E. SMITH CONTRIBUTE TO THIS DEBATE BY REPORTING AND ANALYZING EVIDENCE FROM 15 EXPERIMENTAL SURVEYS IN OHIO IN 1991-93. THEY CONCLUDE THAT THE DISTRIBUTION OF PARTISAN LOYALTIES WILL GENERALLY BE THE SAME WHETHER ONE USES THE GALLUP OR THE SRC QUESTION AND, CONTRARY TO THE FINDINGS OF PAUL R. ABRAMSON AND CHARLES W. OSTROM, THE GALLUP FORM IS NO MORE RESPONSIVE TO SHORT-TERM POLITICAL FORCES THAN ITS SRC COUNTERPART. IN RESPONSE, ABRAMSON AND OSTROM AGREE THAT DURING MANY TIME PERIODS THERE WILL BE LITTLE DIFFERENCE BETWEEN AGGREGATE LEVELS OF MACROPARTISANSHIP REGARDLESS OF WHICH MEASURE IS USED. HOWEVER, THEY ARGUE THAT DURING PERIODS OF POLITICAL VOLATILITY THE GALLUP APPROACH WILL ACCENTUATE DIFFERENCES, WHILE THE SRC VERSION WILL ATTENUATE THEM.

To test whether response effects in public opinion surveys due to response order, question form, wording, or context variations are greatest among respondents (Rs) who are least involved with an issue, meta-analysis was applied to data from 15 split-ballot experiments from telephone surveys conducted in the greater metropolitan area of Cincinnati, Ohio, 1978-1983. Findings support earlier research by Jon A. Krosnick & Howard Schuman ("Attitude Intensity, Importance, and Certainty and Susceptibility to Response Effects," Journal of Personality and Social Psychology, 1988, 54, 940-952) that response effects of counterargument or middle alternative wording in survey questions are larger for less-involved Rs. However, it is also found that issue involvement specifies some response effects, but not others, eg, response & question order are unrelated to involvement. 2 Tables, 15 References. M. Malas

A series of experiments conducted as part of the Greater Cincinnati (Ohio) Survey, a multipurpose random-digit-dialed telephone survey, conducted by the U of Cincinnati's Behavioral Science Laboratory (N not given), explore how question forms affect responses. Findings show that: (1) people are significantly more likely to select the middle-response alternative on an issue when it is explicitly offered to them as part of the question than when it is omitted; & (2) merely mentioning that there is a middle alternative, in the preface to the question, makes it more likely that Rs will select it, even though it is not offered to them as an explicit choice. It is also shown that the order in which the middle alternative is presented -- in the middle or last position -- can make a significant difference in the results, but that a recency bias does not occur invariably or consistently. Finally, people who select a middle-response alternative when it is offered would not necessarily answer the question in the same way as other Rs, if forced to choose sides on the issue. 2 Tables, 1 Appendix, 2 References. Modified AA

It has often been suspected, but seldom researched, that respondents to surveys give opinions on topics to which they have given little or no thought. Telephone survey data from 631 respondents & interview data from 1,218 respondents are used to assess the magnitude of this problem. In the absence of an explicit filter question, about 33.34% of respondents had opinions about a fictitious Public Affairs Act, while filter questions reduced this to 5-10%. People with opinions on other issues are more likely to express opinions on this fictitious issue. Effects of education & race on response rate are identified; age, sex, & income do not have substantive effects. Low interpersonal trust is associated with high response rates & negative responses, while the reverse is true for high interpersonal trust. This pattern is reflected also in responses to real issues. 7 Tables. W. H. Stoddard.