Suchergebnisse

As Howard Schuman noted in his presidential address in 1986, survey respondents expect to be asked "ordinary questions," but our questions often make extraordinary demands. The impact of the difficult tasks we give to respondents on their motivation to provide complete and accurate information is an aspect of the infrastructure of measurement in surveys that has received little attention. When questions are particularly difficult for some groups in a target population, the comparability of measurement for all units in a population is reduced. For measurement to be comparable across all groups in a population, questions must be equally accessible—and equally motivating—to all. Some challenges to improving the practice of measurement in surveys arise because—under the constraints of limited time, budget, and staff—we borrow questions, give them authority based on their source, assert that we need to replicate them to make comparisons, and fail to fully examine claims that they have been validated. Although the constraints are real, these practices limit our ability to take full advantage of advances in research about instrument design to improve the quality of measurement in surveys.

A reanalysis of an experiment that asked for the absolute frequency respondents meant when they said that they were bored or excited, "very often," "pretty often," or "not too often" (see SA 28:1/80K4034). Interview data from a national probability sample of households with quotas (N = 1,172) are analyzed to determine whether conclusions about race, sex, education, & age differences vary for absolute & relative frequencies, & whether the absolute frequency implied by a relative frequency differs according to these grouping variables. With these data, examination of relative vs absolute frequencies can affect conclusions about race differences. Furthermore, the average absolute frequency implied by the relative frequency phrase varies by race, education, & age: very often & pretty often mean a higher absolute frequency for white, better educated, & younger respondents than for the comparison groups. 4 Tables, 1 Appendix, 53 References. Modified AA

To analyze a question wording experiment from the 1980 General Social Survey, an experiment was conducted using two versions of a series of items that asked for evaluations of qualities for children (items adapted from those used by Mel Kohn in Class and Conformity [Homewood, Ill: Dorsey Press, 1969]). The standard version of the items used the noun "child" & the pronoun "he"; the experiment version used the noun "child" consistently. Approximately 1,000 Rs in a national probability sample participated. There were no differences between the two wordings in the way the traits were evaluated. The lack of a question wording effect can be attributed partly to the fact that both wordings are biased in favor of Ms. Question wording interacted with R sex in affecting the sex of the child that Rs claimed to be considering when answering the trait items. The best rewording of the items would specify both the age & sex of the child that Rs should consider when making evaluations. 3 Tables. Modified AA.

In recent decades, research about survey questions has emphasized decision-based approaches. Current research focuses on identifying and systematizing characteristics of questions that are key in researchers' decisions. We describe important classes of decisions and question characteristics: topic, type of question (e.g., event or behavior, evaluation or judgment), response dimension (e.g., occurrence, frequency, intensity), conceptualization and operationalization of the target object (e.g., how to label the object being asked about and the response dimension), question structure (e.g., use of a filter question, placement in a battery), question form or response format (e.g., yes–no, selection from ordered categories, choice from a list, discrete value), response categories, question wording, and question implementation. We use the framework of question characteristics to summarize key results in active research areas and provide practical recommendations. Progress depends on recognizing how question characteristics co-occur, using a variety of methods and appropriate models, and implementing study designs with strong criteria.

▪ Abstract  Survey methodologists have drawn on and contributed to research by cognitive psychologists, conversation analysts, and others to lay a foundation for the science of asking questions. Our discussion of this work is structured around the decisions that must be made for two common types of inquiries: questions about events or behaviors and questions that ask for evaluations or attitudes. The issues we review for behaviors include definitions, reference periods, response dimensions, and response categories. The issues we review for attitudes include bipolar versus unipolar scales, number of categories, category labels, don't know filters, and acquiescence. We also review procedures for question testing and evaluation.

Abstract
Interviewers' ratings of survey respondents' health (IRH) are a promising measure of health to include in surveys as a complementary measure to self-rated health. However, our understanding of the factors contributing to IRH remains incomplete. This is the first study to examine whether and how it matters when in the interview interviewers evaluate respondents' health in a face-to-face survey, in an experiment embedded in the UK Innovation Panel Study. We find that interviewers are more likely to rate the respondent's health as "excellent" when IRH is rated at the end of the interview compared to the beginning. Drawing from the continuum model of impression formation, we examined whether associations between IRH and relevant covariates vary depending on placement in interview. We find that across several characteristics of interviewers and respondents, only the number of interviews completed by interviewers varies by IRH assessment location in its effect on IRH. We also find evidence that interviewer variance is lower when IRH is assessed prior to compared to after the interview. Finally, the location of IRH assessment does not impact the concurrent or predictive validity of IRH. Overall, the results suggest that in a general population study with some health questions, there may be benefits to having interviewers rate respondents' health at the beginning of the interview (rather than at the end as in prior research) in terms of lower interviewer variance, particularly in the absence of interviewer training that mitigates the impact of within-study experience on IRH assessments.

National interview data obtained in 1994 from 646 respondents are used to test two hypotheses regarding public perspectives on federal data sharing: (1) For individuals who are informed about data-sharing processes, confidence in the US Census Bureau's claim of confidentiality is more important than trust in government as a predictor of compliance to share data. (2) For uninformed individuals, trust in government is the more significant predictor. Although the hypotheses were only partially confirmed by the results, it was found that respondent compliance with federal data sharing predicted willingness to give Social Security numbers to private businesses; further, this willingness was not significantly related to general attitudes toward government, supporting the notion that the latter dictate data-sharing willingness. 1 Table, 1 Appendix, 5 References. Adapted from the source document.