Collaborative Research: Motivated Underreporting

本文研究了测量误差的三种形式，并探讨了它们之间的关系。第一种形式的测量误差影响设计用于确定符合特定调查条件的人口成员（例如，65岁以上的人）的问题。几项研究发现，在筛选访谈中，合格人群的成员被低估了。虽然没有一项调查能完全覆盖其目标人群，但针对特定亚人群的调查似乎特别容易掩盖该特定人群。测量误差的第二种形式涉及过滤问题。这些问题，取决于他们如何回答，要么导致额外的后续问题，要么导致被调查者跳过后续项目。许多调查研究人员认为，为了避免后续问题，受访者可能会对过滤问题给出错误的答案。因此，许多调查在问卷开始时询问过滤问题，然后管理后续问题，而不是将过滤问题和后续问题交织在一起。测量误差的最后一种形式涉及到条件作用或样本时间效应。在过去的四十年中，许多调查研究人员认为，在持续的小组调查中，受访者在小组调查的各个阶段和日记调查的各个时期报告的相关事件较少。这三种现象似乎有一个共同点，那就是为了减少完成问卷所需的努力而少报。但目前尚不清楚这些形式的错误是由采访者的行为造成的，还是被调查者的行为造成的，还是两者兼而有之。拟议的研究使用新的实验和分析现有的数据，试图找出这些影响的轨迹（采访者与受访者），并探索不同方法的有效性，以减少这些错误。该项目将有助于改进从受这些问题影响的调查项目中得出的各种国家统计数字。该项目还将进一步培训研究生，并为两个机构的调查研究人员的专业培训作出贡献。作为支持调查和统计方法研究的联合活动的一部分，该研究得到了方法、测量和统计计划和联邦统计机构联盟的支持。

This research examines three forms of survey measurement error and investigates the relations among them. The first form of measurement error affects questions designed to identify members of the population eligible for a given survey (for example, persons over 65 years old). Several studies find that members of the eligible population are underreported in screening interviews. Although no survey perfectly covers its target population, surveys aimed at specific subpopulations seem especially prone to undercover that particular population. The second form of measurement error involves filter questions. These are questions that, depending on how they are answered, either lead to additional follow-up questions or to the respondent's skipping out of the follow-up items. Many survey researchers believe that respondents are likely to give false answers to the filter questions in order to avoid the follow-up questions. As a result, many surveys ask the filter questions at the beginning of the questionnaire and administer the follow-up questions later on rather than interleaving the filter and follow-up questions. The final form of measurement error involves conditioning, or time-in-sample, effects. Over the last forty years, many survey researchers have suggested that respondents in ongoing panel surveys report fewer relevant events across waves of the panel survey and across time periods in a diary survey.What the three phenomena appear to have in common is underreporting motivated by the desire to reduce the effort needed to complete the questionnaire. But it is not clear whether these forms of error result from something the interviewers do, something the respondents do, or both. The proposed studies use both new experiments and analyze existing data to try to pinpoint the locus of these effects (interviewers versus respondents) and to explore the effectiveness of different methods for reducing these errors. The project will contribute to the improvement of various national statistics that are derived from survey items affected by these problems. The project also will further the training of graduate students and contribute to the professional training of survey researchers at both institutions. The research is supported by the Methodology, Measurement, and Statistics Program and a consortium of federal statistical agencies as part of a joint activity to support research on survey and statistical methodology.