Doctoral Dissertation Research: Overreporting of Socially Desirable Behavior in Surveys

点击翻译按钮获取中文摘要

SES-0824759John DelamaterPhilip BrennerUniversity of Wisconsin-MadisonThis project studies the overreporting of socially desirable behavior in traditional sample surveys. Often ignored or simply assumed away, measurement errors -- like overreporting -- can lead to inaccurate and misleading findings. However, when directly investigated, they can provide researchers with an opportunity to better understand human behavior. Estimating overreporting as the difference between measures from two different survey modes, this study examines the survey artifact using religious service attendance as a sample case. Recent work has challenged the validity of common survey measures of church attendance and raised questions about how US citizens compare to rest of the developed world: are Americans alone in their overreporting of church attendance, or do Canadians and Europeans exaggerate the frequency of these behaviors as well? Using multiple datasets spanning five decades and fourteen countries, this study compares trends in attendance to shed light on these questions. More importantly, it will pursue the cause of differential overreporting between countries in light of three related hypotheses: (1) the nature of traditional survey questions lead to identity priming during the survey interview, which (2) allows overreporting to be explained as a function of the importance of socially desirable identities that (3) explains the variation in overreporting between nations as the difference in religious identity importance from country to country. American religious exceptionalism will then be redefined not in terms of actual behavior, but rather in terms of continuing high levels of religious identity importance. These high levels of importance will be argued to be the cause of exceptionally high levels of American overreporting. This approach also contributes to the development of sociological methodology by: (1) using time diary data in a novel way, thereby highlighting the time diary as a flexible and powerful, although underutilized, resource; and (2) applying matching methods in an innovative way to allow analysis at the individual level, combining estimates from the two modes.Broader impacts. This study's focus on measurement error will illuminate one corner of an oft-ignored but potentially critical problem. Invalid measures can lead not only to faulty statistical inferences, but also poor policy decisions. Therefore, improvements in survey measurements, like those investigated in this study, benefit not only socialscientists, but also policy makers and stakeholders by improving the quality of data being used in decision making. Research findings from this dissertation will be disseminated using multiple venues and formats. Findings will be of potential interest to academic and non-academic audiences across the spectrum of social science disciplines.