Increasing nonresponse error by reducing nonresponse rates? Investigating the biasing effect of methods and procedures aimed at increasing response rates in Web-based access panel surveys

基于网络的访问小组调查已经成为一种很有前途的数据收集方法。然而，人们对参与和不参与这种在线调查的人之间的差异了解得还不够。虽然对提高回复率的策略进行了越来越多的研究，但对这些策略可能对样本组成产生不同影响的方式知之甚少。在本研究中，我们将重点放在价值观、人格特质和公民责任取向的差异上，这些变量在社会学、心理学和政治学的调查中经常被研究。我们将在ALLBUS 2008中收集有关这些特征的基线数据。有互联网接入的受访者将被邀请加入由SPP协调员操作的基于网络的接入小组，并参加随后的两次在线调查。在第一个实验中，我们将尝试四种版本的预先通知（通过邮政邮件、短信、电子邮件或没有）和三种版本的反馈（自我导向、他人导向或没有）。在第二次在线调查中，我们将把四种预先通知条件与五种形式的奖励（即时反馈的彩票、预付奖励、承诺奖励、捐赠给慈善机构或没有）结合起来。我们期望这些策略对不同特征的被调查者产生不同的影响。研究结果将帮助使用在线调查的研究人员认识到提高回复率的策略组合，这些策略可能会鼓励（或阻碍）具有特定特征和价值观的受访者，从而使调查估计有偏差。

Web-based access panel surveys have emerged as a promising data collection methodology. However, not enough is known about differences between people who do and who do not participate in such online surveys. While there is a growing body of research on strategies for increasing response rates, little is known about the ways in which sample composition may be differentially affected by those strategies.In this proposed research, we focus on differences in values, personality traits, and civic duty orientation - variables that are frequently studied through surveys in sociology, psychology, and political sciences. We will collect baseline data about these characteristics within the ALLBUS 2008. The respondents who have Internet access will be invited to join a Web-based access panel to be operated by the SPP coordinator(s) and to participate in two subsequent online surveys. In the first one, we will experiment with four versions of prenotification (by postal mail, SMS, email, or none) and three versions of feedback (self-oriented, other-oriented, or none). In the second online survey, we will combine the four prenotification conditions with five versions of incentives (lottery with immediate feedback, prepaid incentives, promised incentives, donation to charity of choice, or none). We expect differential effects of these strategies on the respondents with different characteristics.The results will help researchers using online surveys to recognize the combinations of strategies for increasing response rates that may encourage (or discourage) respondents with particular traits and values, thus biasing survey estimates.