Responsive Design for Random Digit Dial Surveys Using Auxiliary Survey Process Data and Contextual Data

这项研究开发了一种定制的随机数字拨号(RDD)电话调查设计的框架，该设计响应于无回答偏差研究的结果。非受访者缺乏数据是RDD无响应研究中的一个主要挑战，通过使用两种类型的数据(无论响应行为如何)可以克服这一问题-ParaData和上下文数据。ParaData是调查过程本身创建的数据，包括对每个抽样号码进行的所有呼叫的历史记录(例如，拨打的电话数量、呼叫日期和时间)以及每个号码使用的调查设计特征(例如，预付款、金钱奖励、拒绝转换)。背景数据来自外部来源(例如，十年一次的人口普查数据)，并通过将所有抽样电话号码的地理识别符(例如，人口普查区域、邮政编码)与相应地理级别的外部数据相关联来创建。这些数据包括相应地点的各种特征，如人口统计和社会经济，这些特征被假定为与居住在该地点的个人特征大致相同。本研究认为调查反应行为是一个随机过程，同时受到样本特征、调查特征、情境以及这些因素的重要性感知的影响。响应行为使用ParaData中的变量和上下文数据及其交互进行建模。拟合的模型用于预测响应行为如何在给定假设的呼叫时间表和调查设计特征的情况下发生变化，从而允许为任何后续调查量身定做设计。对偏差指标的估计可以分别针对受访者和非受访者进行计算。通过比较这些估计，可以诊断无反应偏差的程度。调查数据是有关人口的可量化信息的重要来源，被政府机构、政策制定者以及社会、政治和卫生科学研究人员广泛使用。尽管回复率不断下降，但鉴于这些调查的受欢迎程度，推进目前的区域发展计划调查做法很重要。这项研究将提供一个新的设计框架，对无反应采取全面的方法，并纳入对无反应的理解。设计裁剪的主要元素是，它旨在提高响应率，减少潜在的无响应偏差。这项研究的结果将帮助进行RDD调查的组织改进他们的设计过程。作为支持调查和统计方法研究的联合活动的一部分，这项研究得到了方法学、测量和统计计划和一个联邦统计机构联盟的支持。

This study develops a framework for a tailored random digit dial (RDD) telephone survey design that responds to findings from nonresponse bias studies. The lack of data for nonrespondents, a major challenge in RDD nonresponse studies, is overcome by using two types of data available regardless of response behavior - paradata and contextual data. Paradata are data created by the survey process itself and include the history of all calls made to each sampled number (e.g., number of calls placed, calling dates and times) and the survey design features used for each number (e.g., advance letter, monetary incentives, refusal conversion). Contextual data come from external sources (e.g., decennial census data) and are created by linking the geographic identifier (e.g., census tract, ZIP code) of all sampled telephone numbers to the external data at the corresponding geographic level. These data include various characteristics of the corresponding location, such as demographics and socio-economics, which are assumed to approximate the characteristics of individuals residing in that location. The study regards survey response behaviors as a stochastic process influenced simultaneously by the traits of the sample, the survey features, situational circumstances, and the perceived importance of these factors. Response behavior is modeled with variables in the paradata and contextual data and their interactions. The fitted model is used to predict how response behaviors change given hypothetical calling schedules and survey design features, allowing design tailoring for any subsequent surveys. An estimate of a bias indicator can be calculated for respondents and nonrespondents separately. The magnitude of nonresponse bias can be diagnosed by comparing these estimates.Survey data are a vital source for quantifiable information about the population and are widely used by government agencies, policy makers, and social, political, and health science researchers. Advancing the current RDD survey practice is important given the popularity of these surveys in spite of ever-decreasing response rates. This study will provide a new design framework that takes a holistic approach to nonresponse and incorporates the understandings of nonresponse. The major element of the design tailoring is that it aims to increase response rates and decrease potential nonresponse bias. The results of this study will help organizations conducting RDD surveys improve their design process. 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.