Augmenting Surveys with Paradata, Administrative Data, and Contextual Data.

Abstract

Over the last decades, there has been growing interest in augmenting survey data with alternative data sources, such as paradata, administrative data, and contextual data. Paradata, for instance, refers to data related to the process of collecting survey data during the field period, which are not directly derived from respondents’ answers to survey questions, but rather are a byproduct of the data collection process (Couper 1998; Kreuter 2013). This may include data from call records, keystroke data in computer-administered surveys, interviewer observations, and more. Administrative data refers to externally created process data that are often linked to individual respondent records by matching personal information (Calderwood and Lessof 2009). For instance, social surveys may link respondents’ interview data (conditional on consent) to tax, insurance, voter registration, and other government databases. Finally, contextual data comprises external sources of information that measure various aspects of the respondent’s physical, social, or informational environment (Fortin-Rittberger et al. 2016). This could involve aggregate data on the demographic composition of a respondent’s neighborhood or organizational characteristics of their place of work, as well as data on respondents’ behaviors, environments, and social networks from wearables, sensors, apps, and digital traces from social media or web browsing. Survey researchers are increasingly utilizing these data sources to enhance their substantive and methodological research and address complex research questions that are difficult (or impossible) to answer using survey data alone. Paradata, for instance, are employed in survey production to monitor fieldwork, increase data collection efficiencies, investigate measurement errors, and assess and correct for nonresponse errors (Biemer et al. 2013; Wagner et al. 2012; Yan and Olson 2013). Likewise, linked administrative data are