The Impact of Parenthetical Phrases on Interviewers� and Respondents� Processing of Survey Questions

Abstract

Many surveys contain sets of questions (e.g., batteries), in which the same phrase, such as a reference period or a set of response categories, applies across the set. When formatting questions for interviewer administration, question writers often enclose these repeated phrases in parentheses to signal that interviewers have the option of reading the phrase. Little research, however, examines what impact this practice has on data quality. We explore whether the presence and use of parenthetical statements is associated with indicators of processing problems for both interviewers and respondents, including the interviewer's ability to read the question exactly as worded, and the respondent's ability to answer the question without displaying problems answering (e.g., expressing uncertainty).Data are from questions about physical and mental health from 355 digitally recorded, transcribed, and interaction-coded telephone interviews. We implement a mixed-effects model with crossed random effects and nested and crossed fixed effects. The models also control for some respondent and interviewer characteristics. Findings indicate respondents are less likely to exhibit a problem when parentheticals are read, but reading the parentheticals increase the odds (marginally significant) that interviewers will make a reading error.

The Effects of Respondent and Question Characteristics on Respondent Answering Behaviors in Telephone Interviews

“…Additionally, question characteristics appear as a package (Dykema et al in press), and as such, we cannot disentangle how different question features interact or moderate each other (e.g., attitudinal questions with different types of response options). Despite these limitations, much of the work on respondent-interviewer interaction is conducted in face-to-face surveys and looks at one question at a time (e.g., Suchman and Jordan 1990;Dykema et al 1997), and less is known about telephone surveys, especially those conducted in a contemporary context (but see Dykema et al 2016). While this study cannot assess some question types and question characteristics, it extends our knowledge about respondent-interviewer interaction to contemporary telephone surveys across multiple items.…”

“…Surprisingly, questions where there was a response task or concept mismatch between questions and response options had a higher rate of adequate answers and fewer requests for clarification. From this analysis, we do not know whether interviewers preemptively changed the question wording in order to address these task and conceptual misalignments (similar to an interviewer decision to read parentheticals; Dykema et al 2016). Future research will examine the interviewer-respondent interaction on mismatched questions in more detail.…”

“…Previous studies have demonstrated the joint effects of multiple question characteristics on question reliability and validity (e.g., Andrews 1984;Alwin 2007;Saris and Gallhofer 2007). Yet limited systematic attention has been given to examining the effects of multiple question, respondent, and interviewer characteristics across a full questionnaire on respondent answering behaviors (but see Holbrook, Cho, and Johnson 2006;Dykema, Schaeffer, Garbaski, Nordheim, Banghart, et al 2016;Holbrook, Johnson, Cho, Shavitt, Chavez, et al 2016). Respondent behaviors during interviewer-administered surveys partially reveal mechanisms for creating reliable and valid answers.…”

The Effect of Question Characteristics on Question Reading Behaviors in Telephone Surveys

“…The question features summarized here give interviewers discretion over what to read, increasing interviewer burden (Japec 2008). Parenthetical information in a question may be skipped by interviewers, especially if it seems ancillary (Olson and Smyth 2015;Dykema, Schaeffer, Garbarski, Nordheim, Banghart 2016). Similarly, for many battery items reading the question stem and response options is optional (i.e., those appearing after the first or second item) (Ongena and Dijkstra 2007;Olson, Smyth and Cochran 2018).…”

“…Because interviewers may adapt question reading to respondents who are particularly likely to have problems (Dykema et al 2016), we control for common measures of respondent cognitive ability, age, and education (e.g., Belli, Weiss, and Lepkowski 1999;Cannell, Fowler and Marquis 1968). Age in years is included as a grand-mean-centered continuous variable (mean WLT: 61.34; NOP: 54.95).…”