Do interviewers manipulate responses to filter questions? : evidence from a multilevel approach

Contemporary survey designs extensively use filtering techniques in personal (face-to-face) as well as telephone interviews. In general, filter questions serve as forkscircumventing follow-up questions that do not apply. While the purpose of filterquestions is to reduce respondent burden (e.g., Bosley, Dashen, & Fox, 1999), inter-viewers can exploit these questions to cut interviews short. Such interviewer manipu-lation could distort the validity of survey data. Analyzing the conditions thatencourage such interviewer behavior is the aim of our study.Much research has explored how the characteristics of the respondent (e.g., Bollinger& David, 2005; Duan, Alegria, Canino, McGuire, & Takeuchi, 2007; Jensen, Watanabe,& Richters, 1999; Kreuter, McCulloch, Presser, & Tourangeau, 2011; Lucas et al., 1999;Yan & Copeland, 2010), the questions (Duan, Alegria, Canino, McGuire, & Takeuchi,2007; Kessler et al., 1998; Kreuter, McCulloch, Presser, & Tourangeau, 2011), as well asthe interviewer (Kennickell, 2000, 2004; Matschinger, Bernert, & Angermeyer, 2005;Olson & Bilgen, 2011; Olson & Peytchev, 2007; Rosen, Murphy, Peytchev, Riley, L Schnell & Kreuter, 2005), and the situational context of interviewing (e.g.,Groves & Kahn, 1979; Groves, 1987; Hochstim, 1967) influence the responses collected.Only a few studies have examined interviewers influence underreporting to filterquestions (Matschinger, Bernert, & Angermeyer, 2005; Schnell & Kreuter, 2000). Anepidemiological mental health survey (Matschinger, Bernert, & Angermeyer, 2005)found that interviewers’ undesirable behavior regarding filter questions led to dis-torted data validity and consequently flawed results. Schnell & Kreuter (2000)