The current research builds on past evidence to examine the antecedents and consequences of IRS specifically. Available findings suggest that interviewer experience is a significant predictor of overall interview length, such that interviews conducted by experienced interviews are shorter (Olson and Peytchev 2007). This leads to our first hypothesis:
HI: Interviewer experience will be positively associated with IRS.
Interviewer and respondent behavior may also change over the course of an interview. Respondents may become more comfortable with the survey process and interviewers may feel comfortable speeding up in response. Interviewers (and respondents) may become impatient as a long interview progresses and engage in strategies to get through the questions more quickly. For respondents, this may result in behaviors like survey satisficing (Krosnick 1991; Krosnick, et al. 2002). For interviewers, this may be reflected in faster IRS. As a result, we hypothesize that:
H2: IRS will increase as an interview progresses.
Like any conversation, surveys involve turn taking and there are rules and norms that govern turn taking in order to facilitate communication (e.g., Wiemann and Knapp 1975). The survey interview involves fairly frequent hand-offs in turn taking and relatively short periods of talking from both parties, which also establishes conversation-specific norms for turn taking. Talking for a long time without giving the other person a chance to speak may be seen as violating norms (Wiemann and Knapp 1975). Furthermore, interviewers are trying to build rapport with respondents, and talking for too long may undermine this goal (e.g., Leech 2002). As a result, interviewers may be uncomfortable with longer questions, and may want to get through them quickly so that they can give respondents an opportunity to talk. This leads to our next hypothesis:
H3: IRS will be positively associated with question length.
We know that respondents experience discomfort when being asked sensitive questions (Krumpal 2013). We argue that interviewers may be as uncomfortable asking sensitive questions as respondents are answering them and they may therefore read these questions more quickly to minimize the length of their discomfort. This leads us to hypothesize that:
H4: Interviewers will read sensitive questions faster than nonsensitive questions.
Our remaining hypotheses address the potential consequences of interviewer pace. First, we examine whether interviewer pace is associated with response latencies, or the amount of time it takes respondents to answer a question. With regard to interviewer pace, some researchers have argued that interviewers communicate their desired pace of the conversation by the speed at which they speak, and respondents match their speed to the interviewers (e.g., Holbrook, Green, and Krosnick 2003). Consistent with this argument, Fowler (1966) found that faster interviewer pace led respondents to perceive that answering quickly was more important than answering completely or accurately. This argument leads to the following hypothesis:
H5a: Faster IRS will be associated with shorter response latencies.
However, one could also make the counterargument that faster interviewer pace may increase respondents' cognitive burden and reduce the speed with which they are able to answer. Faster IRS may make the task of listening to and comprehending the survey question more difficult, reducing respondents' capacity to start the process of retrieving relevant information from memory or constructing a response. As a result, faster IRS may slow respondents' answering process, leading to the following competing hypothesis:
H5b: Faster IRS will be associated with longer response latencies.
One of the key reasons why researchers have studied interviewer pace is because it may be associated with data quality. There is some evidence from other contexts that a fast speaking speed can lower a listener's comprehension (e.g., Conrad 1989; Griffiths 1990). Following this logic, we hypothesize that:
H6: Faster IRS will be associated with more comprehension difficulties.
Faster IRS could also increase the extent to which respondents show mapping difficulties, at least in part because it affects comprehension. It could do so by affecting respondent ability to properly recall all of the response categories when forming judgments. Therefore, we hypothesize that:
H7: Faster IRS will be associated with more mapping difficulties.
The data we analyzed included questions that varied along a number of dimensions (see Figure A17.1 in the online supplemental materials). We controlled for these characteristics in our analyses predicting IRS. Because question characteristics can influence the response process (see Tourangeau, Rips, and Rasinski 2000), we also included these variables in our analyses exploring the potential consequences of IRS.