Text Messaging as a Survey Interviewing Mode: A Deeper Look

Abstract Text message communication is nearly universal—at least in many places—but its use for collecting survey data is still emerging. Respondents in text message interviews have been shown to satisfice less than respondents in telephone (voice) interviews, and when given a choice, opt to be interviewed by text more than by voice. This study explores how the properties of text message communication work together to produce these and other advantages for text message surveys: texting is asynchronous, allowing respondents to invest as much time as needed to answer thoughtfully; text messages are persistent, that is, remain available until it is convenient to respond; and text messages are noticeable, that is, various notifications increase respondents’ awareness they have been texted. The data in the current study were collected in two text and two voice interview modes, distinguished by whether a human or automated agent asked the questions. A higher proportion of invited sample members started text than voice interviews, reflecting the noticeability and persistence of the invitations. Text interviews took longer to complete than voice interviews, and text interview response times (RTs) were more variable, suggesting that respondents were able to invest more thought in their answers as warranted by particular questions. We attribute the reduced satisficing previously observed in text interviews to asynchrony: as RTs increased, the frequency of rounded numerical answers (one satisficing indicator) decreased; the opposite pattern was observed for voice interviews. Although individual text interviews were longer than corresponding voice interviews, the overall field period was shorter for text. We show this is due to much quicker recruitment for text than voice interviews, which we attribute to the noticeability and persistence of texted invitations. The result is that data from text interviews can be released quickly, suggesting the mode may be well-suited for time-sensitive measurement. The findings make a strong case for considering text interviewing when designing a study.