Interviewer versus automation in preelection telephone surveys: evidence from South Korea's 2024 National Assembly Election

초록

This study investigates intramodal variation within telephone surveys by comparing computer-assisted telephone interviewing (CATI) and automated response system (ARS), which differ in the presence or absence of a human interviewer, using preelection polling data from 6,021 respondents across 11 electoral districts in and around Seoul ahead of South Korea's 2024 National Assembly Election. Despite using identical questionnaires and sampling frames, notable differences emerged between the two modes. ARS produced relatively lower response rates, and its respondents were disproportionately older and male. Linear probability model analyses revealed that ARS respondents were more likely to report active past voting experience and voting intention even after controlling for demographics, suggesting that self-selection disproportionately attracts politically engaged respondents to ARS. At the same time, CATI respondents were substantially less likely to express a preferred candidate or party even after controlling for political engagement, pointing to social desirability bias as a potential mechanism suppressing political preference disclosure in interviewer-administered surveys. These findings highlight that intramodal variation within telephone surveys represents a nontrivial source of bias, with interviewer presence shaping survey outcomes through distinct pathways of nonresponse and measurement error.

키워드