Random News

eriinfo

Africa: Are We There Yet? Interview Duration and Respondent Attitudes

eriinfo.com07 mins
Africa: Are We There Yet? Interview Duration and Respondent Attitudes


Key findings

  • Average interview duration increased by 15 minutes from Round 7 to Round 9, while the number of questions asked increased by an average of 17.8 items.
  • The average interview in Round 9 took 33 minutes longer than the 45-minute estimate, with most interviews (96%) lasting beyond this benchmark.
  • In Round 9, interviewers assessed 17% of respondents as impatient based on a quantitative scale.
  • The proportion of impatient respondents per country varied widely, from 5% in Niger and São Tomé and Príncipe to 65% in Mauritius.
  • As with respondent evaluations, the proportion of Round 9 interviews containing interviewer comments varied considerably across countries, from 1% in Morocco to 45% in Cameroon.
  • Results from topic modelling of Round 9 interviewer comments show evidence of increasing negative respondent attitudes in response to longer interview times.
  • Results from Round 7 also reflect negative demeanours observed in respondents, although their concerns seem to centre on the interruption of daily activities caused by the interview rather than its length.

Effective survey design is essential to ensure high-quality data. One critical aspect of this process is managing the overall length of a survey. Research shows that respondents are more likely to complete shorter surveys (Yan, Conrad, Tourangeau, & Couper, 2011), while longer interviews and more complex questions can impose a greater cognitive burden on participants, potentially leading to less accurate responses (Alwin & Beattie, 2016).


Keep up with the latest headlines on WhatsApp | LinkedIn

Respondent fatigue (or survey fatigue) refers to a decline in survey responses’ completeness, quality, or accuracy when participants become tired, bored, or disengaged during an interview (Jeong et al., 2023; Andreadis & Kartsounidou, 2020). These effects are not uniform across all respondents; individual differences in attitudes – such as varying levels of patience – can influence how participants engage with a survey. Impatient respondents may be less attentive and engaged during a long interview, thereby reducing the likelihood of capturing comprehensive or reliable answers – even if they complete the full questionnaire.

Afrobarometer surveys have grown over the years both in the number of questions asked and in their overall duration, with the average interview taking significantly longer than the 45-minute estimate communicated to respondents.

To evaluate whether and how longer interviews compromise data quality as a result of respondent impatience, we must determine effective ways to measure this disposition and to identify potential sources of impatience beyond fixed personality traits. Afrobarometer’s current approach is to capture respondents’ attitudes through interviewers’ observations. While this measure exhibits significant variation in impatience levels across countries, it shows no consistent association with interview duration. These patterns raise questions about the reliability of this approach for assessing respondent attitudes.

This methods note proposes an alternative measurement approach: topic modelling, a natural language processing methodology designed to uncover latent (i.e. hidden) themes within large collections of texts. Using interviewer comments from four Afrobarometer country surveys, we apply topic modelling to explore how these qualitative data capture negative respondent attitudes – including impatience – and how such attitudes are perceived and recorded by interviewers. This method is particularly advantageous because it automatically identifies patterns across unlabelled text, thus significantly reducing the manual effort typically required for qualitative analysis. Moreover, it enables researchers to organise and search documents based on broader thematic content, rather than relying solely on pre-defined keyword matching.