Can survey design reduce anchoring bias in recall data? Evidence from smallholder farmers in Malawi

Abstract

Recall biases in retrospective self-reported survey data have important implications for empirical research. We leverage the survey design literature and test three strategies to attenuate mental anchoring in retrospective data collection: question ordering, retrieval cues and aggregate (community) anchoring. We focus on maize production and happiness reports among smallholder farmers in Malawi. Asking for retrospective before concurrent data on average reduces recall bias (i.e. the deviation of the recalled versus the concurrent outcome reported in the previous period) by 34 per cent for maize production, a meaningful improvement with no increase in data collection costs. Retrieval cues are less successful and community anchors can exacerbate the bias. None of the strategies help to ease the recall bias for happiness reports.