‘Update Bias’: Manipulating past information based on the existing circumstances

Abstract

Many panel surveys elicit information about past events multiple times. It is, however, unclear whether respondents manipulate their past information and update it according to their current circumstances in the later rounds of the panel. We term such a systematic bias in reporting past information as “update bias” in this study. We systematically test the presence of update bias in panel data by comparing teenage religiosity obtained from adults first in 2019 and later in 2022 in the Netherlands. Respondents who become more (less) religious in 2022 than 2019 are likelier to report a higher (lower) teenage religiosity in 2022. Even when we use data with a narrower gap (2019 and 2020 survey waves), we still obtain similar results. Overall, the analysis provides strong evidence for update bias. We suggest that the theory of cognitive dissonance best explains our findings; individuals manipulate their teenage religiosity to minimize dissonance between the past and current religious state and to obtain a higher satisfaction. Unlike predominant existing literature that argues people modify their current beliefs according to previous anchors, we provide contrary evidence where people manipulate their past beliefs following their current circumstances.