Sources of bias in the recall of self-generated data: The role of anchoring

This study examines two issues. First, we assess the reliability of employment data self-generated by respondents using a survey panel of 280 firms surveyed in 1999 and again in 2002. Our results show that there is a close correlation between self-generated and archival secondary data. Second, we test for bias in the recall of previous years' employment levels and the sources of such bias. We assess three reasons for recall bias: respondent-level factors; firm-level factors; and anchoring. Our regression results indicate that: owners and new respondents; those in firms that changed legal status and/or location; those in smaller firms; and those in growing or declining firms were biased in their recall of prior employment. In addition, growing firms tended to underestimate their growth, which points to anchoring as an explanation for these biases. These biases imply that we have to be careful in delineating performance measures and cautious about how we interpret self-generated managerial information.