In this study, we examine whether the use of artificial intelligence (AI) can reduce the effect of independence conflicts on audit firm liability. In two experiments, we manipulate (1) whether procedures are performed by a human auditor or with AI and (2) whether the audit firm was careful in maintaining the appearance of independence from the audit client. Results of both experiments indicate that the use of AI significantly reduces the impact of the appearance of independence conflicts on jurors' judgments of audit firm liability. When concerns relating to the appearance of independence conflicts are present, the use of AI helps maintain the perceived objectivity of the auditor, which results in jurors maintaining higher overall trust in the audit process. Our study contributes to literature on determinants of auditor litigation risk and how technological change that is likely to grow in prominence might affect audit firm liability.

Les études précédentes indiquent que les analystes n’ajustent pas totalement leurs évaluations en fonction du biais général à la baisse des annonces de résultats prévisionnels faites par les directions d’entreprises. Les auteurs rapportent les résultats de deux expériences visant à déterminer comment la feuille de route des entreprises en ce qui a trait à la publication de résultats prévisionnels et les enjeux motivant les analystes expliquent de concert la mesure dans laquelle ces derniers ajustent leurs évaluations en fonction du biais des annonces de résultats prévisionnels. Ces résultats d’expérience semblent indiquer que les analystes pour qui l’enjeu est celui de l’exactitude ajustent leurs évaluations selon la feuille de route de la direction en ce qui a trait à la publication de résultats prévisionnels biaisés à la baisse lorsque le biais est relativement modeste (un cent), mais que les analystes pour qui l’enjeu est celui de leur relation avec la direction s’en abstiennent. Au surplus, la différence dans l’ajustement est plus importante lorsque la feuille de route de l’entreprise en matière de biais est irrégulière que lorsqu’elle est régulière. Aussi, lorsque le biais des annonces de résultats prévisionnels est plus important que moins (deux cents par rapport à un cent), les analystes pour qui l’enjeu est celui de la relation avec la direction procèdent à un ajustement partiel, paraissant ainsi rechercher un équilibre entre exactitude et volonté de plaire à la direction. Ces constatations ont des répercussions pour les investisseurs, les autorités de réglementation et l’interprétation des études précédentes.

Prior research indicates that analysts do not fully adjust for the general downward bias in earnings guidance issued by management. We report the results of two experiments designed to investigate how guidance track record and analysts incentives jointly explain the extent to which analysts adjust for guidance bias. Our results suggest that analysts with accuracy incentives adjust for managements track record of downwardly biased guidance when the bias is relatively small (one cent), but those with relationship incentives do not. Furthermore, the difference in adjustment is larger when the bias track record is inconsistent than when it is consistent. Also, when guidance bias is larger (two cents) relative to smaller (one cent), analysts with relationship incentives partially adjust, as they appear to strike a balance between accuracy and their desire to please management. These findings hold implications for investors, regulators, and the interpretation of prior research.

Prior research indicates that analysts do not fully adjust for the general downward bias in earnings guidance issued by management. We report the results of two experiments designed to investigate how guidance track record and analysts’ incentives jointly explain the extent to which analysts adjust for guidance bias. Our results suggest that analysts with accuracy incentives adjust for management’s track record of downwardly biased guidance when the bias is relatively small (one cent), but those with relationship incentives do not. Furthermore, the difference in adjustment is larger when the bias track record is inconsistent than when it is consistent. Also, when guidance bias is larger (two cents) relative to smaller (one cent), analysts with relationship incentives partially adjust, as they appear to strike a balance between accuracy and their desire to please management. These findings hold implications for investors, regulators, and the interpretation of prior research.

This paper reports an experiment demonstrating that MBA students overrely on old earnings performance when predicting future earnings performance in a laboratory setting. In the experiment, MBA students relied too heavily on old annual ROE information to predict future annual ROE. The experiment shows how a common cognitive error (overreliance on unreliable information) interacts with the structure of the earnings time series to create particular patterns of prediction errors. The results also suggest directions for research on two well‐known anomalies, long‐run overreactions (De Bondt and Thaler 1985, 1987) and post‐earnings‐announcement drift (Bernard and Thomas 1990).