演讲人：黄海涛（南京大学数字经济与管理学院 助理教授）

时间：2026 年4月21日（周二）12:00-13:00

地点：同济大厦A楼505室

同步#腾讯会议：334199003

会议密码：612490

链接：https://meeting.tencent.com/dm/nAKscIEIU69H

摘要：When a firm’s management holds expectations about its industry’s future that diverge from those of its peers, does this deviation signal superior private information or cognitive error? We construct a novel measure of managerial expectation divergence using Large Language Models (LLMs) to analyze forward-looking disclosures in annual reports of Chinese A-share listed firms. We document two central findings. First, expectation divergence predicts deteriorating future fundamentals, consistent with managerial miscalibration. Second, divergence positively predicts future risk-adjusted returns, consistent with an information uncertainty premium demanded by investors facing elevated ambiguity about firm prospects. Both effects are driven primarily by pessimistic divergence and concentrate among firms that subsequently correct their beliefs toward the consensus. This study provides micro-level evidence that LLM-extracted managerial beliefs contain incremental information for firm performance beyond traditional textual analysis.

个人简介： 黄海涛，现任南京大学数字经济与管理学院助理教授。2021年博士毕业于美国佐治亚州立大学罗宾逊商学院。 其研究领域主要涵盖金融计量经济学、金融风险管理、数字金融等。研究成果发表于Journal of Econometrics等国际学术期刊。