Overnight Reversal and the Asymmetric Reaction to News

Abstract

News released overnight has a significant directional impact on individual shares’ opening prices, i.e., the market tends to open higher (lower) when news with positive (negative) sentiment is published. However, the market opening is not fully efficient due to over- or underreactions of market participants to the news, resulting in a predictable pattern of returns on the following trading day. In particular, we find that large daytime returns followed by overnight news with strong sentiment lead to a predictable return reversal during the subsequent trading day. This predictable reversal is present independent of the polarity of the news sentiment. Without overnight news, large previous-day returns only have marginal predictive power.

Keywords: Overnight Returns, Return Predictability, Text Mining, Sentiment Analysis, Machine Learning, BERT

JEL Classification: G10, G11, G12, G14