Limited Attention and Market Pricing of Earnings in a High Frequency World
53 Pages Posted: 25 Jul 2015 Last revised: 14 Apr 2020
Date Written: April 13, 2020
A stream of literature shows that human attention constraints affect asset pricing in predictable ways. When traders are distracted, stock prices tend to initially underreact to earnings news and then gradually incorporate the news over subsequent weeks. In modern markets, however, the majority of trades are conducted by machines using pre-programmed algorithms. We examine whether limited attention still affects stock prices in today’s markets, where high-frequency traders (HFTs), which use fully automated algorithmic trading, play a large role. On one hand, machines should not suffer from human attention constraints. On the other hand, if HFTs primarily trade with non-HFTs or their algorithms reflect human biases, their presence may not alleviate attention effects. Using multiple proxies of attention constraints and a dataset that identifies HFTs’ trades, we find that price inefficiencies are reduced by 65% to 100% when HFTs trade following low-attention earnings announcements: Initial price responses are larger and post-earnings-announcement drift is reduced. Our results are not driven by firm size or announcement time-of-day patterns. We control for possible endogeneity using instrumental variable analysis and two event studies of exogenous shocks to HFTs, all of which suggest that HFTs causally reduce low-attention effects. By highlighting one channel through which machine traders affect asset prices in modern markets, we present important new evidence on the interplay between limited attention and the incorporation of earnings news into stock prices. This is the first evidence that a human bias in asset pricing is mitigated in the presence of HFTs.
Keywords: Limited attention, price efficiency, earnings announcements, high-frequency trading
JEL Classification: G02, G10, G14, M40, M41
Suggested Citation: Suggested Citation