Momentum, Information, and Herding

Abstract

This study investigates the potential explanations to the momentum effect on the equity market. We primarily discuss the underreaction hypothesis, the overreaction hypothesis, and the impact of herding behavior. We find that the momentum effect disappeared after decimalization in all size deciles, which does not support the underreaction hypothesis. We also find that momentum profits do not exist in any intangible assets or R&D expenses deciles, which is not consistent with the continuous overreaction hypothesis. We further investigate the impact of herding behavior on the momentum effect. Using a new firm-level herding measurement, we find that investors require higher returns in high herding stocks and they require even higher returns in high herding stocks among previous losers, indicating that investors herd against the previous losers while they herd toward the winners.

Keywords:

• Momentum
• Underreaction
• Continuous overreaction
• Tick size
• Transaction costs
• Ambiguity information
• Herding behavior

JEL:

• G11
• G12
• G14
• G40
• G41

Notes

1 According to www.thebalance.com, a meme stock is a stock that has seen an increase in volume not because of the company's performance, but rather because of hype on social media and online forums like Reddit. For this reason, these stocks often become overvalued, seeing drastic price increases in just a short amount of time.

2 One can argue that the disappeared momentum profits may be a result of the poor proxies for ambiguous information. To address the concern, we use a more consistent proxy for information ambiguity, book-to-market ratio, to examine the momentum effect in three periods: January 1990 to April 1997 when tick size is 1/8, July 1997 to December 2000 when tick size is 1/16, and May 2001 to December 2015 when tick size is penny (Daniel and Titman, 1999). We do find that momentum profits decreased with information ambiguity when tick size is 1/8 and disappear after tick size decreased.

3 According to Rothschild & Co Wealth Management, 2021. https://www.rothschildandco.com/siteassets/publications/rothschildandco/private_wealth/2021/quarterly_letter/en_wm_quarterly_letter_january_2021_straying_from_the_herd.pdf

4 The study by Daniel, Hirshleifer, and Subrahmanyam (1998) mentioned that momentum should be stronger for stocks that are difficult to value, such as those with high R&D expenses and intangible assets. The proposed theory motivated us to use these two proxies to conduct our analyses.

5 We merge stock returns obtained from CRSP with COMPUSTAT dataset by CUSIP.

6 The market capitalization is calculated six months before the start of the ranking period.

7 The momentum effect disappeared after decimalization across all size deciles are not driven by any particular stock exchanges. Table IA1 presents momentum strategies for the period of January 1980 to December 2015 for NYSE, AMEX and NASDAQ in Panel A, Panel B and Panel C, separately. We conduct the tests for two reasons. First, the average stock size in NASDAQ and AMEX is smaller than that in NYSE. Second, transaction costs for stocks in NASDAQ are higher than those in NYSE (Huang and Stoll, 1996; Goldstein and Kavajecz, 2000). The difference in transaction costs between NASDAQ and NYSE became smaller after tick size changed (Bessembinder 1999; Bessembinder 2003). We separate stock exchanges to address these concerns. Tables IA2 to 4 further present the results for the momentum effect for three subsample periods, January 1980 to April 1997, July 1997 to December 2000, and May 2001 to December 2015. To address the concern that buying the top 30 percent (P1) stocks and selling the bottom 30 percent (P3) stocks seems to be arbitrary and might not be sufficient to capture the momentum effect. We redo the tests by sorting stock returns into 10 groups. Table IA5 presents the results when buying the top 10 percent (P1) stocks and selling the bottom 10 percent (P10) stocks. Although the momentum profits in Table IA5 are higher than those in Table 3, they exhibit very similar patterns. Before tick size decreased from 1/8 dollar to 1/16 dollar, momentum profits decreased with sizes, supporting the underreaction hypothesis. After tick size decreased from 1/8 dollar to 1/16 dollar, momentum profits increased with size, contradicting the underreaction hypothesis. After decimalization, the momentum effect disappeared across all size deciles, supporting the friction hypothesis and the idea that transaction costs can no longer limit arbitrage (Korajczyk and Sadka 2004).

8 For robustness check, we use book-to-market as a proxy for information ambiguity to examine the overreaction hypothesis for three time periods: January 1990 to April 1997 when tick size is 1/8, July 1997 to December 2000 when tick size is 1/16, and May 2001 to December 2015 when tick size is penny. Table IA6 presents the results. Panel A presents the results for the whole sample period, and Panel B, Panel C and Panel D present the results for the three subperiods. We find that momentum profits decrease with book-to-market from January 1980 to April 1997 when tick size is 1/8, which is consistent with the findings in Daniel and Titman (1999). However, after tick size decreased, momentum profits exist only in the smallest book-to-market deciles, which is consistent with the results presented in Table 3 and Table 4. Table IA6 and Figure IA1 further support the results that momentum profits disappeared after tick size decreased, and the continuous overreaction hypothesis can no longer explain the momentum effect.

9 Since we do not find solid results using R&D expense and intangible assets as proxies for information ambiguity, we decide to use book-to-market ratio in this section.