Momentum, Reversals, and Fund Manager Overconfidence

• Published date: 01 August 2016
• Author: Xuemin (Sterling) Yan,Biljana N. Adebambo
• Date: 01 August 2016
• DOI: http://doi.org/10.1111/fima.12128

Momentum, Reversals, and Fund

Manager Overconfidence

Biljana N. Adebambo and Xuemin (Sterling) Yan∗

This paper examines the role of investor overconfidence and self-attribution bias in explaining

the momentum effect. We develop a novel measure of overconfidence based on characteristics

and trading patterns of US equity mutual fund managers. Stocks held by more overconfident

managers experience greater momentum profits and stronger return reversals than stocksheld by

less overconfident managers. The difference in momentum profits is not compensation for risk nor

is it attributable to stock characteristicsthat influence momentum. Our results are consistent with

Daniel, Hirshleifer, and Subrahmanyam (1998) who argue that momentum results from delayed

overreaction caused by overconfidence and biased self-attribution.

In this paper, weexamine whether investor overconfidence, combined with self-attribution bias,

contributes to the momentum effect of Jegadeesh and Titman(1993). The momentum effect, or the

tendency of recent winners to outperform recent losers over the subsequent 3 to 12 months, is the

most prominent anomaly unexplained by the Fama-French three-factor model (Fama and French,

1996). Moreover, the momentum effect has been documented around the world (Rouwenhorst,

1998; Griffin, Ji, and Martin, 2003) across many asset classes (Asness, Moskowitz, and Pedersen,

2013) and remains significant after its initial discovery (Jegadeesh and Titman, 2001).

Although several theoretical and empirical papers offerrational explanations for the momentum

effect, the literature has primarily focused on behavioral explanations due to the magnitude

of momentum profits (Chui, Titman, and Wei, 2010).1For instance, Barberis, Shleifer, and

Vishny (1998) and Hong and Stein (1999) propose models in which investors’ conservatism and

the slow diffusion of news cause initial under-reaction to information and lead to momentum.

Alternatively, Daniel et al. (1998) develop a model in which investor overconfidence and biased

self-attribution generate delayed overreaction to information and result in momentum. In this

paper, we empirically examine the Daniel et al. (1998) explanation for the momentum effect.

This paper is written in partial fulfillment of the requirements for a Doctorate of Philosophy from the University of

Missouri. We thank Marc Lipson (Editor), an anonymous referee, Debarati Bhattacharya, Stephen Ferris, Grace Hao,

John Howe, Dong Lou, Ulrike Malmendier, Shawn Ni, Shastri Sandy, and seminar participants at the University of

Memphis, University of Missouri, University of San Diego, University of Wyoming, Miami University, the 2011 FMA

Annual Meeting, the 2011 FMA Doctoral Consortium, and the 2012 AFA Annual Meeting for helpful comments. Xuemin

(Sterling) Yanacknowledges the support of Chinese National Science Foundation (Grant #71432008).All errors are our

own.

∗Biljana N. Adebambo is an Assistant Professor at the School of Business at the University of San Diego in San Diego,

CA. Xuemin (Sterling) Yanis a Professor in the Robert J. Trulaske, Sr. College of Business at the University of Missouri

in Columbia, MO.

1Rational momentum models include Berk, Green, and Naik (1999), Johnson (2002), and Sagi and Seasholes (2007).

As discussed in Chui et al. (2010), the biggest challenge for these models is to explain the magnitude of momentum

profits without assuming extreme levels of risk aversion. Among empirical papers, Conrad and Kaul (1998) and Chordia

and Shivakumar (2002) provideevidence that momentum is related to expected returns. However, Jegadeesh and Titman

(2001), Grundy and Martin (2001), and Cooper, Gutierrez, and Hameed (2004) challenge the robustness of their findings.

Financial Management •Fall 2016 •pages 609 – 639

610 Financial Management rFall 2016

Daniel et al. (1998) study an informed, but overconfident investor whooverreacts to his private

signal. If subsequent public information confirms this signal, it triggers further overreaction due

to self-attribution bias resulting in stock price momentum. In the long run, as more information

becomes available,prices gradually move to fundamentals reversing the initial overreaction. Thus,

if overconfidence drives the momentum effect, we expect both short run momentum and long

run reversal to be stronger for stocks predominantly owned by overconfident investors.

Several recent empirical papers offer evidence consistent with the hypothesis that overconfi-

dence impacts momentum. Cooper, Gutierrez, and Hameed (2004) find that momentum profits

exist only in periods following prolonged market gains. Although aggregate overconfidence

should be greater following market gains, market state by itself is not a measure of overconfi-

dence.2Chui et al. (2010) confirm that momentum prof its are higher in countries with stronger

individualism. While the authors argue that individualism is correlated with overconfidence and

self-attribution bias, they acknowledge that “. . . it does not directly measure the behavioral biases

suggested in the momentum literature.” (p. 362). In this paper, we focus on the overconfidence

of mutual fund managers and use it as a conditioning variable to provide evidencesuppor ting the

Daniel et al. (1998) hypothesis.3

To develop the overconfidence measure, we use a comprehensive sample of mutual fund

managers. In developing the measure, we focus on mutual fund managers’ overconfidence for

four reasons. First, the psychology literature suggests that overconfidence should be stronger

among professional investors (Heath and Tversky, 1991; Griffin and Tversky,1992). In addition,

Daniel et al. (1998) model overconfidence as an investor’s overestimationof the precision of their

private information and professional investors, such as mutual fund managers, are more likelyto

possess private information. Moreover, mutual funds hold a large and growing fraction of the US

stock market. According to the Investment Company Institute Fact Book (2015), mutual funds

held 24% of the US stock market at the end of 2014. Finally, detailed characteristics and holdings

data are readily available for mutual funds and their managers.

However, overconfidence is not directly observable. To overcome this challenge, we construct

an overconfidence index by combining six overconfidence and self-attribution bias proxies sug-

gested in the prior literature. Specifically, our overconfidence index includes manager’s gender,

manager’s tenure, portfolio turnover, portfolio concentration, prior performance, and idiosyn-

cratic risk. The index approach has three distinct advantages. It is parsimonious, it reduces the

noise associated with individual proxies, and, most importantly, it allows us to capture multiple

dimensions of overconfidence and the self-attribution bias.

To provide robustness to the analysis, we construct two versions of the overconfidence index.

The first version of the overconfidence index equally weights each of the six proxies. The second

version of the index is the first principal component of the six proxies. Since overconfidence is a

manager-level characteristic, while momentum is a stock-level anomaly, for each version of the

index, we compute the stock-level overconfidence index as the weighted average overconfidence

index of all fund managers who hold the stock, using the manager’s holdings of the stock as a

weight. We use the two resulting stock-level overconfidence indexes as conditioning variables in

the analysis.

2In a related paper, Asem and Tian (2010) find that momentum profits are stronger when markets continue in the same

state than when they transition to a different state, thereby supporting Daniel et al. (1998) and rejecting Hong and Stein

(1999), as well as Sagi and Seasholes (2007). However, once again, the continuation of the market state is not a direct

measure of overconfidence.

3For ease of exposition, we use the term “overconfidence” to refer to “dynamic overconfidence due to self-attribution

bias.” Our overconfidence measure captures both overconfidence and the self-attribution bias.