Limited attention and portfolio choice: The impact of attention allocation on mutual fund performance

• Author: Ankur Pareek,Swasti Gupta‐Mukherjee
• Published date: 01 December 2020
• Date: 01 December 2020
• DOI: http://doi.org/10.1111/fima.12294

DOI: 10.1111/fima.12294

ORIGINAL ARTICLE

Limited attention and portfolio choice: The impact

of attention allocation on mutual fund

performance

Swasti Gupta-Mukherjee1Ankur Pareek2

1Quinlan School of Business, LoyolaUniversity

Chicago, Chicago, Illinois

2LeeBusiness School, University of Nevada, Las

Vegas , Neva da

Correspondence

SwastiGupta-Mukherjee, Quinlan School of

Business,Loyola University Chicago, 16 E Pearson

St,Chicago, IL 60611.

Email:sguptamukherjee@luc.edu

Abstract

This study proposes that the performance of mutual fund managers

is linked to how efficiently they allocate attention across assets in

their investment set. Motivated by existing models of optimal port-

folio choice and rational inattention, we posit that the efficiency

of attention allocation increases when a manager chooses larger

(smaller) active positions in assets that need more (less) informa-

tion acquisition effort to resolve uncertainty about future payoffs.

Weshow that the efficiency of attention allocation has a significantly

positive impact on future fund performance. Efficient attention allo-

cation has a lesser impact on performance as the total demands on a

manager’s limited attention increase.

KEYWORDS

attention allocation, limited attention, mutual funds, performance

evaluation

[Fidelity Contrafund] lagged the S&P 500 from 1994 to 1997 …and held a mammoth number of stocks: 700.

Danoff [manager of Fidelity Contrafund] says he came to realize that therewas a limit to the number of compa-

nies he could follow…. So he cut the number of names down to 400.

Fidelity News & Insights (July 16, 2013)1

1INTRODUCTION

The issue of whether mutual fund managers possess skill has been the topic of vigorous debate in the academic

literature. Various studies posit that a typical actively managed U.S.equity fund underperforms by earning negative

after-cost alphas (see, e.g., Carhart, 1997; Fama & French, 2010; French, 2008; Gruber,1996). Others have refuted

c

2019 Financial Management Association International

1Nellie S. Huang, “7 Best Large-Company Mutual Funds,” Kiplinger Personal Finance, July 16, 2013, https://finance.yahoo.com/news/7-best-large-company-

mutual-040001707.html.

Financial Management. 2020;49:1083–1125. wileyonlinelibrary.com/journal/fima 1083

1084 GUPTA-MUKHERJEEAND PAREEK

the evidence on underperformance, with some studies focusing on identifying skilled managers (see, e.g., Amihud &

Goyenko, 2013; Baker,Litov, Wachter, & Wurgler, 2010; Berk & van Binsbergen, 2015; Cremers & Petajisto, 2009;

Kacperczyk & Seru, 2007; Kacperczyk, Sialm, & Zheng, 2005, 2008; Koijen, 2014; Kosowski, Timmermann, Wermers,

& White, 2006; Pastor & Stambaugh, 2002; Pástor, Stambaugh, & Taylor,2015; Wermers, 2000). Although much of

the literature has focused on measuring the performance of actively managed funds and what it implies about skill,

less is known about the portfolio-level actions of managers that generateperformance and lead at least some funds to

underperform.

The rational inattention paradigm suggests that, because information acquisition and processing can impose sig-

nificant costs in terms of required effort and expenses, among the most crucial tasks facing fund managers could be

selecting how to allocate their limited attention.2In this article, we shed light on the distribution of effort (or attention)

to learning about different assets as an important channel by which a manager’s portfolio-levelactions could translate

to investorreturns in delegated portfolios. Although effort and attention allocation are unobservable, there are at least

two reasons why we expect fund managers to put different levelsof effort into analyzing different assets. First, assets

vary in the uncertainty of their valuation, and thus the effort required to estimate their true value. Second, a manager’s

incentive to put effort into analyzing an asset should depend on how influential the asset is in driving the fund’s perfor-

mance relative to the benchmark. Prefaced by these ideas that suggest that there is variation in how much attention

a manager allocates to specific assets, we use an extensive sample of actively managed U.S. equity mutual funds to

test the hypothesis that fund performance increases with the efficiency with which managers allocate their attention

across assets.

Anecdotal evidence from actively managed funds, such as Fidelity Contrafund’s manager Will Danoff believing he

could not realistically follow many companies, indicates that fund managers are subject to limited attention. Neverthe-

less, little evidence exists linking the managing of limited attention to the real outcomes of activeinvestors. In viewing

the efficient allocation of finite attention resources across specific tasks as a facet of investmentskill, our article is most

closely related to Kacperczyk, VanNieuwerburgh, and Veldkamp (2014, 2016) (hereafter KVV). KVV (2014, 2016) are

mainly interested in the time-varying attention fund managers allocate between markettiming and stock picking tasks

in different states of the economy.Our article differs in that we focus on how fund managers allocate attention across

the cross-section of stocks that vie for their attention during stock picking and monitoring tasks at a snapshot in time.

To test our centralhypothesis, we develop a measure called attention gap, where the efficiency of attention allo-

cation is interpreted as the inverse of the attention gap measure. The theoretical motivation for this measure comes

from studies that model optimal portfolio choice under rational inattention, building on the seminal works on limited

attention by Kahneman (1973) and Sims (2003). Specifically, Peng (2005) posits that investorsshould optimally allo-

cate more information acquisition effort to learning about the assets with more uncertain payoffs—that is, assets that

areharder to value—to minimize total portfolio uncertainty. Additionally, VanNieuwerburgh and Veldkamp (2010) pro-

pose that investorsshould optimally hold larger active positions in assets on which they spend more information acqui-

sition (i.e., learning) effort, where active positions captured by deviations from the benchmark represent a manager’s

“learning portfolio.”3Takentogether, the implication is that it is optimal for an investor to put more effort into learn-

ing about assets with more uncertain payoffs, and make larger activeallocations to these assets because the investor

spends more effort on them. Hence, portfolio choices converge toward optimal when the sizes of the active positions

in assets are proportional to the uncertainty in their payoffs.4In our framework, the efficiency of attention allocation

2Fortheoretical and empirical studies on rational inattention in financial markets, see Hong, Torous, and Valkanov(2007), Peng and Xiong (2006), and Huang

andLiu (2007).

3Otherrelated studies include Mondria (2010), Gabaix & Laibson (2005), and Gabaix, Laibson, Moloche,and Weinberg (2006).

4A similar prediction emerges from cost–benefit models of attention allocation (see Gabaix & Laibson, 2005; Gabaix et al., 2006). Given the structure of

commonly used portfolio management contracts linking the compensation of fund managers to their excess performance overa passive benchmark, fund

managershave an incentive to allocate larger active positions to stocks requiring more attention that contribute most to their portfolio uncertainty and, thus,

mostinfluence their performance relative to the benchmark.

GUPTA-MUKHERJEEAND PAREEK 1085

increases, and the attention gap decreases, when the manager chooses larger (smaller) active positions in assets that

need more (less) information acquisition effort to resolve uncertainty.

To construct our baseline measure of attention gap, we follow a four-step process. First, for each fund in a quar-

ter, we measure the information acquisition effort associated with each stock in the fund’s investment set based on

the uncertainty in its valuation.5We view this as a proxy for optimal attention allocation to the stock, or its “attention

demand.” We assume that stocks that are more volatile, havea higher dispersion of analyst earnings forecasts, or that

are linked to more diversified businesses require more information acquisition effort (see Cohen & Lou, 2012;Zhang,

2006). Second, we measure the manager’s investment choice in each stock in the fund’s investment set based on its

active weight (see Cremers & Petajisto, 2009). Third, we calculatethe attention gap for each stock in the fund’s invest-

ment set in each quarter from the disparity in the optimal attention allocation and the investment choice, which equals

the absolute value of the difference between the stock’s attention demand relative to all the stocks in the investment

set, and its active weight relative to all the stocks in the investment set. Finally,we sum the attention gap across the

stocks in the investment set and divide it byone-half to obtain the fund’s attention gap, where the fund’s attention gap

ranges from zero to one in theory.

We construct three alternative measures of attention gap with analyst forecast dispersion, stock return volatility,

or firm complexity as alternativestock-level proxies for attention demand. We find that there is substantial dispersion

in attention gap across funds and that smaller and more active funds (i.e., with higher Active Share)tend to have lower

attention gaps.6The results are consistent with diseconomies of scale being associated with less efficient attention

allocation by the manager of a bigger fund. Also, in view of existing studies such as Berk and Green (2004) and Cre-

mers and Petajisto (2009) that posit that evidence of skill is likely to be found in smaller and more active funds, these

resultsforeshadow the notion that the efficiency of attention allocation is a dimension of skill. Additionally, we find that

the efficiency of attention allocation increases significantly with the total demands on the manager’s limited attention

(“total attention demand”) for lower levels of total attention demand but becomes less sensitive to increases in the

total attention demand at higher levels of such demand. Overall, the correlations between fund and portfolio charac-

teristics and our attention gap measures are quite low,and attention gap contains information distinct from other fund

and portfolio attributes. Additionally, we find strong persistence in funds’ attention gap overtime, indicating that the

attention gap is driven primarily by stable fund-levelfactors.

Further, our evidence supports the prediction that fund performance decreases with attention gap, or increases

with the efficiency of attention allocation. Forthe three alternative specifications of attention gap, the after-cost four-

factor alphas of Carhart (1997) are 1.94–3.45% higher per year for the funds in the lowest attention gap decile (Decile

1) relative to the funds in the highest attention gap decile (Decile 10), where funds are sorted on their lagged attention

gap. A main driverof the results is the underperformance of funds in Decile 10 of attention gap, which generate average

four-factor alphas of –1.67% to –2.36% per year.The results are similar when abnormal fund performance is measured

relative to comparable index funds (Berk & van Binsbergen, 2015) or relativeto benchmark portfolios based on stock

characteristics (Daniel, Grinblatt, Titman, & Wermers, 1997). Additionally, there is strong persistence in the positive

Decile 1−10 return spread up to about 2 years following the ranking of funds into deciles.

To rule out an important alternative interpretation that a fund’s attention gap simply proxiesfor omitted pricing

factors linked to the stock attributes used to compute attention gap, the tests are sharpened by augmenting the

factor models used to evaluate abnormal performance. The standard factor models are augmented with severalfactor

mimicking portfolios, which capture the return spread associated with the cross-sectional variation in the following

stock attributes: average activeweight allocated to a stock by funds, active weight dispersion (see Jiang & Sun, 2014),

5A fund’s investment set is defined as the universeof stocks likely to compete for the manager’s attention, assumed to be the stocks currently in the fund’s

portfolioand the set of stocks in the fund’s benchmark index.

6Tothe extent that attention gap reflects the efficiency of attention allocation, these results are similar to KVV (2014) who find that smaller and more active

fundsexhibit better time-varying attention allocation between stock picking and market timing skill, conditional on the state of the economy.