Mutual Fund Stock‐Picking Skill: New Evidence from Valuation‐ versus Liquidity‐Motivated Trading

• Date: 01 June 2018
• Author: Martin Rohleder,Dominik Schulte,Janik Syryca,Marco Wilkens
• Published date: 01 June 2018
• DOI: http://doi.org/10.1111/fima.12198

Mutual Fund Stock-Picking Skill: New

Evidence from Valuation- versus

Liquidity-Motivated Trading

Martin Rohleder, Dominik Schulte, Janik Syryca,

and Marco Wilkens∗

Wepropose a novel TradeMotivation Matrix that allows differentiating funds’ valuation-motivated

(VM) and liquidity-motivated (LM) trades on single trade level. It thus enables analyses of

stock-picking skill on three levels: trade, stock, and fund. On trade level, we find significant

outperformance of VM buys and significant underperformance of VM sells, indicating manager

stock-picking skills, especially during illiquid market periods. VM trades outperform LM trades,

confirming negative performance effects due to flow risk, especially when marketliquidity is low.

On stock level, collective VM buying explains high futurestock returns while collective VM selling

is related to futurelosses, indicating wisdom of the crowd. On fund level, higher trading discretion,

measured by a higher degree of VM trading, is observed for smaller, older funds holding higher

cash buffers. Finally, higher trading discretion is related to higher future fund alpha, especially

during illiquid times.

When assessing the stock-picking skill of professional investors such as mutual funds, it is vital

to distinguish valuation-motivated(VM) trades from liquidity-motivated (LM) trades. Only trades

based on valuationsallow judging managers’ stock-picking skill, while forced trades based on fund

holders’ liquidity demands may be thought of as noise trading and do not represent skill (Edelen,

1999). Based on portfolio holdings, we propose a novel Trade Motivation Matrix (TMM), which

is the first to differentiate between single holdings-based VM and LM trades. With the TMM,

it is therefore possible to run analyses on three different levels—individual trades, individual

stocks, and individual funds—whereas previous research remains on an aggregated trade level

(Alexander, Cici, and Gibson, 2007). Thus, our model enables more precise measurement of

stock-picking skill and of the costs of liquidity provision to fund investors due to flow risk

(Rakowski, 2010), a very relevant matter to which the Securities Exchange Commission (SEC)

We are grateful for helpful comments and suggestions by Bing Han (Editor), an anonymous referee, Katja Ahoniemi,

Gordon Alexander,Nicole Branger, John Broussard, Jeffrey Busse, TeodorDyakov, Ralf Elsas, Oliver Entrop, Joachim

Grammig, Jennifer Koski, Mingshen Li, Markus Natter, Alexandra Niessen-Ruenzi, Hendrik Scholz, Erik Theissen, and

participants of the 2016 Financial Management Association Europeanconference, the 2016 Conference of the German

Academic Association for Business Research, the 2016 workshop on financial management of the German Operations

Research Association, the 2015 Southern Finance Association Meeting, the 2015 PhD seminar of the German Finance

Association, and the 2015 HypoVereinsbank˗UniCredit Group doctoralseminar. Significant parts of this research were

conducted while Marco Wilkenswas a visiting scholar at the University of Sydney. Weacknowledge financial support by

the Research Center Global Business Management of the University of Augsburg.All remaining errors are our own.

∗Martin Rohleder is a post-doc at the Chair of Finance and Banking at the Universityof Augsburg in Augsburg, Germany.

Dominik Schulte is a portfolio managerat Tecta Invest in Munich, Germany.Janik Syryca is a PhD student at the Chair of

Finance and Banking at the University of Augsburgin Augsburg, Germany. Marco Wilkens is a full professorand holder

of the Chair of Finance and Banking at the University of Augsburg in Augsburg,Germany.

Financial Management •Summer 2018 •pages 309 – 347

310 Financial Management rSummer 2018

recently turned their attention.1Moreover, it allows investigating which stocks are traded based

on VM and LM and whether funds’ VM trading is related to future stock and fund performance.

Applying the TMM to a sample of more than 4.7 million trades results in several contributions.

First, on trade level, wecontribute to the literature on stock-picking skill by f inding that VM buys

have on averagehigher retur ns and VM sells havesignif icantlylower returns than their respective

benchmarks, consistent with stock-picking skill. Additionally, by being the first to conduct such

an analysis during different marketliquidity regimes, we show that VM trading decisions are more

successful during times of low market liquidity. Such times are usually associated with higher

pricing uncertainty (Chordia, Roll, and Subrahmanyam, 2008, 2011) during which fundamental

or private signals may have higher informational value, consistent with reduced information

efficiency of the market. This creates more opportunities for VM trading (Sadka and Scherbina,

2007; Dong, Feng, and Sadka, 2017; P´

astor, Stambaugh, and Taylor, 2017), whichis also reflected

in an overproportionally high number of VM trades during illiquid times. It is also consistent with

the finding that manager skill is time varying and depends on the overall economic conditions

(Kacperczyk, van Nieuwerburgh, and Veldkamp, 2014).

Second, we are the first to consider different benchmark universes to measure trade perfor-

mance. Specifically, we use all Center for Research in Security Prices (CRSP) stocks to measure

trade performance relative to stocks with similar characteristics (Daniel et al., 1997, henceforth

DGTW). However, for sells, this assumes unrestricted short selling, which is not allowed by the

SEC and therefore seldom done (Chen, Desai, and Krishnamurthy, 2013). Therefore, we alter-

natively use the respective fund’s holdings at the time of the trades to measure if trades improve

portfolio quality. Withthis distinction, we are the f irst to showspecif ically that funds’ VM trades

overall improve portfolio quality, while LM trades do not.

Third, we contribute to the literature on flow risk as the TMM facilitates a more detailed

analysis of funds’ LM trading compared to previous research. Contrasting VM and LM trades

clearly shows that funds are forced to make disadvantageous trading decisions if investors’ and

managers’ investment strategies are not aligned. This represents strong empirical evidence that

mutual fund performance suffers significantly from investor-induced flow risk. It also indicates

that previous trade-based approaches to measuring skill are biased by LM trading. We show that

this adverse effect of investor flows also worsens during illiquid times.

Fourth, on stock level, we contribute to the general understanding of fund manager’s trading

preferences. Specifically, the TMM allowsdetailed analyses of the characteristics of stocks traded

by mutual fund managers based on VM and LM. Weshow that with their VM buys, fund managers

prefer smaller over bigger companies and valueover growth stocks—that is, fund managers chase

size and value premiums (Fama and French, 1993). Moreover, if funds have a clear valuation,

they are prepared to accept higher market risk exposure as well as higher illiquidity risk, even

during times of low market liquidity. If managers are forced to trade without clear valuations

(LM buys and sells), they prefer to engage in momentum trading (Jegadeesh and Titman, 1993;

Carhart, 1997), sell very liquid stocks, and reduce risk with their LM sells.

Fifth, we find that the collective VM trading decisions of mutual funds in single stocks—

that is, the ratio of VM buys (sells) to all buys (sells) in a specific stock during a certain

quarter—represents wisdom of the fund manager crowd (Chalmers, Kaul, and Phillips, 2013;

Jiang, Verbeek,and Wang, 2014; Sias, Turtle, and Zykaj, 2016). Specifically,we show that stock-

specific VM buying ratios are significantly related to positive future stock performance over

horizons at least up to 12 months, while VM selling ratios are significantly related to negative

future stock performance.

1https://www.sec.gov/news/pressrelease/2015-201.html.

Rohleder et al. rValuation- versus Liquidity-Motivated Trading 311

Sixth, on fund level, we contribute to the literature on differences between mutual funds by

analyzing the fund characteristics associated with a higher degree of VM trading—that is, the

ratio of VM trades to all trades by a fund during a certain quarter. We show that inflows and

higher cash increase buying discretion, while outflows and low cash decrease buying discretion

(Simutin, 2014). Moreover, younger and smaller funds have higher buying discretion, consistent

with the prior literature on diseconomies of scale (Berk and Green, 2004; Chen et al., 2004; Pollet

and Wilson, 2008; P´

astor, Stambaugh, and Taylor, 2015). Fee structures such as expense ratios

and load fees have no effecton the deg ree of VM trading. Turnover as a measure of overalltrading

is positively related to VM trading, consistent with P´

astor, Stambaugh, and Taylor (2017).

Seventh, we show that a higher degree of VM trading is significantly related to funds’ future

Carhart (1997) alpha and thus translates directly into benefits for investors, especially during

illiquid market periods. It also confirms our previous finding that stock-picking skill is valuable

primarily in periods with low market efficiency and high valuation uncertainty.

Our work is thus related to various popular streams of mutual fund research. The flow-

performance relation and the potentially adverse effect of investor flows on the discretion of

open-end mutual fund managers—that is, flow risk—was first empirically investigated byEdelen

(1999). He finds that the general underperformance of actively managed funds compared to

passive alternatives can partly be explained byLM trading. Dubofsky (2010) as well as Fulkerson

and Riley (2017) confirm the strong relation between investor gross flows and aggregated mutual

fund trading during later periods.2Therefore, the SEC recently turned their attention to flow risk

and mutual fund liquidity, considering new regulation to protect buy-and-hold investors from

negative effects of LM trading caused by purchasing and redeeming investors (Hanouna et al.,

2015). Within this literature, the TMM builds particularly on the study by Pollet and Wilson

(2008) who investigate mutual fund behaviorin reaction to growth. They argue that mutual funds

may react to investor flows by means of two alternative strategies: scaling and diversification.

On the one hand, a manager without new investment ideas or valuations uses investor flows to

scale her existing holdings, thereby maintaining her old portfolio allocation. On the other hand,

a manager possessing new investment ideas and valuations may utilize investor flows to alter her

allocation and to invest in new stocks.

The TMM is also related to the fund trading literature, which attempts to assess manager

performance directly from the success of buying and selling decisions. The first study to use such

an approach is Grinblatt and Titman (1993) which documents a significantly positive covariance

between mutual fund holdings-weight-changes and subsequent stock returns based on quarterly

holdings.3Chen, Jegadeesh, and Wermers(2000) use quar terlyholdings and DGTW benchmark-

adjusted stock returns and find that stocks bought by mutual funds significantly outperform stocks

sold.4Using the same approach, Dyakov, Jiang, and Verbeek (2017) report that the informational

advantage leading to this pattern turned negative after 2001. However, none of these studies

consider trade motivation and thereby potentially underestimate skill. Moreover, these studies do

2Further empirical studies confirming the existence of flow risk are, for example, Coval and Stafford (2007), Frino,

Leopone, and Wong (2009), Cherkes, Sagi, and Stanton (2009), Rakowski(2010), and Rohleder, Schulte, and Wilkens

(2017).

3There are studies using actual mutual fund trades from the Abel Noser Corp. ANCERNO database (Puckett and Yan,

2011; Eisele, Nefedova, and Parise 2017; Busse et al., 2017). However, this database includes only 8% of total trading

volume in US stocks and 10% of total trading volume byUS domestic equity funds. Thus, these data are very valuable for

specific types of studies such as those studying the transaction costs of mutual funds (Busse et al., 2017) but inadequate

for large-scale studies on the mutual fund universe.

4Further studies using holdings-based mutual fund trades include Pinnuck (2003), Baker et al. (2010), Cullen, Gasbarro,

and Monroe (2010), Brown, Wei,and Wermers (2014), and Wei, Wermers, and Yao(2014).