Do actively trading institutional investors trade on value uncertain stocks?

(1)

Do Actively Trading Institutional Investors Trade on Value

Uncertain Stocks?

Elbert Wessemius

University of Amsterdam, Amsterdam Business School MSc Finance, track Corporate Finance

10279148 15 August, 2017

Thesis supervisor: E. Zhivotova

Abstract

This study investigates whether trading behavior from active institutions predict returns in sentiment-driven, overvalued stocks. I hypothesize that most of the return institutions earn would be earned from trading activity in an environment of optimistic market sentiment. Sell activity in high VU stocks would be earning the most profit due to arbitrage limits. From the theory, the combination of short-sale impediments and market-wide sentiment increases the occurrence of overpriced stocks. Analysis of the results shows however that the results in this research are contradictory. Stocks that have high VU, as measured by 1/firm age, market-to-book equity ratio, market capitalization, and return volatility, earn positive subsequent returns when they are sold by active institutions.

(2)

Statement of Originality

This document is written by Student Elbert Sieme Wessemius who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1. Introduction

Recent research has found that anomalies in stock return are often due to overpricing in en-vironments with an optimistic market sentiment. Stambaugh, Yu, and Yuan (2012) illustrated that 11 anomalies cause higher returns in stocks, which investors are able to profit from by selling short during market optimism in particular. In the study by Stambaugh et al. (2012) sentiment was measured using the approach of Baker and Wurgler (2006), who also illustrated that when sentiment is high, stocks with specific characteristics tend to be mispriced.

Yang, Goh, and Chiyachantana (2016) were the first to show in the literature that the profits institutions earn are mainly from selling stocks. They found this result while investigating the returns for stock with specific characteristics sold by institutions. They illustrated that selling overvalued securities with high disagreement during periods of high sentiment earn institutions a significant return above stock-assigned benchmarks.

In my paper I follow Yang et al.’s (2016) approach in order to answer the question of why they are the first to show that institutions majority of profits from trading activity are gained from selling. I replicate Yang et al.’s method (2016) of investigating institutional trading behavior.

In this study I make adjustments to which type of institutional trades I investigate from the start. I only use the trading of active institutions. This highlights the profitability of active trading in periods of different sentiments. The results may have implications for managers, investors, and traders with ability to pick stock.

Active institutions are the focus due to the fact that passive institutions play a large role in the large amount of trading activity by institutions. These passive institutions trade to match a certain portfolio or a certain index. The results of Yang et al. (2016) from difficult-to-value stocks sold by institutions in an optimistic sentiment environment could be due to reasons other than return predictability.

Active traders are more likely to have return predictability and profit, as they have incentives to sell overpriced stocks and need profits to compensate for the trading costs. However, without stock-picking ability or return predictability active traders would not be able to profit from large amounts of activity. Due to their extensive trading activity they would also have the most economies of scale in gathering of information and monitoring.

Albeit there is evidence that institutional investors are informed as a group, there is less re-search investigating ability or informational advantage by institutional type and two sentiment environments. With this paper, we can determine whether institutions possess and trade on pri-vate information and profit from it. Following the results of Baker and Wurgler (2006), Stambaugh et al. (2012), and Yang et al. (2016), I investigate whether active institutions still make abnormal profit from detecting and selling overpriced stocks in two sentiment environments.

The results from Yang et al. (2016) and other recent studies on overpricing are surprising. Overpricing and profit from it should not occur due to unlimited competition, information efficiency, prices would reflect all information available to market participants. However, theories in rational and behavioral asset pricing suggest that the fact that underpricing does not seem to occur can be

(5)

reconciled. And when limits on short-selling are introduced to the theory, it is possible that stocks trade for a price higher than the present value of the cash flows.

Since institutions often own stocks, they usually only have long positions, where return pre-dictability can only be shown for sell trades. Showing return prepre-dictability with buying of under-priced stocks is excluded from the theory even when the concept of pessimistic sentiment environ-ments is introduced. However, there remains an open question of why Yang et al. (2016) are only the first in research on institutional trades to show that the return predictability of institutions stems mainly from sell trades.

One of the reasons might be that since 1990s more institutions do not have active trading strategies. Investigating all trading behavior of institutions a group for a long time period might show different results for the return predictability than when investigating institutions which have an active strategy. More institutions than before 1990s passively follow a benchmark, while in-stitutions that show return predictability commonly have holdings that are different (Cremers & Petajisto, 2009). It might be possible for research to show that return predictability of institutions comes from selling of stocks by distinguishing between institutions.

On the other hand, it could also be that there are fewer institutions able to earn a profit from stock analysis since the 1990s. This suggests that proving that institutions mainly derive profits from selling high VU stocks during market optimism would be difficult due to the difficulty of finding informed institutions. The competition between fund managers has dramatically increased since the 1990s. Institutional ownership has increased and high-frequency trading together might have made profiting from stock analysis more difficult. The time period showing performance of trading in a more recent period is however more interesting for current investors.

Institutions who are the most informed also have the most incentive to trade frequently to make use of their informational advantage. When taking into account that institutions differentiation in stock ownership rose after the 1990s and that abnormal profits might have become less common as a result, the focus must be on the trading of the most informed institutions. Actively trading institutions have the most economies of scale and naturally high trading volumes and henceforward are supposedly the most informed.

Using the method of Yang et al. (2016) this study seeks to determine the reason for their results, and therefore I focus in particular on sentiment-driven overvaluation in stocks with high disagreement (VU), examining trading from active institutions since the 1990s. I use four of eight of the proxies for VU from Yang et al. (2016) to come up with a reason for the suggested overpricing. Because of the focus on active institutions, I calculate quarterly adjusted returns instead of the yearly assigned benchmark stock returns used by Yang et al. (2016).

The empirical results from my study are quite surprising because the results are contradictory. Stocks that have high VU, as measured by firm age, book-to-market equity ratio, market capi-talization, and return volatility, earn positive subsequent returns when they are sold. The results concerning a pessimistic market sentiment environment suggest that active institutional investors do not have stock-picking skills and buy overpriced stocks as can be seen from the fact that these

(6)

stocks earn negative subsequent returns. .

One suggested reason for the contradictory result is that the excess returns are calculated for quarterly updated portfolios. The advantage of quarterly updating is the ability to account for changing stock characteristics over the year, quarterly updating intends to measure stock-picking capabilities of institutions while excluding momentum trading from the results.

The remainder of the paper is organized as follows: Section II outlines the data and how the data is used in the methodology. Section III reviews the literature on the subject of overpricing and active institutions. Section IV illustrates the empirical implications that we can test from the theory. Section V reports results on the impact of sentiment on the overvaluation of high-VU stocks. The conclusion is in Section VI.

2. Empirical method and data

2.1. The sample

Following the methods of Yang et al. (2016), the active institutions trade sample needs to consist solely of common stocks listed on NYSE, AMEX, and NASDAQ. Hence, firms that do not have a share code in CRSP equal to 10 or 11 are excluded,and the sample is modified to contain solely those stocks that have an exchange code of 1, 2 or 3. The COMPUSTAT/CRSP merged database provides quarterly fundamental values and stock prices. Daily and monthly returns stem from CRSP data. Quarterly institutional holdings and changes are taken from the institutional (13F) database.

The quarters in the sample need to be split into two market sentiment environments to distin-guish between an environment that is optimistic or pessimistic. To follow Yang et al. (2016) I use the approach of Baker and Wurgler (2006).1

The sample period begins in 1990 and ends in 2012 because of the availability of the institutions classifications by Bushee (2001).

2.2. Descriptive statistics

Table I illustrates summary statistics of the sample used in investigating whether active insti-tutions investors earn an excess return on trading stocks with high VU during market optimism.

1

Professors Malcolm Baker and Jeffrey Wurgler made their sentiment index available to researchers at http://pages.stern.nyu.edu/∼jwurgler/

(7)

Table 1: Summary statistics

Table 1 presents summary statistics of the main variables used in the analysis. ACT IV E is the the fraction of total shares outstanding owned by active institutions. ∆ACT IV E is the change in level of ownership by transient institutions compared to the previous quarter. The table also reports statistics concerning the construction of VU variables. Respectively, F IRM AGE is the number of months for which data are available on a stock in the CRSP database at time t. RET U RN V OL is the average weekly returns over the past 52 weeks and is shown in percentages. M ARKET CAP is the end-of-the-quarter market price times the shares outstanding in millions of dollars. BOOK M is the book-to-market equity ratio and is constructed by dividing total assets minus total liabilities by the market capitalization. EXCESSRET U RN is the three months DGTW-adjusted subsequent return in percentage. The total number of observations for stocks available measured over all quarters in the dataset is 187941. The data ranges from the first quarter in 1990 to the last quarter in 2012.

#Obs

Mean

Median

Std.Dev

Pct. 25

Pct. 75

ACT IV E

187941

0.125

0.094

0.116

0.037

0.181 ∆ACT IV E

187941

0.001

0.000

0.049 -0.012

0.013 F IRM AGE

187941

241.810

193

194.408

94

328 BOOK M

187941

1.523

0.527

330.138

0.331

0.781 M ARKET CAP

187941

3602.958

468.973 15802.62

141.871 1774.33

RET U RN V OL

187941

5.657

5.060

2.843

3.678

6.949 EXCESSRET U RN

184542

-1.210

-1.502

18.078 -10.522

7.315

Active institutions hold on average 12.4% of the common shares outstanding for stocks held (ACT IV E). The change in holdings (∆ACT IV E) is near equal to zero in the mean and median, but the 25 and 75 percentile show that this statistic is nonequal to zero for many stocks in the dataset. It also shows from the standard deviation in the change of quarterly active institutional holding of 4.9%. Summary statistics for the VU measures shows in the table 1 as well.

When compared to the statistics for institutions as a group (see for instance Yang et al. (2016)), The statistics on F IRM AGE and M ARKET CAP show that active institutions hold older firms or have more a longer time of information in the CRSP database and larger market capitalization, and they therefore hold stocks that are on average lower VU. The return volatilitys mean of 5.657% (RET U RN V OL) shows the same interpretation on the statistics of the stocks in active institution’s portfolio. The return volatilitys mean is lower for the stocks held by active institutions. Because a higher value uncertainty value aligns with higher return volatility, active institutions seem to hold on average stocks that are by their fundamental characteristics less difficult-to-value

2.3. General regression equation

The raw return on the stocks with specific characteristics that form institutional trading portfo-lios is adjusted with a benchmark return. This means that for each stock a comparable benchmark

(8)

is found as well as a benchmark index to control for the effects size, the book-to-market equity ratio, and momentum have on the return stocks earn.

Instead of yearly updating the benchmark like Yang et al. (2016), I quarterly update the benchmark to which a single stock belongs. This means that I am taking a different approach to measure DGTW-adjusted returns. In this way, each return is adjusted for momentum effects that change per quarter. With the quarterly adjusted DGTW return calculated, I can more reliably assess the stock-picking ability of institutions, separating excess returns from momentum trading. The time series average of the excess returns is calculated. Just as with Yang et al. (2016), the average is calculated for stocks that satisfy certain conditions. For each combination of sentiment environment; low, medium, or high VU; and active institutional trading portfolio, the excess return that institution earns is calculated. We find the excess returns that can be earned from trading on sentiment-driven overvaluation in high VU stocks. To be consistent, the t-statistic of the average excess return is adjusted by Newey-West serial correlation up to four lags.

2.4. Variable construction 2.4.1. Institutional holdings

In the 13-F holdings from Thomson-Reuters, each institutional holding is listed with a CU SIP number, an eight-character ID for each stock. The number of shares each institution holds of a particular stock is given at file date F DAT E. I determine the P ERM N O code for each eight-digit CU SIP in the 13-F holdings from Thomson-Reuters by matching the CU SIP number to the historical N CU SIP code in CRSP.

The data are then matched by P ERM N O and quarter to CRSP data to calculate the level of institutional ownership with the amount of total stocks outstanding. The SHROU T in the 13-F data also provides data on shares outstanding. However, these data in 13-F on shares outstanding are found to be inaccurate, many numbers missing.

To calculate the institutional stock ownership for active institutions, I use the classifications of Bushee (2001). I match the 13-F data by M GRN O and year with the file from Bushee’s website and keep those entries that corresponding to ownership from transient institutions.

2.4.2. Sentiment

The top-down approach of Baker and Wurgler (2006) is used to measure market sentiment.2 The median of the quarterly sentiment measure presents serves as cutoff point to split the sample into two environments. A value above the median (0.1995) qualifies a period as optimistic.

2_{Professors Malcolm Baker and Jeffrey Wurgler made their sentiment index available to researchers at}

(9)

2.4.3. Portfolios with institutional active trades

To use active trades when calculating changes, I will use the classifications made by Bushee (2001). Transient institutions are the institutions of interest. To follow the same approach as in Yang et al. (2016) for calculating changes in the level of ownership of institutions, the stocks in the data are subdivided into quantiles with the level of institutional stock holdings held by active institutions in the previous quarter divided by the total shares outstanding.

For each quantile stock holdings based on initial holdings, five new quantiles are formulated according to the changes in active institutional holdings over the quarter. The total change over the quarter, divided by shares outstanding, divides each initial quantile in five quantiles, amounting to a total of 25 portfolios. This approach to form institutional holdings portfolios was been based on the approach taken in Nofsinger and Sias (1999).

There are two portfolios of interest with stocks that experienced the largest increase or decrease in institutional stock ownership. The portfolios have stocks that are most voluminously bought or sold. In my research I call these portfolios the extreme buy and extreme sell trades portfolio. Yang et al. (2016) called these portfolios Intense Buy and Intense Sell. To be clear, stocks that had the greatest decrease in institutional ownership when controlled for initial stock ownership are allocated to the extreme sell trade.

2.4.4. Excess returns

The excess return is calculated based on benchmark assignments. Firm’s raw return over three months is subtracted with DGTW return to obtain the excess return.3 I follow the approach taken by Wermers (2004). The original DGTW benchmarks were researched by Daniel, Grinblatt, Titman, Wermers (1997). The benchmarks can be found by sorting stocks into five groups based on size, book-to-market ratios, and momentum on a quarterly basis. Adjusting raw returns with the method of DGTW is simple, and wide-spread used in investigating trades of institutions. The benchmark return adjusts for characteristics stocks have that are found in the empirical asset pricing literature to be paramount to a return on a stock. However, as we will later see, I will use a small adjustment as compared to the benchmarks used in Yang et al. (2016).

To calculate DGTW benchmark returns, the stocks listed on the NYSE, AMEX or NASDAQ are used. Firms are excluded that are Canadian and other non-U.S. Incorporated, ADRs, REITs, HOLDRs, and Primes and Scores. First, I stocks are assigned in a size portfolio based on formed NYSE market capitalization quintiles. The size quintiles breakpoints are based on the NYSE to make sure that stocks are compared in a fair manner. Each of the five portfolios is then subdivided into five new portfolios based on industry-adjusted book-to-market quintiles (Wermers, 2004). Based on 48 industry classifications for industry SIC codes from CRSP, calculated are the average market-to-book ratio for quarter t. 4 The standard deviation of the difference between the

3

Professor Russ Wermers has portfolio assignment and benchmark returns available at http://terpconnect.umd.edu/∼wermers/ftpsite/Dgtw/coverpage.htm

4

(10)

book-to-market ratio of stock i and the average in each of the 48 industry is calculated.

Each of the 25 portfolios based on the market capitalization and the industry-adjusted book-to-market ratio is then divided into five portfolios with the past year’s return. One hundred and twenty-five portfolios are therefore formed. Each portfolio has stocks with characteristics that are similar. The benchmark return is the average in the quarter of these stocks’ returns in each of the 125 portfolios. In order to account for the disproportional impact that small stocks can have on the benchmark return, the value-weighted average return is taken. The market value of equity at time t is the weight for each stock in its benchmark. The value-weighted benchmark return is then the sum of the returns and weights. Each stock has benchmark consisting of a portfolio of firms in same size, industry-adjusted market-to-book ratio, and past one year return. Just as in one of the sections of Wermers (2004), the benchmarks based on stock characteristics are updated on a quarter-to-quarter basis. The benchmark used in Yang et al. (2016) were assigned to a stock once a year.

The benchmark-adjusted DGTW return for each stock is the raw return minus the value-weighted DGTW benchmark return. The raw return is subtracted by the matching DGTW char-acteristic benchmark for all quarters for each stock. The holding period of 3, 6 or 12 months DGTW-adjusted return is calculated compounding the subsequent quarterly DGTW-adjusted re-turns. Formally calculating the excess return is shown in equation 2:

Excess Returnit= Rit− Rpt (1)

The excess return in equation (2) is the benchmark-adjusted return for stock i in quarter t. The excess return is calculated using benchmark portfolio return p, subtracting it from the raw return of stock i (DGTW, 1997; Wermers, 2004). I report the time series average of these excess returns for different degrees of Value Uncertainty in section 5. In the following section, I note what the results would be according to the theory.

3. Literature review

3.1. Institutions

Institutions are not always allowed to trade to earn an excess return on overpriced stocks. A large share of institutions are not allowed to take a short position by their regulations. Thus, institutions need to own stock in order to sell and make profit if solely overpricing provides profit opportunities. Mandates of institutions could therefore constitute one of reasons that impairs the ability to correct mispricing during optimistic market sentiment from the viewpoint of active and passive institutions. Agency problems institutions have are generally a major impairment to sell short overvalued assets more frequently (He Xiong, 2013).

The institutions that are allowed to short-sell and conduct short-selling have an important

(11)

contribution to stock prices during an environment of optimistic sentiment (Nagel, 2005). It is increasing welfare in society as a whole. The impact of sentiment namely increases when insti-tutional stock holding decreases on the probability of a stock market crisis (Zouaoui, Nouyrigat, Beer, 2011). Trading on sentiment-driven overvaluation is, therefore, potentially important for it could increase the amount of information and research reflected in prices.

The number of institutional investors with stock-picking ability able to trade on private in-formation and monitor stocks have decreased. Active trading strategies became less popular in the recent years (Cremers Petajisto, 2009). Short selling might have become less popular as well because of less informational advantage and skill that institutions can use.

Institutions might have different mechanisms in which they are beneficial to price efficiency. Hedge funds seem to contribute the most during normal periods to price efficiency. During a liquidity crisis however the increase in price efficiency because of the trading of these hedge funds stops and can even reverse Cao, Liang, Lo, Petrasek, 2014. Informed traders seem often to have incentives to trade in a way that minimizes information leakage. There might exist incentives for institutions to short-sell less often based on the workings of price discovery (Boehmer Wu, 2013). The existing research findings are mixed on what the impact is from short-term trading on market efficiency (Cremers Pareek, 2015).

Yang et al. (2016), who test institutional ability across types, follow Bushee (2001) in classifying institutions on past portfolio turnover and diversification of institutions. The authors use a sample from the first quarter of 1980 to the second quarter of 2013. The ability of the different types of institutions is, however, not tested for sentiment-driven overvaluation in highly VU stocks.

3.2. Short selling

Stambaugh, et al. (2012) show that 11 anomalies commonly discussed in the asset pricing literature present more profit opportunities in a high sentiment environment. This is evidence that most of the time, share prices reflects optimistic valuations in an optimistic environment. Rational and pessimistic investors tend to not longer hold these stocks during such a period. Stambaugh, et al. (2012) explain that anomalies during high sentiment in stock returns exist because of the limits on short-selling.

Short-selling limits tend to withdraw pessimistic valuations from being reflected in prices during optimistic market sentiment. If the mispricing in stocks persists, it leads to an inefficient allocation of capital and resources. Less efficient stock prices guide firms to make worse-informed investment and financing decisions.

The existence of mispricing needs us to the reject the efficient market hypothesis. An equity price should be efficient due to competition and unlimited arbitrage opportunities. Even if there is a lot of demand from optimistists in the market, the efficient market hypothesis argues that the demands should be met by arbitrageurs. The large demand of optimists would thus have no significant impact on prices. Mispricing in an environment of optimistic market sentiment is therefore in conflict with the assumptions on which the efficient market hypothesis is built.

(12)

A less strong variant of the efficient market hypothesis could explain overpricing. The costs of arbitrage could be too high to trade overvalued stocks in market environments with too high optimism. Selling short requires a counterparty willing to lend the asset and is sometimes the main mechanism by which rational or pessimistic investors can trade on overvaluation. Although there are often no restrictions on short-selling in the developed markets around the world, it is often difficult to sell short.

To investors, short-sell limits mean that when supply of a stock is relatively small, only opti-mistic investors buy and hold shares during high sentiment. The consequence is that the market price of a stock can be above its fair value. Short-sellers make up 20% of the trading volume and tend to have access to relevant information important to price efficiency. This emphasizes the fact that short sellers play an important role in price efficiency (Boehmer, Jones & Zhan, 2008).

The key insight into short-selling is that in the presence of arbitrage constraints, optimistic valuations tend to be overrepresented in stocks with high disagreement (Miller 1977). Overpricing would then most often appear and remain in stocks with high disagreement or value uncertainty (VU). Institutions have less short-sale constraints than individuals. Institutions can also monitor many stocks when considering buying and are the most likely to short-sell salient stocks.

Mispricing is, by definition, the result of uninformed demand shocks in the presence of a binding arbitrage limit, i.e. short selling (Miller, 1997; Baker & Wurgler, 2006). I investigate the role of institutions in clearing the market of mispricing by trading on sentiment-driven overvalued VU stocks. Mispricing predicts that sentiment does not have the same effect on all stocks traded in the market. Some stocks might have more arbitrage limits due to difficulties with short selling or costs of arbitrage compared to stocks with less VU.

3.3. Sentiment

By definition, sentiment means that during a period of time beliefs in the market can become irrational, wide-spread shared, and significantly impacting prices. As Baker and Wurgler (2007) note, unlike they would have a few decades ago, today the question is not quote ”whether investor sentiment affects stock prices, but rather how to measure investor sentiment and quantify its ef-fects.” In terms of the quantification of sentiment effects, the present research investigates for which stocks the effect of sentiment is the strongest. This suggests that whether investor sentiment does affect stock prices significantly is debatable. Keynes (1936) argued that investors can be driven by emotions, and this affects financial markets. The concept of investor sentiment has already been a subject of discussion for a long time.

The bottom-up approach in the literature considers the reactions of individual investors. Indi-vidual overreaction or underreaction to past returns and fundamentals is then applied to determine the expected return or possible overpricing. The other approach is to measure sentiment in a macroeconomic rigour. The advantage of this so-called ”top-down approach” is the intuition it has. The macroeconomic approach has a large scope of applicability and it fits alongside bubbles and crashes.

(13)

The top-down approach of Baker and Wurgler (2006) to sentiment is often adopted in research. Their index is formed on the basis of market-based measures. They find that when sentiment is optimistic, stocks with specific characteristics tend to be mispriced. Stocks that seem most affected by sentiment are those around which exists most disagreement about their value. This is because value uncertain stocks are attractive to optimists and difficult to arbitrage, and thus experience a lower return compared to their price in a period of optimistic market sentiment environment.

The classic argument against an the effect of sentiment on prices is that overpricing should be eliminated by rational traders seeking to exploit the profit opportunities created by mispricing. So therefore, although some authors might not agree with sentiment driving prices, mispricing exists because of limits on short-selling. If rational traders cannot trade on overpricing, the probability of sentiment effects increases.

Sentiment-driven investors are likely to participate more heavily in the stock market during high-sentiment periods, creating more mispricing (e.g. Stambaugh et al. 2012). Investor sentiment has thus potential to influence prices in the market, which are commonly with a tendency it being above the value of future cash flows in an optimistic environment.

3.4. Value uncertainty

High beta stocks tend to have both higher short demand and more binding short sale constraints. The model of Hong and Sraer (2016) illustrates the impact of shorting limits on the Security Market Line. A riskier stock is shown in their study to have a higher return only until the SML kinks. They discover that some stockholders are not rewarded for higher risk as would be implied from a positive relationship between risk investors expose themselves to and return.

High beta stocks are stock that tend to be in uncertain or immature industries, many of them are technology sensitive or recently founded. It might be uncertain for high beta stocks what signals are particularly important to the future payoff a stock. When individuals cannot figure out or agree how much a signal need to be weighted to value a stock appropriately, asymptotic beliefs could emerge. Market participants have the most difficult agreeing on their models in value uncertain stocks (Acemoglu, Chernozhukov, Yildiz, 2009).

Overconfidence can make professional traders believe too strongly in signals. In particular during optimistic environments, the probability of self-attribution bias leading to overconfidence is high. When the signal of an investor is confirmed by other optimistic investors during such a optimistic market environment, overconfidence increases (Glaser, Langer and Weber 2013).

4. Empirical implications

In this research, I investigate whether overpricing exists in stocks with high disagreement. Investors may hold different beliefs due to different interpretation of knowledge, overconfidence in signals, or ideas about the workings of the market of assets.

(14)

short-sale constraints. In other words, if not all investors can buy or sell stock on the basis of their beliefs, the price would increase. For instance, investors with a strong pessimistic belief about the value of a stock would want to sell. The fact that pessimistic investors would need to have the stock in order to sell, could lead to mispricing if short-selling constraints exist.

Value Uncertain stocks are more easily subject to being priced higher than their fundamental value, and, as previously stated, this is likely to be affected by market sentiment. Therefore, I apply the concept of market sentiment index widely used in the literature (Baker and Wurgler 2006). In low sentiment periods, it is less likely that beliefs about the value of a stock are optimistic. More pessimistic investors hold stocks, and pessimistic valuations are most likely fully reflected in prices. Investors with higher optimistic valuations can buy the stock and are allowed to reflect their valuation. However, in high sentiment periods, if short selling is costly, pessimistic traders may be less likely to trade, and, as a result, biasedness emanate in prices.

I use trading of active institutions to test the hypothesis that stocks with high VU under short-sale constraints tend to be overpriced in periods of optimistic market sentiment. I compare stocks with high disagreement (high value uncertainty) with stocks with low disagreement (low value uncertainty) in optimistic and pessimistic market sentiment environment.

I distinguish active institutions from the start. Bushee (2001) made classifications to seperate institutions. I use these classifications made by Bushee to find which institutions are active. The empirical results and the discussion on the validity of the implication is under section 5.

Implication 1. Active institutions should show the ability to predict stock returns. Active investors are most likely to respond to firm-specific information because of the incentives to monitor and to gather and generate information. Sentiment-driven mispricing in highly VU stocks should have been spotted and the return predicted by active traders.

Implication 2. Institutions should show a decrease in return predictability compared to a more distant past. This research covers the period 1990 to 2013. It should demonstrate less overpricing in VU stocks from the increased competition for profits. Competition increases price efficiency and, as a consequence, the results might show that institutions have a decrease in ability to predict stock returns.

Implication 3. The institutions should show return predictability in optimistic sentiment environ-ments. Limits on short-selling are shown to drive stock price above fundamental value in optimistic environments. Optimistic investors are most likely be the holders of stocks during these periods, which would restrict the reflection of pessimistic or rational investor information and valuation in equity prices. Overpricing mainly in an optimistic sentiment environment would be in line with previous research (Stambaugh et al. 2012; Yang et al. 2016).

Implication 4. Institutions are expected to show return predictability confined to selling. Any profit from analysis on stocks, as supposed by Miller (1977), would be due to selling securities with high VU. Yang et al. (2016) were the first to reveal that institutional selling is the main driver in their ability to have return predictability.

(15)

Implication 5. The ability to trade and detect overpriced stocks should be high for active institu-tions. Yang et al. (2016) illustrate that transient institutions and quasi-indexers earn more profit from overpriced stocks than dedicated institutions. Generally, the incentives for active institutions are bigger to detect and trade on overpricing.

5. Results of the extreme buy and sell trades

The degree of disagreement is captured by four of the eight variables created by Yang et al. (2016). I only calculate the book-to-market equity ratio, firm age, market capitalization, and return volatility as measures of Value Uncertainty.

5.1. Firm age

F IRM AGE is constructed using the number of months since the first entry on fundamentals about a stock in CRSP. A history of business fundamentals provides institutional investors more information on the firm. CRSP provides more information about fundamentals the longer the firm is included in the database. Firms earlier in the past tend to be included at an earlier date. Younger, or newly listed firms, have less information in CRSP. Therefore, firm age is a proxy for how much information on a firm exist. As a rule, younger firms tend to be more value uncertain with less information available. I take the number of months to allow interpreting a higher value for the variable as higher value uncertainty.

When the inverse of a firm’s age is used as proxy for VU, the stocks with high VU institutions buy during optimistic sentiment underperform the DGTW benchmark after three months by -4.76%. The results are shown in Table 3. The stocks with low and medium VU within the same subsequent holding period also earn an abnormal negative return of -2.163% and -2.965% after the stocks have been bought. After 6 months the returns on holding the stock approximately double, while after 12 months the decrease is less substantial when compared to the returns after a holding period of 6 months.

(16)

Table 2: Value Uncertainty measured as 1/firm age

Table 2 reports the excess return that active institutional investors earn on extreme buy and sell trades for stocks with low, medium, and high disagreement as measured by 1/F IRM AGE. The total change in active institutional stock ownership puts stocks each quarter in either the extreme buy or extreme sell portfolio. The DGTW-adjusted average returns for these total of 6 portfolios are given for 3 months in columns (1), (2), and (3); 6 months in columns (4), (5), and (6); and 12 months in columns (7), (8), and (9) after the stocks are respectively bought or sold. These averages are based on the fact that trades took place in either an optimistic (Panel A) or a pessimistic market sentiment environment (Panel B) between 1990-2012. The returns are reported in percentages. T-statistics is corrected for NeweyWest serial correlation up to four lags. ** and * denote statistical significance at the 1% and 5% level, the number in the brackets below each excess return coefficient is the corresponding t-statistic.

Panel A: Optimistic market sentiment environment

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VU low VU medium VU high VU low VU medium VU high VU low VU medium VU high Extreme buy trades -2.163** -2.965** -4.755** -4.370** -6.359** -8.950** -5.058** -7.695** -9.355** [-9.91] [-11.41] [-18.98] [-14.12] [-17.35] [-24.80] [-14.18] [-18.05] [-21.06] Extreme sell trades 1.687** 1.421** 1.470** 2.790** 2.086** 2.525** 1.966** 1.413** 2.440**

[7.70] [5.92] [5.91] [8.58] [6.24] [7.03] [5.23] [3.52] [5.57]

Panel B: Pessimistic market sentiment environment

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VU low VU medium VU high VU low VU medium VU high VU low VU medium VU high Extreme buy trades -2.703** -4.114** -4.221** -4.830** -7.247** -7.877** -4.982** -7.453** -8.355

[-10.49] [-14.49] [-14.09] [-13.11] [-17.97] [-18.11] [-11.38] [-15.22] [-13.74]

Extreme sell trades 0.489* 0.577* 0.433 0.739* 0.833* 1.216** 0.639 0.295 1.072*

[2.22] [2.32] [1.66] [2.34] [2.40] [3.02] [1.76] [0.66] [2.31]

Stocks that have an extreme increase in level of ownership by active institutions, the sell trades portfolio, have returns that are significantly positive during an optimistic sentiment environment across all stock characteristics and holding periods. The return on high VU stocks sold by institu-tions is 1.470% after 3 months, 2.525% after 6 months, and 2.440% after 12 months.

This does not confirm the hypothesis that it is chiefly stocks with high disagreement that tend to be overpriced and sold by active institutional investors during high sentiment. In fact, the opposite seems to be true. Similar to stocks with little information in CRSP, stocks that have large amounts of information in CRSP earn positive subsequent returns when sold.

The results for the firm age as proxy for Value Uncertainty during an environment of pessimistic sentiment suggest that stocks bought by active institutional investors earn an approximately equal negative return as for the result on optimistic market sentiment. It suggests that active institutional investors neither have stock picking skills during market pessimism, and buy stocks that are being overpriced in the market. Since the high and medium VU stocks earn an negative DGTW-adjusted return of -7.877% and -7.247% as compared to -4.830% for low VU, respectively, it can be argued that this is evidence for larger overvaluation in stocks with higher Value Uncertainty. However, since the sell trades show that stocks with a lower firm age would earn chiefly a positive significant return after 12 months of 1.072% (t-statistic = 2.31), it would suggest undervaluation exists in these

(17)

type of stocks. Undervaluation during pessimistic market sentiment was typically precluded from asset pricing models, and it is not confirming of recent work on Value Uncertainty and sentiment. 5.2. Market-to-book equity ratio

In periods of high sentiment, stocks would tend to be overvalued as compared to their funda-mentals. The financial measure book-to-market equity is capturing stages of firm’s growth. Low book-to-market ratio are usually growth firms. A book-to-market ratio with a value greater than 1 means the market has a higher valuation on firms equity value and are more mature. The COM-PUSTAT/CRSP merged database provides quarterly fundamental values and stock prices. Higher VU firms have a higher inverse value of the equity book-to-market ratio. Therefore we reported the results as market-to-book to measure VU.

Table 3: Value Uncertainty measured as market-to-book ratio

Table 3 reports the excess return that active institutional investors earn on extreme buy and sell trades for stocks with low, medium, and high disagreement as measured by 1/BOOK M . The total change in active institutional stock ownership puts stocks each quarter in either the extreme buy or extreme sell portfolio. The DGTW-adjusted average returns for these total of 6 portfolios are given for 3 months in columns (1), (2), and (3); 6 months in columns (4), (5), and (6); and 12 months in columns (7), (8), and (9) after the stocks are respectively bought or sold. These averages are based on the fact that trades took place in either an optimistic (Panel A) or a pessimistic market sentiment environment (Panel B) between 1990-2012. The returns are reported in percentages. T-statistics is corrected for NeweyWest serial correlation up to four lags. ** and * denote statistical significance at the 1% and 5% level, the number in the brackets below each excess return coefficient is the corresponding t-statistic.

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VU low VU medium VU high VU low VU medium VU high VU low VU medium VU high Extreme buy trades -0.530* -2.351** -6.642** -1.939** -5.125** -11.893** -2.433** -5.910** -12.746** [-2.32] [-10.48] [-26.07] [-6.04] [-16.43] [-33.89 [-6.34] [-16.25] [-30.95]

Extreme sell trades 2.745** 1.958** -0.304 3.807** 3.160** 0.113 3.034** 2.873** -0.261

[13.37] [9.04] [-1.13] [12.67] [10.29] [0.30] [8.54] [7.77] [-0.59]

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VU low VU medium VU high VU low VU medium VU high VU low VU medium VU high Extreme buy trades -1.036** -3.763** -6.242** -2.617** -6.216** -11.067** -2.312** -6.308** -11.936** [-3.69] [-14.24] [-19.48] [-6.57] [-15.52] [-25.45] [-3.76] [-13.09] [-23.18]

Extreme sell trades 1.543** 0.351 -0.407 2.205** 0.620 0.003 2.277** 0.668 -0.889*

[6.23] [1.47] [-1.55] [6.10] [1.92] [0.01] [5.25] [1.62] [-2.02]

The results of institutional trading activity during optimistic environment in panel A of table 3, report large losses on the stocks bought. After three month of holding high VU stocks, the return is -6.642%. Prices of these stocks show further decline from the 6 months subsequent excess return of -11.893%. The stocks with lower VU show higher returns. Stocks with low VU report to have a return on average of -2.433%. Interestingly, the stocks sold do seem to show more rational trading

(18)

from institutions. The foregone negative losses are smaller and the stocks sold with high VU have (insignificant) negative returns after 12 months.

Panel B of table 3 presents the excess returns institutions have from trading activity during pessimistic market environment. The stocks bought with high disagreement also decline in price. After 12 months the return is also in negative double digits (-11.936 %, t-statistic=-23.18). From the sell activity during this environment we see comparable results to Panel A. High VU stocks sold earn in this case a significant negative excess return of -0.889% (t-statistic=-2.02).

5.3. Market capitalization

M ARKET CAP is a measure for disagreement about the value of a stock because of the fact that larger firms tend to have lower valuation uncertainty. Larger firms have wider news coverage, more feedback from its customers, a history of business conduct, and more stock market and accounting information available from databases.

Table 4: Value Uncertainty measured as market capitalization

Table 4 reports the excess return that active institutional investors earn on extreme buy and sell trades for stocks with low, medium, and high disagreement as measured by M ARKET CAP . The total change in active institutional stock ownership puts stocks each quarter in either the extreme buy or extreme sell portfolio. The DGTW-adjusted average returns for these total of 6 portfolios are given for 3 months in columns (1), (2), and (3); 6 months in columns (4), (5), and (6); and 12 months in columns (7), (8), and (9) after the stocks are respectively bought or sold. These averages are based on the fact that trades took place in either an optimistic (Panel A) or a pessimistic market sentiment environment (Panel B) between 1990-2012. The returns are reported in percentages. T-statistics is corrected for NeweyWest serial correlation up to four lags. ** and * denote statistical significance at the 1% and 5% level, the number in the brackets below each excess return coefficient is the corresponding t-statistic.

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VU low VU medium VU high VU low VU medium VU high VU low VU medium VU high Extreme buy trades -2.549** -3.762** -3.927** -4.881** -7.462** -7.779** -5.161** -8.144** -9.206** [-12.57] [-15.84] [-14.29] [-16.69] [-22.71] [-19.97] [-15.18] [-20.96] [-19.71] Extreme sell trades 1.601** 1.658** 1.181** 2.994** 2.934** 1.416** 3.430** 2.661** 0.023

[6.59] [7.36] [5.19] [8.91] [9.06] [4.34] [8.60] [6.75] [0.06]

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VU low VU medium VU high VU low VU medium VU high VU low VU medium VU high Extreme buy trades -1.596** -4.251** -5.215** -3.555** -7.448** -9.035** -3.273** -7.599** -9.961** [-6.17] [-15.39] [-15.52] [-9.18] [-19.78] [-18.79] [-7.21] [-15.85] [-15.15] Extreme sell trades 0.927** 1.180** -0.564* 2.521** 1.431** -0.973** 2.417** 1.491** -1.660**

[4.31] [4.68] [-2.11] [7.88] [3.88] [-2.56] [6.63] [3.61] [-3.41]

The same pattern holds in the return for some of the results when measuring Value Uncertainty with M ARKET CAP , the corresponding table is table 5. The first 3, 6, and 12 months after that stocks are bought extremely by active institutions during an optimistic sentiment, earn more

(19)

negative significant return as compared to their DGTW benchmark. Medium and high VU stocks earn a larger negative than the low VU stocks. A large share of negative excess return is incurred after 6 months, only the high VU stocks further decrease in return by more than 1.5% when comparing the 12 month holding return (-9.206%, t-statistic=-19.71) with the 6 month holding return (-7.779%, t-statistic=-19.97).

The extreme sell trades during optimistic market sentiment environment show a new result when M ARKET CAP is used. The high VU stocks extreme sold earn a lower return than the low and medium VU stocks, the return after 12 months becomes insignificant for high VU stocks. Some caution needs to be made when interpreting the results of M ARKET CAP , since the measure is confounded with other characteristics of stocks than it is intended to measure.

Stocks that are traded during pessimistic sentiment show a somewhat equal pattern, except for the fact that high VU stocks sold earn significant negative return across all holding periods. After three months the return is -0.564% (t-statistic=-2.11), 6 months equals -0.973% (t-statistic=-2.56), and 12 months has an even lower return of -1.660%.

5.4. Return volatility

Stock volatility of the return captures how uncertain the fundamental value of firm is. RET U RN V OL increases when there is a lot of volatility in the valuation investors made on a stock. RET U RN V OL is defined as the past year weekly returns and are calculated with the daily returns from the CRSP database. A higher RET U RN V OL intends to correspond to higher Value Uncertainty.

As we saw previously, the results for RET U RN V OL as proxy for VU in table 6 show, that stocks extremely much bought by active institutions during market optimism earn a negative excess return, and higher VU earning a lower return than medium and low VU. Most of the negative return occurs 6 months after the quarter that stocks are bought. While stocks that are sold have a positive significant excess return.

(20)

Table 5: Value Uncertainty measured as weekly return volatility

Table 5 reports the excess return that active institutional investors earn on extreme buy and sell trades for stocks with low, medium, and high disagreement as measured by RET U RN V OL. The total change in active institutional stock ownership puts stocks each quarter in either the extreme buy or extreme sell portfolio. The DGTW-adjusted average returns for these total of 6 portfolios are given for 3 months in columns (1), (2), and (3); 6 months in columns (4), (5), and (6); and 12 months in columns (7), (8), and (9) after the stocks are respectively bought or sold. These averages are based on the fact that trades took place in either an optimistic (Panel A) or a pessimistic market sentiment environment (Panel B) between 1990-2012. The returns are reported in percentages. T-statistics is corrected for NeweyWest serial correlation up to four lags. ** and * denote statistical significance at the 1% and 5% level, the number in the brackets below each excess return coefficient is the corresponding t-statistic.

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VU low VU medium VU high VU low VU medium VU high VU low VU medium VU high Extreme buy trades -2.162** -2.965** -4.755** -4.370** -6.359** -8.950** -5.058** -7.695** -9.355** [-9.91] [-11.41] [-14.18] [-14.12] [-17.35] [-24.80] [-14.18] [-18.05] [-21.06] Extreme sell trades 1.687** 1.421** 1.470** 2.790** 2.086** 2.525** 1.966** 1.413** 2.440**

[7.70] [5.92] [5.91] [8.58] [6.24] [7.03] [5.23] [3.52] [5.57]

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VU low VU medium VU high VU low VU medium VU high VU low VU medium VU high Extreme buy trades -2.703** -4.114** -4.221** -4.830** -7.247** -7.877** -4.982** -7.453** -8.355** [-10.49] [-14.49] [-14.09] [-13.11] [-17.97] [-18.11] [-11.38] [-15.22] [-13.74]

Extreme sell trades 0.489* 0.577* 0.434 0.739* 0.833* 1.216** 0.639 0.295 1.072*

[2.22] [2.32] [1.66] [2.34] [2.40] [3.02] [1.76] [0.66] [2.31]

Panel B in Table 6 displays the results of trades completed in a pessimistic sentiment environ-ment and shows similar returns on purchased stocks. Higher VU stocks earn a lower excess return than other stocks for all holding periods for extreme buy trades.

Extreme sell trades during pessimistic market sentiment show new findings. High VU stocks sold during this sentiment earn insignificant excess returns after three months. Furthermore, significant positive excess returns in high-VU stocks decrease from 1.216% at 6 months to 1.072% after 12 months.

Institutions seem to take a loss however from selling these stocks. Low VU and medium VU stocks show no significant excess return after 12 months. This could mean that during market pessimism stocks with low value uncertainty have a premium due to their perceived less risk. However, this could not be conclusively determined using a comparison with other results.

5.5. Discussion

The results from the portfolios for the three levels of VU are inconsistent with the hypothesis. Instead of stocks with high VU earning lower subsequent returns when sold by institutions during market optimism, most of the results in this research illustrate that the excess returns, when

(21)

compared to DGTW benchmarks, are positive for 3, 6, and 12 months.

In this study, value uncertainty was measured using a few fundamental values pertaining to a stock. These values show consistency in the results for stocks traded in an optimistic market sentiment environment.

The implications of the results concerning the active trading from institutions are striking. The results of extreme buy trades show that institutions buy stocks that will earn them negative profits. In addition, a substantial amount of the stocks sold earn higher return than their respective benchmark, which would also mean that institutions do not demonstrate stock-picking capability.

The results are not directly comparable to the results of Yang et al. (2016), who used a different time period for their research on institutions and did not show the results of transient institutions trading in two different sentiment environment periods. There is a clear disjunction between their results and those of this study. Extreme sell trades earn a significant negative DGTW-adjusted return, especially for stocks traded that have high VU.

Comparing my results to those of Yang et al. (2016), we can see some opposite findings. The higher VU stocks are shown to earn larger absolute returns than medium and low VU stocks, however, the sign of the returns suggests that a different relationship exists. Institutions would actually earn less from trading on high-VU stocks as opposed to more. My results in the different sentiment periods therefore do not agree with the findings of Yang et al. (2016).

I attempted to answer the following research question: Are the returns of sentiment-driven overvaluation in high-VU stocks still predicted by active institutional trading activity? The first empirical implication proposed by the theory, that active institutions show ability to predict returns, is not supported by my findings. The stocks that institutions extremely sell perform better than stocks extremely bought.

Therefore, although the results contradict the first implication, the second implication is sup-ported. More importantly, institutions not only show a decrease in return predictability, but they do not show any return predictability for the stocks they have extremely bought or sold.

The third and fourth implications can only hold when institutions show some amount ofreturn predictability, and since we can conclude from the result that they do not, both implications are not found to hold in this study.

Conclusions on the basis of the tables in this section must be made with caution due to the possible existence of problems with the DGTW-adjusted returns. The low returns of stocks bought by active institutions might be the reason that the results are so different. If there is a mistake made in the return, the results are easily explained when compared to the benchmarks.

The average raw returns, benchmark returns, and the difference are shown in Table 2. In con-trast to Yang et al. (2016), I calculated benchmark assignments and returns from CRSP data, up-dated the portfolios quarterly, and allocated stocks to benchmarks. The advantage of this approach to update quarterly is being able to account for changing momentum and other characteristics of a stock. Updating yearly would mean that stocks have benchmarks in the data based on funda-mentals from the current quarter but also from 3, 6, or 9 months ago. Since active or transient

(22)

institutions are often trading on momentum, it is important to quarterly update the benchmark portfolios. The results from my paper more closely evaluate the stock-picking abilities of fund managers. Quarterly updating ensures that the results do not show momentum trading but rather actual stock-picking ability.

Further research might want to change the frequency of the benchmark assignments from yearly to quarterly, or in this research’s case from quarterly to yearly or biannually, and check the accuracy of the results with these new results. Investigating trades of active institutions may require updating DGTW benchmarks quarterly.

In addition, it seems more reasonable to separate investors into groups by how much a portfolio differs from a benchmark index. By applying Cremers and Petajistos (2009) methodology, a new measure to predict fund performance and measure active management called ”Active Share” could be used, which they show to be superior to the conventional portfolio turnover ratio.

6. Conclusion

In my research I have attempted to categorize institutions according to the method of Bushee (2001) due to the conciseness. Investigating the results of different types of institutional investors across different market sentiments, the methodology of my research was largely based on the methodology of Yang et al. (2016).

Yang et al. (2016) were the first to discover that the return predictability institutions show is mainly from the stocks sold. I tried to come up with an answer as to why they were the first to achieve this in the research. They used the approach to measure sentiment of Baker and Wurgler (2006) and found that the sell trades significantly predict returns in periods of market optimism.

Yang et al. (2016) did not report their results of different types of institutional investors for two sentiment environments. My contribution to the research underscored the results of active trading in different sentiment periods. I hypothesized that one of the reasons previous research had not agreed with the results by Yang et al. (2016) is that they had done little to separate institutions according to their trading strategy. I hypothesized that active institutions would show more significant return predictability due to their incentives.

The results in my research however draw a different conclusion. I used quarterly benchmark assignments in order to determine the excess returns that active institutions earn from trading on momentum. Quarterly updating would show stock-picking ability but also has drawbacks.

Further research might also want to find institutional trades from active institutions by dif-ferent classifications. Other approaches to find active institutions might have advantages over the classification used in this research (Cremers & Petajisto, 2009).

(23)

7. Bibliography

Acemoglu, D., Chernozhukov, V., & Yildiz, M. (2009). Fragility of asymptotic agreement under Bayesian learning. Unpublished paper, Department of Economics, Massachusetts Institute of Technology.[342, 343].

Baker, M., & Wurgler, J. (2006). Investor sentiment and the crosssection of stock returns. The Journal of Finance, 61 (4), 1645-1680.

Baker, M., & Wurgler, J. (2007). Investor sentiment in the stock market. The Journal of Economic Perspectives, 21(2), 129-151.

Boehmer, E., & Wu, J. J. (2013). Short selling and the price discovery process. Review of Financial Studies, 26 (2), 287-322.

Bushee, B. J. (2001). Do institutional investors prefer nearterm earnings over longrun value?. Contemporary Accounting Research, 18 (2), 207-246.

Cao, C., Liang, B., Lo, A. W., & Petrasek, L. (2014). Hedge fund holdings and stock market efficiency (No. 2014-36). Board of Governors of the Federal Reserve System (US).

Cremers, M., & Pareek, A. (2015). Short-term trading and stock return anomalies: Momentum, reversal, and share issuance. Review of Finance, 19(4), 1649-1701.

Cremers, K. M., & Petajisto, A. (2009). How active is your fund manager? A new measure that predicts performance. Review of Financial Studies, 22 (9), 3329-3365.

Daniel, K., Grinblatt, M., Titman, S., & Wermers, R. (1997). Measuring mutual fund perfor-mance with characteristicbased benchmarks. The Journal of finance, 52 (3), 1035-1058.

Glaser, M., Langer, T., & Weber, M. (2013). True overconfidence in interval estimates: Evidence based on a new measure of miscalibration. Journal of Behavioral Decision Making, 26 (5), 405-417. He, Z., & Xiong, W. (2013). Delegated asset management, investment mandates, and capital immobility. Journal of Financial Economics, 107 (2), 239-258.

Hong, H., Sraer, D. A. (2016). Speculative betas. The Journal of Finance, 71 (5), 2095-2144. Keynes, J. M. (1936). The General Theory of Employment, Interest and Money, Harcout Brace and World. Inc, New York.

Miller, E. M. (1977). Risk, uncertainty, and divergence of opinion. The Journal of finance, 32 (4), 1151-1168.

Nagel, S. (2005). Short sales, institutional investors and the cross-section of stock returns. Journal of Financial Economics, 78 (2), 277-309.

Nofsinger, J. R., & Sias, R. W. (1999). Herding and feedback trading by institutional and individual investors. The Journal of finance, 54 (6), 2263-2295.

Stambaugh, R. F., Yu, J., & Yuan, Y. (2012). The short of it: Investor sentiment and anomalies. Journal of Financial Economics, 104(2), 288-302.

Wermers, R. (2003). Is money really’smart’ ? New evidence on the relation between mutual fund flows, manager behavior, and performance persistence.

Yang, L. Z., Goh, J., & Chiyachantana, C. (2016). Valuation uncertainty, market sentiment and the informativeness of institutional trades. Journal of Banking & Finance, 72, 81-98.

(24)

Zouaoui, M., Nouyrigat, G., & Beer, F. (2011). How does investor sentiment affect stock market crises? Evidence from panel data. Financial Review, 46 (4), 723-747.

Do actively trading institutional investors trade on value uncertain stocks?