“Attention – Grabbing Stocks and Individual
Holdings”
University of Amsterdam
Faculty of Economics and Business
Master in Business Economics: Finance
Aug 2014
Supervisor:
Florian Peters
Di Shen
10605053
Abstract
Barber and Odean (2008) study the trading behavior of individual investors in response to attention-attracting events: existence in news, abnormal trading volume and abnormal stock return. Repeated events make the stocks appear in front of the public more and people tend to be more familiar with them. This thesis is trying to find out whether individual investors, with limited time and resources, tend to buy stocks that stand out more in a time period. After doing sorts and regressions, this thesis proves that individual investors are more likely to buy stocks that are experiencing abnormal trading volume and abnormal stock return in that time period.
TABLE OF CONTENTS
I. INTRODUCTION ... 1
II. LITERATURE REVIEW ... 3
A. Irrational trading behavior ... 3
B. Limited attention ... 4
C. Media and financial market... 5
D. Advertisement and financial market ... 5
III. DATA... 6
A. Trading data ... 6
B. Institutional holdings data ... 7
C. News impact ... 8
D. Data screening ... 8
IV. METHODOLOGY AND HYPOTHESIS... 9
A. Analysis and hypothesis ... 9
B. Sort methodology...10 C. Regression methodology...12 V. RESULTS...13 A. Sorts...13 B. Regression ...22 VI. DISCUSSION ...25 VII. CONCLUSION ...26 REFERENCES ...29
1 I. Introduction
Behavioral finance is attracting increasing attention. The neoclassical theories do give us perfect frames to analyze economical and financial problems. The imperfection in the real world, however, makes it less likely to adjust the forecasted result to the real world.
In financial market, we usually assume that investors have full knowledge of each firm and are totally rational and informed when they are making decisions on buying and selling. People are supposed to invest in stocks that are most valuable. However, this is usually not the case in the real world. On the one hand, as individual investors only have limited energy, resources and time to learn about each firm, they tend to pay more attention to stocks that they are more familiar with when choosing which one to buy or which one to learn about. On the other hand, even if they are aware of their current irrationality, they are reluctant to make more effort to learn more about the firms they are going to invest. For individual investors, as their cost of investigating all the companies is much larger than their investment and potential benefit, it is rational for them to take into account only a limited number of firms. Gabaix and Laibson (2003) report the first empirical test of a cost-benefit model of the endogenous allocation of attention. They assume that economic agents only have finite mental processing speeds and are unable to analyze all of the elements in complex problems. This assumption, although different from standard economic models, is much more realistic. In addition, individual investors are more suffered from short-selling constraints, and only sell stocks that they already own (because of short-selling ban or personal preference). As a result, individual investors with limited time and money, only pay attention to stocks that attract their attention. What is worse, they only pick up stocks from this subsample to buy, while not short-selling any of the stocks about which they hold negative views.
Previous researches focus on people’s immediate response to attention-attracting affairs, such as news coverage, advertisements, abnormal trading volume or abnormal stock return. When the stocks are showing extreme trading volume or stock return, covered by news, or appear in front of the public more often through advertisements, individual investors ttend to pay attention to them. Although, these affairs do attract individual investors’ attention, they are
2
also sending information to the public. For example, extremely high stock return is probably attractive itself as a measure of stock performance or positive expectation from the market. So it is hard to distinguish whether it is the very existence of the attention-attracting events or the information people get from those events that make individual investors tend to buy those stocks.
This thesis is trying to find out whether individual investors tend to buy stocks that stand out more. The main idea is that this research is using the data of a longer term, in order to get rid of the influence of information. As for the information part, investors usually respond to the new information they get by trading immediately. So the relatively long term effect of being outstanding is just strengthening people’s familiarity about a stock and thus attracting more individual investors. The similar thing could happen among superstars. A superstar may try to make things happen, no matter good things or bad things, trying to appear in front of the public as much as possible. It is possible that firms use the similar strategy to make themselves known by more investors. This research can be helpful not only in explaining people’s trading behavior and the movement of stock prices but also in suggesting more reasonable individual trading. It is also a new direction for supervisors to consider when analyzing firms’ inappropriate behaviors and making suggestions to protect individual investors.
Barber and Odean (2008) prove that individual investors only buy stocks that have recently caught their attention. They are net buyers of stocks in the news, stocks experiencing high abnormal trading volume and stocks with extreme one-day returns. Institutional investors, compared to individual ones, have more sources and more power in studying the stocks, are less constrained by short-selling bans and thus are more rational when making investment decisions. So the percentage of institutional share holdings of each stock is a good measure of how attractive the stock is to an individual investor compared to an institutional investor. This thesis uses a similar method to their research in terms of the effect of abnormal trading volume and abnormal stock return, but tr ies to find out the long term impact of them. Getting information and getting attracted by something happen just immediately, but getting familiar
3
with something unconsciously because of repeated appearance is a longer term effect.
This thesis tries to find out whether the very existence of standing out influences the individual demand for a firm’s stock. Following the general method of Barber and Odean (2008), I sort the data by abnormal trading volume and abnormal stock return for each quarter from 1990 to 2010, finding out the change in institutional holdings percentage in each decile. As the long term effect is much less significant than the immediate effect, which makes the result not clearly shown by the sort tables or the figures, this thesis also does regressions of institutional holdings percentage change on deciles with time fixed effect.
The paper is organized as follows. The next section reviews the related literature. Section 3 describes the data, screening rules and variable definitions. Section 4 describes the hypothesis and research method. Section 5 examines the long term effect of outstanding, including the procedures and results. Section 6 analyzes the results and discusses. Finally, section 6 concludes.
II. Literature Review
A. Irrational trading behavior
A large number of researches examine investors’, especially individual investors’, irrational trading behavior and biases from standard model assumptions in the real world. Barberis and Thaler (2003) discuss the two building blocks of behavioral finance, limits to arbitrage and psychology, and show many behavioral finance applications. Hirshleifer (2001) sets up a framework for understanding decision biases and proves that investor psychology p lays an important role in security pricing. Baker, Pan and Wurgler (2012) find that target’s past peak prices offers reference points, and thus influence bidder’s offer price, deal success, merger waves and bidder announcement effects. Odean (1998b) states that many investors trade too much because they are overconfident about the quality of the information they get. Investors tend to hold losing investments too long and sell winning investments too soon. This tendency is labeled the disposition effect by Shefrin and Statman (1985). Odean (1998) tests the disposition effect by analyzing trading records for 10,000 accounts and proves it to be true.
4
Moreover, individual investors’ trading decisions tend to be influenced by their mood (Bollen, Mao and Zeng, 2011) and financial illiteracy is prevalent among investors (Van Rooij, Lusardi and Alessie, 2011).
B. Limited attention
In behavioral finance area, the sub-area concerning the effect of limited attention on financial market attracts attention of many researchers. According to Barber and Odean (2008), individual investors are net buyers of attention-grabbing stocks. The paper argues that individual investors are more influenced by attention when buying - where they search across thousands of stocks - than when selling - where they usually choose from the stocks they already own. In addition, individual investors are more suffered from short-selling constraints. In that paper, they measure attention by news, abnormal trading volume and extreme one-day returns and measure individual investor’s trading behavior by buy-sell imbalance. Examining people’s immediate trading response to attention-attracting events, they prove their proposition to be true. Seasholes and Wu (2004) find that, on the Shanghai Stock Exchange, individual investors are net buyers the day after a stock hits an upper price limit. They argue that it is because the event of hitting an upper price limit attracts the attention from individual investors, especially first-time buyers. Kaniel and Mingelgrin (2001) find that stocks experiencing a high trading volume tend to appreciate in the following month because of increased visibility. Managers take advantage of the effect of attention on financial market and investors’ distraction by releasing worse news on Fridays (Della and Pollet (2006)). As individual investors are more subject to short selling constraints than institutional investors, only optimistic views of individual investors can be reflected by the movement of stock price, which leads to overpricing. Miller (1977) argues that individual investors with pessimistic views are prevented from short-selling. Lee (1992) examines trading activity around earnings announcements for 230 stocks over a one-year period and finds that small traders, who place market orders that are less than $10,000, are net buyers following not only positive earnings surprises but also negative ones. This thesis also tries to find out the long-term effect of sudden decrease in stock return on individual investors’ trading behavior. If individual investors do, in the long term, have the tendency to buy stocks experiencing extreme stock
5
return, even extremely negative ones, they are influenced by the salience of the event, not the information they get from the event.
C. Media and financial market
Many studies have investigated the relationship between media and financial market. Similar to attention-catching events, media is a good way of attracting investors’ attention. In the short term, media coverage brings about information that is significant for trading and makes people pay attention to stocks in the news. In the long term, the public is more familiar with stocks that are mentioned more by the news. Engelberg and Parsons (2011) find that local media coverage predicts local trading and that the timing of the local reporting is also related to the local trading. By comparing the behaviors of individual investors who have different access to different media because of geography factors, keeping the information event the same, they find that media has a causal impact in financial market. They exploit the impact of extreme weather and micro-level variation in the timing of a story’s publication to identify media effects on people’s trading behavior. This paper sets up a model of investor demand considering the influence of media on financial market. This thesis could use this model to better explain the theory. The model analyzing the impact of media can be used to analyze the impact of attention-attracting events. Peress (2011) investigates this causal impact by exploiting exogenous media blackouts resulting from national newspaper strikes in several countries to see how the difference in time when investors get information influences their trading behavior.
D. Advertisement and financial market
A large number of papers also try to find out the relationship between advertisement, another way firms attract attention from investors, and financial market. However, as advertising is a strategy of firms, the information provided by advertisements can be biased, exaggerated or even total lies. Although the public knows this fact, they still tend to buy products or stocks from firms that are more advertised. Grullon, Kanatas and Weston (2004) document that firms who have higher advertising expenditure have more institutional and individual investors because advertising can increase investors’ familiarity with the firm and investors tend to pay
6
attention to familiar firms and buy them. This thesis expands this research in the effect of advertising to that of abnormal trading volume and abnormal stock return. Even though the information from advertisements is unreliable, people still tend to buy stocks that are more advertised because of higher familiarity. So it can also be true that no matter what the attention-attracting event is, even the sudden decrease in stock return, the more often stocks stand out among thousands of stocks, the more individual investors purchase them because of higher familiarity. Lou (2009) provides evidence that managers exploit the effect of advertising on short-term stock prices by adjusting firm advertising expenditures. Chemmanur and Yan (2009) find that as advertising could help a firm to attract investors’ attention, a larger amount of advertising is associated with a larger stock return in the advertising year but the stock return decreases in the subsequent year as the attracted attention wears out over time.
III. Data
A. Trading Data:
This thesis tries to follow the method of measuring attention in Barber and Odean (2008), using abnormal trading volume and abnormal stock return. In Barber and Odean (2008), they examine individual investors’ immediate response to attention-grabbing events using data for each day. However, as this thesis is studying the long term effect of standing out on people trading behavior, both the trading volume and stock return are quarterly data. Specifically, monthly trading volume (vol) and monthly holding period return (ret) from CRSP (Center for Research in Security Prices) dataset are used. In monthly files, “vol” is the sum of the trading volumes during that month and is expressed on hundred shares. Stock return is the change in total value of an investment in a common stock over a time period, per dollar of initial investment. CRSP uses the following formula to get holding period return:
r(t) = [p(t)f(t) + d(t) p(t′) ] − 1
where: 𝑡0=time of last available price, usually one period before t r(t)=return on purchase at 𝑡0, sale at t
7
d(t)=cash adjustment factor for t f(t)=price adjustment factor for t
p(𝑡0)=last sale price or closing bid/ask average at time of last available price
The time period that the data is covering is from January 1990 to December 2010. Instead of studying all the companies, the trading data of only S&P 500 companies is used. There are two reasons. Firstly, a too large sample can bring bias to the result of hypothesis tests. Even if there is no relationship between two variables, with a huge sample, the coefficient between them can be significant. Secondly, S&P 500 (Standard & Poor ’s 500) companies are large companies having common stocks listed on the NYSE or NASDAQ. The S&P 500 index has diverse constituency and is considered one of the best representations of the U.S. stock market. A large number of papers use S&P 500 firms as their research sample. So this thesis is also using these 500 companies as research sample.
Next, in order to get the quarterly trading volume data and quarterly holding period return data, monthly trading volume and monthly holding period return are added up respectively for the three months in the quarter.
B. Institutional holdings data
This thesis uses the percentage of institutional investors holding of each stock to measure individual investors’ trading behavior. The data of institutional holdings is quarterly data including the percentage of institutional stock holdings (inst) of each quarter and the change of percentage of institutional stock holdings from quarter to quarter (Dstake_inst) from 1990 to 2010. Specifically, when the percentage of institutional investors’ stock holding decreases, which means the percentage of individual investors’ stock holding increases, individual investors are net buyers in that time period. When the percentage of institutional investors’ stock holding increases, which means the percentage of individual investors’ stock holding decreases, individual investors are net sellers in that time period. For the same reason, the data of only S&P 500 companies is used.
8
C. News impact
In addition to abnormal trading volume and abnormal stock return, news plays an important role in attracting individual investors’ attention. Barber and Odean (2008) try to find out its impact in individual investors’ trading by comparing the buy-and-sell imbalance for stocks that are covered and not covered in the news for each day. However, the information effect mentioned before is even stronger for news. News brings much information about the firms. Is the buy-and-sell imbalance for stocks that are covered by news caused by the information or the very existence of news coverage? It is hard to distinguish because they are testing people’s immediate trading response to news. However, as the time period of news data available is different from the time period of institutional stock holdings data, it is not possible to test its influence in the way of testing the impact of abnormal trading volume and abnormal stock return. It has been proved many times and is also obvious that the more frequently a stock is covered by news, the higher the trading volume is. As a result, the trading volume is also an effective measure of news coverage. It is the combination of pure accidentally high trading volume and the high trading volume caused by news coverage. So this thesis is not going to test the impact of news separately, but includes it in the study of abnormal trading volume.
D. Data screening
After calculation of quarterly data using monthly data, the total number of observations is 40,853. This number is smaller than 42,000, which it is supposed to be (500(companies)*21(years)*4(quarters)). The reason is that for some companies, the data of some time periods is missing. After excluding observations with missing Dstake_inst data and missing Aret_q and Avol_q (the calculation of these two data is included in the method part, which will be introduced later) data, the number of observations becomes 28,738. According to the histogram of Dstake_inst got from current data, excluding Dstake_inst data that is larger than 0.2 or more negative than -0.2 is most helpful for getting rid of the influence of outliers. After this step, there are 28,462 observations left. This is the final sample for analysis.
9 IV. Methodology and Hypothesis
A. Analysis and hypothesis
To better explain the research and the hypothesis, a model from a research in a related area is used.
Many papers have tried to analyze the impact of media coverage in financial market. But few have established a model until Joseph (2011). He formalizes this problem using the model of investor demand D: D(X, M(X, Y)) (1) 𝑑𝐷 =𝜕𝐷 𝜕𝑋𝑑𝑋 + 𝜕𝐷 𝜕𝑀∗ 𝜕𝑀 𝜕𝑋𝑑𝑋 + 𝜕𝐷 𝜕𝑀∗ 𝜕𝑀 𝜕𝑌𝑑𝑌 (2)
Where M is media coverage, X is a set of characteristics that potentially determine both media coverage and investor demand, and Y is a set of characteristics that only influence media coverage. Y pertains to factors that influence only the media’s objective function, but otherwise have no bearing on the behavior of traders. In most of the previous research including this research by Joseph, people try to reject the null hypothesis that D=D(X). They try to find out whether media has a (causal) impact in the demand of each stock, which influence the financial market. Further, in (2) they empirically separate the second and third terms - the media effects - from the market’s reactions to the underlying events (first term). However, at the end of this paper, Joseph mentioned that “a number of papers have shown that news stories in national newspapers are associated with substantial price responses” and that “Less explored is the possibility that the story’s very existence- a media effect – may generate a response independent of these channels. One could imagine, for example, decomposing the “news response coefficients” estimated in such studies into ‘media effects’ and ‘content effects’ ”.
This thesis is studying a similar area to this less explored question, decomposing the “abnormal trading volume response coefficients” and “abnormal stock return coefficients” into “outstanding effects” and “information effects”. More specifically, this thesis tries to find out whether individual investors tend to buy stocks that are outstanding more. For individual
10
investors with only limited resources and time, going through all of the stocks is impossible, not to mention learning well about them. So they tend to pick up several stocks from a subsample that they are more familiar with to analyze. As what is mentioned before, the information is normally reflected in the price through investors’ immediate trading response. So using the data of a longer time period is appropriate to explain the question in this thesis.
As a result, this thesis is trying to test the following hypothesis:
Hypothesis 1: The more a stock experiences abnormal trading volume, the less institutional holding of the stock is, the more individual investors tend to hold the stock.
Hypothesis 2: The more a stock experiences abnormal stock return, the less institutional holding of the stock is, the more individual investors tend to hold the stock.
Hypothesis 3: For stocks that are experiencing negative stock return, the larger the negative stock return is, the less institutional holding of the stock is, the more individual investors tend to hold the stock.
B. Sort Methodology
Barber and Odean (2008) test the extent to which individual investors tend to buy stocks on days when the trading volume is abnormally high, the stocks return is abnormally high the last day or the stocks appear in the news. They are testing individual investors’ immediate trading response. However, individual investors are also trading in response to the information they get from the events not only to the stocks’ salience. For a longer time, such as a quarter, if the more a stock attracts individual investors’ attention, the more individual investors buy them, it is less likely that this kind of preference is caused by new information. People trade as soon as they get some new information until the information is reflected by the price. It is impossible that people are still trading and the stock price is still changing because of an event a month ago. Consequently, if people tend to buy stocks which stand out a lot in that quarter, this purchase tend to be caused by people’s higher familiarity to the stocks at that time period.
This thesis also uses sort methodology to measure the difference in the stock holdings allocation change among individual and institutional investors for stocks that are catching
11
attentions frequently and “quiet stocks”.
(1) Volume sorts
During the quarter when the stocks experience abnormally large trading volume, either caused by more news coverage or for no reason, individual investors will surely pay more attention to them. This thesis studies how the change in stock holding allocation different for stocks with different abnormal trading volume rate (the ratio of the stock’s trading volume that quarter to its average trading volume over the previous one year (i.e.,4 quarters)). Following the method in Barber and Odean (2008), abnormal trading volume rate 𝐴𝑉𝑖𝑡 is defined with the following formula:
𝐴𝑣𝑜𝑙𝑖𝑡=𝑣𝑜𝑙𝑖𝑡 𝑣𝑜𝑙𝑖𝑡
Where 𝑣𝑜𝑙𝑖𝑡 is the trading volume for stock i in quarter t.
𝑣𝑜𝑙𝑖𝑡=𝑣𝑜𝑙𝑖,𝑡−1+ 𝑣𝑜𝑙𝑖,𝑡−2+ 𝑣𝑜𝑙4 𝑖,𝑡−3+ 𝑣𝑜𝑙𝑖,𝑡−4
For each quarter, stocks are sorted into ten deciles according to their abnormal trading volume rate (𝐴𝑣𝑜𝑙𝑖𝑡). For each decile, the mean change in percentage of institutional investors’ stock holdings (i.e., the mean of Dstake_inst𝑖𝑡) is calculated. To further avoid the bias caused by outliers, in addition to the mean of Dstake_inst, the thesis uses the median value of Dstake_inst series for each decile.
(2) Return sorts
Caused by some news or not, extreme stock returns can also attract much attention from individual investors. However, as not all news can be recognized as positive or negative and people have different opinion upon different information, the impact of news in stock return can be weaker compared to that in trading volume. The thesis is studying the long term impact of standing out. Individual investors tend to buy stocks with extreme stock returns, positive returns or even negative ones, as long as the extent to which the stock returns are salient is
12
large enough to attract attention from individual investors and make the investors more familiar with the stocks during the time period. The thesis sort stocks based on quarterly returns and calculate the average change in percentage of institutional investors’ stock holdings for each decile of each quarter. In Barber and Odean (2008), they calculate imbalances for the day following the extreme returns instead of the same day. But here, as a longer time period is studied, using the data of the following quarter is inappropriate. It is because people’s attention is attracted by new events continuously. They are unlikely to keep attracted by some event a quarter ago while disregarding recent events. So the average change in percentage of institutional investors’ stock holdings of the exactly same quarter as abnormal stock returns is used.
For each quarter t, all stocks are sorted into 10 deciles according to their abnormal return. Similar to sorts based on abnormal volume, abnormal return for stock i on quarter t, 𝐴𝑟𝑒𝑡𝑖𝑡 is defined as
𝐴𝑟𝑒𝑡𝑖𝑡=𝑟𝑒𝑡𝑖𝑡 𝑟𝑒𝑡̅̅̅̅̅̅ 𝑖𝑡 Where 𝑟𝑒𝑡𝑖𝑡 is the stock return for stock i in quarter t.
𝑟𝑒𝑡𝑖𝑡 =𝑟𝑒𝑡𝑖,𝑡−1+ 𝑟𝑒𝑡𝑖,𝑡−2+ 𝑟𝑒𝑡4 𝑖,𝑡−3+ 𝑟𝑒𝑡𝑖,𝑡−4
For each quarter, stocks are sorted into ten deciles according to their abnormal stock return (𝐴𝑟𝑒𝑡𝑖𝑡). For each decile, the mean change in percentage of institutional investors’ stock holdings (i.e., the mean of Dstake_inst𝑖𝑡) is calculated. Again, to further avoid the bias caused by outliers, in addition to the mean of Dstake_inst, the median value of Dstake_inst series for each decile is used.
C. Regression Methodology
Compared to Barber and Odean (2008), however, individual investors’ tendency to buy stocks that attract attention more in a longer time period is not clearly shown in the sort table or paragraph. There are two reasons. Standing out influences people’s trading behavior through
13
two ways. Firstly, people get some information from the events. For example, abnormal trading volume of a stock may be caused by the news coverage. Individual investors hold different opinion on the same news and trade in different directions. Even the very existence of abnormal trading volume makes people feel that the stock is kind of “hot” in that period and people buy the stocks because they want to follow the trend. Secondly, the very existence of the events attracts attention from individual investors. The abnormal trading volume make the stock stand out among thousands of stocks and appear in front of the public. Barber and Odean (2008) study people’s immediate trading response to these events. Because the effect through either of the two ways can happen immediately, especially the first one, that paper cannot distinguish them. This thesis tries to find out whether the effect of second way really exists. In other words, whether repeatedly attracted attention strengthens individual investors’ familiarity to a stock and thus attracts more individual investors. A longer term study is thus more appropriate here, getting rid of the impact of information from events. However, this part of impact is significant. So after excluding it from the study, the effect of events is less significant. The second reason is that it is immediate trading response that grabs most part of the effect, not only the effect of information part, but also part of the effect of the very existence of events. For a longer time, all kinds of characters can have an effect on the result. There is much more noise for a longer time. So, in addition to sort, regression method is also used to do the research. Specifically, the independent variable is decile (1 represents the lowest abnormal trading volume rate or lowest abnormal return while 10 represents the highest ones.). The dependent variable is the average value and median value of change of percentage of institutional investors’ stock holdings (Dstake_inst𝑖𝑡).
V. Results A. Sorts
(1) General sorts:
14
Figure 1: Institutional ownership change for stocks general sort by abnormal trading volume The figure shows the general relationship between deciles and the mean value or median value of Dstake_inst𝑖𝑡 in each decile. This first sort disregards the time and company, is only a general relationship between deciles sorted by abnormal trading volume and change in percentage of institutional ownership. The thesis gets the general level by calculating the mean or median for each time series of a decile.
Figure 2: Institutional ownership change for stocks general sort by abnormal stock return The figure shows the general relationship between deciles and the mean value or median value of Dstake_inst𝑖𝑡 in each decile. This first sort disregards the time and company, is only a general relationship between deciles sorted by abnormal stock return and change in percentage of institutional ownership. The thesis gets the general level by calculating the mean or median for each time series of a decile.
0.0000% 0.1000% 0.2000% 0.3000% 0.4000% 0.5000% 0.6000% 0.7000% 1 2 3 4 5 6 7 8 9 10 C ha ng e i n pe rc e nt ag e of ins ti tut ion al o w n e rs h ip mean-by volume median-by volume
15
We can see that in the first figure, there is not any obvious relationship between the mean of change in percentage of institutional ownership and the deciles. The deviation is large. There is a slightly negative relationship between the median value of change in percentage of institutional ownership and deciles. The second method gets rid of the bias from outliers. In the second figure, no matter in which method, there is a strongly relationship between these two variables, change in percentage of institutional ownership and deciles.
(2) Volume sorts:
The stocks of each quarter are sorted into ten deciles according to their abnormal trading volume rate. Table 1 and Figure 3 show the general relationship between deciles and the mean value or median value of Dstake_inst𝑖𝑡 in each decile. The mean or median for each time series of a decile is calculated to get the general relationship. Accordingly, there are four combinations of the sorting: the mean value of Dstake_inst𝑖𝑡 (mean inst-by volume) with the mean value of “mean inst-by volume” (mmean inst-by volume (1)), the mean value of Dstake_inst𝑖𝑡 (mean inst-by volume) with the median value of “mean inst-by volume” (mdmean inst-by volume (2)), the median value of Dstake_inst𝑖𝑡 (median inst-by volume) with the mean value of “median inst-by volume” (mmedian inst-by volume (3)), the median value of Dstake_inst𝑖𝑡 (median inst-by volume) with the median value of median inst-by volume (mdmedian inst-by volume (4)).
-0.6000% -0.4000% -0.2000% 0.0000% 0.2000% 0.4000% 0.6000% 0.8000% 1.0000% 1.2000% 1 2 3 4 5 6 7 8 9 10 C ha ng e i n pe rc e nt ag e of ins ti tut iona l ow ne rs hi p mean-by return median-by return
16
Table 1: Institutional ownership change for stocks sorted by abnormal trading volume
Stocks are sorted quarterly into deciles based on the abnormal trading volume in that quarter. Abnormal trading volume is calculated as the ratio of current quarter’s trading volume (the sum of three monthly trading volumes in the quarter, monthly trading volume is reported in the CRSP) divided by the average trading volume over the previous 4 quarters. The change in percentage of institutional stock holdings is reported quarterly. For each quarter each decile, I calculate the median or mean of the change in percentage of institutional stock hold ings (mean inst-by volume or median inst-by volume). The table reports the mean or median for each time series for a decile. Four combinations of the sorting: the mean value of Dstake_inst𝑖𝑡 (mean inst-by volume) with the mean value of “mean inst-by volume” (mmean inst-by volume (1)), the mean value of Dstake_inst𝑖𝑡 (mean inst-by volume) with the median value of “mean inst-by volume” (mdmean inst-by volume (2)), the median value of Dstake_inst𝑖𝑡 (median inst-by volume) with the mean value of “median inst-by volume” (mmedian inst-by volume (3)), the median value of Dstake_inst𝑖𝑡 (median inst-by volume) with the median value of “median inst-by volume” (mdmedian inst-by volume (4)).
decile (1) (2) (3) (4) Observations 1 0.1040% 0.4709% 0.1136% 0.3272% 2,852 2 0.4022% 0.4838% 0.3463% 0.3178% 2,859 3 0.5537% 0.5657% 0.4755% 0.4472% 2,860 4 0.4298% 0.4551% 0.3813% 0.3388% 2,862 5 0.3938% 0.3722% 0.3468% 0.4129% 2,856 6 0.3489% 0.0157% 0.3220% 0.0995% 2,854 7 0.3908% 0.5161% 0.2759% 0.3365% 2,859 8 0.3441% 0.2022% 0.2121% 0.0256% 2,846 9 0.2194% 0.2018% 0.1640% 0.0635% 2,838 10 0.3952% 0.3513% 0.2612% 0.0537% 2,776
Figure 3: Institutional ownership change for stocks sorted by abnormal trading volume
The figure shows the general relationship between deciles and the mean value or median value of Dstake_inst𝑖𝑡 in each decile. I get the general relationship by calculating the mean or median for each time series of a decile.
17
We can see from the sort table and figure that, although with large deviation, the change in percentage shows slightly downward trend no matter in which way the general level is calculated (mean or median). There is slightly negative relationship between abnormal trading volume and institutional ownership change.
(3) Return sorts:
The stocks of each time period are sorted into ten deciles based on their abnormal stock return. Table 2 and Figure 4 present the general relationship between deciles and the mean value or median value of Dstake_inst𝑖𝑡in each decile. Similar to volume sort, I get the general relationship by calculating the mean or median for each time series of a decile. Again, there are four combinations of the sorting: the mean value of Dstake_inst𝑖𝑡 (mean inst-by return) with the mean value of “mean inst-by return” (mmean inst-by return (1)), the mean value of Dstake_inst𝑖𝑡 (mean inst-by return) with the median value of “mean inst-by return” (mdmean inst-by return (2)), the median value of Dstake_inst𝑖𝑡 (median inst-by return) with the mean value of “median inst-by return” (mmedian inst-by return (3)), the median value of Dstake_inst𝑖𝑡 (median inst-by return) with the median value of “median inst-by return” (mdmedian inst-by volume (4))
0.0000% 0.1000% 0.2000% 0.3000% 0.4000% 0.5000% 0.6000% 1 2 3 4 5 6 7 8 9 10 C ha ng e i n pe rc e nt ag e of ins ti tut iona l ow ne rs hi
p mmean inst-by volume
mdmean inst-by volume mmedian inst-by volume mdmedian inst-by volume
18
Table 2: Institutional ownership change for stocks sorted by abnormal stock return
Stocks are sorted quarterly into deciles based on the abnormal stock return in that quarter. Abnormal stock return is calculated as the ratio of current quarter ’s stock return (the sum of three monthly stock return in the quarter, monthly stock return is reported in the CRSP) divided by the average stock return over the previous 4 quarters. The change in percentage of institutional stock holdings is reported quarterly. Four combinations of the sort: the mean value of Dstake_inst𝑖𝑡 (mean inst-by return) with the mean value of “mean inst-by return” (mmean inst-by return (1)), the mean value of Dstake_inst𝑖𝑡 (mean inst-by return) with the median value of “mean inst-by return” (mdmean inst-by return (2)), the median value of Dstake_inst𝑖𝑡 (median inst-by return) with the mean value of “median inst-by return” (mmedian inst-by return (3)), the median value of Dstake_inst𝑖𝑡 (median inst-by return) with the median value of “median inst-by return” (mdmedian inst-by return (4)).
decile (1) (2) (3) (4) Observations 1 0.0509% 0.3521% 0.0711% 0.0290% 2,848 2 -0.3387% -0.3223% -0.3004% -0.2371% 2,846 3 0.0789% -0.0179% 0.0392% -0.0072% 2,845 4 0.1266% 0.2226% 0.0267% 0.0642% 2,846 5 0.2226% 0.3183% 0.1482% 0.1746% 2,845 6 0.4632% 0.4585% 0.3472% 0.2866% 2,847 7 0.5552% 0.4719% 0.4723% 0.3150% 2,846 8 0.6796% 0.8507% 0.5881% 0.6013% 2,845 9 0.8481% 0.8511% 0.7787% 0.7125% 2,846 10 0.8961% 0.8349% 0.7763% 0.6589% 2,848
Figure 4: Institutional ownership change for stocks sorted by abnormal stock return
The figure shows the general relationship between deciles and the mean value or median value of Dstake_inst𝑖𝑡 in each decile. I got the general relationship by calculating the mean or median for each time series of a decile.
19
This time, we can see that there is a strongly positive relationship between abnormal stock return and institutional ownership change.
(4) Negative return sorts
People tend to buy stocks that have experienced abnormally high stock return historically because they suppose these stocks are more likely to have high stock return in the future. However, as what this thesis studies is the influence of the very existence of standing out on individual investors’ trading behavior, it does not matter whether the stock return is positive or negative. In other words, individual investors are even more likely to buy stocks that experience abnormally poor stock return than to buy stocks that are “quiet” if the third hypothesis is true. So this thesis also analyses the impact of negative stock returns in institutional ownership change. Specifically, the observations with positive stock returns are excluded. Data with positive abnormal return rate is also deleted because sudden decrease in stock return from positive ones to negative ones are more likely to attract individual investors’ attention. Also, although individual investors tend to pay attention to this kind of stocks and buy them, as soon as they find that one stock is experiencing negative and decreasing stock return all the time, no one would still hold it. After all, no one likes losing money all the time. So this thesis just studies the impact of sudden decrease in stock return. After deleting the observations with positive stock returns or positive abnormal stock return, I sort the stocks of
-0.6000% -0.4000% -0.2000% 0.0000% 0.2000% 0.4000% 0.6000% 0.8000% 1.0000% 1 2 3 4 5 6 7 8 9 10 C ha ng e i n pe rc e nt ag e of ins ti tut iona l ow ne rs hi
p mmean inst-by return
mdmean inst-by return mmedian inst-by return mdmedian inst-by volume
20
each time period into ten deciles based on their abnormal stock return. Table 3 and Figure5 present the general relationship between deciles and the mean value or median value of Dstake_inst𝑖𝑡in each decile. Similar to volume sort, the mean or median for each time series of a decile is calculated. The four combinations of the sort: the mean value of Dstake_inst𝑖𝑡 (mean inst-by nreturn) with the mean value of “mean inst-by nreturn” (mmean inst-by nreturn (1)), the mean value of Dstake_inst𝑖𝑡 (mean inst-by nreturn) with the median value of “mean inst-by nreturn” (mdmean inst-by nreturn (2)), the median value of Dstake_inst𝑖𝑡 (median inst-by nreturn) with the mean value of “median inst-by nreturn” (mmedian inst-by nreturn (3)), the median value of Dstake_inst𝑖𝑡 (median inst-by nreturn) with the median value of “median inst-by nreturn” (mdmedian inst-by nreturn (4)).
Table 3: Institutional ownership change for stocks sorted by abnormal negative stock return Stocks are sorted quarterly into deciles based on the negative abnormal stock return in that quarter. Abnormal stock return is calculated as the ratio of current quarter ’s stock return (the sum of three monthly stock return in the quarter, monthly stock return is reported in the CRSP) divided by the average stock return over the previous 4 quarters. The observations with negative stock returns only include observations with negative stock return and negative abnormal stock return rate. The change in percentage of institutional stock holdings is reported quarterly. For each quarter each decile, I calculate the median or mean of the change in percentage of institutional stock holdings (rmean inst or rmedian inst). The table reports the mean or median for each time series for a decile. Four combinations of the sort: the mean value of Dstake_inst𝑖𝑡 (mean inst-by nreturn) with the mean value of “mean inst-by nreturn” (mmean inst-by nreturn (1)), the mean value of Dstake_inst𝑖𝑡 (mean inst-by nreturn) with the median value of “mean inst-by nreturn” (mdmean inst-by nreturn (2)), the median value of Dstake_inst𝑖𝑡 (median inst-by nreturn) with the mean value of “median inst-by nreturn” (mmedian inst-by nreturn (3)), the median value of Dstake_inst𝑖𝑡 (median inst-by nreturn) with the median value of “median inst-by nreturn” (mdmedian inst-by nreturn (4)).
21 decile (1) (2) (3) (4) Observations 1 -0.2672% -0.2241% -0.2845% -0.3739% 663 2 -0.6120% -0.3601% -0.3951% -0.2307% 661 3 -0.2584% 0.0108% -0.2694% 0.0000% 662 4 -0.6534% -0.6350% -0.5977% -0.3620% 662 5 -0.2773% -0.2220% -0.3260% -0.2207% 661 6 -0.3057% -0.4925% -0.2365% -0.3127% 661 7 -0.1984% -0.3394% -0.2580% -0.2117% 663 8 0.0398% -0.1411% -0.0445% -0.1179% 661 9 0.1647% 0.3423% 0.0764% 0.1358% 662 10 0.1158% 0.1240% 0.0612% -0.0027% 663
Figure 5: Institutional ownership change for stocks sorted by abnormal negative stock return The figure shows the general relationship between deciles and the mean value or median value of Dstake_inst𝑖𝑡 in each decile. I get the general relationship by calculating the mean or median for each time series of a decile.
From the table and figure, we can see that there is a positive relationship between abnormal negative stock return and institutional ownership change.
-0.8000% -0.6000% -0.4000% -0.2000% 0.0000% 0.2000% 0.4000% 1 2 3 4 5 6 7 8 9 10 C ha ng e i n pe rc e nt ag e of ins ti tut iona l ow ne rs hi
p mmean inst-by nreturn
mdmean inst-by nreturn mmedian inst-by nreturn mdmedian inst-by nreturn
22
B. Regression:
To further prove that the negative and positive relationships do exist and are significant, regression method is used to analyze the data. The independent variable is decile (1 represents the lowest abnormal trading volume rate or lowest abnormal return while 10 represents the highest ones.). The dependent variable is the average change of percentage of institutional investors’ stock holdings. I do the panel data regression for both volume sort decile and stock return sort decile with time fixed effect. The time frequency, as mentioned before, is quarterly.
(1) Volume regression
The independent variable is decile based on abnormal trading volume sort. The dependent variable is the mean or median of change in percentage of institutional investors’ stock holdings. As the coefficient of decile is negative, the higher decile, the more abnormal trading volume, the lower the average change of percentage of institutional investors’ stock holdings. This means that the more abnormal trading volume, the higher the percentage of individual ownership of stocks.
(2) Stock return regression
The independent variable is decile based on abnormal stock return sort. The dependent variable is the mean of change in percentage of institutional investors’ stock holdings. As the coefficient of decile is negative, the higher decile, the more abnormal stock return, the lower the average change of percentage of institutional investors’ stock holdings. This means that the more abnormal stock return, the higher the percentage of individual ownership of stocks.
(3) Negative stock return regression
Specifically, observations with positive stock returns are excluded. Data with positive abnormal return rate is also excluded because sudden decrease in stock return from positive ones to negative ones are more likely to attract individual investors’ attention. Also, although individual investors may pay attention to this kind of stocks and buy them, as soon as they find that one stock is experiencing negative and decreasing stock return all the time, no one
23
would still hold it. After all, no one likes losing money all the time. So I just study the impact of sudden decrease in stock return. Table 4 summarizes the result of the three regressions. We can see that the coefficients are all statistically significant in 1% level. As abnormal trading volume increases, abnormal stock return, overall or only negative one, decreases, change in percentage of institutional ownership decreases. The three hypothesis are strongly proved statistically.
24
Table 4: Results of the regression of institutional ownership change
Stocks are sorted quarterly into deciles by the abnormal trading volume (decile-by volume), abnormal stock return (decile-by return) and negative abnormal stock return (decile-by nreturn). The observations with negative stock returns only include observations with negative stock return and negative abnormal stock return rate. The change in percentage of institutional stock holdings is reported quarterly. For each quarter each decile, I calculate the median or mean of the change in percentage of institutional ownership (mean inst or median inst). The standard errors are clustered by institution and shown in parentheses.
variables Abnormal trading volume Abnormal stock return Negative Abnormal stock return
meanv inst medianv inst meanr inst medianr inst meannr inst mediannr inst
decile-by volume -0.00005** -0.00016*** (0.00021) (0.00002) decile-by return 0.00135*** 0.00122*** (0.00002) (0.00002) decile-by nreturn 0.00097*** 0.00078*** (0.00006) (0.00006) constant 0.00387*** 0.00379*** -0.00386*** -0.00374*** -0.00756*** -0.00654*** (0.00011) (0.00009) (0.00013) (0.00009) (0.00040) (0.00038) observations 28,462 28,462 28,462 28,462 6,619 6,619 Adj R-squared 0.8834 0.9040 0.8913 0.9209 0.6881 0.7041
25 VI. Discussion
When the data is sorted based on abnormal trading volume , the relationship between decile and the change in percentage of institutional stock holdings is negative. This fact is shown not only in sort table and figure but also in negative coefficient. This means that, as decile changes from 1 to 10, the abnormal trading volume grows, institutional investors tend to hold less of the stocks, individual investors tend to hold more of the stocks. The coefficient between decile and the change in percentage of institutional stock holdings is not only negative but also statistically significant, which means that my first hypothesis is strongly proved. The more a stock experiences abnormal stock return, the less institutional holding of the stock is, the more individual investors tend to hold the stock.
When the data is sorted based on abnormal stock return, the relationship between decile and the change in percentage of institutional stock holdings is positive. This fact, again, is shown strongly not only in sort table and figure but also in negative coefficient. As decile changes from 1 to 10, the abnormal stock return grows, institutional investors tend to hold more of the stocks, individual investors tend to hold less of the stocks. The coefficient between decile and the change in percentage of institutional stock holdings is not only positive but also statistically significant. This result, however, is completely opposite to the second hypothesis. This is mainly because the study is using longer term data. In a short term, the stock return can fluctuate because of some noise. In the longer term, however, the stock return is still a good reflection of how the firm is performing. So, compared to individual investors, institutional investors hold more when the stocks are good ones. The fact that in a longer term, the stock return is a good measurement of the stocks’ performance is also proved in previous researches. We can see that a lot of researches try to prove irrational trading behavior and bias in stock price by showing disappointing stock return in a later time. As a result, the strongly positive relationship between percentage of institutional holding and stock return and negative relationship between percentage of individual holding and stock return is a good proof of the rationality of institutional investors.
26
decile and the change in percentage of institutional stock holdings is positive. This fact is shown not only in sort table and figure but also in positive coefficient. This means that, as decile changes from 1 to 10, the negative abnormal trading volume becomes smaller, institutional investors tend to hold more of the stocks, individual investors tend to hold less of the stocks. We can also see that, the coefficient between decile and the change in percentage of institutional stock holdings is not only positive but also statistically significant, which means that my third hypothesis is strongly proved. The fact that individual investors hold more of stocks with smaller negative stock return is obvious. As shown above, stock return is a good measure in a longer term. What is surprising is that individual investors hold more of stocks with larger negative stock return. This fact has strongly proved the third hypothesis. The more a stock experiences abnormal stock return (negative ones), the more it stands out, the less institutional holding of the stock is, the more individual investors tend to hold the stock. However, based on previous research, there are other characters of individual investors that can lead to this kind of result. For example, individual investors tend to hold losers and sell winners (Shefrin and Statman (1985), Odean (1998)).
VII. Conclusion
Barber and Odean (2008) study the impact of attention in individual investors’ trading behavior by analyzing their immediate response to attention-catching events in terms of news, abnormal trading volume and abnormal stock return. However, people also trade immediately in response to information they get from the events. In addition, people hold different opinion about the same events. Accordingly, as long as there is something happening about a stock, people tend to do some trading on the stock. This kind of attention is more likely caused by the information brought by the events and people’s different opinion on the information. This thesis tries to find out the influence of a slightly different kind of attention, or more accurately, familiarity. When searching among thousands of publicly traded stocks, individual investors tend to pick up stocks that they are more familiar with. In most cases, they themselves even cannot explain why they pick up those stocks. Familiarity, actually, is caused by repeated attention paid to the stock. It is not hard to understand. In our daily life, if we could see some people frequently in one time period, we are more likely to remember their name, appearance
27
or even habit. As for people we hardly see in a period, although we can be familiar about them during a period in the past, we could forget them gradually as time goes by. This kind of mechanism also applies to financial market. In time period during which the stocks grab attention more from people, such as more abnormal trading volume, more news, more abnormal stock returns, individual investors, with limited time and resources, tend to buy them. During that time period, these stocks stand out more often from thousands of stocks and thus, are more familiar to individual investors.
In this thesis, as the mechanism is in a longer term, the frequency of the data used is quarterly. Similar to the method in Barber and Odean (2008), the data in each quarter is firstly sorted into deciles based on abnormal trading volume and abnormal stock returns. Next, the thesis calculates the average change in percentage of institutional stock holdings for each decile in each quarter. In addition to sort method, panel data regression is used with deciles and the average change in percentage of institutional stock holdings.
Hypothesis 1 and hypothesis 3 are both strongly proved. The result of the test for hypothesis 2 is opposite to my guess. It is probably because this study is a long term one and in a long term, stock return is a good reflection of stocks’ performance. In addition, the result of test for hypothesis 3 could also be caused by other factors, such as disposition effect (Shefrin and Statman (1985), Odean (1998)). Future research could try to find a better way to exclude this kind of bias. But generally, people tend to buy stocks that stand out more in a time period. This is not only reasonable in theory but also proved statistically to some extent.
This research is not only meaningful statistically, but also, more importantly, helpful in exploring better financial market supervision. If, no matter what the events are, people tend to buy stocks with more trading volume and more extreme stock returns, firms may try to manipulate the financial market and increase stock price by making things happen, trying to draw attention from the public as much as possible, making individual investors more familiar with their stocks and thus attracting more individual investors. In that case, the stock prices in financial market would be strongly biased and the whole market would not be efficient. In
28
addition, this research is also meaningful for establishing a better model to analyze characters that can influence the financial market. In a standard model, too many assumptions are unrealistic, which makes the result of analysis make no sense in the real world. Taking into people’s, especially individual investors’ irrationality is a big move in improving the research model. I believe that, in the future, behavior finance will become increasingly important for analysis in financial market.
29 References:
Barber, Brad M., and Terrance Odean, 2008, All that glitters: The effect of attention and news on the buying behavior of individual and institutional investors, Review of Financial Studies 21, 785-818.
Barberis, Nicholas, and Richard Thaler, 2003, A survey of behavioral finance, Handbook of
the Economics of Finance 1, 1053-1128.
Baker, Malcolm, Xin Pan, and Jeffrey Wurgler, 2012, The effect of reference point prices on mergers and acquisitions, Journal of Financial Economics 106(1): 49-71.
Bollen J, Mao H, and Zeng X, 2011, Twitter mood predicts the stock market, Journal of
Computational Science 2, 1-8.
Chemmanur, Thomas, and An Yan, 2009, Advertising, attention, and stock returns, Working paper, Fordham University.
DellaVigna, S., and Pollet, J., 2006, Investor Inattention, Firm Reaction, and Friday Earnings Announcements, Working Paper.
Engelberg, Joseph, and Christopher A. Parsons,2011, The causal impact of media in financial markets, The Journal of Finance 66, 67-97.
Gabaix, Xavier, David Isaac Laibson, Guillermo Moloche and Weinberg Stephen, 2003, The allocation of attention: theory and evedence, Munich Personal RePEc Archive Paper.
Gervais, Simon, Ron Kaniel, and Dan H. Mingelgrin, 2001, The high-volume return premium,
The Journal of Finance 56(3): 877-919.
Grullon, Gustavo, George Kanatas, and James P. Weston, 2004, Advertising, breadth of ownership, and liquidity, Review of Financial Studies 17, 439-61
Hirshleifer, David, 2001, Investor psychology and asset pricing, The Journal of Finance 56(4): 1533-1597.
Lee, Charles M.C., 1992, Earnings news and small traders, Journal of Accounting and
Economics 15, 265-302.
Lou, Dong, 2009, Attracting investor attention through advertising, Working paper, London School of Economics.
Miller, Edward M., 1977, Risk, uncertainty, and divergence of opinion, Journal of Finance 32: 1151-68.
30
Odean, Terrance, 1998, Do investors trade too much? Available at SSRN 94143
Odean, Terrance, 1998, Are investors reluctant to realize their losses? The Journal of finance 53(5): 1775-1798.
Peress, Joel, 2011, The media and the diffusion of information in financial markets: Evidence from newspaper strikes, Working paper, INSEAD.
Seasholes, Mark, and Guojun Wu, 2004, Profiting from predictability: Smart traders, daily price limits, and investor attention, Working paper, University of California, Berkeley.
Shefrin, Hersh and Meir Statman, 1985, The disposition to sell winners too early and ride losers too long: Theory and evidence, The Journal of finance 40(3): 777-790.
Van Rooij M, Lusardi A, and Alessie R, 2011, Financial literacy and stock market participation, Journal of Financial Economics 101, 449-472
