Reactions of the Institutional and Individual Investors in the Volatile Stock Market

Reactions of the Institutional and Individual Investors in the Volatile Stock Market

1 Brief introduction of this paper
This paper just discusses the reactions of the institutional and individual investors in the volatile stock market. They are interested in who sells during large market swings. The authors address this question by investigating the relationship between the ownership structure and returns of firms on days when the absolute value of the market's return is two percent or more. They find that a firm's abnormal return on these days is related to the percentage of institutional ownership, that there is abnormally high turnover in the firm's shares on these days, and that this abnormal turnover is significantly related to the percentage of institutional ownership in the firm. Taken together, they come to a conclusion that institutions herd together and trade with the momentum of the market on days when there are large moves in the stock market, particularly mutual and pension funds.
2 Discussion about the research problem
2.1 The propose of the question
The authors asked a very simple question at the beginning of the paper: who sells when there is a large drop in the stock market, institutional or individual investors? The arguments take divergent paths when answering this question. One path explains that individuals are less sophisticated and more risk averse than institutions, so individual is the one who sells during a sharp market drop. The other path explains that institutions have short horizons and sell during large market decline. In this paper, the authors tried to identify who on earth sells or buys during market swings based on empirical evidences.
2.2 Theoretical background-Herding Theory
The issue discussed in this paper is closely related to herding behavior, a phenomenon which is frequently mentioned in behavioral finance. Herding is a group of investors trading in the same direction over a period of time. A growing literature has been exploring herding behavior in the stock market since the 1990s’.
There are several theoretical models interpreting institution managers’ herding. For example, managers may rationally choose to focus only on information that pays off in the short run (Froot, Scharfstein, and Stein (1992)); they may herd for reputation concerns (Scharfsein and Stein (1990)); they may herd if the information they have revealed sequentially (Banerjee (1992)).
2.3 The focus of this research
Though extant studies have explored a lot about both individual and institutional investors’ herding in the stock market, this paper adds to the literature by focusing on institutions’ herding behavior during extremely volatile market (defined as the overall movement of market exceeds two percent in one day). The basic premise is that institutional shareholders react strongly to large market price changes by herding together and moving prices. Though this idea is not theoretically and empirically new, the authors test the premise in a different way from prior studies. Unlike prior studies using quarterly or annual data, it used daily data for days of sharp market swing, and also observed the half-year post-event performance for all firms.
2.4 Importance of the problem
The answer to the question raised in this paper is important because it helps us to understand the dynamics associated with large swings in stock prices as well as the sources of market volatility, in other words, which type of the individual and institutional investors contributes more to the market volatility.
3 Research Method of this paper
3.1 the methodology—Empirical method
This paper just uses the empirical method to do the study. After a brief review of the theoretical basis and empirical evidence for institutional herding the authors test this premise in a very different way from prior studies. The methodology of this paper is sound. As we know, the empirical method is generally characterized by the collection of a large amount of data before much speculation as to their significance, or without much idea of what to expect, and is to be contrasted with more theoretical methods in which the collection of empirical data is guided largely by preliminary theoretical exploration of what to expect. Just as the authors said at first:”Who sells when there is a large drop in the stock market, institutional or individual investors? The answer is not clear.”So in this study problem, the empirical method is the best choice.
3.2 the research design
3.2.1 the hypothesis
Based on the theoretical analysis and empirical evidence, the authors proposed the idea that institutions herd together and trade with the momentum of the market on days when there are large moves in the stock market. So they stated their hypothesis in the empirical parts, that is: Institutional investors react to large market price swings. The authors told us if this hypothesis is tested all right, we will see larger price movements for stocks with significant institutional ownership. So the authors use the regression model to test whether the cross-sectional distribution of the returns of individual firms on event-days is a function of the level of institutional ownership. After the empirical process, although the regression results are supportive of the hypothesis, the result may also be caused by other factor—the volume of trade. So the authors propose another hypothesis that the cross-sectional distribution of abnormal turnover on the event-days is related to both the level and composition of a firm's institutional shareholder base. At last, the model (2)’ results are consistent with it so the authors’ idea is tested all right and gets robust.
3.2.2 Data
For brevity, we will skip the data range and its resources, and only discuss the event-day-choosing criteria.
The authors chose days when absolute value of market return is greater than two percent, in both equal- and value-weighted method, as volatile market event-days. There are two issues raised in this criteria- the cutoff of market return, and the calculate method for market return.
Firstly, the two percent cutoff is not arbitrary. The authors calculated the mean and standard deviation of daily returns for the CRSP equal- and value-weighted NYSE/AMEX/Nasdaq portfolios during the research period and select days that are roughly three standard deviations above or below the mean. While the choice of what constitutes an extreme market price change day is somewhat arbitrary, the results obtain for days when the return is two standard deviations above or below the research period mean.
Secondly, both equal- and value-weighted samples have the potential issue of outliers. When using equal-weighted sample, large positive or negative returns for several small firms could produce an extreme portfolio return when the majority of firms have returns of the opposite sign. Similarly, when using the value-weighted sample, large positive or negative returns for several large firms could generate a large portfolio return, and the market price change does not reflect a broad market shift. In both cases, the authors calculated the ratio of firms with negative (positive) returns to those with positive (negative) returns. The results indicate that the samples can reflect the whole direction of market movement, and the movement is not outlier driven.
3.2.3 Empirical process –return
In the empirical part, in order to find the relationship between returns of individual firms and the level of institutional ownership, the authors at fist discuss the univariate results and then the regression results. The authors technically calculated the minimum, first quartile, median, mean, third quartile, maximum, and standard deviation of each independent variable. They also partitioned firms into high and low institutional ownership subsamples so to facilitate afterwards comparison. In the descriptive statistics analysis, they mainly focus on the io and return variables in table II. They find the return difference and clustering of returns for portfolios with high institutional ownership are supportive of the hypothesis. In order to get a stronger conclusive result, the authors constructed a model using the approach of Fama and MacBeth to get the regression evidence. They included five independent variables in the model and gave the reasons of choosing them. Then they decomposed institutional ownership into four different types and investigated the relationship deeply. They analyze the regression results and give the explanation. Then they presented another explanation for the result and gave their analysis. However, we think their analysis is not strong and reasonable. They just used the result of another paper and compared the difference between mutual fund and pension fund. They didn’t improve their model or explore their prior data further. In order to make the result robust statistically, they implied the robustness tests. After the test the result has been more convincing.
3.2.4 Empirical process –turnover
The authors tested the relationship between abnormal turnover and institutional ownership on the event-days in the forth part of the paper, using very similar steps as in part III. In addition to descriptive statistic, they performed a T-test in the univariate comparison part and showed there was significant difference of abnormal turnover between high IO firms and the low IO ones. This is supportive for the hypothesis, and for further conclusive evidence, the authors still use regression to address in the possibility that difference in abnormal turnover is the result of a concentration of larger firms in the high institutional ownership portfolio. In the regression, the authors used variance and firm size as controls, while institutional ownership is the essential independent variable. They also estimate the model with institutional ownership decomposed in to four sub categories by type (mutual funds, pension funds and endowments, insurance companies, and banks). The regression results suggested that the cross-sectional distribution of abnormal turnover on the event-days is related to both level and composition of a firm’s institutional shareholder base. In the robustness test, the authors use the same orthogonalization process as in part III to remove the impact of correlation between independent variables, they also estimated the model with alternate specifications of the independent variables (namely, substitute market value of equity with book value of assets, or substitute variance with total return variance). The robustness test draw quantitatively and qualitatively similar conclusion as the regression part.
3.2.5 Postevent Performance testing
The idea that institutions herd together when there are large moves in the stock market can’t simply and necessarily implies they contribute to market volatility. In order to improve and perfect prior result, the authors compute postevent cumulative returns for six months (125 trading days) starting the day after the event-day for each stock. They analyzed step by step. At first get the raw result, then controlled the probably disturbing facts, such as capitalizations (by constructing the postevent abnormal returns) and overlapping time periods problem( by using only observations that are at least six months apart),etc. After that, they also did the robustness check to make the result strong. Fortunately, the results presented in Panel A weren’t in accord with the prediction and the authors gave the explanation, but it didn’t sound reasonable and convincing.
3.3 The paper’s conclusions
This paper has four main findings:
Firstly, after controlling for risk and liquidity, the percentage of institutional ownership in a firm is inversely related to that firms return on great market decline days and there is similar result in up market days.
Secondly, types of institutional ownership have something to do with the firm’s abnormal return. Ownership by mutual funds, pension funds and endowments is positively (negatively) related to the abnormal return on up (down) days, while ownership by banks has the opposite effect.
Thirdly, the abnormal turnover is positively related to the level of institutional ownership when there is sharp market increase (decrease).
Finally,abnormal returns following the event of a large market drop is positive (negative) for stocks with high (low) levels of institutional ownership. The reason for this pattern may be that institutions are driving the stock prices below their fundamental values on the event day of market decline.
All in all, institutions react more strongly than individuals when sharp market swings take place.
4 The paper’s contribution and limitation
4.1 Contribution
This research contributes to the extant herding studies by answering the question that who contributes to short term market volatility.
4.2 Limitation

1. Not classify the stocks. As we know, the funds share an aversion to stocks that have recently dropped significantly in price. So if looking at subgroups of stocks, we may find differential levels of herding in small and big stocks.
2. In the analysis of the empirical results the authors didn’t compare the data in the up and down market. They just analyzed the relationships among items so they ignored to analyze the institutions’ herding behavior when buying versus selling stocks.
3. The authors just test the postevent performance and didn’t consider the prior conditions of stocks. Maybe there exists different levels of herding in stocks with high versus low past return.

5 Our suggestion for improvement or extension
The research contributes to the literature of herding in the stock market. Yet, there are still some issues to be taken into account according to this study. In the following part of our discussion, we list some of our concerns.
5.1 Observing the herding into industrial level
Recent studies show that herding behavior, to a large extent, happens in industrial level. If there is severe herding in a specific industry, it leads to partial market volatility. Maybe herding is more observable if the authors breaks down the sample into industrial level. If the authors wants to look into a volatile market, partial market volatility is also important.
5.2 Identifying the holding motivation of institutions
The research has a potential assumption that all institutions hold firms’ stock for speculative purpose. This is not always true in real world. Some institutions hold stocks of promising companies as long-term investment. They become the principal shareholder of the companies and maybe even participate in the corporate governance of the companies. These non-speculative institutional ownerships should be identified, because the institutions holding these stocks have different tendency to herd. The authors decomposed the institutional ownership into four subgroups (mutual funds, pension funds and endowments, insurance companies, and banks), and found different herd intention for these four groups. Holding motivation to some extent can explain this finding as holding motivation differs in the four groups, but it is still not the same thing as type of institution.
5.3 Describing the distribution for institutional ownership of all firms
The authors chose the median of IO to divide the sample into two subgroups, while in postevent performance study, using the quintile to divide the sample into five subgroups. This is a little confusing method for sample division, because if we do not know the distribution of IO for the whole sample, we cannot evaluate if the median or quintile is a reasonable cutoff. For instance, if most samples are distributed very close to the median, than using median as a partition of high and low IO is not very compellent.
5.4 Broadening the criteria of defining volatile market
The authors used days of absolute value of market return over two percent as extreme volatile market, but they are extreme cases, with only about thirty days over eight years. What about institutions herding in gentle market swings, namely, days of absolute value change only exceed one and a half percent? This could be another extension for the research.