Herding Behavior in Indonesian Investors

This research attempts to investigate the herding behavior of the companies that invested in IDX LQ45 Index during 2014 through 2016. Herd behavior is the tendency of investors to follow other investors’ actions in the market. LQ45 was chosen as it comprises the most heavily-traded stocks of the Indonesian Stock Exchange. This research used Vector Autoregressive model to determine the effects of size and market return on the herding behavior. The Granger causality test suggests that there are dynamic interactions: (i) between size and herding behavior; and (ii) between market return and herding behavior. In addition, Variance Decomposition and Impulse Response reveal that market capitalization (size) has variable of the greater role in defining herding behavior, compared to that of market return.

This research attempts to investigate the herding behavior of the companies that invested in IDX LQ45 Index during 2014 through 2016. Herd behavior is the tendency of investors to follow other investors' actions in the market. LQ45 was chosen as it comprises the most heavily-traded stocks of the Indonesian Stock Exchange. This research used Vector Autoregressive model to determine the effects of size and market return on the herding behavior. The Granger causality test suggests that there are dynamic interactions: (i) between size and herding behavior; and (ii) between market return and herding behavior. In addition, Variance Decomposition and Impulse Response reveal that market capitalization (size) has variable of the greater role in defining herding behavior, compared to that of market return.   Herd behavior happens when an investor makes an investment decision that follows other investors in the market, rather than through an informed analysis (Szyszka, 2013). Herding investors believe that they follow those with superior information and consequently, they think that they made a less risky decision. This behavior can be labeled as irrational. Furthermore, herding investors may also cause instability in a financial market due to ignorance of important and fundamental information (Baddeley, 2012).
According Chang, Cheng, and Khorana (2000), herd behavior causes stock price fluctuation and mispricing in equity valuation. This is caused by a biased expected return and risk perception, which leads to a significant difference between stock price and its fundamental value. Kremer and Nautz (2013)

Herding Behavior
Herding is a psychological condition where investors ignore their own abilities and beliefs, and choose to follow others without proper contemplation (Devenow & Welch, 1996). Banerjee (1992)  Return is an important aspect of an investment.
The higher the return, the more attractive an investment becomes. When the market is soaring, investors would share the news of their gains; words would spread, and this would trigger herding (Lan & Lai, 2011). Ozsu (2015)

Return
Return is the result generated from an investment.
This result would be considered by other investors in performing investing decisions. Investors would predict future returns by reflecting at past returns.
The following is the formula to calculate return (Jogiyanto, 2000): Composite Index for period t; and IHSG t-1 is Indonesia Composite Index for the period one year prior to t.

Size (Market capitalization)
Market capitalization is generally an important measure for the success of a public corporation.
It is also related to the market demand for a  (Christie & Huang, 1995).
where R i,t is the stock return i for period t and R m,t is the average cross-section return. Chang, Cheng, and Khorana (2000) would then modify equation (1); using CSAD as proxy for herd behavior where R m is the proxy for expected market return, expressed in the following regression: where | R m,t | is the absolute aggregate market return and gR 2 m,t is the square of market return. g signifies the nonlinear relationship between squared market return and CSAD; a negative and significant value indicates herding. Using the crosssectional absolute deviation of returns (CSAD) as the measure of dispersion, it's demonstrate that rational asset pricing models predict not only that equity return dispersions are an increasing function of the market return but also that the relation is linear. If market participants tend to follow aggregate market behavior and ignore their own priors during periods of large average price movements, then the linear and increasing relation between dispersion and market return will no longer hold. Instead, the relation can become non-linearly increasing or even decreasing.
On Eviews, ordinary least squares would be utilized to detect herd behavior. As for describing the dynamic relationship, vector autoregression (VAR) would be used. VAR is a non-theoretical model used to predict a system with time-series variables. It is a model suited for economic models. VAR includes several stages: stationarity testing, lag order determination, cointegration testing, model stability testing, Granger causality testing, variance decomposition, and impulse response function. Table 2  RETURN shows a 0.000375 average, whereas SIZE has a 1.068841 average.

Regression Analysis
The Tables 3 and 4 display the results of crosssectional regression.

b. Lag Order
Maximum lag can be estimated by determining lag structure and AR roots table. Based on roots of characteristic polynomial, maximum lag is obtained from modulus value that is lower than 1.
As shown on Table 6, maximum lag is determined from the values of LR, FPE, AIC, SC, and HQ.
Optimal lag length is determined based on the lowest AIC, which is indicated by the star (*) sign.
There are three lags with (*); they are lag 1, 2, and 4. Lag 4 has the most (*), therefore lag 4 would be used for further analysis. Figure 1 shows that using lag 4, the VAR model is stable (stationer) because the roots possess modulus placed in the circle with values less than one. The greatest modulus value is 0.990203, which is lower than 1.

c. Cointegration test
Cointegration test was conducted to determine whether a relationship exists between variables.   Whereas VAR is stable, whose modulus values are less than one. Table 8 shows the VECM model to be unstable. This is seen from 2 moduli showing the values of 1, as well as the "VEC specification imposes 2 unit root(s)" statement that was generated by the Eviews software.
These tests show VECM to be unstable and VAR to be stable.

e. Granger causality test
Granger test was conducted to determine causal relationship between variables in the VAR model.  Table 9 displays the output of the Granger test.
From Table 9, it can be determined that there are the following: A two-way relationship between RETURN and CSAD, and a two-way relationship between SIZE and CSAD, and a one-way relationship from SIZE to RETURN. Figure 2 shows the relationships mentioned in Table 9.

Variance Decomposition
This analysis is used to establish the percentage of error variance of each variable explained by shocks to the other variables in the VAR model.
Tables 10 through 12 display the results of variance decompositions. Meanwhile, variations in CSAD were also shown to be influenced by SIZE (1% to 7%). As for RETURN, it did not really influence in the variations of CSAD.  Table 9. Granger Causality Test of the Variables Table 12 shows that variations in RETURN were greatly influenced by SIZE at 83-85%. They were also influenced by RETURN itself at 13-14%.
However, CSAD did not really influence the variations in RETURN.

Impulse Response Function (IRF)
This analysis is used to identify the response of   The response was, at the beginning, positive for CSAD. At period 4, the response fell down into the negatives at -0.03% level. It rose and became relatively stable from period 6.

Discussion
Prior to forming the VAR model, the data were deemed stationary, which means that the  These stocks are highly-demanded, which make them highly-priced as well. The higher the price in the future means the higher return for investors.
However, if the price and demand drop, then the return would move in the same direction.
The two-way causation between return and herd behavior may be explained by several factors.
High return would trigger herd behavior when the This would attract other investors to purchase the same instruments with the intention of gaining yield, thus triggering herding behavior among investors. This is consistent with the findings of Chang, Cheng, & Khorana (2000).
Herd behavior also affects market return. The higher the intensity of the herd behavior would result in lower market return. According to Harris (2003), there are informed investors and uninformed investors. Informed investors possess information and technical skills; they herd and speculate to maximize their portfolio. Whereas uninformed investors are those with limited information and resources. The act of herding is carried out by speculators (informed investors) and then followed by uninformed investors. These private investors would trade at a later time, which cause them either not to gain as much return as the speculators, or even lose their money.
Finally, the unidirectional causality from size to herd behavior could stem from the investors assuming security when they invest in large capstocks: should there be any organizational crises, the corporation would not just collapse right away.
Another illustration is to consider commodity firms: the price of the product is determined by the market and not the firm itself. Consequently, large market capitalization would trigger herd behavior.
More herding investors means larger market capitalization. Conversely, when these investors simultaneously sell their stocks, it would cause a decline in the market capitalization of the stock.