With this perspective, the current study investigates the presence of herding behavior in the South African stock market. The results indicate a significant presence of herding behavior among investors on the Johannesburg Stock Exchange (JSE).
INTRODUCTION
- CONTEXT OF RESEARCH
- MOTIVATION
- OBJECTIVES
- METHODS OF ESTIMATION AND DATA
- ORGANISATION OF THE STUDY
To indirectly test the presence of herd behavior by examining whether stock returns in the JSE show volatility clustering or not. To investigate whether herd behavior in the context of the South African stock market varies with market conditions or not.
LITERATURE REVIEW
THEORETICAL LITERATURE ON HERD BEHAVIOUR IN THE FIELD OF SOCIAL PSYCHOLOGY
- LITERATURE ON INFORMATIONAL SOCIAL INFLUENCE
- LITERATURE ON NORMATIVE SOCIAL INFLUENCE
In the same way, normative social influence can be achieved through processes of identification or compliance. It has been shown that normative social influence can lead people to make false statements (Asch, 1956; Deutsch and Gerard, 1955; and Milgram, Bickman, and Berkowitz, 1969), to use illegal drugs (Maxwell, 2002), or fail to react to an imminent danger (Latané and Darley, 1970).
THEORETICAL LITERATURE ON HERD BEHAVIOUR IN THE FIELD OF BEHAVIOURAL FINANCE 5
- IRRATIONAL HERDING
- MOMENTUM-INVESTMENT AND/ OR POSITIVE-FEEDBACK STRATEGY Momentum-investment consists of buying past winner and selling past loser stocks, whilst
- SHARED-AVERSION
- RATIONAL HERDING
- INFORMATION-BASED HERDING OR INFORMATION CASCADE
- REPUTATION-BASED HERDING
- COMPENSATION-BASED HERDING
Bikhchandani and Sharma (2001) explained that regardless of the signal being randomly noisy, the probability of starting a cascade is greater than 0.93 after the first four individuals invest. Roll (1992) emphasized that when the agent's compensation is linked to his/her performance, as evaluated against the performance of the benchmark, this can cloud his/her investment decisions and lead to herding behavior.
EMPIRICAL LITERATURE ON HERD BEHAVIOUR
- LAKONISHOK SHLEIFER AND VISHNY (LSV) MEASURE OF HERDING Earlier empirical studies on herding widely used the measure of herding based on
- PORTFOLIO-CHANGE MEASURE (PCM) OF CORRELATED TRADING One of the shortcomings of the LSV measure is the inability to take into consideration the
- THE CONCEPT OF BETA HERDING TO MEASURE HERD BEHAVIOUR In their attempt to solve the inability of some of the models to differentiate between a rational
Therefore, by averaging BMHi t, we obtain BMHi t, which is the average measure of herding on the buy side of the market for all stocks i in all quarters t in a given period. Second, the LSV measure only takes into account the number of investors who traded, without taking into account the volume of shares traded by investors.
1 Etbri tb
- MEASURING HERDING USING CROSS-SECTIONAL STANDARD DEVIATION (CSSD)
- TESTING THE PRESENCE OF HERDING USING CROSS-SECTIONAL ABSOLUTE DEVIATION (CSAD)
- HERD BEHAVIOUR IN EMERGING MARKETS
- HERD BEHAVIOUR IN SOUTH AFRICA
- SUMMARY
The idea behind the CSSD measure is that when there is herding in the market, investors tend to converge towards the market consensus ie. Dt takes the value of unity if the market return on day t is located in the extreme upper tail of the distribution and is equal to zero otherwise. This husbandry measure was also designed under the conditional version of the CAPM.
Conversely, when herding is present in the market, the relationship becomes negative and non-linear. Using Huang and Salmon's (2004) herd measurement8, Zaharyeva (2011) found evidence of a herd in the Ukrainian market. Three models were considered to detect the herd in the market as a whole: (1) herd beta measure, (2) CSSD, and (3) CSAD.
THE JOHANNESBURG STOCK MARKET
CHARACTERISTICS OF THE JSE
- NUMBER OF REGISTERED COMPANIES
The changes led to rapid growth in the JSE, which is reflected in the increase in the number of listed companies, the increase in market capitalization and the increase in the value of the broad share index. It is seen that there is a decrease in the total number of listed companies from 427 companies in 2003 to 385 companies in 2013. The decrease in the total number of listed companies can be attributed to various reasons.
The reader should be reminded that despite the fact that the number of companies listed on the stock exchange has slightly decreased, Figure 3.2 shows that the number of foreign companies listed on the stock exchange has increased significantly over the same period. The number of foreign listings increased from 22 to 49; this increase is an indication of foreign investor confidence in South African markets. SETS are state-of-the-art, flexible and robust trading platforms that have significantly contributed to improving liquidity and ensuring more efficient operations.
STOCK EXCHANGE NEWS SERVICE (SENS) 11
In terms of the SENS framework, price-sensitive information is any “unpublished information that, if published, would reasonably affect a company's share price” (Mabhunu, 2004: 17). However, as long as the information remains confidential, possession of price-sensitive information does not necessarily obligate the company to disclose it. For example, companies should avoid consulting with material shareholders before other shareholders on price-sensitive issues.
If there is suspicion that they are being misinterpreted, or wrongly accused of publishing price-sensitive information, companies. When journalists press for unpublished price-sensitive information, companies should be ready to give a 'no comment' response. In cases where it is likely that sufficient price-sensitive information has been collected for a story to be 'broadly' accurate, a company is under an obligation to ensure that the information is disclosed through SENS and in the press to ensure that the correct information is widely available.
SUMMARY
DATA AND METHODOLOGY
DISCRIPTIVE ANALYSIS OF DATA
The summary statistics of the variable used in this study, namely Cross-Sectional Absolute Deviation (CSAD), the absolute value of weighted market return (|Rm,t|) under different market conditions and the return of All Share Index( RALSI) are given in table 4.1. These summary statistics are the sample mean, maximum, minimum, median, standard deviation, skewness, and Jarque-Bera tests (with their P value). From Table 4.1 it can be seen that all these statistics are non-normal and have leptokurtic distribution, features common to most financial data (Chinzara, 2006).
All variables except the return for the All Stock Index are positively skewed, meaning that most of the actual series are below the mean. In addition, for all variables, the kurtosis is greater than three, indicating that the distributions are thin and long-tailed (leptokurtic). The leptokurtic distribution is confirmed by the Jarque–Bera normality test, which shows a significantly low P-value for all variables considered.
STATIONARITY
- DICKEY-FULLER (DF) AND AUGMENTED DICKEY-FULLER (ADF) UNIT ROOT TESTS
The DF test assumes that the error term in Equations 4.2, 4.3 and 4.4 is independently and identically distributed. The ADF test consists of supplementing the initial DF regressions by the lagged dependent variables (ΔYt-1). The number of lagged difference terms to be included should be determined empirically by including enough terms so that the error term in the above equations is serially uncorrelated (Gujarati 2003: 817).
Similar to the DF test, the ADF tests the null hypothesis of δ = 0, and follows the same asymptotic distribution as the DF test. Under the null hypothesis that δ = 0, the PP Zt and Z statistics have the same asymptotic distributions as the ADF t statistic and normalized bias statistics. As mentioned above, an advantage of the PP test over the ADF test is that it is robust against general forms of heteroscedasticity in the error term t.
METHODOLOGY
- TESTING THE PRESENCE OF VOLATILITY CLUSTERING
- GENERALISED AUTOREGRESSIVE CONDITIONAL HETEROSCEDASTICITY (GARCH)
- GLOSTEN, JAGANNATHAN AND RUNKLE GARCH (GJR GARCH)
- EXPONENTIAL GARCH (EGARCH)
- TESTING THE PRESENCE OF HERDING USING CROSS-SECTIONAL ABSOLUTE DEVIATION (CSAD)
- TESTING THE PRESENCE OF HERDING IN VARIOUS MARKET CONDITIONS
- HERD BEHAVIOUR DURING INCREASING AND DECREASING MARKET RETURNS
- HERD BEHAVIOUR DURING HIGH AND LOW TRADING VOLUME
- HERD BEHAVIOUR DURING HIGH AND LOW MARKET VOLATILITY Finally, the effects of volatility on investors herding behaviour were studied, using a similar
- TESTING FOR LAGS IN HERDING 14
S is the sample variance of the residual least squares, t is a consistent estimate of 2 and the Newey-West long-run variance estimate of t using t as a consistent estimate of 2. Besides accounting for asymmetric effects, another advantage of EGARCH (1,1,1) is the fact that when modeling ln(t21), t2 will remain positive even in cases where the parameters are negative . When this is the case, the cross-sectional standard deviation of stock returns is more likely to decrease or increase at a decreasing rate (Al-shboul, 2012).
Another shortcoming is the fact that the CCK model assumes that risk is constant over time, as Tan et al explained, "the characterization of time-varying risk requires specifying an appropriate time window over which to measure risk. Rm tUP, is the equilibrium portfolio return at time t when the market is rising and CSADUPt , is the CSAD at a time corresponding to Rm t,. The following formula is used to calculate standard deviations:. where Ri t, is individual return at time t, R is the average return and n is the number of the observation.
SUMMARY
ANALYSIS OF EMPIRICAL RESULTS
- STATIONARITY TESTS
- TESTING VOLATILITY CLUSTERING USING GARCH-TYPE MODELS Before running GARCH-type models, the mean equation, as expressed in Equation 4.15 was
- TESTING THE PRESENCE OF HERDING USING CROSS-SECTIONAL ABSOLUTE DEVIATION (CSAD)
- ASYMMETRIC EFFECT ON HERDING
- HERD BEHAVIOUR DURING INCREASING AND DECREASING MARKET RETURNS
- HERD BEHAVIOUR DURING HIGH AND LOW TRADING VOLUME
- HERD BEHAVIOUR DURING HIGH AND LOW MARKET VOLATILITY In order to test the asymmetric effect of herding during high and low market volatility the
- TESTING FOR LAGS IN HERDING
- SUMMARY
The current study replicated Equation 4.25 by suppressing the constant and by adding a time trend to the CCK regression equation. The present study used the modified CCK model, which takes care of the asymmetric investor behavior under different market conditions, as specified in Equation 4.27.see Table 5.6. To test asymmetric effect on herding during bear and bull markets in the JSE, the current study used a modified CCK model as expressed by Equation 4.29 and Equation 4.30.
The current study tested the asymmetric effects of high and low trading volumes on livestock, as formulated in Equations 4.30 and 4.31. The results are presented in table 5.15. To measure the speed of adjustment after the swarm has occurred, the error correction coefficient ECMt-1 is calculated and shown in Table 5.28. The ECMt-1 error correction coefficient in Table 5.28 is negative and statistically significant; this means that the series is non-explosive and that the long-run equilibrium ie.
SUMMARY, CONCLUSION AND POLICY IMPLICATIONS
SUMMARY AND MAJOR CONCLUSIONS
The results of GARCH-type models indicated the presence of herding behavior at the JSE, given the fact that stock returns in the JSE show volatility clustering. It has also been revealed that herding behavior is more common during bull markets than during bear markets, that it is more intense during low trading volume than during high trading volume and that it is more common during low market volatility than during high market. volatility. Finally, using the ARDL approach to cointegration, it has been shown that herding behavior occurs over time.
However, the results revealed a high rate of adaptation and it was concluded that herding behavior on the JSE is a short-lived phenomenon.
POLICY AND INVESTMENT IMPLICATIONS
LIMITATIONS AND SUGGESTED AREA FOR FURTHER RESEARCH It was observed in Chapter 2 that there is no direct link between the theoretical arguments on
“Generalized Autoregressive Conditional Heteroscedasticity”, Journal of Econometrics Self-monitoring and Product Conpicuousness on Reference Group Influence”, Advances in Consumer Research. “A study of herding behavior in the stock markets: an international perspective”, Journal of Banking and Finance. 2009): “Dynamic Returns Linkages and Volatility Transmission between South African and World Major Stock Markets”, Journal of Studies in Economics and Econometrics.
"A Study of Normative and Informational Social Influences on Individual Judgment," Journal of Abnormal Social Psychology. "Volatility Pooling in the Greek Futures Market: Curse or Blessing?", International Research Journal of Finance and Economics. 2008): “A behavioral perspective on customer distrust of financial services”, Journal of Financial Service Professionals Quantile Regression Analysis of Dispersion of Stock Return - Evidence of Herding? Keskustelualoitteita,57.
