CAPÍTULO I. INTRODUCCIÓN
1.1. Descripción de la empresa
The current section reviews and highlights the chronological development of key methods proposed to test the existence of herding behaviour in financial markets. The section first focusses principally on the methodological improvements in detecting herding behaviour. In addition, an assessment of related empirical studies on herding behaviour in the Johannesburg Stock Exchange is made.
The methodological trajectory in testing herding behaviour has been ongoing for almost three decades now. Many other alternative models have recently been proposed to solve identified weaknesses in the conventional approaches. However, findings using these advanced models also contradict findings of the conventional models as well as the recently proposed ones. An attempt to reconcile the empirical findings using the various methodologies remains a critical subject to stakeholders in industry and academia.
First, one of the most influential studies in the empirical herding literature has been the analysis by Lakonishok et al. (1992) (hereafter, LSV). The proposed approach of Lakonishok et al. (1992) mainly analyses investors‟ portfolio holdings for testing for herding behaviour in financial markets. Given a specified time distance, their method primarily seeks to measure whether more investors are trading more on either the buy side or sell side of the market than would normally be expected if market participants traded independently. Thus, when investors actively trade, their decisions regarding their trades (i.e. buying/selling) may cause stock prices to deviate from their fundamental economic levels. The LSV measure apart from following the hypothesis of an efficient market also utilises trading data in its estimation.
The measure introduced by LSV is calculated as the difference between the number of investors who buy (sell) an asset in a specified time frame against their theoretical values. Lakonishok et al. (1992), evidence of correlated trading by investors shows the presence of herding behaviour in a financial market. Similarly, other alternative models of testing for herding behaviour proposed by Welch (2000) and Walter & Moritz Weber (2006) also follow the LSV principles.
Even though the LSV measure has been adopted in a great number of studies (see, for example, Choi & Sias (2009) for South Korea, Kyrolainen & Perttunen (2003), for Finland, Voronkova & Bohl (2005), for Poland, Wylie (2005), for the UK, Lobao & Serra (2007), for Portugal, Walter & Moritz Weber (2006), for Germany), the approach has been criticised both from the theoretical and the empirical front.
Critics of the LSV statistic (fsee, for example, Bikhchandani & Sharma, 2000; Wylie, 2005) argue that the use of a binary measure of buys versus sells, rather than the size of the transaction and the lack of an inter-temporal dimension of herding may lead to spurious results. For example, Grinblatt et al. (1995) and Wermers (1999), Andreu et al. (2009) and Hu et al. (2008) provide alternative but slightly different means in addressing some identified drawbacks with the LSV approach. For instance, to account for different transaction size, many studies have strongly advocated that researchers should concentrate only on transactions of sufficient size when employing the LSV approach.
Christie & Huang‟s (1995) proposed model for testing for herding behaviour in financial markets is indistinguishable from that of Chang et al. (2000). The two models primarily employ the Ordinary Least Squares Regression model (hereafter, OLS), to study the linear and non-linear relationship between equity dispersion and market return as a method of detecting investors‟ herding behaviour. While the former tests for the linear relationship between equity dispersion and market return using dummy regression analysis, the latter investigates the non-linear relationship between the two market metrics (i.e. equity dispersion and market return) in analysing herding behaviour in financial markets using OLS.
Chang et al. (2000) updated Christie & Huang (1995) model citing the sensitivity of the former model to outliers as indicated by Economou et al. (2011). Unlike the Lakonishok et
al. (1992) indicator which focuses on the number of transactions by investors with respect to a specific asset, Chang et al. (2000) and Christie & Huang (1995) models primarily focus on the degree of dispersion of an investor with respect to securities‟ returns. Both models make use of data on the aggregate market data for testing evidence of the market phenomenon in financial markets.
While Lakonishok et al. (1992 herding statistic is considered as the current standard in the literature on fund‟s herding behaviour, the two other conventional methodologies proposed by Christie & Huang (1995) and Chang et al. (2000 are considered as the current benchmarks in the literature on stock market herding. However, several studies have confirmed the superiority of the CSAD over the CSSD model (Yao et al., 2014; BenSada et al., 2015; Litimi et al., 2016). For example, while the CSSD is linear and principally tests evidence of herding behaviour under the condition of extreme return only, the CSAD is nonlinear and hence influenced by several factors (Lux, 1995).
However, Chiang et al. (2010) argue that herding behaviour apart from being dominant during the period of market stresses as suggested by Christie & Huang (1995) may also exhibit traces of the market phenomenon at the entire return distribution. More so, other inherent drawbacks of the estimation method of the CSSD methodology include the definition of extreme return which is arbitrary. The cut-off point commonly used in literature defining the extreme market movements is respectively 1%, 5% and 10%. To capture the equity dispersions, Chang et al. (2000) proposed the Cross-Sectional Absolute Standard Deviation (hereafter, CSAD), as a measure of return dispersion. The return-dispersion-based models such as CSSD and CSAD tend to follow the CAPM specification of returns. According to Chang et al. (2000), investors primarily depend on the overall market conditions in taking decisions to achieve their investment objectives. They observed that investors rely on their private beliefs and information, which are mostly different in taking investment decisions, especially during the normal market period. Thus, during periods when the market is either rising or falling, investors suppress their private beliefs and conform to the prevailing market consensus and in so doing deviate from their investment strategies. Under such market condition, they argue that individual asset returns cluster around the market return, leading to herding becoming more prevalent.
On the other hand, Chang et al. (2000)posit that because of asymmetry, especially during periods of market stress, there is an inverse linear relationship between equity dispersion and market returns. Lao & Singh (2011) have provided evidence for asymmetric effects of herd behaviour patterns in financial markets. This suggests that investors tend to herd more intensively during either an upward movement or a downward movement of the market.
Alternatively, in 2001, Hwang and Salmon (hereafter, HS) introduced a new measure for testing herding behaviour. The motivation of their approach was derived from the CAPM. The HS statistic is connected to the CAPM theory and hence its validity is also
conditioned to the validity of the CAPM operates under restricted assumptions such as the efficient market hypothesis, rational expectations among others. According to Hwang and Salmon, a change of asset betas from their equilibrium position measures the level of herding behaviour in financial markets.
Thus, the HS measure is based on the cross-sectional dispersion of the factor sensitivity of assets within a given market. They further posit that the pioneering methods of testing for evidence of herding behaviour centred mainly on the cross-sectional variation in equity returns. Their proposed technique, however, focused on the risk-return relationship. Contrary to the pioneering models, Hwang & Salmon (2001) used assets betas as a proxy to measure herding behaviour. They observe that when investors follow the collective market consensus, the dispersion of asset betas tends to be smaller.
Despite the identified disadvantages of the traditional methods of detecting herding behaviour among investors, the alternative approach suggested by Hwang & Salmon (2001) was also not without weaknesses. One key drawback of using their methodology was that it was not easy to obtain expected results due to insignificant coefficients. This has motivated the continuous search for alternative models in testing for evidence of herding behaviour in financial markets.
Subsequently, other alternative statistical techniques have been explored in testing for evidence of herding behaviour (see, for example, Quantile Regression (Saastamoinen, 2008; Chiang et al., 2010; Vo & Phan, 2016), Rolling Regression method (Chianga et al., 2012), State-Space Model (Fu, 2010), Non-Parametric Kernel Regression (Mahmud & Tiniç, 2015) in many studies.
Saastamoinen (2008) is among the few researchers who have used alternative techniques to test for evidence of herding behaviour especially in emerging financial markets. The study employed quantile regression in analysing the stock returns dispersion of large capital companies on the Helsinki Stock Exchange. The results from the quantile regression indicated that stock returns dispersion increases in a less-than-proportional rate with the market return in the lower tail of stock return distribution, while in the upper tail, the rate of increase was nonlinearly increasing. Specifically, the author concluded that the results of the study suggest evidence of herding, but added that the evidence was not a conclusive proof of herding behaviour.
Likewise, Chiang et al. (2010) used the quantile regression method to examine herding behaviour in A-and B-type stock markets in the Chinese stock markets using daily data from 1996 to 2007. In addition to the quantile regression approach, the study also employed the CSSD and modified CSAD methodologies to test the market phenomenon. By applying Quantile Regression Analysis, the authors find supporting evidence of herding in both Exchange conditional on the return dispersion in the lower quantile region. Using OLS methods, on the other hand, they found strong evidence of herding behaviour within both the Shanghai and Shenzhen A-share markets, but no evidence is reported within both B-shares. A-share investors‟ display herding in both up and down market.
In addition, Vo & Phan (2016) recently examined the existence of herding behaviour in Vietnam stock market using daily data. The authors used the quantile regression analysis to test evidence of herding behaviour employing the metric outlined by Chang et al. (2000) as a measure of equity dispersion. They found evidence of herding behaviour during periods of market stress. However, no evidence is found in the higher quantile of return dispersion distribution. Thus, the asymmetric effect of herding exists in Vietnam equity market with the prevalence of this market phenomenon in bear market phase than in bull market phase.
Similarly, Tan et al. (2008) employed the CSSD and modified CSAD methodology to test existence of herding behaviour in the Chinese market. The study used daily data from 1994 to 2003. The results reveal the presence of herding within Chinese stock market as well as asymmetric effect in terms of stock returns, trading volume and volatility.
Fu (2010) on the other hand used two conventional methods; namely, the CSSD and CSAD to examine the Chinese stock market for evidence of herding behaviour. The study used monthly data from 2006 to 2009. They found no evidence of herding behaviour in the Chinese stock market, but the asymmetric effect is found with the prevalence of herding in up-market than in down-market.
Chianga et al. (2012) employed the rolling regression method to estimate herding equations in the Chinese capital market. The model coefficients were based on a one-year window. They found evidence of herding behaviour in both markets (i.e. A-and B-type stock markets) and showed that herding behaviour is correlated to global markets as well. On another front, Fu (2010) employed the state-space model and found herding during lower extreme values of market return in the Chinese stock market.
Also, Mahmud & Tiniç (2015) report new evidence of herding behaviour in the Chinese A-type and B-Type markets by employing the non-parametric kernel regression. They found statistically significant evidence of herding in A-type market under both extremely high and low market returns but only indicated a weak evidence of herding behaviour in the B-type market.
Apart from the Chinese stock market, selected empirical studies on herding behaviour in other single stock markets are catalogued in Table 3.1.
Table 3.1: A summary of empirical evidence on herding in financial markets
Note: LSV, CH, CCK, HS and NS respectively refer to models proposed by Lakonishok et al. (1992), Christie & Huang
(1995), Chang et al. (2000), Hwang & Salmon (2001) and Nofsinger & Sias (1999).
Source: Adapted from Vo & Phan (2017) and updated
Author(s) Period Model(s) Findings
Quarterly data LSV The study finds no evidence of substantial herding
(1985-1989) by US fund managers, except in small stocks.
Nofsinger & Sias (1999) Monthly data NS The results show the presence of herding in both US (1977-1996) institutional investors and individual investors. However,
institutional herding impacts prices more than herding by individual investors.
Wermer (1999) Quarterly data LSV The authors find little herding by mutual funds in the (1975-1994) average stock but much higher levels in trades of small
stocks in the US.
Jiao & Ze (2014) Quarterly data LSV The findings indicate evidence of US mutual fund herding (2000-2007) and the associatedprice destablisation effects. Moreover,
the results reveal that mutual funds herd into or out of stocks following the herd of hedge funds, not vice-versa. Bernales et al. (2016) Daily data CSAD They report significant herding effects in option returns, using
(1996-2012) their factors their cross-sectional dispersion, conditional on a set of systematic factors related to periods of market stress. Wylie (2005) Semi-annual data LSV Significant amount of fund manager herding is found in the
(1986-1993) largest and smallest UK stocks after adjusting for the positive bias in the LSV herding measure.
Caparrelli et al. (2010) Daily data CSSD, CSAD Herding behaviour is evident during extreme market
(1988-2011) LSV conditions in accordance to CSSD model.
Caporale et al. (2008) Daily/weekly/monthlyCSSD, CSAD Herding exist in the Athens stock market. data (1998-2007)
Walter & Weber (2006) Data (1998-2002) LSV The results provide evidence of herding by German mutual fund managers. Moreover, the authors find that a significant portion of herding detected in the German market is associated with spurious herding.
Kremer, S. & Nautz, D. (2013) High-frequency data Panel RegressionThe results show that institutions exhibit herding behaviour
(2006-2009) on a daily basis.
Gavriilidis et al. (2013) Quarterly data Sias model The results provide evidence that Spanish institutional (1995-2008) industry herds intentionallyfor most sectors which are
underperformed; thus, generating high volatility and volume Kim & Nofsinger (2005) Monthly data NS The authors find the presence of herding in Japan with a
(1975-2001) large impact on price movements. In addition, the effects and behaviour of institutional herding depend on the economic condition and regulatory environment. Choe et al. (1999) Daily data LSV Herding is found by foreign investors in Korea before
(1996-1997) economic crisis. However, herding falls during the crisis and no destabilisation impacts on Korea stock market over the entire sample.
Joen & Moffett (2010) Daily data Sias model The study finds evidence of strong impact of foreign investors (1994-2003) herding on stock returns. Moreover, the results also indicate
the opposite direction in buying and selling shares between oreign and domestic investors in the herding years. Choi et al. (2015) Daily data VAR They found that while domestic individual investors
(1995-2005) systematically apply aggressive contrarian trades, foreign and some domestic institutions are mostly trend-chasers
Germany
Table 2.1: A summary of empirical evidence on herding in single market setting
Spain Japan Korea USA UK Italy Greece Lakonishok et al. (1991)
Table 3.2: A summary of empirical evidence on herding in financial markets cont‟d
Note: LSV, CH, CCK, HS and NS respectively refer to models proposed by Lakonishok et al. (1992), Christie & Huang
(1995), Chang et al. (2000), Hwang & Salmon (2001) and Nofsinger & Sias (1999).
Source: Adapted from Vo & Phan (2017) and updated
Although the literature in this field is plentiful, empirical studies on the JSE are less numerous. In the existing literature, barring a few exceptions, most of the empirical studies (Gilmour & Smit, 2002; Seetharam & Britten, 2013; Sarpong & Sibanda, 2014; Niyitegeka & Tewari, 2015) on the JSE have predominantly employed the conventional methods in testing for herding behaviour. The findings of these studies support other related studies on emerging markets in general (see, for example, Chang et al., 2000; Lobao & Serra, 2007; Tan et al., 2008; Goodfellow et al., 2009; Zhou & Lai, 2009; Demirer et al., 2010; Lao & Singh, 2011; Economou et al., 2011; Holmes et al., 2013; Gavriilidis et al., 2013).
Gilmour & Smit (2002) tested for the “institutional” herding phenomenon in the unit trust industry in South Africa. They found that herding is present for unit trusts at a certain level of volatility. They concluded that the greater the volatility, the greater the herding behaviour in the unit trust industry.
Lakshman et al. (2013) Daily data HS The study shows that herding is observed in India stock
(1996-2008) market but not very severe.
Vo & Phan (2016) Daily data CSAD Herding is reported in Vietnam stock market, particularly (2005-2015) in the median and lower quantile of return dispersion
distribution. The results also reveal that herding is more pronounced in down market than up market.
Dang, H. V. & Lin, M. (2016). Daily data CSSD/CSAD They document compelling evidence of herd behaviour in
(2007-2015) the Ho Chi Minh Stock Exchange (HOSE) of Vietnam.
Galariotis et al. (2016) Daily data CSAD Overall, the study found no evidence of investor herding either (2000-2013) before or after the EU crisis. Further tests indicate that durring the EU crisis period macroeconomic information induces bond market investor herding.
Andrikopoulos et al. (2017) Daily data CSSD, CSAD The results show intraday herding to be significant before, (2002-2010) during and after the 2007–09 financial crisis period, with its
presence appearing the least strong during the crisis. Mobarek et al. (2014) Daily data CSAD While they report insignificant results for the whole period,
(2001-2012) they document significant herding behaviour during crises and asymmetric market conditions.
Galariotis et al. (2016) Daily data CSAD The initial results show no evidence of herding but when they (2000-2015) condition on the liquidity of stocks they found significant
evidence of herd behavior for high liquidity stocks, for most countries
Galariotis et al. (2015) Daily data CSAD The results indicate that US investors tend to herd during days (1989-2011) when important macro data are released, and that there
have been herding spill-over effects from the US to the UK during earlier financial crises. Further results reveal more differences in herding behavior between the two markets: in the US they find that investors herd due to both fundamentals and non-fundamentals during different crises, when in the UK there is herding only due to fundamentals and only during the Dotcom bubble burst
Eurozone/Euronext
UK/USA G5 Markets
Vietnam India
Seetharam & Britten (2013) examined herding behaviour among investors using all listed stocks as well as the All Share Index (ALSI) for the period 1995 to 2011 on the Johannesburg Stock Exchange (hereafter, JSE). They used three different methods proposed by Christie & Huang (1995), Chang et al. (2000) and Hwang & Salmon (2001) to analyse their data. They found evidence of herding behaviour during bear market periods only. However, it was absent overall.
Sarpong & Sibanda (2014) investigated herding behaviour among 41 domestic equity mutual fund managers and the performance of mutual funds that trade against the herd in South Africa. The herding measure of trading of Lakonishok et al. (1992) was employed to test herd behaviour in mutual funds over the period 2006 to 2012. They reported evidence of herding behaviour among mutual fund managers. They concluded that institutional investors in South Africa are prone to the behavioural bias of herding and this