Influence of Price Volatility on Herding Behaviour: A Study of Nigerian Stock Market

Price volatility in the stock market could lead to irrational behaviour of investors which might tantamount result into herding behaviour. This study evaluated the influence of price volatility on herding behaviour in Nigerian stock market with focus on Consumer goods, Financial services, Health care and Industrial goods sectors. Monthly data of stock prices for fifteen years from 2001 – 2015 were used and 97 companies' stocks from the four sectors were considered. OLS model was used to determine the existence and extent of herding behaviour in these sectors. The results showed that price volatility had influence on herding behaviour, but there was no evidence of herding noticed in any of the sectors, except Financial services sector which was not statistically significant. The study recommended that NSE should make information available to all market participants in order to boost their confidence in making investment decisions.

performances and their reputations by ignoring voluntarily their own analyses and to reproduce another manager who possesses a source of more reliable information or the analysis competencies of more eminent decisions (Ouarda, et al., 2013).

Theoretical Review
This study is based on the capital asset pricing model (CAPM). It was developed in mid-1960s by Sharpe (1964), Lintner (1965) and Mossin (1966). Consequently, the model is often referred to as Sharpe-Lintner-Mossin Capital Asset Pricing Model. CAPM was developed when the theoretical foundations of decision making under uncertainty were moderately new and empirical facts about return and risk in the stock markets were unknown (Perold, 2004). Oseni and Olanrewaju (2017) described the CAPM as one of the early risk management models which has remained a principal ornament for modelling modern financial economics. Rossi (2016) described it as a useful tool for estimating the cost of capital for firms and the returns that investors require in investing in a company's assets. The CAPM explains the trade-off between assets' returns and their risks, measuring the risk of an asset as the covariance of its returns with returns on the overall market.
The model (CSAD) for detecting herding behaviour was derived by Chang, Cheng and Khorana (2000) from the conditional version of CAPM. Chiang and Zheng (2010) stated that according to the Capital Asset Pricing Model (CAPM), there is a linear relationship between return dispersion of individual company's stock and return on market portfolio, but when different market conditions exist, investors may react in a more uniform manner, exhibiting herding behaviour which brings about a non-linear market return.

Empirical Studies
The existence and measure of herding in stock markets is distinguished on two categories of measuring herding based on the nature of the defined data. The basis for the first measures is investors' portfolio's composition and investors' transaction flow, while the second category focuses on herding as a whole and this indicates collective behaviour of all market participants.
Lakonishok, Shleifer and Vishny (1992) empirically examined the patterns of trading of institutional investors by concentrating on the frequency of herding and positive-feedback trading, which are related to the general notion that institutional investors disrupt stock prices. 769 all-equity tax-exempt funds which is predominantly pension funds were sampled and evaluated; these funds were managed by 341 institutional money managers in the United State. The result shows that there is a little evidence of herding among pension fund managers when trading in large stocks (those in the top two quintiles by market capitalization), which indicates over 95% concentration of their trading. Evidence of herding was found in smaller stocks, but the extent is far from dramatic. Also, the estimations of the study of Walter (2006) based on 60 mutual funds specialized in shares German declare that herding is a little higher than the ones gotten from other developed financial markets. Christie and Huang (1995) made used of Cross-Sectional Standard Deviation (CSSD) which they derived to measure herding. Data on daily returns of stocks listed on the NYSE and Amex for July 1962 to December 1988 were used and the results show that herding occurs when the market is under stress, i.e. when an individual investor possibly ignore their own information and evaluation and go with the market consensus. Chang et. al. (2000) also made used of their derived technique which is the Cross-Sectional Absolute Deviation (CSAD) and studied markets in the U.S., Hong Kong, South Korea, Taiwan and Japan. They discovered that herding does not take place in the U.S. and Hong Kong, little trace of herding in Japan, but significant proof of herding in South Korea and Taiwan. Chiang and Zheng (2010) used daily data of industrial stock returns to study herding behaviour of 18 countries which are United States, Australia, France, Germany, Hong Kong, Japan, the United Kingdom, Argentina, Brazil, Chile, Mexico, China, South Korea, Taiwan, Indonesia, Malaysia, Singapore, and Thailand for May 25, 1988to April 24, 2009. Contrary to previous studies that evidence of herding was not seen in advanced markets (Chang et al. 2000;Demirer & Kutan, 2006), they discovered significant evidences supporting the existence of herding in all the national markets except the US and Latin America.
Lakshman, Basu and Vaidyanathan, (2011) concluded that Indian investors are better informed and behave rationally since there is no significant evidence of herding in Indian stock markets. In contrast to Christie and Huang (1995), they suggested that herding can be more pronounce before market stress, rather than during market stress since market crisis can lead to market equilibrium. Contrary to the results of Chang et al. (2000) who discovered herding in emergent economies such as South Korea and Taiwan; Prosad, Kapoor and Sengupta (2012) used daily data of fifty (50) from the period of April 2006 to March 2011 and did not find herding in the Indian stock market. Nevertheless, individual tests for bull and bear periods of the market show that herding is detected in larger degree in bull period. These findings are in support of the results of Lao and Singh (2011). Ahsan and Sarkar (2013) examined herding in Dhaka Stock Exchange (DSE) in Bangladesh for a period of seven years. Daily and monthly returns for all stocks listed on the Dhaka Stock Exchange were used and there was no existence of herding in Dhaka Stock Exchange for the period studied.

Foundation of Estimated Model
The model for analysing herding behaviour during period of price volatility was adopted from the work of Ouarda, et al. (2013) which is in line with the study made by Chiang and Zheng (2010). The Cross-Sectional Absolute Deviation (CSAD) which measures returns dispersion was used to identify herding behaviour. The CSADt is stated below: (1) Where Ri,t is the observed stock return of industry i at time t, Rm,t is the cross-sectional average stock of N returns in the portfolio at time t and N is the number of firms in the portfolio. The model (CSAD) for detecting herding behaviour was derived by Chang, et al. (2000) from the conditional version of CAPM which stipulated that there is a linear relationship between the return dispersion of individual company stock and the return of market portfolio. This model was adapted by Ouarda, et al. (2013). For this study, the adapted model by Ouarda, et al. (2013) will be adopted to determine the existence of herding behaviour in Nigerian stock market.

Herding Behaviour and Price Volatility
According to Gleason, Mathur and Peterson (2004), herding is more pronounced during periods characterized by abnormal volatility. This perception is supported by Ouarda, et al. (2013) who discovered that herding behavior is more likely to be spread during months characterized by strong volatilities in European financial market. The potential effects of asymmetric herding behaviour in relation to price volatility is measured by using the understated approach.
CSADi,t = γ0 + γ1 D Hvolatilty │Rm,t│+ γ2 (1 -D Hvolatilty )│Rm,t│+ γ3 D Hvolatilty R 2 m,t + γ4 (1 -D Hvolatilty ) R 2 m,t + ε (2) Where D Hvolatilty is a dummy variable which takes the value 1 during the month of high volatility and 0 otherwise. Market volatility is assumed to be high if it exceeds the weighted average of the volatilities of six months preceding our study period and vice versa.

Data
To determine the existence of herding behaviour during the price volatility, the study made used of 97 stocks' returns from four sectors in the Nigerian stock exchange with a monthly frequency from January, 2001 to December, 2015. The criteria for choosing the 97 stocks from the total of 118 stocks listed in these sectors are the stocks that are consistently listed on the Nigerian stock exchange, companies that are still actively trading on the floor of the Nigerian stock exchange, those that traded most on the Nigerian stock market and they have more than 50% of total market capitalization. The methodology stated above are applied on the group of stocks on the basis of sector classification in the Nigerian stock exchange. The 97 company stocks are already classified into 4 sectors by the Nigerian stock exchange as Consumer Goods, Financial Services, Healthcare and Industrial Goods. The monthly stock returns are determined by applying the formula Ri,t = Pi,t -P0,t P0,t respectively. Pi,t represents the monthly closing prices of stock i at time t while P0,t represents the monthly opening prices of stock 0 at time t. The returns of market portfolio are calculated based on equally weighted portfolio of all companies in each sector classification.  Table 1 summarized the descriptive statistics respectively for average monthly market returns and dispersion returns of market portfolio. The average monthly returns of market portfolio range from a minimum incremental rate of 0.02 for the three sectors except for consumer goods with a maximum of 0.05. The average monthly return volatility varies between a maximum of 0.20 for Consumer goods sector to a minimum of 0.09 for Financial services sector. The observations on Table 1 show that Consumer goods sector show higher variations of 20% compared to the other sectors; nevertheless, the percentage of these variations are still negligible.

Descriptive Statistics
The descriptive statistics of CSAD for the sectors show mean values of 0.16, 0.12, 0.14, and 0.13 for Consumer goods, Financial services, Health care and Industrial goods respectively showed that Consumer goods sector has the highest mean value, followed by Health care, then Industrial goods and Financial services. These results depict that consumer goods sector has the highest market variation across industrial returns compared to others. The values of the standard deviation compared to the mean values of some of the sectors show that these sectors do not experience unusual variations, except for Consumer goods compared to their mean values. Thus, showing that Consumer goods experienced unusual variation due to unexpected news or shocks. This decision was reached based on the work of Chiang and Zheng (2010) which says that a higher mean value suggests significantly higher market variations across industrial returns for one industry compared to others, while a higher standard deviation suggests that the market had unusual cross-sectional variation due to unexpected news or shocks.    Figure 5. It showed that at the early 2007, consumer goods sector experienced the most increase in market return which showed that due to the anticipated change of government, investors bought more stocks from consumer goods sectors compared to other sectors. The values in the parentheses () are p-value Table 2 showed the estimated results of the possible effect of price volatility on herding behaviour for all the four sectors are shown. The coefficients of a non-linear market return during high and low volatility showed a negative sign of -0.33 and -1.06 for Financial Services, while other sectors showed positive signs. This imply that during high and low volatility, Financial service showed evidence of herding behaviour, but the level of herding was not statistically significance, while herding was not noticed in the other three sectors. The p-values of almost all the sectors were lower than 0.05 level of significance, except for Industrial goods (0.25) during high price volatility. This implied that price volatility had effect on herding behaviour in all the three sectors, but did not have in Industrial goods sector during high price volatility.

Regression Results
These findings contradicted the work of Ouarda, et al. (2013) who discovered high level of herding behaviour in Financial services during low volatility, and also some levels of herding in the other three sectors of European market during high volatility.

Conclusion
The purpose of this paper was to empirically study the existence of herding behaviour during price volatility in four sectors of the Nigerian stock exchange for a period of 15 years, that is, from January 2001 to December 2015. For the different sectors identified, the study showed empirical evidences in support of the existence of herding behaviour that is not statistically significant during price volatility only in Financial services sector, while evidence of herding behaviour was not noticed in the remaining three sectors. In the stock market, it is theoretically believed that the presence of any market condition would influence investors' behaviour, but this study has shown that the influence of price volatility on herding behaviour did not determine significant level of herding in any of the sectors. It is therefore concluded that price volatility influences herding behaviour, but the influence did not determine significance level of herding in any of the sectors. The study recommended that the existence of herding in Financial services sector should be investigated and NSE should also make all information available to all market participants in order to boost their confidence in making investment decisions since information is regarded as power and a well-informed investor has an edge over others.