Investment Strategy for International Investors in Asian Emerging Markets PDF

Preview

Summary

Article Asia-Pacific Journal of Management Investment Strategy for International Research and Innovation 13(1) 1–17 Investors in Asian Emerging Markets © 2017 Asia-Pacific Institute of Management SAGE Publications sagepub.in/home.nav DOI:10.1177/2319510X17740043 Surender Kumar1 http://apjmri.sagepub...

Description

Article

Investment Strategy for International Investors in Asian Emerging Markets Surender Kumar1 Moon MoonHaque2

Asia-Pacific Journal of Management Research and Innovation 13(1) 1–17 © 2017 Asia-Pacific Institute of Management SAGE Publications sagepub.in/home.nav DOI:10.1177/2319510X17740043 http://apjmri.sagepub.com

Abstract Incorporating the extra risky stocks in portfolios can lead to realisation of extra returns. Due to financial liberalisation, there are huge opportunities of investment in Asian stock markets as these markets have gone through a considerable expansion. In large number of empirical studies of risk return analysis, it is observed that economic stability and good perspectives have been key assets for the development of emerging markets. Thus, emerging capital markets are becoming increasingly important for institutional as well as individual investors. Liberalisation of financial systems in these emerging Asian markets has attracted domestic and foreign institutional investors to diversify their funds across the markets and reduce their portfolio risk. To examine the important aspects of investment strategy under risk and uncertainty, we have used stock returns of five major stock markets of emerging economies, namely, China, South Korea, Taiwan, India and Malaysia. The daily stock prices for the period January 2011 to July 2016 is used to study the impact of European sovereign-debt crisis and Chinese economic reform as well as currency devaluation on selected emerging markets. In this research, we have carried out a detail autoregressive conditional heteroskedasticity (ARCH) and its generalised models to estimate conditional and asymmetric volatilities. Engle’s Lagrange Multiplier (LM) test is used to confirm long periods of time with no evidence of ARCH effects as a diagnostic testing of fitted models.

Keywords Emerging market, financial liberalisation, investment strategy, unexpected volatility, ARCH effect

Introduction In the recent past, international portfolio investments in emerging markets have become a matter of interest. In case of an efficient capital market, it has been researched and observed that it reflects the available information pertaining to stocks, resulting in investors having homogeneous expectations of the stocks’ performance which helps investors to value the stocks, taking into account the risk and return prospects (Mossin, 1966; Sharpe, 1964). These situations thwart investors from realising abnormal returns by utilising the inherent information in stock prices. Investors realise extra risk premium only by exposing their portfolios to unexpected variations in stock prices when efficient capital market hypothesis holds true and documents the random walk movements in stock prices. Substantial empirical work supports outcomes in developed as well as

1  Jaipuria

developing stock markets. The area has great potential for research in emerging stock markets. Choice of time period in this study is made in such a manner where it is just starting of the impact of European sovereign-debt crisis. In 2011, the world noticed the latest financial crisis and investors started pulling out from the places such as Greece, Spain and other places. Impact on investment was all over the world. If we recall, this is the time when Indian rupee started to sink from `44 against the dollar and reached around `56 against the dollar within a year. If we consider the case of China, there were many social and economic reforms during this period which ultimately influenced the other economies. This period also includes currency devaluation and IMF warning to China for their rapid debt run-ups. Similarly, there are many political and economic issues that emerge in these countries also. In our research, we are trying to study the investment

Institute of Management, Noida, India. Institute of Management, New Delhi, India.

2  Asia–Pacific

Corresponding author: Surender Kumar, Jaipuria Institute of Management, A-32A, Sector 62, Noida 201309, India. E-mail: [email protected]

2

Asia-Pacific Journal of Management Research and Innovation 13(1)

opportunities in Asian region even after these events, whether it is positive or negative for economies. Due to the ability to generate superior returns in global portfolios emerging markets have achieved significant importance. There are many researchers (Bekaert, Erb, Harvey, & Viskanta, 1998; Harvey, 1995; Peter & Kannan, 2007) who have observed superior return in emerging markets as compare to developed markets. To reduce portfolio risk and realising the optimum diversification of funds across the markets, it has become very important for researchers and policymakers to examine the dynamic relationships among the emerging Asian markets. There are many researchers who have investigated the integration of stock markets (Johnson & Soenen, 2002; Mukherjee & Mishra, 2007; Nath & Verma, 2003). Accelerating trade relations and liberalisation of financial systems integrated the Asian stock markets. Large number of researchers have investigated the seasonal structural in stock returns in developing as well developed markets and recommended a seasonal pattern by identifying the autocorrelation in returns of these markets (Aggarwal & Rivoli, 1989; Jarrett & Kyper, 2005a; 2006; Johnson & Soenen, 2002, 2003; Lee, 1992). Generalised autoregressive conditional heteroskedasticity (GARCH) (1, 1) model was found to be a better predictor of volatility in case of the US and European stock market data (Akgiray, 1989; Brooks, 1998; Corhay & Rad, 1994). In another study, the sudden change in volatility was found in the emerging stock markets, and it was observed that high volatility leads to a sudden change in variance (Aggarwal, Inclan, & Leal, 1999). Time varying relationship was found in stock market returns and high correlation during high volatility time was identified (King & Wadhwani, 1990; Schwert, 1990). To study the dynamic relationship across the stock markets using daytime and overnight returns, many researchers have used a two-stage GARCH model (Balaban & Kan, 2001; Gurmeet, 2016; Hamao, Masulis, & Ng, 1990; Kumar & Mukhopadyay, 2002). McClure, Clayton and Hofler (1999), Hu (2000) and Frank and Frans (2001) in their studies had observed strong interdependence among the stock markets. There are a number of studies where co-integration and Granger causality tests were used to understand the role of the US stock market in the world stock market integration (Cheung & Mak, 1992; Karolyi & Stulz, 1996; Masih & Masih, 2001). In many empirical works, researchers have identified the dynamic relationship between different markets across the globe, but there are some who did not find any such dynamic relationship (Cheung & Lee, 1993; Ewing, Payne, & Sowell, 1999; King, Enrique, & Wadhwani, 1994; McClure, Clayton, & Hofler, 1999). There are a number of studies which investigate volatility spillover, market efficiency and financial volatility that have significant influence on the economy growth and the decisionmakers (Kalsie, 2012; Mishra, 2015; Pandey, 2014; Poon & Granger, 2003; Verma & Kumar, 2015). There is a

significant association between high volatilities in market and investor’s confidence to invest. Due to these reasons, volatility models are of great importance to explain the movements in financial markets (Kamal, Hammmad-UlHaq, & Khan, 2012). In empirical finance, one of the major and interested areas of research is prediction of returns on stocks and currencies. This research is not just limited to predicting the expected returns on stocks and currencies but in associated volatility of returns also. There are enough evidences of time varying risk which means existence of relative volatile periods alternatively followed by smoothed ones. Variance, as central parameter, plays an important role in the understanding of any distribution; as we know that variance is the only unknown and necessary in option pricing (Black & Scholes, 1973). As per the Basel Committee on Banking Supervision, 2001, adequacy in banking capital be subject to the measurement of the value at risk depend on the volatility of returns of the portfolio of assets held by banks, which regulates financial institutions. Following which many financial regulators have adopted the risk metrics methodology (Mina & Yi Xiao, 2001). In this method, volatility estimate is based on an exponential smoother. There are many linear models which can be used in empirical finance that explains returns and volatility. But with the passage of time, we found several researches in which non-linear models exhibit non-linear characteristics of financial time series more appropriately (Franses & van Dijk, 2002). In financial time series data, large returns occur more frequently and tend to appear in clusters. It has been seen that large negative returns occurred more frequently than positive ones in stock markets. Because of non-linearity in the variance, non-linear model are very useful and used frequently in the research in the area finance economics. Considering these features of financial time series data, autoregressive conditional heteroskedasticity (ARCH) (Engle, 1982) and GARCH (Bollerslev, 1986) models are popular due to their capability of describing volatility clustering as well as other characteristics such as excess kurtosis or fat-tailedness, etc. As ARCH and GARCH do not capture leverage effects in that case, asymmetric models include exponential GARCH (Nelson, 1991). ARIMA technique was used for seasonal adjustment as well as decomposition of the time series data into its mechanism such as trend, seasonal and noise with the assumption of the Gaussian ARIMA model (Hillmer & Tiao, 1982). In one of the research, volatility modelled and forecasted in GSE by taking an individual index and using the models or specifications such as RW, GARCH (1,1), EGARCH (1,1) and TGARCH (1,1) (Magnus & Fosu, 2006). Rafique and Kashif-ur-Rehman (2011) studied the volatility clustering, excess kurtosis and heavy tails of the time series of KSE using ARCH and GARCH, and it was found that GARCH (1,1) has done the best to fully capture the persistence in volatility.

Kumar and MoonHaque 3 The ‘leverage effect’ was successfully overcome by EGARCH (1,1) specification in KSE-100 index. Rodriguez and Ruiz (2009) also studied the theoretical characteristics of a few and most trendy GARCH specifications having the component of leverage effect. Malkiel (1979) and Pindyck (1984) studied the cause of decline in the US stock prices during the 1970s and 1980s, and found that upward trend in market was a major reason for the same. In one of the study, Poterba and Summers (1986) revealed that if the shocks to volatility persist over a long time, then it leads to significant impact of volatility on the stock price. Another study conducted by French, Schwert and Stambaugh (1987) identified positive unexpected change in volatility which resulted in increased future expected risk premium and at the same time decreased the current stock prices while investigated the relationship between stock returns and volatility of NYSE for the period of 1928 to 1984. Wang and Gunasekarage (2005) examined the return and volatility spillover from the US and Japan to three Southeast Asian capital market, namely, India, Pakistan and Sri Lanka. They found a return spillover from the US and Japan to all the three markets. Mukherjee and Mishra (2005) using daily data also reported that the Indian stock market was not integrated with the aforementioned developed nations. Similarly, other researches on the integration of the Indian stock market with that of the emerging markets also provided mixed results. Mukherjee and Mishra (2005) discovered that the Indian stock market was integrated with the emerging Asian markets of Indonesia, Malaysia, Philippines, Korea and Thailand. The aforementioned literature reviews reveal that not much work has been carried out on these selected stock market indices return volatility; therefore, the present study makes an attempt to measure the stock market indices volatility and may be useful to make investment decision by investors across all these emerging markets. There are various models used by different researchers to explain and forecast patterns in volatility. The common measure of volatility such as variance and standard deviation are not sufficient to distinguish time-varying and clustering patterns of asset volatility (Olowe, 2009). One of the most useful models to identify time-varying volatility was introduced by Engle (1982) as ARCH model and further improved by Bollerslev (1986) into GARCH model. These models are very useful to conditional (varies over time) and the unconditional volatility (remains constant) of stochastic process (McMillan & Thupayagale, 2010).

Research Methodology Stock markets data of major five economies of Asia is used for modelling volatility. In our investigation, we selected major five Asian markets which are included in The Morgan Stanley Capital International (MSCI) as an emerging market index which is one of the best captures of

emerging markets. According to the list of MSCI, those top five emerging markets of Asia are as follows: China (SSE composite index), South Korea (KOSPI composite index), Taiwan (TSEC weighted index), India (S&P BSE sensex) and Malaysia (FTSE Bursa Malaysia KLCI). The daily stock prices for the period January 2011 to July 2016 is used to estimate the volatility; data is adjusted for all public holidays. Daily stock prices (adjusted close) are used to calculate the returns. Daily stock returns variations are responsible of volatility in stock market. Suppose St and St–1 are stock price at time period t and t – 1, respectively, rate of return Rt in ‘t’ time period would be as follows: R t = Ln (S t) - Ln (S t - 1) . Generally, return consists of two components; expected return E(Rt) and unexpected return ‘ft’ as follows: R t = E (R t ) + f t . Unexpected rise or decline in return reflects the impact of good news or bad news. ARCH procedure is applied to identify the conditional variance (function of past error term) (Engle, 1982). In this approach, past news is considered to be observed by conditional variance which can be used to forecast the volatility.

ARCH Model Before the ARCH model was introduced by Engle (1982), the most common way to forecast volatility was to determine volatility using a number of past observations under the assumption of homoscedasticity. However, variance is not constant. Hence, it was inefficient to give same weight to every observation considering that the recent observations are more important. ARCH model, on the other hand, assumes that variance is not constant and estimates the weight parameters, and it becomes easier to forecast variance by using the most suitable weights. Mean function of ARCH (1) is a simple first order auto regression: R t = c + bR t - 1 + f t and the conditional variance equation is as follows: v 2t = ~ + af 2t - 1.

GARCH Model The GARCH model was developed by Bollerslev (1986) and Taylor (1986) independently. In GARCH (1,1) model, conditional variance depends on previous own lag. The mean equation of GARCH (1,1) is: R t = c + bR t - 1 + f t

4

Asia-Pacific Journal of Management Research and Innovation 13(1)

and the variance equation is: v 2t = ~ + af 2t - 1 + bv 2t - 1 where ω is constant, f 2t - 1 is the ARCH term and v 2t - 1 is the GARCH term. As we can see, today’s volatility is a function of yesterday’s volatility and yesterday’s squared error.

The GJR-GARCH Model–GJR-GARCH (1,1) This model was proposed by Glosten, Jagannathan and Runkle (1993). Conditional variance is given by: v 2t = ~ + af 2t - 1 + bv 2t - 1 + cf 2t - 1 I t - 1 It–1 = 1   ft–1 < 0   It–1 = 0 where, if and otherwise.

The Exponential GARCH Model: EGARCH Nelson’s (1991) EGARCH (1,1) model’s variance equation is as follows:

Empirical Results A detail summary of the stock return is depicted in Table 1. Flatness and peakedness of any data series can be identified on the basis of its kurtosis value. Kurtosis value more than 3, equal to 3 and less than 3 represents peakedness,

standard normal and flatness in distribution. Higher kurtosis value shows that the data series is peaked; moreover in the data series selected in our analysis, excess kurtosis is varied between 5.76 (SSE) and 1.53 (BSE). It means the maximum variance is due to infrequent extreme deviations; and in other words, return distribution is inconsistent with the assumption of normality which leads to large magnitude returns occurrence more frequently than in the case of a normal distribution. As per this observation, these emerging markets seem to be very attractive to those investors who want to realise extra return. In addition to this, we can observe from the Table 1 that Indian market followed by Malaysian and Taiwan are more consistent as compared to Chinese and South Korean. This can be a better strategy to those who are moderate risk taker. Overall, as per descriptive statistics, it seems that investment in these emerging markets would be a profitable affair to any international investor, if invested with proper planning. Table 1 also reveals the result of test of normality using Jarque–Bera (JB) test, for the sampled data set at 1 per cent significant level rejects the hypothesis of normality.

Stationarity Test of the Stock Return The most prevalent approach, augmented Dickey–Fuller (ADF) test of the stationarity, given by Dickey and Fuller in 1979 is used to examine the stationarity of daily stock return. We have also applied Philips–Perron (PP) test to reconfirm the stationarity results. The ADF and PP test results are shown in Table 2. At 1 per cent level of

Table 1. Descriptive Statistics of Return of Different Market Statistic Mean

SSE

KOSPI

TSEC

BSE

KLSE

6.18985E-05

–1.51508E-05

3.91875E-08

0.000224466

5.84771E-05

Standard error

0.000424

0.000273458

0.00026399

0.00028304

0.000161152

Median

0.000487779

0.000031748

–0.000428321

0.000236062

0.000237375

Standard deviation

0.015497754

0.010147475

0.00976407

0.010483959

0.00596699

Sample variance

0.00024018

0.000102971

9.53371E-05

0.000109913

3.5605E-05

Kurtosis

5.76540718

4.382025058

2.8669656

1.530677506

2.61927263

–0.895966263

–0.377597942

0.398914439

–0.149797345

–0.276201153

Skewness Range

0.162852471

0.113202643

0.102016106

0.098231727

0.060602663

Minimum

–0.088729056

–0.064202455

–0.04459425

–0.061197108

–0.027380294

Maximum

0.074123415

0.049000188

0.057421856

0.037034619

0.033222369

Sum

0.082696351

–0.020862699

5.36085E-05

0.307967029

0.080172118

1,336

1,377

1,368

1,372

1,371

250.3737894

–669.7634532

249162.7897

46.70627286

102.0397634

2,012.083

1,124.215

499.87.26

137.3255

405.1513

0

0

0

0

0

Count Coeff. of Var. Jarque–Bera Prob. Source: Authors’ own calculation.

Kumar and...