Impact of News on Indian Stock Market: A Periodic Study with Asymmetric Conditional Volatility Models

Abstract

The article analyzes the leverage effect for Indian stock market. The effect of good and bad news on volatility during the period of growth, recession and the post-recession period is investigated taking American recession of 2008–2009 as the benchmark. Bombay Stock Exchange (BSE) 30 index was used as the proxy for Indian stock market. Using suitable asymmetric generalized autoregressive conditional heteroscedasticity (GARCH) models based on information criterions to study the asymmetry for the period and sub-periods, it was found that Indian stock market reacts differently to positive and negative news confirming leverage effect in all periods considered. However, the leverage effect was seen prominent during the period of gloom than other periods shown by the news impact curve.

Keywords

Leverage effect asymmetric news impact conditional volatility

Introduction

Stock market plays an important role in the economy by providing a platform where investors and capital seekers meet and flow of capital takes place. Investors take their investment decisions based on prospects of the performance of the company. The investors need to predict accurately the future performance of the company’s stock price based on the current prices. For accurate prediction of the future stock price, volatility is required to be estimated. Information plays an important role in deciding the level of volatility. As per leverage theory, volatility in the stock market is more for bad news than good news of similar intensity (Black, 1976). This is used to study the effect of stock prices in response to macroeconomic news (Bekaert & Wu, 2000; Cappiello, Engle & Shephard, 2006; Koutmos & Booth, 1995) and business cycle (Adams, McQueen & Wood, 2004; Conrad, Cornell & Landsman, 2002; Flannery & Protopapadakis, 2002; McQueen & Roley, 1993). It is to be noted that although good news brings positivity in the market while bad news creates negativity about the company performance, however, it is the news (either good or bad), which induces volatility. Volatility is natural for the stock market, but this brings uncertainty in the market when it is high like the recent US recession, which significantly impacted the world economy. With globalization, liberalization and development of information technology, the impact of news has increased manifold.

Liberalization of Indian economy started in 1991; policies were relaxed for foreign investors to make investments conducive in Indian stock market. However, with this Indian market was made vulnerable to external shocks, which were developed in other economies. The great recession was one such shock faced by the Indian markets due to the liberalized economy, which changed the behaviour of stock market towards the information perceived by the investors. Good and bad news are frequently seen in any stock market and to this news, investors react differently creating volatility in the stock market. However, the impact of news on volatility depends on the economic environment existing during the period (Hamilton & Susmel, 1994), which can act as a cushion for investors in reacting to good or bad news. This article studies the leverage effect of Indian stock market using asymmetric volatility models during the period of growth, recession, post-recession and whole period (i.e., from liberalization to April 2014) by applying suitable asymmetric volatility models and news impact curve.

Keeping the great American recession as benchmark, the Indian economy can be divided into three phases:

Growth Phase: The growth phase witnessed reaping of benefits by Indian economy as a result of the liberalization. The period witnessed growth in foreign investment and the digital boom in the 1990s. The period reached its conclusion in the form of negative impact due to the great American recession.

Recession Phase: The period started with the sub-prime crisis in US, ripple effects of which were witnessed by Indian economy due to integration with the world economy. With the withdrawal of foreign institutional investors (FIIs) from the markets, Indian economy faced some serious shocks. With a reduction in export and investment, economic growth was hampered. This was the period of gloom that impacted the performance of stock market.

Post-recession period: During this period, the stock market restored to the path of high growth. The economy created confidence in the mind of investors by the way it handled recession. The period witnessed FII investments flowing into the market, exports rising, thereby putting the economy on track.

Literature Review

The development of autoregressive conditional heteroscedasticity (ARCH) model by Engle (1982), later extended to generalized autoregressive conditional heteroscedasticity (GARCH) model by Bollerslev (1986), has revolutionized the research on volatility. Based on these two models, several models have extended to asymmetry models, such as Glosten–Jagannathan–Runkle GARCH (GJR-GARCH), threshold GARCH (TGARCH) and exponential GARCH (EGARCH), which capture the leverage effect. Since these methodologies have been evolved in developed economies, the research in this regard is majorly done on these economies. McQueen and Roley (1993) found good news to result in lower stock price during the high economic period and high stock prices during low economy. They concluded that this could be due to hope arising in the market during bad times. Braun, Nelson and Sunier (1995) studied leverage effect in stock returns with EGARCH model. They found leverage effect in volatility but absence in conditional beta. Flannery and Protopapadakis (2002) and Adams et al. (2004) found the same results for the macroeconomic announcement as found by McQueen and Roley (1993). Cappiello, Engle and Sheppard (2006) investigated the presence of asymmetry in conditional volatility of equity and bond returns applying the asymmetric version of dynamic conditional correlation. Batra (2004) analyzed the time-varying variation in Indian stock market for the period of 1979–2003. Using monthly returns and asymmetric GARCH methodology, he found that volatility persistence has increased in India with liberalization. Alberg, Shalit and Yosef (2008) compared the forecasting performance of conditional volatility for Tel Aviv Stock Exchange. They found that the asymmetric GARCH model forecasted better than symmetric GARCH models; however, among asymmetric GARCH models, EGARCH with skewed Student’s t distribution is the most successful. Petr Seďa (2011) studied the effect of good and bad news on volatility for Czech and Polish stock markets before and during recession period of 2008–2009. Applying EGARCH and TGARCH, he found asymmetry effect in both the stock markets. Liu, Wong, An and Zhang (2014) studied the asymmetric characteristics of returns and volatilities of various Chinese commodity futures. They found that the asymmetric threshold stochastic volatility model outperforms the corresponding symmetric stochastic volatility models in forecasting.

In the studies focusing on emerging stock markets like India, the literature on asymmetric volatility is limited. Karmakar (2007) studied risk–return relationship applying EGARCH model for 14-year data. The author did not find any relation between the two. Tripathy (2010) found leverage effect in Indian stock market for the period from January 2005 to January 2010 when the volume is taken as the proxy for news. Goudarzi and Ramanarayanan (2011) studied leverage effect in Indian stock market proxied by Bombay Stock Exchange (BSE) 500 returns with EGARCH and TGARCH model for a 10-year period. They found asymmetric response of volatility for good and bad news. Mittal, Arora and Goyal (2012) studied symmetric and asymmetric volatility of Indian stock market for the period from 2000 to 2010. They concluded GARCH for symmetry and PGARCH to best estimate asymmetric volatility.

Methodology

Development of conditional heteroscedastic models by Engle (1982) captures the time-varying volatility, which is common than constant volatility. As per ARCH, conditional volatility is a simple quadratic function of the lagged values of the innovations.

σ_{t}^{2} = α_{0} + \sum_{i = 1}^{q} α_{i} ε_{t - i}^{2}

where ε_t = σ_tz_t and z_t is i.i.d. random variable with zero mean and constant variance.

This ARCH model was extended by Bollerslev (1986) to overcome the requirement of many parameters and higher order q. A general GARCH model is

h_{t} = ω + \sum_{i = 1}^{q} α_{i} ε_{t - 1}^{2} + \sum_{j = 1}^{p} β_{j} h_{t - j}

where ω is the mean, ε²_t-1 is the news about volatility from the previous period measured as lag of the squared residual (ARCH term) and h_t_-j is past period estimated variance.

These models were further modified to capture the informational asymmetry and leverage effect. The EGARCH was developed by Nelson (1991); the asymmetric power ARCH (APARCH) model developed by Ding (1993) and the GJR-GARCH developed by Glosten (1993) allow for leverage effect.

This study uses daily closing observation of BSE 30 index from the period 6 May 1992 to 30 April 2014 a total of 5,352 observations divided into three sub-periods discussed earlier. To study the asymmetry effect during the period of boom and gloom for Indian stock market, R software version 3.1.2 was used.

Results and Discussion

Table 1 shows compiled descriptive analysis of BSE Sensex logarithmic returns for all the four periods, that is, the whole period of study from FY 1992–1993 to FY 2013–2014. The second period is of boom for Indian stock market where BSE has experienced growth with IT boom and the effect of liberalization policy adopted by the government, which is from May 1992 to November 2007. The third time period is the recession period, which started with the bankruptcy of Lehman brothers. Its duration is from December 2007 to 30 June 2009. The last period is the post-recession period, the time period is from July 2009 to 30 April 2014. The division of the periods is based on the information in Wikipedia ‘When did recession end?’

Table 1.

Descriptive Statistics of Returns for Period and Sub-period

	Whole Period	Pre-recession Period	Recession Period	Post-recession Period
Minimum	–0.11809	–0.11809	–0.1160	–0.06008
Maximum	0.15990	0.10589	0.1599	0.03703
Mean	0.000344	0.00045	–0.0007	0.00035
Median	0.00085	0.00106	–0.00015	0.00053
Sum	1.8398	1.6934	–0.3019	0.42572
Variance	0.00027	0.00027	0.00076	0.00012
Std dev.	0.01655	0.0164	0.0277	0.01125
Skewness	–0.1253	–0.2927	0.3291	–0.08829
Kurtosis	5.08076	3.3914	3.2285	1.3415

Source: Authors’ calculation.

Whole Period (1992–2014)

The ARCH-Lagrange multiplier (LM) test with 12 lags for heteroscedasticity was conducted for the whole period, which shows the presence of ARCH effect; the p-value of the test is highly significant showing the presence of heteroscedasticity for the time period. The presence of the ARCH guides that the data for this period are to be modelled by ARCH or higher models. GARCH (1, 1), EGARCH (1, 1), GJR-GARCH (1, 1) and APARCH (1, 1) models have been used to test for best-suited model for the data. Table 2 shows the comparative value for Akaike information criterion (AIC), Bayesian information criterion (BIC), Shibata and Hannan–Quinn statistics for each applied model for the whole time period.

Table 2.

Comparative Information Criteria for Whole Period

	Garch11	Egarch1	Gjrgarch11	Aparch11.1
Akaike	–5.625021	–5.630559	–5.632872	–5.624120
Bayes	–5.620100	–5.624407	–5.626720	–5.617968
Shibata	–5.625022	–5.630561	–5.632873	–5.624122
Hannan–Quinn	–5.623302	–5.628410	–5.630723	–5.621971

Source: Authors’ calculation.

The value of AIC, BIC, Shibata information criteria (SIC) and Hannan–Quinn test is smallest for GJR-GARCH (1, 1) model, so this model fits for the data for the whole period. So GJR-GARCH (1, 1) is used to model the volatility and study leverage effect for BSE Sensex for the period mentioned above.

Table 3 shows the estimated parameters for GJR-GARCH (1, 1); the estimates are significant with t-value greater than |2|. γ in the output is positive and significant which shows the leverage effect. Negative news produces more volatility in the return than positive during the period for BSE.

Table 3.

Estimated Parameters for GJR-GARCH (1, 1), Whole Period

Parameter	Estimate	Std Error	p-value
μ	0.00056	0.00018	0.00167
ω	0.00005	0.00002	0.02734
α	0.07313	0.01009	0.00000
β	0.87081	0.01198	0.00000
γ	0.07857	0.01615	0.00000

Source: Authors’ calculation.

Sign bias test (Table 4) examines the squared normalized residual by some variable observed in the past (Engle & Victor). The table shows sign bias test for GJR-GARCH model. The sign bias, negative sign bias and positive sign bias have t-values less than 2 and are statistically insignificant. So this is evidence towards the robustness of the model.

Table 4.

Sign Bias Test GJR-GARCH (1, 1), Whole Period

	t-value	p-value
Sign bias	0.4277	0.6689
Negative sign bias	1.1020	0.2705
Positive sign bias	0.2267	0.8207
Joint effect	3.7719	0.2872

Source: Authors’ calculation.

Pre-recession

The ARCH-LM test with 12 lags was conducted for the recession period to test for heteroscedasticity; it shows the presence of ARCH effect which should be modelled by ARCH or higher models. In this period also, GARCH (1, 1), EGARCH (1, 1), GJR-GARCH (1, 1) and APARCH (1, 1) models have been used to test for the best-suited model for the data. Table 5 shows the result obtained for selection criteria.

Table 5.

Comparative Information Criteria for Pre-recession Period

	Garch11	Egarch11	Gjrgarch11	Aparch11.1
Akaike	–5.563174	–5.565688	–5.569976	–5.557748
Bayes	–5.556548	–5.557406	–5.561694	–5.549466
Shibata	–5.563176	–5.565691	–5.569979	–5.557751
Hannan–Quinn	–5.560818	–5.562743	–5.567031	–5.554803

Source: Authors’ calculation.

The values of AIC, BIC, SIC and Hannan–Quinn tests are smallest for GJR-GARCH (1, 1) model, so this model fits for the data for the whole period. So GJR-GARCH (1, 1) is used to model the volatility and study the leverage effect for BSE Sensex for the period mentioned above.

Table 6 shows the estimated parameters for GJR-GARCH (1, 1); the estimates are significant with t-value greater than |2|. γ in the output is positive and significant which shows the leverage effect. Negative news produces more volatility in the return than positive during the period for BSE.

The sign bias test (Table 7) for the model is statistically insignificant. The t-values of sign bias, positive and negative sign bias are less than 2 and are insignificant, providing evidence towards the GJR-GARCH (1, 1) model.

Table 6.

Estimated Parameters, GJR-GARCH (1, 1), Pre-recession

Parameter	Estimate	Std Error	p-value
μ	0.00076	0.00025	0.00070
ω	0.00001	0.00000	0.00000
α	0.08789	0.00647	0.00000
β	0.828764	0.00956	0.00000
γ	0.08924	0.01621	0.00000

Source: Authors’ calculation.

Table 7.

Sign Bias Test, GJR-GARCH (1, 1), Pre-recession

	t-value	p-value
Sign bias	0.9574	0.3384
Negative sign bias	0.9535	0.3404
Positive sign bias	0.3190	0.7497
Joint effect	4.5592	0.2071

Source: Authors’ calculation.

Recession Period

The ARCH-LM test with 12 lags was conducted for the recession period to test for heteroscedasticity. It shows the presence of ARCH effect (p = 0.043) which should be modelled by ARCH or higher models. In this period also, GARCH (1, 1), EGARCH (1, 1), GJR-GARCH (1, 1) and APARCH (1, 1) models have been used to test for best-suited model for the data. Table 8 shows the result obtained for selection criteria.

The values of AIC, BIC, SIC and Hannan–Quinn for APARCH (1, 1) model are minimum. It is the best-suited model for the recession period to model the volatility (Table 9).

Table 8.

Comparative Information Criteria for Recession Period

	Garch11	Egarch11	Gjrgarch11	Aparch11.1
Akaike	–4.400850	–4.432257	–4.406708	–4.441748
Bayes	–4.359618	–4.380716	–4.355167	–4.390207
Shibata	–4.401066	–4.432592	–4.407043	–4.442083
Hannan–Quinn	–4.384494	–4.411812	–4.386263	–4.421303

Source: Authors’ calculation.

Table 9.

Estimated Parameters, APARCH (1, 1), Recession Period

Parameter	Estimate	Std Error	p-value
μ	–0.00632	0.00124	0.61025
ω	0.00160	0.00664	0.01552
α	0.11672	0.03358	0.00051
β	0.85144	0.03629	0.00000
γ	0.43511	0.20342	0.03244

Source: Authors’ calculation.

In this model also, γ is positive and statistically significant which shows the leverage effect in the sensex return volatility. The volatility in return during recession period is more impacted by negative news than positive news.

The sign bias test (Table 10) for the model is statistically insignificant. The t-value of sign bias, positive and negative sign bias are less than 2 and are insignificant, providing evidence towards the APARCH (1, 1) model.

Table 10.

Sign Bias Test, APARCH (1, 1), Recession Period

	t-value	p-value
Sign bias	0.4072	0.6841
Negative sign bias	0.2502	0.8026
Positive sign bias	0.3912	0.6959
Joint effect	0.5122	0.9162

Source: Authors’ calculation.

Post Recession

In this period also, GARCH (1, 1), EGARCH (1, 1), GJR-GARCH (1, 1) and APARCH (1, 1) models have been used to test for the best-suited model for the data. Table 11 shows the result obtained for selection criteria.

The values of AIC, BIC, SIC and Hannan–Quinn for GJR-GARCH (1, 1) model are minimum. It is the best-suited model for the recession period to model the volatility.

Table 11.

Comparative Information Criteria for Post-recession Period

	Garch11	Egarch11	Gjrgarch11	Aparch11.1
Akaike	–6.222841	–6.250724	–6.251920	–6.227059
Bayes	–6.205942	–6.229601	–6.230796	–6.205935
Shibata	–6.222863	–6.250759	–6.251954	–6.227093
Hannan–Quinn	–6.216477	–6.242769	–6.243965	–6.219104

Source: Authors’ calculation.

Table 12 shows the estimated parameters for GJR-GARCH (1, 1); the estimates are significant with t-value greater than |2|. γ in the output is positive and significant which shows the leverage effect. Negative news produces more volatility in the return than positive during the period for BSE.

Table 12.

Estimated Parameters, GJR-GARCH (1, 1), Post Recession

Parameter	Estimate	Std Error	p-value
μ	0.000231	0.000279	0.40687
ω	0.000003	0.000000	0.00000
α	0.00000	0.003890	1.00000
β	0.918792	0.009560	0.00000
γ	0.112611	0.019136	0000000

Source: Authors’ calculation.

The sign bias test for the model is statistically insignificant. The t-value of sign bias, positive and negative sign bias are less than 2 and are insignificant, providing evidence towards the GJR-GARCH (1, 1) model (Table 13).

Table 13.

Sign Bias Test, GJR-GARCH (1, 1), Post Recession

	t-value	p-value
Sign bias	0.2724	0.7853
Negative sign bias	0.8357	0.4035
Positive sign bias	1.1811	0.2378
Joint effect	2.1164	0.5486

Source: Authors’ calculation.

The findings of the GJR-GARCH model are supported by the news impact curve. From the graph, it can be interpreted that the future volatility is high for bad news than good news in the current scenario. This period has experienced growth due to IT boom, foreign policy impact, globalization, etc. The period of recession was because of the American crisis and in the post-recession period investors were sceptical about their investments. For pre-recession period, bad news has produces almost similar volatility when compared to the whole period but volatility caused by positive news is less than that for whole period. This can be due to the length of the boom period which is longest when compared to other time periods. For the recession period, the volatility caused by bad news is more than whole period, pre-recession and post-recession period because of the global gloom prevailing during that period. Also positive news created less volatility during this period when compared to other periods due to recession itself. The condition after recession has become almost constant as per the news impact curve; however, the intensity of volatility caused by the negative news is much lower than for other time periods. Post recession, the behaviour of investors has inclined to less riskier options, such as bank deposits and debts; this may have influenced the Indian investors creating sceptism in the mindset and leading to opt for safer options. Bad news post recession has least influence on volatility than any of the periods mainly due to the bitter experience during recession (Figure 1.1, 1.2, 1.3 and 1.4).

Conclusion

The volatility of Indian stock market was investigated taking BSE 30 index daily returns as proxy which was modelled using the suitable assymetric GARCH models with normal distribution of errors based on AIC, Hannan–Quinn information criteria, BIC and SIC for the period from 1992 to 2014 to study the leverage effect for the period and sub-periods. The sub-periods were divided on the basis of economic gloom prevailing due to great American depression of 2008–2009. The period prior to recession witnessed growth due to growth in IT and economic liberalization, and in the post-recession period, investors were sceptical about their investments because of losses arising from the recession. Sign bias test, news impact curve and asymmetric GARCH output confirm the leverage effect for the period and sub-periods. It was observed that the leverage effect for the recession period was more than other periods corresponding to the economic gloom during the period, while the post-recession period shows constant volatility on positive news arrival from its news impact curve.

Since liberalization, the Indian stock market has attracted investors from throughout the world, which has brought wealth. This wealth integrates Indian financial market with outside economy also, which can pose risks like those observed in the sub-prime financial crisis. The asymmetric effect poses great risk by making the investments riskier at the time of financial gloom. This study will assist investors, traders and regulators in taking informed decision by estimating expected volatility more precisely considering the leverage effect.

References

Adams

, McQueen

, & Wood

, (2004). The impact of inflation news on high frequency stock returns. The Journal of Business, 77(3), 547–574.

Alberg

, Shalit

, & Yosef

(2008). Estimating stock market volatility using asymmetric GARCH models. Applied Financial Economics, 18(15), 1201–1208.

Batra

(2004). Stock return volatility patterns in India (pp. 1–34). Working Paper, 124. India: Indian Council for Research on International Economic Relations.

Bekaert

, & Wu

(2000). Asymmetric volatility and risk in equity markets. Review of Financial Studies, 13(1), 1–42.

Black

(1976) Studies of stock market volatility changes. Proceedings of the American Statistical Association, Business and Economic Statistics Section, pp. 177–181, Washington DC, US.

Bollerslev

(1986). Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics, 31(3), 307–327.

Braun

P.A.

, Nelson

D.B.

, & Sunier

A.M.

(1995). Good news, bad news, volatility, and betas. The Journal of Finance, 50(5), 1575–1603.

Cappiello

, Engle

R.F.

, & Sheppard

(2006). Asymmetric dynamics in the correlations of global equity and bond returns. Journal of Financial Econometrics, 4(4), 537–572.

Conrad

, Cornell

, & Landsman

W.R.

(2002). When is bad news really bad news? The Journal of Finance, 57(6), 2507–2532.

10.

Ding

, Granger

C.W.

, & Engle

R.F.

(1993). A long memory property of stock market returns and a new model. Journal of empirical finance, 1(1), 83–106.

11.

Engle

(1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica, 50(5), 987–1007.

12.

Flannery

, & Protopapadakis

(2002). Macroeconomic factors do influence aggregate stock returns. The Review of Financial Studies, 15(3), 751–782.

13.

Glosten

L.R.

, Jagannathan

, & Runkle

D.E.

(1993). On the relation between the expected value and the volatility of the nominal excess return on stocks. Journal of Finance, 48(5), 1779–1801.

14.

Goudarzi

, & Ramanarayanan

C.S.

(2011). Modeling asymmetric volatility in the Indian stock market. International Journal of Business and Management, 6(3), 221.

15.

Hamilton

J.D.

, & Susmel

(1994). Autoregressive conditional heteroskedasticity and changes in regime. Journal of Econometrics, 64(1), 307–333.

16.

Karmakar

(2007). Asymmetric volatility and risk-return relationship in the Indian stock market. South Asia Economic Journal, 8(1), 99–116.

17.

Koutmos

, & Booth

G.G.

(1995). Asymmetric volatility transmission in international stock markets. Journal of International Money and Finance, 14(6), 747–762.

18.

Liu

, Wong

, An

, & Zhang

(2014). Asymmetric information and volatility forecasting in commodity futures markets. Pacific-Basin Finance Journal, 26, 79–97.

19.

McQueen

, & Roley

V.V.

(1993). Stock prices, news and business conditions. The Review of Financial Studies, 6(3), 683–707.

20.

Mittal

A.K.

, Arora

D.D.

, & Goyal

(2012). Modelling the volatility of Indian stock market. Gitam Journal of Management, 10(1), 224–243.

21.

Nelson

D.B.

(1991). Conditional heteroskedasticity in asset returns: A new approach. Econometrica, 59(2), 347–370.

22.

Seďa

(2011). Asymmetric conditional volatility modeling: Evidence from Central European stock markets. Proceedings of Finančnířízenípodniků a finančníchinstitucí 2011, pp. 375–383, Ostrava, Czech Republic.

23.

Tripathy

(2010). The empirical relationship between trading volumes & stock return volatility in Indian stock market. European Journal of Economics, Finance and Administrative Sciences, (24), 1450–2275.