Stock Market Crash and Stock Return Volatility: Empirical Evidence from Dhaka Stock Exchange

Transcription

1 Bangladesh Developmen Sudies Vol. XXXVIII, Sepember 015, No. 3 Sock Marke Crash and Sock Reurn Volailiy: Empirical Evidence from Dhaka Sock Exchange K. M. ZAHIDUL ISLAM * SAYED FARRUKH AHMED ** This paper empirically invesigaes he impac of sock marke crash on he volailiy of Dhaka Sock Exchange sock reurn of Bangladesh wih GARCH-ype framework by using daa of daily closing sock price indices of (DSE General Index) over he period from 9 November 004 o 31 July 013. The resuls of GARCH-M (1,1) model conclude ha condiional sandard deviaion is negaively relaed o he level of reurns. While his resul is no consisen wih he heory of a posiive risk premium on sock indices, in special circumsances invesors may no claim higher risk premium if hey are compeen enough o bear risk a imes of specific volailiy. Moreover, he model also confirms ha sock marke crash affeced he volailiy of DSE General Index reurn and here is a propensiy for he volailiy o erode over ime. Keywords: Dhaka Sock Exchange, Volailiy, Crash, GARCH, Bangladesh JEL Classificaion: C, C5, G1, E3 I. INTRODUCTION Sock marke crash is an inense and unanicipaed decline of sock marke prices for a shor period of ime which makes marke paricipans panicked and causes significan invesmen losses o invesors. A sock marke crash has unpleasan impac on capial which may have long erm or permanen effec on he invesmen climae of he counry. I may erode public confidence gradually o inves in sock marke. When sock markes crash, everyone becomes very worried, everyone wans o sell and nobody wans o buy in anicipaion of furher fall in prices. A he individual level, many invesors may remain under consan panic and a sense of severe anguish may work in hem. When an invesor observes ha her porfolio has dropped in value, she may decide o pull * Associae Professor, Insiue of Business Adminisraion, Jahangirnagar Universiy, Savar, Dhaka, Bangladesh. ** Assisan Professor, Faculy of Business & Economics, Daffodil Inernaional Universiy, Dhaka, Bangladesh.

2 Bangladesh Developmen Sudies ou of socks alogeher and focus on safer, lower-risk invesmen securiies like bonds and noes commonly known as fixed income securiies. Ahmed, Uchida and Islam (01) sae ha Bangladesh has experienced wo major sock marke crashes one in 1996 and he oher in 010. There was a big bubble in 1996 followed by a caasrophic crash. The index poin of Dhaka Sock Exchange (DSE) jumped from abou 800 poins in June 1996 o around 3,600 poins in November 1996.Afer ha, he index began o fall. Large invesors lef he marke wih huge profi, leaving he small invesors in he siuaion of sharp price fall wih many losing heir capial. The burs of he sock marke bubble in 1996 creaed a negaive impression in he mind of invesors. The marke winessed anoher severe bou of decline in 010 ha resuled in wiping ou even he iniial capial invesmen made by housands of wreched and illfaed invesors. The afershocks from he marke crash of lae 010 coninued o be fel wih he index poin falling 3,03 poins in 011 and a furher 1,038 poins in 01. Figure 1: DSE daily DGEN index beween Ocober 010 o February 011 9,000 8,500 8,000 7,500 7,000 6,500 6,000 5,500 5,000 DGEN On 5 h December 010, DSE General Index (DGEN) reached is all-ime high of 8, poins following a sharp rise in Ocober-November 010. Subsequenly, he DSE winessed a dramaic decline. Choudhury (013) poined ou ha lack of resricion on opening beneficiary owner (B/O) accouns, speculaion hrough omnibus accouns, wrong placemen of IPOs, violaion of banking ac, rumour spread by he brokers and he dealers, wrong mehod used in face value deerminaion and lack of monioring in he

3 Islam & Ahmed: Sock Marke Crash and Sock Reurn Volailiy 7 share marke were behind he sock marke crash in Bangladesh. In he wake of he dramaic crash of December 010, a commiee commonly known as The Probe Commiee (Khaled 011) was formed o invesigae he sock marke scam. The commiee accused 60 influenial individuals o be responsible for he crash. I found an array of irregulariies, including he exisence of omnibus accouns ha allowed key marke players o make excessive profis a he cos of he reail invesors. In Bangladesh, a few sudies (Choudhury 013, Ahmed, Uchida and Islam 01) have been conduced o explore he reasons and consequences of sock marke crashes. Empirical sudies on he relaionship beween sock marke crash and sock price volailiy is scarce. In his conex, he presen sudy aims o answer he following research quesion: Does sock marke crash cause sock price volailiy? Using daily daa of closing sock price indices of DSE for he period from 9 November 004 o 31 July 013, his paper invesigaes he impac of he sock marke crash on he volailiy of DSE sock marke reurns in Bangladesh. This is he firs sudy of is kind for he Bangladesh sock marke and we expec ha he resuls of he sudy will be of ineres o invesors, policymakers and academics. II. LITERATURE REVIEW A number of sudies have been conduced around he world regarding sock marke crash and sock marke reurn. Choudhry (1996) invesigaed he volailiy of sock markes of Argenina, Greece, India, Mexico, Thailand and Zimbabwe by using monhly daa ranging from January 1976 o Augus Using GARCH-M model, he sudy found differences in he ARCH parameer, risk premium and exisence of volailiy before and afer he 1987 crash. The sudy concluded ha he changes vary from marke o marke. According o Mishkin and Whie (00), sock markes infrequenly experience srong and sudden fall of prices, which may ake place afer a period of srong sock price increases. Their sudy concluded ha he key reasons behind he sock marke crash are illogical marke behaviour, failure of regulaory auhoriy o supervise he marke suiably, excessive liquidiy in he marke, defecive lising sysem, issuance of righ shares and preference shares a high price, sock manipulaions by insider rading, lack of knowledge of he invesors abou he marke, avoidance of company basics by he invesors and excessive greed of invesors o make profi in shor duraion of imes.

4 8 Bangladesh Developmen Sudies Using he daily daa of S&P 500 index (closing prices) (American Sock Marke Index) which is based on he marke capialisaion of 500 large companies having common sock lised on he NEW York Sock Exchange (NYSE) or Naional Associaion of Securiies Dealers Auomaed Quoaions (NASDAQ) of he period ranging from 4 January 198 o 13 March 009, Nikolaos and Scharler (009) observed a move ino high volailiy quie some ime before he crash, which persiss for a considerable period of ime even afer he crash. Singh and Makkar (014) examined he relaionship beween he financial crisis and sock reurns volailiy of he Indian banking secor by using daily daa of closing sock prices spanning from January 1, 004 o December 31, 01. They applied a GARCH model o capure he impac of he crisis on he volailiy of banks. Their resuls found significan impac of crisis on he sock volailiy of he Indian banking secor and hey claimed ha sock reurn volailiy has significanly changed during he pre- and pos- crisis period. Kishor and Singh (014) examined he impac of he news of he US sock marke on he sock reurns volailiy of he BRICS sock markes hrough a GARCH model by using daily daa of he period saring from 1 January 007 o 31 December 013. Based on he esimaed resuls, he auhors deduced ha he news of he US sock marke has significanly affeced some of he BRIC counries (e.g., India and Souh Africa). Some sudies found significan impac of sock marke crash on sock reurn volailiy while ohers highlighed he imporance of favourable economic condiions in relaion o he sock marke crash. Some sudies even concluded ha news abou he sock marke significanly affec he sock reurn volailiy. In his backdrop, he presen sudy aemps o empirically invesigae he impac of marke crash on he volailiy of DSE sock marke reurn in Bangladesh. III. DATA AND EMPIRICAL MODEL Daily closing sock price indices of DSE in erms of DSE General Index (DGEN) for he period from 9 November 004 o 31 July 013 consising of oal observaions of,074 are used in he sudy. Afer reaching he all-ime highes poin (8, ) on 5 December 010, DSE General Index (DGEN) winessed successive falls of index poins in laer periods (Bhuiyan, 011). So, he period from 6 December 010 o 31July 013 has been considered as crash period. The sock marke reurn (RDGEN) is calculaed as he naural log difference of

5 Islam & Ahmed: Sock Marke Crash and Sock Reurn Volailiy 9 DGEN: PI R =ln 100, where PI is he price index a ime and PI -1 is he PI-1 price index a ime -1. The daa of DGEN have been colleced from differen publicaions of DSE. This paper examines he impac of sock marke crash on he volailiy of sock marke reurn of DSE. We employ a model of variances using he generalised ARCH (GARCH) formulaion for he sudy. Bollerslev (1986) inroduced he GARCH model ha considers he condiional variance as a funcion of he preceding period s squared errors and condiional variances o explain he volailiy evoluion of sock-reurn series. The GARCH models can recognise he endency for volailiy clusering in financial daa. Volailiy clusering in sock reurn indicaes ha large price changes are followed by large price changes and small price changes are followed by small price changes (Brooks 008). GARCH-M model, inroduced by Engle, Lilien and Robins (1987), allows he condiional mean o depend on is own condiional sandard deviaion. Therefore, he GARCH-M (1,1) model has he following form: Mean Equaion: r = + + u (1) Variance Equaion: = where, r = Index reurn on day ; = Variance of he error erm; 1 u 1 = Previous period s volailiy (he ARCH erm); 1 = Previous period s variance (he GARCH erm); u Di () D i = Dummy variable equal o 1 for observaion during crash period and 0 for he res of he period. This dummy variable for sock marke crash is added as an explanaory variable in he condiional variance equaion o invesigae he impac of sock marke crash on he volailiy of he DSE sock marke reurn (RDGEN). The presence of in equaion (1) helps us o sudy he rade-off beween risk and expeced reurn. The coefficien of,, helps o capure he influence of volailiy on sock reurns. The coefficien,, represens he risk premium parameer. A significan posiive coefficien,, indicaes ha he invesors are

6 30 Bangladesh Developmen Sudies rewarded wih higher reurns for bearing higher risks (Brooks, 008). Conversely, a significan negaive coefficien,, indicaes ha he invesors are penalised for aking risks. Glosen, Jagannahan and Runkle (1993) saed ha boh posiive and negaive relaionships beween curren reurns and curren risks are possible. From equaion (), i can be said ha he curren fied variance,, is a weighed funcion of a long-erm average value (relian on 0 ), informaion 1 abou volailiy during he prior period ( 1 u ), he fied variance from he model during he prior period ( 1 1 ) and crash dummy ( Di ).The consrains on he coefficiens 1 0 and 1 0 are imposed o make sure ha he condiional variance is posiive (Poon 005). IV. EMPIRICAL FINDINGS Table I provides he resuls of he descripive saisics of he variables used in he sudy. From Table I, i is eviden ha he average daily reurn for RDGEN is 0.044% during sample period, which is 0.113% before crash period and % during crash period. The reurn ranges from -9.33% o 0.38% during sample period, from -7.36% o 0.38% before crash period and from -9.33% o 14.48% during crash period. The risk (sandard deviaion) per day for RDGEN is 1.71% during sample period, 1.33% before crash period and.35% during crash period. RDGEN has posiive skewness wih es saisic of 0.80 during sample period,.3 before crash period and 0.0 during crash period. TABLE I DESCRIPTIVE STATISTICS OF RDGEN Saisics RDGEN (Full sample period) RDGEN (Before crash period) RDGEN (During crash period) Mean (%) Maximum (%) Minimum (%) Sd. Dev. (%) Skewness Kurosis Jarque-Bera a (p-value) Observaion, Noe: a Under he null hypohesis of a normal disribuion, he p-values indicae rejecion of he null hypohesis, hence deviaion from normaliy.

7 Islam & Ahmed: Sock Marke Crash and Sock Reurn Volailiy 31 The es saisic of kurosis of RDGEN is 18.0 during he sample period, before he crash and 6.89 during he crash period (which exceed he benchmark of 3). The Jarque-Bera es shows deviaions from normaliy in all cases. In order o deec he saionariy of he variable, RDGEN, wo popular uni roo ess, namely he Augmened Dickey-Fuller (ADF) and Phillips-Perron (PP) ess, have been applied. The resuls are presened in Table II. TABLE II THE AUGMENTED DICKEY-FULLER (ADF) AND PHILIPS-PERRON (PP) TESTS RESULTS Variable ADF Tes (Level) PP Tes (Level) Saionary Drif Drif and Trend Drif Drif and Trend RDGEN *** *** *** *** Yes Noe: ***Denoes he rejecion of null hypohesis ha he ime-series is non-saionary a 1% significance level. Boh ADF es and PP es confirm ha RDGEN is saionary in level a 1% significance level. This means ha he variable is suiable for he analysis. TABLE III ARCH-LM TEST Variable Obs*R-squared p-value ARCH effec RDGEN *** Yes Noe: *** denoes he rejecion of null hypohesis of no ARCH effec a 1% significance level The ARCH-LM es of RDGEN (see Table III) o deec he ARCH effec in he residuals of mean equaion confirms ha here exis ARCH effecs in he residuals of RDGEN. Figure : Residuals of RDGEN

8 3 Bangladesh Developmen Sudies From Figure, i is apparen ha here is presence of volailiy clusering, which indicaes ha he curren level of volailiy ends o be posiively correlaed wih is level during he immediae preceding periods. Furher, here exis several spikes wih high volailiy which persis for a number of periods. The resuls of GARCH-M (1,1) model are presened in Table IV. The esimaed coefficien of condiional sandard deviaion erm on he mean equaion is found o be negaive and saisically significan a 1% significance level, which indicaes ha condiional sandard deviaion used as a proxy for risk of reurn is negaively relaed o he level of reurns. This resul is no consisen wih he heory of a posiive risk premium on sock indices (Brooks 008). One plausible explanaion of such finding is ha in special circumsances invesors may no claim higher risk premium if hey are beer able o bear risk a imes of specific volailiy. This resul is similar o wha Glosen e al. (1993) repored in heir sudy. Mean Equaion C RDGEN(-1) Variance Equaion C TABLE IV RESULTS OF GARCH-M (1,1) MODEL ON RDGEN 1 1 GARCH-M(1,1) (0.0790) a *** (0.08) a *** (0.0617) a (0.0468) a *** (0.0418) a *** (0.0415) a *** (0.1065) a Diagnosic Checking ARCH LM es [0.791] b Normaliy es of residuals (0.000) b Ljung-Box Q Saisic Significan ** Observaions,074 Noes: a Sandard errors are in parenheses; b p-values are in brackes; ** and *** denoe significance a 5% level and 1% level respecively. Moreover, he presence of dummy variable in he condiional variance equaion is saisically significan and posiive by which one can conclude ha sock marke crash affeced he volailiy of DSE General Index reurn. Besides

9 Islam & Ahmed: Sock Marke Crash and Sock Reurn Volailiy 33 his, he coefficiens on boh he lagged squared residual and lagged condiional variance erms in he condiional variance equaion are highly saisically significan. The resuls show ha he sum of he esimaed coefficiens on he lagged squared error and lagged condiional variance (approximaely 0.88) is lower han uniy which implies a endency for he volailiy response o erode over ime. The LM es confirms ha here is no ARCH effec afer GARCH-M (1,1) esimaion as he null hypohesis of no ARCH effec canno be rejeced a 5% level of significance. The normaliy es of residuals indicaes ha residuals are no normally disribued and he Ljung-Box Q saisic es confirms ha residuals are no free from serial correlaions. V. CONCLUSION AND POLICY IMPLICATIONS This paper has empirically invesigaed he impac of sock marke crash on sock reurn volailiy of Dhaka Sock Exchange wih GARCH-ype framework and he persisence of shocks o volailiy based on daily closing sock price indices of DSE (DSE General Index) over he period from 9 November 004 o 31 July 013. The resuls of GARCH-M (1,1) model conclude ha condiional sandard deviaion used as a proxy for risk of reurn is negaively relaed o he level of reurns. One plausible explanaion of such finding is ha in paricular condiions invesors may no claim higher risk premium if hey are capable enough o olerae risk a imes of specific volailiy. Moreover, he presence of dummy variable in he condiional variance equaion is saisically significan, which indicaes ha sock marke crash affeced he volailiy of DSE General Index reurn and here is a endency for he volailiy response o erode over ime. The implicaion of his sudy is ha he regulaory auhoriies such as Bangladesh Securiies and Exchange Commission may acively monior he reasons for sock marke crash in Bangladesh and formulae and implemen rules and regulaions o avoid such collapse in fuure. The sock marke crash desiued a large number of ill-faed small invesors who los almos whole of heir invesmen. Proper remedial incenives o he affeced invesors and sric legal acions agains he conviced people behind he crash may recover he fragile confidence of he invesors abou sock marke. The sudy focused on he co-movemen of sock marke volailiy and sock marke crash wih limied daase of abou en years only. The analysis relied on only one variable like sock index price, while here are many oher facors ha may have profound influence on volailiy of sock marke. Fuure research may be conduced applying co-inegraion es and granger causaliy es o idenify

10 34 Bangladesh Developmen Sudies he relaionship beween sock marke volailiy and sock marke crash along wih some addiional explanaory variables like poliical insabiliy, srikes, liquidiy crisis and oher macroeconomic variables ha are likely o influence sork price volailiy. REFERENCES Ahmed, S. U., S. Uchida and M. Z. Islam. 01. Sock Marke Crashes: Reasons Implicaions. KEIEI TO KEIZAI, 91(4): Bhuiyan, K Small invesors anger may urn ino big issue! Accessed June 19, 015, from hp://bangladeshinside.com/ invesmen/small-invesors-anger-mayurn-ino-big-issue-63 Bollerslev, T Generalized Auoregressive Condiional Heeroscedasiciy. Journal of Economerics, 31: Brooks, C Inroducory Economerics for Finance ( nd ed.). New York: Cambridge Universiy Press. Choudhry, T Sock Marke Volailiy and he Crash of 1987: Evidence from Six Emerging Markes. Journal of Inernaional Money and Finance, 15(6): Choudhury, M. A. H Sock Marke Crash in 010: An Empirical Sudy on Reail Invesor s Percepion in Bangladesh. ASA Universiy Review, 7(1): Engle, R. F., D. M. Lilien and R. P. Robins Esimaing Time Varying Risk Premia in he Term Srucure: The ARCH-M model. Economerica, 55: Glosen, L., R. Jagannahan and D. Runkle On he Relaion beween he Expeced Value and he Volailiy of Nominal Excess Reurn on Socks. Journal of Finance, 48: Khaled, K. I Invesigaion Repor of Probe Commiee. Dhaka. Kishor, N. and R. P. Singh Sock Reurn Volailiy Effec: Sudy of BRICS. Transnaional Corporaions Review, 6(4): Mishkin, F. S. and E. N. Whie. 00. U.S. Sock Marke Crashes and Their Afermah: Implicaions for Moneary Policy. NBER Working Paper No Nikolaos, A. and J. Scharler Volailiy, Informaion and Sock Marke Crashes. Working Paper No Johannes Kepler Universiy of Linz. Poon, S. H. 005: A Pracical Guide o Forecasing Financial Marke Volailiy. Wiley Finance, England. Singh, S. and A. Makkar Relaionship beween Crisis and Sock Volailiy: Evidence from Indian Banking Secor. IUP Journal of Applied Finance, 0():