EMPIRICAL ANALYSIS OF STOCK RETURN VOLATILITY WITH REGIME CHANGE: THE CASE OF VIETNAM STOCK MARKET. Vietnam Development Forum Tokyo Presentation

Transcription

1 EMPIRICAL ANALYSIS OF STOCK RETURN VOLATILITY WITH REGIME CHANGE: THE CASE OF VIETNAM STOCK MARKET. Vienam Developmen Forum Tokyo Presenaion By Vuong Thanh Long Deparmen of Economic Developmen and Policies Graduae School of Inernaional Cooperaion Sudies (GSICS) Kobe Universiy Keywords: ARCH model, GARCH model, Sock Marke, Sock Reurn Volailiy, Regime Change, Financial Liberalizaion. ii

2 SUMMARY This hesis sudies he characerisic of he sock reurn volailiy in he Vienam sock marke (VSM) and shows how i changes when regime changes are aken ino consideraion. Moreover, he effecs of financial liberalizaion on volailiy are also examined and analyzed. Firs, his hesis ess he commonly known hypohesis of highly persisen volailiy of sock reurn by using a (generalized) auoregressive condiional heeroskedasiciy (ARCH/GARCH) model. The VSM sock reurn volailiy is found o be eviden. Nex, he ieraed cumulaive sums of squares (ICSS) algorihm is used o idenify he break poins (shocks) ha lead o he changes in he sock reurn rae variances. Following ha, he conrol for he VSM reurn volailiy regimes is incorporaed ino he above-menioned ARCH/GARCH model by using a se of dichoomous dummy variables. The volailiy regimes and heir lifeimes are repored. All he evens ha are possibly relaed o hese volailiy regimes are idenified, and heir causes and effecs are analyzed. Specifically, i is found ha financial liberalizaion has a negaive influence on he volailiy of sock reurn in VSM. Because he las financial liberalizaion in VSM also coincided wih a large increase in he number of IPOs, i is hard o clearly separae he influence of financial liberalizaion and ha of he growing number of IPOs in he domesic equiy marke. The observaion period, herefore, needs o be coninued for a longer ime span so ha jusified conclusions can be drawn. iii

3 TABLE OF CONTENTS SUMMARY...ii TABLE OF CONTENTS...iv I. THEORETICAL FRAMEWORK... A. Lieraure Review... B. Daa for Analysis...3 C. The Economeric ARCH/GARCH Model...5 D. Deecing and Incorporaing he Regime Change: The Combined Model...6 II. EMPIRICAL RESULTS...8 A. Empirical Resuls of he ARCH/GARCH Model wihou Regime Change...8 B. Deecion of Break Poins and Volailiy Regime Effec Examinaion...9 C. Empirical Resuls of he Combined Model...3 III. CONCLUSION AND POLICY IMPLICATIONS... A. Summary of Vienam Sock Marke Volailiy... B. Policy Implicaions for Vienam Based on he Empirical Resuls....5 REFERENCE:...7 APPENDIX A: TABLES...9 Table A: Correlogram and Q-saisic of Vnindex Daily Raes of Reurn....9 Table A: Uni Roo Tes of Vnindex Daily Raes of Reurn... Table A3: Regression Resul of he ARCH () Model wihou Regime Change... Table A: Regression Resul of he Combined ARCH () Model wih Regime Change.... Table A5: ARCH Tes Resul of he ARCH () Model wihou Regime Change... Table A6: ARCH Tes Resul of he Combined ARCH () Model wih Regime Change...3 APPENDIX B: FIGURES... Figure B: VSM Reurn Condiional Sandard Deviaion wihou Regime Change.... Figure B: VSM Reurn Condiional Sandard Deviaion wih Regime Change... Figure B3: Plo of Sandardized Dk from k= o k=3...5 Figure B: Plo of Sandardized Dk from k= o Figure B5: Plo of Sandardized Dk from k=3 o end of series...6 Figure B6: Plo of Sandardized Dk from k=3 o Figure B7: Plo of Sandardized Dk from k=8 o end of series...7 Figure B8: Plo of Sandardized Dk from k=8 o Figure B9: Plo of Sandardized Dk from k=8 o Figure B: Plo of Sandardized Dk from k=66 o Figure B: Plo of Sandardized Dk from k=6 o end of series...9 Figure B: Plo of Sandardized Dk from k=3 o Figure B3: Plo of Sandardized Dk from k= o k=5...3 Figure B: Plo of Sandardized Dk from k=66 o end of series...3 Figure B5: Vnindex and Trading Volume of HoSTC from July, o May, Figure B6: Hisogram of Residuals from Equaion (6) Regression wihou Regime Change...3 Figure B7: Hisogram of Residuals from Equaion (8) Regression wih Regime Change...3 iv

4 I is widely known in finance research ha sock reurn volailiy is highly persisen, especially in he developing markes. Many researchers have provided evidence concerning his characerisic of sock reurn volailiy using he class of ARCH/GARCH model (e.g., Engle 98, Bollerslev 986). When regime changes are aken ino consideraion, his highly persisen volailiy reduces. In addiion, while globalizaion is becoming a mainsream, economiss have by no means agreed upon wheher financial liberalizaion is a good choice for developing counries. As reviewed by Fry (995, Par I), here are wo differen wellknown houghs in he lieraure. The Keynesians discourage he financial liberalizaion, being afraid of is desabilizing effecs. On he conrary, he financial liberalizaion heoriss, namely McKinnon (973) and Shaw (973), advocae liberalizaion because hey hink i helps o produce he long-erm sabiliy. Vienam Sock Marke (VSM) is growing rapidly and has been underaking wo financial liberalizaion seps. Hence, i provides an ineresing example o review he wo conrasing perspecives concerning sock reurn volailiy and he effecs of financial liberalizaion. I has been years since Vienam sared is well-known Doi Moi policy, which means renovaion, o open is economy o he res of he globe. The counry has made considerable achievemens from his imely decision. In he beginning of he 99s, he former Sovie Union and Easern European communis counries underwen abrup changes ha were followed by upheavals in erms of boh economic and social srucure. Forunaely, Vienam had overcome ha period wih relaive sabiliy, while gradually adaping o he new world order and enjoying relaively high economic growh. The GDP growh rae of Vienam has been around 7 percen annually over he pas decade. High economic growh commonly leads o accompanying changes in he economic srucure. The financial secor is he skeleon of he economy, and i is usually required ha i be adjused well in advance, so ha changes in oher economic secors can follow. The banking sysem in Vienam, formerly meaning only he cenral bank and is four affiliaes, firs underwen fundamenal reform, signaling he change owards marke mechanism. These affiliaes were separaed during period and renamed he sae-owned commercial banks. Following ha, he financial secor began o develop wih ens of privae and join sock banks. These new banks have been improving and diversifying heir services o cach up wih modern banking faciliy and services required by Vienam s inegraion ino he world economy. However, recenly he domesic capial marke has reached a criical juncure where he supply of domesic bank loans canno mee he growing demand for medium o long-erm finances (wihou facing he mismach risk). Meanwhile, even hough he counry has been running small curren accoun and budge deficis, he imperaive o raise more funds o suppor economic growh may, however, resul in an increasing budge defici in he near fuure. Furhermore, he equiizaion/privaizaion process requires he parallel emergence and developmen of a more user-friendly equiy marke. As a resul, Vienam Sock Marke (VSM) came ino being as a maer of course. The preparaion process for he birh of he sock marke sared wih he esablishmen of he Sae Securiies Commission (SSC) in he middle of he 99s. The firs securiies rading cener of VSM was launched in July in Ho Chi Minh Ciy (hereinafer called HoSTC) as a pilo projec. The second securiies rading cener was esablished in Hanoi (herein afer called HaSTC) five years laer in March 5. From July 3, foreign invesors were allowed o buy up o 3 percen of he value of a newly privaized enerprise. This figure was hen raised by 9 percen o 9 percen a he end of he hird quarer of 5. I is he deerminaion o speed up he process of privaizaion in erms of boh pace and scale ha financial They are: Bank for foreign rade of Vienam (VCB), Bank for invesmen and developmen of Vienam (BIDV), Vienam bank for agriculural and rural developmen (VBARD) and he indusrial and commercial bank of Vienam (ICB).

5 liberalizaion was underaken. Likewise, in he domesic marke, more sae-owned enerprises are being urged o submi legal documens o sar he process of going public. As a maer of fac, he ransacion volume and marke capializaion ha were comparaively small for several iniial years has recenly shown hecic growh. Marke capializaion is esimaed o be roughly percen of GDP in May 7. The high and rapid growh of he sock marke is, of course, very appealing o domesic and foreign invesors. Saisical sources repor ha here are abou, accouns regisered o make ransacions in he sock brokerage companies a he end of May 7, and his figure is increasing daily (Vienam Daa Communicaion Company, 7). The presence of he foreign invesors, he high growh poenial of Vienam economy, he idle public capial volume (esimaed a USD8 billions), and he yearly remiance from he Vienamese overseas workers and relaives amouning o around USD-3 billions have all conribued o make VSM become really ho (Vienam Minisry of Indusry, 7). Ye, he sock reurn rae has no been sable for any exended period, and he marke had more han once experienced a recession and hen a specacular recovery showing volailiy clusering in he sock reurns. Having been operaing for almos 7 years and undergoing wo seps of financial liberalizaion, VSM provides an ineresing example o examine, ) he characerisics of sock reurn volailiy in a developing equiy marke, ) how robus is his characerisic of volailiy when regime change is aken ino accoun, and 3) how financial liberalizaion process or oher economic/poliical evens affec he sock reurn volailiy. This hesis will invesigae hese quesions and analyze wheher he VSM case suppors he financial liberalizaion heoriss who propose financial liberalizaion, or he Keynesians view ha suppors financial conrols. To examine he characerisics of he VSM reurn volailiy, his hesis uses an ARCH/GARCH model. Specifically, we es he hypohesis of highly persisen volailiy of VSM raes of reurn. Then, he regime change will be idenified endogenously by a saisical mehod ha applies he ieraed cumulaive sums of squares (ICSS) algorihm. Afer he regime changes are deermined, he ARCH/GARCH model is revisied o analyze how he volailiy alers for differen periods separaed by hese break poins. I will be examined o see wheher here are any coincidences beween he poins of change of sock reurn variance and he seps of financial liberalizaion. If indeed here are any, an analysis will be underaken from boh he financial liberalizaion heoriss perspecive and from he Keynesians view. The res of his hesis is organized as follows. Secion I sars wih he lieraure review. The daa used in he analysis is nex presened ogeher wih he resuls of diagnosic saisical ess. I also describes he model o be applied in his hesis and shows how he ICSS algorihm is employed o deec he break poins in he daa. Secion II presens he regression resuls, deeced volailiy regimes, and he oher findings. Secion III summarizes he discussion by confirming he findings and providing some suggesions for policy making. I. Theoreical Framework A. Lieraure Review A well-known sylized fac of financial ime-series is ha hey are ofen found o be nonsaionary, and mos of hem exhibi phases of relaive ranquiliy followed by periods of high volailiy. These differen phases of volailiy imply a ime-varying variance in he daa. Pagan, and William (989) summarize ha volailiy can be broken down ino predicable and unpredicable componens, and research ineres has largely been placed on he deerminans of he predicable par: he condiional variance of ha seriesσ. For invesmen in financial asses, his concern wih he predicable componen of volailiy is moivaed by he fac ha he risk premium is a funcion of i.

6 To capure he volailiy in financial ime-series, several models of condiional volailiy have been proposed. An ousanding class of model was firs inroduced by Engle (98) which was known as he ARCH (auoregressive condiional heeroskedasiciy) model. This model was laer generalized by Bollerslev (986) o GARCH (generalized ARCH) model by including he lags of condiional variance iself. In hese models, one very common finding in he financial asse is ha shocks o volailiy are ofen highly persisen. Besides he high persisence in volailiy of sock reurns, he consequences of financial liberalizaion for sock markes in he developing counries have been under close observaion from many economiss. Their findings are, however, by no means consisen. While financial liberalizaion heory proponens, namely McKinnon (973) and Shaw (973), argue ha his process will increase he savings and invesmen in developing counries leading o more sable and high economic growh raes, Keynesians argue ha his is doubful. In he opinion of he liberalizaion heoriss, even if volailiy increases, his may no be damaging o he real economy. Lamoureux and Lasrapes (99) poined ou ha he resul of increased informaion flow in urn could make he marke more efficien. On he conrary, a Keynesian view is ha he opening of he financial marke will inroduce more volailiy because of he increased volume and pace of ransacions, which is considered o be a desabilizing facor in economic developmen. However, in he case of rapidly developing markes, he increased volume and pace of ransacions can be caused by eiher foreign invesors or he growing number of domesic IPOs. For his reason, i would be ineresing if he conribuion of hese wo can be assessed. An increase or decrease in he volailiy of he VSM reurn following financial liberalizaion may be due o, ) foreign facors: many more foreign invesors increase he volume and pace of ransacions, or ) domesic facors: oo many domesic companies going IPO wihin a very shor ime period. For sock marke volailiy, in acualiy, several auhors have found ha when he regime changes are aken ino accoun, he above-menioned highly persisen ARCH/GARCH effecs are reduced. Susmel () showed evidence for reduced swiching volailiy in he US, Canada, he UK and Japan using he swiching ARCH (SWARCH) model formerly developed by Hamilon and Susmel (99). Malik e al. (5) use he ICSS algorihm o deec he regime changes and show he decrease in he volailiy for he Canadian sock reurns afer hese changes are considered. Concerning he influence of financial liberalizaion on financial volailiy, he evens ha occurred in Eas Asian counries in he closing years of he las millennium seem o favor he Keynesians view. This financial crisis showed ha foreign invesors used o inves ino he emerging equiy markes in search of higher raes of reurn in somewha of a speculaive mood. They ended o wihdraw he invesed capial under herd behavior hus causing a capial fligh in a ime of difficuly. The empirical evidence abou he effecs of financial liberalizaion is, however, limied and mixed, as saed by Kassimais (). He applied he exponenial GARCH (E-GARCH) model proposed by Nelson (99) o derive he news impac curves and showed ha sock reurn volailiy has fallen in Argenina, India, Souh Korea and Taiwan, while i has risen in Pakisan and in he Philippines. Some oher researchers, such as Grabel (995) and Aiken (996) provided he evidence of he increasing volailiy of he sock marke afer financial liberalizaion. B. Daa for Analysis The daily sock index daa of VSM was colleced from he websie of HoSTC (7). These are he closing marke index values (Vnindex) for he period from July hrough May 7. A oal of,57 daily observaions were obained. I should be noed ha, from he launch of HoSTC in July o he end of February, he rading sessions were held 3

7 only once every wo working-days. Since March, securiies rading ransacions have been conduced every working day. The Vnindex of HoSTC is chosen as represenaive for he Vienam marke index o be sudied in his hesis since HoSTC was launched firs and has a hisory of almos years longer han ha of HaSTC. The reurn rae of sock marke index is compued on he basis of he daily indexes in percenage scale as: ( P P ) * RR =, P where RR is he rae of reurn of sock on he day, P is he sock price on he day, and P is he sock price on he (-) day. I should be noed ha his hesis uses he ne reurn rae and no excess reurn rae, which should include dividends. This applicaion is accepable since he daa under invesigaion is compued on a daily basis so ha he dividends are small enough o be reasonably ignored wihou causing much disorion. Before modeling he sock marke reurn and is volailiy, he diagnosic saisical ess are underaken o check he characerisics of he aforemenioned daa. The summary of descripive saisics is shown in Table. The saisics shows ha he average daily reurn rae of he marke is abou.7 percen, saisically differen from zero a he percen level. The kurosis of he VSM reurn rae is abou 6.96, meaning i is lepokuric, a phenomenon which has been widely documened in he lieraure of sock marke reurns. According o Adrangi e al. (999), his kurosis, having a fa ail, ogeher wih he auocorrelaion among he squared residuals ells us ha reurn rae variance may follow an ARCH/GARCH process. The Jarque-Bera saisic is well above he criical Chi-squared value wih wo degree of freedom a he percen level of saisical significance. This allows he rejecion of he hypohesis of normal disribuion of he VSM sock reurn, and shows anoher indicaion of he excess kurosis as jus concluded. Moreover, he Ljung-Box (LB) and Augmened Dickey-Fuller (ADF) es saisics are calculaed o check for auocorrelaion and saionariy of he VSM reurn rae. The resuls are repored in Table A, A of Appendix A and Table below. As we can see, here is a significan auocorrelaion beween sock reurns in he Table A where he auocorrelaion funcion lengh is aken up o 36 lags. This means ha VSM is no an efficien one like ha noiced by Fama (965) in his efficien marke hypohesis (EMH). When EMH holds, changes in sock prices should no show any specific paern since if here was a paern, some raders (arbirageurs) would spo i and make profi from is idenificaion. However, in our case, since he VSM reurn raes show clear auocorrelaion, i seems ha he curren or pas values can provide some predicion abou fuure reurns. Table : Descripive Saisics of he VSM Daily Sock Rae of Reurn Saisics Vnindex Daily Reurn Rae Mean.68*** Sandard Deviaion.678 Skewness -.37 Kurosis Jarque-Bera 9.99 p-value a). Observaions 56 Noe: *** means saisically significan a he percen level under he null hypohesis: mean daily sock reurn rae is zero. a) p-value of Jarque-Bera saisic follows he Chi-square disribuion wih df.

8 8 6 Vnindex Reurn Rae (%) HoSTC Trading Session Figure : HoSTC Daily Sock Raes of Reurn Volailiy Clusering. The ADF es for he VSM price index in level form indicaes ha i is a non-saionary ime-series, bu his es for sock rae of reurn shows he resul o be saionary. Therefore, he VSM sock rae of reurn can be used as inpu for he model o capure he sochasic properies of Vienam sock marke. The VSM sock reurn rae shows ha i seems o conain clear volailiy clusering. Figure above provides an overview of he VSM sock reurn rae. Table : Uni Roo Tes of Vnindex Daily Sock Rae of Reurn -Saisic p-value* Augmened Dickey-Fuller es saisic Tes criical values: % level -3.3 Noe: Null Hypohesis: RR has a uni roo. *One-sided p-values. ADF es includes a consan erm. Lag lengh is chosen as 5 based on Schwarz Informaion Crierion. C. The Economeric ARCH/GARCH Model To es he hypohesis ha he VSM reurn rae volailiy is highly persisen, his sudy uilizes he ARCH/GARCH model, firs inroduced by Engle (98) and Bollerslev (986). The ARCH model can be described as follows: k () RR = α + α i RR i + ε i= () ε ~ N(, h ), Ω i (3) h = γ + γ iε i. q i= In equaion (), RR is he reurn rae of sock marke index a ime, RR is reurn rae of sock marke index a ime -, α,α i are he inercep erm and coefficiens of he lagged reurn raes of sock marke index. The number of lag k in equaion () is usually deermined using he Box-Jenkins approach. In equaion (3), h is he condiional variance for he curren 5

9 ime, γ is a consan erm, and ε i (he ARCH erm) represens he news abou volailiy from he previous period, measured as he lags of he squared residual from equaion (). The GARCH model inroduces one more erm ino he righ-hand side of (3): q p λ j i= j = (3 ) = + γ iε i + h γ h, here h j (he GARCH erm) indicaes news of he las periods forecas condiional variance. In he above process, N represens he condiional normal disribuion wih a mean of zero and a variance h, and Ω is he informaion available up o ime (-). The above q and p are he number of lags of he residuals from equaion () and he number of lags of he variance h from equaion (3) consecuively. These are o be specified based on boh he Box- Jenkins approach and he value of γ, or γ + λ which should be less han one. Since h is he forecas variance based on pas informaion i is called he condiional variance. The ousanding feaure of his ARCH/GARCH model is ha he condiional variance of he error erm is smaller han he uncondiional one, since i akes ino accoun he known curren and pas realizaion of he series, and i is hus preferable. Unlike ARCH, GARCH allows for boh auoregressive and moving average componens in he heeroskedasic variance. According o Enders (995, Ch.3), γ in ARCH or γ + λ in GARCH model should be less han one so ha he equaion for he condiional variance is guaraneed o be sable. Therefore, in his hesis, a model needs o be specified so ha γ or γ + λ is no over one, for he sake of model sabiliy in esimaion of he condiional variance. D. Deecing and Incorporaing he Regime Change: The Combined Model In addiion o he high persisence in volailiy in he sock reurns, invesors are also concerned wih regime changes caused by poliical, social or economic evens. These changes can raise or reduce he volailiy drasically. Thus, he idenificaion of volailiy due o regime changes plays an imporan role in he successful performance of he sock marke. I can assis in advising invesors on decisions concerning porfolio invesmen, and can assis policy-makers in he financial policy making process. In realiy, a more volaile marke causes curren and poenial invesors o lose confidence, making i hard for currenly lised companies o raise necessary capial. Consequenly, he decreasing ransacion volume may lead o a weakly performing sock marke. This, in urn, resrics he capial mobilizing funcion of equiy marke and he fuure growh of he economy. As noed in he lieraure review, many researchers have claimed ha, by aking regime change ino consideraion, he high persisence of volailiy is reduced. This means he models ha incorporae regime changes provide a more precise condiional variance periodically, and as a resul beer help forecas fuure values for he ime series under invesigaion. In acualiy, as will be shown laer, he volailiy is esimaed more accuraely for each volailiy regime afer aking accoun of he regime change by using dummy variables. To consider a sudden change in regime, many research endeavors have been underaken. Hamilon and Susmel (99) developed he swiching auoregressive condiional heeroskedasiciy (SWARCH) model. The exponenial GARCH (EGARCH) model proposed by Nelson (99) can be used o deec he level of curren volailiy generaed by he pas news. However, as Malik e al. (5) poined ou mos sudies impose regime shifs based j For proof, refer o Enders (995, pp.39-). 6

10 on a priori grounds and hus are subjec o he problem ha he break daes are correlaed wih he daa imposing regimes in his manner may inroduce serious biases ino he analysis. To overcome his difficuly, his hesis applies he ieraed cumulaive sums of squares (ICSS) algorihm, firs proposed by Inclan and Tiao (99). By using his echnique, he regime changes are incorporaed and found endogenously. Firs, le he ε series in equaion () be a series wih zero mean, and an uncondiional varianceσ. The variance wihin each volailiy regime is assumed o be homogeneous and denoed by τ j, j =,,... NT, where N T is he oal number of variance changes in T observaions, and a se of < κ < κ <... < κ N < T are he se of break poins. T < < κ σ = τ = τ κ < < κ = κ < T τ NT Second, denoec k = N < T k = ε, k =,,... T as he cumulaive sum of squares (ICSS) from he firs observaion o he k-h poin in ime. Then, he D k saisic is defined as below: Ck k Dk =, k =,,..., T Wih D = D T = CT T Inclan and Tiao (99) show ha he plo of D k oscillaes around zero for series wih homogeneous variance. When here is a sudden change or break poin, he plo of D k will exend beyond he specified boundaries wih high probabiliy. Under he null hypohesis of consan or homogeneous variance, his behavior leads o he possibiliy of he search for a variance break poin upon observing he max absolue D k. Firs, define k* as he poin in ime a which he maximum absolue value of D is reached. Then, if he absolue value of D k T / (sandardized disribuion of D k ) a k * falls ouside he specified boundaries, he null hypohesis can be rejeced. Inclan and Tiao (99) compued and poined ou criical values of.358 being he 95 h percenile and.68 being he 99 h percenile of he asympoic disribuion of max sandardized D k. In fac, o beer capure break poins, Malik e al. (5) adjused ICSS mehodology as firs suggesed by Sanso e al. (). They used he criical value of.6 (correcing for he kurosis) in heir analysis where under he same assumpions, he sandard Inclan and Tiao criical value is.358. They claimed ha researchers are likely o find more spurious srucural breaks if criical values are no properly adjused, i.e. he null hypohesis will be over-rejeced. For he safe deecion of regime change, his hesis uses he criical value of.6 as he boundary since he VSM raes of reurn plo also shows a clear lepokuric case. Inclan and Tiao (99) also claim ha using he D k saisic o deec he break poins simulaneously may be difficul when he daa under invesigaion has muliple variance changes. This is because of he masking effec. This masking effec may occur since a major break poin followed by a moderae-sized break poin can be overlooked by he D k funcion. To avoid his masking effec, Inclan and Tiao (99) suggesed using he D k funcion o sysemaically deec he break poins a differen pars of he series under concern. Firs, he D k funcion is applied o he whole samples o deec he firs possible break poin. Second, we again apply he D funcion o examine each of he wo sub-series divided by he k k 7

11 firs possible break poin. We coninue his process unil he maximum Sandardized D k is no longer larger han he criical value.6. Afer all he break poins are deeced, he las procedure is o check he exisence and convergence of possible break poins. For example, he sandardized D k for he par from k= o he second break poin will be examined o confirm he exisence of he firs break poin. In oher words, his is o check wheher he maximum Sandardized D k a he firs break poin is sill larger han he criical value. The same procedure is hen done for he par from he firs break poin o he hird break poin, and so on. Finally, afer finishing he deecion of he break poins ha cause regime changes, he ARCH/GARCH model specified by equaions () o (3) is be modified as follows: k () RR = α + α i RR i + ε i= () ε ~ N(, h ), Ω i () h = + dd + d D + + d ndn + γ iε i + γ... λ h, q p i= j= where D, D,..., Dn are a se of dummy variables conrolling for regime changes, aking a value of one from each poin of sudden change of variance onwards, and zero elsewhere. The number of D in he se represens he number of break poins deeced. γ is he consan erm ha sands for he average volailiy of he firs volailiy regime, ignoring any effec of pas residuals on he condiional variance. The coefficiens of D, D,..., Dn show, wihou he effecs of previous news as represened by he lags of residuals and/or condiional variance, how he subsequen regimes volailiy is differen from he firs regime volailiy and how he volailiy varies beween differen regimes (if hese Ds are found o be saisically significan). The resul of his regression is o be repored and analyzed in he nex secion. Addiionally, one very imporan par of our ask is o provide possible explanaion for he differen volailiy regimes found. II. Empirical Resuls A. Empirical Resuls of he ARCH/GARCH Model wihou Regime Change The GARCH (p, q) model as inroduced in Secion I is firs applied o examine he characerisic of he VSM reurn volailiy. To deermine he level of k, p, and q, his hesis uses he Box-Jenkins echniques as noed by Gujarai (3, Ch.). Afer rying various lag levels of k ino he mean equaion () and differen p, q levels ino variance equaion (3), only he ARCH () process wih AR () specificaion in equaion () proves o be appropriae for VSM case. In previous sudies of sock reurn volailiy, he GARCH (, ) model is commonly applied for many sock markes. For VSM, however, perhaps he reurn volailiy persisence is oo high so ha he oher levels of k, p, q, which are larger han, are no applicable. This is because he oher levels of k, p and q show ha he value of γ or γ + λ is larger han one, which desroys he sabiliy of he model. Therefore, he ARCH () process wih AR () specificaion in equaion () is chosen. The ARCH () regression resul is as follows (for deails, see Table A3 in Appendix A). (5) RR = RR + ε ε Ω i ~ N(, h ), (.9)*** (.76)*** (6) h = ε (.6)*** (.68)*** j j 8

12 I should be noed ha he above ARCH () process is he one ha has no ye aken ino accoun any regime changes. 3 The regression resul shows ha he coefficiens in all he equaions are saisically significan a he 5 percen level. As he coefficien of he previous rae of reurn in equaion (5) is saisically significan a he percen level, he presen sock reurn rae seems o be predicable from is pas value. I is known ha a value of coefficien of ε equal one indicaes an inegraed ARCH process in which shocks will have a permanen effec on volailiy. I is shown in our case ha he coefficien of he pas residualε in he condiional variance equaion (3) is.959, which is very close o. This resul implies he effec of a shock dies down slowly, and signals a high persisence of VSM reurn volailiy hroughou differen periods. Wihou aking regime changes ino consideraion, he esimaed condiional variance of equaion () shows a high persisence in volailiy and his formula is applied o compue he condiional variance hroughou various periods in he life span of VSM. However, when looking a he plo of sock reurn raes in Figure, i can be seen ha using his esimaion for all periods of he VSM developmen may no be reasonable. Hence, he nex ask is o deec he break poins and show how he condiional variance alered over differen regimes. B. Deecion of Break Poins and Volailiy Regime Effec Examinaion In his par, he residuals obained from regression in par A are uilized, wih he applicaion of he ICSS algorihm, o compue Dk o deec he break poins and volailiy regimes. Figure plos he obained sandardized D k. The resul shows ha VSM may be a case of muliple changes in variances. The possible firs break poin is a k*=3 corresponding o December 6 h,. However, when he secion from k= o k=3 is examined, anoher break poin is found a k*=9. A furher search of he secion from k= o k=9 shows he exisence of no possible break poin. The firs break poin is herefore deermined a k*=9 corresponding o he rading session 3 on June h,. The secion from k=9 o he end of he series is hen checked, and his is repeaed unil all break poins are found. Table 3 liss he deeced break poins and he regime changes following hem. The deails of break poin deecion are repored in he Figures from B3 o B in Appendix B. The deeced regime changes seem o coincide wih he changes ) in he sock marke operaing mechanism such as increase in he ransacion frequency, or ) in he macro policy concerning financial liberalizaion in Vienam: e.g., allowing more foreign ownership of socks and shares, or c) poliical evens around ha ime. The evens ha are possibly relaed o he deeced breaking poins are also repored in Table 3. Table 3: Break Poins and Regime Changes Possibly Relaed Evens or Sandard Mean Time period Duraion Policy Changes Deviaion Daily RR Jul 8 h, - From he launch of VSM o he firs June h **, monhs change poin. June 3 h, - 6 monhs The Congress IX of Vienam Communis In his and following regressions, he sandard errors are repored in he parenheses. In all he regressions, he null hypohesis is: he coefficiens are zero. (***), (**) and (*) show he saisical significance level of percen, 5 percen, and percen, respecively. k=9 corresponds o he HoSTC rading session 3 as he iniial observaion is los when compuing he Vnindex reurn raes, and he firs sock rae of reurn is used as he degree of freedom in equaion (). 9

13 Dec. 6 h, Pary was held from 9 h o 3 rd of April. Dec. 8 h, - 3 Trading sessions sared o be held Nov. h, 3 monhs everyday from March s, ** Nov. h Three monhs afer foreign invesors are, 3-6 Feb. 6 h allowed o buy up o 3 percen of he, 6 monhs sock value of a privaized enerprise ** Feb. 7 h, 6 -. Three monhs afer he foreign invesors >5 Presen (o be are allowed o buy up o 9 percen of monhs coninued) he sock value of a privaized enerprise ** Noe: ** means saisically significan differen from zero from he 5 percen level. The sandard deviaions are considered o be homogeneous wihin each volailiy regime and are also compued and summarized for he purpose of comparison. The firs period is equivalen o ime span from he launch of HoSTC in July, o June h,, lasing monhs. The sandard deviaion for his period,.978, reveals a comparaively volaile regime, ranking hird compared o all volailiy regimes. While he volailiy is no he highes, his period s daily reurn rae is.98 percen, saisically significan a he percen level. This is he highes reurn compared o any oher subsequen periods. This marked he successful birh of he sock marke in Vienam. However, he second regime from 3 h June, o 6 h December,, whose lifeime is exremely shor lasing only 6 monhs, creaes a record high volailiy wih he sandard deviaion being.6 percen. The poliical even which may have been relaed o he sar of his regime was he Vienamese Communis Pary Congress IX held in he las half of April of he same year. In spie of he fac ha he volailiy is found o be exraordinarily high in his period, he highes compared o any oher volailiy regimes, he average daily reurn rae is desperaely low a -.58 percen. This reurn rae is, however, no saisically significanly differen from zero a he percen level. The high volailiy and low daily reurn rae in his period emphasized he necessiy for furher reform o enhance he efficiency of VSM. There may be wo oher explanaions for his surge in he VSM volailiy. Firs, he firs volailiy regime had a lower risk level and high raes of reurn because VSM was hen relaively small Sandardized Dk HoSTC Trading Session Sand.Dk BOUNDARY Figure : Plo of Sandardized Dk for full sample. Max Sandardized Dk a k*=3. Moreover, alhough domesic invesors had no ye become familiar wih sock marke, hey srongly believed ha i would be a profiable new marke. Afer nearly a year, more people

14 sared recognizing he profiabiliy of lised companies and of he sock marke and were ready o inves in i. In addiion, he IX Communis Pary Congress, confirming he coninuaion of marke liberalizaion by saing ha he marke mechanism is he arge, increased he public confidence concerning he fuure rapid developmen of he sock marke. All of hese caused he firs sock and share price boom and domesic invesors rushed o buy securiies. The sar of his boom could be observed paricularly in he firs monh afer he congress bu i cooled down. This realiy may be responsible for he exraordinary volailiy during he second regime. Simply, domesic invesors in he second volailiy regime recognized ha heir expecaion for change resuling from he Pary Congress had been oo high. In fac, he public had o wai nearly anoher years for he governmen o open he financial marke o overseas invesors and 3 years for he governmen o release is decree no. 87 (6 h November, ) which promoed he equiizaion of he public secor. A characerisic in erms of he operaing procedures during his ime needs o be menioned. Throughou hese firs wo regimes, he rading sessions were held only every wo working-days. Hence, he small number of ransacions migh appear as signs of weak performance of he marke, resuling in domesic invesors disappoinmen. The hird volailiy regime sars from January, and mainains a raher long life of 3 monhs. In fac, so as o make HoSTC a more acive equiy marke, from March, he sock marke has been operaing every weekday. Invesors began showing posiive responses since hey became o rus ha VSM would soon carry ou is real funcion as a direc channel for mobilizing capial. Socks and shares became more aracive invesmen ools for domesic invesors. The increase of he volume and pace of ransacions enabled equilibrium of he supply and demand of socks and shares o be more easily reached. The volailiy, which is revealed by he magniude of sandard deviaion for he period from January unil November 3, shows a dramaic decrease. This volailiy level is, in fac, he lowes one compared o all he oher regimes, and shows a sharp decline from he second regime, wih sandard deviaion of The mean daily reurn rae also recovered o -.6 percen and is saisically significan a he percen level. However, here may also be anoher explanaion for his mild volailiy of VSM during his period. Figure B5 of he Appendix shows ha he ransacion volume in his period is sill low compared o he nex period. This means he low volailiy may be he produc of a sill weakly performing marke wih low ransacion volume, where he governmen has no ye implemened he simulaing policies necessary o promoe is developmen. The fourh volailiy regime, lasing he longes 6 a monhs, begins in early November 3. Afer more han 3 years of operaion, he marke capializaion appeared o be very modes and hus he governmen decided o do more o develop VSM. For he firs ime, foreign invesors were invied o ener. Effecive from Augus 3, according o decision no. 6/3/QĐ-TTg (7 h July, 3), foreign invesors were allowed o buy up o 3 percen of he sock value of an equiized enerprise. This new policy appeared o affec he volailiy raher quickly. Three monh afer he decision no. 6/3/QĐ-TTg, he volailiy showed a sign of increase. The volailiy increases almos 35 percen compared o he previous regime, being.737 percen during he fourh regime. However, i may no be a desperae siuaion since he sock rae of reurn recovered is posiive sign a.6 percen on average. Though his is only slighly above zero, i is saisically significan a he percen level. The realiy shows ha he increase of he VSM reurn volailiy afer he iniial liberalizaion can be parly due o he domesic invesors percepion ha he paricipaion of foreign invesors in he marke would make i more unsable. Foreign invesors are generally more powerful and hus hey have more influence over he marke s operaion. I should also be noed ha, in he middle of his regime, he Vienam governmen endeavored o increase he size of he sock marke. By is decree no. 87//NĐ-CP (6

15 November ), he governmen ried o accelerae he equiizaion. The decree allows all he former sae-owned enerprises which are no necessarily owned percen by he Sae o go public. Acually, companies of his sor accoun for almos all of he curren saeowned enerprises. In essence, by his decree, he governmen aimed a hree arges. Firsly, he governmen sen a clear message ha enerprises would need o become more ransparen as required by he sock marke. Secondly, he governmen expeced ha equiizaion could expand he budge available (via selling hese sae-owned enerprises) for concenraing on he consrucion of he counry s infrasrucure. This is an essenial sep o prepare he grounds for aracing more foreign invesmen. Lasly, i is a furher progress owards promoing he marke mechanism o fulfill is role in he Vienamese economy by allowing, on a larger scale, a variey of ownership. This decree really brough abou he hope ha more goods he socks - would be provided in a marke which was hen highly shor of supply. However, hings were no able o change as quickly as expeced. I ook anoher years for he policy o be pu ino pracice on a large scale. The las volailiy regime sared from he middle of February 6 and coninues up o he presen day. Seemingly recognizing he posiive response from he firs opening up of he equiy marke, in he hird quarer of 5, he governmen refreshed he equiy marke again wih is furher liberalizing decision no. 38/5/QĐ-TTg (9h Sepember, 5) ha allowed foreign invesors o purchase up o 9 percen of he value of shares of a join sock company. Like he firs liberalizaion, he news was absorbed relaively quickly. Three monhs afer he decision was made, he fourh change poin or he fifh volailiy regime occurred. Sandard deviaion for his regime,.978 percen, shows a nearly double increase compared o he previous regime. However, he average daily sock rae of reurn also showed significan improvemen. This figure,.39 percen, is also more han double ha of he previous regime alhough i is lower han ha of he iniial regime. Abou in he middle of his regime, a large number of sae-owned enerprises wen public. Especially, in December 6, fify one companies in HoSTC and sixy eigh ohers in HaSTC wen public. These IPO creaed a real boom in he growh of VSM. The single capializaion of HoSTC alone (compued a he end of 7) accouned for more han 5 percen of GDP in 5. Up o his poin, i can be seen ha in boh occasions of opening up he sock marke o foreign invesors, he volailiy of VSM increases. Is he increase in sock reurn volailiy purely due o he financial liberalizaion? In oher words, does he Keynesians view prevail over ha perspecive of McKinnon (973) and Shaw (973)? The incidens ha follow he wo liberalizaion seps a VSM seem o suppor his idea. However, i is imporan o be careful when inerpreing his phenomenon in order o capure he norm/essence. The fac migh be ha he sock marke volailiy increased because of boh more sae-owned companies had been privaized and more foreign invesors were coming and began acively paricipaing in he ransacions. Therefore, he cause of sock reurn volailiy increase may no be purely a consequence of he financial liberalizaion (he opening of VSM o foreign invesors). In realiy, VSM has no been in operaion long enough for us o evaluae separaely he effecs of financial liberalizaion and he growh of VSM caused by domesic IPOs. In oher words, domesic raders are also possibly responsible for he increasing volailiy. Furher sudy on he phenomenon should be underaken over a longer ime period before a jusified conclusion can be made.

16 Vnindex Daily Reurn Rae (%) RR Upper 95% Lower 95% HoSTC Trading Session Figure : Daily reurn rae of HoSTC wih regime volailiy bands (July May 7) Noe: Bands a +/- sd. deviaion, where break poins are esimaed using he ICSS algorihm. C. Empirical Resuls of he Combined Model Having idenified he break poins, he nex ask is o incorporae hese break poins ino he model in equaions (), (5) and (6). Using a se of dichoomous dummy variables aking values of from each break poin of variance onwards and elsewhere, we can now conrol for he regime changes in he esimaion. The regression resul is as follows (for deails, see he Table A of Appendix A). (7) RR = RR + ε ε Ω i ~ N(, h ), (.) (.)*** (8) h D.5858D.D D +.6ε = (.8)*** (.83)*** (.3)*** (.)* (.365)*** (.559)*** As is easily seen, full incorporaion of he regime changes significanly reduces he esimaed persisence of volailiy from.959 in equaion (6) o.6. The consan erm of he mean equaion (7) now becomes -.88, bu i is no saisically significan even a he percen level. In he combined ARCH () model, he finding implies ha abnormally high persisence of volailiy, if esimaed wihou aking consideraion of regime changes, may be misleading. Wih his new regression resul, i can be saed ha he condiional variance is esimaed more accuraely and hus he forecasing process is able o produce a beer resul. The plo of he ARCH sandard deviaion ha is provided in Figures B and B of Appendix B shows he differen volailiy regimes and heir volailiy levels. The consan erm of he condiional variance,.8333, shows a mildly low level of volailiy during he firs regime jus afer he sock marke was launched. The coefficien of D, 8., is exremely high and saisically significan a he percen level, showing how highly volaile he second regime is. The saisical significance of D, D, D 3, and D s coefficiens ells us ha, ignoring he effecs of he pas shocks on condiional variance, here is a real difference in he volailiy of differen regimes. The coefficien of D, even hough much lower han ha of D, is much higher han ha of he firs regime. I is esimaed a.696 percen and saisically significan a he percen level. Boh he coefficiens 3

17 of D and D3 have negaive signs which mean he lower volailiy in hese regimes compared o regime one. The coefficien of D is saisically significan a he percen level, bu ha of D3 shows saisical significance only a he percen level. This is no a surprise considering he plo of he VSM rae of reurn. The residual es of he combined ARCH () model provides furher evidence of he improvemen in he saisical characerisics of he regression ha ake ino accoun he regime changes. The hisogram of normaliy es of residuals in Figure B7 of Appendix B shows a significan decrease in boh residual s skewness and kurosis compared o hose of Table B6 in he same Appendix. Hence, he Jarque-Bera saisic declines a grea deal. The regression shows ha he volailiy persisence is reduced. The LM ARCH es also finds ha TR-squared, he indicaion of auocorrelaion in he squared residuals, shows a significan decline (for deails, see Table A5 and A6 in Appendix A). Consequenly, he probabiliy of no commiing auocorrelaion in he squared residuals increases from 9.5 percen o 9.6 percen.. III. Conclusion and Policy Implicaions A. Summary of Vienam Sock Marke Volailiy Wihou considering he regime changes, he ARCH () model shows a saisically significan high persisence of volailiy in he VSM sock reurn rae. When he algorihm of ieraed cumulaive sums of squares (ICSS) is employed, i is found ha he highly persisen volailiy of he VSM sock reurn rae is reduced. This finding suggess ha when invesigaing he volailiy of any sock marke reurn, i is beer o ake ino accoun he regime changes by a cerain approach. Considering he regime change helps o improve he esimaion of boh he volailiy persisence and condiional variances for differen periods. In he iniial regime when he VSM was jus esablished, he volailiy was no oo high while sock rae of reurn was he highes compared o any subsequen periods. However, he second regime provides a very grey picure of a weakly performing sock marke, in which volailiy is very high in spie of he negaive daily mean rae of reurn. Forunaely, his period is relaively shor and he similar high level of volailiy has no been repeaed in any oher subsequen volailiy regimes. I is, however, difficul o evaluae exacly he cause of he high volailiy in hese firs wo regimes since he marke was sill very young. Afer more han a year of operaion, along wih he improvemen in he sock marke operaing procedures (higher frequency of ransacions) he operaion of he marke has improved. This is revealed by he lower volailiy level found in all he subsequen volailiy regimes. Alhough i was more sable, he presence of only domesic invesors did no seem o make he marke perform well enough. For his reason, he governmen decided o open VSM o he ouside world by allowing overseas invesors o ener. More invesors were coming in, bu he volume of securiies - he goods of VSM - appeared no o have been sufficien. The decree no.87 (6 h November, ) helped by puing more join sock companies ino operaion. Volailiy increases in his fourh regime and furher increases in he fifh regime, when he equiy marke becomes more open, because of he second liberalizing decision of he governmen. A he end of 6, he large number of sae-owned companies simulaneously going public migh also have been he cause for he increase in sock reurn volailiy in he fifh regime. Alhough he volailiy is geing higher, he sock marke reurn rae is also showing signs of recovery and has remained posiive in boh of he fourh and fifh regimes, including he curren ime period.

18 As for he effecs of financial liberalizaion, he regression resuls show ha from boh he wo break poins when he equiy marke becomes more open, here are increases in sock reurn volailiy. However, we need furher examinaion o draw he final conclusion since he las financial liberalizaion of VSM coincided wih he growing densiy of he marke s IPOs, which may have also been he cause of he increasing ransacion volume and pace. B. Policy Implicaions for Vienam Based on he Empirical Resuls. Following he analysis and conclusions presened above, some suggesion concerning policy making on sock marke may be proposed. Firs, financial liberalizaion is a good way o arac foreign direc invesmen. Alhough he volailiy is geing higher, invesors are reaping higher daily reurn raes. Vienam has long been an ODA receiver bu ODA has many pros and cons, as widely poined ou by economiss in he field. Moreover, ODA is argeed o less developed counries, while Vienam is growing fas and expeced o graduae from he lis in. Thus, besides he domesic sources of savings and invesmen, aracing more foreign capial via he sock marke is a good choice. However, how far and how fas his process of financial liberalizaion should be implemened poses very ineresing quesions. VSM is geing bigger, bu i is sill a small equiy marke even for he Eas Asian region. As a resul, he opening of he marke needs o be conduced wih exreme cauion. I should be done in a manner whereby he pace of change is harmonized wih he increase in he size of he marke. The reason is o keep he securiies prices from suffering any bad effecs possibly brough abou from neighboring markes. Second, he sabiliy of he marke is dispensable o is fuure susainable growh for VSM and mus be considered as he firs prioriy. Even hough VSM has proven o be a successful case so far, i is ye a young, small marke and hence can be affeced easily by shocks, boh good and bad ones, and by herd behavior. To mainain is sabiliy, i is necessary o keep he operaion of he marke smooh wihou changes in operaing echniques. For insance, here are wo ypes of order maching procedures currenly employed a HoSTC and HaSTC and one should be chosen so as o unify he marke operaion. 5 Moreover, a oo high frequency in he number of enerprises going public may no be advisory. Though he governmen of Vienam is eager o push forward he ransiion oward a marke economy, and his is parly done by speeding up he privaizaion process, i is no always fruiful o push oo hasily. Vigilan observaion of marke performance needs o be coninued as more and more enerprises are privaized in an increasingly liberalizing scenario. If any oher sudden increase in he number of IPOs occurs, he domesic invesors may ge confused ino hinking ha he sock marke is ballooning because of he governmen s will. Consequenly, i creaes a source of fuure volailiy because he confidence is los. The pas half year showed a marke wih increasing volailiy, whose level may be somewha comparaive o he second period. Some pause in developmen and some self-sabilizaion by he law of invisible hand by he marke iself are boh necessary and good. Domesic invesors surely can learn much from such lessons and hey will become more maure. A he end of 7, i has been scheduled by he governmen ha some big enerprises, especially in he field of finance, such as he commercial sae-owned banks, will go public. The experiences from he las regimes of volailiy provide a lesson ha a wave of IPO should no conain all he big insiuions going IPO simulaneously, because he marke canno have enough ime o adjus invesors responses o mainain sabiliy. The IPOs his ime also provide anoher precious chance o see how he sock marke in Vienam reacs o a shock in ransacion volume. 5 These are coninuous order maching sysem in HaSTC and hree-ime-a-day maching sysem in HoSTC. 5