STOCK MARKET EFFICIENCY IN NEPAL

Transcription

1 40 Vol. Issue 5, May 0, ISSN ABSTRACT STOCK MARKET EFFICIENCY IN NEPAL JEETENDRA DANGOL* *Lecurer, Public Youh Campus, Tribhuvan Universiy, Nepal. The paper examines random-walk behaviour and weak-form marke efficiency on daily marke reurns of All Share Price Index (ASPI) and Sensiive Index (SI) on he Nepal Sock Exchange (NEPSE) using Lo and MacKinlay s (988) variance-raio ess and runs ess for he period beween Sepember 3, 006 o May 3, 00. The sudy finds ha he randomwalk hypohesis is srongly rejeced for boh indices. There is no evidence for weak-form efficiency in eiher series. I implicaes ha marke paricipans have opporuniies o predic fuure price and earn abnormal reurns from he Nepalese sock marke. KEYWORDS: Marke efficiency, Random walk hypohesis, Runs ess, Variance raio es. I. INTRODUCTION In an efficien sock marke, share prices reflec all informaion available o marke paricipans and ha, by implicaion, share prices canno be prediced, hus precluding any abnormal profi reurns. From view-poin of marke paricipans, he sock price behaviour is very imporan o deermine fuure abnormal reurns. Thus, his paper inends o measure he behaviour of he sock reurns in he emerging Nepalese sock marke. Once he behaviour of he sock reurns is deermined, hen one can beer undersand he marke and he economy. I makes sock prices reflec he rue picure of he company as well as he condiion of he overall economy. I can provide beer confidence o decision-makers on heir invesmen decisions and help in reducing he level of risk. The prior assumpion of he sudy is ha he marke is efficien and he series follow a random-walk. If his is rue, hen pas informaion including pas prices are irrelevan in predicing fuure sock prices for he companies lised in he Nepalese sock marke. If successive reurns are independen, hen, he marke is said o be efficien in is weak-form. The Secion II of he paper briefly surveys he relaed lieraure. Secion III conains deails abou he daa, mehodology and empirical resuls, while Secion IV consiss of he conclusion and implicaions. II. LITERATURE REVIEW For weak form ess, informaion can include only pas hisory of securiy prices. Tess for weak-form marke efficiency are, more generally, referred o as es of reurn predicabiliy (Fama 99). The weak-form of marke efficiency is invesigaed by examining wheher sock prices in equiy markes exhibi specific paerns, which allow fuure prices o be prediced. For a marke o be efficien in weak-form hen no such paerns should exis and

2 4 Vol. Issue 5, May 0, ISSN prices should follow a random walk. The weak-form inefficiency of he sock marke provides an opporuniy o he raders for predicing he fuure prices and earning abnormal profis. Fama and French (988) repored ha NYSE has negaive serial correlaion (mean revering) in marke reurns over observaion inervals of hree o five years, i.e., sock reurns are predicable. They argued ha auocorrelaion may reflec marke inefficiency or ime-varying equilibrium expeced reurns generaed by raional behaviour. On conrary o Fama and French (988) sudy resuls; Lo and MacKinlay (988) using a simple volailiy based specificaion es, concluded ha he NYSE-AMEX reurn indices showed posiive serial correlaion in marke reurns and he random-walk model is srongly rejeced. They argued ha he negaive serial correlaion in Fama and French s (988) sudy for long (hree- o five-years) holding-period reurns was, on purely heoreical grounds, no necessarily inconsisen wih posiive serial correlaion for shorer holding-period reurns. They also claimed ha he sum of a random-walk and mean-revering process canno be a complee descripion of sock-price behaviour. Similarly, Lo and MacKinlay (988) opined ha he rejecion of he random-walk model does no necessarily imply he inefficiency of sock price formaion. Jegadeesh (990) showed ha he monhly reurns on individual socks exhibied significan negaive firs-order serial correlaion and significanly posiive higher-order (longer lags) serial correlaion. The sudy also showed ha he reurn in January was significanly differen from oher monhs. The sock reurns showed a specific paern. I is a srong evidence of predicable behaviour of securiy reurns. I indicaes he rejecion of he hypohesis ha he sock prices follow a random-walk. The auhor poined ou ha he predicabiliy of sock reurns can be aribued eiher o marke inefficiency or o sysemaic changes in expeced sock reurns. Fama and French (988), Lo and MacKinlay (988), and Jegadeesh (990), he sudies in he developed markes, showed he predicabiliy of fuure reurns and concluded ha he marke was inefficien in weak-form, i.e., price formaion is dependen on or follow specific paerns. Bu, hese sudies did no explain he economic implicaion of he inefficien markes. Similarly, using monhly index prices in local currency, Urruia (995) esed he efficiency of Lain American counries: Argenina, Brazil, Chile and Mexico. The ime series behaviour of sample Lain American equiy prices did no seem o fi mean-revering processes eiher, since variance-raios larger han uniy imply posiive reurn auocorrelaion. Thus, resuls of he variance raio es rejeced he random walk hypohesis for all sample equiy markes. However, findings from he run ess indicae ha he four Lain American equiy markes are weak-form efficien. Auhor argue ha boh he economy and he capial markes of developing counries have been growing a an unusually rapid pace, and i is likely ha posiive auocorrelaions are indicaors of economic growh raher han evidence agains he efficien marke hypohesis. Thus, based on he resuls of he run es, he auhor concluded ha he four Lain American emerging equiy markes were weak-form efficien. Urruia (995) is successful o link beween he marke inefficiency and economy, which are lacking in he previous sudies; for example, Fama and French (988) and Lo and MacKinlay (988).

3 4 Vol. Issue 5, May 0, ISSN Recenly, Worhingon and Higgs (009) examined efficiency in he Ausralian sock marke for long period of,59 daily and,575 monhly observaions. They repored ha he monhly Ausralian sock reurns followed a random-walk, bu daily reurns did no because of shor-erms auocorrelaion in reurns. In he conex of Nepal, using auocorrelaion and runs ess, Bhaa (00) and Dangol (00) found ha he Nepalese sock marke did no follow random-walk hypohesis as well as inefficiency in weak-form for daily, weekly and monhly marke reurns series. On he conrary, Pradhan and KC (00) repored inconclusive resuls of heir sudy regarding he random-walk hypohesis and weak-form of marke efficiency using auocorrelaion and runs ess for weekly sock prices of 6 individual companies on hree years period beween mid- July 005 and mid-july 008. The previous sudies show ha he mixed resuls regarding he random-walk hypohesis and weak-form of marke efficiency. The reasons for inefficiencies are observed due o auocorrelaion srucures in heir reurns series. The developed markes show auocorrelaion on is reurns series, may be sysemaic changes in expeced sock reurns or raional behaviour of he invesors. On he oher hand, he majoriy of he emerging equiy markes provide posiive auocorrelaion indicaing unusual rapid economic growh. III. DATA, METHODOLOGY AND RESULTS The paper employed daily marke reurns of he wo indices, namely, All Share Price Index (ASPI) and Sensiive Index (SI) of Nepal Sock Exchange (NEPSE) from Sepember 3, 006 o May 3, 00. The paper used he above said periods because o capure he following poliical and non-poliical evens: () NEPSE jus move o semi-auomaion process from radiional rading sysem (Augus 4, 007); () poliical uncerainies, for example, elecion of Consiuion Assembly, he Maois led firs republican governmen afer he Consiuion Assembly and was replaced afer nine monhs by he UML pary; (3) increased lending as well as deposi ineres raes by financial insiuions; (4) experience of liquidiy crisis in financial insiuions; (5) increased capial gain ax from 0 o 5 per cen and hen reduced i from 5 back o 0 per cen; (6) mandaory provision o declare income source while rading on shares worh one million Nepalese rupees and above; (7) huge size of righs-share and iniial public offerings; and (8) capuring of boh up (bullish) and down (bearish) marke rends 3. The sudy used sock marke reurns as an individual ime-series variable. Marke reurns are calculaed from he daily price indices. Daily Marke reurns (P ) are calculaed from he price indices as follows: The ASPI is based on marke prices of all socks lised wih he NEPSE. A presen 73 companies are lised in NEPSE. The SI is based on he marke prices of group A shares lised wih NEPSE. The Companies should fulfill he following crieria for group A share: () he number of shareholders mus be a leas 000; () he company mus be in profi since las hree years; (3) he paid up capial of he company mus be a leas Rs 0 million; (4) he book value per share mus no be less han is paid up value; and (5) submission of he financial saemen wihin six monhs from he closure of he fiscal year. A presen 94 companies are classified under group A caegory in Nepal. Nepal formally implemened he sensiive index from Sepember 3, Highes NEPSE index of poins on Augus 3, 008, afer ha he marke has showed downward rend ill dae.

4 43 Vol. Issue 5, May 0, ISSN P = Ln PI PI -... () Where, P refers o marke reurn in period ; PI, price index a period ; PI -, he price index a period - and Ln refers o naural log. The reasons o ake logarihm reurns are jusified by boh heoreically and empirically. Theoreically, logarihmic reurns are analyically more racable when linking reurns over longer inervals. Empirically, logarihmic reurns are more likely o be normally disribued, which is a prior condiion of sandard saisical echniques (Srong, 99). To es he weak-form of marke efficiency, he paper has firs examined he normal disribuion of sock reurns. If sock reurns series follow a normal disribuion, i belongs o he assumpion of random-walk model; hence he marke is acceped as having he weakform of efficiency. The paper ess normaliy using he skewness, kurosis and Jarque-Bera saisic. Descripive saisics can be inerpreed o es he informaional efficiency of sock marke. Generally, values for zero skewness and kurosis a hree represen ha he observed disribuion is normally disribued. Table shows he descripive saisics of daily reurns of All Share Price Index (ASPI) and Sensiive Index (SI). The disribuion of daily sock reurns have slighly negaive-skewed bu i is highly lepokuric (peaked). Therefore, skewed and lepokuric frequency disribuion of daily marke reurns series indicae ha he disribuions are no normal. Jarque-Bera es also rejecs he null hypohesis of normal disribuion for boh indices; ASPI and SI. I gives evidence ha he frequency disribuion is no normal. Bu he posiive mean reurn and low variance indicae ha he Nepalese sock marke involves low risk. TABLE : DESCRIPTIVE STATISTICS OF DAILY RETURNS ASPI SI Observaions Mean Median Maximum Minimum Sandard Deviaion Skewness Kurosis Jarque-Bera

5 44 Vol. Issue 5, May 0, ISSN Probabiliy VARIANCE-RATIO TESTS OF RANDOM-WALK The paper uses he variance-raio mehod of Lo and MacKinlay (988) o es for randomwalk in he Nepal sock exchange (NEPSE). The idea behind he variance-raio es is ha if he naural logarihm of a ime series Y is a pure random-walk, he variance of is q- differences grows proporionally wih difference q. Tha is, he variance of he incremens in a random-walk is linear in he sampling inerval. Therefore, if a ime series follows a randomwalk process, he variance of is q-differences should by q imes he variance of is firs differences. The variance- raio, VR(q), is defined as: VR(q) = σ ( q )...() σ () Where, σ (q) is /q he variance of he q-differences and σ () is he variance of he firs differences. According o Lo and MacKinlay (988), formulas for he calculaion of σ (q) and σ () are as follows: σ (q) = q( - q +)( - q ) q ( Y Y q q μ)...(3) and σ () = ( ) ( Y Y μ)...(4) where, μ ( Y Y0 ) Y 0 and Y are he firs and las observaions of he ime series. The es is performed under boh homoskedasic and heeroskedasic specificaions. Under homoskedasiciy, he asympoic variance of he variance raio is expressed as follows: (q) = (q )( q 3q( ) )...(5) Under heeroskedasiciy, he asympoic variance can be expressed as: q ( q k *(q) = δ( k)...(6) q k

6 45 Vol. Issue 5, May 0, ISSN ( Y Y μ) ( Y k Y k k where, δ( k ) ( Y Y μ) μ) The homoskedasiciy and heeroskedasiciy consisen Z-saisics are denoed by Z(q) and Z*(q) and expressed as follows: Z(q) = VR ( q) ( q) ~ N(0,).... (7) and Z*(q) = VR ( q) * ( q) ~ N(0,). (8) Under a single variance-raio es, he null hypohesis is ha VR(q) = or ha he chosen index follows a random-walk. If he null hypohesis is rejeced and VR(q) >, hen he compued Z(q) and Z*(q) are posiive and reurns are posiively serially correlaed. If he null hypohesis is rejeced and VR(q) <, hen he compued Z(q) and Z*(q) are negaive and reurns are negaively serially correlaed, i.e., mean revering. Table repors he variance-raio ess, which are compued for inerval q =, 4, 8, 6 daily observaion inerval. The rejecion of he random-walk hypohesis under homoskedasiciy is no sufficien on is own, as i could be due o heeroskedasiciy or auocorrelaion in he examined series. Hence, i is imporan o focus mainly on heeroskedasiciy consisen Z- saisics. As per Table of he variance-raio es, Z-saisics are negaive wih saisically significan for boh homoskedasiciy and heeroskedasiciy in ASPI and SI. Hence, he random-walk hypohesis is srongly rejeced for boh indices. Similarly, he empirical findings reveal ha he null hypohesis of random-walk for boh seleced indices canno be acceped for all levels of inerval q a he one per cen level of significance. The boh sample indices: ASPI and SI, variance-raio values below one and hey decrease wih he inerval q increases. I indicaes negaive serial correlaion in he reurns and poenial mean reversion. In oher words, if sock price-reurns do rever means, hey should be negaively serial correlaed, and he variance raio should ge smaller and smaller han uniy as he inerval q increases. This ype of behaviour is generally observed in emerging financial markes ha may suffer a bubble effec (Summers, 986).

7 46 Vol. Issue 5, May 0, ISSN TABLE : VARIANCE-RATIO TEST Indices Q VR(q) Z(q) Z*(q) ASPI * -4.47* * -6.54* * -6.4* * -5.6* SI * * * * * * * * Noe: Number of observaions: 838 for each index. The aserisk denoes saisical significance a he 0.0 level wih a criical value equal o.57. Similarly, he observaions are spli up ino wo equal sub-samples and he resuls of variance-raio ess are repored no differen wih overall sample observaions. The paper provides he evidence of variance-raios lesser han one suggesing negaive reurns auocorrelaion. The resuls end o disagree wih Lo and MacKinlay (988) who find posiive auocorrelaion (variance-raio larger han one) for he New York Sock Exchange (NYSE) and American Sock Exchange (AMEX). Similarly, he resuls of negaive auocorrelaion are conradiced wih Worhingon and Hinggs (009) in Ausralia who finds value of variance-raio are larger han one showing posiive auocorrelaion. Bu he meanrevering process of sock reurns is documened by Fama and French (988), and Jegadeesh (990). TESTS OF WEAK-FORM OF MARKET EFFICIENCY The paper has also esed for weak-form marke efficiency using run es, which is a nonparameric es used for deecing he frequency of he changes in he direcion of a ime series. The run es is accomplished by compuing he expeced runs and he acual runs for he sample reurns. The expeced number of runs is represened as under: E(R) = n n n n...(9) Where n represens he oal number of observaions, n and n represen observaions ha equals or above and below he sample mean (or median), and R represens he observed number of runs. The sandard error can be wrien as under:

8 47 Vol. Issue 5, May 0, ISSN σ(r) = n n (n n n ( n ) n)...(0) The asympoic (and approximaely normal) Z-saisic can be wrien as follows: Z(R) = R E( R) σ( R)... () Table depics ha he boh series are no normally disribued. Thus, non-parameric runs es, which ess for independence beween successive evens (in a series) wihou requiring normaliy of disribuion, is used o es for weak-form efficiency of he seleced indices ASPI and SI. Table 3 presens he ess of independence, i.e., runs es. The negaive Z-values for boh reurns series indicaes ha he acual number of runs falls shor of he expeced number of runs under he null hypohesis of reurns independen a he 0.0 level. The negaive Z values for he boh indices indicae posiive serial correlaion. The run es shows ha he successive reurns for boh indices under sudy are no independen a he one per cen significan level. There is no evidence for weak-form efficiency in eiher series. TABLE 3: RUNS TESTS OF DAILY MARKET RETURNS Indices Mean N n n E(R) R Z(R) p-value ASPI SI Noe: The resuls of runs es are based on mean value. The runs ess were also performed wih he median as base and i also repored similar resuls. Similarly, he observaions are spli up ino wo equal sub-samples and he resuls of runs ess are repored no differen wih overall sample observaions. VI. CONCLUSION AND IMPLICATIONS This paper examines he random-walk hypohesis and weak-form marke efficiency in he Nepalese sock marke employing variance-raio and runs ess. The ess used on wo imporan daily marke indices; namely, ASPI and SI. Since he variance-raio is less han uniy, he random-walk hypohesis for boh indices is srongly rejeced. Similarly, he runs ess have rejeced he independence of sock price movemens ha indicaed he Nepalese sock marke as inefficien in weak-form. The Nepalese sock marke is inefficien in daily reurns series suggesing ha pas movemens in sock prices can be used o predic heir fuure movemens. I provides marke players bring he possibiliy of earning higher reurns han expeced. The presence of a random-walk in he sock daa has an imporan implicaion for porfolio invesors, he allocaion of capial wihin an economy and hence overall economic developmen. I is, herefore, relevan o sugges ha here should be an effecive regulaory framework and is implemenaion; and a more effecive role by all he sakeholders should be helpful in making he marke reflecive of a rue picure of he economy.

9 48 Vol. Issue 5, May 0, ISSN The possible reasons for he marke inefficiency in he emerging sock marke like Nepal may be he poor insiuional infrasrucure, weak legal framework, lack of supervision, slow developmen of he marke infrasrucure, poor corporae governance and accounabiliy, low level of capaciy of major marke players and lack of ransparency of marke ransacion. The sudy provides he ime series behaviour of a less developed marke of Nepal. The processing of new informaion in Nepal is raher weak, i may resul from persisen large number of non-acively raded shares, limied role of muual fund and lack of professionally managed invesmen and broker houses. The absence of qualified analyss, insiuional invesors and invesmen-friendly environmen is a well-known consrain in he emerging marke like Nepal. The main challenges of he Nepalese sock marke include frequenly changing governmen policies, policy-level inervenions, ussles of regulaory auhoriies, lack of commimens on economic agenda from poliical paries, and slow economic growh. Oher challenges are o widen he use of auomaion, srenghen regulaions and supervision, and educae invesors. Similarly, funcional auonomy of SEBON (regulaory body), and NEPSE (implemenaion body) are also he need of he day. Furher research is recommended o assess wheher his paper s findings are verifiable using alernaive economeric procedures. REFERENCES Bhaa, G. P. (00). Does Nepalese Sock Marke Follow Random Walk?, SEBON Journal, Volume 4, p Dangol, J. (00). Efficien Marke Hypohesis and he Emerging Capial Marke in Nepal, In Efficien Marke Hypohesis: Few Empirical Evidences from Nepalese Sock Marke, Kahmandu: Ques Publicaion Fama, E. F. (99). Efficien Capial Markes: II, Journal of Finance, Volume 46, Issue 5, p Fama, E. F. and French, K. R. (988). Permanen and Temporary Componens of Sock Prices, Journal of Poliical Economy, Volume 96, Issue, p Jegadeesh, N. (990). Evidence of Predicable Behavior of Securiy Reurns, Journal of Finance, Volume 45, Issue 3, p Lo, A. W. and MacKinlay, A. C. (988). Sock Marke Prices Do No Follow Random Walks: Evidence from a Simple Specificaion Tes, The Review of Financial Sudies, Volume, Issue, p Pradhan, R. S. and KC, S. (00). Efficien Marke Hypohesis and Behaviour of Share Prices: The Nepalese Evidence, SEBON Journal, Volume 4, p Summer, L. H. (986). Does he Sock Marke Raionally Reflec Fundamenal Values?, Journal of Finance, Volume 4, Issue 3, p Urruia, J. L. (995). Tess of Random Walk and Marke Efficiency for Lain American Emerging Markes, Journal of Financial Research, Volume 8, Issue 3, p Worhingon, A. C. and Higgs, H. (009). Efficiency in he Ausralian Sock Marke, : A Noe on Exreme Long-Run Random Walk Behaviour, Applied Economics Leers, Volume 6, p