Comparative Systematic Risk Analysis: Evidence on the Banking Sector in the United States, Western Europe and South East Asia

Transcription

1 Nawazsh Mrza and Danel Danny Smatupang 149 Comparatve Systematc Rsk Analyss: Evdence on the Bankng Sector n the Unted States, Western Europe and South East Asa Nawazsh Mrza and Danel Danny Smatupang * I. Introducton The bass for asset prcng n fnancal markets was provded by Bacheler (1900) n hs magnfcent dssertaton Théore de la Spéculaton submtted at Sorbonne (Unversté de Pars). Although from today s perspectve, the mathematcs and economcs he appled were flawed, yet the great genus, Markowtz, declares ths early work as an nspraton for hs own classcal paper of Portfolo Selecton. The rsk return relatonshp has always been a debatable ssue n fnancal theory. Portfolo Selecton came up wth a meanngful measure of quantfyng the rsk assocated wth nvestment; the varance of returns. The equlbrum model of Captal Asset Prcng (CAPM) (Sharpe 1964, Lntner 1965, Mossn 1966) further classfed the rsk as relevant and rrelevant rsk. Accordng to the CAPM, the relevant rsk s the systematc rsk or non dversfable rsk. The systematc rsk s the volatlty of returns of a partcular stock to the market returns. Hstorcally, the bankng sector s not that actve n captal markets. The nvestors lke to nvest n deposts and savng accounts more than they would lke to go for the stocks of a bank or a fnancal nsttuton, prmarly because of hgher rsk nvolved n stock markets. Ths could be a possble explanaton why the bankng sector has less representaton n stock markets as compared to other sectors. However, the absence of publc equty also ncreases the rsk of a bank. The major chunk of assets and labltes n a bank are of a fnancal nature. They are subject to nterest rate changes and respond quckly to the volatlty n the economy. The equty or the share captal acts as a cushon n case of bank s default on ts oblgatons. Due to ths utmost mportance of equty n a bank, the regulatory authortes have * The authors, have done Masters n Busness Admnstraton (MBA) from Ecole Supéreure des Affares de Llle, France. They are thankful to Prof Jöel Petey of Ecole Supéreure des Affares de Llle and Prof Edward Altman of Stern School of Busness, New York Unversty.

2 150 The Lahore Journal of Economcs, Vol.9, No.1 set a standard captal adequacy rato (see Basle 1988, 1996). Ths mnmum rato s a measure to nsure the credtors aganst any mshap or at least to mnmse the possble losses. Smlarly, every credt rsk management model (Credt Metrcs 1, KMV 2 etc) ncorporates the shareholders equty as an mportant component. The sleepng nature of bankng stocks makes them an alen n the fnancal markets. Ther senstvty to economc events makes them more volatle as compared to other ndustres. Dversfcaton s a tool to mnmse the rsk and consequently maxmse returns. Ths dversfcaton 3 could nvolve nvestng n dfferent ndustres or even nternatonally. Bankng stocks could be possble canddates for ncluson n a dversfed portfolo. Thus the problem arses as to how these stocks respond to the stock markets and what level of systematc rsk they are exposed to n dfferent markets, gven certan economc crcumstances. The stock markets n the Unted States, Western Europe 4 and South East Asa 5 are the sgnfcant stock markets n the world, and based on ther dfferent geographc locaton and economc crcumstances, they could be a test case for observng the comparatve rskness of bankng stocks n these three dfferent regons. Hstorcally, the fnancal sector has been blamed for bubbles, pancs and shocks. Leadng from Tulp mana to the South East Asan currency crss, the weaknesses n the bankng sector have always been regarded as the man cause for the spread of the fnancal crses. All these factors make the bankng sector more hostle for any nvestment n ther stocks. There are vrtually hundreds of studes on systematc rsk and ts mpact on dfferent prcng models, yet the lterature s not that vast for systematc rsk n nternatonal markets. In ths study we wll estmate the systematc rsk n an nternatonal envronment and test for the rskness (market rsk) of the bankng stocks. The paper s organsed as follows. Secton-II wll provde a theoretcal background on systematc rsk, CAPM and related concepts. Secton-III wll descrbe the possble bas n betas due to emergng market 1 See J P Morgan Credt Metrcs Techncal Document 2 KMV Rsk Management based on Black, Scholes and Merton s Opton Prcng Models. 3 In ths artcle, we refer to dversfcaton only n stocks. 4 The countres that adopted Euro as common currency. 5 We wll refer to Indonesa, Malaysa, Thaland, Phlppnes and Sngapore.

3 Nawazsh Mrza and Danel Danny Smatupang 151 phenomenon and beta correcton methods. Secton-IV wll descrbe the research methodology, data and model estmaton. Secton-V wll represent data fndngs and results and Secton-VI wll conclude the study. II. Theoretcal Perspectve a. Systematc Rsk The systematc rsk s the volatlty of a partcular stock or a portfolo to the market. It can be measured by the degree to whch returns a gven stock tends to move up or down wth the market. Ths tendency of the stock s reflected n ts beta coeffcent. The beta determnes how the stock affects the rskness of a dversfed portfolo, so t s theoretcally the most relevant measure of any stock s rsk. The concept of systematc, non dversfable rsk or beta was frst dscussed under the frame work of captal asset prcng model (CAPM), presented by Sharpe. The CAPM framework s very smple under deal condtons. The model states that the expected returns of an asset are a postve functon of three varables: beta, the rsk free rate and the expected market return. A smple CAPM equaton can be wrtten as R = R ( R R ) (1) f m f The above equaton of CAPM can be wrtten as a smple tme seres model that s normally used to estmate betas n the CAPM context. Ths regresson nterpretaton s R R = α γ e..(2) t ft t t where γ = R R and s known as rsk premum. t mt ft From the above equaton, t s evdent that systematc rsk attrbutable to ts senstvty to macroeconomc factors s reflected n ; non-systematc rsk, the unexpected component due to unexpected events that are relevant only to the securty, s reflected n e. The expected return on an asset depends only on ts systematc rsk. No matter how much total rsk an asset has, only the systematc porton s relevant n determnng the expected return on that asset (Corrado and Jordan [2000], p.524).

4 152 The Lahore Journal of Economcs, Vol.9, No.1 Another popular model of estmatng betas, s the market model or sngle ndex model. The studes of stock prces behavour show that when the market, as measured by a market ndex, rses most stocks prces tend to ncrease. Smlarly when the market s on a downsde, the stocks n general lose ther value. Ths observaton suggests that the reason the stock returns are correlated mght be because of common response to the stock market. Ths correlaton could be obtaned by relatng the return on stock to return on market ndex. Mathematcally ths could be expressed as R = α R e.. (3) m The α and e are the components of return of securty, and are ndependent of the market. They are random varables representng the returns nsenstve or ndependent of markets. Beta s a measure of rsk n equlbrum n whch nvestors maxmse a utlty functon that depends on the mean and varance of returns of ther portfolo. The varance of returns s a questonable measure of rsk for at least two reasons: Frst, t s an approprate measure of rsk only when the underlyng dstrbuton of return s symmetrc. Second, t can be appled straghtforwardly as a rsk measure only when the underlyng dstrbuton of returns s normal. However, both the symmetry and the normalty of stock returns are serously questoned by the emprcal evdence on the subject. b. Systematc Rsk and CAPM The systematc rsk or the beta has been n the lmelght snce ts ncepton n the 1960s. For the last 30 years academcans and practtoners have been debatng the merts of CAPM, focusng on whether beta s an approprate measure of rsk. Moreover, the stablty of beta has always been a concern n emprcal studes. The test of CAPM s the observaton of exstence of a postve lnear relatonshp between beta and returns. Although the model postulates a postve trade off between beta and expected returns, researchers n general always found a weak but postve relatonshp between beta and returns over the sample perod. Hence, they clamed that the results are nconsstent wth the postve lnear relatonshp between beta and returns as prescrbed by CAPM and the valdty of CAPM s n queston, questonng beta as an approprate measure of systematc rsk. Fama and MacBeth (1973) tested the valdty of CAPM usng a three step approach. In the frst perod, ndvdual stocks betas are

5 Nawazsh Mrza and Danel Danny Smatupang 153 estmated and portfolos are formed accordng to these estmated betas. In the second perod, betas of portfolos that are formed n the frst perod are estmated. In the fnal step, usng data from a thrd tme perod, portfolo returns are regressed on portfolo betas (obtaned from the second perod) to test the relatonshp between beta and returns. They found a sgnfcant average excess return of 1.30% per month, for the perod 1935 through 1968, a postve relatonshp exsts between beta and monthly returns. They concluded that results support the CAPM n the US stock market and consequently beta s a vald measure of systematc rsk. However, Schwert (1983) suggested that Fama and MacBeth (1973) only provded a very weak support for a postve rsk return trade off snce the postve rsk return relatonshp found s not sgnfcant across sub perods. Furthermore, when consderng seasonal behavour of ther results, the t-statstcs for the study perod becomes hghly suspect and the basc rsk return trade off vrtually dsappears. Fama and French (1992) studed the monthly average returns of NYSE stocks and found an nsgnfcant relatonshp between beta and average returns. They concluded that CAPM cannot descrbe the last 50 years of average stock returns and only market captalsaton and the rato of book value to market value have sgnfcant explanatory power for portfolo returns. The above mentoned studes gve evdence aganst beta as a useful measure of rsk. However, Pettengll et al. (1995) developed a condtonal relatonshp between beta and realsed returns by separatng perods of postve and negatve market excess returns. Usng US stock market data n the perod 1936 through 1990, they found a sgnfcant postve relatonshp between beta and realsed returns when market excess returns are postve and a sgnfcant negatve relatonshp between beta and realsed returns when market excess returns are negatve. Ths sgnfcant relatonshp s also found when data are dvded by months n a year. Furthermore, they found support for a postve rsk return relatonshp. Isakov (1999) followed the approach of Pettengll et al. (1995) and examned the Swss stock market for the perod He found supportng results that beta s statstcally sgnfcantly related to realsed returns and has the expected sgn. Hence, Isakov (1999) concluded that beta s a good measure of rsk and s stll alve. Most of the studes relatng to systematc rsk have been usng the domestc markets. Thus a logcal queston arses whether the relatonshp

6 154 The Lahore Journal of Economcs, Vol.9, No.1 between beta and returns can also be appled to nternatonal markets. Does beta have an explanatory power n nternatonal equty markets? To the best of our knowledge, no study (except one) has nvestgated ths ssue. Fletcher (2000) examned the relatonshp between beta and returns n nternatonal stock markets between January 1970 and January 1998 usng the approach of Pettengll et al. (1995). Usng monthly returns of Morgan Stanley Captal Internatonal (MSCI) equty ndces of 18 countres and the MSCI world ndex, Fletcher (2000) found that a consstent result exsts. There s a sgnfcant postve relatonshp between beta and returns n perods when the world market excess returns are postve and a sgnfcant negatve relatonshp n perods when the world market excess returns are negatve. Besdes, ths relatonshp s symmetrc and there s a postve mean excess return on the world ndex on an average. Fletcher (2000) also found that the sgnfcant condtonal relatonshp n January exsts only n perods of postve market excess returns and the relatonshp s nsgnfcant n perods of negatve market excess returns. The results dffer from those obtaned from Pettengll et al. (1995) on the US market data. III. Bas n Beta Coeffcent The estmaton of beta usng the CAPM framework or market model s not dffcult. However, there are some ssues related to the goodness of the measure. The beta estmates usng the above mentoned models wll be a sutable measure only f the stocks are actvely traded. The actve tradng n the market helps the beta coeffcent to explan the rsk assocated wth the partcular stock. One mportant pont to note s that t s not only the stock that has to be traded actvely, but also the markets should be actve. If, on the contrary, the stock s not actvely traded or the markets are thn tradng markets, the estmated beta wll not be a good estmaton of the systematc rsk of the stock. Ths requres correcton of estmated betas. Beta commonly s estmated by usng the Ordnary Least Square (OLS). In the OLS model, hstorcal returns on a gven securty are regressed aganst the concurrent returns of the market. Bascally, such estmaton has a dsadvantage because t gves unstable and based Beta (Scott and Brown [1980]). Based Beta usually happens n a thn-tradng market. Thn-tradng phenomenon that results n based Beta s dentcal wth non-synchronous tradng that s caused by nfrequent tradng. In ths sense, there mght be some sleepng stocks. Non-synchronous tradng

7 Nawazsh Mrza and Danel Danny Smatupang 155 problems arse n securtes due to the tme lag between the settng of market clearng prces for securtes and the market ndex computed at the end of a dscrete tme nterval, known as the ntervallng effect (Arff and Johnson [1990], p.85). Upon pros and cons, the potental for bas n the OLS due to non-synchronous tradng has been recognsed. For securtes traded wth tradng delays dfferent than those of the market, OLS estmates are based. Lkewse, for securtes wth tradng frequences dfferent from those of the market ndex, OLS estmates are based (Peterson [1989]). The adjustment to Beta values for non-synchronous tradng actvtes s necessary. Most of the non-synchronous tradng phenomenon happens n emergng stock markets because n those markets trade s low (thn). In most practces, not all securtes are traded n the same nterval, and some of them are not traded for a perod of tme. If there s no securty transacton on a certan day, the securty closng prce for that day s actually the prce from the prevous day, whch was the prce at the last tme the securty was traded. It could be two days ago, three days ago, or may be weeks ago. When the prce s used to calculate the market ndex of a day, the market ndex actually reflects the tradng value of ts prevous days. If Beta s calculated usng returns of a securty and returns of a market ndex formed from securty returns from dfferent tradng perods, the Beta wll be serously based (Hartono and Suranto [2000]). Ths phenomenon happens n almost all the emergng Stock Exchanges. The major problem s that shares lsted on these exchanges are thnly-traded, thus leadng to the problem of non-synchronous tradng where the market s prces at the end of a perod cannot be accurately matched wth the prces of a thnly-traded share (Lantara [2000], p.18-19). Consequently, estmates of systematc rsk of these shares wll be based. If the estmate of α and are based, the estmate of e wll also be based, and the extent of the bas wll be more serous for more thnly-traded shares (Arff and Johnson [1990], p.82). a. Beta Correcton Methods In a perfect stock market where prces are contnuously formed, the problem of non-synchronous tradng should not exst as every stock n the market would have regstered a market clearng prce at the dscrete tme of observng the market ndex, whch s the average of all prces at that nstant. A sgnfcant proporton of the stocks n a market, however, trades

8 156 The Lahore Journal of Economcs, Vol.9, No.1 so nfrequently that prces may be cleared on a few days n a typcal month. Ths s the general behavour n developng countres. Consequently, the measured market prce (and the market return, R m ) devates from the actual returns had there been contnuous tradng. Non-synchronous tradng makes beta based. If the market Beta value obtaned from the weghted average of ndvdual Beta values s not equal to one, the adjustment to the Beta values s obvously necessary. There have been many methods suggested by experts to adjust or correct the based Beta. However, we wll use Fowler Rorke method of beta estmaton. Fowler and Rorke (1983) developed a based Beta correctng method by scalng the coeffcents wth approprate weghts. The weghtng factors to multply n perods of regresson coeffcents are calculated as follows: w n 1 ρ 1 ρ 2... ρ n 1 ρ = 1 2 ρ 1 2 ρ ρn The values for ρ n are generated from a regresson equaton as follows: R mt = α j ρ 1 Rmt 1 ρ 2Rmt 2... ρnrmt n The corrected Beta values usng Fowler-Rorke method s ganed from:... n n t = w n j... w1 j j w1j w nj McInsh and Wood (1986) examned the adjustment technques proposed by Fowler-Rorke and found that these technques reduce a porton of the bas n arsng from thn tradng and delays n prce adjustments. For some researchers, partcularly those who do research n emergng captal markets, the Fowler-Rorke method s beleved to be the strongest one n reducng the bas. Arff and Johnson (1990) used Fowler-Rorke s three lags and three leads n estmatng the corrected Betas at the Sngapore Sock Exchange. Hartono and Suranto (2000) found that Fowler-Rorke s four lags and four leads s the best method n correctng Betas on the Jakarta Stock Exchange, after dong several tests wth dfferent lags and leads each. n e t

9 Nawazsh Mrza and Danel Danny Smatupang 157 IV. Research Methodology As mentoned earler, our research wll test the comparatve rskness of bankng stocks n an nternatonal context. The market rsk wll be measured usng the portfolo of bankng stocks n each geographc market. The three markets that we have chosen are that of the Unted States, Western Europe (Euro zone) and South East Asa. These stock markets are the sgnfcant stock markets n the world, and based on ther dfferent geographc locaton and economc crcumstances, they could be a test case for observng the comparatve rskness of bankng stock. Unted States fnancal markets are consdered to be safer for nvestment prmarly because of two reasons. The markets are close to strong form effcency and the tendency of extraordnary proft makng s lower. The players n the fnancal markets can be regarded as somewhat ratonal compared to other countes. Secondly, the government regulatons are strct enough to provde a far game. They are weaker n emergng markets, especally n Asan markets where ether there are no proper gudelnes for market tradng, or else not practsed. Although there have been ups and down n US stock markets such as the Market Crashes of 1929 and 1987, whch could be a strong challenge to the beng safer proposton, yet the over all stuaton s better than many of the emergng stock markets. The pancs and bubbles have occurred every where but the recovery of the US stock market has been tremendous n response to these pancs and crashes. The Unted States bankng sector s also more developed than those n other countres and the banks are more actvely traded both on organsed and over the counter markets. The monetary ntegraton of Western Europe has helped to homogense the economes n these countres. The Euro was launched n the year 2001 but the preparaton for a common European currency started from the early 90s. The bankng sector n these countres s well montored and the fnancal markets are no less than those n the Unted States. The ndustral gants such as Germany and France rely heavly on ther bankng sectors. However, the evdence shows that n the Euro zone the number of banks lsted on the stock exchange, s much lower than that of the Unted States and all of them that are lsted, are not even traded very actvely. The bankng sector n the Euro zone s less actve n stock markets, than that of the Unted States but s far better than that of Asa. Thus the comparatve analyss of rsk measure and the performance of bankng stocks n the Euro Zone should gve us an nsght about these stocks n an nternatonal

10 158 The Lahore Journal of Economcs, Vol.9, No.1 reference. An mportant pont to note s that we exclude the Unted Kngdom and Swtzerland because frstly they have not adopted the Euro as a common currency and secondly the bankng sector n these two countes have dfferent characterstcs than that of Euro countres. The fnancal sector n Swtzerland s more complcated and t requres more factors to analyse t than just systematc rsk. Hence, to consder the performance on a regonal bass, t could be only possble n the countres that have dentcal systems and characterstcs. The thrd geographcal regon we mentoned s South East Asa 6. Apart from Japan, generally Asan economes are consdered to be a lot more rsky. Government nterventon n Asan markets, to obtan desred results, s nothng new. The economc polces have a very short lfe. Personal lkngs and dslkes of the rulng government translates nto regulatons wthout lookng nto ts outcome. All ths makes foregn captal escape from a partcular economy. An excepton to ths could be Japan, whch s the second bggest economy after the Unted States. Apart from Japanese markets the second opton could be South East Asa. The South East Asan countres have long been known as Asan Tgers. These Asan Tgers share common features n ther economes. The bankng sector n these countres s weaker than that of the Unted States or Europe but could be consdered better than that of other underdeveloped countres. The countres from South East Asa suffered the worst crss n ther hstory n The crss orgnated by the devaluaton of Thaland s currency, the Baht and the perod followng the devaluaton of the Tha Baht wtnessed a sudden and unprecedented collapse n asset prces, corporate and fnancal fraglty, and a drastc economc slowdown n East Asan markets. In just over 12 months, the regon's stock markets, once among the largest n the world, saw ther market captalsaton shrnk by as much as 85% n US dollar terms. Smlarly, East Asan currences deprecated sharply beyond the levels needed to mantan export compettveness, wth some currences fallng by 50-80% aganst the US dollar by end The rapd deprecaton n East Asan currences, coupled wth a plunge n asset prces n these countres, led to a fall n real purchasng power as nflatonary pressures took root. Concurrently, there was a marked slowdown n economc growth: Asa's real GDP growth declned to 5.8% n 1997 from 6.6% n 1996, wth a further declne to 4.1% n Emergng markets took on an ncreasngly hgh-rsk low- 6 ASEAN countres & excludng Japan.

11 Nawazsh Mrza and Danel Danny Smatupang 159 return profle, as rsng volatlty and the deteroraton n economc fundamentals led to the outflow of captal from these markets. The fnancal sector was the most to suffer n ths crss. Many banks were lqudated, taken over by the government or smply were forced to merge. Ths made the performance of the bankng sector n South East Asa more questonable and nvestors began to avod nvestment thereafter n the fnancal sector. The economes started to recover n 1999, and so also the stock markets but the poston of the fragle bankng sector s stll questonable. Many researchers blame the weaknesses n the fnancal system for ths crss that spread lke a contagon. Gven these facts, the queston arses as to how the stock market s performng now, has nvestor confdence developed agan? The best observaton of the bankng sector and ts rskness should be measured by observng the perods n pre crss, durng crss and post crss. To observe the mpact of the South East Asan currency crss on the fnancal sector and the stock markets, and to avod a bas n market rsk estmaton, we need to observe the study perod n three perods. These three perods should be pre crss, durng crss and post crss. In the pre crss perod the fnancal markets n South East Asa were performng well and hence ther rskness should be close to that of the Unted States and Western Europe. Durng the perod when the crss was at ts peak, the rsk must have ncreased sgnfcantly and consequently market performance must have been affected. In the post crss perod, the fnancal sector started to recover, but maybe nvestor confdence has not fully revved so the markets n South East Asa mght stll be rsker as compared to the other two. 1. Types and Sources of Data In every study of nternatonal dversfcaton, the frst concern s dfferences n currency. One cannot compare drectly a return or a rsk from a country s portfolo wth another one, f both portfolos are stll denomnated n each of ther own currency (Fletcher [2000]). For ths reason, we transform all currences n the Europe Zone and Asa portfolos nto the US dollar at hstorcal spot exchange rate. Hence, we observe the bankng portfolos from the pont of vew of an Amercan nvestor (Elton and Gruber [1995]). We wll use secondary data, extracted from Data Stream. The data conssts of daly closng prces for the selected bankng companes and the closng market ndces of Morgan Stanley Captal Index (MSCI). The am s to have one base of market return, so that the return of each portfolo

12 160 The Lahore Journal of Economcs, Vol.9, No.1 wll be reflected from ts relaton wth one specfc market. To homogense the returns n one currency.e. the US dollar, we wll also use the exchange rates of the local currences, for each of the South East Asan countres, aganst the US dollar. Smlarly durng pre Euro perod, we wll use the exchange rate for every Euro zone country s orgnal currency aganst the US dollar and lastly the US dollar-euro exchange rate. All these rates wll be on a daly bass. 2. Crtera Lmtaton Durng sample selecton, we observed there are not many banks lsted on the stock exchanges of the countres under consderaton. Moreover, some banks have strategc busness unts that are also lsted as a separate entty. These busness unts are dfferent from the man bank. They normally have small captalsaton and are thnly traded. Ther selecton n the data could possbly result n bas due to frm sze effect. So we exclude all such busness unts from our research. Hence only the parent bankng company s selected from the whole bankng group. Ths lmtaton was more ntense whle selectng stocks n the bankng sector from Euro zone countres. In the case of the Unted States and South East Asa the constrant was somewhat lmted. The banks that are more nto advsory servces and nvestment bankng lke Morgan Stanley, Goldman Sachs and Merrll Lynch are also excluded from the research despte ther hgher captalsaton that s, at tmes, greater than some of the European and Asan banks. To be elgble for ncluson as a sample, each company had to meet the followng crtera: 1. The company must be a publc lmted bankng company, lsted on NYSE, organsed exchanges of Euro zone and organsed exchanges of South East Asa from March 1994 untl March The company must be transacted on the above mentoned stock exchanges and must have a complete record of daly prces. The perod of contnued lstng s mportant because we dvde our study nto three perods, namely pre crss, crss and post crss, for beta estmaton and ts comparson. Many banks from South East Asa went bankrupt durng the perod of crss and hence are excluded from the sample. The fnal sample ncluded 30 bankng companes from each regon.

13 Nawazsh Mrza and Danel Danny Smatupang 161 All these banks have smlar captalsaton n purchasng power party terms. Smlarty n captalsaton s requred to avod a possble bas that mght arse due to frm sze effect. The research has shown that sze to book value rato s an mportant factor n determnng the rsk profle of a company. Durng the fnancal crss n South East Asa some hgh performng banks were forced to be lqudated as they were not able to sustan the economc pressure. There were about a hundred banks that were lqudated durng the crss. Though these banks performed very well n the pre crss era but as they ceased to exst after the crss. They are excluded from the sample. 3. Econometrc Lmtatons Whenever beta s estmated there are certan conceptual problems assocated wth the estmaton. We wll present three most basc econometrc ssues related wth betas. 1. The systematc rsk or beta estmates are based on ex-ante rsk premums, whch are not drectly observable. These estmates are based on ratonal expectatons for an nvestor. Under ratonal expectatons, the realsed rates of return on assets n a gven tme perod are drawngs from the ex-ante probablty dstrbutons of returns on those assets. However, no logcal justfcaton can be gven that nvestors wll be ratonal over tme. 2. Betas are normally estmated usng lnear regresson. The underlyng assumpton for these estmates s the normal dstrbuton of returns. However, n realty the normalty of returns s not necessary. 3. The thrd major problem relates to the observaton of the proxy of market portfolo. In fact, many assets are not marketable and the proxes used for return on market portfolos exclude major classes of assets such as human captal, prvate busnesses and prvate real estate. The most common assumpton used to overcome ths problem s by assumng that the dsturbance terms from regressng the asset returns, on the return of the market proxy portfolo, are uncorrelated wth the true market portfolo and that the proxy portfolo has a unt beta. If the market proxy s a portfolo constructed from the ndvdual assets or portfolos contaned n the test sample, ths assumpton s equvalent to assumng that the market proxy s the mnmum varance unt beta portfolo of the set of all feasble portfolos constructed from the assets n the test sample.

14 162 The Lahore Journal of Economcs, Vol.9, No.1 4. Estmaton of Beta Coeffcents We wll use two methods to estmate the beta coeffcent. The frst one n by usng Ordnary Least Square (OLS) regresson or the sngle ndex model, and the second one s wth the Fowler-Rorke method. These two methods, however, are not appled to all portfolos. The OLS wll only be appled to Europe and the Unted States portfolo, because both markets are matured markets and have no sgnfcant sleepng stocks phenomenon. The Fowler- Rorke method wll be appled exclusvely for the Asan market, due to consderatons of non-synchronous tradng, sleepng stocks, and emergng markets. Elton and Gruber (1995) clearly explore that n a good portfolo, the Alpha and Beta respectvely, must be statstcally sgnfcant, equal to zero and one. Thus, frst we wll test the betas and alphas generated by both methods for all portfolos. Ths s amed to have a clearer overvew about the robustness of results. The sngle ndex model used for estmaton of beta wll be smlar to equaton (3), R = α R e. m The returns R and R m wll be calculated usng the logarthmc approach. The daly returns wll be of the form Pd R d = ln where P d and P d-1 are prces on day d and d-1. Pd 1 MSCIregX, d R md = ln where MSCI regx,d s Morgan Stanley Captal MSCIregX, d 1 Index on day d and d-1 for ether regon Europe, Asa or the Unted States. For South East Asan bankng stocks we wll use the Fowler-Rorke based Beta correcton method, wth several tests usng: (1) three lags and three leads, (2) four lags and four leads, and (3) fve lags and fve leads. The fnal objectve s to determne the crteron that best estmates market Beta value.e. closest to one (see Fowler and Rorke 1983).

15 Nawazsh Mrza and Danel Danny Smatupang 163 The Fowler-Rorke method can be establshed by frst calculatng the weghtng factors. The formula s as follows: n n n wn ρ ρ ρ ρ ρ ρ ρ = The values for ρ n are generated from a regresson equaton as follows: t n mt n mt mt mt e R R R R = ρ ρ ρ α The corrected Beta values usng Fowler-Rorke method s generated from: n n n n t w w w w = The n value s among three, four, and fve to test: (1) three lags and three leads, (2) four lags and four leads, and (3) fve lags and fve leads. The crteron that gves a market Beta value closest to one wll be used n calculaton. 5. Hypotheses Accordng to the general percepton the hypothess to be tested s that, based on systematc rsk, Asan bankng stocks are the rskest followed by that of Western Europe and the Unted States. The hypotheses (alternatve) tested are» For the perod March 1994 June 1997 (Pre Crss Perod) 0 : 0 : 0 : A US US E A E H H H» For the perod July 1997 December 2000 (Durng Crss) 0 : 0 : 0 : A US US E A E H H H

16 164 The Lahore Journal of Economcs, Vol.9, No.1» For the perod January 2001 March 2003 (Post Crss) H H H : : : E E US A US A The subscrpts n the term sgnfy countres: US for Unted States, E for Euro zone and A for South East Asa. V. Data Fndngs and Results The returns and beta coeffcents are estmated accordng to the sub-perods of hypothess: 1. March 1994 untl end of June 1997 for the pre-crss perod. 2. July 1997 untl end of 2000 for the crss perod. 3. January 2001 untl end of March 2003 for the post-crss perod. The followng Table presents the mean beta for all the perods usng Fowler Rorke for Asan portfolo whle OLS for Euro Zone and Unted States. Table 1: Beta for all Regons and all Perods One-Sample Statstcs N Mean Std. Devaton Std. Error Mean BETAEUR BETAEUR BETAEUR BETAUS BETAUS BETAUS BASIA1FR BASIA2FR BASIA3FR Note that n Table 1 above, the perod 1, 2, and 3 sgnfy the pre-crss, crss, and post-crss perods, respectvely, and EUR, US, and ASIA sgnfy the regons. The Fowler Rorke method used was 5 Leads and 5 Lags nstead of 3

17 Nawazsh Mrza and Danel Danny Smatupang 165 Leads and 3 Lags and 4 Leads and 4 Lags because the calculatons resulted n beta sgnfcantly close to 1 n 5 Leads and 5 Lags than the other two. The followng table presents the mean dfference hypotheses for systematc rsk usng t test approach. Table 2: Mean Systematc-Rsk Dfferences Sgnfcance Test Panel A: Beta Europe mnus Asa BT1EAF R F Sg. T Sg. (2-taled) Mean Dfference Std. Error E E E E E BT2EAF R E-10 BT3EAF R Panel B: Beta Unted States mnus Asa BT1AUF R BT2AUF R BT3AUF R F Sg. T Sg. (2-taled) Mean Dfference Std. Error 8.29E E E E E E E

18 166 The Lahore Journal of Economcs, Vol.9, No.1 Panel C: Beta Europe mnus Unted States F Sg. t Sg. (2- taled) Mean Dfference Std. Error BETA1E U BETA2E U s BETA3E U *Note that 1, 2, and 3 sgnfy the perods. As we can see from Table 2, all results lead us to reject our null hypotheses. The negatve mean dfference n Panel A of Table 2 shows that the rsk of Asan bankng stocks s greater than European ones, and all are statstcally sgnfcant at p = 1%. The same result s also establshed n Panel B, where rsk of Asan regon bankng stock s greater than that of the Unted States market, and all are sgnfcant. The comparson between Europe and Unted States markets gves a rsk n European portfolo rather than n that of the Unted States but sgnfcance s very low. The F-test shows great sgnfcance for Panel A and Panel B. Ths means that varances of beta n Panel A and Panel B are statstcally not equal and each beta stands on ts own varance. However, the F-test n panel C s not sgnfcant n comparson between Europe and the Unted States n the crss and the post crss perod. The Mann-Whtney U non-parametrc test confrms all these results.

19 Nawazsh Mrza and Danel Danny Smatupang 167 Table 3: Mann Whtney Non Parametrc Test Panel A Europe and Asa BT1EAFR BT2EAFR BT3EAFR Mann-Whtney U Wlcoxon W Z Asymp. Sg. (2-taled) where EAFR s Europe and Asa Fowler Rorke. Panel B Asa and Unted States BT1EAFR BT2EAFR BT3EAFR Mann-Whtney U Wlcoxon W Z Asymp. Sg. (2-taled) where AU s Asa FR and Unted States. Panel C. Europe and Unted States BT1EAFR BT2EAFR BT3EAFR Mann-Whtney U Wlcoxon W Z Asymp. Sg. (2-taled) where EU s Europe & Unted States From the Mann Whtney U test t s evdent that European and US s bankng stocks are sgnfcantly less rsky than that of Asa, for all perods, pre, crss and post. Whle the US Bankng sector appeared to be less rsky than that of the Euro Zone s, but ths fact s not sgnfcant n statstcal terms. These fndngs are n lne wth those we observed before. Thus we reject our null hypotheses and conclude that Asan stocks are the most rsky followed by European and US stocks. These fndngs are sgnfcant at the 5% level of sgnfcance. In order to test whether the three mean betas (ndependent sample) are not equal we further perform the Kruskal Walls Test. The mean betas

20 168 The Lahore Journal of Economcs, Vol.9, No.1 were found to be sgnfcantly dfferent from each other. The followng table summarses the result for the Kruskal Walls non parametrc test. Table 4: Kruskal Walls Ch Square Test Ranks REGIONS N Mean Rank BETA1 Europe Asa Unted States Total 90 BETA2 Europe Asa Unted States Total 90 BETA3 Europe Asa Unted States Total 90 BETA1 BETA2 BETA3 Ch-Square df Asymp. Sg where 1, 2 and 3 represent the three perods as used prevously. Our test statstcs resulted n rejecton of our null hypotheses and support our alternatve hypotheses. The beta of bankng stocks from Asa was found sgnfcantly hgher than that of Western Europe and the Unted States. Ths s agan smlar to what we obtaned ntally. The Ch square statstcs clearly specfy that there s no mean dfference between the three samples. However, the sgnfcance s qute low n the case of the Unted States vs. Western Europe. All the beta estmates were sgnfcant at p = 1%. It would be nterestng to observe the beta values of the bankng portfolo for the three perods under consderaton. The mean beta for the Euro Zone durng the pre crss perod was around Now f we consder the CAPM framework the beta estmate, whch s close to zero, makes the portfolo return equal to a rsk free rate. The equlbrum return n the presence of such beta would be slghtly hgher than that of rsk free

21 Nawazsh Mrza and Danel Danny Smatupang 169 rate prevalng n the economy. Smlarly beta n the US for ths perod was 0.06, gvng a smlar observaton of bankng portfolo for the perod. In Asa, as demonstrated by our statstcs, the beta was sgnfcantly greater than one. One possble explanaton for ths phenomenon could be better prudental practces n the Unted States and Western Europe for the bankng sector than n South East Asa. In the perod, when the crss was at ts peak, the beta for all three markets ncreased. Some studes have shown that the crss n South East Asa not only affected the domestc economes but the fear of the spread of crss lke a contagon produced concerns n foregn economes ncludng the Unted States and Western Europe. The beta of Asan bankng stocks was around 3 and t s a clear pcture of what was happenng to the bankng sector n the ASEAN countres. In the post crss era, as s evdent from the emprcs above, the Asan stocks recovered. However the stocks of the Unted States and Western Europe became more rsky. In fact ths perod was the post September 11 era so nvestment was rsky. Ths perod could be termed as the crss perod for especally the Unted States and to some extent Western Europe. Lke all of Wall Street, bankng stocks also suffered and a hgh rsk profle emerged. VI. Concluson Ths study was amed at testng the comparatve rskness of bankng stocks n three dfferent geographc markets. The theory of fnance suggests that the systematc rsk s the only relevant rsk for whch the nvestor s rewarded. There are many factors that contrbute to the systematc rsk both at the macro and mcro levels. In an nternatonal envronment the systematc rsk becomes more relevant as t also ncludes country specfc factors such as country rsk, exchange rate rsk etc. The bankng sector s lke the backbone of the economy of any country and surprsngly the number of banks traded, on organsed or over the counter markets, s low. They also face the phenomenon of thn tradng due to nvestors low nterest n ther shares. The analyss was based on the performance of bankng stocks n the stock market. We compared the systematc rsk for three regons. South East Asa was ht by one of the worst fnancal crss n Ths crss badly affected the economy of the Asan regon. The fnancal sector was the most to suffer. Many banks were lqudated or were taken over by the government as they were unable to

22 170 The Lahore Journal of Economcs, Vol.9, No.1 sustan the pressure created by the crss. Some of the banks lterally went bankrupt overnght. The falure of central banks to handle the crss added fuel to the fre. However, durng ths era, the bankng stocks n EU and the US were performng normally. Hence we dvded the study perod nto three dfferent perods. The pre crss era when there was no abnormalty, the era of crss when the South East Asan sector was on fre and the post crss era when the bankng sector started to recover. In the post crss era the events of September 11 took place makng Western markets more rsky for nvestments. Our emprcal results, as reported, support our noton of rsk profles of the three regons. We reject our null hypothess for Asa vs. US and Asa vs. Europe concludng that Asan bankng stocks were more rsky than those of EU and the US. However, the results for Europe vs. US were not sgnfcant. These observatons were supported by Kruskal Walls Ch square and Mann Whtney test at the 5% level of sgnfcance. The mean beta estmates for the three perods ndcate that bankng portfolo of Asan stocks, durng the crss, was thrce as rsky as that of the market. Moreover the beta estmate for the US and EU ncreased sgnfcantly n the post crss perod. The possble explanaton for ths phenomenon could be the events of WTC makng nvestment n the stock market more rsky. The less than one beta portfolo n the Unted States and Western Europe make them a strong canddate for nvestment. However we feel that n order to have more nsght nto the ssue, performance of the stocks usng event study methodology, must be observed. Do these stocks outperform or underperform the market n the three perods and to what extent? Ths phenomenon requres further explanaton and could be a possble ssue for further research.

23 Nawazsh Mrza and Danel Danny Smatupang 171 References Altman, Jacqullat and Levassuer, 1974, Comparatve Analyss of Rsk Measures: France and Unted States, Journal of Fnance 29, No 5, December, pp Amhud, Yakov, and Ham Mendleson, 1986, Asset Prcng and the Bd- Ask Spread, Journal of Fnancal Economcs 17, No.2, December, pp Bacheler, L., 1900, "Théore de la Spéculaton", Annales de l'ecole Normale Supéreure de Pars. Baesel, Jerome B, 1990, On the Assessment of Rsk: Some Further Consderaton, Journal of Fnance 29, No. 5, December, pp Banz, Rolf W., 1981, The Relatonshp between Returns and Market Value of Common Stocks, Journal of Fnancal Economcs 9, No. 1, March, pp Barnes, Paul, 1986, Thn Tradng and Stock Market Effcency: The Case of the Kuala Lumpur Stock Exchange, Journal of Busness Fnance and Accountng, Wnter, pp Bartholdy, Jan, and Allan Rdng, 1994, Thn Tradng and the Estmaton of Betas: The Effcacy of Alternatve Technques, Journal of Fnancal Research 17, No. 2, Summer, pp Basu, Sanjoy, 1983, The Relatonshp between Earnngs Yeld, Market Value, and Return for NYSE Common Stocks: Further Evdence, Journal of Fnancal Economcs 12, No. 1, June, pp Berglund, Tom, Eva Lljeblom, and Anders Loflund, 1995, Estmatng Betas on Daly Data for a Small Stock Market, Journal of Bankng and Fnance, March, pp Bergstrom, Gary L., 1975, A new route to hgher returns and lower rsks, Journal of Portfolo Management, pp

24 172 The Lahore Journal of Economcs, Vol.9, No.1 Berry, Mchael A., George W. Gallnger, and Glenn V. Henderson Jr., 1987, Adjustng for Nonsynchronous Data: How Important s It?, Akron Busness and Economc Revew, Wnter, pp Bhardwaj, Rawnder K. and LeRoy D. Brooks, 1992, Stock Prce and Degree of Neglect as Determnants of Stock Returns, Journal of Fnancal Research 15, No. 2, Summer, pp Black, Fscher, 1973, Yes, Vrgna, There Is Hope: Tests of the Value Lne Rankng System, Fnancal Analysts Journal 29, No. 5, Sept/Oct, pp Blume, E. Marchall, 1971, On the Assessment of Rsk, Journal of Fnance 6, No. 1, March, pp Blume, Marshall E., and Robert F. Stambaugh, 1983, Bases n Computed Returns: An Applcaton to the Sze Effect, Journal of Fnancal Economcs 12, No. 3, November, pp Brown, K.C., W.V. Harlow, and S.M. Tnc, 1988, Rsk Averson, Uncertan Informaton, and Market Effcency, Journal of Fnancal Economcs 22, pp Corrado, C. and Jordan, B., 2000, Rsk Averson, Uncertan Informaton, and Market Effcency: Re examnng the Evdence, Revew of Quanttatve Fnance and Accountng 8, pp Dmson, Elroy, 1979, Rsk Measurement When Shares are Subject to Infrequent Tradng, Journal of Fnancal Economcs 7, June, pp Erb, Claude, Campbell, Harvey & Vskantas, 1994, Forecastng Internatonal Correlaton». Fnancal Analysts Journal, 50, No.6 (Nov/Dec). pp Fama, Eugene F., 1970, Effcent Captal Markets: A Revew of Theory and Emprcal Work, Journal of Fnance 25, No. 2, May, pp Fama, Eugene F., 1991, Effcent Captal Markets: II, Journal of Fnance 46, No. 5, December, pp

25 Nawazsh Mrza and Danel Danny Smatupang 173 Fama, Eugene F., and Kenneth R. French, 1988, Permanent and Temporary Components of Stock Prces, Journal of Poltcal Economy 96, Aprl, pp Fama, Eugene F, and Kenneth R. French, 1992, The CAPM s wanted, Dead or Alve, Journal of Fnance 54, pp Fama, E., and MacBeth, J., 1973, Rsk, Return and Equlbrum: emprcal tests. Journal of Poltcal Economy, 81, pp Feltz, Bruce, 1974, Indrect versus Drect Dversfcaton, Fnancal Management 3, No. 4 (Wnter) pp Fletcher, J., 2000, On the Condtonal Relatonshp between Beta and Return n Internatonal Stock Markets. Internatonal Revew of Fnancal Analyss, 9, pp Fowler, Davd J., and C. Harvey Rorke, 1983, Rsk Measurement When Shares are Subject to Infrequent Tradng: Comment, Journal of Fnancal Economcs 12, August, pp Grubel, G. Herbert., 1968, Internatonally Dversfed Portfolos : Welfare Gans & Captal Flows, Amercan Economc Revew 58, pp Hartono, Jogyanto, and Suranto, 2000, Bas n Beta Values and Its Correcton: Emprcal Evdence from the Jakarta Stock Exchange, Gadjah Mada Internatonal Journal of Busness 2, No. 3, September, pp Isakov, D., 1999, Is Beta Stll Alve? Conclusve Evdence from the Swss Stock Market. The European Journal of Fnance, 5, pp Jaffe, Jeffrey, Donald B. Kem, and Randolph Westerfeld, 1989, Earnngs Yeld, Market Values, and Stock Returns, Journal of Fnance 44, No. 1, March, pp Klemkosky, Robert, and John Martn, 1975, The Adjustment of Beta Forecasts, Journal of Fnance 10, No. 4, September, pp

26 174 The Lahore Journal of Economcs, Vol.9, No.1 Krtzman, Mark, 1991, What Practtoners Need to Know About Regressons, Fnancal Analysts Journal 47, No. 3, May/June, pp Lakonshok, Josef, Andre Shlefer, and Robert W. Vshny, 1994, Contraran Investment, Extrapolaton, and Rsk, Journal of Fnance 49, No. 5, December, pp Lakonshok, J., and Shapro, A. C., 1986, Systematc Rsk, Total Rsk, and Sze as Determnants of Stock Market Returns. Journal of Bankng and Fnance, 10, pp Levy, Ham and Marshall Sarnat., 1970, Internatonal dversfcaton of nvestment portfolos. Amercan Economc Revew, 60, pp Lntner, J., 1965, The valuaton of rsk assets and the selecton of rsky nvestments n stock portfolos and captal budgets, Revew of Economcs and Statstcs 47, Markowtz, H., 1952, Journal of Fnance, Portfolo Selecton, March pp Mossn, J., 1966, Equlbrum n a captal asset market, Econometrca 34, Pettengll, G. N., Sundaram, S., and Mathur, I., 1995, The Condtonal Relaton between Beta and Returns. Journal of Fnancal and Quanttatve Analyss, 30, pp Poterba, James, and Lawrence Summers, 1988, Mean Reverson n Stock Prces: Evdence and Implcatons, Journal of Fnancal Economcs 22, No. 1, October, pp Renganum, Marc R., 1981, A new Emprcal Perspectve on the CAPM. Journal of Fnancal and Quanttatve Analyss, 16, pp Scholes, Myron, and Joseph Wllams, 1977, Estmatng Betas from Nonsynchronous Tradng, Journal of Fnancal Economcs 5, December, pp Schwert, G. Wllam, 1992, Sze and Stock Returns, and Other Emprcal Regulartes, Emprcal Research n Captal Markets, ed: G.