The day of the week effect patterns on stock market return and volatility: Evidence for the Athens Stock Exchange

Transcription

1 Neapolis Universiy HEPHAESTUS Reposiory School of Economic Sciences and Business hp://hephaesus.nup.ac.cy Conference papers 005 The day of he week effec paerns on sock marke reurn and volailiy: Evidence for he Ahens Sock Exchange Kenourgios, Dimiris hp://hdl.handle.ne/1178/707 Downloaded from HEPHAESTUS Reposiory, Neapolis Universiy insiuional reposiory

2 The day of he week effec paerns on sock marke reurn and volailiy: Evidence for he Ahens Sock Exchange Dimiris F. Kenourgios Universiy of Ahens, Faculy of Economics, 5 Sadiou Sree, Office 115, 1056 Ahens, Greece Tel: dkenourg@econ.uoa.gr Ariseidis G. Samias Universiy of he Aegean, Deparmen of Business Adminisraion, Business School 6 Chrisou Lada Sree, Office 15, Ahens, Greece asamias@econ.uoa.gr Spyros Papahanasiou School of Social Sciences, Hellenic Open Universiy, 140 Asklipiou Sr., GR Ahens s.papahanasiou@prelium.com Proceedings of he nd Applied Financial Economics (AFE) Inernaional Conference on Financial Economics, Samos island, Greece, July 15-17, 005 1

3 Absrac This paper invesigaes he day of he week effec in he Ahens Sock Exchange (ASE) General over a en year period divided ino wo subperiods: and Five major indices are also considered: Banking, Insurance, and Miscellaneous for he firs subperiod, and FTSE-0 and FTSE-40 for he second subperiod. Using a condiional variance framework, which exends previous work on he Greek sock marke, we es for possible exisence of day of he week variaion in boh reurn and volailiy equaions. When using he GARCH (1,1) specificaion only for he reurn equaion and he Modified-GARCH (1,1) specificaion for boh he reurn and volailiy equaions, findings indicae ha he day of he week effec is presen for he examined indices of he emerging ASE over he period However, his sock marke anomaly seems o loose is srengh and significance in he ASE over he period , which migh be due o he Greek enry o he Euro-Zone and he marke upgrade o he developed. Jel classificaion: G10; G1; G Keywords: Day of he week effec; mean sock reurns; volailiy; GARCH Inroducion Securiy price anomalies have araced he ineres of academic economiss, saisicians and marke professionals for many years. Since he seminal work of Fama (1965), a vas number of sudies have been made and many books have been wrien on his subjec. Some of hese anomalies are broadly known as calendar effecs. Calendar effecs in sock marke reurns have puzzled financial economiss for over 50 years. The mos imporan calendar effecs sudied are he day of he week effec (significanly differen reurns on some day of he week; usually higher Friday reurns and lower Monday reurns), he monhly or January effec (relaively higher January reurns), he rading monh effec (reurns higher over he firs fornigh of he monh) and he holiday effec (reurns higher on he days before vacaions). Thaler (1987a, 1987b) provides an early and parial survey, while Mills and Cous (1995) and Cous e al. (000) provide selecive and more recen inernaional references. For he day of he week effec in sock marke reurns, French (1980), Lakonishok and Levi (198), Rogalski (1984) and Keim and Sambaugh (1984) demonsrae he presence of his phenomenon. Oher sudies have examined he ime series sock price behaviour in erms of volailiy by using generalized auoregressive condiional heeroskedasiciy (GARCH) models (French e al., 1987; Hamao e al., 1990, Nelson, 1991; Campbell and Henschel, 199; and Glosen e al., 1993). For example, French e al. (1987) suppor ha unexpeced sock marke reurns are negaively correlaed o he unexpeced changes in volailiy, while Campbell and Henschel (199) found ha an increase in volailiy raises he required rae of reurn on common shares and hence lowers sock prices.

4 Generally, all hose sudies repor ha reurns in sock markes are ime varying and condiionally heeroskedasic. In a decision-making process, a raional financial decision maker mus ake ino accoun no only reurns bu also he variance (risk) or volailiy of reurns. I is imporan o idenify wheher here are variaions in volailiy of sock reurns and wheher a high (low) reurn is associaed wih a high (low) volailiy for a given ime. If cerain paerns in sock reurn volailiy can be idenified, hen invesors would make invesmens decisions based on boh reurn and risk easier. Uncovering cerain volailiy paerns in reurns migh also benefi invesors in valuaion, porfolio opimizaion, opion pricing and risk managemen. This paper aims o exen previous sudies for he ASE by providing evidence for he day of he week effec no only for he reurn equaion by using he GARCH (1,1) specificaion, bu also for boh he reurn and volailiy equaions by using he M-GARCH (1,1) specificaion. The ASE General index and hree major indusry indices (Banking, Insurance and Miscellaneous) are considered over he period , while he General and Banking Indices along wih he FTSE-0 and FTSE-40 indices are also considered over he period These wo ime periods are he mos recen periods ever invesigaed and include some of he mos imporan macroeconomic, poliical and sock marke evens ook place in Greece. I is worh noing ha only a few sudies concerning seasonaliies in he Ahens Sock Exchange are repored in he finance lieraure. Specifically, Alexakis and Xanhakis (1995) ake ino accoun ha he variance is dependen over ime while an EGARCH-M model invesigaes he volailiy. During he period from January 1985 o February 1994, a posiive reurn is found for Mondays, while Tuesdays show negaive reurns. Mills e al. (000) examine no only baske indices bu also consiuen socks of he Ahens Sock Exchange General index from 1986 o In accordance wih oher sudies, hey find significan evidence for higher reurns on Fridays and lower reurns on Tuesdays and Wednesdays. Moreover, hey suppor he exisence of he monhly, rading monh and holiday effecs, and he significan variaion of hese calendar regulariies across he consiuen shares of he General. Finally, Cous e al. (000) invesigae he exisence of securiy price anomalies for four indices (General, Banking, Insurance and Leasing) over he period Their finding ha he Friday reurns are always posiive and highes is consisen wih ha of Alexakis and Xanhakis (1995). Specifically, hey suppor he exisence of his anomaly for he general and bank indices, bu no for he insurance and leasing indices. They also provide evidence for a weekend effec, a significan January effec and he exisence of he holiday effec as he mos significan anomaly in he ASE. The res of he paper is organized as follows: Secion presens he daa se along wih he reasons for choosing he wo examined periods. Secion 3 discusses he mehodology employed in our invesigaion of he day of he week effec in he ASE. The resuls are presened in Secion 4, in relaion o he day of he week effec for he reurn equaion and for boh he reurn and volailiy equaions for he general and five indices (banks, insurance, miscellaneous, FTSE-0 and FTSE-40) of he ASE. Finally, in Secion 5, we draw conclusions concerning he exisence of he day of he week effec in he ASE. 3

5 Daa The daa consis of closing values of he general index of he Ahens Sock Exchange as well as he values of hree secor indices (banks, insurance, miscellaneous, FTSE-0 and FTSE-40), covering a en-year period 1. They are daily observaions beween January 1995 and 31 December 000 for he general, bank, insurance and miscellaneous indices, and 4 January 001 and 31 December 004 for he general, bank FTSE-0 and FTSE-40 indices. These wo periods were chosen for a number of reasons. Firs, hey simply updae earlier work ha has no considered periods beyond Second, hey cover some ineresing periods of sock marke behaviour and he Greek economy as evidenced by i) hree general elecions, ii) he worldwide crash in Hong-Kong in 1997, ii) he enry of Greece o he European Exchange Raes Mechanism II (1998), iii) he readjusmen of is macroeconomic variables in order o achieve he crieria o become he 1 h member of he Euro Zone, iv) he enry of Greece o he Euro Zone (001) v) he ASE insiuional reform of 1995 in an aemp o ease illiquidiy problems and foser an increased volume of ransacions, and vi) he characerizaion of he Greek sock marke as a developed marke since 001. The close o close daa does no conain informaion abou he paymen of dividends on socks. However, he exclusion of dividend paymens should no necessarily invalidae our resuls. Many researchers have discovered ha heir conclusions remain essenially unchanged wheher hey adjused heir daa for dividends or no (e.g., Lakonishok and Smid, 1988 and Fishe e al., 1993). Hence, hey sugges ha any dividend bias, which occurs from no employing dividend adjused reurns, is relaively small and is no sufficien o eliminae he calendar effecs or o have any impac on heir saisical significance. The daily sock reurns for day ( R ) are calculaed as 100 ln( P / P 1 ), where P is he index value on day and P 1 is he index value on day -1 (he previous day he ASE was open). Mehodology Mos of he sudies repored in he finance lieraure invesigae he day of he week effec in mean reurns by employing he convenional OLS mehodology on appropriaely defined dummy variables. However, his mehodology has wo drawbacks. Firs, he error erms may no be whie noise due o auocorrelaion and heeroskedasiciy problems resuling o misleading inferences. To address his drawback, we include lagged values of he reurn variable in a model wih he following sochasic equaion: R n ΤΤ TH ΤΗ F F + ai R i + I = i = α 0 ε (1) 1 Since 3/10/94, he miscellaneous Companies index has been subsiued by hree indices for holding companies, he consrucion enerprises and he miscellaneous companies index. The calculaion of he miscellaneous index has been ended in March 001. The FTSE-40 index focuses on 40 companies of middle capializaion. The FTSE-0 index is a large capializaion index which includes he 0 larges companies (blue chips) lised in ASE. 4

6 where R represens reurns on a examined index, M, TH, and F are he dummy variables for Monday, Tuesday, Thursday and Friday a ime, and n is he lag order which is specified by using he final predicion error crieria (FPEC). The second drawback is ha error variances may no be consan over ime. To address his second drawback, we allow variances of errors o be ime dependen o include a condiional heeroskedasiciy ha capures ime variaion of variance in sock reurns. The following GARCH (p,q) model proposed iniially by Engle (198) and furher developed by Bollerslev (1986) is used in analyzing he behaviour of he ime series over ime: h q p + β j ε a j + j= 1 j= 1 = α γ γh () jb j Thus, error erms have a mean of zero and a ime changing variance of h [ ε ~ (0, h )] 3. We consider various models o invesigae he day of he week effec in boh reurn and volailiy equaions. Our firs model is he GARCH (1,1) specificaion of he following form: R h n = ΤΤ TH ΤΗ F F + ai R i + λh + I = i = + β1aε 1 + γ 1bh 1 α 0 ε (1a) α (a) where λ is a measure of he risk premium, as i is possible ha he condiional variance, as proxy for risk, can affec sock markes reurns. If λ is posiive, hen he risk averse agens mus be compensaed o accep higher risk 4. In our second model, we include some exogenous variables ino he GARCH specificaion. This modificaion has been suggesed by a few sudies in he lieraure. For example, Karolyi (1995) includes he volailiy foreign sock reurns o explain he condiional variance of home counry sock reurns for he case of he Unied Saes and Canada, Hsieh (1988) includes he day of he week effec in volailiy for various exchange raes, and Kiymaz and Berumen (003) include he day of he week effec ino he volailiy equaion for Canada, Germany, Japan, Unied Kingdom and Unied Saes. Following he above sudies, we model he condiional variabiliy of sock reurns by incorporaing he day of he week effec ino he volailiy equaion. Thus, he consan erm of he condiional variance equaion is allowed o vary for each day. Therefore, our second model is he M-GARCH (1,1) specificaion of he following form: Following Kiymaz and Berumen (003), we exclude Wednesday s dummy variable from he equaion o avoid he dummy variable rap. The FPEC deermines n such ha i eliminaes auocorrelaion in he residual. 3 The GARCH (p,q) specificaion requires ha β j ε j + γ jbγh j p1 in order o saisfy he nonexplosiveness of he condiional variances. Furhermore, each α, jα, and jb has o be posiive o saisfy he nonnegaiviy of condiional variances for each given ime. 4 Here, we ake ino accoun he possibiliy ha he lagged values of he squared residuals and he condiional variances migh be oo resricive. q j= 1 a p j= 1 5

7 R n ΤΤ TH ΤΗ F F + ai R i + λh + I = i = + δ + δ ΤΤ + δ TH ΤΗ + δ F F + β jε 1 + γ 1bh 1 = α 0 ε (1a) h α (b) Finally, he parameers of he wo differen ypes of specificaions for he reurn and volailiy equaions are esimaed following he quasi-maximum likelihood (QML) esimaion inroduced by Bollerslev and Wooldridge (199) 5. Empirical resuls Various descripive saisics (sample mean, sandard deviaion, skewness and kurosis) for he sample of he ASE indices, as far as he day of he week effec is concerned, are examined (no repored here due o space limiaions). Examinaion of he means indicaes ha neiher reurns were consan hroughou he week nor he reurns on Monday were negaive, as suggesed by he day of he week effec. By examining he skewness for he reurn series of each index under consideraion, we find ha all sample disribuions are negaively skewed, indicaing ha hey are nonsymmeric. Furhermore, hey all exhibi high levels of kurosis, indicaing ha hese disribuions have hicker ails han normal disribuions. These iniial findings show ha daily reurns are no normally disribued and are characerised as lepokuric and skewed. We use Barle s es o examine wheher he consancy of he variances can be rejeced. The es (no repored here) rejecs he null hypohesis ha he variances are he same across differen days of he week. Tables 1 and repor he day of he week effecs and sock marke volailiies (reurns only and reurns and volailiies respecively) for he six indices under consideraion. Panel A of Table 1 displays he esimaes for reurn equaion. The FPEC suggess ha he order of reurn equaion is one for all he examined indices. The esimaed coefficiens of he Monday s dummy variables for he general index (period ), he miscellaneous index ( ), and he FTSE-0 and FTSE-40 indices ( ) are negaive and saisically significan a 5%, 5%, 5% and 1% respecively. The esimaed coefficiens for he general and he insurance indices ( ) are lowes and saisically significan a 10% and 1% respecively on Tuesdays, a finding which is consisen o Alexakis and Xanhakis (1995) and Mills e al. (000). The highes and saisically significan reurns for he general and bank indices ( ) are observed on Fridays, while he esimaed coefficiens of he dummy variables for he bank index ( ) are all insignifican. The coefficien of he condiional sandard deviaion of he reurn equaion (risk) is posiive for all he examined indices. However, i is 5 One disadvanage of using he GARCH (1,1) wih he relevan dummies for each anomaly is he possibiliy of being oo resricive. In order o assess he condiional variance beer, we include addiional erms in he condiional variance equaion. Specifically, we include (a) addiional lag values for he ARCH erm [GARCH (1,)], (b) addiional lag values for he GARCH coefficien [GARCH (,1)], and (c) hreshold GARCH (1,1) values for he innovaion effec. The resuls for all indices of he ASE are robus wih our previous findings and hese findings are no abulaed and repored. 6

8 saisically significan only for he miscellaneous index. Using he Wald es, he null hypohesis ha he day of he week dummy variables are joinly equal o zero is rejeced for he general, bank, insurance and miscellaneous indices (period ), while is acceped for he bank index (period ) a 1% and 5% level. Hence, he day of he week effec is presen for he examined indices of he period , while for he indices of he period , his effec looses is srengh for he general and FTSE-0 indices and srongly exiss only for he FTSE-40 index. In Panel A of Table 1, we also repor he esimaes of he GARCH (1,1) coefficiens. The esimaed coefficien of he consan erm for he condiional variance equaion is α, while and are he esimaed coefficien of he lagged value of he squared residual erm and he lagged value of he condiional variance respecively. Each of hese coefficiens is saisically and posiive for each index under consideraion. Also, he sum of he and coefficiens is less han one. Thus, our resuls sugges ha condiional variances are always posiive and no explosive in our samples. Panel B of Table 1 repors he Ljung-Box Q saisics for he normalized residuals and Engle s (198) ARCH-LM es a 5-, 10-, 15-, and 0-day lags. Almos none of hese coefficiens are saisically significan. Therefore, we canno rejec he null hypohesis ha he residuals are no auocorrelaed. Furhermore, here is no significan ARCH effec in any of he sampled indices. This finding indicaes ha he sandardized residuals erms have consan variances and do no exhibi auocorrelaion. The condiional variance of he reurns is hen allowed o change for each day of he week by modeling he condiional variance of reurn equaion as a modified GARCH. This is done o deec he presence of a day of he week effec in volailiy. In his framework, we reexamine boh he reurns and he condiional variance equaions. Findings are repored in Table. The esimaed coefficiens of he Monday s dummy variables are similar o he previous findings repored in Table 1 for he general index and he FTSE-0 and FTSE-40 indices of he period (lowes and saisically significan a 1%, 5% and 1% respecively). The same finding is observed for he bank index of he period The esimaed coefficiens for he general, bank, insurance and miscellaneous indices of he period are lowes (negaive) and saisically significan on Tuesdays. The coefficiens of he condiional sandard deviaion of he reurn equaion (risk) are posiive and saisically significan for he general, bank, insurance and miscellaneous indices of he period These resuls would indicae ha invesors wan o be compensaed wih higher reurns for holding riskier asses. The esimaed volailiy coefficiens for he consan erms, as well as he slope erms, are posiive and saisically significan. This finding saisfies he nonnegaiviy of he condiional variances. The resuls for condiional variance equaion are repored in Panel A of Table (lower par). The highes volailiy occurs on Mondays for he general index (period ), he bank index (period ), and he insurance and miscellaneous indices ( ), and on Fridays for he bank index ( ). However, wih he excepion of he esimaed coefficiens of he general and bank indices ( ), which are significan bu very close o zero, he oher esimaed values are saisically insignifican. Furhermore, he esimaed coefficiens indicaing he highes volailiy on 7

9 Mondays for he general index ( ) and he FTSE-0 index ( ) and on Thursday for he FTSE-40 index ( ) are negaive, alhough significanly close o zero for he firs wo indices. The lowes volailiy occurs on Fridays for he general index ( ), bank, FTSE-0 and FTSE-40 indices ( ), on Thursdays for general index ( ), bank and insurance indices ( ), and on Tuesdays for he miscellaneous index ( ). Wih he excepion of ha of he general and insurance indices on Thursdays, all he resuls are saisically significan, alhough very close o zero. The significanly highes and lowes volailiy seems o be spi among indices, where he general index for he period has significanly higher volailiy on Mondays, and he bank index ( ) on Fridays, while he FTSE-40 and he miscellaneous indices have significanly lower volailiy on Fridays and Tuesdays respecively (alhough very close o zero). The saisical evidence clearly suggess he presence of he day of he week effec on sock marke reurn volailiy in he ASE indices. By using he Wald es, we rejec he null hypohesis ha here is no day of he week effec in he condiional variance equaion for all he examined indices, excep for he bank and he FTSE-0 indices of he period (1% and 5% level). Hence, we confirm ha he day of he week effec is presen in boh he mean (reurn) and variance (volailiy of risk) equaions for he general, bank, insurance and miscellaneous indices of he period , and he general and FTSE-40 indices of he period On he conrary, his effec is no srongly presen in boh he mean and variance equaions for he bank index and he FTSE-0 index (weak evidence) of he period Panel B of Table repors he auocorrelaion Q saisics and ARCH- LM ess. The Q es indicaes ha here is no auocorrelaion for all indices under consideraion, excep he cases of he general, bank and FTSE-40 indices of he period Engle s ARCH-LM es saisics can rejec he null hypohesis of no ARCH effec for all indices excep general, bank and miscellaneous indices of he period Conclusions The day of he week effec anomaly is sudied and documened exensively in finance lieraure. This sudy invesigaes he day of he week effec on sock marke volailiy for major sock indices of he Ahens Sock Exchange using a condiional variance mehodology. When using daily closing values of he general, bank, insurance, and miscellaneous indices for he period , and he general, bank, FTSE-0 and FTSE-40 indices for he period , we documen he exisence/non-exisence of he day of he week effec in boh reurn and volailiy equaions The empirical analysis discussed in he previous secion is summarized and abulaed in Table 3. I clearly emerges from he able ha (i) he day of he week effec is presen in mean reurns for he ASE general index and he hree secor indices of he period , which is in par consisen o he evidence provided by Cous e al. (000), and Mills e al. (000), (ii) here is srong evidence for he day of he week effec in boh reurn and volailiy equaions for he examined indices of he period , which is in line wih he inernaional evidence of Kiymaz and Berumen (003), and (iii) i seems ha his sock marke anomaly no srongly exiss in boh reurn and 8

10 volailiy equaions for he indices covering he period , excep he case of he general and FTSE-40 indices. The day of he week effec paerns in reurn and volailiy migh enable invesors o ake advanage of relaively regular marke shifs by designing and implemening rading sraegies, which accoun for such predicable paerns. The findings of his paper suppor ha his poenial advanage of invesors due o he day of he week effec anomaly is presen in he emerging Greek sock marke of he period , bu seems o lose is srengh and significance afer he enry of Greece o he Euro-Zone and he upgrade o a developed marke. References Alexakis, P., and Xanhakis, M. (1995). Day of he week effec on he Greek sock marke. Applied Financial Economics, 5, Bollerslev, T. (1986). A generalized auoregressive condiional heeroscedasiciy. Journal of Economerics, 31, Bollerslev, T., and Wooldridge, J.M. (199). Quasi-maximum likelihood esimaion and inference in dynamic models wih ime varying covariances. Economeric Reviews, 11, Campbell, J.Y., and Henschel, L. (199). No news is good news: An asymmeric model of changing volailiy in sock reurns. Journal of Financial Economics, 31, Cous, J.A., Kaplanidis, C., and Robers, J. (000). Securiy price anomalies in an emerging marke: The case of he Ahens Sock Exchange. Applied Financial Economics, 10, Engle, R. (198). Auoregressive condiional heeroskedasiciy wih esimaes of he variance of Unied Kingdom inflaion. Economerica, Fama, E.F. (1965). The behavior of sock marke prices. Journal of Business, 34, French, K. (1980). Sock reurns and he weekend effec. Journal of Financial Economics, 8, French, K., Schwer, G., and Sambaugh, R. (1987). Expeced sock reurns and volailiy. Journal of Financial Economics, 19, Glosen, L.R., Jagannahan, R., and Runkle, D.E. (1993). On he relaion beween he expeced value and he volailiy of he nominal excess reurns on socks. Journal of Finance, 48, Hamao, Y., Masulis, R., and Ng, V. (1990). Correlaions in price changes and volailiy across inernaional sock markes. Review of Financial Sudies, 3, Hsieh, D.A. (1988). The saisical properies of daily foreign exchange raes: Journal of Inernaional Economics, 4, Karolyi, A.G. (1995). A mulivariae GARCH model of inernaional ransmission of sock reurns and volailiy: The case of he Unied Saes and Canada. Journal of Business and Economic Saisics, 13,

11 Keim, D.B., and Sambaugh, F. (1984). A furher invesigaion of weekend effecs in sock reurns. Journal of Finance, 39, Kiymaz, H., and Berumen, H. (003). The day of he week effec on sock marke volailiy and volume: Inernaional evidence. Review of Financial Economics, 1, Mills, T.C., Siriopoulos, C., Markelos, R.N., and Harizanis, D. (000). Seasonaliy in he Ahens Sock Exchange. Applied Financial Economics, 10, Nelson, D.B. (1991). Condiional heeroskedasiciy in asses reurns: A new approach. Economerica, 59, Rogalski, R.J. (1984). New findings regarding day of he week reurns over rading and non-rading periods: A noe. Journal of Finance, 39, Thaler, R.H. (1987a). Anomalies: The January effec. Journal of Economic Perspecives, 1, Thaler, R.H. (1987b). Anomalies: seasonal movemens in securiy prices II: Weekend, holiday, urn of he monh and inra-day effecs. Journal of Economic Perspecives, 1,

12 Table 1. The day of he week effec in reurn equaion Panel A: Esimaes of reurn equaion and variance Reurn equaion General Bank Insurance Miscellane ous FTSE 0 FTSE 40 ( ) Consan (0.0005) Monday (0.0003) Tuesday *** Thursday Friday *** Reurn * (0.079) Risk 0.19 (0.0719) Wald es [0.0068] ( ) ( ) (0.0017) (0.0005) ** (0.0010) (0.0011) (0.0005) 4.64E-05 (0.0011) (0.0011) * (0.036) (0.1353).3103 [0.0561] ** (0.0005) * (0.077) (0.0668).734 [0.08] [0.70] [0.005] [0.0017] ( ) ( ) ( ) (0.008) (0.0013) (0.0005) ** (0.0013) (0.0007) (0.0005) * * (0.0014) (0.0007) (0.0005) (0.0014) (0.0007) (0.0005) -8.50E (0.0015) (0.0007) (0.0005) * * * (0.0361) (0.096) (0.059) * (0.18) (0.1453) (0.0659) ( ) (0.001) ** (0.001) (0.001) (0.001) (0.001) * (0.0345) (0.1550) [0.0771] ( ) (0.0015) * (0.0011) (0.0011) (0.001) (0.0013) * (0.0369) (0.1091) [0.0047] Volailiy α 1.30E-06 * (1.74E-07) * (0.0144) 0.83 * (0.0116) 4.84E-06 * (1.53E-06) * (0.0136) * (0.0173) 9.30E-07 * (.10E-07) * (0.019) * (0.0109).14E-05 * (5.6E-06) * (0.0180) * (0.0311) 1.1E-05 * (.16E-06) 0.11 * (0.0175) * (0.0360) 4.8E-07 * (8.30E-08) * (0.0067) * (0.0066) 3.06E-06 * (1.09E-06) * (0.0113) * (0.01) 6.4E-06 * (1.6E-06) * (0.0168) * (0.0176) 11

13 Panel B: Auocorrelaion Q saisics and ARCH-LM ess for various lags Q(5) ARCH(5) Q(10) ARCH(10) Q(15) ARCH(15) Q(0) ARCH(0) General ( ).669 [0.811].8040 [0.730] [0.798] [0.869] [0.853] 7.70 [0.934] [0.943] [0.98] Bank ( ) ( ) [0.159].4051 [0.791] * [0.0049] [0.913] [0.3] [0.546] ** [0.0387] [0.987] [0.154] [0.79] [0.1067] [0.855] [0.175] [0.93] [0.1919] [0.866] ( ) [0.40] [0.1008] [0.458] [0.3675] [0.054] [0.439] [0.07] [0.647] Insurance ( ) [0.313] [0.8595] [0.444] [0.4630] [0.776] [0.497] [0.866] [0.7368] Miscellaneous ( ) FTSE 0 ( ) FTSE 40 ( ) [0.568] [0.170] ** [0.030] [0.4] * [0.0046].7990 ** [0.0161] [0.703] [0.06] [0.088] 7.49 [0.703] [0.0708] [0.0963] [0.655] [0.065] [0.058] [0.8756] [0.1859] [0.1695] [0.739] 7.67 [0.117] 9.66 [0.076] [0.8817] [0.3174] [0.481] Noe: *, **, *** denoe significance a 1%, 5% and 10% respecively. Sandard errors are repored in parenheses and p values in brackes. This noe also applies o Table. 1

14 Table. The day of he week effec in reurn and volailiy equaions Panel A: Esimaes of reurn and volailiy equaions Reurn equaion General Bank Insurance Miscellaneou s FTSE 0 FTSE 40 Tuesday Thursday (0.0005) Friday * Reurn * (0.056) Risk * (0.1103) (0.001) 4.86E-05 (0.001) (0.0011) * (0.0379) (0.1507) ( ) Consan (0.0007) Monday (0.0005) ( ) (0.000) * (0.0011) ( ) ** (0.0005) ** ** * (0.058) 0.74 * (0.08) ( ) (0.0031) ** (0.0013) (0.0015) (0.0015) (0.0014) * (0.0365) (0.1894) ( ) ( ) ( ) ( ) ** (0.0010) (0.0009) (0.003) (0.001) *** ** * (0.0006) (0.0005) (0.001) (0.0013) * * (0.0006) (0.0006) (0.0013) (0.0014) (0.0006) (0.0006) (0.0014) (0.0014) (0.0006) * (0.037) * (0.1107) * (0.0005) * (0.053) *** (0.1113) 4.17E-05 (0.0013) * (0.0365) (0.1548) (0.0014) * (0.037) (0.1191) Wald es [0.0000] [0.05] [0.0006] [0.1030] 1.68 [0.000] [0.0001].0309 [0.0880] [0.0007] Volailiy equaion α 3.46E-06 * (.55E-06) * (0.0173) * (0.0345) Monday.55E-05 * (3.50E-06) Tuesday -4.40E-06 (3.50E-06) Thursday 1.8E-05 * (4.13E-06) Friday -7.31E-06 ** (.98E-06) 4.18E-05 * (1.45E-05) * (0.017) * (0.0350) -1.01E-05 (1.87E-05) -.8E-05 (.1E-05) -4.35E-05 (.37E-05) -4.11E-5 ** (1.84E-05) -9.51E-06 * (.80E-06) 0.04 * (0.0185) * (0.0191) 4.16E-05 * (4.44E-06) 7.98E-0 (4.34E-06).67E-05 * (4.3E-06) 8.46E-06 * (3.31E-06) 6.43E-05 * (.7E-05) * (0.049) * (0.0471) 4.80E-07 (.9E-05) -3.15E-05 (3.5E-05) -3.7E-05 (3.55E-05) -7.34E-05 * (.8E-05).19E-05 * (3.98E-06) * (0.017) * (0.0335) 5.71E-06 (4.07E-06) -1.85E-05 * (3.83E-06) -6.1E-06 (6.59E-06) -.08E-05 * (3.55E-06) 1.93E-05 * (3.81E-06) * (0.0179) * (0.094) 1.67E-07 (4.57E-06) -9.64E-06 * (4.41E-06) -1.65E-05 * (6.39E-06) -1.48E-05 * (4.09E-06) 4.63E-05 ** (1.87E-05) * (0.0171) * (0.065) -.64E-05 (.37E-05) -3.99E-05 (.70E-05) -5.07E-05 (3.08E-05) -6.18E-05 * (.33E-05) 9.94E-05 * (1.0E-05) * (0.075) * (0.0344) -4.4E-05 * (1.34E-05) -7.80E-05 * (1.41E-05) * (1.79E-05) -8.00E-05 * (1.70E-05) 13

15 Panel B: Auocorrelaion Q saisics and ARCH-LM ess for various lags Q(5) ARCH(5) Q(10) ARCH(10) Q(15) ARCH(15) Q(0) ARCH(0) General ( ) [0.534] [0.674] [0.555] [0.836] [0.757] 8.59 [0.898] [0.866] [0.969] Bank ( ) ( ) 6.40 [0.69] 3.35 [0.650] * [0.0001] [0.348] [0.404] [0.193].3863 * [0.0085] [0.08] [0.71] [0.363].1886 * [0.0055] [0.137] 4.9 [0.30] [0.631] [0.0798] [0.1] ( ) Insurance ( ) Miscellaneous ( ) FTSE 0 ( ) FTSE 40 ( ) [0.471] 6.11 [0.86] [0.559] [0.349] 11.1 ** [0.047] [0.31] [0.460] 3.63 * [0.009] * [0.0000] * [0.0000] [0.449] [0.403] [0.563] [0.46] [0.19] 0.78 [0.646] [0.5657].993 * [0.0009] * [0.0001] * [0.0001] 5.98 ** [0.046] [0.713] [0.47] 0.86 [0.14] [0.134] [0.58] [0.6451].751 * [0.0003].8469 * [0.000] * [0.0001] 30.0 [0.067] [0.80] [0.79] 6.00 [0.166] [0.139] [0.784] [0.873].417 * [0.0005].048 * [0.0049].5391 * [0.000] Table 3. Summary of he day of he week effec in reurn and volailiy equaions Day of he week effec in reurn Day of he week effec in reurn and volailiy General Srong Srong Bank Srong Srong Insurance Srong Srong Miscellaneous Srong Srong General Weak Srong Bank None None FTSE 0 FTSE 40 Weak Srong Weak Srong 14