CALENDAR ANOMALIES AND CAPITAL MARKET EFFICIENCY: LISTED PROPERTY TRUST INVESTMENT STRATEGIES ABSTRACT VINCENT PENG Universiy of Wesern Sydney This sudy focuses on he efficien marke hypohesis (EMH) and inspecs he exisence of calendar anomalies in he Ausralian capial markes. I is he firs sudy ha provides a comprehensive examinaion of calendar anomalies in he Ausralian capial markes, paricularly in he Ausralian Lised Propery Trus (LPT) marke. The sudy reveals ha calendar anomalies exis in boh he Ausralian LPT marke and he broader equiy marke, providing evidence of marke inefficiency. I is also found ha some of hese reurn irregulariies are diminishing and dissipaing over ime, suggesing he move owards a higher level of efficiency. The resuls from his sudy provide suppor o acive invesing. Moreover, his sudy quanifies he poenial for profiing on he calendar anomalies by acive invesing in a pracical sense and provides significan pracical implicaions for LPT porfolio managers and invesors. Keywords:, EMH, calendar anomalies, invesmen sraegy. INTRODUCTION The efficien marke hypohesis (EMH) has significan implicaions for capial marke invesmen sraegies. If capial markes are perfecly efficien, invesmen sraegies should simply concenrae on he consrucion of a porfolio ha accommodaes he risk-reurn preferences of he invesors, because no abnormal reurns can be realised if all securiies are already fully priced and any aemps o ouperform he marke would simply be of limied value. In such a case, any ouperformance by acive managemen is purely luck raher han skill. If he EMH does no hold, here exis opporuniies for porfolio managers o exploi abnormal reurns in he marke, supporing acive invesing. The EMH suggess ha all securiies prices reflec all relevan informaion. Based on he infusion of such informaion, marke efficiency akes hree forms: he weak form, he semi-srong form and he srong form (Fama, 1970). Deails on he hree forms of marke efficiency and heir pracical implicaions are demonsraed by Peng (2004a). Pacific Rim Propery Research Journal, Vol 11, No 1 65
Table 1: Three forms of marke efficiency Efficiency Form Level of Informaion Infusion Pracical Implicaion Weak Form Curren prices fully reflec all pas informaion concerning he securiies. Fuure price changes are random. Relaing o 'Price Predicabiliy' and "echnical Analysis' -- If his level of efficiency holds, securiies prices should follow a random walk, and prices in he fuure canno be prediced by sudying hisorical prices, quesioning he relevance of Technical Analysis. Semi-srong Form Securiies prices reflec all public informaion. Relaing o 'Evens' and 'Calendar Anomalies' -- If his level of efficiency holds, prices should only changes wih he infusion of new marke informaion such as announcemens, and here should no exis paerns of anomalous regulariies, such as calendar anomalies. Srong Form Securiies prices reflec all informaion including privae informaion Relaing o 'Insider Trading' -- If his level of efficiency holds, here should no exis abnormal reurns from insider rading. Peng (2004a) examines EMH for he Ausralian capial markes by invesigaing daily closing prices of Lised Propery Truss () and All Ordinaries. Based on he resuls of uni roo ess and co-inegraing analysis, Peng (2004a) provides evidence ha he Ausralian capial markes are efficien in he weak form. This sudy focuses on he semi-srong form EMH and invesigaes wheher calendar anomalies exis in he Ausralian LPT marke, by means of a comprehensive examinaion of he daily closing prices of for he period of 1 June 1992 o 31 May 2003. For comparison purposes, his sudy also ess he EMH for he broader Ausralian equiy marke. Numerous sudies have invesigaed he exisence of calendar anomalies for common socks and evidence of several calendar anomalies has been well documened by he finance lieraure. Evidence of he day-of-he-week effec, where reurns for socks are normally lower on Monday han he oher days, has been documened; for example, Wang, Li and Erickon (1997); Abraham and Ikenberry (1994); Admai and Pfleiderer (1989); Flannery and Proopapadakis (1988); Harris (1986); Jaffe and Weserfield (1985); Keim and Sambaugh (1984); Gibbons and Hess (1981); French (1980); Cross (1973). Evidence of he monhly effec, where reurns are higher in a cerain monh or monhs han he oher monhs, has been documened by Roll (1983), Keim (1983, 1985), and Rozeff and Kinney (1976). Since January is normally he monh wih higher reurns, he monh effec is also commonly known as he January effec. Brown, Keim, Kleidon and Marsh (1983) find ha average reurns o mos Ausralian socks are subsanially higher, no only in January, bu also in July, compared wih he oher en monhs. 66 Pacific Rim Propery Research Journal, Vol 11, No 1
Ariel (1987) and Ogden (1990) have provided evidence of he urn-of-he-monh anomaly, which implies ha reurns are greaer on he urn-of-he-monh rading days han oherwise. Ariel (1990) has also documened he pre-holiday effec, wih reurns being higher on rading days before holidays han he oher days of he year. In general, sudies on common socks have shown ha abnormal reurns can be earned a differen imes of a week, a monh or year, conrary o he implicaions of he EMH. In he real esae markes, a few sudies have invesigaed he exisence of calendar anomalies in he US REIT marke. Friday and Higgins (2000) provided evidence of he day-of-he-week effec. Colwell and Park (1990) and Liu and Mei (1992) noiced evidence of he January effec. Redman, Manakyan and Liano (1997) conduced an exensive examinaion of calendar anomalies for US REITs and provided evidence ha here exis he January effec, he urn-of-he-monh effec, he day-of-he-week effec and he pre-holiday effec in he REITs marke. However, in he lieraure search, no previous sudies have been found invesigaing he calendar anomalies in he Ausralian securiised propery marke. This sudy provides a comprehensive examinaion of calendar anomalies in he Ausralian capial marke, paricularly in he LPT marke. In he field of real esae sudies, i pus he examinaion of calendar anomalies ino he conex of EMH inspecion. The remainder of his paper is srucured as follows. Secion wo briefly describes he daa and hen inroduces and jusifies he mehodology used in his sudy. Secion hree provides resuls and analysis of he examinaion of calendar anomalies. Pracical implicaions are illusraed and discussed in secion four, and he las secion provides concluding commens. DATA AND METHODOLOGY Daa Daa used in his sudy are he marke daily closing price indices for he LPT 300 and All Ordinaries over he period of 1 June 1992 o 31 May 2003, sourced from he Ausralian Sock Exchange (ASX). There are oally 2786 rading days over his period. Simple daily (as opposed o logarihmic reurns) are calculaed for each of he 2786 days and used in his sudy, o conform o he common pracices in he markeplace (Peng, 2004b). Table 2 provides some descripive analysis of he daa used for his sudy. Pacific Rim Propery Research Journal, Vol 11, No 1 67
Table 2: Summary of descripive saisics (1 June 1992-30 May 2003) Daily Reurns (Day-of-he-Week Effec) DAY Observaions Mean Sandard Deviaion Mean Sandard Deviaion Monday 535-0.00027 0.00615-0.00004 0.00932 Tuesday 564 0.00074 0.00648 0.00015 0.00805 Wednesday 567 0.00027 0.00640 0.00048 0.00858 Thursday 567 0.00014 0.00606 0.00045 0.00744 Friday 553-0.00001 0.00603 0.00008 0.00802 All 2786 0.00018 0.00623 0.00023 0.00829 Daily Reurns (Monhly Effec) Monh Observaions Mean Sandard Deviaion Mean Sandard Deviaion January 222 0.00003 0.00650 0.00050 0.00828 February 222-0.00012 0.00637 0.00002 0.00754 March 240 0.00001 0.00620-0.00018 0.00835 April 209 0.00026 0.00571 0.00140 0.00870 May 244 0.00029 0.00493-0.00007 0.00705 June 225 0.00019 0.00584 0.00012 0.00719 July 244 0.00109 0.00626 0.00015 0.00792 Augus 243-0.00057 0.00623-0.00028 0.00775 Sepember 236 0.00022 0.00691-0.00062 0.00946 Ocober 243-0.00019 0.00767 0.00015 0.01029 November 236 0.00040 0.00570 0.00039 0.00810 December 222 0.00055 0.00599 0.00142 0.00822 All 2786 0.00018 0.00623 0.00023 0.00829 Daily Reurns (Pre-Holiday Effec) Pre-Holiday Observaions Mean Sandard Deviaion Mean Sandard Deviaion No 2725 0.00015 0.00625 0.00022 0.00834 Yes 61 0.00138 0.00510 0.00056 0.00611 All 2786 0.00018 0.00623 0.00023 0.00829 Daily Reurns (Turn-of-he-Monh Effec) Turn-of-he-Monh Observaions Mean Sandard Deviaion Mean Sandard Deviaion No 2277 0.00007 0.00612 0.00005 0.00828 Yes 509 0.00066 0.00669 0.00102 0.00830 All 2786 0.00018 0.00623 0.00023 0.00829 Mehodology This sudy invesigaes he exisence of calendar anomalies in he Ausralian LPT marke and he broader equiy marke, including a comprehensive examinaion of he day-of-he-week effec, he monhly effec, he urn-of-he-monh effec and he pre-holiday effec. The exisence of hose calendar anomalies would sugges ha 68 Pacific Rim Propery Research Journal, Vol 11, No 1
he infusion of new informaion is no he only facor ha changes securiies prices, violaing he semi-srong form EMH. The mehods are deailed as follows. Day-of-he-week effec Equaion (1) is he regression used o examine he day-of-he-week effec, wih dummy variables represening he days of he week. R = a1 + a2d2, + a3d3, + a4d4, + a5d5, + (1) ε where: R is he daily reurn on day D 2,...D 5, is 1 if day is Tuesday Friday, and 0 oherwise ε is he error erm. In equaion (1), he inercep measures he average daily reurn on Monday. A posiive and significan inercep implies ha he average reurn on Monday is significanly greaer han zero. The coefficiens a 2 hrough a 5 are he pairwise comparison beween he average reurn on Monday and he average reurn on Tuesday hrough Friday. A posiive and significan a 2 indicaes ha he average reurn on Tuesday is significanly higher han ha on Monday. The coefficiens for he remaining hree dummy variables are inerpreed similarly. The F-value from equaion (1) measures he join significance of he coefficiens. In addiion o he parameric es, a nonparameric van der Waerden es (Conover, 1980) is conduced o es he equaliy of reurns across he days of he week. A significan F-value and van der Waerden es would rejec he hypohesis ha reurns are equal across days, providing evidence of he day-of-he-week effec. In addiion o esing for he join hypohesis, his paper also ess wheher he average daily reurn on Monday or Tuesday is significanly differen from he res of he week as a whole, using he following equaion: R ε = a1 + a2 D2, + (2) where: D 2, is 1 if day is Monday, and 0 oherwise in he es for Monday; is 1 if day is Tuesday, and 0 oherwise in he es for Tuesday; and all oher variables are as defined immediaely above. In equaion (2), he inercep measures he average daily rae of non-mondays (or non-tuesdays) as a whole, and he coefficien a i is he pairwise comparison beween he average reurn on Monday and ha for he res of he week as a whole. Pacific Rim Propery Research Journal, Vol 11, No 1 69
Monhly effec Equaion (3) is he regression used o examine he day-of-he-week effec, wih dummy variables represening he monhs of he year. R = a1 + a2d2, + a3d3, + a4d4, +... + a11d11, + a12d12, + (3) ε where: D 2,...D 12, is 1 if day falls in he monhs of February December, and 0 oherwise; and all oher variables are as defined immediaely above. In equaion (3), he inercep measures he average daily reurn in January. The coefficiens a 2 hrough a 12 are he pairwise comparison beween he average reurn in January and he average reurn in February hrough December. In addiion o esing for he join hypohesis, his paper also ess wheher he average daily reurn in July is significanly differen from he res of he year as a whole, using he following equaion: R ε = a1 + a2d2, + (4) where: D 2, is 1 if day falls in July, and 0 oherwise; and all oher variables are as defined immediaely above. In equaion (4), he inercep measures he average daily reurn in he monhs oher han July as a whole, and he coefficien a 2 is he pairwise comparison beween he average reurn in July and ha for he res of he year as a whole. All coefficiens and es saisics are analysed and inerpreed in a similar way o ha for he day-of-he-week effec. Pre-holiday effec This sudy compares he reurns on rading days before eigh Ausralia naional holidays (New Year s Day, Ausralia Day, Good Friday, Easer Monday, Anzac Day, Queen s Birhday, Chrismas Day and Boxing Day) o he reurns on non-preholiday rading days. Equaion (5) is used o examine wheher he pre-holiday reurns o and All Ordinaries are significanly differen from he non-pre-holiday reurns. R ε = a1 + a2 D2, + (5) 70 Pacific Rim Propery Research Journal, Vol 11, No 1
where: D 2, is 1 if day is a pre-holiday rading day, and 0 oherwise; and all oher variables are as defined immediaely above. A significan inercep suggess ha he reurns on non-pre-holiday rading days are significanly differen from zero. A posiive and significan coefficien implies ha pre-holiday reurns are significanly higher han non-pre-holiday reurns, providing evidence of he pre-holiday effec. A significan F-value and van der Waerden es also suppors he exisence of a pre-holiday effec. Turn-of-he-monh effec To analyse he urn-of-he-monh effec, rading days are classified ino urn-of-hemonh rading days (he final rading day of he previous monh and he firs hree rading days of he curren monh) and non-urn-of-he-monh rading days, a definiion adoped from Ogden (1990) and Redman, Manakyan and Liano (1997). Equaion (6) is used o compare urn-of-he-monh and non-urn-of-he-monh reurns. R ε = a1 + a2 D2, + (6) where: D 2, is 1 if day is a he urn-of-he-monh, and 0 oherwise; and all oher variables are as defined immediaely above. In equaion (6), he inercep measures he average daily rae of reurn on non-urnof-he-monh rading days. A posiive and significan coefficien indicae ha urnof-he-monh rading raes of reurn are significanly higher han non-urn-of-hemonh rading reurns, providing evidence of he urn-of-he-monh effec. A significan F-value and van der Waerden es also indicaes he presence of a urnof-he-monh effec. The ordinary leas square regressions of equaions (1) o (6) assume equal variances and no serial correlaion. The Durbin-Wason (DW) saisic is used o check for he presence of serial correlaion. The Levene es (Brown and Forsyhe, 1974) is used o check for he equaliy of variances. A significan Levene es suggess he presence of heeroscedasiciy. In he absence of serial correlaion and heeroscedasiciy, he -saisics are he sandard OLS -saisics. In he presence of heeroscedasiciy, he Whie (1980) adjused -saisics are repored. In he presence of serial correlaion or boh serial correlaion and heeroscedasiciy, he -saisics will be correced using he echnique of Hansen (1982). Pacific Rim Propery Research Journal, Vol 11, No 1 71
RESULTS AND ANALYSIS Day-of-he-week effec 1 A significan van der Waerden saisic and F-value would rejec he join hypohesis ha reurns are equal across days, providing evidence of he day-of-heweek effec. In Table 3, van der Waerden saisic and F-value are significan for bu no for All Ordinaries, suggesing ha daily reurns are no equal across days in marke, bu his inequaliy is no evidenced in he broader sock marke. Table 3: Day-of-he-week effec for LPT and sock markes (1 June 1992-30 May 2003) Coefficien -Saisics Coefficien -Saisics Consan -0.00027-1.00735-0.00004 [-0.09454] Tuesday 0.00101 2.67970 *** 0.00019 [0.35786] Wednesday 0.00054 1.43691 0.00052 [0.96076] Thursday 0.00041 1.10277 0.00049 [0.95960] Friday 0.00026 0.69838 0.00012 [0.23221] DW 1.98149 1.92358 Levene 0.01327 4.22105 *** van der Waerden 8.12513 * 1.04630 F-Value 1.98446 * 0.42606 *** Significan a 1% level; ** Significan a 5% level; * Significan a 10% level; -saisics in brackes are he Whie [1980] adjused -saisics. The consan is he average daily reurn earned on Monday. The coefficiens for Tuesday, Wednesday, Thursday and Friday are he excess daily reurns (agains ha on Monday) earned on hese days respecively. The average daily reurn on Monday for boh and All Ordinaries are negaive. However, his negaive reurn is no saisically significan. The posiive coefficiens for Tuesday, Wednesday, Thursday and Friday indicae ha average daily reurns on hese days are higher han ha on Monday. However, he posiive excess reurns earned on hese days are no significan, excep for he excess reurn on Tuesday in he LPT marke. The coefficien for Tuesday is srongly significan (a 1% level) wih a magniude of 10.1 basis poins, which implies ha in he LPT marke, he average daily reurn earned on Tuesday is abou 10 basis poins higher han ha earned on Monday. 1 In Table 2 hrough Table 10, he values of -saisics in brackes are he Whie [1980] adjused - saisics. 72 Pacific Rim Propery Research Journal, Vol 11, No 1
Over he 11-year sudy period, here are 531 pairs of rading Monday and Tuesday (i.e., a rading Monday followed by a rading Tuesday), ranslaing ino an average of 4.9% excess reurn per annum on Tuesday agains he average reurn on Monday, if here were no ransacion coss. In Table 3, i appears ha when compared wih Monday individually, he higher reurn on each of he non-mondays is insignifican excep for Tuesday in he LPT marke. I would be ineresing o es wheher he average daily reurn on non- Mondays, collecively, is significanly higher han ha on Monday. In oher words, wheher Monday has a significanly lower reurn han he res of he week as a whole. Similarly, i would be ineresing o es wheher reurn on Tuesday is significanly higher han ha on he res of he week as a whole. Table 4 shows he resuls of reurns on Monday compared wih he res of he week as a whole. Table 4: Monday effec for LPT and sock markes (1 June 1992-30 May 2003) Coefficien -Saisics Coefficien -Saisics Consan 0.00029 2.17859 0.00029 [1.73317] * Monday -0.00056-1.85992 * -0.00033 [-0.75846] DW 1.98118 1.92348 Levene 0.02518 9.41279 *** van der Waerden 3.23690 * 0.23018 F-Value 3.45931 * 0.69047 *** Significan a 1% level; ** Significan a 5% level; * Significan a 10% level; -saisics in brackes are he Whie [1980] adjused -saisics. Again, he van der Waerden saisic and F-value are significan for bu no for All Ordinaries, suggesing ha he reurns are no equal beween Monday and non-mondays in he LPT marke, bu his inequaliy is no evidenced in he broader equiy marke. Also, in he LPT marke, he average daily reurn on Monday is significanly lower han ha on non-mondays for abou 5.6 basis poins. Wih 535 rading Mondays over he 11-year sudy period (Table 2), buying on Mondays in he LPT marke would have he poenial o realise an excess reurn of 2.7% per annum over he sudy period if here were no ransacion coss. Table 5 compares he average daily reurns on Tuesday and non-tuesdays. Pacific Rim Propery Research Journal, Vol 11, No 1 73
Table 5: Tuesday effec for LPT and sock markes (1 June 1992-30 May 2003) Coefficien -Saisics Coefficien -Saisics Consan 0.00004 0.28520 0.00025 [1.40942] Tuesday 0.00070 2.37829 ** -0.00010 [-0.25994] DW 1.98121 1.92300 Levene 0.01091 4.80904 ** van der Waerden 6.33307 ** 0.07365 F-Value 5.65624 ** 0.06463 *** Significan a 1% level; ** Significan a 5% level; * Significan a 10% level; -saisics in brackes are he Whie [1980] adjused -saisics. Again, he van der Waerden saisic and F-value are significan for, bu no for All Ordinaries, suggesing ha he reurns are no equal beween Tuesday and non-tuesdays in he LPT marke, bu his inequaliy is no evidenced in he broader equiy marke. Also, in he LPT marke, he average daily reurn on Tuesday is significanly higher han ha on non-tuesdays for abou 7.0 basis poins. Wih 564 rading Tuesdays over he 11-year sudy period (Table 2), selling on Tuesday could poenially earn an excess reurn of 3.6% per annum if here were no ransacion coss. In summary, here is evidence ha he day-of-he-week effec exiss in he LPT marke, bu no he broader sock marke. In he LPT marke, he average daily reurn on Monday is significanly lower han ha on non-mondays collecively, providing poenial for abnormal reurns by buying on Mondays. Also in he LPT marke, he average daily reurn on Tuesday is significanly higher han ha on non- Tuesdays collecively, providing poenial for abnormal reurns by selling on Tuesdays. Monhly effec Table 6 shows he resuls of he ess for he January effec for and All Ordinaries. In Table 6, he van der Waerden saisic and F-value are insignifican for boh and All Ordinaries, suggesing ha he null hypohesis can no be rejeced and here is no evidence ha he daily reurns are no equal across monhs in and he broader sock markes. The insignifican consan (alhough posiive) and he posiive coefficiens (alhough no all significan) for some monhs also sugges he non-exisence of a January effec in and he broader equiy marke in Ausralia. 74 Pacific Rim Propery Research Journal, Vol 11, No 1
Table 6: January effec for LPT and sock markes (1 June 1992-30 May 2003) Coefficien -Saisics Coefficien -Saisics Consan 0.00003 [0.07194] Consan 0.00050 [0.90090] February -0.00015 [-0.25195] February -0.00048 [-0.64187] March -0.00002 [-0.04059] March -0.00068 [-0.87499] April 0.00023 [0.39609] April 0.00090 [1.09347] May 0.00026 [0.48771] May -0.00058 [-0.80324] June 0.00016 [0.26502] June -0.00038 [-0.51612] July 0.00106 [1.78704] * July -0.00035 [-0.46561] Augus -0.00060 [-1.00878] Augus -0.00078 [-1.04079] Sepember 0.00019 [0.30103] Sepember -0.00112 [-1.35167] Ocober -0.00022 [-0.34016] Ocober -0.00036 [-0.41247] November 0.00037 [0.63664] November -0.00011 [-0.14022] December 0.00052 [0.86784] December 0.00092 [1.17290] DW 1.99061 DW 1.93058 Levene 0.02100 ** Levene 2.21276 ** van der Waerden 12.88802 van der Waerden 15.52733 F-Value 1.05037 F-Value 1.23283 *** Significan a 1% level; ** Significan a 5% level; * Significan a 10% level; -saisics in brackes are he Whie [1980] adjused -saisics. In he LPT marke, i appears ha he average daily reurn in July is significanly higher han in January, suggesed by he significanly posiive coefficien. Thus, i would be ineresing o examine wheher reurns in July are significanly differen form hose in oher monhs in he LPT marke. Two hypohesis ess are performed. The firs one is a join es similar o ha for January effec, bu he consan would represen for July raher han January. The focus of his es would be on he coefficiens and relevan -saisics, because he Durbin-Wason, Levene and van der Waerden saisics, and he F-value would be expeced o remain he same as hose in he column in Table 6. Tha is, we focus on he comparison of reurns beween July and each of he oher monhs. The second es is o compare he average daily reurn o July wih ha o all he oher monhs as a whole; a es similar o hose used in he secion assessing he day-ofhe-week effec. Table 7 presens he resuls of hese wo ess. The join es suggess ha he average daily reurn in July is posiive and in he amoun of 10.9 basis poins, a figure consisen wih ha in Table 1. This posiive reurn is significan a he 1% level. All coefficiens are negaive, indicaing ha he average daily reurns in each Pacific Rim Propery Research Journal, Vol 11, No 1 75
of he oher monhs are lower han ha in July. Of hese negaive coefficiens, January, February, March, Augus and Ocober are significan. Table 7: July effec for LPT marke (1 June 1992-30 May 2003) Join Hypohesis July Coefficien -Saisics Coefficien -Saisics Consan 0.00109 [2.71935] *** Consan 0.00009 0.74185 January -0.00106 [-1.78704] * July 0.00100 2.39151 ** February -0.00121 [-2.06950] ** March -0.00108 [-1.91090] * April -0.00083 [-1.46729] May -0.00080 [-1.56046] June -0.00090 [-1.61670] Augus -0.00166 [-2.92447] *** Sepember -0.00087 [-1.44302] Ocober -0.00128 [-2.02057] ** November -0.00069 [-1.27021] December -0.00054 [-0.95826] DW 1.99061 DW 1.99061 Levene 0.02100 ** Levene 0.46826 van der Waerden 12.88802 van der Waerden 5.67671 ** F-Value 1.05037 F-Value 5.71929 ** *** Significan a 1% level; ** Significan a 5% level; * Significan a 10% level; -saisics in brackes are he Whie [1980] adjused -saisics The second es suggess ha he average daily reurn in July is significanly higher han ha in he non-july monhs, based on he significan (a 5% level) -saisic, van der Waerden saisic and F-value. Compared wih he non-july monhs as a whole, July has an average excess daily reurn of 10 basis poins. Wih 244 days in July over he 11-year sudy period, his will ranslae ino an excess reurn of 2.3% per annum over ha period, if here were no ransacion coss. In summary, here is no evidence ha he monhly effec exiss in he broader equiy marke. In he LPT marke, here is no January effec ha is evidenced by previous sudies on common socks. However, i appears ha July has consanly araced significanly higher average daily reurns compared wih he oher monhs as a whole as well as wih some of he oher monhs individually. This July effec in he LPT marke provides poenial for abnormal reurns by selling in July. Pre-holiday effec Table 8 presens he resuls of he pre-holiday effec. 76 Pacific Rim Propery Research Journal, Vol 11, No 1
As shown in Table 8, for he enire sudy period, he insignificance of he - saisics, F-value and van der Waerden saisic suggess ha he pre-holiday effec does no exis in he broader equiy marke. Table 8: Pre-holiday effec for LPT and sock marke (1 June 1992-30 May 2003) Coefficien -Saisics Coefficien -Saisics Consan 0.00015 1.27513 0.00022 [1.39232] Pre-Holiday 0.00123 1.51930 0.00033 [0.42209] DW 1.98385 1.92309 Levene 1.57235 4.74501 ** van der Waerden 3.10106 * 0.12977 F-Value 2.30828 0.09701 *** Significan a 1% level; ** Significan a 5% level; * Significan a 10% level; -saisics in brackes are he Whie [1980] adjused -saisics. For, while he -saisics and F-value are insignifican, suggesing no preholiday effec, he nonparameric es (van der Waerden) is significan suggesing he exisence of such an effec. This may be simply because ha, in he LPT marke, here exis differences in he daily reurns beween pre-holiday rading days and he oher rading days, bu for he enire sudy period, his effec is no srong enough for he -saisic and F-value o be significan. To examine wheher his is he case, are furher examined and i is found ha significan pre-holiday effec does exis in he early days in he 11-year sudy period, bu his effec diminishes over ime and dissipaes afer 31 May 2001. Table 9 presens he resuls. As shown in Table 9, he firs sub-sample (sub-sample 1) akes he period of 1 June 1992 o 31 May 1994. The sub-sample is hen exended by a wo-year inerval unil 31 May 2000, i.e., sub-sample 2 (1 June 1992 o 31 May 1996), sub-sample 3 (1 June 1992 o 29 May 1998) and sub-sample 4 (1 June 1992 o 31 May 2000). The las sub-sample (sub-sample 5) ends up wih 31 May 2003, where he pre-holiday effecs disappear. Pacific Rim Propery Research Journal, Vol 11, No 1 77
Table 9: Pre-holiday effec for LPT and sock markes (sub-samples) (Sub-sample 1) (Sub-sample 2) (Sub-sample 3) (1 June 1992-31 May 1994) (1 June 1992-31 May 1996) (1 June 1992-29 May 1998) Coefficien -Saisics Coefficien -Saisics Coefficien -Saisics Consan 0.00013 0.48274-0.00004-0.21064 0.00017 1.10381 Pre-Holiday 0.00490 2.63928 *** 0.00245 2.05670 ** 0.00202 1.97352 ** DW 2.00052 2.00919 2.04829 Levene 1.01655 0.75941 1.53134 van der Waerden 7.30068 *** 4.95415 ** 5.13824 ** F-Value 6.96579 *** 4.23000 ** 3.89479 ** (Sub-sample 4) (Sub-sample 5) (1 June 1992-31 May 2000) (1 June 1992-31 May 2001) Coefficien -Saisics Coefficien -Saisics Consan 0.00010 0.69963 0.00010 0.77789 Pre-Holiday 0.00166 1.70364 * 0.00152 1.67122 * DW 2.00604 1.98536 Levene 1.99081 0.89082 van der Waerden 3.75061 * 3.76616 * F-Value 2.90240 * 2.79299 * *** Significan a 1% level; ** Significan a 5% level; * Significan a 10% level; -saisics in brackes are he Whie [1980] adjused -saisics. In Table 9, no only he significance level of he -saisics, F-values and van der Waerden saisics is declining wih he exension of esing period, he magniude of he coefficien for pre-holiday is also decreasing. This implies ha he preholiday effec has been significan in he early days of he sudy period, bu he significance of his effec is declining over ime, unil disappearing oally, suggesing he move owards higher efficiency. In summary, for he enire sudy period, here is no evidence of he pre-holiday effec eiher in he LPT marke or he broader sock marke. However, in he LPT marke, pre-holiday effec does exis in he early days of he sudy period. The evidence of a diminishing pre-holiday effec may well sugges a move owards higher levels of efficiency. Turn-of-he-monh effec Table 10 shows he resuls for he urn-of-he-monh effec. The resuls are consisen for and All Ordinaries. The -saisics, F-values and he van der Waerden saisics are all significan, suggesing he exisence of significan urn-of-he-monh effec in boh and he broader equiy markes. 78 Pacific Rim Propery Research Journal, Vol 11, No 1
Table 10: Turn-of-he-monh effec for LPT marke and December effec for sock marke (1 June 1992-30 May 2003) *** Significan a 1% level; Coefficien -Saisics Coefficien -Saisics Consan 0.00007 [0.55650] 0.00005 0.30333 Turn-of-he-Monh 0.00059 [1.82415] * 0.00097 2.38512 ** DW 1.98478 1.92706 Levene 2.94511 * 0.60032 van der Waerden 3.12388 * 5.42771 ** F-Value 3.72099 * 5.68881 ** ** Significan a 5% level; * Significan a 10% level; -saisics in brackes are he Whie [1980] adjused -saisics. The average daily excess reurns on he urn-of-he-monh rading days are 5.9 basis poins and 9.7 basis poins for and All Ordinaries respecively, compared wih he average daily reurns on he oher rading days. Wih 509 days (Table 2) falling ino he urn-of-he-monh rading days over he 11-year sudy periods, his implies he poenial for obaining an average excess reurn of 2.7% and 4.5% per annum in and he broader sock markes respecively, if here were no ransacion coss. PRACTICAL IMPLICATIONS The invesigaions on wheher he EMH holds for Ausralian and he broader equiy markes are imporan o deermining he managemen syles beween acive and passive managemen/invesing. Managers who are supporive of acive managemen/invesing argue ha markes are inefficien and herefore ouperformance can resul from heir skills and experise in he specific markes. However, if markes are efficien, hen buying and selling securiies in an aemp o ouperform he marke will effecively be a game of luck raher han skill. Financial markes are flooded wih inelligen, well-educaed and well-paid invesmen analyss and porfolio managers who are seeking under or over-valued securiies o buy or sell in order o achieve ou-performance. In heory, he compeiions beween hese invesmen professionals should eliminae any speculaive and arbirage opporuniies, resuling in marke efficiency. In an efficien marke, acive managemen will become a zero-sum game, because curren prices reflec all informaion and he only way an invesor can profi will be purely by luck and a he expense of he loss from anoher less forunae acive paricipan. In such a case, he primary role of a porfolio manager, raher han seeking o bea Pacific Rim Propery Research Journal, Vol 11, No 1 79
he marke, should focus on ailoring a porfolio o invesors risk preferences, ax consideraions ec. The presence of calendar anomalies suggess ha and he broader equiy markes are inefficien in he semi-srong form, because prices are no only driven by informaion alone, bu also some oher facors. While only he urn-of-he-monh effec exiss in he broader equiy marke, here are he day-of-he-week effec, monhly effec and he urn-of-he-monh effec evidenced in he LPT marke. As illusraed in Secion hree, he accumulaed excess reurns ha may be realised from hese effecs are significan. For example, ignoring ransacion coss, he poenial for profiing from he day-of-he-week effec may be as large as 4.9% per annum in he LPT marke and here is he poenial for profiing 4.5% per annum from he urn-of-he-monh effec in he broader equiy marke. Taking ino accouns of ransacion coss, which are in he range of 10 o 30 basis poins for insiuional invesors in he Ausralian markes, he exploiable opporuniy from he calendar anomalies are cerainly limied. However, if porfolio managers who are planning o make a shif in he porfolio composiions can ime he rades o ake advanage of hese anomalous regulariies, hey will be able o save a few basis poins on each rade. Over ime, a few basis poins aggregaed from a large number of ransacions can significanly impac reurns. This sudy also provides suppor o he EMH heory, in ha compeiion from raional invesors will eliminae arbirage opporuniies and drive markes owards marke efficiency or higher levels of efficiency. For example, here exiss he preholiday effec in he early days of he sudy period in he LPT marke. However, boh he magniude and he level of significance of his effec in LPT markes are diminishing over ime and dissipaing evenually. This may well sugges a move owards a higher level of efficiency driven by compeiion from raional invesors, supporing he heory of EMH. Moreover, i appears ha he broader equiy marke is more efficien han he LPT marke. For example, for he four calendar anomalies esed in his sudy, only he urn-of-he-monh effec exiss in boh markes, and he day-of-he-week and monhly effecs are evidenced in he LPT marke bu no he broader equiy marke. This may simply be because ha LPT marke has a shor hisory and is less maure han he broader equiy marke; again, suggesing he move owards higher levels of efficiency over ime. All hese imply ha he poenial for profiing from he anomalous regulariies would no las forever, and porfolio managers should invesigae and ake advanages of hose opporuniies sooner raher han laer. The resuls from his sudy provide evidence ha and he broader equiy markes are inefficien in he semi-srong form, and suppor acive managemen/invesing. An ineresing phenomenon regarding he EMH is ha 80 Pacific Rim Propery Research Journal, Vol 11, No 1
marke efficiency depends on he belief ha markes are no efficien, and is driven by he aemp o ouperform he marke by marke paricipans who believe he marke is inefficien. If every invesor believed ha a marke was efficien (a case for passive managemen/invesing), hen he marke would no be efficien because no one would analyse securiies and i would be unlikely for securiies prices o fully reflec all available informaion. In fac, he weak form efficiency in he Ausralian LPT and he broader equiy markes, as evidenced in Peng (2004a) may simply resul from he compeiions by acive managers and raional invesors seeking ou performance in he presence of semi-srong form inefficiency such as he calendar anomalies evidenced in his sudy. CONCLUSION This sudy inspecs he semi-srong form of efficien marke hypohesis (EMH). I provides a comprehensive examinaion of calendar anomalies in he Ausralian capial markes, paricularly he LPT marke. In he field of real esae sudies, i pus he examinaion of calendar anomalies ino he conex of EMH inspecion; more specifically, semi-srong form EMH inspecion. The resuls of his sudy provide evidence ha calendar anomalies exis in he LPT and he broader equiy markes. This suggess ha he and broader equiy markes are inefficien in he semi-srong form, which provides opporuniies for profiing abnormal reurns. This paper furher quanifies he abnormal reurns ha could be poenially achieved by acive managers who would uilise he findings of he sudy. This paper also illusraes how he opporuniies o obain abnormal reurns could diminish and dissipae over ime, and suggess ha porfolio managers should invesigae and ake advanage of hese opporuniies sooner raher han laer. Finally, i provides srong reasoning for supporing acive managemen and invesing as opposed o passive managemen and invesing. REFERENCES Abraham, A. and Ikenberry, D. (1994), The individual invesor and he weekend effec. Journal of Financial Quaniaive Analysis, 29:2, 263-277. Admai, A. and Pfleiderer, P. (1989), Dividend and conquer: a heory of inraday and day-of-he-week mean effecs. The Review of Financial Sudies, 2:2, 189-223. Ariel, R. (1987), The monhly effec in sock reurns. Journal of Financial Economics, 18, 161-174. Pacific Rim Propery Research Journal, Vol 11, No 1 81
Ariel, R. (1990), High sock reurns before holidays: exisence and evidence on possible causes. The Journal of Finance, 45:5, 1611-1626. Brown, P., Keim, D., Kleidon, A. and Marsh, T. (1983), Sock reurn seasonaliies and he ax-loss selling hypohesis: analysis of he argumens and Ausralian evidence. Journal of Financial Economics, 12, 105-127. Brown, M. and Forsyhe, A. (1974), Robus ess for he equaliy of variances. Journal of he American Saisical Associaion, 69, 364 67. Colwell, P. and Park, H. (1990), Seasonaliy and size effecs: he case of real-esaerelaed invesmen. Journal of Real Esae Finance and Economics, 3, 251-259. Conover, W. (1980), Pracical Nonparameric Saisics (Second Ediion). John Wiley & Sons: New York. Cross, F. (1973), The behaviour of sock prices on Fridays and Mondays. Financial Analyss Journal, November-December, 67-69. Fama, E. (1970), Efficien capial markes: a review of heory and empirical work. The Journal of Finance, 25, 383-417. Flannery, M. and Proopapadakis, A. (1988), From -bills o common socks: invesigaing he generaliy of inra-week reurn seasonaliy. The Journal of Finance, 43:2, 431-450. French, K. (1980), Sock reurns and he weekend effec. Journal of Financial Economics, 9, 55-69. Friday, H. and Higgins, E. (2000), The day of he week effec in real esae invesmen russ. Journal of Real Esae Porfolio Managemen, 6:3, 273-282. Gibbons, M. and Hess, P. (1981), Day of he week effecs and asse reurns. Journal of Business, 1981, 54:4, 579-596. Hansen, L. (1982), Large sample properies of generalized mehod of momens esimaors. Economerica, 50, 1029 1054. Harris, L. (1986), A ransacion daa sudy of weekly and iner-daily paerns in sock reurns. Journal of Financial Economics, 15, 99-117. Jaffe, J. and Weserfield, R. (1985), The week-end effec in common sock reurns: he inernaional evidence. The Journal of Finance, 40:2, 433-454. 82 Pacific Rim Propery Research Journal, Vol 11, No 1
Keim, D. (1985), Dividend yields and sock reurns: implicaions of abnormal January reurns. Journal of Financial Economics, 14, 473 490. Keim, D. (1983), Size-relaed anomalies and sock reurn seasonaliy: furher empirical evidence. Journal of Financial Economics, 12, 13-32. Keim, D. and Sambaugh, R. (1984), A furher invesigaion of he weekend effec in sock reurns. The Journal of Finance, 39:3, 819-835. Liu, C. and Mei, J. (1992), The predicabiliy of reurns on equiy REITs and heir co-movemen wih oher asses. Journal of Real Esae Finance and Economics, 5:4, 401-418. Ogden, J. (1990), Turn-of-monh evaluaions of liquid profis and sock reurns: a common explanaion for he monhly and January effecs. The Journal of Finance, 45:4, 1259-1272. Peng, V. (2004a), Price predicabiliy and capial marke efficiency: lised propery rus invesmen sraegies. Pacific Rim Propery Research Journal, 10:4, 451-466. Peng, V. (2004b). Seleciviy, iming and he performance of lised propery russ: implicaions for invesmen sraegies. Pacific Rim Propery Research Journal, 10:2, 235-255. Redman, A., Manakyan, H. and Liano, K. (1997), Real esae invesmen russ and calendar anomalies. Journal of Real Esae Research, 14: 1, 19-28. Roll, R. (1983), The urn of he year effec and he reurn premium of small firms. Journal of Porfolio Managemen, 9, 18-28. Rozeff, M. and Kinney, W. (1976). Capial marke seasonaliy: he case of sock reurns. Journal of Financial Economics, 3, 379-402. Wang, K., Li, Y. and Erickon, J. (1997), A new look a he Monday effec. The Journal of Finance, 52:5, 2171-2186. Whie, H. (1980), A heeroscadasiciy-consisen covariance marix esimaor and a direc es for heeroscadasiciy. Economerica, 48, 817-838. Pacific Rim Propery Research Journal, Vol 11, No 1 83