ACADEMY OF ECONOMIC STUDIES DOCTORAL SCHOOL OF FINANCE AND BANKING DISSERTATION PAPER The day of the week effect on stock market return and volatility: International evidence Student: Sorin Stoica Supervisor: Professor Moisa BUCHAREST, JULY 2008
Jan 03, 2016
ACADEMY OF ECONOMIC STUDIESDOCTORAL SCHOOL OF FINANCE AND BANKING
DISSERTATION PAPER
The day of the week effect on stock market return and volatility:
International evidence
Student: Sorin Stoica Supervisor: Professor Moisa Altar
BUCHAREST, JULY 2008
Contents
1. Introduction
2. Literature review
3. Data and Model description
4. Empirical results
5. Conclusions
6. Bibliography
1. Introduction
A market is efficient if prices fully and instantaneously reflect all available information and no profit opportunities are left unexploited. In an efficient situation, new information is unpredictable, so stock market returns cannot be predicted and there is therefore no trading pattern, which an investor can follow in order to make unexpected profits.
*(The efficient-market hypothesis was developed by Professor Eugene Fama at the University of Chicago Graduate School of Business as an academic concept of study through his published Ph.D. thesis in the early 1960s at the same school)
The day of the week effect refers to the existence of a pattern of stock returns during the week, a seasonal «anomaly», which contradicts the «Efficient Market Hypothesis» *
2. Literature review
Cross (1973) and French (1980) were the first to observe a specific seasonality in stock returns during the week, that was named «Day of the Week Effect». According to this phenomenon, the average stock market return on the last trading day of the week (Friday) is positive and is the highest across all days of the week and the return on the first trading day of the week (Monday) is negative and is the lowest across the same period.
French (1987) examine the relationship between stock prices and volatility and report that unexpected stock market returns are negatively related to the unexpected changes in volatility. Campbell and Hentschel (1992) report similar results and argue that an increase in stock market volatility raises the required rate of return on common stocks and hence lowers stock prices. Glosten (1993) and Nelson (1991), on the other hand, report that positive unanticipated returns reduce conditional volatility whereas negative unanticipated returns increase conditional volatility.
Chen (2001) examine the day of the week effect in the stock markets of China for the recent years. The conclusion is consistent with the efficient market;
Kiymaz and Berument (2003) investigate the day of the week effect on the volatility and return of major stock markets (German, Japan, US, Canadaand United Kingdom) for the time period from 1998 to 2002. Their findings are consistent with the day of the week effect both for returns and volatility.
Patev (2003) examine the presence of the day-of-the-week effect anomaly in the Central European stock markets during the period 1997 to 2002. Their results indicated that the Czech and Romanian markets have significant negative Monday returns while the Slovenian market has significant positive Wednesday returns and has non-significant negative returns on Fridays. The Polish and Slovak markets have no day-of-the week effect anomaly.
Cabello and Ortiz (2004) investigate the day of the week and month of the yeareffect for Latin America stock markets. The paper supports the existence of calendar anomalies. They find the lowest and negative returns on Mondays and high returns on Fridays.
Hui (2005) examines the day of the week effect at Asian-Pacific markets during the period of Asian financial crisis and also tests the presence of weekend effect in developed stock markets of US and Japan. The paper supports no evidence of the day of the week effect in capital markets for the recent years, in both Asian Pacific and US capital markets.
3. Data and Model description3.1 Data
Period ROMANIABET 10
FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
Downtrend 1/10/2001-28/3/2003
1/10/2001-28/3/2003
1/10/2001-28/3/2003
1/10/2001-28/3/2003
1/10/2001-28/3/2003
1/10/2001-28/3/2003
(390) (390) (390) (390) (390) (390)
Uptrend 31/3/2003-20/6/2008
31/3/2003-20/6/2008
31/3/2003-20/6/2008
31/3/2003-20/6/2008
31/3/2003-20/6/2008
31/3/2003-20/6/2008
(1365) (1365) (1365) (1365) (1365) (1365)
The data set used in this study consists of six European Index values obtained from Bloomberg.For econometric reasons, for working days that the stock markets did not open and of course the indices did not change, the value of the previous day has been used.
The returns used in each of the time series are computed as follows:
day workingprevious in theindex theof valuethe:
index theof valuethe:
returnday the:
ln
1
1
t
t
t
t
tt
P
P
R
P
PR
Notes: Numbers in parentheses depict observations used in each period
0
5000
10000
15000
20000
25000
30000
35000
01:10 02:01 02:04 02:07 02:10 03:01
BETCACDAX
FTSEMADRIDMIBTEL
DOWNTREND PERIOD
0
10000
20000
30000
40000
50000
2003 2004 2005 2006 2007
BETCACDAX
FTSEMADRIDMIBTEL
UPTREND PERIOD
0
10000
20000
30000
40000
50000
2002 2003 2004 2005 2006 2007
BETCACDAX
FTSEMADRIDMIBTEL
WHOLE PERIOD
3.3 Model description
tR represents returns on a selected index
tttt FandHTM ,, are the dummy variables for Monday, Tuesday, Thursday, and Friday at time t
is a measure of the risk premium
tttt FandHTM ,, are the dummy variables for Monday, Tuesday, Thursday, and Friday at time t
12
t the GARCH term
12
t the ARCH term
The first GARCH-M (1, 1) model investigate the day of the week effect in stock return and it consists of the following two equations:
12
112
102
110
ttt
ttttFtHtTtMt RFHTMcR
tR represents returns on a selected index
tttt FandHTM ,, are the dummy variables for Monday, Tuesday, Thursday, and Friday at time t
is a measure of the risk premium
tttt FandHTM ,, are the dummy variables for Monday, Tuesday, Thursday, and Friday at time t
12
t the GARCH term
12
t the ARCH term
In both models the Wednesday dummy variable is excluded to avoid the dummy variable trap
The mean equation allows for an autoregression of order 1 in the mean of returns since most of the returns data exhibit a small but significant first order autocorrelation
0 the mean
tR represents returns on a selected index
tttt FandHTM ,, are the dummy variables for Monday, Tuesday, Thursday, and Friday at time t
is a measure of the risk premium
tttt FandHTM ,, are the dummy variables for Monday, Tuesday, Thursday, and Friday at time t
12
t the GARCH term
12
t the ARCH term
The second GARCH-M (1, 1) model investigate the day of the week effect in both stock return and volatility and it consists of the following two equations:
tR represents returns on a selected index
tttt FandHTM ,, are the dummy variables for Monday, Tuesday, Thursday, and Friday at time t
tttt FandHTM ,, are the dummy variables for Monday, Tuesday, Thursday, and Friday at time t
12
t the GARCH term
12
t the ARCH term
The quasi-maximum likelihood estimation (QMLE) method introduced by Bollerslev and Wooldridge (1992) is used to estimate parameters
The mean equation allows for an autoregression of order 1 in the mean of returns since most of the returns data exhibit a small but significant first order autocorrelation
12
112
102
110
tttFtHtTtMt
ttttFtHtTtMt
FHTM
RFHTMcR
0 the mean 0 the mean
Notes: p values are reported in brackets; ** denotes significance at the 1% level of significance
• The return series are nonsymmetric and leptokurtic compared to the normal distribution
• According to Augmented Dickey - Fuller test all return series are stationary
Whole period (Returns)
ROMANIABET 10
FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
Mean 0.001142 6.77E-05 0.000250 -4.66E-05 0.000311 3.15E-05
Median 0.000115 0.000102 0.000576 0.000000 0.000481 0.000309
Maximum 0.145016 0.070023 0.075527 0.068219 0.067222 0.064038
Minimum -0.119056 -0.070774 -0.074335 -0.059332 -0.078393 -0.053131
Std. Dev. 0.016653 0.013626 0.015120 0.011948 0.012338 0.011510
Skewness 0.144307 -0.008740 -0.036558 -0.102100 -0.038203 -0.082341
Kurtosis 10.33795 6.783804 6.536587 6.345886 6.627896 5.730291
Jarque-Bera 3941.297 1046.369 914.4765 821.2117 962.3224 546.7811
Observations 1754 1754 1754 1754 1754 1754
ADF (returns) -39.11451**[0]
-43.93095**[0]
-44.38366**[0]
-27.51158**[0]
-44.40952**[0]
-44.00585**[0]
4. Empirical results
4.1 Testing the series
Down trend(Returns)
ROMANIABET 10
FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
Mean 0.001547 -0.000982 -0.001337 -0.000911 -0.000454 -0.000608
Median 0.000000 -0.001053 -0.001204 -0.001179 0.000000 -0.000648
Maximum 0.145016 0.070023 0.075527 0.068219 0.056942 0.064038
Minimum -0.044481 -0.060448 -0.063360 -0.059332 -0.052006 -0.050102
Std. Dev. 0.016518 0.021481 0.024335 0.017495 0.018810 0.017979
Skewness 1.901538 0.250631 0.196448 0.144311 0.300625 0.217018
Kurtosis 17.59871 3.769944 3.393660 4.263171 3.185388 3.180879
Jarque-Bera 3688.783 13.68108 5.013823 27.21225 6.416413 3.583733
Observations 389 389 389 389 389 389
ADF Test -19.18971**[0]
-19.84382**[0]
-20.90961**[0]
-21.38023**[0]
-20.41756**[0]
-20.14965**[0]
Notes: p values are reported in brackets; ** denotes significance at the 1% level of significance
• The return series are nonsymmetric and leptokurtic compared to the normal distribution
• According to Augmented Dickey - Fuller test all return series are stationary
Notes: p values are reported in brackets; ** denotes significance at the 1% level of significance
• The return series are nonsymmetric and leptokurtic compared to the normal distribution
• According to Augmented Dickey - Fuller test all return series are stationary
Up trend(Returns)
ROMANIABET 10
FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
Mean 0.001027 0.000367 0.000703 0.000200 0.000529 0.000214
Median 0.000182 0.000449 0.000924 0.000253 0.000765 0.000786
Maximum 0.089371 0.058335 0.057610 0.044623 0.067222 0.038619
Minimum -0.119056 -0.070774 -0.074335 -0.056277 -0.078393 -0.053131
Std. Dev. 0.016695 0.010342 0.011155 0.009805 0.009735 0.008841
Skewness -0.338816 -0.376683 -0.239517 -0.306060 -0.532929 -0.558735
Kurtosis 8.304512 6.575000 6.731557 5.784750 9.571692 5.869344
Jarque-Bera 1626.456 759.1779 805.0084 462.3666 2520.881 539.2817
Observations 1365 1365 1365 1365 1365 1365
ADF (returns) -34.09442**[0]
-40.82585**[0]
-39.29837**[0]
-42.64936**[0]
-40.17416**[0]
-40.08223**[0]
Whole period
Mean MO TU WE TH FR F-stat Prob
BET 10 0.005918 0.005890 0.005843 0.005785 0.005734 0.001340 1.0000
CAC 40 0.000432 0.000421 0.000394 0.000257 0.000227 0.004023 1.0000
DAX 30 0.001320 0.001305 0.001190 0.001116 0.001093 0.003717 1.0000
FTSE 100 -0.000137 -0.000128 -0.000228 -0.000308 -0.000385 0.007723 0.9999
MADRID 0.001653 0.001620 0.001588 0.001492 0.001568 0.001945 1.0000
MIBTEL 0.000248 0.000228 0.000150 5.13E-05 7.60E-05 0.004465 1.0000
Whole period
Std. Dev. MO TU WE TH FR Levene Prob
BET 10 0.041494 0.039676 0.038668 0.035554 0.037045 1.063576 0.3730
CAC 40 0.030611 0.029600 0.030619 0.025393 0.025475 1.640912 0.1614
DAX 30 0.034507 0.033402 0.033373 0.029319 0.030010 1.478357 0.2062
FTSE 100 0.026093 0.023406 0.024967 0.021326 0.021358 1.860181 0.1149
MADRID 0.028230 0.025554 0.027220 0.023941 0.023496 0.974811 0.4201
MIBTEL 0.026436 0.025032 0.025754 0.022853 0.022933 0.725194 0.5747
The descriptive statistics for each day of the week
The F-Stat refers to the F-Statistic of the Equality of means test.If p-value < 0.050, then the hypothesis of equal means is rejected
The L-Value refers to the Levene Value of the Equality of variance test. If p-value < 0.050, then the hypothesis of equal variances is rejected
Down Trend
Mean MO TU WE TH FR F-stat Prob
BET 10 0.007724 0.007595 0.007658 0.007856 0.007857 0.000832 1.0000
CAC 40 -0.004991 -0.004796 -0.004768 -0.005616 -0.005471 0.005675 0.9999
DAX 30 -0.006612 -0.006367 -0.007044 -0.007342 -0.007490 0.000461 1.0000
FTSE 100 -0.004284 -0.004305 -0.004432 -0.005108 -0.005291 0.014669 0.9996
MADRID -0.002345 -0.002299 -0.002143 -0.002791 -0.002249 0.003187 1.0000
MIBTEL -0.003127 -0.003062 -0.003210 -0.003838 -0.003443 0.005131 0.9999
Down trend
Std. Dev. MO TU WE TH FR Levene Prob
BET 10 0.034672 0.036623 0.038090 0.033666 0.035559 0.105995 0.9804
CAC 40 0.047271 0.049090 0.051749 0.039945 0.038637 0.730156 0.5718
DAX 30 0.053515 0.053612 0.053847 0.045270 0.045831 0.934238 0.4431
FTSE 100 0.036625 0.036621 0.039259 0.030501 0.029112 0.615906 0.6514
MADRID 0.041747 0.038969 0.042352 0.036434 0.033663 1.010731 0.4017
MIBTEL 0.041197 0.040986 0.041212 0.034844 0.034272 0.668828 0.6140
The F-Stat refers to the F-Statistic of the Equality of means test.If p-value < 0.050, then the hypothesis of equal means is rejected
The L-Value refers to the Levene Value of the Equality of variance test. If p-value < 0.050, then the hypothesis of equal variances is rejected
Up trend
Mean MO TU WE TH FR F-stat Prob
BET 10 0.005409 0.005409 0.005332 0.005201 0.005135 0.002722 1.0000
CAC 40 0.001961 0.001892 0.001850 0.001914 0.001834 0.001591 1.0000
DAX 30 0.003557 0.003469 0.003512 0.003501 0.003514 0.006821 0.9999
FTSE 100 0.001032 0.001050 0.000958 0.001046 0.000999 0.001116 1.0000
MADRID 0.002780 0.002725 0.002641 0.002700 0.002645 0.002172 1.0000
MIBTEL 0.001201 0.001156 0.001098 0.001148 0.001068 0.002043 1.0000
Up trend
Mean MO TU WE TH FR Levene Prob
BET 10 0.043267 0.040546 0.038883 0.036107 0.037495 1.114507 0.3480
CAC 40 0.023817 0.021000 0.021095 0.019209 0.020094 1.912600 0.1059
DAX 30 0.026548 0.024640 0.024378 0.022458 0.023280 0.930642 0.4461
FTSE 100 0.022193 0.017942 0.019046 0.017755 0.018433 2.442196 0.0450
MADRID 0.023016 0.020199 0.021081 0.018918 0.019658 0.532459 0.7119
MIBTEL 0.020454 0.018185 0.019297 0.018037 0.018497 0.532061 0.7122
The F-Stat refers to the F-Statistic of the Equality of means test.If p-value < 0.050, then the hypothesis of equal means is rejected
The L-Value refers to the Levene Value of the Equality of variance test. If p-value < 0.050, then the hypothesis of equal variances is rejected
Whole Period
Return ecuation
ROMANIABET 10
FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
Constant -0,0003 0,000706 0,000878 0,000173 0,001168 0,001476*
(0.001708) (0.000869) (0.000926) (0.000765) (0.000821) (0.000725)
Monday -0,00031 -0,000594 -0,000242 0,000356 -0,00114 -0,001503*
(0.001108) (0.000724) (0.000823) (0.000633) (0.000729) (0.000676)
Tuesday 0,001601 -0,00078 -0,000862 0,000148 -0,00106 -0,001336*
(0.001138) (0.000708) (0.000767) (0.000668) (0.000694) (0.000613)
Thursday 3,06E-05 1,58E-04 -3,39E-05 7,46E-04 -1,83E-04 -8,49E-04
(0.001089) (0.00073) (0.000787) (0.000669) (0.000677) (0.000649)
Friday 6,61E-05 1,05E-04 -2,41E-04 9,46E-04 -2,99E-04 -6,52E-04
(0.001057) (0.000709) (0.000767) (0.000639) (0.000724) (0.000625)
Return(t-1) 5,53E-02 -6,60E-02** -4,93E-02* -8,99E-02** -3,58E-02 -5,85E-02**
(0.034143) (0.023936) (0.024818) (0.024974) (0.025588) (0.02407)
Risk 0,092859 0,015239 0,028418 -0,020148 0,023416 -0,016238
(0.113662) (0.076608) (0.071814) (0.073613) (0.074867) (0.074664)
4.2 The Results of the regressions
Notes: Standard errors are reported in parentheses; ** denotes significance at the 1% level of significance
The day of the week effects in returns for whole period
VolatilityROMANIA
BET 10FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
Mean 3,01E-05** 1,93E-06** 2,07E-06* 1,46E-06** 2,14E-06** 1,15E-06*
(0.0000111) (0.000000745) (0.000000964) (0.000000524) (0.000000831) (0.000000508)
ARCH 0,177732** 0,090911** 0,089462** 0,089707** 0,097743** 0,073798**
(0.057288) (0.019176) (0.021683) (0.014662) (0.024015) (0.016404)
GARCH 0,721278** 0,897015** 0,899845** 0,899368** 0,887776** 0,916383**
(0.068387) (0.018995) (0.022146) (0.014838) (0.023285) (0.017282)
Whole Period
Ljung–Box Q
statisticsROMANIA
BET 10FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
5 2,5467 8,3985 8,2389 6,1078 3,7531 5,7284
[0.769] [0.136] [0.144] [0.296] [0.585] [0.334]
10 7,2557 12,559 11,504 6,9764 6,6237 9,4085
[0.509] [0.128] [0.175] [0.539] [0.578] [0.309]
15 24,409 17,954 17,336 12,63 19,498 12,492
[0.058] [0.265] [0.299] [0.631] [0.192] [0.642]
20 31,865 24,681 22,088 20,86 23,096 17,725
[0.045] [0.214] [0.336] [0.405] [0.284] [0.606]
25 37,99 25,995 24,727 28,881 25,632 18,442
[0.046] [0.408] [0.478] [0.269] [0.427] [0.823]
Whole Period
ARCH-LM test
ROMANIABET 10
FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
5 0,382733 1,727216 1,129006 0,69928 4,075016 0,933572
[0.860877] [0.125118] [0.342759] [0.624008] [0.00111] [0.458033]
10 0,381393 0,927862 0,649215 0,850661 2,106105 0,714608
[0.955162] [0.5062] [0.772107] [0.579599] [0.021217] [0.711419]
15 0,319659 1,106179 0,835782 1,612083 1,900103 1,063579
[0.993643] [0.344869] [0.637904] [0.063357] [0.019338] [0.386045]
20 0,280678 0,937869 0,794883 1,348491 1,953253 0,978515
[0.999309] [0.537914] [0.722357] [0.138103] [0.007002] [0.48565]
25 0,764475 1,090896 0,796243 1,160234 1,746724 1,119203
[0.790675] [0.344074] [0.750802] [0.265791] [0.012622] [0.310641]
The conditional variances are always positive and are not explosive in our samples. According to the Ljung–Box Q statistics we can not reject the null hypothesis that the residuals are not autocorrelated. The ARCH-LM test does not indicate the presence of a significant ARCH effect in any of the sampled markets except MADRID.
Whole period
Return ecuation
ROMANIABET 10
FRANCECAC 40
GERMANY
DAX 30UK
FTSE 100SPAIN
MADRIDITALY
MIBTEL
Constant -0,00078 0,000767 0,000794 0,000224 0,001116 0,001341
(0.001663) (0.000862) (0.000962) (0.000761) (0.00084) (0.000714)
Monday -0,00034 -0,000613 -0,00025 0,00036 -0,00119 -0,00151*
(0.001088) (0.000721) (0.000843) (0.000632) (0.000745) (0.000683)
Tuesday 0,00139 -0,000804 -0,0009 0,000141 -0,00109 -0,00131*
(0.001079) (0.000709) (0.000786) (0.000665) (0.000712) (0.000617)
Thursday 2,38E-05 2,15E-04 -3,07E-05 7,93E-04 -2,94E-04 -8,24E-04
(0.00109) (0.00073) (0.000795) (0.000667) (0.000681) (0.000647)
Friday 2,48E-04 1,85E-04 -1,75E-04 9,84E-04 -2,73E-04 -6,65E-04
(0.001025) (0.000708) (0.000776) (0.000638) (0.000721) (0.000623)
Return(t-1) 6,06E-02 -6,63E-02** -5,00E-02*-8,89E-
02** -3,62E-02 -6,06E-02**
(0.034595) (0.023934) (0.024656) (0.024993) (0.025303) (0.023909)
Risk 0,122603 0,009604 0,034378 -0,025356 0,031732 -0,00411
(0.116125) (0.075484) (0.073363) (0.072933) (0.074814) (0.076502)
The day of the week effects in returns and volatilities for whole period
* Statistically significant at the 5% level.** Statistically significant at the 1% level.
Whole Period
VolatilityROMANIA
BET 10FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
Mean -9,87E-06 4,92E-06 1,32E-05 -2,95E-07 1,13E-05 1,42E-06
(0.0000562) (0.0000109) (0.0000141) (0.00000802) (0.000011) (0.0000076)
ARCH 0,210745** 0,093075** 0,087229** 0,089355** 0,094191** 0,072936**
(0.059474) (0.019896) (0.021338) (0.014525) (0.023343) (0.01512)
GARCH 0,662341** 0,896094** 0,901471** 0,900508** 0,892052** 0,915054**
(0.076557) (0.01932) (0.021912) (0.014675) (0.022591) (0.016479)
Monday 7,72E-05 -1,42E-05 -1,48E-05 -4,13E-06 -1,35E-05 1,25E-05
(0.000058) (0.0000147) (0.0000198) (0.0000111) (0.0000157) (0.0000125)
Tuesday 1,00E-04 -9,59E-08 -1,69E-05 8,89E-06 -1,42E-05 -1,51E-05
(0.0000969) (0.0000215) (0.0000284) (0.0000153) (0.0000217) (0.0000153)
Thursday 3,89E-05 9,30E-07 -1,73E-05 3,18E-06 -2,06E-05 5,68E-06
(0.0000778) (0.0000159) (0.0000201) (0.0000127) (0.0000159) (0.0000123)
Friday 2,48E-05 -2,07E-06 -6,38E-06 5,01E-07 2,03E-06 -3,61E-06
(0.0000577) (0.0000153) (0.0000182) (0.0000114) (0.0000158) (0.0000124)
The conditional variances are always positive and are not explosive in our samples.
Whole Period
Q statROMANIA
BET 10FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
5 3,0236 8,5064 8,7964 6,0607 4,0458 5,5976
[0.696] [0.13] [0.117] [0.3] [0.543] [0.347]
10 8,2793 12,759 12,084 7,024 6,8198 9,1098
[0.407] [0.12] [0.148] [0.534] [0.556] [0.333]
15 25,043 18,238 18,178 12,551 20,135 12,019
[0.049] [0.25] [0.253] [0.637] [0.167] [0.678]
20 31,952 25,255 22,881 20,821 23,846 17,078
[0.044] [0.192] [0.295] [0.408] [0.249] [0.648]
25 38,21 26,578 25,594 28,619 26,714 17,848
[0.044] [0.377] [0.429] [0.28] [0.37] [0.849]
The Ljung–Box Q statistics for the normalized residuals at 5-, 10-, 15-, 20-, and 25-day lags
None of these coefficients are statistically significant. Therefore, we cannot reject the null hypothesis that the residuals are not autocorrelated.
Whole Period
ARCH-LM test
ROMANIABET 10
FRANCECAC 40
GERMANYDAX 30
UK FTSE 100
SPAINMADRID
ITALYMIBTEL
5 0,420936 1,680777 1,201455 0,692022 4,505728 0,980701
[0.834402] [0.135968] [0.306014] [0.629501] [0.000439] [0.428102]
10 0,601501 0,88257 0,691385 0,911325 2,343972 0,759711
[0.813712] [0.548923] [0.73335] [0.521666] [0.009571] [0.668039]
15 0,505872 1,074607 0,877044 1,590531 2,057816 1,222953
[0.938819] [0.375135] [0.590322] [0.068863] [0.009606] [0.246415]
20 0,474225 0,920121 0,822288 1,341716 2,094818 1,113194
[0.976242] [0.561059] [0.688144] [0.141982] [0.003109] [0.327823]
25 0,799504 1,062295 0,804766 1,148274 1,863541 1,209985
[0.746558] [0.379769] [0.739655] [0.278409] [0.005968] [0.217401]
Engle’s ARCH-LM for whole period
Engle’s ARCH-LM test does not indicate the presence of a significant ARCH effect in any of the sampled markets except MADRID. This finding indicates that the standardized residual terms have constant variances and do not exhibit autocorrelation except MADRID.
The results for downtrend period
The day of the week effects in returns for downtrend period
There is no coefficient of dummy’s variables statistically significant. Thus, we don’t find the evidence for the existence of the classical day of the week effect.
The estimated coefficients for BET 10, MADRID and MIBTEL are lowest on Mondays but they are statistically insignificant. The coefficient of the conditional standard deviation of the return equation (risk) is positive for BET10 (0,347303), CAC 40 (0,115031), DAX 30 (0,058048), FTSE 100 (0,178102), MADRID (0,224529) and MIBTEL (0,218751). However, the estimated coefficients are not statistically significant.
The conditional variances are always positive and are not explosive in our samples. According to the Ljung–Box Q statistics we cannot reject the null hypothesis that the residuals are not autocorrelated. The ARCH-LM test does not indicate the presence of a significant ARCH effect in any of the sampled markets. This finding indicates that the standardized residual terms have constant variances and do not exhibit autocorrelation.
The day of the week effects in returns and volatilities for downtrend period
The estimated coefficients for dummy’s variables in volatility equation are not statistically significant except the ones from Monday and Tuesday for BET10, the one from Tuesday for DAX 40 and the one from Friday for FTSE 100 who are statistically significant.
Not only we don’t find strong evidence for the existence of the classical day of the week effect, but there is no any obvious pattern in coefficient’s significances.
The coefficients of the conditional standard deviation of the return equation (risk) are positive for all markets. However, the estimated coefficients are not statistically significant except BET10. The conditional variances are always positive and are not explosive in our samples According to the Ljung–Box Q statistics we cannot reject the null hypothesis that the residuals are not autocorrelated. The ARCH-LM test does not indicate the presence of a significant ARCH effect in any of the sampled markets
The results for uptrend period
The day of the week effects in returns for uptrend period
The estimated coefficient of the Mondays’ dummy variable for MIBTEL (-0,001503) is negative and statistically significant at the 1% level, suggesting that Mondays’ returns are smaller than those of Wednesdays. Also the estimated coefficient of the Tuesdays’ dummy variables for MIBTEL (-0,001316) is negative and statistically significant at the 1% level, suggesting that Tuesdays’ returns are smaller than those of Wednesdays. All the rest of dummy’s coefficients are not statistically significant.
The coefficient of the conditional standard deviation of the return equation (risk) is positive for CAC 40 (0,09187), DAX 30 (0,158252), MADRID (0,108172), MIBTEL (0,012795) and it is negative for BET10 (-0,08354), FTSE 100 (-0,005897), However, the estimated coefficients are not statistically significant.
There is no classical version of the day of the week effect and no substantial day effect for the developed stock markets.
The conditional variances are always positive and are not explosive in our samples. According to the Ljung–Box Q statistics we cannot reject the null hypothesis that the residuals are not autocorrelated. ARCH-LM test does not indicate the presence of a significant ARCH effect in any of the sampled markets except MADRID.
The estimated coefficients for dummy’s variables in volatility equation are not statistically significant. Thus, there is no evidence of a day of the week in volatility.
The coefficients of the conditional standard deviation of the return equation (risk) are positive for all markets except BET10 (-0,071027) and FTSE100 (-0,008442) who are negative. However, the estimated coefficients are not statistically significant.
The estimated coefficient of the Mondays’ dummy variable in the return equation for MIBTEL (-0,00143) is negative and statistically significant at the 1% level, suggesting that Mondays’ returns are smaller than those of Wednesdays. Also the estimated coefficient of the Tuesdays’ dummy variables in the return equation for MIBTEL (-0,00129) is negative and statistically significant at the 1% level, suggesting that Tuesdays’ returns are smaller than those of Wednesdays.
The conditional variances are always positive and are not explosive in our samples. According to the Ljung–Box Q statistics we cannot reject the null hypothesis that the residuals are not autocorrelated. ARCH-LM test does not indicate the presence of a significant ARCH effect in any of the sampled markets except MADRID.
The day of the week effects in returns and volatilities for uptrend period
5. The Conclusions
Finally, the conclusion of this study is that the phenomenon of the «Day of the Week Effect» seems to be weaker than it was in previous decades as a result of investor’s behavior. Investors are more mature, well educated, with more professional attitude, characteristics that help stock markets to become more efficient.
The phenomenon of the «Day of the Week Effect» seems to disappears from the developed stock markets and not to have a specific pattern in general.
Nowadays, the stock markets are more liquid than ever and seem to be more efficient that the previous decades because of the easiest capital transmission, the technological changes and the changes in the stock market microstructure. So, it is logical for investors to react more mature, something that induces less inefficient results.
1. Joshi, N.K., F.K.C. Bahadur (2006). “The Nepalese Stock Market: Efficiency and Calendar Anomalies”. Economic Review 17.2. Chiaku Chukwuogor-Ndu (2006).” Stock Market Returns Analysis, Day-of-the-Week Effect, Volatility of Returns: Evidence from European Financial Markets 1997-2004.” ISSN 1450-2887 International Research Journal of Finance and Economics3. Berument, H., & Kiymaz, H. (2001). The day of the week effect on stock market volatility. Journal of Economics and Finance, 25, 181–193.4. Halil Kiymaza, Hakan Berument “The day of the week effect on stock market volatility and volume: International evidence” Review of Financial Economics 12 (2003) 363–3805. Lyroudi, K., D. Subeniotis, G. Komisopoulos (2002). “Market Anomalies in the A.S.E.: The Day of the Week Effect”. European Financial Management Association, London Meetings.
6. Bibliography
6. Lakonishok, J., S. Smidt, (1988), “Are seasonal anomalies real? A ninety-year perspective”, Review of Financial Studies, 1, 403-25.7. Morton B. Brown and Alan B. Forsythe (1974), “Robust Tests for the Equality of Variances” Journal of the American Statistical Association, Vol. 69, No. 346, (Jun., 1974), pp. 364 -3678. Keim, B.D., R. F. Stambaugh, (1984), “A further investigation of the weekend effect in stock returns”, Journal of Finance, 39, 819-840.9. Dimitris Balios, Sophia Stavraki (2007), “Stock Market Trends, Day of the Week Effect and Investor’s Behavior after the September’s 2001 Attacks”, European Journal of Economics, Finance and Administrative Sciences ISSN 1450-2887 Issue 8 (2007)10. Green, William H. (2000), Econometric Analysis, New York University