The Week Effect In The Ftse350 Technology

Published Date: 02 Nov 2017

Rayda Alomar (120182701)

Sara Ghabra

Qianling He (120182446)

Priya Rajmane

Rachana Pramod Dhadiwal (120182457)

Contents

Abstract

This research paper examines the existence of the day-of-the-week effect among two indices from different sectors, technology and retail. Daily close prices are selected from 2004 until 2010using the Yahoo Finance UK. After the estimation of individual returns of each index on five dummy variables, findings suggest evidence of the existence of calendar effect in the FTSE350 Technology index but not for the FTSE350 Retail index. Significant positive Wednesday average returns are found in the FTSE350 Technology index, which provides evidence against the weak form of efficient market hypothesis. Whereas, the estimation’s findings for the FTSE350 Retail support the weak form of the efficient market hypothesis that historical information are reflected in the share price. Once the risk factor is incorporated within the model, part of the seasonality disappeared and the R2figure dramatically increased. Furthermore, when the constant risk assumption is relaxed using an interactive dummy variable, the seasonality effect almost disappeared but still exists to a minimal degree. However in this research paper, transaction costs are not accounted for due to time constraints. If transaction costs are included, the calendar effect within the FTSE350 technology index would be expected to disappear completely.

Introduction

Efficient market hypothesis is considered as a one of the most important concepts in finance. According to the theory of the efficient market hypothesis, it is impossible to "beat the market’’ as stock prices reflect all available information as soon as it is released. Owing to this the investors can’t predict the trends in the market by using security analysis techniques.

Prior to 1950, it was generally believed that one could beat the market by using either technical or fundamental analysis as they also believed that stock prices reflect the available information gradually and not instantly. But in 1953, when Maurice Kendall presented a paper on the behaviour of the stock and commodity prices to the royal statistical society, he found no particular pattern in the stock prices and the concept of "random walk" came into the light. After this, research started to provide the evidence of the behaviour of the stock prices and in 1970, Professor Eugene Fama, the father of Efficient Market Hypothesis (EMH) theory, came up with concept of EMH. He defines EHM as, "the current price [of an investment] should reflect all available information…so prices should change only based on unexpected new information", (Piper, 2013).

There are three forms of market efficient hypothesis:

First form of EMH is the weak form. When all the available historical information is reflected in the current share prices, it is known as the weak form of the EMH. In an efficient market it is impossible to constantly predict the future stock prices by using the past information as prices follow a random walk. To examine the weak form of EMH researchers employed statistical tests and measured profitability of the firms to find out patterns in the stock prices. After all the tests they found very few patterns in the weekly returns.

In the semi-strong form of EMH, share prices reflect not only the historical information but also all the publically available information. In an efficient market all the publically available information immediately reflects in the stock prices. Publically available information includes a new issue of the stock, announcement of the last quarter’s earnings, merger or acquisition of two companies etc. To examine this form of hypothesis, researchers analysed how instantly the market responds to the news like takeover news, dividend announcements and other macroeconomic information. After the research, they conclude that when a firm publishes any new information like earnings or dividends within 10 to 15 minutes of this news, share prices reflect or adjust to this information. In addition, when the company makes announcement of a takeover, the share prices increase rapidly in the expectation of the takeover’s premium.

Strong form of EMH states that prices reflect all the historical, public, and private information and if the markets are efficient it is impossible to predict the market. It is impossible to outperform consistently and earn the profits above average as stock prices reflect the information as soon as it is available. However, a number of researchers have shown that by using insider trading combined with financial ratios, one can earn more profits than average. Hence there is no substantial proof to support this form of the EMH.

The remaining part of the paper is segregated into sections talking about the literature review followed by the formulation of the Hypothesis for this study. The section thereafter focuses on the Data and Methodology employed for the study, followed by the discussion of Results obtained. The Conclusions and Limitations of this study are summarised in the last section of the paper.

Literature Review

The calendar effect can be defined as an observed systematic pattern in the stock market which the investors could use as a tool in order to obtain abnormal returns. In particular, day-of-the-week effect indicates that positive or negative stock returns can be observed on a particular day of the week in the stock market. Various studies have been conducted on the stock market indices of different countries all over the world, spanning different time periods; to study the day of the week anomaly. However, when the transaction costs and risk variations on different days-of-the-week are considered, the results may suggest evidence against the anomaly. Chris Brooks and Gita Persand (1999) in their paper, investigated whether the calendar effect, specifically the day-of-the-week-effect exists within the Southeast Asian Stock Markets during the 1990s. They closely examined five different Southeast Asian Stock Markets of South Korea, Malaysia, The Philippines, Taiwan, and Thailand. The paper extended the research further in considering the risk-return relationship using the CAPM framework, in order to explain the abnormal returns in a certain day-of the week with respect to the risk variations. They argue that a higher (lower) return observed on a certain day could be explained by the higher (lower) risk associated with the particular day. The study used three different equations to test the data. In the first equation, five intercept dummy variables were assigned for each day of the week. Then, in the second equation, the risk- return relationship using the CAPM framework was accounted for by incorporating the Market Risk Proxy from the World market index. The final equation included an interactive slope dummy variable multiplied by the Market Risk Proxy to relax the assumption of constant risk across all days of the week. The results were significant for only three out of five Southeast Asian Stock Markets; Thailand, Malaysia, and Taiwan and with the incorporation of the Market Risk Proxy within the equation, very little abnormal returns could be actually explained. However, with the relaxation of the assumption of constant variance for all days of the week, abnormal returns could be explained further. The study further suggested that some of the remaining calendar effects could be explained by other risk factors such as unexpected inflation, unexpected changes in the exchange rate, default risk premium, or release of new information on a particular day of the week.

A study conducted by Kontonikas A. (2004) examined the day of the week anomaly on the FTSE 100 index of UK for a period of 11 years spanning between 1986 and 1997. The main focus of the paper was to assess whether the day of the week effect still exists when transaction costs are accounted for. The transaction costs were accounted for by considering the bid-ask spreads and not the brokerage fees as shown in the previous literature reviews. He further extended his research to study the time variability property of the stock returns by using the GARCH model. The GARCH supported the view that transaction costs eventually let the day of the week effect fade away in the UK stock market. For the study, daily prices of the FTSE 100 Index were collected for the period between 1st January 1986 and 31st December 1997, totalling 3131 observations after excluding holidays. The data prior to 1986 could not be collected due to unavailability of prices and bid-ask spreads. The period studied included some major events like the deregulation of the London Stock Exchange (LSE) in 1986, the market crash of October 1987 & the Asian Financial Crisis of 1997. This inclusion indicated that if a day of the week effect exists during a time span covering some major economic events, then the day of the week effect is a strong irregularity in the stock market that becomes persistent over time. The data was tested using the conventional methodology similar to one of the methods employed in the study by Chris Brooks and Gita Persand (1999). The daily returns were regressed on five dummy variables, each dummy variable representing a day in the week to see whether the returns are different on each of the days of the week. The study confirmed that in the UK, the results for the day of the week effect were significant and stated that the stock returns were higher on the last day of the trading week but lowest on the first day of the week. The study indicated that Monday returns, as opposed to the other days’ returns, exhibit the lowest mean and highest standard deviation. The greater variance could be attributable to the higher trading volume due to the existence of the anomaly. The possible explanations which could be attributed to this phenomenon are the ‘market settlement procedures’, (Gibbons and Hess, 1981), ‘measurement errors in stock prices’, (Keim & Stambaugh, 1984) and the usual tendency of the firms to release unfavourable information after trading closes for the weekend. However, once the transaction costs have been accounted for, though the returns on Monday are negative, they are insignificant at the desired level, confirming that the day of the week effect tends to fade away when transaction costs are accounted for. These results are in line with the results of the study conducted in the previous paper. A prior study on the US stock markets also confirmed the existence of the day of the week effect; however the effect disappeared after 1975.

Unlike in the study conducted by Chris Brooks and Gita Persand (1999), Kontonikas (2004) used the GARCH model to study the time varying volatility in the stock returns. The conditional mean component of the model showed that the mean Monday returns were negative and significant and the conditional variance component showed that Mondays were associated with higher stock return conditional volatility. This clearly showed that the day of the week effect existed in the UK stock market. However, on taking into consideration the transaction costs, the anomaly tended to fade away. The study showed that the returns on Mondays were negative and significantly different from other days of the week, clearly indicative of the fact that stocks were cheaper to buy on Mondays. This led to the creation of the irregularity in the stock market, since the investors would prefer buying stocks on Mondays, thereby creating a day of the week effect. However, accounting for the transactions costs revealed that the returns on Mondays were not significantly different from the other days in the week. This clearly indicated that stocks were no longer cheaper to buy on Mondays as thought earlier, causing the day of the week anomaly to disappear. This further paved its way in support of the weak form efficient markets.

The studies discussed above mainly cover short time intervals to study the day of the week anomaly. However, Arsad and Coutts (1997) conducted a study on the security price anomalies in the London International Stock exchange, across a large time span of 60 years on the Financial Times Industrial Ordinary Shares Index. They found evidence supporting the stock market anomalies; the weekend effect, January effect and the holiday effects in UK. The study focused on analysing the security market regularities using UK data, in order to test and verify the existence of the three calendar anomalies in the London International Stock Exchange. The period studied covered most of the major economic events of the twentieth century, unlike a few covered in the study conducted by Kontonikas (2004). However, the ‘size-effect’ (Mills and Coutts, 1995) and lack of information regarding the payment of dividends might have introduced a bias in the study. The study made use of the regression equations involving dummy variables, just like the ones used in the studies mentioned above. The 60-year sample was divided into 12 sub-samples and a regression was run on each sub sample as well as the whole sample of 60 years. The results showed that the mean return for Mondays were all negative and significant for the whole sample and the 12 sub-samples individually. The results are very much consistent for the calendar anomalies with those obtained for other countries. The study showed that the various calendar anomalies did exist in the data set to some extent. Since most of the investors are aware of the existence of the weekend effect, they would try to profit from this irregularity in the stock market. However, the existence of the illiquidity in the markets, transaction costs, the offset between the profits earned from the irregularity and the costs associated with the same, would make any investment strategy unprofitable. Hence, the findings are in sync with the market efficiency hypothesis, since no strategy employed would result in offering consistent abnormal returns.

The studies mentioned above have been more specific in nature since they cover major indices in particular countries. However, a more global study was conducted by Amelie Charles (2007) which illustrated whether the removal of asymmetries had any substantial impact on the day of the week effect. The study also investigated the impact of day of the week anomaly, empirically, using GARCH family models, in some of the major international stock markets. The study was performed on five international indices; FTSE100 (UK), DAX30 (Germany), CAC40 (France), DJIA (US) and NIKKEI 225(Japan), reflecting the largest firms, spanning the period between 1987 and 2007. The result showed that all the indices had a positive average daily return with the exception of NIKKEI 225 (Japan) index, which exhibited negative returns. The results also focussed on the variance factor, with the highest variance in all the indices for Mondays and the lowest variances on Fridays for France, Germany and Japan alone. In line with the previous studies, Amelie Charles (2007) also focussed on the property of variance in the stock markets. Two different tests were applied to the tests the null hypothesis that the variance is constant across all the days of the week. The Brown–Forsythe test, rejected the null hypothesis for all the time series except for the FTSE 100 index and showed that the series were non-normal. Another test; the Kruskal–Wallis test was applied to examine the existence of the day-of-the- week effect in the world’s developed equity markets. This test also rejected the null hypothesis that the mean is constant over the week for all series except for the DJIA index. The study also made use of the GARCH and GJR- GRACH models to study the effect of asymmetries on the day of the week anomaly. The research was further extended in checking the use of seasonal effects for forecasting purposes. Five forecasting models; ARMA–GARCH without dummies, ARMA–GJR–GARCH without dummies, ARMA–APARCH without dummies, ARMA–GJR–GARCH with dummies and ARMA–APARCH with dummies were used in the study. Each of these models was used to estimate each series in the ‘in-sample’ estimation period. The models were used to obtain the ‘one-step’ forecasts of the conditional variance. The results clearly illustrated the impact of the chosen model on the volatility factor of the day of the week anomaly. The seasonal effects remain unaffected by the asymmetry factor, implying that the day of the week effects on volatility fails to improve the volatility forecast.

The literature review covered comprises of studies conducted on some of the largest stock exchanges in the world to test the day-of-the-week anomaly. The studies covered focused on the effect of different factors like volatility, risk variations and transaction costs on the irregularities of the stock markets. The studies find out whether the anomaly persists in the presence of the factors mentioned or the anomaly tends to die away when these factors are taken into consideration. One of the study shows that when risk variations are taken into account the abnormal returns which could not be explained otherwise, can now be explained in greater details. The variations in risk were explained with the help of different models like the CAPM & GARCH model. However, other risk factors like the unexpected change in inflation rates, exchange rates, release of new information could also be responsible for the calendar anomalies to some extent. Another study which focussed on transaction costs showed that inclusion of transaction costs led the day-of-the-week anomaly to fade away. A global study conducted on some of the largest stock indices in the world focussed on the impact of asymmetries on the day of the week anomaly. The study made use of GARCH family models and the Brown–Forsythe & the Kruskal- Wallis tests to examine the effect of asymmetries on the anomaly. The study also focussed on the use of GARCH family models in forecasting the conditional variance in the stock markets. It clearly indicated that the choice of the model had an effect on the volatility factor of the day of the week anomaly. All these are indicative of the fact that risk plays a major part in the existence of the calendar effect. However, other risk factors and independents variables should be accounted for in the model, to explain the remaining seasonality found in the Stock Markets.

Hypothesis

After analysing past studies, two indices, FTSE350 Technology and FTSE350 Retail, are chosen to examine the day-of-the-week effect. The reasons behind choosing these two particular indices are that: each index is specialized in a different sector, and each index has different level of risk. Since each index carries different level of risk, one might expect that different positive or negative returns could be observed on different days to a different degree. In general, the technology sector tends to be more risky in nature; hence higher beta coefficients may be observed across trading days within the FTSE350 Technology index rather than the FTSE350 Retail index.

Hypothesis I: The day-of-the-week effect is more likely to exist within the FTSE350 Technology index rather than the FTSE350 Retail index due to the higher level of risk found in the technology sector.

Data and Methodology

Daily closed prices are selected from September 22, 2004 to December 31, 2010 for both indices using the Yahoo Finance UK website. The total number of observations is 1552; which excludes weekends as well as public events such as bank holidays, Easter, and Christmas. The selected period covers pre and post the US financial crisis; since risk could be considered as a critical factor. The returns on the indices are calculated using the logarithms of index prices. The methodology of Brooks and Persand (1999) is used to conduct this research; which consists of three equations.

Equation 1:

It consists of five slope dummy variables for each trading day of the week: Monday, Tuesday, Wednesday, Thursday and Friday. The constant term is excluded to prevent the dummy variable trap. Individual returns for each index are regressed against five dummy variables for each trading day using Eviews statistical software.

Equation 2:

It extends Equation 1 further to incorporate the risk factor using the CAPM- type framework, in order to explain the calendar effect found in FTSE350 Technology index earlier. The higher (lower) average returns can be explained by the higher (lower) level of risk; which is a fundamental concept in finance. The market risk is calculated using the return on the S&P500 index, since S&P500 is one of the largest indices in the world.

Equation3:

In this equation an interactive dummy variable is created in order to allow the risk to vary across trading days; which will be a better way to explain the positive significant Wednesday average returns found in the FTSE350 Technology index. The higher (lower) average returns which are found on a certain day are dependent on the level of risk on that particular day.

Results

Table1.1 and Table1.2 highlight the results of the estimation for both FTSE350 Technology and FTSE350 Retail indices using Equation 1. According to the results found in Table1.1, FTSE350 Technology index has significant positive Wednesday average returns; indicating that the index has significant calendar effect. The coefficient of Wednesday average returns in the FTSE350 Technology is 0.001843 (t-ratio 2.06) and is significant at the 5% level. The R2 figure is at a low value of 0.27%; indicating that the five dummy variables are able to explain very little of the variation in the model. Further, independent variables should be added to the model in order to explain more of the variation. However, the results in Table1.2 indicate that the FTSE350 Retail index does not have any significant calendar effect. Therefore, the null hypothesis in this case cannot be rejected and that there are no signs of seasonality found in the FTSE350 Retail index. The findings tend to support the weak form of efficient market hypothesis that historical data are reflected in the share price for the FTSE350 Retail index. In addition, there are no systematic patterns found in the FTSE350 Retail index which investors could use to obtain abnormal returns. Since that is the case, the next two equations will mainly focus on the FTSE350 Technology index.

Furthermore, Table2.1 highlights the results for the FTSE350 Technology index after incorporating the return on S&P500 as the market risk. The findings suggest that the significant positive Wednesday average returns still exist even after the addition of another independent variable to the model. The coefficient for Wednesday average returns decreased by -13.29% to 0.001598(t-ratio 2.01) and still significant at the 5% level. However, the R2 figure estimated using Equation 2 increased dramatically from 0.27% to 21.35%; which is an overall increase of 79.07%. The high R2 figure can be explained by the highly significant coefficient for S&P500 average returns which stands at 0.509837 (t-ratio 20.37). In comparison with the FTSE350 Retail index, the results in Table2.2 also indicate highly significant coefficient for S&P500 average returns but much lower than the coefficient found in the FTSE350 Technology index. This implies that the FTSE350 Retail index holds lower level of risk.

Moreover, Table3.1 represents the results for the FTSE350 Technology index when the assumption for constant risk is relaxed by creating an interactive dummy variable allowing the risk to vary across the trading days. Results suggest that the significant calendar effect almost disappeared within the index. The coefficient for Wednesday average returns decreased slightly to 0.001560 (t-ratio 1.98) but still significant at the 5% level. Furthermore, the R2 figure increased slightly by 5.10% and stood at 22.46%. In addition, the beta coefficients for all trading days are highly significant at the 5% level. The highest betas are observed on Monday, Wednesday and Friday at 62.67%, 58.82%, and 58.85% respectively; whereas, the lowest beta is observed on Thursday at 32.07%. The high risk found on Wednesday can explain the significant positive Wednesday average returns. Whereas, it is not the case for Monday or Friday, which might be due to the fact that a lot of observations for Mondays and Fridays are missing, due to public events, specifically bank holidays and Easter. In comparison with the FTSE350 Retail index, the results in Table3.2 show that overall the beta coefficients are also highly significant but lower than the beta coefficients found in the FTSE350 Technology index. The findings suggest that the FTSE350 Technology index is riskier than the FTSE350 Retail index.

Conclusions

The analysis of the literature review helped in the formation of the research question covered in the study. The main focus of the study was to examine the existence of seasonality within different sectors; Technology and Retail. Hence two indices; the FTSE350 Technology and the FTSE350 Retail were selected owing to the different functionalities the sectors operate under, involving different levels of risk. While the literature review covered time spans within the 19th century, this study covered more relevant time periods (2004-2010). Also the earlier literatures were conducted mainly on some of the largest stock indices around the world, while in this study the size of the index was not the main focus. The finding in the literature review suggested that the seasonality existed across most of the trading days within a week. However this study’s findings showed that the seasonality was only significantly positive for the Wednesday average returns in the FTSE350 Technology index. The R2 value estimated by the model stood at a value as low as 0.27%. The five dummy variables failed to explain the variation in the model used and hence more independent variables should be added to explain the rest. However, by adding the risk factor using the S&P500 average returns, the coefficient for the average returns and the t ratio reduced slightly and the R2 value increased dramatically from 0.27% to 21.35%. This is indicative of the fact that the risk factor plays an important role in explaining the variation within the model. Subsequently, interactive dummy variables were created to relax the assumption of constant risk across all the trading days in the week. This led to a further decrease in the coefficient value and the t-ratios for the positive Wednesday average returns, but still significant at the 5% level. This provides evidence against the weak form of the Efficient Market Hypothesis; that the historical prices are reflected in the share prices. Whereas, the results for the FTSE350 Retail showed no trace of seasonality, favouring the weak form Efficient Market Hypothesis.

An important factor that the study did not take into account was the transaction costs. Literature review suggests that accounting for transaction costs and risks caused the seasonality to gradually fade away.

Limitations

The group faced quite a few limitations when performing this study. Firstly, there is always a scope for human error while preparing dummy variables especially in a large sample. Secondly, couple of missing trading days are observed due to public events such as Christmas, Easter and bank holidays. These days are automatically excluded, because returns are zero on the respective days; hence, the returns would not account for any dummy variable. Thirdly, to account for the risk factor, the US stock index (S&P500) is used as the world index. Thus, the trading days for both S&P500 and the FTSE350 Retail and the FTSE350 Technology indices did not match to some extent; hence observations are added to the S&P500 index with a return of zero in order to fix the mismatch. Fourthly, the sample size is limited to eight years owing to the limited resources such as data and time constraints. Lastly, the transaction costs could not be accounted for, due to the unavailability of the bid-ask spread data.

Appendix

Eviews results

Equation 1:

Table1.1 FTSE350 Technology index

Dependent Variable: RETURN_TECHNOLOGY

Method: Least Squares

Date: 04/15/13 Time: 15:10

Sample: 9/22/2004 12/31/2010

Included observations: 1552

Variable

Coefficient

Std. Error

t-Statistic

Prob.  

MONDAY

-0.000130

0.000935

-0.138636

0.8898

TUESDAY

-0.000384

0.000909

-0.421900

0.6732

WEDNESDAY

0.001843

0.000893

2.062461

0.0393

THURSDAY

-0.000192

0.000905

-0.212473

0.8318

FRIDAY

0.000602

0.000909

0.661749

0.5082

R-squared

0.002669

    Mean dependent var

0.000363

Adjusted R-squared

0.000090

    S.D. dependent var

0.016032

S.E. of regression

0.016031

    Akaike info criterion

-5.425365

Sum squared resid

0.397570

    Schwarz criterion

-5.408137

Log likelihood

4215.083

    Hannan-Quinn criter.

-5.418958

Durbin-Watson stat

2.012830

Table1.2_FTSE350_Retail_index

Dependent Variable: RETURN_RETAIL

Method: Least Squares

Date: 04/15/13 Time: 15:12

Sample: 9/22/2004 12/31/2010

Included observations: 1552

Variable

Coefficient

Std. Error

t-Statistic

Prob.  

MONDAY_R

-0.000276

0.000778

-0.354567

0.7230

TUESDAY_R

0.000521

0.000757

0.689016

0.4909

WEDNESDAY_R

-0.000275

0.000744

-0.370037

0.7114

THURSDAY_R

-0.000220

0.000753

-0.292048

0.7703

FRIDAY_R

0.000698

0.000757

0.922509

0.3564

R-squared

0.001035

    Mean dependent var

9.05E-05

Adjusted R-squared

-0.001548

    S.D. dependent var

0.013335

S.E. of regression

0.013345

    Akaike info criterion

-5.792153

Sum squared resid

0.275498

    Schwarz criterion

-5.774926

Log likelihood

4499.711

    Hannan-Quinn criter.

-5.785746

Durbin-Watson stat

2.091544

Step 2

Equation 2:

RME= Return on S&P500

Table2.1 FTSE350 Technology and S&P500 indices

Dependent Variable: RETURN_TECHNOLOGY

Method: Least Squares

Date: 04/15/13 Time: 16:09

Sample: 2 1554

Included observations: 1552

Variable

Coefficient

Std. Error

t-Statistic

Prob.  

RETURN_S_P500

0.509837

0.025029

20.36980

0.0000

MONDAY

-0.000138

0.000830

-0.166370

0.8679

TUESDAY

-0.000615

0.000807

-0.761227

0.4466

WEDNESDAY

0.001598

0.000794

2.013424

0.0442

THURSDAY

-0.000108

0.000804

-0.133910

0.8935

FRIDAY

0.000834

0.000807

1.033173

0.3017

R-squared

0.213703

    Mean dependent var

0.000363

Adjusted R-squared

0.211160

    S.D. dependent var

0.016032

S.E. of regression

0.014239

    Akaike info criterion

-5.661823

Sum squared resid

0.313445

    Schwarz criterion

-5.641151

Log likelihood

4399.575

    Hannan-Quinn criter.

-5.654135

Durbin-Watson stat

2.354389

Table2.2 FTSE350 Retail and S&P500 indices

Dependent Variable: RETURN_RETAIL

Method: Least Squares

Date: 04/15/13 Time: 16:11

Sample: 2 1554

Included observations: 1552

Variable

Coefficient

Std. Error

t-Statistic

Prob.  

RETURN_S_P500

0.374268

0.021448

17.45019

0.0000

MONDAY_R

-0.000282

0.000712

-0.396603

0.6917

TUESDAY_R

0.000352

0.000692

0.508266

0.6113

WEDNESDAY_R

-0.000455

0.000680

-0.668828

0.5037

THURSDAY_R

-0.000158

0.000689

-0.229201

0.8187

FRIDAY_R

0.000869

0.000692

1.255752

0.2094

R-squared

0.165419

    Mean dependent var

9.05E-05

Adjusted R-squared

0.162719

    S.D. dependent var

0.013335

S.E. of regression

0.012202

    Akaike info criterion

-5.970654

Sum squared resid

0.230164

    Schwarz criterion

-5.949982

Log likelihood

4639.228

    Hannan-Quinn criter.

-5.962966

Durbin-Watson stat

2.294386

Equation 3:

Table3.1 FTSE350 Technology index with an interactive dummy variable

Dependent Variable: RETURN_TECHNOLOGY

Method: Least Squares

Date: 04/15/13 Time: 16:12

Sample: 2 1554

Included observations: 1552

Variable

Coefficient

Std. Error

t-Statistic

Prob.  

MONDAY

-0.000140

0.000826

-0.169691

0.8653

TUESDAY

-0.000589

0.000803

-0.732828

0.4638

WEDNESDAY

0.001560

0.000789

1.976268

0.0483

THURSDAY

-0.000139

0.000799

-0.173954

0.8619

FRIDAY

0.000870

0.000803

1.083063

0.2789

RETURN_S_P500*MONDAY

0.626701

0.050949

12.30048

0.0000

RETURN_S_P500*TUESDAY

0.452269

0.051309

8.814606

0.0000

RETURN_S_P500*WEDNESDAY

0.588231

0.056999

10.32006

0.0000

RETURN_S_P500*THURSDAY

0.320695

0.054580

5.875630

0.0000

RETURN_S_P500*FRIDAY

0.588523

0.069674

8.446817

0.0000

R-squared

0.224612

    Mean dependent var

0.000363

Adjusted R-squared

0.220087

    S.D. dependent var

0.016032

S.E. of regression

0.014158

    Akaike info criterion

-5.670641

Sum squared resid

0.309096

    Schwarz criterion

-5.636186

Log likelihood

4410.417

    Hannan-Quinn criter.

-5.657827

Durbin-Watson stat

2.340105

Table3.2 FTSE350 Retail index with an interactive dummy variable

Dependent Variable: RETURN_RETAIL

Method: Least Squares

Date: 04/18/13 Time: 16:06

Sample: 2 1554

Included observations: 1552

Variable

Coefficient

Std. Error

t-Statistic

Prob.  

MONDAY_R

-0.000284

0.000710

-0.400596

0.6888

TUESDAY_R

0.000365

0.000690

0.529055

0.5968

WEDNESDAY_R

-0.000436

0.000678

-0.642147

0.5209

THURSDAY_R

-0.000172

0.000687

-0.250707

0.8021

FRIDAY_R

0.000881

0.000691

1.275382

0.2024

RETURN_S_P500*MONDAY_R

0.496994

0.043788

11.35012

0.0000

RETURN_S_P500*TUESDAY_R

0.344507

0.044097

7.812539

0.0000

RETURN_S_P500*WEDNESDAY_R

0.334402

0.048987

6.826386

0.0000

RETURN_S_P500*THURSDAY_R

0.287856

0.046908

6.136567

0.0000

RETURN_S_P500*FRIDAY_R

0.400019

0.059880

6.680350

0.0000

R-squared

0.172157

    Mean dependent var

9.05E-05

Adjusted R-squared

0.167325

    S.D. dependent var

0.013335

S.E. of regression

0.012168

    Akaike info criterion

-5.973607

Sum squared resid

0.228306

    Schwarz criterion

-5.939152

Log likelihood

4645.519

    Hannan-Quinn criter.

-5.960793

Durbin-Watson stat

2.256909