Printer Friendly

Calendar anomalies: evidence from the Colombo Stock Exchange.

The primary objective of this study is to examine the possible presence of widely documented Calendar Anomalies in the securities markets of Sri Lanka as any evidence of such presence may suggest beneficial trading strategies and provide insights as to the form of the market efficiency. This paper further intends to examine the robustness of the previous finding by Elyasiani et al. (ibid.) whose study provides evidence in support of the turn-of the-year effect in the CSE. The invalidation of the finding by Elyasiani et al. (ibid.) reinforces the finding by Abeysekara (2001) that 'substantial autocorrelation' that is present in the CSE returns. In turn, it implies the importance of incorporating specific autocorrelation schemes on the CSE. This study further reveals that the assertion of Meneu and Pardo (2004) stressing the irrelevancy of incorporating any specific autocorrelation scheme to the classical regression model is seemingly not appropriate to the Sri Lankan scenario.

The structure of the paper is as follows: The following section discusses the existing literature. The second section discusses the empirical evidence of anomalies from the CSE. The Section three contains the data and methodology whilst Section four provides the statistical analysis of the results. The last Section draws some conclusions and suggests directions for future research.

1. Existing Literature

Over the past two decades a number of calendar anomalies have been discovered in stock returns in markets ranging from developed through emerging to less developed markets. (1) Some of the well admitted anomalies include the month-of-the-year effect (Aggarwal and Rivoli, 1989; Arsad and Coutts, 1997; Kuan and Tat, 1998; Gao and Kling, 2005), the day-of-the-week effect (Balaban, 1995; Lakonishok and Smidt,1998; Tan and Tat, 1998; Ian and Chen, 2004; Gao and King, 2005), turn-of-the-year, and turn-of-the financial year effects (Elyasiani et al., 1996; D'Mello et al., 2003), and turn-of-the-month-effect (Lakonishok and Smidt, 1988; Boido and Fasano, 2005). These anomalies in stock returns are of particular interest because their manifestation leads to the denial of the efficient market

hypothesis (EMH) which asserts the fact that security prices reflect all available information and hence there are no opportunities for abnormal returns. The research findings documenting the presence of anomalies in fact make these anomalies disappear and contribute to create more efficient markets.

2. Empirical Evidence on the CSE

Even though a number of empirical studies investigating comprehensively the phenomenon of the EMH pertaining to CSE exist, only two studies as to calendar anomalies can be unveiled hitherto. These two studies have also been carried out as sub sections of the studies investigating the broader area of the EMH as any finding as to the existence of seasonalities enables the researchers to deny the weak form of the market efficiency.

The study by Elyasiani et al. (1996) examines the presence of a number of seasonal effects documented in the literature of calendar anomalies. The sample they select includes 8 years from 1986 to 1993 and utilizes both the SI and the ASPI, the two main price indices of the CSE. (2) They predominantly make use of non parametric univariate procedures as such procedures are free from restrictive distributional assumptions. Simultaneously the results are further endorsed by the use of the multivariate regression analysis.

The subsequent study by Abeysekara (2001) (3) investigates the day-of-the week effect and the month-of-the-year effect commencing a period from January 1991 to November 1996. The SI over ASPI is considered for the sample as close correlation between the two indices and thin trading hypothesis experienced by most stocks in the ASPI may hinder the accuracy of the results. He constructs his own time series using 40 stocks (4) and employs multivariate regression analysis.

The results of both studies, except the turn-of-the-year effect none of the other effects are not supported by the empirical findings of the study by Elyasiani et al. (1996), whereas the study by Abeysekara (2001) concludes that neither day-of-the-week effect nor turn-of-the-year is present on the CSE. A swift glance at the outcome may let somebody to contemplate that the conclusions seems contentious as one study acknowledges the existence of the January effect while other does not. This can for the most part be attributable to the discrepancies' in their definitions. Elyasiani et al. (1996, 69) consider the last five days of a certain calendar year and the first five days of the calendar year as the turn-of-the-year and employ the multivariate regression model. On the other hand, Abeysekara (2001) assigns the January effect to the phenomenon that January stock returns are, on average, higher than other months. (5) He employs the most frequently used model (Pandey, n.d.; Cheung and Coutts, 1999; Kuan and Tat, 1998; Arsad and Coutts, 1997; Andrew, Kaplandis and Jennifer, 2000; Lei and Gerald, 2005) to test the monthly seasonality.

As mentioned above, Elyasiani et al. (1996) admit the existence of the turn-of-the-year anomaly while denying the turn-of-the-financial-year effect on the CSE. This finding is somewhat interesting as it leads to a notable conclusion which asserts the fact that 'since the turn-of-the-financial year effect is not found at the CSE it is not reasonable to associate tax-loss selling hypothesis to the turn of the year effect' (Elyasiani et al., 1996, 70).6 This is a situation which goes against the most popular explanation of tax loss selling hypothesis. However, Elyasiani at al. (1996) present a number of other possible causes for the turn-of-the year effect. One such explanation is the habits of the institutional investors who influence the flow of funds in and out of the market. Another explanation can be attributable to the popular explanation of window dressing in which case portfolio managers clean up their portfolios before their performance evaluation deadlines to avoid any irksome positions. Moreover they elaborate the reasons for the meager extant of seasonality in the Sri Lankan market. Institutional features such as low volume and inadequate trading prevalent in the country, lack of strong connection between Sri Lanka and the markets in the industrialized world which prevents the transmission of calendar anomalies to Sri Lanka are some of the such reasons presented.

Finally it is noteworthy that both studies demonstrate some inadequacies as to their models. Abeysekara (2001) utilizes the basic multivariate regression model disregarding the fact that it is based on some restrictive assumptions. Although his study lucidly specifies the presence of 'substantial autocorrelation' (2001, 253), no attempt is made to modify the basic regression model. This statement is common to Elyasiani et al. (1996) as well. However the application of the non parametric procedures with no restrictive distributional assumptions by Elyasiani et al. (1996) should be acclaimed. Moreover the use of a period of 6 years by Abeysekara (2001) for a study involving time series data cannot be acquiesced. All these inadequacies on one hand remind the fact that 'theoretical models have yet to be developed to explain the presence and the absence of these anomalies in any part of the world' (Elyasiani et al., 1996, 75).

3. Data and Methodology

The data set consists of daily closing values of the ASPI and the SI of the CSE. (7) The period under examination spans over the period 01 January 1985 through 31 December 2005. The entire sample is divided further into two sub samples of approximately 10 years (1985-1994 and 1995-2005) as such enables to test the presence of any possible effect over short periods of time and most importantly will indicate whether the detected effect is persistent. Besides, partitioned data sets are imperative in examining the robustness of the previous findings on the CSE.

3.1 Index Returns

Following the standard methodology, the daily return is calculated as the logarithmic change in the value of the index i compared to previous day's closing value.

[R.sub.it] = ln([P.sub.t]/[P.sub.t-1]) (1)

Where [R.sub.it] is the change in index i on day t, [P.sub.t] represents the last value of index i on day t, and [P.sub.t-1] represents the last value of index i on day t-1.

It should also be noted that this study utilizes 'non-dividend adjusted' stock returns. Vast majority of previous studies have not captured the effect of dividends in their analysis as they argue that the exclusion of dividends does not necessarily invalidate the results of a study. (8)

3.2 Adjustment to Control the Influence of Other Major Events

Over the past two decades it has become customarily for public media including CSE reports to announce abnormal stock price movements during the periods of presidential and general elections as well as during other periods where the country was stumbling upon several snags caused by the conflict in the north and external crisis situations. In order to isolate the true anomalous behavior, the influence of other major events on stock returns needs to be eliminated. Hence an independent study is carried out by utilizing a list including major events that are alleged to have had triggered a momentous upshot in the CSE performance. (9) As detailed in the Annexure these major events are categorized into 3 main groups namely, elections, other local events (examples, assassination of political leaders, criminal acts by terrorist groups and milestones in the peace process.) and international crisis situations. In order to examine their impact on the stock market performance the following regression model is estimated.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)

where [R.sub.it] is the change in daily value of index i in period t and [D.sub.elec(Tn-1 &Tn+1]), [D.sub.local] and [D.sub.crises] represents dummy variables for Elections (with 1 for combinations of n- & n+ where n ranges from 1 to 5 and the day of the election if is a trading day is included in the [T.sub.n-] period), other local events and international crisis situations respectively. [R.sub.t-I] represents a lag variable and et stands for the random error term for the day t.

3.3 Adjustment for the Presence of Autocorrelation

With the view of detecting any anomalous behavior in the stock returns of the CSE, this study basically employs the multivariate regression model. Zero autocorrelation is one of the most problematic restrictive assumptions innate to this classical multivariate regression models whose violation may lead to spurious results. (10) The stock returns in most cases tend to violate the restrictive assumption of Zero Autocorrelation. Hence the classical model urges an adjustment to overcome this limitation. With a view of overcoming the problem of autocorrelation, following Kamas and Berument (2003), and Balbina and Martins (2002) lagged values of the return variable are embedded into the classical model.

Subsequent to this adjustment, the Durbin-Watson test for detecting autocorrelation is used to test whether the autocorrelation is still present. As a rule of thumb if D-W statistic (the first order autocorrelation coefficient) is found to be 2 in an application one may assume that there is no first-order autocorrelation. Hence adjustment for lag values is continued until the D-W statistic settles around 2. The regression models applied in connection with these two adjustments provide significant contribution to the available literature.

3.4 Day-of-the-Week Effect

Day-of-the week refers to the phenomenon that returns on Mondays are on average largely negative and those on Fridays are largely positive. In the attempt to find the evidence for the existence of the day-of-the-week effect the following standard model with adjustments for autocorrelation is estimated.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)

Regression coefficients for [D.sub.M] through [D.sub.F] are the mean returns for Monday through Friday which are set equal to 1 for the trading day on which the return is observed and Zero otherwise

3.5 Month-of-the-Year Anomaly (January Effect)

January effect refers to the phenomenon that 'January stock returns are, on average, higher than other months' (Kuan and Tat 1998, p. 457). The following conventional regression model with an adjustment for autocorrelation is applied in this paper with the view of testing whether there are any statistically significant differences between mean index returns for the different months of the year.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4)

Regression coefficients for [D.sub.Jan] to [D.sub.Dec] are the mean returns for January through December respectively.

3.6 Turn-of-the-Year and Turn-of-the-Financial-Year Effects

Previous findings in the literature have publicized the fact that the returns in the period between the last few trading days of a tax year and the first few days of the following month are significantly higher than the remaining period.

In most countries, the year-end month, December is identical to the tax-year end. Based on that most of the previous studies have reached the conclusion that the presence of the turn-of-the-year anomaly is consistent with the taxloss selling hypothesis. However, in Sri Lanka, the financial/tax-year end is different from the month of December where the financial year begins on April, 01 and ends on March, 31. Hence in order to comprehend the popular tax-loss selling hypothesis in the Sri Lankan context, it is imperative to examine the presence of this anomaly as to both the turn-of-the year and the turn-of-the-financial year periods. At the same time it is worth mentioning that different researchers from different countries have delineated the term "few trading days" in numerous ways. (11) But it is worthy to follow the definition by Elyasiani et al. (1996) as it is an objective of this study to examine the robustness of the previous finding by Elyasiani et al. (ibid.) whose study is the only comprehensive study carried out so far in the Sri Lankan context. Hence for this study following Elyasiani et al. (ibid., 69), 'the last five days of a certain calendar year (financial year) and the first five days of the calendar year(financial year) are considered to be the turn-of-the-year'.

In this paper, the following regression model is applied to test whether the turn-of-the-year (Financial year) period returns are significant enough (null hypothesis being the equality of Mean returns) to accept the presence of the turn-of-the-year (financial year) anomaly.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)

[D.sub.TY] ([D.sub.TFY]) is a dummy variable which is set equal to one for the last five trading days in a particular financial year and the first five trading days in the following financial year and zero otherwise.

3.7 Turn-of-the-Month Effect

Even though different researchers from different countries have defined the term, 'turn-of-the-month' in numerous ways following Elyasiani et al. (1996, 71) and Lakonishok and Smidt (1998) the turn-of-the-month period is defined as the 1st through the 15th calendar day of a particular month and if the 15th is not a trading day, then the next trading day up to the 19th is included. Tan and Tat (1998), on the other hand considers the period from the last trading day of the previous calendar month to the first three trading days of the current calendar month. Additionally, Bildik (n.d.) who documented the presence of the turn-of-the-month effect for the Istanbul Stock Exchange defined the turn-of-the month period as the last five trading days of the month and the first five trading days of the following month. This paper attempts to detect this anomaly for CSE by taking up his definition to the term as well.

Similar to the turn-of-the-year (financial year) effect, the following regression model is applied to test whether the return coefficient relevant to the first half of the month is significant enough to accept the presence of this anomaly in the CSE.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (6)

[D.sub.TM] is a dummy variable taking the value 1 for the first half of the trading month and zero otherwise.

4. Results

4.1 Adjustment to Control the Influence of Other Major Events

Table 1 details the results of the regression equation (2) which is estimated to examine the influence of other major events on the CSE stock returns. Largely negative and significant regression coefficients relating to international crisis situations suggests that such events have triggered an adverse impact on the stock market performance over the period under study. As clearly indicated in the same Table, particular combinations of trading days in relation to both general and presidential elections appear to be significant with varying effects. However, even though the regression coefficients relating to local events appear to be negative they are not significant.

In connection with both indices this special adjustment results in an elimination of 2.7% of data points from the set of total data points.

4.2 Day-of-the-Week Effect

Table 2 illustrates the details the results of the regression equation (3) for the entire sample as well as for each of the two sub samples.

In the entire sample period large, positive and significant mean daily returns can be observed for ASPI on Wednesdays through Fridays where the Friday returns are the largest and are significant at the 1% level. However no Monday seasonality cannot be observed for ASPI as Monday returns are positive and insignificant. On contrary, Monday effect can be observed for SI with large and negative mean returns on Mondays. For SI although Friday returns are positive, they are small and insignificant. Hence for the entire sample period ASPI and SI show Friday and Monday daily effects respectively. When we consider the first sub-period (1985-1994), for ASPI although return coefficient for Mondays are significant at the 10%, they appear to be positive. Apart from this Monday seasonality both the indices do not reflect any significant daily seasonality. Hence the results for the aforesaid period are seemingly in agreement with Abesekara (2001) whose results do not support any day-of-the-week in the 1991-1996 period (a period somewhat identical to the first sub sample).

It is noted that the second sub-period returns for both indices (although in varying degrees) are in strong agreement with the conventional definition of the day-of-the-week-effect. The daily mean returns for ASPI on Mondays and Tuesdays appear to be large and negative and are significant at the 1% level. At the same time Friday returns appear to be large and positive and are too significant at the 1% level. This observation is common to the SI but additionally it shows an increasing positive effect from Wednesday through Friday. The results in fact are identical to the study by Tan and Tat (1998) who observed that the mean returns on Mondays and Tuesdays are lower than those of Wednesdays through Friday. The Figures 1(a) and 1(b) clearly indicate that the returns for both indices on Mondays are highly negative whereas on Fridays are largely positive. (12)

[FIGURE 1 OMITTED]

Hence the empirical results of this paper contribute to the vast majority of previous research findings that daily anomalies in stock markets are an international phenomenon by providing evidence from an emerging stock market of a developing country. The existence of an anomalous behavior in stock returns in the CSE in turn suggests that some form of inefficiency has occurred over the period under study. (13)

4.3 Month-of-the-Year Anomaly (January Effect)

Table 3 details the results of the regression equation (4) for the entire sample period as well as for the two sub-sample periods. It shows that 'on average' for the entire sample period as well as for the two-sub sample periods return coefficients are not statistically significant. Even so, in the entire sample with respect to ASPI for the month of February the return coefficient seems to be large and positive. Yet it is marginally significant at the 10% level. Again in the first sub sample (1985-1994) with respect to ASPI, the return in the month of February too appears to be large and positive, but is marginally significant at the 10% level. However such presence cannot be regarded as significant to acknowledge any monthly seasonality. Simultaneously it is worth mentioning that there is no any detectable January effect. The mean returns for January are although positive for the entire sample as well as for the first sub-sample are small and insignificant. With regard to the second sub sample mean returns are negative and insignificant. Hence it seems that the regression results do not confirm the accepted behavior with higher January returns. Similarly all other months do not reflect any monthly seasonality. (14)

Conclusively neither January Effect nor any monthly seasonality is present in the CSE. Thus the findings of this study are in strong contrast to previous international evidence documented for vast majority of equity markets around the world. The results are in conflict with the studies by Aggarwal and Rivoli (1989), Kuan and Tat (1998), Arsad and Coutts (1997), Gao and Kling (2005), Pandey (n.d.), Kuan and Tat (1998), Coutts et al. (2000). However the results are in agreement with Lakonishok and Smidt (1988), Cheung and Coutts (1997) and Coutts et al. (2002) as their studies do not provide any evidence of January effect in their respective markets. Most importantly this study provides further evidence in support of Abesekara (2001) who found no evidence of a month-of-the-year effect on the CSE.

4.4 Turn-of-the-Year and Turn-of-the-Financial-Year Effects

Tables 4 and 5 show the results of the regression model (5) for the entire sample as well as for the two sub samples. As indicated by the results there appears to be no turn-of-the year effect or a turn-of-the-financial year effect as both the turn-of-the-year and the turn-of-the-financial year coefficients although positive are not found to be statistically significant. (15)

As detailed in Table 5 with few exceptions the corresponding results for the turn-of-the-financial-year period are almost equivalent to the turn-of-the-year results. Conclusively the results indicate that neither of these effects exist in the CSE.

The findings of this paper conflict with those of Elyasiani et al. (1996) as they have provided evidence as to the existence of turn-of-the-year effect in the CSE. The results are not consistent even for the first sub period which is somewhat comparable to the sample period.

With the intention of finding out whether the contradiction is due to the statistical model selected and/or the adjustments (exclusion of data points with the view of controlling the influence of other events on stock returns) made to the original data set the work of Elyasiani et al. is replicated and presented in Tables 6.1, 6.2, 6.3, and 6.4.

Table 6.2 shows regression results for the identical sample period that of Elyasiani et al. (ibid.), but using the regression equation (5). In fact it is interesting to note that the results are not in support of the turn-of-the-year effect. However when we use the non-adjusted data set and the same statistical model applied by Elyasiani et al. (ibid.) not surprisingly the results provide evidence in great support of the turn-of-the year effect with coefficients statistically significant at the 1% level (conclusion identical to Elysiani et al. (ibid.)). (16) At this occurrence it is worthwhile to examine the D-W statistic (first order auto correlation coefficient) as well as the F-stat. As mentioned in the methodology section as a rule of thumb if D-W stat. is found to be 2 in an application one may assume that there is no first order autocorrelation and at the same time a large value for F-statistic indicates that most of the variation in Y is explained by the regression equation and that the model is useful. As Table 6.2 shows D-W statistic (which is 1.105 and 1.465 respectively for ASPI and SI) is far away from 2 which indicates the presence of strong autocorrelation and F- statistic is smaller (although significant). On contrary the regression estimates presented in Table 6.1 (estimated using equation 05) show that the estimated D-W statistic is closer to 2 and F-statistic is significantly larger. Hence it suggests that the equation (5) is somewhat better and improved. (17)

In addition to results, the appropriateness of the equation (5) can be validated by the finding of Abeysekara (2001).18 Furthermore, Meneu and Pardo (n.d., 6) in their study on pre holiday effect although asserts the fact that 'the incorporation of specific autocorrelation schemes does not affect the inferences', this assertion does not seem appropriate in the Sri Lankan scenario. (19)

Decisively it can be claimed that turn-of-the-year anomaly does not exist in the CSE. Hence the contention of Elyasiani et al. (1996) is merely a "statistical artifact." The invalidation of the finding of Elyasiani et al. (ibid.) on one hand strengthen the assertion of Abeysekara (2001) that "substantial autocorrelation is present in the CSE returns" and, in turn, suggests the importance of incorporating specific autocorrelation schemes for any future study on the CSE. Additionally this reveals that the declaration by Meneu and Pardo (ibid.) is irrelevant in the Sri Lankan scenario.

4.5 Turn-of-the-Month Effect

Table 7 presents the estimates of the regression equation (6) for the entire sample as well as for the two sub samples. As per the results the turn-of-themonth coefficients are not statistically significant for the entire sample or for the two sub samples and except for the first sub sample they appear to be negative.

Similarly as shown in Table 8, the application of the definition of Bildik (n.d.) does not provide any evidence in support. Hence the results indicate that the turn-of-the-year anomaly does not exist in the CSE.

Although the findings are consistent with Elyasiani et al. (1996) they are inconsistent with the findings of the studies by Lakonishok and Smidt (1988), Tan and Tat (1998), Bildick (n.d.) as such provides evidence in support the turn-of-the-month anomaly in their respective stock markets. Results however are consistent with Chang (1998, cited by Tan and Tat) whose test does not reveal any turn-of-the-month effects in Hong Kong, Malaysia and Taiwan.

5. Conclusions

The empirical results of this paper apparently confirm the fact that, besides the presence of the day-of-the-week effect over the period, 1995-2005, all other anomalies are not spotted in the CSE. The following concluding remarks can be made with reference to the finding of the day-of-the-week effect. For the entire period the ASPI reports large and negative daily mean returns (significant at the 1% level) on Mondays and Tuesdays whereas the SI reports large and positive mean daily returns (significant at the 1% level) on Fridays. However it is interesting to note that the second sub period (1995-2005) returns for both indices (although in varying degrees) are in strong agreement with the conventional definition of the-day-of-the week. Apart from the general explanations of the strategic planning by institutional investors and risk posed due to unclear expectations, more peculiar explanations such as limited and the narrowed down role played by the stock brokering firms in Sri Lanka and the frequently reported manipulations over the past few years will better clarify the presence of this daily seasonality in the CSE. The existence of an anomalous behavior in stock returns in the CSE in turn suggests that some form of inefficiency has occurred during the last 10 years. Concurrently it advocates the fact that as daily returns do not follow a random walk (implied by the daily regularities) it suggests that the investors can take benefits by timing their trades.

Although this study do not find any evidence in support of the month-ofthe-year effect the extended study on the same topic decisively proves the presence of the turn-of-the-year documented by Elyasiani et al. in 1996 on the CSE is merely a statistical artifact. This extended study further confirms the fact that the model employed in this study is rather advanced in explaining calendar anomalies than the model exploited by the invalidation of the finding of Elyasiani et al. (ibid.) on one hand strengthen the finding by Abeysekara (2001) which asserts the fact that substantial autocorrelation is present on the CSE. Thus, the latter finding emphasizes the importance of incorporating specific autocorrelation schemes for future studies on the CSE. Also, this study rebuts the assertion of Meneu and Pardo (ibid.)--irrelevancy of incorporating any specific auto correlation schemes--in the Sri Lankan context. Conclusively amidst others the empirical findings of this paper as to the daily seasonality contribute to the vast majority of previous research findings by providing evidence from an emerging stock market of a developing country.

REFERENCES

Abeysekara, Sarath, P. (2001), "Efficient Market Hypothesis and Emerging Capital Market in Sri Lanka: Evidence from the Colombo Stock Exchange," Journal of Business Finance and Accounting 28 (1-2): 249-261.

Al-Loughani, Nabeel, and David Chappell, (2001), "Modeling the Day-of-the-Week Effect in the Kuwait Stock Exchange: A Nonlinear GARCH Representation," Applied Financial Economics 11: 353-359.

Arsad, Zainudin, and Andrew Coutts, (1997), "Security Price Anomalies in the London International Stock Exchange," Applied Financial Economics 7: 455-464.

Balaban, Ercan, and Meliha Bulu (1996), "Is There a Semi-Monthly Effect in the Turkish Stock Market?" Working Paper 9606, The Central Bank of the Republic of Turkey, at http://www.tcmb.gov.tr/yeni/evds/teblig/96/9606.html

Balbina, Miguel, and Nuno C. Martins (2002), "The Analysis of Calendar Effects on the Daily Returns of the Portuguese Stock Market: The Weekend and Public Holiday Efects," Economic bulletin, Banco de Portugal.

Bildik, Recep, (n.d.), "Calendar Anomalies in Istanbul Stock Exchange," at http://mfs.rutgers.edu/MFC/MFC10/mfcindex/files/MFC-092%20Bildik.pdf

Boido, Fasano, and Claudio Antonio (2005), "Calendar Anomalies: Daylight Savings Effects," at www.fma.org/Siena/Papers/340347.pdf

Cheung, Kwong, C., Andrew J. Coutts (1999), "The January Effect and Monthly Seasonality in the Hang Seng Index: 1985-97," Applied Economic Letters 5: 121-123.

Coutts, Andrew, J., and Christos Kaplanidis, Jennifer Roberts (2000), "Security Price Anomalies in an Emerging Market," Applied Financial Economics 10: 561-571.

Coutts, Andrew, J., and Mohamed, A. Sheikh (2002), "The Anomalies that Aren't There: The Weekend, January and Pre-holiday Effects on the Gold Index on the Johannesburg Stock Exchange: 1987-1997," Applied Financial Economic 12: 863-871.

Davidson, Russel, and James G. MacKinnon (1993), Estimation and Inference in Econometrics. New York: Oxford University Press, 364.

D'Mello, Ranjan, Stephen P. Ferris, and Chuan Yang Hwang, (2003), "The TaxLoss Selling Hypothesis, Market Liquidity and Price Pressure around the Turn-ofthe-Year," Journal of Financial Markets 6: 73-98.

Elyasiani, E., Perera, P., Puri, T. (1996), "Market Efficiency and Calendar Anomalies in Emerging Capital Markets: Evidence from the Colombo Stock Exchange," Journal of International Financial Markets Institutions and Money 6(7): 59-77.

Gao, Lei, and Gerald King (2005), "Calendar Effects in Chinese Stock Market," Analysis of Economics and Finance 6: 75-88.

Joshi, Joshi, Nayan Krishna, and K.C. Fatta Bahadur (n. d.), "The Nepalese Stock Market: Efficiency and Calendar Anomalies,'" Economic Review 17(17), at http://ssrn.com/abstract=743666

Kuan, Seow Ruth, and Wong Nee Tat, (1998), "The Diminishing Calendar Anomalies in the Stock Exchange of Singapore," Applied Financial Economics 8: 119-125.

Kiymaz, Hail, and Berument Hakman (2003), "The Day of the Week Effect on Stock Market Volatility and Volume: International Evidence," Review of Financial Economics 1(12): 363-380.

Kunkel, Robert, A., William S. Compton, Scott Beyer (2003), "The Turn-of-the-Month Effect Still Lives: The International Evidence," International Review of Financial Analysis 12(2): 207-221.

Lakonishok, Josef, and Seymour Smidt, (1998), "Are Seasonal Anomalies Real? Ninety-Year Perspective," The Review of Financial Studies 1(4): 403-425.

Lian, Kim Kok, and Yoke Wong Chen (2004), "Seasonal Anomalies of Stocks in Asian Equity Markets," Sunway College Journal 1-11.

Lucey, Brian, M. (2000), "Anomalous Daily Seasonality in Ireland?" Applied Economics Letters 7: 637-640.

Madureira, Leonardo, L., and Ricardo P.C. Leal (2001), "Elusive Anomalies in the Brazilian Stock Market," International Review of Financial Analysis 10: 123-134.

Nabeel, Al-Loughani, and David Chappell, (2001), "Modeling the Day-of-the-week Effect in the Kuwait Stock Exchange: A Nonlinear GARCH Representation," Applied Financial Economics 11: 353-359.

Pardo, Angel, and Vicenty Meneu (2004), "Pre-holiday Effect, Large Trades and Small Investor Behavior," Journal of Empirical Finance 11: 231-246.

Panday, I.M., (n.d.), "Is There Seasonality in the Sensex Monthly Returns?" at http://www.iimahd.in/~impandey/

Sears, Stephen, R., and Trennepohl, Gary, L. (1993), Investment Management. New York: Dryden Press, 213-218.

Yanxiang Gu, Anthony (2003), "The Declining January Effect: Evidences from the U.S. Equity Markets," The Quarterly Review of Economics and Finance 43: 395-404.

NOTES

(1.) A number of research papers reviewed taking from different markets (See, for examples, Balbina et al., 2002; D'Mello et al., 2003; Kunkel et al., 2003; Lakonishok et al., 1998; Madureira et al., 2001) directed towards testing calendar anomalies in emerging market of Sri Lanka.

(2.) The MPI was introduced in 1999 to replace the Sensitive Price Index (SI). Therefore, the ASPI and the MPI are the two main price indices now available in the CSE.

(3.) The study by Abeysekara (2001) has principally been carried out to examine the behavior of stock returns on CSE with a view to determine its consistency with the weak form of the EMH.

(4.) A fourty Share Total Return Index--the stocks that missed less than 100 days of trading over the selected study period--with the objective of taking dividends into account (as SI excludes dividends).

(5.) See, Kuan and Tat 1998, 457 for further evidence.

(6.) Capital gains are tax free from March 2002. Consequently tax motivated selling is not relevant to Sri Lanka. However the implication of this cannot be acknowledged fully as the regression model used by Elyasiani et al. (ibid.) is limited by the innate biases. The application of the same definition to the 'improved model' of this study with adjustments to some of these limitations and to control impact of other events (equation 5) is worthy to decide whether such effect actually exist or is just an outcome of the nature of the model used and on the basis of those to determine whether the rejection of the popular tax-loss selling hypothesis is reasonable.

(7.) The ASPI represents the share price movements of all listed companies in the market where as the Milanka Price Index (MPI) and the Sensitive Price Index (SI) of the CSE represent price movements of 25 companies selected based on their market capitalization and liquidity.

(8.) Please refer Arsad and Coutts, 1997; Tan and Tat, 1998; Cheung and Coutts (1999) for illustrations.

(9.) See, Annexure 1 for further details.

(10.) Autocorrelation is usually defined as the 'correlation between members of series of observations ordered in time as in time series data or space as in cross sectional data' (Davidson and Mackinnon, 1993, 364).

(11.) For instance, Bildik (n.d.) defines the turn-of-the year period as the fifteen trading days between the last five days of December where the tax-year ends and the first ten days of January. On the contrary Rozeff and Kinney (cited by Boido and Fasano, 2005) define the first 15 days of January as the turn-of-the-year.

(12.) The descriptive statistics related to the period from 1985 to 1994 indicates that mean returns are negative on Mondays but are notably positive on Fridays.

(13.) Further research is required to examine the reasons for the presence of daily seasonal ties in the CSE. Even though strategic planning of institutional investors and unclear expectations for the following week could be considered as reasonable explanations, it is worthwhile to examine the causes more peculiar to the CSE. Apart from the institutional changes to the operations of the CSE, severe manipulations by market participants seem capable of further clarifying the presence. Over the past few years although there has been an upsurge in the share market investments, the brokering firms in Sri Lanka still appear to play a limited role. Notwithstanding the continuous emphasis on the importance of brokering firms diversifying their operations and moving away from being agency brokers, not much emphasis has been made upon. Further the research content in these firms appears to play an insignificant role. In effect, the inefficient and the inadequate role played by stock brokering firms for the most part seem capable of explaining the existence of the day-of-the-week anomaly.

(14.) Descriptive statistics also do not indicate wide variations in mean returns across months for three sample sets.

(15.) Descriptive statistics show that the mean daily returns in the turn of the year period are on average than the average return in out of turn of the year period by about two folds for both ASPI and SI in the overall period as well as in the two sub periods. However, two folds do not sound significant enough to provide evidence as to the existence of the turn of the year (financial year) effects. Bildic (n.d.) at the time of detecting the turn-of-the-year anomaly in the Istanbul Stock Exchange found that daily average return in the turn-of-the-year period is approximately four-fivefold of the average return in out of turn-of-the-year period.

(16.) Extending further, an attempt has been made to determine whether the disappearance of the effect is either due to the inclusion of the lag values (to eliminate serial correlation) or is just a result of the exclusion of data points that are presumed as a result of the influence of other major events (including political and international crisis situations) or a combination of both. As indicated in Tables 6.3 and 6.4 the results indicate that a combination of these two adjustments can be made responsible for this disappearance. However the regression coefficient estimated for SI using the unadjusted data but with lag values is significant at the 5% level. Hence it seems that with regard to SI adjustment done for serial correlation can explain a major portion of the disappearance of the effect.

(17.) Although adjusted for serial autocorrelation the model still suffers from two main restrictive assumptions that are innate to multivariate regression models, namely normality and constant variance in disturbances.

(18.) The study by Abeysekara (2001, 253) apparently indicates the presence of 'substantial autocorrelation' in the CSE while implying a need for an adjustment of the model for serial correlation and the study the Treasury bill rates and the Stock market performance in Sri Lanka (2001) indicates the significant political impact on market conditions while implying the need for an adjustment to control their impact.

(19.) This may be due to the presence of substantial auto correlation as the inclusion has created a drastic effect on the inferences.

WEERAKOON BANDA YATIWELLA

weerakonyatiwella@yahoo.com

University of Sri Jayewardenepura,Nugegoda

J.L.N. DE SILVA

nimali @ gmail.com

University of Melbourne
Table 1 Adjustment to control the influence of other major
events (Regression Results, 1985-2005)

                                        ASPI

Variable                     Coefficient      t-stat

Constant                                      2.899 ***

[P.sub.T5-] & [P.sub.T2+]      -0.043        -1.226
[P.sub.T4-] & [P.sub.T4+]       0.047         0.886
[P.sub.T3-] & [P.sub.T1+]       0.015         0.400
[P.sub.T2-] & [P.sub.T3+]      -0.035        -0.835
[P.sub.T2-] & [P.sub.T1+]       0.022         0.490
[P.sub.T1-] & [P.sub.T5+]       0.070         2.154 **
[P.sub.T1-] & [P.sub.T4+]      -0.202         -3424 ***
[P.sub.T1-] & [P.sub.T2+]       0.242         6.107 ***
[P.sub.T1-] & [P.sub.T1+]      -0.099        -2.257 **
[G.sub.T5-] & [G.sub.T1+]       0.028         0.872
[G.sub.T4-] & [G.sub.T4+]      -0.055        -1.042
[G.sub.T3-] & [G.sub.T3+]       0.036         0.787
[G.sub.T2-] & [G.sub.T4+]       0.052         0.805
[G.sub.T2-] & [G.sub.T2+]      -0.054        -0.830
[G.sub.T1-] & [G.sub.T5+]      -0.032        -0.983
[G.sub.T1-] & [G.sub.T3+]      -0.049        -0.921
[G.sub.T1-] & [G.sub.T2+]       0.111         1.721 *
[G.sub.T1-] & [G.sub.T1+]      -0.016        -0.530
[D.sub.local]                  -0.008        -0.623
[D.sub.crisis]                 -0.031        -2.357 **
1st lag                         0.338        25.359 ***
[R.sup.2]                       0.124
F-stat.                        34.768 ***
D-W stat.                       1.969

                                         SI

Variable                     Coefficient      t-stat

Constant                                      2.571 **

[P.sub.T5-] & [P.sub.T3+]      -0.034        -0.883
[P.sub.T4-] & [P.sub.T1+]       0.025         0.579
[P.sub.T3-] & [P.sub.T4+]      -0.146        -2.888 ***
[P.sub.T3-] & [P.sub.T1+]       0.128         2.377 **
[P.sub.T2-] & [P.sub.T2+]      -0.001        -0.034
[P.sub.T1-] & [P.sub.T5+]       0.045         1.348
[P.sub.T1-] & [P.sub.T3+]      -0.049        -1.041
[P.sub.T1-] & [P.sub.T2+]       0.320         6.820 ***
[P.sub.T1-] & [P.sub.T1+]      -0.247        -5.516 ***
[G.sub.T5-] & [G.sub.T5+]      -0.100        -2.339 **
[G.sub.T5-] & [G.sub.T1+]       0.097         2.072 **
[G.sub.T4-] & [G.sub.T3+]       0.002         0.035
[G.sub.T3-] & [G.sub.T2+]       0.003         0.081
[G.sub.T2-] & [G.sub.T1+]      -0.026        -0.792
[G.sub.T1-] & [G.sub.T4+]       0.048         1.127
[G.sub.T1-] & [G.sub.T3+]      -0.010        -0.191
[G.sub.T1-] & [G.sub.T2+]       0.047         1.004
[G.sub.T1-] & [G.sub.T1+]      -0.034        -0.749
[D.sub.local]                  -0.002        -0.159
[D.sub.crisis]                 -0.027        -1.984 **
1st lag                         0.256        18.745 ***
[R.sup.2]                       0.083
F-stat.                        22.622 ***
D-W stat.                       1.999

 ***, ** and * denote statistical significance at the
1%, 5% and 10% levels respectively.

Table 2 The Day-of-the-Week Effect (Regression Results)

                      1985-2005                  1985-1994

Variable          ASPI         SI            ASPI           SI

Constant         0.000        0.001 **      0.000         0.000
               (-1.250)      (2.321)       (0.299)       (0.501)
[D.sub.Mon]      0.007       -0.031 *       0.042 *       0.038
                (0.433)     (-1.765)       (1.840)       (1.514)
[D.sub.Tue]                  -0.043 **
                            (-2.437)
[D.sub.wed]      0.037 **    -0.014         0.001        -0.002
                (2.166)     (-0.805)       (0.032)      (-0.071)
[D.sub.Thu]      0.036 **                  -0.001         0.015
                (2.119)                   (-0.035)       (0.607)
[D.sub.Fri]      0.080 ***    0.021         0.027         0.005
                (4.704)      (1.228)       (1.204)       (0.180)
1st lag          0.338 ***    0.027 ***      0.49 ***      0.29 ***
                (25.06)      (19.26)       (27.53)       (14.80)
[R.sup.2]        0.117        0.073         0.245         0.086
F-stat          130.27 ***    77 47 ***     152.8 ***     44.33 ***
D-W stat         1.966        2.008         2.003         2.040

                      1995-2005

Variable          ASPI           SI

Constant         0.001        -0.001 *
                (1.534)      (-2.172)
[D.sub.Mon]      -0.07 ***    -0.010
               (-2.895)      (-0.410)
[D.sub.Tue]     -0.065 ***
                (0.007)
[D.sub.wed]                    0.057 **
                              (2.361)
[D.sub.Thu]      0.006          0.07 ***
                (0.264)       (2.887)
[D.sub.Fri]      0.066 ***      0.12 ***
                (2.726)       (5.008)
1st lag          0.245 ***      0.24 ***
                (12.76)       (12.57)
[R.sup.2]        0.073         0.071
F-stat           40.84 ***     39.19 ***
D-W stat         1.949         1.982

T-statistics are in parentheses.

***, ** and * denote statistical significance at the 1%,
5% and 10% levels respectively.

Table 3 The Month-of-the-Year Effect (Regression Results)

                      1985-2005                   1985-1994

Variable          ASPI           SI            ASPI            SI

Constant         0.000         0.000          0.000          0.001
               (-0.426)      (-0.020)       (-0.362)        (1.188)
[D.sub.Jan]      0.008         0.009          0.015         -0.010
                (0.457)       (0.500)        (0.605)       (-0.366)
[D.sub.Feb]      0.022         0.013          0.043 *        0.018
                (1.262)       (0.704)        (1.813)        (0.699)
[D.sub.Mar]                                                -0.013
                                                          (-0.501)
[D.sub.Apr]      0.022         0.018          0.024          0.006
                (1.257)       (1.034)        (1.012)        (0.234)

[D.sub.May]      0.017         0.012          0.015         -0.015
                (0.972)       (0.681)        (0.618)       (-0.565)
[D.sub.Jun]      0.028         0.027          0.019         -0.007
                (1.565)       (1.470)        (0.801)       (-0.247)
[D.sub.Jul]      0.031 *       0.012          0.037
                (1.719)       (0.654)        (1.512)
[D.sub.Aug]      0.008        -0.013          0.016         -0.029
                (0.450)      (-0.698)         (.659)       (-1.083)
[D.sub.Sep]      0.030         0.022          0.012         -0.006
                (1.637)       (1.207)        (0.517)      (-10.227)
[D.sub.Oct]      0.020         0.020          0.017         -0.004
                (1.079)       (1.073)        (0.703)       (-0.139)
[D.sub.Nov]     -0.002        -0.011          0.004         -0.022
               (-0.112)      (-0.622)        (0.184)       (-0.833)
[D.sub.Dec]      0.021         0.007          0.027         -0.011
                (1.189)       (0.368)        (1.147)       (-0.417)
1st lag          0.334 ***      0.291 ***     0.492 ***      0.292 ***
               (24.753)      (21.125)       (27.230)       (14.645)
[R.sup.2]        0.112         0.086          0.243          0.083
F-stat          52.427 ***    38.632 ***     63.464 ***     18.518 ***
D-W stat         1.968         1.973          1.999          2.039

                       1995-2005

Variable          ASPI          SI

Constant         0.001        -0.001
                (0.919)      (-0.844)
[D.sub.Jan]     -0.021         0.021
               (-0.824)       (0.816)
[D.sub.Feb]     -0.021         0.007
               (-0.824)       (0.263)
[D.sub.Mar]     -0.021
                (0.801)
[D.sub.Apr]     -0.006         0.029
               (-0.237)       (1.159)

[D.sub.May]      0.002         0.033
                (0.072)       (1.293)
[D.sub.Jun]      0.013         0.056
                (0.529)       (2.207)
[D.sub.Jul]      0.002         0.026
                (0.094)       (1.020)
[D.sub.Aug]     -0.023         0.004
               (-0.883)       (0.136)
[D.sub.Sep]      0.025         0.045
                (0.952)       (1.730)
[D.sub.Oct]                    0.035
                              (1.340)
[D.sub.Nov]     -0.030        -0.009
               (-1.165)      (-0.343)
[D.sub.Dec]     -0.005         0.021
               (-0.212)       (0.830)
1st lag          0.239 ***     0.237 ***
               (12.363)      (12.214)
[R.sup.2]        0.058         0.059
F-stat          13.922 ***    13.998 ***
D-W stat         1.955         1.944

T-statistics are in parentheses.

 ***, ** and * denote statistical significance at the 1%,
5% and 10% levels respectively

Table 4 Turn-of-the-Year Effect (Regression Results)

                      1985-2005                   1985-1994

Variable          ASPI          SI            ASPI          SI

Constant         0.000 ***     0.000 **      0.000 **      0.001 **
                (2.858)       (2.396)       (2.406)       (2.335)
[D.sub.TY]       0.013         0.020         0.009         0.012
                (0.980)       (1.445)       (0.487)       (0.616)
1st lag          0.336 ***     0.266         0.495 ***     0.293 ***
               (24.891)      (19.182) **   (27.491)      (14.754)
[R.sup.2]       0.0113         0.071         0.245         0.086
F-stat         311.032 ***   185.996 ***    378.79 ***   109.320 ***
D-W Stat         1.968         2.001         2.001         2.039

                      1995-2005

Variable          ASPI           SI

Constant         0.000         0.000
                 (1.52)       (1.017)
[D.sub.TY]       0.015         0.028
                (0.763)       (1.427)
1st lag          0.242         0.240
               (12.572)      (12.381)
[R.sup.2]        0.058         0.058
F-stat           79.52 ***     78.34 ***
D-W Stat         1.954         1.986

T-statistics are in parentheses.

***, ** and * denote statistical significance at the 1%,
5% and 10% level respectively.

Table 5 Turn-of-the-Financial-Year Effect (Regression results)

                      1985-2005

Variable         ASPI            SI

Constant        0.000 ***       0.000 **
               (3.138)         (2.513)
[D.sub.TFY]    -0.005           0.012
              (-0.389)         (0.891)
1st lag         0.336 ***       0.267 ***
              (24.912)        (19.214)
[R.sup.2]       0.113           0.071
F-stat        310.576 ***     185.300 ***
D-W stat        1.968           2.009

                       1985-1994

Variable         ASPI            SI

Constant        0.000 **        0.001 **
               (2.306)         (2.162)
[D.sub.TFY]     0.017           0.029
               (0.059)         (1.48)
1st lag         0.495 ***       0.293 ***
              (27.519)        (14.712)
[R.sup.2]       0.245           0.086
F-stat        379.244 ***     110.275 ***
D-W stat        2.002           2.040

                       1995-2005

Variable         ASPI            SI

Constant        0.000 *         0.000
               (1.784)         (1.236)
[D.sub.TFY]    -0.025          -0.005
              (-1.304)        (-0.234)
1st lag         0.242 ***       0.242 ***
              (12.575)        (12.479)
[R.sup.2]       0.059           0.058
F-stat         80.444 ***      77.920 ***
D-W stat        1.953           1.942

T-statistics are in parentheses.

***, ** and * denote statistical significance at the 1%,
5% and 10% level respectively.

Table 6 Turn-of-the-year Effect: January 1986-May 1993 (The period
corresponding to Elysiani et al., 1996) Regression Results for Model 5

                ASPI           SI          ASPI           SI

Variable        6.1                         6.2

Constant       0.000         0.000         0.001 ***     0.001 **
{D.sub.TY]     0.005         0.016         0.063 ***     0.073 ***
              (0.223)       (0.694)       (2.642)       (1.856)
1st lag        0.484 ***     0.318 ***       --            --
             (22.965)      (13.936)
[R.sup.2]      0.233         0.102         0.003         0.005
F-stat       264.060 ***    97.727 ***     6.981 ***     9.460 ***
D-W stat       2.014         2.040         1.105         1.465

             [R.sub.u] = [a.sub.u] +       [R.sub.it] = [a.sub.0] +
             [a.sub.1] [D.sub.TY] +        [a.sub.1] [D.sub.TY] +
             [n.summation over (i)]        [e.sub.t]
             [R.sub.t-1] + [e.sub.t]

             The data set is not adjusted for major events that are
             alleged to have had triggered a momentous upshot in the
             CSE performance.

             where [R.sub.it] is the change in daily value of index i
             in period t, [D.sub.TY] is a dummy variable taking the
             value of 1 for the last five trading days in a particular
             year and the first five trading days in the following
             year, and zero otherwise. [R.sub.t-1] represents a lag
             variable and et stands for the random error term for the
             day t.

               ASPI            SI         ASPI          SI

Variable        6.3                        6.4

Constant       0.000 **      0.001        0.001 ***    0.001
{D.sub.TY]     0.028         0.050 **     0.018        0.020
              (1.287)       (2.182)       0.759       (0.882)
1st lag        0.448 ***     0.268 ***      --           --
             (20.919)      (11.604)
[R.sup.2]      0.202         0.075        0.000        0.000
F-stat        223.16 ***    72.420 ***    0.576        0.778
D-W stat       2.005         2.023        1.032        1.347

             [R.sub.it] = [a.sub.0] +     [R.sub.it] = [a.sub.0] +
             [a.sub.1] [D.sub.TY] +       [a.sub.1] [D.sub.TY] +
             [n.summation over (i)]       [e.sub.t]
             [R.sub.t-1] + [e.sub.t]

             The data set is not          The data set is adjusted
             adjusted for major events    for major events that are
             that are alleged to have     alleged to have had
             had triggered a momentous    triggered a momentous
             upshot in the CSE            upshot in the CSE
             performance.                 performance.

             where [R.sub.it] is the change in daily value of index i
             in period t, [D.sub.TY] is a dummy variable taking the
             value of 1 for the last five trading days in a particular
             year and the first five trading days in the following
             year, and zero otherwise. [R.sub.t-1] represents a lag
             variable and et stands for the random error term for the
             day t.

T-statistics are in parentheses; ***, ** and * denote statistical
significance at the 1%, 5% and 10% levels respectively

Table 7 Turn-of-the-month effect - Definition 01
(Elyasiani et al., 1996)

                   1985-2005                   1985-1994

Variable       ASPI          SI            ASPI          SI

Constant      0.001         0.001         0.000         0.001
             (2.534) **    (2.238) **    (1.231)       (1.367)
DTM          -0.007        -0.006         0.014         0.011
            (-0.486)      (-0.435)       (0.783)       (0.542)
1st lag       0.336         0.267          0494 ***     0.294 ***
            (24.925) **   (19.232) **   (27.461)      (14.765)
[R.sup.2]     0.113         0.071         0.245         0.086
F-stat      310.624 ***   184.974 ***   379.041 ***   109.273 ***
D-W stat      1.968         2.009         2.001         2.039

                   1995-2005

Variable       ASPI          SI

Constant      0.001 *       0.001
             (1.917)       (1.609)
DTM          -0.022        -0.021
            (-1.158)      (-1.063)
1st lag       0.241 ***     0.242 ***
            (12.521)      (12.478)
[R.sup.2]     0.059         0.058
F-stat       80.645 ***    78.491 ***
D-W stat      1.954         1.942

T-statistics are in parentheses

***, ** and * denote statistical significance at the 1%,
5% and 10% level respectively

Table 8 Turn-of-the-month effect--Definition 02 (bildik, n.d.)

                   1985-2005                   1985-1994

Variable       ASPI          SI            ASPI          SI

Constant      0.000 ***     0.001 ***     0.000 **      0.000
             (2.375)       (2.235)       (1.264)       (1.144)
DTM          -0.003        -0.006         0.013         0.017
            (-0.247)      (-0.429)       (0.743)       (0.857)
1st lag       0.336 ***     0.267 ***     0.495 ***     0.294 ***
            (24.911)      (19.218)      (27.525)      (14.771)
[R.sup.2]     0.113         0.071         0.245         0.086
F-stat      310.526 ***   184.971 ***   379.001 ***   109.514 ***
D-W stat      1.968         2.009         2.002         2.009

                   1995-2005

Variable      ASPI            SI

Constant      0.001 *       0.001 ***
             (1.872)       (2.008)
DTM          -0.018        -0.029
            (-0.933)      (-1.496)
1st lag       0.242 ***     0.240 ***
            (12.557)      (12.368)
[R.sup.2]     0.058         0.058
F-stat        79677 ***    78.446 ***
D-W stat      1.955         1.987

T-statistics are in parentheses.

***, ** and * denote statistical significance at the 1%,
5% and 10% level respectively
COPYRIGHT 2011 Addleton Academic Publishers
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2011 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Yatiwella, Weerakoon Banda; de Silva, J.L.N.
Publication:Economics, Management, and Financial Markets
Article Type:Report
Geographic Code:9SRIL
Date:Jun 1, 2011
Words:8719
Previous Article:Response to Hartmann on political economy.
Next Article:Social networks and innovation (handicraft industry in Bantul, Yogyakarta).
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters