The January effect in the BRIC countries
Thesis
Master of Science in Business Administration
Specialisation: Finance
University of Groningen
Faculty of Economics and Business
Author: Vladimir Chilingirov
Student number: 2020831
Supervisor: Dr. Peter Smid
2 Abstract
This study investigates the existence of the January effect in Brazil, Russia, India and China (the BRIC countries) by using the Threshold Generalized Autoregressive Conditional Heteroskedasticity in Mean (TGARCH-M) model. The dataset includes the daily returns of four small market capitalization indices for the period of 2003-2011. The results indicate that there is no January effect in any of the countries when the return for the whole month of January is examined. However, when the daily pattern of the returns is investigated, results displaying significant positive returns for the first few days of January in Brazil, Russia and India are obtained.
3 Table of Contents 1. Introduction ... 4 2. Literature review ... 5 2.1 Theoretical background ... 5 2.2 International evidence ... 8 2.3 Emerging markets ... 8 2.4 Hypotheses ... 10 3. Methodology ... 11
4. Data and descriptive statistics ... 15
5. Empirical Results ... 19
6. Conclusion ... 27
Reference list ... 29
4 1. Introduction
The January effect is present when the stock market returns in January are significantly higher than the stock returns during the rest of the year. In this paper I will focus on the January effect in Brazil, Russia, India and China.
Despite the vast quantity of articles regarding the effect, the majority of them are concentrated only in the United States or other countries with a developed market, and there are not many papers, which examine the January effect in a country which is considered to represent an emerging market. According to Bowers and Dimson (1988) international comparisons enable researchers to examine whether factors which are supposedly important in one particular economy, are also important in other economies. What is more, such comparisons are of course particularly useful, because stock market irregularities and their explanations may be market/economy specific. An explanation of an irregularity which is supported by empirical evidence from only one country, could simply be a consequence of that particular country, or the market microstructure of the particular stock market under examination (Coutts and Sheikh (2000)). This gives me a reason to believe that testing for whether or not there is a significant January effect in various different countries would give us a valuable further understanding of this phenomenon.
The acronym BRIC countries is first introduced by O’Neil (2001) and it refers to Brazil, Russia, India and China. These countries are considered to have the most important emerging markets as they currently represent approximately 42% of the world population and 26% of the world territory and by the year 2050 are expected to overtake G7 in terms of gross domestic product by all of the BRIC countries being in top 6 in terms of gross domestic product, with China, India, Brazil and Russia being first, third, fourth and sixth respectively.
5
The main goal of this study is to investigate for the presence of the January effect in the BRIC countries. Consistent with some special conditions in China and India I also expect a possible seasonality in the months of February, March and April. In contrast to other authors who apply monthly returns I am using daily data which allows me to examine the first few days within the month of January, which are the days in which the effect is expected to have the strongest presence.
I am testing for the January effect in the BRIC countries by investigating small market capitalization indices, which increases the probability of detecting the possible January effect. The indices that are examined are: FTSE Brazil Small Cap (Brazil), MICEX Start Cap (Russia), INDIA BSE SMALL CAP (India) and SHENZHEN SE SMEB (China). The data is collected from Datastream and from the official website of MICEX stock exchange1 and it spreads in the period: 01.04.2003 – 29.04.2011. According to Zhang et al. (2008), Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models have a number of advantages in modelling financial anomalies, which include their capability of capturing the stylized facts of financial data and incorporating heteroskedasticity into the modelling process. Their construction can also be flexible enough to allow for various dynamic structures of conditional variance. The GARCH model is also suitable for processing data with non-normal distribution. All these factors are beneficial for avoiding some of the shortcomings of a simple Ordinary Least Squares (OLS) regression. The model that I estimate is the TGARCH-M model, as it is the most suitable model known to the author for explaining anomalies in index returns.
This paper is organized as follows. Firstly, Section 2 reveals the theoretical background on which the paper is based and the empirical results of previous studies. Section 3 explains the methodology that is used. Section 4 analyses the data. Section 5 analyses the empirical results. Finally, Section 6 concludes the paper and gives recommendations for further research.
2. Literature review
2.1 Theoretical background
The first academic study which documented the January effect was written by Wachtel (1942). He believes that additional research is required and states that ―the findings of this brief study have barely scratched the surface of a little-explored subject: seasonal movements in security prices‖. However, the study did not receive appropriate attention and it takes more than three decades for a new study on the subject to be published. Thirty-four years later Rozeff and Kinney (1976) brought the January effect to light. They presented empirical evidence that at the New York Stock Exchange for the period of 1904-1974 the return in January was significantly higher in comparison to the other months of the
1
6
year. Although since then many papers examining the January effect have been written, there is still no consensus about its origin. Ligon (1997) and Moller and Zilca (2007) have summarized well the main theories and the evolution of the effect.
The researches of Keim (1983) and Reinganum (1983) reveal that the January effect is attributable mostly to stocks with small market capitalization. What is more, most of the risk premium of small stocks, which is called the size effect, is attributable to January. Another paper shows that stocks with low price and bad December performance are likely to rebound in the following January, contributing to the January effect (Branch and Chang (1990)). By testing if most people overreact to dramatic news events (the overreacting hypothesis) De Bondt and Thaler (1985, 1987) receive results showing that stocks with bad performance in the past outperform stocks with good performance in the past, in the next few years, most of all by realizing large positive excess returns in January. In contrast to the overreaction hypothesis, which implies that the investors’ irrational behavior can have a permanent influence on the market, there are also theories that suggest that investors still behave rationally but the risk factors are not measured in a proper way, like the research of Reinganum (1981) and Chan et al. (1985) reveals. Rogalski and Tinic (1986) add that the main reason for the January effect is the increased risk in stocks with small market capitalization in January. However, it remains unclear why these unpriced risk factors, generating excess return are concentrated only in January, and Seyhun (1993), after using a stochastic dominance approach, claims that the January effect cannot be attributed to omitted risk factors.
Arguably the most popular explanation of the January effect is the tax-loss selling hypothesis, which was suggested by the person who first spotted the January effect, Wachtel (1942). It states that individuals sell bad performing stocks before year-end to recognize capital losses and reinvest at the beginning of the next year (Branch (1977), Dyl (1977), Keim (1983), Reinganum (1983), Ritter (1988)). A recent research of Chen and Singal (2004) shows that the tax-loss selling hypothesis is the most important cause of the January effect. However, other surveys reveal that the effect exists also in countries where the tax year for individual investors is different from the calendar year (Brown et al. (1983), Fountas and Segredakis (2002)) and even in countries or time periods with no capital gain tax (Berges et al. (1984), Kato and Schallheim (1985), Jones and Wilson (1989)). What is more, the validity of the tax-loss selling hypothesis suggests that December returns should be as negative as January returns are positive, but in general, they are not (Tinic and West (1984)).
7
with small market capitalization (Riepe (1998)). Haugen and Lakonishok (1988) as well as Athanassakos and Schnabel (1994) and Porter et al. (1996) are in support of the hypothesis.
The liquidity hypothesis is consistent with market efficiency and it states that year-end transfers of cash, like employee bonuses, the funding of pension fund contributions and others, produce the January effect (Ogden (1990)). Ligon (1997) also claims that excess individual liquidity is the prime reason for the effect. The fact that individual investors are more likely to purchase small risky stocks than institutional investors is consistent with the liquidity hypothesis and can explain why the January effect is displayed most significantly in stocks with small market capitalization.
The information hypothesis is suggested by Rozeff and Kinney (1976). According to them: ―January marks the beginning and ending of several potentially important financial and informational events. As examples of the latter, January is the start of the tax year for investors, the beginning of the tax and accounting years for most firms and the period during which preliminary (and in many cases final) announcements of the previous calendar (fiscal) year’s accounting earnings are made. It is possible that seasonality is in some way associated with these accounting events.‖2
Another theory suggested by Wachtel (1942) is that the Christmas holidays create positive emotions, excitement and optimism whose impact can continue in the month of January. This can reflect to the stock market in the way that investors who are overoptimistic about the future performance of their stocks are likely to make a lot of stock purchases.
The seasonal money hypothesis is also connected with the Christmas holidays. Chen and Fishe (1994) claim that in late November or early December the Federal Reserve starts to emit money in order to accommodate the increased demand for money provoked by the holidays. However, excess money is a symbol of inflation to investors and can cause a decline in stock prices. Nevertheless, the Federal Reserve has not always withdrawn the holiday-specific money on time and sometimes excess money remains in the economy. In this environment, equity prices may be expected to decline gradually as more holiday-specific money is produced. When the holiday-specific money is withdrawn, which usually begins in the first week of the year, equity values rebound as the threat of inflation is removed, being the main cause for the January effect.
Riepe (1998) and Gu (2003) suggests that publicity and the expanding derivatives market have negative influence on the January effect and its presence on the New York Stock Exchange is declining over the years. Mehdian and Perry (2002) obtain similar results for the main United States markets, unable to find the January effect in post-1987 years. However, Moller and Zilca (2007) argue that the fact that Mehdian and Perry (2002) and Gu (2003) rely solely on monthly data in their researches is the main reason for their results, as the majority of the effect is concentrated in the beginning of January (Keim (1983)). Hence, Moller and Zilca (2007) investigate the daily pattern of
8
the effect, concluding that in the period of 1995-2004 in comparison to 1965-1994 the January effect is present for a shorter period but it is more intense.
2.2 International evidence
International evidence for the effect also exists. Gultekin and Gultekin (1983) investigate for seasonality in 17 developed countries and find significant results for the January effect in 12 of them. In Australia and the United Kingdom they also test for respectively July and April effect as the tax year in Australia starts in July and the tax year in UK starts in April. Brown et al. (1983) investigate the Australian market by examining stocks divided in size deciles and find significant positive January and July effects. Draper and Paudyal (1997) report both January and April effect in UK.
Raj and Thurston (1994) search for seasonality in the New Zealand market, where the tax year starts in April and find neither January nor April effects.
Cheung and Coutts (1999) are unable to find any persistent monthly seasonality in the Hang Seng index in the Hong Kong Stock Exchange.
Berges et al. (1984) discover a significant January effect in Canada both prior and post the implementation of capital tax gain in 1973. In addition, the connection between the effect and the potential of tax-loss selling is insignificant, a result which is not supportive of the tax-loss selling hypothesis.
Kato and Schallheim (1985) find the January effect in Japan.
2.3 Emerging markets3
In comparison to the literature considering the January effect in US or other countries with developed stock market the literature examining countries with emerging markets4 is not that abundant and the author of this paper is not aware of any official papers with the January effect subject in Brazil or Russia.
Fountas and Segredakis (2002) investigate 18 emerging stock markets (currently 15 of them are classified as emerging by Standard & Poor’s), finding evidence for the January effect in only four: Chile, Greece, Nigeria and Turkey. One of the examined countries is India, where the tax year starts in April, and no January or April effect is found there.
Balaban (1995) investigates the Turkish market and finds a January effect, even though there is no capital gain tax in Turkey.
3
Summarizing table with the empirical studies regarding countries with emerging markets is available in the Appendix.
4
see The S&P GBMI (2010) for country classification-
9
Coutts and Sheikh (2000) are unable to find the January effect in the All Gold Index on the Johannesburg Stock Exchange in South Africa.
Maghayereh (2003) searches for seasonality in the Amman Stock Exchange in Jordan, without obtaining any significant results, including for the January effect.
The Kuwait market is examined for seasonality by Al-Saad and Moosa (2005) and significantly high return is found in July but not in January.
Chen et al. (2010) investigate three Asian countries for the January effect using a stochastic dominance approach. One of the investigated countries is Taiwan, which is the only one in the research classified as emerging market by S&P, and the results do not display a significant January effect.
Wong et al. (1990) research the Malaysian stock market for seasonality using the Gregorian, Chinese and Muslim calendars. The authors suggest that the Wachtel’s (1942) hypothesis for the optimistic feeling that the Christmas holidays generate might be relevant for other holidays as well. They find significantly high returns in the months of: January, the Chinese New Year and the Muslim holiday Aidilfitri.
According to Zhang et al. (2008), there is no evidence for the January effect in the Chinese stock market. They believe that the market conditions in China are very different from the market conditions in developed countries. The lack of capital gain tax makes the tax-loss selling hypothesis, which is usually considered the most important cause of the effect, irrelevant. What is more, there is no general practice for Chinese firms to publish information in the end of the year, therefore the information hypothesis can be rejected and fund’s managers do not have special incentives to rebalance the portfolios under their management in the end of the year in particular. In addition, the authors claim that fund managers are not a significant factor in the Chinese market in comparison to the interventions of the Chinese government. Their theory is that the window dressing hypothesis may still be adequate, however it is caused by government actions in the political high season in March when the yearly National Plenary Conferences of People’s Representatives and of People’s Political Consultation, known as the Two Great Conferences, are held. At this time the effect of bad news in the stock market is magnified so the government takes actions to prevent it, creating a March effect. The results of Zhang et al. (2008) are consistent with their theory, being insignificant for the months of January and the month in which the Chinese New Year usually occurs – February and being positive and significant in March.
10 2.4 Hypotheses
From the literature review, it can be observed that there are various different hypotheses for the January effect and each of them is supported by some studies and rejected by others. What is more, the presence or absence of the effect varies with the different markets, time periods and methodology used. While the majority of papers regarding US and other countries with developed markets give evidence in favor of the January effect the same cannot be said about the articles for countries with emerging markets. The papers examining a BRIC country give inconclusive results as well. They also have the shortcoming of universal usage of monthly returns. In this research I will use daily returns as they will enable me to investigate for the January effect within the month of January. At this starting point it is difficult for sensible predictions about the cause of the possible January effect in the BRIC countries to be made. Therefore, this paper will focus mainly on the detection of the possible January effect as well as a few other possible monthly seasonalities which are consistent with the existing theories. In addition to the January effect in each of the countries, in China I will test for the Chinese New Year effect in the month of February and the effect of the political high season in March and in India I will test for an April effect as the beginning of the tax year is on the 1st of April and this is consistent with the tax-loss selling hypotheses. I will also compare the returns in the four countries and if the power of the January effect among them is different. The hypotheses that I will test are whether or not:
1) The average daily return in January is higher than the average daily return for the remaining of the year in FTSE Brazil Small Cap index for the tested period.
2) The average daily return in January is higher than the average daily return for the remaining of the year in MICEX Start Cap index for the tested period.
3) The average daily return in January is higher than the average daily return for the remaining of the year in INDIA BSE SMALL CAP index for the tested period.
4) The average daily return in April is higher than the average daily return for the remaining of the year in INDIA BSE SMALL CAP index for the tested period.
5) The average daily return in January is higher than the average daily return for the remaining of the year in SHENZHEN SE SMEB index for the tested period.
6) The average daily return in February is higher than the average daily return for the remaining of the year in SHENZHEN SE SMEB index for the tested period.
7) The average daily return in March is higher than the average daily return for the remaining of the year in SHENZHEN SE SMEB index for the tested period.
11 3. Methodology
In order to test for the January effect various regressions are estimated and cumulative abnormal return methodology is used (see Moller and Zilca (2007)).
According to Mandlebrot (1963), Fama (1965) and Brooks (2008), it is likely that financial time series will be characterized by a leptokurtic distribution, which is presented by fatter tails and higher peak at the mean in comparison to a normally distributed random variable with the same mean and variance. In addition, financial asset return series are regarded to possess the feature of volatility clustering, which is the tendency of large changes (of either sign) in asset prices to follow large changes and small changes (of either sign) to follow small changes. Brooks (2008) also believes that it is unlikely for financial time series to have error variances being constant, therefore suggesting that heteroscedasticity with non-constant variance of the errors should be expected. In addition, the negative correlation between current returns and future volatility, causing volatility to rise more following a large price fall than following a price rise of the same magnitude is known as the leverage effect (Bollerslev et al. (1992), Brooks (2008)) and is likely to be found in financial asset return (Choudhry (2001)). Classical linear regression models are unable to deal with any of these issues and are likely to produce biased results.
In order to avoid these shortcomings all of the regressions in this research are first estimated by OLS (not reported) but afterwards tested for Autoregressive Conditional Heteroskedasticity (ARCH) effects5. As was expected, in all of the regressions the ARCH test is strongly significant indicating that the usage of a non-liner ARCH type of model is required. Due to the fact that the ARCH model has some limitations, which make its implementation difficult in practice (Brooks (2008), Zhang et al. (2008)), the more commonly used (Choudhry (2001), Maghayereh (2003), Zhang et al. (2008)) GARCH model is estimated.
A Moving Average (MA (1)) term is included in the mean equation in order to capture the effect of negative serial correlation induced by nonsynchronous trading as suggested by Susmel and Engle (1994) and as it is used by Choudhry (2001), Maghayereh (2003) and Zhang et al. (2008).
Given the properties of the examined data series, the error distribution is assumed to be conditionally heteroskedastic and non-normal. Therefore, the model is adjusted for the Student’s-t distribution (Choudhry (2001), Zhang et al. (2008)).
According to Brooks (2008), the standard GARCH model does not take into account the leverage effect and enforces a symmetric response of volatility to positive and negative shocks. In order to examine whether the volatility is indeed symmetric, sign and size bias tests of Engle and Ng (1993)
5 The only exceptions are regressions (10) and (11), which are estimated by simple Panel Least Squares (PLS),
12
are estimated in all of the regressions in the paper. The null hypothesis for symmetric volatility is rejected in all regressions, suggesting that an asymmetric GARCH model is required. Therefore, maybe a Glosten, Jagannathan, Runkle (GJR) (also known as TGARCH) extension to the basic GARCH model is necessary ((Choudhry (2001)).
Finally, to account for one of the most important concepts in finance that higher return should be related to higher volatility the TGARCH model is converted to TGARCH in mean (TGARCH-M) model. The conditional variance enters the mean equation directly implying a time-varying risk premium.
The mean equation of the model is specified in the following way:
(1)
Where is the daily stock return, is an explanatory variable, is the conditional variance, is the lagged error term and it represents the moving average term, is an error term and , , and are parameters. If is positive and statistically significant, then increased risk, given by an increase in the conditional variance, leads to a rise in the mean return, therefore, it can be interpreted as risk premium. A significant is evidence for autocorrelation in the returns.
The conditional variance equation is the following:
(2)
Where is a leverage dummy variable which takes the value of one when innovations to returns
are negative, and zero otherwise. A positive and significant indicates that negative
innovations have a larger effect on returns than positive innovations ((Choudhry (2001)).
In the literature regarding the January effect there is no consistent methodology and variable set that is used universally other than return as dependent variable and dummies for January or dummies for the months from February to December as independent variables. Some of the papers (Cheung and Coutts (1999), Coutts and Sheikh (2000)) do not use additional control variables. However, the control variables that are commonly used account for size and risk. In this research there is no size variable included due to the omission of various stocks or portfolios. The paper focuses only on small market capitalization indices, which makes the presence of huge differences in companies’ sizes unlikely. The risk factor is accounted for by the mean addition to the standard TGARCH model.
The daily rate of return of the indices is calculated using the closing value of the indices at the end of each day by the standard formula:
(3)
13
First, each of the indices is examined by simply testing for the presence of the January effect. Using the TGARCH-M model specified by equations (1) and (2) the following regression is estimated:
(4)
Where D1 is a dummy variable taking a value of 1 if t is in January and 0 otherwise.
Model (4) is a test of the average daily return in January relative to the average daily return for the rest of the year, with α0 measuring the mean daily percentage return for all months other than January and α1 measuring the difference in the daily percentage return for January relative to the daily return in the other months of the year. The test for the January effect is that α1 is significantly greater than zero. In India I also test for seasonality in April by an analogous regression to model (4), estimated using a dummy which takes the value of 1 in April and 0 otherwise. In China I will also test for seasonality in February and March, by estimating two analogous to model (4) regressions - the first using a dummy which takes the value of 1 in February and 0 otherwise and the second using a dummy which takes the value of 1 in March and 0 otherwise.
(4a) (4b)
(4c)
In addition, for each index, I investigate the return within January, by calculating the average daily return (for the whole period), the average abnormal return and the cumulative abnormal return (the following methodology is based on Moller and Zilca (2007)).
The average daily return for the sample period is:
(5)
Where ADR is the average daily rate of return for the sample period, Rt is the rate of return at day t and n is the number of observations in the whole sample period.
In order to calculate the average abnormal return I sort all the observations by their consecutive trading day number during the year. Afterwards I calculate the average for each of the trading days:
(6)
Where TDAtd is the average rate of return at trading day td, is the sum of the returns at trading day td of all years in the sample period and ntd is the number of observations at trading day td in all years in the sample period.
The average abnormal return for each trading day is estimated in the following way: (7)
Where AARtd is the average abnormal rate of return at trading day td and ADR is the average daily rate of return for the sample period.
14
(8)
Where CARtd is the cumulative abnormal rate of return at trading day td. In order to clarify the process an example is presented - at trading day 2, CAR2 = AAR1 + AAR2. Examining the maximum of the cumulative abnormal rate of return in the beginning of the year allows me to display on which day in the year the possible January effect is at its peak. As it is possible that the January effect keeps its presence for a certain time even after the end of January, Moller and Zilca (2007) believe that the first 80 trading days are relevant. In the current paper results for all trading days are presented.
No statistical evidence is obtained from equation (8) but it is used for illustrative purposes and mainly as a guideline for regression (9).
According to Keim (1983), 26.3% of the small firm premium in his research is realized in the first five trading days of the year and 10.5% is realized in the first trading day of the year. Therefore, the first trading days of the year are considered to be extremely important and the next step in my searching for the January effect is to test whether they have higher returns than the rest of the year. The exact number of days to be examined in each country is determined individually as the consecutive number from the beginning of the year of the trading day with the maximum CAR in the first 22 trading days of the year. The following regression is estimated for each index:
(9)
Where D2 is a dummy variable taking a value of 1 if t is within the first day of the year and the chosen number of days depending on the CAR and 0 otherwise.
Model (9) examines the average daily rate of return in the first trading days of January relative to the average daily rate of return for the rest of the year, with α0 measuring the mean daily percentage retum for all days other than the first trading days in each year and α1 measuring the difference in the daily percentage return for January relative to the daily return in the other months of the year. The first trading days of the year have significantly higher rate of return than the rate of return during the rest of the year if α1 is significantly greater than zero.
Next, the observations for all BRIC countries are examined together testing for possible country based determinants of the effect which allow us to compare the power of the effect among the countries, by the following PLS regression:
(10)
15
return in January in a country to the average daily return during the rest of the year in the country and the relation of the average daily return in January in the rest of the countries to the average daily return during the rest of the year in rest of the countries. α1 is measuring the average daily return in Brazil from February to December, α2 is measuring the average daily return in Russia from February to December, α3 is measuring the average daily return in India from February to December, α4 is measuring the average daily return in China from February to December, α5 is measuring the difference in the daily percentage return for January relative to the daily return in the other months of the year in Brazil, α6 is measuring the difference in the daily percentage return for January relative to the daily return in the other months of the year in Russia, α7 is measuring the difference in the daily percentage return for January relative to the daily return in the other months of the year in India, α8 is measuring the difference in the daily percentage return for January relative to the daily return in the other months of the year in China.
Finally, similar PLS regression to equation (10) is estimated, comparing the power of the January effect among the BRIC countries, but using only the first few days of January as a dummy based on equations (8) and (9) instead of a simple dummy for January.
(11)
Where D2 is a dummy variable taking a value of 1 if t is within the first day of the year and the chosen number of days depending on the CAR and 0 otherwise. Based on the CAR the number of days included is chosen individually for each country.
Regressions (10) and (11) are estimated by PLS due to the inability of Eviews 7 to process Panel Data by ARCH/GARCH type of models. The standard errors are corrected for heteroskedastcity by White’s cross-section adjustment (Brooks (2008)) and the coefficients are unlikely to be biased. However, the results should be interpreted with caution as some of the shortcomings of OLS models still remain.
4. Data and descriptive statistics
According to Ligon (1997) and Gu (2003), using indices instead of a portfolio of individual stocks avoids issues associated with portfolio formation, such as size–beta correlation, size–price correlation and survivorship for example. The indices that I use are: FTSE Brazil Small Cap (Brazil), MICEX Start Cap (Russia), INDIA BSE SMALL CAP – price index (India) and SHENZHEN SE SMEB – price index (China). These are all value-weighted, small market capitalization indices and I obtained the data from Datastream and from the official website of MICEX stock exchange.6 All indices are
6
16
price indices which means that the dividend yield is not taken into account. Although there is certain data available for total return small capitalization indices in Brazil, China and India, its implementation will significantly limit the number of observations and what is more will be inconsistent in terms that there is no data available for a total return small market capitalization index for a sufficiently long period in the Russian market.
In addition, according to Coutts (1997), Cheung and Coutts (1999), Coutts and Sheikh (2000), Choudhry (2001) and Zhang et al. (2008), it is very unlikely that the exclusion of dividend payments invalidates the results of anomaly studies. His claim is in support of the researches of Lakonishok and Smidt (1988), Fishe et al. (1993) and Draper and Paudyal (1997), who employ a variety of data sets, and conclude that any bias which occurs from not employing dividend adjusted returns is so small, as to be negligible. Gultekin and Gultekin (1983) have dividend data available for only part of the sample period. However, after comparison of the adjusted and unadjusted index they also conclude that the omission of dividend payments is insignificant for the results. Therefore, I believe that using price indices, unadjusted for dividend yield, is not inconsistent with the previous literature and it will not bias the results of the study.
The collected daily data is from the period: 23.09.2003 - 29.04.2011 in Brazil, 01.12.2005 – 29.04.2011 in China, 01.04.2003 – 29.04.2011 in India and 11.01.2005 – 29.04.2011 in Russia. Since calculation of small capitalization indices is a relevantly new occurrence in the BRIC countries stock exchanges, observations for earlier periods are not available to the author. However, I believe that the current data is still adequate for a regression, as other researches use periods with similar length as well (see e. g. Balaban (1995), Draper and Paudyal (1997), Fountas and Segredakis (2002) and Maghayereh (2003), the last of whom estimates GARCH, EGARCH and TGARCH models, which have similar requirements for the data quantity as the model that is used in the current research). Descriptive statistics about the returns in the indices in Brazil, Russia, India and China for the sample period is presented in table 1:
Table 1 Table 1a Brazil
17 August 0.001035 0.000896 0.013864 -0.617407 4.458383 23.58355*** 155 September 0.000875 0.002738 0.019188 -0.487596 14.59669 852.1071*** 151 October -0.000844 0.00162 0.025685 -0.265313 9.28297 281.6141*** 170 November 0.002511 0.003866 0.016334 -0.28817 4.690721 20.73959*** 156 December 0.002454 0.003493 0.012341 -0.902603 6.63255 109.6946*** 160 All 0.000961 0.001668 0.0164 -0.324876 9.741526 3593.178*** 1880 Table 1b Russia
Month Mean Median Std. Dev. Skewness Kurtosis JB Observati ons January -0.000108 0.000463 0.024695 -0.669922 29.13148 3137.977*** 110 February 0.004075 0.002466 0.018571 0.350899 5.100449 27.38293*** 134 March 0.001319 0.001425 0.013135 -0.341051 4.310759 13.64596*** 150 April 0.000698 -0.000587 0.014095 0.329676 4.770747 22.16555*** 149 May 0.000556 0.001100 0.018303 -0.413443 5.215025 27.95037*** 120 June 0.000095 0.000886 0.017011 -1.544212 11.21889 398.2911*** 124 July 0.001190 0.002476 0.013607 -0.556636 3.877733 11.05384*** 132 August 0.001583 0.001531 0.011230 0.148645 4.154852 7.880609** 133 September 0.001751 0.003582 0.020725 -2.214580 24.88355 2679.475*** 129 October -0.003930 0.000602 0.020824 -2.175159 10.93674 450.5445*** 132 November 0.001985 0.002307 0.014362 -1.180585 6.800288 102.5887*** 123 December 0.002882 0.001450 0.011610 1.432783 15.33448 875.2487*** 131 All 0.001028 0.001403 0.016834 -1.054175 19.8689 18869.62*** 1567 Table 1c India
18 Table 1d China
Month Mean Median Std. Dev. Skewness Kurtosis JB Observations January 0.000892 0.003393 0.022927 -0.442489 3.893218 7.44399** 113 February 0.002824 0.006004 0.023354 -1.000826 6.259692 59.74816*** 98 March 0.000934 0.003583 0.01799 -0.998726 6.449791 88.72405*** 134 April 0.002152 0.005715 0.024075 -0.262071 4.237469 9.256025*** 123 May 0.002666 0.005799 0.025657 -0.704049 4.029566 11.91743*** 94 June -0.002786 0.001122 0.027359 -0.982203 4.239029 23.37435*** 104 July 0.003294 0.004476 0.020101 -0.654392 4.18256 14.39006*** 111 August -0.001075 0.005198 0.023925 -0.674908 3.71943 10.72309*** 110 September -9.87E-05 0.001857 0.020938 0.268986 6.861131 63.95717*** 101 October -0.001844 0.001053 0.020423 -0.852519 4.565909 19.42719*** 87 November 0.00276 0.004828 0.023927 -0.286543 4.027946 6.175233** 107 December 0.003611 0.005227 0.018491 -0.65501 3.844003 13.25553*** 131 All 0.001199 0.003927 0.022443 -0.636518 4.81602 269.0856*** 1313
Where the daily rate of return is calculated from the daily closing price of the indices according to equation (3) and the standard deviation is estimated based on a sample by the following formula:
( ) (12)
Where, STD is the standard deviation, Rt is the rate of return at day t, ADR is the average daily rate of return for the relevant month (or for the last row of each table the whole sample period) and n is the number of observations in the relevant month (or for the last row of each table the whole sample period).
19 5. Empirical Results
Table 2 displays the results of regression (4). Considering the majority of papers investigating emerging markets and the descriptive statistics of the data, it is not puzzling that the coefficient of the January dummy is negative in all of the countries, suggesting that there is a negative relation between the occurrence of January and the returns. However, the relation is not significant in Russia, India and China and only marginally significant in Brazil. In Brazil, Russia and India the negative coefficient of the conditional variance addition to the mean equation shows a negative relation between the conditional variance and the return and the positive relation in China is consistent with the Capital Asset Pricing Model (CAPM), however the relation is not significant in any of the BRICs. In all countries the significant positive MA (1) term indicates positive autocorrelation, which means that days with positive returns are more likely to be followed by days with positive return and days with negative returns are more likely to be followed by days with negative return. The positive and significant in Brazil and India is evidence for the leverage effect that bad news have greater impact on volatility than good news.
Table 2
Test for the January effect TGARCH-M
(4) (2)
Brazil Russia India China
20
***, ** and * denotes significance at 1%, 5% and 10% level, respectively. The t-statistics are reported in parentheses.
The results for February and March in China and April in India are presented in Table 3 and are not significant, not being in favor of the Chinese New Year effect and the window dressing hypothesis for the political season in China and the tax loss selling hypothesis in India. is again an evidence for positive autocorrelation and indicates a leverage effect in India.
Table 3
Test for February, March and April effect TGARCH-M (4a) (4b) (4c) (2) China February(4a)
China March(4b) India April(4c)
21
In order to examine the possible daily presence of the January effect the cumulative abnormal returns during the year, calculated by equation (8) are displayed in figures 1a, 1b, 1c and 1d. The cumulative abnormal returns in the first 22 trading days of the year is displayed in table 4.
22 Figure 1c
Figure 1d
Table 4
Cumulative abnormal return in the first 22 trading days of the year
Brazil Russia India China
23 4 0.008872 0.013773 0.016596 0.02039 5 0.012574 0.018137 0.00707 0.020481 6 0.011277 0.022142 0.002933 0.021918 7 0.013453 0.037667 -0.01105 0.019031 8 0.016378 0.035061 -0.02156 0.033498 9 0.005158 0.005372 -0.01461 0.043553 10 0.004419 -0.01015 -0.01554 0.03766 11 0.00575 -0.01904 -0.02308 0.045048 12 -0.00049 -0.0155 -0.02788 0.054112 13 -0.01041 -0.01225 -0.02436 0.044265 14 -0.01994 -0.01014 -0.04064 0.036147 15 -0.01985 -0.01256 -0.06797 0.012361 16 -0.0203 -0.01376 -0.08508 0.008638 17 -0.014 -0.00699 -0.07809 0.019065 18 -0.02057 -5.9E-05 -0.07844 0.021783 19 -0.02121 -0.00369 -0.08082 0.013643 20 -0.02901 -0.00515 -0.08178 0.015475 21 -0.03244 0.003615 -0.09217 0.022907 22 -0.03004 0.021607 -0.09978 0.01638
The maximum CAR in each country is highlighted.
24
negative return on the 7th trading day. It is possible that the TGARCH-M model will take that into account and it is less likely to display a significant result in favour of the January effect when 7 trading days are used. Therefore, I believe it is more appropriate to use 6 days in the regression considering Russia. The results from the regressions are presented in table 5.
Table 5
Test for the first days of the year TGARCH-M
(9)
Where D2 is a dummy variable taking a value of: 1 if t is within the first 3 days of the year and 0 otherwise in Brazil, 1 if t is within the first 6 days of the year and 0 otherwise in Russia, 1 if t is within the first 2 days of the year and 0 otherwise in India and 1 if t is within the first 12 days of the year and 0 otherwise in China.
(2)
Brazil Russia India China
25 (34.19577) (27.96999) (27.40975) (29.38125) 0.126007*** (4.535075) 0.034887 (0.812115) 0.085372** (2.347794) 0.000313 (0.008189) ***, ** and * denotes significance at 1%, 5% and 10% level, respectively. The t-statistics are reported in parentheses.
Investigating the daily pattern of the returns, in terms of the January effect completely different results in comparison to the results in regression (4) are obtained. The difference between the first few trading days of the year and the rest of the year is positive in all of the countries and significant at at least the 10% level in Brazil, Russia and India. As was expected the extension presents similar results to regression (4), with no significant negative coefficients in Brazil, Russia and India and positive coefficient in China. As in the previews regression positive autocorrelation is present in all of the countries and the leverage effect is significant in Brazil and India.
Table 6
Test for country based determinants of the effect Panel Least Squares
(10)
(11)
Where D2 is a dummy variable taking a value of: 1 if t is within the first 3 days of the year and 0 otherwise in Brazil, 1 if t is within the first 6 days of the year and 0 otherwise in Russia, 1 if t is within the first 2 days of the year and 0 otherwise in India and 1 if t is within the first 12 days of the year and 0 otherwise in China.
26 (3.897308) (2.624997) 0.001227* (1.897308) 0.000937 (1.459671) -0.00207 (-1.49367) 0.006489* (1.879407) -0.00122 (-0.51284) 0.00383** (2.068918) -0.005*** (-2.77664) 0.014924*** (7.526248) -0.00034 (-0.14965) 0.004771** (2.063403) Adjusted R2 0.001009 0.001962
***, ** and * denotes significance at 1%, 5% and 10% level, respectively. The t-statistics are reported in parentheses.
Equation (10) with all the countries in a single regression is estimated by PLS and the results are presented in table 6. The significant α1, α2, α3 and α4 coefficients indicate that there is a significant difference in the return among all of the countries and as expected the average daily return between February and December is positive. Similarly to regression (4) all of the coefficients regarding the month of January are negative. What is more, α7 indicates that India is the only country in which the difference between the return in January and the return during the rest of the year is both negative and significant and is the highest in comparison to the other countries.
27 6. Conclusion
In this paper I investigate indices of stocks with small market capitalization, in the BRIC countries, for the last eight years, by using the TGARCH-M model, for the presence of the January effect. Considering the negative, although not significant relation between the presence of January and the returns in each country it is hard to argue in favor of the January effect. This result is consistent with most of the papers regarding the January effect in countries with emerging markets. However, when we observe the daily pattern of the returns within January the first few days have a significant positive return in comparison to the rest of the year in Brazil, Russia and India. This is indeed consistent with Moller and Zilca (2007) who argue that the presence of the effect has become shorter over time. However, there are two main reasons that do not allow me to consider this result in favor of the January effect. Whereas Moller and Zilca (2007) investigate the US market for which numerous papers have documented the effect, the current paper investigates emerging markets for which only limited information is publicly known. Therefore, it is hard to conclude that the duration of the effect has declined over time as it is arguable if the effect has ever been present. What is more, Moller and Zilca (2007) claim that although the duration of the effect has declined the return realized in the first few days of January has increased and the power of the effect is the same as before. This is clearly not the case in this research as the average January returns are one of the lowest in comparison to the other months in all of the countries.
Although the paper does not aim to provide a thorough explanation for the cause of the January effect it can be concluded that the hypothesis that tax-loss selling is the main cause for the effect is not supported as the country with the highest and most significant coefficient for the dummy for the first few days of January is India where the tax year ends in March. What is more, no positive and significant April return is indicated there. The hypotheses for the Chinese New Year effect and window dressing in the politically intensive March in China are not supported as well.
It is unlikely that investors can profit from the effect by buying stocks in December and selling them at a particular day in January as it is possible that the obtained results are specific for the examined indices. All of them are small market capitalization indices in emerging markets, therefore, the stocks included in them are likely to have very high transaction costs which will eliminate the profits. What is more, Moller and Zilca (2007) suggest that the daily pattern of the January effect evolves over time, which is an obstacle for choosing the optimal number of days for which to hold the stocks.
28
countries to provide different results. Due to the unavailability of earlier data for the examined indices a period of only 8 years is examined which does not allows the author to compare different time periods.
What is more, a big part of the examined period includes the years of the global financial crisis. Despite that, with the exception of Russia, the other BRIC countries have not been influenced by the crisis as strongly as the US, it is possible that it has affected the distribution of the return during the year and therefore the January effect.
In regressions (10) and (11), comparing the January effect among the countries, a simple PLS model is used which is not consistent with the TGARCH-M model used in the other regression and it is possible to provide results that are biased.
Since this study indicates that that the first days of the year have significantly higher return than the rest of the year, the author recommends for future studies to investigate datasets including longer periods in the BRIC countries focusing again on the daily pattern of the return. In this case it will be possible to separate the observations in different time periods and observe the evolution of the daily returns and the possible January effect.
29 Reference list
Al-Saad, K. and Moosa, I. A. (2005), Seasonality in stock returns: evidence from an emerging market, Applied Financial Economics, 15, 63 – 71.
Athanassakos, G. and Schnabel, J. A. (1994), Professional portfolio managers and the January effect: Theory and evidence, Review of Financial Economics, 4, 79 – 91.
Balaban, E. (1995), Juanuary effect, yes! What about mark twain effect?, The Central Bank of the Republic of Turkey- Research Department, Discussion Paper No. 9509.
Berges, A., McConnell, J. J. and Schlarbaum, G. G. (1984), The Turn-of-the-Year in Canada, Journal of Finance, 39, 185 - 92.
Bollerslev, T., Chou, R. and Kroner, K. (1992) ARCH modeling in finance, Journal of Econometrics, 52, 5 – 59.
Bowers, J. and Dimson, E. (1988), Introduction, in Stock Market Anomalies (Ed.) E. Dimson, Cambridge University Press, Cambridge, pp. 3 - 15.
Branch, B. (1977), A tax-loss trading rule, Journal of Business, 50, 198 - 207.
Branch, B. and Chang, K. (1990), Low price stocks and the January effect, Quarterly Journal of Business and Economics, 29, 90 - 118.
Brooks, C. (2008), Introductory Econometrics for Finance, Second edition, Cambridge University Press.
Brown, P., Keim, D. B., Kleidon, A. W. and Marsh, T. A. (1983), Stock return seasonalities and the tax-loss selling hypothesis: Analysis of the arguments and Australian evidence, Journal of Financial Economics, 12, 105 - 127.
Chan, K. C., Chen, N. and Hsieh, D. A. (1985), An Exploratory Investigation of the Firm Size Effect, Journal of Financial Economics, 14, 451 - 471.
Chen, H. and Singal, V. (2004), All things considered, taxes drive the January effect, Journal of Financial Research, 27, 351 - 372.
Chen, L. and Fishe, R. P. H. (1994), Seasonal Money Movements and the January Effect, Atlantic economic journal, 22, 26 – 42.
Chen, W., Hsueh, Y., Chen, Y. and Chang, C. (2010), Investigating Seasonal Anomalies in Asian Stock Market Prices: A Stochastic Dominance Approach, International Journal of Management, 27, 144 – 152.
Cheung, K. C. and Coutts, J. A. (1999), The January effect and monthly seasonality in the Hang Seng index: 1985-97, Applied Economics Letters, 5, 121 — 123.
30
Cooper, M. J., McConnell, J. J. and Ovtchinnikov, A. V. (2006), The other January effect, Journal of Financial Economics, 82, 315 – 341.
Coutts, J. A. (1997), Discussion of microstructure and seasonality in the UK equity market, Journal of Business Finance and Accounting, 24, 1205 - 1215.
Coutts, J. A. and Sheikh, M. A. (2000), The January effect and monthly seasonality in the All Gold Index on the Johannesburg Stock Exchange 1987-1997, Applied Economics Letters, 7, 489 - 492. De Bondt, W. F. M. and Thaler, R. (1985), Does the stock market overreact? Journal of Finance, 40, 793 - 805.
De Bondt, W. F. M. and Thaler, R. (1987), Further evidence on investor overreaction and stock market seasonality, Journal of Finance, 42, 557 - 581.
Draper, P. and Paudyal, K. (1997), Microstructure and seasonality in the UK equity market, Journal of Business Finance and Accounting, 24, 1177 - 1204.
Drew, M. E., Naughtong, T. and Veeraraghavan, M. (2003), Firm size, Book-to-Market Equity and Security Returns: Evidence from the Shanghai Stock Exchange, Australian Journal of Management, 28, 119 – 140.
Dyl, E. A. (1977), Capital gains taxation and year-end stock market behavior, Journal of Finance, 32, 165 - 175.
Engle, R. F. and Ng, V. K. (1993), Measuring and Testing the Impact of News on Volatility, Journal of Finance, 48, 1749 – 1778.
Fama, E. (1965), The behaviour of stock market prices, Journal of Business 38, 34 – 105.
Fishe, R. P. H., Gosnell, T. F. and Lasser, D. J. (1993), Good news, bad news, volume and the Monday effect, Journal of Business Finance and Accounting, 20, 881 - 892.
Fountas, S. and Segredakis, K. N. (2002), Emerging stock markets return seasonalities: The January effect and the tax-loss selling hypothesis, Applied Financial Economics, 12, 291 - 299.
Gu, A. Y. (2003), The declining January effect: Evidence from the US equity markets, Quarterly Review of Economics and Finance, 43, 395 – 404.
Gultekin, M. N. and Gultekin, N. B. (1983), Stock market seasonality: International evidence, Journal of Financial Economics, 12, 469 – 481.
Haugen, R. A. and Lakonishok, J. (1988), The Incredible January Effect: The Stock Market's Unsolved Mystery (Dow Jones-Irwin, Homewood, III).
Jones, S. P. and Wilson, J. W. (1989), An analysis of the January effect in stocks and interest rates under varying monetary regimes, The Journal of Financial Research, 12, 341 – 353.
31
Keim, D. B. (1983), Size-related anomalies and stock return seasonality: Further empirical evidence, Journal of Financial Economics, 12, 13 - 32.
Lakonishok, J. and Smidt, S. (1988), Are seasonal anomalies real? : a ninety year perspective, Review of Financial Studies, 1, 403 - 425.
Ligon, J. A. (1997), A simultaneous test of competing theories regarding the January effect, Journal of Financial Research, 20, 13 - 32.
Maghayereh, A. (2003), Seasonality and January effect anomalies in an emerging capital market, The Arab Bank Review, 5, 25 – 32.
Mandlebrot, B. (1963) The variance of certain speculative prices, Journal of Business, 36, 394 – 419. Mehdian, S. and Perry, M. J., (2002), Anomalies in US equity markets: A reexamination of the January effect, Applied Financial Economics, 12, 141 – 145.
Moller, N. and Zilca, S. (2008), The evolution of the January effect, Journal of Banking & Finance, 32, 447 – 457.
O’Neil, J. (2001), Building Better Global Economic BRICs, Global Economics Paper, 66.
Ogden, J. P. (1990), Turn-of-month evaluations of liquid profits and stock returns: A common explanation for the monthly and January effects, Journal of Finance, 45, 1259 - 1272.
Porter, D. C., Powell, G. E. and Weaver, D. G. (1996), Portfolio Ownership, Rebalancing, Institutional and The Small Firm-January Effect, Review of Financial Economics, 5, 19 – 29.
Raj, M. and Thurston, D. (1994), January or April? Tests of the turn-of-the-year effect in the New Zealand stock market, Applied Economics Letters, 1, 81 – 83.
Reinganum, M. R. (1981), Misspecification of capital asset pricing: Empirical anomalies based on earnings' yields and market values, Journal of Financial Economics, 9, 19 – 46.
Reinganum, M.R. (1983), The anomalous stock market behaviour of small firms in January: Empirical tests for tax-loss selling effects, Journal of Financial Economics, 12, 89 - 104.
Riepe, M. W. (1998), Is publicity killing the January effect?, The Journal of Financial Planning, 11, 64 – 70.
Ritter, J. R. (1988), The buying and selling behavior of individual investors at the turn of the year, Journal of Finance, 43, 701 - 717.
Rogalski, R. J. and Tinic, S. M. (1986), The January Size Effect: Anomaly or Risk Mismeasurement?, Financial Analysts Journal, November- December, 63 - 70.
Rozeff, M. S. and Kinney, W. R. Jr. (1976), Capital market seasonality: the case of stock returns, Journal of Financial Economics, 3, 379 - 402.
32
Susmel, R. and Engle, R. F. (1994), Hourly volatility spillovers between international equity markets, Journal of International Money and Finance, 13, 3 – 25.
Tinic, S. M. and West, R. R. (1984), Risk and return: January vs. the rest of the year, Journal of Financial Economics, 13, 561 - 574.
Wachtel, S. (1942), Certain observations on seasonal movements in stock prices, Journal of Business, 15, 184 - 193.
Wong, P. L., Neoh, S. K., Lee, K. H. and Thong, T. S. (1990), Seasonality in the Malaysian Stock Market, Asia Pacific Journal of Management, 7, 43 – 62.
Zhang, Z., Sun, W. and Wang, H. (2008), A new perspective on financial anomalies in emerging markets: the case of China, Applied Financial Economics, 18, 1681 – 1695.
Electronic Resources
http://www.micex.com/marketdata/indices/shares/capital The S&P Global Broad Market Index, 31 December 2010
33 Appendix
Table 7
Summary of empirical papers regarding countries with emerging markets
Authors Research question Methodology Investigated
period Investigated country Significant January Effect Other relative findings Wong et al. (1990) Is there January, Chinese
New year or Aidifitry effects?
U-test, using the Gregorian, Chinese lunar and Muslim lunar calendars.
1970-1985 Malaysia yes Significant Chinese New Year effect and Aidifitry effects. Balaban (1995) Is there a monthly
seasonality?
T-test and Two-sample test. 1988-1993 Turkey yes Coutts and Sheikh
(2000)
Is there a monthly seasonality?
OLS regression. 1987-1997 South Africa no
Fountas and
Segredakis (2002)
Is there a monthly seasonality?
OLS regression. 1987-1996 Argentina
34 Drew et al. (2003) Is the beta coefficient of
the CAPM the sole measure of risk?
OLS regression of multifactor model.
1993-2000 China no Not significant Chinese New Year effect. Maghayereh (2003) Is there a monthly
seasonality?
GARCH, EGARCH and
TGARCH regressions.
1994-2002 Jordan no
Al-Saad and Moosa (2005)
Is there a monthly seasonality?
Structural Time Series Model. 1984-2000 Kuwait no Zhang et al. (2008) Is there a monthly
seasonality?
GARCH-M. 1993-2003 China no Not significant Chinese
New Year effect and significant March effect.
Chen et al. (2010) Is there a monthly seasonality?
Stochastic dominance approach and Davidson and Duclos test.
