The January Effect during the financial crisis; a comparison
between developed and emerging economies.
Bachelor Thesis Finance & Organization July 1th 2014
Name: Boyd Ridder
Student number: 10003659 Field: Finance
Supervisor: I. Sakalauskaite
Abstract
In this study a test for the January effect in emerging economies during the time period 2001-2012 is conducted. Also the comparison is made for the January effect in two different time periods. On the one side 2001-2006 represents a time period without financial crisis, while on the other side 2007-2012 represents a time period during a financial crisis. Last, a
comparison is made for the January effect between emerging and developed economies during a financial crisis. Information about these differences will give valuable information for investors on the major stock markets. Knowing when and where the January effect is on his strongest, gives the investors insights when and where to spend their money to yield the highest profits. Results showed that the January effect was present in the emerging economies during the time period 2001-2012. Second, there was not enough evidence to proof that the January effect was stronger during the financial crisis. Also, there was not enough evidence to proof that the January effect was stronger for emerging economies than for developed economies during the financial crisis.
Table of contents
1. Introduction………4
2.Literature Review………...5
2.1 Reasons for the January effect………...5
2.2 The January effect outside the U.S.………...7
2.3 Other finding about the January effect………..8
2.4 Link to Hypotheses………....9
3. Data, Methodology & Results………...11
3.1.1 Data and Sample………..11
3.1.2 Test for correlation in control variables………...12
3.2 Methodology, Descriptive Statistics and Results for the first hypothesis……..13
3.2.1 Methodology………....13
3.2.2 Descriptive Statistics………14
3.2.3 Results………..15
3.3 Methodology and results for the second and third hypothesis...16
3.3.1 Methodology...16
3.3.2 Descriptive Statistics...18
3.3.3 Results...20
3.4 Robustness test for Model 1………22
4. Conclusion and Limitations………23
4.1 Conclusion………...23 4.2 Limitations………...24 4.3 Further Research………..25 References………26 Appendix………..29 3
1. Introduction
Although the efficient market hypothesis is widely accepted, science discovers more and more anomalies which make it possible to obtain abnormal returns in the stock market. Rozeff and Kinney (1976) first presented to us the January effect, a seasonal anomaly which shows that the security prices in January increase more than in every other month of the year. This is contrary to the efficient market hypothesis, which excludes making profit in the financial markets due to price differences in securities.
Since the discovery of the January effect, the anomaly was taken further into research. For instance, Gultekin and Gultekin (1983) found international evidence for this phenomenon with their work. In their study they took the theory outside the U.S. and proved that the January effect was also present in other capital markets around the world. However, the most research on this effect was conducted between 20 and 30 years ago. Also, they focused mostly on the January effect in developed economies. This is the case while it is just interesting for investors to know if the January effect is still present as the way it was 30 years ago. Also it is interesting for investors to know if the emerging economies experience the same January effect as the developed economies. If, for example, the January effect is much bigger for the emerging economies it would be wise for investors to invest their money there, instead of in the developed economies, to take full advantage of the January effect with their investing strategy. Such a strategy, for example, would be to buy these low-priced stocks in other months than January. Then, in January, these investors could sell their high-priced stocks and earn a lot of money via this difference in price. Therefore, the first goal of this research is to look if the January effect is also present in the emerging economies. It does this for the time period 2001-2012.
Furthermore, the time period 2001-2012 can be divided into two sub-periods, namely: One where there was no financial crisis (2001-2006) and one time period during a financial crisis (2007-2012). Past research did not make this difference, while it just would be good for investors to know if their investing strategy with respect to the January effect still works in times of the financial crisis. Therefore, the second goal of this research is to investigate for both emerging and developed economies if there is a difference for the January effect before and during a financial crisis.
Finally, there is still ambiguity if there is a difference in magnitude of the January effect for developed and emerging economies. Thus, the third and last objective of this study is to look if there is a difference in magnitude for the January effect between the emerging and developed economies in times of financial crisis.
This paper continues with an overview of the most important and relevant literature from the past about the January effect in section 2. In this section the hypotheses for this study will also be presented. Then in section 3 the data, methodology and results for this study are discussed. In section 4 conclusions are drawn and limitations of this study are given. The section ends with proposals for further research.
2. Literature review
This study is about the January effect in emerging and developed economies during times of financial crisis. Past research about the January effect has focused mostly on the reasons why the January effect exists, the size of the January effect and on which sort of stocks the January effect has this effect.
2.1. Reasons for the January effect
The first paper that presented the January effect to the world was that from Rozeff and Kinney (1976). They proved that for the period 1904-1974 a seasonal anomaly existed on the New York Stock Exchange (NYSE). For this period, there were abnormal large January mean returns on stocks compared to other months, and this difference in means proved to be
significant. When, in the same study, the same test was applied to the Australian stock market, they found something else. In the Australian stock market the stock returns were highest in the month July, with the month January coming on the second place. Possible explanations for this phenomenon were not tested in this paper, but this paper showed that the January effect was not globally present. It is possible that in other countries this effect on stock prices is present, but not in the month January.
The first possible explanation for the existence of the January effect was tested by Brown et al. (1983). They tested if the January effect was caused by a phenomenon called the tax-loss-selling principle. This principle implies that, at the end of the year, stockholders sell those stocks that declined in price during the year. In this way investors benefit from the opportunity to write-off capital losses against their normal income when computing their own payable taxes. According to this principle the pressure to sell stocks causes the prices of these stocks to decline. Then after December, so after the tax-year has ended, this pressure to sell stocks disappears, which causes the price of these stocks to returns to their normal level. This sudden jump in price in the month January could then be called ‘the January effect’. However,
Brown et al. (1983) failed to find enough evidence for this tax-loss-selling principle for both the Australian and the American stock market. They conclude that ‘the relationship between the U.S. tax year and the January seasonal may be more correlation than causation’ (Brown et al., 1983).
The same applies to other papers which test the relationship between the January effect and the tax-loss-selling principle. For instance, Berges et al. (1984) test this relationship for the Canadian stock market for the period 1951-1980. They distinguish two time periods, namely 1951-1973 and 1973-1980. The first period describes a period without a capital gains tax and the second a period after the introduction of this tax. However, both periods
experience a January effect, which leads to the conclusion that the tax-loss-selling principle cannot be the whole explanation of the January effect. The same conclusion was reached in the research of Tinic et al. (1987), which concluded that the tax-loss-selling hypothesis was only a part of the explanation of the January effect, and Jones et al. (1987), who showed that the January effect was already present in the U.S. before the income taxes were introduced.
Another explanation of the January effect was given by Ogden (1990). He proposed the turn-of-month liquidity hypothesis. This hypothesis states that investors receive a certain amount of cash receipts at the end of each month or year. With this extra cash in their pockets their demand for stocks increases at the beginning of the next month (or year), which drives the prices of these stocks up. Their hypothesis was confirmed, but again, the turn-of-the-month liquidity hypothesis is only a partly explanation for the January effect.
Finally, the most recent explanation for the January effect is the window dressing hypothesis, which was proposed by Dbouk et al. (2013). This hypothesis suggests that by the end of the year portfolio managers sell their most risky securities with the means to make their other bonds appear less risky. Then, the next month (January), they purchase the most risky securities to achieve the highest returns. This hypothesis was supported by their research which led to the conclusion that the January effect can partly be explained by the window dressing hypothesis.
Other possible explanations for this effect would then be the tax-loss-selling
hypothesis and the reversal effect. The reversal effect implies that stocks reverse course after a specific time period. So, the underperformers of the first period become the top performers in the second period and vice versa. In this case January would then be a period where this reversal of course takes place for stocks which were underperforming in the month December.
2.2 The January effect outside the U.S.
One of the first studies to test for the January effect outside of the U.S. was done by Gultekin and Gultekin (1983). They tested if the January effect was also present in eighteen other major industrialized countries. They found evidence of strong seasonality in stock market returns in most of these researched capital markets. In other words, the stock returns followed a certain pattern, which means that the highest stock returns were always on the same time of the year. This time was January for most countries, which showed evidence in favor of the presence of the January effect outside the U.S.
Another study which searched for evidence in favor of the January effect in another developed economy was that from Kato and Schallheim (1985). They conducted their research on the Japanese stock market and found that the January effect, together with the small size effect (see 2.3), was also present in the Japanese stock market. These two effects proved to be highly similar in the U.S. and Japan, which ‘may be indicative for well-integrated markets on an international scale’ (Kato and Schallheim, 1985). However, the similarities in the January effect for both countries are also a sign that the tax-loss-selling hypothesis is not the complete explanation for its existence, because the tax regimes of both countries are not similar.
The first study to examine if the January effect was also present in emerging stock markets was the study of Fountas and Segredakis (2002). They tested for seasonality in stock returns in eighteen emerging stock markets. They found evidence for the presence of a monthly seasonality in stock returns, but this was not in favor of the January effect. Also, no evidence was found for the tax-loss-selling principle. This implies that no abnormal returns can be gained from the difference in the price of securities over the months, which is in line with the efficient market hypothesis. However, this study was conducted for the relatively peaceful period 1987-1995, and does not take a financial crisis, like the one started in 2007, into account.
A study which does take a financial crisis into account is the one conducted by Balint and Gica (2012). They examine if the January effect is present on the Romanian stock market, and if so, if there is a difference for the January effect in times of financial crisis compared to a time period without financial crisis. The results are that the January effect was present on the Romanian stock market before the financial crisis. During the crisis, negative stock values were obtained and the January effect disappeared. Only for the firms with the smallest
capitalization the January effect was still present, which again can be linked to the small firm effect (see 2.3). However, this study was only concentrated on the Romanian stock market, so
the results about the January effect in times of financial crisis cannot be generalized for stock markets around the rest of the world.
In general, the major shortages of the most recent papers about the January effect are: They are outdated, they do not take a financial crisis into account and the results cannot be generalized across different countries or types of economies. These shortages will not be found in this study. This study works with data from the most recent period (2001-2012), which includes a time of financial crisis, and makes the comparison between emerging and developed economies (both represented by ten countries).
2.3 Other findings about the January effect
Reinganum (1983) proved in his research that the January effect is the most present for small firms. This is called the small firm effect, and is in line with the findings of Banz (1981) that small firms experience higher returns than big firms. This implies that firms with smaller market capitalization experience the highest rise in price in the month January. Reinganum (1983) also shows that these large returns in January are mainly concentrated in the first few trading days of the month. Because also prior year winning stocks experience these big returns in the month January, again the tax-loss-selling hypothesis is not the full explanation for the January effect.
On the other hand, another conclusion was reached by De Bondt and Thaler (1985). They found that prior year losing stocks experience a big January effect, while prior year winning stocks do not. In their opinion, the small firm effect is not present, and should be called ‘the losing firm effect’ (De Bondt and Thaler, 1985). Therefore their conclusion is that small market capitalization is not related with the January effect. Only if a firm is making a loss determines if it experiences the January effect. A possible explanation for the phenomena is the overreaction hypothesis, which states that investors overreact on dramatic news. A losing firm would then expect higher positive returns than a winning firm. However, this explanation was not tested in this paper.
Another interesting finding was reached by Moller and Zilca (2008). They discovered that the duration of the January effect was shorter for the period 1995-2004 than it was in the past (1965-1994). On the one hand, the abnormal returns in the second part of January were lower than in the past, while on the other hand the abnormal returns in the first part of January had risen. These two opposite effects kept the total magnitude of the January effect
unchanged. Finally, they discovered that the total trading volume in the second part of
January declined substantially, while the total trading volume in the first part of January remained unchanged.
2.4 Link to Hypotheses
Given the research of Gultekin and Gultekin (1983) and Kato and Schallheim (1985), it is clear that the January effect is present in most developed economies around the world. However, Fountas and Segredakis (2002) failed to find enough evidence in favor of the January effect for the emerging economies. This would imply that strategies for investors to take advantage of the January effect will only work on stock markets of developed economies. Such a strategy, for example, would be to buy these low-priced stocks in other months than January. Then, in January, these investors could sell their high-priced stocks and earn a lot of money via this difference in price. However, following Fountas and Segredakis (2002), this strategy could only be used in developed economies.
The research of Fountas and Segredakis (2002) is about the period 1987-1995. This time period differs significantly with the current time period. According to Mullin (1993) and Diacogiannis and Segredakis (1996) the emerging markets’ growth rate was much higher than the growth rate from the developed economies in the 1990’s and 2000’s. A logical
consequence would be that in this period the capital markets of these economies also have developed. In other words, the emerging economies are closing in on the developed economies. Since the developed economies experience a January effect it is likely that the emerging economies are now also experiencing this for the most recent time period. However this was not studied yet in previous research. Therefore the first hypothesis of this study is:
Hypothesis 1: The January effect is present in emerging economies for the period 2001-2012.
The second objective of this study is to compare the January effect before and during the financial crisis. This comparison will be made for both developed and emerging economies. Balint and Gica (2012) have performed a rather similar study on the Romanian stock market. They found that the January effect was not present during the financial crisis, while it still was before the crisis. Only for firms with the smallest market capitalization the January effect was still present during the financial crisis. However, these results only apply to the Romanian stock market. Other stock markets around the world differ with the Romanian stock market with respect to trading volumes, trading laws, trading options etc. Therefore, the results of the study of Balint and Gica (2012) do not have to apply to other stock markets around the world.
If economic reasoning is used, the conclusion about the January effect in times of financial crisis will be different from the conclusions of Balint and Gica (2012). It is certain that during times of financial crisis the risk of bankruptcy of firms will be higher. In other words, the general risk of investing will be higher in times of financial crisis. Tinic and West (1984) show in their paper that the month January is the only month of the year which contains a significant positive relationship between risk and return. The other months do not. This theory implies that when the risk goes up in times of financial crisis, this will lead to an increase in return in the month January. In the other months the returns do not rise with an increase in risk caused by the financial crisis. This increase in return in January, due to the increase in risk caused by the financial crisis, will cause the January effect to rise as well. Therefore the second hypothesis of this paper will be:
Hypothesis 2: The January effect will be larger during the financial crisis for both developed
and emerging economies than it was in times without financial crisis.
The third objective of this paper is to compare the possible January effect between developed and emerging economies during times of financial crisis. Again, here the economic theory applies. During times of financial crisis, the risk of investing increases. Reinganum (1983) found that the January effect is more present with small firms than with big firms. This could be explained by the lack of information about the prospects of small firms. Lack of
information increases the risk of investing in small firms which increases the returns of the stocks of those firms in January (Tinic and West, 1984). This situation is not different when developed and emerging economies are compared with each other. Because there is less information about stocks form emerging markets, this increases the risk of investing in them. This increase in risk than increases the returns from those stocks when compared to stocks from developed markets. This reasoning leads to the third hypothesis of this paper:
Hypothesis 3: The magnitude of the January effect will be larger during the financial crisis
for securities from emerging economies than for securities from developed economies.
3. Data, Methodology & Results
3.1.1 Data and sample
The purpose of this study is to compare a possible January effect between emerging and developed economies. This is done by analyzing the stock returns from two stock portfolios, whereby the first portfolio represents the developed economies and the second portfolio represents the emerging economies. These portfolios are constructed by taking the ten largest stock indexes from both the developed countries and the emerging countries via classification by the MSCI (2014). The ten stock indexes which represent the developed economies are: NYSE (U.S.), Nikkei 225 (Japan), FTSE 100 (U.K.), TSX 60 (Canada), CAC 40 (France), S&P/ASX 200 (Australia), DAX 30 (Germany), HSI (Hong Kong), STI (Singapore) and SMI (Switzerland). On the other hand, the ten stock indexes which represent the emerging
economies are: SSE 50 (China), BSE 30 (India), MICEX 30 (Russia), FTSE/JSE 40 (South Africa), BOLSA (Mexico),FTSE Bursa Malaysia KLCI 30 (Malaysia),IDX (Indonesia), BUX (Hungary), KOSPI (South Korea) and IGPA (Chile). Originally, the Ibovespa (Brazil) was also chosen to represent the emerging economy portfolio, but because data about daily stock returns from this index were unavailable the Ibovespa was replaced by the IGPA. Data about the daily stock prices from these twenty stock indexes was gathered through DataStream. This was done for the period 2001-2012. Next, this time period can be split up into two separate time periods. On the one hand 2001-2006 (which represents a time period without financial crisis), and on the other hand 2007-2012 (which represents a time period of financial crisis). The year 2007 was chosen as the begin of the financial crisis, because in this year the U.S. housing bubble burst. This had a negative impact on the global economy and was followed by a time period of global recession, and a time period of global recovery.
Data from monthly inflation rates in all chosen countries were collected via the World Bank. Geske and Roll (1983) prove that there is a general relationship between inflation rates and stock returns. Similar results were found in Japan by Hiraki (1985) and in Brazil by Pimentel and Choudhry (2014). However, the exact nature of this relationship between monthly stock returns and monthly inflation rates is not yet known. For this reason, monthly inflation rates are only taken as a control variable in the used model of this study. This is done to control for potential omitted variable bias in the regression.
Third, data about real interest rates per year was collected via the World Bank. Real interest data was used, because real interest is corrected for inflation. Therefore, real interest gives a better view of how high the interest rate actually was and how this really was
experienced by the investors. Pimentel and Choudhry (2014) show that for emerging markets, in this case Brazil, interest rates do have an effect on changes in stock returns. This effect is mainly present in times of high inflation. Again, data about real interest rates is only taken as a control variable in the model of this study.
Finally, data about changes in exchange rate of the chosen country was gathered from OANDA. Bartram et al. (2012) show that exposure of changes in exchange rate have a significant impact on stock returns. So, because changes in exchange rate can have an impact on stock returns, a control variable for changes in exchange rate was created within the used model of this study. Because the U.S. dollar is the most exchanged currency in the world, data about changes in exchange rate with respect to the dollar are used in the model of this study.
3.1.2 Test for correlation in control variables
In Table 1 the correlation-test between the control variables of Model 2 is shown. This is done to check for potential imperfect/perfect multicollinearity. Perfect multicollinearity means that two or more independent variables in the regression are perfectly linear, which makes it impossible to estimate the regression. Imperfect multicollinearity, on the other hand, is a situation where two independent variables are highly (but not perfectly correlated). High imperfect multicollinearity causes the estimators of the model to be estimated with less precision. This means that the estimators of the variables in the model are less close to their true value, than they would be if there was no imperfect collinearity.
Table 1 shows that all control variables are correlated with each other, although no single correlation is significantly high enough to cause a very high imperfect
multicollinearity. However, because all control variables are correlated with each other the results of the regression of model 2 have to be analyzed with caution. The multicollinearity (however how small) causes the estimators of the model to be estimated with larger standard errors. This results in estimators with less precision. However, the outcome of the regression model is not affected by the multicollinearity.
Table 1
3.2 Methodology, Descriptive Statistics and Results for the first hypothesis
3.2.1 Methodology
To see if the January effect is present in emerging economies the rate of return of all stock indexes of these countries are needed. According to Reinganum (1983) and Keim (1983) the January effect is the strongest on the first trading days of this month. Lakonishok and Schmidt (1984) take this even a step further and show that the biggest returns of the year were
generated on the last day of December and the first four days of January. Moller and Zilca (2007), on their turn, show that the focus of the January effect lays in the first part of January, while the January effect in the second part of January is diminishing. Because previous research indicates that the January effect is mainly due the high returns in the first days of January, this research calculates the returns of the stock indexes only for the first trading week of each month. The returns of the stock indexes are calculated via the following formula:
Rt= 1 1 − − − t t t P P P (DeMarzo 2007) Where: = t R Return in month t
Pt = price of index in month t on the last day of the first trading week
Pt-1 = price of index in month t on the first day of the first trading week
Important to note here is that the calculated returns are not cumulative returns, but absolute returns. So, the returns are calculated by calculating the percentage difference in price between the first and last day of the first trading week each month.
exeme -0.0284 -0.1224 0.1798 -0.1820 -0.0687 1.0000 exdev 0.3027 0.0082 -0.0086 -0.0006 1.0000 emerii -0.2310 -0.3795 0.1395 1.0000 devrii -0.1635 -0.2504 1.0000 emeinf 0.1378 1.0000 devinf 1.0000 devinf emeinf devrii emerii exdev exeme
For testing for the January effect in emerging economies the same method is used as in Jones et al. (1987). This study uses the following model (Model 1) in an OLS-regression to test for a potential January effect:
𝑅𝑅𝑅𝑅 = 𝛽𝛽0 + 𝛽𝛽1𝐽𝐽𝐽𝐽𝐽𝐽 + 𝜀𝜀 Where:
Rt = stock returns in month t
JAN is a dummy variable with JAN =1 if month is January and 0 if otherwise ε = error term
To see if the January effect was present in the emerging economies of this study for the period 2001-2012 the following hypothesis is tested:
H0: β1 = 0, H1: β1>0
If it is proven that β1 in this model is significantly greater than zero, this would mean that the month January had a positive effect on stock returns, and thus that a potential January effect is present in the emerging economies.
For this model data from multiple entities and multiple time periods are used, which makes it panel data. To control for influences from omitted variables that differ between entities, while being constant over time, fixed effects are added to the regression.
3.2.2 Descriptive Statistics
First, the mean returns of every month of the period 2001-2012 for every chosen country were calculated. The results are shown in Table 2:
Table 2
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
China 0.69 1.66 0.04 0.70 0.01 -1.16 0.03 -0.30 0.15 -0.43 0.4 1.64 India 1.85 0.64 -1.31 1.54 1.09 0.25 -0.05 0.34 0.53 -0.14 0.85 1.21 Russia 1.59 -0.3 0.77 1.05 -0.41 0.70 0.11 -0.25 -0.64 -1.28 0.08 1.28 South-Africa 0.78 -0.30 0.15 0.67 0.44 -0.26 -0.74 -0.62 -0.76 -0.27 0.44 -0.31 Mexico 1.26 -0.49 0.15 0.65 0.76 -0.38 -0.18 -0.73 -0.24 -0.53 0.76 1.32 Malaysia 0.71 0.5 -0.52 -0.17 0.55 0.17 0.95 -0.70 -0.08 0.38 -0.2 0.29 Indonesia 1.6 -0.64 0.55 0.82 0.3 0.21 0.26 -1.58 0.69 0.26 0.55 1.06 Hungary 0.73 -0.89 -0.22 1.95 0.07 0.14 0.31 -0.99 -0.64 -0.12 0.13 0.66 South-Korea 2.86 -0.44 -0.05 1.41 -0.03 -0.56 0.55 -1.2 -0.74 -0.15 0.86 0.56 Chile 0.56 -0.18 0.31 0.42 0.36 -0.06 -0.48 -0.91 -0.81 -0.38 -0.64 0.09 14
A first glance at this table leads to a couple of insights about the mean returns in emerging economies. First, only in India, Russia, South Africa, Indonesia, South Korea and Chile, January has the highest mean returns compared to other months. Having the highest return does not have to imply directly that a January effect is present. A test is needed to test if January has a significant positive impact on the mean index returns. This test is done via OLS-regression and will be discussed further up this section. Second, it can be seen that January always ends on the first or second place when it comes to mean returns for all countries except China. This also raises the expectations of a potential January effect for the emerging economies. This is also illustrated in Table 3:
Table 3
Here again, it is shown that January has the highest mean return overall for developing countries (see also Appendix A). Further, it shows that the month January has the highest maximum and minimum return of all months. This could also be an indicator for a potential January effect.
3.2.3 Results
Results for the regression of Model 1 can be found in table 4 below. For all ten emerging economies monthly data was used for the period 2001-2012. This means that for every economy 144 results were used. It was done for all ten countries and at different time periods, which makes it panel data. To control for influences from variables outside the model which change over time, the model is estimated with a fixed effects regression. In this way it is
december 10 .7786 .6240113 -.311 1.641 november 10 .3227 .4852445 -.643 .858 october 10 -.2657 .4578343 -1.278 .383 september 10 -.2538 .5600625 -.81 .69 august 10 -.6982 .5394361 -1.587 .34 july 10 .0752 .4900378 -.747 .95 june 10 -.095 .5238657 -1.162 .705 may 10 .3166 .4331836 -.415 1.09 april 10 .9071 .6098764 -.178 1.95 march 10 -.0116 .5863911 -1.31 .773 february 10 .0797 .7638131 -.898 1.662 january 10 1.2668 .7278066 .562 2.866 Variable Obs Mean Std. Dev. Min Max
possible to cope with panel data. Furthermore, robust standard errors are used in this regression to deal with heteroskedasticity.
To test for a potential January effect in the emerging economies, this dummy variable for January should be significantly greater than zero. Looking at the p-value for the dummy JANUARY (0.000), the conclusion is that the dummy variable for January is significantly greater than zero for the 5% and 1% confidence level. This means that H0 was rejected at these confidence levels. This, in turn, leads to the conclusion that the month January had a positive effect on stock returns for the emerging economies, or in other words, that the January effect was present in the emerging economies for the period 2001-2012.
For economically significance the following can be concluded: The coefficient for JANUARY was significant for the 1% confidence level, which means that the month January had a significant influence on stock returns. This means for the coefficient that the month January causes the stock return to rise with 1.15%, when compared to other months of the year. In other words, January has a positive economically significant effect on stock returns.
Table 4
Return Coefficient Robust Std. Error t-value p-value 95% Confidence Interval JANUARY 0.0115425 0.0021812 5.29* 0.000 0.0066083 – 0.164767 Constant 0.001121 0.0001818 6.17* 0.000 0.0007098 – 0.0015322 * Significant at the 1% confidence level
3.3 Methodology and results for the second and third hypothesis
3.3.1 Methodology
After the test for a potential January effect in the emerging economies for the period 2001-2012, this paper studies two other aspects of this effect. First, this study looks if there is a difference in the January effect before and during the financial crisis. In particular, it studies if the January effect was larger is times of financial crisis. It does this for both developed and
emerging economies. Second, it compares the magnitude of the January effect between developed and emerging economies, or to be specific, whether the magnitude of the January effect was bigger for emerging economies than for developed economies in times of the financial crisis. The model which examines these two aspects is generated via an OLS-regression and looks as follows:
𝑅𝑅𝑅𝑅 = 𝛽𝛽0 + 𝛽𝛽1𝐶𝐶𝑅𝑅𝐶𝐶 + 𝛽𝛽2𝐶𝐶𝑅𝑅𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 + 𝛽𝛽3𝐶𝐶𝐽𝐽𝐼𝐼 + 𝛽𝛽4𝐶𝐶𝐽𝐽𝐼𝐼 + 𝛽𝛽5𝐶𝐶𝐸𝐸𝐶𝐶𝐸𝐸 + 𝜀𝜀 Where:
Rt = stock return in January
CRI= dummy variable with CRI= 1 for period 2007-2012 and 0 otherwise EME= dummy variable with EME= 1 for emerging countries and 0 otherwise INF= control variable for inflation
INT= control variable for interest rate EXCH= control variable for exchange rate 𝜀𝜀 = error term
This model will first look if the January effect was larger for both developed and emerging economies in times of financial crisis. It does this by testing the following hypothesis:
H0: β1= 0 H1: β1>0
If β1proves to be positively significant this implies that the January effect was positively influenced by the financial crisis.
Next, model 2 looks if, in times of financial crisis, the January effect was bigger for emerging economies than for developed economies. The following hypothesis will be tested to
investigate this:
H0: β2= 0 H1: β2>0
If β2 proves to be positively significant this indicates that the January effect is positively
influenced by the fact that the economy is an emerging one. In other words, if β2 is positively 17
significant, this means that the January effect was bigger for emerging economies than for developed economies in times of financial crisis.
For Model 2 data from multiple entities and multiple time periods are used, which makes it panel data, just like Model 1. To control for influences from omitted variables that differ between entities, while being constant over time, again fixed effects are added to the regression.
3.3.2 Descriptive Statistics
The second and third hypotheses make the comparison between sort of economies and time periods for the January effect. In Table 5 are the mean returns for every month given for the developed economies like they were presented for the emerging economies in section 3.2.2. When looking at Table 5 and 6, the following things stand out. First, that January only has the highest mean return for three developed economies (U.K., Germany and Singapore) instead of the six economies that had the highest mean return in January for the emerging economies. When looking at mean returns for developed indexes overall, Table 6 also shows that April has a higher mean return than January (see also Appendix A). This could be an indicator for a possible April effect in developed economies, instead of a January effect.
Table 5
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
U.S. 0.73 -0.76 -0.09 0.64 -0.4 -0.1 -0.53 -1.65 -0.37 -0.37 0.77 0.88 Japan 0.85 -0.23 -0.02 2.09 0.32 -0.44 -0.06 -1.04 -0.73 -0.38 1.08 0.74 U.K. 0.99 -0.17 -0.15 0.9 0.24 -0.41 -0.25 -1.01 -0.98 0.49 0.87 0.35 Canada 0.86 -0.22 0.2 0.27 0.91 -0.08 -0.56 -0.68 -0.29 -1.27 0.99 0.06 France 0.54 -1.28 -0.24 0.88 -0.54 -0.66 -0.62 -1.33 -1.15 0.13 0.83 0.48 Australia 0.24 -0.08 -0.13 1.47 -0.03 -0.14 -0.35 -0.79 -0.87 0.03 0.23 0.09 Germany 1.28 -0.74 0.04 1.6 -0.53 -0.4 0.05 -1.85 -1.18 0.14 1.04 0.58 Hong-Kong 2.16 0.4 -1.15 1.77 1.23 -0.07 0.52 -1.13 -0.49 0.32 0.3 0.83 Singapore 2.13 0.09 -0.14 1.95 1.25 0.54 0.77 -1.8 -0.61 0.61 -0.38 0.98 Switzerlan d 0.87 -0.93 -0.69 0.89 0.18 -0.31 -0.54 -1.97 -0.55 0.04 0.63 0.38 18
Table 6
Table 7 shows the results of the regression in Model 1 for developed economies. This was done to make sure that the developed economies experienced a January effect at all. Knowing that both emerging and developed economies experienced this effect makes it easier to
compare them, and to test hypotheses 2 and 3 of this study.
Here, it can be seen that the dummy variable for the month January has a significant positive effect for the returns of the indexes (p-value of 0.000). This implies that the January effect was also present in the developed economies for the period 2001-2012.
Table 7
Return Coefficient Robust Std. Error t-value p-value 95% Confidence Interval January 0.0109869 0.0015635 7.03* 0.000 0.0074501 – 0.0145237 Constant -0.000388 0.0001303 -2.98** 0.016 -0.0006827 - -0000932 * Significant at the 1% confidence level
** Significant at the 5% confidence level
Further, table 8 shows some descriptive statistics with respect to the control variables of model 2. It shows that the mean inflation in January for emerging economies is higher (1.98)
dec 10 .537 .3219748 .06 .98 nov 10 .636 .460198 -.38 1.08 oct 10 -.026 .5417092 -1.27 .61 sep 10 -.722 .3141054 -1.18 -.29 aug 10 -1.325 .4647162 -1.97 -.68 jul 10 -.157 .4805564 -.62 .77 jun 10 -.207 .3261578 -.66 .54 may 10 .263 .6780372 -.54 1.25 apr 10 1.246 .61085 .27 2.09 mar 10 -.237 .3949135 -1.15 .2 feb 10 -.392 .5160276 -1.28 .4 jan 10 1.065 .6306831 .24 2.16 Variable Obs Mean Std. Dev. Min Max
than for developing economies (1.49). This means that the average price increase per month for goods was higher in the emerging economies than in the developing economies. The mean real interest rate per year, on the other side, is slightly higher for developed economies than for emerging economies (3.08 to 2.97), while this is also the case for the change in exchange rate with respect to the U.S. dollar in the month January (0.65 to 0.18).
Table 8
Variable Mean Std. Dev. Min Max
DEVINF 1.49225 1.31374 -1.6 6 EMEINF 1.975833 3.72561 -0.8 20.7 DEVRII 3.078833 2.360725 -2.5 11.7 EMERII 2.971667 3.914897 -7.3 13.1 DEVEX 0.6537037 2.228802 -4.5 6.9 EMEEX 0.1833333 3.061258 -12 7.7 3.3.3 Results
Results of the regression of Model 2 for hypothesis 2 of this study are shown in table 9:
Table 9
Return Coefficient Robust Std. Error t-value p-value 95% Confidence Interval CRI -0.0091054 0.0037571 -2.42* 0.026 -.0169692 - -0.0012417 INF 0.0003178 0.0017379 0.18 0.857 .0033197 -0.0039554 RII -0.001089 0.0005001 -2.18 0.042 -.0021358 - -0.0000423 EX -0.0002488 0.000648 -0.38 0.705 -.001605 - .0011075 Constant 0.018536 0.0046301 4.00 0.001 .008845 -.0282269
*Significant at the 5% confidence level
Where CRI represents the dummy variable for the financial crisis and INF,RII and EX represent the control variables for inflation, real interest rate and exchange rate.
It shows if the January effect was stronger in times of financial crisis for developed and emerging economies. However, it can be seen that the coefficient for the variable for the financial crisis is a negative value (-0.0091). This can indicate that the financial crisis had a negative impact on the January effect for both sort economies. In other words, during the financial crisis the January effect was 0.91% lower. Looking at the p-value for CRI (0.026) shows that this negative relationship is significant at the 5% confidence level, but not at the 1% confidence level. Hypothesis 2 of this study can now be rejected, or in other words, there is not enough evidence to show that the financial crisis had a positive effect on the January effect. Instead there was found enough evidence to proof a negative relationship between the financial crisis and the January effect at the 5% significance level. This negative relationship is valid for both emerging and developed economies.
Results of the regression of Model 2 for hypothesis 3 of this study are shown in table 10:
Table 10
Return Coefficient Robust Std. Error t-value p-value 95% Confidence Interval CRI -0.008343 0.0030597 -2.73 0.013 -0.0147469 - -0.001939 CRIEME -0.0014838 0.0071648 -0.21 0.838 -.0164799 - .0135122 INF 0.0002833 0.0017526 0.16 0.873 -.0033851 - .0039516 RII -0.0010676 0.0004956 -2.15 0.044 -.0021049 - -.0000303 EX -0.0002312 0.0006421 -0.36 0.723 -.0015752 - .0011127 Constant 0.0185143 0.0046236 4.00 0.001 .0088369 - .0281917 21
Where CRI represents the dummy variable for the financial crisis, CRIEME represents the interaction effect between the dummies for the financial crisis and the emerging economies and INF,RII and EX represent the control variables for inflation, real interest rate and exchange rate.
The same reasoning applies for the third hypothesis of this study. Here it was expected that the dummy variable for emerging economies would take a positive value, which implied that the January effect was bigger, in times of financial crisis, for emerging economies than for developed economies. A look at table 10 shows that this coefficient is negative(-0.0014), which means that the January effect was bigger for developed economies than for emerging economies during the financial crisis. In other words, during the financial crisis the January effect was 0.14% lower for the emerging economies than for the developed economies. However, looking at the p-value for the variable for emerging economies (0.838), yields the following conclusion: There is not enough evidence at the 10% confidence level to proof a significant negative relationship between the classification as an emerging economy and the January effect during the financial crisis. For the third hypothesis of this study this means that there was not enough evidence for H1 to be true, and that H1 should be rejected in favor of H0.
3.4 Robustness test for Model 1
In this section the robustness of the results for the tests of the first hypothesis are tested. It can be seen in Table 2 that the mean returns for the indexes in emerging economies are the highest in the months January, April and December. In this section it will be tested if there was also an April and December effect present in the period 2001-2012 for the emerging economies. The results for these tests can be found in table 11 and table 12.
Table 11 (test for April effect)
Return Coefficient Robust Std. Error t-value p-value 95% Confidence Interval April 0.0076257 0.002003 3.81 0.004 0.0030947 - 0.0121568 Constant 0.0014474 0.0001669 8.67 0.000 0.0010698 - 0.001825 22
Table 12 (test for December effect)
Return Coefficient Robust Std. Error t-value p-value 95% Confidence Interval December 0.0062688 0.0017123 3.66 0.005 0.0023954 -0.0101423 Constant 0.0015604 0.0001427 10.94 0.000 0.0012377 - 0.0018832
It can be seen that the April effect and December effect are also present in the emerging economies at the 1% confidence level. This shows that the January effect was not as clearly present as section 3.2.3 showed. This means that the conclusions taken in section 3.2.3 should be analyzed with caution. Further research, with more detailed data, should investigate if these effects are also present when more detailed data is used. It is possible that the January effect disappears in favor of a potential April of December effect, when other data was used. This implies that the results obtained in section 3.2.3 are still true, but that it should be considered that the January effect was only as present as the way it was due the dataset used. An actual April or December effect for the emerging economies during the period 2001-2012 should definitely not be excluded yet.
4. Conclusion and Limitations
4.1 Conclusion
This study began to look at the mean returns of the ten largest stock indexes of the largest emerging economies following MSCI. The first goal was to look if the January effect was present in these emerging economies for the period 2001-2012. The data from these ten different stock indexes showed via an OLS-regression that January had a significant positive effect on the stock returns of these indexes. So, in this study there was enough evidence found for the presence of a January effect for these emerging economies in the time period 2001-2012. However, as the results of the robustness tests show, also the months April and December proved to have a significant positive relationship with the stock returns of the
emerging stock indexes. This means that in the time period 2001-2012 also an April effect and a December effect were present in the emerging economies.
Further, this paper studied if the January effect was stronger in times of financial crisis for both sort of economies. The results of this research were not as expected beforehand. The regression showed that the coefficient for the dummy variable for the financial crisis was negative. This would indicate that an increase in risk (due to the financial crisis) is not
followed by an increase in return (a stronger January effect). Actually, the negative coefficient for the dummy variable for the financial crisis would imply that there is a negative
relationship between the financial crisis and the January effect, or in other words, that the January effect was stronger in times when there was no financial crisis. This is in line with the finding of Balint and Gica (2012), who found similar results on the Romanian stock market.
Finally, this study investigated if the January effect was stronger for emerging
economies than for developed economies during the financial crisis. Again, the results of this test were surprising. The coefficient of the dummy variable for the emerging economies was negative. This would imply that emerging economies experienced a weaker January effect than developed economies during the financial crisis. The difference in available information between small and big firms is thus not the same as the difference in available information between emerging and developed economies. The increase in risk due to less information was not followed by a stronger January effect. However, the negative relationship between the dummy variable for emerging economies and the January effect was not significant. Thus, there is not enough evidence to proof that being an emerging economy leads to a decrease in the January effect, when compared to a developed economy during the financial crisis.
4.2 Limitations
This study has some limitations, which will be mentioned in this section. First, data was gathered only from the ten biggest stock indexes from both the emerging and developed economies. Reinganum (1983) found that the January effect was mostly present for small firms instead of big firms. However, in this study only the stock prices of firms of the biggest stock indexes are included. These firms are typically not the smallest firms. Thus, the small firms are not enough represented in the sample of this study. This leads to a biased estimator for the returns of the stock indexes.
Also both models of this study potentially suffer from omitted variable bias. There are several other factors which influence the stock returns of the chosen indexes. These factors are not included in the models. Because they are omitted from the regression, the
results of the regressions for the OLS-estimator for stock return are biased. For this to be true, the omitted variables have to be correlated with a regressor which was included in the
regression.
Also, the estimated relationships in this study do not have to be linear. When this is actually not the case (for example when the relationships were exponential), this would imply misspecification of the functional form of the regression function. This leads to biased OLS-estimators. However, further tests are needed to find evidence for misspecification of the functional form of the regression functions.
4.3 Further research
First, future research should solve the underrepresentation of the small firms that was present in this study. When stock prices of small firms are also included in the sample this would lead to better results for the study of the January effect for both the developed and emerging economies. It would give a clearer view of the real situation about the presence of the January effect for this period and these economies.
Also, further research should focus on the difference in magnitude of the January effect for different time periods. In this study only one time period was used to investigate the January effect, namely the first trading week of each month. Further research should
investigate if the January effect differs for other time periods (for example the first-two-trading-week period, or the last trading week period of each month). This would give insights about when the January effect is on its strongest, and how long it lasts. Also, further research should investigate if this differs in times of financial crisis.
Another option for further research is to look for possible explanations for the
differences in the January effect between emerging and developed economies. Also, it is still not clear why the January effect differs in magnitude in times of financial crisis, compared to times when there was no crisis.
References
Balint, C. and Gica, O. (2012). ‘Is the January effect present on the Romanian capital market?’ Procedia- Social and Behavioral Sciences, vol. 58, pp. 523-532.
Banz, R.W. (1981). ‘The relationship between return and market value of common stocks’ Journal of Financial Economics, vol. 9, pp. 3-18.
Bartram, S.M., Brown, G.W. and Minton, B.A. (2010). ‘Resolving the exposure puzzle: The many facets of exchange rate exposure’ Journal of Financial Economics, vol. 95, pp. 148-173.
Berges, A., McConnell, J.J. and Schlarbaum G.G. (1984).’The turn-of-the-year in Canada’ Journal in Finance, vol.39, pp. 185-192.
Brown, P., Keim D.B., Kleidon, A.W. and Marsh, T.A. (1982). ‘Stock return seasonalities and the ‘tax-loss-selling hypothesis’: Analysis of the arguments and Australian evidence’ Journal of Financial Economics, vol 12.
Dbouk, W., Jamali, I. and Kryzanowksi L. (2013). ‘The January effect for individual corporate bonds’ International Review of Financial Analysis, vol 30, pp. 69-77.
Diacogiannis, G.P. and Segredakis, K.N. (1996) ‘The Athens stock exchange and the emerging equity markets’ Hellenic Bank Association, pp. 132-141.
De Bondt, W. F. M. and Thaler, R.H. (1985). ‘Further evidence on investor overreaction and stock market seasonality’ Journal of Finance, vol. 42, pp. 557-581.
Fountas, S. and Segredakis, K.N. (2002). ‘Emerging stock markets seasonalities: the January effect and the tax-loss-selling hypothesis’ Applied Financial Economics, vol 12, pp. 291-299.
Geske, R. and Roll, R. (1983). ‘The fiscal and monetary linkage between stock returns and inflation’ Journal of Finance, vol. 38, pp. 1-33.
Gultekin, M.N. and Gultekin N.B. (1983). ‘Stock market seasonality: International evidence’ Journal of Financial Economics, vol. 12, pp. 469-481.
Hiraki, T. (1985). ‘Testing the proxy effect hypothesis of inflation on stock returns for the Japanese market’ Quarterly Journal of Business and Economics, vol. 24, pp. 73-87.
Jones, C.P., Pearce, D.K. and Wilson, J.W. (1987). ‘Can tax-loss selling explain the January effect? A Note.’ Journal of Finance, vol. 42, pp. 453-461
Kato, K. and Schallheim, J.S. (1985). ‘Seasonal and size anomalies in the Japanese stock market’ Journal of Financial and Quantitative Analysis, vol. 20, pp. 243-260.
Keim, D.B. (1983). ‘Size-related anomalies and stock return seasonality’ Journal of Financial Economics, vol. 12, pp. 13-32.
Lakonishok, J. and Smidt, S. (1984). ‘Volume and turn-of-the-year behavior’ Journal of Financial Economics, vol. 13, pp. 435-455.
Moller, N. and Zilca, S. (2008). ‘The evolution of the January effect’ Journal of Banking & Finance, vol. 32, pp. 447-457.
MSCI (2014, May 23). Developed market classification. Retrieved from:
http://www.msci.com/products/indexes/country_and_regional/dm/
MSCI (2014, May 23). Emerging market classification. Retrieved from:
http://www.msci.com/products/indexes/country_and_regional/em/
Mullin, J. (1993). ‘Emerging equity markets in the global economy’ Federal Reserve Bank of New York Quarterly Review, pp. 54-83.
OANDA (2014, May 23). Historical exchange rates. Retrieved from:
http://www.oanda.com/currency/historical-rates/
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, vol. 42, pp. 1259-1272.
Pimentel, R.C. and Choudhry, T. (2014). ‘Stock returns under high inflation and interest rates: Evidence from the Brazilian market’ Emerging Markets and Finance Trade, vol. 50, pp. 71-92.
Reinganum, M.R. (1983). ‘The anomalous stock market behavior of small firms in January: Empirical tests for tax-loss selling effects’ Journal of Financial Economics, vol. 12 pp. 89-104.
Rozeff, M.S. and Kinney, W.R. (1976). ‘Capital market seasonality: The case of stock returns’ Journal of Financial Economics, pp. 379-402.
Thaler, R.H. (1987). ‘Anomalies, the January Effect’ Economic Perspectives, vol. 1, pp. 197-201.
Tinic, S.M., Barone-Adesi, G. and West, R.R. (1987).’Seasonality in Canadian Stock Prices: A test of the tax-loss-selling hypothesis’ Journal of Finance and Qualitative Analysis, vol. 22, pp. 51-63.
Tinic, S.M. and West, R.R. (1984). ‘Risk and Return: January vs. the rest of the year’ Journal of Financial Economics, vol. 13, pp. 561-574.
World Bank (2014, 23 May). Inflation rates. Retrieved from:
http://data.worldbank.org/indicator/NY.GDP.DEFL.KD.ZG
World Bank (2014, 23 May). Real interest rates. Retrieved from:
http://data.worldbank.org/indicator/FR.INR.RINR
Appendix A
Mean monthly index returns of emerging economies
Mean monthly index returns of developed economies
-. 5 0 .5 1 1 .5
mean of january mean of february mean of march mean of april mean of may mean of june mean of july mean of august mean of september mean of october mean of november mean of december
-2
-1
0
1
mean of jan mean of feb
mean of mar mean of apr
mean of may mean of jun
mean of jul mean of aug
mean of sep mean of oct
mean of nov mean of dec
