Turn-of-the-Year Effect in Asia Pacific Stock Markets: New Evidence

This paper examines a well-known seasonal anomaly - the turn-of-the-year (TOY) effect in fifteen Asia Pacific stock indices by using an updated dataset and forward-looking methods. The analysis utilizes daily dataset that spans from January 2000 to December 2018. Applying the Ordinary Least Square (OLS) regression and EGARCH approach, the results of this paper suggest that the TOY effect becomes detectable again after the GFC in developed stock markets with tax year not ending in December. Oppositely, the magnitude of this anomaly has diminished in the emerging financial markets after the GFC, which is consistent with the EMH. The evidence of the leverage effect in the unconditional volatility is proposed that volatility in negative shocks is considerably higher than that of positive shocks across examined stock indices. This phenomenon is more conspicuous in mature stock indices compared to emerging indices. The positive connection between the leverage effect and stock market volatility is propositioned as diminishing magnitude of this effect during the stable market condition after the GFC. Our findings lend reinforcement to the conclusion that some Asia Pacific stock markets satisfy the weak form of the EMH.


Introduction
In modern finance theory, one of the keystones is the Efficient Market Hypothesis (EMH). The strongest form of EMH indicates that stock prices fully reflect all relevant information in an impartial manner (Fama, 1970). In an efficient market, it is very difficult for investors to obtain abnormal returns consistently. However, stock market anomalies are one of the violations of the EMH. The presence of stock market anomalies indicates that investors can obtain abnormal profit from predictable patterns in stock returns (Hoang, Phan & Ta, 2020). Stock market calendar anomalies or seasonal anomalies represent the occasions that abnormal returns emerge from certain periods in a calendar year (Officer, 1975). The turn-of-the-year (TOY) effect or January effect exposes that stock return in the first month of the tax year is relatively higher than other months of the year (Rozeff & Kinney, 1976). In other words, this phenomenon indicates a systematic predictable pattern of stock returns. This pattern contradicts EMH that future stock price movements are unpredictable (Fama, 1970). Exclusively, this anomaly has been investigated by a large number of theoretical and empirical papers around the globe from the 1980s, which encourages investors to exploit the abnormal returns from the mispricing prospects. As a consequence, the existence of this anomaly on stock return is inconsistent with traditional asset pricing models as patterns of stock return could not be rationalised by the Efficient Market Hypothesis (Fama, 1970) and the Capital Asset Pricing Model (CAPM). Wachtel (1942) is considered as the first scholar that acknowledged the evidence of the January effect. Another early example of this anomaly is Rozeff and Kinney (1976), whereby stock returns in January are higher than other months of the year. Subsequently, a large number of studies has been composed to support the presence of the TOY effect in consequence of its inconsistency with the EMH.
The findings of this anomaly mostly concentrate in the developed stock markets, especially in the U.S market such as Tinic and West (1984), De Bondt and Thaler (1987), and Haugen and Jorion (1996).

5
The structure of this study is as follows. Section 2 provides the data descriptions and methodology for further analyses. The empirical analyses are reported in Section 3. Finally, the last section presents a brief conclusion.

Data and Methodology
To extend the blooming finance literature in seasonal anomaly, this paper examines the TOY effect in 15 stock market indices in the Asia Pacific, including 12 national and 4 regional indices. The detail of examined stock indices is reported in table 1. To examine the TOY effect on value-weighted index returns, this study will construct two methods, including Ordinary Least Squares (OLS) regression and Exponential Generalized Autoregressive Conditional Heteroskedastic -EGARCH. The EGARCH model is firstly developed by Nelson (1990) and then extended by Nelson and Cao (1992) and McAleer and Hafner (2014). This model also captures a stylized fact of volatility clustering in financial time series as it can empirically capture the circumstance that negative shocks at time t -1 have more significant impacts on the variance at time t than the positive shocks. The mean equation for the OLS and EGARCH. is as follow for Where, Rt is the monthly buy-and-hold returns of all stock market indices. The " represents the coefficient of the monthly returns except for the first month of the tax year. D %& is the dummy variable for month i with the estimated coefficient β " , which is the first month of the tax year. The dummy variable equal to 1 for returns of the first month of the tax year and 0 for the returns of the other eleven months.
The error term of the regression model is taken by ε ' . The variance equation of the EGARCH model is constructed as follow: The intercept and error term are and the ! ~ N(0, ! ( ), respectively. D %& is the dummy variable for month i with the estimated coefficient β " , which is the first month of the tax year. The spill over effect or the association between preceding and current variance in absolute value is captured by the ARCH term ( " ). The asymmetry or leverage effect is captured by the leverage term ( ). The leverage effect is identified if is negative when the negative shocks are followed by higher volatility. This indicates that the negative shocks (bad news) have more impact on volatility than positive shocks (good news) of the same size (Chang & McAleer, 2017). The asymmetry is the leverage effect as the risks from growing leverage embrace the negative shocks (Tsay, 2005  For five developed markets, the average returns in the first month of the tax year are significantly higher than other months in Australian, Hong Kong, Japan, and New Zealand markets. This disposition also presents in Taiwan, Thailand, Philippines, South Korea, and four regional stock indices. This inclination is also related to the TOY effect those in stock markets. For the other stock indices, the average return in January is negative and relatively lower than other months, which signifies the absence of the January anomaly. The results of the Unit root test (The Augmented Dickey-Fuller and Phillips-Perron Tests) suggest that data series are stationary (p-value <0.01), which means that variables follow a random walk. Therefore, the data series are appropriate for further empirical time-series analyses in the succeeding section. Using the EGARCH model, we also consider the asymmetry and leverage effect on the association between stock return shocks and previous shocks to volatility (Giovanis, 2009). As seen in Panel A of These findings are still consistent when we regress on the first sub-period in Panel B. However, the leverage effect is also testified for the New Zealand index at 1% significant level. Further, the leverage effect evaporates in Australian, Japanese, and New Zealand markets when we consider the second subperiod.  Considering the asymmetry and leverage effect in emerging stock markets by using EGARCH model in Table 4, the leverage effect is captured in Malaysia, Taiwan, South Korea, and Philippines stock markets. The negative and insignificant coefficients (δ) are also recorded for Indian and Thailand indices.

TOY effect in emerging stock markets
Interestingly, the leverage effect is invisible in the Taiwan market and becomes significant in Indian market when we examine the first sub-period. After the GFC, the leverage effect is undetectable in all six examined emerging stock markets.   Taking into account variations reflecting conditions across regions and market cap segments, the TOY effect is examined by using four regional market indices. The results of the full sample and two subperiods are tabulated in Table 5. The results indicate that the January anomaly is unobservable in four regional indexes regardless of market capitalization and inspected periods. However, the January returns moderately decrease after the GFC as the intercept of two regression model increase while we can obtain more negative values of the slope. The leverage effect is captured in three indices except for the small cap index of developed markets. We also obtain comparable results when examining in sub-periods in the Australian market after the GFC. This study also cannot detect the January effect in the Singaporean stock market, which is consistent with Wong at el. (2006) who suggests seasonal anomaly is disappearing in the Singaporean stock market. Regarding the Japanese market index, our results support the absence of the January effect. This result corroborates the finding of Li and Gong (2015) related to the deteriorating movement of the January effect after the Japanese economic recession during the 1990s.

Summary of the results
The empirical results of six emerging stock markets (except Taiwan) confirm the absence of the January effect during the full examined period. However, this anomaly is visible in the Philippines, Thailand, Taiwan, and South Korean markets before and during the GFC. This finding is inconsistent with Tangjitprom (2011) and Tong (1992) that this anomaly is not witnessed in both Thailand, South Korea, and Taiwan equity markets. The discrepancy can be rationalised by the variations in sample periods and methodology. The TOY effect becomes invisible in all emerging stock markets and four regional stock indices after the GFC. Our findings also reconcile with Raj and Kumari (2006) findings for the Indian market and Ali, Nassir, Hassan and Abidin (2009) and Ali Ahmed and Haque (2009) for the Malaysian market, who document the absence of January effect in these two markets.
The leverage effect denotes that the negative shocks (bad news) have more power on volatility than positive shocks (good news) at the same magnitude (Chang & McAleer, 2017;Nguyen & Nguyen, 2019), which are reported in Table 6 (Panel B). For developed markets, we depict evidence of leverage in the unconditional volatility in three out of five examined indices, including Australia, Hong Kong, and Singapore. When we regress on two sub-period, the leverage is reported in four indices (except Japan) before and during the GFC, while it is relatively faded away after this event. Evidence of leverage effect in emerging market indices also denotes the same condition. The number of stock indices that experienced leverage effect reduces from four to one after the GFC.
We also observe convincing evidence for the disappearance of the leverage effect after the GFC in 19 four examined regional stock indices. This finding reveals that the significance of the leverage effect in unconditional volatility has depreciated overtime after the global financial turmoil. This finding is strongly consistent with a prior finding of Campbell and Hentschel (1992) that suggests the positive association between the leverage effect and overall stock market volatility during the crisis. Also, the leverage effect is more prominent in the case of developed and large-cap stock indices. It corroborates the findings of Jayasuriya, Shambora and Rossiter (2009), Talpsepp and Rieger (2010), and Kayal and Maheswaran (2018) who posit the magnitude of leverage effect in unconditional volatility is more significant in mature markets compared to emerging markets.

Conclusion
Given the prospective effect of the TOY effect on the theories of modern finance and the inadequate research in this field, this study provides a comprehensive examination of the existence of the TOY effect in fifteen stock market indices of the Asia Pacific region. Our findings indicate that the TOY effect becomes visible again in three developed stock markets with tax year not ending in December after the GFC. In contrast, this seasonal anomaly becomes invisible in emerging stock markets after the GFC, which is consistent with the EMH. Generally, our findings demonstrate that the magnitude of this anomaly has diminished in the emerging markets, but it remains prevalent in some developed markets in recent years. This finding also moderately tolerates the argument for the weakening of stock market anomalies over time, since investors progressively exploit this effect (Wong at el., 2006;Lu & Gao, 2016, Huynh, 2020. The evaporation of this effect would lend encouragement to the supposition that some Asia Pacific stock markets satisfy the weak form of the EMH. It also has significant inferences for the trading behaviours of investors in the stock markets.

20
We also find evidence of the leverage effect in the unconditional volatility that volatility after negative shocks is significantly higher than that of positive shocks across examined stock indices.
However, this effect is more conspicuous in mature stock indices than emerging indices. It is explicable as there is incidence large number of investors in developed stock markets compared to that of emerging markets. Our findings also propose the positive connection between the leverage effect and stock market volatility as the magnitude of this effect has weakened during the stable market condition after the GFC.
Our study notes significant implications for fund managers and investors to take into consideration this anomaly to create higher rates of abnormal profit. The presence of the leverage effect plays a critical role in financial risk controlling, hedging approaches, and options pricing. It also supports investors in their investment decision-making in the stock market. Another contribution is for researchers, this study sheds light on the current trend of seasonality effect in stock returns. This investigation advances our understanding of stock market anomalies and asset pricing theories.