The excess returns for stock price in January have long been documented in literatures. Numerous studies have been carried out with the attempt to explore the root cause of this January anomaly. In this paper, three categories of explanations, namely price-pressure hypothesis, omitted risk factors and transaction costs, mismeasurement problems, are presented with both supportive and opposing arguments from empirical evidences. It can be concluded from the review of those studies that the existence of January Effect cannot be attributed to a single factor, and the reason for the presence as well as persistence of this anomaly is still a puzzle worth investigating.
The January effect, which refers to the phenomenon that stocks exhibit higher returns in January, has been a heated issue in the academic field, since it presents regularities in stock returns, seems to afford profitable opportunities and contradict to the efficient market hypothesis.
With regard to the root cause of the January anomaly, three categories of explanations are comparatively representative and convincing. The first group of explanation, consisting of tax-loss selling hypothesis, window-dressing hypothesis and so on, focuses on investor behavior around year end and emphasizes selling or buying pressure exerted on stock prices. The second one assumes that the abnormal returns in January are compensation for underlying risk, thus the effect does not contradict to the efficient market hypothesis. The last type of explanation generally suggests that January effect is an artificial result due to data-mining and measurement problems, or cannot be exploited in the presence of transaction costs.
Drawing on empirical evidences, a sensible conclusion can be made that neither of these explanations are able to justify the existence as well as persistence of January anomaly completely, since all of them are tested in various studies with conflicting results. Consequently, we conclude in the paper that the reason for the existence of January effect is still a puzzle in the academic world and further investigations are in need to draw a clearer picture regarding this well-known anomaly existed in the financial market.
Introduction
In recent years, a growing body of academic literatures has documented the anomalous regularities in security return at turn of the year. These anomalies are generally referred to as the January effect. According to Seyhun (1993), the January effect (also called the turn-of-the-year effect) can be defined as the phenomenon that stocks exhibit large and positive abnormal returns during the first few weeks in January. The first terminal paper brought the January effect to the attention of modern finance is Rozeff and Kinney (1976). They study the existence of seasonal patterns in an equal-weighted index of New York Stock Exchange prices over the period of 1904-1974 using both non-parametric tests (Kruskal-Wallis test (Conover, 1971)) and parametric analysis (Bartlett's (1937) test). The statistically significant higher mean of return in January found in the test confirms the existence of January effect.
Explanations for the January anomaly have invoked heated debate in the academic field. The ongoing discussion regarding the root cause of January effect falls into three broad categories. Firstly, price-pressure hypothesis attempts to explain the turn-of-the-year effect from the perspective of investor behavior, such as tax-loss selling hypothesis and window-dressing hypothesis. Secondly, explanations based on risk factors suggest that January anomaly does not contradict to the efficient market hypothesis since higher returns in January are merely a compensation for higher risk. The last possible explanations for January effect are attributed to data-mining, mismeasurment problems and bid-ask spread bounce, which suggest that the anomaly is empirical illusory or cannot be exploited as profitable opportunities. In this paper, I will introduce each of these explanations and present empirical evidence for and against these arguments, while attempting to draw a sensible conclusion regarding the cause of January effect.
Price-pressure hypothesis
2.1 Tax-loss selling hypothesis:
The tax-loss selling hypothesis is the most controversial explanation for the January effect since Branch (1977), who asserts that stocks exhibiting high returns in January have experienced decline in prices during the previous year. According to Thaler(1987), the tax-loss selling hypothesis indicates that shortly before the year end, investors will sell stocks with decline in prices during the previous year in order to realize capital loss for tax purpose. Then, in January, prices of those stocks will rebound in the absence of selling pressure.
Numerous studies have been conducted to prove the validity of this hypothesis. Reinganum (1983) finds evidence to support the tax-loss selling hypothesis and argues that January returns are higher for previous-year 'losers' and the excess returns cannot be observed for small 'winners' in the first five trading days in January. Schultz (1985) and Steven et al. (1991) also provide evidence in support of the hypothesis by analyzing the seasonal effect around the introduction of personal income taxes in 1917. Ritter (1988) explains the price movements at turn of the year at the level of individual investors and concludes that the return anomaly is associated with buying and selling habits of "small" investors who tend to sell poorly-performed stocks in December to realize loss and reinvest in a broad range of small-cap stocks in January with cash infusion from end-year bonus or capital gains from the sale of large-cap stocks. He also offers a "parking the proceeds" hypothesis as to why higher returns are largely concentrated in small firm stocks, especially those who have experienced price decrease in the prior year.
In addition, Griffiths and White (1993) justify the tax-induced hypothesis by examining Canadian and U.S. data with the purpose of discriminating between tax-motivated and other year-end effects. Poterba and Weisbenner (2001) find the changes in capital gains tax rules have significant influence on the January effect. Chen and Singal (2004) examine the hypothesis by controlling the effect of other factors such as window dressing and bid-ask bounce. Starks et al. (2006) use municipal bond closed-end funds to examine the validity of tax-loss selling hypothesis because these funds are mainly held by tax-sensitive individual investors. All of these empirical studies provide evidence in support of the tax-loss selling hypothesis. Besides, Hang and Hirschey (2006) report that tax-loss selling by small investors rather than institutional investors contributes to the anomalous pattern of stock returns at turn of the year, which consistent with the study of D'Mello et al. (2003).
However, according to Thaler (1987), international evidence suggests that tax may partially explain the abnormal returns in January, but it is not the entire explanation. For instance, Japan with no capital gains and loss offset exist can also observe January effect (Kato and Schallheim, 1985). Canada had a January effect before 1972 when capital gains tax is still not implemented (Berges et al. 1984). Great Britain and Australia have the turn-of-the-year effect although their tax years begin on April 1 and July 1, respectively (Thaler , 1987).
Despite the fact that overwhelming empirical evidence have been presented in support of the tax-loss selling hypothesis, opposite arguments still exist. Roll (1983) regards the tax-loss selling hypothesis as a patently absurd and argues that despite the fact that some investors may trade for tax purpose, others would have already bought the stocks in anticipation of the higher returns in January, leading to the disappear of January effect. Reinganum (1983) shows that abnormal returns can also be observed in small firm stocks without prior-year price declining. Constantinides (1984) oppose the hypothesis by indicating that delaying the loss realization until December is not an optimal tax trading strategy. Jones et al. (1987) report that before the imposition of income taxes, the January effect had already existed in U.S. Moreover, Hang and Hirschey (2006) have proved persistence of January effect in their test after the Passage Tax Reform Act of 1986, when all seasonal tendencies associated with tax-motivated selling by institutional investors should have been eliminated at calendar year end, indicating that tax-loss selling hypothesis for institutional investors should be rejected. Thus, empirical evidences regarding the tax-loss selling hypothesis provide mixed results.
Window-dressing hypothesis:
Another competing explanation for the January effect is the window-dressing hypothesis proposed by Haugen and Lakonishok (1988) and Lakonishok et al. (1991). Specifically, the hypothesis indicates that portfolio managers are evaluated on both their investment performance and consistency of their investment philosophy. They tend to include risky and small firm stocks in their portfolios to earn higher returns and sell them before the calendar year end to avoid revealing them in year-end holding. In January, they would reverse this process by reinvesting in those small firm stocks, typically including some past 'losers'.
Empirical evidences concerning this hypothesis provide conflicting opinions. Musto (1997) concludes that the window-dressing hypothesis contributes to the January effect to some extent because he finds turn-of-the-year effect can be observed in money market in which instruments do not generate capital losses through tax effect. Maxwell (1998) asserts that for noninvestment grade bonds window dressing is a major factor behind the January anomaly. Chopra and Ritter (1989) and Meier and Schaumburg (2004) also find evidence consistent with the window-dressing hypothesis. However, Hang and Hirschey (2006) claim that it is reasonable to assume that window dressing by large institutional investors would be a large-cap firm phenomenon. Since most abnormal returns in January are concentrated in small-cap firms, window-dressing hypothesis has limited relevance with the turn-of-the-year effect.
Nevertheless, it is noteworthy that both the tax-loss selling hypothesis and window-dressing hypothesis rely on year-end selling pressure and provide same prediction regarding the excess returns at turn of the year. Therefore, it is difficult to differentiate the influence of the two hypotheses and determine which one has derived the January effect (Steven et al., 1991), Chen and Singal (2004) and Starks et al. (2006)). Due to this testing problem, several studies have been designed aiming to distinguish the two hypotheses. For example, controlled tests conducted by Sias and Starks (1997) evaluate these two hypotheses separately and find tax-loss selling hypothesis provides stronger power in explaining January effect. Furthermore, base on the idea that institutional investors will window dress their portfolio more than once in a year when necessary, Chen and Singal (2004) study the stock return behavior around semi-annual closing when interference of tax-motivated selling is eliminated. Both tests tend to reject the validity of window-dressing hypothesis. Hence, empirical evidences are inclined to support tax-related hypothesis more than window-dressing hypothesis, but they are not necessarily mutually exclusive (Steven et al., 1991).
2.3 Differential information hypothesis
In accordance with Chen and Singal (2004), the excess returns in January can be attributed to the influence of significant information releases at turn of the year. There are different versions of information hypothesis relative to the abnormal returns in January.
Barry and Brown (1984) assert that lack of information is generally considered a higher risk indicator, resulting in the inclination of investors to regard those firms with less information as bearing higher risk than those with comparatively adequate information although the beta risk of both types of companies are equal. Due to the fact that returns exhibit total risk involved in the stocks whereas asset pricing model merely price systematic risk alone, the excess return for less-information firms may be treated as abnormal returns. For testing this hypothesis, they use listing period as a proxy for obtainability of information and find differential production of information among firms can partially explain the January effect.
Merton (1987) proposes the investor recognition hypothesis as another way of interpreting the information hypothesis. Specifically, since January is the common period for firms to publish their information, investors would tend to buy the stocks when they become more aware of the companies through those newly published information, leading to observed January effect.
Chen and Singal (2004) test the differential information hypothesis based on stock returns and turnover. To be specific, if the hypothesis holds, not only in January but also in other months should higher returns for small firm stocks be observed when listed firms submit their accounting information quarterly in accordance with the requirement of Securities Act of 1933. Furthermore, less trading volume in December is expected since investors would wait until the publishing of new information. However, results from both aspects unanimously suggest that information hypothesis cannot be the primary driver of January effect.
2.4 Turn-of-month liquidity hypothesis
Ogden (1990) argues that the January effect in U.S. market can be, at least in part, due to the standardization in payment system, which leads to a concentration of cash flow at turn of the year. Specifically, substantial cash receipts at turn of the year enable investors to invest in a wide range of stocks resulting in higher demand of stocks and a surge in stock returns in January. He tests the implication of this hypothesis based on value-weighted and equally-weighted stock index returns and concludes that this hypothesis can be a partial explanation for the January effect.
Risk-based explanations
The Risk-based explanations suggest that higher risk is the root cause of abnormal returns at turn of the year. For example, Chan et al. (1985) employ a multi-factor pricing equation as an explanation of the firm size effect and concludes that the size anomaly can be justified by additional risk factors involved in small-cap firms.
Base on the same idea, Rogalski and Tinic (1986) attempt to explain the turn-of-the-year effect in terms of risk. Specifically, they argue that previous studies generally assume that systematic risks and equilibrium rate required on stocks returns remains constant over calendar months. However, this assumption is neither required by equilibrium asset pricing models in theory nor consistent with stock price movements in practice. As a result, they correct this problem by examining nonstationarities in total, systematic and diversifiable risks of common stocks over calendar months and claim that a pronounced increase of small firm beta is observed at the beginning of the year; hence higher returns in January are risk premium but not anomalies.
However, a contrary conclusion is drawn by Seyhun (1993). He tests the hypothesis that omitted risk factors contribute to the large January returns, employing a stochastic dominance approach whose results do not depend on risk-return tradeoffs generated by asset pricing models. Findings in the paper suggest that excess returns in January are too high to be equilibrium returns, thus omitted risk factors in explaining January returns are less plausible. Moreover, Keim(1983) opposes the risk hypothesis advanced by Brown, Kleidon and Marsh by arguing that even if those unspecified risk factors are able to explain part of the size effect, they are insufficient to justify those January behaviors since the return premium is observed in the same month each year.
Bias, transaction costs and mismeasurment problems
4.1 Bid-ask spread bounce
The bid-ask spread bounce documented by Keim (1989) suggests that Investors tend to trade at the bid in December and at ask in January, thus the bid-ask bounce gives an impression of large positive January returns, when the fact to be not the case.
Base on this assumption, Blume and Stambaugh (1983) report that stock returns computed by closing prices are imparted an upward bias, primarily due to bid-ask effect. Bhardwaj and Brooks (1992) demonstrate that the turn of the year switch from bids to asks leads to a positive bid-ask bias in calculated returns estimated at approximately 1% and this bias is large and significant for low-price stocks. Hence, the positive bid-ask bias in returns has overestimated the January anomaly. Griffiths and White (1993) also conclude that a systematic shift from seller-initiated transactions at bid price to buyer-initiated trades at ask price at turn of the year drive the emergence of January effect.
Due to this concern, specific measures have been taken in many empirical studies in order to control the effect of bid-ask bias when analyzing the January effect. For example, Cox and Johnston (1998) correct the bid-ask bounce by excluding low-price stocks while Chen and Singal (2005) use midpoint of the bid-ask price for calculating returns. Most of these studies suggest that the bid-ask spread bounce may upwardly bias the excess returns in January but is not the entire explanation.
4.2 Transaction costs
Transaction costs contain commission fee, bid-ask spread and the market impact of the transaction (Seuhun, 1993). A reasonable question would be if transaction costs are responsible for the existence of January effect since such costs prevent arbitrageurs from exploiting anomalous patterns and eliminating price inefficiency.
Empirical tests are inclined to believe that transaction costs eliminate profitable opportunities from January anomaly, resulting in the persistence of abnormal returns. Bhardwaj and Brooks (1992) examine returns after transaction costs on stocks within different price ranges, aiming to determine whether typical investors are able to obtain excess returns. The result of the study is time-dependant. For the first ten-year period of 1967-1976, low-price stocks outperform their counterparts after considering transaction cost. However, in the following ten years, a reversal is observed when high-price stocks dominate low-price stocks. Thus, economically and statistically significant excess returns cannot be earned by typical investors from trading around January. This finding is consistent with the argument presented by Haugen and Jorion (1996), namely turn-of-the-year effect is not a indication of market inefficiency since the abnormal returns are not exploitable for typical investors. Seyhun (1993) argues that investors with transaction costs lower than 300-basis-points, typically large investors, are able to profit from strategy derived from January effect while for those who bear transaction costs higher than 500 basis points, a buy and hold strategy would be more ideal.
Hence, empirical evidences generally suggest that transaction costs prevent typical investors profiting from those anomalies and contribute to the persistence of January effect to some extent.
4.3 Data-mining and mismeasurment problems
Lokonishok and Smidt (1988) present three generic considerations regarding the reliability of January effect evidence. The first one is called boredom factor, which stresses the danger that academic field attaches undue significance to studies documenting anomalies due to a selection bias (Merton, 1985). Then, noise in security returns emphasizes difficulties in distinguishing anomalies with noise in the market (Fischer Black, 1986). The last consideration is called data snooping, which represents the act of using the same data to discover or test hypothesis in different studies. All of these problems may lead to biased results for stock return patterns.
After taking these factors into account, Lokonishok and Smidt (1988) still find evidence for the persistence of anomalous returns at turn of the year. This result strongly refutes the hypothesis that seasonal anomalies are product of sampling error and data mining (Lokonishok and Smidt, 1988). Gultekin and Gultekin(1983) also proves that January effect is not merely a statistical artifact. However, Steven et al. (1991) prove in their study that different methods of computing returns may lead to different results of empirical tests. Therefore, measurement problems may impact the study results of January anomaly to some extent, but are not primary explanations.
To conclude, it can be seen from empirical evidences that explanations concerning the existence of January effect are mixed results. Price-pressure hypotheses based on investor behavior can afford partial explanation to the January anomaly but studies hardly reach a consensus. Risk-based explanations are sensible intrinsically since most January effects are small firm phenomenon and small firms are generally considered to bear higher risk. However empirical tests cannot draw a definite conclusion for this explanation either. The last category of explanation generally suggests that anomalies do not exist in the first place. They are artificial results due to data-mining and bid-ask bias or cannot be exploited by arbitrageurs due to transaction costs. In this sense, it is unlikely to present a single explanation with regard to the January anomaly (Lokonishok and Smidt, 1988). More studies are expected to resolve the puzzle in the future.
