اثر دیگر ژانویه : شواهد برابر کارایی بازار؟

The Other January Effect (OJE), which suggests positive (negative) equity market returns in January predict positive (negative) returns in the following 11 months of the year, underperforms a simple buy-and-hold strategy before and after risk-adjustment. Even the best modified OJE strategy, which benefits from several ex-post adjustments, does not generate statistically or economically significant excess returns. When the OJE is tested with a method that is consistent with investor experience it is clear the OJE is no more profitable than an 11-month strategy that uses November or December as the conditioning month.

The Other January Effect,1 which suggests positive (negative) returns in January predict positive (negative) returns in the remaining 11 months of the year, is shown to be a remarkably simple yet powerful prediction tool in recent studies by Brown and Luo, 2006, Cooper et al., 2006 and Sturm, 2009.2 We build on these important papers by investigating whether the Other January Effect (hereafter OJE) can be implemented by investors to earn risk-adjusted excess returns. In doing so, we are effectively considering whether the OJE is evidence against the efficient market hypothesis. As Schwert (2003, p. 942) notes, “if anomalous return behavior is not definitive enough for an efficient trader to make money trading on it, then it is not economically significant”. Fama (1991, p. 1575) describes this version of the efficient market hypothesis, which dates back to Jensen (1978), as “economically more sensible.” To be sure, we are not criticizing previous OJE papers. These show that, on average, 11-month returns following positive Januaries are larger than 11-month returns following negative Januaries, and that this “spread” cannot be explained by standard asset pricing models. We verify this result before turning our attention to the question of whether the OJE can be used to earn economic profits, which is something these earlier OJE papers have not addressed. If the OJE market timing technique can be used to exploit market inefficiency, then portfolio managers, individual equity investors, the management of listed companies looking to raise additional equity and private companies considering an IPO should all take the sign of the January return (and the corresponding 11-month OJE return prediction) into account when making their decisions. Alternatively, if the OJE does not generate economic profits then those currently using the OJE should reconsider their faith in this timing tool. Our approach is consistent with many other strands of the return predictability literature,3 where an original study which documents predictability is followed by subsequent work which considers implementation issues. For example, Cooper et al. (2005) test whether the predictability of book-to-market equity, size, momentum, and beta, as shown by previous authors, can be used by an investor to form a portfolio that outperforms a passive index. Moreover, Lesmond et al. (2004) show an investor who attempts to exploit the momentum predictability documented by Jegadeesh and Titman (1993) incurs large transaction costs, which erode the majority of profits. On the face of it, the OJE would appear to be compelling evidence against market efficiency. Previous papers report average differences in 11-month returns following positive and negative Januaries that are frequently in excess of 10%. Moreover, unlike many “anomalies”, the OJE is easy to implement. It only gives one signal per year so transaction costs are considerably lower than those in many quantitative strategies. Most interpret it to relate to market indices so short positions can be easily created. Finally, the information required to open a position, namely the January return, is readily available. These last two features imply the gross profits generated by the OJE need not be very large to offset the costs incurred in implementing it. We therefore suggest it is somewhat surprising to discover, as we do, that the OJE is not evidence against market efficiency. We show the OJE cannot be implemented to earn risk-adjusted excess returns. OJE returns are neither economically nor statistically significantly different to buy-and-hold returns, and OJE strategy Sharpe Ratios are inferior to buy-and-hold Sharpe Ratios. It seems clear that followers of the OJE interpret it as implying that a negative January indicates an 11-month return that is negative rather than one that is simply less than the 11-month return following positive Januaries. Hensel and Ziemba (1995a, p. 188) quote Hirsch (1986), who appears to have been the first to propose the OJE, as follows “The supposition is that: If the market rises in January, then it will also rise during the rest of the year; but if it falls in January, then there will be a decline during the rest of the year.” This implies the most logical way to exploit the OJE is to observe the January return and take an 11-month long (short) position following positive (negative) Januaries. It also seems clear the OJE is intended to be implemented by remaining out of the equity market in January while the January return is being observed.4 However, so as not to be accused of testing a “straw man” version of the OJE, we also test three major variations of the standard long T-bills in January then long (short) 11-month equity market position following positive (negative) Januaries OJE strategy. These include: (1) staying long (short) the equity market for 12 months (February–January) following an OJE signal, (2) always being long the equity market in January and long (short) the equity market for February–December based on the actual January return, and (3) always being long the equity market in January, being long the equity market for February–December following a positive January, and being long T-bills for 11 months following negative Januaries. Variation one and two helps the OJE by letting it sometimes and always respectively capture the January return, which tends to be positive on average due to the January Effect. Variation three further improves the OJE by limiting losses on its short positions. We acknowledge these variations, which are made after observing how the OJE performed, involve data mining bias5 but we feel their inclusion strengthens our argument regarding the OJE not being evidence against market efficiency.6 Neither the standard OJE strategy nor any of its modified versions generate returns on that are statistically or economically significantly different from buy-and-hold returns. It is also not possible for an investor to profitably adopt the OJE in international equity indices. There are four factors that contribute to the underperformance of the OJE compared to a simple passive buy-and-hold strategy despite the predictive ability documented previously in the literature based on the simple spread approach. Firstly, the simple spread is not consistent with the average return earned by an investor over February–December periods. For instance, assume that during a 3-year period January returns are positive in the first two years and the 11-month (February–December) returns are 8% for both these years. Assume the third year has a negative January return and an 11-month return of 1%. Based on these numbers the simple spread is 7% (8% − 1%) but anyone adopting the OJE would experience an average return, or weighted spread of 5% ([8% × 2/3 − 1% × 1/3]). We show the actual average 11-month return earned by an OJE investor is lower than the 11-month spread. Secondly, the simple spread approach does not consider the January return which a passive investor earns, but an OJE investor foregoes. Thirdly, the OJE gives inaccurate signals to short the market. Eleven-month returns following negative Januaries (i.e. periods when the OJE is short the market) are positive on average so a passive investor who is always long the market earns these returns but anyone adopting the OJE would incur losses on their short positions during these periods. Even though they are positive on average, 11-month returns following negative Januaries are smaller than 11-month returns following positive Januaries. This raises the possibility of an investor being able to benefit from a modified OJE strategy which is always long the equity market in January, long the equity market for February–December following positive Januaries, and long T-bills for 11-month periods following negative Januaries.7 This heavily modified OJE strategy is identical to the passive buy-and-hold approach at all times other than February–December periods following negative Januaries. The modified OJE investor is long T-bills during these periods while the passive investor is long the equity market. We find this modified strategy also does not out-perform the passive strategy. This represents the fourth factor behind the OJE not out-performing. Namely, 11-month T-bill returns have only been marginally larger than 11-month equity market returns during periods following negative Januaries so an investor following the modified OJE strategy would not have received profits that are either statistically or economically significantly larger than those earned by a buy-and-hold investor. Moreover, the investor following this modified OJE strategy would also not have earned statistically significant excess risk-adjusted returns. After completing our analysis we become aware of a recent paper by Cooper et al. (2010)8 that also considers whether the OJE can be implemented to earn abnormal returns. There are numerous differences between our papers, including the conclusion. They conclude (p. 18) “the January Barometer does appear to provide useful information for would-be investors, or, at least historically, it would have contained useful information”, whereas we conclude the OJE has not been a useful tool for investors. This difference in conclusions appears to be due to at least two factors. Firstly, we consider the statistical significance of the difference in raw and risk-adjusted returns to the data-mined modified OJE/T-bill strategy versus the buy-and-hold approach and find there is no statistical significance. Secondly, we have different interpretations of the economic significance (or lack thereof in our case) of the less than 0.5% p.a. (pre-transaction cost adjustment) excess returns generated by the best of the modified OJE strategies. Other differences include our inclusion of international results, which demonstrate the OJE cannot be profitably implemented internationally. We also use the Manipulation-Proof Performance Measure, which Ingersoll et al. (2007) demonstrate overcomes deficiencies in Sharpe Ratio analysis. Our paper also contains a novel approach of comparing the profitability of the OJE with 11-month strategies based on other conditioning months, from an investor perspective, which demonstrates the OJE is no more profitable than an 11-month strategy using November or December as the conditioning month. Finally, we show the OJE does not add value when applied to individual stocks. We compare and contrast our results with those of Cooper et al. (2010) throughout the paper. The remainder of this paper is organized as follows. Section 2 provides evidence of the current focus on the OJE in the popular press and a discussion of the academic papers that have investigated the OJE. Our methodology and results are presented in Section 3. Section 4 concludes the paper.

The Other January Effect (OJE) suggests positive (negative) returns in January are said to predict positive (negative) February–December returns. We consider if the OJE market timing technique can be implemented to earn risk-adjusted excess returns. This issue should be of interest to portfolio managers, individual equity investors, and the management of listed companies looking to raise additional equity and private companies considering an IPO. If the OJE market timing technique can be used to profit from market inefficiency, then they should all pay attention to the sign of the January return before making their decision. On the other hand, if the OJE does not generate economic profits then proponents of the OJE should reconsider their faith in this indicator. We show the OJE cannot be implemented to earn statistically and economically significant risk-adjusted excess returns. The OJE is therefore not evidence against market efficiency. Consistent with earlier work, we find 11-month returns are, on average, larger following positive Januaries than negative Januaries (i.e. the simple spread is positive), but there are four factors behind these translating into superior profits for an investor adopting the OJE. Firstly, the simple spread between the average 11-month returns following positive and negative Januaries does not represent the 11-month returns earned by an OJE investor. We demonstrate the weighted spread (which is consistent with investor experience) between 11-month returns following positive and negative Januaries is smaller than the simple spread. Secondly, the OJE requires the January return to be observed prior to a signal being generated to go long or short the market. January returns tend to be larger than the monthly returns of other months in the US so remaining out of the market in January incurs a relatively large opportunity cost. Thirdly, 11-month returns following negative Januaries are positive on average. The OJE is short the market during these periods so losses are incurred. This results in it underperforming a simple strategy that is long the equity market every February to December. Fourthly, a data-mined OJE strategy of always remaining long the market in January, taking a 11-month long position following positive Januaries, and investing in T-bills (or a combination of T-bills and the equity market) following negative Januaries still does not significantly (either statistically or economically) out-perform a 12-month passive buy-and-hold strategy. This is due to 11-month T-bill returns only being marginally larger than 11-month equity returns on average following negative Januaries. Our results are robust. The OJE cannot be implemented to earn risk-adjusted excess returns in portfolios of stocks of different sizes or international indices. Moreover, we show that when the OJE is tested with techniques that are consistent with investor experience it is not a superior 11-month return predictor to strategies based on November or December as the conditioning month. In summary, we suggest practitioners should think twice before using the OJE as a market timing tool. There is no evidence that it provides useful information. The academic community should also be aware that the OJE is not able to earn excess returns and is therefore not evidence against the commonly expressed version of efficient market hypothesis which recognizes the importance of economic significance.