# A November Effect? Revisiting the Tax-Loss-Selling Hypothesis.

We document the existence of another seasonality in stock returns: a November effect. The uniqueness of this study is that the November effect is observed only after the passage of the Tax Reform Act of 1986. We document a unique and significant relationship between excess returns and the potential for tax-loss selling and conclude that the November effect is explained by the tax-loss-selling hypothesis. We also show that the January effect in the post-Act period is stronger than in the pre-Act period. This result is likely due to the Act's elimination of the preferential treatment for capital gains. The evidence presented in this paper suggests that tax-loss selling is a dominant explanation for the seasonality of stock returns.* Seasonalities in US stock returns have been well documented in the finance literature. Wachtel (1942), Rozeff and Kinney (1976), and Dyl (1977) first observed the prevalence of significant excess returns and volume during January and suggested that tax-loss-selling at year-end was a likely cause for this anomaly. [1] Subsequent research by Banz (1981), Keim (1983), and Reinganum (1983) suggests that small firms are the most likely candidates for tax-loss-selling and that tax-loss-selling is just one of the explanations for the January effect. Schultz (1985) and Jones, Lee, and Apenbrink (1991) provide further evidence for tax-loss-selling by examining the January effect around the introduction of individual taxes in 1917.

For the past two decades, financial researchers have focused their attention on several other non-mutually exclusive hypotheses to explain the January effect. DeBondt and Thaler (1985, 1987) and Chopra, Lakonishok, and Ritter (1992) argue that investor overreaction causes the January effect. Roll (1981 and 1983a,b) and Blume and Stambaugh (1983) argue that the January seasonality may be due to data and statistical biases. Chan, Chen, and Hsieh (1985) and Chan and Chen (1988a,b) argue that the January effect is due to misspecification of systematic risk. Ball, Kothari, and Shanken (1995) suggest that the effect may be explained by low-priced stocks trading within a relatively wide bid-ask spread.

It also has been argued that the excess January returns could possibly be the effect of significant information releases that occur in January. As summarized by Jones and Lee (1995), the information hypothesis involves an adverse selection problem that may explain how seasonal selling results in price pressure that survives arbitrage. Indeed, earlier work by Rozeff and Kinney (1976), Brown, Keim, Kleidon, and Marsh (1983), Berges, McConnell, and Schlarbaum (1984), and Maxwell (1998), among others, point toward price pressure from both seasonality of information and tax-loss selling.

This paper exploits a unique opportunity, provided by the Tax Reform Act of 1986 (TRA), to evaluate the tax-loss-selling hypothesis. The TRA was the most comprehensive change in the US tax system in recent times and had a significant impact on the decisions of individuals and corporations. The changes of particular relevance to this study are: 1) tax year-end for all mutual funds to realize capital gains and losses is changed to October 31 from December 31, which was the tax year-end for most funds, and 2) the elimination of the preferential treatment of capital gains. The potential effects of these changes on individual investors' trading behavior and on stock prices have not yet been documented.

The contribution of this study is twofold. The paper documents a November effect and a stronger January seasonality following the TRA. We argue that both results are the obvious product of peculiarities of the TRA that support the tax-loss-selling hypothesis. The TRA requires mutual fund managers to distribute at least 98% of the realized capital gains income generated during the 12-month period ending October 31. [2] Any undistributed earnings are subject to a 4 % excise tax. Mutual funds have an incentive to sell losing stocks before October 31 to reduce the taxable capital gains they are required to pass on to shareholders. Similarly, mutual funds are likely to postpone the sale of winners until after the tax year-end. This creates an incentive for significant tax-loss-selling since mutual funds are dominant traders of equities. [3] Because November is likely to be free of the sort of significant information releases observed in January, the passage of the TRA provides a unique opportunity to conduct testing of the tax-loss-selling hypothesis without a confounding information effect. The tax-loss-selling hypothesis implies neither a November return nor a trading volume seasonality prior to the TRA. We would, however, expect a post-TRA November effect. To the extent that tax selling is documented in November, this paper presents evidence supporting an explanation of stock seasonalities based on selling pressure. Though there is some mention in the financial press that such an effect may exist, there is no study yet that documents its existence. [4] Also, the TRA eliminated the 60% deduction for long-term capital gains, effectively resulting in an increase in the capital gains tax rate. Thus, there is a greater incentive for investors to realize losses in the post-TRA period. [5] Consequently, one should observe an increase in January excess stock returns and trading volume in the post-Act period if tax-loss selling is the dominant explanation for the seasonality of stock returns.

We document a pattern of November stock returns during the post-TRA period and thus call it "the November effect." This effect is exclusive to the period following the TRA and thus likely to be the result of tax-loss-selling. Since there is no reason to believe that November is characterized as a month of significant dissemination of information, we argue that the November effect is not due to information effects. We further argue that the November effect is not solely a statistical artifact since the methodology and measurement used in this study are identical for the pre- and post-TRA periods. We document a significant relation between both stock returns and trading volume activity and the tax-loss selling potential of stocks. Based on these findings, we argue that the tax-loss selling can by itself lead to seasonality in stock returns.

We also find that the magnitude of the January effect has increased significantly since the enactment of the TRA and that November trading volume for these firms with the greatest potential for tax-loss selling increased after the enactment of the TRA. Thus, the evidence presented in this paper indicates that the tax-loss-selling hypothesis is likely to be a dominant explanation for the seasonality of stock returns. This paper contributes to the evolution of the literature in this field by documenting a price pressure effect in November and a significant upward shift in prices in January after the passage of TRA in 1986.

The remainder of the paper is organized as follows. Section I describes the data and methodology. Section II presents the empirical evidence of the November and January effects and distinguishes between pre- and post-TRA results. The paper concludes in Section III.

I. Data and Methodology

The data are from the CRSP NYSE/AMEX monthly returns and master files. Monthly returns are analyzed over the period January 1980 through December 1994. The data are split into a pre-TRA period (1980 through 1986) and a post-TRA period (1988 through 1994). Stock returns for 1987 are omitted due to the stock market crash of October 1987. [6] For a firm to be included in the sample, it must have complete data on price, returns, and the number of shares outstanding for the year. This paper includes two analyses: stock return seasonality and trading volume activity.

Portfolio raw and excess risk-adjusted returns are calculated based on size and the potential for tax-loss selling. Risk-adjusted excess returns are calculated using the Scholes-Williams procedure to account for the nonsynchronous trading problem. For each of the years 1980 through 1994, we formed portfolios based on size and tax-loss-selling potential in October of each year. In every year, all securities with available returns are divided equally into 10 size-based portfolios; portfolio 1 contains the smallest stocks and portfolio 10 contains the largest. The size of a firm is defined as the market value of equity at the end of October.

The potential tax-loss-selling measure is determined by computing the stock returns for a six- month period preceding the end of October. Monthly excess returns are computed for November to examine the November effect. Stocks in each size decile are divided into quartiles based on the tax-loss selling measure. Thus, there are a total of 40 portfolios for each year. The portfolios are updated each year. A similar procedure is followed to reexamine the January effect. Portfolios for the January effect are formed in December of the preceding year. [7]

We examine changes in trading volume during the pre- and post-TRA periods for both October and December. Portfolio excess volume is calculated using methodology similar to Dyl (1977). Relative volume for firm i at month t, [RVOL.sub.i,t], is determined by dividing the monthly volume by the mean monthly volume of the preceding 12 months. The relative volume for the firm is regressed with the mean relative market (all NYSE/AMEX stocks) volume, [RVOL.sub.m,t], for month t. A 36-month interval is used to estimate the regression coefficients. The OLS model is of the form:

[RVOL.sub.i,t] = [[alpha].sub.i] + [[beta].sub.i]([RVOL.sub.m,t]) + [[epsilon].sub.i,t] (1)

where [[alpha].sub.i] and are the coefficients for the regression model. These estimates are used to determine monthly excess volume, [EVOL.sub.i,t], for the preceding October and December using the following relationship:

[EVOL.sub.i,t] = [RVOL.sub.i,t] - ([[alpha].sub.i] + [[beta].sub.i]([RVOL.sub.m,t])) (2)

An excess volume of 0.01 indicates that the volume is 1% above the expected level. Similar to the return analysis, firms are grouped into four portfolios formed on the basis of potential tax-loss selling (PTS) on October 31 of each year.

II. Empirical Evidence

This section outlines the empirical evidence inferred from applying the tax-loss-selling hypothesis to the November Effect.

A. Evidence for the November Effect

The tax-loss-selling hypothesis refers to the downward price pressure induced by investors selling stocks to realize losses that offset taxable gains. This pressure is relieved at the beginning of the new tax year. Prior to 1986, this pressure was manifested at the end of December leading to a January effect. The TRA mandated October 31 to be the tax-year end for all mutual funds. Since December is the most common tax-year-end, this created a de facto change in tax-year-end for mutual funds from December to October. This would result in selling pressure in the month of October. The existence of a November return seasonality would provide support for the tax-loss-selling hypothesis.

In this section, we test for the presence of a November effect in stock returns. Evidence for the tax-loss-selling hypothesis implies no return seasonality prior to the TRA and a pattern similar to that observed in January during the post-TRA period. Previous research on the seasonality of stock returns has used size portfolios as a means of testing for information effects. In the same spirit, we examine portfolios split by the potential for tax-loss selling and size. We start by analyzing returns in the pre-TRA period of 1980-1986 using dummy variable regressions by size decile. The dependent variable for the regression is the monthly excess return for a security in November. Dummy variables are used for the potential tax-loss-selling quartiles PTS2, PTS3, and PTS4. The coefficient of the intercept term represents the PTS1 quartile. The dummy variable regression coefficients for the pre-TRA period are reported in Table 1.

There appears to be no clear pattern of November excess returns during the pre-TRA period. The coefficient for PTS1 for the smallest size decile (FS1) is negative though not significantly different from zero.

Further, the coefficient for PTS1 is not significantly different from the coefficient for PTS4 in the same size decile. The results are as expected since there is no reason to observe a tax-induced November effect during the pre-TRA period. However, if tax-loss selling is a significant explanation for seasonality of returns, we should expect to observe a November effect in the post-TRA period. The analysis for the post-TRA period, from 1988 through 1994, is presented in Table 2.

The November excess returns show a generally declining pattern with positive coefficients on PTS1 (portfolios with the largest potential for tax-loss selling) and negative coefficients on PTS4 (portfolios with the smallest tax-loss selling potential). Further, the coefficients on PTS1 for the two smallest size portfolios (FS1 and FS2) are positive and significant while the coefficients on PTS4 for the seven largest portfolios (FS4 through FS10) are negative and significant. [8] The post-TRA November excess returns exhibit a pattern similar to that observed in previous studies of the January effect; [9] positive returns for the smallest-size portfolios and negative returns for the largest-size portfolios. The exceptions are portfolios FS3 through FS10 in the PTS1 quartile. Five of the seven portfolios exhibit significant positive returns. This pattern is consistent with tax-loss selling if mutual funds hold and sell (in October) firms that are large in size but low in price. [10] In an analysis not reported in the paper, we assign firms in the PTS1 quartile to share price deciles and find that the highest returns are concentrated in the fifth and sixth deciles. The mean prices for these deciles are $9.60 and $12.21 respectively, far smaller than the mean price of $48.72 for the highest price decile and the average stock price for S&P 500 firms. These results indicate that lower-priced stocks are generating the high returns in the PTS1 quartile. This is consistent with the evidence in Bhardwaj and Brooks (1992), who show that the high January returns in PTS1 quartile are concentrated in stocks with low stock prices.

Next, we use ordinary-least-squares regression analysis to determine whether the differences between pre- and post-TRA November excess returns by tax-loss-selling quartiles are statistically significant. The November excess return is the dependent variable and a period dummy, that takes the value of 1 for the post-TRA period and 0 for the pre-TRA period, is the independent variable. The intercept of this model represents the pre-TRA November excess returns. The slope coefficient represents the difference between pre- and post-TRA November excess returns. The results of this analysis are presented in Table 3.

The coefficients on dummy are significant in three out of the four quartiles. The coefficient for the first quartile is 0.0159 and is significantly different from zero at the 0.01 level. This indicates that, on average, the post-TRA November excess return is 1.59% greater than the pre-TRA November excess return for the portfolio of firms with the greatest tax-loss-selling potential. This is consistent with price pressure from selling losers prior to the end of October. For the third and fourth quartiles, the dummy coefficients are negative and significant indicating that the post-TRA November excess returns are lower than the pre-TRA November excess returns. We argue that the lower post-TRA November excess returns reflect mutual funds delay in selling prior winners until November to avoid gains distribution. The results support the hypothesis that the TRA altered the pattern of November returns by changing the tax year-end for mutual funds from December to October. Since this effect is only observed after the enactment of the TRA, we can properly term this pattern the November effect.

We also analyze the trading volume activity of portfolios classified by their PTS measure. The October mean excess volume for each PTS portfolio is presented in Table 4.

Excess trading volume for the loser portfolio (PTS1) during the month of October for the pre- and post-TRA period is 0.95% and 3.91%, respectively. As expected, the excess volume for the losers is not significantly different from zero in the pre-TRA period. However, the difference in excess trading volume between pre- and post-TRA periods is 2.96% and is significantly different from zero at the 0.01 level. Further, the excess trading volume for the winner portfolio (PTS4) declined from 16.69% to 8.7%. Again this decline is significant at the 0.01 level. This evidence is consistent with the tax-loss selling hypothesis. If the TRA caused increased tax-loss selling by mutual funds in October, one should observe greater post-TRA trading volume activity in the firms with the greatest potential for tax-loss-selling.

The above analyses document a unique and statistically significant relationship between post-TRA November excess returns and the potential for tax-loss selling. The importance of the month of November is due to the change in year-end for mutual funds to recognize capital gains and losses affected by the TRA. Since the November seasonality is observed only in the post-TRA period, we argue that the November effect is the result of tax-loss selling.

We argue that the November effect documented here does not suffer from confounding problems caused by information releases in the same month. Unlike January, November is not a month of significant disclosure of information and there is no reason to expect that the November effect is primarily due to information releases, especially, when no pattern was detected in the pre-TRA period returns. In any event, previous studies have documented that the information effects have a greater impact on small-firms as they are more likely to be information-poor firms and would benefit more from information releases. Thus, a test for confirming the absence of information effects in November can be performed by analyzing the pattern of stock returns across size deciles. The results presented in Table 1 indicate that there is no systematic pattern of November excess returns across size deciles during the pre-TRA period. We observe positive, though not significant, returns among the largest size deciles, while a majority, four out of six, of the smallest deciles (FS1 through FS6) have negative though insignificant returns. In general, the absence of a pre-TRA-period size effect is consistent with the view that there are no information effects in the November returns. The results indicate that there is no evidence of information effects of the type suggested by Rozeff and Kinney (1976), Williams (1987), Seyhun (1988), and Brauer Chang (1990), nor is there evidence of the existence of the "window dressing" posited by Haugen and Lakonishok (1988).

Finally, we argue that the results of this study are not likely to be related or influenced by the potential biases attributed to earlier work on the January effect. In this paper, the same methodology is used for the pre- and post-TRA period analyses of returns. The fact that the November effect is only present in the post-TRA period suggests that the results are independent of other non-tax explanations, such as the statistical artifact of Roll (1981, 1983a,b) and Blume and Stambaugh (1983), the systematic risk misspecification of Chan, Chen, and Hsieh (1985) and Chan and Chen (1988a,b), and the bid-ask spread explanation of Ball, Kothari, and Shanken (1995). The above explanations may contribute to the January effect, but they are not likely to have a role in the November seasonality documented in this paper.

B. Evidence for the January Effect

The TRA also provides us with an opportunity to investigate further the tax-loss explanation of the January effect. The TRA effectively increased the capital gains tax rate by eliminating the preferential rate for capital gains. This implies that there is a greater incentive for investors, specially, individual investors, to realize losses at the year-end after the enactment of the TRA. [11] If the turn-of-the-year seasonality is mainly a tax-induced anomaly, we would expect to observe an even stronger reaction after the passage of the TRA. [12] We proceed to investigate this hypothesis by analyzing January stock returns for the pre- and post-TRA periods. OLS regression coefficients for the pre-TRA period for January excess returns are reported in Table 5.

The smallest loser portfolios (FS1 and FS2) have significant positive coefficients on PTS1, which generally decline with the potential tax-loss-selling measure, suggesting the same decreasing relationship between excess returns and size deciles found in previous studies. The OLS regression coefficients for January excess returns during the post-TRA period are presented in Table 6.

Similar to the pre-TRA period, the two smallest loser portfolios (FS1 and FS2) have significant positive coefficients on PTS1. Generally, the largest portfolios across tax-loss-selling measures have significant negative excess returns. Reinganum (1983), DeBondt and Thaler (1985), and Chopra, Lakonishok, and Ritter (1992) find a similar pattern of January excess returns. To determine whether the behavior of the January returns after the passage of the TRA is different from the pattern of January returns observed during the pre-TRA period, we run an ordinary least squares regression for each tax-loss-selling quartile. The returns for January are regressed on a dummy variable that takes the value of 1 for the post-TRA period and 0 for the pre-TRA period. The results of this analysis are presented in Table 7.

The coefficient on dummy is significant for three out of the four quartiles. For the first quartile, the coefficient is 0.0331, which is significantly different from zero at the 0.01 level. This indicates that the post-TRA January excess returns are, on average, significantly greater than the pre-TRA January excess returns for the greatest tax-loss-selling portfolio. For the third and fourth quartiles, the dummy coefficients are negative and significant suggesting that the post-TRA returns are actually lower than the pre-TRA returns. The returns follow the expected pattern under the tax-loss-selling framework. Similar to the analysis for the November effect, we determine the excess trading volume in December. The results for mean excess volume by PTS quartiles are reported in Table 8.

The losers portfolio exhibits a significant increase in trading volume from the pre- to the post-TRA period. The difference in excess volume is 7.97%, which is significant at the 0.01 level. In contrast, the winner portfolio has a significant decline in excess volume over the same period of 8.40%. The results presented in this section are consistent with the hypothesis that the TRA altered the pattern of January returns by eliminating the differential taxation of capital gains.

III. Summary and Conclusions

The Tax Reform Act of 1986 effectively changed the tax-year-end for mutual funds from December to October and eliminated the preferential treatment of capital gains. Thus, the TRA provides a unique opportunity to evaluate the tax-loss-selling hypothesis separately from the information and other hypotheses posited in the extant literature.

If tax-loss selling is a potential explanation for the January effect, then one may observe a November effect given that mutual funds are dominant traders of equities. We document a significant pattern in November returns following the change in tax laws, which did not exist prior to the TRA, and we term this pattern the November effect.

We document a unique and significant relationship between excess returns and the potential for tax-loss-selling and conclude that the November effect is explained by the tax-loss-selling hypothesis. Similar results are obtained when an analysis of trading volume is performed. Since the methodology and data measurement are identical for the pre- and post-TRA periods, we argue that the effects documented here are not the result of differences in methodologies. Moreover, we hypothesize that the elimination of the preferential treatment for capital gains should result in a greater January effect, if tax-loss selling influences returns. We find evidence consistent with this hypothesis. Also, an analysis of trading volume indicates that firms with the greatest potential for tax-loss selling experience an increase in volume following the enactment.

Harjeet S. Bhabra is an Assistant Professor at Concordia University. Upinder S. Dhillon is Professor at Binghamton University. Gabriel G. Ramirez is Professor at Virginia Commonwealth University.

The authors thank seminar participants at Condoria and Binghamton Universities, the 1997 Financial Management Association International Meetings, and especially two anonymous referees and the Editors for helpful comments.

(1.) Tax-loss-selling refers to a downward price pressure induced at the end of the tax year when investors sell stocks that have experienced losses during the year. The incentive to sell resides in the ability to utilize capital losses in the current tax year to offset capital gains. This selling pressure is relieved at the beginning of the next tax year and prices bounce back creating positive excess returns.

(2.) Section 852.(3).(c) of the Internal Revenue Code states "the amount of net capital gain for a taxable year shall be determined without regard to any net capital loss attributable to transactions after October 31 of such year, and any such net capital loss shall be treated arising on the first day of the next taxable year."

(3.) According to the NYSE Fact Book 1994, mutual funds currently hold about 10% of all equity securities outstanding. By the end of 1998, there were approximately 12,000 mutual funds, suggesting an increasingly important role for these funds as a distinctive class of institutional investor.

(4.) "Investors see signs that 'January Effect' is already here" Wall Street Journal, November 30, 1992, and Barrons, November 4, 1996.

(5.) This incentive might, to a certain extent, be mitigated because the TRA put a "cap" on the maximum rate for capital gains.

(6.) For the analyses of the January effect data for 1987 are included. Bolster, Lindsey, and Mitrusi (1987) document tax selling in December 1986 and a weakened January effect in 1987. It is possible that because investors were aware of the pending capital gains tax hike they may have begun to act upon it in October. If this were the case, inclusion of 1986 year in the pre-TRA period would create a bias against a significant finding. However, exclusion of 1986 from the analysis does not alter the results and thus we use 1986 in our analysis as part of the pre-TRA period.

(7.) The number of stocks in each portfolio varies from 141 to 377 for the pre-TRA period and from 177 to 651 for the post-TRA period.

(8.) Nonparametric tests were performed on median excess returns for each of the 40 portfolios to determine whether the results are driven by outliers. The evidence suggests that parametric test results are not due to outliers.

(9.) This is confirmed in an analysis, not reported in the paper, where the PTS quartiles are collapsed to examine the size effect.

(10.) A referee suggested this explanation.

(11.) Dyl and Maberly (1992) study odd-lot purchases and sales around the turn of the year and find evidence to support the hypothesis that trading by individual investors is the main reason for the January effect.

(12.) Brauer and Chang (1990) report that in discussions with fund managers, they found that most funds had no capital gains realization and distribution policies prior to the Act. Thus, we should not expect to observe a decline in the January effect due to the shifting of capital gains and losses realizations.

References

Ball, R., S.P. Kothari, and J. Shanken, 1995, "Problems in Measuring Portfolio Performance: An Application to Contrarian Investment Strategies," Journal of Financial Economics (May), 79-107.

Banz, R.W., 1981, "The Relationship between Return and the Market Value of Common Stocks," Journal of Financial Economics (March), 3-18.

Berges, A., J.J. McConnell, and G.G. Schlarbaum, 1984, "The Turn-of-the-Year in Canada," Journal of Finance (March), 185-192

Bhardwaj, R.K. and L.D. Brooks, 1992, "The January Anomaly: Effects of Low Share Price, Transaction Costs, and Bid-Ask Bias," Journal of Finance (June), 553-575.

Blume, M.E. and R.F. Stambaugh, 1983, "Biases in Computed Returns: An Application to the Size Effect," Journal of Financial Economics (November), 387-404.

Bolster, P., L. Lindsey, and A. Mitrusi, 1989, "Tax-Induced Trading: The Effect of the 1986 Tax Reform Act on Stock Market Activity," Journal of Finance (June), 327-344.

Brauer, G.A. and E.C. Chang, 1990, "Return Seasonality in Stocks and their Underlying Assets: Tax-Loss Selling versus Information Explanations," Review of Financial Studies (Summer), 255-280.

Brown, P., D.B. Keim, A.W. Kleidon, and T.A. Marsh, 1983, "Stock Return Seasonalities and the Tax-Loss Selling Hypothesis: Analysis of the Arguments and Australian Evidence," Journal of Financial Economics (June), 105-128.

Chan, K.C. and N. Chen, 1988a, "An Unconditional Asset Pricing Test," Journal of Finance (June), 309-325.

Chan, K.C. and N. Chen, 1988b, "Business Cycles and the Returns of Small and Large Firms," University of Chicago Unpublished Manuscript.

Chan, K.C., N. Chen, and D. Hsieh, 1985, "An Exploratory Investigation of the Firm Size Effect," Journal of Financial Economics (September), 451-471.

Chopra, N., J. Lakonishok, and J.R. Ritter, 1992, "Measuring Abnormal Performance: Do Stocks Overreact?" Journal of Financial Economics (April), 235-268.

DeBondt, W.F.M, and R. Thaler, 1985, "Does the Stock Market Overreact?" Journal of Finance (July), 793-805.

DeBondt, W.F.M. and R. Thaler, 1987, "Further Evidence on Investor Overreaction and Stock Market Seasonality," Journal of Finance (July), 557-582.

Dyl, E.A., 1977, "Capital Gains Taxation and Year-End Stock Market Behavior," Journal of Finance (March), 65-98.

Dyl, E.A., 1979, "A State Preference Model of Capital Gains Taxation," Journal of Financial and Quantitative Analysis (September), 529-536.

Dyl, E.A. and E. D. Maberly, 1992, "Odd-Lot Transactions Around the Turn of the Year and the January Effect," Journal of Financial and Quantitative Analysis (December), 591-604.

Haugen, R.A. and J. Lakonishok, 1998, The Incredible January Effect: The Stock Market's Unsolved Mystery, Homewood, IL, Dow-Jones, Irwin.

Jones, S.L. and W. Lee, 1995, "Evidence on the Behavior of Bid and Ask Prices at the Turn of the Year: Implications for the Survival of Stock Return seasonality," Journal of Financial Research (Winter), 383-400.

Jones, S.L., W. Lee, and R. Apenbrink, 1991, "New Evidence on the January Effect before Personal Income Taxes," Journal of Finance (December), 1909-1925.

Keim, D.B., 1983, "Size-Related Anomalies and Stock Return Seasonality: Further Empirical Evidence," Journal of Financial Economics (June), 12-32.

Lakonishok, J. and S. Smidt, 1984, "Volume and Turn-of-the-Year Behavior," Journal of Financial Economics (September), 535-556.

Lakonishok, J. and S. Smidt, 1986, "Volume for Winners and Losers: Taxation and Other Motives for Stock Trading," Journal of Finance (September), 951-974.

Maxwell, W., 1998, "The January Effect in the Corporate Bond Market: A Systematic Examination," Financial Management (Summer), 18-30.

Reinganum, M.R., 1981, "Misspecification of Capital Asset Pricing: Empirical Anomalies Based on Earnings Yield and Market Values," Journal of Financial Economics (March), 19-46.

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 (June), 89-104.

Roll, R., 1981, "A Possible Explanation of the Small Firm Effect," Journal of Finance (September), 879-888.

Roll, R., 1983a, "Vas Ist Das? The Torn of the Year Effect and the Return Premia of Small Firms," Journal of Portfolio Management (Winter), 18-28.

Roll, R., 1983b, "On Computing Mean Returns and the Small Firm Effect," Journal of Financial Economics (November), 371-386.

Rozeff, M.S. and W. R. Kinney, 1976, "Capital Market Seasonality: The Case of Stock Returns," Journal of Financial Economics (October), 379-402.

Schultz, P., 1985, "Personal Income Taxes and the January Effect and Small Firm Stock Returns Before the War Revenue Act of 1917: A Note," Journal of Finance (March), 333-343.

Seyhun, H.N., 1988, "The January Effect and Aggregate Insider Trading," Journal of Finance (March), 129-141.

Wachtel, S., 1942, "Certain Observations on Seasonal Movement in Stock Prices," Journal of Business (April), 184-193.

Williams, J., 1987, "Financial Anomalies Under Rational Expectations: A Theory of the Annual Size and Related Effects," New York University Working Paper.

Ordinary-Least-Squares Regression Estimates for November: Pre-TRA Period (1980-86)

This table estimates OLS regressions to test for the presence of the November Effect in the Pre-TRA Period (1980-86). Regression equations are estimated for each decile based on the market capitalization of the firm. The regression equation is of the form:

[R.sub.j] = [[beta].sub.1] + [[beta].sub.2][D.sub.2] + [[beta].sub.3][D.sub.3] + [[beta].sub.4][D.sub.4] + [[epsilon].sub.j]

where [R.sub.j] = excess return for security j in November, and [D.sub.j] (j=2,3,4) are dummy variables based on the quartile for potential tax-loss-selling (PTS). PTS is based on a six-month holding period return prior to October 31, the portfolio formation date. PTS1 is the quartile that represents losers with the greatest potential for tax-loss-selling while PTS4 represents winners. The dummy variable [D.sub.2] one if the firm belongs to the PTS2 quartile and zero otherwise. Similarly, dummy variables [D.sub.3] and [D.sub.4] are one if the firm belongs to the PTS3 and PTS4 quartiles respectively. By construction, the intercept term [[beta].sub.1] measures the excess returns for firms in the PTS1 quartile. t-values are in parentheses.

Size Decile [[beta].sub.1] [[beta].sub.2] [[beta].sub.3] [[beta].sub.4] FS1 -0.0053 0.0210 [***] -0.0023 -0.0024 (smallest) (0.723) (4.621) (0.156) (0.171) FS2 0.0016 -0.0005 0.0081 -0.0039 (0.250) (0.053) (0.678) (0.388) FS3 0.0034 -0.0046 -0.0104 -0.0061 (0.500) (0.460) (0.948) (0.638) FS4 -0.0026 -0.0075 0.0131 0.0049 (0.426) (0.816) (1.290) (0.561) FS5 -0.0101 0.0037 0.0121 0.0108 (1.598) (0.422) (1.286) (1.254) FS6 -0.0061 0.0043 0.0048 -0.0001 (1.039) (0.534) (0.561) (0.009) FS7 0.0130 [**] -0.0085 -0.0141 -0.0076 (2.079) (1.030) (1.578) (0.944) FS8 0.0033 0.0004 0.0013 0.0106 (0.531) (0.045) (0.146) (1.273) FS9 -0.0011 -0.0005 0.0036 0.0001 (0.181) (0.063) (0.442) (0.016) FS10 0.0087 0.0083 -0.0054 0.0021 (largest) (1.370) (1.040) (0.648) (0.258) Size Decile Adj. [R.sup.2] FS1 0.0004 (smallest) FS2 -0.0022 FS3 -0.0022 FS4 0.0012 FS5 -0.0006 FS6 -0.0026 FS7 -0.0005 FS8 -0.0006 FS9 -0.0029 FS10 0.0008 (largest) (***.)Significant at the 0.01 level. (**.)Significant at the 0.05 level.

Ordinary-Least-Squares Regression Estimates for November: Post-TRA Period (1988-94)

This table estimates OLS regressions to test for the presence of the November Effect in the Post-TRA Period (1988-94). Regression equations are estimated for each decile based on the market capitalization of the firm. The regression equation is of the form:

[R.sub.j] = [[beta].sub.2] + [[beta].sub.2][D.sub.2] + [[beta].sub.3][D.sub.3] + [[beta].sub.4][D.sub.4] + [[epsilon].sub.j]

where [R.sub.j] = excess return for security j in November, and [D.sub.j] (j=2,3,4) are dummy variables based on the quartile for potential tax-loss selling (PTS). PTS is based on a six-month holding period return prior to October 31, the portfolio formation date. PTS1 is the quartile that represents losers with the greatest potential for tax-loss selling while PTS4 represents winners. The dummy variable [D.sub.2] one if the firm belongs to the PTS2 quartile and zero otherwise. Similarly, dummy variables [D.sub.3] and [D.sub.4] are one if the firm belongs to the PTS3 and PTS4, quartiles respectively. By construction, the intercept term [[beta].sub.1] measures the excess returns for firms in the PTS1 quartile. t-values are in parentheses.

Size Decile [[beta].sub.1] [[beta].sub.2] [[beta].sub.3] FS1 0.0155 [**] -0.0199 -0.0333 (smallest) (2.010) (1.305) (1.665) FS2 0.0160 [**] 0.0176 -0.0177 (2.262) (1.636) (1.504) FS3 0.0104 -0.0082 -0.0186 (1.729) (0.942) (0.985) FS4 0.0198 [***] -0.0241 [***] -0.0284 [***] (3.309) (2.969) (3.440) FS5 0.0117 -0.0175 -0.0177 (1.629) (1.767) (1.793) FS6 0.0248 [***] -0.0330 [***] -0.0278 [***] (4.271) (4.127) (3.542) FS7 0.0124 -0.0148 -0.0226 [***] (1.838) (1.638) (3.542) FS8 0.0181 [***] -0.0154 [**] -0.0204 [***] (3.162) (2.070) (2.738) FS9 0.0117 [**] -0.0123 -0.0216 [***] (2.267) (1.896) (3.452) FS10 0.0142 [***] -0.0172 [***] -0.0214 [***] (largest) (3.065) (3.040) (3.994) Size Decile [[beta].sub.4] Adj. [R.sup.2] FS1 -0.0098 0.0006 (smallest) (0.549) FS2 -0.0189 0.0010 (1.596) FS3 -0.0138 -0.0005 (1.469) FS4 -0.0222 [***] 0.0083 (2.671) FS5 -0.0291 [***] -0.0046 (2.986) FS6 -0.0478 [***] 0.0252 (5.944) FS7 -0.0478 [***] 0.0052 (5.944) FS8 -0.0349 [***] 0.0168 (4.886) FS9 -0.0234 [***] 0.0107 (3.745) FS10 -0.0304 [***] 0.0227 (largest) (5.638) (***.)Significant at the 0.01 level. (**.)Significant at the 0.05 level.

Ordinary-Least-Squares Regression Estimation for the Change in November Effect Between the Pre- and Post-TRA Period

This table estimates an OLS regression to test for the change in the November returns between the pre- and the post-TRA periods. Regression equations are estimated for firms belonging to the different quartiles formed on the basis of the potential tax-loss-selling (PTS) measure. The regression is of the form:

[R.sub.j] = [[beta].sub.0] + [[beta].sub.t] DUMMY + [[epsilon].sup.j]

where:

[R.sub.j] is the monthly excess return in November for firm j

DUMMY equals 1 for post-Act period (1988-1994) and 0 for the pre-Act period (1980-1986) t values are in parentheses.

Potential Tax-Loss Selling Measure [[beta].sub.0] [[beta].sub.1] PTS1 -0.0003 0.0159 [***] (Losers) (0.109) (3.951) PTS2 0.0022 -0.0047 (1.072) (1.703) PTS3 0.0016 -0.0076 [***] (0.821) (3.176) PTS4 0.0015 -0.0132 [***] (Winners) (0.754) (4.901) Potential Tax-Loss Selling Measure Adj. [R.sup.2] PTS1 0.0025 (Losers) PTS2 0.0003 PTS3 0.0018 PTS4 0.0040 (Winners) (***.)Significant at the 0.01 level.

Mean Portfolio Excess Trading Volume for October Due to Potential Tax-Loss Selling

Portfolio excess trading volume is calculated for the month of October using a methodology similar to Dyl (1977). Relative volume for a firm is determined by dividing the monthly trading volume by the mean monthly trading volume for the preceding 12 months. The relative trading volume for the firm is regressed on the mean market volume for the month. A 36-month interval, preceding October, is used to estimate the regression coefficients. These estimates are used to determine monthly excess trading volume for October. Firms are grouped into four portfolios formed on the basis of the potential tax-loss selling (PTS) measure on October 31 of each year.

Pre-TRA Period Post-TRA Period Difference Portfolios (1980-1986) (1988-1994) PTS1 (Losers) 0.0095 0.0391 0.0296 [**] PTS2 -0.0998 -0.0363 0.0635 [***] PTS3 -0.0352 -0.0641 -0.0289 [**] PTS4 (Winners) 0.1669 0.0871 -0.0798 [***] (***.)Significant at the 0.01 level. (**.)Significant at the 0.05 level.

Ordinary-Least-Squares Regression Estimates for January: Pre-TRA Period (1980-87)

This table estimates OLS regressions to test for the presence of the January Effect in the Pre-TRA Period (1980-87). Regression equations are estimated for each decile based on the market capitalization of the firm. The regression equation is of the form:

[R.sub.j] = [[beta].sub.1] + [[beta].sub.2][D.sub.2] + [[beta].sub.3][D.sub.3] + [[beta].sub.4][D.sub.4] + [[epsilon].sub.j]

where [R.sub.j] = excess return for security j in November, and D. (j=2,3,4) are dummy variables based on the quartile for potential tax-loss selling (PTS). PTS is based on a six-month holding period return prior to December 31 of the previous year, the portfolio formation date. PTS1 is the quartile that represents losers with the greatest potential for tax-loss selling while PTS4 represents winners. The dummy variable [D.sub.2] is one if the firm belongs to the PTS2 quartile and zero otherwise. Similarly, dummy variables [D.sub.3] and [D.sub.4] are one if the firm belongs to the PTS3 and PTS4 quartiles, respectively. By construction, the intercept term [[beta].sub.1] measures the excess returns for firms in the PTS1 quartile. t-values are in parentheses.

Size Decile [[beta].sub.1] [[beta].sub.2] [[beta].sub.3] FS1 0.0429 [***] -0.0079 -0.0057 (smallest) (6.143) (0.664) (0.422) FS2 0.0226 [***] -0.0139 0.0012 (3.687) (1.460) (0.119) FS3 0.0076 -0.0043 -0.0058 (1.207) (0.470) (0.604) FS4 0.0041 -0.0007 -0.0166 (0.690) (0.083) (1.887) FS5 0.0013 -0.0016 -0.0033 (0.231) (0.199) (0.406) FS6 0.0157 [**] -0.0210 [**] -0.0249 [***] (2.386) (2.400) (2.886) FS7 -0.0114 0.0142 -0.0098 (1.883) (1.764) (1.219) FS8 -0.0120 [**] 0.0167 [**] 0.0008 (1.973) (2.092) (0.101) FS9 0.0063 -0.0158 [**] -0.0178 [**] (1.025) (2.055) (2.358) FS10 -0.0110 0.0070 0.0099 (largest) (1.799) (0.895) (1.346) Size Decile [[beta].sub.4] Adj. [R.sup.2] FS1 -0.0243 -0.0002 (smallest) (1.668) FS2 -0.0225 [**] 0.0035 (2.241) FS3 -0.0183 [**] 0.0010 (1.980) FS4 -0.0189 [**] 0.0041 (2.171) FS5 -0.0127 0.0002 (1.596) FS6 -0.0264 [***] 0.0072 (3.068) FS7 0.0029 0.0065 (0.372) FS8 -0.0103 0.0105 (1.340) FS9 -0.0152 [**] 0.0028 (1.994) FS10 -0.0009 0.0010 (largest) (0.123) (***.)Significant at the 0.01 level. (**.)Significant at the 0.05 level.

Ordinary-Least-Squares Regression Estimates for January: Post-TRA Period (1988-94)

This table estimates OLS regressions to test for the presence of the January Effect in the Post-TRA Period (1988-94). Regression equations are estimated for each decile based on the market capitalization of the firm. The regression equation is of the form:

[R.sub.j] = [[beta].sub.1] + [[beta].sub.2][D.sub.2] + [[beta].sub.3][D.sub.3] + [[beta].sub.4][D.sub.4] + [[epsilon].sub.j]

where [R.sub.j] = excess return for security j in November, and [D.sub.j] (j=2,3,4) are dummy variables based on the quartile for potential tax-loss selling (PTS). PTS is based on a six-month holding period return prior to December 31 of the previous year, the portfolio formation date. PTS1 is the quartile that represents losers with the greatest potential for tax-loss selling while PTS4 represents the winners. The dummy variable [D.sub.2] is one if the firm belongs to the PTS2 quartile and zero otherwise. Similarly, dummy variables [D.sub.3] and [D.sub.4] one if the firm belongs to the PTS3 and PTS4 quartiles, respectively. By construction, the intercept term [[beta].sub.1] measures the excess returns for firms in the PTS1 quartile. t-values are in parentheses.

Size Decile [[beta].sub.1] [[beta].sub.2] [[beta].sub.3] FS1 0.1877 [***] -0.0998 [**] -0.1235 [***] (smallest) 9.542) (2.368) (2.616) FS2 0.0471 [***] -0.0257 [**] -0.0395 [***] (6.575) (2.342) (3.376) FS3 -0.0035 0.0043 0.0125 (0.537) (0.448) (1.334) FS4 0.0021 0.0067 -0.0079 (0.290) (0.697) (0.794) FS5 -0.0109 -0.0006 -0.0071 (1.255) (0.045) (0.595) FS6 -0.0071 0.0047 -0.0133 (1.028) (0.504) (1.427) FS7 -0.0152 [**] -0.0021 -0.0094 (2.283) (0.241) (1.072) FS8 -0.0093 -0.0177 [**] -0.0207 [**] (1.425) (2.074) (2.434) FS9 -0.0246 [***] 0.0062 -0.078 (3.979) (0.804) (1.017) FS10 -0.0099 -0.0241 [***] -0.0223 [***] (largest) (1.475) (3.008) (2.872) Size Decile [[beta].sub.4] Adj. [R.sup.2] FS1 -0.1095 [**] 0.0080 (smallest) (2.320) FS2 -0.0357 [***] 0.0096 (2.841) FS3 -0.0096 0.0017 (0.962) FS4 -0.0219 [**] 0.0054 (2.181) FS5 -0.0131 -0.0011 (1.078) FS6 -0.0268 [***] 0.0092 (2.850) FS7 -0.0175 [**] 0.0020 (2.035) FS8 -0.0261 [***] 0.0060 (3.172) FS9 -0.0153 [**] 0.0075 (2.023) FS10 -0.0278 [***] 0.0080 (largest) (3.586) (***.)Significant at the 0.01 level. (**.)Significant at the 0.05 level.

Ordinary-Least-Squares Regression Estimation for the Change in January Effect Between the Pre- and Post-TRA Period

This table estimates a OLS regression to test for the change in January returns between the pre- and the post-TRA periods. Regression equations are estimated for firms belonging to the different quartiles formed on the basis of the potential tax-loss-selling (PTS) measure. The regression is of the form:

[R.sub.j] = [[beta].sub.0] + [[beta].sub.1]DUMMY + [[epsilon].sub.j]

where:

[R.sub.j] is the monthly excess return in January for firm j

DUMMY equals 1 for the post-TRA period (1988-1994) and 0 for the pre-TRA period (1980-1987) t-values are in parentheses.

Potential Tax-Loss-Selling Measure [[beta].sub.0] [[beta].sub.1] PTS1 (Losers) 0.0118 [**] 0.0331 [***] (2.292) (4.672) PTS2 0.0036 -0.0059 (1.635) (1.922) PTS3 -0.0026 -0.0108 [***] (1.465) (4.365) PTS4 (Winners) -0.0098 [***] -0.0126 [***] (4.782) (4.406) Potential Tax-Loss-Selling Measure Adj. [R.sup.2] PTS1 (Losers) 0.0035 PTS2 0.0005 PTS3 0.0030 PTS4 (Winners) 0.0031 (***.)Significant at the 0.01 level. (**.)Significant at the 0.05 level.

Mean Portfolio Excess Trading Volume for December Due to Potential Tax-Loss Selling

Portfolio excess trading volume is calculated for the month of December using a methodology similar to Dy1 (1977). Relative trading volume for a firm is determined by dividing the monthly trading volume by the mean monthly trading volume for the preceding 12 months. The relative trading volume for the firm is regressed on the mean market trading volume for the month. A 36-month interval, preceding October, is used to estimate the regression coefficients. These estimates are used to determine monthly excess trading volume for October. Firms are grouped into four portfolios formed on the basis of potential tax-loss-selling (PTS) measure on October 31 of each year.

Pre-TRA Period Post-TRA Period Difference Portfolios (1980-1987) (1988-1994) PTS1 (Losers) 0.0956 0.1753 0.0797 [***] PTS2 -0.0657 -0.0428 0.0229 PTS3 -0.0754 -0.1048 -0.0294 [**] PTS4 (Winners) 0.0749 0.0091 -0.0840 [***] (***.)Significant at the 0.01 level. (**.)Significant at the 0.05 level.

The prevalence of significant excess returns and volume during the month of January (called the "January Effect") is well documented in extant finance literature. One of the earliest explanations for this phenomenon is tax-loss-selling, which refers to the downward price pressure induced by investors selling stocks to realize losses that offset taxable gains. This pressure is relieved at the beginning of the new tax year. However, in recent years, several researchers have proposed alternative explanations such as investor overreaction, information seasonality, misspecification of systematic risk and statistical and data biases.

This paper exploits a unique opportunity, provided by the Tax Reform Act of 1986, and provides support for the tax-loss-selling hypothesis. The Tax Reform Act of 1986 mandated October 31 to be the tax-year end for all mutual funds and eliminated the preferential treatment of capital gains. Since December is the most common tax-year end, this created a defacto change in tax year-end for mutual funds from December to October. The contribution of this study is twofold. First, we document significant excess stock returns and volume during the month of November. This phenomenon is observed only in the post-Act period and we call it the "November Effect." Second, we argue that the elimination of the preferential treatment of capital gains would lead to a stronger "January Effect." We document that the magnitude of the January effect has increased significantly since the enactment of the Tax Reform Act.

Based on the significant relation between stock returns and trading volume activity, and the tax-loss-selling potential of stocks found in this study, we argue that the tax-loss-selling can by itself lead to seasonality in stock returns. This paper contributes to the evolution of the literature in the field of stock seasonalities by documenting a price pressure effect in November and a significant increase in the January effect after the passage of Tax Reform Act in 1986.

The tax-loss-selling hypothesis implies November excess returns and trading volume seasonality following the Act. We find that excess stock returns for firms with the highest (lowest) potential for tax-loss-selling are significantly positive (negative) during the month of November. We also find a significant increase (decrease) in trading volume activity during the month of November for firms with the highest (lowest) potential for tax-loss-selling. These effects are exclusive to the period following the Tax Reform Act, and are therefore likely to be the result of tax-loss-selling. This is so because the Tax Reform Act provides mutual funds with an incentive to sell loosing stocks before October 31 to reduce the taxable gains they are required to pass on to shareholders. Mutual funds are also likely to postpone the sale of winners until after the tax year-end. Since there is no reason to believe that November is characterized as a month during which there is significant dissemination of information, we argue that the November effect is not due to information seasonality, an alternative hypothesis in the literature. We further argue that the November effect is not solely due to statistical or data biases, since the methodology and measurement used in this study is identical for the pre- and post-Act periods.

We also find that post-Act January excess returns and trading volume are, on average, significantly greater than the pre-Act January excess returns and volume for firms with the greatest tax-loss-selling potential. The Tax Reform Act effectively eliminated preferential treatment of long-term capital gains, resulting in an increase in the capital gains tax rate, thus, there is a greater incentive for investors to realize losses in the post-Act period. The price pressure and trading activity documented in this study are consistent with the tax-loss-selling hypothesis.

Printer friendly Cite/link Email Feedback | |

Author: | Bhabra, Harjeet S.; Dhillon, Upinder S.; Ramirez, Gabriel G. |
---|---|

Publication: | Financial Management |

Article Type: | Statistical Data Included |

Geographic Code: | 1USA |

Date: | Dec 22, 1999 |

Words: | 8321 |

Previous Article: | Open Market Stock Repurchase Signaling. |

Next Article: | The Information Content of Dividend and Capital Structure Policies. |

Topics: |