The turn-of-the-month (TOM) effect is observed as one of the seasonalities in many markets. The author examines the TOM effect in the KOSDAQ market and finds that the effect is significant. The TOM effect in the KOSDAQ market is not due to size, turn-of-the-year, turn-of-the-quarter or index rebalancing effect. The author also finds that individual and institutional traders do not trade and buy more stocks at the TOM than on the rest days, not consistent with existing explanations of the increased liquidity by individual investors or institutional window-dressing activity. When the author investigated the net buying volume and net turnover of each investor, the net volume and turnover of individual investors at the TOM were significantly lower than those on the other days, rejecting the hypothesis of their increased demand. Interestingly, net foreign volumes at the TOM are significantly higher than on the other days. Finally, using panel regressions, the author finds that stocks with a higher net buying volume of foreigners for the TOM period tend to have higher returns, while stocks with a higher net buying volume of individual traders for the TOM period are likely to have lower returns. The results confirm that the TOM effect is not due to the increased demand of individual investors. Instead, higher net buying volume by foreigners may partially cause the TOM effect. Therefore, this study contributes to the literature by revealing the presence of the TOM effect in the KOSDAQ market and the foreign role in the anomaly in the market even mainly traded by retail investors.
1. Introduction
The turn-of-the-month (TOM) effect, one of the return seasonal anomalies, is defined as the tendency of stock returns to surge during a period encompassing the last trading day of the month and the first few trading days of the next month. This effect is observed in most markets [1]. Ariel (1987) and Lakonishok and Smidt (1988) report the TOM effect on the US stock market, and Cadsby and Ratner (1992) find the significant TOM effect in 6 of 10 international markets. Agarwal and Tandon (1994) document the tendency of stock returns to surge at the TOM in 18 developed markets. Compton (2002) also reports the TOM effect in Pacific Rim countries, and Aziz and Ansari (2017) show it in Asia–Pacific markets.
I examine the TOM effect in the KOSDAQ market, and there is little literature on the TOM effect in the KOSDAQ market. For Korean stock markets, Yun and Kim (2014), Hong et al. (2014), Aziz and Ansari (2017), and Lee and Kim (2022) study the TOM effect in the KOSPI market, and they find the significantly high returns on the last and first days of the month. Only Hong et al. (2015) investigate the TOM anomaly in the KOSDAQ.
Another reason why I examine the TOM in the KOSDAQ is that this market has a proper circumstance to explore the hypotheses for this anomaly. Although many studies confirm the TOM anomaly, the reason for this anomaly is not pronounced. The representative explanation is the increased liquidity of investors around the TOM period. Ogden (1990) argues that as individual investors get usually paid at the month-end, the liquidity and demand from them around the TOM are increased. Booth et al. (2001) also show that the trading volumes are larger over the TOM in the Finish market. However, McConnell and Xu (2008) and Maher and Parikh (2013) do not find the result supporting this explanation [2]. As another explanation, Thaler (1987) and Barone (1990) argue that institutional investors trade more stocks at the end of the month to boost their performance as window dressing. The KOSDAQ market has a far higher proportion of individual trading than other markets do. On average, individual trading volumes account for 89.7% of total trading volume in the KOSDAQ market from 2000 through 2020, the sample period in this paper [3]. If the increased liquidity or demand from individual investors induces the TOM effect, I may observe a strong TOM anomaly in the KOSDAQ market. Further, I use the total trading and net buying volumes of individuals, institutions and foreigners in this study, thus, it is possible to investigate the impact of trading volume by each investor on the anomaly. If the increased liquidity and demand of individual or institutional investors cause the TOM effect in the KOSDAQ market, I can observe their higher buying or trading volume around the TOM.
Using value- and equal-weighted portfolios with all stocks in the KOSDAQ, I find the significantly positive TOM effect. The returns on the first and last trading days of the month are the highest and significantly positive. The average returns on the four days between the last and the first three days of the month which many prior researchers define as the TOM period are also significantly positive in the market. When I divide the whole stocks into small, medium and big stocks based on the firm size, the evidence indicates that the TOM anomaly is not confined to small firms. I also find that this significant TOM effect is not due to other calendar effects such as the turn-of-the-year effect (Rozeff and Kinney, 1976; Roll, 1983; Lakonishok and Smidt, 1984), the turn-of-the-quarter effect (Bernhardt and Davies, 2005; Carhart et al., 2002), or the index rebalancing effect (Carhart et al., 2002).
As existing hypotheses suggest the surge of trading volume by individual or institutional traders at the TOM, I examine the total trading volume and share turnover of all stocks in the market to test these explanations. The results show that they are not higher at the TOM than at the rest-of-the-month (ROM), but lower. When I calculate trading volumes of individual, institutional and foreign investors, trading volumes by the individual and institutional traders at the TOM are insignificantly lower than those at the ROM. Further, the foreign trading volume at the TOM is insignificantly different from those at ROM. These results indicate that the TOM effect in the KOSDAQ market is not due to the increased liquidity of investors.
I also calculate net buying volumes and net turnovers of individual, institutional and foreign investors, and find that individual net buying volume and net turnover at the TOM of two days between the last day of the month and the first day of the next month are significantly lower than those at the ROM. As a result, the hypothesis of the increased demand of individual investors for the TOM effect is rejected in this study. As almost institutional net buying volume and net turnover at the TOM are also insignificantly different from those at ROM, the explanation that the TOM effect is associated with the increased liquidity of institutional investors as a result of the institutional window-dressing activity is not appropriate in the KOSDAQ market as well. Interestingly, foreign net buying volume and net turnover at the TOM starting the last day of the month and ending the first day of the next month are significantly higher.
Finally, I examine the relationship between the stock return, the TOM period, and net buying volume by investors using panel regressions with the firm- and day-fixed effects. I find that excess returns on individual stocks are higher during the TOM period after controlling variables related to stock returns such as firm size and market-to-book ratio and the turn-of-the-year effect. Further, stocks with a higher net buying volume of foreigners for the TOM period tend to have higher returns, while stocks with a higher net buying volume of individual traders for the TOM period are likely to have lower returns. The results confirm that the TOM effect is not due to the increased demand of individual investors. Instead, higher net buying volume by foreigners may partially cause the TOM effect.
Lee and Kim (2022) also find significantly higher market returns at the TOM than at the ROM, in the Korean market from 2020 through 2020, but they use KOSPI stocks. The spread in this study using value-weighted portfolios of KOSDAQ stocks is 1.8 times higher than their spread. I find significantly higher net buying volume and net turnover of foreign investors at the TOM period of two days between the last and first days of the month [−1, +1], although they do insignificantly higher net turnover of foreigners at the same TOM period. Further, the TOM effect for large firms is not different from the effect for small firms in Lee and Kim (2022), when I use KOSDAQ stocks that tend to be less-established small- and medium-sized firms, the TOM effect for large firms is significantly stronger than that for small firms, at the TOM period of two days between the last and first days of the month [−1, +1]. Lee and Kim (2022) do not find what induces the TOM effect, and I perform additional analyses to reveal the reason for the anomaly.
As I find foreign traders and a large-size effect on the TOM anomaly in the KOSDAQ market, I run panel regressions using only a sample at the TOM, excluding data for the ROM, and find large firms with higher foreign net buying volume have higher TOM returns. Additionally, I perform panel regressions using stock-day observations with zero foreign trading volume and find significantly higher returns at the TOM. However, when I add the interaction terms between the net buying volume of investors or size and the TOM dummy, the coefficients of TOM dummies become insignificant. As a result, I confirm that foreign trading partially causes the TOM anomaly, and find that after eliminating the foreign trading effect, the TOM anomaly is associated with institutional trading, size and the turn-of-the-year effect.
This study contributes to the literature by revealing the presence of the TOM effect in the KOSDAQ market and the foreign role in the anomaly in the market even mainly traded by retail investors. Prior studies using Korean stocks examine only whether there is the TOM effect in the market. However, this study calculates the trading volume and net buying volume for each investor type and shows which investors trade more stocks at the TOM to potentially affect the TOM anomaly. The results in this paper indicate that the existing explanation that a surge of trading volume by individual or institutional traders at the TOM causes the anomaly is rejected.
The rest of the paper is organized as follows. Section 2 describes the data and defines the TOM period. Section 3 shows the significant TOM effect in the KOSDAQ market. Section 4 examines trading volume, share turnover and net buying volume by each investor type at the TOM. Section 5 runs panel regressions of excess returns on the TOM period and net buying volume by investors. Section 6 concludes the study.
2. Data and TOM period
I use all stocks listed on the KOSDAQ market from January 2000 to December 2020. I exclude data over two years before firms are delisted. Consequently, 1,986 firms are included in this study. All financial data used in this paper, including returns, trading volumes, shares traded by investors and firm fundamentals are extracted from DataGuide.
Although there is no consensus regarding the precise TOM period, most studies such as Lakonishok and Smidt (1988), Agarwal and Tandon (1994) and McConnell and Xu (2008) define the TOM period as the four days between the last and first three days of the month [–1, +3] where −1 denotes the last trading day of the month, +1 is the first trading day of the month, +2 is the second trading day of the month and so on. For consistency with these prior studies, I shall construe the TOM as encompassing day −1 through day +3 as the common TOM period, which is expressed as TOM [–1, +3]. Consequently, ROM [–1, +3] is defined as the rest of the month (ROM) except the TOM period. Additionally, as Yun and Kim (2014), Aziz and Ansari (2017) and Lee and Kim (2022) show that the TOM period with significantly positive average daily returns in the KOSPI market is the last and first days of the month, [–1, +1], I use the interval [–1, +1] of the month as an alternative TOM period, which is expressed as TOM [–1, +1]. ROM [–1, +1] is the rest of the month except TOM [–1, 1].
3. Returns at the turn of the month
First of all, to see the daily return pattern over the month, I plot the value- and equal-weighted average daily returns by day of the month from day −10 through day +10, in Figure 1. The figure shows that the value-weighed (VW) and equal-weighted (EW) average returns are the highest on day +1 and the second-highest on the day −1 in the KOSDAQ market. VW and EW returns for the interval [+2, +3] are also positive, hence returns for the TOM [–1, +3] would be higher than those for the ROM [–1, +3]. Overall, returns for the first half of the month [+1, +10] are positive, but returns for the latter half of the month [–10, −1] are negative.
Panel A of Table 1 presents the numerical values of the VW and EW average daily returns from day −9 through +9. The VW returns are significantly positive only on days −1 and +1. The highest VW return is 0.472% on day +1, which is higher than the VW return (0.312%) in the KOSPI market reported by Lee and Kim (2022). The VW returns on days +2 and + 3 are also positive but insignificant. The EW average returns on days −1, +1, and +2 at the TOM period are significantly positive. The significantly positive EW returns are clustered around the TOM and the first half of the month.
Returns around the turn of the month
| −9 | −8 | −7 | −6 | −5 | −4 | −3 | −2 | −1 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Panel A: Average returns by trading day of the month | ||||||||||||||||||
| VW | −0.097 (−0.80) | 0.036 (0.33) | −0.356 (−3.36) | −0.218 (−1.79) | −0.194 (−1.54) | −0.057 (−0.51) | −0.280 (−2.82) | 0.109 (0.94) | 0.400 (3.95) | 0.472 (4.09) | 0.087 (0.93) | 0.024 (0.21) | 0.121 (1.08) | 0.149 (1.42) | −0.049 (−0.45) | −0.080 (−0.75) | 0.045 (0.37) | 0.185 (1.72) |
| EW | −0.068 (−0.59) | 0.099 (0.96) | −0.307 (−2.85) | −0.289 (−2.55) | −0.183 (−1.49) | −0.032 (−0.29) | −0.174 (−1.67) | 0.174 (1.61) | 0.374 (3.91) | 0.527 (4.84) | 0.255 (2.95) | 0.151 (1.41) | 0.204 (2.02) | 0.189 (1.92) | 0.076 (0.73) | 0.029 (0.28) | 0.189 (1.82) | 0.237 (2.41) |
| TOM [–1, +3] | ROM [–1, +3] | Diff. | TOM [–1, +1] | ROM [–1, +1] | Diff. | |
|---|---|---|---|---|---|---|
| Panel B: Returns at the TOM and ROM | ||||||
| VW | 0.243*** (3.46) | −0.049 (−1.35) | 0.292*** (3.69) | 0.431*** (5.25) | −0.037 (−1.03) | 0.468*** (5.22) |
| EW | 0.331*** (5.02) | 0.006 (0.17) | 0.325*** (4.33) | 0.455*** (5.97) | 0.028 (0.82) | 0.427*** (5.11) |
Note(s): This table reports returns around the turn of the month. Panel A presents the value-weighted (VW) and equal-weighted (EW) average returns on the KOSDAQ market by trading day of the month. Panel B presents average daily returns at the TOM and ROM, and their differences. t-statistics are in parentheses. In Panel A, numbers in italic denote statistical significance at the 5% level. In Panel B, *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. The sample period is from 2000 to 2020
Panel B of Table 1 shows the differences between returns at the TOM and ROM. The VW and EW average daily returns at the TOM [–1, +3] and TOM [–1, +1] are significantly positive, but those for the ROM [–1, +3] and ROM [–1, +1] are insignificantly negative. TOM [–1, +1] returns are higher than TOM [–1, +3] as expected from Figure 1. Differences between returns at the TOM and ROM in columns 3 and 6 in Panel B are significantly positive. The spread of the VW returns between TOM [–1, +3] and ROM [–1, +3] is 0.292%, and the return spread of TOM [–1, +1] and ROM [–1, +1] is 0.468%. The evidence indicates that the TOM effect in the KOSDAQ market is significant.
While existing studies such as Sharma and Narayan (2014) argue that the TOM effect is limited to small firms, McConnell and Xu (2008) document that the effect is not confined to small firms. To examine the size effect on the TOM anomaly, I divide the whole stocks into small, medium, and big stocks based on the market capitalization of stocks, and calculate the VW average daily returns at the TOM and ROM in Table 2. The evidence indicates that there is the TOM effect in all-size tertiles. The average daily TOM [–1, +3] returns for small firms and large firms is 0.319% (t-value is 5.02) and 0.242% (t-value is 3.43). The average daily TOM [–1, +1] returns for small firms and large firms is 0.327% (t-value is 4.39) and 0.468% (t-value is 5.52). The differences in TOM returns between small firms and large firms are not significant, thus high returns at the TOM are not confined to small firms. Rather, TOM [–1, +1] returns for large firms are higher than those for small firms. On the other hand, ROM returns for small firms are significantly higher than those for large firms; consequently, the spreads of TOM and ROM for large firms are higher than those for small firms. That is, the TOM effect is not due to the small size effect. Instead, for the TOM [–1, +1], the anomaly in large firms is stronger than that in small firms.
Turn-of-the-month returns by firm size
| TOM [–1, +3] | ROM [–1, +3] | Diff. | TOM [–1, +1] | ROM [–1, +1] | Diff. | |
|---|---|---|---|---|---|---|
| Small | 0.319*** (5.02) | 0.062* (1.77) | 0.256*** (3.53) | 0.327*** (4.39) | 0.089*** (2.64) | 0.238*** (2.91) |
| Medium | 0.293*** (4.29) | −0.029 (−0.75) | 0.322*** (4.11) | 0.372*** (4.74) | −0.003 (−0.07) | 0.374*** (4.33) |
| Big | 0.242*** (3.43) | −0.060 (−1.63) | 0.302*** (3.80) | 0.468*** (5.52) | −0.051 (−1.40) | 0.519*** (5.62) |
| Small – Big | 0.077 (1.24) | 0.122*** (2.40) | −0.046 (−0.48) | −0.141 (−1.25) | 0.140*** (2.82) | −0.281** (−2.57) |
Note(s): This table reports returns at the turn of the month and the rest days, for firm size tertiles. t-statistics are in parentheses. *, ** and *** indicate significance at the 10%, 5%, and 1% levels, respectively. The sample period is from 2000 to 2020
Other calendar effects may contribute to the TOM effect. The turn-of-the-year (TOY) effect is defined as the tendency of outperformance of stocks around the end and the start of the year (Rozeff and Kinney, 1976; Roll, 1983; Lakonishok and Smidt, 1984; Jones et al., 1987; Haugen and Lakonishok, 1988). The TOY can lead to the TOM anomaly. Similarly, Carhart et al. (2002), Bernhardt and Davies (2005), and Carhart et al. (2002) argue that the turn-of-the-quarter effect that the high return on the last trading day of a quarter may cause the TOM effect. Otherwise, when major indexes like the MSCI's indexes are rebalanced, especially foreign portfolio managers whose funds or ETFs track those indexes may trade more stocks. Their heavy trading volume can lead to the TOM effect.
To examine the influence of several calendar effects on the TOM anomaly, I calculate the VW average of daily returns at the TOM and ROM except for the turn-of-the-year (TOY), turn-of-the-quarter (TOQ) or index rebalancing (IR) TOM and report them in Table 3. I roughly define the TOY period as encompassing the last half of December through the first half of January in the next year, that is, day −10 through day +10 relative to the new year. In the same manner, the TOQ period is defined as encompassing the last half of March, June or September, through the first half of April, July or October. The turn of the fourth quarter (the last half of December and the first half of January) is excluded in the TOQ period in this study because this period is already included in TOY. Lastly, the MSCI indexes are rebalanced on the last business day of February, May, August or November. Therefore, I define the index rebalancing period as encompassing the last half of February, May, August or November, through the first half of March, June, September or December [4].
Turn-of-the-month returns for various calendar categories
| TOM [–1, +3] | ROM [–1, +3] | Diff. | TOM [–1, +1] | ROM [–1, +1] | Diff. |
|---|---|---|---|---|---|
| Panel A.1: returns for the TOY only | |||||
| 0.855*** (3.09) | 0.040 (0.20) | 0.815** (2.34) | 1.275*** (6.24) | 0.067 (0.31) | 1.208*** (3.62) |
| Panel A.2: returns excluding the TOY | |||||
| 0.193*** (2.79) | −0.047 (−1.49) | 0.240*** (3.25) | 0.362*** (4.27) | −0.044 (−1.44) | 0.406*** (4.66) |
| Panel B.1: returns for the TOQ only | |||||
| 0.108 (0.58) | 0.069 (0.95) | 0.039 (0.19) | 0.303 (1.34) | 0.037 (0.56) | 0.266 (1.13) |
| Panel B.2: returns excluding the TOQ | |||||
| 0.279*** (3.82) | −0.076* (−1.78) | 0.355*** (4.20) | 0.469*** (5.45) | −0.058 (−1.38) | 0.527*** (5.50) |
| Panel C.1: returns for the MSCI IR TOM only | |||||
| 0.176 (1.64) | −0.042 (−0.90) | 0.219** (2.16) | 0.345** (2.21) | −0.039 (−0.90) | 0.383*** (3.05) |
| Panel C.2: returns excluding the MSCI IR TOM | |||||
| 0.275*** (3.27) | −0.030 (−0.66) | 0.305*** (3.39) | 0.483*** (5.29) | −0.024 (−0.54) | 0.507*** (5.37) |
Note(s): This table reports daily value-weighted average returns at the turn of the month and rest days for various calendar categories. Panels A.1 gives returns for the turn of the year (TOY) only. Panels A.2 gives returns for all turns-of-the-month excluding the TOY. Panels B.1 gives returns for the turn of the quarter (TOQ) only. Panels B.2 give returns for turns-of-the-month that are not the TOQ. Panels C.1 gives returns for the MSCI index rebalancing (IR) turn of the month only. Panels C.2 gives returns for all turns-of-the-month excluding the IR TOM. t-statistics are in parentheses. *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. The sample period is from 2000 to 2020
Panel A.1 in Table 3 shows the so strong TOY effect in the KOSDAQ market. The VW average daily return for the last day of December and the first day of January is 1.275% and significant. This value is about three times larger than the return for all turns-of-the-month excluding the TOY (0.362%). This TOY effect in the KOSDAQ market is stronger than in the KOSPI market reported by Lee and Kim (2022). They find the TOY VW return of 0.568%. Nevertheless, the TOM returns are still significantly higher than the ROM returns after excluding the TOY period in Panel A.2. This suggests that the TOM anomaly is not completely caused by the high returns at the TOY. Panel B.1 and B.2 show the TOM effect is not due to returns at the TOQ. For the TOQ only, the TOM returns are never significantly higher than the ROM returns. When I investigate returns excluding the TOQ, the VW average returns at the TOM are significantly higher than those at the ROM. I also examine the impact of the index rebalancing on the TOM effect. According to Panels C.1 and C.2, the TOM returns and differences between returns at the TOM and ROM are significantly positive both for the MSCI index rebalancing TOM and in non-index-rebalancing TOM. Further, TOM returns and differences between TOM and ROM returns for non-index-rebalancing months are higher than those for index rebalancing months. Therefore, the TOM effect is not owing to the large trading of portfolio managers whose funds track the MSCI indexes.
4. Trading and net buying volume around the turn of the month
The literature explains that the TOM effect can be caused by the larger trading volume or buying pressure by investors than during the rest days. For example, Ogden (1990) proposes the increased liquidity from retail investors who get usually paid at the month-end as the reason for the TOM anomaly. On the other hand, Barone (1990) argues that institutional investors trade more stocks at the end of the month as window dressing activity. However, some studies such as McConnell and Xu (2008), Maher and Parikh (2013), and Lee and Kim (2022) do not find a higher trading volume at the TOM. To investigate these hypotheses and their counterview for the KOSDAQ market, I calculate trading and net buying volume by investors around the turn of the month.
4.1 Trading volume and turnover by investor types
As most explanations imply the larger trading volume around the TOM, I first calculate the total trading volume and turnover in the market. Figure 2 presents the average total trading volume and turnover by day of the month from day −10 through day +10. The average trading volume in Panel A is calculated as the average of the sum of trading volume on sample stocks in the market on a day, in billions of wons. Share turnover in Panel B is the average of the total shares traded in the market on a day divided by the number of shares outstanding. The evidence indicates that trading volume and turnover around the TOM are not higher, rather the lowest. The average trading volume and turnover on days +1 and −1 are the highest and second-highest, respectively. This suggests that the liquidity in the market is not increased contrary to literature. This result is different from Lee and Kim (2022) who show that trading volume on day −1 in the KOSPI market is higher than the average. I also perform a t-test for the differences in trading volume and turnover between the TOM and ROM, but the results are not reported. According to the results, trading volume and turnover at the TOM are insignificantly lower than those at the ROM. This is following the results in McConnell and Xu (2008) that NYSE trading volume is not higher at the TOM than on other trading days.
To examine which investors' trading activities affect the TOM effect, I compute the average daily trading volume and turnover for individual, institutional and foreign investors, and the results are reported in Figure 3 and Table 4. Trading volume by each investor is the sum of their buying and selling volume. Turnover of each investor is calculated as the sum of their buying and selling shares scaled by total shares. Trading volume is in billions of wons, and turnover is rescaled by multiplying 1,000. Figure 3 presents the average trading volume and turnover of each investor by day of the month from day −10 through day +10. I find trading volumes and turnovers by individual and institutional traders are lower at the TOM than at the ROM. Only foreign volume on day −1 is higher than on other days, however, their volume on the remained days at the TOM is lower and foreign turnover at the TOM is also lower. While Lee and Kim (2022) show that higher trading volume on day −1 in the KOSPI market is due to foreigners' high volume, foreign volume in the KOSDAQ market has little effect on the total volume because foreign volumes in the KOSDAQ market account only for 4.6% of the total trading volume.
Trading volume and turnover at the TOM and ROM by investors
| TOM [–1, +3] | ROM [–1, +3] | Diff. | TOM [–1, +1] | ROM [–1, +1] | Diff. | |
|---|---|---|---|---|---|---|
| Panel A: Trading volume | ||||||
| Individual | 5,020 | 5,049 | −28.503 (−0.10) | 4,895 | 5,052 | −156.397 (−0.55) |
| Institution | 190 | 199 | −9.07 (−0.67) | 188 | 198 | −10.36 (−0.77) |
| Foreigner | 307 | 304 | 3.55 (0.09) | 315 | 302 | 12.996 (0.32) |
| Panel B: Turnover | ||||||
| Individual | 53.946 | 57.081 | −3.135 (−1.57) | 52.408 | 56.905 | −4.497** (−2.26) |
| Institution | 0.699 | 0.727 | −0.028 (−0.84) | 0.683 | 0.726 | −0.042 (−1.30) |
| Foreigner | 1.396 | 1.458 | −0.062 (−0.53) | 1.392 | 1.452 | −0.060 (−0.50) |
Note(s): This table reports the average daily trading volume and turnover at the turn of the month and the rest days, by investors. Trading volume by investors is the sum of their buying and selling volume. Turnover of investors is calculated as the sum of their buying and selling shares scaled by total shares. Trading volume is in billions of wons, and turnover is rescaled by multiplying 1,000. t-statistics are in parentheses. *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. The sample period is from 2000 to 2020
Table 4 shows the trading volumes of all investors at the TOM are not significantly different from those at the ROM. In Panel A, individual and institutional volumes at the TOM are insignificantly lower than those at the ROM, while foreign volumes at the TOM are insignificantly higher than those at the ROM. In Panel B, turnovers of all investors at the TOM are lower than those at the ROM. Especially, individual turnover at the TOM is significantly lower. As a result, the evidence does not support the hypothesis of the increased liquidity of individual investors. This is consistent with the results in Maher and Parikh (2013). They find that there is no evidence of a surge in the retail volumes of trading at month-end in India. Further, institutional investors do not trade more stocks at the TOM, not supporting the explanation of the increased trading of institutions as a result of their window-dressing activity. Foreigners do not also provide a significantly larger volume.
4.2 Net buying volume and net turnover by investor types
To test whether the larger buying pressure by investors around the turn of the month induces the TOM anomaly, I calculate net buying volume and net turnover by the individual, institutional and foreign investors. Net buying volume of the investor is computed as the difference between buying and selling volume by the investor. Net turnover is calculated as the difference between the investor's buying and selling shares scaled by total shares. The results are presented in Table 5, in which net buying volume is in millions of wons, and net turnover is rescaled by multiplying 1,000. Panels A and B show that net buying volumes and net turnovers by individual traders at all TOM periods are significantly lower than at the ROM. That is, individual investors trade and buy stocks less at the TOM than on the other days. Therefore, the hypothesis of the increased demand of individual investors for the anomaly is rejected in the KOSDAQ market.
Net buying volume and Net turnover at the TOM and ROM by investors
| TOM [–1, +3] | ROM [–1, +3] | Diff. | TOM [–1, +1] | ROM [–1, +1] | Diff. | |
|---|---|---|---|---|---|---|
| Panel A: Net buying volume | ||||||
| Individual | 0.554*** (2.74) | 0.832*** (5.23) | −0.278 (−1.08) | 0.188 (0.77) | 0.841*** (5.68) | −0.653** (−2.28) |
| Institution | −0.471*** (−3.45) | −0.450*** (−3.76) | −0.021 (−0.12) | −0.293* (−1.90) | −0.472*** (−4.22) | 0.179 (0.63) |
| Foreigner | 0.382*** (3.22) | 0.146* (1.73) | 0.236 (1.62) | 0.521*** (3.19) | 0.159** (1.99) | 0.362** (1.99) |
| Panel B: Net turnover | ||||||
| Individual | 0.101*** (9.85) | 0.133*** (17.07) | −0.037** (−2.54) | 0.080*** (6.80) | 0.132*** (17.20) | −0.052*** (−3.68) |
| Institution | −0.045*** (−7.94) | −0.050*** (−11.48) | 0.005 (0.69) | −0.034*** (−4.93) | −0.050*** (−11.81) | 0.017** (2.10) |
| Foreigner | −0.014** (−2.07) | −0.027*** (−5.47) | 0.014* (1.65) | −0.007 (−0.87) | −0.026*** (−5.34) | 0.019** (1.96) |
Note(s): This table reports the average daily net buying volume and net turnover at the turn of the month and the rest days, by investors. The net buying volume of the investor is the difference between buying and selling volume of the investor in millions of wons. Net turnover is calculated as the difference of buying and selling shares scaled by total shares and rescaled by multiplying 1,000. t-statistics are in parentheses. *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. The sample period is from 2000 to 2020
Interestingly, the net buying volume and net turnover of foreign traders at the TOM are significantly higher than those at the ROM, although their volume at the TOM is insignificantly different from that at the ROM. Therefore, foreigners rather than individual investors may have a positive impact on the TOM effect. For institutional traders, almost all their net buying volume and net turnover at the TOM are not significantly different from those at the ROM. Only their net turnover at TOM [–1, +1] is significantly higher than that at the ROM [–1, +1]. As they are net sellers on average both at the TOM and ROM, institutional traders may sell fewer stocks at the TOM [–1, +1] in the market.
5. Panel regression analysis
5.1 Variables
I have found that the net buying volumes of foreigners rather than individual or institutional investors are larger at the TOM than at the ROM. Then, does their larger net buying cause the TOM effect in the market? To investigate this question, I run panel regressions with the firm- and day-fixed effect. The dependent variable is the excess return on individual stocks as a proxy of risk premium. Excess returns are stock returns minus the risk-free rate measured by the CD91 rate. To examine whether stock returns are higher during the TOM period, I include the TOM dummy in the regressions as an explanatory variable, which takes one at the TOM (encompassing day −1 through day +3 or encompassing day −1 through day +1) and 0 otherwise. The net buying volume of each investor (NetBuyVj) is also used as the explanatory variable. NetBuyVj is defined as the difference between the natural log of 1 plus buying volume by investor j and the natural log of 1 plus selling volume by investor j. j is indi (individual investor), inst (institutional investor) or forg (foreign investor). I also use the interaction terms between the TOM dummy and the net buying volume of investors, TOM × NetBuyVj, to explore whether their net buying at the TOM has an impact on stock returns. Since there is the turn-of-year effect in the market, I create the TOY dummy to capture the effect, which takes one in the last half of December and the first half of January, and 0 otherwise. Then, the TOY dummy and the interaction term between the TOM and TOY dummies are included in the regressions. As control variables associated with stock returns, LnSize, MB and Turnover are added. LnSize is the log of the market capitalization of the common stock on a given day. MB is the market-to-book ratio, calculated as the market value of the common stock (Size) on a given day, scaled by the book value of common equity. The book value of equity is from the latest available accounting statement. Turnover is shares traded divided by the number of shares outstanding. Finally, the interaction term between the TOM dummy and LnSize is added.
5.2 Result
5.2.1 Full sample
Table 6 shows the results from panel regressions of excess return. In regressions 1 and 2, the TOM dummy takes the value of one at TOM [–1, +3]. In regressions 3 and 4, the TOM dummy takes one at TOM [–1, +1]. In all regressions, the coefficients of the TOM dummy are significantly positive after controlling the turn-of-the-year effect, size effect, net buying volume of each investor and variables related to stock returns. Thus, the TOM effect is likely to be not confined to small or large stocks, and not confined to calendar year-ends. For the investors' net buying impact on the TOM anomaly, the interaction terms between the TOM dummy and net foreign volume have a significantly positive relationship with excess returns, whereas the coefficients of the interaction terms with net individual volume are negative. This indicates that stocks largely bought by foreigners, not individual investors, have higher returns at the TOM in the KOSDAQ market, which is consistent with Lee and Kim (2022). Note that foreigners buy stocks more at the TOM. The results in this paper imply that foreign trading during the TOM period may partially cause the TOM effect. Furthermore, the explanation that increased demand of individual investors induces the effect is rejected in the KOSDAQ market. Net institutional volume at the TOM has a significantly positive relation with excess returns only at the TOM [–1, +1]. For the TOM [–1, +3], the coefficient of TOM × NetBuyVinst is not significant.
Panel regression analysis: full sample
| TOM [–1, +3] | TOM [–1, +1] | |||
|---|---|---|---|---|
| (1) | (2) | (3) | (4) | |
| TOM | 0.216*** (43.62) | 1.363*** (12.31) | 0.291*** (44.32) | 1.519*** (10.63) |
| TOM × NetBuyVindi | −0.039*** (−3.00) | −0.079*** (−4.25) | ||
| TOM × NetBuyVinst | 0.001 (0.35) | 0.004* (1.95) | ||
| TOM × NetBuyVforg | 0.006*** (3.30) | 0.006*** (2.61) | ||
| TOM × TOY | 0.763*** (34.30) | 0.760*** (34.19) | 0.887*** (34.00) | 0.882*** (33.79) |
| TOM × LnSize | −0.046*** (−10.47) | −0.050*** (−8.67) | ||
| NetBuyVindi | −0.662*** (−17.36) | −0.654*** (−17.32) | −0.661*** (−17.36) | −0.654*** (−17.27) |
| NetBuyVinst | 0.032*** (21.73) | 0.032*** (21.20) | 0.032*** (21.69) | 0.032*** (21.28) |
| NetBuyVforg | 0.102*** (56.71) | 0.101*** (54.50) | 0.102*** (56.62) | 0.101*** (55.27) |
| LnSize | 0.079*** (19.36) | 0.089*** (20.67) | 0.079*** (19.31) | 0.084*** (20.04) |
| MB | 0.002*** (3.64) | 0.002*** (3.64) | 0.002*** (3.64) | 0.002*** (3.64) |
| Turnover | 9.001*** (46.87) | 9.001*** (46.87) | 9.003*** (46.88) | 9.003*** (46.88) |
| TOY | −0.042*** (−5.23) | −0.042*** (−5.20) | 0.016** (2.20) | 0.016** (2.22) |
| Constant | −2.198*** (−21.87) | −2.422*** (−23.07) | −2.180*** (−21.69) | −2.299*** (−22.37) |
| Observations | 4,875,325 | 4,875,325 | 4,875,325 | 4,875,325 |
| Clusters (firms) | 1,986 | 1,986 | 1,986 | 1,986 |
| R2 | 0.034 | 0.034 | 0.034 | 0.034 |
Note(s): This table reports the results of panel regressions of excess stock returns. In regressions (1) and (2), TOM takes one at the TOM that encompasses day −1 through day +3, and 0 otherwise. In regressions (3) and (4), TOM takes one at the TOM that encompasses day −1 through day +1, and 0 otherwise. Standard errors are clustered by firm. t-statistics are presented in parentheses. *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. The sample period is from 2000 to 2020
5.2.2 Sample at the TOM
To reveal which stocks have higher returns during the TOM, I perform equivalent panel regressions with the firm- and day-fixed effects using only a sample at the TOM, excluding data for the ROM. Thus, the TOY dummy and the interaction terms with the TOY dummy are excluded in the regressions. Since large firms have a significantly higher spread between TOM [–1, +1] and ROM [–1, +1] than small firms in Table 2 and foreigners partially affect the TOM effect in the KOSDAQ market, I add the interaction term between the net buying volume of foreigners and firm size, each investor (NetBuyVforg × LnSize). Results are reported in Table 7. In regressions 1 and 2, the TOM period is defined as the interval [–1, +3]. In regressions 3 and 4, the TOM period is the interval [–1, +1].
Panel regression analysis: sample at the TOM
| TOM [–1, +3] | TOM [–1, +1] | |||
|---|---|---|---|---|
| (1) | (2) | (3) | (4) | |
| NetBuyVindi | −0.703*** (−16.93) | −0.686*** (−16.85) | −0.750*** (−17.04) | −0.733*** (−16.95) |
| NetBuyVinst | 0.033*** (16.70) | 0.034*** (17.17) | 0.036*** (14.50) | 0.036*** (14.89) |
| NetBuyVforg | 0.107*** (46.08) | −0.939*** (−12.36) | 0.107*** (38.49) | −0.898*** (−10.41) |
| NetBuyVforg × LnSize | 0.042*** (13.67) | 0.040*** (11.61) | ||
| LnSize | 0.042*** (5.84) | 0.042*** (5.87) | 0.030*** (3.16) | 0.030*** (3.14) |
| MB | 0.002** (2.37) | 0.002** (2.33) | 0.003** (2.04) | 0.003** (1.98) |
| Turnover | 9.517*** (36.21) | 9.511*** (36.22) | 9.176*** (27.94) | 9.171*** (27.94) |
| TOY | 0.719*** (36.62) | 0.721*** (36.72) | −0.035 (−1.61) | 0.901*** (35.40) |
| Constant | −1.055*** (−6.03) | −1.058*** (−6.06) | −0.663*** (−2.88) | −0.658*** (−2.86) |
| Observations | 946,601 | 946,601 | 473,884 | 473,884 |
| Clusters (firms) | 1,986 | 1,986 | 1,986 | 1,986 |
| R2 | 0.039 | 0.040 | 0.039 | 0.039 |
Note(s): This table reports the results of panel regressions of excess stock returns using only a sample at the TOM, excluding data for the ROM. In regressions 1 and 2, the TOM period is defined as the interval [–1, +3]. In regressions 3 and 4, the TOM period is the interval [–1, +1]. Standard errors are clustered by firm. t-statistics are presented in parentheses. *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. The sample period is from 2000 to 2020
Results for the net buying volume are consistent with Table 6 using the whole sample. In regressions 1 and 3, stocks with lower individual net buying volumes and higher institutional and foreign net buying volumes have higher TOM returns. However, in regressions 2 and 4, the coefficients of NetBuyVforg × LnSize are significantly positive, but the coefficients of NetBuyVforg are significantly negative. This result shows large firms with higher foreign net buying volume have higher TOM returns. The result of firm size (LnSize) confirms that large firms have a stronger TOM effect than small firms. Additionally, TOM returns during TOY are significantly higher after controlling investors' net buying volume and other stock characteristics.
5.2.3 Stocks not traded by foreign investors
If the high net buying volume of foreign investors affects the TOM anomaly in the KOSDAQ market, is not there the TOM effect for stocks that foreign investors do not trade? I replicate the previous panel regression analyses using stock-day observations with zero foreign trading volume. Since the stocks with zero foreign trading volume have zero foreign net buying volume, NetBuyVforg is excluded in these regressions. Results are reported in Table 8. In regressions 1 and 2, the TOM dummy takes the value of one at TOM [–1, +3]. In regressions 3 and 4, the TOM dummy takes the value of one at TOM [–1, +1].
Panel regression analysis: stocks that foreign traders do not trade
| TOM [–1, +3] | TOM [–1, +1] | |||
|---|---|---|---|---|
| (1) | (2) | (3) | (4) | |
| TOM | 0.283*** (29.74) | −0.101 (−0.35) | 0.416*** (32.40) | −0.528 (−1.50) |
| TOM × NetBuyVindi | 0.000 (0.01) | 0.011 (0.57) | ||
| TOM × NetBuyVinst | 0.010** (2.41) | 0.022*** (3.99) | ||
| TOM × TOY | 1.098*** (26.77) | 1.100*** (26.73) | 1.026*** (22.02) | 1.027*** (21.99) |
| TOM × LnSize | 0.016 (1.33) | 0.040*** (2.68) | ||
| NetBuyVindi | −0.111*** (−8.78) | −0.111*** (−9.07) | −0.111*** (−8.77) | −0.112*** (−9.09) |
| NetBuyVinst | 0.007*** (2.69) | 0.005* (1.74) | 0.006*** (2.58) | 0.004* (1.67) |
| LnSize | 0.123*** (14.20) | 0.120*** (12.81) | 0.123*** (14.21) | 0.120*** (13.32) |
| MB | 0.003** (2.55) | 0.003** (2.55) | 0.003** (2.55) | 0.003** (2.55) |
| Turnover | 6.989*** (21.87) | 6.989*** (21.87) | 6.985*** (21.82) | 6.984*** (21.82) |
| TOY | −0.277*** (−18.90) | −0.277*** (−18.90) | −0.165*** (−12.21) | −0.165*** (−12.22) |
| Constant | −3.145*** (−15.38) | −3.070*** (−13.89) | −3.131*** (−15.32) | −3.039*** (−14.38) |
| Observations | 1,422,216 | 1,422,216 | 1,422,216 | 1,422,216 |
| Clusters (firms) | 1,684 | 1,684 | 1,684 | 1,684 |
| R2 | 0.008 | 0.008 | 0.007 | 0.007 |
Note(s): This table reports the results of panel regressions of excess stock returns using stocks with zero foreign trading volume. In regressions (1) and (2), TOM takes one at the TOM that encompasses day −1 through day +3, and 0 otherwise. In regressions (3) and (4), TOM takes one at the TOM that encompasses day −1 through day +1, and 0 otherwise. Standard errors are clustered by firm. t-statistics are presented in parentheses. *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. The sample period is from 2000 to 2020
In regressions 1 and 3, the coefficients of TOM dummies are significantly positive, which demonstrates the significant TOM effect on stocks that foreign investors. However, when I add the interaction terms between the net buying volume of investors or size and the TOM dummy, the coefficients of TOM dummies become insignificant. Large stocks and stocks with higher institutional net buying volume have significantly higher returns at the TOM. TOM returns at the TOY are also significantly higher after excluding stocks that foreign investors' trade and controlling investors' net buying volume and other stock characteristics. These results confirm that foreign trading partially causes the TOM anomaly. After eliminating the foreign trading effect, the TOM anomaly is associated with institutional trading, size and the TOY effect.
Overall, the evidence indicates that the hypothesis of the increased demand of individual investors is rejected in the KOSDAQ market and foreign trading is likely to influence the TOM effect. Maher and Parikh (2013) suggest that the window dressing activity by institutional traders causes the TOM anomaly, finding an increase in the foreign and domestic institutional volumes at month-end in the Indian market. For the Korean market, Lee et al. (2014) show fund managers (especially foreign managers) are tempted to distort fund performance through portfolio pumping at year-end, which supports the window dressing explanation for the TOM effect.
6. Conclusion
I examine the TOM effect using KOSDAQ stocks from 2000 through 2020 and find a significantly positive TOM effect in the market. The VW and EW average daily returns on the first and last trading days of the month are the highest and significantly positive. When I test whether the TOM effect is due to the size effect or other calendar effects such as the TOY, the TOQ or the IR effect. When I divide the whole stocks into small, medium and big stocks based on the firm size, the TOM anomaly is not confined to small firms. Further, the anomaly is significant after excluding the turn-of-the-year, the turn-of-the-quarter or the index rebalancing TOM.
I investigate the trading volume and turnover of individual, institutional and foreign investors because the explanations such as the increased liquidity of individual investors and the institutional window-dressing activity highlight the increasing individual or institutional trading volume at the TOM. The results show that trading volumes and turnovers of all investors are not significantly higher at the TOM than at the ROM. Moreover, individual turnover at the TOM [–1, 1] is significantly lower than those at the ROM. When I test net buying volume and net turnover by each investor, the evidence indicates that the net buying volume and net turnover of individual traders at the TOM are significantly lower than those at the ROM. The net buying volume of institutional traders at the TOM is not significantly different from those at ROM. This implies that the hypothesis of the increased demand of individual investors is rejected in the KOSDAQ. The TOM effect may not be due to the increased net buying volume of institutions. Interestingly, only foreign net buying volume and net turnover at the TOM are significantly higher.
Finally, using panel regressions, I find that excess returns on stocks are higher during the TOM period after controlling variables related to stock returns. Moreover, stocks with a higher net buying volume of foreigners for the TOM period tend to have higher returns, while stocks with a higher net buying volume of individual traders for the TOM period have lower returns. The results confirm that the TOM effect is not due to the increased demand of individual investors. Instead, higher net buying volume by foreigners may partially cause the TOM effect.
This study does not reveal why foreign investors have a higher net buying volume at the TOM. Maher and Parikh (2013) suggest that the window dressing activity by institutional traders causes the TOM anomaly, finding an increase in the foreign and domestic institutional volumes at month-end in the Indian market. For the Korean market, Lee et al. (2014) show fund managers (especially foreign managers) are tempted to distort fund performance through portfolio pumping at year-end, which supports the window dressing explanation for the TOM effect. Thus, foreign investors may have a higher net buying volume at the TOM for window dressing, which further research is needed to study.
Notes
See Ariel (1987), Lakonishok and Smidt (1988), Jacobs and Levy (1988), Pettengill and Jordan (1988), Odgen (1987, 1990), Cadsby and Ratner (1992), Cinar and Vu (1991), Ziemba (1991), Cadsby and Ratner (1992), Agarwal and Tandon (1994), Hensel and Ziemba, 1996, Martikainen et al. (1995), Booth et al. (2001), Compton (2002), Kunkel et al. (2003), Nikkinen et al. (2007) and McConnell and Xu (2008).
McConnell and Xu (2008) show that New York Stock Exchange (NYSE) trading volume is not higher at the TOM. Maher and Parikh (2013) do not find an increase in the retail volumes at month-end in the Indian market.
During the same period, institutional and foreign trading volumes in the KOSDAQ market account for 3.8 and 4.6% of the total trading volume, respectively.
I also consider the rebalancing effect of the KOSPI's indices such as KOSPI 200 because almost all domestic funds or ETFs track these indices. However, since rebalances in the KOSPI indices usually occur on the second Friday of June and December, not at the TOM, I do not include their rebalancing effect in this analysis.
