«Equity Style Investing RONG, WU How to cite: Equity Style Investing, Durham theses, Durham University. RONG, WU (2013) Available at Durham E-Theses ...»
This table reports the corresponding compound returns of the winner and loser styles during 1980-2004. The 9 style portfolios are small-cap growth (SG), small-cap blend (SB), small-cap value (SV), mid-cap growth (MG), mid-cap blend (MB), mid-cap value (MV), large-cap growth (LG), large-cap blend (LB), large-cap value (LV).
To better understand the return patters of the P1-P9 style portfolios, Table 4-5 computes the fraction that a particular style is in long or short side for all monthly strategies that rank styles according to the prior 3- and 12-month returns. Separate statistics are reported for hedge portfolios that contain two or four extreme winner and loser styles. The distribution of winner and loser styles suggests that overall value styles dominate the winners and growth styles dominate the losers. Specifically, it is shown that in most cases investors tend to favour SV styles and dislike LG styles for PC- and BM-based categorisation. For DY-based style classification, it is found that LV is the in-favour investment style, and again LG is the out-of-favour style.
In summary, the empirical findings in this session would suggest that, consistent with the literature, overall value styles tend to be winner styles and growth styles tend to be loser styles. But once interacted with the size dimension, the winners and losers may change alone the size axis, suggesting that style momentum portfolios need active rebalancing. To illustrate this, Figure 4-4 depicts the stock migration rate (%) in winner and loser styles based on 12-month ranking period and the use of 2 extreme styles in hedge portfolios. The negative sign represents the short side (loser style). The migration rate represents the percentage of stocks that will be moved in and out the winner or loser styles based on new ranking. The number will be 100 in general should the winner and loser be changed completely, and it would be between 0-100 once the previous winner or loser continue to be the winner and loser but with some new stocks moved in or out. To complement Figure 4-4, Table 4-6 reports the average migration rate (%) of stocks between the same styles, i.e. the average percentage of stocks that are likely to be moved in or out for styles that are continue to be the winner or loser in the next period.
Figure 4-4 reveals that, even for 12-month ranking period, the winner and loser changes quite frequently at both long and short side, while for styles that are continue to be the winner or loser in the next period, on average there are about 5% of the stocks that will be reclassified and move in or out from where they used to be. It is suggested that such rebalance of style momentum portfolios would introduce nontrivial transaction costs. Arguably, rebalance is needed when (1) the winner and loser styles changed; (2) a stock moves in and out of the winner or loser styles and; (3) a stock demonstrates exceptional high cross-sectional volatility and thus style portfolio needs rebalancing.
Obviously, the shorter the rank period is, the more rebalance may be needed, and therefore the more transaction cost occurred. Hence from a practical investment perspective, financial practitioners should assess whether style momentum is able to generate economically positive profit once transaction costs are considered. Chen and De Bondt (2004) propose that such strategy is most useful for asset allocation experts who direct fund flows or used to enhance passive investing such as indexation strategy.
Table 4-5 The composition of style momentum portfolios Every month between January 1982 and October 2003, based on firm characteristic variable PC, BM and DY respectively, 9 style portfolios (P1-P9) are ranked by their returns for the prior 3 or 12 months. The style momentum hedge portfolios are formed to buy winners (one or two style portfolios with the best past performance) and to sell losers (one or two style portfolios with the worst past performance). This table reports the percent of portfolio replications that either on long or short side.
Figure 4-4 Average stock migration rate % for winner and loser style The figure below illustrates the percentage of stocks that will be moved in or out of winner and loser styles in next time period based on the current identification of winner and loser according to 12-month ranking period and the use of 2 extreme styles in hedge portfolios. The number will be 100 in general if the winner or loser is changed completely, and it would be between 0-100 once the previous winner or loser continues to be the winner and loser but with new stocks included or excluded.
Stock Migration Rate % in Winner-loser Styles (J = 12, PC-based styles)
Table 4-6 Average migration rate (%) for stocks in consecutive extreme styles This table reports the average migration rate (%) of stocks between the same winner or loser styles, i.e. the percentage of stocks that are likely to be moved in or out for styles that are continue to be the winner or loser in the next period.
4.6 Style, price and industry momentum While section 5 has found the profitability of style momentum strategy in the U.K. stock market, one may well argue that such profits are simply the miracle of the price momentum of Jegadeesh and Titman (1993) or the industry momentum of Moskowitz and Grinblatt (1999) documented in the literature. This is because those stocks in current winner (loser) styles may also be categorised into the winner (loser) portfolios based on past individual stock returns, or winner (loser) industry portfolios. Therefore the style continuations may be simply due to a concentration of winner (loser) stocks within winner (loser) styles whose returns persist in the test periods. Grundy and Martin (2001) argue that price momentum strategy loads investors up on factors that perform well recently. Thus the return of price momentum captures the investors’ sentiment about the firm’s future perspective.
Similarly, as Chen (2003) argues, industry momentum contains the changes of business sentiment about the industry’s perspective.
Hence it is important to disentangle style, price and industry effects.
Following Chen and De Bondt (2004), three methods are applied. First, the style momentum returns are calculated after adjusting price and industry momentum at the firm level. Next, a two-way independent sorting is implemented to investigate whether style momentum is independent from price and industry momentum. Finally, monthly cross-sectional regressions are tested by regressing expected returns for individual stocks on style momentum (SM), price momentum (PM) and industry momentum (IM) indicators to examine the explanatory ability of the three underlying momentum effects.
Every month, for each characteristic variable PC, BM and DY respectively, 9 style portfolios (P1-P9) are ranked according to their past 3- and 12-month returns starting from 1982 to 2003. Meanwhile, all stocks in P1-P9 styles are ranked into 9 quintiles according to (1) the past 3- or 12-month of the style portfolio returns to which they belong; (2) their own past 3- or 12-month total returns; and (3) the past 3- or 12-month industry portfolio returns to which they belong.23 Under this procedure, each stock will be properly plotted in a 3-D space with the information of style, price and industry momentum rankings. A pair of two ranking information will be examined and style momentum and price momentum are to buy the best quintile stocks and to sell the worst quintile stocks, while the industry momentum only buys and sells P1-P9 stocks that belong to the top and bottom of two industry portfolios whose ranking is based on all P1-P10 stocks in the universe.
Table 4-7 reports the value weighted average raw returns in the test periods up to 36 months as well as style, price or industry adjusted returns. The raw returns are adjusted on the individual stock level by deducting the contemporaneous value weighted returns of control portfolios. The control portfolios are either the industry momentum portfolios based on all stocks (P1-P10), or price momentum portfolios and style portfolios of based on stocks in P1-P9. Note that the style momentum returns reported in Table 4-7 are different from those in Table 4-3 because of the different weighting schemes used. The returns in Table 4-7 are based on value weighted scheme, and hence are smaller than those presented in Table 4-3 where equally-weighted scheme is used.
Table 4-7 suggests that it is difficult to disregard style momentum.
Especially for stocks sorted on PC and DY and based on 12-month ranking and with holding period 6 and 9 months, the raw payoffs of style momentum have similar magnitude to PM- and IM-adjusted returns. It is also shown that SM, PM and IM are interacted. For example, once adjusting for PM effect, SM payoffs tend to decline.
Similarly, PM effect tends to decrease when adjusting for IM or SM.
The industry classification follows the Datastream variable INDC3. There are 14
industries identified altogether, i.e. BASIC,CYCGD, CYSER, GENIN, ITECH, NCYCG, NCYSR, OTHEQ, RESOR, SUSEQ, TOTLF, UNCLS, UQEQS and UTILS.
While IM also declines when adjusted for PM, it tends to increase the performance once SM is adjusted (except for IM-DY based on 3-month ranking). This would suggest a strong interaction between PM and IM effects, which is consistent with the literature. For example, prior studies such as Moskomitz and Grinblatt (1999) find that after controlling for industry effects price momentum disappears. Lee and Swaminathan (2001) show that adjusting for industries effects weakens the individual price momentum return from 12.5% to 10.1% per annum, and Grundy and Martin (2001) argue that industry momentums captures half of the size of price momentum effect. More recently, Lewellen (2002) and Chordia and Shivakumar (2002) also find individual momentum effect is still present after controlling for industry momentum.
Table 4-7 reveals some interesting findings. First, value weighted SMBM returns are less persistent based on current sample data, and PMDY demonstrates short term reversals (although not significant). Next, significance alone, the ranking of SM, PM and IM returns based on PC, BM and DY varies, suggesting that these firm attributes may capture different information affecting SM, PM and IM effects. For a (12, 12) strategy, it shows that IM-PC tends to have highest returns followed by PM-PC and SM-PC. In addition, IM-BM tends to have higher returns than SM-BM, and IM-DY tends to have highest returns followed by SM-DY, while PM-DY has the lowest performance. Overall, it should be safe to conclude that style momentum is a different phenomenon as compared to price and industry momentum.
It is necessary to further examine the interaction of style, price and industry momentum using an independent two-way sorting. Such two-way independent sorting avoids the problems criticised by Berk (2000) when distinguishing the explanatory power for future returns from two variables that are perceived to be correlated. Following Chen and De Bondt (2004), every month, for each variable BM, DY and PC, 9 style portfolios are first ranked either by 3- or 12-month past period returns. Style portfolios in the top three are labelled #1 (winners) while portfolios in the bottom three are labelled #3 (losers). Style portfolios in the middle range are labelled #2. Next, all stocks in P1-P9 styles are sorted into 3 quintiles according to their prior 3- or 12-month performance. The winner quintile is labelled #1 and the loser quintile is labelled #3 while the middle is labelled #2. Finally, 14 industry portfolios defined by the Datastream variable INDC3 are ranked by prior 3- or 12-month returns. Industry portfolios in the top 4 are labelled #1 (winners) while industries in the bottom 4 are labelled #3 (losers). Industry portfolios in between are labelled #2. Following These procedures, every stock in P1-P9 will be assigned to a 3 dimensional space containing the information of style, price and industry ranking.
Table 4-8 reports the equally weighted average monthly raw returns for the long, short and hedge momentum portfolio returns. Panel A is based on 3-month ranking and Panel B are the results for 12-month ranking period. Regardless which characteristics are used to define styles, it is demonstrated that once capitalising on the interaction with style effect, the price and industry momentum are significantly enhanced and the durations of return continuation are extended up to 2 years and possibly longer. It is evident that stocks in winner styles continue to outperform stocks in loser styles regardless whether they have been classified as price winner or losers, or whether they are in winner industries or loser industries. Moreover, the magnitude of return spreads for stocks in extreme styles but in the same times also classified into different price or industry performance categories are quantitatively similar. This indicates that style momentum plays more important role in affecting the structure of equity returns dynamics than price momentum or industry momentum does.
To disentangle style, price and industry momentum effects, Table 4-9 applies Fama-MacBeth (1973) cross-sectional multivariate regressions.