1
Master Thesis
Finance
The Post-Cost Profitability of Momentum Trading
Strategies in the Dutch Stock Market
By Gijs Kouwenberg
Abstract
In this paper I investigate the momentum strategies in the Dutch stock market over a time span of 1996 to 2014. I present evidence that momentum returns are observable in the Dutch stock market and I analyse the stock concentration, the turnover ratio and the trading costs of those strategies. After subtracting the trading costs, no strategy yields a significant positive return as a result of high trading costs. An alternative strategy using only large-cap stocks reduces costs dramatically but also lowers the returns. The trade-off however is positive to exclude small-cap stocks because the gains of lower trading costs are higher than the loss in returns. I also show that the majority of the returns of the momentum strategy are generated by the winner portfolio, this is contrary to other studies done in the field of momentum trading strategies.
JEL classification: G10, G11, G14,
Keywords: Momentum, turnover ratios, Bid-ask spreads, trading costs, market efficiency Author: Gijs Sebastiaan Pieter Kouwenberg
Mail: Gijskouwenberg@gmail.com Phone: +31642258421
Student number: S1811681
Place and date: Groningen, 27 May 2014 Supervisor: Dr A. Plantinga
2
1. Introduction
Momentum: “ every object in motion will stay in motion until acted upon by an external force” Isaac Newton 1687
This paper examines and investigates the post-cost profitability of momentum trading strategies in the Dutch stock market over a time span of 1996 to 2014. Investors and portfolio managers are always looking for a strategy, which yields them significant returns on a steady basis. If only such a strategy would exist. Henry Mintzberg1 from the McGill University formulates strategy as “a pattern in a stream of decisions”. Max Mckeown argues that strategy is all about shaping the future and is the human attempt to achieve desirable results with available means. Dr Vlamdimir Kvint argues that strategy is a system of finding, formulating and developing a doctrine that will ensure long-term success if followed strictly2. Most of these explanations overlap and come to the same conclusion: strategy is a high level plan to achieve one or more goals under conditions that are uncertain. Strategy is important because resources that are available for reaching the goals are usually limited and therefore require a strategy to achieve these goals.
De Bondt and Thaler (1985) show in their paper that following a contrarian strategy, which implies selling “winning” stocks and buying “losing” stocks, yields abnormal return. Jegadeesh and Titman (1993) present evidence of the anomaly called momentum. This strategy implies the buying of “winners” and the selling of “losers”. Contrarian and momentum differ in the fact that contrarian strategies are based on the reversal of stock returns and momentum strategies are based on the continuation of stock returns. The empirical literature suggests that a successful contrarian strategy has a long time horizon; momentum strategies involve a short or medium term time horizon (Jegadeesh and Titman, 1993; De Bondt and Thaler, 1985). In the last three decades stock market anomalies, such as momentum and contrarian investment strategies, have attracted much attention from both investors and researchers in the finance literature. Because rational risks models, adopted in finance, do not fully explain the anomalies like momentum and contrarian strategies, behavioural models are taken more into consideration. Barberis, Schleifer and Vishny (1998), Daniel, Hirshleifer and Subrahmanyam (1998) and Hong and Stein (1999) all present
1Mintzberg, Henry and, Quinn, James Brian (1996). The Strategy Process: Concepts, Contexts, Cases. Prentice Hall. ISBN 978-0-132-340304
.
3
behavioural models that are based on the idea that wrong or delayed information interpretation of investors lies at the heart of these anomalies. These models imply that the abnormal returns, achieved using momentum and contrarian strategies, are due to the fact of a delayed overreaction of information that leads to stock prices which are higher or lower than their long-term values.
Anomalies violate the view of efficient markets as proposed by Fama (1965). The concept of efficient markets, which is the backbone of many financial theories, states that markets reflect all information available directly in the stock prices. Following this theory it cannot be that historical return paths of stocks have a predicting element for future returns. Different studies already prove the existence of momentum strategies, which yields abnormal returns for different markets around the world. These various studies are addressed in greater detail in the literature section.
4
This paper examines momentum anomalies, the transaction costs and the profitability of momentum strategies for the Dutch stock market over a period from 1996 to 2014. Although much research is done in the field of momentum strategies, little research is done in other than the US and UK markets. As stated before, transaction costs have the capability to completely destroy all momentum profit. Transaction costs are sensitive to different investment horizons and styles, but are also sensitive to the exchange where is traded (Keim and Madhaven, 1997; Barber and Odean, 2000). On this basis, it is possible that studies, conducted on other than the Dutch stock market, may not be accurate and conclusive for the momentum profits generated in the Dutch stock market. Available literature states that there is an inverse relationship between the size and trading costs of stocks. Size is measured by the market capitalization of a company. Next to these findings it is also argued that stocks, that are more illiquid, are subject to higher trading costs. It is clear that the Dutch stock market is less intensive in terms of trading than the US stock market.
Research Question
Based on the observations stated above, the main research question of this paper is determined as follows:
What is the post-cost profitability of various momentum investment strategies in the Dutch stock market?
In order to answer this question I will investigate if the Dutch market shows significant momentum returns. Also I will examine the turnover ratios and the intensity of holding momentum investment strategies. To answer the main research question, I will further examine the trading costs associated with implementing and maintaining these momentum investment strategies.
5
The results show that the momentum anomaly is observable in the Dutch stock market, and that momentum investment strategies yield significant positive returns. This paper shows that the majority of the returns observed are mostly generated by the winner portfolio, which is in contrast with studies done on foreign stock markets. When, however the momentum profits are adjusted for trading costs, the momentum returns are destroyed. This is due to high portfolio turnover. A sub-sample is checked in order to reduce the trading costs; the trading costs are significantly reduced but the exclusion of stocks also leads to less momentum return. The sub-sample shows no significant positive return when corrected for trading costs.
The remainder of this paper is structured as follows way: the second section is attributed to the literature review which provides an overview of the relevant studies done on the subject of momentum investing and market related anomalies. Section three describes the methodology and the data used in order to be able to carry out the research. Section four provides an overview of the results and critically evaluates those empirical results. The fifth and final section contains the conclusion, which answers the main question. This section also provides a discussion on the restrictions found and proposals for a follow-up research/study.
2. Literature review
This part explains what momentum is and how it contradicts the popular view of efficient markets. Different studies regarding momentum investing are discussed; also the different specific characteristics of stocks and how these characteristics influence the momentum returns are described.
2.1 Momentum
6
The main findings of Jegadeesh and Titman (1993) are that significant momentum returns are observed in the period of 1965 to 1989. Jegadeesh and Titman (2001) confirm their prior research.
2.2 Efficient markets
Fama (1965) suggests that markets should behave in an efficient; way meaning that stock prices at all times should fully reflect all available information. This theory lies at the heart of many financial theories and models. Fama (1965) considers three relevant information subsets of market efficiency, namely the weak form, the semi-strong form and the strong form. Prices of stocks should already incorporate all information available under the assumption of market efficiency. It follows that future returns cannot be predicted by historical returns. Momentum strategies imply exactly the opposite and therefore receive much attention in literature. Momentum cannot exist even in the weakest form of market efficiency. Although a lot of studies have pursued the goal of explaining momentum profits, little conclusive answers have been given. The literature in which momentum received attention will be further elaborated in the next section.
Although much research is done, the excess returns earned by momentum strategies are still mostly unexplained and therefore interesting to investigate. This view is supported by Chan et al. (1996) when he writes “In the absence of an explanation, the evidence on momentum stands out as a major unresolved puzzle.” Or by Fama (1998) who identifies the momentum phenomenon as “an open puzzle”.
2.3 Literature
Momentum strategies consist of buying recent winner- and selling recent loser-stocks (Jegadeesh and Titman, 1993). Many empirical studies show that following a momentum strategy yields significant higher returns. Table 1 presents an overview of the different studies of momentum strategies.
7
Table 1
This table presents literature done in the field of momentum investment strategies; it shows the authors, year of publishing, the name of the article and the markets researched by the paper.
Author (year published) Name article Market(s)
Jegadeesh and Titman(1993, 2001)
Returns to buying winners and selling losers: implications for stock market efficiency.
Profitability of momentum strategies: an evaluation of alternative explanations.
US Stock market
Chan, Jegadeesh and Lakonishok (1996,1999)
Momentum strategies.
The profitability of momentum strategies
US Stock market
Doukas and McKnight(2005)
European momentum strategies, information diffusion, and investor
Conservatism
8 different European markets
Froner and Marhuenda(2003)
Contrarian and momentum strategies in the Spanish stock market
Spanish Stock market
Lui, Strong and Xu(1999) The profitability of momentum investing Uk Stock market
Hon and Tonks(2003) Momentum in the UK stock market Uk Stock market Dische (2002) Dispersion in analyst forecasts and the profitability of earnings
momentum strategies
European stock markets
Antoniou (2005) Contrarian profits and the overreaction hypothesis: The case of the Athens Stock Exchange
Rouwenhorst (1998, 1999)
International momentum strategies.
Local return factors and turnover in emerging markets
International stocks
Griffin, Ji and Spencer (2003)
Momentum investing and business cycle risk: evidence from pole to pole
International stocks
Where significant returns of contrarian strategies suggest overreaction to information, momentum suggests underreaction towards information that is publically available. Jegadeesh and Titman (1993) find evidence in favour of return persistence in the US stock market using data from 1965 to 1989. They find that momentum strategies yield abnormal returns in the 3-12 month time period following the construction of a momentum portfolio. Doukas and McKnight (2005) also find significant momentum returns in more than half of the European markets. Other studies that yield similar results are those of Hon and Tonks (2003) Forner and Marhuenda (1999).
8
been made to understand this phenomenon. Conrand and Kaul (1998) describe the “momentum premium” as a premium that is received because of the higher risk momentum strategies bear. Black (2003) and MacKinlay (1995) state that the momentum premium is a result of data mining. Other authors argue that the momentum premium is illusory and economically insignificant (Lesmond et al, 2004; Hanna and Ready, 2005).
The arguments cited by Black (2003), MacKinlay (1995) and Conrad and Paul (1998) however do not fully explain the existence of the “momentum premium” (Jegadeesh and Titman, 2002; Grundy and Martin, 2001; Lui et al, 1999). There are also several behavioural explanations for the momentum premium in the literature. Daniel, Hirshleifer and Subrahmanyam (1998) provide evidence for investor overconfidence, which can lead to misinterpretation of information. Barberis Shleifer and Vishny (1998) present evidence for investors’ underreaction towards information on prices of securities over time horizons of one to twelve months. Overconfidence, misinterpretation of information, and underreaction of information lead to anomalies like the “momentum premium”. In the efficient market theory such an anomaly should be arbitraged away. This implies that when post-cost momentum premiums are found, the anomaly is not fully exploited and therefore the market is not as efficient as stated by Fama (1965).
Momentum strategies are subjected to high turnover ratios and therefore high trading costs. Several studies show that, although trading costs for momentum profits are high, strategies that involve longer holding periods still yield significant returns (Agyei-Ampomah, 2007; Lesmond et al. 2004). For these studies the abnormal returns of the anomalies are not explained by the trading costs
2.4 Stock specific elements.
9
stocks, that significant returns are made after subtracting the costs (Aygei-Ampomah, 2007). These findings are enforced by the study of Rouwenhorst (1998). He finds that, in all size classes, recent outperformers keep outperforming and recent “losers” keep losing. Findings done by Li (2009) impose that even if the stocks, which are illiquid and have high returns, cannot be selected for momentum portfolios; momentum strategies still yield high returns. They prove this for the UK stock market. The Swiss stock market is also investigated regarding the profitability of momentum strategies; Rey and Schmid (2007) only use a sample of the largest stocks traded in Switzerland. They achieve large profits -up to 44%-, by investing, using momentum strategy for single stock investments.
Following from the literature review, the next hypothesis is formulated:
Significant positive momentum returns are observable in the Dutch stock market, but when those returns are corrected for trading costs, the momentum returns are destroyed.
3. Data and methodology
This section will elaborate on the data and methods used to examine momentum returns in the Dutch stock market. Also the methods and equations used to determine the turnover ratios, trading costs and stock concentration in the portfolios are shown and explained.
3.1 Data
The data analysis begins with extracting information from all stocks traded on Euronext Amsterdam from Datastream, including the time period from May 1996 to May 2014. I exclude investment trusts and warrants. Trading costs are calculated over the period from May 2001 till May 2014 based on the bid - and offer - price.
10
3.2 Methodology
I use the same technique as Jegadeesh and Titman (1993), which is the standard in the literature. In order to make the winner and loser portfolio, stocks are ranked based on their return in the formation period, J. The return per stock is calculated using equation 1. The best 16 performing stocks are assigned to the winner portfolio and the 16 worst performing stocks are assigned to the loser portfolio. I choose for this amount because the average traded stocks are 164 and the portfolios are a decile of the entire sample, which is common in literature. This paper evaluates five different formation or J-periods; these J-periods are one, three, six, nine and twelve months. J-3 means that the return of a stock is assessed over a 3 month period, starting from t=-3 to t=0. I want to examine whether momentum is also observable in a short time frame; therefore I choose to include a 1-month formation period in my analysis. The returns are calculated using an arithmetic approach. In that way this research can be compared to other papers, including Jegadeesh and Titman (1993), Aygei-Ampomah (2007), Rouwenhorst (1998), and Chan et al. (1996).
(1)
Where stands for the total return of stock i at time t and stands for the returns of
stock i at time t and j stands for the formation period. Total return stands for the total growth in value of a stock assuming that dividends are used for repurchasing additional stocks3. Following equation 1 the following portfolios are made; portfolios consisting of the winner stocks and the portfolios consisting of the loser stocks . For each portfolio, I
evaluate a holding period of K months. I evaluate five different holding periods, one, three, six, nine and twelve months. K-3 means that the return of the stock is assessed over three months starting from t=1 to t=4. The holding period begins 1 month later than the formation period in order to mitigate any market microstructure effects. In this paper a strategy of, for example, six month formation period and a nine month holding period will be abbreviated to J;K 6*9.
The returns of the stocks in the portfolios are calculated using the same formula for the formation period, 1. Except for that j is now replaced by k, which stands for the holding period.
11
The total returns of the portfolios are calculated by the following equation:
(2)
Where (j;k) stands for the formation and holding period respectively, and for the winner
or loser portfolio at time t. N stands for the amount of stocks in the portfolio, which in my study are 16 per portfolio. The stocks are all equally weighted in the portfolio.
The combined performance of the winner and loser portfolio is calculated using the following equation:
(3)
is the overall performance of the momentum portfolio of the J;K strategy, where
the J and K stand for the formation and holding respectively. Whenever stocks are delisted during the formation period, they are not taken into account for the holding period. If stocks are delisted during the holding period, the return of these stocks is then set to zero. This is done irrespective of the cause of delisting. Academic research shows that assigning 0 to delisted stocks, irrespectively of the cause of delisting, reduces the profits of momentum. However, the reduction is negligible and the profits remain virtually unchanged (Liu et al. 2009).
The momentum portfolio, which consists of long positions in the past winners and a short position of the past losers, can be seen as a zero sum investment. There are no costs affiliated with implementing such a strategy, besides the trading costs, which will be elaborated later in this paper4. The portfolios are calculated on a rolling basis. The rolling basis method increases the power of the statistical analyses. To assure the robustness of the results of the winner portfolio, the loser portfolio and the performance of the overall results are tested for significance, using a student T-test.
(4)
√ ⁄
Where – is the sample mean, is the tested value, s is the standard deviation of the sample and n is the amount of observations.
12
I also examine whether Dutch momentum concentrates in different classes of stocks. Jegadeesh and Titman (1993, 2002) suggest that, for example, momentum profits are larger among small stocks, measured in market capitalization. These findings are also confirmed by Liu et al, (1999) and Rouwenhorst (1998). The loser and winner portfolios are tested for these concentrations. This will be done by a Chi-Squared test. The question is whether the two portfolios are disproportionate weighted towards certain characteristics of stocks. These characteristics are price, liquidity and market-cap. This disproportionate weighting can influence the overall profitability of the strategy. For instance: small cap and illiquid stocks can suffer from higher bid-offer spreads and higher commission fees.
The Chi-squared test is performed in the following way. First all stocks, traded in the sample time frame, are divided in different quintiles for price, market cap and liquidity. Every quintile consists of 20% of the sample size at that time. The quintiles are denoted as qi, it follows that every qi is 0.2. Then the amount of stocks, assigned to each quintile, is calculated (pi) for portfolio P at time t. The Chi-Test is used to test whether the distribution of the stocks in the loser and winner portfolio differ significantly from the total sample distribution.
The Chi- squared statistic is estimated as:
(5)
∑
13
costs associated with these formations and holding periods of the momentum portfolios; I link the turnover ratio to the relative performance of the portfolios. The turnover ratios are estimated by comparing the different compositions of the portfolios over the holding periods. At the end of the holding period, the portfolio composition is compared to the new composition of the portfolio. The portfolio turnover is therefore estimated by measuring the amount of stocks, which are new in the portfolio, divided by the total amount of stocks in the portfolio.
6
Where TR is the turnover ratio, So is the amount of stocks which are new in the portfolio and N stands for the total amount of stocks in the portfolio, which is 16.
In order to know the actual profitability of the momentum strategies, the different costs associated with the implementation of the formation and holding periods must be taken into account. Some of these costs, like the brokerage commissions, can easily be estimated. Other costs, and in particular the bid-offer spreads, cannot easily be estimated. This is because the impact that a trade has on the different prices is hard to perceive when the trade is not actually done (Roll, 1984). Different studies in literature suggest robust estimators of the trading costs. This paper follows the method used by Lesmond et al. (2004). Lesmond et al. (2004) suggest three different methods. This paper follows the method that calculates the quoted spreads, which involves estimating the spread by using the average spread in a 12-month observation, starting 18 months before the trade is made until six months before. Because of the lack of data on ask prices and bid prices in Datastream, this analysis is done from June 2002 until the end of the sample period. Note that in equation 7 the spreads in excess of 100% and time frames where no bid prices and ask prices are available have been left out.
7 ∑
Where PA is the ask-price and PB is the bid price.
I use a fixed commission fee of 0.67%; based on information from ABN AMRO, who state that this is a competitive rate for private clients. The total amount of costs, called the round-trip costs, is calculated using formula 8.
14
4. Empirical Results
Section 4.1 begins with a discussion about the results of the analyses and regressions of the full sample, named the unrestricted sample. After the presentation and discussion of these results, a sub-sample of stocks is analysed in section 4.2. This sub-sample is called the restricted sample; this sample is restricted in the way that small-cap stocks are excluded. Section 4.3 discusses the risk-adjusted returns of the unrestricted sample.
4.1 Unrestricted sample results
15
Rouwenhorst (1998) check the European stock market for momentum and found a strategy, which generate 16.2% per year. In more recent studies Agyei-Ampomah (2007) show major returns in the UK stock market with even returns above 44% per year, Rey and Schmid (2007) show more or less the same results for Sweden.
16
Table 2Average annualised portfolios returns for the different J-K strategies
Every month the stocks in the sample are ranked, based on their past performance in formation period J, assigned to their portfolio, winner or loser. Pw is the winner portfolio (best performing 16 stocks) and Pl is the loser portfolio (worst performing 16 stocks). The stocks are equally weighted in the portfolio and the overlapping returns are calculated, averaged and annualised for the various holding periods. The (t-values) tests whether the returns of the portfolio significantly differ from zero. Price describes the average share price in Euro, size describes the average market capitalisation in million euro and liquidity describes the average shares traded, per month, per million. A,B and C denotes 1%, 5% and 10% significance based on the (t-values).
Holding Period
Formation
Period Portfolio 1 M 3 M 6 M 9 M 12 M Price Size Liquidity
17
Table 2 Continued.
Holding Period
Formation
period Portfolio 1 M 3 M 6 M 9 M 12 M Price Size Liquidity
18
The averages for the different characteristics in table 2 could suggest, given the market sample averages, that the loser portfolio contains more expensive stocks. Table 2 also shows that, on average, the loser portfolio consists of stocks that belong to small-cap companies; the liquidity appears to be the same between the loser and winner portfolio. In order to fully assess the differences between the portfolios, table 3 shows the extent of the concentration of stocks within the loser and winner portfolios, based on the size, price and liquidity. Panel A presents the concentration of stocks based on their price. Table 3 panel A shows that the loser portfolio, on average, has a higher concentration of stocks with low prices, ranging from 36.2% for the one month formation strategy to 54.3% for the 12 months formation strategy. The significance level shows the amount of rejections of the Chi-squared test of no concentration. The significance level of loser portfolios is high for every formation period, except for the 1-month formation period where the significance level does not exceed the 90%. This proves that the loser portfolio is more biased towards stocks that have a lower price with a significance level of 10%.
If the concentration of stocks, based on price of the loser portfolio, is compared to the winner portfolio, the winner portfolio shows a more equal distribution. The concentration of the stocks with the smallest price in the winner portfolio ranges from 23.2% for the 1-month formation period to 165% in the 12 months formation period. The percentage that the Chi-squared test rejects is lower compared to the loser portfolio, ranging from 69.5% with the 1-month formation period to 15.47% with the 12 1-months formation period. The rejection rate never exceeds the 90% level; the distribution in the winner portfolio is not significantly different than the sample distribution. Panel B table 3 shows the distribution based on the market-cap of stocks. Panel B shows that loser portfolios are not equally distributed and have a higher concentration of small-cap stocks. For formation periods 6, 9 and 12 months the rejection rate of the Chi-squared statistic lays above 90%, which indicates that the concentration of the loser portfolio is significantly different from the concentration of the sample. Once again, the stocks in the winner portfolio are more evenly distributed.
19
Table 3Concentration of stocks in the winner and loser portfolio, unrestricted sample.
Every month the stocks in the sample are ranked based on their past performance in formation period J, and assigned to their portfolio, winner or loser. Stocks in the total sample in the specific time frame are categorised into quintiles, based on market capitalisation (size) panel B, shares traded per month, Panel C (liquidity), and price of the share, Panel A. The proportion of each quintile is therefore denoted as qi, 0.2. The proportion of the stocks in the loser and winner portfolio that belong to each of the quintiles, denoted as pi, is calculated. The chi-squared statistic, χ2, is calculated using the following formula:
∑
Where N is the portfolio size.
The χ2 is calculated for each month and averaged over the sample period. Chi values are the average χ2 values and the % significance is the percentage of times χ2 was more than the critical value, thus rejected. The number of degrees of freedom for this chi-test in each panel is 4 (I-1, where I is the number of groups).
1 - Month 3 - Months 6 - Months 9 - Months 12 - Months
Quintile Loser Winner Sample Loser Winner Sample Loser Winner Sample Loser Winner Sample Loser Winner Sample
20
Table 3 Continued
1 - Month 3 - Months 6 - Months 9 - Months 12 - Months
Quintile Loser Winner Sample Loser Winner Sample Loser Winner Sample Loser Winner Sample Loser Winner Sample
21
22
Table 4Portfolio Turnover unrestricted sample
This table shows the average percentage turnover of both the winner and loser portfolios for the different J-K periods. For each row, the top number is the average turnover per holding period; the number between brackets [] is the corresponding percentage annualised turnover.
Lesmond et al. (2004) show turnover ratios for the winner and loser portfolio of 170% and 155% respectively; Aygei-Ampomah (2007) show turnover ratios over 153% and 169% for the loser and winner portfolio respectively. Although the two papers research a bigger sample, they both show the same turnover ratios as I do. This implies that sample size does not matter for the turnover ratios. Figure 1 shows the turnover ratios of the J;K 6*6 strategy for each month over time. The turnover ratios, associated with this strategy, ranges from 50% to 100% turnover for the loser portfolio and 47% to 100% for the winner portfolio. Table 4 presents the substantial differences between the turnover ratios of the different strategies. In all cases the turnover ratio of the winner portfolio is slightly higher than the turnover ratio of the loser portfolio; this findings are also confirmed by Agyei-Ampomah (2007). The portfolios with the highest turnover ratios are those associated with the shortest
Holding period
Formation Period Portfolio 1-Month 3-Months 6-Months 9-Months 12-Months
23
holding period. Table 4 shows that if the formation period is longer, the turnover ratios are lower. The conclusion can be drawn that if a stock is judged over a longer formation period on its return, these returns are more robust and the stock is likely to meet the return criteria for the next holding period and therefore can stay in the portfolios. This lowers the turnover ratio. The only exception for this conclusion is the strategies involving a 12 months holding period.
Figure 1, Portfolio turnover for the J;K 6*6 strategy
Table 4 shows large differences between the turnover ratios of various strategies. High trading frequencies lead to high trading costs. For this reason the trading costs of the various strategies strongly differ from each other.
For every portfolio Pw or Pl the trading costs are calculated as the average of the round-trip cost for that portfolio at time t. The average round-trip costs of the Pw or Pl of time t are multiplied with the turnover ratio of that specific portfolio at month t. This study therefore shows the estimated trading costs based on actual turnover ratios. The trading costs for each portfolio are calculated each month for every strategy; these trading costs are then average and annualised. To estimate quoted spread, the bid and ask prices are needed, starting from 18 months untill 6 months before the portfolio formation. Datastream has accurate data starting from 2001; therefore the costs associated with momentum strategies are only calculated for portfolios starting after July 2002. Table 5 panel A reports the average trading costs per strategy, the total of trading costs are made up as stated in formula 8 in the methodology. In table 5 panel A W-L stands for the total costs of buying winners and
24
selling the losers of that particular strategy, the costs associated with the different strategies are almost equal to the findings made by Agyei-Ampomah (2007). Panel B reports the returns of the various strategies net of trading costs. This paper shows that the costs associated with momentum strategies are very high and substantial; this even leads to the conclusion that the reported strategies are not viable since their returns are negative. Table 5 reports on average higher trading costs for the winner portfolios; which is due to the slightly higher turnover ratios of the winner portfolio. Table 5 panel B does show that the returns become less negative if the holding and ranking periods increase. The best, or less worst, result is reached by the J;K 3*12 strategy. The worst performance is of the J;K 1*1 strategy, yielding a negative return of almost 145%.
The results of the unrestricted sample of the Dutch stock market differ from other papers; Aygei-Ampomah (2007) show significant positive returns for strategies involving longer holding and formation periods. Korajczyk and Sadka (2004) find that some momentum strategies, net of trading costs, are profitable and therefore do not explain the momentum anomaly. My research shows the opposite; trading costs do explain the momentum anomaly observed in the Dutch stock market since it destroys all excess return. Therefore the anomaly is only observed virtually and cannot be exploited since implementing a momentum strategy yield significant losses.
25
Table 5Portfolio implementation costs and net returns based on the quoted spread plus commission.
Panel A shows the average trading costs, based on the quoted spread plus the commissions. W-L describes the total cost of the momentum strategy of both the winners and the losers portfolio. Panel B shows the post-cost returns of the different J-K strategies, after trading costs. A,B and C denotes the significance at 1%,5% and 10% respectively.
Holding period Formation
period Portfolio 1-month 3-months 6-months 9-months 12-months
Panel A: Trading costs (%) based on the quoted spreads plus commissions and taxes
1 Loser 71.45% 22.97% 11.51% 7.77% 5.73% Winner 72.76% 23.53% 12.30% 8.29% 6.22% W-L 144.21% 46.50% 23.81% 16.07% 11.95% 3 Loser 46.75% 24.21% 12.46% 8.46% 6.92% Winner 48.93% 25.88% 13.15% 8.84% 6.72% W-L 95.68% 50.09% 25.61% 17.30% 13.64% 6 Loser 34.94% 18.36% 12.48% 8.56% 7.43% Winner 37.37% 19.82% 13.40% 9.12% 7.13% W-L 72.31% 38.19% 25.89% 17.68% 14.56% 9 Loser 28.90% 14.85% 10.28% 8.21% 7.46% Winner 31.26% 16.57% 11.40% 9.25% 6.95% W-L 60.16% 31.41% 21.68% 17.46% 14.41% 12 Loser 27.99% 14.06% 9.64% 7.75% 8.16% Winner 27.97% 14.84% 10.42% 8.37% 6.97% W-L 55.96% 28.90% 20.05% 16.12% 15.13%
Panel B: Returns of the winners minus the losers after trading costs
26
4.2. Restricted sample
This section analyses the returns of the restricted sub-sample of large-cap and liquid stocks. The analysis for the sub-sample is an exact copy of the analysis as described above for the unrestricted sample. Agyei-Ampomah (2007) shows that, when the small-cap and illiquid stocks are rejected from their sample, the spreads dramatically lower. The authors present evidence that trading in large and liquid stock is easier and cheaper. As stated above, the sub-sample includes only stocks that belong in around the top 30%, based on their market cap. If the market cap falls below the required rate during the holding period, the stocks are not excluded once the stocks are in the sample. Due to the limited amount of stocks available, the amount of stocks in the winner and loser portfolio of the restricted sample will decrease to five.
27
Table 6Return restricted sample.
Average annualised portfolios returns for the different J-K strategies
Every month the stocks in the sample are ranked, based on their past performance in formation period J, assigned to their portfolio, winner or loser. Pw is the winner portfolio (best performing16 stocks) and Pl is the loser portfolio (worst performing16 stocks). The stocks are equally weighted in the portfolio and the overlapping returns are calculated, averaged and annualised for the various holding periods. The (t-values) tests whether the returns of the portfolio significantly differ from zero. . A,B and C denotes the significance at 1%,5% and 10% respectively.
28
Figure 2 presents the difference between the trading costs of the winner portfolio of the restricted and unrestricted sample of the J;K 6*6 strategy. It shows that there is indeed a substantial difference between the trading costs. The turnover ratios for the different strategies do not differ substantial from the unrestricted sample.
Figure 2 Trading costs winner portfolio J;K 6*6 strategy of the restricted and unrestricted sample
The restricted sample shows average trading costs of 3.32%, whereas table 5 presents averaged and annualised trading costs of 13.40%. The results prove that the roundtrip costs and therefore the bid–ask spreads are lower for the unrestricted sample. This is due to the exclusion of small-cap and illiquid stocks and is in line with the findings made by Agyei-Ampoma (2007). Those findings show similar reduced spreads and trading costs for the restricted sample. Although the spreads and trading costs are substantially less for the restricted sample, there are no strategies that yield significant positive returns. Figure 2 shows that the debt crisis in 2008 caused spreads to rise. Table 7 shows the results net of transaction costs; as for the unrestricted sample also the restricted sample is evaluated over a time frame ranging from 2002 untill 2014. The presented results are on average better than the results of the unrestricted sample, but still negative. The worst performing strategy is again J;K 1*1 and the best, or least worst, performing strategy is J;K 12*12 which yields a negative return of 4.22%. The conclusion is, when small-cap stocks are excluded from the sample, the returns of the various momentum strategies are no longer significantly positive. The trading costs associated with the strategies however also drop substantially, therefore the net effect is that the restricted sample has better results on average than the unrestricted sample, although both negative. Therefore the gains of excluding small-cap
29
stocks, in terms of trading costs, are higher than the returns they generate for the momentum strategy.
Table 7
Net profit momentum investment strategy of the restricted sample
This table shows the post-cost returns of the different J-K strategies, after trading costs. A,B and C denotes the significance at 1%,5% and 10% respectively.
This paper also considers the risk-adjusted abnormal returns of the winner and loser portfolio of the various momentum strategies. To do this, all the various returns are adjusted using the Fama and French (1996) three-factor model. This is only done for the unrestricted sample since the three-factor model considers size effects, and a size effect adjustment is not suitable for a sample of only large stocks (Fama and French, 1998). The three-factor model is presented by the following equation:
(9) Rpt – RFt =αp + β1P (RMt – RFt ) + β2PSMB + β3PHML + εPt
Rpt stands for the portfolio return and RFt stands for the risk-free rate, SMB stands for the Fama and French size factor, and HML stands for the book-to-market value. The values needed to perform this regression, are downloaded from the Fama website. The European values are used; considering they are a good proxy since the Dutch stock market is part of European market and is strongly correlated with other European countries. The returns, presented in table 2, should disappear if adjusted for the common risk factors if those returns were mere a compensation for the higher risk the portfolios carry. Table 8 shows that the unrestricted sample displays significant risk-adjusted returns for almost all strategies when the returns of the loser portfolio are subtracted from the returns of the Formation
Period Portfolio 1-Month 3-Months 6-Months 9-Months 12-months
Returns of the winners minus the losers after trading costs
30
31
Table 8Risk-adjusted portfolio returns
This table reports the risk-adjusted abnormal returns for the winner portfolio, loser portfolio and the winner-loser portfolio (w-l). These abnormal returns are based on the Fama French three-factor model. The (t-values) shows the significance of the returns, and A,B and C denotes the significance at 1% 5% and 10% respectively.
Holding period Formation
Period Portfolio 1-Month 3-Months 6-Months 9-Months 12-Months
32
5. Conclusion
This paper examines the momentum anomalies in the Dutch stock market, and the post-cost profitability of such anomalies. The Dutch stock market, during 1996-2014, shows significant positive returns for most of the examined strategies. The composition of the different portfolios, their turnover ratios, and trading costs are also examined. Although momentum profits are reported, it shows that when those returns are corrected for trading costs, the returns are fully destroyed and that most momentum strategies are showing significant negative results. This is in line with the hypothesis in the literature part. Examining the composition of the portfolios shows that the loser portfolio mostly consists of small-cap stocks. These stocks are known for being too expensive to trade (Ampyei-Ampomah, 2007; Rouwenhorst 1999). Having these stocks in a portfolio results in high trading costs. The distribution of the stocks in the winner portfolio shows a less skewed distribution. The momentum returns in the Dutch stock market as presented, are different from momentum returns in other research in the way that momentum returns in the Dutch stock market are less dependent on the short position in the loser portfolio; this is of interest for investors who are excluded from short positions (Agyei-Ampomah, 2007; Jegadeesh and Titman, 1993; Grinblatt and Moskowitz, 2003; Doukas and McKnight, 2005).
Further, this study shows the intensity of momentum investing strategies in the Dutch stock market by presenting the turnover ratios for the different strategies. Portfolio turnover ratios as high as 1000% per year were reported for the strategies involving short holding periods and as low as 85% per year for strategies involving long holding periods. High turnover ratios lead to excessive trading costs, higher than the trading costs implied by other research, such as the 0.5% by Jegadeesh and Titman (1993). The skewed distribution towards small-cap stocks in the loser portfolio also increases the trading costs because of large bid-ask spreads. By forming a sub-sample that excludes small-cap stocks, I tried to lower the costs of trading in order to find a significant positive momentum investment strategy. The restricted sample shows that much lower trading costs are achieved, but also that by excluding the small-cap stocks a large part of the momentum return is lost. The gain of having lower trading costs is higher than the loss of the returns of the stocks that are excluded from portfolio; this is also reported in the study of Agyei-Ampomah (2007)
33
those alphas were even higher than the non-risk-adjusted returns. However those alphas are not high enough to be profitable when corrected for trading costs.
In conclusion, the results suggest that after accounting for trading costs, the presented momentum returns are not high enough to be profitable after trading costs. Even so, every strategy investigated shows substantial and mostly significant negative returns. A conclusion can be that the Dutch stock market is more efficient than the markets where post-cost momentum returns are reported.
5.1 Propositions for further research
This study examines the short and medium formation and holding periods, and shows that no profitable momentum strategy is found. However, with the increase of the formation and holding periods the returns also increase. Therefore it could be interesting whether longer formation and holding periods do reveal positive strategies. Further research could be interesting for practitioners.
34
6. References
Agyei‐Ampomah, S., 2007. The Post‐Cost Profitability of Momentum Trading Strategies: Further Evidence from the UK. European Financial Management,134, 776-802. Antoniou, A., Galariotis, E. C., and Spyrou, S. I., 2005. Contrarian profits and the overreaction
hypothesis: The case of the Athens stock exchange. European Financial Management, 111, 71-98.
Barber, B. M., and Odean, T., 2001. Boys will be boys: Gender, overconfidence, and common stock investment. The Quarterly Journal of Economics, 1161, 261-292.
Barberis, N., Shleifer, A., and Vishny, R., 1998. A model of investor sentiment. Journal of financial economics, 493, 307-343.
Bondt, W. F., and Thaler, R., 1985. Does the stock market overreact?. The Journal of finance, 403, 793-805.
Chan, K. C., 1988. On the contrarian investment strategy. Journal of Business,612, 147. Chan, L. K., Jegadeesh, N., and Lakonishok, J., 1996. Momentum strategies. The Journal of
Finance, 515, 1681-1713.
Chan, L. K., Jegadeesh, N., and Lakonishok, J., 1999. The profitability of momentum strategies. Financial Analysts Journal, 80-90.
Conrad, J., and Kaul, G., 1998. An anatomy of trading strategies. Review of Financial Studies, 113, 489-519.
Daniel, K., Hirshleifer, D., and Subrahmanyam, A., 1998. Investor psychology and security market under‐and overreactions. the Journal of Finance, 536, 1839-1885.
De Bondt, W. F., and Thaler, R. H., 1987. Further evidence on investor overreaction and stock market seasonality. Journal of finance, 557-581.
Dische, A., 2002. Dispersion in analyst forecasts and the profitability of earnings momentum strategies. European Financial Management, 82, 211-228.
Doukas, J. A., and McKnight, P. J., 2005. European momentum strategies, information diffusion, and investor conservatism. European Financial Management, 113, 313-338.
Fama, E. F., 1965. The behavior of stock-market prices. Journal of business, 34-105.
Fama, E. F., 1998. Market efficiency, long-term returns, and behavioural finance. Journal of financial economics, 493, 283-306.
35
Fama, E. F. and French, K. R., 1998. Value versus growth: The international evidence. The Journal of Finance, 536, 1975-1999.
Forner, C., and Marhuenda, J., 2003. Contrarian and momentum strategies in the Spanish stock market. European Financial Management, 91, 67-88.
Frazzini, A., Israel, R., and Moskowitz, T. J., 2012. Trading costs of asset pricing anomalies. Unpublished working paper. University of Chicago, Chicago, IL. Griffin, J. M., Ji, X., and Martin, J. S., 2003. Momentum investing and business cycle risk:
Evidence from pole to pole. The Journal of Finance, 586, 2515-2547.
Grinblatt, M., and Moskowitz, T. J., 2004. Predicting stock price movements from past returns: The role of consistency and tax-loss selling. Journal of Financial Economics, 713, 541-579
Grundy, B. D., and Martin, J. S., 2001. Understanding the nature of the risks and the source of the rewards to momentum investing. Review of Financial Studies, 141, 29-78. Hon, M. T., and Tonks, I. 2003., Momentum in the UK stock market. Journal of Multinational
Financial Management, 131, 43-70.
Hong, H., and Stein, J. C., 1999. A unified theory of underreaction, momentum trading, and overreaction in asset markets. The Journal of Finance, 546, 2143-2184.
Jegadeesh, N., and Titman, S., 1993. Returns to buying winners and selling losers: Implications for stock market efficiency. The Journal of Finance, 481, 65-91.
Jegadeesh, N., and Titman, S., 2001. Profitability of momentum strategies: An evaluation of alternative explanations. The Journal of Finance, 562, 699-720.
Jegadeesh, N., and Titman, S., 2002. Cross‐Sectional and Time‐Series Determinants of Momentum Returns. Review of Financial Studies, 151, 143-157.
Keim, D. B., 2003. The cost of trend chasing and the illusion of momentum profits. Rodney L. White Center for Financial Research
Keim, D. B., and Madhavan, A., 1997. Transactions costs and investment style: an inter-exchange analysis of institutional equity trades. Journal of Financial Economics, 463, 265-292.
Korajczyk, R. A., and Sadka, R., 2004. Are momentum profits robust to trading costs? The Journal of Finance, 593, 1039-1082.
Lesmond, D. A., Schill, M. J., and Zhou, C., 2004. The illusory nature of momentum profits. Journal of Financial Economics, 712, 349-380.
36
MacKinlay, A. C., 1995. Multifactor models do not explain deviations from the CAPM. Journal of Financial Economics, 381, 3-28.
Marhuenda, M. A., Marhuenda, Y., and Morales, D., 2003. On the computation of the exact distribution of power divergence test statistics. Kybernetika, 391, 55-74.
Odean, T., 1998. Do investors trade too much? Available at SSRN 94143.
Rey, D. M., and Schmid, M. M., 2007. Feasible momentum strategies: Evidence from the Swiss stock market. Financial Markets and Portfolio Management, 213, 325-352. Roll, R., 1984. A simple implicit measure of the effective bid‐ask spread in an efficient
market. The Journal of Finance, 394, 1127-1139.
Rouwenhorst, K. G., 1999. European equity markets and the EMU. Financial Analysts Journal, 57-64.
37
Acknowledgements.
