1
Country and industry momentum strategies using
indexes and Exchange Traded Funds
Otto Bazuin∗ Master’s Thesis Finance University of Groningen
Abstract
This paper investigates momentum strategies across countries and industries. Indexes and Exchange Traded Funds (ETFs) are used to analyze whether an individual investor is able to exploit the momentum effect. Winner-minus-loser portfolios are created using a calendar time procedure and the returns are risk-adjusted using the Fama-French three-factor model. Using indexes, significant results are found for the existence of the momentum effect in countries and industries. No evidence is found for momentum in country and industry ETFs. However, if it is assumed that ETFs will generate the same returns as indexes in the long run, an individual investor is able to yield a net annualized risk-adjusted return of 4.77% using country ETFs and 5.59% using industry ETFs.
Key words: Momentum effect, behavioral finance, individual investor JEL codes: G12, G14
I.
Introduction
Buy past winners and sell past losers. Such a simple strategy is one of the most researched topics in the academic field of finance.1 Individual stocks across the world, country indexes and industries have been
extensively investigated for momentum. Momentum is the tendency of asset prices to exhibit persistence in their performance, first documented by Jegadeesh and Titman (1993). Investments that have performed relatively well continue to perform relatively well; those that have performed relatively poorly continue to perform relatively poorly. Strategies that exploit the momentum effect are called momentum strategies or relative strength strategies. A momentum strategy involves constructing a long-short portfolio; it
∗Student at the Faculty of Economics and Business, University of Groningen , The Netherlands.
E-mail: obazuin@gmail.com, student number: s2213788.
1 See for example Jegadeesh and Titman (1993); Carhart (1997); Rouwenhorst (1998); Moskowitz and Grinblatt (1999); Jegadeesh and Titman
2
purchases securities with strong performance and sells securities with poor performance in the previous period.
Existing literature (Jegadeesh and Titman, 1993; Rouwenhorst, 1998; Moskowitz and Grinblatt, 1999) has a focus on employing momentum strategies with an investment universe consisting of many individual securities. For an individual investor it is cumbersome to implement such momentum strategies because it requires buying and selling of many individual stocks. Besides that, a lot of research is based on stock indexes which are not real tradable assets. An investor has to buy and sell all the stocks in the index separately to implement such a strategy. However, someone can buy an index fund tracking these indexes, but do strategies using index funds yield the same returns?
Bhojraj and Swaminathan (2006) argue that country momentum profits can be earned by buying stock market indexes of countries with high past returns and selling country indexes with low past returns. However, these stock market indexes are not tradable, so the investor still has to buy and sell many different stocks. Moskowitz and Grinblatt (1999) find that returns earned by individual stock momentum strategies can be captured by following industry momentum strategies. Individual stocks are grouped by industry classifications. An investor implementing such a strategy would face high transaction costs because all individual stocks should be bought and sold separately. O’Neal (2000) concludes that past winner industry mutual funds over past loser industry mutual funds earn excess returns after correcting for fees. The downside of this long-short strategy is that it is barely possible to sell short mutual funds for the retail investor. Therefore, it is impossible to implement this strategy for the retail investor. O’Neal (2000) concludes that a long-only strategy does not earn significant excess returns. Andreu et al. (2013) focus on country and industry momentum. They argue that an investor would have been able to exploit the momentum effect with an excess return of 5% per annum. Andreu et al. (2013) also analyze Exchange Traded Funds (ETFs) but do not find significant abnormal returns.
3
these indexes. After that, momentum strategies using ETFs are examined to see if the same returns can be achieved.
This paper focuses on the perspective of the individual investor. Is an individual investor able to take advantage from the momentum anomaly? And is the performance still significant after controlling for trading costs? This results in the following research question:
Is an individual investor able to exploit the momentum anomaly using Exchange Traded Funds?
The contribution of this paper is two-fold. Firstly, this paper analyzes ETFs which can be relatively easily traded, whereas many studies use non-tradable securities. Secondly, the perspective of this paper is from an individual investor; therefore, all trading costs are taken into account which an individual investor has to deal with. Furthermore, the worse performance of momentum strategies during the period 2009 and thereafter will be discussed. Finally, this will result in a conclusion whether an individual investor is able to implement a profitable momentum strategy in real-life.
4
II.
Literature review
This section discusses the relevant literature. I start with a discussion of the empirical evidence for the momentum effect. Secondly, the momentum effect for ETFs and related trading costs are presented. Lastly, different theories explaining the momentum effect are described.
2.1 Empirical evidence
Since the seminal paper of Jegadeesh and Titman (1993), the momentum phenomenon has been studied extensively. Jegadeesh and Titman (1993) document the performance of momentum strategies for US common stock returns. They define momentum strategies as follows. The length of the period over which past returns are calculated, J, is referred as the formation period. The length of time the position is held, K, is referred as holding period. “Winners” are those firms with the highest formation-period returns and “losers” are those stocks with the lowest formation-period returns. Stocks with the top 10% ranking-period returns are defined as “winners” and stocks with the lowest 10% ranking-ranking-period returns are defined as “losers”. A zero-cost strategy is created by taking a long position in the top decile and a short position in the lowest decile. Jegadeesh and Titman (1993) use monthly data, so J and K are measured in months. Momentum strategies are denoted as (J,K) strategies. For example, a strategy with a formation and holding period of six months is denoted as (6,6) strategy. Jegadeesh and Titman (1993) consider strategies based on stock returns over the past 3, 6, 9 or 12 months and then hold them for 3, 6, 9 or 12 months. They also implement a waiting period of one week between the formation and holding period due the short-term reversal effect (Jegadeesh and Titman, 1991). Jegadeesh and Titman (1993) conclude that all, except for one, zero-cost momentum portfolios generate statistically significant positive returns. The most profitable strategy is a strategy with a formation period of 12 months, a waiting period of one week and holding period of three months. Jegadeesh and Titman (2001) report continuing efficacy of the momentum strategies after the time of the publication of their original paper.
Some studies implement the momentum strategy at the end of the formation period and hold it for K months. Other studies implement a waiting period between the formation period and holding period, a so called waiting period and denoted by its length S. Andreu et al. (2013) apply a waiting period of one month and state that the results are similar to results without a waiting period.
5
that of Jegadeesh and Titman (1993) for US equity markets. Bhojraj and Swaminathan (2006) examine momentum and reversals for 38 international stock market indexes. They find that winners outperform losers in the first 3 to 12 months after the formation period. After this period winners underperform losers in the subsequent 2 years.
Hameed and Kusnadi (2002) investigate six (Hong Kong, Malaysia, Singapore, South Korea, Taiwan and Thailand) emerging Asian stock markets. Unlike studies of the U.S. and European markets, they find little evidence of momentum in the Asian markets. Applying different holding periods and formation periods, the strategies consistently result in insignificant profits. Hameed and Kusnadi (2002) conclude that the factors that contribute to the momentum phenomenon in the United States are not present in Asian markets. Kang, Liu and Ni (2002), however, do find significant abnormal profits for momentum strategies in the Chinese stock market.
Moskowitz and Grinblatt (1999) analyze individual stock momentum and industry momentum. They create 20 monthly industry portfolios during the period July 1963 to July 1995. They report a strong momentum effect in industry components of stock returns which accounts for much of the individual stock momentum anomaly. They argue that when individual stock momentum strategies are controlled for industry momentum, the strategies are less profitable. The results suggests that individual stock momentum strategies are not very well diversified because the winners and losers tend to be from the same industry. Industry momentum strategies (buy past winner and sell past loser industry portfolios) appear to be highly profitable. The results are still significant after controlling for individual stock momentum, size and book-to-market equity.
Momentum strategies across asset classes has received less attention in the existing literature. Blitz and van Vliet (2008) examine momentum for global tactical asset allocation (GTAA) strategies across 12 different asset classes in the period 1974-2007. They find significant excess returns for long-short portfolios constructed with a formation period of twelve months and a holding period of one period. The returns cannot be explained by potential structural biases towards asset classes with high risk premiums, nor the Fama-French and Carhart factors.
Fama and French (1993) report that momentum returns based on individual stock momentum cannot be explained by the Fama-French 3-factor model.
2.2 Momentum in Exchange Traded Funds
6
Funds, the latter which can be traded directly. Both are analyzed separately using the methodology from Jegadeesh and Titman (1993). For non-tradable assets U.S. industry momentum is investigated over the period 1926-2009. The (6,6) momentum strategy delivers a return of 0.43% per month and a risk-adjusted return of 0.62% per month (significant at 1% level). The returns reported are similar to the returns found by Moskowitz and Grinblatt (1999), who investigate the period 1965 to 1995. Andreu et al. (2013) state the Fama-French three-factor model cannot explain industry momentum returns. Analyzing momentum industry with tradable assets (ETFs), Andreu et al. (2013) find that returns are of similar size compared to those of non-tradable industries. The (6,6) momentum strategy generates an abnormal return of 0.84% per month and a risk-adjusted return of 1.16% per month during the period December 1998 to December 2009. These returns are not statistically significant. Andreu et al. (2013) argue that this is caused by the relatively short sample period that these ETFs are available to investors.
Country momentum strategies have been even more profitable. Andreu et al. (2013) investigate country index momentum during the period 1970-2009. Country index momentum strategies earn about 8% per annum excess return. These returns could also have been captured by trading ETFs, however the results for the ETF sample are not statistically significant. They argue that given long run statistical significance of country index momentum an investor may be confident of the existence of the effects. To see if the ETFs are successful in tracking an index, they calculate the tracking error for the ETFs. The tracking errors are not reported but they argue that the tracking error is generally below 1%. Therefore, they conclude that their study does not suffer from interaction of country or industry momentum with active portfolio management. This might be the case in the paper of O’Neal (2000), who uses sector mutual funds.
7
Korajczyk and Sadka (2004) argue that the estimated excess returns on momentum strategies disappear when a portfolio size of $4.5 to $5.0 billion is used in momentum strategies.
De Jong and Rhee (2008) analyze contrarian and momentum strategies using ETFs. The sample period ranges from March 1996 to December 2005. The sample starts with 19 ETFs in 1996 and ends with 217 ETFs in 2005. They argue that ETFs provide economically and statistically significant abnormal returns for momentum strategies with a formation and holding period from 4 to 39 weeks. De Jong and Rhee (2008) report annualized momentum risk-adjusted returns ranging from 8.4 to 13.5%, significant at the 1% level. De Jong and Rhee (2008) also calculate transaction costs and apply this to the momentum strategy with a formation and holding period of six months. The annualized return of this strategy is 12.83%. The total transaction costs, consisting of bid-ask spread and broker commissions, are 8.25% per year. This reduces the annualized risk-adjusted return to 4.58% per year. De Jong and Rhee (2008) conclude that after taking trading costs into account, abnormal momentum returns are not illusionary but economically and statistically significant. The results of De Jong and Rhee (2008) are in contrast with Lesmond, Schill, and Zhou (2004) who report that momentum abnormal returns are illusory due to high trading costs. However, De Jong and Rhee (2008) argue that ETFs are much less costly than individual equities because they have smaller bid-ask spreads and more liquidity which reduces the price impact of large trades. Therefore, abnormal returns are achievable by investors using ETFs, which are ideal instruments to implement a momentum strategy.
2.3 Drivers of the momentum effect
The empirical findings are a source of debate in the literature regarding market efficiency. The momentum effect challenges the weak form of market efficiency. The weak-form of efficient market hypothesis states that past price performance cannot be used to predict future performance. Abnormal returns cannot be earned by using investment strategies based on historical data. Thus, momentum may be associated with market inefficiency. Two different explanations are described in literature to explain this. One explanation based on transaction costs and one explanation based on behavioral theories.
8
because investors hold on to it for too long. In the long-run, the market value of the stock will move to the intrinsic value of the stock, resulting in higher positive returns for winning stocks and higher negative returns for losing stocks.
Daniel et al. (1998) and Barberis et al. (1998) state that momentum and reversals are the result of under- and overreaction to fundamental news. The theories predict that momentum and reversals are not independent effects but are related to each other. Initial momentum is followed by subsequent reversals. First, security prices underreact to information resulting in momentum and then overreact to information leading to reversals. Daniel et al. (1998) argue that security under- and overreaction is based on two psychological biases: investor overconfidence; and biased self-attribution, which causes asymmetric shifts in investors’ confidence as a function of their investment outcomes
Barberis et al. (1998) present a model of how investors form expectations of future earnings. The model is motivated by the evidence that in making forecasts people pay too much attention to the strength of the evidence they are presented with and too little attention to its statistical weight. For example, corporate earnings announcements represent information that is of low strength but significant statistical weight. This yields the prediction that stock prices underreact to earnings announcements and similar events. Also, patterns of news, such as series of good earnings announcements, represent information that is of high strength and low weight. This yields that a prediction that stock prices overreact to consistent patterns of good or bad news.
Bhojraj and Swaminathan (2006) argue that the theories predicting the momentum and reversal effect, under- and overreaction, are not only limited to individual stocks, but that this is also observable across country indexes. In case of country indexes, the relevant fundamental news for under- and overreaction is not news about earnings but macroeconomic news.
Moskowitz and Grinblatt (1999) use the model of Daniel et al. (1993) to explain industry momentum. Investors who have overconfidence and attribution biases may exhibit more overconfidence and self-attribution in certain types of industries over time. The difficulty of valuing new or changing industries may cause greater overconfidence among investors who are employed in these sectors or know these sectors very well, which aggravates industry mispricing.
9
post holding period returns tends to depend on the composition of the sample, the sample period, and sometimes whether the post holding period returns are risk adjusted. Thus, positive momentum returns are sometimes associated with post holding period reversals and sometimes not. This suggests that behavioral models provide only a partial explanation for the momentum anomaly.
2.4 Momentum crashes
10
III.
Data
This section describes the data used in this paper. Two samples are created for both countries and industries. Two samples based on non-tradable indexes (MSCI indexes) and two samples based on real tradable assets, ETFs. In section one, ETFs in general are discussed. Section two describes country samples and section three discusses industry samples.
3.1 Exchange Traded Funds
An ETF is an index fund that combines the diversified portfolio features of a mutual fund with the trading possibilities of an individual security. An index fund is a collective investment scheme that aims to mimic the movements of an index of a certain market. An ETF trades like a stock on an exchange. It can bought and sold intraday, bought on margin and sold short (De Jong and Rhee, 2008). Salomon and Tarasow (2000) describe that ETFs are easier to sell short than stocks because ETFs are exempt from the uptick rule. Ordinary mutual funds can only be bought and sold by market orders for end-of-day prices, and cannot be bought on margin or sold short. The first ETF was introduced in 1993 and by the end of 2012 the total number of ETFs has grown to 1,194 with total net assets of over $ 1.3 trillion.2
3.2 Country sample
Following Richards (1997) and Bhojraj and Swaminathan (2006), I select 16 developed countries to test for country momentum. I use Morgan Stanley Capital International (MSCI) indexes with total returns in U.S. dollars. U.S. dollar returns are obtained because the corresponding ETFs also trade in U.S. dollars. The indexes are composed of high capitalization, heavily traded stocks and are unlikely to be subject to bid-ask spread biases (Richards, 1997). Also, the data availability for these MSCI indexes is the longest. The sample period ranges from January 1971 to December 2012. For the ETF sample, the corresponding ETFs that track the MSCI country indexes are selected. The MSCI country indexes are tracked by iShares Inc. The country ETF sample ranges from January 2000 to December 2012. For each ETF and MSCI index the total return index (RI) data is obtained from DataStream to account for dividend payouts and splits. The data availability of the ETFs is relatively short, therefore first momentum is analyzed across the MSCI indexes.
Returns are calculated as follows,
11
where ,௧ is the monthly simple return for index (ETF) i in month t and ,௧ is the total return index for
index (ETF) i in month t. Table 1, Panel A, summarizes the descriptive statistics of the country index sample, Panel B presents the descriptive statistics of the country ETF sample. The average return across all country indexes is 1.07%, the average return of all country ETFs is 0.61%. The average standard deviation of country indexes is 6.76% and for country ETFs this is 7.17%.
Table 1: Descriptive statistics of country returns
This table presents the descriptive statistics of the country index sample and country ETF sample. The country index sample ranges from January 1971 to December 2012. The country ETF sample ranges from January 2000 to December 2012. The statistics are calculated over simple monthly returns, based on total return index data. Panel A presents the country index sample, Panel B presents the country ETF sample. Means are calculated as arithmetic means.
Panel A: Country indexes Mean (%) St. dev. (%) Skewness Kurtosis Min (%) Max (%)
Australia 1.03 7.12 -0.60 4.18 25.51 -44.51 Austria 0.93 7.02 -0.12 4.16 32.67 -34.26 Belgium 1.10 6.07 -0.45 4.58 26.76 -35.55 Canada 0.98 5.90 -0.47 2.31 23.30 -25.77 France 1.03 6.70 -0.11 1.23 26.83 -23.18 Germany 1.01 6.52 -0.27 1.16 23.78 -22.00 Hong Kong 1.70 10.21 0.90 11.09 87.86 -43.44 Italy 0.72 7.55 0.16 0.72 30.99 -22.57 Japan 0.93 6.27 0.20 0.66 24.26 -19.38 Netherlands 1.13 5.73 -0.49 2.22 25.68 -24.62 Singapore 1.28 8.39 0.42 5.30 53.27 -41.34 Spain 0.96 6.94 -0.04 1.64 26.72 -27.31 Sweden 1.34 7.14 -0.13 0.72 22.91 -22.72 Switzerland 1.07 5.47 -0.16 1.33 24.58 -17.64 United Kingdom 1.03 6.55 1.11 10.40 56.41 -21.53 United States 0.88 4.63 -0.41 1.96 17.79 -21.22
Panel B: Country ETFs Symbol Mean (%) St. dev. (%) Skewness Kurtosis Min (%) Max (%)
Australia EWA 1.18 7.37 -0.21 2.33 -23.85 28.61 Austria EWO 1.06 8.51 -0.58 3.09 -34.31 31.70 Belgium EWK 0.51 7.35 -1.03 3.16 -33.83 19.11 Canada EWC 0.94 6.90 -0.46 1.71 -25.16 21.62 France EWQ 0.42 7.20 -0.34 0.80 -23.01 20.36 Germany EWG 0.61 8.18 -0.27 1.02 -23.22 26.63
Hong Kong EWH 0.69 6.86 -0.16 1.02 -21.55 21.78
Italy EWI 0.29 7.81 -0.23 0.66 -23.72 19.51 Japan EWJ -0.03 5.65 0.13 0.44 -15.66 18.38 Netherlands EWN 0.40 7.24 -0.56 1.28 -25.89 20.97 Singapore EWS 0.89 7.74 -0.07 1.61 -25.28 29.74 Spain EWP 0.67 8.00 -0.07 1.00 -25.03 24.04 Sweden EWD 0.85 9.22 -0.01 0.63 -24.32 27.29 Switzerland EWL 0.61 5.79 -0.30 1.95 -16.34 21.48
United Kingdom EWU 0.36 5.84 -0.26 1.64 -20.71 17.18
United States SPY 0.29 5.00 -0.44 1.36 -16.33 15.66
3.3 Industry samples
12
consumer non-durables, consumer durables, manufacturing, energy, business equipment, telecom, wholesale and retail, healthcare, utilities and other. The index industry sample ranges from January 1928 to December 2012. The data is obtained from Kenneth R. French’s website3. Further on, this sample is referred to as the industry index sample.
The selection of the ETF sample is based on Select Sector SPDRs. Select Sector SPDRs are ETFs that divide the S&P 500 into nine industry index funds, which together they represent the S&P 500. The nine industries are consumer discretionary, consumer staples, energy, financials, health, industrials, information technology, materials, and utilities. Select Sector SPDR are ETFs that track Select Sector Indexes. Select Sector Indexes are designed to measure the performance of Global Industry Classification Standard (GICS®) sectors, the broadest level of industry classification.4 In total GICS divides sectors into
10 sectors. There are only nine ETFs because two of the GICS sectors, Information Technology and Telecommunication Services, have been combined to form the Select Sector Technology Index. The reason to choose for Select Sector ETFs is that these are available for the longest time period.
The sample for ETFs ranges from January 2000 to December 2012. For each ETF the total return index (RI) data is obtained from DataStream to account for dividend payouts and splits. Table 2, Panel A, summarizes the descriptive statistics of the industry index sample, Panel B presents the descriptive statistics of the industry ETF sample.
The average return across all industry indexes is 0.99 %, the average return of all industry ETFs is 0.52%. As with the country sample, the average return of the ETF sample is lower than the index sample. The average standard deviation of industry indexes is 6.06% and for industry ETFs this is 6.06%. The standard deviation for industry indexes and industry ETFs is coincidentally the same.
13
Table 2: Descriptive statistics of industry returns
This table presents the descriptive statistics of the industry index sample (January 1928 - December 2012) and industry ETF sample (January 2000 to December 2012). The statistics are calculated over simple monthly returns, based on total return index data. Panel A presents the industry index sample, Panel B presents the industry ETF sample. Means are calculated as arithmetic means.
Panel A: Industry indexes Mean (%) St. dev. (%) Skewness Kurtosis Min (%) Max (%)
Business Equipment 1.08 7.38 0.27 5.93 53.43 -33.83 Consumer non-durables 0.98 4.65 -0.05 5.74 34.41 -24.48 Consumer durables 1.09 7.82 1.12 13.91 79.71 -34.78 Energy 1.05 6.02 0.19 3.01 33.50 -25.98 Healthcare 1.07 5.70 0.16 7.20 38.66 -34.74 Manufacturing 1.01 6.36 0.87 11.89 57.42 -29.80 Other 0.89 6.53 0.85 12.85 58.74 -29.99 Telecom 0.85 4.64 -0.01 3.06 28.16 -21.56 Utilities 0.88 5.61 0.07 7.72 43.16 -32.96
Wholesale and retail 0.98 5.84 -0.03 5.46 37.05 -30.15
Panel B: Industry ETFs Symbol Mean (%) St. dev. (%) Skewness Kurtosis Min (%) Max (%)
Consumer discretionary XLY 0.61 6.19 -0.40 1.34 -20.00 19.26
Consumer staples XLP 0.46 3.65 -0.83 1.39 -11.91 9.50 Energy XLE 0.99 7.00 -0.82 1.14 -19.99 19.36 Financials XLF 0.24 7.74 -1.23 5.43 -32.10 28.13 Industrials XLI 0.55 6.25 -0.69 1.85 -20.66 19.08 Information technology XLK 0.08 7.72 -0.44 0.95 -22.75 21.90 Health XLV 0.45 4.65 -0.77 2.46 -16.70 12.99 Materials XLB 0.74 6.96 -0.40 1.68 -21.46 26.10 Utilities XLU 0.53 4.40 -1.07 2.55 -15.75 12.57
14
IV.
Methodology
This section discusses the methodology for portfolio construction, introduces the models for risk-adjusted returns and describes the different trading costs applied.
4.1 Portfolio construction
The methodology for portfolio construction is based on Jegadeesh and Titman (1993). This methodology is based on a ranking method. Returns are calculated over the last J months (formation period) and the performance is evaluated over the next K months (holding period). Following existing literature, J and K contain values of 3, 6, 9 and 12. Consequently, in total 16 strategies are tested.
At the end of month, t, indexes (ETFs) are ranked in descending order on the basis of their returns in the past J months. Based on these rankings, a winner portfolio and a loser portfolio is created. The winner portfolio consists of the best performing indexes (ETFs). The loser portfolio consists of the worst performing indexes (ETFs). The winner-minus-loser (WML) portfolio is a combination of these two portfolios, which is a zero-investment portfolio. A long position is taken in the winner portfolio and short position is taken in the loser portfolio. The portfolios are equally weighted and subsequently held for the next K months. With a holding period of one month, a portfolio is terminated after one month and a new portfolio is constructed. However, when the holding period is longer than one month, the strategy has to wait K months before buying and selling new portfolios. Jegadeesh and Titman (1993) create overlapping portfolios to avoid the waiting period and to increase the power of the tests.
The procedure of overlapping portfolios is as follows. At the end of every month a new portfolio is constructed. This portfolio consists of a long position in the winner portfolio and a short position in the loser portfolio. The weight of a portfolio is 1/K due overlapping holding periods. Subsequently this portfolio is held for the next K months. The return of a strategy is then measured at the end of the month t+1. For example, the return of a strategy with J=3 and K=3 consists of three parts. A position carried over from the investment at the end of month 3, and two positions from a similar investment in month t-2 and t-1. At the end of the month the position from month t-3 is liquidated. This position is replaced by a new investment in portfolio month t.
15
of the bottom 1, 2 or 3 indexes (ETFs). Indexes and ETFs are equally weighted in a portfolio. The returns presented in this paper are from WML portfolios which are zero-investment strategies with a long position in the winner portfolio and short position in the loser portfolio. Figure 1 provides an graphical representation of the calculations for a momentum strategy with a formation and holding period of three months.
Portfolio 1 Formation Holding
Portfolio 2 Formation Holding
R etu rn Ju ly 2 0 0 5 Portfolio 3 Formation R etu rn A u g . 2 0 0 5 Portfolio 4 Formation R etu rn S ep 2 0 0 5 Portfolio 5 Formation
Portfolio 6 Formation Holding
Jan-05 Feb-05 Mar-05 Apr-05 May-05 Jun-05 Jul-05 Aug-05 Sep-05 Oct-05 Nov-05 Figure 1: Portfolio formation for a (3,3) momentum strategy
Using the calendar time procedure, time series of monthly returns are obtained from the momentum portfolios. To check whether the results are significantly different from zero, a t-test is performed.
=
/ √,
(2)
where is the t-statistic, is the average monthly simple return, is the standard deviation and is the degrees of freedom.
4.2 Risk-adjusted returns
To evaluate the risks in the returns of momentum strategies risk adjustments are made. Risk-adjusted returns are calculated using Fama-French’s three-factor model (Fama and French, 1993):
௧=+ − + ௧+ ℎ ௧+ ௧, (3)
where ௧ is the return of the WML portfolio i in month t, ௧ is the Treasury bill (risk-free) rate in
month t, is the return on a value-weighted market proxy (all CRSP firms incorporated in the US and
listed on the NYSE, AMEX, or NASDAQ that have a CRSP share code of 10 or 11 at the beginning of month t)5.
௧ is the difference between the returns on diversified portfolios of small stocks and big
stocks, and ௧ is the difference between the returns on diversified portfolios of high book-to-market
(value) stocks and low book-to-market (growth) stocks (Fama and French, 2012). ௧ is the regression
16
error term and is calculated for returns on WML portfolios to measure abnormal performance. Monthly
data is obtained from the website of Kenneth R. French6. No adjustments are made to the risk factor premiums because all ETFs are traded in dollars and during U.S. market hours (De Jong and Rhee, 2008). Zhong and Yang (2005) also state that returns of international ETFs are greatly influenced by U.S. risk factors.
4.3 Trading costs
In real-life an individual investor has to deal with several trading costs when implementing a trading strategy. Three types of trading costs are considered; the bid-ask spread, transaction costs and short selling costs.
The bid-ask spread is calculated following Korajczyk and Sadka (2004):
= ଵ − ଶ ( + )
, (4)
The daily spread is calculated for all ETFs in a sample and the average of this taken as estimate for bid-ask spread. De Jong and Rhee (2010) estimate that transaction costs are 0.13% by combining Scotts trade $7 per trade with in total an estimated account balance of $ 250,0007. Andreu et al. (2013) assumes that fixed trading costs are negligibly small and therefore they do not impose transaction costs. Following De Jong and Rhee (2010), I assume that transaction costs are 0.13% per trade. In addition, following De Jong and Rhee (2010) and Andreu et al. (2013), no short selling costs are taken into account.
6 http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
17
V.
Results
This section presents the empirical results for country and industry momentum strategies. Non-tradable indexes are analyzed to see if momentum is present in country and industry indexes. After analyzing indexes, real tradable funds (ETFs) are analyzed. This section starts with country momentum strategies. Secondly, industry momentum strategies are analyzed. Lastly, the returns are controlled for trading costs. The returns presented are from WML portfolios, which are zero-investment portfolios. A long position is taken in the winner portfolio and a short position is taken in the loser portfolio. The winner portfolio consists of top 1, 2 or 3 indexes (ETFs), equally weighted. The loser portfolio consists of the bottom 1, 2 or 3 indexes (ETFs), equally weighted. The results are presented in tables and are grouped based on different formation periods (3, 6, 9 and 12 months) and holding periods (3, 6, 9 and 12 months). A t-test is performed to see whether raw returns are significantly different from zero. Risk-adjusted returns are obtained by adjusting raw returns for Fama-French risk factors.
5.1 Country index momentum
Table 3 reports the raw returns, the risk-adjusted returns and their associated t-statistics for country index momentum strategies. Momentum strategies holding in the winner and loser portfolio two or three indexes, yield the highest risk-adjusted returns (Table 3, panel B and panel C). Within that, highly significant alpha is found for strategies with a holding and formation period of six or nine months. I find the highest monthly average return (0.73%) for a (6,9) strategy with the winner and loser portfolio consisting of two indexes. I find the highest risk-adjusted return (0.73%) for a (9,3) strategy holding two indexes in the winner and loser portfolio, which is statistically significant at 5 percent level.
18
Table 3: Returns from country index momentum strategies
This table presents the monthly raw and risk-adjusted returns for country momentum strategies using 16 country indexes in the period January 1971 to December 2012. The returns presented are from winner-minus-loser portfolios which are zero-investment strategies with a long position in the winner portfolio and short position in the loser portfolio. The winner portfolio consists of the top 1, 2 or 3 indexes (equally weighted). The loser portfolio consists of the bottom 1, 2 or 3 indexes (equally weighted). The first column shows the monthly raw return and the second column presents the risk-adjusted return based on the Fama-French three-factor model (alpha). Associated t-statistics are presented next to the returns. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Formation period (J)
Holding period (K)
3 6 9 12
Return *** Alpha *** Return *** Alpha *** Return *** Alpha *** Return *** Alpha ***
Panel A: Winner and loser portfolio holding one index
3 -0.03% -0.08 0.08% 0.22 0.26% 0.91 ** -0.13% -0.44 0.52% 2.03 ** 0.53% 2.02 ** 0.33% 1.44 * 0.29% 1.22 6 0.39% 0.98 0.41% 1.04 0.67% 1.93 ** 0.27% 0.75 0.67% 2.16 ** 0.65% 2.02 ** 0.48% 1.67 ** 0.42% 1.45 9 0.41% 1.02 0.45% 1.10 0.62% 1.68 0.18% 0.48 0.56% 1.64 ** 0.56% 1.60 0.46% 1.47 * 0.49% 1.53 12 0.52% 1.32 * 0.52% 1.29 0.58% 1.60 * 0.15% 0.39 0.63% 1.89 ** 0.62% 1.83 * 0.46% 1.45 * 0.47% 1.46 Panel B: Winner and loser portfolio holding two indexes
3 0.19% 0.68 0.33% 1.17 0.28% 1.23 0.35% 1.54 0.47% 2.26 ** 0.34% 1.46 0.38% 1.96 ** 0.39% 2.01 ** 6 0.49% 1.65 ** 0.54% 1.78 * 0.69% 2.54 *** 0.72% 2.59 ** 0.69% 2.74 *** 0.69% 2.48 *** 0.50% 2.18 ** 0.49% 2.05 ** 9 0.70% 2.25 ** 0.73% 2.29 ** 0.73% 2.53 *** 0.71% 2.41 ** 0.67% 2.47 *** 0.67% 2.24 ** 0.50% 1.97 ** 0.51% 1.96 * 12 0.60% 1.89 ** 0.63% 1.93 * 0.58% 1.99 ** 0.62% 2.07 ** 0.56% 2.01 ** 0.58% 1.93 * 0.38% 1.44 * 0.45% 1.65 * Panel C: Winner and loser portfolio holding three indexes
3 0.13% 0.59 * 0.28% 1.23 0.28% 1.47 ** 0.35% 1.54 0.43% 2.48 *** 0.34% 1.47 0.36% 2.23 ** 0.39% 2.32 ** 6 0.34% 1.43 0.46% 1.90 * 0.58% 2.55 *** 0.72% 2.59 ** 0.58% 2.72 *** 0.69% 2.48 ** 0.42% 2.18 ** 0.43% 2.19 ** 9 0.46% 1.84 ** 0.57% 2.22 ** 0.61% 2.49 *** 0.71% 2.41 ** 0.53% 2.30 ** 0.67% 2.24 ** 0.40% 1.87 ** 0.45% 2.02 ** 12 0.42% 1.58 ** 0.51% 1.88 * 0.45% 1.79 ** 0.18% 0.61 0.41% 1.69 ** 0.14% 0.47 0.28% 1.23 0.37% 1.57
In general momentum strategies consisting of four (two in the winner portfolio and two in the loser portfolio) or six country indexes and have a formation period of six to nine months perform best. The returns achieved cannot be explained by the Fama-French three-factor model; therefore, it can be concluded that country indexes exhibit momentum. The next step is to see whether these significant returns can also be obtained by real tradable assets (ETFs).
5.2 Country ETF momentum
Indexes are not real tradable assets. If an investor wants to mimic an index it should buy all stocks in the index separately. This is cumbersome and involves high transaction costs. Instead an investor could buy an investment fund (ETF) tracking this index. So can the significant risk-adjusted returns observed in country indexes also observed across country ETFs?
19
Table 4: Returns from country ETF momentum strategies
This table presents the monthly raw and risk-adjusted returns for country ETF momentum strategies using 16 country ETFs in the period January 2001 to December 2012.The returns presented are from winner-minus-loser portfolios which are zero-investment strategies with a long position in the winner portfolio and short position in the loser portfolio. The winner portfolio consists of top 1, 2 or 3 ETFs (equally weighted). The loser portfolio consists of the bottom 1, 2 or 3 ETFs (equally weighted). The first column shows the monthly raw return and the second column presents the risk-adjusted return based on the Fama-French three-factor model (alpha). Associated t-statistics are presented next to the returns. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Formation period (J)
Holding period (K)
3 6 9 12
Return ** Alpha ** Return ** Alpha ** Return *** Alpha ** Return ** Alpha **
Panel A: Winner and loser portfolio holding one ETF
3 -0.31% -0.78 -0.38% -0.98 -0.13% -0.43 -0.14% -0.45 -0.07% -0.30 -0.10% -0.41 -0.09% -0.43 -0.16% -0.75 6 -0.05% -0.12 -0.21% -0.44 -0.04% -0.10 -0.12% -0.31 0.12% 0.34 0.02% 0.07 0.14% 0.46 0.04% 0.12 9 -0.38% -0.72 -0.58% -1.14 -0.06% -0.13 -0.16% -0.38 0.30% 0.76 0.21% 0.53 0.38% 1.10 0.26% 0.78 12 -0.48% -0.92 -0.59% -1.15 0.11% 0.24 0.00% 0.00 0.40% 1.00 0.29% 0.72 0.53% 1.57* 0.40% 1.18 Panel B: Winner and loser portfolio holding two ETFs
3 -0.33% -1.04 -0.43% -1.36 -0.26% -1.12 -0.32% -1.37 -0.12% -0.59 -0.16% -0.84 -0.06% -0.37 -0.13% -0.75 6 -0.18% -0.54 -0.28% -0.84 -0.05% -0.16 -0.09% -0.31 0.12% 0.45 0.05% 0.19 0.15% 0.69 0.08% 0.36 9 -0.07% -0.18 -0.20% -0.55 0.04% 0.13 -0.06% -0.17 0.26% 0.86 0.16% 0.54 0.34% 1.26 0.24% 0.90 12 -0.25% -0.67 -0.41% -1.10 0.19% 0.54 0.10% 0.30 0.42% 1.35* 0.34% 1.11 0.40% 1.47*** 0.32% 1.19 Panel C: Winner and loser portfolio holding three ETFs
3 -0.13% -0.49 -0.21% -0.79 -0.12% -0.59 -0.14% -0.45 -0.03% -0.19 0.02% 0.10 -0.01% -0.08 -0.05% -0.35 6 -0.25% -0.88 -0.31% -1.10 -0.10% -0.41 -0.12% -0.31 -0.01% -0.03 0.03% 0.15 0.03% 0.14 -0.02% -0.12 9 0.02% 0.07 -0.08% -0.26 0.07% 0.24 -0.16% -0.38 0.21% 0.84 0.25% 1.00 0.25% 1.15 0.18% 0.83 12 -0.10% -0.32 -0.21% -0.70 0.20% 0.69 0.00% 0.00 0.37% 1.46* 0.44% 1.73 0.34% 1.52* 0.27% 1.23
20
Table 5: Returns from country ETF and index momentum strategies
This table presents the raw returns for country ETF and index momentum strategies. The country index sample consists of 16 country indexes and the country ETF consists of 16 ETFs. Both sample periods range sample from January 2001 to December 2012. The returns presented are from winner-minus-loser portfolios which are zero-investment strategies with a long position in the winner portfolio and short position in the loser portfolio. The winner portfolio consists of top 3 ETFs (indexes) (equally weighted). The loser portfolio consists of the bottom3 ETFs (indexes) (equally weighted). The first column shows the monthly raw return of country ETF strategies and the second column shows the shows the monthly raw return of country index strategies. The third column presents t-statistic of the difference between the returns.
Formation period (J)
Holding period (K)
3 6 9 12
ETF Index t-stat *** ETF Index t-stat ** ETF Index t-stat ** ETF Index t-stat 3 -0.13% -0.30% 0.43 -0.12% -0.09% -0.09 -0.03% -0.11% 0.13 -0.01% -0.01% 0.00 6 -0.25% -0.41% 0.46 -0.10% -0.18% 0.24 -0.01% -0.12% 0.19 0.03% 0.01% 0.06 9 0.02% -0.25% 0.67 0.07% -0.05% 0.33 0.21% 0.00% 0.36 0.25% 0.14% 0.38 12 -0.10% -0.28% 0.47 0.20% 0.11% 0.37 0.37% 0.17% 0.50 0.34% 0.23% 0.35
5.3 Industry index momentum
Now country strategies have been discussed, this subsection discusses industry momentum strategies for indexes, the subsection hereafter discusses industry ETFs. Table 6 presents returns for industry index momentum strategies. In almost all combinations significant risk-adjusted returns are found. I find highly significant alpha (0.80%) for the (12,3) strategy with holding one index in the winner portfolio and loser portfolio. The results are robust across different combinations of formation and holding periods. Only for strategies with a formation period of three months and holding period for three to six months no statistically significant returns are found. Whereas country index strategies perform best holding two or three indexes in the winner and loser portfolio, industry index strategies perform best holding one index in the winner and loser portfolio.
21
Table 6: Returns from industry index momentum strategies
This table presents the monthly raw and risk-adjusted returns for industry momentum strategies using 10 industry indexes in the period from January 1928 to December 2012. The returns presented are from winner-minus-loser portfolios which are zero-investment strategies with a long position in the winner portfolio and short position in the loser portfolio. The winner portfolio consists of top 1,2 or 3 indexes (equally weighted). The loser portfolio consists of the bottom 1,2 or 3 indexes (equally weighted). The first column shows the monthly raw return and the second column presents the risk-adjusted return based on the Fama-French three-factor model (alpha). Associated t-statistics are presented next to the returns. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Formation period (J)
Holding period (K)
3 6 9 12
Return *** Alpha *** Return *** Alpha *** Return *** Alpha *** Return *** Alpha ***
Panel A: Winner and loser portfolio holding one index
3 0.19% 1.23 0.32% 2.16 ** 0.15% 1.11 0.10% 0.72 0.18% 1.55 * 0.33% 2.87 *** 0.20% 1.92 ** 0.35% 3.49 *** 6 0.37% 2.07 ** 0.53% 2.98 *** 0.42% 2.60 *** 0.35% 2.13 ** 0.40% 2.73 *** 0.55% 3.79 *** 0.28% 2.11 ** 0.44% 3.35 *** 9 0.54% 2.86 *** 0.69% 3.64 *** 0.52% 2.98 *** 0.46% 2.59 *** 0.39% 2.38 *** 0.55% 3.48 *** 0.23% 1.50 * 0.39% 2.64 *** 12 0.60% 3.17 *** 0.80% 4.29 *** 0.45% 2.54 *** 0.40% 2.23 ** 0.24% 1.40 * 0.44% 2.68 *** 0.11% 0.70 0.30% 1.99 ** Panel B: Winner and loser portfolio holding two indexes
3 0.17% 1.43 * 0.29% 2.59 *** 0.16% 1.56 * 0.13% 1.31 0.19% 2.13 ** 0.30% 3.50 *** 0.20% 2.56 *** 0.31% 4.16 *** 6 0.28% 2.07 ** 0.42% 3.20 *** 0.32% 2.62 *** 0.31% 2.47 ** 0.31% 2.81 *** 0.42% 3.94 *** 0.23% 2.26 ** 0.35% 3.60 *** 9 0.43% 3.07 *** 0.54% 3.91 *** 0.42% 3.29 *** 0.41% 3.18 *** 0.33% 2.76 *** 0.46% 3.98 *** 0.20% 1.81 ** 0.33% 3.08 *** 12 0.48% 3.38 *** 0.65% 4.71 *** 0.34% 2.55 *** 0.32% 2.32 ** 0.22% 1.73 ** 0.39% 3.21 *** 0.09% 0.80 0.26% 2.32 ** Panel C: Winner and loser portfolio holding three indexes
3 0.19% 1.92 ** 0.31% 3.24 *** 0.15% 1.80 ** 0.13% 1.31 0.16% 2.17 ** 0.26% 3.65 *** 0.16% 2.50 *** 0.26% 4.22 *** 6 0.23% 2.08 ** 0.37% 3.41 *** 0.25% 2.41 *** 0.31% 2.47 ** 0.26% 2.87 *** 0.37% 4.10 *** 0.20% 2.41 *** 0.31% 3.86 *** 9 0.33% 2.79 *** 0.43% 3.63 *** 0.30% 2.79 *** 0.41% 3.18 *** 0.23% 2.34 ** 0.34% 3.60 *** 0.14% 1.49 * 0.25% 2.90 *** 12 0.36% 3.05 *** 0.52% 4.61 *** 0.26% 2.37 *** 0.32% 2.32 ** 0.17% 1.58 * 0.32% 3.25 *** 0.06% 0.64 0.22% 2.36 **
5.4 Industry ETF momentum
22
Table 7: Returns from industry ETF momentum strategies
This table presents the monthly raw and risk-adjusted returns for industry ETF momentum strategies using 9 industry ETFs in the period from January 2001 to December 2012. The returns presented are from winner-minus-loser portfolios which are zero-investment strategies with a long position in the winner portfolio and short position in the loser portfolio. The winner portfolio consists of top 1, 2 or 3 ETFs (equally weighted). The loser portfolio consists of the bottom 1, 2 or3 ETFs (equally weighted). The first column shows the monthly raw return and the second column presents the risk-adjusted return based on the Fama-French three-factor model (alpha). t-statistics are presented below returns. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Formation period (J)
Holding period (K)
3 6 9 12
Return *** Alpha ** Return *** Alpha ** Return *** Alpha ** Return *** Alpha **
Panel A: Winner and loser portfolio holding one ETF
3 -0.24% -0.48 -0.29% -0.62 -0.05% -0.12 0.01% 0.03 0.09% 0.22 0.04% 0.09 0.20% 0.53 0.30% 0.84 6 0.05% 0.08 0.02% 0.04 0.41% 0.74 0.45% 0.86 0.41% 0.75 0.33% 0.60 0.40% 0.74 0.51% 1.00 9 0.48% 0.75 0.45% 0.76 0.43% 0.67 0.50% 0.82 0.34% 0.55 0.28% 0.45 0.29% 0.50 0.40% 0.70 12 0.42% 0.65 0.35% 0.58 0.46% 0.71 0.52% 0.84 0.39% 0.63 0.33% 0.51 0.41% 0.69 0.50% 0.88 Panel B: Winner and loser portfolio holding two ETFs
3 -0.15% -0.41 -0.12% -0.33 -0.06% -0.18 0.01% 0.05 0.06% 0.19 0.11% 0.37 0.13% 0.46 0.20% 0.72 6 0.07% 0.16 0.07% 0.18 0.14% 0.35 0.19% 0.51 0.14% 0.38 0.19% 0.51 0.18% 0.51 0.26% 0.75 9 0.16% 0.37 0.17% 0.43 0.13% 0.32 0.20% 0.49 0.19% 0.45 0.25% 0.62 0.12% 0.32 0.21% 0.57 12 0.15% 0.34 0.14% 0.34 0.19% 0.44 0.26% 0.61 0.20% 0.47 0.27% 0.64 0.09% 0.22 0.16% 0.42 Panel C: Winner and loser portfolio holding three ETFs
3 -0.05% -0.19 -0.04% -0.16 0.00% -0.01 0.03% 0.11 0.05% 0.23 0.07% 0.31 0.09% 0.45 0.13% 0.62 6 -0.01% -0.04 0.00% 0.00 0.03% 0.09 0.08% 0.27 0.06% 0.19 0.09% 0.32 0.10% 0.37 0.16% 0.58 9 0.11% 0.32 0.14% 0.43 0.07% 0.22 0.12% 0.39 0.13% 0.40 0.17% 0.53 0.02% 0.06 0.06% 0.23 12 -0.06% -0.17 -0.07% -0.19 -0.02% -0.05 0.03% 0.08 0.02% 0.06 0.06% 0.19 -0.07% -0.24 -0.04% -0.13
Table 8 presents the results of industry index momentum strategies in the period 2001-2009. As for ETF momentum strategies, no significant alpha is found for index strategies. As mentioned in section 3.3, the industry ETFs track a different index than the indexes analyzed in the industry index sample; therefore, for some combinations the results deviate a bit. It can be concluded that no industry momentum is present in the period 2001-2012.
Table 8: Returns from industry index momentum strategies
This table presents the monthly raw and risk-adjusted returns for industry momentum strategies using 10 industry indexes in the period from January 2001 to December 2012. The returns presented are from winner-minus-loser portfolios which are zero-investment strategies with a long position in the winner portfolio and short position in the loser portfolio. The winner portfolio consists of top 1 index (equally weighted). The loser portfolio consists of the bottom 1 index (equally weighted). The first column shows the monthly raw return and the second column presents the risk-adjusted return based on the Fama-French three-factor model (alpha). t-statistics are presented next to the returns. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Formation period (J)
Holding period (K)
3 6 9 12
Return Alpha ** Return Alpha ** Return Alpha ** Return Alpha **
3 -0.14% -0.30 -0.24% -0.58 0.20% 0.50 0.24% 0.65 0.23% 0.57 0.31% 0.84 0.19% 0.54 0.26% 0.77 6 0.72% 1.11 0.68% 1.17 0.64% 1.09 0.75% 1.35 0.46% 0.77 0.58% 1.04 0.25% 0.44 0.35% 0.64 9 0.58% 0.87 0.54% 0.90 0.40% 0.63 0.46% 0.78 0.20% 0.32 0.30% 0.51 -0.14% -0.24 -0.08% -0.14 12 0.44% 0.69 0.45% 0.77 0.31% 0.50 0.40% 0.67 0.01% 0.01 0.10% 0.17 -0.19% -0.33 -0.12% -0.21
23
momentum strategy deteriorates. The ETF in the loser portfolio (short position) recovers more than the ETF in the winner portfolio (long position); therefore, the WML portfolio performs poorly in the period thereafter. Interestingly, this process continues until the last month of the sample.
Figure 2: The performance of loser, winner and WML portfolio for a (6,6) momentum strategy
24
Table 9: Risk-adjusted returns for the (6,6) industry ETF momentum strategy
This table presents the risk-adjusted returns for the (6,6) industry ETF momentum strategy, holding one ETF in the winner and loser portfolio. The sample is divided into two periods; January 2001 to February 2009 and March 2009 to December 2012. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Alpha t-value Market t-value SMB t-value HML t-value Jan. 2001 – March 2009 Winner 0.73% 1.83 * 0.82 8.94 *** 0.09 0.61 0.26 1.73* Loser -0.02% -0.03 1.50 12.44 *** -0.05 -0.27 0.28 1.38* WML 0.74% 1.05 -0.68 -4.23 *** 0.15 0.55 -0.02 -0.08* April 2009 – Dec. 2012 Winner -0.17% -0.35 0.91 7.44 *** -0.11 -0.52 0.08 0.39* Loser 0.52% 0.99 1.18 9.10 *** -0.40 -1.73 -0.01 -0.04* WML -0.69% 0.32 -0.27 -1.55 0.29 0.93 0.09 0.30*
5.5 Returns net of trading costs
For country and industry ETFs, I do not find significant alpha. This could be caused by the short sample period. But what if it is assumed that the ETFs are able to generate the same abnormal returns as the indexes analyzed above, in the long run? Is an individual investor then able to implement such momentum strategies profitable? This subsection analyzes the most significant returns of country and industry index samples and adjusts them for trading costs.
Following De Jong and Rhee (2008), alphas are first annualized by multiplying the monthly risk-adjusted return by 12. Then, the annual trading costs incurred by a strategy are calculated and subtracted from the annualized return, this results in the net annualized return.
25
The total trading costs a strategy incurs per year are calculated as follows. Every month the portfolio is rebalanced for 1/K part of the portfolio. This number is multiplied by the number of ETFs that winner and loser portfolio hold. Lastly, this is multiplied by 12 months and by two since every month ETFs are sold (bought back) and new one are bought (sold short). For example, a (6,6) industry momentum strategy holding two industry ETFs in the winner and loser portfolio trades in total 16 ETFs per year. Every month 1/6 of the portfolio is rebalanced, since the winner and loser portfolio hold both two ETFs this is multiplied by four. In total this strategy trades 16 (= 1/6 * 4 * 12 * 2) times a year. Table 10 presents the net annualized returns for country index momentum strategies (two indexes in winner and loser portfolio) and industry index momentum strategies (one index in the winner and loser portfolio). The trading costs for country strategies are higher than for industry strategies. This is due to a combination of higher bid-ask spread and to the fact that the country strategies consists of a double number of indexes. The best performing strategy is the (6,9) strategy for countries with a net annualized abnormal return of 4.77 % per year. The (12,3) strategy is the best performing strategy for industries, returning a net annualized abnormal return of 5.59% per year. The results are comparable with De Jong and Rhee (2008) who find a net annualized return of 4.58% with a formation and holding period of six months using a sample of all ETFs available. It can be concluded, that if in the long run the risk-adjusted returns of indexes can be obtained by ETFs, then it is possible for the individual investor to exploit the momentum effect profitably.
Table 10: Annualized risk-adjusted returns net of trading costs
This table presents annualized risk-adjusted returns net of trading costs for the most profitable momentum strategies. The first column presents annualized risk-adjusted returns. The second column presents the trading costs (TC) associated with the momentums strategies. The third column presents the net annualized risk-adjusted returns. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Formation period (J)
Holding period (K)
3 6 9 12
Return TC Net return *** Return TC Net return Return TC Net return Return TC Net return Panel A: Country index momentum strategies holding two indexes in the winner and loser portfolio
3 3.93% 10.56% -6.63% 4.25% 5.28% -1.03% 4.06% 3.52% 0.54% 4.74% ** 2.64% 2.10% 6 6.47% * 10.56% -4.09% 8.59% ** 5.28% 3.31% 8.29% *** 3.52% 4.77% 5.83% ** 2.64% 3.19% 9 8.76% ** 10.56% -1.80% 8.56% ** 5.28% 3.28% 8.03% ** 3.52% 4.51% 6.09% * 2.64% 3.45% 12 7.57% * 10.56% -2.99% 7.43% ** 5.28% 2.15% 6.97% * 3.52% 3.45% 5.37% * 2.64% 2.73% Panel A: Industry index momentum strategies holding one index in the winner and loser portfolio
26
VI.
Conclusions
This paper investigated whether or not is possible for the individual investor to exploit the momentum anomaly using country and industry Exchange Traded Funds (ETFs). First, the presence of momentum in countries and industries was examined. Second, it was analyzed if momentum is present in country and industry ETFs.
Momentum country index strategies earn significant risk-adjusted returns over the period January 1971 - December 2012. The best performing country momentum strategies are strategies holding two or three indexes in the winner and loser portfolio. These strategies generate risk-adjusted returns of about 0.70% per month. The results are robust for different holding and formation periods and are consistent with Bhojraj and Swaminathan (2006). Momentum is also present in industries during the period January 1929 - December 2012. Industry index momentum strategies that hold one index in the winner and loser portfolio perform best. The strategies yield about 0.45% alpha per month. This is consistent with the findings of Moskowitz and Grinblatt (1999) and Giannikos and Ji (2007).
The significant results observed in country and industry indexes cannot be observed across country and industry ETFs. I do not find statistically significant risk-adjusted returns for country and industry ETFs during the period January 2001 - December 2012. The index samples are, in addition, analyzed for this period but also for these samples no significant results are found. Therefore, it can be concluded that this is not an issue of ETFs but that this depends on the sample period. The finding of insignificant results are in line with Andreu et al. (2013), who also do not find significant returns for country and industry ETFs. So why is there no momentum observed during the sample period 2001-2012? In 2008, the stock market crashed. After this market crash ending in March 2009, all momentum strategies started to deteriorate in performance. This is caused by a combination of a large negative beta of the winner-minus-loser portfolio and a sharp recovery of the market. This implies that momentum is not always present in every time period and steady momentum profits could be interrupted by large momentum reversals.
27
The results of the momentum strategies imply that past performance of securities can be used to predict future performance. Therefore, the results I find reject the weak-form of efficient market hypothesis. Furthermore, the trading costs are lower than the risk-adjusted returns which implies that the momentum returns cannot be explained by trading costs. However, a limitation here is that it is assumed that momentum ETF strategies will achieve the same returns as momentum index strategies in the long run. Furthermore, the observed evidence of momentum does support the theory of underreaction. Which implies that security prices underreact to news and react slowly to new public information resulting in consistent price momentum (Doukas and McKnight, 2005).
As mentioned before, the presence of momentum is very robust across industry and country indexes. However, a limitation of this study is that the obtained results are based on U.S. dollar returns. For a Dutch investor holding a bank account in euros, the returns should be converted to euros which could lead to different returns.
28
Appendix A
Figure A.1 : 20-day moving average of quoted spread industry ETFs
29
References
Agarwal, V., Wang, L., 2007. Transaction costs and value premium. Unpublished working paper. Georgia State University, Atlanta.
Andreu, L., Swinkels, L., Tjong-A-Tjoe, L., 2013. Can exchange traded funds be used to exploit industry momentum? Financial Markets and Portfolio Management 27, 127-148.
Asness, C., Moskowitz, T., Pedersen, L., 2009. Value and momentum everywhere. Unpublished working paper. University of Chicago, Chicago.
Barberis, N., Shleifer, A., Vishny, R., 1998. A model of investor sentiment. Journal of Financial Economics 49, 307-343.
Bhojraj, S., Swaminathan, B., 2006. Macromomentum: returns predictability in international equity indexes. The Journal of Business 79, 429-451.
Carhart, M., 1997. On persistence in mutual fund performance. Journal of finance 52, 57-82.
Corwin, S., Schultz, P., 2011. A simple way to estimate bid‐ask spreads from daily high and low prices. Unpublished working paper. University of Notre Dame, Notre Dame.
Daniel, K., Hirshleifer, D., Subrahmanyam, A., 1998. Investor psychology and security market under‐and overreactions. Journal of Finance 53, 1839-1885.
Daniel, K., Moskowitz, T., 2013. Momentum crashes. Unpublished working paper. Columbia Business School, New York.
Doukas, J., Mcknight, P., 2005. European momentum strategies, information diffusion, and investor conservatism. European Financial Management 11, 2005.
De Jong, J., Rhee, S., 2008. Abnormal returns with momentum/contrarian strategies using exchange-traded funds. Journal of Asset Management 9, 289-299.
Fama, E., French, K., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3-56.
Fama, E., French, K., 1996. Multifactor explanations of asset pricing anomalies, Journal of Finance 51, 55-84.
30
Fong, W., Wong, W., Lean, H., 2005. International momentum strategies: a stochastic dominance approach. Journal of Financial Markets 8, 89-109.
Giannikos C., Ji, X., 2007. Industry momentum at the end of the 20th century. International Journal of Business and Economics 6, 29-46.
Grinblatt, M., Han, B., 2005. Prospect theory, mental accounting, and momentum. Journal of Financial Economics 78, 311-339.
Grundy, B., Martin, J., 2001. Understanding the nature of the risks and the source of the rewards to momentum investing. Review of Financial Studies 14, 29-78.
Jegadeesh, N., Titman, S., 1991. Seasonality in stock price mean reversion: evidence from the U.S. and the U.K. Journal of Finance 46, 1427–1444.
Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers: implications for stock market efficiency. Journal of Finance 48, 65-91.
Jegadeesh, N., Titman, S., 2001. Profitability of momentum strategies: an evaluation of alternative explanations. Journal of Finance 56, 699-720.
Kang, J., Liu, M., Ni, S., 2002. Contrarian and momentum strategies in the China stock market: 1993– 2000. Pacific-Basin Finance Journal 10, 243-265.
Korajczyk, R., Sadka, R., 2004. Are momentum profits robust to trading costs? Journal of Finance 59, 1039-1082.
Lesmond, D., Ogden, J., Trzcinka, C., 1999. A new estimate of transaction costs. Review of Financial Studies 12, 1113-1141.
Lesmond, D., Schill, M., Zhou, C., 2004. The illusory nature of momentum profits. Journal of Financial Economics 71, 349-380.
Maginn, J., Tuttle, D., McLeavey, D., Pinto, J., 2007. Managing investment portfolios: a dynamic process. John Wiley & Sons Inc., Hoboken.
Moskowitz, T., Grinblatt, M., 1999. Do industries explain momentum? Journal of Finance 54, 1249-1290. Muga, L., Santamaría, R., 2009. Momentum, market states and investor behavior. Empirical Economics 37, 105-130.
31
Rouwenhorst, K., 1998. International Momentum Strategies. Journal of Finance 53, 267-284.
Richards, A., 1997. Winner-loser reversals in national stock market indexes: can they be explained? Journal of Finance 52, 2129-2144.
Salomon, R., Tarasow, D., 2000. The virtues of ETFs. Fobes 166, 206.
