Three essays on exchange traded funds

... international and country funds, and bond funds, I contribute to the literature by demonstrating momentum/contrarianism in a changing asset allocation setting which includes U.S stocks, U.S bonds,... & Lakonishok (1996), Hong & Stein (1999)), expectation extrapolation, conservatism in expectation updating (Barberis, Shleifer, Vishny (1998)), biased self attribution or investor overconfidence... MOMENTUM/CONTRARIAN ABNORMAL RETURNS AND EXCHANGE TRADED FUNDS Abstract: Investing in portfolios of exchange traded funds (ETFs) provides risk adjusted abnormal returns that exceed transactions costs

THREE ESSAYS ON EXCHANGE TRADED FUNDS A DISSERTATION SUBMITTED TO THE GRADUATE DIVISION OF THE UNIVERSITY OF HA WAIT IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN INTERNATIONAL MANAGEMENT August 2007 By Jack C. De Jong Jr. Dissertation Committee: S. Ghon Rhee, Chairperson Rosita P. Chang (Victor) Wei Huang Qianqiu Liu Sang Hyop Lee Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI N um ber: 3288099 INFORMATION TO USERS The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed-through, substandard margins, and im proper alignm ent can adversely affect reproduction. In the unlikely event that the author did not send a complete m anuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. ® UMI UMI Microform 3288099 Copyright 2008 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ABSTRACT Essay #1 shows that investing in portfolios of U.S. exchange traded funds (ETFs) provides risk adjusted abnormal returns that exceed transactions costs. Short formation and holding periods of one day to one week provide abnormal contrarian returns as past losers become winners and past winners become losers. Medium formation and holding periods of four weeks to thirty nine weeks provide abnormal momentum returns as past winners keep winning and past losers keep losing. Abnormal returns result for portfolios of ETFs when returns are adjusted for risk using the Capital Asset Pricing Model and Fama & French’s three factor model. Essay #2 shows that investing in portfolios of international ETFs provides risk adjusted abnormal returns using either Lo & MacKinlay’s or Jegadeesh & Titman’s weighting methodologies. A short formation and holding period of one week provides abnormal contrarian returns, while medium formation and holding periods of four weeks to twenty six weeks provide abnormal momentum returns. Abnormal returns result for portfolios o f international ETFs when returns are adjusted for risk using the Capital Asset Pricing Model and Fama & French’s three factor model. High trading volume increases the momentum abnormal returns for formation and holding periods of 4 weeks and 26 weeks; low trading volume increases the contrarian abnormal returns for a formation and holding period of 1 week. Essay #3 shows that investing in portfolios of non-U.S. ETFs from Australia, Canada, France, Germany, Hong Kong, Japan, and the U.K. provides risk adjusted abnormal returns. Short formation and holding periods o f 1 day to 8 weeks provide iii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. abnormal contrarian returns, while longer formation and holding periods of 16 weeks to 52 weeks provide abnormal momentum returns. Abnormal contrarian returns result for portfolios of non-U.S. ETFs when returns are adjusted for risk using international versions of both the Capital Asset Pricing Model and Fama & French’s three factor model; however, abnormal momentum returns result only when returns are adjusted for risk using an international version o f CAPM. iv Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. TABLE OF CONTENTS Abstract List of Tables Essay #1: Essay #2: Essay #3: Introduction Methodology Description of Data Outline of Model Results Robustness of Results Conclusion References Tables Introduction Methodology Description of Data Outline of Model Results Robustness of Results Conclusion References Tables Introduction Methodology Description o f Data Outline of Model Results Robustness of Results Conclusion References Tables Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. iii vi 1 9 12 12 13 20 29 31 34 52 59 64 65 66 76 83 86 90 117 124 126 127 128 136 149 151 154 LIST OF TABLES Table 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1.10 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 2.11 2.12 2.13 2.14 3.1 3.2 3.3 Growth of Exchange Traded Funds Annualized Raw Returns Annualized Abnormal Returns Abnormal Returns Using CAPM Abnormal Returns Using Fama & French’s Three Factor Model Annualized Abnormal Returns with Extra Time between Formation and Holding Periods Annualized Abnormal Returns Net o f Transactions Costs for WML with 26 Week Formation and Holding Period Abnormal Returns Divided Into Two Periods: 03/21/96-12/31/00 vs. 01/01/01 - 12/31/05 Abnormal Returns Using CAPM and Fama & French’s Three Factor Model Abnormal Returns Using 26 week Formation and Holding Periods Annualized Raw Returns: All International ETFs Annualized Abnormal Returns: All International ETFs Abnormal Returns Using CAPM: All International ETFs Abnormal Returns Using Fama & French’s Three Factor Model: All International ETFs Annualized Raw Returns: All International ETFs by High and Low Trading Volume Annualized Abnormal Returns Using CAPM: All International ETFs by High and Low Trading Volume Annualized Abnormal Returns Using Fama & French’s Three Factor Model: All International ETFs by High and Low Trading Volume Annualized Abnormal Returns with Extra Time between Formation and Holding: All International ETFs Abnormal Returns Using Different Betas in Up and Down Markets Abnormal Returns Divided Into Two Periods: 03/21/96 - 12/31/00 vs. 01/01/01 - 12/31/05 Annualized Raw Returns: Chan, Hameed, & Tong’s Country ETFs Annualized Abnormal Returns: Chan, Hameed, & Tong’s Country ETFs Abnormal Returns Using CAPM: Chan, Hameed, & Tong’s Country ETFs Abnormal Returns Using Fama & French’s Three Factor Model: Chan, Hameed, & Tong’s Country ETFs Growth of Non-U.S. Exchange Traded Funds Annualized Raw Returns Annualized Abnormal Returns vi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Page 34 35 36 37 39 41 43 44 46 49 90 91 93 95 97 99 101 103 105 108 110 111 113 115 154 155 156 LIST OF TABLES Table 3.4 3.5 3.6 3.7 3.8 3.9 3.10 Abnormal Returns Using International Version of CAPM Abnormal Returns Using International Version of Fama & French’s Three Factor Model Annualized Abnormal Returns with Extra Time between Formation and Holding Periods Abnormal Returns Divided Into Two Periods: 01/01/00 - 12/31/03 vs. 01/01/04 - 03/31/07 Abnormal Returns Using International Versions of CAPM and Fama & French’s Three Factor Model and Annual Dummy Variables Abnormal Returns Using International Version o f CAPM and Country Dummy Variables Abnormal Returns Using International Version o f Fama & French’s Three Factor Model and Country Dummy Variables vii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Pace 158 160 162 164 166 168 172 ESSAY #1 MOMENTUM/CONTRARIAN ABNORMAL RETURNS AND EXCHANGE TRADED FUNDS Abstract: Investing in portfolios of exchange traded funds (ETFs) provides risk adjusted abnormal returns that exceed transactions costs. Short formation and holding periods o f one day to one week provide abnormal contrarian returns as past losers become winners and past winners become losers. Medium formation and holding periods of four weeks to thirty nine weeks provide abnormal momentum returns as past winners keep winning and past losers keep losing. Abnormal returns result for portfolios of ETFs when returns are adjusted for risk using the Capital Asset Pricing Model and Fama & French’s three factor model. 1. Introduction: Previous studies show stock prices exhibit medium term momentum, as buying portfolios o f recent winners and shorting portfolios of recent losers results in abnormal returns that may or may not exceed transactions costs. One strand of the literature investigates whether the momentum anomaly exists in various markets; various studies find a momentum strategy generates abnormal returns in U.S. equities (Jegadeesh & Titman (1993,2001), Hong, Lim, & Stein (2000)), in U.S. mutual funds (Grinblatt, Titman, & Wermers (1995), Carhart (1997), Wermers (2003), Sapp & Tiwari (2004)), in U.S. industries (Moskowitz & Grinblatt (1999)), in international equity markets (Rouwenhorst (1998), Chan, Hameed, & Tong (2000), Balvers & Wu (2006)), and in foreign exchange markets (Okunev & White (2001)). Gebhardt, Hvidkjaer, & Swaminathan (2002) find no evidence of momentum among investment grade corporate bonds, but find contrarian returns as the bonds experienced significant reversals as well as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. a momentum spillover from stocks to bonds o f the same firm as past equity returns are good predictors of future bond rating changes. Momentum/contrarianism is an anomaly in the sense that it violates market efficiency in its weak form as investors can use technical analysis o f past prices and returns to form portfolios of winners and losers and so earn arbitrage profits on a zero investment portfolio that buys the winners and shorts the losers or buys the losers and shorts the winners, respectively. Momentum is also an anomaly in the sense that the standard empirical asset pricing model using Fama & French’s three factor model cannot explain the medium term return continuation of momentum. Another strand of the literature accepts the momentum anomaly as a stylized fact and seeks to explain what causes momentum using behavioral finance theory as rational, market efficiency has difficulty explaining its existence. Behavioral finance posits various explanations o f momentum, including investor overreaction or underreaction (Chan, Jegadeesh, & Lakonishok (1996), Hong & Stein (1999)), expectation extrapolation, conservatism in expectation updating (Barberis, Shleifer, Vishny (1998)), biased self attribution or investor overconfidence (Daniel, Hirshleifer, & Subrahmanyam (1998)), disposition effects (Grinblatt & Moskowitz (2004)), selective information conditioning, and herding behavior by investors (Jordan (2004)) and mutual fund managers. Clearly, as anomalies, momentum and contrarianism require further research, as current studies agree that a momentum/contrarian investment strategy generates abnormal returns, but disagree as to what causes a momentum/contrarian strategy to be successful and disagree as to whether the abnormal profits are realistically attainable by 2 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. an actual investor or arbitrageur after recognizing appropriate transactions costs, like bid ask spreads, brokerage commissions, the price impact of large trades, and the higher capital gains taxes associated with increased trading. Other studies focus on the persistence of performance by mutual funds and how mutual funds’ performance relates to their use of momentum strategies. This strand of the literature is especially important in my study of ETFs, since ETFs are similar in some aspects to mutual funds as well as having many of the trading characteristics of individual stocks. Grinblatt, Titman, and Wermers (1995) find that 77% of the mutual funds are momentum investors, buying stocks that are past winners, but most do not systematically sell past losers; they find a positive relationship between mutual fund positive performance and the use of a momentum trading strategy. Carhart (1997) finds that a momentum strategy of buying last year’s winner decile mutual funds and shorting last year’s loser decile mutual funds generates an abnormal return of 8% per year, where differences in market value and momentum o f stocks held explain 4.6%, differences in expense ratios explain 0.7%, and differences in transactions costs explain 1%. Sapp & Tiwari (2004) use Carhart’s four factor model to determine that mutual fund investors are not “smart money” even though the mutual funds that receive positive cash inflows tend to be the best performing funds. When Sapp & Tiwari use Fama & French’s three factor model, a strategy of buying the positive cash inflow funds and shorting the negative cash outflow funds generates positive a ' s indicating a “smart money” effect, but when they use Carhart’s four factor model, the same strategy generates a ' s that are insignificantly different from zero indicating no “smart money” effect. Sapp & Tiwari conclude that 3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. mutual fund investors are “naively chasing past returns, not identifying skilled fund managers” (Sapp & Tiwari p. 14) who follow momentum style strategies. Wermers (2003) studies mutual fund performance persistence and finds that part of the abnormally positive performance o f the winner funds is due to their holding of winner stocks providing momentum returns like Carhart found, but mutual fund investors strongly invest cash into last year’s top returning funds and only weakly disinvest in last year’s poorest returning funds and winner fund managers use their large cash inflows to implement momentum strategies more strongly than loser fund managers who tend to hold their loser stocks which continue to be losers. Wermers found that investor cash inflows to winner funds continue for two to four years, which is longer than the one year period attributed to momentum effects, and thus the high cash inflow funds continue to perform well due to the manager’s flow related trades chasing stocks with high past returns. Chan, Hameed, & Tong (2000) find momentum returns in 23 country index returns during 1980 to 1995 of at least 1% per month for formation and holding periods of 1 week, 2 weeks, or 4 weeks. They use Lo & MacKinlay’s (1990) weighting scheme where portfolio weights reflect the country’s past performance relative to the average past performance o f all 23 countries; above average performers are purchased and below average performers are shorted, but all countries may have a non-zero weight in the winner minus loser portfolio. This weighting scheme differs from Jegadeesh & Titman’s usual methodology of buying long the top decile of winners and shorting the bottom decile of losers. Chan, Hameed, & Tong find that 80 to 90% of the momentum profits are 4 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. due to equity return predictability and only a small portion is due to exchange rate predictability, and they find that momentum profits are larger for country indices following an increase in trading volume. Their momentum profits would be difficult to obtain after transactions costs as country indices are not directly investable and some of their markets restrict short selling. Balvers & Wu (2006) jointly consider momentum and mean reversion for 18 developed country index returns from 1970 to 1999. Their model generates a signal that identifies the winner and loser countries by its indicator score incorporating both momentum and mean reversion information, resulting in excess returns o f 1.1 - 1.7% per month, which outperforms both pure momentum and pure mean reversion strategies. Balvers & Wu find a strong negative correlation of - 35% between momentum and mean reversion effects, which explains why controlling for mean reversion effects can improve momentum returns. The studies of both Chan, Hameed, & Tong (2000) and Balvers & Wu (2006) can be improved by using ETFs rather than country indices since ETFs indexing various country stock indices are readily investable and shortable which are necessary conditions to realistically implement a momentum or contrarian strategy. My motivation in this study is to extend the domain of momentum/eontrarianism to a relatively new investment vehicle, namely, exchange traded funds or ETFs. ETFs are powerful and flexible investment vehicles that combine the diversified portfolio features of mutual funds with the trading possibilities of individual securities. Currently, ETFs function similarly to passively managed index mutual funds, as they are composed o f a portfolio of stocks or bonds that track a particular index, thus providing diversification 5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. within the portion o f the market tracked by that index. Four general categories of ETFs include: (1) broad based domestic indices like the S&P 500, the NASDAQ 100, the Dow Jones U.S. Total Market, the Russell 3000, the Wilshire 5000, some style specific indices in both a “value” and a “growth” version, and size based indices including large cap, mid cap, small cap, and micro cap, (2) sector indices including consumer, energy, financial, health, natural resources, real estate, utilities, and technology, (3) international indices including global stock indices, regional indices, and country specific indices, and (4) bond indices including three of the Lehman Treasury bond indices, two different corporate bond indices, and the Lehman TIPS index. What differentiates an ETF from a mutual fund is an ETF trades on an exchange (most on the AMEX) like a stock, enabling an ETF to be: purchased or sold at intraday market prices, purchased on margin, sold short, and traded via stop orders and limit orders. Ordinary mutual funds can only be purchased and sold by market orders for end of day prices, and cannot be purchased on margin or sold short, which prevents the usual zero investment momentum and contrarian portfolios o f buying the winners and shorting the losers or of buying the losers and shorting the winners. Also, many mutual funds have redemption fees and other constraints to discourage or prevent the short term trading necessary to implement a momentum or contrarian strategy. For implementing a momentum or contrarian strategy, purchasing or shorting an ETF gives the arbitrageur or investor a diversified portfolio of stocks while incurring only one bid ask spread and one round trip commission, clearly a cost advantage over assembling a portfolio of individual winner and loser stocks, which entails many bid ask spreads and many commissions. 6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. My research questions are: (1) whether a momentum investing strategy of buying winners and shorting losers generates abnormal returns in the ETF market, (2) whether a contrarian investing strategy of buying losers and shorting winners generates abnormal returns in the ETF market, and (3) which formation period, holding period is optimal for momentum investing and for contrarian investing. My results contribute to the literature an affirmative answer to the first research question about momentum, as buying the winner decile o f ETFs and shorting the loser decile of ETFs provide statistically significant abnormal returns for formation and holding periods of 4 weeks, 8 weeks, 12 weeks, 16 weeks, 20 weeks, 26 weeks, and 39 weeks with risk adjustment by either CAPM or Fama & French’s three factor model. The annualized momentum abnormal returns range from 8.2% to 22.1% under the Capital Asset Pricing Model and range from 8.4% to 13.5% under Fama & French’s three factor model. My results contribute to the literature an affirmative answer to the second research question about contrarianism, as buying the loser decile o f ETFs and shorting the winner decile of ETFs provide statistically significant contrarian abnormal returns for formation and holding periods of 1 day and 1 week. The annualized contrarian abnormal returns range from 17.0% to 86.9% and hold true for risk adjustment by the Capital Asset Pricing Model and Fama & French’s three factor model. Following the classic approach of Jegadeesh & Titman (1993), I find for question (3) that a 26 week formation and holding period provides the highest annualized abnormal returns of 22.1% to an ETF momentum strategy using CAPM to adjust for risk, while a 20 week formation and holding period provides the highest annualized abnormal returns of 13.5% to an ETF momentum strategy using Fama 7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. & French’s three factor model to adjust for risk. Also, a 1 day formation period and a 1 day holding period provide the highest abnormal returns to an ETF contrarian strategy, with annualized abnormal returns of 85.2% using CAPM and 86.9% using Fama & French’s three factor model. These research questions are important, because momentum and contrarianism are both wide spread anomalies identified by researchers as well as investment strategies used by practitioners like mutual funds and individual investors to attempt to earn abnormal returns. With the growth in ETFs from their introduction in 1993 with the SPDR Trust Series tracking the S&P 500 Index to 2005’s assortment of 217 ETFs consisting of 80 broad based domestic, 82 domestic sector, 49 global/international equity, and 6 bond ETFs, ETFs are on a growth path which should soon surpass the dollar amount invested in equity index mutual funds. From the Investment Company Institute’s (a mutual fund trade organization) December 2005 statistics, ETFs (excluding Merrill Lynch’s HOLDRS) represent a market value of $296.02 billion, which represents over 5% of the $5,504.50 billion invested in stock and hybrid mutual funds. Considering that about 10% of stock mutual fund investments are in indexed investments as opposed to actively managed funds, ETFs represent a significant portion (almost 35%) of the U.S. wealth invested in passively managed, index type investment vehicles. With the growing popularity of ETFs by traders and investors such an innovative financial product merits further study, especially when it can generate abnormal returns via a momentum or contrarian strategy. Most momentum/contrarian studies to date look at U.S. stocks, U.S. bonds, U.S. domestic mutual funds, and international stocks or country indices separately. By finding 8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. abnormal returns in momentum/contrarian portfolios of ETFs including broad domestic funds, sector or industry funds, international and country funds, and bond funds, I contribute to the literature by demonstrating momentum/contrarianism in a changing asset allocation setting which includes U.S. stocks, U.S. bonds, foreign stocks, as well as sector or industry funds. To my knowledge, this study is the first to identify momentum/contrarianism in such an asset allocation setting. Also, I contribute to the literature by finding that ETFs are an ideally suited investment with which to implement a momentum or contrarian trading strategy, since ETFs permit the purchase or sale of a diversified portfolio of securities for one commission and one bid ask spread, resulting in abnormal returns that exceed transactions costs. These results contradict Lesmond, Schill, & Zhou’s (2004) characterization of momentum profits as illusory and further contribute to the existence o f momentum/contrarian profits that are realistically attainable by investors and arbitrageurs utilizing ETFs, which provides further evidence against the theory o f market efficiency. 2. Methodology: Since Jegadeesh & Titman (1993) establish the generally accepted methodology for researching the momentum anomaly, I follow their methodology with appropriate adjustments to accommodate my study of ETFs. Since ETFs represent diversified portfolios designed and passively managed to track various domestic, sector, international, and bond indices, forming portfolios of ETFs is less necessary than Jegadeesh & Titman’s method o f forming decile portfolios of individual stocks based on 9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. formation period performance. However, when Moskowitz & Grinblatt (1999) study industry momentum they define the winner and loser portfolios as the top or bottom 3 out o f 20 industries, respectively, when ranked by formation periods of 1, 6, or 12 months, and measured holding period returns for periods of 1, 6, 12,24, or 36 months. Most momentum studies of mutual funds also group winner and loser funds into portfolios; at least with ETFs it is possible to short the losers while shorting mutual funds is not possible. With the relative newness of ETFs such long periods as 24 or 36 months would restrict the sample sizes to be rather small, as the end of 1999 found only 32 ETFs in existence, 4 broad based domestic ETFs, 11 sector ETFs, 17 foreign or country ETFs, and 0 bond ETFs. The year 2000 brought the largest number of new ETFs as 57 ETFs were introduced during 2000, resulting in 89 ETFs in existence by year end, consisting of 29 broad based domestic ETFs, 35 sector ETFs, 25 foreign or country ETFs, and 0 bond ETFs. Table 1.1 shows the annual growth in ETF offerings from 1993 to the present. Following Jegadeesh & Titman’s methodology, I define the winner ETFs as the top performing decile over various formation periods and the loser ETFs as the poorest performing decile over various formation periods, and then form the momentum portfolio that buys the winner ETFs and shorts the loser ETFs over various holding periods. Also, adapting Jegadeesh & Titman’s methodology, I form the winner minus loser portfolio each week to increase the power of my tests; I equally weight the appropriate winner and loser ETFs in the portfolios formed each week during the sample period and held for the indicated amount o f time. Also, with ETFs tracking four different types o f indices, broad based domestic, sector or industry, foreign or country, and bond, I measure momentum 10 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. with all the various ETF types pooled together to consider a strategy where market performance may favor one of the four types over the other three, which can determine whether a momentum-based asset allocation strategy exists, as well as to maximize my sample size. My sample period runs from March 20,1996 to December 31,2005, a period of 483 weeks, with 19 ETFs available in 1996, so that the top and bottom deciles begin with 2 ETFs each as winners and losers, respectively. In 2005,217 ETFs are available so that the deciles o f winners and losers both include 21 ETFs. I consider formation periods o f 1 day, 1 week, 2 weeks, 4 weeks, 8 weeks, 12 weeks, 16 weeks, 20 weeks, 26 weeks, 39 weeks, and 52 weeks coupled with holding periods of the same length as the formation period for a total of 11 different momentum strategies. Following Jegadeesh & Titman’s methodology, I also consider the above 10 weekly momentum strategies with a one week lag between portfolio formation and holding period to allow a more realistic time for an investor to determine the winner and loser ETFs and form the appropriate portfolios while avoiding some price pressure and perhaps minimizing the transactions costs compared to a hurriedly assembled portfolio. For the 1 day, 1 day momentum/contrarian strategy, I use an extra one day lag rather than an extra week lag. For each combination of formation and holding period, I compute annualized abnormal returns for the winner ETFs, the loser ETFs, and the zero investment momentum portfolio of buying the winner ETFs and shorting the loser ETFs using CAPM and Fama & French’s three factor model to adjust for risk. Clearly, the zero investment contrarian portfolio of buying the loser ETFs and shorting the winner ETFs simply reverses the sign of the zero investment momentum portfolio. 11 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3. Description of Data: Since most ETFs are listed on the AMEX and others are listed on the NYSE and NASDAQ, the daily return data is available on the Center for Research in Securities Prices (CRSP) database. Most of the previous studies use CRSP data which is of excellent quality when defining returns using close to close security prices. Momentum/contrarian portfolio risk characteristics like factor loadings on various risk measures like the market risk premium, RM - R F, small firm size minus big firm size, SMB, and high book to market value minus low book to market value, HML are calculated to confirm that abnormal momentum/contrarian returns are not due to different risk levels, or to different firm characteristics, like size or book-to-market ratios, or to different industry compositions, or to different value or growth measures. Daily data on the factor mimicking portfolios for the three zero investment factor mimicking portfolios, i.e., Rm - R f , SMB, and HML are available on Kenneth R. French’s website at: http://mba.tuck.dartmouth.edu/pages/facultv/ken.french/data library.html. 4. Outline of Model: To evaluate the various ETFs in terms of risk levels, I use CAPM and Fama & French’s three factor model. Thus, the factor models used are: Rit CAPM: - Rfl = or, + b: •RMRFt + e„ Fama & French’s three factor model: Rit - RFi = or, + b, ■RMRF, + s, ■SMB, + h, ■HML, + e„ 12 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. where Rlt is the return on portfolio i in period t, RFl is the Treasury bill rate in period t, RMRF is the excess return on a value-weighted market proxy, and SMB and HML are the returns on zero investment, value-weighted, factor-mimicking portfolios for firm size and book-to-market, respectively. Thus, using both models, I calculate the crosssectional a ’s for the excess return on the winner ETFs, the excess return on the loser ETFs, as well as the winner minus loser momentum portfolio to measure abnormal positive or negative returns. Although the three risk factors, RMRF, SMB, and HML are all calculated relative to U.S. equities, no adjustments need be made to use the above factor risk premiums for the foreign ETFs since all the ETFs are traded in dollars, traded during U.S. market hours, and function as perfect substitutes for the other three categories o f ETFs. Also, Zhong & Yang (2005) find that the prices of international ETFs are greatly influenced by U.S. risk factors. 5. Results: A. Raw Returns: Table 1.2 shows the annualized mean raw returns and t-statistics for the 11 momentum strategies with information for the winner minus loser portfolio as well as the winner and loser portfolios separately reported. The momentum returns for the formation and holding periods of 2 weeks to 39 weeks are economically significant, ranging from annualized winner minus loser returns of 5.8% to 18.6%. Also, the winner portfolios keep winning and some of the losers keep losing while others reverse with winner annualized returns ranging from 8.2% to 16.4% and with loser annualized returns ranging 13 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. from - 2.8% to 2.3%. Momentum returns are maximized at 18.6% with a 20 week formation period and a 20 week holding period. Contrarian returns are evident in the short formation and holding periods of 1 day and 1 week as well as in the long formation and holding periods o f 52 weeks with mean annualized winner minus loser returns ranging from - 2.8% to - 86.4%. Thus, by reversing our momentum strategy to a contrarian strategy of buying the past losers and shorting the past winners, economically significant returns are possible with annualized loser returns ranging from 7.6% to 55.9% and annualized winner returns ranging from - 30.5% and 4.8% for the 1 day, 1 week, and 52 week formation and holding periods. The long formation and holding periods of 39 weeks and 52 weeks are not clearly momentum or contrarian strategies as both winners and losers continue to win but with less economic significance than the short formation and holding periods. However, all the raw returns are statistically insignificant as the largest t-statistic is only 0.49. These results are consistent with Balvers & Wu (2006) who find economically significant but statistically insignificant returns for their model which combines momentum and mean reversion for 16 international ETFs from April 1996 to December 2003. Balvers & Wu attribute the statistical insignificance to the shortness of the available sample period which may also be the problem in my study. Henker, Martens, & Huynh (2006) find statistically insignificant momentum returns for U.S. stocks in the 1993 - 2004 period due to the poor performance of momentum strategies during the 2001 - 2004 subperiod. The statistical insignificance could also be a sample specific result from the added volatility in stock returns during the late 1990s and early 14 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2000s. Thus, I next use both CAPM and Fama & French’s three factor model to adjust my momentum strategies for risk and identify statistically significant abnormal returns. 5. Results: B. Risk Adjusted Momentum Returns: Table 1.3 shows that the annualized momentum abnormal returns for winner minus loser portfolios are very statistically significant at the 1% level of significance for strategies of formation and holding periods from 4 weeks to 39 weeks with risk adjusted by both CAPM and Fama & French’s three factor model. The winner minus loser annualized momentum abnormal returns range from 8.2% to 22.1% over the strategies of formation and holding periods from 4 weeks to 39 weeks with risk adjusted by CAPM and Fama & French’s three factor model. In general, the annualized momentum abnormal returns are larger for CAPM than for Fama & French’s three factor model, which is the expected result, as CAPM makes less complete risk adjustments. In general, the annualized momentum abnormal returns are maximized for the 20 week or 26 week formation and holding period strategy with momentum returns of 22.1% under CAPM for 26 weeks and 13.5% under Fama & French’s three factor model for 20 weeks. From the results for the excess returns above the appropriate periodic Treasury bill rate for the momentum winner and loser ETF portfolios, I find that the losers drive the winner minus loser results under both CAPM and Fama & French’s three factor model. The loser annualized momentum abnormal returns are very significant at the 1% level for all formation and holding periods from 4 weeks to 39 weeks with magnitudes 15 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ranging from - 8.2% to - 14.4%, while the winner annualized momentum abnormal returns are statistically insignificant for the formation and holding periods of 4 weeks, 8 weeks, 12 weeks, and 39 weeks with statistically significant magnitudes ranging from 3.5% to 6.3% for the formation and holding periods of 16 weeks, 20 weeks, and 26 weeks. Annualized contrarian abnormal returns are very statistically significant at the 1% level of significance for the short formation and holding periods o f 1 day and at the 5% level o f significance for the formation and holding periods of 1 week with risk adjusted by both CAPM and Fama & French’s three factor model. The winner minus loser returns range from - 17.0% to -86.9%. Thus, by reversing our momentum strategy to a contrarian strategy o f buying the past losers and shorting the past winners, statistically significant returns are possible with abnormal annualized loser returns for the 1 day formation and holding period of 45.2% under CAPM and 45.4% under Fama & French’s three factor model, and abnormal annualized winner returns for the 1 day formation and holding period o f —40.1% under CAPM and - 41.6% under Fama & French’s three factor model. The long formation and holding periods of 39 weeks and 52 weeks under CAPM and Fama & French’s three factor model are not clearly momentum or contrarian strategies as both winners and losers continue to lose with negative abnormal annualized returns. Thus, a formation and holding period of 1 day maximizes our contrarian abnormal returns, while a formation and holding period of 20 weeks or 26 weeks maximizes our momentum abnormal returns. 16 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5. Results: C. Risk Adjustment under CAPM: Table 1.4 shows the winner minus loser portfolio has a negative beta or systematic risk measure with significantly negative coefficients on the market risk premium term for all formation and holding periods except 20 weeks. Excess returns on all winner and loser portfolios have very significant positive betas with positive coefficients on the market risk premium and p-values of less than .0001 for all formation and holding periods. The losers have higher betas than the winners across all formation and holding periods, indicating that the loser portfolios have a higher level of systematic risk than the winners. Clearly, the resulting momentum abnormal returns to the winner minus loser ETFs for formation and holding periods from 2 weeks to 39 weeks are not due to a higher level of risk in the winner ETFs than the loser ETFs. Also, the higher level of systematic risk in the loser portfolios explains some but not all of the contrarian abnormal returns as a beta difference of 0.19 to 0.28 is too small to generate the magnitude of the contrarian returns. With risk adjustment under CAPM, momentum strategies of buying the winner ETFs and shorting the loser ETFs with formation and holding periods from 2 weeks to 39 weeks generate significant positive abnormal returns and contrarian strategies of buying the loser ETFs and shorting the winner ETFs with formation and holding periods from 1 day to 1 week generate significant positive abnormal returns. 17 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5. Results: D. Risk Adjustment under Fama & French’s three factor model: Table 1.5 shows the winner minus loser portfolio has a negative beta or systematic risk measure with significantly negative loadings on the market risk premium for all formation and holding periods except 1 week, 20 weeks, and 26 weeks. All winner and loser portfolios have very significant positive betas with positive factor loadings on the market risk premium and p-values of less than .0001 for all formation and holding periods. The losers have higher factor loadings on the market risk premium than the winners across all formation and holding periods, indicating that the loser portfolios generally have a higher level of systematic risk than the winners. The winner minus loser portfolio has positive factor loadings on HML, the value minus growth risk premium, for all formation and holding periods except 39 weeks, with about half o f the formation and holding periods significant at the 5% level and the rest insignificant. In general, the winner portfolios have positive factor loadings on HML suggesting the winner ETFs contain value stocks with higher book to market ratios; however, most formation and holding periods are insignificant at the 10% level. In contrast, the loser portfolios have negative factor loadings on HML with most significant suggesting the loser ETFs contain growth stocks with lower book to market ratios. Thus, book to market levels significantly distinguish winner and loser ETFs, as loser ETFs contain more growth stocks than winner ETFs. The winner minus loser portfolio has a positive factor loading on SMB, the small firm minus big firm risk premium for all formation and holding periods from 2 weeks to 18 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 39 weeks, with all being significant except 2 weeks and 39 weeks. The positive factor loadings suggest the winner ETFs contain smaller cap stocks than the loser ETFs for these time periods corresponding to momentum abnormal returns. The winner minus loser portfolio has a negative factor loading on SMB, for the 1 day, 1 week, and 52 week formation and holding periods corresponding with contrarian abnormal returns, but only 52 weeks is significant, which again suggests that the loser ETFs contain larger cap stocks than the winner ETFs. The winner ETFs load significantly positive on SMB for all formation and holding periods except the 1 week, 2 week, and 52 week winners which are positive but insignificant. The loser ETFs’ factor loadings on SMB are all insignificantly different from zero except for the 1 day and 52 week formation and holding period which are very significantly positive with p-values less than .0001 and of larger magnitude than the corresponding winner ETFs, indicating that the loser ETFs for both the 1 day and 52 formation and holding periods have significantly smaller cap stocks on average than the corresponding winner stocks, which are also small cap but not as small on average as the loser ETFs. Thus, small cap stocks in the ETFs account for both the positive momentum abnormal returns and the positive contrarian abnormal returns with the momentum returns driven by winners in the formation period and the contrarian returns driven by the losers in the formation period. Clearly, the resulting momentum returns to the winner minus loser ETFs are not due to a higher level of risk in the winner ETFs than the loser ETFs. The losers have higher systematic risk than the winners, the winners have higher book to market ratios than the losers, suggesting winners include more value stocks while losers include more 19 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. growth stocks, and the ETFs with positive abnormal returns during the holding period, i.e., contrarian losers and momentum winners, are generally smaller cap than the ETFs with negative abnormal returns during the holding period, i.e., contrarian winners and momentum losers. With risk adjustment under Fama & French’s three factor model, momentum strategies o f buying the winner ETFs and shorting the loser ETFs with formation and holding periods from 4 weeks to 39 weeks generate significant positive abnormal returns and contrarian strategies of buying the loser ETFs and shorting the winner ETFs with formation and holding periods from 1 day to 1 week generate significant positive abnormal returns. 6. Robustness of Results: A. Extra Time Between Formation and Holding Periods: If ETF returns are autocorrelated, then a momentum strategy that benefits from continuation o f returns would appear to be profitable. Two possible explanations are nonsynchronous trading as the underlying stocks in some of the international ETFs may trade over different time periods than the actual ETFs trade on the U.S. markets and bid ask bounce from pricing pressure attempting to buy past winners or to short past losers. Jegadeesh & Titman’s usual methodology is to allow an extra week between the portfolio formation period where the winners and losers are identified by past performance and the portfolio holding period where the winners are purchased and the losers are shorted. I adapt Jegadeesh & Titman’s methodology to add an extra day between the formation and holding period for the 1 day 1 day strategy since an extra week seemed excessive; for all 20 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. other formation and holding periods from 1 week to 52 weeks I add the usual week between the formation and holding periods. Such an adjustment should reduce the ability of a momentum strategy to take advantage o f the autocorrelation in ETF returns. Since contrarian strategies benefit from reversals of returns rather than continuations, it is not clear apriori what impact the extra week between formation and holding periods should have on the contrarian abnormal returns. Table 1.6 shows adding extra time between the formation and holding period does not eliminate the momentum and contrarian returns identified earlier in Table 1.3. Momentum strategies o f buying past winner ETFs and shorting past loser ETFs continue to generate significant abnormal returns for all formation and holding periods from 2 weeks to 39 weeks when adjusting for risk using both CAPM and Fama & French’s three factor model. The extra time generally only slightly reduces the abnormal momentum returns for most holding periods from 4 weeks to 39 weeks. However, the extra day removes about 72% of the contrarian abnormal returns for the formation and holding period o f 1 day, but a significant but smaller contrarian return still remains. The extra week converts the formation and holding period of 1 week from generating a significant contrarian abnormal return to generating an insignificant momentum return and it increases the magnitude and significance of the momentum abnormal returns for a formation and holding period o f 2 weeks under both risk adjustment models. The 52 week contrarian returns remain insignificant under both CAPM and Fama & French’s three factor model. 21 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6. Robustness of Results: B. Transactions Costs: Another issue is whether the momentum and contrarian abnormal returns are real or illusory after considering transactions costs, since both momentum and contrarian strategies require a significant amount of trading to implement, potentially costing the investor or arbitrageur the bid ask spread, brokerage commissions, and price impact for large orders. Lesmond, Schill, & Zhou (2004) characterize the momentum profits with individual stocks identified by Jegadeesh & Titman in 1993 and 2001 and Hong, Lim, & Stein in 2000 as illusory since the momentum profits net o f their transactions cost estimates are insignificantly different from zero. However, ETFs are much less costly to trade than individual equities with smaller bid ask spreads and more liquidity to reduce the price impact of large trades. Table 1.7 shows the momentum returns net of transactions costs for the formation and holding period of 26 weeks, which is the usual recommended momentum strategy in most previous studies. The actual transactions were tabulated by domestic and bond, sector, and international ETF over the 9.29 years studied. On average, the winner ETFs consisted of 10.64% domestic and bond ETFs, 46.52% sector ETFs, and 42.84% international ETFs, while the loser ETFs consisted o f 8.08% domestic and bond ETFs, 54.97% sector ETFs, and 36.95% international ETFs. I estimated the quoted bid ask spreads as the higher o f those identified in Huang & Wei (2004) and Salomon Smith Barney (2002), resulting in estimates of 0.33% for domestic and bond ETFs, 0.62% for sector ETFs, and 0.867% for international ETFs. Brokerage commissions were estimated 22 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. at 0.13% by combining Scottrade’s $7.00 per trade with an estimated account balance of $125,000 invested long in the winner ETFs and $125,000 invested short in the loser ETFs. Quoted bid ask spreads and brokerage commissions total 8.25% per year which reduced the net annualized abnormal returns to 13.81% under CAPM and 4.58% under Fama & French’s three factor model. Huang and Wei estimated the effective bid ask spread to be about 30% less than the quoted bid ask spread as transactions often take place between the quoted bid and ask prices; using effective bid ask spreads reduces the transactions cost to 5.82% per year. With effective spreads, the net annualized abnormal returns are 16.24% under CAPM and 7.01% under Fama & French’s three factor model. Clearly, the momentum abnormal returns under CAPM and Fama & French’s three factor model are not illusory, but represent economically viable risk adjusted returns even when reduced by transactions costs. 6. Robustness of Results: C. Growth in Number of ETFs: Another issue is the rapid growth in the number of ETFs. Were the momentum and contrarian returns determined by the early years when a smaller number of ETFs existed? Table 1.8 considers this question by dividing the sample period into the early period from March 20,1996 to December 31,2000 with 19 to 89 ETFs in existence and the later period from January 1, 2001 to December 31, 2005 with 113 to 217 ETFs in existence. I distinguish these two sample periods for portfolios with a formation and holding period o f 26 weeks with risk adjusted by CAPM and Fama & French’s three 23 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. factor model. Significant momentum abnormal returns occur under both risk adjustment models for the early period from 1996 to 2000. However, during the later period from 2001 to 2005, significant momentum abnormal returns occur under CAPM, but significant contrarian abnormal returns occur under Fama & French’s three factor model. Most o f the factor loadings remain the same sign and significance in both periods as well as similar to the sign and significance for the entire period from 1996 to 2005 as denoted in Tables 1.4 and 1.5 previously. For some portfolios the abnormal returns are larger in the earlier period and for others they are larger in the later period. One clear difference between the two periods is that winner ETFs included growth stocks, with low book to market ratios, during the 1996 - 2000 period but winner ETFs included value stocks, with high book to market ratios, during the 2001 - 2005 period. Dividing the sample period into two shorter periods illustrates some periodic differences between winner and loser ETFs in the two periods. However, such differences do not indicate a problem with the number o f ETFs in existence, but more likely reflect the differing economic conditions in the different sub-periods as 1996 to early 2000 reflected a roaring bull market, followed by the bear markets o f later 2000 through 2002, and the milder bull markets of 2003 to 2005. 6. Robustness of Results: D. Portfolio Rebalancing: Another issue is portfolio rebalancing. The usual Jegadeesh and Titman methodology rebalances the winner and loser portfolios monthly, resulting in the 6 month 24 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. winners and losers consisting o f six equally weighted portfolios with one sixth formed from today’s winners and losers and one sixth formed from each of the previous one, two, three, four, and five month’s winners and losers. My methodology used equal weightings of the top and bottom decile ETF performers to form the winner and loser portfolios like Jegadeesh and Titman, but formed a new portfolio each week during the sample period and held it for the indicated period of time without rebalancing. Table 1.9 compares the abnormal returns for portfolios with a formation and holding period of 26 weeks, with risk adjustment by CAPM and Fama & French’s three factor model, using my methodology and 4 week rebalancing similar to Jegadeesh and Titman. Rebalancing six portfolios over 4 week periods necessitates using a 24 week formation and holding period rather than a 26 week formation and holding period. The results are qualitatively the same for abnormal returns under both CAPM and Fama & French’s three factor model. Annualized abnormal return magnitudes are 22.1% (no rebalancing) vs. 22.2% (4 week rebalancing) under CAPM and 12.8% (no rebalancing) vs. 16.1% (4 week rebalancing) under Fama & French’s three factor model, while 4 week rebalancing has slightly lower statistical significance than no rebalancing. 6. Robustness of Results: E. Asset Allocation vs. Type of ETF: My results so far have not identified the source of the abnormal momentum or contrarian returns since I formed all winner and loser portfolios of ETFs from the asset pool that included all four types of ETFs, namely, domestic, sector, international, and 25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. bond. My study purposefully chooses all four types of ETFs simultaneously to extend momentum/contrarian studies to the asset allocation domain as well as to maximize the length of the sample period studied, since ETFs are a relatively new investment vehicle. On average, about 51% of the ETFs in the winner and loser portfolios are sector ETFs and about 40% are international ETFs, with the remaining 9% from the domestic and bond ETFs. To control for the effects of including sector and international ETFs, I augmented CAPM and Fama & French’s three factor model with dummy variables identifying whether or not the winner, loser, and WML portfolios included sector or international portfolios. Specifically, the augmented models are: CAPM: Rit - RFl = a t + bi ■RMRFt + c, • Dum St + d t - Dum It + ejt Fama & French’s three factor model: Rlt - RFt = a , + bt ■RMRFt + st ■SMB, + ht ■HML t + c, • Dum St + d t • Dum lt + eit where: DumSt = 1 if the portfolio at time t includes at least one sector ETF, and 0 otherwise, DumIt = 1 if the portfolio at time t includes at least one international ETF, and 0 otherwise, and all the other variables are defined as before. In the augmented model, or, measures the abnormal momentum/contrarian returns in portfolios consisting of domestic and bond ETFs only, while c, measures the difference in abnormal momentum/contrarian returns in portfolios including sector ETFs relative to those consisting of domestic and bond ETFs, and dt measures the difference in abnormal momentum/contrarian returns in portfolios including international ETFs relative to those consisting of domestic and bond ETFs. 26 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.9 shows clearly large and very significant abnormal momentum returns for WML portfolios formed from domestic and bond ETFs only with annualized abnormal momentum returns ranging from 44.7% to 101.8% with larger magnitudes for risk adjustment by augmented CAPM than Fama & French’s augmented three factor model and larger magnitudes using 4 week rebalancing than no rebalancing. These large abnormal momentum returns are driven by the loser portfolios of domestic and bond ETFs which continue to lose a very significant annualized abnormal momentum return of - 33.2% to - 56.6%. Clearly, the economic times when the top and bottom 26 week performance deciles are dominated by domestic and bond ETFs rather than the more volatile and focused sector and international ETFs are excellent times to implement a momentum strategy for the next 26 weeks. At more typical economic times when the top and bottom 26 week performance deciles include either sector or international ETFs or both, the abnormal momentum returns over the next 26 weeks are reduced, with the result being driven by the sector and international ETF losers losing significantly less than their domestic and bond ETF counterparts. Table 1.10 frames the analysis of Asset Allocation vs. Type of ETF differently, by comparing the 26 week formation and holding period WML, Winner, and Loser ETF performance over a shorter sample period from January 1, 1999 to December 31,2005 to allow a sufficient number o f ETFs in each category to meaningfully define winners and losers. With risk adjustment by CAPM, the asset allocation pool including all four types of ETFs generates very significant abnormal momentum returns as do all four ETF types separately as well. Both the sector ETFs and domestic ETFs generate larger abnormal 27 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. momentum returns than the asset allocation pool, but the bond ETFs and international ETFs both generate smaller abnormal returns than the asset allocation pool primarily because the bond loser abnormal returns are insignificantly different from zero and the international loser abnormal returns are positive rather than negative. With risk adjustment by Fama & French’s three factor model, the asset allocation pool including all four types o f ETFs generates a somewhat significant abnormal contrarian return, but the bond ETFs generate a significant abnormal momentum return and the domestic, international, and sector ETFs generate abnormal contrarian returns that are insignificantly different from zero. Only the international winner ETFs generate a positive significant abnormal return. The bond loser ETFs generate a negative very significant abnormal return, but the loser ETFs for the asset allocation pool and the domestic, international, and sector generate positive very significant abnormal returns. Since losing 1996, 1997, and 1998 from the sample reduces the magnitude and significance of the abnormal momentum/contrarian returns across all five ETF pools, some of the advantage of the asset allocation pool of all four types of ETFs is due to the longer available sample period. However, this analysis indicates some differences between the holding period performance of the four different types of ETFs which may lead to improvements over a naive momentum/contrarian strategy. 28 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7. Conclusion: This study extends Jegadeesh & Titman’s momentum/contrarian anomaly to a new domain, portfolios of ETFs that either buy the winners and short the losers or buy the losers and short the winners, respectively. Currently, all U.S. ETFs are passively managed to track an index, not actively managed to time the market or “beat the market” by loading up on high momentum stocks. Yet, in spite of this disadvantage to actively managed mutual funds, ETFs provided economically and statistically significant abnormal returns to contrarian strategies of buying the loser ETFs and shorting the winner ETFs with formation and holding periods of 1 day and 1 week, and to momentum strategies o f buying the winner ETFs and shorting the loser ETFs with formation and holding periods from 4 weeks to 39 weeks. This study is also the first to demonstrate momentum in a changing asset allocation setting which includes U.S. stocks, U.S. bonds, foreign stocks, as well as sector or industry funds. In contrast to Lesmond, Schill, & Zhou, I find that momentum/contrarian abnormal returns are not illusory, but are achievable by investors and arbitrageurs using realistic estimates of bid ask spreads and brokerage commissions. ETFs are ideal instruments with which to implement a contrarian and momentum strategy, since ETFs allow the purchase or short sale of a diversified portfolio o f securities for one commission and one bid ask spread with minimal price impact. Using CAPM and Fama & French’s three factor model to adjust for risk, I find that the contrarian and momentum abnormal returns available with ETFs can not be explained by rational differences in risk and so I provide further evidence of the momentum/contrarian anomaly’s attack on market efficiency. Such findings 29 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. contribute support to the behavioral explanation o f why momentum occurs as investors underreact to relevant information and so prices only gradually reflect relevant news and information. In the future, as the sample period of ETFs’ data is longer, I plan to check the longer term performance for 18, 24, and 36 months after portfolio formation, to determine the duration o f momentum/contrarian performance and anticipate the normal long-term reversals as documented by De Bondt & Thaler (1985, 1987), thus extending their results to a new domain of ETFs. Also, I anticipate that ETFs will continue to grow and expand their array o f offerings where eventually I anticipate actively managed ETFs to more directly compete with mutual funds and yet appeal to larger investors and investors that trade more actively than permitted by mutual funds. With actively managed ETFs, I expect the usual hot performance chasing behavior by investors, resulting in ETF managers that utilize momentum strategies thus magnifying the momentum performance from the passively managed ETFs studied in this paper. ETFs will continue to adapt and evolve to provide more flexible investment vehicles for investors and traders, as well as future questions to research and investigate relative to momentum/contrarian strategies as new ETFs and more years o f data become available. 30 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. REFERENCES Balvers, Ronald J. & Yangru Wu, 2006. Momentum and Mean Reversion Across National Equity Markets. Journal o f Empirical Finance, 13: 2 4 -4 8 . Barberis, Nicholas, Andrei Shleifer, & Robert Vishny, 1998. A Model of Investor Sentiment. Journal o f Financial Economics, 49: 3 0 7 -3 4 3 . Berkowitz, Stephen A., Dennis E. Logue, & Eugene A. Noser, 1988. The Total Costs of Transactions on the NYSE. Journal o f Finance, 43: 9 7 -1 1 2 . Carhart, Mark M., 1997. On Persistence in Mutual Fund Performance. Journal o f Finance, 52: 57 - 82. Chakrabarty, Bidisha & Charles Trzcinka, 2003. Momentum Strategies and Financial Databases: An Investigation of Intraday Pattern in Price Momentum. Unpublished Working Paper, Saint Louis University, University of Indiana. Chan, Kalok, Allaudeen Hameed, & Wilson Tong, 2000. Profitability of Momentum Strategies in the International Equity Markets. Journal o f Financial and Quantitative Analysis, 35: 153 - 172. Chan, Louis K. C., Narasimhan Jegadeesh, & Josef Lakonishok, 1996. Momentum Strategies. Journal o f Finance, 51: 1681-1713. Conrad, Jennifer & Gautam Kaul, 1998. An Anatomy of Trading Strategies. Review o f Financial Studies, 11: 489 - 519. Daniel, Kent, David Hirshleifer, & Avanidhar Subrahmanyam, 1998. Investor Psychology and Security Market Under- and Overreactions, Journal o f Finance, 53: 1 8 3 9 - 1886. De Bondt, Werner F. M. & Richard Thaler, 1985. Does the Stock Market Overreact? Journal o f Finance, 40: 793 - 805. De Bondt, Werner F. M. & Richard Thaler, 1987. Further Evidence On Investor Overreaction and Stock Market Seasonality. Journal o f Finance, 42: 557 - 581. Fama, Eugene F. & Kenneth R. French, 1993. Common Risk Factors in the Returns on Stocks and Bonds. Journal o f Financial Economics, 33: 3 - 56. Fama, Eugene F. & Kenneth R. French, 1996. Multifactor Explanations of Asset Pricing Anomalies. Journal o f Financial Economics, 51: 5 5 -8 4 . 31 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. REFERENCES Gastineau, Gary L., 2002. The Exchange-Traded Funds Manual. (John Wiley & Sons, Inc., New York). Gebhardt, William R., Soeren Hvidkjaer, & Bhaskaran Swaminathan, 2002. Stock and Bond Market Interaction: Does Momentum Spill Over? Unpublished Working Paper, Axia Energy Europe, Ltd. London, U.K., University of Maryland, & Cornell University. Grinblatt, Mark & Tobias J. Moskowitz, 2004. Predicting Stock Price Movements from Past Returns: the Role of Consistency and Tax-Loss Selling. Journal o f Financial Economics, 71: 541 - 579. Grinblatt, Mark, Sheridan Titman, & Russ Wermers, 1995. Momentum Investment Strategies, Portfolio Performance, and Herding: A Study of Mutual Fund Behavior. American Economic Review, 85: 1088 - 1105. Henker, Thomas, Martin Martens, & Robert Huynh, 2006. The Vanishing Abnormal Returns of Momentum Strategies and “Front-Running” Momentum Strategies. Unpublished Working Paper. Hong, Harrison, Terrance Lim, & Jeremy C. Stein, 2000. Bad News Travels Slowly: Size, Analyst Coverage, and the Profitability of Momentum Strategies. Journal o f Finance, 55: 265 - 295. Hong, Harrison & Jeremy C. Stein, 1999. A Unified Theory of Underreaction, Momentum Trading and Overreaction in Asset Markets. Journal o f Finance, 54: 2143-2184. Huang, Chi - Hsiang & Peihwang Wei, 2004. Bid Ask Spreads and Holding Periods of Exchange Traded Funds. Unpublished Working Paper. Jegadeesh, Narasimhan & Sheridan Titman, 1993. Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency. Journal o f Finance, 48: 6 5 -9 1 . Jegadeesh, Narasimhan & Sheridan Titman, 2001. Profitability of Momentum Strategies: An Evaluation of Alternative Explanations. Journal o f Finance, 56: 699 - 720. Jordan, Steve, 2004. Is Momentum A Self-Fulfilling Prophecy? Yale International Center for Finance Working Paper No. 04 - 25. 32 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. REFERENCES Lesmond, David A., Michael J. Schill, & Chunsheng Zhou, 2004. The Illusory Nature of Momentum Profits. Journal o f Finance, 71: 349 - 380. Lo, Andrew W. & A. Craig MacKinlay, 1990. When Are Contrarian Profits Due to Stock Market Overreaction? Review o f Financial Studies, 3: 175 - 205. Moskowitz, Tobias J. & Mark Grinblatt, 1999. Do Industries Explain Momentum? Journal o f Finance, 54: 1249 - 1290. Okunev, John & Derek White, 2001. Do Momentum Based Strategies Still Work In Foreign Currency Markets? Unpublished Working Paper, B.T. Funds Management, Sydney, Australia, University of New South Wales. Poterba, James M. & John B. Shoven, 2002. Exchange Traded Funds: A New Investment Option for Taxable Investors. Massachusetts Institute of Technology Department of Economics Working Paper #02-07. Rouwenhorst, K. Geert, 1998. International Momentum Strategies. Journal o f Finance, 53: 2 6 7 -2 8 4 . Salomon Smith Barney, 2002. Snapshot Study Performed 40 Random Snapshots of all ETFs in January 2002. Sapp, Travis & Ashish Tiwari, 2004. Does Stock Return Momentum Explain the “Smart Money” Effect? Journal o f Finance, 59: 2605-2622. Scott, Maria Crawford & Jean Henrich, 2004. Mutual Funds: The Individual Investor’s Guide to Exchange-Traded Funds. American Association o f Individual Investors Journal, 26: No. 9,13 - 29. Wermers, Russ, 2003. Is Money Really “Smart”? New Evidence on the Relation Between Mutual Fund Flows, Manager Behavior, and Performance Persistence. Unpublished Working Paper. University of Maryland. Zhong, Maosen & Hui Yang, 2005. Risk Exposures and International Diversification: Evidence from iShares. Journal o f Business Finance & Accounting, 32: 737 771. 33 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 124 134 167 217 Total ETFs CM0> o> o> CMO) CO CM CO CO to in in CD CO ■M M- 58 82 CMCMCMco o> CO ■M - 05 CMCO CO CO o CO CO O o O o CMCM 2005 25 CO in O) in CO 'M- 'M- o o o o o O) Broad Based Sector o *5 (0 o> o o o CM 2001 2002 Q I II. h- N. 1994 1995 1996 1997 1998 1999 o> 1 o or Country Bond c O o O h- 85 ITABLE 1.1 : Growth of Exchange Traded Funds: O o O O o O o o O '3' CD CD CO Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.2: Annualized Raw Returns Exchange Traded Funds and Momentum Retturns: March 21,1996 - December 31, 2005 1day 1day 1wk 1wk 2wk 2wk 4wk 4wk 8wk 8wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk WML: Mean -86.4250% -18.8344% -0.0955 -0.1920 t - statistic 10.7016% 0.0720 13.1781% 0.1278 12.3149% 0.1562 9.6802% 0.1495 15.3358% 0.2567 18.5864% 0.3307 17.9328% 0.3527 5.8472% 0.1081 -2.8178% -0.0593 Winner: -30.5250% Mean -0.0739 t - statistic -1.4872% -0.0086 12.7686% 0.1090 12.3734% 0.1389 11.0318% 0.2044 10.2531% 0.2009 14.6988% 0.3076 16.4421% 0.3754 15.1296% 0.4879 8.1793% 0.3187 4.8153% 0.2050 Loser: Mean t - statistic 17.3524% 0.0825 2.0696% 0.0133 -0.8047% -0.0071 -1.2838% -0.0144 0.5729% 0.0076 -0.6367% -0.0098 -2.1442% -0.0380 -2.8032% -0.0521 2.3321% 0.0438 7.6331% 0.1646 Notes: 55.9000% 0.1234 (1) Annualized mean raw returns for the period indicated. Periodic returns are annualized by mul tiplying by eil her 250 trad ing days for 1 day 1 day strategy or by 52 divided by the number of w eeks in the 10 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period of time. (4) The WML is the zero net - inves tment portfo io created by buying the winner ETFs and by shorting the loser ETFs for t ie indicated holding period. 35 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.3: Annualized Abnormal Returns: Exchange Traded Funds and Momentum Relturns: March 21,1996 - December 31,2005 12wk12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39 wl 52wk 52wk 1day 1day 1wk 1wk 2wk 2wk 4wk 4wk 8wk 8wk CAPM: R ^ ~ R & = a , + b, • R M R F , + e tt WML: 8.1920% -1.0904% Constant -85.2250% -16.9780% 13.5174% 15.5623% 15.4895% 12.1745% 17.2520% 19.3531% 22.0604% 0.0014 0.0002 0.0442 0.0001 0.0000 0.0000 0.0000 0.0018 0.0000 0.0473 0.6590 p - value Winner: 3.7937% 1.5139% -0.5009% 5.6134% 3.1798% 2.0579% 6.3106% -0.0535% -2.7969% Constant -40.0750% -11.3152% 0.7597 0.0204 0.4264 0.4841 0.3820 0.0002 0.0000 0.9562 0.0461 0.0010 p - value 0.0000 Loser: 45.1500% 5.6576% -10.3402% -13.5031% -13.9757% -12.6754% -13.4579% -13.7394% -14.4270% -8.2455% -1.7065% Constant 0.0002 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.3722 0.0366 0.3786 p - value 0.0000 Fama - French's 3 Factor Model: R,, - R Fl = a , + bf • R1 dRF , + j, • S M B . + h, • HML . + e WML: 8.4255% -0.0062% Constant -86.9250% -19.7964% 11.0838% 11.6922% 13.0897% 10.2860% 13.2135% 13.5426% 12.8348% 0.0164 0.0006 0.0019 0.0000 0.0000 0.0000 0.1056 0.0083 0.0000 0.0213 0.9985 p - value Winner: 0.3237% -0.3315% 3.5321% 4.6792% 3.8318% -0.6001% -3.9998% 2.4336% 0.1768% Constant -41.5500% -14.6744% 0.8364 0.0292 0.0264 0.9543 0.8435 0.0015 0.0019 0.5465 0.0344 0.0000 0.5395 p - value Loser: 7.5140% -8.6502% -11.5154% -12.7660% -10.6175% -9.6811% -8.8634% -9.0030% -9.0256% -3.9937% 45.3750% Constant 0.0001 0.0002 0.0001 0.0002 0.0002 0.0020 0.0008 0.1332 0.0000 0.2342 0.0981 p - value (1) Annualized abnormal returns for the period indicated. Periodic abnormal returns, i.e. alpha intercepts, are annualized by Notes: multiplying by either 250 trading days for 1 day 1 day strategy or by 52 divided by the number of w eeks in the 10 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period of time. (4) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for t ie indicated holding period. (5) p-values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using the Newey-W est adjustment (1987). 36 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.4: Abnormal Returns Using CAPM Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 1day 1day 1wk 1wk WML: Constant p - value M kt-RF p - value Adj. R-Sq. Winner: Constant p - value M kt-RF p - value Adj. R-Sq. Loser: Constant p - value M kt-RF p - value Adj. R-Sq. 2wk 2wk 4wk 4wk 8wk 8wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39 wl 52wk 52wk 0.0741 0.0281 0.0002 -0.3813 0.0005 0.0408 -0.2960 0.0164 0.0208 0.0744 0.0000 -0.1256 0.3339 0.0015 0.0016 0.4841 1.0315 0.0023 0.3820 0.9468 -0.0012 0.7597 1.1113 0.0117 0.0204 1.1484 0.0216 0.0002 1.2076 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.4653 0.4477 0.5023 0.5789 0.5814 0.5353 0.5471 -0.0040 0.0366 1.3421 -0.0104 0.0002 1.3929 -0.0215 -0.0293 -0.0414 -0.0528 -0.0721 0.0000 0.0000 0.0000 0.0000 0.0000 1.1558 0.0011 0.3722 1.2467 1.4462 1.4926 1.4444 1.3332 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5057 0.5321 0.5405 0.5617 0.4901 0.4773 0.4521 0.3939 -0.0033 0.0473 -0.2837 0.0146 0.0295 0.0052 0.0442 -0.4255 0.0003 0.0582 0.0120 0.0014 -0.3614 0.0002 0.0442 0.0238 0.0001 -0.4994 0.9667 -0.0022 0.0461 0.9631 0.0012 0.4264 0.9167 0.0000 0.0000 0.4248 0.0018 -0.0034 0.0000 -0.1891 0.0004 0.0133 -0.0016 0.0000 0.0000 0.0000 37 0.0531 0.0000 0.1103 0.0614 0.0018 -0.4827 0.0161 0.0292 -0.0109 0.6590 -0.4256 0.0099 0.0241 0.9491 -0.0004 0.9562 1.0003 -0.0280 0.0010 1.0603 0.0000 0.0000 0.0000 0.5137 0.5703 0.6276 1.4430 -0.0618 0.0001 1.4830 -0.0171 0.3786 1.4859 0.0000 0.0000 0.0000 0.3921 0.3006 0.3343 0.0000 -0.4491 0.0018 0.0427 0.0316 0.0000 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.4: Abnormal Returns Using CAPM Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: (1) CAPM: Dependent variable is periodic return for the WM L which is the winner R» ~ R F , = a i vb, R M RF minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or e x c e ss periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic Treasury bill rate over the listed formation and holding period. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sam ple period and held for the indicated period of time. (4) RMRF is the e x c e ss return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson A ssociates) over the listed formation and holding periods. (5) p-values are computed with robust standard errors corrected for heteroskedasticilty and autocorrelation us ng the Newey-W est adjustmen (1987). 38 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.5: Abnormal Returns Using Fama & French's Three Factor Model Exchange Traded Funds and Momentum Relturns: March 21,1996 - December 31, 2005 1day1day 1wk 1wk WML: Constant p - value Mkt - RF p - value HML p - value SMB p - value Adj. R-Sq. Winner: Constant p - value M kt-RF p - value HML p - value SMB p - value Adj. R-Sq. Loser: Constant p - value Mkt - RF p - value HML p - value SMB p - value Adj. R-Sq. -0.0035 0.0000 -0.1027 0.0336 0.2287 0.0435 -0.0451 0.5242 0.0175 2wk 2wk 4wk 4wk 8wk 8wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39 wl 52wk 52wk -0.0038 0.0213 -0.1471 0.1402 0.3671 0.0586 -0.0570 0.7585 0.0434 0.0043 0.1056 -0.3163 0.0026 0.2941 0.1246 0.0118 0.9418 0.0639 0.0090 0.0164 -0.2629 0.0044 0.4036 0.0214 0.3598 0.0225 0.0701 0.0201 0.0006 -0.5391 0.2350 0.1122 0.5484 0.0004 0.1188 0.0237 0.0019 -0.4408 0.0002 0.1712 0.1890 0.4965 0.0015 0.0773 0.0009 0.5395 0.9248 0.0001 0.9543 1.0525 0.0050 0.8435 0.8985 -0.0008 0.8364 1.0169 0.0000 1.0200 -0.0028 0.0264 0.9827 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1036 0.1309 0.2726 0.0781 0.4964 0.1035 0.4683 0.4581 0.0656 0.5272 0.1461 0.1239 0.4503 0.1855 0.0670 0.2965 0.0074 0.5206 0.1126 0.0652 0.3746 -0.0381 0.5512 0.2881 0.0014 0.2342 1.1397 -0.0033 0.0981 1.2411 -0.0089 0.0020 1.3154 -0.0017 0.0000 0.0000 0.4334 0.0018 0.0000 1.1228 0.0000 0.0000 0.0000 0.0000 -0.1252 0.1478 0.3177 -0.2718 0.0531 0.1608 0.2224 0.5450 -0.2285 0.0828 0.1342 0.2567 0.5494 -0.2181 0.0470 -0.0633 0.5327 0.5645 0.0000 0.5209 0.0407 0.0521 0.0000 0.0000 -0.1869 0.0977 0.3991 0.0089 0.2896 0.0582 0.0467 0.0648 0.6024 0.5558 0.0001 0.3417 0.0482 0.0504 0.0180 0.0015 1.1550 0.0000 0.0498 0.4950 0.3397 0.0192 0.0019 0.9686 0.0000 0.0000 0.6250 0.6039 0.0109 0.0292 1.0901 0.0000 0.0027 0.9734 0.2480 0.0017 0.5476 -0.0196 0.0001 1.4376 0.0000 -0.1224 0.2572 -0.1739 0.1590 0.4922 -0.0245 0.0002 1.4577 -0.0298 0.0001 1.2770 39 0.0642 0.0000 -0.2173 0.1044 0.6994 0.0632 0.0083 -0.5644 0.0049 -0.0824 0.5682 0.2071 0.3222 0.0293 -0.0001 0.9985 -0.2791 0.0919 0.1283 0.4658 -0.6845 0.0004 0.0617 -0.0400 0.0001 1.1184 0.1675 0.0003 0.4006 -0.0045 0.5465 0.9206 0.0000 -0.0442 0.4470 0.3461 0.0000 0.0000 0.0000 0.5688 0.5696 0.6117 0.0944 0.1065 0.0730 0.1950 0.6311 -0.0341 0.0002 1.0902 -0.0450 0.0002 1.1859 0.0000 -0.5319 -0.0677 0.0008 1.4851 -0.0399 0.1332 1.3975 0.0000 0.0000 0.0000 -0.2093 0.0306 -0.2084 0.0994 0.4831 -0.3964 -0.5060 0.0000 0.5045 0.0058 0.1431 0.0000 0.0000 0.0000 0.0000 -0.0417 0.7197 0.4679 -0.0020 0.9875 0.4269 -0.1039 0.4983 0.4279 0.0000 0.0000 0.0000 0.0382 0.7374 0.1390 0.4541 0.2990 -0.0338 0.7966 0.7575 0.0000 0.3814 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.5: Abnormal Returns Using Fama & French's Three Factor Model Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: SMB, + • HML t + e it (1) Fama & French's three factor model: R U ~ R Ft ~ a t +bj ■RI\IRF , + s, • Dependent variable is periodic return for the WML which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or ex c ess periodic return for the winner or loser portfolio above the risk free rate as proxied by the appropriate periodic Treasury bill rate over the listed formation and holding period. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sam ple period and held for the indicated period of time. (4) RMRF is the e x c e ss return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson Associates) over the listed formation and holding periods. (5) HML (High Book to Market Minus Low Book to Market) is the average return on the two value portfolios minus the average return on the two growth portfolios, i.e. HML = 1/2 (Small Value + Big Value) -1 /2 (Small Growth + Big Growth). (6) SMB (Small Size Minus Big Size) is the average return on the three small portfolios minus the average return on the three big portfolios, SMB = 1/3 (Small Value + Small Neutral + Small Growth) -1 /3 (Big Value + Big Neutral + Big Growth). (7) p-values are computed with robust standard errors corrected for heteroskedasticily and autocorrelation using the Newey-W est adjustment (1987). 40 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.6: Annualized Abnormal Returns with Extra Time between Formation and Holding Periods: Exchange Traded Funds and Momentum Relturns: March 21,1996 - December 31,2005 1day 1day 1wk 1wk CAPM: WML: Constant p - value Winner: Constant p - value Loser: Constant p - value 2wk 2wk 4wk 4wk 8wk 8wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39 wl 52wk 52wk , R H ~ R F t = a , + b • RMRF , + e,t -24.5250% 0.0036 12.3656% 0.2017 17.1808% 0.0092 13.3276% 0.0055 -16.9000% 0.0011 -1.5444% 4.0820% 0.2785 0.2574% 0.9285 14.4944o/0 12.1299% 0.0001 0.0001 18.4022% 0.0000 19.3378% 0.0000 21.5366% 0.0000 7.4336% 0.0046 -1.5986% 0.5188 4.7258% 0.0048 5.8950% 0.0001 6.1390% 0.0000 -0.4219% 0.6644 -2.9541% 0.0005 7.6250% -13.9152% -13.0988% -13.0702% -13.6130% -12.5658% -13.6767% -13.4430% -14.0114% 0.2460 0.0085 0.0003 0.0000 0.0000 0.0466 0.0000 0.0000 0.0000 -7.8555% 0.0002 -1.3556% 0.4853 12.4942% 0.0000 7.7425% 0.0147 -0.4364% 0.8938 3.7066% 0.0026 -1.0445% 0.2798 -4.1208% 0.0000 -8.7876% 0.0003 -8.7871% 0.0011 -3.6844% 0.1644 0.7835 0.8814% 0.6134 -0.4359% 0.7972 Fama - French's 3 Factor Model: R il ~ RFt = a t + bj ■R14RF , + Sj • SMB , + h , • HML , + e WML: Constant -24.6250% 11.4868% 13.8242% 10.7848% 12.3526% 10.3363% 14.2090% 13.5361% 0.0384 0.0045 0.0202 0.0010 0.0000 p - value 0.2266 0.0013 0.0000 Winner: 2.4882% -1.1011% -0.0897% -0.3800% Constant -17.7500% -1.5808% 4.1031% 4.9379% 0.8191 0.0007 0.7777 0.5015 0.7130 0.9565 0.0125 p -v a lu e 0.0009 Loser: Constant 6.8750% -13.0624% -11.3360% -11.8859% -12.4423% -10.7163% -10.1059% -8.5982% 0.0294 0.0011 0.0001 0.0001 0.0000 0.0004 p - value 0.3118 0.0589 41 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.6: Annualized Abnormal Returns with Extra Time between Formation and Holding Periods: Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: (1) Annualized abnormal returns for the period indicated. Periodic abnormal returns, i.e. alpha intercepts, are annualized by multiplying by either 250 trading days for 1 day 1 day strategy or by 52 divided by the number of w eeks in the 10 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period of time. (4) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for the indicated holding period. (5) An extra day is added between the formation and holding period for the 1 day 1 day strategy. For all other strategies with formation and holding periods from 1 w eek to 52 w eeks, an extra week is added between the formation and holding period s. (6) p-values are computed with robust standard errors corrected for heteroskedasticily and autocorrelation using the Newey-W est adjustmen (1987). 42 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.7: Annualized Abnormal Returns Net of Transactions Costs for WML with 26 Week Formation and Holdin g Period: Exchange Traded Funds and Momentum Retturns: March 21,1996 - December 31, 2005 Model: CAPM Fama & French's three factor model Notes: Annualized Abnormal Return 22.06% 12.83% Transactions Costs 8.25% 8.25% Net Annualized Abnormal Return 13.81% 4.58% (1) Annualized abnormal returns for the period indicated. Periodic abnormal returns, i.e. alpha in ercepts, are annualized oy multiplying by two to convert the 26 week abnormal return to an annualized abnormal return. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period of time. (4) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for the indicated holding period. (5) Actual transactions were tabulated by domestic, sector, or international ETF over the 9.29 years studied. (6) Quoted bid ask spreads were estimated as the higher of those identified in Huang & Wei (2004) and Salomon Smith Barney (2002) resulting in estim ates of 0.33% for domestic ETFs, 0.62% for sector ETFs, and 0.867% for international ETFs. (7) Comm issions were estimated at 0.13% by combining Scottrade's $7.00 per trade with an estimated account balance of $125,000 invested long in the winner ETFs and $125,000 invested short in the loser ETFs. Scottrade's flat commission ap plies to both market and limit orders regardless of trade frequency, account balance, or number of shares in the transaction. 43 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.8: Abnormal Returns Divided into Two Periods: 03/21/96 -12/31/00 vs. 01/01/01 -12/31/05 Exchange Traded Funds and Momentum Rel urns: 26 week formation and holding period. Fama & French's Three Factor Model: ICAPM: 1996-2000 2001-2005 1996-2000 2001-2005 WML: WML: 0.1254 -0.0253 Constant 0.0613 0.1768 Constant 0.0013 p - value 0.0000 0.0000 0.0000 p - value -0.1589 -0.3750 Mkt RF -0.6999 -0.3490 Mkt - RF 0.0000 p - value 0.5710 0.0000 0.0011 p - value 1.4277 HML 0.7123 0.0804 0.0609 Adj. R-Sq. 0.0000 0.0000 p - value 0.5614 SMB 0.7772 p - value 0.0001 0.0000 0.1365 0.5730 Adj. R-Sq. Winner: Winner: Constant 0.0301 0.0111 0.0216 0.0455 Constant 0.0052 0.0059 p - value 0.0000 0.0181 p - value 1.1170 0.6749 Mkt - RF 0.7671 1.1733 Mkt - RF 0.0000 0.0000 p - value 0.0000 0.0000 p - value -0.0641 0.4656 HML 0.8224 0.3102 Adj. R-Sq. 0.2307 0.0000 p - value 0.2607 0.3770 SMB 0.0014 0.0000 p - value 0.3399 0.8850 Adj. R-Sq. Loser: Loser: 0.0365 Constant -0.0953 -0.0128 -0.1376 Constant 0.0000 0.0000 0.0432 p - value 0.0000 p - value 1.4919 Mkt - RF 0.8339 1.5357 Mkt - RF 1.4849 0.0007 p - value 0.0000 0.0000 0.0000 p - value -0.9621 -0.7764 HML 0.2924 0.7856 Adj. R-Sq. 0.0000 p - value 0.0000 -0.1844 -0.5165 SMB 0.0325 p - value 0.0020 0.3568 0.8750 Adj. R-Sq. 44 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.8: Abnormal Returns Divided into Two Periods: 03/21/96 - 12/31/00 vs. 01/01/01 • 12/31/05 Exchange Traded Funds and Momentum Relturns: 26 week formation and holding period. Notes: a t + bt •R k (RLt'+eit (1) CAPM: R it R Ft ~ (2) Fama & French's three factor model: R « - R f , = a , + bt ■Ri\4RF, + s t ■SMB, + h • HM L, + e (/ (3) Dependent variable is periodic return for the WML which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the 26 week formation and holding periods, or ex c ess periodic return for the winner or loser portfolio above the risk free rate as proxied by the appropriate periodic Treasury bill rate over the 26 week formation and holding period. (4) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (5) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period of time. (6) RMRF is the ex c ess return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson Associates) over the listed formation and holding periods. (7) HML (High Book to Market Minus Low Book to Market) is the average return on the two value portfolios minus the average return on the two growth portfolios, i.e. HML = 1/2 (Small Value + Big Value) -1 /2 (Small Growth + Big Growth). (8) SMB (Small Size Minus Big Size) is the average return on the three small portfolios minus the average return on the three big portfolios, SMB = 1/3 (Small Value + Small Neutral + Small Growth) -1 /3 (Big Value + Big Neutral + Big Growth). (9) p-values are computed with robust standard errors corrected for heteroskedasticily and autocorrelation using the Newey-W est adjustmen (1987). 45 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abnormal Returns Using CAPM and Fama & French's Three Factor Model Table 1.9: Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 CAPM: WML: Constant p - value Mkt- RF p - value HML p - value SMB p - value DumSect p - value Dumlnt p - value Adj. R-Sq. Winner: Constant p - value Mkt- RF p - value HML p - value SMB p - value DumSect p - value Dumlnt p - value Adj. R-Sq. 26wk 26wk 24wk 24wk FF3F: 4wk Rebalancing No Reba ancing No Dum. Incl. Dum. No Dum. Incl. Dum. 0.0170 0.0175 -0.5611 0.0029 0.0783 0.0000 -0.5578 0.0030 0.0427 -0.1075 0.0006 -0.1311 0.0028 0.0937 0.0887 -0.0195 0.2918 -0.0500 0.0055 0.0985 0.0316 0.0000 0.9491 0.0000 0.0142 0.3616 1.0049 0.0000 0.0034 0.3603 1.0876 0.0000 -0.0067 0.5921 1.0901 0.0000 0.5137 0.0370 0.0062 -0.0066 0.6232 0.5240 0.6070 0.0048 0.5866 0.0078 0.5052 0.6024 0.1103 0.0000 -0.4491 0.0018 0.3035 0.0000 -0.6105 0.0000 24wk 24wk 26wk 26wk 4wk Rebalancing No Reba ancing No Dum. Incl. Dum. No Dum. Incl. Dum. 0.0642 0.0000 -0.2173 0.1044 0.6994 0.0000 0.5045 0.0058 0.1431 0.0192 0.0019 0.9686 0.0000 0.1675 0.0003 0.4006 0.0000 0.5696 46 0.2234 0.0000 -0.5151 0.0000 0.9840 0.0000 1.1623 0.0000 -0.2585 0.0000 -0.0022 0.9587 0.2930 0.0124 0.0935 -0.3915 0.0273 0.6155 0.0082 0.5375 0.0450 -0.0112 0.4922 0.9595 0.0000 0.1938 0.0000 0.4317 0.0000 -0.0042 0.8057 0.0341 0.0415 0.5705 0.0014 0.7150 1.1392 0.0000 0.2606 0.0368 0.3357 0.0027 0.1343 0.6301 0.0545 0.0037 -0.4279 0.0168 0.6431 0.0097 0.6535 0.0193 -0.0321 0.1039 -0.0217 0.2752 0.1544 -0.0145 0.2714 1.1450 0.0000 0.3157 0.0317 0.3739 0.0049 -0.0026 0.7996 0.0184 0.1556 0.6294 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abnormal Returns Using CAPM and Fama & French's Three Factor Model Table 1.9: Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2(>05 CAPM: Loser: Constant p - value Mkt- RF p - value HML p - value SMB p - value DumSect p - value Dumlnt p - value Adj. R-Sq. FF3F: 24wk 24wk 26wk 26wk 4wk Rebalancing No Rebalancing No Dum. Incl. Dum. No Dum. Incl. Dum. -0.0721 0.0000 1.4430 0.0000 -0.1661 0.0000 1.6171 0.0000 0.3921 0.1351 0.0000 0.0012 0.9480 0.4402 -0.0136 0.0147 1.6487 0.0000 -0.0428 0.0052 1.6716 0.0000 0.5913 0.0267 0.0659 0.0137 0.1195 0.6006 26wk 26wk 24wk 24wk 4wk Rebalancing No Reba ancing No Dum. Incl. Dum. No Dum. Incl. Dum. -0.0450 0.0002 1.1859 0.0000 -075319 0.0000 -0.1039 0.4983 0.4279 -0.2380 0.0000 1.3995 0.0000 -0.8358 0.0000 -0.7177 0.0000 0.2444 0.0000 0.0569 0.0033 0.5391 -0.0110 0.0689 1.5307 0.0000 -0.3549 0.0407 -0.2018 0.3483 0.5954 -0.0435 0.0037 1.5561 0.0000 -0.4209 0.0118 -0.3241 0.1181 0.0329 0.0262 0.0135 0.1019 0.6115 ...................i .... ....... 47 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.9:j Abnormal Returns Using CAPM and Fama & French's Three Factor Model Exchange Traded Funds and Momentum Returns: IVarch 21,1 996 - December 31, 2005 Notes: (1) CAPM: R-it R-Ft d-a, +b, R MRIf+e,, (2) Fama & French's three factor model: R„ - R F, = a + b, ■RM i W t + sr S,\tB ,+hr HML,+eu (3) Dependent variable is periodic return for the WML which is the winner minus loser porltfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or e x c e s s periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic Treasury bill rate over the listed formation and holding period. (4) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (5) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with eq ual weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period of time for the No Rebalancing. (6) The 24 week, 4 week Rebalancing portfolios represent six equally weighted portfolios formed from the winner and loser ETFs from today, 4 w eeks ago, 8 w eeks ago, 12 w eeks ago, 16 w eeks ago, and 20 w eeks ago. Thus, the winner and loser returns represent 4 week returns. (7) RMRF is the e x c e ss return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson A ssociates) over the listed formation and holding periods. (8) HML (High Book to Market Minus Low Book to Market) is the average return on the two value portfolios minus the average return on the two growth portfolios, i.e. HML = 1/2 (Small Value + Big Value) -1/ 2 (Small Growth + Big Growth). (9) SMB (Small Size Minus Big Size) is the average return on the three small portfolios minus the average return on the three big portfolios, SMB = 1/3 (Small Value + Small Neutral + Small Growth) - 1/3 (Big Value + Big Neutral + Big Growth). (10) DumSect is a dummy variable that equals 1 if the WML, Winner, or Loser portfolio includes at least one Sector ETF, and it equals 0, otherwise. (11) Dumlnt is a dummy variable that equals 1 if the WML, Winner, or Loser portfolio includes at least one i International ETF, and it equals 0, otherwise. l (12) p-values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using the Newey-W es [adjustment (1987). 48 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abnormal Returns Using 26 week Formation and Holding Periods Table 1.10: Exchange Traded Funds and Momentum Returns: January 1 , 1999 - December 31,2005 CAPM: WML: Constant p - value Mkt - RF p - value Adj. R-Sq. Winner: Constant p - value Mkt-RF p - value Adj. R-Sq. Loser: Constant p - value Mkt- RF p - value Adj. R-Sq. Types of ETFs Includ ed Domestic Internat’l All Types Bond Sector 0.0715 0.0000 -0.4067 0.0000 0.0883 0.0205 0.0004 -0.1696 0.0002 0.0764 0.0787 0.0000 -0.2586 0.0237 0.0343 0.0070 0.3447 0.1572 0.0167 0.0199 0.1138 0.0000 -0.6491 0.0000 0.0956 0.0370 0.0000 1.1298 0.0000 0.7507 0.0231 0.0000 -0.0692 0.1022 0.0088 0.0444 0.0000 1.0236 0.0000 0.5681 0.0517 0.0000 1.2614 0.0000 0.6307 0.0306 0.0000 1.1514 0.0000 0.6147 -0.0290 0.0000 1.5480 0.0000 0.7480 0.0038 0.2017 0.1037 0.0008 0.0438 -0.0309 0.0000 1.3041 0.0000 0.7711 0.0426 0.0000 1.1137 0.0000 0.6545 -0.0744 0.0000 1.8204 0.0000 0.5956 49 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1.10: Abnormal Returns Using 26 week Formation and Holding Periods Exchange Traded Funds and Momentum Returns: January 1 , 1999 - December 31, 2005 FF3F: Types of ETFs Includ ed All Types Bond Domestic Internat'l Sector WML: 0.0177 -0.0077 Constant -0.0169 -0.0065 -0.0230 p - value 0.0880 0.0176 0.3985 0.4869 0.1296 -0.0160 -0.4614 0.0980 Mkt - RF -0.5199 -0.6963 p - value 0.0000 0.8779 0.0000 0.1951 0.0000 -0.0597 0.4910 HML 0.5668 0.0781 0.9995 0.6054 p - value 0.0000 0.0000 0.2383 0.0000 SMB 1.2073 -0.2130 1.3176 0.2713 1.6787 0.0000 0.1134 p - value 0.0000 0.0296 0.0000 0.4194 0.0872 0.3951 0.0351 0.4223 Adj. R-Sq. Winner: 0.0071 -0.0098 0.0034 0.0195 Constant 0.0016 0.1467 p - value 0.1979 0.6395 0.0188 0.8314 Mkt- RF 1.1181 0.3147 1.2354 0.9048 1.2738 0.0012 p - value 0.0000 0.0000 0.0000 0.0000 0.2284 HML 0.4040 0.2149 0.2336 0.3492 0.0000 0.0000 0.0069 p - value 0.0001 0.0000 SMB 0.6932 0.4002 -0.4391 0.4502 0.1982 p - value 0.0001 0.0012 0.0001 0.0013 0.1333 0.7923 0.6629 Adj. R-Sq. 0.2491 0.6548 0.6570 Loser: Constant 0.0240 -0.0276 0.0110 0.0260 0.0247 0.0000 p - value 0.0018 0.0060 0.0001 0.0433 0.3307 1.1374 Mkt - RF 1.6380 1.3662 1.9700 p - value 0.0000 0.0000 0.0000 0.0000 0.0000 HML -0.3384 0.4636 -0.2761 0.1556 -0.6503 p - value 0.0000 0.0000 0.0001 0.0104 0.0000 SMB -0.6244 -0.8070 -0.2261 0.1789 -1.4805 p - value 0.0000 0.0002 0.0000 0.1060 0.0000 0.2818 0.8384 0.6648 Adj. R-Sq. 0.8229 0.7460 50 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abnormal Returns Using 26 week Formation and Holding Periods Table 1.10: Exchange Traded Funds and Mlomentum Returns: January 1 , 1999 - December 31, 2005 Notes: (1) CAPM: R-u Rpt ~ a , +br R.\4RIf+ elt (2) Fama & French's three factor model: Ru = « , + b, •RMRF, + s, ■SMB , +h, ■HMi + e u (3) Dependent variable is periodic return for the WML which is the winner minus loser poiitfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or e x c e s s periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic Treasury bill rate over the listed formation and holding period. (4) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (5) The winners represent the top decile of ETF returns available during the formation period; the losers represen the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period of time. (6) All sample periods include January 1, 1999 through December 31, 2005 except for the Bond ETFs which include the sample period from July 27, 2002 to December 31, 2005. (7) RMRF is the e x c e ss return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson A ssociates) over the listed formation and holding periods. (8) HML (High Book to Market Minus Low Book to Market) is the average return on the two value portfolios minus the average return on the two growth portfolios, i.e. HML = 1/2 (Small Value + Big Value) - 1/ 2 (Small Growth + Big Growth). (9) SMB (Small Size Minus Big Size) is the average return on the three small portfolios minus the average return on the three big portfolios, SMB = 1/3 (Small Value + Small Neutral + Small Growth) - 1 / 3 (Big Value + Big Neutral + Big Growth). (10) p-values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using the Newey-West adjustment (1987). 51 ESSAY #2 MOMENTUM ABNORMAL RETURNS AND INTERNATIONAL EXCHANGE TRADED FUNDS Abstract: Investing in portfolios of international exchange traded funds (ETFs) provides risk adjusted abnormal returns using either Lo & MacKinlay’s difference from average performance weights or Jegadeesh & Titman’s equal weights within the top and bottom deciles. A short formation and holding period o f one week provides abnormal contrarian returns as past losers become winners and past winners become losers. Medium formation and holding periods of four weeks to twenty six weeks provide abnormal momentum returns as past winners keep winning and past losers keep losing. Abnormal returns result for portfolios of international ETFs when returns are adjusted for risk using the Capital Asset Pricing Model and Fama & French’s three factor model. High trading volume increases the momentum abnormal returns for formation and holding periods of 4 weeks and 26 weeks; low trading volume increases the contrarian abnormal returns for a formation and holding period of 1 week. 1. Introduction: Momentum, the process of buying portfolios of recent winners and shorting portfolios of recent losers to earn abnormal returns, continues to be an anomaly in many markets and continues to be an investment strategy followed by a variety of investors including individuals, mutual funds, and other money managers. Since Jegadeesh & Titman’s (1993) groundbreaking documentation of medium term momentum profits for U.S. stocks over 3 to 12 month holding periods, numerous studies have confirmed momentum profits in many markets including: U.S. equities (Jegadeesh & Titman (1993, 2001), Hong, Lim, & Stein (2000)), U.S. mutual funds (Grinblatt, Titman, & Wermers (1995), Carhart (1997), Wermers (2003), Sapp & Tiwari (2004)), U.S. industries 52 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (Moskowitz & Grinblatt (1999)), U.S. exchange traded funds (ETFs) (De Jong (2007a)), international equity markets (Rouwenhorst (1998, 1999), Chan, Hameed, & Tong (2000), Balvers & Wu (2006)), and foreign exchange markets (Okunev & White (2001)), but excluding U.S. investment grade corporate bonds Gebhardt, Hvidkjaer, & Swaminathan (2002). As an anomaly, momentum merits further study since such future return predictability in past return patterns violate the weak form of market efficiency, and since such medium term return continuation confounds the benchmark asset pricing model, namely Fama & French’s three factor model. The momentum anomaly serves as a new battleground between the long time academic forces of rational market efficiency and the more recent but growing forces of behavioral finance. Both sides claim victory in explaining why the momentum anomaly occurs. Behavioral finance posits various explanations o f momentum, including investor overreaction or underreaction (Chan, Jegadeesh, & Lakonishok (1996), Hong & Stein (1999)), expectation extrapolation, conservatism in expectation updating (Barberis, Shleifer, Vishny (1998)), biased self attribution or investor overconfidence (Daniel, Hirshleifer, & Subrahmanyam (1998)), disposition effects (Grinblatt & Moskowitz (2004)), selective information conditioning, and herding behavior by investors (Jordan (2004)) and mutual fund managers (Grinblatt, Titman, & Wermers (1995), Wermers (2003)). Rational, market efficiency supporters fire back with the alternative explanations o f momentum returns being unrealizable due to transactions costs (Lesmond, Schill, & Zhou (2004)) or being fair compensation for risk (Conrad & Kaul (1998)) or being the product of data mining. Clearly, the momentum anomaly requires 53 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. further research, as current studies agree that a momentum investment strategy generates abnormal returns, but disagree as to what causes a momentum strategy to be successful and disagree as to whether the abnormal profits are realistically attainable by an actual investor or arbitrageur after recognizing appropriate transactions costs, like bid ask spreads, brokerage commissions, the price impact of large trades, and the higher capital gains taxes associated with increased trading. Part of the motivation of researching momentum in international stocks and country indices is to dispel the notion that momentum is the product of data mining in the U.S. securities’ data sets by identifying momentum abnormal returns in international markets responsive to non-U.S. local factors and controlled by differing government policies and financial institutions and structures. While country indices provide potential measures of investment returns in various countries, unfortunately, they do not provide readily investable benchmark returns attainable by a normal investor or arbitrageur when considering transactions costs, restrictions on short selling, and restrictions placed on foreign investors, unless there exists an ETF designed to passively match the corresponding country index. Chan, Hameed, & Tong (2000) find momentum returns in 23 country index returns during 1980 to 1995 o f at least 1% per month for formation and holding periods o f 1 week, 2 weeks, or 4 weeks. They use Lo & MacKinlay’s (1990) weighting scheme where portfolio weights reflect the country’s past performance relative to the average past performance of all 23 countries; above average performers are purchased and below average performers are shorted, but all countries may have a non-zero weight in the 54 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. winner minus loser portfolio. This weighting scheme differs from Jegadeesh & Titman’s usual methodology of buying long the top decile of winners and shorting the bottom decile o f losers. Chan, Hameed, & Tong find that 80 to 90% of the momentum profits are due to equity return predictability and only a small portion is due to exchange rate predictability, and they find that momentum profits are larger for country indices following an increase in trading volume. Their momentum profits would be difficult to obtain after transactions costs as country indices are not directly investable and some of their markets restrict short selling. Balvers & Wu (2006) jointly consider momentum and mean reversion for 18 developed country index returns from 1970 to 1999. Their model generates a signal that identifies the winner and loser countries by its indicator score incorporating both momentum and mean reversion information, resulting in excess returns o f 1.1 - 1.7% per month, which outperforms both pure momentum and pure mean reversion strategies. Balvers & Wu find a strong negative correlation of - 35% between momentum and mean reversion effects, which explains why controlling for mean reversion effects can improve momentum returns. The study o f Chan, Hameed, & Tong (2000) can be improved by using ETFs rather than country indices since ETFs indexing various country stock indices are readily investable and shortable which are necessary conditions to realistically implement a momentum or a contrarian strategy. My motivation in this study is to extend the domain of momentum in international equities and country indices to a relatively new investment vehicle, namely, exchange traded funds or ETFs. ETFs are powerful and flexible investment vehicles that combine the diversified portfolio features of mutual funds with the trading possibilities of 55 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. individual securities. Currently, ETFs function similarly to passively managed index mutual funds, as they are composed of a portfolio of stocks or bonds that track a particular index, thus providing diversification within the portion of the market tracked by that index. Four general categories of ETFs include: (1) broad based domestic indices like the S&P 500, the NASDAQ 100, the Dow Jones U.S. Total Market, the Russell 3000, the Wilshire 5000, some style specific indices in both a “value” and a “growth” version, and size based indices including large cap, mid cap, small cap, and micro cap, (2) sector indices including consumer, energy, financial, health, natural resources, real estate, utilities, and technology, (3) international indices including global stock indices, regional indices, and country specific indices, and (4) bond indices including three o f the Lehman Treasury bond indices, two different corporate bond indices, and the Lehman TIPS index. What differentiates an ETF from a mutual fund is an ETF trades on an exchange (most on the AMEX) like a stock, enabling an ETF to be: purchased or sold at intraday market prices, purchased on margin, sold short, and traded via stop orders and limit orders. Ordinary mutual funds can only be purchased and sold by market orders for end of day prices, and cannot be purchased on margin or sold short, which prevents the usual zero investment momentum and contrarian portfolios of buying the winners and shorting the losers or of buying the losers and shorting the winners, respectively. Also, many mutual funds have redemption fees and other constraints to discourage or prevent the short term trading necessary to implement a momentum or contrarian strategy. For implementing a momentum or contrarian strategy, purchasing or shorting an ETF gives the arbitrageur or investor a diversified portfolio of stocks while incurring only one bid ask spread and one 56 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. round trip commission, clearly a cost advantage over paying the many bid ask spreads and many commissions to assemble a portfolio of individual winner and loser stocks to replicate the winner and loser country indices. My research questions are: (1) whether a momentum investing strategy of buying winners and shorting losers generates abnormal returns in the international ETF market, (2) whether the differing methodologies of Lo & MacKinlay (1990) and Jegadeesh & Titman (1993) affect the abnormal momentum returns in the international ETF market, (3) which formation period, holding period is optimal for momentum investing in the international ETF market, and (4) whether trading volume affects the abnormal momentum returns in the international ETF market. My results contribute to the literature an affirmative answer to the first research question about momentum, as buying the winner international ETFs and shorting the loser international ETFs provide statistically significant momentum abnormal returns for formation and holding periods of 4 weeks, 12 weeks, and 26 weeks with risk adjustment by either CAPM or Fama & French’s three factor model. The annualized momentum abnormal returns range from 5.1% to 11.8% under the Capital Asset Pricing Model and range from 4.7% to 10.2% under Fama & French’s three factor model. The formation and holding period of 1 week provides statistically significant contrarian abnormal returns, for buying the loser international ETFs and shorting the winner international ETFs, with annualized magnitudes o f 12.2% to 18.9%. My results contribute to the literature a negative answer to the second research question about methodologies, as both Lo & MacKinlay and Jegadeesh & Titman’s weighting schemes of winners and losers identify statistically 57 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. significant risk-adjusted momentum abnormal returns for the same formation and holding periods with identical signs and similar magnitudes. Following both methodologies, I find for question (3) that a 4 week formation and holding period provides the highest annualized abnormal returns of 11.1 % or 11.8% to an ETF momentum strategy using CAPM to adjust for risk and annualized abnormal returns of 9.2% or 10.2% to an ETF momentum strategy using Fama & French’s three factor model to adjust for risk. Also, a 1 week formation and holding period provides the highest abnormal returns to an ETF contrarian strategy, with annualized abnormal returns of over 12.2% using Lo & MacKinlay’s methodology and over 18.3% using Jegadeesh & Titman’s methodology. For question (4), my results show that high trading volume improves the risk-adjusted momentum abnormal returns for 4 week and 26 week formation and holding periods, but low trading volume improves the risk-adjusted contrarian abnormal returns for a 1 week formation and holding period. These research questions are important, because momentum and contrarianism are both wide spread anomalies identified by researchers as well as investment strategies used by practitioners like mutual funds and individual investors to attempt to earn abnormal returns. With the growth in ETFs from their introduction in 1993 with the SPDR Trust Series tracking the S&P 500 Index to 2005’s assortment of 217 ETFs consisting of 80 broad based domestic, 82 domestic sector, 49 global/international equity, and 6 bond ETFs, ETFs are on a growth path which should soon surpass the dollar amount invested in equity index mutual funds. From the Investment Company Institute’s (a mutual fund trade organization) December 2005 statistics, ETFs (excluding Merrill Lynch’s 58 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. HOLDRS) represent a market value of $296.02 billion, which represents over 5% of the $5,504.50 billion invested in stock and hybrid mutual funds. Considering that about 10% of stock mutual fund investments are in indexed investments as opposed to actively managed funds, ETFs represent a significant portion (almost 35%) of the U.S. wealth invested in passively managed, index type investment vehicles. With the growing popularity of ETFs by traders and investors such an innovative financial product merits further study, especially when it can generate abnormal returns via a momentum or contrarian strategy. Clearly, international ETFs represent not only a low cost way to diversify a domestic portfolio into international markets, but also a viable trading vehicle to implement a momentum or contrarian investment strategy after using past returns to identify the winners and losers. 2. Methodology: I use two methodologies in parallel that dominate the momentum/contrarian literature: Jegadeesh & Titman’s methodology (1993) which defines the winner as the top performing decile and the loser as the poorest performing decile over the various formation periods, and Lo and MacKinlay’s (1990) methodology which defines winners and losers relative to the average performance of the equally weighted market index with winners outperforming the average and losers underperforming the average over the various formation periods. Since most studies use one or the other methodologies, but not both, my consistent use o f both methodologies in parallel should illustrate any differences between the weighting schemes and provide more powerful results supported 59 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. by two different methodologies. Since Jegadeesh & Titman exclude the 80% o f the middle performing securities and focus on the 10% performance extremes, while Lo & MacKinlay include all securities at non - zero weights proportional to the individual security’s deviation from the average performance, a priori I expect Jegadeesh & Titman’s deciles to generate larger magnitude abnormal returns while I expect Lo & MacKinlay’s better diversified portfolios to generate less variability in the abnormal returns resulting in more statistically significant abnormal returns. Although Lo & MacKinlay’s weights are theoretically preferable in certain studies that decompose the momentum/contrarian abnormal returns into different components, the Jegadeesh & Titman weights are clearly easier to implement as equal weights within the top and bottom decile, as well as less costly to implement since investing in only the top and bottom decile reduces the number of transactions and therefore the transactions’ costs. Since international ETFs represent diversified portfolios designed and passively managed to track global stock indices, regional indices, and country specific indices, forming portfolios o f ETFs is less necessary than Jegadeesh & Titman’s method of forming decile portfolios o f individual stocks based on formation period performance. However, when Moskowitz & Grinblatt (1999) study industry momentum they define the winner and loser portfolios as the top or bottom 3 out of 20 industries, respectively, when ranked by formation periods of 1, 6, or 12 months, and measured holding period returns for periods of 1, 6,12, 24, or 36 months. Most momentum studies of mutual funds also group winner and loser funds into portfolios; at least with ETFs it is possible to short the losers while shorting mutual funds is not possible. With the relative newness of ETFs 60 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. such long periods as 24 or 36 months would restrict the sample sizes to be rather small, as the end of 1999 found only 17 international ETFs in existence. The years 2000 and 2001 brought the largest number of new international ETFs as 8 international ETFs were introduced during 2000, resulting in 25 international ETFs in existence by year end, and as 10 international ETFs were introduced during 2001, resulting in 35 international ETFs in existence by year end. With continued annual growth in the international ETF offerings, 45 international ETFs were in existence during 2004 and 2005. For the results following Jegadeesh & Titman’s methodology, I define the winner ETFs as the top performing decile over various formation periods and the loser ETFs as the poorest performing decile over various formation periods, and then form the momentum portfolio that buys the winner ETFs and shorts the loser ETFs over various holding periods. Also, adapting Jegadeesh & Titman’s methodology, I form the winner minus loser portfolio each week to increase the power of my tests; I equally weight the appropriate winner and loser ETFs in the portfolios formed each week during the sample period and held for the indicated amount of time. My sample period runs from March 19, 1996 to December 31,2005, a period of 483 weeks, with 17 international ETFs available in 1996, so that the top and bottom deciles begin with 1 or 2 ETFs each as winners and losers, respectively. In 2005,45 international ETFs are available so that the deciles of winners and losers both include 4 or 5 ETFs. For the results following Lo & MacKinlay’s methodology, I define the winner and loser ETFs relative to the average performance of an equally-weighted index of all available international ETFs during the formation period; winner ETFs outperform this 61 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. average performance and loser ETFs underperform this average performance. Since these international ETFs are all traded in the U.S. markets in dollars, no exchange rates are necessary, and all the ETFs are investable, liquid, and shortable, even on a downtick. With N international ETFs available in formation period t-1, the momentum portfolio 1 weight for ETF i in holding period t is: wu = — • N 1 N - Rml ]) where Rml } = — ■£ N is average return of the equally-weighted index formed from the N international ETFs available during formation period t-1. The Lo & MacKinlay weights are positive, indicating a long position, for the winner ETFs, whose returns during formation period t1 outperform the average return, i.e., (f?/M > Rml^ ), and negative, indicating a short position, for the loser ETFs, whose returns during formation period t-1 underperform the average return, i.e., (i?,M < Rmt_x). Since the portfolio weights are proportional to the difference between the return on ETF i and the average return of all available ETFs during period t-1, the ETFs whose returns most exceed or underperform the average in period t-1 will account for greater weight in the winner or loser portfolio, respectively. By construction, the winner minus loser (WML) portfolio is a zero-cost portfolio as the Lo & MacKinlay portfolio weights, in total, sum to zero, i.e., =o The total V 0: ^ mner = ^ ( w ,( Rit) and Inv\ i =1 i =1 w,.>0 w„>0 W in n e r N => HPRTnner = f L1^ , and (2) for losers where w„,)• RMRF, + e„ where: Dt is a dummy variable that equals one if the CRSP value-weighted market return is positive in holding period t and it equals zero otherwise, and all the other variables are defined as before. Panel A of Table 2.9 shows the results for the winner minus loser portfolio of international ETFs as well as the excess returns to both the winner and loser portfolios separately. My results are consistent with Rouwenhorst’s concern as the winners have higher betas than the losers in up markets and lower betas than the losers in down markets for all 5 formation and holding periods, but the market dependent betas do not eliminate the abnormal returns as all WML a 's remain statistically significant at the 5% level except for the formation and holding period of 2 weeks. Comparing Panel A o f Table 2.9 to CAPM’s results in Table 2.3, the market 78 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. dependent betas generally slightly reduce the magnitude and statistical significance of the abnormal returns. Panel B o f Table 2.9 shows the results allowing market dependent betas for the winner minus loser portfolios of international ETFs partitioned by high and low trading volume. The results are very similar to Table 2.6’s risk-adjustments by CAPM partitioned by high and low trading volume. For high volume WML portfolios, the momentum abnormal returns remain very significant at the 1% level for formation and holding periods of 4 weeks and 26 weeks. For low volume WML portfolios, the market dependent betas cause the momentum abnormal returns to be insignificant when using the Jegedeesh & Titman weights, but under the Lo & MacKinlay weighting, statistically significant contrarian abnormal returns remain for the 1 week formation and holding period and statistically significant momentum abnormal returns remain for the 26 week formation and holding period. 6. Robustness of Results: C. Growth in Number of ETFs: Another issue is the rapid growth in the number of ETFs. Were the momentum and contrarian returns determined by the early years when a smaller number of ETFs existed? Table 2.10 considers this question by dividing the sample period into the early period from March 20, 1996 to December 31, 2000 with 17 to 25 international ETFs in existence and the later period from January 1, 2001 to December 31, 2005 with 35 to 45 international ETFs in existence. I distinguish these two sample periods for portfolios 79 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. with a formation and holding period o f 26 weeks with risk adjusted by CAPM and Fama & French’s three factor model. My results support De Jong’s (2007a) findings with all four categories of ETFs; I find larger magnitude and more significant momentum abnormal returns occur under both risk adjustment models for the early period from 1996 to 2000 using international ETFs. However, during the later period from 2001 to 2005, significant momentum abnormal returns, at the 1% level with Lo & MacKinlay weights and at the 10% level with Jegadeesh & Titman weights, occur under CAPM, but contrarian abnormal returns occur under Fama & French’s three factor model which are insignificant with Lo & MacKinlay weights and significant at the 10% level with Jegadeesh & Titman weights. The 1996 to 2000 momentum abnormal returns are clearly driven by the return continuation of the loser portfolios, but during 2001 to 2005 the loser international ETFs exhibit very significant return reversals resulting in positive abnormal returns which reduce or eliminate the momentum abnormal returns generated by the winner international ETFs strong return continuations. One clear difference between the two periods is the increased level of systematic risk during the 2001 - 2005 period, especially in the winner ETFs. Also, winner ETFs include growth stocks, with low book to market ratios, during the 1996 - 2000 period but winner ETFs include value stocks, with high book to market ratios, during the 2001 - 2005 period. During the later period, loser ETFs shift to contain more small-cap stocks than the winner ETFs, rather than more large-cap stocks than the winners as during the early period. Dividing the sample period into two shorter periods illustrates some periodic differences between winner and loser ETFs in the two periods. However, such differences do not indicate a problem with the 80 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. number of ETFs in existence, but more likely reflect the differing economic conditions in the different sub-periods as 1996 to early 2000 reflected a roaring bull market, followed by the bear markets of later 2000 through 2002, and the milder bull markets of 2003 to 2005. 6. Robustness of Results: D. Chan, Hameed, & Tong’s (2000) Study of 23 Country Indices: Part of the motivation of this study was to improve on previous momentum studies o f country indices, since ETFs are readily investable at lower transactions costs than replicating country indices, ETFs have no restrictions on short-selling, and are readily available to U.S. investors without risks of currency fluctuations or the problems of non-synchronous trading in markets open different hours. However, ETFs are a newer investment security which shortens the sample period of available international ETF data. Tables 2.11 - 2.14 replicate Tables 2.1 - 2.4 using only the ETFs for 19 of the 23 country indices studied by Chan, Hameed, & Tong (2000) for which ETFs are available. The Diamonds ETF, which tracks the Dow Jones Industrial Average, is added to the iShares ETFs matching the MSCI country indices for Australia, Austria, Belgium, Canada, France, Germany, Hong Kong, South Korea, Italy, Japan, Netherlands, South Africa, Spain, Singapore, Switzerland, U.K., Taiwan, and Malaysia; unfortunately, no ETFs are available for Denmark, Norway, Thailand, and Indonesia. By eliminating 26 international ETFs while adding 1 large-cap domestic ETF, it is not clear a priori what the impact 81 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. should be on the momentum/contrarian abnormal returns relative to my whole sample of the 45 international ETFs covered by CRSP. The results using Chan, Hameed, & Tong’s (2000) 19 country ETFs are qualitatively the same as my previous results using my whole sample of 45 international ETFs. Table 2.11 shows annualized raw momentum/contrarian returns that are economically significant but statistically insignificant with the largest t-statistic of 0.57. For the winner minus loser portfolio, Table 2.12 shows statistically significant, at the 5% level, contrarian abnormal returns for the 1 week holding and formation period and momentum abnormal returns for holding and formation periods of 4 weeks and 26 weeks; these abnormal returns occur for both weighting methodologies and both risk-adjustment models. Generally, the abnormal returns are larger magnitude using the 19 country ETFs as compared to the full sample of international ETFs and the largest momentum abnormal returns occur for the 26 week formation and holding period rather than the 4 week one. Both Table 2.13 using CAPM and Table 2.14 using Fama & French’s three factor model show a reduced level o f systematic risk for both winner and loser ETFs, under both weighting schemes, for almost all formation and holding periods using the 19 country ETFs. Also, Table 2.14, using the 19 country ETFs, shows all the winner portfolios and almost all the loser portfolios have a greater concentration of value stocks with high book to market ratios; for the longer formation and holding periods, the winner portfolios have less small-cap stocks while the loser portfolios have more small-cap stocks as compared to the whole sample of 45 international ETFs. 82 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7. Conclusion: This study extends the momentum/contrarian anomaly to a new domain, portfolios o f international ETFs that either buy the winners and short the losers or buy the losers and short the winners, respectively. Currently, all U.S. ETFs are passively managed to track an index, not actively managed to time the market or “beat the market” by loading up on high momentum stocks. Yet, in spite of this disadvantage to actively managed mutual funds, ETFs provide economically and statistically significant abnormal returns to contrarian strategies of buying the loser ETFs and shorting the winner ETFs with formation and holding periods of 1 week, and to momentum strategies of buying the winner ETFs and shorting the loser ETFs with formation and holding periods from 4 weeks to 26 weeks. As compared to country indices or individual international equities, ETFs are ideal instruments with which to implement a contrarian and momentum strategy, since ETFs allow the purchase or short sale o f a diversified portfolio of securities for one commission and one bid ask spread with minimal price impact. Using CAPM and Fama & French’s three factor model to adjust for risk, I find the contrarian and momentum abnormal returns available with international ETFs can not be explained by rational differences in risk and so I provide further evidence of the momentum/contrarian anomaly’s attack on market efficiency. The evidence I present is more powerful since my results are true using both Lo & MacKinlay’s difference from the average performance weights as well as Jegadeesh & Titman’s equal weights within the top and bottom deciles. For both methodologies, the 4 week formation and holding 83 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. period maximizes the momentum abnormal returns and the 1 week formation and holding period maximizes the contrarian abnormal returns. Risk-adjusted abnormal returns can be increased by either using trading volume as an indicator or allowing an extra week between the formation and holding periods. High trading volume increases the momentum abnormal returns for formation and holding periods of 4 weeks and 26 weeks; low trading volume increases the contrarian abnormal returns for the 1 week formation and holding period. Also, adding an extra week between the formation and holding periods, converts the contrarian abnormal returns o f the 1 week formation and holding period to momentum abnormal returns, as well as increasing the magnitude and statistical significance of the momentum abnormal returns for the other formation and holding periods of 2 weeks, 4 weeks, 12 weeks, and 26 weeks. In the future, as the sample period of international ETFs’ data is longer, I plan to check the longer term performance for 18,24, and 36 months after portfolio formation, to determine the duration of momentum/contrarian performance and anticipate the normal long-term reversals as documented by De Bondt & Thaler (1985,1987), thus extending their results to a new domain o f international ETFs. Also, I anticipate that ETFs will continue to grow and expand their array of offerings where eventually I anticipate actively managed ETFs to more directly compete with mutual funds and yet appeal to larger investors and investors that trade more actively than permitted by mutual funds. With actively managed ETFs, I expect the usual hot performance chasing behavior by investors, resulting in ETF managers that utilize momentum strategies thus magnifying 84 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the momentum performance from the passively managed ETFs studied in this paper. ETFs will continue to adapt and evolve to provide more flexible investment vehicles for investors and traders, as well as future questions to research and investigate relative to momentum/contrarian strategies as new ETFs and more years of data become available. 85 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. REFERENCES Balvers, Ronald J. & Yangru Wu, 2006. Momentum and Mean Reversion Across National Equity Markets. Journal o f Empirical Finance, 13: 2 4 -4 8 . Barberis, Nicholas, Andrei Shleifer, & Robert Vishny, 1998. A Model of Investor Sentiment. Journal o f Financial Economics, 49: 307 - 343. Blume, Lawrence, David Easley, & Maureen O’Hara, 1994. Market Statistics and Technical Analysis: The Role of Volume. Journal o f Finance, 49: 153-181. Campbell, John Y., Sanford J. Grossman, & Jiang Wang, 1993. Trading Volume and Serial Correlation in Stock Returns. Quarterly Journal o f Economics, 108: 905 939. Carhart, Mark M., 1997. On Persistence in Mutual Fund Performance. Journal o f Finance, 52: 57 - 82. Chan, Kalok, Allaudeen Hameed, & Wilson Tong, 2000. Profitability o f Momentum Strategies in the International Equity Markets. Journal o f Financial and Quantitative Analysis, 35: 153 —172. Chan, Louis K. C., Narasimhan Jegadeesh, & Josef Lakonishok, 1996. Momentum Strategies. Journal o f Finance, 51: 1681-1713. Conrad, Jennifer, Allaudeen Hameed, & Cathy Niden, 1994. Volume and Autocovariances in Short-Horizon Individual Security Returns. Journal o f Finance, 49: 1305 —1329. Conrad, Jennifer & Gautam Kaul, 1998. An Anatomy of Trading Strategies. Review o f Financial Studies, 11: 489 - 519. Daniel, Kent, David Hirshleifer, & Avanidhar Subrahmanyam, 1998. Investor Psychology and Security Market Under- and Overreactions, Journal o f Finance, 53: 18 3 9 - 1886. De Bondt, Werner F. M. & Richard Thaler, 1985. Does the Stock Market Overreact? Journal o f Finance, 40: 793 - 805. De Bondt, Werner F. M. & Richard Thaler, 1987. Further Evidence On Investor Overreaction and Stock Market Seasonality. Journal o f Finance, 42: 5 5 7 -5 8 1 . De Jong, Jr., Jack C., 2007. Momentum/Contrarian Abnormal Returns and Exchange Traded Funds. Unpublished Working Paper, University of Hawaii. 86 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. REFERENCES Fama, Eugene F. & Kenneth R. French, 1993. Common Risk Factors in the Returns on Stocks and Bonds. Journal o f Financial Economics, 33: 3 - 56. Fama, Eugene F. & Kenneth R. French, 1996. Multifactor Explanations o f Asset Pricing Anomalies. Journal o f Financial Economics, 51: 5 5 -8 4 . Gagnon, Louis & G. Andrew Karolyi, 2007. Information, Trading Volume, and International Stock Return Comovements: Evidence from Cross-listed Stocks. Unpublished Working Paper, Fisher College of Business at Ohio State University. Gastineau, Gary L., 2002. The Exchange-Traded Funds Manual. (John Wiley & Sons, Inc., New York). Gebhardt, William R., Soeren Hvidkjaer, & Bhaskaran Swaminathan, 2002. Stock and Bond Market Interaction: Does Momentum Spill Over? Unpublished Working Paper, Axia Energy Europe, Ltd. London, U.K., University o f Maryland, & Cornell University. Grinblatt, Mark & Tobias J. Moskowitz, 2004. Predicting Stock Price Movements from Past Returns: the Role of Consistency and Tax-Loss Selling. Journal o f Financial Economics, 71: 541 - 579. Grinblatt, Mark, Sheridan Titman, & Russ Wermers, 1995. Momentum Investment Strategies, Portfolio Performance, and Herding: A Study of Mutual Fund Behavior. American Economic Review, 85: 1088 - 1105. Henker, Thomas, Martin Martens, Sc Robert Huynh, 2006. The Vanishing Abnormal Returns o f Momentum Strategies and “Front-Running” Momentum Strategies. Unpublished Working Paper. Hong, Harrison, Terrance Lim, & Jeremy C. Stein, 2000. Bad News Travels Slowly: Size, Analyst Coverage, and the Profitability of Momentum Strategies. Journal o f Finance, 55: 265 - 295. Hong, Harrison & Jeremy C. Stein, 1999. A Unified Theory of Underreaction, Momentum Trading and Overreaction in Asset Markets. Journal o f Finance, 54: 2143-2184. Jegadeesh, Narasimhan & Sheridan Titman, 1993. Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency. Journal o f Finance, 48: 6 5 -9 1 . 87 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. REFERENCES Jegadeesh, Narasimhan & Sheridan Titman, 2001. Profitability of Momentum Strategies: An Evaluation of Alternative Explanations. Journal o f Finance, 56: 699 - 720. Jordan, Steve, 2004. Is Momentum A Self-Fulfilling Prophecy? Yale International Center for Finance Working Paper No. 04 - 25. Lee, Charles M. C. & Bhaskaran Swaminathan, 2000. Price Momentum and Trading Volume. Journal o f Finance, 55: 2017-2069. Lesmond, David A., Michael J. Schill, & Chunsheng Zhou, 2004. The Illusory Nature of Momentum Profits. Journal o f Finance, 71: 3 4 9-380. Llorente, Guillermo, Roni Michaely, Gideon Saar, & Jiang Wang, 2002. Dynamic Volume - Return Relation of Individual Stocks. Review o f Financial Studies, 15: 1005-1047. Lo, Andrew W. & A. Craig MacKinlay, 1990. When Are Contrarian Profits Due to Stock Market Overreaction? Review o f Financial Studies, 3: 175 - 205. Moskowitz, Tobias J. & Mark Grinblatt, 1999. Do Industries Explain Momentum? Journal o f Finance, 54: 1249 - 1290. Okunev, John & Derek White, 2001. Do Momentum Based Strategies Still Work In Foreign Currency Markets? Unpublished Working Paper, B.T. Funds Management, Sydney, Australia, University of New South Wales. O’Neil, William J., 1995. How to Make Money in Stocks: A Winning System in Good Times or Bad. (McGraw-Hill, Inc., New York). Poterba, James M. & John B. Shoven, 2002. Exchange Traded Funds: A New Investment Option for Taxable Investors. Massachusetts Institute of Technology Department of Economics Working Paper #02-07. Rouwenhorst, K. Geert, 1998. International Momentum Strategies. Journal o f Finance, 53: 2 6 7 -2 8 4 . Rouwenhorst, K. Geert, 1999. Local Return Factors and Turnover in Emerging Stock Markets. Journal o f Finance, 54: 1439 - 1464. Sapp, Travis & Ashish Tiwari, 2004. Does Stock Return Momentum Explain the “Smart Money” Effect? Journal o f Finance, 59: 2605 - 2622. 88 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. REFERENCES Scott, Maria Crawford & Jean Henrich, 2004. Mutual Funds: The Individual Investor’s Guide to Exchange-Traded Funds. American Association o f Individual Investors Journal, 26: No. 9,13 - 29. Wermers, Russ, 2003. Is Money Really “Smart”? New Evidence on the Relation Between Mutual Fund Flows, Manager Behavior, and Performance Persistence. Unpublished Working Paper. University of Maryland. Zhong, Maosen & Hui Yang, 2005. Risk Exposures and International Diversification: Evidence from iShares. Journal o f Business Finance & Accounting, 32: 737 771. 89 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.1: Annualized Raw Returns: All International ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Lo & MacK nlay Difference from Average Return Weights: 12wk 12wk 26wk 26wk 1wk 1wk 2wk 2wk 4wk 4wk Jegadeesh & Titman Decile Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk WML: Mean t - statistic -12.7504% -0.1051 7.3476% 0.0797 11.3516% 0.1699 4.1080% 0.0976 10.2466% 0.2661 -19.3752% -0.1153 5.9852% 0.0493 10.3740% 0.1158 5.5081% 0.0939 10.6002% 0.2146 Winner: Mean t - statistic 5.6472% 0.0379 13.9074% 0.1242 15.4245% 0.1918 11.5280% 0.2653 16.0038% 0.5115 4.3056% 0.0257 13.5226% 0.1128 16.2643% 0.1768 13.2041% 0.2676 17.4706% 0.4962 Loser: Mean t - statistic 18.4028% 0.1149 6.5598% 0.0597 4.0729% 0.0481 7.4200% 0.1290 5.7572% 0.1418 23.6808% 0.1264 7.5374% 0.0580 5.8903% 0.0576 7.6960% 0.1105 6.8704% 0.1388 Notes: (1) Annualized mean raw returns for the period indicated. Periodic returns are annualized by multiplying by 52 divided by the number of w eeks in the 5 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegadeesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sam ple period and held for the indicated period. (5) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for t le indicated ho ding period under both weighting strategies. 90 7.8416% 0.0625 11.8287% 0.0002 5.0800% 0.0142 10.3590% -3.4268% 0.4676 4.5500% 0.2384 5.5848% 0.0354 1.5704% 0.2391 6.9264% 8.8296% 0.0726 -3.2916% 0.3365 -6.2426% 0.0240 -3.5100% 0.0921 R Ft £ II I Fama - French's 3 Factor Model: WML: 7.2462% Constant -12.2148% 0.0895 p - value 0.0269 Winner: Constant -5.4704% 1.9890% 0.2584 0.5987 p - value Loser: Constant 6.7444% -5.2572% 0.1623 p - value 0.1980 ~ + -12.2512% 0.0213 II • RM RF , R -U R CAPM: WML: Constant p - value Winner: Constant p - value Loser: Constant p - value 5** Lo & MacKinlay Difference from Average Return Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk at, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.2: Annualized Abnormal Returns: All International ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31,2005 Jegadeesh & Titman Decile Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk + e tt -18.3092% 0.0126 6.8198% 0.2097 11.0838% 0.0085 6.8068% 0.0170 10.5710% -4.6748% 0.4314 4.1548% 0.3346 6.0684% 0.0602 3.0502% 0.0703 8.2194% 0.0000 -3.5360% 0.0220 13.6396% 0.0271 -2.6650% 0.5491 -5.0141% 0.1637 -3.7566% 0.1671 -2.5460% 0.2147 0.0000 + 0.0000 0.0000 bi ■RA m , +•*, • SMB r + h, • HML r + e 10.1725% 0.0019 4.6761% 0.0449 7.8956% 0.0006 -18.9280% 0.0129 6.1698% 0.2809 9.2209% 0.0353 7.3723% 0.0220 7.0144% 0.0198 2.5948% 0.3286 -0.3948% 0.7746 4.2232% 0.0002 -7.1240% 0.2437 1.9266% 0.6553 3.1564% 0.3484 1.7793% 0.3220 5.3974% 0.0001 -7.5777% 0.0110 -5.0713% 0.0209 -3.6724% 0.0431 11.8040% 0.0711 -4.2458% 0.3860 -6.0645% 0.1113 -5.5926% 0.0411 -1.6172% 0.5201 91 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.2: Annualized Abnormal Returns: All International ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: (1) Annualized abnormal returns for the period indicated. Periodic abnormal returns, i.e. alpha intercepts, are annualized by multiplying by 52 divided by the number of w eeks in the 5 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegadeesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period. (5) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for t he indicated holding period under both weighting strategies. (6) p-values are computed with robust standard errors correcl ed for heteroskedasticity and autocorrelation using the Newey-W es adjustment (1987). 92 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.3: Abnormal Returns Using CAPM: All International ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31,2005 Lo & MacKinlay Difference from Average Return Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk WML: Constant p - value Mkt - RF p - value Adj. R-Sq. Winner: Constant p - value Mkt - RF p - value Adj. R-Sq. Loser: Constant p - value Mkt - RF p - value Adj. R-Sq. -0.0024 0.0213 -0.0761 0.2098 0.0040 0.0030 0.0625 -0.0747 0.2027 0.0029 0.0091 0.0002 -0.0723 0.1657 0.0024 0.0117 0.0142 -0.1486 0.0335 0.0134 -0.0007 0.4676 0.8478 0.0000 0.4905 0.0018 0.2384 0.8816 0.0043 0.0354 0.9594 0.0000 0.0000 0.4526 0.5322 0.0036 0.2391 0.9895 0.0000 0.6315 -0.0013 0.3365 0.9564 -0.0048 0.0240 1.0317 -0.0081 0.0921 1.1381 0.0000 0.0000 0.0000 0.5041 0.5534 0.5523 0.0017 0.0726 0.9239 0.0518 Jegadeesh & Titman Decile Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk -0.0035 0.0126 -0.1628 0.0742 0.0123 0.0026 0.2097 -0.1261 0.1516 0.0060 0.0085 0.0085 -0.1076 0.1091 0.0034 0.0157 0.0170 -0.1986 0.0467 0.0121 0.0032 0.9826 -0.0022 0.9946 -0.0009 0.4314 0.8334 0.0016 0.3346 0.8830 0.0047 0.0602 1.0134 0.0070 0.0703 1.0195 0.0411 0.0000 1.0252 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5477 0.3745 0.3958 0.4538 0.5207 0.4604 -0.0177 0.0220 1.0326 0.0026 0.0271 0.9961 -0.0010 0.5491 1.0092 -0.0039 0.1637 1.1210 -0.0087 0.1671 1.2181 -0.0127 0.2147 1.0542 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.4748 0.3477 0.4282 0.4404 0.4485 0.3712 0.2446 0.0000 -0.0122 0.9158 -0.0021 0.0346 0.0000 93 0.0529 0.0000 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.3: Abnormal Returns Using CAPM: All International ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: • RM RIf\ + e tt (1) CAPM: Dependent variable is periodic return for the WML which is he winner 5 T - R F t= a , ' minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or e x c e s s periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic Treasury bill rate over the listed formation and holding period. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegadeesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each w eek during the sample period and held for the indicated period. (4) RMRF is the ex c e ss return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson A ssociates) over the listed formation and holding periods. (5) p-values are computed with robust standard errors correcl ed for heteroskedasticity and autocorrelation using the Newey-W es adjustment (1987). 94 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.4: Abnormal Returns Using Fama & French's Three Factor Model: All International ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Jegadeesh & Titman Decile Weights: |Lo & MacKinlay Difference from Average Return Weights: 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk 1wk 1wk 12wk 12wk 26wk 26wk 2wk 2wk 4wk 4wk 1wk 1wk WML: 0.0024 0.0071 -0.0036 0.0170 0.0351 0.0108 0.0395 0.0078 -0.0023 0.0028 Constant 0.0129 0.2809 0.0353 0.0220 0.0006 0.0198 0.0019 0.0449 0.0895 0.0269 p - value -0.1457 -0.1014 -0.0764 -0.2332 0.1389 -0.1416 0.0720 -0.0463 -0.0931 -0.0528 Mkt - RF 0.5382 0.1023 0.2599 0.3396 0.0386 0.3470 0.4580 0.0648 0.1193 0.4472 p - value 0.0578 0.0743 0.1878 -0.0632 0.3602 0.0411 0.2426 0.1662 0.0673 -0.0316 HML 0.5739 0.1396 0.5654 0.6403 0.0015 0.5858 0.0046 0.5311 0.0923 0.7403 p - value 0.0282 0.2464 0.0793 0.0263 0.1074 0.2274 0.0915 0.0403 0.1087 0.0301 SMB 0.8207 0.5284 0.0408 0.8747 0.5550 0.5851 0.0139 0.7146 0.2394 0.7488 p - value 0.0097 0.0028 0.0140 0.0090 0.0154 0.0204 0.0102 0.0106 0.0063 0.0001 Adj. R-Sq. Winner: 0.0007 -0.0014 0.0024 0.0041 0.0211 0.0270 -0.0009 0.0008 0.0020 -0.0011 Constant 0.2437 0.6553 0.3484 0.3220 0.0001 0.7746 0.0002 0.5987 0.3286 0.2584 p - value 0.9296 0.9565 1.0802 1.0196 1.0224 1.0429 1.0368 1.0166 0.9192 0.9761 Mkt - RF 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 p - value 0.2437 0.1244 0.1877 0.2797 0.3008 0.1872 0.3124 0.1985 0.2178 0.2895 HML 0.0201 0.0037 0.0316 0.0014 0.0000 0.0138 0.0004 0.0000 0.0041 0.0155 p - value 0.1117 0.1603 0.3040 0.2002 0.4721 0.2727 0.2198 0.4133 0.1271 0.1559 SMB 0.0047 0.1099 0.0063 0.0000 0.0000 0.3779 0.0014 0.0000 0.1104 0.2244 p - value 0.3827 0.4039 0.4756 0.5313 0.6073 0.5190 0.5588 0.6542 0.4989 0.4649 Adj. R-Sq. Loser: -0.0047 -0.0184 0.0023 -0.0016 -0.0129 -0.0081 -0.0117 -0.0058 0.0013 -0.0020 Constant 0.0711 0.3860 0.0411 0.5201 0.0431 0.1113 0.0209 0.0110 0.1980 0.1623 p - value 1.0579 1.0754 1.1566 1.2528 0.9041 1.1582 0.9505 1.0831 1.0123 1.0289 Mkt - RF 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 p - value 0.1694 0.1131 0.1876 -0.0550 0.2220 -0.1730 0.1462 0.1574 0.2494 0.2222 HML 0.0974 0.0318 0.2010 0.0701 0.0761 0.0177 0.4355 0.0206 0.0024 0.0031 p - value 0.0324 0.1321 0.0577 0.1739 0.3647 0.3218 0.0453 0.1796 0.0472 0.0970 SMB 0.7476 0.1840 0.5390 0.0726 0.0281 0.0866 0.0119 0.5092 0.5877 0.2542 p - value 0.4430 0.4479 0.3755 0.2647 0.3655 0.4315 0.5542 0.4816 0.5605 0.5116 Adj. R-Sq. 95 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.4: Abnormal Returns Using Fama & French's Three Factor Model: All International ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 | a , + b, ■ i^M R b, + s, • S M B , + h, ■H M L , + e it Notes: (1) Fama & French's three factor model: R-it RFt — Dependent variable is periodic return for the WML which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or ex cess periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic Treasury bill rate over the listed formation and holding period. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegad eesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period. (5) RMRF is the e x c e s s return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson Associates) over the listed formation and holding periods. (6) HML (High Book to Market Minus Low Book to Market) is the average return on the two value portfolios minus the average return on the two growth portfolios, i.e. HML = 1/2 (Small Value + Big Value) -1 /2 (Small Growth + Big Growth). (7) SMB (Small Size Minus Big Size) is the average return on the three small portfolios minus the average return on the three big portfolios, SMB = 1/3 (Small Value + Small Neutral + Small Growth) -1 /3 (Big Value + Big Neutral + Big Growth). (8) p-values are computed with robust standard errors correct ed for heteroskedasticity and autocorrelation using the Newey-W es adjustment (1987). 96 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.5: Annualized Raw Returns: All International ETFs by High and Low Trading Volume Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31,2005 Lo & MacKinlay Difference from Average Return Weights: High Volume 4wk 4wk 12wk 12wk 26wk 26wk 2wk 2wk 1wk 1wk Jegadeesh & Titman Decile Weights: High Volume 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk 1wk 1wk WML: Mean t - statistic -0.4524% -0.0028 7.6830% 0.0581 11.9093% 0.1272 3.8190% 0.0618 12.6816% 0.2999 -11.1332% -0.0502 9.3886% 0.0582 16.4047% 0.1517 0.9876% 0.0135 15.3522% 0.3089 Winner: Mean t - statistic 7.7428% 0.0509 13.0598% 0.1096 15.8210% 0.1923 10.4160% 0.2201 14.3184% 0.4469 8.4240% 0.0442 14.6926% 0.0962 18.0232% 0.1798 10.0200% 0.1748 17.3968% 0.4510 Loser: Mean t - statistic 17.8776% 0.1046 6.3804% 0.0553 3.5594% 0.0424 6.5147% 0.1077 2.2866% 0.0592 19.5572% 0.0880 5.3066% 0.0362 1.6185% 0.0157 9.0320% 0.1139 2.0444% 0.0422 Lo & MacK nlay Difference from Average Return Weights: Low Volume 4wk 4wk 12wk12wk 26wk 26wk 2wk 2wk 1wk 1wk Jegadeesh & Titman Decile Weights: Low Volume 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk 1wk 1wk WML: Mean t - statistic -23.7328% -0.1367 6.0840% 0.0527 5.7590% 0.0617 0.7393% 0.0123 8.8428% 0.1711 -18.4340% -0.0871 1.6926% 0.0115 0.3848% 0.0033 2.7040% 0.0376 5.6706% 0.0991 Winner: Mean t - statistic 2.1996% 0.0137 10.2206% 0.0964 13.0689% 0.1597 11.3000% 0.2662 15.3138% 0.4593 2.4232% 0.0125 10.2206% 0.0829 11.0136% 0.1120 13.5291% 1.2180 16.0318% 0.4100 Loser: Mean t - statistic 15.3088% 0.0921 6.2894% 0.0529 6.5013% 0.0676 11.1549% 0.2005 8.0400% 0.1828 20.8572% 0.0939 8.5280% 0.0571 10.6288% 0.0805 10.8251% 0.1440 10.3612% 0.1865 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.5: Annualized Raw Returns: All International ETFs by High and Low Trading Volume Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: (1) Annualized abnormal returns for the period indicated. Periodic abnormal returns, i.e. alpha intercepts, are annualized by multiplying by 52 divided by the number of w eeks in the 5 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegadeesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period. (5) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for the indicated holding period. (6) Volume is measured a s the percentage change in trading volume from the previous periodic formation period. High (Low) Volume is greater than (less than or equal to) the median percentage change in trading volume relative to the previous periodic formation period. 98 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.6: Annualized Abnormal Returns Using CAPM: All International ETFs by High and Low Trad ng Volume Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Lo & MacK nlay Difference from Average Return Weights: 2wk 2wk 4wk 4wk 1wk 1wk 12wk 12wk 26wk 26wk High Volume: WML: 0.4727% Constant 0.9475 p - value Winner: -1.4716% Constant 0.7620 p - value Loser: 8.1744% Constant 0.1314 p - value Low Volume: WML: -23.5820% Constant 0.0021 p - value Winner: -6.8224% Constant 0.2007 p - value Loser: 5.7876% Constant 0.2752 p - value 8.0028% 0.1822 11.8118% 0.0078 2.7317% 0.3803 9.3240% 3.7206% 0.3758 6.5923% 0.0228 0.7718% 0.6253 6.3542% -2.9770% 0.4471 -5.8422% 0.0478 6.6976% 0.1935 Jegadeesh & Titman Decile Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk -10.3116% 0.2909 9.5680% 0.1775 16.6712% 0.0011 1.7238% 0.6301 13.6574% -0.5824% 0.9348 5.0024% 0.3932 8.7529% 0.0268 0.6357% 0.7831 9.2360% 0.0000 -3.2851% 0.1727 -5.0006% 0.0050 9.7292% 0.2297 -4.5682% 0.4073 -7.9183% 0.0510 -1.0877% 0.7512 -5.3606% 0.0241 6.9550% 0.1207 2.2499% 0.4212 10.9438% -17.1912% 0.0702 3.2864% 0.6103 2.1541% 0.6898 5.0254% 0.1422 7.2340% 0.0026 1.4976% 0.6888 3.3241% 0.1925 2.0878% 0.1384 7.2054% -6.6924% 0.3551 1.6302% 0.7322 1.2415% 0.7328 4.6254% 0.0166 7.9010% 0.0000 -3.3774% 0.3890 -3.4463% 0.3030 0.8537% 0.6647 -0.7032% 0.6565 10.4988% 0.1869 -1.6562% 0.7591 -0.9126% 0.8533 -0.4000% 0.8923 1.2898% 0.5587 0.0000 0.0000 99 0.0000 0.0000 0.0000 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.6: Annualized Abnormal Returns Using CAPM: All International ETFs by High and Low Trading Volume Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: (1) Annualized abnormal returns for the period indicated. Periodic abnormal returns, i.e. alpha intercepts, are annualized by multiplying by 52 divided by the number of w eeks in the 5 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegadeesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each w eek during the sample period and held for the indicated period. (5) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for the indicated holding period. (6) Volume is measured a s the percentage change in trading volume from the previous periodic formation period. High (Low) Volume is greater than (less than or equal to) the median percentage change in trading volume relative to the previous periodic formation period. (7) p-values are computed with robust standard errors correc! ed for heteroskedasticity and autocorrelation using the Newey-W es adjustment (1987). 100 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.7: Annualized Abnormal Returns Using Fama & French's Three Factor Model: Al International ETFs by High and Low Volume Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Lo & MacKinlay Difference from Average Return Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk High Volume: WML: 1.5392% Constant 0.8369 p - value Winner: -3.2344% Constant 0.5118 p - value Loser: 6.1048% Constant 0.2924 p - value Low Volume: WML: -24.3932% Constant 0.0025 p - value Winner: -8.8036% Constant 0.1136 p - value Loser: 3.5516% Constant 0.5225 p - value Jegadeesh & Titman Decile Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk 9.3574% 0.1247 9.2716% 0.0387 1.5483% 0.6622 11.6614% 0.0000 -9.3080% 0.3702 10.0048% 0.1727 13.8437% 0.0076 1.6441% 0.6710 13.8354% 0.0000 1.9656% 0.6399 3.5308% 0.2092 -1.4205% 0.3599 2.8252% 0.0103 -1.6380% 0.8264 3.1304% 0.5976 5.6290% 0.1446 -1.8464% 0.4047 6.4338% 0.0000 -5.3950% 0.2057 -6.4701% 0.0430 -5.0483% 0.0475 -7.1110% 0.0001 7.6700% 0.3721 -6.8744% 0.2358 -8.2147% 0.0615 -3.4901% 0.3287 -7.4014% 0.0040 4.5292% 0.4128 5.7824% 0.2548 2.9514% 0.3844 5.3086% 0.0883 -16.3800% 0.0927 1.4950% 0.8290 0.9633% 0.8658 7.6943% 0.0574 3.5444% 0.3153 -1.7082% 0.6449 0.5382% 0.8514 0.3337% 0.8246 5.0396% 0.0001 -8.2888% 0.2698 -1.4170% 0.7649 -1.9357% 0.6220 3.9364% 0.0571 5.7542% 0.0005 -4.8438% 0.2592 -5.2611% 0.1390 -1.4716% 0.5062 -0.4898% 0.8184 8.0912% 0.3293 -2.9120% 0.6305 -2.8990% 0.5720 -3.7579% 0.2683 2.2098% 0.4599 101 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.7: Annualized Abnormal Returns Using Fama & French's Three Factor Model: Al International ETFs by High and Low Volume Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: (1) Annualized abnormal returns for the period indicated. Periodic abnormal returns, i.e. alpha intercepts, are annualized by multiplying by 52 divided by the number of w eeks in the 5 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegadeesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each w eek during the sample period and held for the indicated period. (5) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for the indicated holding period. (6) Volume is measured a s the percentage change in trading volume from the previous periodic formation period. High (Low) Volume is greater than (less than or equal to) the median percentage change in trading volume relative to the previous periodic formation period. (7) p-values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using the Newey-Wes adjustment (1987). 102 Lo & MacKinlay Difference from Average Return Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk CAPM: WML: Constant p - value Winner: Constant p - value Loser: Constant p - value Jegadeesh & Titman Decile Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk R tt — R Ft = a , + b , • RMRF . + 13.0052% 0.0182 11.2606% 0.0115 12.3240% 0.0001 5.2056% 0.0122 10.4442% 5.0076% 0.2968 5.7070% 0.1306 5.4483% 0.0300 1.6818% 0.2247 6.9620% -7.9924% 0.1021 -5.5536% 0.1288 -6.8757% 0.0141 -3.5234% 0.0936 -3.5070% 0.0244 a II «*S I Fama - French's 3 Factor Model: WML: 13.5044% 10.1062% Constant 0.0227 0.0164 p - value Winner: 2.9120% 2.8236% Constant 0.4175 0.5624 p - value Loser: -10.6808% -7.1942% Constant 0.0359 0.0676 p - value as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.8: Annualized Abnormal Returns with Extra Time between Formation and Holding: All International ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 0 .0 0 0 0 0 .0 0 0 0 a i + bi ■RA rR F t +s,-& M B f + 15.2256% 0.0487 8.0444% 0.1678 11.7481% 0.0049 6.7691% 0.0151 10.8304% 4.4304% 0.4459 4.5032% 0.2879 6.9823% 0.0242 3.1248% 0.0722 8.2776% -10.7952% 0.0917 -3.5412% 0.4601 -4.7658% 0.1916 -3.6443% 0.1788 -2.6694% 0.1943 0 .0 0 0 0 0 .0 0 0 0 • HML , + e 11.2346% 0.0004 4.5704% 0.0504 7.8418% 0.0008 15.6364% 0.0475 7.2852% 0.2288 10.4988% 0.0148 6.7518% 0.0321 7.1938% 0.0169 2.6754% 0.3061 -0.4585% 0.7492 4.1862% 0.0002 -2.4648% 0.6769 2.1372% 0.6175 4.2562% 0.1985 -1.5604% 0.3982 5.4254% 0.0001 -8.5592% 0.0039 -5.0288% 0.0221 -3.6556% 0.0480 -13.1716% 0.0453 -5.1480% 0.3180 -6.2426% 0.1017 -5.1913% 0.0736 -1.7684% 0.4860 103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.8: Annualized Abnormal Returns with Extra Time between Formation and Holdinc : All International ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: (1) Annualized abnormal returns for the period indicated. Periodic abnormal returns, i.e. alpha intercepts, are annualized by multiplying by 52 divided by the number of w eek s in the 5 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegad eesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sam ple period and held for the indicated period. (5) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for the indicated holding period. (6) For all 5 weekly strategies with formation and holding periods from 1 week to 26 w eeks, an extra week is added between the formation and holding periods. (7) p-values are computed with robust standard errors correcl ed for heteroskedasticity and autocorrelation using the Newey-W es adjustment (1987). 104 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.9: Abnormal Returns Using Different Betas in Up and Down Markets Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31,2005 Panel A: Vl/hole Sample of International ETFs: Lo & MacKinlay Difference from Average Return Weights: 12wk 12wk 26wk 26wk 1wk 1wk 2wk 2wk 4wk 4wk WML: Constant p - value Beta-Up p - value Beta-Down p - value Adj. R-Sq. Winner: Constant p - value Beta-Up p - value Beta-Down p - value Adj. R-Sq. Loser: Constant p - value Beta-Up p - value Beta-Down p - value Adj. R-Sq. 0.0355 0.0003 0.4833 Jegadeesh & Titman Decile Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk -0.0034 0.0166 0.0381 0.7009 -0.3216 0.0185 0.0276 0.0023 0.2654 0.2018 0.1159 -0.3816 0.0006 0.0461 0.0085 0.0073 0.2353 0.0199 -0.3569 0.0421 0.0168 0.0107 0.1512 0.1719 -0.4952 0.0008 0.0487 0.0014 0.3895 1.0845 0.0047 0.0560 1.2371 0.0073 0.0623 1.0997 0.0344 1.1493 -0.0008 0.4980 1.0109 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 -0.0022 0.0296 0.1107 0.0876 -0.2239 0.0001 0.0307 0.0028 0.0802 0.1533 0.1109 -0.2524 0.0001 0.0362 0.0091 0.0002 0.1858 0.0166 -0.2599 0.0001 0.0418 0.0128 0.0068 0.1868 0.0102 -0.4331 0.0000 0.0802 -0.0006 0.5384 0.9937 0.0016 0.2746 1.0337 0.0043 0.0327 1.1639 0.0040 0.1961 1.1020 0.0000 0.0000 0.0000 0.0000 0.0000 0.7323 0.7632 0.8107 0.8941 0.7821 0.6930 0.7261 0.8506 0.9515 0.7397 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5011 0.4621 0.5491 0.6379 0.5647 0.3868 0.4109 0.4691 0.5225 0.4850 0.0017 0.0763 0.8830 -0.0012 0.3668 0.8804 -0.0048 0.0241 0.9781 -0.0088 0.0687 0.9152 -0.0059 0.4756 0.6660 0.0026 0.0278 0.9728 -0.0009 0.5995 0.8827 -0.0095 0.1329 0.9485 0.0019 0.8625 0.5594 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.9562 1.0156 0.0000 0.5553 1.0706 1.3272 1.5360 1.0146 1.1077 -0.0039 0.1634 1.0019 0.0000 1.2076 1.4467 1.6788 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5525 0.4897 0.4061 0.4273 0.4447 0.4511 0.3859 0.3052 0.0000 0.5040 0.0000 -0.7539 0.0001 0.1350 0.0297 105 0.0000 0.0325 0.0102 0.6336 0.0000 -0.9391 0.0001 0.1321 0.0000 1.2331 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.9: Abnormal Returns Using Different Betas in Up and Down Markets Exchange Traded Funds and Momentum Returns: March 21,1996 • December 31, 2005 Panel B: Sample of International ETFs Partitioned by High and Low Trading Volume: Lo & MacKinlay Difference from Average Return Weights: 12wk 12wk 26wk 26wk 1wk 1wk 2wk 2wk 4wk 4wk Jegadeesh & Titman Decile Weights: 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk 1wk 1wk High Trading Volume: WML: 0.0028 0.0003 Constant 0.2227 0.8488 p - value 0.0907 0.2063 Beta-Up 0.2252 0.4113 p - value -0.2535 -0.3139 Beta-Down 0.0005 0.0069 p - value 0.0304 0.0191 Adj. R-Sq. 0.0091 0.0068 0.3533 0.0031 -0.2287 0.0359 0.0321 0.0075 0.2737 0.6521 0.0000 -0.2295 0.1299 0.0717 0.0460 0.0001 0.6926 0.0000 -0.0354 0.7778 0.0809 -0.0019 0.3102 -0.0005 0.9973 -0.2281 0.1915 0.0051 0.0033 0.2193 0.3315 0.0385 -0.3088 0.0137 0.0250 0.0129 0.0010 0.2292 0.0453 -0.2391 0.0132 0.0140 0.0049 0.5565 0.2440 0.0555 -0.4339 0.0086 0.0267 0.0618 0.0000 0.5744 0.0001 -0.3680 0.0000 0.0512 Low Tradin g Volume: WML: 0.0024 -0.0045 Constant 0.0025 0.2238 p - value 0.0391 0.0900 Beta-Up 0.7229 0.3751 p - value -0.1123 -0.2005 Beta-Down 0.0930 0.5061 p - value 0.0089 0.0014 Adj. R-Sq. 0.0054 0.1203 -0.0633 0.6276 -0.2889 0.0184 0.0178 0.0055 0.3962 -0.1157 0.4035 -0.3455 0.0747 0.0203 0.0344 0.0165 0.3274 0.0300 -1.0281 0.0027 0.1063 -0.0032 0.0777 -0.0659 0.6501 -0.2943 0.0141 0.0132 0.0011 0.6609 -0.0935 0.4004 -0.3629 0.0426 0.0225 0.0017 0.6767 0.0975 0.5149 -0.5579 0.0000 0.0463 0.0122 0.1203 -0.1146 0.4603 -0.6049 0.0059 0.0453 0.0135 0.3756 0.5179 0.0001 -1.1242 0.0013 0.1164 106 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.9: Abnormal Returns Using Different Betas in Up and Down Markets Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: - D t)Rty[RFt +eit Dependent variable is periodic return for the (1) Model: = a , + t T ■Dt RMR} WML which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or e x c e ss periodic return for the winner or loser portfolio above the risk free rate as proxied by the appropriate periodic Treasury bill rate over the listed formation and holding period. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegad eesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period. (4) RMRF is the e x c e ss return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson A ssociates) over the listed formation and holding periods. (5) The D, is a dummy variable that is one if CRSP value-weighted market return is positive in holding period t and zero otherwise. (6) Volume is measured a s the percentage change in trading volume from the previous periodic formation period. High (Low) Volume is greater than (less than or equal to) the median percentage change in trading volume relative to the previous periodic formation period. (7) p-values are computed with robust standard errors correcl ed for heteroskedasticity and autocorrelation using the Newey-W es adjustment (1987). 107 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abnormal Returns Divided into Two Periods: 03/21/96 ■12/31/00 vs. 01/01/01 - 12/31/05 Table 2.10: Exchange Traded Funds and Momentum Returns: 26 week formation and holding period. FF3F: Lo & MacK nlay Wts. Jegadeesh & Titman W ts. Lo & MacKinlay Wts. Jegadeesh & Titman Wfts. CAPM: 1996-2000 2001-2005 1996-2000 2001-2005 1996-2000 2001-2005 1996-2000 2001-2005 WML: WML: -0.0085 Constant 0.0811 0.0901 -0.0145 0.0123 0.1082 0.0199 0.0932 Constant 0.0000 0.0000 0.1333 0.0909 p - value 0.1019 0.0001 0.0000 0.0000 p - value 0.0044 0.0200 0.1032 0.3195 Mkt - RF 0.0954 -0.2122 0.0013 -0.1311 Mkt - RF 0.0077 0.9884 0.9343 0.0000 0.1665 p - value 0.9754 0.3476 0.4748 p - value 1.0521 0.2110 0.7575 0.3015 0.0047 HML 0.0016 -0.0043 -0.0008 Adj. R-Sq. 0.0802 0.1025 0.0000 0.0000 p - value 0.4171 -0.5077 SMB 0.2922 -0.1366 0.0615 0.0853 0.0473 0.0001 p - value 0.0063 0.3022 0.0131 0.2789 Adj. R-Sq Winner: Winner: 0.0167 0.0173 0.0223 0.0260 0.0744 Constant 0.0622 0.0093 0.0082 Constant 0.0017 0.0392 0.0008 p value 0.0623 0.0000 0.0000 0.3133 0.2709 p - value 0.7105 0.7759 1.1519 1.2301 Mkt - RF 1.2633 0.8725 1.1845 0.8749 Mkt - RF 0.0000 0.0000 0.0000 0.0000 p - value 0.0000 0.0000 0.0000 0.0000 p - value -0.1477 0.7649 HML -0.0808 0.7050 0.6446 0.7142 0.3619 0.4567 Adj. R-Sq. 0.0946 0.0000 0.0076 0.0000 p - value 0.1923 0.3750 0.2021 0.3987 SMB 0.0122 0.0015 0.0000 0.0023 p - value 0.7321 0.8096 0.3886 Adj. R-Sq 0.4781 Loser: Loser: -0.0644 -0.0678 0.0258 0.0406 Constant 0.0610 0.0421 -0.0936 -0.0819 Constant 0.0000 0.0000 0.0003 0.0000 p - value 0.0000 0.0000 0.0000 0.0000 p - value 1.0487 0.7061 0.9106 Mkt RF 0.7559 1.1828 1.1917 1.0935 1.0126 Mkt - RF 0.0002 0.0066 0.0000 0.0000 p - value 0.0000 0.0000 0.0000 0.0000 p - value -0.2918 -0.0525 -0.4493 -0.2872 HML 0.1787 0.6972 0.8322 0.2448 Adj. R-Sq. 0.0041 0.0068 0.5075 0.0100 p - value -0.0999 0.5116 -0.2150 0.9064 SMB 0.0000 0.2174 0.0000 p - value 0.4199 0.2541 0.8611 0.1959 0.7750 Adj. R-Sq 108 Notes: - 12/31/05 II i 01 (1) CAPM: R,, ~ = a , + b, ■R I a t + b r i IM R F , + ■SMB t + ht • HML t + e it (2) Fama & French's three factor model: (3) Dependent variable is periodic return for the WML which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the 26 week formation and holding periods, or e x c ess periodic return for the winner or loser portfolio above the risk free rate as proxied by the appropriate periodic Treasury bill rate over the 26 week formation and holding period. (4) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (5) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (6) The Jegadeesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period. (7) RMRF is the e x c e ss return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson A ssociates) over the listed formation and holding periods. (8) HML (High Book to Market Minus Low Book to Market) is the average return on the two value portfolios minus the average return on the two growth portfolios, i.e. HML = 1/2 (Small Value + Big Value) - 1/2 (Small Growth + Big Growth). (9) SMB (Small Size Minus Big Size) is the average return on the three small portfolios minus the average return on the three big portfolios, SMB = 1/3 (Small Value + Small Neutral + Small Growth) - 1/3 (Big Value + Big Neutral + Big Growth). (10) p-values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using the Newey-W es [adjustment (1987). of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.10: Abnormal Returns Divided into Two Periods: 03/21/96 - 12/31/00 vs. 01/01/01 Exchange Traded Funds and Momentum Returns: 26 week formation and holding period. 109 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.11: Annualized Raw Returns: Chan, Hameed, & Tong's Country ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Lo & MacKinlay Difference from Average Return Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk Jegadeesh & Titman Decile Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk WML: Mean t - statistic -12.9532% -0.1064 2.9770% 0.0630 9.9931% 0.1408 2.7023% 0.0594 11.0704% 0.2631 -20.8468% -0.1210 6.5884% 0.0501 10.6730% 0.1153 3.9667% 0.0638 13.9324% 0.2603 Winner: Mean t - statistic 6.5312% 0.0446 6.8640% 0.1232 14.9695% 0.1909 10.6656% 0.2525 16.6192% 0.5558 6.3024% 0.0362 14.9812% 0.1152 15.5792% 0.1776 11.1440% 0.2309 20.0134% 0.5736 Loser: Mean t - statistic 19.4844% 0.1239 3.8844% 0.0730 4.9764% 0.0584 7.9634% 0.1372 5.5486% 0.1315 27.1492% 0.1443 8.3928% 0.0646 4.9062% 0.0485 7.1773% 0.0994 6.0808% 0.1233 Notes: (1) Annualized mean raw returns for the period indicated. Periodic returns are annualized by multiplying by 52 divided by the number of w eek s in the 5 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegadeesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period. (5) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for the indicated holding period under both weighting strategies. (6) The sample includes only the ETFs for 19 of the 23 country indices studied by Chan, Hameed, and Tong (2000). The Diamonds ETF indexing the Dow Jones Industrial Average is added to iShares ETFs matching the MSCI Indices for Australia, Austria, Belgium, Canada, France, Germany, Hong Kong, South Korea, Italy, Japan, Netherlands, South Africa, Spain, Singapore, Switzerland, U.K., Taiwan, and Malaysia. Unfortunately, no ETFs are available for Denmark, Norway, Thailand, and Indonesia. 110 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Annualized Abnormal Returns: Chan, Hameed, & Tong's Country ETFs Table 2.12: Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Lo & MacKinlay Difference from Average Return Weights: 12wk 12wk 26wk 26wk 2wk 2wk 4wk 4wk 1wk 1wk CAPM: WML: Constant p - value Winner: Constant p - value Loser: Constant p - value Jegadeesh & Titman Decile Weights: 2wk 2wk 4wk 4wk 1wk 1wk 12wk 12wk 26wk 26wk 7?^ ~ 7? p , - a , + b t • R M R F , + .e u __ -12.6048% 0.0180 6.1282% 0.1564 10.6288% 0.0017 3.7384% 0.0914 11.5824% 0.0000 -20.4308% 0.0074 7.1474% 0.2301 11.5024% 0.0084 5.2152% 0.0830 13.9018% 0.0000 -2.1944% 0.6511 4.6878% 0.2426 5.7226% 0.0400 1.3485% 0.3586 8.0754% 0.0000 -2.6104% 0.6842 5.8656% 0.2377 6.2114% 0.0618 1.6878% 0.3483 11.1440% 0.0000 10.4104% 0.0431 -1.4404% 0.6901 -4.9062% 0.1022 -2.3894% 0.2879 -3.4722% 0.0367 17.8204% 0.0085 -1.2818% 0.7913 -5.2910% 0.1682 -3.5273% 0.2428 -2.9868% 0.1558 Fama ■French's 3 Fact or Model: WML: -12.7504% 5.1298% Constant 0.0203 0.2309 p - value Winner: 1.6016% -4.6956% Constant 0.3414 0.6810 p - value Loser: 8.0548% -3.5256% Constant 0.3624 0.1389 p - value R il ~ R f, = a i + b t • R k 'RF. + s, • S M B , + h t • HML , + e 8.5943% 0.0109 3.6729% 0.1283 10.1186% 0.0001 -21.1016% 0.0069 5.5562% 0.3697 9.4393% 0.0334 5.6658% 0.0751 11.9838% 0.0002 1.9240% 0.4780 -1.2008% 0.4212 5.5342% 0.0000 -5.6472% 0.3919 2.8730% 0.5715 2.4206% 0.4786 -0.2942% 0.8702 8.4540% 0.0000 -6.6716% 0.0365 -4.8733% 0.0356 -4.5852% 0.0226 15.4596% 0.0303 -2.6806% 0.6082 -7.0200% 0.0775 -5.9601% 0.0483 -3.5298% 0.1671 111 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.12: Annualized Abnormal Returns: Chan, Hameed, & Tong's Country ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: (1) Annualized abnormal returns for the period indicated. Periodic abnormal returns, i.e. alpha intercepts, are annualized by multiplying by 52 divided by the number of w eeks in the 5 weekly strategies. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegad eesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each w eek during the sam ple period and held for the indicated period. (5) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for the indicated holding period under both weighting strategies. (6) The sam ple includes only the ETFs for 19 of the 23 country indiced studies by Chan, Hameed, and Tong (2000). The Diamonds ETF indexing the Dow Jones Industrial Average is added to iShares ETFs matching the MSCI Indices for Australia, Austria, Belgium, Canada, France, Germany, Hong Kong, South Korea, Italy, Japan, Netherlands, South Africa, Spain, Singapore, Switzerland, U.K., Taiwan, and Malaysia. Unfortunately, no ETFs are available for Denmark, Norway, Thailand, and Indonesia. (7) p-values are computed with robust standard errors correct ed for heteroskedasticity and autocorrelation using the Newey-W es adjustment (1987). 112 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abnormal Returns Using CAPM: Chan, Hameed, & Tong's Country ETFs Table 2.13: Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Lo & MacKinlay Difference from Average Return Weights: 12wk 12wk 26wk 26wk 1wk 1wk 2wk 2wk 4wk 4wk WML: Constant p - value Mkt - RF p - value Adj. R-Sq. Winner: Constant p - value Mkt - RF p - value Adj. R-Sq. Loser: Constant p - value Mkt - RF p - value Adj. R-Sq. -0.0039 0.0074 -0.0635 0.4679 0.0001 0.0027 0.2301 -0.0843 0.3170 0.0010 0.0088 0.0084 -0.1256 0.0717 0.0049 0.0120 0.0830 -0.1908 0.0496 0.0096 0.9006 -0.0005 0.6842 0.8233 0.0023 0.2377 0.8448 0.0048 0.0618 0.8880 0.0039 0.3483 0.9128 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.4594 0.5414 0.4919 0.3368 0.3076 0.3826 0.4349 0.9579 0.0000 0.4118 -0.0038 0.1022 0.9660 -0.0055 0.2879 1.0499 -0.0174 0.0367 0.9846 0.0034 0.0085 0.8868 -0.0005 0.7913 0.9291 -0.0041 0.1682 1.0136 -0.0081 0.2428 1.1037 -0.0149 0.1558 0.9929 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.4785 0.3954 0.2934 0.3358 0.3725 0.3733 0.2830 0.2185 -0.0024 0.0180 -0.0534 0.3756 0.0010 0.0024 0.1564 -0.0258 0.6556 -0.0014 0.0082 0.0017 -0.0964 0.0825 0.0049 0.0086 0.0914 -0.1584 0.0474 0.0129 -0.0004 0.6511 0.7942 0.0018 0.2426 0.8334 0.0044 0.0400 0.8696 0.0031 0.3586 0.8916 0.0404 0.0000 0.0000 0.0000 0.4443 0.4086 0.0020 0.0431 0.8476 -0.0006 0.6901 0.8592 0.0000 0.4752 0.0000 0.4394 Jegadeesh & Titman Decile Weights: 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk 0.0579 0.0000 -0.0557 0.6564 -0.0012 0.0000 113 0.0695 0.0000 0.0033 0.9821 -0.0022 0.0557 0.0000 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.13: Abnormal Returns Using CAPM: Chan, Hameed, & Tong's Country ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31, 2005 Notes: (1) CAPM: Dependent variable is periodic return for the WML which is he winner R-tt Rpt = a , + b , ■R M R I f+ e „ minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or e x c e ss periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic Treasury bill rate over the listed formation and holding period. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegadeesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each w eek during the sam ple period and held for the indicated period. (4) RMRF is the e x c e ss return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson A ssociates) over the listed formation and holding periods. (5) The sample includes only the ETFs for 19 of the 23 country indices studied by Chan, Hameed, and Tong (2000). The Diamonds ETF indexing the Dow Jones Industrial Average is added to iShares ETFs matching the MSCI Indices for Australia, Austria, Belgium, Canada, France, Germany, Hong Kong, South Korea, Italy, Japan, Netherlands, South Africa, Spain, Singapore, Switzerland, U.K., Taiwan, and Malaysia. Unfortunately, no ETFs are available for Denmark, Norway, Thailand, and Indonesia. (6) p-values are computed with robust standard errors correct ed for heteroskedasticity and autocorrelation using the Newey-W es adjustment (1987). 114 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abnormal Returns Using Fama & French's Three Factor Model: Chan, Hameed, & Tong's Country ETFs Table 2.14: Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31,2005 Lo & MacKinlay Difference from Average Return Weights: Jegadeesh & Titman Decile Weights: 4wk 4wk 12wk 12wk 26wk 26wk 1wk 1wk 2wk 2wk 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 26wk 26wk WML: 0.0066 0.0085 0.0506 -0.0041 0.0021 -0.0025 0.0020 0.0073 Constant 0.0131 0.0599 0.0203 0.2309 0.0109 0.1283 0.0001 0.0069 0.3697 0.0334 p - value 0.0751 0.0002 0.0111 -0.0359 -0.1211 0.0022 -0.0387 -0.0157 -0.0596 -0.0592 -0.1444 Mkt - RF 0.1172 0.8719 0.5824 0.1564 0.9867 0.6644 0.8621 0.3115 0.4747 p - value 0.1713 0.4543 0.1544 0.1128 0.2156 0.0149 0.0739 0.1896 HML -0.0043 0.2206 -0.0336 0.2494 0.0384 0.8478 0.1020 0.5808 0.9646 0.3150 0.2003 0.0693 p - value 0.7489 0.0402 -0.1141 -0.0392 0.0915 0.0495 0.1501 0.0416 0.0235 0.1288 SMB -0.2260 -0.1405 0.1162 0.3021 0.5953 0.3035 0.8168 0.7351 0.8563 0.2952 p - value 0.1342 0.5088 0.0153 0.0137 0.0016 -0.0031 0.0006 -0.0023 0.0018 0.0089 Adj. R-Sq. 0.0145 0.0056 Winner: 0.0277 0.0006 0.0015 -0.0028 -0.0011 0.0011 Constant -0.0009 0.0019 -0.0007 0.0423 0.4212 0.0000 0.3414 0.6810 0.4780 0.3919 0.5715 p - value 0.4786 0.8702 0.0000 0.9692 0.8887 0.9515 0.9888 0.9519 0.9392 0.9535 1.0079 Mkt - RF 0.9888 1.0228 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 p - value 0.0000 0.0000 0.2672 0.1959 0.3407 0.2788 0.3529 0.4051 0.3369 HML 0.4046 0.2113 0.2155 0.0010 0.0000 0.0000 0.0001 0.0086 0.0060 p - value 0.0033 0.0000 0.0004 0.0005 0.1292 0.2518 0.1432 0.2978 0.1613 0.1577 0.1406 0.2471 0.0594 SMB 0.2773 0.0052 0.0105 0.0000 0.1471 0.2383 0.1920 0.2070 0.0274 p - value 0.4058 0.0017 0.4957 0.5321 0.4276 0.5661 0.3494 0.3196 0.4568 0.4105 0.4439 Adj. R-Sq. 0.4401 Loser: -0.0112 0.0015 -0.0014 -0.0051 -0.0229 0.0030 -0.0010 -0.0054 Constant -0.0138 -0.0176 0.3624 0.0365 0.0356 0.0226 0.0303 0.6082 0.1389 0.0775 p - value 0.0483 0.1671 0.9404 1.0247 1.0903 0.9498 0.9483 0.9779 0.9693 1.0670 Mkt - RF 1.1332 0.9056 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 p - value 0.0000 0.0000 0.0000 0.0000 0.2832 0.2400 0.1895 0.2523 0.0415 0.2668 0.1473 0.1840 HML 0.2448 -0.0340 0.0003 0.5867 0.0224 0.0013 0.0011 0.0056 0.1516 0.0459 p - value 0.0321 0.7252 0.2574 0.0797 0.1017 0.3369 0.1197 0.1342 0.1184 SMB 0.0491 0.2854 0.4177 0.0202 0.0146 0.2594 0.5933 0.3277 0.1575 0.1725 0.2187 p - value 0.0076 0.0118 0.4127 0.4839 0.4831 0.3096 0.3411 0.3743 0.3757 0.4498 0.2942 Adj. R-Sq. 0.2378 115 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2.14: Abnormal Returns Using Fama & French's Three Factor Model: Chan, Hameed, & Tong's Country ETFs Exchange Traded Funds and Momentum Returns: March 21,1996 - December 31,2005 Notes: (1) Fama & French's three factor model: R u ~ R f, = a f + b r l iM RF t + s t • SMB , + h, ■HML, + e it Dependent variable is periodic return for the WML which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or e x c ess periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic Treasury bill rate over the listed formation and holding period. (2) Formation period includes daily returns from Thursday through Wednesday; holding period includes daily returns from Friday through Thursday. (3) The Lo & MacKinlay winners represent the ETFs w hose returns outperform the average return of all ETFs during the formation period; the losers represent the ETFs w hose returns underperform the average return of all ETFs during the formation period. Both winner and loser ETFs are weighted proportionally to their deviation from the average ETF return and held for the indicated period. (4) The Jegadeesh & Titman winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sam ple period and held for the indicated period. (5) RMRF is the e x c e ss return on the market portfolio, which is the value-weighted return on all NYSE, AMEX, and NASDAQ stocks from CRSP minus the periodic Treasury bill rate (from Ibbotson Associates) over the listed formation and holding periods. (6) HML (High Book to Market Minus Low Book to Market) is the average return on the two value portfolios minus the average return on the two growth portfolios, i.e. HML = 1/2 (Small Value + Big Value) -1 /2 (Small Growth + Big Growth). (7) SMB (Small Size Minus Big Size) is the average return on the three small portfolios minus the average return on the three big portfolios, SMB = 1/3 (Small Value + Small Neutral + Small Growth) -1 /3 (Big Value + Big Neutral + Big Growth). (8) The sample includes only the ETFs for 19 of the 23 country indices studied by Chan, Hameed, and Tong (2000). The Diamonds ETF indexing the Dow Jones Industrial Average is added to iShares ETFs matching the MSCI Indices for Australia, Austria, Belgium, Canada, France, Germany, Hong Kong, South Korea, Italy, Japan, Netherlands, South Africa, Spain, Singapore, Switzerland, U.K., Taiwan, and Malaysia. Unfortunately, no ETFs are available for Denmark, Norway, Thailand, and Indonesia. (9) p-values are computed with robust standard errors correcl ed for heteroskedasticity and autocorrelation using the Newey-W es [adjustment (1987). 116 ESSAY #3 MOMENTUM/CONTRARIAN ABNORMAL RETURNS AND NON-U.S. EXCHANGE TRADED FUNDS Abstract: Investing in portfolios of non-U.S. exchange traded funds (ETFs) from Australia, Canada, France, Germany, Hong Kong, Japan, and the U.K. provides risk adjusted abnormal returns. Short formation and holding periods of 1 day to 8 weeks provide abnormal contrarian returns as past losers become winners and past winners become losers or at least generate smaller positive returns than the past losers. Longer formation and holding periods o f 16 weeks to 52 weeks provide abnormal momentum returns as past winners keep winning and past losers generate smaller positive returns than the past winners. Abnormal contrarian returns result for portfolios o f non-U.S. ETFs when returns are adjusted for risk using international versions o f both the Capital Asset Pricing Model and Fama & French’s three factor model; however, abnormal momentum returns result only when returns are adjusted for risk using an international version of CAPM. 1. Introduction: Momentum and contrarian investment strategies result in abnormal returns for stocks, mutual funds, country indices, and U.S. exchange traded funds (ETFs), providing evidence of anomalies to both the theory of market efficiency and the benchmark asset pricing models. Since Jegadeesh and Titman’s (1993) seminal study unearthing medium term momentum profits for U.S. stocks over 3 to 12 month holding periods, numerous studies have confirmed momentum profits in many markets including: U.S. equities (Jegadeesh & Titman (1993,2001), Hong, Lim, & Stein (2000)), U.S. mutual funds (Grinblatt, Titman, & Wermers (1995), Carhart (1997), Wermers (2003), Sapp & Tiwari (2004)), U.S. industries (Moskowitz & Grinblatt (1999)), U.S. exchange traded funds (ETFs) (De Jong (2007a, 2007b)), international equity markets (Rouwenhorst (1998, 1999), Chan, Hameed, & Tong (2000), Balvers & Wu (2006)), and foreign exchange 117 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. markets (Okunev & White (2001)). Gebhardt, Hvidkjaer, & Swaminathan (2002) find no evidence o f momentum among investment grade corporate bonds, but find contrarian returns as the bonds experienced significant reversals as well as a momentum spillover from stocks to bonds of the same firm as past equity returns are good predictors of future bond rating changes. Momentum/contrarianism is an anomaly in the sense that it violates market efficiency in its weak form as investors can use technical analysis of past prices and returns to form portfolios of winners and losers and so earn arbitrage profits on a zero investment portfolio that buys the winners and shorts the losers or buys the losers and shorts the winners, respectively. Momentum is also an anomaly in the sense that the standard empirical asset pricing model using Fama & French’s three factor model cannot explain the medium term return continuation of momentum. Other studies accept the momentum/contrarian anomaly as a stylized fact and seek to explain what causes momentum/contrarianism using behavioral finance theory as rational, market efficiency has difficulty explaining its existence. Behavioral finance posits various explanations of momentum, including investor overreaction or underreaction (Chan, Jegadeesh, & Lakonishok (1996), Hong & Stein (1999)), expectation extrapolation, conservatism in expectation updating (Barberis, Shleifer, Vishny (1998)), biased self attribution or investor overconfidence (Daniel, Hirshleifer, & Subrahmanyam (1998)), disposition effects (Grinblatt & Moskowitz (2004)), selective information conditioning, and herding behavior by investors (Jordan (2004)) and mutual fund managers (Grinblatt, Titman, & Wermers (1995), Wermers (2003)). Rational, market efficiency supporters fire back with the alternative explanations of momentum 118 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. returns being unrealizable due to transactions costs (Lesmond, Schill, & Zhou (2004)) or being fair compensation for risk (Conrad & Kaul (1998)) or being the product of data mining. Clearly, the momentum anomaly requires further research, as current studies agree that a momentum investment strategy generates abnormal returns, but disagree as to what causes a momentum strategy to be successful and disagree as to whether the abnormal profits are realistically attainable by an actual investor or arbitrageur after recognizing appropriate transactions costs, like bid ask spreads, brokerage commissions, the price impact of large trades, and the higher capital gains taxes associated with increased trading. Chan, Hameed, & Tong (2000) find momentum returns in 23 country index returns during 1980 to 1995 o f at least 1% per month for formation and holding periods of 1 week, 2 weeks, or 4 weeks. They use Lo & MacKinlay’s (1990) weighting scheme where portfolio weights reflect the country’s past performance relative to the average past performance of all 23 countries; above average performers are purchased and below average performers are shorted, but all countries may have a non-zero weight in the winner minus loser portfolio. This weighting scheme differs from Jegadeesh & Titman’s usual methodology of buying long the top decile of winners and shorting the bottom decile of losers. Chan, Hameed, & Tong find that 80 to 90% of the momentum profits are due to equity return predictability and only a small portion is due to exchange rate predictability, and they find that momentum profits are larger for country indices following an increase in trading volume. Their momentum profits would be difficult to obtain after transactions costs as country indices are not directly investable and some of 119 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. their markets restrict short selling. Balvers & Wu (2006) jointly consider momentum and mean reversion for 18 developed country index returns from 1970 to 1999. Their model generates a signal that identifies the winner and loser countries by its indicator score incorporating both momentum and mean reversion information, resulting in excess returns o f 1.1 - 1.7% per month, which outperforms both pure momentum and pure mean reversion strategies. Balvers & Wu find a strong negative correlation of - 35% between momentum and mean reversion effects, which explains why controlling for mean reversion effects can improve momentum returns. The studies of both Chan, Hameed, & Tong (2000) and Balvers & Wu (2006) can be improved by using ETFs rather than country indices since ETFs indexing various country stock indices are readily investable and shortable which are necessary conditions to realistically implement a momentum or contrarian strategy. My motivation in this study is to extend the momentum/contrarian anomaly to the non-U.S. ETFs listed and traded in Australia, Canada, France, Germany, Hong Kong, Japan, and the United Kingdom. Confirming momentum/contrarian abnormal returns in non-U.S. ETFs helps to dispel the myth that this anomaly is merely the product of data mining within the U.S. securities’ data sets. ETFs are powerful and flexible investment vehicles that combine the diversified portfolio features of mutual funds with the trading possibilities o f individual securities. Currently, all U.S. ETFs and most non-U.S. ETFs function similarly to passively managed index mutual funds, as they are composed of a portfolio o f stocks or bonds that track a particular index, thus providing diversification within the portion of the market tracked by that index. Four general categories of ETFs 120 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. include: (1) broad based domestic indices, some style specific indices in both a “value” and a “growth” version, and size based indices including large cap, mid cap, small cap, and micro cap, (2) sector indices including consumer, energy, financial, health, natural resources, real estate, utilities, and technology, (3) international indices including global stock indices, regional indices, and country specific indices, and (4) bond indices of varying maturities and credit risk. What differentiates an ETF from a mutual fund is an ETF trades on an exchange like a stock, enabling an ETF to be: purchased or sold at intraday market prices, purchased on margin, sold short, and traded via stop orders and limit orders. Ordinary mutual funds can only be purchased and sold by market orders for end of day prices, and cannot be purchased on margin or sold short, which prevents the usual zero investment momentum and contrarian portfolios of buying the winners and shorting the losers or of buying the losers and shorting the winners. Also, many mutual funds have redemption fees and other constraints to discourage or prevent the short term trading necessary to implement a momentum or contrarian strategy. For implementing a momentum or contrarian strategy, purchasing or shorting an ETF gives the arbitrageur or investor a diversified portfolio of stocks while incurring only one bid ask spread and one round trip commission, clearly a cost advantage over assembling a portfolio of individual winner and loser stocks, which entails many bid ask spreads and many commissions. My research questions are: (1) whether a momentum investing strategy of buying winners and shorting losers generates abnormal returns in the non-U.S. ETF market, (2) whether a contrarian investing strategy of buying losers and shorting winners generates abnormal returns in the non-U.S. ETF market, and (3) which formation period, holding 121 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. period is optimal for momentum investing and for contrarian investing in the non-U.S. ETF market. My results contribute to the literature an affirmative answer to the first research question about momentum, as buying the winner decile of non-U.S. ETFs and shorting the loser decile o f non-U.S. ETFs provide statistically significant abnormal returns for formation and holding periods of 16 weeks, 20 weeks, 26 weeks, 39 weeks and 52 weeks with risk adjustment by an international version of CAPM; however, with risk adjustment by an international version o f Fama & French’s three factor model, no statistically significant abnormal returns occur since both the winners and losers generate positive abnormal returns that offset in the winner minus loser portfolio. The annualized momentum abnormal returns range from 5.4% to 17.6% under an international version of the Capital Asset Pricing Model. My results contribute to the literature an affirmative answer to the second research question about contrarianism, as buying the loser decile of non-U.S. ETFs and shorting the winner decile of non-U.S. ETFs provide statistically significant contrarian abnormal returns for formation and holding periods of 1 day, 1 week, 2 weeks, and 8 weeks with risk adjustment by an international version of CAPM and for formation and holding periods of 1 day, 1 week, 2 weeks, 4 weeks, 8 weeks, and 12 weeks with risk adjustment by an international version of Fama & French’s three factor model. The annualized contrarian abnormal returns range from 10.3% to 63.7% under an international version o f the Capital Asset Pricing Model and from 5.7% to 65.6% under an international version of Fama & French’s three factor model. Following the classic approach o f Jegadeesh & Titman (1993), I find for question (3) that a 26 week formation and holding period provides the highest annualized abnormal returns o f 17.6% 122 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. to a non-U.S. ETF momentum strategy using an international version of CAPM to adjust for risk, while no formation and holding period provides a statistically significant abnormal return to a non-U.S. ETF momentum strategy using an international version of Fama & French’s three factor model to adjust for risk. Also, a 1 day formation period and a 1 day holding period provide the highest abnormal returns to a non-U.S. ETF contrarian strategy, with annualized abnormal returns of 63.7% using an international version of CAPM and 65.6% using an international version of Fama & French’s three factor model. These research questions are important, because momentum and contrarianism are both wide spread anomalies identified by researchers as well as investment strategies used by practitioners like mutual funds and individual investors to attempt to earn abnormal returns. With the growth in ETFs from their introduction in 1973 with the Canadian Pacific Ltd. HOLDRS Fund to 2007’s assortment of 450 ETFs in the 7 countries studied, consisting of 3 listed and traded in Australia, 32 in Canada, 140 in France, 198 in Germany, 11 in Hong Kong, 6 in Japan, and 60 in the U.K., ETFs are on a growth path which should soon surpass the amount invested in equity index mutual funds. With the growing popularity of ETFs by traders and investors such an innovative financial product merits further study, especially when it can generate abnormal returns via a momentum or contrarian strategy. Most momentum/contrarian studies to date look at U.S. stocks, U.S. bonds, U.S. domestic mutual funds, international stocks, or country indices separately. By finding abnormal returns in momentum/contrarian portfolios o f non-U.S. ETFs from Australia, Canada, France, Germany, Hong Kong, Japan, and the U.K., I contribute to the literature 123 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. by demonstrating momentum/contrarianism in a changing country allocation setting which includes non-U.S. ETFs following various stock and bond indices, country, regional, and world indices, as well as sector or industry indices. Such results compliment and enrich the literature, since momentum/contrarian abnormal returns in non-U.S. ETFs both strengthen the case that the anomaly is not the product of data mining U.S. securities’ returns as well as indicate a potential tool for international portfolio managers to boost risk-adjusted returns by using past securities’ returns. 2. Methodology: Since Jegadeesh & Titman (1993) establish the generally accepted methodology for researching the momentum anomaly, I follow their methodology with appropriate adjustments to accommodate my study of non-U.S. ETFs. Since ETFs represent diversified portfolios designed and passively managed to track various domestic, sector, international, and bond indices, forming portfolios of non-U.S. ETFs is less necessary than Jegadeesh & Titman’s method of forming decile portfolios o f individual stocks based on formation period performance. However, when Moskowitz & Grinblatt (1999) study industry momentum they define the winner and loser portfolios as the top or bottom 3 out of 20 industries, respectively, when ranked by formation periods of 1, 6, or 12 months, and measured holding period returns for periods of 1, 6,12,24, or 36 months. Most momentum studies o f mutual funds also group winner and loser funds into portfolios; at least with ETFs it is possible to short the losers while shorting mutual funds is not possible. With the relative newness of ETFs such long periods as 24 or 36 months 124 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. would restrict the sample sizes to be rather small, as the end of 2000 found only 8 nonU.S. ETFs in existence in my 7 country study, 4 in Canada, 1 in Germany, 1 in Hong Kong, and 2 in the U.K. The year 2006 brought the largest number of new ETFs in my 7 country study, as 190 ETFs were introduced during the year, resulting in 384 ETFs in existence by year end, consisting of 6 listed and traded in Australia, 27 in Canada, 126 in France, 150 in Germany, 11 in Hong Kong, 6 in Japan, and 58 in the U.K. Table 3.1 shows the annual growth in non-U.S. ETF offerings by country from 1996 to the present. Following Jegadeesh & Titman’s methodology, I define the winner ETFs as the top performing decile over various formation periods and the loser ETFs as the poorest performing decile over various formation periods, and then form the momentum portfolio that buys the winner ETFs and shorts the loser ETFs over various holding periods. Also, adapting Jegadeesh & Titman’s methodology, I form the winner minus loser portfolio each week to increase the power of my tests; I equally weight the appropriate winner and loser ETFs in the portfolios formed each week during the sample period and held for the indicated amount o f time. Also, with ETFs listed and traded in all 7 countries but returns converted to U.S. dollars, I measure momentum with all 7 countries’ ETFs pooled together to consider a strategy where market performance may favor one country’s ETFs over the other 6 countries’, which can determine whether a momentum-based country allocation strategy exists, as well as to maximize my sample size. My sample period runs from January 1, 2000 to March 31,2007, a period of 378 weeks, with 8 ETFs available in 2000, so that the top and bottom deciles begin with 1 ETF each as winners and losers, respectively. In 2007, 450 ETFs are available so that the deciles of winners and losers 125 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. both include 45 ETFs. I consider formation periods of 1 day, 1 week, 2 weeks, 4 weeks, 8 weeks, 12 weeks, 16 weeks, 20 weeks, 26 weeks, 39 weeks, and 52 weeks coupled with holding periods o f the same length as the formation period for a total of 11 different momentum strategies; these formation and holding period combinations parallel the methodology of De Jong 2007a using U.S. ETFs. Following Jegadeesh & Titman’s methodology, I also consider the above 10 weekly momentum strategies with a one week lag between portfolio formation and holding period to allow a more realistic time for an investor to determine the winner and loser ETFs and form the appropriate portfolios while avoiding some price pressure and perhaps minimizing the transactions costs compared to a hurriedly assembled portfolio. For the 1 day, 1 day momentum/contrarian strategy, I use an extra one day lag rather than an extra week lag. For each combination of formation and holding period, I compute annualized abnormal returns for the winner ETFs, the loser ETFs, and the zero investment momentum portfolio of buying the winner ETFs and shorting the loser ETFs using an international version of CAPM and an international version of Fama & French’s three factor model to adjust for risk. Clearly, the zero investment contrarian portfolio of buying the loser ETFs and shorting the winner ETFs simply reverses the sign of the zero investment momentum portfolio. 3. Description of Data: Since the non-U.S. ETFs are listed and traded on the 7 countries’ stock exchanges, the daily adjusted price data is available from Datastream, as are the Morgan Stanley Capital International country and world indices and the Reuters currency 126 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. exchange rates. Momentum/contrarian portfolio risk characteristics like factor loadings on various risk measures like the market risk premium, RM- R F, small firm size minus big firm size, SMB, and high book to market value minus low book to market value, HML are calculated to confirm that abnormal momentum/contrarian returns are not due to different risk levels, or to different firm characteristics, like size or book-to-market ratios, or to different industry compositions, or to different value or growth measures. Daily data on the zero investment factor mimicking portfolios for firm size and book-tomarket ratios, i.e., SMB and HML, are available on Kenneth R. French’s website at: http://mba.tuck.dartmouth.edu/pages/facultv/ken.french/data library.html. Daily data on the zero investment factor mimicking portfolio for the market risk premium, RM- R F, combines the daily returns of the MSCI World Index in U.S. dollars from Datastream with the appropriate U.S. Treasury interest rates from Kenneth R. French’s website above, which he attributes to Ibbotson Associates. 4. Outline of Model: To evaluate the various ETFs in terms of risk levels, I use an international version of CAPM and an international version of Fama & French’s three factor model. Thus, the international versions of factor models used are: CAPM: Rit - R Ft = a , + b, ■RMRF,* + e„ Fama & French’s three factor model: R„ - RFi = a ,+ b, ■R M R F ' + s, • SMB, + h, ■HML, + e„ 127 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. where Rjt is the dollar return on portfolio i in period t, RFt is the U.S. Treasury bill rate in period t, RMRF* is the excess dollar return on the value - weighted MSCI World Index above the U.S. Treasury bill rate, and SMB and HML are the dollar returns on zero investment, value-weighted, factor-mimicking portfolios computed for U.S. securities for firm size and book-to-market, respectively. Balvers & Wu (2006) use a similar model blending a world market excess dollar return with Fama and French’s U.S. size and book-to-market factors to adjust for risk with non-U.S. securities. Thus, using both models, I calculate the cross-sectional a 's for the excess return on the winner ETFs, the excess return on the loser ETFs, as well as the winner minus loser momentum portfolio to measure abnormal positive or negative returns. For each non-U.S. ETF, the daily Datastream adjusted closing price in its local currency is converted to U.S. dollars using the midpoint o f the Reuters’ bid and offered currency exchange rates available for the same time period as the adjusted closing price. I compute daily holding period returns in dollars from the daily adjusted closing prices converted to U.S. dollars for each non-U.S. ETF. 5. Results: A. Raw Returns: Table 3.2 shows the annualized mean raw returns and t-statistics for the 11 momentum strategies with information for the winner minus loser portfolio as well as the winner and loser portfolios separately reported. The momentum returns for the formation and holding periods o f 16 weeks to 52 weeks are economically significant, ranging from 128 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. annualized winner minus loser returns of 2.8% to 17.2%. Also, most of the winner portfolios keep winning and the losers reverse to generate positive returns, with winner annualized returns ranging from 17.6% to 25.1%, and with loser annualized returns ranging from 7.9% to 15.0%. Momentum returns are maximized at 17.2% with a 26 week formation period and a 26 week holding period. Contrarian returns are evident in the shorter formation and holding periods from 1 day to 8 weeks with mean annualized winner minus loser returns ranging from - 4.4% to - 63.6%. Thus, by reversing our momentum strategy to a contrarian strategy of buying the past losers and shorting the past winners, economically significant returns are possible with annualized loser returns ranging from 17.0% to 47.4% and annualized winner returns ranging from - 16.2% to 12.6% for the 1 day to 8 week formation and holding periods. The 12 week formation and holding period is not clearly a momentum or contrarian strategy as both winners and losers continue to win and nearly offset, resulting in less economic significance in the winner minus loser portfolio. However, all the raw returns are statistically insignificant as the largest t-statistic is only 0.43. These results are consistent with Balvers & Wu (2006) who find economically significant but statistically insignificant returns for their model which combines momentum and mean reversion for 16 international ETFs from April 1996 to December 2003. Balvers & Wu attribute the statistical insignificance to the shortness of the available sample period which may also be the problem in my study. Henker, Martens, & Huynh (2006) find statistically insignificant momentum returns for U.S. stocks in the 1993 - 2004 period due to the poor performance of momentum strategies 129 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. during the 2001 - 2004 subperiod. The statistical insignificance could also be a sample specific result from the added volatility in stock returns during the late 1990s and early 2000s. Thus, I next use international versions of both CAPM and Fama & French’s three factor model to adjust my momentum strategies for risk and identify statistically significant abnormal returns. 5. Results: B. Risk Adjusted Momentum Returns: Table 3.3 shows that the annualized momentum abnormal returns for winner minus loser portfolios are very statistically significant at the 1% level of significance for strategies of formation and holding periods from 20 weeks to 52 weeks with risk adjusted by an international version o f CAPM, with the winner minus loser annualized momentum abnormal returns ranging from 7.5% to 17.6%. However, no combinations of formation and holding periods yield a statistically significant momentum abnormal return using an international version of Fama & French’s three factor model to adjust for risk. The magnitudes and statistical significances of the annualized momentum abnormal returns are larger for the international version of CAPM than for the international version o f Fama & French’s three factor model, which is the expected result, as the international version of CAPM makes less complete risk adjustments. The annualized momentum abnormal returns are maximized for the 26 week formation and holding period strategy under the international version of CAPM with momentum returns of 17.6% and for the 52 week formation and holding period strategy under the international version of Fama & 130 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. French’s three factor model with momentum returns of 2.7% that are not significant at the 10% level. From the results for the excess returns above the appropriate periodic Treasury bill rate for the momentum winner and loser ETF portfolios, I find that the winners drive the winner minus loser results under both the international versions of CAPM and Fama & French’s three factor model for the longer formation and holding periods from 20 weeks to 52 weeks, while the losers drive the winner minus loser results for the shorter formation and holding periods from 1 day to 8 weeks. The winner annualized momentum abnormal returns are very significant at the 1% level for all formation and holding periods except 1 week and 2 weeks under the international version of CAPM, with magnitudes ranging from -18.8% to 21.9%, and for the formation and holding periods of 1 day, 20 weeks, 26 weeks, 39 weeks, and 52 weeks under the international version of Fama & French’s three factor model, with magnitudes ranging from -24.1% to 4.8%. The loser annualized momentum abnormal returns are very significant at the 1% level for all formation and holding periods under the international version of CAPM and for the shorter formation and holding periods from 1 day to 12 weeks under the international versions o f Fama & French’s three factor model with magnitudes ranging from 4.3% to 44.8%. Annualized contrarian abnormal returns are very statistically significant at the 1% level of significance for the formation and holding periods of 1 day, 1 week, and 8 weeks with risk adjusted by international versions of both CAPM and Fama & French’s three factor model. Other formation and holding periods produce annualized contrarian 131 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. abnormal returns that are statistically significant at the 5% or 10% level, with 2 weeks at the 10% level under the international version of CAPM, and with 2 weeks and 4 weeks at the 5% level and 12 weeks at the 10% level under the international version of Fama & French’s three factor model. The winner minus loser returns range from -5.7% to -65.6%. Thus, by reversing our momentum strategy to a contrarian strategy of buying the past losers and shorting the past winners, statistically significant returns are possible with abnormal annualized loser returns for the 1 day formation and holding period of 44.8% under the international version of CAPM and 41.6% under the international version of Fama & French’s three factor model, and abnormal annualized winner returns for the 1 day formation and holding period of -18.8% under the international version of CAPM and -24.1% under the international version of Fama & French’s three factor model. Thus, a formation and holding period of 1 day maximizes our contrarian abnormal returns, while a formation and holding period of 26 weeks or 52 weeks maximizes our momentum abnormal returns. 5. Results: C. Risk Adjustment under International Version of CAPM: Table 3.4 shows the winner minus loser portfolio has a negative beta or systematic risk measure with significantly negative coefficients on the market risk premium term for all formation and holding periods except 1 week and 2 weeks. Excess returns on all winner and loser portfolios have very significant positive betas with positive coefficients on the market risk premium and p-values of less than .0001 for all formation and holding 132 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. periods, except for 2 weeks’ loser with a p-value of .0002. The losers have higher betas than the winners across all formation and holding periods, indicating that the loser portfolios have a higher level o f systematic risk than the winners. Clearly, the resulting momentum abnormal returns to the winner minus loser ETFs for formation and holding periods from 16 weeks to 52 weeks are not due to a higher level of risk in the winner ETFs than the loser ETFs. Also, the higher level of systematic risk in the loser portfolios explains some but not all of the contrarian abnormal returns for formation and holding periods from 1 day to 8 weeks as a beta difference of 0.07 to 0.29 is too small to generate the magnitude of the contrarian returns. With risk adjustment under the international version of CAPM, momentum strategies of buying the winner ETFs and shorting the loser ETFs with formation and holding periods from 16 weeks to 52 weeks generate significant positive abnormal returns and contrarian strategies of buying the loser ETFs and shorting the winner ETFs with formation and holding periods of 1 day, 1 week, 2 weeks, and 8 weeks generate significant positive abnormal returns. 5. Results: D. Risk Adjustment under International Version of Fama & French’s three factor model: Table 3.5 shows the winner minus loser portfolio has a negative beta or systematic risk measure with significantly negative loadings on the market risk premium for all formation and holding periods except 1 day, 1 week, and 2 weeks. All winner and loser portfolios have very significant positive betas with positive factor loadings on the market 133 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. risk premium and p—values of less than .0001 for all formation and holding periods. The losers have higher factor loadings on the market risk premium than the winners across all formation and holding periods, indicating that the loser portfolios have a higher level of systematic risk than the winners. The winner minus loser portfolio has positive factor loadings on HML, the value minus growth risk premium, for all formation and holding periods except 1 week, with over half of the formation and holding periods significant at the 10% level and the rest insignificant. All of the winner portfolios have positive factor loadings on HML suggesting the winner ETFs contain value stocks with higher book to market ratios; also, all formation and holding periods are very significant at the 1% level except 1 week which is insignificant and 2 weeks which is significant at the 5% level. In contrast, the loser portfolios have lower positive factor loadings than the winners or negative factor loadings on HML with most significant suggesting the loser ETFs contain less value stocks than the winners or growth stocks with lower book to market ratios. Thus, book to market levels significantly distinguish winner and loser ETFs, as loser ETFs contain less value stocks or more growth stocks than winner ETFs. The winner minus loser portfolio has a positive factor loading on SMB, the small firm minus big firm risk premium for all formation and holding periods from 1 week to 26 weeks, with all being significant except 12 weeks, 16 weeks, and 20 weeks. The positive factor loadings suggest the winner ETFs contain smaller cap stocks than the loser ETFs for these time periods. The winner minus loser portfolio has a negative factor loading on SMB, for the 1 day, 39 week, and 52 week formation and holding periods with 134 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. both 39 and 52 weeks significant at the 1% level, which again suggests that the loser ETFs contain larger cap stocks than the winner ETFs. The winner ETFs load significantly positively on SMB for all formation and holding periods except the 12 week winners, which are positive but insignificant. The loser ETFs’ factor loadings on SMB are significantly positive for the 1 day and 12 week to 52 week formation and holding periods which are very significantly positive with 1 week to 8 week formation and holding periods insignificantly different from zero. In general, the winner ETFs have larger positive factor loadings on SMB indicating that the winner ETFs have significantly smaller cap stocks on average than the corresponding loser stocks, which are also small cap but not as small on average as the winner ETFs. Thus, small cap stocks in the ETFs account for the positive contrarian abnormal returns with the contrarian returns driven by the smaller cap stocks in the winners than the losers during the formation period. Clearly, the resulting abnormal returns to the winner minus loser ETFs are not due to a higher level of risk in the winner ETFs than the loser ETFs. The losers have higher systematic risk than the winners, the winners have higher book to market ratios than the losers, suggesting winners include more value stocks than losers, and the winners are generally smaller cap than the losers for all formation and holding periods from 1 week to 26 weeks. With risk adjustment under the international version of Fama & French’s three factor model, contrarian strategies of buying the loser ETFs and shorting the winner ETFs with formation and holding periods from 1 day to 8 weeks generate significant positive abnormal returns. No momentum strategies of buying the winner and shorting the loser generate statistically significant abnormal returns as both the winners and losers generate 135 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. positive abnormal returns for formation and holding periods o f 12 weeks to 52 weeks and so tend to offset one another in the winner minus loser portfolio. 6. Robustness of Results: A. Extra Time Between Formation and Holding Periods: If ETF returns are autocorrelated, then a momentum strategy that benefits from continuation of returns would appear to be profitable. Two possible explanations are nonsynchronous trading as the underlying stocks in some o f the non-U.S. ETFs may trade over different time periods than the actual ETFs trade on their home markets and bid ask bounce from pricing pressure attempting to buy past winners or to short past losers. Jegadeesh & Titman’s usual methodology is to allow an extra week between the portfolio formation period where the winners and losers are identified by past performance and the portfolio holding period where the winners are purchased and the losers are shorted. As in De Jong (2007a), I adapt Jegadeesh & Titman’s methodology to add an extra day between the formation and holding period for the 1 day strategy since an extra week seemed excessive; for all other formation and holding periods from 1 week to 52 weeks I add the usual week between the formation and holding periods. Such an adjustment should reduce the ability o f a momentum strategy to take advantage of the autocorrelation in ETF returns. Since contrarian strategies benefit from reversals of returns rather than continuations, it is not clear apriori what impact the extra week between formation and holding periods should have on the contrarian abnormal returns. 136 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.6 shows adding extra time between the formation and holding period does not eliminate the momentum and contrarian returns identified earlier in Table 3.3. Momentum strategies of buying past winner ETFs and shorting past loser ETFs continue to generate significant abnormal returns for all formation and holding periods from 16 weeks to 52 weeks when adjusting for risk using the international version of CAPM; when using the international version of Fama & French’s three factor model, the momentum abnormal returns continue to be statistically insignificant. The extra time generally has the largest impact on the shorter formation and holding periods, which is not surprising as the extra time represents a larger percentage change for the shorter formation and holding periods. The extra day removes about 41% of the contrarian abnormal returns for the formation and holding period o f 1 day, but a significant but smaller contrarian return still remains. The extra week removes the statistical significance o f the contrarian abnormal returns for formation and holding periods of 1 week and 2 weeks and generally decreases the magnitude and significance of the contrarian abnormal returns for formation and holding period o f 4 weeks and 8 weeks under both risk adjustment models. Surprisingly, the momentum abnormal returns increase in magnitude and statistical significance for formation and holding periods of 16 weeks, 20 weeks, and 26 weeks with risk adjustment by the international version of CAPM; for formation and holding periods of 39 weeks and 52 weeks, the magnitudes and significance levels of the momentum abnormal returns are essentially unchanged. 137 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6. Robustness of Results: B. Growth in Number of ETFs: Another issue is the rapid growth in the number of ETFs. Were the momentum and contrarian returns determined by the early years when a smaller number of ETFs existed? Table 3.7 considers this question by dividing the sample period into the early period from January 1, 2000 to December 31, 2003 with 8 to 41 non-U.S. ETFs in existence and the later period from January 1, 2004 to March 31, 2007 with 95 to 450 non-U.S. ETFs in existence. I distinguish these two sample periods for portfolios with a formation and holding period o f 26 weeks with risk adjusted by the international versions of both CAPM and Fama & French’s three factor model. Very significant momentum abnormal returns at the 1% level occur during both periods with risk adjustment by the international version of CAPM; however, the magnitude drops from an annualized momentum abnormal return of 18.4% during 2000 - 2003 to 4.7% during 2004 - 2007. However, with risk adjustment by the international version of Fama & French’s three factor model, statistically insignificant momentum abnormal returns over the entire period change to statistically significant annualized contrarian abnormal returns of 6.3% during 2000 - 2003 and very statistically significant annualized momentum abnormal returns of 17.1% during 2004 - 2007. Most of the factor loadings change dramatically from the early period to the later period; also, the factor loadings for 2000 - 2003 are very similar to the sign and significance for the entire period from 2000 to 2007 as denoted in Tables 3.4 and 3.5 previously. Clearly, the losers have a higher level of systematic risk during the early period, while the winners have a higher level of systematic risk during the later 138 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. period. Another difference between the two periods is that winner ETFs include value stocks, with high book to market ratios, during the 2000 - 2003 period but loser ETFs include value stocks, with high book to market ratios, during the 2004 - 2007 period. Also, during the 2000 - 2003 period, both the winners and losers include small cap stocks with the winners behaving like smaller cap stocks than the losers, while, during the 2004 - 2007 period, cap size is not a defining characteristic of either the winners or the losers. Dividing the sample period into two shorter periods illustrates some periodic differences between winner and loser ETFs in the two periods. However, such differences do not indicate a problem with the number of ETFs in existence, but more likely reflect the differing economic conditions in the different sub-periods as 1996 to early 2000 reflected a roaring bull market, followed by the bear markets of later 2000 through 2002, and the milder bull markets of 2003 to 2007. Table 3.8 segregates the momentum and contrarian abnormal returns by year using a 26 week formation and holding period by incorporating dummy variables for each year from 2000 to 2005 while using 2006 and the first quarter of 2007 as the base period. To control for the effects of forming portfolios during the different years, I augment the international versions o f CAPM and Fama & French’s three factor model with dummy variables identifying whether or not the winner, loser, and WML portfolios are formed during the years 2000, 2001, 2002, 2003, 2004, or 2005. Specifically, the augmented models are: 139 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. International Version of CAPM: Rit - RFt = or, + b, ■R M R F ’ + c, • Bum 2000 t + d r Bum 2001, + f , • Bum 2002, + g t • Bum 2003, + j, • Bum 2004, + kt ■Bum 2005, + eu International Version o f Fama & French’s three factor model: Rit - RFt = a t + bt ■R M R F ’ + si ■SMBt + hl ■HMLt + ct • B um 2000, + d i - B u m 2001, + / • Bum 20021+ g t - Bum 20031 + y, • Bum 2004, + k: ■Bum 20051+ eit where: BumYEARt = 1 if the portfolio at time t is formed during YEAR, and 0 otherwise, for YEAR = 2000, 2001, 2002, 2003, 2004, and 2005, and all the other variables are defined as before. In the augmented model, a t measures the abnormal momentum/contrarian returns in portfolios formed during 2006 or the first quarter of 2007 only, while c,, d t , f : , g n y ,, and kt measure the difference in abnormal momentum/contrarian returns in portfolios formed during 2000, 2001,2002, 2003, 2004, and 2005, respectively, relative to those formed during 2006 or the first quarter of 2007. With risk adjustment by the international version of CAPM, the base period annualized momentum abnormal return o f 14.9% is increased by portfolios formed in 2000 and 2001, decreased by portfolios formed in 2002, and flips to a contrarian abnormal return by portfolios formed in 2003 and 2004. With risk adjustment by the international version of Fama & French’s three factor model, the momentum abnormal returns remain statistically insignificant at the 10% level except for portfolios formed during 2003 and 2004, which result in very statistically significant contrarian abnormal returns at the 1% level. Table 3.8 supports the differing economic conditions argument raised above in 140 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.7, also providing evidence that the growth in number of non-U.S. ETFs is not problematic in measuring the momentum or contrarian abnormal returns. 6. Robustness of Results: C. Country Allocation: My results so far have not identified the source of the abnormal momentum or contrarian returns by country since I formed all winner and loser portfolios of non-U.S. ETFs from the asset pool that included ETFs from all seven countries studied, namely, Australia, Canada, France, Germany, Japan, Hong Kong, and the U.K. My study purposefully chooses all seven countries’ ETFs simultaneously to extend momentum/contrarian studies to the country allocation domain as well as to maximize the length o f the sample period studied, since ETFs are a relatively new investment vehicle. On average, about 4% of the ETFs in the winner and loser portfolios are Australian ETFs, about 20% are Canadian ETFs, about 17% are French ETFs, about 30% are German ETFs, about 8% are Japanese ETFs, about 10% are Hong Kong ETFs, and about 11% are U.K. ETFs. To control for the effects of including ETFs from the various countries, I augmented the international versions of CAPM and Fama & French’s three factor model with dummy variables identifying whether or not the winner, loser, and WML portfolios included ETFs selected from Australia, France, Germany, Hong Kong, Japan, and the U.K.; I treated ETFs from Canada as the base case since Canada had the most stable number of ETFs over the entire period studied. Specifically, the augmented models are: 141 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. International Version of CAPM: Ru ~ Rft = a i + • RMRF* + ai ■DumAustral ia, + f ■DumFrance , + g, ■DumGermany , + j, ■DumJapan , + k, ■DumHK , + u, ■D umUK, + e„ International Version o f Fama & French’s three factor model: Ru ~ Rft ~ a i + hi • R M R F' + s t ■SMB, + h, ■HML, + a, ■DumAustral ia, + f , ■DumFrance , + g, ■DumGermany , + j, ■DumJapan , + k, • D umHK, + u, • D um UK, + e„ where: DumCountry, —1 if the portfolio formed at time t includes ETFs from Country, and 0 otherwise, for Country = Australia, France, Germany, Hong Kong, Japan, and U.K., and all the other variables are defined as before. In the augmented model, a, measures the abnormal momentum/contrarian returns in portfolios consisting of Canadian ETFs only, while a,, f,, g ,, j,, k:, and u, measure the difference in abnormal momentum/contrarian returns in portfolios including ETFs from Australia, France, Germany, Japan, Hong Kong, and the U.K., respectively, relative to those consisting of Canadian ETFs only. Table 3.9 uses the international version of CAPM to adjust for risk and shows clear and very significant differences in how various countries’ ETFs contribute to abnormal momentum or contrarian returns for the eleven combinations of formation and holding periods as segregated by WML, Winner, and Loser portfolios. The inclusion of Canadian ETFs in the WML portfolio contributes significant annualized contrarian abnormal returns o f 24.5% for an 8 week formation and holding period and significant momentum abnormal returns for formation and holding periods of 16 weeks and 26 142 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. weeks, with annualized momentum abnormal returns o f 19.8% and 38.2%, respectively. Also, the inclusion of Canadian ETFs in both the winner and loser portfolios generate significantly positive abnormal returns for all formation and holding periods except the very short ones of 1 day, 1 week, and 2 weeks. The inclusion of Australian ETFs in the WML portfolio contributes significant annualized contrarian abnormal returns of 24.7% for a 4 week formation and holding period and significant momentum abnormal returns for formation and holding periods of 2 weeks, 20 weeks, 26 weeks, and 39 weeks, with annualized momentum abnormal returns ranging from 9.0% to 27.6%. Australian winners become losers and generate statistically significant negative abnormal returns for formation and holding periods of 4 weeks, 8 weeks, 16 weeks, 20 weeks, 26 weeks, and 52 weeks. Australian losers continue to be losers and generate statistically significant negative abnormal returns for formation and holding periods of 8 weeks, 16 weeks, 20 weeks, 26 weeks, 39 weeks, and 52 weeks. The inclusion of French ETFs in the WML portfolio contributes significant annualized momentum abnormal returns for formation and holding periods o f 12 weeks, 16 weeks, 26 weeks, and 39 weeks, ranging from 10.6% to 31.2%. Long term French winners continue to be winners over formation and holding periods of 39 weeks and 52 weeks and generate statistically significant positive abnormal returns, but for a formation and holding period of 26 weeks, French winners become losers and generate statistically significant negative abnormal returns. Long term French losers continue to be losers and 143 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. generate statistically significant negative abnormal returns for formation and holding periods of 39 weeks and 52 weeks. The inclusion o f German ETFs in the WML portfolio contributes significant contrarian abnormal returns for formation and holding periods of 1 day, 16 weeks, and 26 weeks, with annualized contrarian abnormal returns ranging from 16.4% to 47.5%. Long term German winners become losers and generate statistically significant negative abnormal returns for formation and holding periods of 39 weeks and 52 weeks. German losers continue to be losers and generate statistically significant negative abnormal returns for formation and holding periods of 2 weeks, 12 weeks, and 16 weeks. The inclusion o f Japanese ETFs in the WML portfolio contributes significant annualized contrarian abnormal returns ranging from 10.1% to 47.5% for formation and holding periods of 1 day, 20 weeks, and 26 weeks and significant momentum abnormal returns for formation and holding periods of 8 weeks, 39 weeks, and 52 weeks, with annualized momentum abnormal returns ranging from 5.9% to 17.8%. Japanese winners become losers and generate statistically significant negative abnormal returns for formation and holding periods of 1 day, 16 weeks, 20 weeks, 26 weeks, and 39 weeks. Long term Japanese losers continue to be losers and generate statistically significant negative abnormal returns for a formation and holding period of 52 weeks, while 1 day and 26 week losers reverse and become winners, generating significantly positive abnormal returns. The inclusion of Hong Kong ETFs in the WML portfolio contributes significant contrarian abnormal returns for formation and holding periods o f 12 weeks, 16 weeks, 20 144 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. weeks, 26 weeks, and 52 weeks, with annualized contrarian abnormal returns ranging from 8.0% to 29.9%. Long term Hong Kong winners become losers and generate statistically significant negative abnormal returns for formation and holding periods of 39 weeks and 52 weeks. Long term Hong Kong losers continue to be losers and generate statistically significant negative abnormal returns for a formation and holding period of 52 weeks, while 16 week and 26 week losers reverse and become winners, generating significantly positive abnormal returns. The inclusion o f U.K. ETFs in the WML portfolio contributes significant annualized contrarian abnormal returns of 55.0% for a formation and holding period o f 1 day and significant momentum abnormal returns for formation and holding periods of 20 weeks and 52 weeks, with annualized momentum abnormal returns ranging from 6.4% to 13.0%. U.K. winners become losers and generate statistically significant negative abnormal returns for formation and holding periods of 1 day, 12 weeks, 16 weeks, and 20 weeks, while 39 week winners continue to win by generating significantly positive abnormal returns. Most U.K. losers continue to be losers and generate statistically significant negative abnormal returns for formation and holding periods o f 8 weeks, 12 weeks, 16 weeks, 20 weeks, 26 weeks, 39 weeks, and 52 weeks, while 1 day losers reverse and become winners, generating significantly positive abnormal returns. Table 3.10 uses the international version of Fama & French’s three factor model to adjust for risk and also shows clear and very significant differences in how various countries’ ETFs contribute to abnormal momentum or contrarian returns for the eleven combinations of formation and holding periods as segregated by WML, Winner, and 145 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Loser portfolios. The inclusion of Canadian ETFs in the WML portfolio contributes significant annualized contrarian abnormal returns of 33.9% for an 8 week formation and holding period. Including Canadian ETFs in the winner portfolio contributes very significant annualized momentum abnormal returns for formation and holding periods of 39 weeks and 52 weeks. Also, the inclusion of Canadian ETFs in the loser portfolio generates significantly positive abnormal returns for formation and holding periods of 2 weeks, 4 weeks, 8 weeks, 12 weeks, 39 weeks, and 52 weeks, as the Canadian losers reverse and become winners. The inclusion o f Australian ETFs in the WML portfolio contributes significant annualized contrarian abnormal returns of 19.4% for a 4 week formation and holding period and significant momentum abnormal returns for formation and holding periods of 2 weeks, 12 weeks, 20 weeks, 26 weeks, and 39 weeks, with annualized momentum abnormal returns ranging from 12.4% to 28.1%. Australian winners continue to win and generate statistically significant positive abnormal returns for formation and holding periods of 12 weeks, 16 weeks, 20 weeks, 26 weeks, and 39 weeks. Australian losers continue to be losers and generate statistically significant negative abnormal returns for formation and holding periods o f 8 weeks, 16 weeks, 20 weeks, 26 weeks, and 39 weeks. The inclusion of French ETFs in the WML portfolio contributes significant annualized momentum abnormal returns for formation and holding periods o f 4 weeks, 8 weeks, 12 weeks, and 16 weeks, ranging from 19.6% to 26.2%. French winners continue to be winners over formation and holding periods o f 8 weeks, 12 weeks, 16 weeks, 39 weeks, and 52 weeks and generate statistically very significant positive abnormal returns. 146 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. French losers make insignificant abnormal return contributions for all formation and holding periods. The inclusion o f German ETFs in the WML portfolio contributes significant contrarian abnormal returns for formation and holding periods o f 1 day, 39 weeks, and 52 weeks, with annualized contrarian abnormal returns ranging from 6.0% to 50.0%. German winners become losers and generate statistically significant negative abnormal returns for formation and holding periods of 12 weeks, 16 weeks, 39 weeks, and 52 weeks. German losers continue to be losers and generate statistically significant negative abnormal returns for formation and holding periods of 8 weeks and 12 weeks. The inclusion of Japanese ETFs in the WML portfolio contributes significant annualized contrarian abnormal returns ranging from 14.4% to 45.0% for formation and holding periods of 1 day and 26 weeks and significant momentum abnormal returns for formation and holding periods of 8 weeks and 52 weeks, with annualized momentum abnormal returns ranging from 6.3% to 19.0%. Japanese winners become losers and generate statistically significant negative abnormal returns for formation and holding periods of 1 day, 16 weeks, 20 weeks, 26 weeks, and 39 weeks. Japanese losers reverse and become winners and generate statistically significant positive abnormal returns for formation and holding periods of 1 day and 26 weeks. The inclusion of Hong Kong ETFs in the WML portfolio contributes significant contrarian abnormal returns for formation and holding periods of 16 weeks, 20 weeks, and 26 weeks, with annualized contrarian abnormal returns ranging from 10.0% to 25.9%. Long term Hong Kong winners become losers and generate statistically 147 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. significant negative abnormal returns for formation and holding periods of 39 weeks and 52 weeks, while 26 week winners continue to win and generate statistically significant positive abnormal returns. Hong Kong losers reverse and become winners, generating significantly positive abnormal returns for formation and holding periods of 16 weeks, 20 weeks, and 26 weeks. The inclusion o f U.K. ETFs in the WML portfolio contributes significant annualized contrarian abnormal returns of 55.0% for a formation and holding period o f 1 day and significant momentum abnormal returns for formation and holding periods of 20 weeks, 26 weeks, 39 weeks, and 52 weeks, with annualized momentum abnormal returns ranging from 6.3% to 19.1%. Short term U.K. winners become losers and generate statistically significant negative abnormal returns for a formation and holding period of 1 day, while 26 week and 39 week winners continue to win by generating significantly positive abnormal returns. Most U.K. losers continue to be losers and generate statistically significant negative abnormal returns for formation and holding periods o f 8 weeks, 12 weeks, 20 weeks, 26 weeks, 39 weeks, and 52 weeks, while 1 day losers reverse and become winners, generating significantly positive abnormal returns. Clearly, this detailed country analysis indicates some differences between the holding period performances of the ETFs from the seven different countries which may lead to improvements over a naive momentum or contrarian strategy. 148 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7. Conclusion: This study extends Jegadeesh & Titman’s momentum/contrarian anomaly to a new domain, portfolios o f non-U.S. ETFs, which either buy the winners and short the losers or buy the losers and short the winners, respectively, in a country allocation setting, which includes Australia, Canada, France, Germany, Hong Kong, Japan, and the United Kingdom. Currently, most non-U.S. ETFs are passively managed to track an index, not actively managed to time the market or “beat the market” by loading up on high momentum stocks. Yet, in spite of this disadvantage to actively managed mutual funds, non-U.S. ETFs provided economically and statistically significant abnormal returns to contrarian strategies of buying the loser ETFs and shorting the winner ETFs with formation and holding periods of 1 day to 8 weeks, with risk adjustment by either the international version of CAPM or the international version of Fama & French’s three factor model, and to momentum strategies o f buying the winner ETFs and shorting the loser ETFs with formation and holding periods from 16 weeks to 52 weeks, with risk adjustment by the international version of CAPM. Also, this study is the first to demonstrate momentum and contrarianism in a changing country allocation setting using non-U.S. ETFs. ETFs are ideal instruments with which to implement a contrarian and momentum strategy, since ETFs allow the purchase or short sale of a diversified portfolio of securities for one commission and one bid ask spread with minimal price impact, and so ETFs represent a great improvement over previous studies using country indices which are not readily investable securities. Using international versions of both CAPM and Fama & French’s three factor model to adjust for risk, I find that the contrarian and 149 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. momentum abnormal returns available with non—U.S. ETFs can not be explained by rational differences in risk and so I provide further evidence of the momentum/contrarian anomaly’s attack on market efficiency. Such findings contribute support to the behavioral explanation of why momentum occurs as investors underreact to relevant information and so prices only gradually reflect relevant news and information. In the future, as the sample period of non-U.S. ETFs’ data is longer, I plan to check the longer term performance for 18, 24, and 36 months after portfolio formation, to determine the duration of momentum/contrarian performance and anticipate the normal long-term reversals as documented by De Bondt & Thaler (1985,1987), thus extending their results to a new domain of non-U.S. ETFs. Also, I anticipate that non-U.S. ETFs will continue to grow and expand their array of offerings where eventually I anticipate significant numbers of actively managed non-U.S. ETFs to more directly compete with mutual funds and yet appeal to larger investors and investors that trade more actively than permitted by mutual funds. With actively managed non-U.S. ETFs, I expect the usual hot performance chasing behavior by investors, resulting in ETF managers that utilize momentum or contrarian strategies thus magnifying the momentum or contrarian performance from the passively managed non-U.S. ETFs studied in this paper. ETFs will continue to adapt and evolve to provide more flexible investment vehicles for investors and traders, as well as future questions to research and investigate relative to momentum/contrarian strategies as new ETFs and more years of data become available. 150 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. REFERENCES Balvers, Ronald J. & Yangru Wu, 2006. Momentum and Mean Reversion Across National Equity Markets. Journal o f Empirical Finance, 13: 2 4 -4 8 . Barberis, Nicholas, Andrei Shleifer, & Robert Vishny, 1998. A Model of Investor Sentiment. Journal o f Financial Economics, 49: 307 - 343. Carhart, Mark M., 1997. On Persistence in Mutual Fund Performance. Journal o f Finance, 52: 57 - 82. Chan, Kalok, Allaudeen Hameed, & Wilson Tong, 2000. Profitability of Momentum Strategies in the International Equity Markets. Journal o f Financial and Quantitative Analysis, 35: 153 - 172. Chan, Louis K. C., Narasimhan Jegadeesh, & Josef Lakonishok, 1996. Momentum Strategies. Journal o f Finance, 51: 1681-1713. Conrad, Jennifer & Gautam Kaul, 1998. An Anatomy of Trading Strategies. Review o f Financial Studies, 11: 489 —519. Daniel, Kent, David Hirshleifer, & Avanidhar Subrahmanyam, 1998. Investor Psychology and Security Market Under- and Overreactions, Journal o f Finance, 53: 1839-1886. De Bondt, Werner F. M. & Richard Thaler, 1985. Does the Stock Market Overreact? Journal o f Finance, 40: 793-8 0 5 . De Bondt, Werner F. M. & Richard Thaler, 1987. Further Evidence On Investor Overreaction and Stock Market Seasonality. Journal o f Finance, 42: 557-5 8 1 . De Jong, Jr., Jack C., 2007. Momentum/Contrarian Abnormal Returns and Exchange Traded Funds. Unpublished Working Paper, University of Hawaii. De Jong, Jr., Jack C., 2007. Momentum Abnormal Returns and International Exchange Traded Funds. Unpublished Working Paper, University of Hawaii. Fama, Eugene F. & Kenneth R. French, 1993. Common Risk Factors in the Returns on Stocks and Bonds. Journal o f Financial Economics, 33: 3 - 56. Fama, Eugene F. & Kenneth R. French, 1996. Multifactor Explanations of Asset Pricing Anomalies. Journal o f Financial Economics, 51: 5 5 -8 4 . 151 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Gastineau, Gary L., 2002. The Exchange-Traded Funds Manual. (John Wiley & Sons, Inc., New York). Gebhardt, William R., Soeren Hvidkjaer, & Bhaskaran Swaminathan, 2002. Stock and Bond Market Interaction: Does Momentum Spill Over? Unpublished Working Paper, Axia Energy Europe, Ltd. London, U.K., University of Maryland, & Cornell University. Grinblatt, Mark & Tobias J. Moskowitz, 2004. Predicting Stock Price Movements from Past Returns: the Role of Consistency and Tax-Loss Selling. Journal o f Financial Economics, 71: 541 - 579. Grinblatt, Mark, Sheridan Titman, & Russ Wermers, 1995. Momentum Investment Strategies, Portfolio Performance, and Herding: A Study of Mutual Fund Behavior. American Economic Review, 85: 1088 - 1105. Henker, Thomas, Martin Martens, & Robert Huynh, 2006. The Vanishing Abnormal Returns of Momentum Strategies and “Front-Running” Momentum Strategies. Unpublished Working Paper. Hong, Harrison, Terrance Lim, & Jeremy C. Stein, 2000. Bad News Travels Slowly: Size, Analyst Coverage, and the Profitability of Momentum Strategies. Journal o f Finance, 55: 265 —295. Hong, Harrison & Jeremy C. Stein, 1999. A Unified Theory of Underreaction, Momentum Trading and Overreaction in Asset Markets. Journal o f Finance, 54: 2143-2184. Jegadeesh, Narasimhan & Sheridan Titman, 1993. Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency. Journal o f Finance, 48: 6 5 -9 1 . Jegadeesh, Narasimhan & Sheridan Titman, 2001. Profitability of Momentum Strategies: An Evaluation o f Alternative Explanations. Journal o f Finance, 56: 699 - 720. Jordan, Steve, 2004. Is Momentum A Self-Fulfilling Prophecy? Yale International Center for Finance Working Paper No. 0 4 -2 5 . Lesmond, David A., Michael J. Schill, & Chunsheng Zhou, 2004. The Illusory Nature of Momentum Profits. Journal o f Finance, 71: 349 - 380. Lo, Andrew W. & A. Craig MacKinlay, 1990. When Are Contrarian Profits Due to Stock Market Overreaction? Review o f Financial Studies, 3: 175 - 205. 152 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Moskowitz, Tobias J. & Mark Grinblatt, 1999. Do Industries Explain Momentum? Journal o f Finance, 54: 1249 - 1290. Okunev, John & Derek White, 2001. Do Momentum Based Strategies Still Work In Foreign Currency Markets? Unpublished Working Paper, B.T. Funds Management, Sydney, Australia, University of New South Wales. Poterba, James M. & John B. Shoven, 2002. Exchange Traded Funds: A New Investment Option for Taxable Investors. Massachusetts Institute of Technology Department of Economics Working Paper #02-07. Rouwenhorst, K. Geert, 1998. International Momentum Strategies. Journal o f Finance, 53: 2 6 7 -2 8 4 . Rouwenhorst, K. Geert, 1999. Local Return Factors and Turnover in Emerging Stock Markets. Journal o f Finance, 54: 1439-1464. Sapp, Travis & Ashish Tiwari, 2004. Does Stock Return Momentum Explain the “Smart Money” Effect? Journal o f Finance, 59: 2605 - 2622. Wermers, Russ, 2003. Is Money Really “Smart”? New Evidence on the Relation Between Mutual Fund Flows, Manager Behavior, and Performance Persistence. Unpublished Working Paper. University of Maryland. 153 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. o> CM 450 00 00 o CO CO U.K. CM LO CO 194 Total ETFs LO CM CO 85 00 r- S6 CO CO CO CO CO CO CD CO o> M* 00 o O) M- France CM lO 150 LO O 126 Hong Kong Germany CM CM CM CO CO CO 00 r*- CM CM CO CO LO Is- CO CO CM O O O O CM CM CO Is- o o o o CM CM | O) CM 20031 2004 2005 o> | 00 1999 20001 Canada CM Australia |Year I 19961 1997 TABLE 3.1I: Growth of Non-U.S. Exchange Traded Fund LO Is- 154 uedep 0) M " Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.2: Annualized Raw Returns Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31,2 D07 1day 1day 1wk 1wk 2wk 2wk 4wk 4wk 8wk 8wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk WML: -63.5750% -29.3384% -12.8518% Mean -0.1015 -0.1444 -0.1422 t - statistic -4.3771% -10.1088% -0.0452 -0.1448 -0.9516% -0.0172 5.4574% 0.1010 7.5067% 0.1572 17.1914% 0.4334 8.1833% 0.2787 2.8138% 0.1004 Winner: -16.2000% Mean -0.0483 t - statistic Loser: Mean t - statistic Notes: 47.3750% 0.1326 1.5132% 0.0091 7.1916% 0.0706 12.6295% 0.1694 10.2362% 0.2058 14.6337% 0.3470 17.5637% 0.4512 19.3791% 0.5386 25.0546% 0.7499 19.2151% 0.9098 17.7922% 1.2096 30.8516% 0.1900 20.0434% 0.1815 17.0066% 0.2104 20.3450% 0.3332 15.5853% 0.3096 12.1063% 0.2687 11.8724% 0.2954 7.8632% 0.2496 11.0319% 0.3478 14.9785% 0.5192 (1) Annualized mean raw returns in U.S. dollars for the period indicated. Periodic returns are annualized by multiplying by either 250 trading days for 1 day 1 day strategy or by 52 divided by the number of w eeks in the 10 weekly strategies. (2) Formation period includes daily returns in U.S. dollars from Thursday through Wednesday; holding period includes daily returns in U.S. dollars from Friday through Thursday. (3) Daily Datastream adjusted closing price in its local currency is converted to U.S. dollars using the midpoint of the Reuters bid and offered currency exchange rates available for the sam e time period a s the closing price. Daily holding period returns are computed from the adjusted closing prices converted to U.S. dollars for each non-U.S. ETF. (4) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period of time. (5) The WML is the zero net - inves tment portfo io created by buying the winner ETFs and by shorting the loser ETFs for t ie indicated holding period. 155 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.3: Annualized Abnormal Returns: Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31, 2007 iday iday 1wk 1wk 2wk 2wk 4wk 4wk 8wk 8wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk j International Version of CAPM: R/l ~ R Ft = a i + b t. • R MRF * + e l WML: 5.3684% 7.5033% 17.5824% 10.7140% Constant -63.6500% -29.3436% -12.9324% -4.4616% -10.3006% -1.1137% 7.9185% 0.0052 0.3949 0.0093 0.7191 0.0822 0.0000 0.0545 0.0089 0.0000 p - value 0.0000 0.0000 Winner: 4.5188% 9.7825% 7.7578% 11.9271% 14.7547% 16.5602% 21.9092% 15.4744% 13.9450% Constant -18.8000% -1.4508% 0.8457 0.0073 0.0069 0.3263 0.0010 0.0000 0.0000 0.0000 0.0000 p - value 0.0000 0.0000 Loser: 9.3863% 44.8250% 27.8928% 17.4512% 14.2441% 18.0583% 13.0407% 9.0568% 4.3268% 4.7604% Constant 6.0266% 0.0002 0.0000 0.0000 0.0001 0.0000 0.0000 0.0000 0.0000 p - value 0.0005 0.0000 0.0000 International Version of Fama - French's 3 Factor Mode : WML: Constant -65.6250% -31.8396% -15.2074% -13.9490% -13.6780% 0.0171 0.0012 0.0000 0.0056 0.0190 p - value Winner: Constant -24.0750% -7.0356% -2.1736% -1.0426% -0.7176% 0.7512 0.4131 0.6299 0.7811 p - value 0.0011 Loser: 41.5750% 24.8092% 13.0338% 12.9064% 12.9604% Constant 0.0000 0.0012 0.0076 0.0006 0.0000 p - value ~ R n = a , + b, ■R MRF / + s, • SMB , + h, • HML , + e„ -5.7144% 0.0617 -2.5623% 0.3765 1.6471% 0.5460 1.2506% 0.5096 2.6693% 0.2758 2.6918% 0.1743 1.4404% 0.4340 1.3592% 0.3771 3.3340% 0.0179 3.9306% 0.0018 3.9135% 0.0359 4.8259% 0.0032 7.1543% 0.0006 3.9215% 0.0499 1.6869% 0.3780 2.6800% 0.0371 1.2441% 0.2121 2.1340% 0.0275 156 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.3: Annualized Abnormal Returns: Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31, 2007 Notes: (1) Annualized abnormal returns for the period indicated. Periodic abnormal returns, i.e. alpha intercepts, are annualized by multiplying by either 250 trading days for 1 day 1 day strategy or by 52 divided by the number of w eeks in the 10 weekly strategies. (2) Formation period includes daily returns in U.S. dollars from Thursday through Wednesday; holding period includes daily returns in U.S. dollars from Friday through Thursday. (3) Daily Datastream adjusted closing price in its local currency is converted to U.S. dollars using the midpoint of the Reuters bid and offered currency exchange rates available for the sam e time period as the closing price. Daily holding period returns are computed from the adjusted closing prices converted to U.S. dollars for each non-U.S. ETF. (4) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each w eek during the sample period and held for the indicated period of time. (5) The WML is the zero net - investment portfolio created by buying the winner ETFs and by shorting the loser ETFs for t he indicated holding period. (6) p-values are computed with robust standard errors corrected for heteroskedastici ty and autocorrelation using the Newey-W est adjustmen (1987). 157 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.4: Abnormal Returns Using International Version of CAPM Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31,2 307 1day 1day 1wk 1wk 2wk 2wk 4wk 4wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk 8wk 8wk WML: Constant p - value Mkt - RF p - value Adj. R-Sq. -0.0025 0.0000 -0.1151 0.0795 0.0026 -0.0056 0.0052 -0.0683 0.6516 -0.0014 -0.0050 0.0545 -0.1712 0.1791 0.0072 -0.0034 0.3949 -0.3549 0.0066 0.0348 -0.0158 0.0093 -0.2868 0.0467 0.0205 -0.0026 0.7191 -0.5078 0.0000 0.0750 0.0165 0.0822 -0.6768 0.0000 0.1133 0.0289 0.0089 -0.4340 0.0004 0.0486 0.0879 0.0000 -0.5337 0.0000 0.0907 0.0804 0.0000 -1.0223 0.0000 0.4509 0.0792 0.0000 -1.1791 0.0000 0.4841 Winner: Constant p - value Mkt - RF p - value Adj. R-Sq. -0.0008 0.0069 0.6137 0.0000 0.1594 -0.0003 0.8457 0.7796 0.0000 0.2389 0.0017 0.3263 0.7519 0.0000 0.2912 0.0075 0.0073 0.6937 0.0000 0.2384 0.0119 0.0010 0.7475 0.0000 0.3065 0.0275 0.0000 0.6973 0.0000 0.2514 0.0454 0.0000 0.5335 0.0000 0.1379 0.0637 0.0000 0.6618 0.0000 0.2092 0.1095 0.0000 0.5743 0.0000 0.1515 0.1161 0.0000 0.5069 0.0000 0.2136 0.1395 0.0000 0.3571 0.0000 0.1595 Loser: Constant p - value Mkt - RF p - value Adj. R-Sq. 0.0018 0.0000 0.7288 0.0000 0.1979 0.0054 0.0001 0.8479 0.0000 0.2956 0.0067 0.0002 0.9231 0.0002 0.3726 0.0110 0.0000 1.0486 0.0000 0.4643 0.0278 0.0000 1.0344 0.0000 0.3907 0.0301 0.0000 1.2051 0.0000 0.5230 0.0289 0.0000 1.2102 0.0000 0.5204 0.0348 0.0000 1.0958 0.0000 0.4504 0.0216 0.0005 1.1080 0.0000 0.6126 0.0357 0.0000 1.5292 0.0000 0.8428 0.0603 0.0000 1.5362 0.0000 0.7575 158 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.4: Abnormal Returns Using International Version of CAPM Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31, 2007 Notes: (1) International CAPM: R it - R Ft =f a t + -R MRF / + e u Dependent variable is periodic return for the WM L which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or e x c e s s periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic U.S. Treasury bill rate over the listed formation and holding period. (2) Formation period includes daily returns in U.S. dollars from Thursday through Wednesday; holding period includes daily returns in U.S. dollars from Friday through Thursday. (3) Daily Datastream adjusted closing price in its local currency is converted to U.S. dollars using the midpoint of the Reuters bid and offered currency exchange rates available for the sam e time period a s the closing price. Daily holding period returns are computed from the adjusted closing prices converted to U.S. dollars for each non-U.S. ETF. (4) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period of time. (5) RMRF is the e x c e s s dollar return on the market portfolio, which is the dollar return on the value-weighted MSCI World Index from Datastream minus the periodic U.S. Treasury bill rate (from Ibbotson A ssociates) over the listed formation and holding periods. (6) p-values are computed with robust standard errors corrected for heteroskedastici ty and autocorrelation using the Newey-West adjustmen (1987). 159 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.5: Abnormal Returns Using International Version o f Fama & French's Three Factor Model Non-U.S. Exchange Traded Funds and M om entum Returns: January 1, 2000 - March 3 1 ,2 D07 1day 1day 1wk 1wk WML: Constant p - value Mkt - RF p - value HML p - value SMB p - value Adj. R-Sq. Winner: Constant p - value Mkt - RF p - value HML p - value SMB p - value Adj. R-Sq. Loser: Constant p - value Mkt - RF p - value HML p - value SMB p - value Adj. R-Sq. 2wk 2wk 4w k 4wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk 8wk 8wk -0.0469 0.4540 0.2074 0.0964 -0.0892 0.4006 0.0083 -0.0061 0.0056 -0.1242 0.3724 -0.0403 0.8577 0.5316 0.0151 0.0348 -0.0058 0.0190 -0.1788 0.1404 0.0619 0.7519 0.2779 0.0751 0.0156 -0.0107 0.0171 -0.2956 0.0296 0.4896 0.0052 0.6325 0.0013 0.0979 -0.0210 0.0012 -0.4480 0.0009 0.0498 0.7955 0.6530 0.0084 0.0664 -0.0132 0.0617 -0.5156 0.0000 0.2242 0.1516 0.2851 0.2764 0.0836 -0.0079 0.3765 -0.5485 -0.0010 0.0011 0.7346 -0.0014 0.4131 0.7940 -0.0008 0.6299 0.8094 -0.0008 0.7811 0.7959 -0.0011 0.7512 0.7434 0.0000 0.0000 0.0000 0.0000 0.3755 0.0004 0.1485 0.0765 0.1812 0.2213 0.2392 0.5046 0.0019 0.2795 0.3501 0.0115 0.4096 0.0002 0.3388 0.6022 0.0017 0.7815 0.0048 0.0012 0.9182 0.0000 0.1681 0.0450 0.2377 0.0007 0.2073 -0.0026 0.0000 0.0000 0.0000 0.5820 0.5148 0.0232 0.1704 0.5353 0.1474 0.0063 0.5460 -0.3387 0.0029 0.3701 0.0746 0.1397 0.6161 0.0665 0.0063 0.5096 -0.3494 0.0014 0.9417 0.0033 0.4340 0.6051 0.0042 0.3771 0.4842 0.0000 0.0000 0.4306 0.0001 0.5653 0.4328 0.0001 0.8330 0.8000 0.3889 0.0165 0.0006 1.1206 0.0000 0.0200 0.2758 -0.7953 0.0269 0.1743 -1.0473 0.0000 0.0000 0.9763 1.0174 0.0000 0.0000 0.0000 0.6774 0.0068 0.3025 -0.3419 0.0068 0.5753 -0.5894 0.0003 0.5789 0.0128 0.0179 0.6141 0.0197 0.0018 0.6797 0.0294 0.0359 0.5920 0.0483 0.0032 0.3541 0.0000 0.0000 0.0000 0.0000 0.0000 0.5865 0.0001 0.9707 0.5537 0.9136 0.8086 0.6765 0.0000 0.0000 0.0000 0.0000 1.0105 1.0233 0.5172 0.5371 0.0000 0.0000 0.0000 0.0000 0.0000 0.3544 0.4191 0.5149 0.4432 0.3150 0.0121 0.0499 1.0327 0.0065 0.3780 0.9529 0.0134 0.0371 1.0291 0.0093 0.2121 1.3873 0.0213 0.0275 1.4014 0.0000 0.0000 0.3541 0.4003 0.0050 0.0076 0.9882 0.0099 0.0006 1.0916 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.2616 0.0542 -0.0269 0.8245 0.3069 0.2882 0.0135 0.1317 0.1896 0.3854 0.1126 0.3131 -0.0505 0.6883 0.4666 0.3808 0.0043 -0.0878 0.6471 0.4261 0.2086 0.0292 0.5479 0.0007 0.5561 0.0717 0.4620 0.8004 0.1836 0.0906 0.8708 -0.0281 0.5962 0.3459 0.0271 0.6241 -0.1677 0.0046 0.8591 -0.3409 0.0004 1.1264 0.0199 1.9137 160 0.0000 0.0000 0.5795 0.5260 0.0000 0.0000 0.9037 0.8608 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.5: Abnormal Returns Using International Version of Fama & French's Three Factor Model Non-U.S. Exchange Traded Funds and Momentum Returns: January 1,2000 - March 31, 2007 Notes: ■RMRF r+ 5 ,. ■s m B , + h, ■HIdL , + e„ (1) International Version of Fama & French's three factor model: Ru - R-Ft ~ a i Dependent variable is periodic return for the WML which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or e x c ess periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic U.S. Treasury bill rate over the listed formation and holding period. (2) Formation period includes daily returns in U.S. dollars from Thursday through Wednesday; holding period includes daily returns in U.S. dollars from Friday through Thursday. (3) Daily Datastream adjusted closing price in its local currency is converted to U.S. dollars using the midpoint of the Reuters bid and offered currency exchange rates available for the sam e time period a s the closing price. Daily holding period returns are computed from the adjusted closing prices converted to U.S. dollars for each non-U.S. ETF. (4) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sam ple period and held for the indicated period of time. (5) RMRF is the e x c e ss dollar return on the market portfolio, which is the dollar return on the value-weighted MSCI World Index from Datastream minus the periodic U.S. Treasury bill rate (from Ibbotson Associates) over the listed formation and holding periods. (6) HML (High Book to Market Minus Low Book to Market) is the average return on the two value portfolios minus the average return on the two growth portfolios, i.e. HML = 1/2 (Small Value + Big Value) -1 /2 (Small Growth + Big Growth). (7) SMB (Small Size Minus Big Size) is the average return on the three small portfolios minus the average return on the three big portfolios, SMB = 1/3 (Small Value + Small Neutral + Small Growth) -1 /3 (Big Value + Big Neutral + Big Growth). (8) p-values are computed with robust standard errors corrected for heteroskedasticilty and autocorrelation using the Newey-W est adjustmen (1987). 161 1day iday 1wk 1wk 4wk 4wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk 8wk 8wk | International Version of CAPM: WML: Constant -37.8500% -10.5560% 0.0022 0.2719 p - value Winner: 0.4600% 6.1568% Constant 0.9498 0.3280 p - value Loser: 38.3250% 16.7128% Constant 0.0000 0.0155 p - value R„ - Ri, =■a, + b r 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.7: Abnormal Returns Divided into Two Periods: 01/01/00 -12/31/03 vs. 01/01/04 - 03/31/07 Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31, 2007 165 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.8: Abnormal Returns Using International Versions of CAPM and Fama & French's Three Factor Model & Annual Dummy Variables. Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31, 2(>07 Constant p - value Mkt - RF p - value Dum2000 p - value Dum2001 p - value Dum2002 p - value Dum2003 p - value Dum2004 p - value Dum2005 p - value Adj. R-Sq. International Version of CAPM: Loser: Winner: WML: -0.0035 0.0711 0.0746 0.0000 0.5775 0.0000 0.9698 0.0237 0.9935 0.0000 0.0000 0.8881 0.2950 -0.0109 0.3069 0.6415 0.0000 0.0001 0.1058 -0.0308 0.1366 0.1197 0.0081 0.0076 0.0677 0.0081 -0.0596 0.7643 0.0015 0.1009 -0.0024 0.0961 -0.0985 0.9079 0.0001 0.0010 -0.0400 0.0503 -0.0903 0.0000 0.0206 0.0000 0.0097 -0.0168 0.0265 0.5324 0.0835 0.1358 0.3414 0.6810 0.3341 International Version of Fama & French's Three Factor Model: Loser: WML: Winner: 0.0563 0.0145 Constant 0.0418 0.2232 0.1659 0.0198 p - value 0.7024 0.0229 0.6796 Mkt - RF 0.0000 0.0000 0.9259 p - value 0.7830 1.0633 0.2803 HML 0.0005 0.0000 0.0086 p - value 0.4641 1.3309 0.8668 SMB 0.2345 0.0000 0.0000 p - value 0.0914 -0.0813 -0.1727 Dum2000 0.3292 0.0001 0.4043 p - value 0.0677 -0.0910 -0.1588 Dum2001 0.0748 0.0000 0.3531 p - value -0.0114 Dum2002 -0.0283 0.0169 0.7706 0.5396 p - value 0.6095 -0.1027 0.0311 -0.1338 Dum2003 0.0004 0.1114 0.0010 p - value -0.0725 0.0303 Dum2004 -0.1028 0.0031 0.0000 0.0006 p - value -0.0308 -0.0543 Dum2005 0.0235 0.2084 0.0001 p - value 0.4255 0.7274 0.3814 0.5413 Adj. R-Sq. 166 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.8:(Abnormal Returns Using International Versions of CAPM and Fama & French's Three Factor Model & Annual Dummy Variables. Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31, 2(>07 Notes: (1) Augmented International Version of CAPM: Rit —RFl —a i + bt ■R1dRF * + C; • Dum 2000 t + d t - Dwn 2001 , + /] • Dum 20( AA'T , +, Ji : • D u m ZUU*t { T Ktr ’ D M )1 ZUUD t "T i + g ■ • Dum O ZUUJ it (2) Augmented International Version of Fama & French's three factor model: Rlt —RFt — a t + • R1dRF * + ■SMB , + h • HML s + 'i ■Dum 20( )0 t + d r nUftf ,,,,, 1 + J i ' D u m zuuz , + ■U zuu. t t" Ji • D u m ZUUh t + i\> j mDM11 zU ^ )5 , _L + eit L um 2 0 0 1 t (3) Depend ent variable is periodic return for the WML which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the 26 week formation and holding periods, or e x c e ss periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic U.S. Treasury bill rate over the 26 week formation and holding period. (4) Formation period includes daily returns in U.S. dollars from Thursday through Wednesday; holding period includes daily returns in U.S. dollars from Friday through Thursday. (5) Daily Datastream adjusted closing price in its local currency is converted to U.S. dollars using the midpoint of the Reuters bid and offered currency exchange rates available for the sam e time period a s the closing price. Daily holding period returns are computed from the adjusted closing prices converted to U.S. dollars for each non-U.S. ETF. (6) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sample period and held for the indicated period of time. (7) RMRF is the e x c e ss dollar return on the market portfolio, which is the dollar return on the value-weighted MSCI World Index from Datastream minus the periodic U.S. Treasury bill rate (from Ibbotson A ssociates) over the listed formation and holding periods. (8) HML (High Book to Market Minus Low Book to Market) is the average return on the two value portfolios minus the average return on the two growth portfolios, i.e. HML = 1/2 (Small Value + Big Value) -1 /2 (Small Growth + Big Growth). (9) SMB (Small Size Minus Big Size) is the average return on the three small portfolios minus the average return on the three big portfolios, SMB = 1/3 (Small Value + Small Neutral + Small Growth) - 1/3 (Big Value + Big Neutral + Big Growth). (10) DumYEAR is a dummy variable that equals 1 if the portfolio is formed during YEAR, and 0 otherwise, for YEAR = 2000, 2001 , 2002, 2003, 2004, and 2005. B ase period is 2006 and first quarter of 2007. (11) p-values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using the Newey-W est adjustment (1987). 167 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.9: Abnormal Returns Using International Version of CAPM and Country Dummy Variables Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31,2007 1day 1day 1wk 1wk WML: Constant p - value M kt-RF p - value DumAust p - value DumFran p - value DumGerm p - value DumJapn p - value DumHK p - value DumUK p - value Adj. R-Sq. 0.0009 0.4184 -0.1004 0.1246 0.0008 0.3473 -0.0005 0.6357 -0.0019 0.0862 -0.0019 0.0262 -0.0002 0.8551 -0.0022 0.0186 0.0134 -0.0022 0.7811 -0.0606 0.6896 -0.0021 0.5915 0.0061 0.2718 0.0003 0.9603 -0.0034 0.4061 -0.0020 0.7324 -0.0018 0.7364 -0.0012 2wk 2wk -0.0004 0.9610 -0.1744 0.1666 0.0106 0.0189 0.0095 0.1602 -0.0035 0.6423 -0.0041 0.4329 -0.0048 0.5240 -0.0064 0.3299 0.0090 4wk 4wk 0.0057 0.6294 -0.3529 0.0095 -0.0190 0.0107 0.0135 0.1301 -0.0010 0.9158 0.0029 0.7285 -0.0084 0.4115 0.0000 0.9978 0.0349 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk 8wk 8wk -0.0377 0.0322 -0.4137 0.0078 0.0050 0.6095 0.0226 0.1191 -0.0089 0.6232 0.0274 0.0213 -0.0121 0.4490 0.0200 0.1613 0.0427 I 168 -0.0003 0.9888 -0.5797 0.0000 0.0144 0.2091 0.0482 0.0030 -0.0153 0.3676 0.0193 0.2185 -0.0376 0.0828 0.0035 0.8244 0.0940 0.0609 0.0825 -0.7332 0.0000 0.0071 0.6703 0.0961 0.0003 -0.0504 0.0478 0.0212 0.3218 -0.0921 0.0015 0.0094 0.6363 0.1636 0.0679 0.1067 -0.4040 0.0021 0.0487 0.0126 0.0383 0.1834 -0.0341 0.2089 -0.0389 0.0636 -0.0787 0.0141 0.0501 0.0195 0.1006 0.1912 0.0000 -0.2848 0.0393 0.0451 0.0255 0.1260 0.0014 -0.0943 0.0076 -0.1267 0.0000 -0.0724 0.0230 0.0118 0.6556 0.1866 0.0350 0.1189 -1.2985 0.0000 0.1160 0.0000 0.0796 0.0344 -0.0405 0.1505 0.0441 0.0855 -0.0384 0.1144 0.0048 0.8265 0.5084 0.0396 0.1302 . -1.2591 0.0000 -0.0200 0.5244 0.0389 0.4426 -0.0599 0.2248 0.1477 0.0000 -0.0804 0.0162 0.0636 0.0290 0.5489 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.9: Abnormal Returns U sing International Version o f CAPM and Country Dummy Variables Non-U.S. E xchange Traded Funds and Momentum Returns: January 1, 2000 - March 31, 2 D07 1day 1day 1wk 1wk Winner: Constant p - value M kt-RF p - value DumAust p - value DumFran p - value DumGerm p - value DumJapn p - value DumHK p - value DumUK p - value Adj. R-Sq. 0.0008 0.1755 0.6177 0.0000 -0.0002 0.7492 -0.0008 0.3117 -0.0001 0.9126 -0.0012 0.0264 -0.0007 0.1617 -0.0014 0.0068 0.1651 0.0014 0.6620 0.7864 0.0000 0.0003 0.9156 0.0020 0.6253 -0.0002 0.9534 -0.0003 0.9178 -0.0027 0.3207 -0.0026 0.2973 0.2295 2wk 2wk 0.0010 0.7820 0.7566 0.0000 0.0042 0.1695 -0.0001 0.9780 -0.0022 0.6859 -0.0017 0.6245 0.0040 0.2523 -0.0020 0.5636 0.2858 4wk 4wk 0.0094 0.1036 0.7072 0.0000 -0.0091 0.0797 -0.0034 0.5683 0.0044 0.5265 -0.0038 0.5135 0.0037 0.5080 -0.0039 0.4531 0.2335 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk 8wk 8wk 0.0179 0.0126 0.7645 0.0000 -0.0122 0.0378 0.0075 0.4044 -0.0048 0.6147 -0.0086 0.3245 -0.0013 0.8629 -0.0023 0.7093 0.3021 169 0.0430 0.0000 0.7359 0.0000 -0.0092 0.1997 0.0183 0.1100 -0.0186 0.1354 -0.0092 0.3126 -0.0116 0.2274 -0.0176 0.0151 0.2584 0.0725 0.0000 0.6104 0.0000 -0.0189 0.0619 0.0222 0.1746 -0.0150 0.3428 -0.0321 0.0038 -0.0217 0.1106 -0.0241 0.0108 0.1582 0.1011 0.0000 0.7958 0.0000 -0.0164 0.0938 -0.0025 0.8891 0.0061 0.7289 -0.0771 0.0000 -0.0185 0.2201 -0.0424 0.0008 0.2590 0.1585 0.0000 0.7068 0.0000 -0.0277 0.0572 -0.0547 0.0312 0.0185 0.4404 -0.0893 0.0000 -0.0267 0.2346 0.0220 0.2103 0.2195 0.1642 0.0000 0.5063 0.0000 0.0227 0.1762 0.0868 0.0054 -0.0475 0.0537 -0.0367 0.0408 -0.1073 0.0000 0.1184 0.0001 0.3681 0.1937 0.0000 0.5017 0.0000 -0.0480 0.0840 0.1022 0.0032 -0.0543 0.0561 -0.0112 0.6499 -0.1252 0.0000 0.0162 0.7123 0.3052 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.9: Abnormal Returns Using International Version of CAPM and Country Dummy Variables Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31, 2 307 1day1day 1wk 1wk Loser: Constant p - value M kt-RF p - value DumAust p - value DumFran p - value DumGerm p - value DumJapn p - value DumHK p - value DumUK p - value Adj. R-Sq. 0.0007 0.2102 0.7235 0.0000 0.0003 0.4549 -0.0001 0.8668 0.0003 0.6805 0.0017 0.0026 -0.0002 0.6633 0.0010 0.0750 0.2005 0.0046 0.1361 0.8513 0.0000 0.0015 0.5470 -0.0046 0.1770 0.0006 0.8878 -0.0005 0.8592 0.0030 0.2645 0.0006 0.8409 0.2891 2wk 2wk 0.0093 0.0235 0.9301 0.0000 0.0015 0.6508 0.0004 0.9098 -0.0071 0.0959 0.0003 0.9428 0.0018 0.6162 -0.0026 0.4853 0.3711 4wk 4wk 0.0151 0.0093 1.0638 0.0000 0.0047 0.2593 -0.0021 0.7649 -0.0048 0.5035 0.0047 0.3653 -0.0013 0.7900 -0.0074 0.1404 0.4647 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk 8wk 8wk 0.0609 0.0000 1.1355 0.0000 -0.0181 0.0007 0.0159 0.1656 -0.0168 0.1515 -0.0092 0.2508 -0.0095 0.2339 -0.0387 0.0000 0.4463 170 0.0664 0.0000 1.3579 0.0000 -0.0113 0.1627 0.0074 0.5838 -0.0355 0.0064 -0.0067 0.5248 -0.0025 0.7831 -0.0385 0.0001 0.5781 0.0557 0.0000 1.3903 0.0000 -0.0334 0.0078 -0.0100 0.5722 -0.0288 0.0858 -0.0126 0.3570 0.0232 0.0304 -0.0312 0.0036 0.5735 0.0696 0.0000 1.2122 0.0000 -0.0473 0.0003 -0.0060 0.7971 -0.0120 0.5757 0.0010 0.9496 0.0123 0.3040 -0.0569 0.0000 0.5154 0.0184 0.0922 1.0803 0.0000 -0.0346 0.0034 -0.0450 0.1023 0.0081 0.7608 0.0500 0.0005 0.0314 0.0008 -0.0268 0.0124 0.6852 0.0993 0.0000 1.6453 0.0000 -0.0798 0.0000 -0.0597 0.0159 0.0023 0.9258 -0.0225 0.1749 0.0106 0.3287 -0.0658 0.0000 0.8933 0.1700 0.0000 1.6974 0.0000 -0.0845 0.0003 -0.1076 0.0239 0.0454 0.3432 -0.0533 0.0122 -0.0348 0.0714 -0.1464 0.0000 0.8248 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.9: Abnormal Returns Using International Version of CAPM and Country Dummy Variables Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31, 2007 Notes: (1) Augmented International Version of CAPM: a, • DumAustr al ia, + / • DumFranc e , TV ( + 6 it UA + gi ■Dunnjermany , + j l • Durn. Japan t + k , • UumMK t + uf • Dunl. T /?., —R n —oct + bi ■R1 M F * + _i_ (2) Depend ent variable is periodic return for the WML which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or ex cess periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic U.S. Treasury bill rate over the listed formation and holding period. (3) Formation period includes daily returns in U.S. dollars from Thursday through Wednesday; holding period includes daily returns in U.S. dollars from Friday through Thursday. (4) Daily Datastream adjusted closing price in its local currency is converted to U.S. dollars using the midpoint of the Reuters bid and offered currency exchange rates available for the sam e time period a s the closing price. Daily holding period returns are computed from the adjusted closing prices converted to U.S. dollars for each non-U.S. ETF. (5) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sam ple period and held for the indicated period of time. (6) RMRF is the e x c e ss dollar return on the market portfolio, which is the dollar return on the value-weighted MSCI World Index from Datastream minus the periodic U.S. Treasury bill rate (from Ibbotson Associates) over the listed formation and holding periods. (7) DumCountry is a dummy variable that equals 1 if the portfolio formed at time t includes ETFs from Country, and 0 otherwise, for Country = Australia, France, Germany, Hong Kong, Japan, and U.K. B ase country is Canada. (8) p-values are computed with robust standard errors corrected for heteroskedasticily and autocorrelation using the Newey-W est adjustmen (1987). 171 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.10: Abnormal Returns Using International Version of Fama & French's Three Factor Model and Country Dummy Variables Non-U.S. Exchange Traded Fund s and Momentum Returns: January 1, 2000 - March 31, 2007 1day 1day 1wk 1wk WML: Constant p - value M kt-RF p - value HML p - value SMB p - value DumAust p - value DumFran p - value DumGerm p - value DumJapn p - value DumHK p - value DumUK p - value Adj. R-Sq. 0.0009 0.4462 -0.0340 0.5866 0.2011 0.1093 -0.1001 0.3417 0.0008 0.3078 -0.0004 0.6715 -0.0020 0.0632 -0.0018 0.0350 -0.0003 0.7989 -0.0022 0.0184 0.0192 -0.0030 0.7159 -0.1224 0.3830 -0.0480 0.8340 0.5348 0.0176 -0.0013 0.7421 0.0073 0.1811 0.0000 0.9927 -0.0032 0.4482 -0.0016 0.7917 -0.0030 0.5613 0.0254 2wk 2wk -0.0019 0.8310 -0.1831 0.1337 0.0657 0.7318 0.2870 0.0586 0.0108 0.0163 0.0102 0.1329 -0.0037 0.6232 -0.0043 0.4211 -0.0049 0.5179 -0.0058 0.3734 0.0184 4wk 4wk -0.0091 0.4321 -0.3121 0.0269 0.4925 0.0062 0.6540 0.0009 -0.0149 0.0310 0.0173 0.0501 -0.0086 0.3239 0.0009 0.9097 -0.0007 0.9440 0.0066 0.4703 0.0974 8wk 8wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk -0.0522 0.0065 -0.6001 0.0001 0.1507 0.4375 0.7822 0.0008 0.0106 0.2799 0.0302 0.0340 -0.0098 0.5824 0.0292 0.0112 -0.0125 0.4334 0.0223 0.1057 0.1059 -0.0340 0.1766 -0.7300 0.0000 0.2416 0.1298 0.5847 0.0274 0.0285 0.0214 0.0491 0.0033 -0.0122 0.4779 0.0230 0.1427 -0.0328 0.1273 0.0134 0.4022 0.1177 0.0008 0.9792 -0.7535 0.0000 0.5280 0.0160 0.3879 0.1525 0.0253 0.1424 0.0806 0.0012 -0.0314 0.2034 0.0241 0.2449 -0.0797 0.0029 0.0268 0.1673 0.1999 -0.0001 0.9984 -0.4621 0.0011 0.4169 0.0367 0.4297 0.1077 0.0620 0.0020 0.0303 0.2675 -0.0216 0.3894 -0.0276 0.1532 -0.0623 0.0346 0.0712 0.0005 0.1280 -0.0130 0.7027 -0.3833 0.0050 1.0732 0.0000 0.9105 0.0010 0.0869 0.0000 0.0602 0.1014 -0.0450 0.1467 -0.0718 0.0005 -0.0499 0.0683 0.0957 0.0001 0.4166 I 172 -0.0503 0.1735 -1.0174 0.0000 0.9936 0.0000 -0.1107 0.5617 0.1157 0.0000 0.0398 0.2425 -0.0453 0.0402 -0.0025 0.9323 -0.0133 0.5700 0.0473 0.0187 0.6142 -0.0241 0.6581 -1.0691 0.0000 1.0917 0.0000 -0.3034 0.2217 -0.0332 0.2217 -0.0012 0.9791 -0.0770 0.0678 0.0632 0.0673 -0.0482 0.1446 0.0849 0.0080 0.6054 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.10: Abnormal Returns Using International V ersion o f Fama & French's Three Factor Model and Country Dummy Variables Non-U.S. E xch an ge Traded Funds and Momentum Returns: January 1, 2000 - March 3 1 ,2 307 1day 1day 1wk 1wk Winner: Constant p - value M kt-RF p - value HML p - value SMB p - value DumAust p - value DumFran p - value DumGerm p - value DumJapn p - value DumHK p - value DumUK p - value Adj. R-Sq. 0.0006 0.3918 0.7372 0.0000 0.3706 0.0005 0.1439 0.0845 0.0000 0.9931 -0.0006 0.3903 -0.0003 0.6907 -0.0011 0.0304 -0.0007 0.1832 -0.0013 0.0111 0.1863 0.0000 0.9983 0.8009 0.0000 0.2236 0.2436 0.5082 0.0023 -0.0004 0.8666 0.0037 0.3562 -0.0009 0.8120 0.0004 0.8780 -0.0023 0.3735 -0.0028 0.2696 0.2709 2wk 2wk -0.0022 0.5319 0.8144 0.0000 0.3625 0.0115 0.4116 0.0002 0.0041 0.1465 0.0043 0.3823 -0.0056 0.2865 0.0000 0.9940 0.0036 0.2898 -0.0013 0.7042 0.3341 4w k 4wk -0.0011 0.8465 0.8035 0.0000 0.6050 0.0000 0.5942 0.0000 -0.0048 0.2745 0.0063 0.2654 -0.0068 0.3007 -0.0014 0.7860 0.0052 0.3098 0.0006 0.8944 0.3482 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk 8wk 8wk -0.0023 0.7426 0.7357 0.0000 0.4440 0.0000 0.5928 0.0000 -0.0014 0.7860 0.0217 0.0544 -0.0161 0.1926 -0.0004 0.9568 0.0003 0.9631 0.0025 0.6706 0.3958 173 -0.0008 0.9236 0.5599 0.0000 0.4496 0.0000 0.9255 0.0000 0.0153 0.0189 0.0357 0.0041 -0.0268 0.0697 0.0079 0.2814 -0.0070 0.3892 -0.0017 0.8329 0.3933 -0.0068 0.5139 0.4431 0.0000 0.6294 0.0001 1.0676 0.0000 0.0207 0.0072 0.0378 0.0012 -0.0202 0.0590 -0.0142 0.0862 -0.0002 0.9881 0.0073 0.2558 0.3605 0.0038 0.6853 0.5869 0.0000 0.5693 0.0000 1.0655 0.0000 0.0185 0.0289 0.0139 0.3614 0.0021 0.8907 -0.0272 0.0019 0.0095 0.4147 0.0049 0.6239 0.4253 -0.0091 0.5397 0.6722 0.0000 1.0003 0.0000 1.1011 0.0000 0.0280 0.0230 -0.0203 0.2449 0.0101 0.5304 -0.0575 0.0000 0.0509 0.0025 0.0566 0.0002 0.5571 0.0705 0.0012 0.6418 0.0000 0.8283 0.0003 0.1330 0.3619 0.0325 0.0267 0.0766 0.0010 -0.0525 0.0054 -0.0646 0.0038 -0.0342 0.0536 0.0942 0.0077 0.4992 0.1100 0.0000 0.4634 0.0000 0.5026 0.0001 0.4100 0.0003 -0.0445 0.1159 0.1000 0.0030 -0.0541 0.0642 0.0031 0.9118 -0.0932 0.0000 0.0143 0.7423 0.3791 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.10: Abnormal Returns Using International Version of Fama & French’s Three Factor Model and Country Dummy Variables Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31,2 007 1day 1day 1wk 1wk Loser: Constant p - value Mkt - RF p - value HML p - value SMB p - value DumAust p - value DumFran p - value DumGerm p - value DumJapn p - value DumHK p - value DumUK p - value Adj. R-Sq. 0.0005 0.3318 0.7773 0.0000 0.1713 0.0421 0.2390 0.0007 0.0002 0.6130 0.0000 0.9981 0.0004 0.6312 0.0017 0.0025 -0.0003 0.6086 0.0010 0.0843 0.2100 0.0038 0.2381 0.9221 0.0000 0.2654 0.0557 -0.0219 0.8641 0.0017 0.5040 -0.0046 0.1648 0.0005 0.8992 -0.0004 0.8787 0.0030 0.2615 0.0008 0.7827 0.3007 2wk 2wk 0.0069 0.0912 0.9912 0.0000 0.2779 0.0187 0.1202 0.2287 0.0016 0.6048 -0.0003 0.9306 -0.0056 0.1879 0.0005 0.9020 0.0025 0.4687 -0.0027 0.4726 0.3826 4wk 4wk 0.0140 0.0183 1.1037 0.0000 0.0938 0.4039 -0.0729 0.5695 0.0046 0.2585 -0.0021 0.7648 -0.0048 0.5023 0.0045 0.3912 -0.0002 0.9747 -0.0077 0.1274 0.4670 8wk 8wk 12wk 12wk 16wk 16wk 20wk 20wk 26wk 26wk 39wk 39wk 52wk 52wk 0.0532 0.0000 1.2842 0.0000 0.2677 0.0420 -0.2018 0.2478 -0.0182 0.0008 0.0161 0.1538 -0.0206 0.0950 -0.0102 0.2121 -0.0039 0.6041 -0.0350 0.0001 0.4724 174 0.0486 0.0004 1.3052 0.0000 0.1892 0.0491 0.2903 0.0694 -0.0089 0.2803 0.0020 0.8865 -0.0278 0.0327 -0.0043 0.6796 0.0021 0.8173 -0.0326 0.0015 0.5885 0.0197 0.1772 1.1780 0.0000 0.0819 0.4050 0.6573 0.0000 -0.0231 0.0616 -0.0244 0.1985 -0.0088 0.6116 -0.0105 0.4350 0.0329 0.0020 -0.0116 0.2815 0.6046 0.0256 0.1187 1.0484 0.0000 0.1528 0.1710 0.6607 0.0000 -0.0331 0.0057 -0.0173 0.4617 0.0063 0.7665 0.0021 0.8823 0.0224 0.0648 -0.0337 0.0140 0.5491 0.0206 0.1180 1.0450 0.0000 -0.0633 0.2600 0.0510 0.7150 -0.0314 0.0050 -0.0406 0.1486 0.0061 0.8177 0.0477 0.0007 0.0338 0.0006 -0.0285 0.0199 0.6845 0.0618 0.0000 1.4783 0.0000 -0.2080 0.0001 0.6166 0.0000 -0.0343 0.0016 0.0028 0.9071 -0.0214 0.3542 -0.0160 0.3078 0.0099 0.3445 -0.0475 0.0001 0.9125 0.1068 0.0000 1.4499 0.0000 -0.5708 0.0000 0.8966 0.0000 0.0003 0.9875 -0.0001 0.9980 0.0564 0.1028 -0.0044 0.8223 -0.0034 0.8341 -0.1041 0.0000 0.8828 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3.10: Abnormal Returns Using International Version of Fama & French's Three Factor Model and Country Dummy Variables Non-U.S. Exchange Traded Funds and Momentum Returns: January 1, 2000 - March 31, 2 007 Notes: (1) Augmented International Version of Fama & French's Three Factor Model: R M R F ,* + .v • S M B , + / r, • H M 1 . s + a , • DumAi istral ia t r\itvnT7V { t J i • L/Unll1r QrlCc t ~r g ; *UUlllL/c i m any t t J t • UUn\JQ\ IClfl t + k , ' iJUmrlK. R u ~ R n - a , + b, + m, • D u m \jK , + e lt (2) Depend ent variable is periodic return for the WML which is the winner minus loser portfolio, a zero net investment portfolio formed by buying the winners and shorting the losers over the listed formation and holding periods, or ex c ess periodic return for the winner or loser portfolio above the risk free rate a s proxied by the appropriate periodic U.S. Treasury bill rate over the listed formation and holding period. (3) Formation period includes daily returns in U.S. dollars from Thursday through Wednesday; holding period includes daily returns in U.S. dollars from Friday through Thursday. (4) Daily Datastream adjusted closing price in its local currency is converted to U.S. dollars using the midpoint of the Reuters bid and offered currency exchange rates available for the sam e time period a s the closing price. Daily holding period returns are computed from the adjusted closing prices converted to U.S. dollars for each non-U.S. ETF. (5) The winners represent the top decile of ETF returns available during the formation period; the losers represent the lowest decile of ETF returns available during the formation period. The above returns reflect portfolios with equal weightings of the appropriate winner and loser ETFs formed each week during the sam ple period and held for the indicated period of time. (6) RMRF is the e x c e ss dollar return on the market portfolio, which is the dollar return on the value-weighted MSCI World Index from Datastream minus the periodic U.S. Treasury bill rate (from Ibbotson A ssociates) over the listed formation and holding periods. (7) HML (High Book to Market Minus Low Book to Market) is the average return on the two value portfolios minus the average return on the two growth portfolios, i.e. HML = 1/2 (Small Value + Big Value) -1 /2 (Small Growth + Big Growth). (8) SMB (Small Size Minus Big Size) is the average return on the three small portfolios minus the average return on the three big portfolios, SMB = 1/3 (Small Value + Small Neutral + Small Growth) -1 /3 (Big Value + Big Neutral + Big Growth). (9) DumCountry is a dummy variable that equals 1 if the portfolio formed at time t includes ETFs from Country, and 0 otherwise, for Country = Australia, France, Germany, Hong Kong, Japan, and U.K. Base country is Canada. (10) p-values are computed with robust standard errors corrected for heteroskedasticity and autocorrelation using the Newey-W est adjustment (1987). 175 [...]... industry funds, international and country funds, and bond funds, I contribute to the literature by demonstrating momentum/contrarianism in a changing asset allocation setting which includes U.S stocks, U.S bonds, foreign stocks, as well as sector or industry funds To my knowledge, this study is the first to identify momentum/contrarianism in such an asset allocation setting Also, I contribute to the literature... contrarian strategy Most momentum/contrarian studies to date look at U.S stocks, U.S bonds, U.S domestic mutual funds, and international stocks or country indices separately By finding 8 Reproduced with permission of the copyright owner Further reproduction prohibited without permission abnormal returns in momentum/contrarian portfolios of ETFs including broad domestic funds, sector or industry funds, ... a momentum or contrarian trading strategy, since ETFs permit the purchase or sale of a diversified portfolio of securities for one commission and one bid ask spread, resulting in abnormal returns that exceed transactions costs These results contradict Lesmond, Schill, & Zhou’s (2004) characterization of momentum profits as illusory and further contribute to the existence o f momentum/contrarian profits... transactions costs for the formation and holding period of 26 weeks, which is the usual recommended momentum strategy in most previous studies The actual transactions were tabulated by domestic and bond, sector, and international ETF over the 9.29 years studied On average, the winner ETFs consisted of 10.64% domestic and bond ETFs, 46.52% sector ETFs, and 42.84% international ETFs, while the loser ETFs consisted... loser portfolios monthly, resulting in the 6 month 24 Reproduced with permission of the copyright owner Further reproduction prohibited without permission winners and losers consisting o f six equally weighted portfolios with one sixth formed from today’s winners and losers and one sixth formed from each of the previous one, two, three, four, and five month’s winners and losers My methodology used... country specific indices, and (4) bond indices including three of the Lehman Treasury bond indices, two different corporate bond indices, and the Lehman TIPS index What differentiates an ETF from a mutual fund is an ETF trades on an exchange (most on the AMEX) like a stock, enabling an ETF to be: purchased or sold at intraday market prices, purchased on margin, sold short, and traded via stop orders and limit... and shortable which are necessary conditions to realistically implement a momentum or contrarian strategy My motivation in this study is to extend the domain of momentum/eontrarianism to a relatively new investment vehicle, namely, exchange traded funds or ETFs ETFs are powerful and flexible investment vehicles that combine the diversified portfolio features of mutual funds with the trading possibilities... portfolios including sector ETFs relative to those consisting of domestic and bond ETFs, and dt measures the difference in abnormal momentum/contrarian returns in portfolios including international ETFs relative to those consisting of domestic and bond ETFs 26 Reproduced with permission of the copyright owner Further reproduction prohibited without permission Table 1.9 shows clearly large and very significant... hold their loser stocks which continue to be losers Wermers found that investor cash inflows to winner funds continue for two to four years, which is longer than the one year period attributed to momentum effects, and thus the high cash inflow funds continue to perform well due to the manager’s flow related trades chasing stocks with high past returns Chan, Hameed, & Tong (2000) find momentum returns... possibilities of individual securities Currently, ETFs function similarly to passively managed index mutual funds, as they are composed o f a portfolio of stocks or bonds that track a particular index, thus providing diversification 5 Reproduced with permission of the copyright owner Further reproduction prohibited without permission within the portion o f the market tracked by that index Four general categories