Are Momentum Strategies Feasible in Intraday-Trading? Empirical Results from the German Stock Market

Transcription

1 Are Momentum Strategies Feasible in Intraday-Trading? Empirical Results from the German Stock Market Tim A. Herberger a *, Matthias Horn a and Andreas Oehler b Abstract Momentum trading strategies have been proved to be profitable in different asset classes and on various national capital markets in the past. However, there are some indications for eroding momentum profits. Based on the theory of gradual information distribution on capital markets and the technological progress of the last years, we suppose that the momentum effect transformed from a monthly basis to shorter time horizons. With regard to stocks that were listed in the German blue chip index DAX 30 between November 2013 and June 2014, this study is the first to examine, whether such strategies generate market adjusted excess returns on an intraday-trading basis. We analyze 16 momentum strategies, inspired by Jegadeesh/Titman s (1993, 2001) original momentum strategy design (4x4), with ranking and holding periods of 15, 30, 45 or 60 minutes. For each stock, we analyze price observations on a five minutes frequency. From the empirical results we conclude that the momentum strategy does not provide positive excess returns. However, we find indications for price reversals in intraday stock market returns for past loser stocks. Our results are robust on portfolio size (winning as well as loser portfolio) and the duration of a lag between the end of ranking period and beginning the holding period. JEL Classification: G10, G11, G14 Key Words: Behavioral Finance, Intraday-Trading, Momentum Strategies a Department of Finance, Bamberg University, Bamberg, Germany. b Full Professor and Chair of Finance, Bamberg University, Bamberg, Germany. * Address correspondence to Tim A. Herberger, Bamberg University, Department of Finance, Kaerntenstrasse 7, D Bamberg, Germany, tim-alexander.herberger@uni-bamberg.de.

2 Are Momentum Strategies Feasible in Intraday-Trading? Empirical Results from the German Stock Market Abstract Momentum trading strategies have been proved to be profitable in different asset classes and on various national capital markets in the past. However, there are some indications for eroding momentum profits. Based on the theory of gradual information distribution on capital markets and the technological progress of the last years, we suppose that the momentum effect transformed from a monthly basis to shorter time horizons. With regard to stocks that were listed in the German blue chip index DAX 30 between November 2013 and June 2014, this study is the first to examine, whether such strategies generate market adjusted excess returns on an intraday-trading basis. We analyze 16 momentum strategies, inspired by Jegadeesh/Titman s (1993, 2001) original momentum strategy design (4x4), with ranking and holding periods of 15, 30, 45 or 60 minutes. For each stock, we analyze price observations on a five minutes frequency. From the empirical results we conclude that the momentum strategy does not provide positive excess returns. However, we find indications for price reversals in intraday stock market returns for past loser stocks. Our results are robust on portfolio size (winning as well as loser portfolio) and the duration of a lag between the end of ranking period and beginning the holding period. JEL Classification: G10, G11, G14 Key Words: Behavioral Finance, Intraday-Trading, Momentum Strategies

3 1 Introduction The momentum effect has been well documented and confirmed by numerous empirical studies that are based on various asset classes like different international equity markets 1, bonds 2, currencies 3, commodities 4 and real estate 5 or use different indicators for the buying and selling signals 6. In 1993 Jegadeesh/Titman show for the US stock market that significantly positive excess returns can be generated by applying momentum strategies (16 strategies with respective formation periods of 3, 6, 9 or 12 months and with a subsequent holding period of 3, 6, 9, 12 months). Based on the relative strength -concept (Levy (1967)), those stocks are bought that generate the highest return during the ranking period (winner portfolio). At the same time, stocks showing the lowest return during the respective ranking period are sold short (loser portfolio). The total return of the strategy consists of the return of the winner portfolio plus the return of the looser portfolio (momentum portfolio). Consequently, a momentum strategy is a gross zero cost portfolio strategy. The authors conclude that stock prices are auto-correlated, contradicting to the neoclassic efficient market hypotheses (Fama/Fisher/Jensen/Roll (1969)), and that significant excess returns can be generated through exploitation of this existing auto-correlation. Although the success of momentum trading strategies is well documented for the past, there are indications for eroding momentum profits at least for the US-stock market since the early 90s of the last century (Hwang/Rubesam (2013)). Based on a gradual distribution model for new information on the market (Hong/Stein (1999); Hong/Stein (2000); Chan/Jegadeesh/Lakonishok (1996)), we suppose that the technological progress, combined with a more efficient information distribution among market participants, led to a transformation of the momentum effect from a monthly basis to shorter time-horizons. In this sense, our paper contributes the extant literature with an adaption of the original 4x4 framework of Jegadeesh/Titman (1993) based on 1 See Jegadeesh/Titman (1993) for the US stock market and Rouwenhorst (1998) for further national stock markets in Europe; Chan/Hameed/Tong (2000) for different national stock market indices; Andreu/Swinkels/Tjong-A-Tjoe (2013) for various country and industry exchange traded funds. 2 See Jostova/Nikolova/Philipov/Stahel (2013). 3 See Menkhoff/Sarno/Schmeling/Schrimpf (2012). 4 See Miffre/Rallis (2007). 5 See Ro/Gallimore (2014) for real estate mutual funds; Beracha/Skiba (2011) for residential real estate. 6 See Bootra/Hur (2013) for 52week-high and 52week-low. 1

4 monthly stock returns, to an intraday momentum strategy scheme based on 5-minute stock returns. We analyse 16 momentum strategies with ranking and holding periods of 15, 30, 45 or 60 minutes. We provide robustness checks by applying different portfolio sizes for winner and loser portfolios and various skipping periods between the ends of the ranking periods and the beginnings of the holding periods in order to rebalance the portfolios and to avoid bid-ask-biases. Based on test runs, the empirical results show that the momentum strategy does not provide excess returns in intraday-trading. However, we find indications for reversals in intraday stock market returns for past loser stocks. Our results are robust on portfolio size and the duration of the lag between the ends of the formation periods and the beginnings of the holding periods. The remainder of the paper is structured as follows: Section 2 gives a literature review, a description of our contribution and the derivation of our hypotheses. Section 3 explains the data and methodology of our analysis. The results of our analysis are presented and discussed in section 4. Section 5 provides a summary. 2 Literature Review and Contribution The success of momentum strategies has hardly been questioned in recent years, but there are still controversial opinions on the cause of the excess returns that contradict neoclassic finance. Based on the neoclassic theory, Lesmond/Schill/Zhou (2004) argue that the momentum effect is a result of risk or liquidity premiums as well as market frictions. They state that no significant excess returns can be generated through momentum strategies after transaction costs have been taken into account. In contrast, Korajczyk/Sadka (2004) conclude that the profits realized in former studies result only partly on the negligence of transaction costs and thus, this does not constitute a complete explanation for the amount of the drawn excess returns. Conrad/Kaul (1993) explain the profitability of the momentum strategies by a cross-sectional variation of the mean returns. In contrast to that, Moskowitz/Grinblatt (1999) point out that the momentum effect can only be found in few industries. Thus, significant excess returns 2

5 can hardly be attested if the data is adjusted for industry effects. According to Wu (2002) as well as Wang (2003) the differences in expected returns over a specific period of time are the reason for the profitability of the momentum strategy. However, Grundy/Martin (2001) as well as Karolyi/Kho (2004) show that industrial influences and differences in time can only partly account for the success of momentum strategies with regard to expected returns. Chordia/Shivaumar (2002) conclude that macroeconomic variables can predict the momentum effect and the resulting excess returns. Based on a Behavioral Finance approach, the momentum effect is a consequence of investor biases. The investor-sentiment-model by Barberis/Shleifer/Vishny (1998) identifies errors made by investors when forecasting expected profits as cause for the momentum effect. The overconfidence-model by Daniel/Hirshleifer/Subrahmanyam (1998) shows that investors overestimate their own capabilities with respect to the valuation of assets. When interpreting public and private information, investors make errors, which are responsible for the momentum effect. Chan/Jegadeesh/Lakonishok (1996) as well as Hong/Stein (1999) explain the trading success of momentum strategies based on an underreaction of market participants in their trading behavior. They argue that the gradual spread of new information in the market among market participants constitutes the reason for the momentum effect, whereby the stock price only gradually reflects the fair value based on the actual news status. Figure I shows the gradual distribution of new information in the market with the consequent lagged reaction on the stock price. On the left side, a positive company message is published in T1 and the stock price reacts delayed in time until T2 when the price fully reflects all available information about the company. On the right side of the figure, the same phenomenon is shown for the occurrence of bad news. [Insert Figure I about here] In addition, Hong/Lim/Stein (2000) show that the momentum effect is stronger, when less analysts cover a company and its stocks, because this aspect promotes the gradual spread of new information in the market. 3

6 However, recent studies show that profits of technical trading strategies in general (for an overview see Park/Irwin (2004 and 2007)) and momentum strategies in particular (Hwang/Rubesam (2013)) have decreased over time. By combining Fama s Efficient Market Hypothesis with the field of behavioral finance, Lo s Adaptive Market Hypothesis (AMH) (Lo (2004)) delivers a theoretical framework that helps to explain dynamic processes in financial markets and therefore might explain the absence of momentum profits. One implication of this framework is the wax and wane of profit opportunities and their disappearance when exploited by too many market participants. Since classical momentum strategies (those using monthly stock price observations as in Jegadeesh/Titman 1993) have been well documented for decades, they might have become too popular and as a consequence the profit opportunities have vanished. Another implication of the AMH is the necessity of innovations to survive in the markets. Market participants have intuitively known that already. With the help of communication and information technologies, they have built a highly automated market microstructure whose proportion of computer-initiated trades lies over 50% (Viebig 2013). Today, these technologies enable market participants to screen and process information and news much faster than several years ago. Combining the accelerated information processing with the Hong-Stein-Model (1999) as well as Chan/Jegadeesh/Lakonishok (1996), it seems possible that recent studies might not have been able to identify momentum profits due to the low frequent data they use. By focusing on new technologies, the question arises if these quantitative rulebased systems have eliminated human-caused market anomalies like the momentum effect. Yet, findings on trading strategies of automated systems show the opposite. A considerable proportion of the strategies underlying algorithms are designed to initiate (Lopez de Prado 2012) and exploit short term momentum (Gomber et al 2011). The consequent hypothesis is that the momentum profits have drifted from daily/monthly to hourly price movements. This hypothesis is supported by Schulmeister (2007, 2008) who explains the phenomenon of vanishing technical trading strategies profits with the underlying daily data of most studies. He states that the profits of technical trading 4

7 rules have moved from daily to intraday strategies. With respect to both, Schulmeister s findings and Lo s AMH, it seems essential to analyze intraday price movements when trying to find momentum profits. Based on the considerations of the gradual distribution model for new information on the market, combined with the technological developments within the last decades, we suppose that the original monthly framework of the momentum strategies is hardly able to cover the momentum effect in the presence, because the effect should have transformed to shorter time horizons. 3 Data and Methodology Our analysis is based on 5-minutes candle quotations (containing open, close, highest and lowest quotation of each 5-minutes candle) of all 30 DAX 30-stocks being traded via XETRA from 11/01/2013 to 06/09/2014. We focus on a blue chip stock index in order to provide adequate market liquidity. Stock prices are adjusted for increase in capital and dividends. As market proxy, we use an equally weighted stock index consisting of all stocks listed in the DAX 30 during the period under observation (DeBondt/Thaler (1985)). Based on the original framework of Jegadeesh/Titman (1993) we formulate 16 momentum trading strategies, however adopted for intradaytrading. The duration of the ranking periods (J) and the holding periods (K) are either 15, 30, 45 or 60 minutes. To increase the validity of the results, overlapping analysis runs are used as proposed in Fuertes/Miffre/Tan (2009), Jegadeesh/Titman (1993) and Jegadeesh/Titman (2001). By skipping a 5-minutes candle between the end of the ranking period and the beginning of the holding period, we avoid effects documented by Jegadeesh (1990) and Lehmann (1990), e.g. price pressure and lagged effects, which could skew our results. Table I presents the number of 5-minutes-candle observations, which lies between for the setup with the longest ranking and holding periods (60 minutes per period) and for the setup with the shortest periods (15 minutes per period). [Insert Table I about here] 5

8 During the ranking period, stocks are graded according to their generated returns at the end of the ranking period according to Jegadeesh/Titman (1993), Jegadeesh/Titman (2001) and Rouwenhorst The stock with the lowest return represents the loser extreme portfolio that is sold short at the beginning of the holding period. The stock with the highest return represents the winner extreme portfolio and is bought at the beginning of the holding period. The gross return of an test run (cumulative return in a ranking or holding period of T 5-minutes candles in test run i, CR i, T ) is defined as: CR i, T T RW, t R t L, t 0 R, constitutes the 5-minutes return of the winner-portfolio and W t that of the loserportfolio. R L, t The market-adjusted return of an analysis run (abnormal cumulative return in a formation or holding period of T 5-minutes candles in test run i, CAR, i T ) is defined as: CAR i, T T R W, t R t RM t L,, t 0 R, is the 5-minutes return of the winner-portfolio, W t R L, t the 5-minutes return of the loserportfolio. R, is defined as the 5-minutes return of the equally weighted stock index. M t We provide robustness checks by testing our results with different portfolio sizes (3 and 6 stocks in the portfolios whereas the respective portfolios are equally weighted) and different durations (10 and 15 minutes) of the lag between the ends of the ranking periods and the beginnings of the holding periods. 4 Results We analyze the profitability of momentum strategies by calculating the average returns of winner- and loser-portfolios and the momentum-portfolios as a combination 6

9 of both. In our calculations, we vary the length of ranking and holding periods, numbers of stock included in the portfolios, and the number of five-minutes-periods between the calculation of the returns in the ranking period and the composition of the portfolios. We compare the momentum-portfolios gross returns with the consequent marketportfolio to get the excess returns of the momentum strategies. The market return had been positive at the observed period of time. Table II shows the average five-minutes returns for loser- and winner-portfolios that consist of one stock. The first column contains the ranking periods in minutes and the third, fourth, fifth and sixth column contain the returns for holding periods of 15, 30, 45, and 60 minutes, respectively. The average five-minutes return of a strategy that, for example, short sells the worst performing stock of the last 15 minutes for the next 30 minutes would be about percent; meaning that the stock, on average, rises % percent every 5 minutes of the 30 minutes of the holding period. Except for one, returns of all winner-portfolios are positive. But most of them are not able to outperform the market (except some of the winner-portfolios with either 15 minutes ranking or holding period). Furthermore, most of the winners-portfolios average returns are not different from zero at statistical significant levels according to the results of the t-test. Consequently, we conclude that the former winners do not, on average, outperform the market in the following periods. [Insert Table II about here] In contrast, the negative returns of the loser-portfolios are different from zero at least at the 5 percent level. Moreover, the absolute value of the loser-portfolios negative average returns is higher than the respective market returns for all parameter combinations covered in Table II. Returns are most negative for ranking periods of 30 and 45 minutes. With respect to the holding periods, the returns are getting more negative when the length of the holding periods decreases. We can conclude that the former loser stocks outperform the market in the following periods at statistical 7

10 significant levels; buying the worst performing stocks instead of selling them, therefore, would lead to significant positive returns before transaction costs. Since we follow the traditional approach of short selling former losers and buying former winners, the momentum portfolios, which consist of the winner- and loserportfolios, show negative returns at statistical significant levels. As market returns are positive in the observed period of time, the momentum portfolios underperform the market. As gross returns are already negative, we do not adjust them for transaction costs. After calculating returns for the best and worst performing stock of the holding period, we increase the number of stocks in the winner- and loser-portfolios to three and six, respectively. Results are presented in Table III and support our findings from above. Again, the winner-portfolios returns are not different from zero at statistical significant levels, except, for a ranking and holding period of 15 minutes. The average returns for this parameter-combination is more than twice as high as the market return. [Insert Table III about here] In contrast, the loser-portfolios show significant negative returns. Furthermore, the pattern that the winner-portfolios returns are most negative for ranking periods of 30 and 45 minutes and are becoming less negative with longer holding periods is supported for portfolio sizes of three and six stocks. By comparing the loser-portfolios negative returns with respect to the stocks included, we can see that the returns move closer to the market portfolio with the number of stocks included. This leads us to conclude that the negative returns are mainly caused by maximal three stocks with the former greatest losses, which outperform the market in the holding period. Due to the negative returns of the loser-portfolio, the momentum-portfolios returns are negative. Consequently, the momentum strategy underperforms the market. 8

11 We analyze whether the delay between the calculation of the stock returns and the composition of the portfolio restricts the generality of our results. Table IV presents the results for the loser-, winner-, and momentum-portfolios where two and three fiveminutes-candles are skipped between the end of the ranking and beginning of the holding period. Loser- and winner-portfolios contain only one stock. [Insert Table IV about here] The results of the former calculations are basically supported. Results for the winner-portfolios are barely significant according to the t-tests. In contrast, most of the results of the loser-portfolios are different from zero at the one percent level. The former described pattern of decreasing negative returns for longer holding periods can be observed, again. For the most significant ranking periods of 15 and 30 minutes, returns are more negative when only two instead of three candles are skipped. In all cases, momentum-portfolios fail to outperform the market. We interpret the results of Tables II to IV as follows. The winner-portfolios show returns that are neither different from zero at statistical significant levels nor higher than the returns of the market-portfolio. A possible explanation for these results can be found with the positive market sentiment in the analyzed period of time. Since possibly appearing good news are already considered in stock prices, underpriced stocks are barely existent (even after the announcement of good news), making it impossible to gain possible excess returns by buying past winners. This explanation is in line with the findings of Antoniou/Doukas/Subrahmanyam (2011). They show that momentum profits in optimistic periods are primarily driven by returns of the loser-portfolios. Due to the composition of our dataset, we are not able to analyze the momentum strategies within an opposite (negative) market sentiment. It would be interesting to see, if additional support for a transformation of the findings of Antoniou/Doukas/ Subrahmanyam (2011) to an intraday-basis could be found. Our results for the loser stocks are more clearly. In all samples, the former worst performing stocks reverse with an average return that exceeds the respective market 9

12 return in the holding period. In most cases, returns are different from zero at the one percent level according to the t-tests. We explain these findings with an overreaction of stocks in the ranking period, which reversals in the following holding period. Veronesi (1999) explains such overreactions with the occurrence of bad news in times of optimistic sentiment. By picking out the loser-portfolios with the most negative returns, we can quantify the length of such an overreaction period. We state that the overreaction lasts 30 to 45 minutes. The following reversal of the stock price is most sharply in the first 15 to 30 minutes after the portfolio composition. Although this overreaction leads to negative returns for our momentum strategies, it does not necessarily contradict the idea of momentum in general. It seems possible that, due to technical innovations, the momentum-period of the Hong/Stein-model became shorter than 35 minutes, which was the shortest period we analyzed (ranking period of 15 minutes + skip of one five-minutes-candle + holding period of 15 minutes). Therefore, our dataset is possibly not high-frequent enough to identify these short term momentum possibilities. Nevertheless, some market participants who use shorter ranking periods or are able to identify bad news very fast, could be able to ride the stocks overreactions we identified and earn significant momentum profits. We leave this question to further research. Since our momentum-portfolios earned negative returns, we did not adjust for transaction costs. It would be interesting to see, if fast acting market participants like high frequency traders could earn positive net returns by exploiting the described overreactions. As the bid-ask spread should account for the largest part of the transaction costs, it furthermore could be interesting for market makers (who do not pay spreads) to not hedge their inventory when they identify such an overreaction. 10

13 5 Conclusion Based on the theory of gradual information distribution on capital markets and the technological progress, we analyzed the success of intraday momentum strategies based on the original framework by Jegadeesh/Titman s (1993, 2001). We suppose a transformation of the momentum effect from a monthly basis to shorter time horizons in intraday-trading. We analyzed 16 momentum strategies with ranking periods as well as holding periods with 15, 30, 45 or 60 minutes. The empirical results show that the momentum effect, as well as the underreaction phenomenon among market participants, cannot be observed in intraday-trading. Following that, the validity of the gradual information model of Chan/Jegadeesh/Lakonishok (1996) and Hong/Stein (1999) for explaining the success of momentum strategies on a monthly base is questioned based on our results as well. However, we find indications for reversals in intraday stock market returns for loser stocks in the past. Our results are robust on portfolio size and the duration of a lag between the end of ranking period and beginning the holding period. 11

14 Literature Andreu, L. / Swinkels, L. / Tjong-A-Tjoe, L. (2013), Can exchange traded funds be used to exploit industry and country momentum?, in: Financial Markets and Portfolio Management, Vol. 27, No. 2, pp Antoniou, C. / Doukas, J. / Subrahmanyam A. (2011), Sentiment and Momentum, Working Paper. Barberis, N. / Shleifer, A. / Vishny, R. (1998), A Model of Investor Sentiment, in: Journal of Financial Economics, Vol. 49, No. 3, pp Beracha, E. / Skiba, H. (2011), Momentum in residential real estate, in: Journal of Real Estate Finance and Economics, Vol. 43, No. 3, pp Bhootra, A. / Hur, J. (2013), The timing of 52-week high price and momentum, in: Journal of Banking and Finance, Vol. 37, No. 10, p Chan, L. / Jegadeesh, N. / Lakonishok, J. (1996), Momentum Strategies, in: Journal of Finance, Vol. 51, No. 5, pp Chan, K. / Hameed, A. / Tong, W. (2000), Profitability of Momentum Strategies in the International Equity Markets, in: Journal of Financial and Quantitative Analysis, Vol. 35, No. 2, pp Conrad, J. / Kaul, G. (1993), Long-term Overreaction or Biases in Computed Returns, in: Journal of Finance, Vol. 48, No. 1, pp Daniel, K. / Hirshleifer, D. / Subrahmanyam, A. (1998) Investor Psychology and Security Market Under- and Overreaction, in: Journal of Finance, Vol. 53, No. 6, pp DeBondt, W. / Thaler, R. (1985), Does the Stock Market Overreact?, in: Journal of Finance, Vol. 40, No. 3, pp Fama, E. / Fisher, L. / Jansen, M. / Roll, R. (1969), The Adjustment of Stock Prices to New Information, in: International Economic Review, Vol. 10, No. 1, pp Fuertes, A.-M / Miffre, J. / Tan, W.-H. (2009), Momentum Profits, Nonnormality Risks and the Business Cycle, in: Applied Financial Economics, Vol. 19, No. 12, pp Gomber, P. / Arndt, B. / Lutat, M. / Uhle, T. (2011), High-Frequency Trading, Working Paper. Grundy, B. / Martin, S. (2001), Understanding the Nature of the Risks and the Source of the Rewards to Momentum Investing, in: Review of Financial Studies, in: Vol. 14, No. 1, pp

15 Hong, H. / Lim, T. / Stein, J. (1999), A Unified Theory of Underreaction, Momentum Trading, and Overreaction in Asset Markets, in: Journal of Finance, Vol. 54, No. 6, pp Hong, H. / Lim, T. / Stein, J. (2000), Bad News Travels Slowly: Size, Analyst Coverage, and the Profitability of Momentum Strategies, in: Journal of Finance, Vol. 55, No. 1, pp Hwang, S. / Rubesam, A. (2013), The Disappearance of Momentum, in: Social Science Research Network SSRN, No (forthcoming in European Journal of Finance). Jegadeesh, N. (1990), Evidence of Predictable Behaviour of Security Returns, in: Journal of Finance, Vol. 45, No. 3, pp Jegadeesh, N. / Titman, S. (1993), Return to Buying Winners and Selling Losers: Implication for Stock Market Efficiency, in: Journal of Finance, Vol. 48, No. 1, pp Jegadeesh, N. / Titman, S. (2001), Profitability of Momentum Strategies: An Evaluation of Alternative Explanations, in: Journal of Finance, Vol. 56, No. 2, pp Jostova, G. / Nikolova, S. / Philipov, A. / Stahel, C. (2013), Momentum in Corporate Bond Returns, in: Review of Financial Studies, Vol. 26, No. 7, pp Karolyi, G. / Kho, B. (2004) Momentum Strategies: Some Bootstrap Tests, in: Journal of Empirical Finance, Vol. 11, No. 1, pp Korajczyk, R. / Sadka, R. (2004), Are Momentum Profits Robust to Trading Costs, in: Journal of Finance, Vol. 59, No. 3, pp Lehmann, B. (1990), Fads, Martingales and Market Efficiency, in: Quarterly Journal of Economics, Vol. 105, No. 1, pp Lesmond, D. / Schill, M. / Zhou, C. (2004), The Illusory Nature of Momentum Profits, in: Journal of Financial Economics, Vol. 71, No. 2, pp Levy, R. (1967), Relative Strength as a Criterion for Investment Selection, in: Journal of Finance, Vol. 47, No. 4, pp Lo, A. (2004), The Adaptive Markets Hypothesis, in: Journal of Portfolio Management, 30 th Anniversary Issue, pp Lopez de Prado, M. (2012), Advances in High Frequency Strategies, Working Paper. Menkhoff, L. / Sarno, L. / Schmeling, M. / Schrimpf, A. (2012), Currency momentum strategies, in: Journal of Financial Economics, Vol. 106, No. 3, pp Miffre, J. / Rallis, G. (2007), Momentum strategies in commodity future markets, in: Journal of Banking and Finance, Vol. 31, No. 6, pp

16 Moskowitz, T. / Grinblatt, M. (1999), Do Industries Explain Momentum, in: Journal of Finance, Vol. 54, No. 4, pp Rouwenhorst, G. (1998), International Momentum Strategies, in: Journal of Finance, Vol. 53, No. 1, pp Park, C.-H. / Irwin, S. (2004), The Profitability of Technical Analysis: A Review, Working Paper Park, C.-H. / Irwin, S. (2007), What Do We Know About the Profitability of Technical Analysis? In: Journal of Economic Surveys, Vol. 21, No. 4, pp Ro, S. / Gallimore, P. (2014), Real Estate Mutual Funds: Herding, Momentum Trading and Performance, in: Real Estate Economics, Vol. 42, No. 1, pp Schulmeister, S. (2007), The Profitability of Technical Stock Trading has Moved from Daily to Intraday Data, WIFO Working Papers. Schulmeister, S. (2008), Profitability of Technical Stock Trading: Has it Moved from Daily to Intraday Data? WIFO Working Papers. Veronesi, P. (1999), Stock Market Overreaction to Bad News in Good Times: A Rational Expectations Equilibrium Model, in: Review of Financial Studies, Vol. 12, No. 5, pp Viebig, J. (2013), High Frequency Trading, Statistical Arbitrage und die Adaptive Markets Hypothesis, in: Corporate Finance biz, Vol. 2013, No. 8, pp Wang, K. (2003), Asset Pricing with Conditioning Information: A New Test, in: Journal of Finance, Vol. 58, No. 1, pp Wu, X. (2002), A Conditional Multifactor Analysis of Return Momentum, in: Journal of Banking and Finance, Vol. 26, No. 8, pp

17 Figure I: Gradual information distribution among market participants in case of good and bad company news This figure presents the underreation effect of market participants during a new company message is published and the subsequent momentumperiod. On the left side a positive company message is published in T1 and the stock price (P1) reacts delayed in time until T2 when the price fully reflects all information about the company (P2). On the right side a negative company message spreads lagged on the market. P P P2 P1 P1 Underreaction Effect Momentumperiod P2 Underreaction Effect Momentumperiod T1 T2 T T1 T2 T 15

18 Table I: Number of 5-minutes-candle test runs respectively observations The lines represents the different durations (15, 30, 45 and 60 minutes) of ranking period (J) whereas the rows represents the different durations (15, 30, 45 and 60 minutes) of holding period (K). Following that we analyze 16 momentum strategies. Taking into account that between the ranking period and the holding period a 5 minute lag for rebalancing the portfolios is considered we have the maximum of test runs in the 15/15 momentum strategy with observations. Ranking Period in Holding Period in minutes (K) minutes (J)

19 Table II: Average 5-minutes- returns for the previous winner and loser stocks in the period from November 2013 to June 2014 This table presents average 5-minutes-returns in percentages of price momentum portfolio strategies involving DAX 30 stocks from November 2013 to June At the end of each 5-minute-period, all stocks are ranked in ascending order based on their cumulative returns in the respective ranking period. The length of the ranking period in minutes (J) is 15, 30, 45 or 60 minutes. One 5-minutes-period after the ranking took place, the stock with the highest return in the ranking period is bought for the winner portfolio and the stock with the lowest return in the ranking period is short sold in the loser portfolio. Overlapping portfolios are constructed to increase the power of the tests. K represents the length of the different holding periods for the stocks in the portfolio in minutes. Returns of the winner and loser portfolios are computed for different holding periods of 15, 30, 45 and 60 minutes. The momentumportfolio is the zero-cost, winner plus loser portfolio. The market returns are computed with the help of an equal weighted portfolio of all DAX 30 stocks. Market returns are subtracted from the gross returns of the momentumportfolio to compute the momentum-portfolios excess returns. T-statistics for 5-minutes-period returns are shown by the symbols ***, ** and *. The symbols denote statistical significance at the one, five and ten percent level, respectively. The loser-portfolio with a ranking period of 15 minutes and a holding period of 30 minutes, for example, contains a short-position in the stock with the worst return in the ranking period. The average 5-minutes-return of the short-position is percent, which means that the stock-price increased percent per five minutes on average. The average 5-minutes-return is different from zero with a statistical significance at the one percent level. Ranking Period in minutes (J) Holding Period (K) in minutes Portfolio Loser-P. -0,00046* -0,00030*** -0,00018*** -0,00009*** Winner-P. 0,00044* 0, , ,00007** Gross Momentum-P. -0, ,00024** -0,00011* -0,00002 Market Return 0, , , ,00005 Excess Return Momentum-P. -0, ,00033*** -0,00018*** -0,00008* 30 Loser-P. -0,00107*** -0,00053*** -0,00025*** -0,00019*** Winner-P. 0, , , ,00004 Gross Momentum-P. -0,00071*** -0,00054*** -0,00024*** -0,00015*** Market Return 0, , , ,00005 Excess Return Momentum-P. -0,00087** -0,00064*** -0,00031*** -0,00020*** 45 Loser-P. -0,00104*** -0,00051*** -0,00031*** -0,00023*** Winner-P. 0, , , ,00004 Gross Momentum-P. -0,00079** -0,00044*** -0,00025*** -0,00019*** Market Return 0, , , ,00005 Excess Return Momentum-P. -0,00095** -0,00054*** -0,00032*** -0,00024*** 60 Loser-P. -0,00062** -0,00044*** -0,00028*** -0,00020*** Winner-P. 0, , , ,00004 Gross Momentum-P. -0, ,00035*** -0,00021*** -0,00016*** Market Return 0, , , ,00005 Excess Return Momentum-P. -0,00072* -0,00045*** -0,00028*** -0,00021*** 17

20 Table III: Average 5-minutes-returns of momentum portfolios with three and six winner and loser stocks in the period from November 2013 to June 2014 This table presents average 5-minutes-returns in percentages of price momentum portfolio strategies involving DAX 30 stocks from November 2013 to June At the end of each 5-minute-period, all stocks are ranked in ascending order based on their cumulative returns in the respective ranking period. The length of the ranking period in minutes (J) is 15, 30, 45 or 60 minutes. One 5-minute-period after the ranking took place, the three (six) stocks with the highest return in the ranking period are bought for the winner portfolio and the three (six) stocks with the lowest return in the ranking period are short sold in the loser portfolio. Overlapping portfolios are constructed to increase the power of the tests. K represents the length of the different holding periods for the stocks in the portfolio in minutes. Returns of the winner and loser portfolios are computed for different holding periods of 15, 30, 45 and 60 minutes. The momentum-portfolio is the zero-cost, winner plus loser portfolio. The market returns are computed with the help of an equal weighted portfolio of all DAX 30 stocks. Market returns are subtracted from the gross returns of the momentum-portfolio to compute the momentum-portfolios excess returns. T-statistics for 5-minutesperiod returns are shown by the symbols ***, ** and *. The symbols denote statistical significance at the one, five and ten percent level, respectively. The loser-portfolio with six short sold stocks, a ranking period of 15 minutes and a holding period of 30 minutes, for example, contains a short-position in the six stocks with the worst return in the ranking period. The average 5-minutes-return of the short-positions is percent, which means that the price of the six stocks increased percent per five minutes on average. The average 5-minutes-return is different from zero with a statistical significance at the one percent level. Ranking Period (J) in minutes Holding Period (K) for three winner and loser stocks in minutes Holding Period (K) for six winner and loser stocks in minutes Portfolio Loser-P. -0, ,00022*** -0,00015*** -0,00010*** -0, ,00019*** -0,00013*** -0,00010*** Winner-P. 0,00047*** 0, , ,00004* 0,00039*** 0, , ,00003* Gross Momentum-P. 0,00035** -0,00015** -0,00011*** -0,00006*** 0,00026** -0,00013*** -0,00010*** -0,00006*** Market Return 0, , , , , , , ,00005 Excess Return Momentum-P. 0, ,00025*** -0,00018*** -0,00012*** 0, ,00022*** -0,00017*** -0,00012*** 30 Loser-P. -0,00056*** -0,00036*** -0,00020*** -0,00015*** -0,00040*** -0,00029*** -0,00017*** -0,00014*** Winner-P. 0, , , , , , , ,00001 Gross Momentum-P. -0,00052*** -0,00039*** -0,00019*** -0,00014*** -0,00034*** -0,00030*** -0,00017*** -0,00012*** Market Return 0, , , , , , , ,00005 Excess Return Momentum-P. -0,00069*** -0,00049*** -0,00026*** -0,00020*** -0,00050*** -0,00040*** -0,00023*** -0,00018*** 45 Loser-P. -0,00067*** -0,00035*** -0,00023*** -0,00017*** -0,00054*** -0,00030*** -0,00019*** -0,00014*** Winner-P. 0, , , , , , , ,00000 Gross Momentum-P. -0,00061*** -0,00035*** -0,00021*** -0,00016*** -0,00060*** -0,00031*** -0,00019*** -0,00014*** Market Return 0, , , , , , , ,00005 Excess Return Momentum-P. -0,00077*** -0,00045*** -0,00028*** -0,00021*** -0,00077*** -0,00040*** -0,00026*** -0,00020*** 60 Loser-P. -0,00058*** -0,00037*** -0,00023*** -0,00016*** -0,00052*** -0,00030*** -0,00020*** -0,00015*** Winner-P. 0, , , , , , , ,00001 Gross Momentum-P. -0, ,00038*** -0,00022*** -0,00016*** -0,00053*** -0,00030*** -0,00020*** -0,00015*** Market Return 0, , , , , , , ,00010 Excess Return Momentum-P. -0,00072*** -0,00048*** -0,00029*** -0,00021*** -0,00069*** -0,00039*** -0,00027*** -0,00020*** 18

21 Table IV: Average 5-minutes-period returns for the delayed in time winner and loser stocks in the period from November 2013 to June 2014 This table presents average 5-minutes-returns in percentages of price momentum portfolio strategies involving DAX 30 stocks from November 2013 to June At the end of each 5-minute-period, all stocks are ranked in ascending order based on their cumulative returns in the respective ranking period. The length of the ranking period in minutes (J) is 15, 30, 45 or 60 minutes. Two (three) 5-minute-periods after the ranking took place, the stock with the highest return in the ranking period is bought for the winner portfolio and the stock with the lowest return in the ranking period is short sold in the loser portfolio. Overlapping portfolios are constructed to increase the power of the tests. K represents the length of the different holding periods for the stocks in the portfolio in minutes. Returns of the winner and loser portfolios are computed for different holding periods of 15, 30, 45 and 60 minutes. The momentum-portfolio is the zero-cost, winner plus loser portfolio. The market returns are computed with the help of an equal weighted portfolio of all DAX 30 stocks. Market returns are subtracted from the gross returns of the momentum-portfolio to compute the momentum-portfolios excess returns. T-statistics for 5-minutes-period returns are shown by the symbols ***, ** and *. The symbols denote statistical significance at the one, five and ten percent level, respectively. The loser-portfolio, for example, with a ranking period of 15 minutes and a holding period of 30 minutes, which starts three 5-minutes-periods after the ranking period, contains a short-position in the stock with the worst return in the ranking period. The average 5-minutes-return of the short-position is percent, which means that the stock-price increased percent per five minutes on average. The average 5-minutes-return is different from zero with a statistical significance at the one percent level. Ranking Period (J) in minutes Holding Period (K) two periods after ranking Holding Period (K) three periods after ranking Portfolio Loser-P. -0, ,00032*** -0,00015*** -0,00010*** -0,00052** -0,00030*** -0,00011** -0,00011*** Winner-P. 0,00047* 0, ,00010** 0,00008** -0, , ,00009* 0,00006* Gross Momentum-P. 0, ,00020* -0, , ,00059* -0,00024** -0, ,00005 Market Return 0, , , , ,0001 0, , ,00005 Excess Return Momentum-P. -0, ,00030** -0,00012* -0,00007* -0,00075** -0,00034*** -0, ,00010** 30 Loser-P. -0,00103*** -0,00045*** -0,00022*** -0,00018*** -0,00094*** -0,00029*** -0,00019*** -0,00017*** Winner-P. 0, , , , , , , ,00003 Gross Momentum-P. -0,00090*** -0,00048*** -0,00020*** -0,00014*** -0,00114*** -0,00040*** -0,00017*** -0,00014*** Market Return 0, , , , , , , ,00005 Excess Return Momentum-P. -0,00107*** -0,00057*** -0,00027*** -0,00019*** -0,00130*** -0,00050*** -0,00024*** -0,00019*** 45 Loser-P. -0,00085*** -0,00046*** -0,00028*** -0,00022*** -0,00084*** -0,00042*** -0,00027*** -0,00022*** Winner-P. 0, , , ,00006* -0, , , ,00004 Gross Momentum-P. -0,00070** -0,00039*** -0,00021*** -0,00017*** -0,00099*** -0,00043*** -0,00024*** -0,00018*** Market Return 0, , , , , , , ,00005 Excess Return Momentum-P. -0,00087** -0,00049*** -0,00028*** -0,00022*** -0,00116*** -0,00053*** -0,00030*** -0,00024*** 60 Loser-P. -0,00072** -0,00041*** -0,00027*** -0,00021*** -0,00096*** -0,00041*** -0,00028*** -0,00021*** Winner-P. 0, , , , , , , ,00005 Gross Momentum-P. -0,00062* -0,00030** -0,00021*** -0,00016*** -0,00094*** -0,00032*** -0,00023*** -0,00016*** Market Return 0, , , , , , , ,00005 Excess Return Momentum-P. -0,00078** -0,00039*** -0,00027*** -0,00021*** -0,00110*** -0,00042*** -0,00030*** -0,00021*** 19