سری تکانه زمانی، به عنوان یک اثر داخلی و درون صنعتی: مفاهیم برای کارایی بازار
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|13050||2013||22 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Economics and Business, Volume 69, September–October 2013, Pages 64–85
Existing studies on time-series predictability in equity returns base their analysis on the usage of a broad market index or individual stocks showing that trend chasing trading rules have largely been futile. This paper shows that trend continuation is predominantly an intra-industry rather than a market-wide or a single-company effect. After adjusting for data snooping bias, trend chasing trading rules achieve superior predictability for a number of sectors and industries in the 1990s. A simultaneous application of trading rules to each sector or industry individually yields superior predictability on the aggregate market level in the 1990s implying that time-series momentum can also be experienced as an inter-industry effect, i.e., momentum can travel across industries reflecting the phenomenon of sector rotation. Sector and industry portfolios exhibit no predictability in their returns in the 2000s due to a persistent negative autocorrelation in their return series. A sharp and sustained rise in correlations between sectors and industries observed since the early 2000s makes it difficult for actively managed trading strategies to outperform the passive benchmarks.
An extensive body of finance literature examines the degree of predictability of equity returns based on past performance or historical price patterns. Two existing strands of research focus on different aspects of return forecasting: cross-sectional predictability and time-series predictability. The cross-sectional predictability of equity returns is commonly associated with the phenomenon of cross-section momentum where trading strategies that buy past winners and sell short past losers have been shown to earn significant profits (Jegadeesh and Titman, 1993, Jegadeesh and Titman, 2001 and Jegadeesh and Titman, 2011). Meanwhile, the existence of time-series predictability in equity returns has often been associated with evidence of the deviation of a return series from a random walk. Even though such anomaly could be revealed by applying a series of variance ratio tests or other testing techniques, a few questions would still remain. Is it possible to take advantage of serial correlation in returns by utilizing a proper trading strategy? How do you choose such a trading strategy ex-ante? If the profitability of trading strategies has been shown ex-post, can it be legitimately attributed to data snooping bias? Could transaction costs erase any profits generated by trading strategies thus implying that time-series predictability in returns can, in fact, co-exist with market efficiency? An application of technical analysis that comprises mechanical trading strategies which trigger buy and sell signals without regard to market fundamentals or personal judgment could help to answer the posed questions. Its prevalent use by the financial market professionals has been featured in the academic literature (e.g., Menkhoff, 2010) as the expansion of computational powers, introduction of new trading vehicles, and reduction in costs of trading that took place over the last decade has allowed market participants to use a great variety of trading rules and techniques at a low cost. While the focus of cross-section momentum is the relative performance across different securities, time-series momentum1 defines strong predictability of future returns based solely on the security's own past returns considered in isolation from returns on other securities. The existence of time-series momentum is commonly associated with a trending behavior in the asset price caused by positive serial autocorrelations in the asset returns. The existing literature on cross-sectional predictability suggests that time-series momentum can be more prevalent in certain segments of the U.S. equity market. Moskowitz and Grinblatt (1999) show that cross-section momentum effects in the U.S. equity market are mainly driven by industry factors and their results suggest further that serial autocorrelations in industry portfolio returns largely contribute to profits on strategies based on cross-section momentum. Using a different method of decomposing momentum profits into the components, Pan, Liano, and Huang (2004) provide direct support to the findings of Moskowitz and Grinblatt (1999). Significant positive serial autocorrelations in returns of a stock portfolio can be caused by the phenomenon of a lead-lag effect where returns on some subset of stocks in the portfolio lag or follow returns on another subset of stocks. Lo and MacKinlay (1990) document that returns on large stocks lead returns on small stocks since returns of small firms are correlated with past returns of big firms, but not vice versa. Hou (2007) elaborates that the lead-lag effect between large firms and small firms is largely an intra-industry phenomenon by showing that the industry cross-section momentum is predominantly about large firms leading same-industry small firms, which is consistent with the hypothesis that the lead-lag effect is primarily driven by the within-industry news dissemination. Since positive serial autocorrelation is, by definition, the primary factor behind the time-series momentum, its significant contribution to the cross-section momentum in industries shown in the existing literature suggests that time-series momentum can exist in industry portfolios even when there may be no evidence of it on the aggregate market level. The time-series momentum strategies are also more likely to succeed when applied to portfolios of stocks with similar characteristics such as industry affiliation than individual stocks since returns on individual stocks have generally been shown to be negatively correlated (Lo & MacKinlay, 1990). And due to the evidence of lead-lag effects on industry level, industry portfolios are more likely to exhibit time-series momentum in their returns than portfolios based on grouping methods other than industry affiliation.2 Meanwhile, there is no any empirical evidence on the existence or lack of time-series predictability at the level of sector or industry in the literature on the application of active trend following trading strategies as no previous work has investigated the performance of technical trading rules applied to sector or industry portfolios, focusing instead on an aggregate market index or individual stocks.3 Such gap in our knowledge about the nature of time-series momentum is surprising since the existing literature on cross-section momentum offers background to successfully argue that trend continuation generally transpires as an industry-specific rather than a market-wide or a single-company effect. Furthermore, time-series momentum on industry level appears to be more likely to be taken advantage of by applying active trading strategies since Moskowitz and Grinblatt (1999) document a strong short-term (one-month) effect in the cross-section industry momentum and they also show that it significantly contributes to the intermediate-term (six-month) industry momentum effect. The main purpose of this paper is to extend the earlier research on time-series momentum by applying technical trading rules to sector and industry portfolios. While some pairs of sectors and industries have highly correlated returns, some others do not. Thus, it is conceivable that returns on some sectors may be predictable even though returns on the aggregate market portfolio are not. Our contribution to the literature is that we are the first to document the performance of mechanical trading rules applied to sector and industry portfolios as well as to show that equity returns can possess superior predictability on the broad market level when trading signals are applied individually to each sector or industry portfolio rather than to the aggregate weighted portfolio comprised of all sectors or industries. The robustness of our results is confirmed by utilizing the appropriate data snooping bias adjustment techniques, using several data sets of different granularity, and performing the analysis for the exchange tradable securities of sector portfolios. Our results provide confirmation that time-series momentum is predominantly an intra-industry effect. After adjusting for data snooping bias, technical trading rules achieve superior predictability for a number of sectors and industries in the 1990s whereas no evidence of predictability in returns is found when trading rules are applied to the aggregate value-weighted portfolio. A simultaneous application of trend chasing trading rules to each sector or industry individually yields superior predictability on the aggregate market level in the 1990s, implying that time-series momentum can also be experienced as an inter-industry effect, i.e., momentum travels across industries reflecting the phenomenon of sector rotation. Such findings suggest that merely rejecting the predictability of technical analysis on the aggregate level or at the level of individual stocks, as the existing literature does, does not necessarily rule out the existence of its superior performance when the market is split into sectors or industries. Although mechanical trading rules appear to possess significant predictive power on the level of sectors and industries prior to the inception of sector-based exchange-traded funds (ETFs), there is no evidence of predictability after sector ETFs became widely accepted in the 2000s, lending support to the argument that the availability of liquid ETFs fosters market efficiency. The existence of time-series momentum is largely associated with the persistent positive autocorrelation in the return series for shorter lags: contrary to the evidence of positive autocorrelation in returns during the 1990s, returns during the 2000s are characterized by negative autocorrelation. The cross-sector and cross-industry correlations as well as the portion of variability in sector and industry returns explained by the major common factors increased sharply in the 2000s reaching their respective highest levels in the last two decades in 2011. The phenomenon of different stocks and industries moving up and down in almost a lock step for a sustained period of time will make it difficult for actively managed trading strategies to outperform their respective benchmarks. The rest of the paper is organized as follows. Section 2 describes the notion of the sector- and industry-specific nature of time-series momentum. Section 3 presents two employed sets of data. Section 4 introduces the methodology toolkit consisting of the universe of trend chasing trading rules and the econometric techniques developed to test a model's predictive power after adjusting for data snooping bias. Section 5 reports the results of the performance of mechanical trading rules applied to sector and industry portfolios. Section 6 concludes and summarizes the main findings.
نتیجه گیری انگلیسی
Previous literature finds that at conventional levels of significance there is little evidence that the application of trend chasing trading rules to the aggregate market portfolio or to individual stocks was of any value during the 1990s. Up to this point, however, the question of whether equity returns can be predictable when the market portfolio is split into related segments such as sectors or industries has not been answered. If trend continuation is an industry-specific rather than a market-wide or single-stock phenomenon, returns on sectors and industries can be predictable even when returns on the aggregate market portfolio and individual stocks are not. This paper is the first to examine the predictability of returns on sector and industry equity portfolios via the application of mechanical trading rules, providing several new implications concerning the usefulness of technical analysis. To overcome the problem of comparing multiple forecasts to a given benchmark, we employ two tests that adjust the statistical significance of the tested hypothesis of predictability in return series for a possible data snooping bias. There is evidence that technical trading rules are capable of producing superior performance for a number of sector and industry portfolios in the 1990s. The fact that trading strategies are more successful when applied to sectors and industries rather than to the aggregate market portfolio demonstrates that time-series momentum is predominantly an intra-sector and an intra-industry effect. More interestingly, in contrast to the existing literature that rejects the existence of predictive power of mechanical trading rules in the aggregate market after the mid-1980s, significant predictability of returns on the aggregate market level is found when trading rules are applied individually to each sector or industry portfolio instead of the aggregate market composed of all sectors or industries. Thus, merely rejecting the predictive power of trend chasing trading rules on the aggregate level does not necessarily rule out the existence of their superior performance when the market is split into sectors or industries. The latter finding also suggests that time-series momentum can also be an inter-sector or inter-industry effect that is a likely reflection of the phenomenon of sector rotation. The existence of time-series momentum is found to largely be a consequence of the persistent positive autocorrelation in the returns. Contrary to the evidence of positive autocorrelation in sector and industry returns during the 1990s, returns during the 2000s are characterized by negative autocorrelation. Subsequently, there is no evidence that technical trading rules consistently outperform the passive benchmark in the 2000s, after accounting for data snooping bias. Such results also imply that more recently, after the introduction of the exchange-traded funds that intend to replicate the performance of popular sector indexes, the markets have become more efficient, as the combination of inexpensive yet substantially expanded computational powers, lower transaction costs and improved liquidity must have helped to significantly reduce the short-term predictability of equity returns based on the usage of historical price patterns. The proliferation of high-frequency trading and the wide acceptance of ETFs must also have contributed to a sharp and sustained rise in correlations between sectors and industries since the early 2000s, making it difficult for actively managed trading strategies to outperform the passive buy-and-hold approach.