نقش پیش بینی معاملات آتی بزرگ برای بازده شاخص S & P500 : تجزیه و تحلیل داده های COT به عنوان یک سیگنال معاملات اطلاع رسانی
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|13032||2013||25 صفحه PDF||سفارش دهید||12960 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of International Financial Markets, Institutions and Money, Volume 27, December 2013, Pages 177–201
This study examines the information role of large S&P500 futures trades (commercial, noncommercial, dealers, asset managers, and hedge funds) in shaping index returns. Using consolidated data across both standard and E-mini futures contracts, we find that commercial firms’ net trading level appears positively correlated with future index returns but the relationship is not stable across time. Based on more recent data, amongst specialist traders, hedge funds appear superior in terms of access to information and/or trading ability but this advantage is only preserved at high frequency. Therefore, the current weekly Commitment of Traders (COT) report – published with a 3-day delay – prevents timely public access to this type of information. Also, trading signals generated by a popular, position-based sentiment index do not produce significant average returns. Overall, this calls into question the reliability of COT-based trading signals used by market professionals.
The Commodity Futures Trading Commission (hereafter CFTC) is the independent regulatory agency for futures and options markets in the United States. The agency publishes a weekly report, called Commitment of Traders (COT) report, disclosing the open interest positions of large traders in the futures market. In its original format, the report classifies the reportable (large, above-threshold positions) open interest into commercial and noncommercial positions. The former/latter is traditionally considered as being held by hedgers/speculators. In 2006 CFTC released a new, disaggregated COT report that breaks down futures open interest by trader type instead of generic entries such as commercial vs. noncommercial. It therefore gives public access to futures positions such as those of dealers, asset management firms, and hedge funds. The COT reports have attracted attention from both academic and professional communities. De Roon et al. (2000) employ the difference between short and long positions obtained from COT reports as a measure of hedging pressure. They specifically test the hedging pressure hypothesis of Keynes (1930) – stipulating that hedgers pay a risk premium to speculators – and discover that both ‘own-market’ and ‘cross-market’ hedging pressures are significant factors in shaping futures risk premium and therefore futures returns. A series of studies such as Bessembinder (1992), Leuthold et al. (1994), Wang (2003b), and Tornell and Yuan (2012) support the hedging pressure effect and reiterate the predictive value of large traders’ holding positions in commodity futures returns. On the other hand, studies such as Sanders et al. (2004), Bryant et al. (2006), and Gorton et al. (2012) appear to reject the hedging pressure hypothesis. Studies such as Martikainen and Puttonen (1992), Chu et al. (1999), Blasco et al. (2009), and Li (2009) show that futures markets are generally more efficient in pricing newly arrived information. Moreover, it is well-documented that institutional traders are often perceived as the ‘smart money’ ( Chakravarty, 2001, Ke and Petroni, 2004, Schmeling, 2007 and Yan and Zhang, 2009). Hence, it is possible that large S&P500 futures traders’ positions contain private information regarding future index returns ( Wang, 2003a). Whether this is the case or not, practitioners have been extracting trading signals from the weekly COT reports almost from inception. Jiler (1985) finds statistics based on large traders’ positioning data a sensible aid in technical forecasting. 1 Kirkpatrick and Dahlquist (2010) introduce a professional market report that suggests that commercial traders’ net long positions – as a percentage of the total net long positions – have a 3-week lead to cash stock positions. Wang (2003a) introduces an oscillating sentiment index based on large futures traders’ net holding positions and finds noncommercial sentiment to be a ‘price continuation indicator’ of future index returns whereas he finds the commercial sentiment a ‘contrary indicator’. Kirkpatrick and Dahlquist (2010) claim this type of sentiment oscillator to be highly indicative of the informed traders’ beliefs regarding market prospects. The predictive value of large futures traders’ positions for market returns is also investigated in other futures markets. For instance, Sanders et al. (2004) investigate the lead–lag relationship between market returns and traders’ net positions in the energy futures markets and find that market returns lead traders’ net positions but not the other way around. Tornell and Yuan (2012) study the information role of currency futures traders’ positions on spot exchange rates and find that peaks and troughs of traders’ net positions have a significant role in exchange rate forecasting. It follows that, if the public can use large traders’ positions data to consistently predict returns, the efficient market hypothesis is seriously in doubt. In financial futures markets, where the underlying assets are stock indexes, bonds and paper currencies, producers and inventory holders are difficult to conceptualize. Hence, it is possible that the hedging pressure effect may not strongly manifest itself in these markets. Moreover, the role played by large traders in financial futures markets may be different from that in commodity futures markets. Even within the financial futures sector, different contract specifications and market microstructures may lead to a different type of trading behavior. We therefore focus the scope of this research on a specific, individual futures market and elaborate on the information role of large futures traders’ positions in shaping future S&P500 index returns. While several studies exist on the information role of large traders in commodity futures markets, relatively few focus on S&P500 futures. Based on the CFTC's 2011 COT report, the average (consolidated) open interest of S&P500 futures stands at 1,059,666 which far exceeds the average open interest of most other financial futures contracts (except Eurodollar futures). The particularity of S&P500 futures, in addition to its liquidity, is that there are actually two futures markets for trading the index, namely the S&P500 standard and the E-mini markets. 2 Large traders’ positions in either market only reflect a proportion of the reportable futures open interest and therefore positions from both markets should be considered, in order to form a joint measure. De Roon et al. (2000) and Wang (2003a), are among the few studies that document the predictive value of large futures traders’ positions on S&P500 futures index. Given their timing, these studies only consider positions of standard S&P500 futures contracts. Schwarz (2012) considers both S&P500 futures markets but within a separate modeling framework. We therefore propose a new, consolidated measure of S&P500 futures traders’ positions that combines the reportable open interest of both markets and also deals with the situation of cross-market spreads. Using this consolidated measure, we investigate whether its information content has predictive value for S&P500 index returns. The results do not only shed light on the validity of Keynes’ hedging pressure hypothesis in this large market but also suggest a more cautious and rigorous approach to the use of position-based indicators in modeling future market returns. While academics and practitioners use data from COT reports as proxy for hedging pressure, speculative interest, or information advantage, few raise the question of construct validity. And yet the way that COT measurements are formulated and their stability/instability through time may have a major impact upon the validity of one's conclusions. We find that the commercial net positioning level appears to be, prima facie, a significant predictor of S&P500 index returns whereas noncommercial net positioning level appears inversely related to future index returns. Traders’ unexpected net positioning measures, which proxy for positioning responses to recent market innovations, are generally found statistically significant only in the short-term (next-day) prediction. The results suggest that the price pressure effect resulting from large traders’ unexpected net positioning is significant and, during our study period, noncommercial firms (speculators) generally make price concessions to commercial firms (hedgers) in exchange for trading immediacy. Furthermore, a brief investigation into traders’ average earnings for holding a position in the futures market provides additional confirmation that commercial firms (including dealers) generally earn a risk premium from their noncommercial counterparts. However, these results are conditioned on the time period under study. When structural break components – based on the dotcom and subprime crises – are added to our models, significant changes are identified. Our results therefore strongly indicate the presence of structural breaks in the predictive role of COT-based large traders’ positioning measures. Using positioning data from the disaggregated COT report, we find that the mix of commercial/noncommercial positions has changed substantially since the start of the subprime crisis. Particularly, we find that an increasing number of hedge fund positions have since been classified by the CFTC as commercial positions. We argue that, as the mix of traders within the commercial and noncommercial categories evolves, the information role of commercial/noncommercial positions shifts accordingly. Our study calls into question the practitioners’ use of COT-based sentiment indexes of the type investigated in Wang (2003a) for prediction or the use of hedging/speculative demand proxies based on commercial/noncommercial net positioning levels for academic pursuit. Results from a battery of back-testing procedures on position-based trading signals – including futures trader sentiment index of Wang (2003a) and extreme sentiment indicators – do not show reliable statistical significance in favor of the signals. The structure of this study is the following: Section 2 provides the data and methodology, Section 3 details the empirical results and their implications, and Section 4 concludes.
نتیجه گیری انگلیسی
Using consolidated position data, we investigate the information role of large (index futures) traders’ net positioning measures on future S&P500 index returns. Our positioning measures are constructed based on the traditional (commercial/noncommercial) COT report as well as the new disaggregated COT report (dealer, asset manager, and leveraged fund). We depart from previous studies by explicitly considering the potential impact of crisis-related structural breaks on the research outcome. We find that the commercial net positioning level appears to be a short-run significant predictor of future index returns whereas the noncommercial net positioning level appears inversely related to future index returns. However, the presence of the dotcom and subprime crises in our sample significantly impacts on the nature of this predictability, suggesting that the state of the market strongly conditions the results. Specifically, we find that during the dotcom crisis, the link between commercial net positioning level and future (1- and 2-week ahead) returns is strengthened whereas, during the subprime crisis, this link is strongly reversed and thus turns (significantly) negative. This can explain the instability of any attempted prediction from one period to another and therefore the lack of reliability of a COT-based sentiment in practical pursuit. The result is also consistent with the recent studies of Chung et al. (2012) and Wolff (2013) which find the predictive power of their investor sentiment measures to be conditioned by the state of the market. The two crises also affect the nature of the relationship between large traders’ net positioning changes – represented here by the unexpected level component – and next-day index returns. Although the direction is reversed, the two crises have, once more, opposing effect on this relationship. Whereas the norm suggests that commercial traders may receive price concessions perhaps in exchange for short-term liquidity, this norm is reversed during the dotcom episode. Once more, we therefore call into question the current use of traditional COT measures in forecasting. Our evidence does not support Keynes’ (1930) hedging pressure hypothesis or speculators’ superior market timing ability documented by Wang (2003a). On the other hand, our findings are consistent with Fishe and Smith (2012) in suggesting that commercial traders often provide market liquidity at the demand of their noncommercial counterparties whilst being relatively less informed. Furthermore, results from a contemporaneous correlation analysis between specialist traders’ net positioning changes and index returns highlight asset managers and dealers’ role as commercial hedgers who position their trades in response to market returns. As for hedge funds, the lack of significant correlation between hedge funds’ net positioning changes and contemporaneous weekly returns appears to undermine the idea that they are chasing returns and may suggest a relative information advantage at intraday frequency. This is further corroborated by the fact that hedge funds’ unexpected net positioning levels are positively correlated with reporting-day returns. These findings are in line with Fishe and Smith (2012) but the issue of which class of specialist traders is more informed, over precisely what horizon and, more importantly, what is the source of their comparative advantage and its stability through time is far from being settled. As more COT data on specialist trades becomes available, more research is needed in order to answer these questions. Last but not least, we show that trading signals based on a popular position-based sentiment index fail to deliver significant average returns over our period of study.