RT Dissertation/Thesis
T1 Optimal portfolio strategies of cointegrated assets
A1 Tang, Tao
AB Statistical arbitrage, as a quantitative method of speculation, has been increasingly prevalent along with the evolution of computational  nance. One of the most popular statistical arbitrage strategies is called pairs trading, which is widely used by hedge funds and investment banks since the mid-1980s. Pairs trading strategy exploits price spread between paired assets by taking long-short positions. If price spread is temporary according to past price information, a trading opportunity arises and pro ts can be made from price correction process. To capture these opportunities, we focus on assets sharing cointegration relations. This long-term relationship implies that paired assets are exposed to common fundamentals, and hence it guarantees price convergence to the equilibrium level. Therefore, this thesis applies cointegration technique to capture short-term market anomalies and exploits these inefficiencies using pairs trading in order to build optimal portfolio strategies.The thesis consists of three chapters. The first chapter presents an equilibrium frameworkbased on equity commonality explicitly adapted to describe the dynamics of pairs trading.Our methodology, built on the price discovery model of Figuerola-Ferretti and Gonzalo(Journal of Econometrics 2010) exploits price leadership for portfolio replication purposesand shows how pairs trading profitability is linked to the speed of equilibrium reversion.A persistence-dependent trading trigger is introduced to impose higher thresholds on pairswith slower mean reversion. Our model demonstrates that equilibrium price convergenceguarantees positive abnormal pro tability. Applied to STOXX Europe 600 traded equitiesour strategy delivers Sharpe ratios that outperform benchmark rules used in the literature.Portfolio performance is enhanced after firm fundamental factor restrictions are imposed.The second chapter proposes a VECM representation for cointegrated assets in the continuous time framework. This model implies a simple method to check for cointegrationbased on the speed of equilibrium reversion. A pair of cointegrated assets is then identified to derive a dynamically optimal pairs trading portfolio with a risk-free bond. This involves maximizing the portfolio value at terminal time without the requirement of a functional form for investor´s preferences. To this end, we connect the derived optimal portfolio with European-type spread options and in consequence the optimal investment policies can be modeled using the spread option's resulting delta hedging strategies. Our framework is tested empirically using pairs identi ed from the Dow Jones Industrial Average. This analysis requires maximum likelihood estimates on continuous VECM parameters, compared to the benchmark Johansen methodology. We  nd that the proposed optimal strategy delivers consistent profitability in terms of Sharpe ratio and cumulative returns. This supports the usefulness of introducing spread option's deltas as the optimal investment policies for pairs portfolio construction. In addition, our model-implied selection algorithm outperforms the Johansen (1991) methodology commonly applied in the previous literature. Finally, the third chapter examines the performance of pair trading portfolios when sorted by the level of cointegration of their constituents. The supercointegrated portfolio, that is formed by pairs at 1% confidence level of cointegration tests, exhibits a superior out-ofsampleperformance than simple buy-and-hold and passive investments in terms of Sharperatio. We find that the degree of performance of pairs strategy is positively related to thelevel of cointegration among pairs. These evidence are also documented in an international context, from the analysis on the European stock market. The time-varying risk of the pairs strategy is linked to aggregate market volatility. A positive risk-return relationship of the strategy is also found.
YR 2017
FD 2017-05
LK https://hdl.handle.net/10016/25103
UL https://hdl.handle.net/10016/25103
LA eng
DS e-Archivo
RD 22 jul. 2024