?
Copula-CVaR Performance in Dynamic Portfolio Optimization with Different Rebalancing Periods and Weight Constraints: Evidence from ETF Market
The idea of implicitly representing dependencies between multiple assets in a single mathematical entity is particularly useful in portfolio allocation models. Over the last decade, copulas have consistently gained significance in risk modeling. Our study examines the portfolio optimization problem exploiting data of international Exchange Trade Funds from 2015 to 2018 motivated by changing investor attitudes and switching from mutual funds to Exchange Trade Funds after the recession triggered the boom of the latter. We consolidate multidimensional copula model with ARMA(1,1)-GARCH (1,1) pre-whitening method that takes into account inter-temporal dependencies. In our paper we consider 4 copula models: Gaussian, Student, Clayton, and Gumbel. We measure downside market risk by Conditional Value-at-Risk (CVaR) as alternative to traditional volatility and VaR. We test our investing strategy with respect to the 250 days of calibration window and we test results of dynamic optimization on 3 different rebalancing bases: daily, weekly, and monthly. The out-of-sample performance is evaluated and compared against two benchmarks: mean-variance and equal weights portfolio. Our empirical analysis shows that the Copula-CVaR approach is the best performing portfolio and yields higher returns and Sharpe Ratio than the benchmarks. We show how maximum weight constraint that takes place in many risk-managementsareas as the additional risk-control tool affects the portfolio performance under considered strategies. Additionally, we calculate the portfolio turnover and the break-even transaction cost of all optimization approaches in order to investigate if the profits are economically significant.