Bayesian portfolio selection using VaR and CVaR

We study the optimal portfolio allocation problem from a Bayesian perspective using value at risk (VaR) and conditional (CVaR) as measures. By applying posterior predictive distribution for future return, we derive relevant quantities needed in computations of VaR CVaR, express weights terms observed data only. This is contrast to conventional method where solution based on unobserved which are estimated. also obtain expressions global minimum (GMVaR) CVaR (GMCVaR) portfolios, specify conditions their existence. It shown that these portfolios may not exist if level used or computation too low. simulation real market data, compare new approach plug-in by studying accuracy GMVaR analysing estimated efficient frontiers. concluded outperforms one, particular predicting out-of-sample VaR.