We evaluate the impact of covariance matrix misspecification on a portfolio’s value at risk, by means of Monte Carlo simulation. The portfolio’s composition is found using four risk-based allocation methods, the minimum-variance portfolio, the inverse-volatility weighted portfolio, the equal-risk contribution portfolio, and the maximum-diversification portfolio. The covariance matrix is estimated historically and also by EWMA and multivariate GARCH methods. Covariance misspecification is considered in three possible way. We consider the value at risk sensitivity due to misspecification of the variance and covariance terms. We also consider misspecification of both simultaneously. Our results show that multivariate GARCH methods are more accurate to estimate the covariance components and are less sensitive to covariance misspecification.