Estimation Error Using High-Frequency Data on Optimal Portfolio Choice

This article investigates the usefulness of high-frequency data in optimal portfolio choice using a comprehensive listing of major stock market indexes. We construct diversified portfolios considering monthly and high-frequency data in the modelling of volatility and compare the performance of these portfolios in terms of several out-of-sample metrics. The results underscore not only the positive performance of switching from data at lower frequencies to intraday data in the context of optimal asset allocation but also the capability of beating the equally weighted portfolio, even in the presence of transaction costs. Moreover, the results show that it is the reduction of the estimation error in estimating the variance-covariance matrix that offsets the increase in the transaction costs implied by these strategies.