Predicting the Global Minimum Variance Portfolio*

We propose a novel dynamic approach to forecast the weights of global minimum variance portfolio (GMVP) for conditional covariance matrix asset returns. The GMVP are population coefficients linear regression benchmark return on vector differences. This representation enables us derive consistent loss function from which we can infer without imposing any distributional assumptions In order capture time variation in returns’ structure, model through recursive least squares (RLS) scheme as well by generalized autoregressive score (GAS) type dynamics. Sparse parameterizations and targeting toward equally weighted ensure scalability with respect number assets. apply these models daily stock returns, find that they perform compared existing static approaches terms both expected unconditional variance.