Portfolio efficiency with high-dimensional data as conditioning information

In this paper, we build efficient portfolios using different frameworks proposed in the literature and drawing upon several datasets that contain an increasing number of predictors as conditioning information. We carry extensive empirical study to investigate approaches impose sparsity dimensionality reduction, well possible latent factors driving returns risky assets. contrast previous studies made use naive OLS low-dimension information sets, find (i) accounting for large (ii) variable selection, shrinkage methods factor models, such principal component regression partial least squares, provides better out-of-sample results measured by Sharpe ratios, implied higher certainty equivalent (CER). • Reducing a set can generate meaningful portfolios. Few methods, PCR PLS especially, provide good from investor standpoint. There are economic gains when primarily these two techniques used shrink high-dimensional sets. OLS-based be damaging results, even cases not instruments.