Macro Factor-Mimicking Portfolios

Abstract

The estimation of risk factors and their replication through mimicking portfolios are of critical importance for academics and practitioners in finance. We propose a general optimization framework to construct macro-mimicking portfolios that encompasses existing mimicking approaches, such as two-pass cross-sectional regressions (Fama and MacBeth, 1973) and maximal correlation portfolio approach (Huberman et al., 1987). We incorporate machine learning estimation improvements to mitigate the impact of estimation errors in the observed macro factors on mimicking portfolios. We provide an application to the construction of mimicking portfolios that replicate three uncorrelated global macro factors: namely growth, inflation surprises, and financial stress indicators. We show how these machine-learning mimicking portfolios can be used to improve the risk-return profile of a typical endowment asset allocation.

Keywords: factor investing, mimicking portfolios, portfolio optimization, macro risk management, machine learning

JEL Classification: G11, D81, C60