working paper

We empirically proof that investors shy away from assets with recent structural breaks which we use as a proxy for parameter uncertainty.

We use unsupervised machine learning to cluster stocks into equally weighted portfolios which in turn are plugged into minimum variance optimization.

A cross-country factor momentum strategy yields highly sigtnificant risk-adjusted returns and subsumes national

In this paper, we develop a novel, intuitive and objective measure of time-varying parameter uncertainty (PU) based on a simple statistical test.

Ranking assets based on their expected returns and subsequently optimizing portfolios on the reduced amount of information produces pobetter forecasts and significantly improved out-of-sample performance related to the plug-in portfolio.

We quantify performance measures distortions in a cryptocurrency sample truncated by survivorship and delisting bias.

We show that parameter uncertainty based on the turbulence withing each cross-section of factor portfolios produces a significant out-of-sample forecast for six out of seven tested Fama-French risk factors.