Forecasting Total Risk

Greg Anderson, Lisa Goldberg, Alec N. Kercheval, Guy Miller, Kathy Sorge

A global model that forecasts risk for portfolios with holdings across several markets will typically disagree with the prediction of a model specifically adapted to a single market. Given a global model and a collection of single market models, we describe an optimal, consistent way to embed the single market forecasts into the global model. The method involves framing the problem as an optimization over the orthogonal group O(n).