Risk models are an essential tool

It is comparatively easier to improve the Information Ratio of your portfolio by controlling your risk than by finding new sources of alpha, so a good risk model is an important tool for all portfolio managers. Most risk models use either a time series or cross-sectional approach. There are disadvantages to both.

We argue for a hybrid approach to risk modelling

We argue that style risk factors are well suited to a cross-sectional approach, while market, region, sector and macro risk factors are better modelled with a time series approach. The UBS Hybrid Risk Model can incorporate both cross-sectional and time series risk factors.

Expectation Maximisation Algorithm & Bayesian priors

We use the Expectation Maximisation (EM) algorithm to estimate the UBS Hybrid Risk Model. This is guaranteed to be locally monotonically convergent so is a robust solution. By including Bayesian priors we may reduce sampling errors and will speed up the convergence of the EM algorithm.