Asset returns can tell us the economic regime - Conditional returns matter

Asset returns are conditional on the market regime. We define some exogenous regime and then look at what were the returns for different assets during that regime. However, there can be another approach to looking at return data and regime identification. We can look at the pattern of returns and then identify a regime.

The overall return distribution for an asset or set of assets may be a mixed distribution based on returns and risk during different regimes. The unconditional returns are a combination of conditional clusters. There are models that can find the set of return regimes that make up the unconditional returns, for example, the Gaussian Mixture Model (GMM). The researchers at Two Sigma used this type of model to identify four main regimes across asset returns based on the clustering of their information on the 17 different factor returns. See "A Machine Learning Approach to Regime Modeling"

The four regimes identified from this GMM model can be called: crisis, steady state, inflation, and walking on ice (WOI), the period around crises. These names are not based on a review of the economic environment but on the clustering of returns across all factors. After the clusters are identified, they can be named based on a look at the environment. 

This work is important because it tell us that regimes matter and market returns cluster around a few key groupings. A judgment on stocks and its regime cannot be isolated from returns across asset classes and factor returns.