Global Research Insight Mapping Investment Regimes

A key component in our currency investment process is the identification of the " investor regime ". This is a phrase we use to describe the one or two themes that are driving investor behaviour and the pattern of cross-border portfolio flows at any particular point in time. In this note we introduce a more formal process which attempts to identify patterns in cross-border equity flows across 39 countries and regions using State Street's Portfolio Flow Indicator (PFI). The goal is to construct a limited number of homogenous clusters based on the degree of similarity in the patterns of portfolio flows across countries. We find that these clusters or investment regimes are quite persistent, and that most are marked by a unique profile in terms of asset market performance. In effect, knowing this month's regime gives one a better idea of what the future will look like, in addition to the possible evolution of the investment regime cycle. Applying cluster analysis to portfolio flows Over the past several years, a key part of our success in forecasting currencies can be attributed to correctly identifying the investor regime and assessing its duration. We believe that an understanding of what is driving investor behaviour is critical in building the optimal currency portfolio. A portfolio manager must be aware of whether his or her peers are seeking out yield or safety, are biased towards growth or value, or turning trend-chasers or contrarians. In the words of Mark Krtizman, no one wants to be wrong and alone. An understanding of where and why the herd is investing is key to assessing risks and the sustainability of market trends. Using portfolio flows to identify the investor regime is as much an art as a science. However, with a total of 39 countries and regions covered by the Portfolio Flow Indicator (PFI), it becomes nearly impossible for the naked eye to discern subtle shifts in the patterns of flows across countries. Our goal was to remove potential biases from the process in order to generate a more objective and quantitative reading of investor behaviour. We rely on two popular clustering methodologies – the K-means algorithm and the Self-Organizing Map (SOM). The appendix provides a detailed explanation of the methodology. While the statistical techniques are advanced, our goal is very straightforward. We seek to group together months in which the pattern of cross-border equity flows across countries …