Archetypes of collective yield curve movements

Abstract

We examine the historical behaviour of interest rate movements in seven major currencies AUD, CAD, CHF, EUR, GBP, JPY and USD. We apply principle components analysis and hierarchical cluster analysis to illustrate, understand and model the past collective movements of yield curves. We show that simple correlations are not able to capture the complex behaviour observed in the data set. In order to model risk factors that are intimately connected, we propose so-called archetypes of collective movements as building blocks. Thus, we start from collective movements that are coherent from a historical perspective. A set of risk factor forecasts is then generated by adapting an archetype rather than building single risk factor forecasts from scratch. This approach opens the door to integrated, coherent forecasts created from complex building blocks. The methods may be applied within scenario simulations, forecasting, filtering techniques and technical analysis.