Factor‐based models such as Principal Components Analysis (PCA) of yield curve movements are used by some financial firms to assess interest rate risk. While appearing statistically reasonable, these approaches have an important and serious flaw which can mislead firms into unknowingly taking risky descisions.

This document is an example of our backtesting analysis carried out on all economic series. We take our y/e 2024 calibration set and set the starting condition to be the end of (say) 1962. We then project forward 1 year and compare the real world realisation at end of (say) 1963 and compare to the percentile fan of possible outcomes. We do this for all years and plot a Q-Q plot to check the relative distributions. A good fit would be a straight 45 degree line.