The present study contrasts two recursive cascade models to determine the dynamics of the term structure of interest rates in Colombia, these models have heterogeneously persistent factors that mean-revert to the immediately preceding factor. Under certain specifications, the number of parameters is invariant to the number of factors in the cascades. The standard model has an assumption in its volatility that makes it always constant, while the sigma variant model (SV) relaxes this assumption giving it a structure that depends on the cascade's size. Both models show fluctuations in the values of the parameters when increases the cascade's size. In the in-sample fit, the standard model shows more stability in the results, but the SV model in certain dimensions (number of factors) has a better fit. In the out-of-sample forecast, both models have a low performance compared to a random walk, but the SV model seems to be more consistent in sample and out-of-sample than the standard model. In general, it may be more valuable to choose a model with at least three factors for the adjustment and forecast of the Colombian yield curve, although it must be a meticulous choice.
