A No-Arbitrage Framework for Forecasting and Analyzing the Peruvian Yield Curve in Soles

This paper proposes a flexible framework for predicting the yield curve of Peruvian government securities denominated in soles. To this end, the level (L), slope (S), and curvature (C) factors are estimated using Bayesian techniques within an Affine Term Structure (ATSM) model for the period between 2003 and 2025. Subsequently, these factors are integrated into a changing mean and stochastic volatility Bayesian Vector Autoregressive (TVMean-SV) model to make short-term predictions, utilizing a set of relevant variables, including information from the Macroeconomic Expectations Survey. The forecast of these factors, and their subsequent translation into predictions for different maturities of the yield curve, can be done freely or conditioned on a projection scenario. This allows for the construction of medium-term projection densities, which can be graphically visualized in two dimensions as a fan chart. The results show that the data set used is relevant for predicting these factors, and this is validated by comparing the predictions for different periods with ex-post data. Subsequently, in the estimated TVMean-SV model, monetary policy shocks are identified through zero and sign restrictions, taking into account that a framework already exists that validates these restrictions for Peru using different nonlinear models (P´erez Forero, 2024a). This allows for exploring the transmission of monetary policy shocks to different maturities along the yield curve. The results show that the identified monetary policy shocks are effectively transmitted to the short end of the yield curve (two years or less). The effect on medium- and long-term interest rates (5, 10, and 30 years) is in the same direction; however, this effect is not very significant. These results suggest that medium- and long-term interest rates are influenced by additional domestic and external factors beyond monetary policy.