Forecasting Corporate Credit Spreads: Regime-Switching in LSTM

Abstract

Corporate credit spreads are modelled through a Hidden Markov model (HMM) which is based on a discretised Ornstein-Uhlenbeck model. We forecast the credit spreads within this HMM and filter out state-related information hidden in the observed spreads. We build a long short-term memory recurrent neural network (LSTM) which utilises the regime-switching information as a feature to predict the change of the credit spread. The performance of the LSTM is analysed and compared to the accuracy of an LSTM without the regime-switching information. Furthermore, purely utilising the HMM forecast, the prediction of the credit spread is compared to the prediction within the LSTM. The HMM-LSTM model is calibrated on corporate credit spreads from three European countries between 2004 and 2019. Our findings show that in most cases the LSTM performance can be improved when regime information is added.

Keywords: Long Short-Term Memory, Artificial Neural Networks, Hidden Markov Models, Filtering, Regime-switching model, Credit Spread, Forecasting

JEL Classification: C45, C53, C60