Forecasting Swedish Inflation and Policy Rates Using Random Forests and Bullard's Modernized Taylor Rule

Abstract: This paper examines whether the Riksbank could have predicted the historic inflationary surge in Sweden in the aftermath of the Covid-19 pandemic and warned the Swedish public prior to embarking on the most aggressive policy rate-hike cycle since the global financial crisis. I study the matter in two steps. First, I generate true out-of-sample, monthly inflation forecasts for the period beginning June 2018 and ending September 2022 using traditional econometric models as well as a modern machine learning method called Random Forest as proposed by Medeiros et al. (2021) and Araujo & Gaglianone (2023). Then, I approximate the Riksbank's policy rule using Bullard's modernized Taylor rule and generate policy rate forecasts for the entire period. Results indicate that the Riksbank could have both predicted the historic surge in inflation and warned the Swedish public about the looming rate hikes sooner, especially if it had used forecasts generated by a Random Forest algorithm. Variable importance figures reported by the Random Forest algorithms show that changes in the U.S. Federal Funds Rate might be more important than changes in the Swedish policy rate "Styrräntan" in forecasting inflation in Sweden, indicating that the Riksbank might have less control over inflation than is generally assumed.