This paper both narrowly and widely replicates the results of Anundsen et al. (Journal of Applied Econometrics, 2016, 31(7), 1291-1311). I am able to reproduce the same results as theirs. Furthermore, I find that allowing for time-varying parameters of early warning system models can considerably improve the in-sample model fit and out-of-sample forecasting performance based on an expanding window forecasting exercise.
