By E. Foutou
Almost 6 months ago I made my first post here at the Prognostikon blog about real-time US inflation forecasts. Although I promised more frequent updates I was finishing my bachelor's degree and I would ask for your understanding! Now, I am back with a new set of inflation forecasts for the first half of 2024! But before I showcase the latest forecasts, you will ask me how the past forecasts did? Well, Table 1 below does the forecast evaluation and I think that I did pretty well!
Table 1. actual values and forecasts of US CPI inflation, forecasts computed June 2023
With the exception of the forecasts overshooting in June and July, the rest of the months are reasonably close to the actual values. Unfortunately, the standard performance metrics do not favor my forecasts compared to the naive benchmark: the MSE of the lower forecast column gives an RMSE of 0.0058 compared to an RMSE of 0.0047 of the naive benchmark (and similarly for the MAE). However, what if I use the predictive precision coefficient introduced by Professor Thomakos in this post? In that case I find that the predictive precision of the naive benchmark is equal to 0.02591 while that of the lower forecast column is equal to 0.02657 - not a big difference numerically, but as Professor Thomakos illustrate in this post too , the predictive precision coefficient can alter model rankings. Maybe based on the predictive precision coefficient I can claim a small victory?!
Table 2. forecasts of US CPI inflation, forecasts computed January 2024 up to August 2024
The new forecasts up to August 2024 are listed above in Table 2. They seem to imply an almost monotonic drop in inflation, which someone might not find very believable - but then again the future is uncertain. The economic implications of these forecasts, should one believe them, are left to those dealing with policy making - and I hope to provide faster updates in this year!
Figure 1. forecasts of US CPI inflation, forecasts computed January 2024 up to August 2024