My Mercatus Center colleague Jack Salmon put together this excellent chart. It compares the inflation forecasts of various institutions and economists to what has actually happened since the beginning of 2021. Jack has closely followed the predictions of many institutions and economists and tracked much more than what is shown on this chart. Along with these projections came analyses, almost all of which repeatedly declared that “inflation has peaked.” These predictions were also expressed with great confidence.
Yet the performance of these models against reality shows that, although such models can be useful, we must always be careful not to place too much faith in their ability to predict the future. When it comes to modeling, projecting and forecasting, more flexibility is better than less.
While Salmon and I were talking about why and how this happens regularly, he emailed me a possible explanation.
Models/forecasts are often based on historical trends, and so assumptions about inflation or interest rates always show variables returning to trend (which my chart nicely demonstrates the faulty forecasts). For example, economists who have observed low and stable inflation for 4 decades believe that inflation will quickly return to this trend, so these assumptions are often baked into their models.
The same is true for interest rates—last year the CBO forecast that the 10-year Treasury yield would rise to 1.9% in the 3rd quarter of 2022. In reality, the 10-year Treasury yield sits at 3.4%.
The same is true for believing this inflationary path and what it will look like in the 1970s and 80s. It may be that in the near term interest rate hikes to control inflation will be much lower than needed in the 80s. That’s what the Federal Reserve seems to believe. I’m personally more concerned that it won’t happen because the administration continues to spend like a drunken sailor with a death wish.
More broadly though, this set of model failures is one more data point that heightens my skepticism about such models, and I feel compelled to once again warn against the dangers of economic forecasting in another context. For example, the optimism bias in both official and private forecasts for economic data has been well documented for some time. Consider these two revealing paragraphs from one Financial Times article A few years ago:
So economists who make near-consensus predictions are, by definition, unlikely to predict extreme events, while those who correctly predict occasional black swans get everything (or everything else) wrong.
Unfortunately, when it comes to economic predictions, there’s really nowhere to turn, as the consensus view misses even cyclical, non-black swan recessions.
This is certainly true of budget estimates.
The bottom line is that predicting the future with models is difficult.
Veronique de Rugy is a senior research fellow at the Mercatus Center and a syndicated columnist for Creators.