Forecasts in a Time of High Uncertainty

Data can be manipulated to prove any theory true. Distinguishing between outright falsehoods, fallacies of weak induction and robust causal relationships is crucial to analyzing the world and forecasting its future path. But at times like these, when the global economy is undergoing substantial change, that task can be nearly impossible.

As statistician Nassim Taleb writes in The Black Swan, when a situation becomes even modestly more complex, the assumptions required to make accurate predictions must be significantly more precise. It’s obvious that the world is a much more complex place than it was just a few months ago. The bottom line is that economic and financial market forecasts are at best fallacies of weak induction in the current environment. In other words, they’re largely guesses based on limited evidence.

Our team’s simulations regarding the economic fallout from a trade war are as good as it gets. Months after the initial publication of our results, the Bank of Canada released analysis that matched our own almost exactly. But that’s cold comfort in times like these.

By definition, models are simplified versions of the real world that are most useful in analyzing modest disturbances around long-term trends. The statistical relationships that form the basis of such models break down when large shocks occur. Think about the pandemic or the global financial crisis. Some of the most sophisticated models in the world performed embarrassingly badly, with few if any predicting the surge in inflation in 2022 or the economic destruction in 2008–09.

The nature of Donald Trump’s attempt to restructure the global trading system presents an additional problem for forecasters. Not only is it a major structural change, but it’s also one that is largely unprecedented, leaving forecasters with little by way of history to go on. If Trump’s efforts usher in a period of global geopolitical upheaval, it will only work to amplify the complexity of the situation and reinforce the unprecedented nature of the current circumstances. And, of course, models have no way of capturing Trump’s penchant for rapidly changing course.

This is not meant to be an indictment of models, which have many important uses when correctly specified. It’s simply a warning to take any forecasts with the appropriate dose of caution, even those from reliable sources. It’s better to concede that the future is unknown than be led astray by overconfidence in methods not well suited for the current situation.

That said, since central banks rarely heed this caution, our models can help predict their reaction function. Both our model and the Bank of Canada’s point to a period of economic weakness, accompanied by only a modest one-time increase in inflation. It therefore seems likely that central bankers will have an easy decision to cut rates another 25 basis points again next week.