In April 2021, the Federal Reserve Board — an institution that employs more than 400 PhD economists — told the American public that rising inflation was "transitory." By June 2022, headline CPI had hit 9.1%. Those same forecasters had been wrong by an average of 2.57 percentage points through 2021–2023. The preceding decade offered no comfort: they'd overpredicted inflation by an average of 0.54 percentage points from 2012 to 2020. The direction of the error changed. The overconfidence didn't.
To be precise about what the Fed got wrong: it wasn't the cause of the price surge — it was the diagnosis. Supply chains fractured by panicked government responses don't produce monetary inflation. Instead, they result in a lack of hands and machines working together. Higher prices are the logical effect of such a scenario, though they have nothing to do with inflation.
Inflation is currency debasement. Except that the dollar was strong throughout that period. Gold was largely flat. That's not the fingerprint of monetary inflation; that's Adam Smith's pin factory running in reverse. When you destroy the specialization and global coordination that keeps goods cheap, prices rise. That's not a Fed problem. That's a political problem dressed up in economic clothing.
Which makes the Fed's failure more damning, not less. An institution with 400 credentialed economists, decades of data, and a Congressional mandate to maintain price stability looked at a supply shock and called it transitory monetary noise. The December 2021 Summary of Economic Projections showed a median core PCE forecast of 2.7% for 2022. Actual core PCE came in at 4.7% — nearly 75% above their projection. They didn't cause it. They just didn't understand it, said so with authority, and delayed rate hikes until March 2022 while inflation embedded itself in wage contracts, energy pricing, and consumer expectations. The argument that they weren't responsible for the price surge doesn't absolve them of the analytical failure that followed.
If the government, the major banks, every insurance company, brokerage firm, and private equity firm in America collectively employ thousands of credentialed economists — and they do — someone has to ask the obvious question: What exactly are we paying for? I've spent thirty years inside these markets watching smart people with good data arrive at wrong conclusions with regularity. The forecasting record doesn't read like a profession mastering its craft. It reads like one that has gotten very comfortable charging for the exercise.
The "transitory" label wasn't merely a bad guess. It was consequential. The word signaled to markets that urgent tightening wasn't needed. By the time the Fed moved, inflation had already done its work. Researchers continue to debate how much of its eventual retreat reflects Fed policy versus supply chain normalization and expiring fiscal stimulus, but the delay in tightening is not in dispute. The forecasters didn't know. They almost never do.
Heraclitus understood this 2,500 years before anyone had a Bloomberg terminal. The pre-Socratic Greek philosopher argued that the universe is governed by perpetual change — that everything is in a state of continuous flux and that you cannot step into the same river twice. That reads like an ancient metaphor. It's actually a precise description of macroeconomic reality. The models calibrated on 2019 conditions were useless in 2020. The models calibrated on post-pandemic recovery were useless in 2021. I watched both cycles fail in real time from inside credit markets. The river had moved, and the forecasters were still standing at the previous bank.
Seneca made a related point with somewhat less poetry: fixating on an unknowable future wastes the mental energy needed to handle what's in front of you today. The Stoics weren't anti-intellectual — they were anti-pretense. They'd have been entirely comfortable with an economist who admitted uncertainty. They'd have been deeply suspicious of one selling it on a quarterly update schedule.
The unknowable is arriving with more force and more frequency. Nassim Taleb's Black Swan thesis — that low-probability, high-impact events are systematically underweighted because they haven't appeared in prior data — has outperformed most of the models it critiques. The 2008 financial crisis. COVID-19. The energy shock from the U.S.-Israeli conflict with Iran. Each event lived in the tails of every mainstream forecast. Each arrived anyway. The Survey of Professional Forecasters revises quarterly; nowcast models update weekly. The forecasts keep changing because the world keeps refusing to cooperate.
Artificial intelligence has arrived as the profession's consensus solution. AI and machine learning models do improve accuracy under certain conditions — empirical studies show 5 to 50% gains depending on the variable and setup, and Federal Reserve research confirms measurable improvements for GDP and inflation projections, particularly when processing large, high-frequency datasets no human analyst could handle manually. AI detects nonlinear patterns well and adapts to regime changes faster than linear models do.
Real-time performance tells a different story. When researchers tested large language models on actual macroeconomic variables in genuine out-of-sample conditions — not curated retrospective data, but live forecasting — the results showed inaccurate, stale predictions on variables like inflation. The better-than-human benchmark mostly holds in controlled settings. In conditions resembling the actual economy, the gap narrows. Garbage in, garbage out is as true for neural networks as it was for spreadsheets. AI is a powerful complement to human judgment. It's not a solution to the underlying problem, which isn't the model — it's the assumption that the future is knowable at all.
Economic forecasting serves legitimate purposes. Businesses plan capital expenditures. Governments build budgets. Central banks set policy on something more considered than intuition, and that discipline has value. The question isn't whether to forecast. It's whether we've confused the exercise with actual knowledge — and whether a profession that has been wrong in both directions over three consecutive decades deserves the deference it routinely receives without examination. Misidentifying the type of problem you're solving isn't an honest mistake. At this price point, it's a recurring subscription to the wrong diagnosis.
Thirty years of managing capital through environments no model predicted has taught me one consistent lesson: epistemic humility — knowing precisely what you don't know — is more useful than a forecast that conceals its uncertainty in decimal points. The Stoics understood it. Heraclitus understood it. The Fed learned it in 2022, at everyone else's expense.
The economists are a dime a dozen. The wisdom to say "I don't know" is rarer, and considerably more expensive when it's absent.
