The IMF has made 469 recession forecasts for advanced economies since 1992. It correctly predicted exactly four of them.
Four out of 469. A 99.1% miss rate.
If your weather app called the forecast that wrong, you'd uninstall it. So why do markets, governments, and journalists still treat economic forecasts like prophecy?
The forecast that built a career on being wrong
IMF economist Prakash Loungani studied this in a 2001 paper that became famous in the field. He found that economic forecasts in 63 countries, over 18 years, had almost no ability to call recessions in advance.
He repeated the study in 2018 with newer data. The result was the same. Forecasters did not see the 2008 financial crisis coming. They did not see the 2020 pandemic collapse coming. They did not see the 2022 inflation shock coming.
This is not a story about bad forecasters. It's a story about a broken method.
The herd is the bug, not the feature
Most economic forecasters work at banks, governments, or institutions that compete with other banks, governments, and institutions. They publish their numbers in public.
If you're a Wall Street economist and your forecast for next year's US GDP is 0.5% while everyone else says 2.5%, you look reckless. If you say 2.3% and everyone else says 2.5%, you look careful.
Being wrong with the crowd is safe. Being right alone is dangerous. So forecasts cluster — and they cluster around whatever the consensus was last quarter, plus a small adjustment.
The model that always fights the last war
Every macroeconomic model is fitted to past data. That is its strength and its fatal flaw. The model learns what causes recessions by studying recessions that already happened.
The next recession will not look like the last one. The 2008 crisis was housing. The 2020 crisis was a virus. The 2022 inflation shock was a war, supply chains, and pandemic-era fiscal policy unwinding all at once. None of those were in the training data of the models that missed them.
The forecasters are not stupid. They are doing exactly what they're paid to do. The problem is the present keeps refusing to look like the past.
Three big misses, one decade
2008: At the start of the year, the IMF projected global growth of 4.1%. Actual: 3.0%, then minus 0.1% the following year. Lehman fell in September. By December the forecasts had been cut six times.
2020: At the start of the year, the IMF projected 3.3% global growth. Actual: minus 3.1%. A six-point miss in a single year — the largest in IMF history.
2022: At the start of the year, the Fed's median dot-plot showed inflation at 2.6% by year-end. Actual: 6.5%. The Fed funds rate ended the year at 4.5%, having started at near zero.
Three of the most consequential economic events of the last twenty years. All three blindsided every major forecaster.
The three failures behind every bad forecast
You can boil most forecast failures down to three things.
Status quo bias. Forecasters assume tomorrow looks like today, because that's the lowest-cost guess and the safest career move.
Anchoring on the last release. Each new forecast quietly references the previous forecast, even when the world has changed under it.
Career protection. Saying something is unlikely costs nothing if you're wrong. Saying something is likely and being wrong gets you fired.
Notice that none of these are about analytical skill. They are about incentives.
What actually beats forecasts
Markets do. Specifically — markets that pay people for being right and punish them for being wrong.
The Iowa Electronic Markets have run since 1988. Across five US presidential elections, they beat 964 polls 74% of the time, with a mean error of 2.65 percentage points versus 4.49 for polls.
Polymarket in 2024 had Trump favored to win the electoral college from August onward, while traditional polling models were calling it a coin flip into October.
Why? Because traders have no career incentive to stay with the herd. They have a financial incentive to bet against it when the herd is wrong.
That single change in incentive — make it costly to be wrong, make it profitable to be right — fixes most of what's broken in expert forecasting.
What Juno is doing about it
Juno is a prediction market built for the global crowd. Will US Q4 GDP beat 2.5%? Will the next FOMC meeting cut rates? Will inflation be under 3% in twelve months?
Each contract is a live, money-backed answer from the people who care most about being right. It's not advice — it's information you can't get from any single analyst note.
The four-out-of-469 lesson
Economic forecasters are not lying. They are doing their job well by the standards they're measured against. The problem is those standards reward consistency, not accuracy.
The next time you read a confident GDP projection, ask one question — what does the person making it lose if they're wrong?
If the answer is "nothing," you have your answer.