Every few months, a new headline warns that a downturn is coming. A bank raises its odds, a yield curve inverts, a jobs report disappoints, and suddenly the word “recession” is everywhere again. Then, more often than not, nothing happens. The economy keeps growing, hiring keeps rising, and the professionals move quietly on to their next call.
This is not a fluke or a run of bad luck for a handful of unlucky analysts. It is close to the normal condition of macroeconomic forecasting. Study after study shows that a recession forecast, however sophisticated the model behind it, is one of the least reliable products the economics profession puts out. Understanding why helps explain not just past misses, but why the next recession forecast you read deserves a healthy dose of skepticism, whatever number is attached to it.
The Recession Forecast That Almost Never Arrives On Time
The most quoted verdict on this subject comes from Prakash Loungani, an IMF economist who has spent much of his career auditing the profession’s track record. After reviewing forecasts through the 1990s, he concluded that the record of failure to predict recessions is virtually unblemished. That was not a one-off finding. In a 2018 follow-up with co-authors Zidong An and João Tovar Jalles, Loungani examined GDP forecasts across 63 countries between 1992 and 2014 and found the same pattern everywhere: forecasters generally sensed that a recession year would look different, but they consistently underestimated how bad it would get until the year was nearly over, and private-sector and official forecasters were equally poor at it.
The 2008–2009 crisis is the starkest illustration. Of 77 economies studied, 49 fell into recession in 2009. As of April 2008, with Bear Stearns already gone and Northern Rock already nationalized, the consensus forecast was that not a single one of those economies would contract the following year. By the time forecasters caught up, in September 2009, they overcorrected and predicted 54 recessions economy-wide, several more than had actually occurred.
The pattern is not confined to global crises. In the United States specifically, the Philadelphia Fed’s Survey of Professional Forecasters, the oldest running survey of its kind, has tracked professional GDP predictions since 1968. Its “Anxious Index,” which measures the probability panelists assign to a contraction in the coming quarter, tends to spike only once a downturn is already visible in the data, not before it starts. The forecast, in other words, tends to describe the recession rather than predict it.
Anatomy Of A Recession Forecast: The Indicators Everyone Watches
Economists do not pull recession forecasts out of thin air. They lean on a handful of well-known signals, each with a real, if imperfect, historical relationship to downturns.
The most famous is the yield curve, specifically the gap between short-term and long-term Treasury yields. When short-term rates rise above long-term rates, the curve “inverts,” a pattern that has preceded most US recessions since the 1950s, missing only in 1966. That impressive record is exactly what made 2022 so confusing. The 10-year/3-month spread inverted in October 2022 and stayed inverted for roughly two years, one of the longest inversions on record, according to the Cleveland Fed. A recession, by the old rule of thumb, should have arrived within about twelve months. It never came. Analysts at the CAIA Association later noted that trading on the signal over the past century would have produced a hit rate closer to random chance than to reliable prediction.
The second widely cited signal is the Sahm Rule, developed by former Federal Reserve economist Claudia Sahm. It flags a recession when the three-month average unemployment rate rises half a percentage point above its low point of the previous year, a threshold that, per the Congressional Research Service, had never produced a false positive going back to 1970. It triggered again in July 2024. No recession followed. Sahm herself argued in real time that the reading was likely overstating labor-market weakness, since the rise in unemployment was driven largely by a growing labor supply, more immigration and higher participation, rather than by layoffs. The rule reversed within months.
Both indicators remain genuinely useful. Both also just demonstrated, within two years of each other, that “reliable historically” and “reliable right now” are not the same claim.
The 2022–2023 Recession Forecast That Never Came True
No recent episode captures the problem better than the “recession that wasn’t” of 2022 and 2023. In October 2022, Bloomberg Economics put the probability of a US recession within the next year at 100%. That same month, a Bloomberg survey found 70% of economists expecting a downturn within twelve months, and a National Association for Business Economics poll found 58% putting the odds above 50%. The Economist ran a piece explaining why a global recession was “inevitable” in 2023.
Instead, GDP grew roughly 2.5% for the year, and the labor market kept adding jobs. Fed Chair Jerome Powell later noted, with some understatement, that a very high share of forecasters had expected weak growth or recession, and instead the economy turned in a strong year.
What went wrong with the recession forecast? Economists interviewed after the fact pointed to a few specific misses. Households entered the tightening cycle with roughly $2 trillion in pandemic-era excess savings, a cushion that let consumer spending hold up far longer than historical models expected. Corporations had locked in low-rate financing during 2020 and 2021, so the Fed’s rapid rate hikes bit into borrowing costs less than usual. And immigration accelerated labor-force growth just as the Fed was trying to cool the job market, letting employment keep rising without reigniting wage-driven inflation. None of these forces were secret, but their combined size and timing were not well captured in the models forecasters were using, largely because nothing quite like the pandemic recovery had occurred before.
The Data Problem Hiding Inside Every Recession Forecast
Part of the difficulty is structural, not just a matter of missed judgment. GDP, the single number most recession calls hinge on, is not observed directly; it is estimated from incomplete data and revised repeatedly, sometimes years later. Economists call the practice of estimating current conditions from partial, staggered data “nowcasting,” and it exists precisely because the real picture of the economy is not available in real time.
Even the official scorekeeper moves slowly. The National Bureau of Economic Research, the body that formally dates US recessions, deliberately waits for revised, settled data before making a call. Research reviewing the postwar period found that between 1979 and 2021, the NBER announced the start of a recession an average of roughly six to seven months after it had actually begun. In 2008, the announcement came eleven months late. That means anyone demanding a precise, timely recession forecast is effectively asking economists to beat an institution that refuses to rush its own retrospective judgment, using data that institution considers too unreliable to act on quickly.
It Isn’t Just Recessions — Even The Fed’s Own Forecasts Miss
If the Federal Reserve’s staff and policymakers cannot forecast their own economy reliably, it says something about the difficulty of the exercise itself. Four times a year, FOMC members publish the Summary of Economic Projections, the “dot plot” that markets pore over for clues about growth, inflation and rate policy. In December 2021, the median projection for core PCE inflation in 2022 was 2.7%. The actual figure, per Federal Reserve Bank of Chicago research, came in at 4.7%, nearly double. San Francisco Fed research found the same slow-to-adjust pattern documented in private recession forecasts: through the low-inflation years before 2021 FOMC projections drifted steadily lower, and from 2021 onward they drifted steadily higher, in both cases trailing the data rather than anticipating it. If the institution with the deepest staff, the broadest data access and the most direct influence over the outcome still gets blindsided, it is a reasonable prior that everyone forecasting from the outside will struggle too.
Why Forecasters Miss Together, Not Alone
There is also a behavioral layer. Research on forecaster incentives suggests economists face limited reward for breaking early from the consensus and being right, and real reputational risk for breaking early and being wrong. A lone forecaster who calls a recession that does not arrive looks foolish; a forecaster who misses a recession that the whole profession also missed faces little individual blame. That asymmetry pushes forecasts toward the middle of the pack and toward gradual, cautious revisions rather than sharp, early warnings, which helps explain why consensus recession forecasts tend to move only after the evidence is already overwhelming.
What This Means For The Next Recession Forecast You Read
None of this means the indicators are worthless or that forecasting is a waste of time. It means a recession forecast is a probability estimate built on an imperfect read of an economy that is still happening, not a weather report for next Tuesday. As of mid-2026, that uncertainty is on full display: estimates from major banks and surveys of professional forecasters have ranged from roughly 25% to over 40% for the odds of a US recession in the coming year, depending on which institution and which month you check, with prediction markets pricing in their own, often different, numbers.
The more useful habit, for anyone trying to make sense of the next alarming headline, is to treat any single recession forecast as one noisy data point rather than a verdict. Watch whether multiple independent indicators, the labor market, consumer spending, credit conditions, corporate earnings, are moving the same direction at the same time, rather than anchoring to one dramatic signal. And remember that even the professionals with the best data, the most sophisticated models and genuine expertise have historically been the last to see a downturn coming, not the first.
Sources & Further Reading
- Zidong An, João Tovar Jalles & Prakash Loungani, “How Well Do Economists Forecast Recessions?”, IMF Working Paper 2018/039
- Federal Reserve Bank of Philadelphia, Survey of Professional Forecasters
- Federal Reserve Bank of Philadelphia, The Anxious Index
- Axios, “The Big Whiff: Why everyone was so wrong about the 2023 economy”
- Yahoo Finance, “Why the most widely anticipated recession in history never came”
- Federal Reserve Bank of Cleveland, Yield Curve and Predicted GDP Growth
- CAIA Association, “Far From Perfect: Inverted Yield Curves Don’t Reliably Predict Recessions or the Direction of the Markets”
- Congressional Research Service, The Sahm Rule Trigger: Is the United States in a Recession?
- Fortune, “This recession indicator is flashing red, but the ‘Sahm Rule’ creator says ‘this time really could be different'”
- Federal Reserve Bank of Chicago, “Forecasting Inflation During the Pandemic: Who Got It Right?”
- Federal Reserve Bank of San Francisco, “Examining the Performance of FOMC Inflation Forecasts”
- National Bureau of Economic Research, Business Cycle Dating
- Federal Reserve Bank of St. Louis, “All About the Business Cycle: Where Do Recessions Come From?”
- Newsweek, “Calling a Recession: How Long Does It Take?”
- Bloomberg, “US Recession Risk Is Receding as We Move Into 2026”
- RSM US, Economic Outlook for 2026
- U.S. Bank, Monthly Economic Outlook
