The Polymarket "US Recession in 2026" contract is currently trading at 18 cents.
Kalshi's "NBER Recession Declared by 2027" is at 21 cents.
The Conference Board's recession probability index says 36%. A Reuters survey of economists last month said 30%. And the 2y10y yield curve, after un-inverting in March, just re-inverted last week.
Five different measures of "how likely is a recession" — five different answers, ranging from 18% to 36%. Which one should you trust?
Why the prediction markets are lower than economists
Prediction market traders have a structural advantage that economists do not: they get paid for being right. Specifically, if you buy "no recession" at 82 cents and there's no recession, you make 18 cents. If you buy "recession" at 18 cents and there's one, you make 82 cents. The asymmetric payoff disciplines the price.
Economists, on the other hand, face the herd problem we've covered before. The IMF made 469 recession forecasts since 1992 and got 4 right. Most of those misses were "we didn't see it coming." Economists don't get fired for missing a recession. They get fired for predicting one that doesn't arrive.
So the economist consensus tends to drift high when uncertainty is rising and snap low only after the recession has clearly been avoided. Prediction markets move continuously.
What the yield curve actually tells you
The 2-year/10-year Treasury yield curve has predicted every US recession since 1955, with no false positives. When the 2-year yields more than the 10-year (an "inversion"), a recession follows on average 14 months later.
The curve inverted in July 2022. It un-inverted in September 2024. Then it inverted again briefly in March 2025. Then un-inverted. Now it's re-inverting in mid-2026.
That's the most ambiguous yield curve signal in modern history. The textbook says "recession soon." The reality is the curve has been flickering between inverted and normal for two years without delivering one.
Possible explanations: the Fed's quantitative tightening has distorted the long end. The fiscal deficit has bid up short rates. AI capex has shifted growth expectations in ways the model doesn't account for. The yield curve might still work — it's just that "soon" might be longer than 14 months.
The three data series Fed watchers actually use
Non-farm payrolls. The single most-watched data series in macro. The Sahm Rule — recession when the 3-month unemployment average rises 0.5 points above its 12-month low — triggered in summer 2024 and has been close to triggering again. Currently sitting at 0.43.
Initial jobless claims. A leading indicator. Claims spike before payrolls weaken. Currently running around 230k/week, up from 210k at the start of the year but nowhere near the 350k that historically precedes recessions.
Conference Board Leading Economic Index (LEI). A composite of 10 indicators. The LEI has been negative for 27 consecutive months — the longest streak since 1959 — without a recession occurring. Either the model is broken or the recession is very, very late.
Why the answer is structurally uncertain
Every recession in history has been driven by something specific. 1990: oil. 2001: dot-com. 2008: housing. 2020: pandemic. 2022: most economists called for one. It didn't arrive.
The "thing that breaks" in the next recession is, by definition, the thing not currently being talked about. If everyone is watching commercial real estate, the recession won't come from CRE. If everyone is watching tech valuations, the recession won't come from tech.
This is the deep problem with forecasting recessions. You're trying to forecast which surprise will arrive. Surprises don't show up in surveys.
What different probabilities imply for portfolio choices
If recession probability is 18%, you should mostly be invested in risk assets. The expected return of being long stocks is high; the downside is real but discounted.
If recession probability is 36%, you should be tilting defensive. Cash, short-duration bonds, gold, defensive sectors. The upside in risk assets has to be 2x the downside to break even.
If recession probability is genuinely uncertain — somewhere between those numbers, with no clear signal — the right answer is what the markets are doing now. Stay invested, hedge tail risk, keep dry powder.
What the prediction markets are getting right that consensus is missing
Three reasons prediction markets are likely more accurate than the economist consensus right now.
One: they update continuously, not quarterly. By the time the next economist survey comes out, the data has moved twice.
Two: they aggregate across views. Polymarket has hedge fund managers, individual traders, political operatives, and quant systems all priced into one number. Economist surveys aggregate only economists.
Three: they're priced by skin in the game. A trader who thinks recession is 35% likely can buy "recession" at 18 cents and earn the spread. If they're right, they make money. If economists were forced to put money on their 30% number, many wouldn't.
What Juno lets you do with this
Juno turns recession into a tradeable, continuously updated probability. Will the US enter recession by end of 2026? By end of 2027? Will US jobs growth be negative in any month of 2026?
Each contract aggregates the real money belief of traders who care most about being right. It's not advice — it's the closest thing to ground truth that exists for the most-asked question in macro.
The five-number lesson
One question. Five answers. 18%, 21%, 30%, 36%, "inverted again."
The honest answer is that nobody knows for sure. The useful answer is that the markets are the lowest of the five — and the markets have historically been the most accurate.
You can listen to your favorite economist. Or you can watch where the money goes. They tell different stories. Only one of them costs the people telling it when they're wrong.