Every recession is described, after it arrives, as a surprise. The truth is closer to the opposite. Downturns rarely come from nowhere. They build for months through the same handful of channels, and those channels leave a trail in public data long before output turns. The problem is that the warning is quiet. It sits scattered across separate corners of the economy, easy to miss until the headlines make it loud, which is the moment it stops being worth anything.
recessionodds.com exists to make that early signal legible. It publishes a single monthly reading of how likely a U.S. recession is to begin within the next three to twenty four months, built entirely from free public data. The point is not to call a date. It is to show how close the business cycle is to turning, and to show it while there is still time to do something about it.
What are recession odds?
Recession odds are the probability that a new recession begins within a given window of time. Plenty of institutions publish a version of this, usually a single figure for the year ahead. What sets a term structure apart is that it reads the probability across a range of horizons at once, from three months out to two years, so the question shifts from whether a recession is coming to how near it is and how fast the risk is climbing. A number that is low for three months but high for twenty four is telling a very different story than one that is high across the board, and the difference is the whole point.
The five signals that turn before a recession
The model reads the economy through five channels. Each has a history of moving ahead of past downturns, and each captures something the others miss, which is why the reading leans on all five rather than any single favorite.
The yield curve is the most watched of them. It measures the gap between long and short term interest rates on U.S. Treasuries, the ten year against the three month. In a normal economy long rates sit above short ones. When that flips and short rates rise above long, the curve has inverted, a condition that has preceded every modern U.S. recession. It works because an inverted curve is the bond market pricing rate cuts to come, which is the market betting on a slowdown.
The curve's trend adds the piece the level alone leaves out, which is speed. A curve that inverts quickly carries a sharper warning than one that has drifted flat and stayed there for a year. By reading the six month change in the spread, the model catches momentum that a snapshot would flatten out.
Credit spreads track the extra yield lenders demand to hold riskier corporate bonds over safe government debt. When lenders grow nervous, that premium widens, and it widens before the strain shows up in output, because the cost of borrowing tightens for real companies first. A rising six month trend here signals capital getting more expensive across the economy.
Business equipment orders measure the machines and tools firms buy when they expect to grow. That spending is discretionary and forward looking, which makes it one of the first things companies cut when confidence slips. A sustained decline points to businesses quietly bracing before they say so out loud.
The stock market, read across a six month window of the S&P 500, reflects where investors think earnings are heading. On its own it is noisy and prone to false moves, which is exactly why it sits beside four steadier channels instead of carrying the reading alone.
How a recession probability model is built
A signal is only worth trusting if it has been tested honestly. This model was fit across more than fifty years of monthly data spanning eight U.S. recessions, and it was tested out of sample, meaning that at every point in history it saw only the data that existed at the time and never the answer ahead of it. In that testing, the long horizon reading crossed eighty percent in the year before every recession in the window. It is not flawless, and it does not pretend to be. It has produced false alarms, the 2022 to 2024 stretch being the loudest, and it cannot see shocks that begin outside the economy, like the 2020 pandemic. What it reads is the business cycle as it builds, which is where recessions give themselves away in advance.
Why the headlines arrive too late
Most of the numbers that dominate economic coverage, employment and GDP among them, are coincident or lagging. They confirm a recession only once it is already underway. That is useful for the record and useless for preparation. A forward probability works from the leading channels instead, the ones that move first, and reads them as a group to estimate how close the cycle is to turning. It trades the certainty of hindsight for the head start of an early, honest guess.
What the reading actually tells you
The site sorts each reading into three bands. Calm, under thirty percent, is the low cost window to prepare while the sky is still clear. Elevated, thirty to sixty five percent, is the range to shorten commitments and hold expansion in reversible steps. High, over sixty five percent, is a near warning to move decisively once the short term confirms it. The bands turn a bare probability into a sense of what stage the economy is in and what response actually fits.
The value is in the lead time. The cheapest moves, building a little cash, pausing an expansion, shortening a commitment, are all available early, while conditions still look normal, and they vanish once a downturn is undeniable to everyone. A reading that looks three to twenty four months ahead turns that early window from something visible only in hindsight into something a business can act on with time to spare.
The number updates every month as new public data is released, and each signal links back to its source, so anyone can trace the reading down to the data underneath it.