Three recession signals for the US economy

The Sahm rule is a real-time recession indicator proposed by the economist Claudia Sahm: take the three-month moving average of the unemployment rate and subtract the lowest value that same moving average reached over the previous twelve months. A gap of 0.50 percentage points counts as a trigger. It measures momentum in unemployment rather than its level, so it can fire while the unemployment rate is still low by historical standards. The Sahm value shown here is taken directly from the official FRED series SAHMCURRENT rather than recomputed, because the definition admits several reasonable readings at the edges and a locally derived number that disagreed with the FRED page would be impossible for a reader to check. In the Sahm bar chart the shaded band spans ±0.50 and the trigger is its upper edge at +0.50; the lower edge is only the symmetric half of the band and carries no meaning. A three-month average is also computed here from the unemployment rate, but only to draw the chart and to cross-check the pipeline; it never feeds the trigger.

A three-month moving average adds up the latest three monthly values and divides by three, sliding forward one month at a time; it smooths out single-month sampling noise so the trend is easier to see than in the raw monthly figures. The Sahm rule is built on the three-month average of the unemployment rate: the "3-month average" in the tile is the latest value of that average, and the "lowest of previous 12 months" is the lowest point it reached over the prior twelve months, and the gap between the two is the definition of the rule. The Sahm value in the tile is not obtained by subtracting those two numbers, though. The average computed here from the same unemployment series is used only to draw the dashed line in the upper Sahm chart, to fill the "3-month avg" reference column in the monthly table, and to cross-check against the official value; the Sahm value in the tiles, the key facts and the table always comes from the official FRED series SAHMCURRENT, and the two can differ slightly at the edges. A month missing from the source leaves every average that depends on it blank; nothing is filled in.

When the 10-year Treasury yield falls below the 2-year yield the spread turns negative, which is what is meant by an inversion. It reflects how the bond market is pricing future short-term rates: investors expect the central bank to cut later, so long-dated yields are pushed below short-dated ones. An inversion has preceded most US recessions of recent decades, but it is not a timetable — historically the gap between the first inversion and the start of a recession has run roughly 6 to 24 months, has varied widely from episode to episode, and inversions have also occurred without a recession following close behind. The spread series on this page takes the value on the last trading day of each month; while the current month is still running, its point is the latest daily reading, and the count of consecutive negative months follows the same rule. This page shows the raw spread and its sign, and does not infer timing from it.

Recession start and end dates in the United States are set by the Business Cycle Dating Committee of the National Bureau of Economic Research, which weighs the depth, diffusion and duration of a range of monthly measures of income, employment, production and sales — not the popular rule of two consecutive quarters of negative GDP growth. The committee waits until the data are stable enough, so a call typically arrives many months after the turning point and has sometimes taken more than a year. The FRED series USREC flattens that judgement into a monthly 0 or 1: it is 1 from the month after a business cycle peak through the month of the trough, and 0 otherwise. USREC is therefore an authoritative record of the past, not a live alarm.

Because no such number could be checked by the reader. The three signals measure different things on different clocks: the Sahm rule tracks momentum in unemployment, the yield spread is how the bond market prices the future, and the NBER indicator is a verdict reached after the fact. Combining them into a single percentage requires assigning a weight and a base rate to each one, and those coefficients have no published source and cannot be reproduced from anything on this page; change any of them and the headline figure changes while the reader sees none of it. The choice made here is to publish the raw reading and trigger state of each signal side by side and leave the weighing to the reader. Anyone who does want a probability model can use the yield-curve recession probability published by the Federal Reserve Bank of New York, which is public and comes with documented methodology.

Automatically, once every weekday. All four underlying series come from FRED at the Federal Reserve Bank of St. Louis: UNRATE (the unemployment rate, from the Bureau of Labor Statistics), SAHMCURRENT (the Sahm value, computed by the Federal Reserve Bank of St. Louis), T10Y2Y (the 10-year minus 2-year spread, whose two legs come from the Federal Reserve Board H.15 release), and USREC (the NBER reference dates). The series are mostly monthly, but the Bureau of Labor Statistics, the Federal Reserve and the NBER publish on dates spread across the month, so refetching daily gets new observations live as soon as they appear; a run that finds nothing new changes nothing. The full fetched dataset is published as JSON at /data/indicators.json. This is a personal project with no affiliation to FRED, the Federal Reserve, the NBER or any government agency; the page is a data monitoring tool and does not constitute investment advice.