What Is the Yield Curve? Normal, Flat and Inverted Explained

What the line actually plots

Along the bottom is time. It runs from three months on the left out to thirty years on the right. Each step along it asks the same question: how long are you lending for? Up the side is the interest rate the borrower pays for each of those lengths. Put a dot at every maturity, join them up, and you've drawn the yield curve.

The borrower is the same at every dot. When people say "the yield curve" without naming whose, they mean the US government's. It sells debt at every length from four weeks to thirty years, on the same day, to the same market. So on any given afternoon you can read off what it costs the Treasury to borrow for three months and what it costs to borrow for three decades.

Holding the borrower still is the whole trick. Lend to a shaky company and you'd want a fatter rate, because it might not pay you back. That extra is credit risk, and our guide to the types of bonds walks the whole ladder of it. A Treasury curve has none of that in it. Every point on the line is the same borrower with the same chance of paying you, so the only thing changing as you move right is time.

Why it normally slopes up

Think about lending someone money for three months against lending it for thirty years. Over three months you have a fair idea what the world looks like. Over thirty you don't. Inflation could eat a chunk of what you're owed. You might want the cash back for something else and be stuck. So you ask for more.

That's what a normal curve is, and it's what the market looks like today. On 12 August 2026 the US government paid 3.87% to borrow for three months. For ten years it paid 4.68%. For thirty years it paid 5.24%. Every step to the right costs the borrower more than the step before it.

The gap people quote most is the ten-year rate minus the two-year rate. The two-year pays 4.20% right now and the ten-year pays 4.68%, so that gap is plus 0.48 percentage points. Traders call it the 2s10s. Take the three-month bill instead of the two-year and the gap is plus 0.81. Both are positive, which is the normal state of the world.

Five real curves

Here are five days that actually happened, read off the Treasury's own published rates. Step through them. The curve before the one you're on stays behind it as a grey dashed line, so you can watch the shape roll over and then roll back.

Normal, flat and inverted

Curves get sorted into three shapes, and the names say what they mean. A normal curve rises left to right: longer money pays more. A flat curve sits close to level, so lending for ten years pays roughly what lending for two years does. An inverted curve falls: short money pays more than long money.

Flat is usually a stage rather than a destination. A curve on its way from normal to inverted, or back again, spends a while looking flat as it crosses over.

Now the thing that trips people up, and it's worth being blunt about. Normal doesn't mean low, and inverted doesn't mean high. Look at June 2014 and April 2000 in the chart above. June 2014 is the normal one, rising the whole way across. April 2000 is the inverted one. And the normal curve sits below the inverted curve at every single maturity they share. Three-month money paid 0.04% in 2014 against 5.90% in 2000. Thirty-year money paid 3.38% against 5.71%. Normal describes which way the line tilts. It tells you nothing about how high it sits.

What an inverted curve means

An inversion looks wrong the first time you meet one. Someone is being paid more to lend for two years than to lend for ten, which sounds like a trade done backwards. It isn't. Two things are going on underneath, and they push in the same direction.

The market's pricing in rate cuts

A long rate is roughly an average of the short rates the market expects over its life. You could lend for ten years in one go, or you could roll a three-month loan forty times and end up in the same place. Those two routes have to pay about the same, or everyone piles into whichever one pays better until they do.

So a ten-year rate below a two-year rate is saying something specific. The market expects short rates to be lower in a few years than they are now. Short rates are set by the central bank, which in the US is the Federal Reserve, and it cuts them when the economy needs the help.

Put those together and the inversion isn't really forecasting a recession. It's pricing the response to one. The market thinks rates will have to come down, and rates come down when something has gone wrong.

Banks stop lending

The second one isn't a forecast at all. It's a business getting worse in real time. A bank takes in deposits and other short-term money, and lends it back out long, as mortgages and business loans and car finance. It lives on the gap between the two rates.

Invert the curve and that gap closes or turns negative. New lending stops paying. So banks write fewer loans, borrowers on the edge get turned down, credit gets harder to come by, and the economy slows for reasons you could go and point at. That's a mechanism, not a metaphor. An inverted curve doesn't only sit there predicting the slowdown. It helps bring it about.

What the record actually says

You've probably heard that an inverted curve predicts recessions. Here's the actual record it's built on.

Below is every sustained inversion of the 2s10s since the daily series began in 1976, meaning every run of at least twenty trading days below zero. Against each one is the business-cycle peak that the National Bureau of Economic Research later dated. NBER is the committee that officially calls US recessions, and it does its calling long after the fact.

Start with how much evidence that is. Six episodes in fifty years. That's the whole basis for one of the most quoted claims in finance, and six is not many. Everything below is read off a handful of events, each one sitting in a different economy.

First, the lead time. It runs from ten months to eighteen months, and it has never come in the same twice. The shortest was September 1980, ten months ahead of the July 1981 peak. The longest was January 1989, eighteen months ahead of July 1990. Someone who sold everything on the day of an inversion would have sat out most of a year at the very least. In several of those episodes stocks kept making new highs during the wait.

Second, the start dates are softer than a table makes them look. The 2006 episode above begins in August. But the spread had already dipped below zero for twenty-five days between January and March that year, then popped back up. Take the August start and the lead is sixteen months. Take the January flicker and it's twenty-three. This table uses the sustained run, which is a defensible rule, and you should still know a choice was made there.

Third, and this is the row to sit with. The curve inverted on 6 July 2022, and the sustained run ended on 26 August 2024. That's 537 trading days, the longest inversion in the fifty-year series, and it went 1.08 percentage points deep at its worst. The spread then flicked below zero twice more in early September, on the 3rd at −0.04 and the 5th at −0.02, and it has been positive ever since. NBER's most recent dated business-cycle peak is still February 2020.

So no recession has been dated after the longest inversion on record. That's where the record stands, two years after the curve went back to normal. And notice what dating a recession involves. NBER names the peak month well after it's gone by, once the data has settled. A signal you can only confirm years afterward isn't a signal you can act on.

Which two points, though?

There's a wrinkle sitting under all of this. The curve has eleven maturities on it, so "the curve inverted" depends entirely on which two you subtract. Two pairs are in common use: the ten-year minus the two-year, and the ten-year minus the three-month bill. Most of the time they agree with each other. Sometimes they don't, and the gap between them is not small.

2019 is the clean example. On the two-year measure the curve was inverted for three days, on 27, 28 and 29 August 2019, and that was the whole event. On the three-month measure it was inverted for about five months, from May to October. Same market, same year, one headline or a summer of them depending on which pair you picked.

April 2000 manages it inside a single day. That curve, the inverted one in the chart above, reads −0.52 on the ten-year minus the two-year, which is decisively inverted. The same day's ten-year minus three-month reads −0.04, which is flat. Both numbers are right. When someone tells you the curve has inverted, it's fair to ask which two points they're talking about.

What it means for you

A signal whose lead time ranges from ten months to eighteen, and whose longest episode has nothing dated after it, isn't a trigger. There's no version of "the curve inverted, so do this on Tuesday" that the record above supports.

What does change when the shape changes is the arithmetic of what you're being paid. Under an inverted curve, cash and short bills pay more than long bonds do. Under a normal curve like today's, they pay less. Right now that's 3.87% for three months against 4.68% for ten years and 5.24% for thirty.

That's the question the curve genuinely answers for a saver. Long bonds swing in price when rates move, and how hard they swing is measured by duration, which our duration guide covers properly. An inverted curve means you're carrying that swing and being paid less than cash for carrying it. A normal one means you're being paid something. Whether that something is enough depends on what the money is for, and on how long you can leave it alone.

One last thing, because this is a US chart. Gilts, bunds, JGBs and every other government bond market have yield curves, and they're built the same way, from one borrower and many return dates. They usually slope up too. The famous inversion research is all US Treasury data, though, so don't assume the lead times travel. If you're investing somewhere else, look up your own market's curve and its own history.