The Yield Curve Explained: Normal, Flat, Inverted and What Each Means

In February 2005, giving his semiannual monetary policy testimony to Congress, Alan Greenspan observed that long-term interest rates had drifted lower over the preceding months even though the Federal Reserve had raised its short-term target by a percentage point and a half, which ran against almost everything experience suggested should happen. He did not offer an explanation. He said that the behaviour of world bond markets “remains a conundrum,” the word attached itself to the episode permanently, and his successor would later put a name to one candidate explanation, a global surplus of savings looking for somewhere to sit, in a lecture the following month.

Look at what that exchange reveals about the object under discussion. The yield curve is a single line plotting what the US Treasury pays to borrow for every length of time, from four weeks out to thirty years. Its level tells you what money costs. Its slope tells you what the market believes money will cost later, and a Federal Reserve chairman, holding every lever that sets the short end of it, could not account for what the long end was doing.

What the line is made of, and who sets each end

Each point on the curve is a traded yield. It is what a Treasury security of that maturity actually changes hands at on the day. Treasuries are used for this because they carry no meaningful credit risk, so what is left in the line is the price of time by itself, and the instruments involved, from four-week bills to thirty-year bonds, are covered in the Treasury securities guide.

The short end is anchored, and anchored hard. The fed funds target sets the overnight rate, and bills maturing within weeks or a few months trade close to where the market expects that target to sit across their brief lives, because there is very little else for them to be about. How the Federal Reserve works covers how that target is enforced.

The long end is a different creature. A ten-year yield is approximately the average short rate expected over the coming ten years, plus a term premium, which is the extra compensation demanded for the risk that the expectation turns out wrong. Neither of those two components can be observed directly. Both have to be estimated, by models that disagree with each other, which is why any confident statement about what the long end is signalling deserves a question about which model produced it.

The four shapes, and which end moved

Shape	What it looks like	What it usually reflects
Normal	Upward sloping, long yields above short	Growth expected, term premium positive
Steep	Sharply upward sloping	Easy policy now, higher rates or inflation later
Flat	Short and long yields nearly equal	Transition, the market undecided
Inverted	Downward sloping, short above long	Policy tight now and expected to be eased later

The shape changes for two quite different reasons. Separating them is most of the skill. A curve can steepen because long yields rose. That is a bear steepener, and it usually reflects worries about growth or inflation ahead. It can equally steepen because short yields fell, and that is a bull steepener, which normally arrives when the Fed has started cutting. The same distinction applies to flattening, in both directions. Quote the spread alone and you have thrown away which end moved. That is generally the half that mattered.

The two spreads, and why the Fed’s own researchers preferred one

Almost all public discussion of the curve reduces it to one number. It is the difference between two points on the line.

The 2s10s spread is the ten-year yield minus the two-year yield. Take round numbers, purely as an illustration. The ten-year sits at 4.00% and the two-year at 4.50%. The spread is 4.00% - 4.50% = -0.50%, quoted as minus 50 basis points. A basis point is one hundredth of a percentage point. The two-year is the popular choice. It captures where policy is expected to go across the next cycle.

The 3m10y spread takes the ten-year yield minus the three-month bill yield instead. Work done at the Federal Reserve Bank of New York in the 1990s, building on academic research from the previous decade, generally favoured this pair for forecasting recessions, on the reasoning that a three-month bill reflects the policy rate in place with almost no forecast content of its own, which makes the spread a cleaner reading of how tight policy is against long-run expectations.

The two spreads usually agree. When they part company it is normally because the two-year has already priced in cuts that the three-month bill cannot register until the Fed actually delivers them, which means the disagreement is itself informative about timing. Both can be plotted on the yield curve tool.

Why anybody accepts less for longer

Nobody lends for ten years at a lower rate than they can get for two out of civic feeling, and they do it because they expect the two-year rate to be gone.

Consider an investor choosing between a two-year note at 4.50% and a ten-year at 4.00%. Both figures are illustrative. Buying the two-year means reinvesting two years from now at whatever rate exists then. If this investor expects the overnight rate to have been cut to 2.50% by that point and to stay in that neighbourhood, then locking in 4.00% for a decade is plainly the better trade, and the 4.50% on offer is a trap dressed as a bargain. When enough investors reach that conclusion, they buy the ten-year. Its price rises, its yield falls, and the curve inverts.

An inverted curve is therefore a statement about expected policy easing. The economy comes into it only at one remove. It becomes a recession signal because the Fed normally eases when conditions are deteriorating. The chain runs through policy expectations, and every link in it is a forecast made by people who can be wrong, which is why the signal is real and also why it is imprecise in exactly the way that matters most to anyone trying to use it.

The record, the lags, and the thin sample

The 2s10s curve has inverted before each US recession since the late 1960s. That record is unusually consistent for a single indicator, and it is why the curve is watched closely by people who never otherwise look at a bond.

Two things about the record deserve equal weight with the record itself. The lag from inversion to the start of a recession has run roughly six to twenty-four months, a range wide enough that inversion tells you very nearly nothing about when. The signal has produced at least one false positive. That was around 1966, when the curve inverted and the expansion continued regardless.

There is also a sample problem that rarely survives contact with a chart. Counting US recessions since the late 1960s gives fewer than ten observations. Ten observations is a thin basis for confidence in any indicator. The pattern looks clean when it is drawn with the recessions shaded in. The shading is added afterwards by people who already know how the story ended.

A further complication belongs here. Two decades of large-scale asset purchases left central banks holding enormous quantities of long-dated government debt, which compresses the term premium in the part of the curve the signal depends on. Whether that changes what an inversion means is an open question, treated in quantitative easing and tightening, and the honest answer is that nobody has enough post-purchase cycles to say.

What the shape does, and what it can be used for

The curve is also a mechanism. Most commentary leaves that part out. Banks borrow short and lend long. When the curve is steep, that spread is profitable. Credit is easy to obtain. When the curve inverts, lending for ten years at less than the cost of overnight funding stops making sense, credit standards tighten, and the marginal borrower is refused. That tightening is a channel through which an inverted curve helps produce the slowdown it is supposed to predict, which is also why bank shares trade on curve shape, one of the links examined in how interest rates affect stocks.

The practical uses are narrower than the headlines suggest. The curve tells you what the bond market expects from policy, which can be set against what the Fed says it intends, a gap examined in fed funds futures and the dot plot. It tells you whether extending maturity is being paid for at all, since a flat curve offers no additional yield in exchange for the additional price risk described in duration and interest rate risk. And it belongs on a dashboard, alongside jobless claims, the Sahm rule and the leading index, which is how it is handled in recession indicators.

One detail from the historical record inverts the usual instruction. The curve has typically re-steepened before a recession begins. The Fed starts cutting the short end in response to what it is seeing. An investor watching only for inversion therefore sees the signal disappear at roughly the moment the risk is greatest, which is a strange property for a warning light and a good reason to treat it as one input among several.

If the price and yield mechanics underneath any of this feel hazy, go back to how bonds work, then plot the shape for yourself on the yield curve tool and test what stuck with the bonds and yields quiz.