The Yield Curve and Its Spreads

What the shape means, and what it is made of

A yield is two things added together: the average of the short rates expected over the life of the bond, and a term premium for the risk of lending long — the risk that inflation surprises, that rates move against you, and that you may have to sell into an illiquid market. The expectations component is the forecast; the premium is the price of not being able to forecast. Both change, which is why two identical forecasters can disagree about what a steepening curve means: it can be expected rates rising, or the premium rising, and the two have opposite implications for risky assets. The shapes have names and consequences. An upward-sloping curve is the normal state: longer money costs more, banks earn a spread between funding and lending, and the economy is expected to grow. A flat curve compresses that spread and squeezes the business model of lending. An inverted curve prices cuts — it says the market expects the policy rate to be lower in future than it is now, which is a forecast of a slowdown, because a central bank only cuts when growth is weakening or inflation is falling. Historically the inversion has preceded recessions with a long and variable lead, which is why the signal is treated as information about timing rather than a date. Two technical caveats keep this honest. First, the signal has been distorted since the financial crisis by quantitative easing and by the composition of central-bank balance sheets, which hold down long yields and flatten the curve for reasons that have nothing to do with expected growth — so the historical relationship may have weakened without being absent. Second, the curve is built from on-the-run Treasury issues, which carry a convenience premium because they are the collateral of choice: when demand for collateral spikes, that premium shows up as a lower yield rather than as a change in any forecast, and reading it as macro news is a category error. One curve, three readings — 3m10y: −10bp: mildly inverted — the academic recession signal ← · 2s10y: +35bp: positively sloped — the version in the headlines · 1-year rate one year forward: 3.65%: below today’s 1-year, so a cut is priced · 10-year rate ten years forward: 4.73%: the market’s view of long-run term premium The curve you see quoted is often the par yield curve, not a set of zero-coupon rates, and the difference matters for the forward arithmetic. Use zero rates when you compute forwards; use the quoted par curve when you want to know what a bond you can actually buy yields.

Forwards: the part that carries information

A forward rate is what the curve implies about a future period, and it is extracted by comparing two points rather than reading one. The one-year rate one year forward comes from the 1-year and 2-year yields: (1 + y₂)² ÷ (1 + y₁) − 1. With a 1-year at 4.05% and a 2-year at 3.85%, that gives 3.65% — the market is pricing a cut within the year, which is information no single yield contains. The same construction ten years out, using the 10-year and 30-year, gives 4.73%, a long-run rate materially above the current 10-year, which is the market pricing a rising term premium rather than rising growth. Forwards are what a borrower or a lender actually faces, which is why they are the right object for a decision rather than the spot curve. A company deciding whether to fix five-year debt should look at the five-year rate; a company deciding whether to refinance in two years should look at the two-year rate two years forward. And because forwards are built from adjacent points, they are noisier than the curve they come from — two basis points of rounding at the two-year maturity can move the implied forward by more than a basis point, so small forward movements are not news. The other use is as a benchmark for disagreement. If the market’s implied one-year rate a year from now is 3.65% and your own view of the economy implies 3.00%, the trade is not “bonds are expensive” — it is a statement about the path that you can check against the data as it arrives, which is the same discipline the reverse-DCF habit applies to equities. Identify the number you disagree with rather than the asset you dislike. • A forward rate is extracted from two points, and it is what the market will pay for money later. • Forwards are the right object for a financing decision, not the spot curve. • They are noisier than the curve they come from, so small moves are rounding. • Compare your own path with the implied path to make a disagreement checkable. The classic curve trades live here: a steepener buys the long end and sells the short, profiting if the spread widens; a flattener does the reverse. Both are expressions of a view about the difference, so a correct view on the level of rates can still lose money if the shape goes the other way.

The inversion that did not behave

A curve that inverts has been one of the more reliable warning signals in macroeconomics, and the reliability is worth describing precisely because it is usually overstated. In the United States, an inverted curve — the short end yielding more than the long end — has preceded every recession of the past several decades, with a lead time that has varied from roughly six months to well over a year. That record has made the signal famous, and the fame has two costs. The first is that a signal with a variable lead is not useful for timing, only for awareness; being early by eighteen months is indistinguishable from being wrong for a long while. The second is that the relationship was never mechanical — an inversion does not cause a recession, it reflects a market pricing a policy path that is restrictive relative to the long-run neutral rate, and whether that becomes a recession depends on what follows. The episode that should be studied alongside the record is the one that departed from it. After the tightening cycle that began in 2022, the curve inverted and stayed inverted for an unusually long stretch, and the recession that the historical pattern implied did not arrive on the usual schedule, with the economy cooling through disinflation instead. That outcome does not refute the signal; it is a reminder that the mechanism it proxies — a policy rate held above neutral long enough to break something — depends on how much of the economy is sensitive to the rate, on whether credit conditions tighten or loosen independently, and on whether the fiscal impulse is running the other way. When transmission is slow because households and firms have locked in fixed-rate debt, an inverted curve can persist without producing the usual damage. The transferable lesson is to treat the curve as a **description of what is priced**, not as a forecast. An inverted curve says the market expects policy to be easier than it is now, at some horizon, for some reason. Reading the reasons from the forward rates — a real slowdown, a lower inflation path, a lower term premium — is the analysis, and the three are distinguishable with the same curve. What the inversion cannot tell you is which of them the market is pricing, and that is the whole question. • Inversion has preceded recessions with a lead that varies from about six months to over a year — useful for awareness, useless for timing. • The signal reflects a policy path priced to be restrictive versus neutral; it is not a mechanism. • The most recent cycle inverted for a long stretch without the usual recession, because transmission was slow. • Read the forward rates to see which of the three reasons the market is pricing.

One curve, three numbers, and the choice of where to draw it

A curve with twenty points on it has almost all of its movement in three dimensions, and naming them is what makes a curve position legible. The first is the **level**: the whole curve moving up or down together, which is what a policy-rate change usually produces and what the word “rates” usually means. The second is the **slope**: the spread between the long end and the short end, which is the flattener or steepener that the curve trades express, and which contains the policy expectation and the premium stories this lesson has been building. The third is **curvature**: the middle of the curve moving relative to the ends — a butterfly, in the market’s language — which is where the difference between a 2s10s and a 5s30s spread lives, and where a great deal of positioning happens precisely because most participants are watching the first two. Three numbers describe nearly all of the variation, which is why a statement about the curve that specifies only “higher” is under-specified: the same upward move can be parallel, a steepener, or a twist in the belly, and each is a different portfolio outcome. The fourth question is which curve you are describing, and it is not a technicality. The **par yield curve** is built from coupon-bearing notes and bonds; the **spot curve** discounts a single payment at a single date; the **forward curve** is what the first two imply about future periods. Then there is the choice of instrument: government bonds in the cash market, interest-rate futures, or swaps indexed to an overnight rate. Each has its own credit content, liquidity and settlement convention, so the same “5-year rate” differs by a few basis points across four markets — and in a stressed week it can differ by much more, because the spread between a government bond and a swap of the same maturity is itself a risk measure rather than a rounding difference. When a chart of “the yield curve” is shown without saying which one, the shape may be indistinguishable and the level may not be, which is how two people can look at the same day’s data and disagree about whether the curve inverted. Which brings the read back to the term premium, because the decomposition of a yield into expectations and premium is the thing the three factors cannot give you on their own. The level moves, and you cannot tell from the level whether expectations rose or the compensation for holding duration did — and the two have opposite meanings for risk assets. The available devices are all estimates rather than measurements: model-based decompositions published by central banks and researchers, surveys of professional forecasters, and the comparison of yields with the corresponding inflation-linked bond. Each has a stated set of assumptions and each is revised, so the honest use is directional and comparative rather than precise: if the model-based premium has risen while surveys of expected short rates have not, the move in the long end is a term-premium move, and the equity market’s response to a term-premium move is different from its response to an expectations move. That distinction is the one the rest of this subject keeps returning to, and the curve is where it is most visible, because a single number is being asked to carry both a forecast and a price. Level, slope, curvature — and the instrument underneath — Level: The whole curve up or down — usually a policy move re-pricing expectations · Slope: 2s10s or 3m10y: the flattening or steepening the curve trades express ← · Curvature: The belly against the wings — 2s5s10s — where positioning is least watched ← · Par, spot, forward; cash, futures, swaps: Same maturity, different instruments — a few basis points in calm, more in stress The market’s own shorthand is worth learning because it encodes the decomposition: a “bear steepener” is long rates rising faster than short ones — usually a term-premium story — while a “bull flattener” is long rates falling faster, which is usually a growth story. The animal names are the analysis in two words.

Sources

• The expectations hypothesis and the term premiumStandard term-structure literature; Federal Reserve Bank of New York term-premium estimates
• Yield-curve inversions as a recession signalEstrella & Mishkin (1998), “Predicting U.S. Recessions: Financial Variables as Leading Indicators”
• Curve construction and forward ratesStandard fixed-income analytics; Treasury yield-curve methodology