The yield curve, and what an inversion is actually a price of

A curve is a price list for time

A yield curve plots what a single borrower pays across maturities. Normally the longer loan pays more. An inversion is the case where a shorter yield sits above a longer one: lenders accepting less to lock money up for longer, which only makes sense if they expect short rates lower later.

Hold the object still for a second, because most of the confusion comes from forgetting what is being plotted. One borrower — a single government — issues debt maturing at many different dates. Each of those has a price, and each price implies an annual return to someone holding it to maturity. That return is the yield. Put maturity on one axis and yield on the other, join the dots, and you have the curve.

Credit risk is not what the curve is measuring. Every point on it is the same issuer, so the difference between the two-year and the ten-year cannot be about who is borrowing. It is entirely about when. That is what makes the curve the cleanest price of time available anywhere in markets, and it is why a curve article and a credit spread article are about genuinely different things: a credit spread prices who, a curve prices when.

So the whole question of this article reduces to one sentence. Lending for longer normally costs the borrower more. When it stops doing so, what has to be true?

Why the longer loan normally pays more

Two things sit inside a long yield, and it is worth naming them separately because an inversion is a statement about only one of them.

The first is the expected path of short rates. A ten-year loan and a sequence of ten one-year loans are competing ways of lending for a decade. If a lender could earn more by rolling short paper year after year, they would, and the ten-year price would have to adjust until the two were roughly comparable. So a long yield behaves, in large part, like an average of the short rates the market expects over the life of the bond. What sets those short rates is the central bank — the mechanism is in interest rates explained.

The second is the extra compensation demanded for the lock-up itself — conventionally called the term premium. Money committed for ten years cannot be redeployed, and its price swings far more on a given change in rates than short paper does. That sensitivity has a name and a measure: duration. A lender who has to carry that risk generally wants paying for it.

Put those together and the ordinary upward slope stops being a convention and becomes arithmetic. Even if the market expected short rates to sit perfectly flat forever, a positive term premium alone would tilt the curve upward. The upward slope is the default, and it takes something to overcome it.

What it takes to invert

Now the mechanism, which is the entire article in one paragraph.

For a ten-year yield to sit below a two-year yield, the expected average of short rates over the next decade has to be low enough to more than offset whatever compensation lenders want for the lock-up. There is no other way to get there. An inverted curve is the market paying up to lock in today's rate, because the rates it expects to be available later are lower.

That sentence is the honest one, and notice how narrow it is. It says what is priced. It says nothing about whether the price is right, and nothing about what happens next. The market that produced it can be wrong, and frequently has been about far simpler things. A price is a weighted opinion of everyone currently holding a position — a useful thing to be able to read, and not a fact about the future.

There is a trade-off in this framing that most explanations skip. Splitting a long yield into an expectations part and a term premium requires a model, and reasonable models disagree. Neither component is observable; only their sum is printed. So the interpretation you gain — "the curve is pricing a lower short-rate path" — is bought with precision you do not have about how much of the move was expectations and how much was compensation for duration. That is the conventional reading, held as a judgement, not a measurement.

Which is also why the reflexive next question is the wrong one. People ask what an inversion predicts. The answerable question is what it prices, and those are different enough that answering the second well is more useful than answering the first badly.

2s10s: a spread, in basis points

The curve has dozens of points on it and market commentary has compressed it to one number. That number is usually 2s10s: the ten-year yield minus the two-year yield. Nothing more complicated than a subtraction.

It is quoted in basis points — a hundredth of a percentage point, so 100 basis points is 1 percentage point. The convention exists because the moves that matter here are small relative to the yields themselves, and "the spread narrowed by 0.12 percentage points" is a worse sentence than "the spread narrowed 12 bp".

Here is the whole quantity, on numbers that are illustrative and invented for the arithmetic — no market has been quoted:

A negative 2s10s is the inversion. There is no separate test and no threshold; the sign of the subtraction is the whole definition.

Two cautions come free with the arithmetic, and both are routinely dropped.

The pair is a choice, and different pairs can disagree. Ten minus two is a convention, not a law. Ten minus three-month, thirty minus five, ten minus the policy rate — each is a legitimate slice of the same curve, and on any given day some can be negative while others are not. A curve is a shape; any single spread is one chord drawn across it. Anyone who says "the curve inverted" without saying which pair has told you less than they think.

A spread throws away the level. Row C above and a curve at 8.9% and 8.4% produce the identical −50 bp, and they are not the same environment for a borrower, a saver or a company refinancing debt. The subtraction is designed to discard the thing it discards. Useful, as long as you remember it went missing.

The level tells you less than the direction of change

Here is the part that changes what you actually look at. A spread is the difference of two numbers, so the same change in the spread can arrive four ways, and they are not interchangeable.

The market's vocabulary for this uses bull to mean yields falling and bear to mean yields rising — borrowed from bond prices, which move opposite to yields. Combine that with which end moved more, and you get four cases.

Read the third and fourth rows together and the specific mistake becomes obvious. A curve can flatten by 30 bp because the world got gloomier at the long end, or because policy expectations got tighter at the short end. Same 30 bp. Opposite news. The spread alone cannot distinguish them — so a report that quotes the change and not the two legs has handed you a number with its meaning removed.

Note the hedge in that fourth column, and that it is doing real work. Every entry says consistent with, not caused by. A curve move is a price change; attributing it to a cause is interpretation, and on any given day several plausible interpretations survive. The discipline is to name which leg moved first, then argue about why — never to read a cause off the spread and work backwards.

This is why the direction of change carries more information than the level. A curve sitting at −50 bp for a month is one fact repeated twenty times. A curve going from −50 bp to −20 bp is somebody's expectations changing, and the two legs tell you whose.

The un-inversion is the more discussed moment

An inversion, once established, is a state. It can persist. Nothing about it forces a resolution on any schedule, and the absence of a clock is precisely why it is so often misused.

What draws far more market commentary than the inversion itself is the re-steepening — the curve crossing back through zero. That is an observation about how the curve gets discussed rather than a finding about markets. The two legs being discussed, though, are not built alike, and that part is mechanical.

The two legs are not equally free to move. The two-year is dominated by the expected policy path over a short, well-defined horizon, and that path gets repriced the moment anything changes the near-term picture. The ten-year contains the same information diluted across a decade, plus a term premium that answers to slower things — issuance, the appetite of long-duration holders, the long-run inflation view. That is a statement about what each leg is made of, not a claim about which one will move: the front end is the leg with the shortest route from news to price.

Which sets up the distinction that matters, and it is the same four-way table applied to one moment:

• A bull re-steepening — the front end falling away — is the market repricing the near-term policy path downward. Something changed about the next few quarters.
• A bear re-steepening — the long end rising — is about the compensation demanded for duration, which is a different subject entirely and reaches equities by the route traced in US yields and Indian share prices.

Both restore a positive spread. They are not the same event, they do not have the same implications for what a distant rupee is worth today, and a headline reporting only that "the curve has un-inverted" has told you the sign of a subtraction and nothing else.

What none of this licenses is a forecast. Naming which leg moved is reading a price more carefully. It is not knowing what comes after it, and this article does not claim to and cannot.

The recession-signal framing, handled honestly

The curve is best known for one claim: that an inversion warns of a recession. It is worth being exact about what that claim is and is not, because the gap between the two is where most curve commentary lives.

Start with the part that is not in dispute. The mechanism is coherent. If lenders collectively expect short rates to be lower in a few years than they are now, one ordinary reason to expect that is that they expect conditions requiring easier policy. So the inversion and the concern are describing the same underlying expectation. That is not a coincidence, and it is not a discovery either — it is the definition, read back.

Now the parts that get skipped:

It is an empirical claim resting on a small number of episodes. Recessions in any one large economy are rare events. A handful of observations, each with its own policy regime, its own inflation backdrop and its own market structure, is a thin base for a rule — and the sample is small enough that the choice of maturity pair and of lead-time window can change the answer. This article states no count, no lead time and no hit rate, because any of those would need a source, a stated definition of recession and a stated pair before it could mean anything.

An expectation being priced is not the event occurring. The curve tells you what lenders collectively expect. Expectations are wrong routinely; that is why prices move. Reading a priced expectation as a scheduled outcome converts a market observation into a forecast, which is a category error and not a small one.

Neither is it a mechanism for Indian equities. A US curve is a US price. It reaches an Indian share through the discount-rate and flows channels traced elsewhere in this cluster, and through nothing else. That two series moved together over some window is a fact about the window — a correlation is consistent with a channel and is not, on its own, evidence that one exists, since a third thing moving both produces the same picture. The argument has to run from the mechanism to the data every time.

The rupee curve is a different animal on the same skeleton

India has a government bond curve too, and it is built from exactly the two components described above: an expected path for the policy rate, plus compensation for the lock-up. The skeleton is universal.

What differs is the composition of demand. A sovereign curve's term premium reflects who is willing to hold long paper and on what terms, and that is a structural feature of a market rather than a global constant. Where a large share of long-dated government debt is held by institutions with genuinely long liabilities, the compensation those holders require is set by their own matching needs rather than by a trading view — which is a different price-setting process from one dominated by leveraged traders. That is a mechanism, stated with no figure attached, and any figure put on it would need a primary source and an as-of date.

Two consequences follow that are worth carrying.

A rupee curve and a dollar curve can disagree, and neither is malfunctioning. The RBI sets policy for Indian inflation and growth, so the Indian front end is anchored to a domestic path. The two curves are linked through the currency and through global capital, not welded together.

The Macro page carries the US curve, not the Indian one, for a reason that is about role rather than importance. The US curve is the anchor that global capital discounts against, which is what makes it a global input to Indian asset prices. The Indian curve is closer to a domestic output. Both matter; they do different jobs, and the page is a global-conditions screen.

What the curve does inside our composite score — and what it does not

Everything above is readable off any two yields and a subtraction. FNOTrader's Macro page puts the rates channel next to the dollar, energy, credit and volatility, and three things about how it does that are worth stating plainly.

Colour means impact on Indian equities, not the direction of the number. A tile turns green when the reading is supportive for Indian equities through the channel it belongs to. So US yields rising shows red, and — the one that catches everyone — USD/JPY rising shows green too, because a weaker yen leaves the carry trade undisturbed. There is no "up is red" rule on the page. Read the colour as an arrow on the number and you will misread half the tiles.

The composite is a weighted average, and the weights are our judgement. The page scores the set as score = 100 × Σ(w_ic_i) ÷ Σ(w_i), over twenty weighted market inputs whose weights sum to 1.33 — not the tidy half-dozen a glance at the page suggests. The largest components are the dollar index at 0.20, the US 10-year yield at 0.15, Brent at 0.12, and USD/JPY and VIX at 0.10 each; the US 2-year sits at 0.06 alongside USD/INR, high-yield spreads, India VIX and the S&P, with MOVE at 0.05 and a tail of smaller inputs below that. One further contribution is not a tile at all: foreign flow — FII net cash, weight 0.12, saturating at ±₹5,000 crore — is folded into the same score at scoring time, which takes the full set to twenty-one contributions and 1.45 of weight on a day when everything reports. Regime boundaries are drawn at ±20, and a stress override fires at a score of −35, or on a one-day VIX move of +20% or more, or a one-day USD/JPY move of −1.2% or worse. Every one of those numbers is FNOTrader's design choice — a considered view about what matters most to Indian equities, not a measured constant and not a fitted coefficient. A different reasonable weighting gives a different score from identical inputs. The divisor is the sum of the weights that actually scored that day, so 1.45 is a ceiling rather than a constant: an input whose source has not published drops out of both halves of the fraction rather than counting as a zero.

Now the part that is genuinely non-obvious, and it falls straight out of that weight list. The curve is not in the composite at all. The rates channel shows the US 2-year, 5-year, 10-year and 30-year, plus a 2s10s tile the page computes itself by subtracting the first from the third. Of those five, exactly two carry weight: the 10-year at 0.15 and the 2-year at 0.06, both with the sign that treats a higher yield as a headwind for Indian equities. The 5-year, the 30-year and the 2s10s tile itself are displayed and unscored.

Work through what that implies, because the arithmetic is the interesting part. A pure flattening — the two-year rising while the ten-year falls by a comparable amount — moves the spread by the two legs added together, which is a large move by curve standards. Inside the score the same event produces two contributions of opposite sign, so they partly cancel. What survives is the difference between the two weights rather than their sum: the spread got both legs, while the score gets 0.15 minus 0.06, which is 0.09 against an all-present total weight of 1.45. The score edges in the ten-year's direction and otherwise sits still while the curve does something dramatic. The composite is reading the level of global rates. It is not reading the shape of the curve. That is a deliberate design choice, and it is worth knowing about: on a flattening day the score and the curve tile can look like they disagree when they are simply measuring different things.

One mechanical caution on the tile itself. The two legs do not come from one place — the 10-year is a market-data series that moves through the session, while the 2-year comes from FRED, whose daily series typically publish a session or two behind. A spread computed from two feeds on different cadences can therefore appear to move on a day when only one leg has updated. Before reading a small change in a computed spread as news, check that both legs are current; a stale leg produces a real-looking move out of nothing.

How the rates channel sits alongside the other four, and how the tiles are meant to be read together rather than one at a time, is the subject of the macro pillar. This article is the long version of one tile.

Common questions

Continue reading

More in Macro & Markets · App: Options Analytics · Definitions: glossary · Free tools: calculators · All: every article