Term Premium vs Rate Expectations: Why One Yield Move Means Two Things
The 10-year yield is a sum, not a price — and whether it rose because the Fed's path repriced or because investors demanded more compensation for duration changes almost everything downstream.
A forty basis point rise in the 10-year Treasury yield is not one event. It is at least two, and they carry close to opposite implications for equity multiples, for the bond side of a balanced portfolio, and for the credibility of the Fed path priced into the front end. The generalist reads the yield as a single number and asks what it means for stocks. An institutional desk asks a prior question — which component moved? — because the answer determines whether the move is a growth signal with an earnings offset or a pure derating with no offset at all.
The Yield Is a Sum, Not a Price
Start with the identity that organizes everything else. The yield on a long Treasury is, to a very close approximation, the average overnight policy rate investors expect to prevail over the life of the bond, plus a term premium — the additional compensation they require for accepting the risk that the expected path is wrong. The first component is a forecast of the Fed's reaction function stretched over years: where neutral sits, how quickly policy converges to it, how deep the average recession-response cut will be. The second component is a risk price. It is not a forecast of anything.
Why doesn't competition drive that risk price to zero? Because duration risk cannot be diversified away in aggregate. The government issues a stock of long-dated obligations and somebody must hold them; the term premium is the clearing price that induces the marginal holder to do so. That marginal holder is not an abstraction — it is a pension fund with a liability profile, a foreign reserve manager with a currency mandate, a bank managing a securities book against deposit behavior, or a central bank balance sheet that is either absorbing duration or releasing it. When the composition of that buyer base shifts from price-insensitive to price-sensitive, the compensation demanded shifts with it, and no amount of clarity about the Fed's intentions offsets that.
There is a second, orthogonal way to slice the same yield — into a real yield and a breakeven inflation rate — and conflating the two decompositions is one of the more common analytical errors in the space. Both slices contain premia. The real yield embeds a real term premium; the breakeven embeds an inflation risk premium on top of expected inflation. A rise in breakevens can therefore mean the market expects more inflation, or that it has become less certain about inflation and is charging for that uncertainty. Those are different regimes with different equity consequences, and only careful cross-referencing of the two decompositions separates them.
Three Layers of Why the Composition Matters
Layer one is discount-rate arithmetic. An equity is a claim on a stream of cash flows, and the further out the cash flows sit, the more sensitive today's fair value is to the rate used to discount them. This is why the market's longest-duration equities — high-multiple growth, early-stage profitability, anything whose value is dominated by a terminal assumption — behave like leveraged bond substitutes when the long end moves. That much a competent generalist gets right, and it is where most analysis stops.
Layer two is the numerator. Yields driven by expectations rarely move alone. If the market marks up its expected policy path, it is usually because the growth or inflation news that would cause the Fed to tighten has actually arrived. That same news lifts nominal earnings estimates. The discount rate rose, but so did the cash flows, and the net effect on the multiple is partially offset — sometimes fully. A term-premium-driven rise carries no such offset. Nothing about a widening in required compensation for duration risk improves next year's revenue. It is pure denominator, and pure derating. This is the substance behind the desk shorthand of a "good" versus a "bad" rise in yields, and it is why the equity response to identical yield moves can differ so sharply across episodes.
Layer three is the one that changes portfolio construction rather than security selection. The term premium is a price on uncertainty about the path. When it widens, the market is telling you the distribution of macro outcomes has fanned out — more inflation variance, more fiscal-supply variance, more doubt about the reaction function itself. That same widening distribution raises the risk premium investors demand on equities. And critically, it degrades the hedging property of duration: a bond only protects a portfolio if it rallies in the states of the world where equities fall. When the shared driver is uncertainty about the price level or about who will absorb duration supply, both legs fall together. The 60/40 investor discovers that the diversifying asset and the risk asset have merged into one factor — and because volatility-targeting and risk-parity mandates size positions off realized correlation, that discovery tends to produce mechanical selling that amplifies the original move.
Two Regimes From the Historical Record
The mid-2000s tightening cycle is the canonical illustration of an expectations-driven move that did not transmit to the long end. Through a long sequence of quarter-point hikes, the Fed moved the policy rate substantially higher while the 10-year yield went roughly nowhere. The curve flattened dramatically, and the puzzle was famous enough to earn a name — the conundrum — with the usual explanations centering on heavy foreign official demand for Treasuries and unusually subdued macro volatility, both of which compress the compensation required to hold duration. The portfolio lesson is the one that matters: equity multiples held up far better than a naive reading of "the Fed is tightening" would predict, because the discount rate that matters for long-dated cash flows never rose.
The 2013 taper episode ran the other way. The trigger was not a revision to where the funds rate would sit years out; it was a revision to who would be absorbing duration and how much. Long real yields moved sharply, and the assets that de-rated hardest were the ones whose valuations had been built on a compressed term premium — long-duration equity proxies, rate-sensitive real assets, and emerging market debt and currencies that had been financed on the assumption that the duration bid was permanent. Nothing about the growth outlook had improved to offset it.
The 2021–2022 inflation episode is the useful complication, and an honest framework has to include it. That repricing was overwhelmingly about the reaction function: the entire curve moved as the market rebuilt its view of how high and how fast policy would go. Growth multiples compressed violently — and bonds and equities still fell together. So a hostile stock-bond correlation is not the exclusive property of a term-premium shock. Any regime in which the dominant macro uncertainty is about the price level will do it, because inflation is the one shock that damages both legs simultaneously. The decomposition tells you the transmission channel; it does not license a mechanical rule about correlation.
The Strongest Objection: It's a Residual
The serious case against this framework is that the term premium is not observable. Nobody trades it. It is backed out as the difference between an observed yield and a modelled expected policy path, which means every estimate inherits the assumptions of its model. The best-known published decompositions — the New York Fed's affine term structure work and the Board's model in the Kim–Wright tradition, alongside survey-based approaches that substitute forecaster expectations for statistical ones — can disagree materially about the level, and some have produced negative estimates in periods of heavy central bank duration absorption, which is theoretically awkward. If the number you are reasoning from is a residual whose level depends on a specification choice, the objection runs, you are dressing up ignorance as insight.
That objection is correct about the level and wrong about the use. Portfolio decisions almost never hinge on whether the term premium is at one level or another; they hinge on whether it is widening or compressing, and on how much of a given yield move it explains. Directional changes are considerably more robust across specifications than levels are, and — more usefully — the direction has observable corroborants that require no model at all. Compare the move in the 2-year against the move in the 10-year and 30-year over the same window: a long-end-led steepening with a static front end is difficult to explain as a reaction-function revision. Check whether the move came through real yields or breakevens. Look at implied rate volatility, which prices the very path uncertainty the term premium compensates for. Watch the behavior of assets that trade off perceived duration scarcity. Track the realized stock-bond correlation. Where several of those independent readings agree with the modelled residual, the inference is sound; where they conflict, the honest answer is that the move is ambiguous and position sizing should reflect that.
This is why Sterling's Embedded Intelligence treats a headline yield move as unreadable until it has been decomposed, and treats the decomposition as a probabilistic judgment rather than a datum. Sterling's working discipline is to attribute the move first, corroborate it with at least two independent observables, and only then draw the equity or cross-asset conclusion — because the conclusion genuinely reverses depending on the attribution. Nothing else in rates analysis carries that much leverage over the downstream read.
One structural note that keeps this framework durable rather than cyclical. The term premium is where the fiscal and monetary architecture of a sovereign bond market shows up in a price. The maturity composition of government issuance, the direction of the central bank's balance sheet, the regulatory treatment of duration on bank and insurance balance sheets, and the willingness of foreign official holders to hold the currency's long-dated debt all act on the same variable. None of those move on a data-release calendar. All of them move slowly and occasionally in steps. That is precisely why an investor who only watches the expected policy path — the part that the Fed talks about, and that the financial press covers most heavily — can be surprised by a long-end repricing that no policy statement foreshadowed.
How to apply this framework
- Run the front-end/long-end split test on every yield move. Measure the change in the 2-year against the change in the 10-year and 30-year over the identical window. Front-end-led moves are reaction-function news and typically arrive with an earnings offset; long-end-led steepening with a static front end points to duration compensation, where the multiple absorbs the full adjustment.
- Attribute the move to real yields versus breakevens before drawing an equity conclusion. A rise in real yields with stable breakevens is the cleanest discount-rate shock to long-duration equity. A rise concentrated in breakevens raises the further question of whether expected inflation or inflation uncertainty is doing the work — the latter behaves like a term premium widening and hits the bond hedge as well.
- Treat the realized stock-bond correlation as a live diagnostic, not a static assumption. When duration stops rallying on risk-off days, the portfolio's true risk has risen without any position changing, and volatility-targeted mandates elsewhere in the market are being pushed toward the same de-risking trade.
- Ask who the marginal holder of duration is, and whether that is changing. Shifts in issuance maturity, central bank balance sheet direction, and foreign official demand act on the compensation channel rather than the expectations channel — which means they can move the discount rate that matters for equities without any change in the policy path being priced.