A promise of one hundred dollars is not a hundred dollars. It has a date, a person who must keep it and conditions under which the money may fail to arrive. The bond gives that promise a form that can be sold. Its owner need not wait until the date written on the certificate; someone else may pay for the remaining wait. The price is therefore a judgment about a future payment made in the circumstances of the present. Change those circumstances and the price can move while every word of the promise remains the same.
Begin with an imagined bond that pays one hundred in a year and nothing before then. Assume payment is certain, the currency is fixed and there are no transaction costs. If the comparable one-year yield is five percent, its price is about ninety-five dollars and twenty-four cents: that sum grows to one hundred at the stated rate. At ten percent the price falls to about ninety dollars and ninety-one cents. The issuer has promised the same hundred. A new buyer pays less because other uses of money now offer more. The inverse movement of price and yield is the arithmetic of that comparison, rather than evidence that the issuer has suddenly become less honest.
The coupon complicates the schedule without changing the principle. A bond paying five in one year and one hundred and five in two years is worth one hundred when each payment can be discounted at five percent. At ten percent, under the same flat-rate assumption, it is worth about ninety-one dollars and thirty-two cents. Each payment has its own distance from today. The yield to maturity compresses the whole schedule into the single rate that reproduces the price. It does not promise that coupons can be reinvested at that rate, or that a holder who sells early will receive it. The realized return also depends on what happens between purchase and exit.
Time gives the price its sensitivity. In the same illustration, raising the yield from five to ten percent reduces the one-year zero-coupon bond’s price by about four and a half percent. A ten-year claim to the same hundred falls from about sixty-one dollars and thirty-nine cents to thirty-eight dollars and fifty-five cents, a decline of roughly thirty-seven percent. More of its value lies farther away. Yet maturity alone is an incomplete measure: a bond returning substantial cash early is differently exposed from one paying everything at the end. Nor does holding to maturity abolish every risk. It removes the need to sell at an intervening market price if the promise is honored; it does not fix the purchasing power of the money received.
Put such claims beside one another and the scattered promises form a curve. At each maturity there is a price, and from that price a yield. To compare dates properly, the claims must be made comparable in other respects. A corporate bond’s yield may contain compensation for default and illiquidity that a government benchmark does not share; a foreign-currency claim adds another exposure. Even a curve drawn from one issuer requires care about coupons and conventions. The shape is useful because it holds some differences still while allowing the date to vary. It becomes misleading when every difference between securities is assigned to time.
The curve also permits a more precise question. Suppose the one-year yield is two percent and the annualized two-year yield is three percent, both for certain zero-coupon payments in the same currency. The forward rate for the second year is approximately 4.01 percent, because 1.03² divided by 1.02 must equal one plus that rate. It is the rate that makes the two locked-in paths agree under the assumed frictionless trading conditions. Calling it a forecast adds a proposition the arithmetic has not supplied. An investor can agree to an exchange of payments at a forward price while expecting a different future spot rate, because bearing the uncertainty has a value of its own.
The term premium names that additional part of the price of duration. It is not simply the difference between a ten-year yield and today’s overnight rate: expected short rates may change throughout the intervening years. And the premium need not be positive. A long bond that tends to become valuable when the rest of a portfolio suffers can provide insurance worth paying for through a lower expected return. Clarida emphasized this possibility in his 2019 account of bond yields. The same exposure to time that makes one investor demand compensation can help another meet a distant liability or withstand a loss elsewhere. The market prices the promise together with its place in a portfolio.
Whose expectations, then, does the curve encode? The people willing and able to change their holdings at the prevailing prices, acting under their constraints. An insurer matching long liabilities, a leveraged trader facing a funding limit and a household protecting a deposit need not want the same thing. Neither their purchasing power nor their capacity to bear losses is equal. The observed price reconciles their orders; it does not count their forecasts as equal votes. Changes in the supply of long securities, or in the demand to hold them, can move the curve even when the expected path of policy has not changed. Extracting an expectation requires a model of what else the price contains.
This leaves the curve informative without making it an oracle. In their 1996 study, Estrella and Mishkin compared the Treasury ten-year–three-month spread with other indicators in forecasts made beyond each model’s estimation period. The spread supplied useful warnings of U.S. recessions, especially at longer horizons in their sample. The output was a probability, not a dated announcement. Even their successful signals varied in strength across recessions. A warning that changes the odds can be useful while often being followed by no recession; a realization cannot be called impossible merely because its estimated probability was low.
There is a further difficulty: policy listens. A central bank may respond to the weakness that investors foresee, while investors revise their prices in anticipation of that response. Falling long rates can reflect expected easing, a demand for safety, purchases of duration or several of these together. A lower rate produced by a smaller premium need not carry the same message as a lower rate produced by a weaker expected economy. The observed curve is one outcome of this interaction. Reading it sensibly means asking what change would produce the shape, and checking that account against evidence beyond the shape itself.
The bond makes a promise tradable; the curve makes the price of waiting visible without making its reasons transparent.
The certainty in the certificate and the uncertainty in its price belong to different things. The first specifies what someone owes; the second asks what that obligation is worth among the alternatives available now. Expectations enter, but so do risk, liquidity, constraints and the desire to be protected when other promises fail. To read the curve is to interpret that combination, with enough confidence to learn from it and enough discipline to ask what the price alone cannot establish. The argument remains open.