Why the Equity Premium Has a Term Structure
The subject of this series is the term structure of the equity risk premium: how much extra return investors require, relative to a risk-free investment, when an equity payoff arrives at different horizons. This opening asks why that compensation should be treated as a curve rather than as one market-wide number, and why the distinction matters for valuation and for interpreting stock-market evidence. Once the basic picture is clear, the more technical question becomes manageable: how can researchers infer compensation at several horizons from prices, cash flows, and economic conditions?
One equity claim, many payment dates
An equity investment is not a promise to pay one amount on one date. Its value reflects a stream of possible future cash flows: dividends, buybacks, or cash that might instead be retained and reinvested in the business. Some of those payments may arrive soon; others depend on the firm surviving, growing, and remaining profitable for many years. The investor who buys the claim bears uncertainty over the entire stream.
The equity risk premium is the compensation for bearing that uncertainty above what a comparable risk-free investment is expected to provide. “Comparable” matters. A one-year equity premium should be compared with a risk-free return over roughly one year; a premium attached to a much longer investment horizon should be compared with the relevant longer-horizon real rate. If the benchmark changes with maturity, the excess return can change even when the underlying equity claim has not.
This is why an equity-premium term structure is more informative than the phrase “the equity premium” by itself. A term structure is simply a schedule indexed by time to payment or time to investment horizon. At each point on the schedule, the question is different: what compensation is required for equity exposure that is concentrated one year from now, five years from now, or much farther into the future?
That distinction also separates expected returns from realized returns. A realized return is an outcome that has already occurred. It may be high because investors were well compensated, because an unexpectedly favorable event occurred, or because risk that seemed remote did not materialize. An expected or required premium is forward-looking. It is an inference about the compensation embedded in prices and beliefs before the outcome is known. Long-run averages of realized returns can inform that inference, but they are not the same object.
The curve is therefore not a decorative refinement of a familiar statistic. It describes how the market prices the timing of risk. A short-dated cash flow can be exposed to a different set of shocks from a distant cash flow, and the market can demand different compensation for those exposures. The difference may concern the level of the premium, its slope from short to long horizons, or both.
Why timing changes valuation
The connection to valuation is direct. A company’s value is the present value of its expected future cash flows. In the simplest dividend-discount model, one required return is applied to every future dividend. That calculation is useful because it turns a complicated stream into a single rate that can be compared across firms or periods. But it also assumes away an important possibility: the return required for a payment next year is the same as the return required for a payment many years from now.
If required compensation differs by horizon, then two firms with the same total expected cash flow can have different values because their payment schedules differ. One may already generate substantial cash and return it to shareholders. Another may be valued mainly on profits expected later, after a long period of investment and growth. A single discount rate can still summarize either valuation, but the summary may hide how much of the price depends on short-, intermediate-, or long-dated cash flows.
That hidden timing matters for net present value and the cost of capital. An investment project whose benefits arrive quickly is not exposed to the same discounting schedule as one whose benefits arrive after a decade. Applying one market-wide number to both can make the comparison appear precise while silently imposing the same horizon profile on unlike cash flows. The issue is not that one constant rate is always unusable; it is that the rate becomes an average over the stream rather than a description of each part of it.
The same logic changes how we read characteristic premia. Researchers often compare small and large firms, high and low book-to-market firms, or firms with different credit quality. If those groups also have different cash-flow duration, an observed return difference may reflect where their cash flows sit in the term structure, not only a difference in a timeless firm trait. This does not prove that duration explains every value, size, or credit pattern. It does mean that a serious explanation must ask when the relevant cash flows arrive and which horizons carry the extra compensation.
The phrase “long-term equity” can also mislead. A longer horizon does not automatically mean a larger premium, a safer premium, or a premium that is insulated from recessions. It only tells us that the claim places more weight on distant outcomes. The required compensation for those outcomes is an empirical question, and it can vary with the economic state and with the kind of firm issuing the claim.
One subject, three lenses
The central subject remains the equity-premium term structure, but it can be examined through more than one lens. The first is the horizon itself: how does compensation change as the payment date moves outward? The second is composition: does the curve look different for the aggregate market, a portfolio, or a firm with particular size, value, or credit characteristics? The third is state: does the curve change when markets are buoyant, stressed, or in recession?
The two research approaches discussed in this series answer those questions from different information. The 2026 New York Fed study uses observed firm prices and projected firm-level cash flows over a long U.S. panel. It imposes enough common structure across firms to estimate maturity-specific discount rates, while using differences in the timing of expected cash flows to identify the shape of the curve. In that setting, the cross-section is not a nuisance around an average market result; it is the source of information about different horizons.
The 2009 ECB study takes a structural time-series route. It uses one pricing kernel—a rule that discounts uncertain payoffs—to price both government bonds and a broad stock index. Inflation, payout yield, and two latent real-rate factors describe the state of the system. Because the same no-arbitrage framework prices assets at different maturities, it produces a time-varying family of equity premia rather than a single unconditional average.
These routes can produce related insights without measuring exactly the same thing. The cross-sectional approach is designed to learn from differences among firms and to use a long sample for studying characteristics and business-cycle patterns. The affine approach is designed to make stock and bond pricing coherent at each date and to trace how current state variables change expected compensation across horizons. One emphasizes heterogeneous cash-flow clocks; the other emphasizes a common pricing system and changing economic conditions.
That distinction helps make sense of an important empirical tension. The 2026 study reports an upward-sloping aggregate curve, yet finds flatter or even downward-sloping curves among some value and speculative-grade firms. The 2009 model likewise allows the curve at a particular date to take different shapes as the state changes. An aggregate curve can therefore be positive without every firm, portfolio, or market episode sharing that shape. Aggregation is a statement about the weighted combination, not a guarantee that the components move in parallel.
A conventional dividend-discount estimate provides a third perspective. It uses expected dividends and one required return, so it is closer to a yield-to-maturity measure: a rate that prices the whole stream on average. That average can move closely with horizon-specific estimates, especially when the same broad information about expected cash flows and market conditions enters both calculations. Co-movement does not turn the average into a full curve, however. It shows that the measures may respond to common valuation information while retaining different meanings.
The essential discipline is to ask four questions before interpreting any quoted premium: What is the horizon? What risk-free rate is the benchmark? Is the estimate conditional on a firm, portfolio, or economic state? And is the number a realized outcome, a single-rate valuation summary, or a model-implied schedule of required compensation? Those questions keep the equity risk premium attached to the cash flows and risks it is meant to price.
The problem now has a concrete form. Once equity is understood as a stream of dated cash flows, a single rate is no longer the only possible description of its value. The next step is to examine how present-value logic turns those different dates into an estimable curve—and why the information needed to do that must come from somewhere beyond one firm’s price.