The Risk-Free Rate: Theory and Practice

Reading-order note. This article is numbered #5 and drafted first, because it supplies an input the rest of the pillar consumes: the equity risk premium is defined relative to it, CAPM begins with it, and every discounted valuation in Pillars 25 and 27 starts from it. A component cannot be introduced after the model that uses it. All figures in this pillar come from the hub's canonical parameter set and are illustrative teaching values, not market data.

Subtract it from anything and you get the compensation for risk; add a premium to it and you get what a risky thing should return; discount by it plus a premium and you get what a future cash flow is worth today. It is the origin point of the coordinate system. And it does not exist. What exists is a set of reasonable approximations, each of which fails in a specific, knowable way — and knowing how each one fails is more useful than the number itself.

What the concept is actually asking for

In theory the risk-free rate is the return on an asset whose payoff is certain over your holding period. Unpack "certain" and it demands three things at once. No default risk — the issuer pays, with certainty rather than high probability. No reinvestment risk — the return is locked for the whole horizon, so you are not exposed to what rates do in the meantime. And a horizon that matches yours — an asset can only be risk-free relative to a period. That third condition is the one most often dropped, and it is the reason the question "what is the risk-free rate?" has no single answer: it is not a property of an asset, it is a property of an asset paired with a holding period. Conceptually it is also the price of pure time — what someone must be paid to defer consumption when nothing can go wrong. Everything above it, as the risk-and-return trade-off establishes, is payment for bearing something.

What practitioners actually use, and why

The standard proxy is short-dated government debt issued by a sovereign in its own currency — Treasury bills and their equivalents, covered mechanically in Treasury Bills and Short-Term Government Paper. The reasoning is specific rather than patriotic. A government borrowing in a currency it issues faces no involuntary default constraint in the way a company does; these markets are among the deepest and most liquid in existence, so the quoted rate reflects a real transactable price rather than a thin one; and short maturities minimise the price sensitivity that would otherwise make the "safe" asset move around. For longer-horizon work, practice diverges, and the divergence is not a mistake. Valuing a company's cash flows decades out against a three-month rate mismatches the horizon badly, so long-dated government bonds are commonly used instead — accepting some interest-rate risk in exchange for a horizon that fits. Both conventions are defensible and they give different answers, which is the first honest thing to say about a number often quoted as though it were observed rather than chosen. This pillar uses 4.0% throughout, per the hub. It is a teaching value.

The four ways the proxy fails

1. Inflation makes nominal certainty a false certainty. A bill returning 4.0% returns exactly 4.0% of currency. What it returns in purchasing power is unknown when you buy it, because inflation over the period is unknown. With 2.5% inflation, the real return is not 1.5% but 1.46% — the ratio, not the difference, since (1.04 ÷ 1.025) − 1 = 1.4634%. The gap between the two is small here and large when inflation is high, which is why the subtraction shortcut is a habit worth breaking. The asset is nominally risk-free and really risky, and for an investor who eats, the second is the one that matters. Inflation-linked government bonds address this directly and introduce their own complications; Inflation and Purchasing Power covers the underlying mechanism.

2. Rolling short bills is not risk-free over a long horizon. A three-month bill is close to risk-free for three months. Held as a strategy for ten years it is a sequence of unknown future rates, and the uncertainty is reinvestment risk rather than default risk — but it is uncertainty, and it can be large. On the illustration below it costs nearly $2,900 on a $10,000 stake. The safe asset was safe; the plan was not.

3. "Risk-free" is currency-specific and jurisdiction-specific. The proxy works because a sovereign issues the currency it borrows in — which means a euro-based investor using a dollar-denominated bill has swapped default risk for currency risk, and has not reduced total risk at all. It also means the argument weakens for governments borrowing in currencies they do not control, and it does not make sovereign default hypothetical: it has happened, including within the modern era and within developed markets. The correct formulation is "the risk-free rate for a given currency and horizon", never "the risk-free rate" unqualified. Ratings and their limits sit with Market Regulators: The Referees of Finance and the credit-ratings material in Pillar 7.

4. The floor intuition has already failed once. Many readers hold an unstated assumption that a safe rate cannot be negative, since holding cash is always an option. In practice several developed-market government bonds traded at negative yields for extended periods in the last decade, meaning holders accepted a small certain loss rather than the risks of the alternatives — because cash at scale has storage, security, and counterparty costs of its own. Whatever else the risk-free rate is, it is not a law of nature with a floor at zero, and any model that assumes otherwise inherits an assumption the evidence has already contradicted. Central-bank policy is the dominant influence on the short end, treated in QE, QT and Central-Bank Balance Sheets: The Biggest Tool Nobody Can See.

Why a number this soft still matters enormously

Because everything is priced off it, small changes in it move large valuations. Under CAPM — built in this pillar's article on expected return and beta — a market-beta asset expects the risk-free rate plus the full equity risk premium: 4.0% + 5.0% = 9.0% on canonical inputs. Raise the risk-free rate by a single percentage point and that becomes 10.0%, and every future cash flow is discounted harder. The magnitude is easy to underestimate: a cash flow stream of $100 growing at 2% is worth $1,428.57 discounted at 9%, and $1,250.00 at 10% — a 12.5% fall in value from a one-point change in the risk-free input, with nothing about the business having changed. This is the single most important practical consequence of the concept, and it explains a great deal of market behaviour that otherwise looks irrational: when the safe rate moves, everything reprices, because the safe rate is the thing everything is measured against. It also means a valuation is only as defensible as its stated risk-free assumption — which is why serious analysis names the rate, the currency, the maturity, and the date, and why a discounted valuation quoting no risk-free rate at all is not really showing its work.