FINANCEOngoing practice92% confidence

The Equity Risk Premium as a Variable

The equity premium is not pi — it is not a constant and cannot be looked up

Problem it solves

Anchoring to a single historical equity premium number as if it were a physical constant

Best for

Institutional investors, financial regulators, actuaries, and anyone required to state a forward-looking equity premium for valuation or regulatory purposes

Not ideal for

Short-horizon speculators or investors making binary buy/sell decisions where a precise premium estimate is irrelevant

Overview

Why this framework exists

The equity risk premium — the expected excess return of equities over risk-free bonds — is the single most important number in institutional finance. It is used to set regulated utility returns, calibrate pension fund assumptions, and justify target asset allocations. But it behaves nothing like a mathematical constant: it varies over time, across countries, and in response to structural changes in interest rates, market valuations, and economic regimes.

Staunton frames the problem with a memorable analogy: when financial mathematicians want the equity premium, they treat it like pi — expecting a precise decimal they can look up and apply. But pi is fixed by geometry; the equity premium is set by market participants' collective forward-looking expectations, which shift continuously. Over the long run, the DMS data shows that realized excess returns of equities over bonds vary enormously across rolling periods — sometimes strongly positive, sometimes negative — and that only very long time windows (125+ years, 35+ countries) produce estimates that are worth anchoring to at all.

The practical implication is that the equity premium is a starting point for judgment, not a lookup table. When it is used as a regulatory input — setting what a water company or railway is 'permitted' to earn for shareholders — the precision implied by citing '5.2%' or '4.8%' is false. The correct posture is to anchor to the best available long-run global estimate, then explicitly adjust for current structural conditions (interest rate level, valuation multiples, regime shifts) with documented reasoning.

Core principles

5 total
  1. The equity risk premium is a forward-looking expectation, not a backward-looking average that can simply be read off historical data.
  2. Realized returns over any finite period are a noisy signal of the true premium — you need 125+ years across 35+ countries to get a credible estimate.
  3. Structural changes (interest rate regimes, market structure, economic institutions) can shift the premium level across historical eras.
  4. Any use of the equity premium for regulatory or actuarial purposes should be treated as an estimate with wide error bars, not a precise input.
  5. History is the best starting point for calibrating the premium, but it must be adjusted for current structural conditions before application.

Steps

5 steps
  1. Establish the longest credible baseline
    Use the broadest available historical data — ideally 100+ years across multiple countries — as the starting point for any equity premium estimate. Single-country or sub-50-year windows are unreliable because they capture only one or two interest-rate cycles and exclude market-failure events.
    Pro tipThe DMS global database (real equity return ~5.6% annualized for the world index) is the most survivorship-corrected long-run baseline publicly available.
    WarningUS-only data overstates the premium because the US was the 20th century's best-performing major market — a fact knowable only in retrospect.
  2. Identify the current structural regime
    Ask whether present conditions resemble the average conditions of the historical period. If interest rates have spent 40 years declining from 16% to near-zero, the bond return component of the realized premium is inflated and will not repeat. If valuations are at historical extremes, expected forward returns are suppressed.
    Pro tipStaunton's explicit forward-looking statement: investors in the early 21st century should anchor on equity premium numbers lower than the 20th-century average.
    WarningRegulators often resist downward adjustments to the premium because a lower premium means lower allowable returns for the utilities they oversee — a political economy pressure that distorts the estimate.
  3. State the estimate as a range, not a point
    Given the statistical noise in any premium estimate, present it as a range with explicit assumptions about the historical baseline used, survivorship corrections applied, and structural adjustments made. A point estimate of '4.8%' implies precision that does not exist.
    Pro tipDocument the adjustments explicitly — this is the 'log non-obvious decisions' discipline applied to financial modeling.
  4. Recalibrate when structural conditions change
    The premium is not static across decades. The 19th-century US data suggests equities did not earn a large premium over the bonds available at the time; the 20th century was exceptional; early 21st-century evidence suggests a return to lower premia. Each regime shift requires re-estimation, not extrapolation.
    WarningRegulatory and actuarial lock-in creates inertia — once a number is in a regulatory formula, it persists even when market conditions have made it stale.
  5. Separate the premium from the level of absolute returns
    The equity risk premium is the excess of equities over risk-free bonds. When bond yields are 1%, a 4% premium implies 5% equity returns. When bond yields are 6%, a 4% premium implies 10% equity returns. Changes in the risk-free rate affect absolute return forecasts independently of any change in the premium itself.
    Pro tipMany practitioners conflate the two and anchor on a historical nominal equity return without adjusting for the current risk-free rate level.

Checklist

Saved in your browser

Examples

4 cases
Regulated utility rate-setting

Water companies, railway operators, and electricity distributors in multiple countries have their allowable shareholder returns set by regulators who cite the DMS equity premium estimate. The precision implied — a specific annualized percentage embedded in a regulatory formula — contrasts sharply with the wide confidence intervals Staunton and Marsh acknowledge in their own data.

OutcomeRegulatory decisions worth billions of pounds annually rest on a number that the researchers who produced it describe as a starting point for judgment, not a lookup value. When the premium is mis-set, consumers pay too much or investors earn too little — with no mechanism to correct the error until the next regulatory review.
Australia's high equity premium misuse

Australia had one of the highest long-run equity returns in the DMS database. Australian regulators became accustomed to using Australia-specific numbers for setting allowable returns. When challenged by international comparisons, defenders of the high number pointed to Australia's historically superior performance as justification.

OutcomeStaunton explicitly warns against this — individual-country data has high variance and should be subordinated to the world-index estimate. Australia's high return is partly real and partly a product of its exceptional economic circumstances; it is not a reliable guide to future allowable returns.
The 19th-century US equity premium revision

Extending US data back to 1800 — before there was a unified US government bond market — shows that the equity premium was much smaller than the 20th century suggests. Some of the apparent premium in 19th-century data reflects the fact that corporate bonds (not risk-free government bonds) were used as the bond comparator, meaning the equity premium was being measured against a risky baseline.

OutcomeCorrecting for this suggests the 20th century was an exceptional era for equity outperformance, not a permanent structural feature of capitalism — supporting the forward-looking view that 21st-century returns will be lower.
Norway's sovereign wealth fund equity allocation decision

The Norwegian Government Pension Fund Global used the early DMS research (before Triumph of the Optimists was published) to justify moving from cash to bonds to equities over several years. The equity allocation decision was grounded in the long-run premium evidence, not short-term return expectations.

OutcomeNorway now holds approximately 1.5% of every listed company in the world and has generated long-run returns broadly consistent with the DMS global equity premium estimate — demonstrating that the framework can inform trillion-dollar institutional decisions when applied with appropriate humility about uncertainty.

Common mistakes

5 traps
Treating the equity premium like a physical constant
Extracting a single number from historical data and embedding it in regulatory or actuarial models as a fixed parameter ignores both the measurement uncertainty and the structural variation of the premium across time and countries.
Using a short window that coincides with a structural tailwind
UK bond investors who measured returns from 1980–2020 captured the entire decline from 16% interest rates to near-zero. That one-directional tailwind inflates the realized bond return, which depresses the measured equity premium — and cannot repeat. Using this window biases any forward-looking estimate.
Cherry-picking the country with the highest realized premium
Australia's very high long-run equity premium is cited by Australian regulators to justify high allowable returns. But it reflects Australia's unusually favorable economic history — using it as a global baseline is selection bias.
Ignoring the 19th-century evidence
US and UK data back to 1800 shows that equities did not earn a large premium over the bonds available in the 19th century. Including this evidence implies that the 20th century's equity premium was unusually high, not the norm — and that reversion toward lower premia is historically plausible.
Confusing precision with accuracy
Quoting the equity premium to two decimal places — '5.24%' — implies accuracy that the underlying data cannot support. Wide confidence intervals around any historical estimate reflect irreducible uncertainty, not inadequate data collection.

Origin story

How this framework came to be

The DMS team arrived at this framework through the regulatory use of their data. Governments and utility regulators around the world cite the Dimson-Marsh-Staunton equity premium estimates when setting allowable rates of return for infrastructure businesses. Staunton became concerned that users were treating a range of plausible estimates — itself derived from noisy multi-country data — as a fixed constant, extracting a single number and enshrining it in regulatory documents. The pi analogy emerged from his observation that regulators wanted the same kind of certainty from the equity premium that a physicist gets from a physical constant — a certainty the data simply cannot provide.

Source

Traced to primary
Source · PODCAST
The Golden Age of Returns is Over
Mike Staunton & Paul Marsh · 2025
Open source →

Related frameworks

Browse all Finance →