Guide

Equity risk premium explained

Harbor Capital ran a 70/30 U.S. stock-bond strategic allocation for a university endowment using a 7.0% historical equity risk premium baked into every mean-variance forecast. When the committee stress-tested a 20% equity drawdown against their spending rule, the model still projected 6.8% nominal equity returns over the next decade — because the ERP assumption had not been updated since 2012. Re-estimating with a forward-looking implied premium of 4.5% (risk-free near 4.2%, S&P 500 earnings yield ~8.7%) cut expected equity returns by 2.1 percentage points annually and justified shifting to 60/40 without abandoning long-term equity exposure. The equity risk premium (ERP) is the extra return investors expect from stocks over a risk-free benchmark. It sits at the center of CAPM, discounted cash flow valuation, glide-path design, and the stock-versus-bond decision in asset allocation. Yet practitioners routinely disagree on whether to use historical averages, survey forecasts, or market-implied premiums — and a one-point ERP change moves every required return and fair-value estimate. This guide covers definitions, historical vs forward methods, arithmetic vs geometric debates, implied ERP from prices, international spreads, a Harbor Capital worked example, a method decision table, common pitfalls, and an investor checklist.

What the equity risk premium measures

The ERP answers a simple question: how much extra return do I need to hold stocks instead of Treasuries? Formally, it is the expected return on a broad equity market index minus the expected return on a risk-free asset (usually short-term government bills or the 10-year Treasury yield, depending on horizon).

ERP = E(R_equity) − R_f

In CAPM, expected return on any stock is R_f + β × ERP. A β of 1.0 earns the market ERP; higher-beta stocks demand more. In valuation, the ERP feeds the cost of equity in a DCF: a higher ERP means a higher discount rate and a lower fair value for the same cash flows. In asset allocation, a higher ERP tilts mean-variance optimizers toward equities; a lower ERP favors bonds and alternatives. The ERP is not observable — you estimate it from history, surveys, or current market prices.

Historical equity risk premium

The most common starting point is the realized spread between U.S. large-cap stocks and Treasury bills over a long window. Ibbotson/Morningstar data for 1926–2024 shows roughly:

Arithmetic average: ~7–8% per year (simple average of annual spreads)
Geometric average: ~5–6% per year (compound growth difference)

The gap matters. Arithmetic returns are what you use in forward one-period models like CAPM; geometric returns are what long-term investors actually compound. Using a 7% arithmetic historical ERP in a 30-year retirement projection overstates wealth by a wide margin compared to a 5.5% geometric estimate. Historical ERP also varies by country, period start date, and whether you measure against bills or bonds. Japan’s post-1990 experience shows that a century of U.S. data does not guarantee future premiums elsewhere.

Why historical ERP can mislead

Survivorship bias. U.S. markets survived and thrived; failed markets are excluded from “history.”
Starting valuation. High starting P/E compresses subsequent returns even if the long-run ERP is stable.
Regime change. Lower inflation, buyback-driven EPS growth, and demographic shifts may not repeat.
Arithmetic abuse. Quoting 7% historical ERP for a 40-year glide path without geometric adjustment inflates expectations.

Forward-looking and survey ERP

Forward ERP asks what premium is reasonable today, not what happened over the last century. Common sources:

Analyst surveys (e.g., Duke CFO Survey, consensus equity strategists) — typically 4–6% for U.S. equities in normal environments.
Academic supply-side models — link ERP to productivity, payout ratios, and GDP growth.
Building-block forecasts — decompose expected equity return into dividend yield + buyback yield + real EPS growth + inflation.

A building-block example for the S&P 500 in mid-2026: dividend yield ~1.3%, net buyback yield ~1.5%, real EPS growth ~2.0%, inflation expectation ~2.5% → nominal expected return ~7.3%. With a 4.2% risk-free rate, implied ERP ≈ 3.1%. That is far below the historical arithmetic average — and that gap is the central debate in strategic asset allocation today.

Implied ERP from market prices

Implied ERP (popularized by Aswath Damodaran and others) backs out the premium that makes a DCF equal to the current market price. Steps at index level:

Start with trailing or forward index earnings and a cash-flow growth assumption.
Discount cash flows at R_f + ERP and solve for ERP that matches today’s index level.
Alternatively, use earnings yield: E/P ≈ R_f + ERP in a simplified Gordon model.

When P/E ratios are elevated, implied ERP falls — the market is already pricing in lower future returns. When stocks crash and earnings hold, implied ERP rises — equities look cheaper in expected-return terms. Implied ERP is market-consistent but sensitive to growth and payout assumptions; a 0.5% change in long-term EPS growth can swing implied ERP by 1–2 points.

ERP, the term premium, and bond yields

Do not confuse the equity risk premium with the term premium (extra yield on long bonds over short bills). Both affect stock-bond allocation:

Rising term premium: long bonds offer more yield; the opportunity cost of holding equities may increase.
Compressed ERP + high R_f: cash and T-bills compete with stocks for the first time in years — the “TINA” (there is no alternative) era ends.
Negative correlation regimes: when stocks and bonds fall together (2022), both ERP and term-premium assumptions need stress-testing alongside portfolio stress tests.

International and regional ERP

ERP is not universal. Emerging markets often carry higher ERP estimates (6–8%+) for political, currency, and liquidity risk. Developed ex-U.S. markets sit between U.S. and EM ranges. When Harbor Capital added a 20% international equity sleeve, they applied a +0.5% ERP uplift over U.S. for currency risk rather than using identical premiums — a common institutional practice. Home bias often hides the fact that domestic historical ERP may not apply to foreign holdings.

Harbor Capital worked example

Context: $240M endowment, 5% annual spending rule, 70/30 U.S. stocks/intermediate Treasuries, 10-year planning horizon.

Problem: The investment policy statement cited a 7.0% historical arithmetic ERP from 1926 data. Combined with a 4.2% risk-free rate, CAPM implied 11.2% expected return on the equity sleeve (β ≈ 1.0). Monte Carlo paths showed only a 12% probability of spending-rule breach — reassuring but inconsistent with current valuations (S&P 500 forward P/E ~19x).

Re-estimation: The team triangulated three methods:

Building-block: 7.3% nominal equity return − 4.2% R_f = 3.1% ERP
Implied (Damodaran-style index DCF): 4.5% ERP
Survey midpoint (internal CIO poll): 4.8% ERP

They adopted 4.5% forward ERP for planning, cutting expected equity return from 11.2% to 8.7%. Breach probability under the spending rule rose from 12% to 28% at 70/30. Re-optimizing with Sharpe-aware constraints and correlation to bonds produced a 60/40 target with modest alternatives sleeve. The committee kept equities as the growth engine but stopped relying on century-old arithmetic premiums to justify concentration risk.

ERP estimation method decision table

Method	Typical range (U.S.)	Best for	Main weakness
Historical arithmetic	7–8%	Academic baselines, long-horizon CAPM teaching	Overstates multi-period compounding; ignores current valuations
Historical geometric	5–6%	Long-term wealth projections, endowment education	Backward-looking; slow to reflect regime shifts
Building-block forecast	3–5%	Strategic asset allocation, IPS updates	Sensitive to growth and payout assumptions
Implied (market prices)	3–5%	Valuation consistency, index fair value	Circular if growth inputs are arbitrary
Survey consensus	4–6%	Committee governance, blending disparate views	Herding; strategists often anchor to recent returns

When to use which ERP

Strategic asset allocation (10+ years): blend forward building-block and implied ERP; use geometric historical as a sanity bound, not the anchor.
DCF stock valuation: implied or sector-adjusted ERP; document growth and terminal assumptions explicitly.
CAPM cost of equity: match ERP horizon to cash-flow duration; state R_f choice (bill vs bond).
Retirement planning tools: prefer geometric or forward ERP; show sensitivity bands (±1%) rather than a single point.
Crisis rebalancing: implied ERP often spikes after drawdowns — check whether the market is offering a higher premium before de-risking at the bottom.

Common pitfalls

Using arithmetic historical ERP for multi-decade compounding. Compounding follows geometric logic; arithmetic overstates terminal wealth.
Ignoring starting valuations. High P/E environments historically deliver lower subsequent 10-year returns even with a stable long-run ERP.
Mixing R_f definitions. Bill yield, 10-year Treasury, and inflation-linked yields produce different ERPs from the same equity forecast.
Applying U.S. ERP to emerging markets. Add country-risk and currency premia or use local implied estimates.
Treating ERP as constant. Premiums compress in bubbles and expand in panics; static IPS assumptions go stale.
Double-counting risk. If you use a high ERP and haircut cash flows in DCF, you may be too conservative.
Confusing realized and expected premium. A bad stock year does not automatically mean ERP was “wrong” — realized spreads are noisy.

Investor checklist

State whether your ERP is historical, forward, or implied — and which R_f benchmark you pair with it.
Run asset allocation and retirement projections at ERP − 1%, base, and ERP + 1% sensitivity bands.
Use geometric historical ERP as a long-run bound, not the primary planning input.
Update forward ERP at least annually or when index P/E moves more than ~15% from your last assumption.
Document ERP in your investment policy statement alongside expected return and rebalancing triggers.
For international sleeves, apply explicit country or currency risk premia rather than cloning U.S. ERP.
Cross-check DCF valuations: if implied ERP is below 3%, verify growth and terminal value assumptions.
Stress-test stock-bond portfolios when both ERP is low and bond term premium is rising.
Compare your ERP to building-block components (yield + growth) for internal consistency.
Revisit ERP after major market dislocations — implied premium may justify adding risk after crashes.

Key takeaways

ERP is expected extra stock return over risk-free. It drives CAPM, DCF discount rates, and stock-bond mix.
Historical U.S. arithmetic ERP (~7%) overstates forward expectations when valuations are elevated and rates are higher.
Triangulate methods: building-block, implied, and survey ERP rarely agree — blend with explicit assumptions.
Arithmetic vs geometric matters for compounding; use geometric or forward estimates for long horizons.
A one-point ERP change moves every fair value and allocation target — document and sensitivity-test it.