How to Estimate the Equity Risk Premium for Your DCF

The equity risk premium is the most consequential variable in a discounted cash flow model. Not the revenue growth rate, not the terminal multiple, not even the operating margin assumption — the equity risk premium. It lives inside the discount rate, which appears in the denominator of every present value calculation, and because it's a denominator, errors don't add to your valuation. They multiply through it. A one-percentage-point mistake in the ERP shifts a fair value estimate by 15 to 25 percent, depending on a business's duration. On a long-duration compounder, the number is closer to 30. This isn't an error you can smooth out with a more careful revenue forecast.

And yet the ERP is the variable most investors treat as settled. They'll argue for an hour about whether a business deserves a 3% or 5% terminal growth assumption. They'll stress-test operating leverage across three scenarios. Then they'll drop 5.5% into the equity risk premium cell because the textbook said so, or because the Bloomberg terminal defaults to it — and carry that number forward without interrogation. The discount rate gets tested; the ERP embedded inside it does not.

I'll say upfront that I'm less certain about which estimation method to privilege than I'd like to be after years of thinking about this. The equity risk premium cannot be known with precision. What it can be is estimated — through three methods that differ in construction, differ in their weaknesses, and differ in what they require you to believe. The goal isn't to find the right number. It's to understand the range, own your assumptions, and stress-test the result — which is, ultimately, what all honest intrinsic value work asks of you.

What the Equity Risk Premium Does Inside a DCF

The equity risk premium is the additional return investors require for holding equities rather than risk-free government bonds. In the Capital Asset Pricing Model, it enters the discount rate as follows:

Cost of Equity = Risk-Free Rate + β × Equity Risk Premium

Beta scales the premium by a security's sensitivity to broad market swings. A business with a beta of 1.0 gets the full ERP added to the risk-free rate. One with beta of 0.7 — a stable utility or consumer staple with predictable cash flows — gets 70% of it. But the mechanism matters less here than the placement: the ERP sits inside the discount rate, which lives in the denominator of every present value calculation in your DCF model.

In most DCF analyses of durable, growing businesses, terminal value accounts for 60 to 80 percent of total estimated enterprise value. That means an ERP error propagates through the majority of your valuation — with the remainder exposed through your explicit forecast period as well. Spending more time calibrating the ERP than refining a year-four revenue projection is entirely rational. Most analysts do the opposite.

The Historical Method: Reliable Data, Contested Meaning

The most intuitive approach is historical: calculate the average annual excess return of equities over risk-free bonds across a long sample, and use that spread as your forward assumption. The logic is that history at minimum establishes what equity investors have been paid for bearing equity risk. The data is extensive and precise.

Aswath Damodaran at NYU maintains the most widely cited historical series. Over the full period from 1928 to 2025, U.S. equities earned an arithmetic average of approximately 6.1% above 10-year Treasuries annually. The geometric average — compounded annual excess return, which better reflects what a buy-and-hold investor actually experienced — is closer to 4.6%. This gap matters. Arithmetic averaging overstates compounded returns because variance drag is excluded; geometric averaging can understate forward expected returns because it reflects starting and ending valuation levels. Neither is unambiguously correct for your DCF.

Time-period choice amplifies the uncertainty. Starting in 1926 captures the Depression and the post-war expansion. Starting in 1962 gives you a very different equity cycle. Using only the last 30 years includes the 1990s bull run and the 2010s zero-rate expansion. The historical ERP is not a stable number — it's a time-weighted average of a specific era's risk appetite and realized outcomes.

Survivorship bias sits underneath all of it. The U.S. market has had an exceptional run by any cross-country comparison. Investors who applied a 6% ERP derived from American history to, say, an Austrian or Argentine equity market in 1920 would have been using a number with almost no predictive relevance for those economies. The U.S. historical ERP may partly reflect the accident of American economic dominance across the 20th century — a premise that cannot simply be assumed forward.

The Implied Equity Risk Premium: Reading the Market Backward

Rather than asking what equities have returned historically, the implied approach asks what the current market price requires going forward. It's the logic of a reverse DCF applied not to a single stock but to the entire market.

The mechanics: take the current level of the S&P 500, add expected dividends and buybacks over the coming year, project cash flows forward using consensus near-term growth rates and a steady-state terminal rate, then solve for the discount rate that makes the present value of those flows equal to today's index price. Subtract the current 10-year Treasury yield and you have the implied equity risk premium — the additional return the aggregate market currently requires to justify its price.

Damodaran publishes this calculation each January. In his January 2026 analysis, he put the implied ERP at approximately 4.23% above the 10-year Treasury — notably lower than both the long-run arithmetic historical average and most survey estimates. That figure reflects U.S. equities trading near a Shiller CAPE around 39 at year-end 2025: at elevated valuations relative to normalized earnings, the market embeds a compressed premium over the risk-free rate. Either stocks are fairly priced at 4.23% above Treasuries, or the premium is thin and equities are expensive. The implied ERP doesn't resolve that — it tells you what the market currently requires.

The practical implication for your DCF: if you assume a 6% ERP and the market-implied rate is 4.23%, you are applying a discount rate above what the market embeds in current prices. Every stock you evaluate will tend to appear cheap relative to market pricing — not because you found genuine margin of safety, but because you're demanding more compensation than the market currently charges. That divergence is worth knowing explicitly. It doesn't mean adopting the market's implied ERP uncritically; it means naming the assumption and owning the choice.

Survey Estimates and the Limits of Consensus

A third method asks what practitioners actually use. The Graham-Harvey survey, conducted annually through Duke University, has asked CFOs to estimate the expected 10-year excess return of U.S. equities over Treasuries since 2000. Estimates have historically centered around 3.5 to 5.5%, with meaningful variance across regimes — higher in post-crisis environments, lower during expansion peaks.

Survey estimates have one thing going for them: they measure revealed assumptions rather than historical realizations. If most practitioners are using a 5% ERP, that number is actually driving real capital allocation in real companies. That's relevant context, especially if your analysis depends on how market participants collectively value assets.

The structural problem is anchoring. Survey respondents revise toward recent experience — strong equity markets pull estimates up, downturns push them down. This is precisely backwards from what rational updating implies: high past returns mean prices are higher relative to fundamentals, which suggests lower forward expected returns, not higher ones. The survey is a measure of sentiment as much as a forecast of forward risk. This is where I want to be explicit about the limit of my confidence: survey-based estimates are useful context, but they shouldn't anchor your DCF assumption. Use them as a cross-check.

Putting a Number in Your Model

In practice, most serious analysts working on U.S. equities in 2026 use a range of 4.5% to 6.5%, depending on their orientation toward historical versus implied evidence. The implied ERP at 4.23% grounds the low end in current market prices. The long-run arithmetic historical average near 6.1% gives the high end empirical support. Where you sit in that range is a judgment call about how much weight you put on forward-looking versus backward-looking evidence — and on how you read current valuations.

To see what this means concretely, consider Microsoft (MSFT). In its fiscal year 2023 10-K, Microsoft reported free cash flow of approximately $59 billion, having grown at a mid-teens annual rate over the preceding five years. A base-case DCF might assume 10% FCF growth for 10 years, then 3% terminal growth, with a beta of 0.9. With the 10-year Treasury at 4.5%:

• ERP of 4.5% → cost of equity ≈ 8.6% → implied fair value (simplified model): roughly $385 per share
• ERP of 5.5% → cost of equity ≈ 9.5% → implied fair value: roughly $315 per share
• ERP of 6.5% → cost of equity ≈ 10.4% → implied fair value: roughly $265 per share

Same business. Same cash flow projections. Same terminal growth assumption. A 2-percentage-point range in the ERP produces roughly a $120-per-share spread — about 30% of the middle estimate. That range isn't a modeling failure. It's the honest uncertainty band embedded in any DCF where the ERP is genuinely contested.

The right response isn't to narrow that band by picking a single ERP and running with it. It's to understand which end of the range you're using and why. As the fair value calculation post covers, the ERP is one of the inputs where precision is unavailable and a range is more honest than a point estimate. Treat it the same way you treat the terminal growth rate: name it, vary it, and report the spread. Fair value is a compass, not a GPS coordinate — and the ERP is one of the main reasons why.

Key Takeaways

• The equity risk premium lives inside the discount rate and propagates through your entire DCF — most heavily through terminal value, which typically represents 60–80% of total enterprise value. A 1% ERP change shifts a typical fair value estimate by 15 to 25%.
• Three methods give different answers: the long-run U.S. historical arithmetic average (≈6.1%), Damodaran's January 2026 implied ERP from market prices (≈4.23%), and survey-based estimates (typically 3.5–5.5%). No single method is definitively correct.
• The market-implied ERP is a useful cross-check: if your assumed ERP is materially higher than the market-implied rate, you are demanding more risk compensation than the market currently prices in — a divergence worth naming explicitly rather than discovering by accident.
• Run your DCF at three ERP levels (e.g., 4.5%, 5.5%, 6.5%) and report the resulting range. The spread is the honest version of the fair value estimate.