Portfolio and risk · Portfolio theory
Equity risk premium (ERP): the extra return stocks earn for their risk
What the equity risk premium is, three ways to estimate it, and SPY against Treasury bills from 1993 to 2026: 8.2% a year on average, and negative for a whole decade.
On this page
The equity risk premium is the return that stocks earn, or are expected to earn, above a safe investment such as Treasury bills or bonds. It is the price the market pays investors for accepting the possibility of large losses. If stocks offered no more than Treasuries, nobody would have a reason to hold them.
The premium is the most important number in long-term investing, and nobody can observe it. The past premium can be measured, but it varies enormously by period. The future premium, the one that matters for a retirement plan or a company valuation, has to be estimated. The beginner lesson Risk and return: volatility, drawdowns and the equity premium introduces the trade-off. This lesson measures the premium on 33 years of data, shows how wide the uncertainty around it is, and sets out the three ways it is estimated.
The idea
Lend money to the US Treasury for three months and the outcome is known in advance. Buy shares and it is not: the result over a year might be +30% or −30%. Investors, taken together, dislike that uncertainty. They will hold shares only at prices low enough to make the expected return higher than the safe rate. The gap is the premium.
Because it is compensation for risk, the premium cannot be collected reliably. If stocks beat bills every year, they would not be risky and would not need to offer more. The premium exists because there are long periods in which it fails to appear.
Each point on the line is a ten-year holding period. An investor who held SPY for the ten years to February 2009 earned 6.6% a year less than Treasury bills. One who held for the ten years to February 2019 earned 16.1% a year more. Both held the same fund for a decade.
How it works
The arithmetic premium is always the larger of the two, by roughly half the variance of stock returns. The arithmetic figure is the one used in formulas such as the Capital asset pricing model (CAPM): what return a beta should earn. The compound figure describes what an investor's wealth actually did.
A worked example: SPY against Treasury bills
The data: 404 monthly total returns of SPY, March 1993 to October 2026, and the 3-month Treasury bill rate for the same months.
- $1 in SPY with dividends reinvested grew to $31.27. That is a CAGR of 10.8%.
- $1 rolled over in Treasury bills grew to $2.31: a CAGR of 2.5%.
- Compound premium: 10.8% − 2.5% = 8.2% a year (the difference of the unrounded figures).
- The average monthly excess return, times twelve, gives the arithmetic premium: 8.9% a year.
SPY's returns are after its expenses; an investor's taxes are not included.
| Period | SPY a year | Treasury bills a year | Premium |
|---|---|---|---|
| March 1993 – 1999 | 21.6% | 4.7% | +16.8% |
| 2000 – 2009 | −1.0% | 2.7% | −3.7% |
| 2010 – 2019 | 13.4% | 0.6% | +12.9% |
| 2020 – October 2026 | 14.9% | 2.9% | +12.0% |
| Whole period | 10.8% | 2.5% | +8.2% |
The whole-period average of 8.2% was not experienced in any of the four sub-periods. Three were far above it and one far below.
How uncertain is the average?
An average taken from a volatile series is itself uncertain. The standard error of the arithmetic premium is the volatility of excess returns divided by the square root of the number of years: 14.7% ÷ √33.7 = 2.5%. Two standard errors either side of 8.9% gives a range of about 3.9% to 13.8%.
So after a third of a century, the data cannot distinguish a "true" premium of 4% from one of 13%. This is the central difficulty. To narrow the range to plus or minus 2 points would take more than 200 years of data, by which time the economy generating the returns would be a different one.
Against bonds
The premium is often measured against government bonds in place of bills, since bonds are the realistic alternative for long-term money. From September 2002 to October 2026, SPY returned 11.1% a year and the 7–10 year Treasury fund IEF 3.3%: a premium over bonds of 7.8%.
Using it: three ways to estimate the future premium
1. The historical average. Take a long record and assume the future resembles it. For US stocks since the 1920s, the commonly cited premium over bills is in the region of 6% to 8% arithmetic. Dimson, Marsh and Staunton's study of many countries since 1900 found lower figures for the world as a whole, in part because the US was one of the century's most successful markets and a sample of survivors overstates what investors could have expected.
2. The implied premium. Work backwards from today's prices. A simple version says the expected return on stocks is their dividend yield plus the expected growth of dividends:
With illustrative figures: a dividend yield of 1.5%, expected growth of 5.5% and a 10-year Treasury yield of 4.5% give an expected return of 7% and an implied premium of 2.5%. More complete versions add share buybacks to the yield and model growth in stages; Aswath Damodaran publishes a monthly estimate of this kind, which has mostly been between 4% and 6% since 2010. The implied premium is forward-looking and moves with prices: it is high after a crash, when stocks are cheap, and low after a long rise.
3. Surveys. Ask investors, finance officers or academics what they expect. Survey answers have mostly fallen between about 3% and 6% and tend to rise after markets have risen, which is the opposite of what valuations imply.
Fama and French made the case in 2002 that the realized premium of the late twentieth century was higher than investors could have expected in advance, because part of it came from valuations rising, which cannot repeat indefinitely. The same reasoning applies to the 12% premiums of 2010 to 2026.
What the number is used for
- Planning. A retirement projection that assumes stocks earn 8% above cash produces a very different answer from one that assumes 4%. Testing both is more honest than choosing one.
- Valuation. The premium is an input to the discount rate. A higher assumed premium lowers the value of every stock.
- Asset allocation. The case for holding stocks over bonds rests on the premium being positive and large enough to justify the drawdowns.
Where it fails or misleads
- The past premium is one draw. US investors since 1993 lived through a particular history. Other countries and other periods gave much lower figures, including premiums near zero over several decades.
- The start and end dates dominate. A window starting in March 2000 or ending in February 2009 gives a negative premium; one starting in March 2009 gives an enormous one.
- A high past return lowers the expected future one. When part of the realized premium came from rising valuations, the same valuations imply less ahead. Extrapolating the last fifteen years forwards is the most common error.
- "Risk-free" is not free of all risk. Treasury bills can lose purchasing power when inflation exceeds their yield, and bonds lose value when rates rise. See Inflation and real returns: what an investment earns after prices.
- It is a market average. Individual stocks, sectors and countries can have premiums far from the average, including none.
- Time does not guarantee it. A decade below Treasury bills has happened in this sample. Longer horizons improve the odds and do not remove the risk; see Time diversification: are stocks safer in the long run?.
Compared with its neighbours
| Premium | Compares | Typical size discussed |
|---|---|---|
| Equity risk premium | Stocks with Treasury bills or bonds | 3% to 8% a year |
| Term premium | Long-term bonds with short-term bills | 0% to 2% |
| Credit premium | Corporate bonds with Treasuries of the same term | 1% to 4% |
| Size and value premiums | Small with large stocks; cheap with expensive | Disputed; see Factor models: explaining a portfolio’s returns with several betas |
| Inflation risk premium | Nominal bonds with inflation-linked bonds | Small, variable |
How Finabase does it
Finabase does not publish an equity risk premium, historical or implied.
The ingredients for a rough implied estimate are on its pages. The Treasury page lists current yields from one month to thirty years, which is the risk-free side. An index fund's symbol page shows its dividend yield and price-to-earnings ratio; the reciprocal of the P/E is the earnings yield, a common shorthand for the expected real return on stocks. Earnings yield minus the 10-year Treasury yield is a crude premium that takes thirty seconds to compute.
For the historical side, a symbol's History tab provides monthly closing prices to export. Those are prices without dividends, so a premium calculated from them understates the total return by the dividend yield, around 1.5 to 2 points a year for SPY over this period.
Frequently asked questions
What is the equity risk premium today?
Nobody knows; it can only be estimated. Implied estimates from current prices have mostly been in the 4% to 6% range in recent years, lower than the premium realized over the previous fifteen.
Why is it measured against both bills and bonds?
Bills are the purest risk-free asset over a short period. Bonds are the realistic alternative over long periods. The premium over bonds is usually a point or so smaller, because bonds have historically earned more than bills.
Can the equity risk premium be negative?
The realized premium can be, for long periods: SPY trailed Treasury bills over 2000 to 2009. The expected premium should be positive, since investors would otherwise not hold stocks, though it may be small when prices are high.
What is the equity premium puzzle?
Mehra and Prescott's 1985 finding that the historical US premium was far larger than standard economic models of risk aversion could explain. Proposed explanations include rare disasters, investors' sensitivity to short-term losses, and plain good luck in the sample.
Does a long holding period guarantee I earn the premium?
No. It raises the probability. The ten years to February 2009 delivered a premium of −6.6% a year, and other countries have had negative premiums over twenty years or more.
Summary
The equity risk premium is what stocks earn over a safe asset in exchange for their risk. For SPY against Treasury bills from 1993 to 2026 it was 8.2% a year compounded, built from one decade of −3.7% and three periods above 12%. The statistical uncertainty around the average is about five points either way, so history alone cannot settle what to expect. Forward-looking estimates from current prices are lower than the recent realized figures. Plans built on the premium should test a range of values and allow for a decade in which it does not arrive.
Check your understanding
4 questions on this lesson. Each answer comes with the reason, and nothing is scored.
Over a period, stocks compounded at 10.8% a year and Treasury bills at 2.5%. What was the realized equity risk premium over bills?
- About 8.3% a year
- 10.8% a year
- 13.3% a year
Show the answer
A. About 8.3% a year The premium is the stock return minus the risk-free return: 10.8% − 2.5%. It is the reward for bearing risk, not the whole return.
From 2000 to 2009, SPY returned −1.0% a year and Treasury bills 2.7%. What does that decade show?
- The equity risk premium does not exist
- The premium is compensation for risk, and the risk includes a decade or more of earning less than the safe asset
- Treasury bills are riskier than stocks
Show the answer
B. The premium is compensation for risk, and the risk includes a decade or more of earning less than the safe asset If stocks beat bills in every period they would not be risky and would need no premium. Long stretches of shortfall are the risk being paid for.
About 34 years of data give an average arithmetic premium of 8.9% with a standard error of 2.5%. What can be said about the true long-run premium?
- It is 8.9%
- It plausibly lies anywhere from about 4% to 14%
- It must be above 8%
Show the answer
B. It plausibly lies anywhere from about 4% to 14% Two standard errors either side of the estimate span roughly 3.9% to 13.8%. Volatile returns make the average imprecise even over decades.
Stocks have risen sharply for fifteen years, partly because investors are paying more for each dollar of earnings. What does that imply for the expected premium from here?
- Higher, because recent returns were high
- Lower, because higher valuations mean lower yields and less room for the same gain to repeat
- Unchanged
Show the answer
B. Lower, because higher valuations mean lower yields and less room for the same gain to repeat The part of past returns that came from rising valuations cannot be extrapolated. Implied premiums fall as prices rise relative to earnings and dividends.
Sources and further reading
- Rajnish Mehra and Edward C. Prescott, “The Equity Premium: A Puzzle,” Journal of Monetary Economics, vol. 15, no. 2 (1985)
- Elroy Dimson, Paul Marsh and Mike Staunton, Triumph of the Optimists: 101 Years of Global Investment Returns (Princeton University Press, 2002)
- Eugene F. Fama and Kenneth R. French, “The Equity Premium,” Journal of Finance, vol. 57, no. 2 (2002)
- Aswath Damodaran, Equity Risk Premiums: Determinants, Estimation and Implications (annual update, NYU Stern)
- Board of Governors of the Federal Reserve System, 3-Month Treasury Bill Secondary Market Rate (TB3MS), via FRED
- Finabase chart data — monthly bars and dividends of SPY, March 1993 to October 2026, and of IEF from September 2002; statistics computed by Finabase
