When the Equity Risk Premium Turns Thin Against Bonds

2.10%. That was Aswath Damodaran's estimate of the implied equity risk premium on the S&P 500 at the end of 1999, set against a Treasury bond rate of roughly 6.5%. In his paper on estimating equity risk premiums from that period, he described the end-1999 premium as a historical low — the bottom of a slide that had run since the early 1980s. His historical table, revised since, now shows 2.05% for that year-end. Either way, stocks were priced to deliver about two percentage points a year more than a government bond you could hold to maturity.

Two points. For owning the whole risk of the equity market — recessions, earnings collapses, bear markets — instead of a coupon from the Treasury. What followed is well known: the S&P 500 returned an annualized -0.95% for the ten years from January 2000 through December 2009, turning $10,000 into $9,089 over the decade, per a published review of that period's market returns. The lost decade.

I keep coming back to that number because the same question is live again. When the gap between what stocks are expected to earn and what bonds pay narrows sharply, what has it historically told us? My answer, after watching a few of these regimes: it tells you a lot about the likely range of forward returns and almost nothing about when the market turns. What follows is a case study in three episodes — 1999, 2023, and the split reading we have in 2026 — plus the measurement choices that decide whether the premium looks thin at all.

Three Gauges, Three Different Answers

There is no single equity risk premium you can look up. There are several estimates, and they disagree most at exactly the moments when the question matters. The three worth knowing:

Earnings-yield gap (the "Fed model") = S&P 500 earnings ÷ S&P 500 price − 10-year Treasury yield (nominal)

Real yield gap = 1 ÷ Shiller CAPE − 10-year TIPS yield (real)

Implied equity risk premium = the internal rate of return that equates today's index level with expected dividends and buybacks − 10-year Treasury yield

The first is the crudest and the most quoted. Its flaw is a units problem. An earnings yield is closer to a real return — corporate earnings tend to rise with inflation over time — while the Treasury yield is nominal. Comparing the two directly flatters bonds when inflation is high and flatters stocks when it is low. So the Fed model makes equities look cheap in deflationary scares and expensive in inflationary ones, and much of that is arithmetic, not insight.

The second gauge fixes the units. One divided by the cyclically adjusted P/E gives a real earnings yield based on ten years of inflation-adjusted earnings, and the TIPS yield is a real bond return. Real against real. The cost is that CAPE moves slowly and carries every quirk of a ten-year earnings average, which I covered in a calmer look at Shiller P/E and future returns.

The third is the one Damodaran has published at the start of every month since September 2008, with annual estimates going back to 1960. It backs an expected return out of current prices and projected cash returned to shareholders, then subtracts the Treasury rate. It is forward-looking, and it counts buybacks, which the first two gauges ignore. It also depends on a growth assumption, which is where honest people disagree. If you're building a single-company valuation rather than reading the market as a whole, James Whitfield's guide to estimating the equity risk premium for a DCF covers that side of the problem; here the interest is the aggregate number and what it has meant.

What Happened the Last Time the Premium Nearly Vanished: 1999

The historical analogue everyone reaches for is the right one here, for once, because every gauge agreed. The S&P 500 ended 1999 at 1,469.25. Robert Shiller's CAPE reached 44.19 in December 1999, the highest reading in his data going back to 1881, against a long-run average that sources put at around 17. Invert it and the real earnings yield was about 2.26% (1 ÷ 44.19). The implied premium sat at its low. And the 10-year Treasury, by Damodaran's figure, paid around 6.5% in nominal terms.

Look at that last pairing again. An expected equity return of roughly 8.6% (the 6.5% Treasury rate plus the 2.1% premium) against a bond paying 6.5% with no default risk and no earnings risk. The stock market wasn't promising a terrible return in absolute terms. It was promising a terrible return relative to the alternative. That distinction is the whole point of measuring against bonds.

Two lessons from that episode have held up for me.

• The thin premium was a statement about the distribution of returns, not the date of the top. Damodaran's own account has the premium trending lower since the early 1980s. An investor who treated "thin" as a sell signal in the middle of that slide spent years being wrong before being right.
• The adjustment came mostly through prices, not bond yields. The premium rebuilt itself because equities went nowhere for a decade while earnings and dividends kept accruing — the -0.95% annualized return over 2000–2009 was the mechanism by which the gap reopened.

That second point is the valuation channel in its plainest form. When the discount rate investors demand rises relative to what the index offers, either the cash flows grow into the price or the price falls to meet the cash flows. In 2000–2009 it was mostly the latter, with two bear markets doing the work. I walk through the general version of this link in when the connection between interest rates and equity multiples breaks.

2023: When Bonds Became Competitive Again

For most of the 2010s the comparison wasn't close. Treasury yields were low enough that almost any equity yield cleared them, and "there is no alternative" became a slogan. That regime ended in 2022 and 2023. In the first week of October 2023, the nominal 10-year Treasury yield reached 4.88%, its highest level since 2007, according to a BNP Paribas Asset Management market note from that month. Real yields on TIPS peaked the same month, according to the Tipswatch analysis of Treasury data published in April 2024, though the exact peak level differs by data series, so I won't pin a number on it.

Here's what surprised me. The implied premium didn't collapse toward 1999 levels. Damodaran's start-of-2024 estimate, in his Data Update 2 for 2024, was an expected return on stocks of 8.48% and an equity risk premium of 4.60% — implying a Treasury rate of 3.88%. Prices had fallen in 2022, and much of the higher discount rate had already been absorbed by lower multiples. Bonds became a far tougher comparison in 2023, but the cash-flow-based premium stayed close to its long-run norm.

Yet that is the counterweight to the 1999 story. Rising bond yields don't automatically make the equity premium thin. They raise the hurdle, and the outcome depends on whether prices adjust down to clear it. In 2022 they did. In 1999 they hadn't — prices had run ahead while the hurdle sat high.

Reading the Gap in 2026 Without Calling a Top

Now to the reading that makes this topic awkward. In his Data Update 2 for 2026, Damodaran started the year with the S&P 500 at 6,845.5, up from 5,881.6 a year earlier — a 16.39% price gain for 2025. His implied expected return was 8.41%. Subtract the 4.18% 10-year Treasury rate and the implied premium was 4.23%, which he described as "more in the middle of the historic range." Strip out a 23 basis point default spread he now assigns to the U.S. after the May 2025 Moody's downgrade, and the riskless rate drops to 3.95%, lifting the premium to 4.46%. His July 2026 update, with the index at 7,499.36 on July 1, put the implied premium at 4.42%.

Meanwhile the CAPE-based gauge tells a different story. Several readings of Shiller's data in August 2026 put the ratio above 40, the highest since 2000. A CAPE of 40 implies a real earnings yield of 2.5% — within a fraction of a point of the 2.26% of December 1999.

So one gauge says the premium is ordinary and the other says it's nearly as thin as at the dot-com peak. Both are measuring something real. The implied premium credits buybacks and builds in expected earnings growth; CAPE averages a decade of earnings that includes the 2020 dip and was earned by a market with a different sector mix. I'm less sure than I'd like to be about which one deserves more weight now. My instinct says the truth sits between them, and that the cash-flow gauge leans heavily on growth staying strong.

A single company shows how much rides on that growth term. Mastercard's 10-K for fiscal 2025 reported diluted earnings per share of $16.52. At the roughly $563 share price in the week of December 29, 2025, from MoatScope's weekly price history, that's an earnings yield of about 2.93% (16.52 ÷ 563.13) — 1.25 points below the 4.18% Treasury rate at the start of 2026. On the Fed model, the bond wins. But Mastercard's 10-K for fiscal 2022 reported net income of $9.93 billion, and its fiscal 2025 10-K reported $14.97 billion: compound growth of about 14.7% a year ((14.97 ÷ 9.93)^(1/3) − 1). Diluted EPS went from $10.22 to $16.52 over the same three years, helped by a share count that fell from 971 million to 906 million. A low starting yield is a bet that growth like that continues. When it does, the yield looks fine in hindsight. When it doesn't, the low yield is all you were paid.

What to do with all this, short of a forecast? I treat a thin premium as a regime marker that shifts probabilities. Across these episodes, starting from a low premium has lowered the central tendency of the following decade's returns and fattened the left tail; it has not told anyone when the repricing would arrive. It also changes what an equity holding has to do — if bonds are paying something real, a stock has to earn its place through growth and durability rather than simply by beating a near-zero yield. That's part of why quality businesses have tended to hold up better when the discount rate shifts, a pattern I traced in why quality stocks hold up better in inflationary regimes.

Key Takeaways

• The equity risk premium is measured at least three ways — earnings-yield gap, real CAPE yield against TIPS, and Damodaran's implied premium — and they can disagree sharply. Match nominal with nominal and real with real before drawing conclusions.
• The 1999 analogue is the clearest case of a thin premium: an implied premium near 2% against a 6.5% Treasury and a CAPE of 44.19, followed by an annualized -0.95% return for 2000–2009. The premium rebuilt itself through prices, not bond yields.
• Higher Treasury yields raise the hurdle but don't thin the premium by themselves. In 2023 prices had already adjusted, and the implied premium started 2024 at 4.60%.
• In 2026 the gauges split: the implied premium sits near 4.2–4.4%, while CAPE above 40 puts the real earnings yield close to 1999 levels. Treat that as a regime and probability question, not a timing call — and lean on business quality where the market's own cushion is thin.