Dad, Are We in an Equity Bubble?

Every valuation method ever invented answers the same underlying question: will the price I pay today be justified by the cash this business produces in the future — as dividends, as reinvested growth, or as proceeds if the whole business were sold? Everything below is just a different lens on that one idea, and each lens replaced the one before it after a crash exposed its blind spot.

Dividend yield & book value (pre-1929). Before the Wall Street Crash, investors judged a share almost entirely by its dividend yield against a government bond, or by the assets stated on its balance sheet — there was barely any real earnings analysis behind the numbers, since audited financial statements only became mandatory in 1934. Both measures are still tracked today, and both currently sit near an extreme. The S&P 500's dividend yield is about 1.2% right now — within a hair of the all-time low of 1.11% set in August 2000, and well below the roughly 1.8–2.0% that was typical for most of the last two decades. It isn't that companies abandoned dividends: just over half of S&P 500 firms still pay one, barely different from 25 years ago. The real shift is who dominates the index — today's largest constituents are a handful of trillion-dollar technology names that pay little or nothing, preferring buybacks instead. Price-to-book tells a similar story from the other side: the index trades around 5.3–5.8× book value today, versus roughly 4.0× at the 2000 dot-com peak and just 1.78× at the 2009 trough — the richest reading in at least a quarter-century.

Value investing, Graham & Dodd (1934). Two Columbia professors codified the lessons of 1929 into Security Analysis: only buy where careful analysis shows a genuine margin of safety, and treat anything else as speculation rather than investment. Their sharpest warning was that investors kept assuming recent growth would simply continue forever — a mistake that resurfaces in nearly every bubble since, arguably including some of the more aggressive AI-driven earnings forecasts for 2027 and 2028 floating around today. Graham's framework doesn't reduce to one clean number the way a dividend yield or a P/E does, but it survives as a habit of mind: always asking how much cushion you have if your growth assumptions turn out to be wrong.

Growth investing, Phil Fisher (1958). Fisher shifted the emphasis toward companies with durable growth potential, usually from competitive advantage and heavy R&D spending. Even Graham eventually adapted his own formula to pay up for earnings growth. Taken too far, that logic helped inflate the 1972–73 “Nifty Fifty” bubble, when America's favourite blue-chip growth stocks traded near 50× earnings — roughly double where the S&P 500's trailing P/E sits today. It's worth sitting with that comparison: today's mid-20s trailing multiple, elevated as it is by modern standards, is nowhere near what investors once paid for a narrow set of “can't lose” growth names.

Discounted cash flow (late 1980s on). Spreadsheets made it trivial to project a company's future cash flows and discount them back to today's value — and just as easy to copy-paste an unrealistic growth rate ten years into the future without checking whether it was realistic. That habit helped inflate the 2000 technology bubble; Vodafone, for example, traded above 70× earnings at the peak, and the shares still haven't recovered that level today. The same discipline matters right now: a good deal of today's bullish 2027–28 forecasts effectively run a DCF-style extrapolation of AI-driven capital spending and productivity gains. The method isn't wrong — but exactly as in 2000, the entire case rests on whether that growth materialises, not on the elegance of the spreadsheet.

Return on invested capital & the simple P/E (2000s on). As investors grew wary of easily-managed earnings, attention shifted toward return on invested capital (ROIC) — how efficiently a company turns each dollar invested into profit — alongside academic frameworks like the capital asset pricing model, which had the unfortunate side effect of making a company look more attractive simply for carrying more debt and less equity. That helped inflate bank valuations ahead of 2008 (Lehman Brothers traded above 2.3× book value in 2006–07, versus a historical norm near 1× for US investment banks). Corporate profitability itself has genuinely shifted since then: after-tax corporate profits currently run at roughly 12% of US GDP, more than double the 5.6% share recorded at the 2000 dot-com peak. That is a real, measurable change in how much of the economy's output companies actually keep — not a sentiment shift — and it's the single best explanation for why almost every ratio benchmarked against GDP or against history looks stretched today, even though the multiple investors pay for each dollar of profit has not expanded nearly as much.

Four of the numbers I'll keep coming back to in this letter deserve a plainer definition than “a ratio Wall Street uses,” because each one is genuinely simple once you strip out the jargon.

The Shiller CAPE ratio. Take the price of the index and divide it by the average profit companies have earned per share over the last 10 years — adjusted for inflation, so a dollar from 2017 and a dollar from today are counted on equal terms — instead of just the most recent 12 months. Why 10 years? Because judging a company (or a whole market) on one year's profit is like judging a salesperson's typical paycheck from their single best or worst month. Averaging a full decade smooths out both booms and recessions, so you're pricing the stock against something closer to its normal, sustainable earnings power rather than whatever happened to be true last quarter.

Price-to-book. A company's “book value” is simply what's left on paper if you added up everything it owns and subtracted everything it owes — the accountant's answer to “what is this company worth, on paper, today?” Price-to-book compares the stock price to that number. A reading of 1× means you're paying exactly what the balance sheet says the company is worth if it were liquidated tomorrow. Above 1× means you're also paying for things a balance sheet can't capture — a strong brand, patents, a loyal customer base, talented people. Well below 1× is the territory where “cigar-butt” investors go hunting for statistically cheap companies, as I mentioned above.

The Buffett Indicator. Add up the total market value of every publicly traded US company, then divide by the size of the entire US economy (GDP) for the year. If the result is 100%, the stock market is worth exactly one year of everything the country produces. Buffett's logic was straightforward: over a long enough stretch, the stock market can't permanently grow much faster than the real economy generating the profits underneath it — so when this ratio balloons far above its own history, it's a sign the market has gotten ahead of the economy that's supposed to be supporting it. One honest flaw worth knowing: the numerator counts the full global earnings of US-listed companies, while the denominator counts only what's produced inside the United States. US companies now do a lot more business abroad than they used to, so some of the ratio's rise over the decades isn't overvaluation at all — it's US companies earning a growing share of their profits overseas, in economies GDP never counts. That doesn't make today's reading cheap, but it does mean comparing it directly to a reading from 30 or 40 years ago slightly overstates how much richer the market has actually become.

The equity risk premium (ERP). If you lend money to the US government for 10 years, you know almost exactly what you'll earn — that's the “risk-free” yield. Stocks are riskier than a government bond, so investors should demand some extra return on top for taking that risk. The ERP is simply that extra amount: roughly, what stocks are priced to earn you each year, minus what a safe 10-year bond already pays you. A thin ERP means investors are accepting very little extra reward for the risk of owning stocks over bonds; a fat one means they're demanding a lot.

Put those alongside the plain dividend yield and the plain P/E ratio, and you have six different lenses on the same question, each with its own blind spot. Line them all up against their own history and a pattern jumps out immediately: CAPE, the trailing P/E, price-to-book, and the Buffett Indicator are all telling you the market is priced richly — several of them as rich as, or richer than, the 2000 peak. The dividend yield tells the same story from another angle: investors are accepting less cash-in-hand today than at almost any point in the last century, betting on growth instead of income. Only the equity risk premium disagrees — and it disagrees for a specific, checkable reason: it's the only one of the six that weighs today's stock prices against today's bond yields rather than only against history, and today's bond yields are unremarkable rather than near zero the way they were for much of the 2010s and 2020s. That single disagreement matters more than any one ratio on its own, which is exactly why I never trust just one of these six.

Here is the chart I wish more people looked at before calling the market a bubble. It shows the S&P 500's trailing twelve-month earnings per share (EPS) against its price/earnings ratio, every quarter, since just before the 2008 crisis.

Look at the shape of those two lines. The green line — earnings — climbs almost relentlessly, with one sharp dent during the 2008–09 crisis and a brief one in 2020. The gold line — the multiple investors pay for those earnings — bounces around in a much narrower range, roughly 12× to 31×, with no clear multi-decade uptrend of its own. Most of the S&P 500's rise over the last 16 years came from companies actually earning more — not from investors deciding to pay more for the same dollar of profit.

Put in annual terms: earnings compounded at roughly 11% a year over that stretch, while the multiple investors were willing to pay expanded at only around 2% a year. Decompose the market's total climb and about 85% of it is earnings growth; the remaining 15% is investors paying a somewhat richer multiple than they did in 2009.

Here's a useful counterfactual. If the market today still traded at the same 20.4× multiple it did at the end of 2009, applied to today's $298 of trailing earnings, the index would sit roughly 28% below where the current multiple actually puts it. That gap — not the full price level — is the part of today's market you'd call “expensive” relative to 2009. The rest is companies simply earning more.

There's one more twist, and it makes the story stronger still. Everything above uses trailing earnings — what companies have already reported. You asked me to check what analysts expect over the next twelve months, so I did: as of mid-September 2026, the consensus forward-12-month EPS estimate for the S&P 500 is running around $400 (FactSet) — higher than the $365 you'd heard, not lower. Trailing earnings over that same window sit almost exactly at the $300 you mentioned, so the consensus is effectively pricing in roughly 33% earnings growth over the coming year, largely on the back of AI-related capital spending.

Your instinct was exactly right — the multiple really does shrink fast once you look forward instead of backward. Simply annualising the most recent quarter alone pulls the multiple from 25.5× down toward 20.6×, without a single new quarter of data needed. Layer in what analysts actually expect the market to earn over the next year, and the multiple compresses further still, to roughly 19× — close to, or even slightly below, the S&P 500's own 10-year average forward P/E of about 19×. A market that looks rich on trailing earnings can look almost ordinary on forward earnings — that gap is the entire argument for why EPS growth, not sentiment, is doing the heavy lifting.

Understanding that earnings drive most of the price is necessary, but it isn't sufficient on its own — a market can still be a bubble if the earnings themselves are unsustainable, or if investors are also paying up for a fair multiple on top of already-elevated profits. Here are the five checks I actually run, in the order I trust them.

1. Shiller CAPE (≈41–42×, vs. a ≈17× historical average). Price against a 10-year average of inflation-adjusted earnings. Best long-run (decade-ahead) return forecaster; weak at timing.

2. Trailing / forward P/E (≈25.5× / ≈19×, mid-September). The simple ratio from the chart above. Easy to sanity-check, but says nothing about interest rates, and the forward figure only holds up if the ≈33% earnings growth analysts are pencilling in actually shows up.

3. Buffett Indicator (≈220–250% of GDP, vs. an ≈84% median). Total market value against the real economy underneath it — see the table above for how far above its own dot-com peak (≈146–172%) it now sits. Ignores rates entirely and is structurally biased upward by overseas revenue growth and today's much higher corporate profit margins.

4. Equity risk premium (≈4.2%, vs. a ≈4.6–4.7% 30-year average). My favourite, because it asks “expensive compared to what?” — specifically, compared to today's bond yields. Currently close to its own long-run norm.

5. Sentiment & positioning (elevated). Margin debt, put/call ratios, bullishness surveys. Not a valuation tool at all — it measures crowding and leverage, which is usually what actually triggers a sharp drop.

Honestly, Arthur? None of these five metrics reliably calls the top by itself — that's the most important sentence in this letter. They are regime indicators, not alarm bells. What I actually do is triangulate: when several flash at once and sentiment and leverage are stretched too, that combination — not any single number — is what has historically preceded the sharpest falls.

Right now: CAPE, trailing P/E, price/book, and the Buffett Indicator all say “historically rich” — several of them at or above where they sat at the 2000 dot-com peak. The equity risk premium, the one I trust most because it accounts for interest rates, says “not unusually rich once you compare stocks to bonds.” And underneath all of it, roughly 85% of the market's climb since 2009 is explained by companies actually earning more, not by a re-rating — with the forward multiple, on the ≈$400 analysts expect the market to earn next year, sitting close to its own 10-year average rather than at an extreme. That's not a green light to be reckless: earnings growth can slow, that $400 estimate can get cut, and a 25–28× trailing multiple leaves little room for disappointment if it does. But it is a real, measurable difference from 2000, when the P/E was similarly high while earnings were not growing anywhere near fast enough to support it, and investors were accepting almost no extra reward for the risk.

My rule stands above all five metrics: read the annual report, understand the business, and never do equations with Greek letters in them just because they look clever. Valuation tools sharpen judgement. They were never meant to replace it.

Arthur, the first pushback you'll get from anyone I know is: “sure, EPS is growing — but isn't that just Nvidia and friends propping up the whole index?” It's a fair question, so I pulled the numbers apart by company.

The Mag 7's earnings did grow faster than the rest of the market — a lot faster. But “faster” is not the same as “alone,” and the dollar figures make that distinction clear.

Nvidia is the standout: its contribution to index EPS climbed from $6.37 to $29.57 — a combination of its index weight more than doubling (2.99% to 8.08%) and its own per-share earnings rising almost sevenfold. It's now the single largest contributor to S&P 500 earnings, Mag 7 or otherwise. Amazon's per-share earnings nearly quintupled ($2.61 to $12.80) as its profitability genuinely improved. Tesla is the odd one out — its per-share earnings actually fell ($3.40 to $1.09), yet its dollar contribution still rose slightly, an index-weighting quirk rather than a earnings-quality story. And Microsoft is the only one of the seven to have lost index weight over the period, even as its earnings kept growing.