The Index Says No Correction. The Average Stock Disagrees.

The S&P 500 closed Friday at 7,376.86, about 6 percent below the record 7,844 it printed in October. By the definition most market desks use, that's a pullback, not a correction. Yet the Nasdaq fell into correction territory during an AI-led selloff this month, a stretch in which the Dow dropped 1,800 points in a single session and the S&P 500 lost 4.36 percent, as CNBC reported. Through late October, roughly 80 percent of S&P 500 members sat at least 10 percent below their own highs.

Both statements are accurate because they measure different things. A float-weighted index lets a small group of very large companies hold the headline near a record while most of the market works through a drawdown. The question investors keep asking, whether a correction has started, gets answered with the least informative number available.

The Definition Is Tighter Than the Conversation

A correction is a decline of 10 percent or more from a recent peak, measured retrospectively from that peak to the lowest closing price before the index recovers. The 10 percent line separates it from an ordinary pullback, and the 20 percent line separates it from a bear market. Neither threshold is scientific. They hardened into vocabulary because they're easy to remember and easy to put in a headline.

Measurement details change outcomes more than most people expect. Close-to-close counts differ from intraday peak-to-trough counts, and price-only indices differ from total-return series, where dividends shave a point or two off the depth of any decline. The most recent full US episode shows how fine the distinction gets. The S&P 500 fell 18.9 percent from its 19 February 2025 peak to its 8 April close, and that correction was declared finished on 27 June 2025.

Sitting about a percentage point under the 20 percent threshold, that episode gets filed as a correction or a bear market depending on which price series you prefer. Two analysts can examine the same seven weeks and disagree about the category. The label decides whether a client hears weather or a warning, which means the behaviour it triggers is being set by a rounding difference.

Two Frequency Numbers, Both Correct

The fullest count of US equity drawdowns since the Second World War runs to 37 corrections of 10 percent or more. Thirteen of them went past 20 percent. Spread across roughly eight decades, that averages out to a correction every 2.2 years and a bear market every 5.6 years.

Start the clock in 1980 instead and corrections arrive closer to every 1.2 years. Both numbers circulate widely, and people quote them as though one must be wrong. They diverge mostly through bookkeeping. Shorter windows capture different volatility regimes, and counting rules differ: a bear market that rallies 12 percent and rolls over again can register as two correction episodes under one method and one episode under another. Examine a long drawdown chart and you'll find several stretches where the index crosses the 10 percent line more than once inside the same decline.

Each figure answers a different question. The 2.2 year number describes how often the market breaks badly across a very long horizon. The 1.2 year number describes the cadence a 30 year investor is more likely to meet. Planning around the longer interval looks prudent and behaves carelessly, because the shorter one is the number the portfolio will be tested against.

What the Index Average Conceals

This is where the frequency statistics mislead most, and it's the part of the consensus I'd push back on hardest. The counts are computed on a cap-weighted index, and a cap-weighted index in 2026 is a narrow instrument. A handful of the largest companies carry enough weight to pull the headline toward its high while the median constituent falls. That's how the current split formed: an index roughly 6 percent off its record, a Nasdaq already in correction, about 80 percent of S&P 500 members in their own 10 percent drawdowns, and an average member maximum drawdown this year near 26 percent, which Charles Schwab's Liz Ann Sonders has called "bear-market-level weakness" beneath a record-setting index.

The consequence is that a correction every two years describes a portfolio most people don't own. An equal-weighted basket, a concentrated account, or a fund tilted toward smaller companies lived through something closer to a bear market this year while the headline index printed records. Institutional allocators run the same comparison at larger scale, which is why concentration decisions inside big sovereign portfolios get stress tested rather than argued (Norway's is the reference case).

Per-stock drawdowns are also more frequent than index drawdowns, because stock-specific risk sits on top of market risk. A 10 percent fall in a single holding is routine enough to be noise. A 10 percent fall in the index is the thing that gets a name and a frequency estimate. The two get conflated because both get called corrections, and only one of them can be planned around.

One in Three Crosses the Twenty Percent Line

The consensus story is that corrections are short, shallow and buyable. The conditional record is less comfortable. Thirteen of the 37 post-war corrections extended past 20 percent, so a 10 percent drawdown has carried roughly a one in three chance of becoming something larger. The unconditional base rate says a bear market every 5.6 years. The conditional rate says that once the market has already broken by 10 percent, the next 10 points are weighted more heavily toward loss than most planning assumes.

Clustering compounds it. Three bear market corrections of 20 percent or more have occurred in the last five years, against a long-run average that would predict fewer than one. Drawdowns aren't evenly spaced because the conditions that produce them persist across quarters rather than days: credit stress, positioning, valuation resets. Volatility is autocorrelated, and a market that has just de-rated is more likely to de-rate again than a market sitting at a record.

For a household the implication is about sizing rather than forecasting. If nobody can tell the two-in-three correction from the one-in-three bear market while it's unfolding, then the position that survives both is the only one worth holding. That's an argument for sizing to the bad case, not for sitting out.

The Recovery Math Nobody Budgets For

Losses and gains aren't symmetric on the way back, and that asymmetry is what makes the frequency data feel worse in practice than it reads on a chart.

The 14.2 percent row is the one to study. That's the average maximum intrayear decline in the S&P 500 from 1980 to 2025, and over the same span 35 of 46 calendar years finished positive. A typical year therefore contained a double-digit hole and a positive close, and recovering from the average hole required a 16.6 percent gain just to reach level again. The index's long-run return doesn't arrive smoothly. A meaningful share of each year's advance gets spent climbing out of a hole dug earlier in that same year.

Duration matters as much as depth. Japan's Nikkei 225 needed until 2024 to exceed its December 1989 peak, a recovery measured in decades rather than months, and one that no cashflow plan built around a two year correction would have survived.

Liquidity sits at the centre of that math. Someone forced to sell during a 14 percent hole converts a temporary mark into a permanent loss, and the cost of the forced sale usually exceeds the return forgone by holding cash. Keeping planned spending for the next 12 to 24 months outside equities is the cheapest insurance available, and it's bought in calm weeks rather than in the middle of an AI-led selloff (the buffer math is here).

What the Record Asks of You

Two conclusions follow. The 1.2 year cadence since 1980 is the better planning assumption than the 2.2 year figure since 1945, because concentration and passive flows make index-level air pockets faster to form, and the modern record reflects it. And index cadence is the wrong yardstick for anyone who isn't holding the index.

I'd argue the more useful habit is to stop treating a 10 percent decline as an event and start treating it as a budget line. If the market hands you a double-digit hole somewhere between every 18 months and two years, the portfolio that works is the one that can fund its own life through that hole without selling at the bottom of it.

The unresolved variable is whether the compressed volatility of the past few years is structural or temporary. Index concentration, systematic option selling and steady passive inflows have all dampened realised moves, and the S&P 500 set records through 2026 while a Gulf war and a midterm cycle played out. If that dampening holds, the interval between 10 percent corrections lengthens and the eventual decline arrives faster than the last generation of drawdowns did. The measure that settles it is the distance between this October's high and the next 10 percent close on the index, and until that gap prints, the defensible assumption is the faster cadence with the deeper constituent-level drawdown underneath it.

Disclaimer: This article is for informational purposes only and does not constitute financial advice. Past performance is not indicative of future results. Consult a qualified financial advisor before making investment decisions.