Why US Stock-Market Sell-Offs Have Become Shorter — and What Could End the Pattern

The US stock market has delivered an unusual combination over the past 15 years: exceptionally high returns, expensive valuations and relatively little time spent in a deep decline.

For the 15 years ending around July 2026, a published market-structure study estimated that the S&P 500 produced a real annualised total return of 12.09%. “Real” means after inflation. “Total return” means that dividends are reinvested rather than taken out as cash. The exact answer changes slightly with the starting day, ending day and inflation series, but the broad conclusion is robust: investors earned roughly 11% to 12% a year after inflation over this period. The underlying price, earnings, dividend and inflation data are available from Robert Shiller’s historical market dataset. (1, 2)

To understand what a 12.09% real return means, imagine investing \$100 and leaving it invested for 15 years:

\[\text{ending purchasing power} = \text{starting money} \times (1 + \text{real return})^{\text{years}}\] \[\text{ending purchasing power} = \$100 \times (1 + 0.1209)^{15}\] \[\text{ending purchasing power} = \$100 \times 5.5399 \approx \$554\]

The investment would therefore have grown to about 5.5 times its original purchasing power, before tax and investment costs.

This strong return was accompanied by another unusual result: the market spent very little time in a deep drawdown. That raises an important question. Has the market become fundamentally safer, or has modern market structure changed the speed and shape of both sell-offs and recoveries?

The evidence points to a more precise answer: modern markets contain mechanisms that can make a sell-off faster, but they also contain large and recurring sources of demand that can help prices recover. Those mechanisms improve resilience while capital continues to arrive. They cannot guarantee that every recovery will be quick.

1. Deep drawdowns became unusually rare

A drawdown measures how far the market has fallen from its previous high.

The formula is:

\[\text{drawdown} = \left(\frac{\text{current index level}}{\text{previous peak}} - 1\right) \times 100\]

Suppose an index reaches 5,000 and then falls to 4,000:

\[\text{drawdown} = \left(\frac{4{,}000}{5{,}000} - 1\right) \times 100\] \[\text{drawdown} = (0.80 - 1) \times 100 = -20\%\]

Using daily S&P 500 closing prices from 29 July 2011 to 29 July 2026, and measuring each day against the highest closing level previously reached in that window, the index was more than 20% below its high on 74 trading days out of 3,771. (1)

The calculation is:

\[\text{share of days in a deep drawdown} = \frac{74}{3{,}771} \times 100\] \[\text{share of days in a deep drawdown} = 1.962\%\]

Across comparable 15-year periods beginning in 1928, the median was approximately 17.7%, using the rolling-window method described in the same market-structure study. “Median” means the middle observation: half of the historical periods spent more time below the 20% threshold and half spent less. (1)

The recent period still contained two bear markets, meaning declines greater than 20%:

In 2020, the S&P 500 fell about 34% during the pandemic shock.
In 2022, the index fell about 24.5% as inflation rose and the Federal Reserve increased interest rates. S&P Dow Jones Indices records the peak on 3 January 2022, the trough on 12 October 2022 and full recovery in December 2023. (3)

The unusual feature was therefore not the complete absence of bear markets. It was the small proportion of the full 15-year period spent more than 20% below a prior high.

2. Recoveries can be much faster, but the speed is not constant

The 1987 crash and the 2020 pandemic crash provide a useful comparison because their maximum declines were similar.

The 1987 decline was approximately 33.5%. From the trough, the index needed 414 trading sessions to regain its previous high.
The 2020 decline was approximately 33.9%. From the trough, the index needed 103 trading sessions to regain its previous high. (1)

The 2020 recovery was therefore about four times faster when measured from the bottom:

\[\text{relative recovery time} = \frac{414}{103} = 4.02\]

However, recovery speed does not follow a rule that makes each new bear market faster than the last. The 2022 decline took until December 2023 to recover. The right conclusion is narrower: some modern sell-offs have recovered exceptionally quickly, while others have not. (3)

To explain the difference, we need to separate the forces that create selling from the forces that create demand.

3. Indexed investing provides steady ownership, not an automatic rescue

An index fund holds securities according to a published index, such as the S&P 500. It does not normally decide that a company is cheap or expensive. A conventional S&P 500 fund buys each company in approximately the same proportion as that company’s weight in the index.

By June 2026, US domestic-equity mutual funds and exchange-traded funds contained:

\$15.215 trillion in indexed funds
\$8.616 trillion in actively managed funds

The indexed share was:

\[\text{indexed share} = \frac{\$15.215\text{ trillion}} {\$15.215\text{ trillion} + \$8.616\text{ trillion}}\] \[\text{indexed share} = \frac{\$15.215\text{ trillion}}{\$23.831\text{ trillion}} = 63.8\%\]

These figures come from the Investment Company Institute’s June 2026 fund statistics. (4)

This is a major structural change. Regular pension contributions, retirement-plan deductions and monthly investment plans can direct new money into index funds without requiring an investor to make a fresh valuation decision each time.

But the distinction between ownership and buying is essential. An existing index fund does not buy more shares simply because prices fall. It generally buys underlying securities when it receives net new money, and it may sell if investors withdraw money. The Bank for International Settlements explains that passive funds trade in response to investor flows, index rebalancing and changes in benchmark composition. (5)

Passive funds can therefore support the market when contributions continue during a decline. They are not a permanent buyer with unlimited capital.

4. Volatility-sensitive strategies can turn a price fall into more selling

Some investment strategies set their exposure according to market volatility. Volatility measures how widely and rapidly prices move. A quiet market has low volatility; a market making large daily moves has high volatility.

A simplified volatility-targeting rule is:

\[\text{desired exposure} = \frac{\text{target volatility}}{\text{estimated market volatility}}\]

Suppose a fund targets volatility of 10%.

When estimated volatility is 5%:

\[\text{desired exposure} = \frac{10\%}{5\%} = 2.0\]

The fund may hold exposure equal to twice its capital, subject to its leverage limit.

If estimated volatility rises to 15%:

\[\text{desired exposure} = \frac{10\%}{15\%} = 0.67\]

The fund may reduce its exposure to about 67% of capital.

In practice, managers use different estimation periods, limits and adjustment speeds. A single daily fall does not determine the amount they must sell. Nevertheless, the broad mechanism is real: falling prices can raise measured volatility, and higher volatility can lead risk-controlled strategies to reduce positions. The European Central Bank discusses this procyclical behaviour in its analysis of risk-parity and volatility-sensitive funds. (6)

“Procyclical” means moving in the same direction as the market cycle. Prices fall, measured risk rises, the strategy sells, and that additional selling can place further pressure on prices.

5. Leveraged multi-manager funds can transmit losses quickly

A multi-manager hedge fund, sometimes called a pod shop, divides capital among many small investment teams. Each team runs its own positions under strict loss and risk limits. Large examples include Millennium, Citadel, Point72 and Balyasny.

These funds often use leverage, meaning their total positions are larger than the investors’ capital. Reuters reported that multi-strategy hedge funds managed approximately \$428 billion in 2025 and that some operated with gross exposure above 600% of capital, based on Goldman Sachs data. (7)

Gross exposure adds the size of all long and short positions without allowing them to cancel each other. A long position gains when its price rises. A short position gains when its price falls.

Consider a team with \$500 million of capital and gross exposure equal to six times capital:

\[\text{gross exposure} = \$500\text{ million} \times 6 = \$3\text{ billion}\]

If the whole portfolio loses 0.83% of gross exposure:

\[\text{loss} = \$3\text{ billion} \times 0.0083 = \$24.9\text{ million}\]

Measured against the team’s capital:

\[\text{capital loss} = \frac{\$24.9\text{ million}}{\$500\text{ million}} \times 100 = 4.98\%\]

A movement of less than 1% across the gross book can therefore create a loss close to 5% of capital. That is why leveraged funds may cut risk quickly.

The resulting trades are not all sales. Closing long positions creates selling, while closing short positions creates buying. The direction and market impact depend on the actual portfolio, the overlap with other funds and the size of the risk reduction.

The most dangerous situation is crowding. Crowding means many funds own similar positions. If several teams reach their loss limits together, they may all try to exit the same securities at the same time. The problem is then not merely leverage; it is leverage combined with similar portfolios and limited buyers.

6. A sell-off and a recovery can be driven by different participants

A modern market decline can develop in stages:

News, an earnings disappointment or a change in interest-rate expectations causes an initial price fall.
Leveraged funds reduce positions because their losses reach internal limits.
The larger price movement raises measured volatility, causing some volatility-sensitive strategies to reduce exposure.
Market liquidity becomes thinner. “Liquidity” means the ability to trade without moving the price sharply. When buyers step back, even a moderate sale can produce a large price movement.

That explains how a sell-off can become fast. It does not explain the recovery by itself. The later demand can come from several separate sources:

New pension and savings contributions entering funds
Active investors buying companies they now consider undervalued
Companies repurchasing their own shares
Short sellers closing profitable positions, which requires them to buy shares
Options dealers changing their hedges as prices and option exposures change
Investors rebuilding positions after their risk measurements stabilise

The balance between forced sellers and available buyers determines whether the decline reverses quickly or remains deep for a long time.

7. Statistical models can describe the pattern, but they cannot name the trader

Two statistical tools help describe how price behaviour has changed.

GJR-GARCH: how shocks affect later volatility

GJR-GARCH is a model of changing volatility. Its full name is the Glosten-Jagannathan-Runkle generalised autoregressive conditional heteroskedasticity model. The long name describes a simple idea: tomorrow’s expected volatility depends partly on recent volatility and recent market shocks.

The model also allows negative surprises to affect volatility differently from positive surprises. Historically, a large fall tends to raise future volatility more than a similarly sized rise. This asymmetric effect is documented in the original GJR research. (8)

The model can estimate whether shocks have become larger or whether their effects fade more quickly. It cannot identify whether a shock was created by a hedge fund, an index fund, an options dealer or a human investor. That requires separate position, order-flow and fund-flow data.

Variance ratio: whether price movements persist or reverse

A variance-ratio test compares the variance of returns over several periods with the variance expected from one-period returns.

In simplified form:

\[\text{variance ratio} = \frac{\text{variance of multi-period returns}} {\text{expected variance under independent returns}}\]

A result near 1 is consistent with price changes that are independent across time.
A result above 1 suggests that movements tend to continue in the same direction.
A result below 1 suggests that movements tend to reverse at the selected time horizon.

The test was developed by Andrew Lo and Craig MacKinlay in their research on stock-market prices. Its result depends on the chosen period, the data frequency and statistical uncertainty. A value below 1 can provide evidence of short-term mean reversion, but it does not reveal why the reversal occurred. (9)

8. High valuations increase the importance of the support mechanisms

The cyclically adjusted price-to-earnings ratio, usually shortened to CAPE, compares the market’s price with average inflation-adjusted earnings over the previous ten years.

The formula is:

\[\text{CAPE} = \frac{\text{current market price}} {\text{average real earnings over the previous ten years}}\]

Using ten years of earnings reduces the effect of one unusually strong or weak year. A higher CAPE means investors are paying more for each dollar of long-term earnings.

Around July 2026, using Shiller’s published market series:

Standard Shiller CAPE was approximately 40.6 to 41.2.
A total-return version, which adjusts the historical index for reinvested dividends, was approximately 43.4 to 44.0.
The 15-year rolling average of standard CAPE was approximately 29.6, the highest such average in the available history. (1, 2)

These figures do not determine when the market will fall. Valuation is usually more useful for estimating long-term return potential than for predicting the next month or year. But a high starting valuation leaves less room for disappointment. If expected earnings fall or investors demand a higher return, prices may need to adjust more sharply.

Low dividend yields also need to be interpreted carefully. A low yield can reflect a high share price, not the disappearance of dividends. S&P 500 companies paid a record \$664.9 billion in dividends over the 12 months ending September 2025 and repurchased a record \$1.020 trillion of shares. (10)

A share repurchase, or buyback, occurs when a company buys its own shares. With fewer shares remaining, each remaining share represents a slightly larger ownership interest in the company, provided the company does not issue an offsetting number of new shares.

9. Higher bond yields can weaken equity valuations

Stocks compete with government bonds for investors’ capital. When a safe government bond offers a higher yield, investors may require a higher expected return from stocks as well.

The ten-year US Treasury yield was approximately 4.7% to 4.8% around the end of July 2026, according to the Federal Reserve’s ten-year Treasury series. A rise to 6% would be a material change. (11)

A simplified present-value calculation shows why. Present value means what a future payment is worth today.

Suppose an investor will receive \$100 in ten years.

At a 4.75% required return:

\[\text{present value} = \frac{\$100}{(1.0475)^{10}} = \$62.87\]

At a 6% required return:

\[\text{present value} = \frac{\$100}{(1.06)^{10}} = \$55.84\]

The change is:

\[\text{percentage change} = \left(\frac{\$55.84}{\$62.87} - 1\right) \times 100 = -11.2\%\]

This is not a stock-valuation model; companies generate many uncertain future cash flows rather than one fixed payment. It demonstrates the basic mathematics: when the required return rises, the current value of future money falls. Expensive companies whose expected profits lie far in the future are especially sensitive to that change.

10. What could turn a fast recovery into a prolonged decline?

A prolonged bear market becomes more likely when several supports weaken together.

Persistent investor outflows

Index funds can provide recurring demand while workers and savers contribute new money. If withdrawals remain larger than contributions, the same funds become net sellers. Retirement demographics may influence this balance gradually, but there is no automatic 2028 selling deadline. Under current US rules, required minimum distributions generally begin at age 73 and the starting age rises to 75 in 2033. Roth individual retirement accounts do not require distributions during the owner’s lifetime. (12)

The quantity that matters is observable net fund flow: money entering funds minus money leaving them.

High interest rates combined with weaker earnings

Higher bond yields reduce the present value of future company profits. If expected profits are also falling, both parts of the valuation calculation move against share prices at the same time.

Crowded leverage across several strategies

Leverage is most dangerous when many investors hold similar positions. A loss in one crowded trade can cause risk reductions across multiple funds, which can then raise volatility and trigger selling by a second group of strategies.

Reduced corporate buying

Companies are often restricted from repurchasing shares shortly before reporting results. These periods are commonly called buyback blackout windows. A market decline during a period of limited corporate buying may have fewer immediate buyers.

Narrow market leadership

A market-capitalisation-weighted index gives larger companies more influence. “Market capitalisation” is the share price multiplied by the number of shares. If a small number of very large companies drive most of the index’s return, the headline index can remain strong while the average company is weak. A later decline in those leaders can then affect the whole index disproportionately.

11. A practical market-structure dashboard

No single indicator can show whether the next decline will recover quickly. A useful monitoring framework combines seven separate measures:

Drawdown depth and duration: Measure both how far the index has fallen and how long it remains below its prior high.
Market breadth: Compare the capitalisation-weighted S&P 500 with an equal-weighted version in which every company has the same influence. A widening gap shows that performance is becoming concentrated.
Net fund flows: Track actual money entering and leaving equity mutual funds, exchange-traded funds and retirement plans. Fund size alone does not reveal current buying pressure.
Realised and expected volatility: Realised volatility measures movements that have already happened. Expected volatility, often inferred from option prices, reflects the market’s estimate of future movement. A rapid rise in both can increase systematic selling pressure.
Hedge-fund leverage and crowding: Monitor gross exposure, factor positioning and whether many funds own the same trades. “Factor” means a shared characteristic such as momentum, value, size or sensitivity to interest rates.
Valuation and earnings expectations: Track CAPE, ordinary price-to-earnings ratios and changes in analysts’ profit estimates. An expensive market can remain expensive, but expensive prices combined with falling earnings create a weaker foundation.
Bond yields and corporate demand: Track Treasury yields, share-repurchase announcements and blackout periods. These show the competition from safer assets and the availability of corporate buyers.

The indicators should be interpreted together. For example, high volatility is more concerning when leverage is elevated, market breadth is narrow, fund flows are negative and corporate buyers are absent.

Conclusion: resilient does not mean invulnerable

The modern US stock market contains two opposing sets of forces.

On one side, leverage, strict risk limits and volatility-sensitive strategies can turn an initial loss into rapid additional trading. This can make a sell-off unusually sharp.

On the other side, retirement contributions, active investors, company buybacks, short covering and position rebuilding can create powerful demand after prices fall. This can make a recovery unusually fast.

The past 15 years show that these recovery forces have often been strong. The S&P 500 spent only about 1.96% of the period more than 20% below its previous high, compared with a historical median near 17.7% for comparable windows. That is evidence of exceptional resilience.

It is not evidence that risk has disappeared. Volatility is not a substance that is mechanically stored during calm periods and released later. The more concrete risk is coordination: several large groups may be forced to reduce exposure at the same time while the usual buyers have less capital or less willingness to act.

The key question during the next major decline will therefore not be whether the market has fallen before and recovered. It will be whether fresh demand remains larger than forced selling. That balance, rather than a belief that the market must always recover quickly, will determine the depth and duration of the drawdown.