O Market, Where Art Thou?

Professional investors often obsess about beta, and for good reason. It’s one of the quickest ways of assessing risk relative to the market. A stock with a beta above one has historically tended to move more than the market; a stock with a beta below one has tended to move less. A negative beta is something stranger. It means that, over the period being measured, the stock has tended to move in the opposite direction from the market. Stocks like those used to be like four-leaf clovers—normally uncommon. But a recent chart making the rounds suggests things have suddenly gotten weird.¹

We give credit to Goldman for the inspiration, but the perspiration was all Belvest. Our reconstruction of S&P 500 membership and returns finds that fully 18.2% of index constituents had a negative 252-day beta to the S&P 500 Total Return Index at the end of September 2026. That is roughly twice the previous high in our reconstructed history, 9.1% in April 2001. More importantly, this is not entirely a collection of estimates hovering just below zero; about 8% of the index currently has a beta that is statistically significantly negative at a conventional 95% threshold. This is real, and it’s meaningful.

Chart 1

Share of S&P 500 Constituents with a Negative Beta

252-day rolling beta on daily returns against the S&P 500 Total Return Index, December 1990–September 2026

Each month’s cross section is built where we have historical prices. E.g., 183 of 491 names in December 1990, rising to 499 of 503 in September 2026. Beta is a 252-observation OLS regression of daily returns on the S&P 500 Total Return Index. The shaded years are where we can only model 70% of the index or less and are survivor-weighted. September 2026 is 18.2%, against a prior high of 9.1% in April 2001. Source: LucenWorks; Financial Modeling Prep; Belvest Family Wealth calculations.

Others have recently observed the same phenomenon using different methodologies. Richard Bernstein, now at Janus Henderson, highlighted a “near-record” number of S&P 500 companies with negative betas in September, based on trailing 36-month calculations through July 2026.² His calculation is not directly comparable with ours—we use one year of daily returns—but the fact that independent approaches are detecting the same unusual behavior is notable.

So what is happening?

The answer appears to involve two developments that might initially seem contradictory: 1) the S&P 500 has become extremely concentrated, while 2) the stocks inside it have become extraordinarily uncorrelated.

A Market Dominated by a Few Companies

A capitalization-weighted index gives its largest companies the greatest influence over its returns. That has always been true. That’s always been a feature and a bug of index investing; the S&P 500 is largely the apotheosis of all of this. What has changed is the magnitude.

In our data, the ten largest S&P 500 companies accounted for almost 40% of total constituent market capitalization in September. This is up dramatically from just 19.4% at the June 2015 trough. The five largest alone now represent a shade under 30%. (Concentration actually reached 41.3% in October 2025 before retreating a bit—the AI trade didn’t happen yesterday.)

Chart 2

The Ten Largest S&P 500 Companies as a Share of Index Market Cap

Total market capitalization is shares outstanding times price, divided into the ten largest. The names we can’t price were dropped from the index; they tend to be small and are missing from the denominator only, which pushes the measured share up. The band is drawn to depict that bias. Source: LucenWorks; Financial Modeling Prep; Belvest Family Wealth calculations.

For data reasons, we had to make certain compromises. For instance, our series uses total market capitalization rather than the float-adjusted weights employed by S&P, but S&P’s official data tell essentially the same story. As of August 31, its ten largest constituents represented almost 38% of the index.³ Separate research from S&P Dow Jones Indices echoed what we calculated—that the top ten had approached 40% by mid-2025, a degree of concentration not seen since the mid-1960s.⁴

There is an obvious temptation to compare today with the internet bubble at the turn of the 21st century. Our reconstructed top-ten share reached 41.1% in March 2000, seemingly similar to today’s level. Unfortunately, our early data are incomplete: we can measure only about 53% of the index at that point, and the missing companies tend to bias measured concentration upward. We therefore cannot confidently say whether 2000 or today was more concentrated. But, we don’t really need to… the broader conclusion is clear enough: today’s S&P 500 is exceptionally top-heavy by modern historical standards. There is no argument about this.

What does it mean? Beta matters not because it is some magic immutable characteristic embedded in a company. It describes the relationship between a stock and a benchmark. Change the behavior of the benchmark and a stock’s beta can change even when nothing fundamental about the company has.

A Limited Beta Budget

There is a formula at the end, but an equation isn’t required to understand the intuition: the S&P 500’s beta to itself is one. In a fixed-weight portfolio, the weighted betas of its constituents collectively have to reconcile to that one. In that sense, the index has a defined “beta budget.”

If a handful of companies represent a very large share of the index, and those companies also have relatively high betas, they consume a disproportionate portion of that budget. The weighted-average beta of everything else must therefore be lower. In so keeping, increasing concentration can push the beta distribution of the rest of the market downward.

But it does not explain negative beta by itself. No matter how concentrated an index becomes, the remainder could theoretically consist entirely of stocks with positive, if low, betas. For a stock’s beta actually to cross below zero, the stock must have moved negatively, on balance, relative to the benchmark. Which brings us to the more surprising part of the story.

Spaghettification

In cosmology, there is this strange phenomenon whereby tidal forces can pull an object apart even as every piece continues falling inward. It’s actually called spaghettification. That’s sort of what’s happening here: the index is moving in one direction even as its constituents are moving increasingly apart.

The average pairwise correlation among S&P 500 constituents in our analysis has fallen to 0.121, approximately the lowest level we observe in a well-covered cross section since 1990. The December 2017 reading was almost identical at 0.127. But, as usual, the average does not tell the whole story (it hardly ever does). Today, 22.5% of all pairs of S&P 500 constituents have a negative trailing one-year correlation. At the comparable 2017 correlation low, only 14.2% did. In other words, the correlation distribution has not simply shifted downward. It has developed a much larger left tail.

Chart 3

Average Correlation Between S&P 500 Constituents

Mean of all pairwise correlations of daily returns over a trailing 252 days. December 1990–September 2026

The mean of all pairwise correlations of daily returns over a trailing 252 trading days. September 2026’s 499-name cross section average looks across 124,251 individual pairs. The band is the tenth to ninetieth percentile of those pairs. Shaded years cover under 70% of the index and are survivor-weighted. Source: LucenWorks; Financial Modeling Prep; Belvest Family Wealth calculations.

This provides the missing link between concentration and negative beta. Imagine a relatively small company inside the S&P 500. Because the company itself is included in the index, there is automatically a small positive relationship between its return and the index return. But its index weight may be only a few basis points. Most of its measured relationship with the S&P therefore depends on how it moves relative to the hundreds of other companies in the benchmark. And increasingly, the performance of that benchmark is determined by a small number of extremely large companies.

If the smaller company repeatedly moves in the opposite direction from those dominant stocks, that relationship can eventually overwhelm its tiny positive contribution to the index. Its measured beta can fall below zero.

Put more simply: Concentration determines whose movements increasingly define “the market.” Correlation determines how many other stocks move with them.

When a capitalization-weighted benchmark becomes unusually concentrated at precisely the same time that correlations across its constituents are unusually low, a large negative-beta tail becomes much easier to understand.

Concentration and Diversification Can Coexist

There is probably another obscure quantum physics allusion here—perhaps some sort of strange duality.

By one measure, the S&P 500 has become less diversified: almost 40% of its capitalization resides in ten companies. By another, it has become more diversified: the individual companies inside it are behaving less alike. Both can be true.

Diversification depends not only on the number of securities in a portfolio or their respective weights, but also on the relationships among their returns. The Cboe makes this point explicitly in describing its implied-correlation indices: lower correlations can improve diversification benefits, while sharp increases in correlation signify greater systematic risk and less effective diversification.⁵

2026 Cboe data have repeatedly shown this tug of war between concentrated leadership and stock-specific differentiation. In May, for example, mega-cap technology stocks led the market higher while implied correlation fell and dispersion increased. In June, leadership broadened sharply and implied correlation fell further.⁶ Today’s S&P 500 may therefore simultaneously be highly concentrated in what determines the index return and unusually diverse in what its individual companies are actually doing. That is an important distinction and one of the points we want to emphasize.

An investor looking only at the top-ten weight could reasonably conclude that nearly everything in the index is participating in the same trade. Our correlation data suggest something a good bit different. A growing portion of the market is behaving very differently from the companies carrying the greatest weights. And therefore one could say that negative beta is simply an extreme manifestation of that divergence.

But Don’t Get Too Excited

The speed of the recent decline in correlation deserves some qualification. Our calculations use rolling 252-trading-day windows. In March 2026, that window still contained April 3 and 4, 2025, two extraordinary market-wide trading sessions, as well as the subsequent 9.5% rebound. Large moves shared broadly across the market naturally increase measured stock-to-stock correlation. When those observations rolled out of the one-year window, our average correlation fell sharply, from 0.268 in March to 0.157 in April.

That is a window effect. Framing. It would be misleading to interpret the entire drop as evidence that the underlying structure of the market transformed in a month. We see this all the time in how you define the period of interest.

But wait! That’s not the whole story, either! Five additional months of more new observations have pushed average correlation still lower, to 0.121. Meanwhile, the proportion of negatively correlated pairs has continued to rise. The latest reading therefore looks less like an artifact than an unusually dispersed market whose underlying structure had been confounded by the unusual spring 2025 volatility.

What Negative Beta Means

The first implication is that beta is always benchmark-relative.

A utility, healthcare company, consumer-staples business or industrial company does not fundamentally become an inverse equity instrument because its trailing beta moves from +0.2 to −0.1. Instead, the economic character of the benchmark against which it is being measured may have changed.

The S&P 500 is a broad representation of large U.S. companies. But it is also float-adjusted and capitalization weighted, which means the returns of the largest companies exert the greatest influence on the index.³ At today’s concentration levels, that’s a lot. Against such a benchmark, a negative beta can sometimes tell us as much about what the benchmark is as it does about the individual company.

The second implication may be more useful to us as investors going forward. If different segments of the equity market are responding to different economic and fundamental drivers, there may currently be more diversification available within equities than headline index concentration suggests. However… a stock with a negative beta over the past year is not necessarily a hedge against the next bear market. Correlations are regime-dependent. Any option trader can tell you correlations rise during periods of financial stress and that investors typically price higher correlation into downside scenarios than upside ones (that’s the skew).⁵ In other words, the very diversification we observe today may evaporate when you are counting on it the most.

Today’s negative-beta stocks may offer something better described as leadership or rotation diversification than crash protection. They may behave differently when the dominant mega-cap trade weakens or reverses without necessarily appreciating during a generalized market decline.

Still no free lunch.

A Different Kind of Market

I hope we’ve gotten across that there is no mathematical rule saying that extreme market concentration must produce negative betas. Nor can we establish from these three charts that concentration caused correlations to fall. We can emphasize the unusual combination/contradiction, however.

The largest companies account for a share of the S&P 500 rarely seen in modern markets. Average correlations among its constituents are near historical lows. Nearly a quarter of constituent pairs now move negatively with one another over the trailing year. And almost one stock in five has wound up with a negative beta to the index itself.

We reconcile this thusly: no, the S&P 500 has not stopped being a broad equity benchmark. But it has become a benchmark whose day-to-day direction can be determined by a small group of companies while much of the rest of the market follows a very different path. That may be the most useful way to interpret today’s extraordinary number of negative-beta stocks.

They are not necessarily telling us that nearly one-fifth of large U.S. companies have somehow become hedges against equities. They may be telling us that this isn’t your father’s market any longer.

Appendix

Data and methodology endnote

Appendix

Endnotes

Important disclosures