Skewness in Stock Returns: What Vanguard's New Diversification Study Shows
Skewness in stock returns is the reason the typical stock trails the market: a few names multiply many times over and pull the average up, while the majority sits below it. A Vanguard study from October 2026 has run the numbers for all US stocks since 1990. Here is what it measured, why the mean and the median are so far apart, and what follows for diversification in your own portfolio.
What the study examines
Plagge and Lewis (2026) analyze every stock that was part of the Russell 3000 at any point between January 1990 and December 2025. The index covers almost the entire listed US stock market. That adds up to roughly 12,500 securities, many of which were acquired or delisted long ago. Returns are total returns in US dollars, so price gains and distributions together.
The study asks three questions: How many stocks beat the index? How much of the wealth the market created depends on a few companies? And how does a portfolio's outcome change with the number of holdings? The authors work for a large provider of index funds. The paper is labeled a marketing communication for professional investors, but its numbers sit well next to the independent research.
The results in numbers
The picture is clear. Of the roughly 12,500 stocks, 75% trailed the Russell 3000, and 56% did not even beat the risk-free rate. The average return across all stocks was 575%, while the middle stock (the median) lost 6.6%. The average and the typical case are far apart.
The concentration is even stronger in wealth creation. For each stock, the study calculates what became of its market value at the time it entered the index, and adds up the gains of all stocks. Ten companies account for 31% of the total wealth the US market created from 1990 to 2025.
Together 31.31% of the wealth created by all stocks that were ever part of the Russell 3000. Based on Plagge and Lewis (2026).
The names on the list are worth a look. Next to the large technology companies you find a tobacco company, a retailer and an oil company. Which stocks would become the big winners was not foreseeable in 1990: two of the ten were not even listed at the time.
Why the mean is above the median
A stock can lose at most 100%, but it can rise by many thousands of percent. Over long periods, compounding amplifies this asymmetry. The distribution of returns becomes right-skewed (see Skewness and Kurtosis): many small and negative outcomes, a few very large ones. The few outliers carry the mean, but not the median.
A simple model shows the order of magnitude. If a stock's annual returns fluctuate with volatility σ and are independent of each other, its final value after T years falls below the expected value with the following probability:
- cumulative distribution function of the standard normal distribution
- volatility of the stock per year
- holding period in years
Model with independent, normally distributed continuously compounded annual returns. The longer the period and the higher the volatility, the less often a single stock reaches the average.
The model is rough, but it points in the same direction as the study: over 36 years the large majority of individual stocks ends below the average, even though each one has the same expected return. An index holds all stocks and therefore earns the mean, not the median. More on this, with a worked example for whole portfolios, is in the article on Stock Picking.
Missing winners costs more than avoiding losers gains
The study also calculates what happens when individual stocks are missing from the index. It leaves out the best stocks in one run and the worst in another (in each case the single one, the 10, the 50 and the 100 with the largest contribution) and spreads their weight across the rest. The result is not symmetric: without the best stocks the return of the Russell 3000 fell by up to 2.53 percentage points per year, and without the worst it rose by at most 1.03 percentage points per year.
Anyone who picks stocks therefore carries two risks of unequal weight. Holding a loser costs little, because its weight shrinks with its price. Not holding one of the few big winners costs a lot, because its contribution grows over decades. The authors add a caveat themselves: over shorter periods the result depends on whether large or small companies happen to be ahead.
How many stocks a portfolio needs
For the third question the authors build random portfolios of 1 to 10, 50, 100, 200 and 250 stocks from the Russell 3000, with 500 runs per size, reconstituted every quarter. With a single random stock the median return was 13 percentage points per year below the index. Every additional stock shrinks this gap.
- Few holdings: The chance of owning one of the big winners is small. The median outcome is well below the market, while individual portfolios end far above it.
- A few dozen holdings: The portfolio's volatility is already close to the market (see Correlation and Diversification), but the median gap is still visible.
- The whole market: If you hold all stocks, you are certain to own the future winners without having to identify them in advance.
The study therefore reframes diversification. Spreading your money does not only lower risk, it also raises the return an investor can expect in the typical case, because it moves the median closer to the mean.
What the research shows
The finding is not new, the study confirms it with current data. Bessembinder (2018) showed for US stocks since 1926 that the majority returned less over their entire listed life than one-month Treasury bills and that the best 4% of companies explain the entire net gain of the market. Bessembinder et al. (2023) found the same pattern for stocks from around the world.
The experience with active funds fits the skewness. By the arithmetic of Sharpe (1991), all active investors together earn the market return before costs. According to Plagge and Lewis (2026), in the ten years through the end of 2025 only about 30% of active US funds with a broad benchmark beat their index, and that is already before costs. More on this in the article Active vs. Passive.
Limits and context
- US only, 1990 to 2025 only: This period includes the rise of a few technology companies. For other countries and periods the authors point to other work and provide no numbers of their own.
- Size matters: Wealth creation is calculated by market value, so large companies rank higher. If all companies are treated as equally large, the ten best still account for 22% of wealth creation, and seven of the ten names are different.
- No verdict on every selection: Skewness does not mean stock selection never works. It means chance works against small portfolios and that an edge has to be large to make up for it.
- A sender with an interest: A provider of index funds has a reason to recommend broad diversification. The core numbers, however, match the independent research.
- Dead stocks are part of it: The calculation only works because acquired and bankrupt companies count as well. If you look only at today's index members, you underestimate the skewness (see Survivorship Bias).
How to check it in Portfolio Lab
In the Backtest you can put a portfolio of a few individual stocks next to a broad index ETF and see return, volatility and the largest drawdown over the same period. The backtest works with the stocks that still exist today. The result of a selection made with hindsight therefore looks better than it was for an investor who had to choose back then.
The Look-through shows how much an ETF portfolio depends on its largest positions, and Monte Carlo calculates the range of possible paths of a portfolio. How the backtest treats prices, distributions and costs is described in the methodology.
Frequently asked questions
What does skewness mean for stock returns?
The distribution is not symmetric: losses end at −100%, gains have no limit. A few stocks with very high returns pull the average up, and the typical stock sits below it.
How many stocks beat the market?
According to the Vanguard study from October 2026, of roughly 12,500 US stocks since 1990 only 25% finished above the Russell 3000. 56% returned less than the risk-free rate.
Which stocks created the most wealth?
From 1990 to 2025, ten companies accounted for 31% of the wealth created by the US stock market, led by Alphabet, Microsoft and Apple. The list also includes Altria, Walmart and Exxon Mobil.
What does the Vanguard study "The power of diversification" show?
It shows that a few stocks carry most of the market's return, that a missing winner costs more than an avoided loser gains, and that more holdings shrink the typical gap to the market.
Are 20 or 30 stocks enough for a diversified portfolio?
For volatility usually yes, for return not reliably. Because of skewness, even a few dozen holdings leave the risk of not owning the few big winners.
Literature
- Bessembinder, Hendrik (2018): Do stocks outperform Treasury bills?. Journal of Financial Economics 129 (3), pp. 440–457. doi:10.1016/j.jfineco.2018.06.004
- Bessembinder, Hendrik; Chen, Te-Feng; Choi, Goeun; Wei, K. C. John (2023): Long-Term Shareholder Returns: Evidence from 64,000 Global Stocks. Financial Analysts Journal 79 (3), pp. 33–63. doi:10.1080/0015198X.2023.2188870
- Plagge, Jan-Carl; Lewis, Sebastian (2026): The power of diversification: Why skewness in stock returns makes it so difficult to beat the market. Vanguard Research, October 2026; accessed November 14, 2026. Source
- Sharpe, William F. (1991): The Arithmetic of Active Management. Financial Analysts Journal 47 (1), pp. 7–9. doi:10.2469/faj.v47.n1.7
This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.