Key takeaways
- Just 42.6% of 25,967 US common stocks beat one-month Treasury bills over their lifetimes from 1926 to 2016, and the median lifetime return was minus 2.29%.
- Meulbroek estimates an employee sacrifices 42% of company stock market value at a ten-year horizon, rising to 80% for a fully undiversified fifteen-year holder.
- Among Vanguard plans offering company stock, 11% of participants held more than 20% of their balance in it during 2024, down from 28% in 2015.
- Concentrated households beat diversified ones by just under 2 percentage points a year in one 1991-96 study, yet their Sharpe ratio was 0.12 against 0.17.
The email lands on a Tuesday. Your shares have vested, and there's a sell button beside them. You've been at the company four years now. Your salary comes from it, your pension holds it, the staff share plan has quietly stacked up more of it than you'd guess without logging in — and the price has been kind, or you wouldn't be looking. Do you press the button?
Most people don't. Selling feels like a vote against the place that pays you. So before you decide, what does the record say about holding one company's stock?
It says the typical single stock loses money. Of the 25,967 US common stocks in the CRSP database between July 1926 and December 2016, only 42.6% beat one-month Treasury bills over their lifetimes. Just 49.5% made any money at all. The median lifetime buy-and-hold return was minus 2.29%.
That's Hendrik Bessembinder's finding, published in the Journal of Financial Economics in 2018, and it's the most useful single number for anyone whose salary and savings both ride on one company. The equity premium is real. It just isn't evenly spread, and the typical stock never earns it. Your employer's shares are one draw from that distribution, and the middle of it is a loss.
What follows is the decision a concentrated holder faces, not the question of how many stocks make a portfolio broad. In order: the lifetime record on single stocks, the modelled price of concentration, how concentrated people actually are, and the strongest arguments for keeping your finger off the button.
What the lifetime record actually looks like
Bessembinder measured every common stock listed on the NYSE, Amex and Nasdaq from July 1926 to December 2016. Each one runs from its first appearance to its last, with dividends reinvested. The distribution is lopsided in a specific way.
Only 30.8% of those stocks beat the value-weighted market over their lifetimes. Just 26.1% beat the equal-weighted market. The most frequent lifetime outcome, once returns are rounded to the nearest 5%, is a loss of 100%. Listings are short, too. The median stock sat on the database for seven and a half years.
Delisting is where the damage concentrates. Among stocks removed by an exchange, the median lifetime return was minus 91.95%. Only 9.8% finished positive, and only 6.8% beat Treasury bills.
Newer listings fared worse than older ones. Of the stocks that entered the database between 1977 and 1986, just 31.7% beat bills over their lives. The median lifetime return is negative for every entry cohort since 1977. Whatever your employer's founding date, it sits in one of those recent cohorts rather than the 1947-56 group, where 87.0% beat bills.
Bessembinder also ran a bootstrap. A strategy that held one randomly selected stock at a time from 1926 to 2016 underperformed the value-weighted market in 96% of simulations. It underperformed one-month Treasury bills in 73% of them.
That's the base rate your position starts from. It isn't a forecast for any particular company, and it doesn't say your employer is the median stock. It says the median is where the odds sit, and that arriving through a payroll doesn't make a share a better draw than one bought on the open market.
Two caveats belong right next to those figures. The sample is US-only and stops at the end of 2016, so lifetime returns for stocks still trading then are truncated rather than finished. And the mean excess return across stocks is positive. It's the median that's negative. Skewness is doing the work, not a missing risk premium.
The bad outcomes don't announce themselves
Fine, you might think, but you'd see trouble coming. You work there. Would you?
J.P. Morgan's Michael Cembalest has tracked what he calls catastrophic declines since 2004: a stock that falls 70% from its peak and doesn't recover. The October 2024 update looks at decliners since 2021, and the pattern in it is uncomfortable.
Most of the catastrophic losers traded on reasonable forward price-to-earnings multiples at their peak. Most had positive profit margins. Most carried net debt below two times EBITDA. Analyst consensus at those peaks tilted heavily towards buy and strong buy, with around half the names still showing projected upside to price targets. Healthcare and biotech contributed the largest number of decliners.
The names include Moderna, Estee Lauder, PayPal and Silicon Valley Bank. None of them looked distressed at the top, and plenty of people inside those firms were reading the same numbers you'd have been reading. That's the conclusion Cembalest draws: event risk for an individual company is close to impossible to anticipate reliably.
This is a bank's own analysis rather than peer-reviewed work, and the 2024 edition profiles a selected set of names instead of reporting a full-universe tally. It illustrates a mechanism. It doesn't establish a base rate.
Why employer stock concentration isn't just another single-stock risk
A concentrated position in a company you don't work for exposes your savings. A concentrated position in your employer exposes your savings and your income to the same shock. Picture the quarter it goes wrong: the redundancy round, the suspended bonus, the frozen share plan and the falling price all tend to arrive together.
Lisa Meulbroek put a price on that in a Harvard Business School working paper, later published in the Journal of Law and Economics. She asked what company stock is worth to an employee who can't diversify it, against its market price. Her method finds the price that would give an undiversified holder the same Sharpe ratio as the market.
The headline figure is 42%. That's the share of market value given up by an employee holding half their pension in company stock, with pension assets making up half of total wealth, over a ten-year horizon.
Say your plan statement shows a million in company stock. Round figures, purely as an illustration — but in her framing that million is worth roughly 420,000 to you, the person who isn't allowed to spread it. Nothing on the statement says so. That 42% is a modelled private value rather than a realised loss, and no line item anywhere shows it as a deduction, which is exactly why the cost is easy to carry without noticing.
Time held matters more than almost anything else in her numbers. Here's the same model at different exposures and horizons.
| Exposure to company stock | Horizon | Market value sacrificed |
|---|---|---|
| Company stock only | Three years | 33% |
| Company stock only | Ten years | Mean of 68% |
| Company stock only | Fifteen years | Mean of 80% |
| 12.5% of total wealth | Ten years | 27% |
| The same, for an average NYSE firm | Ten years | 16% |
Read down the horizon column and the cost roughly doubles between three years and fifteen. Read the last two rows and a second lever appears: shrinking the position's share of your total wealth does as much work as shortening how long you hold it. Both are things you control. Neither requires a view on the share price.
Two caveats on the model itself. The inputs are 1998 CRSP return data, when the average NYSE firm carried 45% annual volatility against 22% for the diversified market, and internet firms ran roughly five times market risk. And it assumes no dividends across the holding period.
How much employer stock concentration people actually carry
All of that is modelled. So how many people are actually paying it, and where would your own balance sit among them?
Vanguard's How America Saves 2025 covers its defined contribution book through 2024. Only 8% of its plans offer company stock, but large plans dominate, so roughly one in three participants has access. Across all participants, 93% hold none, 5% hold between 1% and 20% of their balance in it, and 2% hold more than 20%.
Inside plans that do offer it, 11% of participants hold more than 20% of their balance in company stock and 2% hold more than 60%. That 11% is down from 28% in 2015. The direction of travel is clear. The tail is still there, and if you're reading this you're probably in it.
Plan design drives much of it. Where the employer contributed in cash, plans averaged 7% of assets in company stock. Where the employer contributed in company stock, the average was 18%. Vanguard's own reading is that participants treat the employer's choice as an endorsement, and that most of them see company stock as safer than a diversified equity fund.
Academic work on the same question is sharper, because it can compare two firms that differ in one respect. Choi, Laibson, Madrian and Metrick found a similar pattern in transaction data from three large plans covering 94,191 participants between 1992 and 2000. At the firm that matched in company stock, holdings averaged 31.5% of the portfolio. At a comparable firm matching in cash, the figure was 8.1%. Their behavioural result is subtler than the usual story: participants chased returns when setting contributions, but sold into strength when they actually traded.
Both datasets are US retirement plans. If you're in the UK you accumulate employer shares through SAYE schemes, share incentive plans and vesting RSUs instead, so the plumbing differs even where the exposure doesn't.
The case against diversifying, taken seriously
So why would anyone hold on? Because there's a real case on the other side, and Bessembinder's own data supports it. He calculates lifetime wealth creation of $34.82 trillion across roughly 25,300 companies. The top 1,092 firms, slightly more than 4%, account for all of the net gain. The top five account for 10.07%. Exxon Mobil alone created $1.004 trillion, or 2.88% of the total.
The critics have a point, then. Broad diversification guarantees you'll hold the 4% in trivial size. Anyone who sold their employer's shares on a schedule through the 1990s at Microsoft gave up an enormous amount, and knew it afterwards.
There's also direct evidence that some concentrated investors earn their concentration. Ivkovic, Sialm and Weisbenner studied 78,000 households at a US discount broker from January 1991 to November 1996. Households holding one or two stocks outperformed households holding three or more by 0.16 percentage points a month after a four-factor risk adjustment, just under two points a year. The gap was zero in S&P 500 names, 50 basis points a month in non-S&P 500 names, and 112 basis points a month in non-S&P 500 stocks headquartered within 50 miles of the household. That looks like an information advantage, and it survives controlling for household fixed effects.
The same paper reports the cost honestly. Concentrated households ran monthly standard deviation 4.5 percentage points higher than diversified ones. Their average Sharpe ratio was 0.12 against 0.17. Households with portfolios under $25,000 showed no significant edge at all. And the sample is one broker over six years in the 1990s.
Two things weaken that objection once you apply it to your own vesting shares. First, the edge Ivkovic and co-authors found sits in stocks a household chose and could research, not in a position that arrived through a match or a vesting date — nobody picked it. Second, the employer-stock literature keeps failing to find predictive skill, because allocations track past returns rather than future ones. Goetzmann and Kumar, working on more than 40,000 accounts at the same broker over 1991 to 1996, found most investors under-diversified naively, assembling holdings without regard to the correlations between them. Ivkovic and co-authors summarise the published version of that paper as concluding investors pay considerable costs for those choices. Where those shares are actually held is a separate question, and in nominee accounts the legal owner is not you.
What a diversification schedule solves that employer stock concentration doesn't
A schedule is a pre-commitment device. It converts a repeated judgement call into a rule you set once, which is the same logic behind what the evidence shows about written investment policy statements. The value isn't in the selling. It's in removing the question of when — the question that turns up in your inbox every vesting date and never gets easier.
That matters because the evidence on holder behaviour isn't flattering. Choi and co-authors show contribution decisions following recent returns. Vanguard finds concentrated holders describing their employer's stock as safer than a diversified fund. Neither group is reasoning from the distribution described above.
Meulbroek's numbers show where the lever sits. Her cost of concentration climbs steeply with time held, from 33% at three years to 80% at fifteen. A schedule that shortens your average holding period does more work than one that shaves a percentage point off the weight.
The constraints are real and they aren't uniform. Vesting cliffs, dealing windows, lock-ups after a listing, tax on disposal and the loss of matched shares all shape what's possible for you. J.P. Morgan's 2024 note runs through the usual mechanics for larger positions — staged sales, hedging, exchange funds, gifting — and the tax treatment of each varies by jurisdiction. We priced those routes against each other separately, modelling what an exchange fund actually costs next to a staged sale under the 2026 to 2027 UK rules.
Concentration is also easy to undercount. Imagine your employer's shares, a sector fund and an index tracker all pointing at the same industry: that's the mechanism behind how five popular funds ended up 39% in ten companies. LedgerTouch reports position weight against total portfolio value for that reason.
What would change the conclusion
Several things, and they're worth stating precisely.
Position size relative to your total wealth changes the arithmetic more than anything else does. Meulbroek's cost falls from 68% of market value at full concentration to 27% at 12.5% of total wealth, both over ten years. A holding worth 5% of your net worth is a different object from one worth 60%, and the second is where the difference between risk tolerance and risk capacity starts to bite.
A demonstrable information edge changes it too. Ivkovic and co-authors found one, concentrated in smaller local names among households with enough capital to diversify if they wanted to. If you can show that edge in your own record over years rather than quarters, the median-outcome argument weakens for you. Most people can't.
Your objective changes it. If what you want is a small chance of a very large outcome rather than a reliable median, the skewed distribution is a feature rather than a flaw. That's a coherent position as long as the downside is survivable, which usually means separating the bet from the money that pays the mortgage and from sizing an emergency fund against income volatility.
New data would change it as well. Bessembinder's sample ends in 2016 and covers only US listings. Meulbroek's volatility inputs come from 1998, near a volatility peak, so her cost estimates are probably high for a stable large-cap employer today. The Ivkovic sample ends in 1996, before decimalisation and online broking reshaped retail trading. Vanguard's figures describe its own recordkeeping book, not the whole market.
What none of that touches is the correlation at the centre of the problem, and it's the one that follows you home. Salary, bonus, share plan and pension balance all keying off one company's cash flows is a structural exposure. It's the one thing you hold that a diversified outside investor doesn't — which is what makes that Tuesday email a different decision from any other trade you'll place.