Do Stocks Beat Treasury Bills? A Hundred Years of the Answer

The mean US stock made a fortune and the median one lost money. Both are true, and the distance between them is the whole case for owning the market rather than picking from it.

41.17%
Of stocks beat T-bills over their own lives
−6.87%
Median lifetime buy-and-hold return
1,082
Firms behind all $90.96T of net wealth

The short answer: most individual stocks do not

Hendrik Bessembinder's One Hundred Years in the U.S. Stock Markets (2026-03) measures every one of the 29,754 common stocks in the CRSP database over 1926-2025, each over its own listed life. The result is the most quoted finding in modern market research, and it is not close:

Both numbers are correct and they describe the same population. The gap between them is the finding: returns to individual stocks are so positively skewed that the average is set by a handful of names and tells you almost nothing about the one you picked.

See the three hurdles charted →

So where did the money come from?

Over the century, US stocks created $90.96 trillion of net shareholder wealth above what Treasury bills would have returned. Sort all 29,081 firms from worst to best and the accounting is stark:

Concentration inside that top group is just as steep: Apple alone is 5.52% of the century's net wealth creation ($5.02 trillion), the top 5 are 21.4%, the top 10 are 29.0%, and the top 30 are 43.7%.

See the waterfall →

How few firms it takes, and how that changed

The table reads: how many firms, ranked by wealth created, you need before you have accounted for each share of the total. The middle column is the same calculation on the paper's earlier sample ending 2016; the last is 2017—2025 on its own.

Share of net wealth creation1926—20251926—20162017—2025
10% of it252
25% of it8204
50% of it468913
75% of it20819577
100% of it1,0821,088411

Half the wealth of a hundred years came from 46 companies. On the sample through 2016 it took 89; in the 9 years since 2017 alone it took 13. The distribution is getting narrower, not wider — Nvidia by itself is 9.32% of everything created between 2017 and 2025.

See the concentration ladder →

Is this just a long-horizon artefact?

No. Bessembinder repeats the exercise over ten non-overlapping decades, pooling every stock present at each decade's start:

DecadeMedian returnMean returnBeat T-billsBeat the market
1926—1935−24.53%+19.39%36.08%33.48%
1936—1945+132.08%+198.32%89.38%57.32%
1946—1955+144.86%+225.37%86.59%30.37%
1956—1965+69.32%+126.84%74.65%43.53%
1966—1975−27.07%−0.75%21.54%33.91%
1976—1985+87.20%+333.44%59.05%48.84%
1986—1995+8.08%+84.67%45.95%30.27%
1996—2005+11.24%+87.52%48.89%36.39%
2006—2015+2.52%+45.29%49.00%36.26%
2016—2025+1.44%+71.20%47.70%24.34%

Across all ten decades pooled (62,348 stock-decades), the median ten-year return is +11.56% while the mean is +112.38%, 48.99% beat bills and 35.62% beat the market. The median stock's ten-year return averages 63.6% across the first 6 decades and 5.8% across the last 4 — and in the decades beginning 1926 and 1966 the median stock lost money outright.

See the decade ledger →

What this does and does not imply

FAQ

Do most stocks beat Treasury bills?

No. Across 29,754 US common stocks measured over their own lifetimes (CRSP, 1926–2025), only 41.17% beat one-month Treasury bills. The median lifetime buy-and-hold return was −6.87%.

How many companies created all the US stock market's wealth?

1,082 of 29,081 firms — 3.72% — account for the entire $90.96 trillion of net wealth created above Treasury bills. The other 27,999 firms cancel out.

If the median stock loses money, why does the index go up?

Because returns are positively skewed: the mean lifetime return is 30,621% while the median is −6.87%. The value-weighted market turned $1 into $15,041 over the century, and an index holds the winners by construction.

All figures from /api/static/bessembinder-2026.json, a transcription of the printed tables in Hendrik Bessembinder, One Hundred Years in the U.S. Stock Markets (2026-03-21), computed on CRSP common stocks, 1926-2025. A static edition: it changes only if the paper is revised.

Further reading