13F Portfolio Analysis — Warren Buffett | Tracking 20 Years of Investing Style

How Warren Buffett’s portfolio style shifted over twenty years

“Warren Buffett is a value investor.” Almost everyone has heard it. But look at his actual portfolio over 20 years and the story turns out a little different.

We split Berkshire Hathaway's 13F into three periods — 2001, 2011 and 2024 — and analyzed each the same way. The results were surprising: Buffett's market beta, value beta and degree of diversification all differed from era to era.

In 2001 he was extremely concentrated in Coca-Cola and American Express; by 2011 the book had broadened across financials, consumer staples and IBM; and in 2024 it concentrated again, this time around Apple.

So did Buffett abandon value investing? It may be the opposite.

What changed was not a particular factor, but “how he puts conviction behind good businesses.”

In this piece we use public 13F filings to track that change in numbers.

US institutional investors above a certain size disclose the US-listed stocks they hold at each quarter-end on Form 13F. That lets us see, every quarter, which US-listed stocks Berkshire Hathaway holds and how much.

On SEC EDGAR you can read Berkshire Hathaway's 13F-HR filings and position-by-position holdings directly, while services like Dataroma and WhaleWisdom reorganize the same data around weights and changes.

But 13F alone doesn't show all of Buffett's assets. Cash, bonds, private companies and foreign-listed shares are left out, and short positions and derivatives don't show up as they are. And because it's a quarter-end snapshot, positions may already have changed by the time it's filed.

So there are limits to treating 13F as “a list for copying Buffett's trades.” Still, there is something 13F does show.

Which companies Buffett was putting how much weight behind at that moment — and what kind of portfolio that made. That is exactly what this analysis focuses on.

To see how Buffett's investing style changed, we picked three points roughly six years (72 months) apart. 2001 was close to classic value-and-quality investing; 2011 was when he held consumer staples and financials broadly and newly added IBM; and 2024 stands out for its Apple concentration and the addition of energy.

For each point we built one portfolio from the 13F holdings at their market-value weights at the time, and fed the month-end adjusted prices of those stocks over the corresponding period into DIVA Quantizer.

There is one important processing rule here. Rather than including every holding in each era, we used only names with a continuous 72-month price history. Companies acquired or split up long ago, or lacking price data for the period, were excluded, and weights were rescaled to 100% among the remaining names.

That's why Era I became a highly concentrated portfolio of just four names, while Era II, at nine, is the most spread out. Now let's compare the three portfolios with the same yardstick.

The Fama–French five-factor model shows how closely a portfolio's returns move with five common factors: market, size, value, profitability and investment. Put Buffett's three eras through the same model and the change is sharp.

Surprisingly, “value” was not strongest in 2001

You'd expect Buffett's textbook value investing to show up most strongly in 2001. Yet value beta (HML) was only +0.16 in 2001. The strongest reading was actually 2011's +0.66.

The 2001 portfolio looked more like a heavy concentration in a few companies with strong brands and competitive strength, such as Coca-Cola and American Express.

In 2011, by contrast, the portfolio broadened to Coca-Cola, Wells Fargo, American Express, P&G, IBM and others — and this is when a value tilt in the academic sense appeared most clearly.

In other words, “Buffett = undervalued stocks” alone can't explain all three eras.

Market beta shows a similar shift: 1.15 in 2001, 0.83 in 2011 and 1.17 in 2024.

The 2011 portfolio was the most defensive — high value beta, low market beta. It's the picture of buying good companies relatively cheaply while swinging less than the market.

In 2024, though, market beta rose back above 1. As Apple's weight grew, the portfolio became more sensitive to market moves, and value beta all but disappeared.

Still, this can't simply be read as “Buffett abandoned value investing.” What this analysis does confirm is that Buffett's portfolio did not stay parked on a single “value factor.”

In 2001 he concentrated in a few quality names; in 2011 the traditional value-and-defensive tilt strengthened; and in 2024 a mega-cap quality concentration centered on Apple emerged.

The factor analysis tells us that Buffett's portfolio changed. But how did actual risk and return change? For that we applied Markowitz analysis.

What's interesting is that the portfolio didn't simply get more diversified over time.

Volatility was 24.4% in 2001. In 2011, as the number of names grew, it fell to 18.8% and the diversification effect rose to 28.8%.

Then in 2024 the Apple concentration intensified again, and the diversification effect slipped to 26.9%, below 2011. Yet the Sharpe ratio was the highest of the three eras, at 0.91.

So the 2024 portfolio is not one whose risk shrank. It's closer to a portfolio that accepts risk, but concentrates it in companies worth carrying it.

Plotting today's Era III portfolio directly on the efficient frontier shows at a glance where Buffett stands.

Put the three periods side by side and a pattern emerges.

2001 — Strong conviction in a few companies

Coca-Cola and American Express made up most of the book. The portfolio was highly concentrated, with a diversification effect of just 20.2%. The window also included the dot-com crash, so the Sharpe ratio ended up low at 0.26.

Conviction in good companies was strong, but the portfolio structure was rough.

2011 — The most textbook value-and-defense portfolio

The portfolio broadened to Coca-Cola, Wells Fargo, American Express, P&G, IBM and others. Value beta was the strongest at +0.66, market beta the lowest at 0.83, and the diversification effect the highest of the three periods at 28.8%.

This was the period when value, defense and diversification combined best.

2024 — Concentrated again, but this time in large-cap quality

Top-holding weights grew again — Apple 31%, American Express 19%, Bank of America 13% and so on. Market beta was 1.17 and value beta +0.09. Yet the Sharpe ratio was the highest, at 0.91.

In other words, Buffett didn't keep adding more names over time — he came back to picking the companies he's convinced of and putting more weight behind them.

The most interesting part of this analysis is, if anything, the metrics where consistency seems to vanish.

Value beta: +0.16 → +0.66 → +0.09.
Market beta: 1.15 → 0.83 → 1.17.
Diversification effect: 20.2% → 28.8% → 26.9%.

None of them moved in a steady line. So what does explain Buffett's investing style?

The common thread this 13F analysis confirms is not a single factor but that he concentrated capital in a small number of companies for the long term.

In 2001 he concentrated in companies with brands and competitive advantages; in 2011 he held value-and-defensive companies more broadly; and in 2024 he concentrated again in mega-cap quality like Apple.

So rather than describing him only as “Buffett buys value stocks,” it is better to put it this way:

“He puts heavy, long-term weight behind quality companies he judges he understands.”

That reading explains all three eras better.

First, 13F is closer to a tool for reading style than for copy-trading. There's a lag of up to 45 days, and positions may change after filing. You can't tell what price Buffett actually paid, either.

What you can see instead is which companies he concentrated capital in, how sensitive the portfolio was to the market, and how tilts like value, profitability and size shifted.

Second, sitting below the efficient frontier doesn't automatically make it a “wrong portfolio.” Markowitz's efficient frontier is a model computed from past returns and volatility.

Especially when the window includes a stretch like the one ending in 2024, when Apple surged, past performance can pull the frontier upward. So a dot sitting below the curve can't, on its own, judge the quality of investment decisions.

What's more, Markowitz looks at past risk and return. It isn't a model that accounts for management, business moats, long-term holding intent or taxes.

Third, this analysis isn't just for Buffett. Apply the same models to public holdings and price data and you can look at your own portfolio the very same way.

You can check how exposed you are to value stocks, how closely you move with the market, and whether — despite holding many names — you're actually piled into similar risks.

In the end, the best way to read a 13F is not to stop at “What did Buffett buy?” but to go on to “Why did the portfolio take this shape?”

Sum Buffett up in one sentence and people usually say “value investor.” But break 20 years of his portfolio down into numbers and a slightly different picture appears.

The 2001 Buffett concentrated hard in a few quality names;
the 2011 Buffett combined value, defensiveness and diversification;
the 2024 Buffett concentrated again — but this time in mega-cap quality.

A lot changed. Market sensitivity, value exposure and the degree of diversification all differed by era. But one thing remained.

Allocating capital, for the long term, to companies he judges he understands.

That's the fun of reading a 13F. Not just seeing what a famous investor bought, but taking apart the structure of their portfolio yourself — and, finally, turning that same yardstick on your own portfolio.

DIVA Quantizer lets you check exactly that process in numbers.

Sources: SEC EDGAR 13F-HR (Berkshire Hathaway) and Berkshire Hathaway annual reports / shareholder letters for the relevant years. Prices are yfinance month-end adjusted closes, analyzed over 72 months each: Era I 1997–2002, Era II 2008–2013, Era III 2019–2024. Factors: Ken French Data Library North American FF5. All factor and risk metrics are outputs of DIVA Quantizer's Fama–French and Markowitz models.

This article was written with the help of AI.