Buying What Buffett Buys — A 17-Quarter Replication Test of the Berkshire Portfolio

Institutional investors disclose their holdings each quarter through 13F filings. We tested the oldest copy-trading idea of all — "buy what Buffett buys and you'll earn what Buffett earns" — against the actual filing data.

Test Method

• Data: SEC EDGAR 13F-HR information tables, Berkshire Hathaway, 17 quarters (2022Q2–2026Q2)
• Replication: Top-10 holdings, value-weighted, rebalanced on the trading day after the filing date (quarter-end + 45 days) — based on when we could actually know the information
• Benchmark: Versus SPY, chained across quarters cumulatively

Results

What These Numbers Say

Buffett 45 days late is not Buffett. Two effects overlap here:

1. A 13F shows quarter-end holdings 45 days after the fact. Prices have already moved between the purchase date and the replication date
2. The past four years were a period in which Berkshire itself underperformed the index. Berkshire's returns include an enormous cash position (a buffer), and replicating only the equity positions loses that structure

The common thread running through the copy-trading series (Parts 1–4): what separates success from failure is not "whose trades you follow" but the disclosure lag and the half-life of the information. If the half-life is shorter than the lag, nothing is left by the time the filing appears.

Limitations

• Four holdings that failed ticker mapping (Alphabet share-class distinctions, etc.) were excluded; given their size, this does not change the conclusion
• A value-weighted top-10 replication differs from Berkshire's full portfolio (which includes cash, private holdings, and overseas assets)

Data Sources

• SEC EDGAR 13F-HR filings (original text); prices from our own U.S. equity price database