Evidence on Buffett's Stock Selection and Subsequent Influence

Bowen, Rajgopal and Venkatachalam find that Berkshire’s listed investments often already match some of Buffett’s publicly stated principles before purchase. Evidence of further changes after investment is weaker.

Source dates and applicability

This article distinguishes the May 31, 2011 working paper from the 2014 journal publication. Public equity holdings cover 1980–2006. As of October 2026, the evidence concerns historical selection and subsequent influence, without testing later management practices or wholly owned subsidiaries. Working-paper-only portfolio results retain their separate version attribution.

The researchers turn Buffett’s publicly stated accounting, governance, and investment principles over the years into testable variables, then examine two questions separately:

1. Before Berkshire buys them, do these companies already look more like the businesses Buffett prefers?
2. After Berkshire becomes a shareholder, do these companies move further toward Buffett’s principles?

The conclusion is clear: Buffett’s stock selection criteria align closely with his actual holdings, but evidence that he transforms companies after investing is weaker.

1. How the Paper Tests This

The paper’s identification strategy has roughly three layers.

First, the authors group Buffett’s public statements into three categories: accounting and disclosure, corporate governance, and investment and financing decisions.

Second, they use Berkshire’s publicly traded stock holdings from 1980 through 2006 to construct a sample of 624 firm-years across 206 companies. They compare this with a control sample to see whether the companies already had more of Buffett’s preferred characteristics before purchase.

Third, they retain only companies in which Berkshire held at least a 5% stake for at least one year, comparing changes before and after investment to identify whether Buffett actually had a lasting influence on corporate behavior.

To reduce the influence of industry and year differences, many regressions include industry and year fixed effects, with standard errors clustered by company.

2. The Core Finding: Strongest at Selection, Weakest at Transformation

The findings can be summarized as follows: Buffett is strongest at selecting companies before investment, while subsequent corporate transformation is weaker, especially in governance.

In the summary in section 9 of the authors’ 2011 working paper, Berkshire’s portfolio companies match Buffett’s preferences on 8/10 investment and financing characteristics. The corresponding figures are 4/7 for accounting and disclosure and 2/6 for executive compensation. Board structure and external monitoring are further removed from the principles Buffett publicly emphasizes.

More specifically, companies Buffett buys often already have these characteristics: higher ROE and owner earnings, more persistent advantages in sales and earnings, lower leverage, lower volatility, and share prices closer to their estimated intrinsic value.

In accounting disclosure, these companies are also more likely to reflect good and bad news promptly, have lower abnormal accruals and more conservative pension assumptions, and voluntarily expense stock options before doing so becomes mandatory. However, they are not significantly less likely to disclose non-GAAP adjusted earnings or issue earnings guidance, and their annual reports are not noticeably more readable.

Among the measured indicators, operating and investment characteristics align more closely with Buffett’s public principles. This does not establish an exact ranking of his personal selection priorities.

3. What This Paper Tells Us

Version note: The mimicking-portfolio and future operating-performance analyses in the next two paragraphs appear in the authors’ May 31, 2011 working paper, but were not found in the cited 2014 published version. They should be read as earlier-draft findings, rather than verified results of the published paper.

The authors also construct a portfolio that imitates Berkshire, using similarity measures to find companies statistically close to Berkshire’s holdings. This portfolio outperforms the market benchmark, but its excess returns under the Fama-French four-factor framework are unstable and, in particular, are not significant from 1997 through 2006.

The portfolio does, however, predict better operating performance over the next 1 to 5 years, with significantly greater improvement in operating income before depreciation still evident after five years. In other words, these signals are better suited to helping investors select companies with higher long-term operating quality; they may not consistently generate excess returns.

The paper also has several clear limitations.

First, it studies publicly traded stock holdings, excluding Berkshire’s many privately held controlling investments and acquisitions of entire companies. Buffett’s influence may be much stronger in wholly owned acquisitions.

Second, the sample ends in 2006, so the conclusions primarily reflect Buffett’s style of investing in publicly traded equities during that period.

Third, most governance variables the authors can quantify are observable institutional arrangements. They struggle to capture the soft information Buffett values most, such as whether management is honest, capital allocation is skillful, or corporate culture is healthy. These factors may be among the most valuable parts of his investment ability.

4. References

1. Robert M. Bowen, Shivaram Rajgopal, and Mohan Venkatachalam (2014), Is Warren Buffett’s Commentary on Accounting, Governance, and Investing Practices Reflected in the Investment Decisions and Subsequent Influence of Berkshire Hathaway?, The Accounting Review 89(5): 1609–1644; university publication record. Section III supports the 624 firm-years and 206 companies; sections IV–VI examine accounting, governance, and investment characteristics; section VII compares selection with subsequent influence. This published version does not contain the earlier draft’s mimicking-portfolio analysis below.
2. Robert M. Bowen, Shivaram Rajgopal, and Mohan Venkatachalam (2011), working paper of the same title, May 31, 2011 draft, hosted by the author at the University of Washington. Section 9 supports the 8/10, 4/7, and 2/6 characteristic counts; section 8.2 and Table 8, Panel C support the mimicking-portfolio returns; section 8.3 and Table 9 support operating performance over the next 1–5 years. Section 3’s first two substantive paragraphs draw on this earlier version, not empirical results in the 2014 published paper.