New research on GAAP versus non-GAAP earnings

One of Saul’s investment philosophies has been that non-GAAP earnings are more useful than GAAP earnings (especially for high growth companies?).

Well, there is a new research paper that has been making the rounds on this very same spirit. The paper is called “Core Earnings: New Data and Evidence”, and can be found here:

(Registration to this site if free if you want to download and read the full text.)

In the paper, the professors found that adjusting for certain non-recurring items in the GAAP earnings makes it far more predictive than GAAP earnings on a company’s future growth prospect and stock performance. They showed that by taking advantage of Wall Street not processing this information correctly, the investor can harness up to double digit annual outperformance over the benchmark.

One caveat is that the non-GAAP figures used by equity analysts and companies’ own quarterlies are not the optimal way to go about it either. Part of the issue is that the right items to strip or add (according to their research), may not be easily accessible from traditional data sources, although one can dig out such information by going through 10-K’s in painstaking detail, such as in the footnotes.

But isn’t that in essence what we do on this board already (although not in a purely quantitative systematical way, but in a focused way)?

If you want to see what kind of adjustment to the reported earnings can enhance predictability, you may want to download the paper and give it a read.



Thanks Bashuzi,
Interesting paper, and from Harvard Business School as well. Nice that it confirms what is obvious (to most of us anyway).


From Bloomberg recently…

In many cases, accounting choices appear to be distorting results. In early 2019, General Electric Co. reported GAAP losses of $2.43 per share; under adjusted figures it earned $0.65 per share. Tesla Inc. reported full-year GAAP losses of $5.72 per share but “non-GAAP” losses were only $1.33 per share.

Read on……

There’s battle lines being drawn …Nobody’s right if everybody’s wrong

From Bloomberg recently…

The Bloomberg article appeared at the METaR board. This was my reply:

When Buffett was campaigning to abolish stock options I wrote

June 11, 2002
Options Math

Clearly a company paying labor with stock options is more profitable than one paying with money because the cost of labor is transferred to the shareholders by way of stock dilution. Investor Buffett hates dilution. So stock options are now “expensed.” The shareholder is hit with both a GAAP loss and stock dilution. It makes sense for investors analyzing stocks to reverse the expensed options.…

Denny Schlesinger


Here are some excerpts I took from the paper.

On pro forma (or “core”) earnings used by analysts and managers: “The usefulness of these core earnings metrics is hampered by the evidence that managers and analysts selectively choose which items to include”. “Nonetheless, prior research shows that non-GAAP earnings tend to be more value relevant in the sense that they seem to better explain movements in firms’ stock prices”

The authors use a noval dataset constructed from “a combination of human analysts and machine learning to collect and classify quantitative disclosures embedded in the 10-Ks”, including “items reported on the income statement and items hidden in the footnotes or the management discussion and analysis section (MD&A)”. Using this dataset, “Core Earnings adds back to GAAP net income the non-core net expenses related to (1) acquisitions, (2) currency devaluations or revaluations, (3) discontinued operations, (4) legal or regulatory events, (5) pension adjustments, (6) restructuring, (7) gains and losses that companies disclose as other, and (8) other unclassified gains and losses deemed non-operating”

Among earnings adjustments frequently discussed on this board, you should know that the last of these adjustment, specifically include “goodwill amortization and employee stock options
expenses”. So the authors are aware of treating stock-based compensations correctly.

The paper shows that GAAP net income based variables are least predictive of future performance; pro forma figures do better; but deeply adjusted Core Earnings are the most predictive. But importantly, they show that “market participants are slow to take into account the implications
of transitory earnings”.

Here are some additional interesting findings:

“Managers are increasingly likely to report non-operating gains on the face of the income statement but to disclose non-operating losses only in the footnotes or the MD&A.”

Since 1998, “per-share value of adjustments has increased over time”, especially the so-called “hidden adjustment” which are only found in footnotes: “The per-share value of hidden total adjustments increased significantly over time, from $0.015 in 1998 to $0.127 in 2018.” “A significant portion (about half) of total adjustments per share are reported only in the footnotes or the MD&A.”

So please keep digging into those footnotes!