From the paper...

"A large corporate finance literature documents a secular shift toward intangible capital and R&D-intensive firms in the U.S. economy (e.g., Falato et al. (2022); Begenau and Palazzo (2021)). Atkeson et al. (2026) provide a macroeconomic perspective on these developments, showing that the decline in measured investment and changes in the composition of capital, including the growing role of intangible investment, mechanically lower earnings yields even when free-cash-flow yields remain stable."

So intuitively this makes sense in some ways to modify R&D expense for the rise of intangible investment.

"Accounting conventions may systematically distort the comparison between historical and contemporary reported earnings, making CAPE appear elevated when comparability has been compromised rather than when fundamentals have shifted or markets have become overvalued. We demonstrate that correcting these measurement distortions reveals current valuations as cyclically elevated within historical ranges, rather than representing either a permanent regime shift or unprecedented overvaluation."

Okay makes sense as well. But do the authors take the next step and reconsider R&D expense across the entire historical range of S&P 500 returns?

So prior to 1975 FASB allowed capitalization of R&D expense, after 1974 FASB mandated immediate expensing of R&D. So only going back to 1974 to remove the R&D expense distortion.

"Second, accounting standards changes in the early 1990s—notably Statement of Financial Accounting Standards (SFAS) 121 (1995) on asset impairments—encouraged larger, more frequent recognition of special items."

So going back to 1995 to remove distortion from more frequent special item recognition, which frequently wrotedown assets on temporary effects that does not recognize the true long-run earning ability of an asset (see Precision CastParts writedown in 2020(?)).

These accounting adjustments seem to make sense as Charlie Munger often opined there are limitations to accounting and it's only a crude approximation of reality.