Lesson 4 of 8
What book value stopped measuring
One company spent 25 billion dollars on research in ten years and carries none of it as an asset. That is the accounting change under the intangibles argument.
By Javier Sanz
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
In one minute
- Spending that builds an internally developed asset is charged against profit as it happens, so it never appears on the balance sheet.
- For a research-heavy company that makes book value a record of what was bought rather than what was built.
- Sorting the market by price to book therefore sorts partly by accounting treatment, which is the core of the intangibles objection to the value premium.
The value premium is usually measured by book-to-market. That measure assumes book value tracks what has been invested in a business. For a company whose investment is research and software, it does not, and the gap is not small enough to ignore.
Ten years of research spending against the book value it produced
- Research spending charged to expense, fiscal 2016 to 202525,094$m
- Total book equity at the end of fiscal 202511,623$m
What the filing shows
Adobe's income statement puts research and development in operating expenses: 4,294 million dollars in fiscal 2025, 3,944 million in fiscal 2024 and 3,473 million in fiscal 2023. None of it becomes an asset. Over the ten fiscal years from 2016 to 2025 the total charged to expense is 25,094 million dollars, which is 2.16 times the entire book equity the company reports at the end of that period.
A company that spent the same money on factories would carry most of it on the balance sheet and depreciate it slowly. The two companies would look completely different by book value while having invested the same amount.
What the sort does to a company like that
For scale, this site's own cohort of 3,300 US companies with filed statements, built 2026-09-15, had a median price to book of 2.04 and a 90th percentile of 9.60. A company moving from 8.8 times to 4.5 times moves from the expensive tenth towards the middle of the market on a convention alone.
The proposed repairs, and their cost
Three answers are offered in the literature and in practice.
What people do about it
-
Capitalise the spending yourself and amortise it. Defensible, and it requires you to invent a useful life that no filing states.
-
Replace book with a measure that does not depend on it. Novy-Marx measured gross profits over assets and found it had roughly the same power to predict average returns as book-to-market, which is a repair by substitution rather than by adjustment.
-
Accept the limitation and use the ratio only where it describes something. Book value still means a great deal for a bank, a property company or a shipping firm, and very little for a software company.
The third is the cheapest and it is also an admission: a value measure that cannot be applied to a large part of the market is a narrower instrument than its reputation.
The rules are still moving
The same Adobe filing notes that the FASB issued an update on internal-use software in September 2025 that modernises the accounting and clarifies when those costs may be capitalised, effective for the company in fiscal 2029. Any argument that rests on where the line between expense and asset falls should carry a date, because the line moves.
Check yourself
Why does research spending not appear on the balance sheet as an asset?
- Because it is not a real cost
- Because it is charged against profit in the year it is incurred, so no asset is recorded
- Because the auditors exclude it
Show the answer
Answer: B.
The spending runs through the income statement as it happens. A company that buys a factory instead carries the asset and depreciates it over years.In the illustrative restatement, what happened to the price to book ratio?
- It rose from 8.8 to 12 times
- It fell from 8.8 times to about 4.5 times
- It did not change
Show the answer
Answer: B.
Adding 11,236 million dollars of unamortised research to equity roughly doubles book value and roughly halves the ratio, with no change to the business.What did Novy-Marx propose in place of book-to-market?
- Dividend yield
- Gross profits over assets, which he found had roughly the same predictive power
- Price to sales
Show the answer
Answer: B.
His measure sidesteps the book value problem rather than adjusting for it, and he reports similar power in predicting the cross-section of average returns.Sources
- Adobe Inc., Form 10-K for fiscal 2025, filed 2026-01-15, accession 0000796343-26-000003, and the fiscal 2016 to 2024 filings for the research expense series. Read from the SEC XBRL company facts API on 2026-09-13; the fiscal 2025 filing text was read for the operating expense table and the note on the September 2025 accounting update for internal-use software.
- Novy-Marx, "The Other Side of Value: The Gross Profitability Premium", Journal of Financial Economics, 2013.
- This site's published price to book distribution, cohort of 3,300 companies with filed statements, built 2026-09-15.
What this leaves out
- The five-year straight-line capitalisation is a convention chosen for this page. No filing states a useful life for research, and a different life gives a different restated book value.
- The restated ratio uses the stored market capitalisation from the date shown, not a live quote.
- One company cannot establish a market-wide effect. It shows the mechanism at a size that can be checked against a filing.
- The price to book cohort is one deduplicated row per company with filed statements on record, with exact zeros excluded as coerced missing data and the tails clamped at the first and ninety-ninth percentiles before the breakpoints were computed.
- Nothing here is a view on this company or on any security.