The missing balance sheet
Accounting expenses research and most of the spending that builds an organization, so the assets they create never appear on a balance sheet. Using the published parameters from Ewens, Peters and Wang (2023), this page rebuilds those assets as capital stocks — and extends them to the present.
Look up a company
The 2,000 largest holders of intangible capital, covering 98% of the total. Figures are millions of nominal dollars; the sparkline runs 2018 to the firm’s last available year. Off-book share is intangible capital as a fraction of that capital plus reported assets — the portion of a firm’s asset base that never reaches its balance sheet. Screen by size before reading the ranking: shell companies with almost no reported assets sit at 100% and mean nothing.
| Company | Industry | Off-book intangibles |
% of assets off-book |
2018–2025 chg in intangibles |
Trend |
|---|
Where the invisible assets are
The published parameters differ sharply across the five Fama-French industries. Health capitalizes half of overhead as organizational investment; consumer firms, a fifth.
Total of knowledge and organizational capital, nominal dollars.
| Industry | δ knowledge | δ organization | γ SG&A share | median off-book |
|---|
δ is the annual depreciation rate; γ is the share of adjusted SG&A treated as investment in organizational capital. Fixed at the published values — not re-estimated here. The final column is the median off-book share among firms with over $200m in reported assets.
A constant panel of firms, 2018 to 2025
Same 3,159 firms in every year, so the change is growth rather than a shifting sample. Nominal dollars overstate it — the deflated series is the honest one, and both are shown.
Forty-five years of assets nobody books
The published EPW series runs to fiscal 2018; everything after is a disclosed extension built here with the same fixed parameters. The firm count moves over time, so read this as the scale of the phenomenon rather than a like-for-like growth rate.
How this was built
The recurrence
For each firm and component, real capital accumulates as K_t = (1 − δ)·K_(t−1) + I_t, with flows deflated by CPI before accumulation and stocks re-inflated to nominal dollars for output. Knowledge investment is R&D. Organizational investment is the industry γ times adjusted SG&A, following the submitted definition including the R&D-in-cost-of-goods overlap rule. Parameters are assigned on the historical SIC, filled from the company header SIC. Regulated, financial and public-service firms are excluded, as in the original construction.
Validation
Before extending anything, the recurrence was used to reproduce EPW's own published stocks for 2011–2018 from Compustat flows. Correlation with the published values is 0.9999 for both components; median absolute error is 0.04% for knowledge capital and 0.01% for organizational capital; the aggregate across all matched firm-years differs by +0.09%. As a further check, the CPI deflator was recovered from the published stocks by solving the recurrence, and matches the FRED CPIAUCSL annual-average series to within 0.14% in every year.
Limitations
The extension covers only firms with a published FY2018 stock — 5,067 of them. Firms that listed after 2018 are absent, because seeding them would require their full spending history. Firms exit the panel as they delist or are acquired, so the all-firms aggregate is not a like-for-like series; the constant-panel chart exists for that reason. Fiscal 2025 is included: 86% of its firms have July–December year-ends, against 85% in 2022–24, so late filers are not systematically missing and it is not a partial year. This is a conditional reproduction of the submitted method, not a claim of full historical replication: original-universe and header-SIC coverage were not fully recovered.
Sources and reuse
Parameters and stocks through 2018: Ewens, Peters and Wang, Measuring Intangible Capital with Market Prices, Management Science, published materials under CC BY-NC 4.0. Deflator: FRED series CPIAUCSL, annual averages. Underlying accounting inputs are licensed and are not redistributed here — only derived capital stocks are published. Firm identifiers are shown by name and ticker rather than by internal database key.