You are working with the Ewens, Peters and Wang intangible capital stocks for U.S. public firms.

DATA: https://github.com/michaelewens/Intangible-capital-stocks/raw/master/intangibleCapital_20260924.csv
One row per Compustat firm and fiscal year. 534,089 rows, 46,082 firms, fiscal years 1975-2026. Release September 24, 2026.
Columns:
  gvkey        Compustat firm identifier (integer)
  fyear        fiscal year
  knowCapital  knowledge capital stock, net, nominal $ millions (capitalized R&D)
  orgCapital   organization capital stock, net, nominal $ millions (capitalized share of SG&A)
  note         blank when computed from reported data; otherwise says that Compustat total assets were not reported (flows reported, interpolated, or missing) or that an earlier year's missing flows leave the stock missing
Stocks are net (after depreciation), so year-on-year changes are net investment. Merge to Compustat on (gvkey, fyear).

PARAMETERS: https://github.com/michaelewens/Intangible-capital-stocks/raw/master/capital_accum_parameters_2023.csv gives, by 4-digit SIC, the knowledge depreciation rate (knowDepr), organization depreciation rate (organDepr, 0.20), the share of SG&A treated as investment (gamma), and the Fama-French 5 industry. Baseline by industry (knowDepr / gamma): Consumer 0.43 / 0.20, Manufacturing 0.50 / 0.21, High-tech 0.42 / 0.37, Health 0.33 / 0.51, Other 0.35 / 0.22.

LOADING: `pip install intangiblecapital` then `import intangiblecapital as ic; stocks = ic.load()` (cached; `ic.load(release="20260924")` pins this release; `ic.merge_compustat(funda)` attaches the stocks to a Compustat frame on gvkey and fyear). Or read the CSV directly with pandas, Stata `import delimited`, or R `read.csv`.

METHOD: perpetual inventory. Knowledge capital: K_t = (1 - knowDepr) K_(t-1) + R&D_t. Organization capital: O_t = 0.8 O_(t-1) + gamma x SG&A_t (SG&A net of R&D). Initial stocks imputed from firm age. Details: https://github.com/michaelewens/Intangible-capital-stocks/blob/master/pipeline/README.md

PAPER: the full author manuscript is in the repository as Markdown: https://github.com/michaelewens/Intangible-capital-stocks/blob/master/paper/Ewens_Peters_Wang_2024_manuscript.md. Read Section 1 for the model and what the stocks measure, Section 2.1 for data and sample, Sections 3 and 4 for how the parameters were estimated and their values, Section 5 for validation against market values, Sections 6.2 and 6.3 for industry and time patterns and the effect on market-to-book and ROE, Section 7 for robustness.

CAVEAT: every release rebuilds all years from current Compustat, so a firm's history can change between releases (industry code revisions, restated financials). Cite the release date. Details: https://github.com/michaelewens/Intangible-capital-stocks/blob/master/STOCK_CHANGES.md

CITATION: Ewens, Michael, Ryan Peters and Sean Wang. "Measuring Intangible Capital with Market Prices." Management Science 71.1 (2024): 407-427. https://doi.org/10.1287/mnsc.2021.02058