Why 10-Ks keep getting longer

Last updated: August 19, 2026

Between 1996 and 2013 the median 10-K roughly doubled in length, and research attributes most of that growth to new disclosure requirements rather than to companies volunteering more information. Three topics account for virtually all of the increase: fair value measurement, internal controls, and risk factors. Filings did not get twice as informative; they got twice as long.

The statistic comes from one of the most thorough studies of 10-K text ever run, and it is worth understanding properly, because it explains why reading a modern annual report feels the way it does.

What did the study actually find?

The study is Dyer, Lang and Stice-Lawrence, The Evolution of 10-K Textual Disclosure: Evidence from Latent Dirichlet Allocation, published in the Journal of Accounting and Economics in 2017 (doi:10.1016/j.jacceco.2017.07.002). The authors ran a topic model, Latent Dirichlet Allocation, across the full text of 10-K filings from 1996 through 2013, letting them measure not just how long filings became but what the added words were about.

The method is what makes the paper more useful than a simple page count. A topic model assigns every passage of every filing to one of 150 machine-learned topics, so the authors could ask not only "did filings get longer" but "which subjects absorbed the new words, and were those words specific or boilerplate." That distinction is the difference between measuring disclosure and measuring paper, and it is why this study, rather than a casual word count, is the right citation for the claim.

The headline findings:

• Length roughly doubled. The median 10-K grew to about twice its mid-1990s length over the sample period.
• Boilerplate and redundancy rose. More of the document became standardized language repeated across companies and across years, and more of it repeated content stated elsewhere in the same filing.
• Readability and specificity fell. By standard textual measures, filings became harder to read, less specific, and relatively lighter on hard, quantitative information.
• Regulation drove it. New FASB and SEC requirements explain most of the added length. Of the 150 topics the model identified, just three account for virtually all of the increase.

What is driving the growth in disclosure?

The three topics behind the growth are not a random sample of accounting. Each maps to a wave of rulemaking:

• Fair value measurement. Fair value accounting standards required companies to explain valuation methods, inputs, and hierarchies, adding dense recurring disclosure to the notes.
• Internal controls. The Sarbanes-Oxley Act's internal-control reporting turned a compliance exercise into pages of annual attestation language.
• Risk factors. The SEC began requiring risk factor disclosure in annual reports in 2005; the section has grown, and standardized, ever since.

The pattern matters as much as the list. Disclosure requirements accumulate: each new standard adds a section, and old sections rarely leave. Individually, every requirement is defensible; collectively, they produce a document whose length is set by the union of every rule that ever applied to it. That growth compounds in the notes to the financial statements, which is where fair value tables and measurement disclosures live, and which already make up roughly half of a typical filing. What the notes are and how to work through them is covered in our guide to reading 10-K footnotes.

Did 10-Ks get harder to read, too?

Yes, and this is the finding that should bother readers more than the page count. The study found the added text was disproportionately boilerplate: standardized, sticky from year to year, and redundant within the filing. Meanwhile specificity and the relative share of hard information declined. In plain terms, the signal did not grow with the document. A reader in 2013 had to process roughly twice the words of a reader in 1996 to extract a comparable amount of firm-specific information, and the trend since then has not reversed.

This is why "just read the whole thing" stopped being practical advice. A sequential read of a modern 10-K spends most of its time in language written to satisfy a rule rather than to inform, and the material that matters (the specific, quantitative, firm-level detail) sits scattered through the notes, reached mostly by following cross-references. The document grew at both ends of that trip: more pages between the reference and the note, and more boilerplate to wade through once you arrive.

Has the growth continued since 2013?

The study's sample ends in 2013, so the doubling figure should be quoted with its window: 1996 to 2013. But nothing since has reversed the mechanism the paper identified, and the requirements have kept accumulating. Since the sample closed, public companies have absorbed, among others:

• New revenue recognition disclosures under ASC 606, effective for public companies in 2018, which added disaggregation tables, contract balance rollforwards, and remaining performance obligation disclosures to the notes.
• New lease disclosures under ASC 842, effective in 2019, which brought operating leases onto the balance sheet along with maturity tables and weighted-average term and rate disclosures.
• Human capital disclosure, required by the SEC's Regulation S-K modernization in 2020.
• Cybersecurity disclosure, a dedicated item in the 10-K since the SEC's 2023 rules on risk management and governance.

Each of these is individually reasonable, and each follows the study's pattern exactly: a new requirement adds a recurring section, the section standardizes across filers, and the document ratchets longer. We are not aware of a published follow-up that extends the full topic-model analysis past 2013, which is why this page quotes the doubling only for the studied window and describes the years since qualitatively.

What does this mean for how you read a 10-K?

Three practical conclusions follow from the research:

1. Read by reference, not by page. When most added text is boilerplate, page order is the wrong traversal. Read the statements, then follow their cross-references into the specific notes that explain the movements you care about.
2. Budget for navigation. Longer filings mean longer round trips between a reference and its note. A full expert read runs on the order of three hours, and a growing share of that is finding things, not evaluating them.
3. Treat boilerplate changes as signal. Precisely because so much language is sticky year to year, the sentences that do change deserve attention. Diffing against last year's filing is more productive than rereading what carried over.

The document is not going to get shorter. The workable response is a reading process built for a long, reference-heavy document, plus tooling that cheapens the round trips. That is the problem SeeNote exists to remove: it makes each cross-reference on EDGAR clickable so the referenced note appears where you are reading. How that compares to working with raw EDGAR, row by row, is in EDGAR vs SeeNote.

How to cite the statistic

If you want to use the doubling finding, cite the paper, not us:

The short form we use on our homepage: median 10-K length roughly doubled between 1996 and 2013, per Dyer, Lang & Stice-Lawrence (Journal of Accounting and Economics, 2017).