Did missing/corrupt dates in COBOL default to 1875-05-20?

Original Article ↗ Hacker News Discussion ↗

Misinformation and Epistemic Exhaustion

  • Several comments start from frustration at how easily a plausible technical story (“COBOL defaults to 1875‑05‑20”) spread widely without scrutiny, even among technically literate people.
  • This is tied to a broader sense that distinguishing truth from falsehood is getting harder in an AI/engagement-driven media environment.
  • There is disagreement over “both-sides” narratives: some argue one US political faction uses systematic disinformation far more; others point to this case as evidence that partisan misinformation exists across the spectrum.

The 1875 COBOL Claim: What’s Actually Known

  • Consensus: COBOL has no built‑in date type and does not default corrupt or missing dates to 1875‑05‑20.
  • Different COBOL environments offer date functions with other epochs (e.g., 1601 or 1582 days-since-epoch integers), but that’s unrelated to the viral 1875 story.
  • SSA’s core files (e.g., NUMIDENT, Master Beneficiary Record) historically store dates as text (CCYYMMDD or variations) or, in some post‑Y2K layouts, as days since 1800—not 1875.
  • The 1875 idea appears to originate from an anonymous, technically error‑filled comment and then got laundered through screenshots and social media dunking posts; no evidence SSA actually uses 1875 as a sentinel or epoch.

Unknown / Bad Dates and Sentinel Values

  • Real legacy systems often encode “unknown” values via sentinels (e.g., 0000‑00‑00, 9999‑01‑01) or out‑of‑range dates; some GIS standards use 1875‑05‑20 as a temporal datum default, but that doesn’t imply SSA does.
  • There’s debate over whether a system should use a realistic fake date vs. an obviously-invalid sentinel; most commenters say fabricating plausible dates is worse, because it pollutes analytics and masks uncertainty.

Musk / DOGE Claims About 150‑Year‑Olds

  • Musk and the new federal “DOGE” effort claimed many Social Security records show non‑deceased people aged 150+, implying massive fraud.
  • Later, Musk posted an age histogram (0–369) for records marked “not deceased,” showing millions of entries over 100 and a long tail of impossible ages—decisively contradicting the neat “1875 epoch” explanation.
  • Commenters note: “not deceased in this database” ≠ “currently receiving checks.” Without payment‑side and eligibility context, the fraud implication is speculative.

Existing SSA Audits and Data Quality

  • Multiple comments link SSA Office of Inspector General reports (2015 “112+” and 2023 “100+” audits):
    • ~18.9M records had age ≥100 with no death info; about 44k of those were receiving SSA payments.
    • SSA and OIG agree that “almost none” of these old records are currently being paid; the larger risk is misuse of uncorrected SSNs elsewhere.
    • SSA declined to mass‑annotate presumed deaths because projected correction cost ($5.5–$9.7M) roughly matched or exceeded the benefit.
  • Many argue this shows: (a) data quality is indeed messy; (b) it’s known and documented; (c) it does not support claims of multi‑trillion‑dollar fraud.

Broader Themes: Legacy Systems, Audits, and Propaganda

  • Several COBOL and government‑IT veterans stress that the hard part is not the language but decades of encoded policy and edge cases; naive “look, crazy data ⇒ fraud” narratives are common rookie mistakes.
  • Some see Musk’s communications as reckless “drive‑by” propaganda: releasing partial numbers without context to erode trust in SSA and justify cutting benefits.
  • Others think DOGE may still surface real waste/fraud, but criticize the rush to tweet dramatic claims before thorough, professional audit work is done.