Data Note 003 — A Computational Substrate Audit of GHG Disclosure across 89 Japanese Prime Market Firms This data note examines whether corporate sustainability disclosure provides a computable substrate for downstream analysis. Using 89 Japanese Prime Market firms, the study constructs a Computability Support Score (CSS) over three verified primary axes: machine readability, unit normalization, and API accessibility. The central finding is diagnostic rather than dismissive. Corporate disclosure is not uniformly non-computable; instead, computability support is unevenly distributed across infrastructure axes. Machine readability and unit normalization are substantially stronger than an initial machine-only classification suggested, with pass rates of 92.1% and 88.8%, respectively. By contrast, API accessibility emerges as the primary bottleneck, with a pass rate of 47.2%. Manual verification of voluntary PDF disclosures corrected 35 false-zero classifications in the unit-normalization axis, demonstrating that a structured-filing-only audit can overstate substrate failure if it ignores evidence available in voluntary reporting artifacts. The resulting three-axis Primary CSS has a conservative lower-bound mean of 0.760 and a median of 0.667, indicating that the median firm passes two of three verified axes while failing on API accessibility. Audit trail is retained only as a secondary lower-confidence diagnostic axis, and time-series retrievability is treated as exploratory rather than quantitative due to mixed evidence classes. The note argues that CSS should be read as a diagnostic instrument that localizes where computational substrate integrity breaks, not as a blanket verdict that disclosure is unusable.
Hiroyuki Kokubu (Thu,) studied this question.