Administrative Data Rules in JSON: Why Does Japan Call It “Rule as a Code”?

July 22, 2026

A Curious Choice of Words

When Japan’s Digital Agency released its administrative data rules on GitHub earlier this year — marking the first time the government had provided rules in a machine-readable format — a curious detail tucked away in an agency infographic went virtually unnoticed: it was written as “Rule as a Code.”

Far from a typo, this seems like a subtle message communicating the agency’s underlying strategy for Rules as Code —though so subtle that it might easily fly under almost everyone’s radar.

The Core Mission: Fixing Administrative Data at the Source

At its heart, this initiative aims to eliminate the massive, hidden tax of manual data wrangling that plagues both public and private sectors. By enforcing machine-friendly formatting at the point of data creation by the government agencies, Japan aims to enable data integration and pave the way for automated AI processing at a national scale.

Crucially, they did not just publish static PDF guidelines. For the first time, it released the underlying validation rules as downloadable JSON and CSV formats—allowing developers and software systems to programmatically verify dataset compliance on the fly.

Two-tiered rules applied to the public sector

  • Statistical Data (e-Stat): These specialized standards, released in 2020, govern official national statistics where analytical precision is necessary.
  • General Administrative Data (Tabular data): This newly formulated guideline applies to the tabular data produced across government offices.

Raising Data Standards in Three Stages (Levels)

The new guideline introduces a pragmatic three-tier maturity model designed to elevate data quality across agencies with the minimum friction:

  • Level 1 (Viewable & Transferable) : Basic operations such as copying and pasting in spreadsheet software are mechanically possible. No merged cells, One data point per cell, No cosmetic padding using spaces or line breaks.
  • Level 2 (Aggregatable & Analyzable) : Automatic aggregation is possible in BI tools without requiring manual pre-processing. No strings like “yen” mixed into numerical data. Standardized categorical and survey responses.:
  • Level 3 (Linkable & Automatable) : Automatic acquisition via programming languages and seamless integration with external datasets are both fully enabled. Long format rather than wide format. Standardised temporal axes and regional/geospatial codes.

Why Call It “Rule as a Code”?

While Rules as Code takes many forms, the agency’s altered moniker, “Rule as a Code,” initially seems perplexing. By reducing the plural ‘rules’ to a singular ‘rule’ (a data specification) published as ‘a code’ (a JSON file), the drafters may simply be acknowledging that this basic data management effort falls short of transformative vision of Rules as Code. This initiative, after all, is a precursor or a foundational data-hygiene layer.

Is It a Trojan Horse?

However, if it is a Trojan horse, it is one that risks confusing its own allies. This muddled branding is too subtle to travel well. Rather than highlighting the strategic importance of data interoperability, there is a real concern that it actively invites the misconception that this basic data formatting exercise is the ultimate realization of Rules as Code. In fact, one government ministry is already introducing the release of these JSON and CSV guidelines as a “Rules as Code initiative” to its counterparts.

Data formatting is the ‘indispensable first mile’. If a government cannot automate its spreadsheet schemas, it will never successfully automate its laws. Because this foundational work is valuable in its own right, there is no need to hide behind subtle puns. I hope to see the agency explicitly articulate its true intent and take the lead in defining its own vision for Rules as Code.

Data integration cannot be achieved overnight; it is forged through pragmatic, incremental breakthroughs like this one.

References
English titles of Japanese sources are informal translations by the author.

  1. Rules on Machine Readability in Administrative Data (Cabinet Secretariat)
  2. Initiatives to Code Rules on Machine Readability in Administrative Data (Digital Agency)
  3. Study Group on AI Transformation in Local Governments (1st Meeting) (Ministry of Internal Affairs and Communications)
  4. Efforts for Administrative Data Utilization in Digital Administrative and Fiscal Reform (Digital Administrative and Fiscal Reform Council)
  5. Formulation of Unified Rules for Machine-Readable Data Notation in Statistical Tables (Ministry of Internal Affairs and Communications)
  6. Code for Rules on Machine Readability in Administrative Data (JSON) (GitHub)
  7. Code for Rules on Machine Readability in Administrative Data (CSV) (GitHub)
  8. “Coding Machine-Readability Rules and What Lies Beyond,” Blog, June 25, 2026 (hrkzz)

コメントを残す