MachiKarte

Methodology (English summary)

This page is an English summary of our analysis methodology. The full Japanese version contains additional statistical notes and historical context. If you rely on our numbers in research or reporting, please refer to both.

1. Data sources

Assembly minutes: Full-text minutes are retrieved from municipal assembly record search systems (kensakusystem.jp etc.) and parsed into question-response pairs. Updated per regular session.
Fiscal data: Revenue, expenditure, and indicators are extracted from the Ministry of Internal Affairs and Communications' annual fiscal status survey (“決算カード”). Updated once per year after the previous fiscal year's publication.
Procurement: Bidder lists and designated manager records are collected from each municipality's public procurement portal and the Tokyo joint e-procurement service.
Furusato nozei (hometown tax) data: Tax revenue outflow and donation counts from each municipality's published reports and the Ministry's current status survey.

2. Scoring indicators

QQS — Question Quality Score

The question quality score (0–100) is a composite of four sub-dimensions, each contributing up to 25 points:

Logical structure: Does the question move from problem statement to evidence to proposal?
Specificity: Does the question cite concrete numbers, dates, or precedents?
Originality: Does the question reference original research, fieldwork, or media reports?
Follow-up attention: Does the question revisit prior commitments or outcomes?

RQS — Response Quality Score

The response quality score (0–100) is a composite of four sub-dimensions:

Directness: Does the response address the question structure?
Specificity: Is the response specific or padded with “we will consider” phrasing?
Data disclosure: Does the response include budget figures, achievement rates, or date-specific references?
Commitment: Does the response include specific timelines or budgeted plans?

AHS — Assembly Health Score

AHS = QQS × 0.4 + RQS × 0.6. The 60% weight on responses reflects the design principle that the material impact of assembly discussion is ultimately determined by how specific the executive's response is, not just by how sharp the question is.

Follow-through rate (“kentou-ritsu”)

The share of responses that use non-committal phrasing such as “we will consider” (“検討します”), “we will look into it” (“検討してまいります”), or “we will study” (“研究してまいります”) instead of committing to a timeline or plan. This is independent from AHS and reported as a separate indicator.

3. Statistical limits

Sample size: Cross-municipal rankings and correlation analyses are statistically weak when few municipalities are analyzed. Interpret with caution until n ≥ 50.
Assembly differences: Session length, committee structure, and minutes recording style vary across municipalities. Scores are relative tendencies, not absolute comparisons.
Style bias: Dialogical (one-question-one-answer) and read-aloud (batch question) styles affect QQS differently. Individual municipality pages include style annotations.
Keyword dependence: Theme classification uses a Japanese-kanji BoW method. Specialized vocabulary may be misclassified. The dictionary is updated regularly.

4. Data access and citation

JSON API

GET https://machikarte.isvd.or.jp/api/score-summaries

CORS-enabled. Returns all analyzed municipality summaries.

Citation format

MachiKarte (ISVD) — https://machikarte.isvd.or.jp/
Retrieved: YYYY-MM-DD

License: MachiKarte's derived analysis (scores, classifications, summaries) is released under CC BY 4.0. Source data remains under respective government terms.

5. Contact and collaboration

For research collaboration, bulk data queries, factual corrections, or methodological discussion:

info@isvd.or.jp

We welcome inquiries from journalists, researchers, public officials, and citizens.

Operated by ISVD (Institute for Social Vision Design) / 一般社団法人社会構想デザイン機構