Seven Years of 'Care' Mentions in Japanese Local Assemblies — The Time Structure of Long-Term Care Insurance Act Revisions and Policy Lag

This note reads year-by-year counts of Japanese local assembly speeches containing kaigo (care), youkaigo (care needs), or kaigo hoken (Long-Term Care Insurance) from 2018 to 2024, using the machikarte corpus (~130 million assembly speech records). The aim is to examine the time structure between the Long-Term Care Insurance Act revision cycle, the Care Insurance Plans, and local assembly responses. Naming individual assemblies or councillors in evaluative ranking terms is outside the article's purpose. The scope of observation is held at national-trend and prefecture-aggregate granularity; no top-or-bottom ranking by municipality name is published here.

Across the seven-year window from 2018 to 2024, mentions of care-related terms in Japanese local assemblies range from about 105,000 to 144,000 speeches per year. The absolute count drops from 141,000 in 2018 to 105,000 in 2019, recovers gradually, and reaches its seven-year peak of 144,000 in 2024.

The absolute counts, however, must be read with care. The number of municipalities recorded in machikarte grew from 1,147 in 2018 to 1,316 in 2024, so part of the rise reflects expanding ingestion coverage rather than substantive growth in assembly discourse. The mention rate, normalised by total speeches per year, moves within a narrower band: 2.39% in 2018, 1.94% in 2019, 1.89% in 2022, and 2.16% in 2024.

When the year-by-year mention rate is laid out, two rising phases come into view. One sits at 2018 — the highest value in the seven-year window observed to date. The other sits at 2024, where the rate climbed for two consecutive years from the 2022 trough (1.89%) to 2.16%. Both rising phases coincide with major points in the Long-Term Care Insurance Act revision cycle and the Care Insurance Plan schedule.

Long-Term Care Insurance Act revision timeline

The Long-Term Care Insurance Act (Act No. 123 of 1997) sets out benefits and service-provision rules for long-term care insurance, with care reimbursement revisions and Care Insurance Plan updates running on a three-year cycle. Within 2018-2024, three milestones plausibly affected assembly discourse:

• 2018: Implementation year of an Act revision, with measures to deepen the community-based integrated care system, introduce kyousei-gata (community-shared) services, and strengthen self-reliance support and severity-prevention. Care insurance fee revision took effect in April.
• 2020: The "Act for Partial Revision of the Social Welfare Act and Related Laws for the Realization of a Community-Inclusive Society" (Act No. 52 of 2020, promulgated June 2020) made bundle revisions across 11 related laws including the Long-Term Care Insurance Act. The revision clarified the responsibilities of national and local governments on dementia policy, positioned care workforce securement within Care Insurance Plans, and established the Social Welfare Collaboration Promotion Corporation system (in force April 2022).
• 2024: Start year of the 9th Care Insurance Plan (FY 2024-2026). This was the triple revision year (simultaneous revision of medical fees, care insurance fees, and disability welfare service fees). The care insurance fee revision generally took effect in April, while services tightly linked to medical care (home-visit nursing, home-visit rehabilitation, outpatient rehabilitation, and home-based medical management guidance) took effect in June under a split-implementation schedule.

Overlaying the Act revisions and plan cycle on the mention rate reveals that the 2018 and 2024 rises align temporally with these milestones. The 2020 revision is not accompanied by a comparable spike in assembly mentions — possibly because pandemic response absorbed assembly time, although surface keyword aggregation alone cannot isolate the cause.

The structural backdrop: care workforce shortage

Behind the policy debate of the 2020s sits a demographic shift and a chronic shortage of care workers. Population projections through 2070 show that the share of the population aged 65 and older keeps rising while the working-age population shrinks.

The care-workforce shortage is positioned as a central issue in the care insurance fee revision, and the 9th Plan threads together higher compensation, technology adoption, and foreign workforce intake in parallel. One plausible reason the mention rate rose in local assemblies in 2024 is that the plan's start year drew municipal response strategies onto the agenda.

What the word "kaigo" captures

The regex aggregation will inevitably pick up uses of "kaigo" beyond Long-Term Care Insurance benefits. The term also appears in references to the family-care leave system in labour law, the medical-care continuum, and family-care consultation services. This article treats them collectively as "care-related mentions"; breaking them down by context is left to a later iteration.

The time structure of statutory revision and assembly discourse

A three-year cycle of Long-Term Care Insurance Act revisions is, by design, a predictable schedule, so an increase in assembly discussion around revision years is structurally expected. What is informative is whether the actual mention rate matches that expected pattern, and where it does not, in which direction it deviates.

The 2018 mention rate of 2.39% is the highest in the seven-year window, consistent with discussion clustering around the revision implementation year. The 2020 revision sits at 2.19% — a recovery from the 2019 trough of 1.94% — but does not match the 2018 peak. The 2024 figure of 2.16% comes after two consecutive years of rises from the 2022 trough (1.89%) and aligns with the 9th Plan's launch.

What this observation suggests is that, while assembly discussion of care synchronises with statutory revisions, the amplitude of the response varies across revision years. The substantive content of each revision, broader public attention at the time, and parallel agenda items may all be reflected in the mention rate.

Reading direction for policy lag

How long it takes for a policy to permeate assembly discussion, or conversely how many years assembly debate lags behind policy formation, is hard to measure without a cross-cutting search base of assembly minutes. The aggregation in this article is intended as raw material for measuring that time structure.

That said, surface keyword aggregation alone cannot distinguish whether assembly speech leads or lags the policy cycle. The 2018 rise, for example, blends preparatory discussion before implementation with case review after it. Monthly resolution of speech timing, combining mention rates with topic classification, and separating speech by speaker (councillor versus administrative respondent) would allow the direction of the time lag to be made explicit.

Existence of prefectural variation

This article focuses on national trends, but the prefectural aggregation also shows variation in care mention rates. Prefectures with different ageing rates, different shares of designated cities, and different care-facility supply ratios may show regional differences in the mention rate. A separate article will take this up, with cross-references to the trends presented here.

Room for cross-lab citation

Long-term care policy sits at the intersection of medical care, welfare, the labour market, and the community-based integrated care system. One connection point with other articles in this lab is the distribution of deferral phrasing in assembly responses (case-sakiokuri-rate); combining the two would let us measure how much "we will consider" hedging appears within care-related discussion. The hollowing of debate, treated elsewhere in the broader ISVD lab network, also offers a structural connection point.

• No semantic classification: Contextual classification of speeches containing "kaigo" (Long-Term Care Insurance benefits / family-care leave / family-care consultation, etc.) is not applied; the aggregation stays at surface keyword level.
• No agenda-portfolio normalisation: The aggregation is not normalised for the agenda portfolio of care-related items, so differences in agenda mix may surface in the mention rate.
• No separation of speakers: Councillor questions and administrative responses are not separated; analysing the leading / following structure of the time lag will require speaker-level aggregation.
• Pre-2017 coverage incomplete: machikarte's coverage thins out for earlier years; before-and-after comparisons across revisions are limited to 2018 and later.
• Adjacent senior-policy keywords: Community-based integrated care, dementia policy, and senior housing keywords are not aggregated here and require separate queries.
• No individual ranking: The article stays at the structural level and does not publish municipality-level or councillor-level top-or-bottom rankings.

These limits will be lifted in later iterations. The next priorities are: (1) monthly resolution to measure the time lag around statutory revisions, (2) speaker-level separation, and (3) refined prefectural aggregation.

The query specification (spec_version v1-care-trends-2026-06) is documented at the end of this article, and the BigQuery aggregation queries will be placed in the machikarte GitHub repository so that third parties can reproduce the results independently.

A corrections contact is provided separately to receive notices about aggregation errors and notation drift. The lab's editorial procedure — including the three-tier rule on councillor-level data publication — is documented in the lab overview note (hypothesis-overview).

Aggregation queries (spec_version v1-care-trends-2026-06)

Numerator — annual count of care-related mentions:

SELECT
  year,
  COUNT(*) AS care_mentions,
  COUNT(DISTINCT municipality_code) AS municipalities
FROM `correlate-workspace.isvd_machikarte.speeches`
WHERE year BETWEEN 2018 AND 2024
  AND REGEXP_CONTAINS(body, r'(介護|要介護|介護保険)')
GROUP BY year
ORDER BY year

Denominator — annual total speeches:

SELECT
  year,
  COUNT(*) AS total_speeches,
  COUNT(DISTINCT municipality_code) AS total_municipalities
FROM `correlate-workspace.isvd_machikarte.speeches`
WHERE year BETWEEN 2018 AND 2024
GROUP BY year
ORDER BY year

Query execution date: 3 June 2026 / mention rate = numerator / denominator × 100.

machikarte — Nationwide Local Assembly Speech Search Platform (Beta) — Institute for Social Vision Design (ISVD). ISVD

machikarte (GitHub) — schema, aggregation queries, licenses (MIT + CC BY 4.0) — Institute for Social Vision Design (ISVD). GitHub

Long-Term Care Insurance Act (Act No. 123 of 1997) — Ministry of Health, Labour and Welfare. e-Gov Laws and Regulations

Long-Term Care and Welfare for Older People (9th Care Insurance Plan portal) — Health and Welfare Bureau for the Elderly, Ministry of Health, Labour and Welfare. MHLW

Projections of Japan's Future Population (2023 Estimate) — National Institute of Population and Social Security Research. IPSS

Verification of Role Classification Methods for Diet Records Using BERT-Based Classifiers (in Japanese) — Miyaki, Y. and Uchida, Y.. Journal of Japan Society for Fuzzy Theory and Intelligent Informatics, Vol. 37, No. 1, pp. 530-534