# Israel labour-force check (v199): what could and could not be tested

Source: ILOSTAT tables for Israel (national source: CBS Labour Force Survey), acquired in CI
(run 41936995440) and archived in `data/processed/israel_unemployed_composition_v199.json`.

## Core test: not run

The planned test compares the AI exposure of the unemployed (by previous occupation) with that of the
employed. ILOSTAT does not publish Israel's unemployment by former occupation through any endpoint tried
(rplumber API 400, bulk file 404, SDMX 404). The series exists in the CBS annual Labour Force Survey
publications, table 5.4 (unemployed who worked in the past 12 months, by industry and occupation of last
job). Acquiring it needs a CBS-specific script, one file per year. No result is claimed.

## What the available tables show (descriptive)

| Measure | 2012 | 2019 | 2024 | 2025 |
|---|---:|---:|---:|---:|
| Mean ILO exposure of the employed (ISCO major groups) | 0.320 | 0.325 | 0.327 | — |
| Employed in high-exposure groups (exposure ≥ 0.35: ISCO major groups 1–4) | 57.8% | 60.5% | 61.7% | — |
| Youth (15–24) share of the unemployed | 29.1% | 26.6% | 23.3% | 23.8% |
| Long-term (12+ months) share of the unemployed | 20.0% | 10.0% | 11.4% | 6.3% |

Reading: employment kept shifting toward exposed occupation groups through 2022 and flattened after.
Neither the youth share nor the long-term share of unemployment rose in 2024–2025. None of this tests
whether exposed workers are over-represented among the unemployed; it only rules out a visible shift in
two aggregate margins. 2023–2025 are confounded by war-time reserve duty and labour-supply shocks.
Exposure is an unweighted mean of ILO two-digit means within each major group.