Dataset

german-job-postings

Jobs & Workforce Download 1 source

german-job-postings is a publicly hosted corpus of roughly 70,584 normalized German-language vacancies drawn from the Bundesagentur für Arbeit Jobbörse API, each row tagged with the KldB-2010 occupation code and machine-assigned ESCO occupation and skill labels. The dataset is distributed under CC-BY-4.0 and is intended as an open alternative to the predominantly English/US job-posting resources available to machine-learning practitioners.

Data preview

Loading preview...

Dataset card

What this dataset contains

The release is a single CSV-style split of 70,584 deduplicated vacancies fetched from the official Bundesagentur Jobbörse search endpoint, covering postings surfaced by a 272-query keyword-by-region matrix plus country-wide sweeps. Each row links a normalized title and employer to a NUTS region, the German KldB-2010 classification code, and zero-shot ESCO labels for occupation and skills, with an `is_green` flag marking green-economy roles. The taxonomy fields are explicitly described by the publisher as machine-generated weak labels rather than hand-verified ground truth.

Structure and fields

Each record carries provenance and licensing fields such as `id`, `source`, `source_url`, `license`, `provenance` and `fetched_at`, alongside posting metadata including `posted_date`, `lang`, `title`, `employer`, `region` and `nuts_code`. Classification outputs are stored in `kldb_2010`, `esco_occupation` (with `uri` and `label` subfields), `esco_skills` (a list of uri/label pairs) and the boolean `is_green`. Parsed-from-text attributes include `seniority` and `salary_range` (with `min_eur` and `max_eur` subfields), while `description_derived` is reserved for a future ToS-safe body summary and is redacted in the current version.

How analysts use it

Equity and macro analysts can aggregate the corpus by `nuts_code` and `kldb_2010` to track occupational demand, skill mix, and the share of green-flagged postings (`is_green`) across German Bundesländer as a labour-market pulse for sector exposure work. Recruiters, workforce-planning teams and quant researchers can train or benchmark token-classification and feature-extraction models on the `title` and `esco_skills` fields, including the companion `jobbert-de` checkpoint referenced in the publisher notes. Salary parsing into `salary_range.min_eur` and `max_eur` offers a coarse pay-benchmark signal, though analysts should weight results accordingly given the publisher's mixed-quality and zero-shot labelling caveats.

Documentation & references 1
  1. mischeiwiller/german-job-postings: listing on Hugging Face

Further reading

All guides
  • What is alternative data?Alternative data in investment research: what the term covers, how it differs from data format, and why provenance, timing, access and validation matter.
  • Job postings as an investment signalWhy hiring data predicts fundamentals: growth intent, expansion geography, technology adoption and cost control, how postings are collected, structured and turned into tradeable indicators.
  • Web scraping and compliance in alternative dataHow scraped datasets are actually collected, where terms of service and computer-muse laws draw lines, what hiQ v. LinkedIn settled and did not, and the diligence checklist for legally fragile data.
  • Free alternative data sourcesThe practical list of free alternative-data sources for investors: public filings, official statistics, satellite and earth observation, on-chain data, web archives and open registries, what each covers, what it costs in engineering effort, and where free data falls short.
  • Alternative data for investingHow funds turn alternative data into an investment edge: signal families, evaluation, backtesting, compliance and the build-vs-buy decision, mapped to the datasets and providers indexed on this encyclopedia.
  • Job market data: datasets and providersThe datasets and providers behind labor-market signals: job postings, payroll, employee profiles and job-change data, what each source measures, its caveats and how to choose one.

Discussion

Questions or observations about this dataset? Join the conversation on X.

Discuss on X

No comments yet. Start the conversation on X.