electricsheepasia/asia-ilo-how-temp-eco-ocu-nb-mean-weekly-hours-actually-worked-per-employed-per

--- license: cc-by-4.0 language: - en task_categories: - tabular-classification - tabular-regression - time-series-forecasting multilinguality: monolingual size_categories: - 100K<n<1M tags: - tabular - asia - ilostat - hours-of-work - ilo - labour - employment pretty_name: "Mean weekly hours actually worked per employed person by economic activity and occupation | Asia (ILOSTAT)" --- # Mean weekly hours actually worked per employed person by economic activity and occupation | Asia (ILOSTAT) 🌏 **136,945 observations** · **33 Asia countries** · **1997–2025** · *Repackaged by [Electric Sheep Asia](https://huggingface.co/electricsheepasia)*      ## TL;DR This dataset contains **136,945 observations** of `Hours of work` data across **33 Asia countries**, spanning **1997–2025**, covering **1 distinct indicators**. ## About the source **ILOSTAT** is the ILO's central statistics database, the leading global source for labour statistics. It compiles indicators across employment, unemployment, wages, working time, child labour, informal economy, social protection, occupational injuries, and SDG decent work targets — drawing on national labour force surveys, household income surveys, establishment surveys, and administrative records. Coverage spans 200+ economies, with the ILO's Department of Statistics responsible for harmonisation. - **Source:** [ILOSTAT](https://www.ilo.org/shinyapps/bulkexplorer/?id=HOW_TEMP_ECO_OCU_NB) - **Publisher:** International Labour Organization (ILO) - **License:** [cc-by-4.0](https://creativecommons.org/licenses/by/4.0/) - **Topic:** Hours of work ## Methodology Data pulled directly from the ILOSTAT REST API at `https://rplumber.ilo.org/data/indicator?id=HOW_TEMP_ECO_OCU_NB` and filtered to Asia ISO3 country codes. ILOSTAT harmonises raw survey microdata using ICLS (International Conference of Labour Statisticians) definitions; sources are flagged in the `source.label` column for traceability. ## Geographic coverage 33 Asia countries · top rows shown below, sorted by row count: | Country | Rows | First year | Last year | |---------|-----:|-----------:|----------:| | `KOR` | 12,325 | 2000 | 2025 | | `CYP` | 10,428 | 1999 | 2024 | | `TUR` | 10,009 | 2000 | 2024 | | `IRN` | 8,324 | 2008 | 2024 | | `VNM` | 7,878 | 2007 | 2024 | | `ISR` | 7,138 | 2012 | 2024 | | `THA` | 6,721 | 2010 | 2024 | | `KHM` | 6,394 | 1997 | 2023 | | `LKA` | 6,269 | 2010 | 2024 | | `KGZ` | 6,088 | 2010 | 2023 | | `PHL` | 5,415 | 2012 | 2023 | | `IDN` | 4,954 | 2007 | 2023 | | `PSE` | 4,183 | 2015 | 2025 | | `PAK` | 4,127 | 2009 | 2025 | | `BRN` | 3,588 | 2014 | 2024 | | ... | _18 more countries_ | | | ## Indicators (sample) - `HOW_TEMP_ECO_OCU_NB` — Mean weekly hours actually worked per employed person by economic activity and occupation ## Schema | Column | Type | Description | Example | |--------|------|-------------|---------| | `ref_area` | `string` | ISO 3166-1 alpha-3 country code | `AFG` | | `ref_area.label` | `string` | Country name in English | `Afghanistan` | | `source` | `string` | ILOSTAT source code (e.g. labour force survey) | `BA:15715` | | `source.label` | `string` | Source name in English | `LFS - Labour Force Survey` | | `indicator` | `string` | ILOSTAT indicator code | `HOW_TEMP_ECO_OCU_NB` | | `indicator.label` | `string` | Indicator name in English | `Mean weekly hours actually worked per…` | | `classif1` | `string` | First classification variable (age, education, status, etc.) | `ECO_SECTOR_TOTAL` | | `classif1.label` | `string` | — | `Economic activity (Broad sector): Total` | | `classif2` | `string` | Second classification variable where applicable | `OCU_SKILL_TOTAL` | | `classif2.label` | `string` | — | `Occupation (Skill level): Total` | | `time` | `int64` | Observation year | `2021` | | `obs_value` | `float64` | Observed indicator value (unit varies — see indicator definition) | `39.6` | | `obs_status` | `string` | Observation status flag (e.g. provisional, unreliable) | `U` | | `obs_status.label` | `string` | — | `Unreliable` | | `note_classif` | `string` | — | `—` | | `note_classif.label` | `string` | — | `—` | | `note_indicator` | `string` | — | `T11:142` | | `note_indicator.label` | `string` | — | `Job coverage: Main job currently held` | | `note_source` | `string` | — | `R1:3513_S3:8` | | `note_source.label` | `string` | — | `Repository: ILO-STATISTICS - Micro da…` | ## Data quality & caveats - Data is annual frequency. Some indicators also publish monthly or quarterly series — those are not included here. - When an indicator has multiple sources for the same country×year, the ILO-selected 'best source' is used. - Disaggregation columns (`sex`, `classif1`, `classif2`) are non-null only when the indicator publishes that breakdown. ## Usage ```python from datasets import load_dataset ds = load_dataset("electricsheepasia/asia-ilo-how-temp-eco-ocu-nb-mean-weekly-hours-actually-worked-per-employed-per") df = ds["train"].to_pandas() print(df.head()) ``` ### Filter to one country ```python indonesia = df[df["ref_area"] == "IDN"] ``` ### Time-series for a single indicator ```python sample = (df[df["indicator"] == "HOW_TEMP_ECO_OCU_NB"] .sort_values("time")) sample.plot(x="time", y="obs_value", title="HOW_TEMP_ECO_OCU_NB") ``` ### Pivot to country × year matrix ```python matrix = (df[df["indicator"] == "HOW_TEMP_ECO_OCU_NB"] .pivot_table(index="time", columns="ref_area", values="obs_value")) print(matrix.tail()) ``` ## Citation ```bibtex @misc{asia_ilo_how_temp_eco_ocu_nb_mean_weekly_hours_actually_worked_per_employed_per_2025, title = {Mean weekly hours actually worked per employed person by economic activity and occupation | Asia (ILOSTAT)}, author = {International Labour Organization (ILO)}, year = {2025}, url = {https://www.ilo.org/shinyapps/bulkexplorer/?id=HOW_TEMP_ECO_OCU_NB}, publisher = {HuggingFace Datasets, repackaged by Electric Sheep Asia}, howpublished = {\url{https://huggingface.co/datasets/electricsheepasia/asia-ilo-how-temp-eco-ocu-nb-mean-weekly-hours-actually-worked-per-employed-per}} } ``` ## License Released under [cc-by-4.0](https://creativecommons.org/licenses/by/4.0/). Original data © International Labour Organization (ILO). When using this dataset, please cite both the original source above and the Electric Sheep Asia repackaging. ## About Electric Sheep Electric Sheep Asia is part of the Electric Sheep mission: a unified, ML-ready data layer for Asia on HuggingFace. We pull data from authoritative open sources, normalize the schemas, package as Parquet, and publish with consistent dataset cards so researchers and developers can use `load_dataset()` to start working in seconds. Browse the full collection: [huggingface.co/electricsheepasia](https://huggingface.co/electricsheepasia) --- _Provenance: ingested 2026-05-27 via the Electric Sheep pipeline. Source URL: https://www.ilo.org/shinyapps/bulkexplorer/?id=HOW_TEMP_ECO_OCU_NB_Mean weekly hours actually worked per employed person by economic activity and occupation | Asia (ILOSTAT)

搜集汇总

数据集介绍

构建方式

该数据集源自国际劳工组织（ILO）的ILOSTAT数据库，通过调用其REST API接口直接抽取“HOW_TEMP_ECO_OCU_NB”指标数据，并依据亚洲国家ISO3代码进行筛选过滤。原始调查微观数据经ILO统计部门依据国际劳工统计学家会议（ICLS）定义进行统一协调处理，数据来源信息在“source.label”字段中加以标注，确保可追溯性。数据以每年为观测频率，涵盖33个亚洲国家1997年至2025年间约13.7万条记录，经Electric Sheep Asia重新打包封装，形成机器学习的即用型数据集。

特点

该数据集的核心特色在于聚焦亚洲地区的劳动力工时统计，包含“按经济活动与职业划分的就业者每周实际工作平均小时数”单一指标，并通过多维度分类变量（如经济活动部门、职业技能等级）实现精细分析。每条观测记录除数值外，还附带了观察状态标识（如临时性、不可靠性）及详尽的注释信息，便于用户进行数据质量评估。数据集的时序覆盖跨度近三十年，国家覆盖面广且各国观测记录数量差异显著，可支持跨国比较与长期趋势研究。

使用方法

该数据集设计为与HuggingFace Datasets库无缝集成，用户可通过一行Python代码`load_dataset("electricsheepasia/asia-ilo-how-temp-eco-ocu-nb-mean-weekly-hours-actually-worked-per-employed-per")`加载数据，并直接转换为pandas DataFrame进行后续操作。支持按国家代码筛选子集、按时间序列绘制特定指标的变化趋势，以及通过透视表构建国家×年份的观测矩阵，便于进行面板数据分析与可视化。数据集以Parquet格式存储，兼容主流的统计与机器学习工作流。

背景与挑战

背景概述

该数据集由国际劳工组织（ILO）通过其ILOSTAT数据库于2025年整理发布，并由Electric Sheep Asia在HuggingFace平台上重新封装。ILO作为全球劳工统计领域的权威机构，长期致力于收集并标准化各国劳动数据，其发布的ILOSTAT数据库涵盖200多个经济体的就业、失业、工资及工时等核心指标。本数据集聚焦亚洲33个国家，时间跨度从1997年至2025年，包含136,945条观测记录，核心指标为按经济活动与职业划分的就业者每周实际工作时长均值。该数据集为亚洲劳动经济学、区域发展研究及跨国比较提供了珍贵的高颗粒度时序数据，尤其助力于分析亚洲各国劳动力市场的结构性变化与工作条件的演变趋势。

当前挑战

数据集所应对的领域问题在于，亚洲各国劳动统计口径不一，普查与调查数据频率和覆盖范围差异显著，导致跨国对比困难。ILOSTAT通过遵循国际劳工统计学家会议（ICLS）定义进行数据协调，但跨时段的可靠性与可比性仍是核心挑战。在构建过程中，数据整合面临多重障碍：原始数据源涉及多项调查类型（如劳动力调查、家庭收入调查），需甄别并优选最佳来源；分类变量（如经济部门、职业技能等级）的标准化解析难度高；观测值中存在不稳定或不可靠的状态标记（如'U'标识），需谨慎处理。此外，时间序列的年度频率限制以及部分指标不包含近年的月度或季度细分，为细粒度建模构成约束。

常用场景

经典使用场景

在劳动经济学与区域发展研究的交汇地带，该数据集（亚洲ILOSTAT周均工时数据）扮演着核心数据枢纽的角色。其经典使用场景聚焦于跨国家、跨行业、跨职业的工时结构分析，研究者可借助136,945条覆盖33个亚洲国家、横跨1997至2025年的观测记录，精细刻画不同经济活动和职业类别下就业人口的周均实际工作时长。通过内置的分类变量（如经济部门与技能等级），用户能够剖析制造业、服务业、农业等领域的工时异质性，或比较高技能与低技能职业的劳作强度差异，从而揭示亚洲劳动市场的时空格局与结构性变迁。

衍生相关工作

围绕该数据集已衍生出一系列具有标杆意义的学术与实践工作。在计量方法论层面，研究者利用其面板结构开发了针对非平衡面板数据的时序分解模型，用以分离亚洲国家工时变动的季节性成分、趋势成分与周期成分；在机器学习领域，该数据被用于训练区域劳动需求预测的梯度提升模型，或作为多任务学习框架中工时、失业率与工资方程的联合估计基础。此外，国际劳工组织的年度《亚洲就业与社会展望》报告将此类数据作为核心输入，构建了反映体面劳动进展的复合指标体系。数据集的电气羊（Electric Sheep Asia）再封装也推动了ILOSTAT与HuggingFace生态的深度融合，催生了标准化亚太劳动数据仓库的雏形。

数据集最近研究

最新研究方向

该数据集聚焦于亚洲地区从业人员按经济活动与职业划分的周平均实际工时，为劳动经济学、劳动力市场动态监测及国际比较研究提供了高粒度的时间序列基础。在亚洲经济高速转型与后疫情时代工作模式重塑的背景下，该数据正日益被用于分析灵活就业、非正规部门工时波动、行业间劳动强度差异等前沿议题。结合国际劳工组织对体面劳动的持续关注，该数据集助力学者精准追踪特定产业（如制造业、服务业）的工时演变，并与AI驱动的自动化、远程办公等热点事件形成交叉验证，为评估全球化背景下亚洲劳动力福祉与政策干预成效提供了量化支撑。