Datasets:
China data elements: data, charts and code
Browse files- README.md +169 -0
- assets/competition_and_exchanges.png +3 -0
- assets/data_in_numbers.png +3 -0
- assets/policy_timeline.png +3 -0
- code/build.py +78 -0
- code/viz.py +128 -0
- data/competition.parquet +3 -0
- data/indicators.parquet +3 -0
- data/places.parquet +3 -0
- data/policy_timeline.parquet +3 -0
README.md
ADDED
|
@@ -0,0 +1,169 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
- zh
|
| 6 |
+
pretty_name: "China's Data Elements: Data as a Factor of Production, 2015 to 2026"
|
| 7 |
+
size_categories:
|
| 8 |
+
- n<1K
|
| 9 |
+
task_categories:
|
| 10 |
+
- tabular-classification
|
| 11 |
+
- time-series-forecasting
|
| 12 |
+
tags:
|
| 13 |
+
- data-economy
|
| 14 |
+
- data-governance
|
| 15 |
+
- china
|
| 16 |
+
- policy
|
| 17 |
+
- data-markets
|
| 18 |
+
- artificial-intelligence
|
| 19 |
+
- accounting
|
| 20 |
+
configs:
|
| 21 |
+
- config_name: policy_timeline
|
| 22 |
+
default: true
|
| 23 |
+
data_files:
|
| 24 |
+
- split: train
|
| 25 |
+
path: data/policy_timeline.parquet
|
| 26 |
+
- config_name: indicators
|
| 27 |
+
data_files:
|
| 28 |
+
- split: train
|
| 29 |
+
path: data/indicators.parquet
|
| 30 |
+
- config_name: competition
|
| 31 |
+
data_files:
|
| 32 |
+
- split: train
|
| 33 |
+
path: data/competition.parquet
|
| 34 |
+
- config_name: places
|
| 35 |
+
data_files:
|
| 36 |
+
- split: train
|
| 37 |
+
path: data/places.parquet
|
| 38 |
+
---
|
| 39 |
+
|
| 40 |
+
# China's Data Elements: Data as a Factor of Production, 2015 to 2026
|
| 41 |
+
|
| 42 |
+
**How did China turn data into a factor of production, alongside land, labour and capital, and what does that look like in numbers?** This dataset follows the "data elements" (数据要素) policy from the first big-data exchange in 2015 to the end of the "Data Element ×" three-year plan in 2026: the laws and institutions, the exchanges, the accounting rules that let companies put data on their balance sheets, the national competition, and the official measures of how much data China produces and trades.
|
| 43 |
+
|
| 44 |
+
*数据要素:从大数据战略到数据作为生产要素,2015年至2026年的政策、市场与数字。*
|
| 45 |
+
|
| 46 |
+

|
| 47 |
+
|
| 48 |
+
*Reading the chart: each row is a law, policy, institution, exchange or competition, placed at its date; colour and shape give the kind, the shaded band the 2024 to 2026 action plan. The legal base comes first (2016 to 2021), then the market (exchanges from 2021), then the State's own data administration and accounting rules (2023).*
|
| 49 |
+
|
| 50 |
+
## Dataset description
|
| 51 |
+
|
| 52 |
+
| | |
|
| 53 |
+
|---|---|
|
| 54 |
+
| Subject | Data as a factor of production in mainland China: policy, law, institutions, markets, accounting, and official statistics on data |
|
| 55 |
+
| Period | April 2015 to October 2026 |
|
| 56 |
+
| Size | 4 configurations: 20 dated events, 24 indicator values, 3 competition years, 8 places |
|
| 57 |
+
| Format | Parquet, with the pipeline in `code/` |
|
| 58 |
+
| Owner | Kukai-7, independent compilation |
|
| 59 |
+
| Created | October 2026, version 1.0 |
|
| 60 |
+
| License | CC BY 4.0 for the compilation; figures belong to their publishers (see sources) |
|
| 61 |
+
| Intended use | A reference timeline and a small table of headline figures for teaching, writing and further research on data governance and the data economy. Not legal advice and not investment advice. |
|
| 62 |
+
|
| 63 |
+
## Key facts
|
| 64 |
+
|
| 65 |
+
- **Data became a factor of production in 2019.** The Fourth Plenum of the 19th Central Committee (31 October 2019) listed data with labour, capital, land, knowledge, technology and management as factors that share in distribution; the March 2020 opinions on factor markets made it the fifth factor market.
|
| 66 |
+
- **A legal base in five years:** the Cybersecurity Law (2016), the Data Security Law (June 2021) and the Personal Information Protection Law (August 2021), then rules on cross-border data flows (March 2024).
|
| 67 |
+
- **A market, then an administration.** Exchanges opened in Guiyang (2015), Beijing and Shanghai (2021), Guangzhou and Shenzhen (2022). The "Twenty Measures on Data" of December 2022 separated the rights to hold, to process and to operate data. The National Data Administration opened in October 2023.
|
| 68 |
+
- **Data on the balance sheet.** Since 1 January 2024 companies may book data resources as intangible assets or inventory. In their 2025 annual reports, 136 A-share companies did so, for 3.79 billion yuan in all: 82% as intangible assets, 17% as development spending, under 1% as inventory. The three telecom carriers alone account for 55% of the amount.
|
| 69 |
+
- **Data production up 27% a year.** China produced 32.85 zettabytes of data in 2023, 41.06 in 2024 and 52.26 in 2025, about 27% of the world's. Data used for AI reached 199 exabytes in 2025, and for the first time more of it went to inference (101 EB) than to training.
|
| 70 |
+
- **A market of over 160 billion yuan.** Data trading exceeded 160 billion yuan in 2024 (up more than 30%), of which more than 30 billion on exchanges, about twice the year before. The digital economy as measured by CAICT reached 59.2 trillion yuan, 43.8% of GDP.
|
| 71 |
+
- **A competition that keeps growing.** The "Data Element ×" competition drew over 19,000 teams in 2024, 23,000 in 2025 and 29,000 in 2026, with over 145,000 participants this year; the 2026 final was held in Qingdao on 23 October 2026, in the last year of the three-year plan.
|
| 72 |
+
|
| 73 |
+

|
| 74 |
+
|
| 75 |
+
*Reading the chart: four separate panels, each with its own unit. Data produced each year (zettabytes, with growth inside the bars); data used for AI in 2025, split between training and inference (exabytes); data trading in 2024, all trading against exchange trading ("+" means "over"); data booked as assets by listed companies in 2025, by balance-sheet item.*
|
| 76 |
+
|
| 77 |
+

|
| 78 |
+
|
| 79 |
+
*Reading the chart: left, teams and participants in the "Data Element ×" competition, with the city of each national final; right, the five main data exchanges (blue) and the three finals (red rings). Activity concentrates in Beijing, Shanghai and the Greater Bay Area.*
|
| 80 |
+
|
| 81 |
+
## The 2026 milestone
|
| 82 |
+
|
| 83 |
+
The third "Data Element ×" final closed in Qingdao on 23 October 2026, organised by 22 bodies led by the National Data Administration, the Cyberspace Administration and the Shandong provincial government. The director of the National Data Administration described a "531" implementation system for market-based allocation of data, and said that data circulation was moving "from points to areas", with data vendors, exchanges and circulation platforms building a single national market; he paired "Data Element ×" with "Artificial Intelligence +" as the two lines of work ahead. The National Data Development Research Institute released an insight report on the competition, which points to a faster-closing institutional loop for data circulation, an active data market, and a "flywheel" between AI and high-quality datasets. These are official statements, recorded here as such; the dataset does not evaluate them.
|
| 84 |
+
|
| 85 |
+
## Load it
|
| 86 |
+
|
| 87 |
+
```python
|
| 88 |
+
from datasets import load_dataset
|
| 89 |
+
|
| 90 |
+
events = load_dataset("Kukai-7/china-data-elements", "policy_timeline", split="train").to_pandas()
|
| 91 |
+
indicators = load_dataset("Kukai-7/china-data-elements", "indicators", split="train").to_pandas()
|
| 92 |
+
|
| 93 |
+
# Every law adopted, in order
|
| 94 |
+
events[events.kind == "Law"][["date", "event", "event_zh", "note"]]
|
| 95 |
+
```
|
| 96 |
+
|
| 97 |
+
## Configurations
|
| 98 |
+
|
| 99 |
+
### `policy_timeline` (default), 20 rows
|
| 100 |
+
|
| 101 |
+
| column | description |
|
| 102 |
+
|---|---|
|
| 103 |
+
| `date`, `year`, `date_precision` | Date of adoption, publication, opening or final; `date_precision` flags the one date known only approximately |
|
| 104 |
+
| `kind` | Law, policy, institution, accounting rule, exchange or competition |
|
| 105 |
+
| `event`, `event_zh` | Short description in English and the Chinese name |
|
| 106 |
+
| `city` | Where it happened (Beijing for national acts) |
|
| 107 |
+
| `note` | Date in force, scope or main content |
|
| 108 |
+
|
| 109 |
+
### `indicators`, 24 rows
|
| 110 |
+
|
| 111 |
+
Long format: `indicator`, `year`, `value`, `unit`, `source`. Data production, growth, storage, share of world production, data used for AI, high-quality datasets, active data, data trading, data enterprises, the digital economy and its share of GDP, and data booked as assets by listed companies. Units containing "lower bound" mark figures published as "over".
|
| 112 |
+
|
| 113 |
+
### `competition`, 3 rows
|
| 114 |
+
|
| 115 |
+
Year, city of the national final, teams registered, participants, projects in the national final, tracks, award-winning teams, source.
|
| 116 |
+
|
| 117 |
+
### `places`, 8 rows
|
| 118 |
+
|
| 119 |
+
Cities of the exchanges and competition finals with coordinates, for mapping.
|
| 120 |
+
|
| 121 |
+
## Dataset characterization
|
| 122 |
+
|
| 123 |
+
**Data collection method:** manual compilation from official releases and press reports, each row with its source. There is no automated scraping of official sites in this version.
|
| 124 |
+
|
| 125 |
+
**Labeling method:** the `kind` of each event was assigned by hand.
|
| 126 |
+
|
| 127 |
+
| Publisher | What it gives |
|
| 128 |
+
|---|---|
|
| 129 |
+
| National Data Administration (国家数据局) | National data resources surveys 2023, 2024 and 2025; competition figures; data work conferences |
|
| 130 |
+
| State Council, CPC Central Committee, National People's Congress | Policy documents and laws |
|
| 131 |
+
| Ministry of Finance | Interim rules on accounting for enterprise data resources (2023) |
|
| 132 |
+
| Cyberspace Administration of China | Rules on cross-border data flows (2024) |
|
| 133 |
+
| China Academy of Information and Communications Technology (CAICT) | Digital economy size and share of GDP |
|
| 134 |
+
| Shanghai Advanced Institute of Finance (SAIF), via Yicai | Data assets booked by listed companies, 2025 annual reports |
|
| 135 |
+
| Xinhua, People's Daily, Science and Technology Daily | Press reports of the above |
|
| 136 |
+
|
| 137 |
+
## How it was built
|
| 138 |
+
|
| 139 |
+
`build.py` holds the curated tables with their sources and writes the Parquet files; `viz.py` draws the charts. There is no correlation analysis: the official series start in 2023 and give three points at most, too few for any.
|
| 140 |
+
|
| 141 |
+
## Limitations
|
| 142 |
+
|
| 143 |
+
- **Official figures, short series.** Most indicators come from government surveys or speeches. The national data resources survey started with 2023; earlier estimates of data production used other methods and are not comparable, so they are left out.
|
| 144 |
+
- **Lower bounds.** Many figures are published as "over" (超过); they are stored as the stated floor and flagged in the unit.
|
| 145 |
+
- **Definitions are new and moving.** "Data trading", "data enterprises", "active data" and "high-quality datasets" are defined by the publishers and may change. The CAICT digital economy measure is broad (it includes the digitalised part of traditional industries) and is not an official national accounts figure.
|
| 146 |
+
- **Data assets are small so far.** 3.79 billion yuan across 136 companies is tiny next to their total assets; booking data remains optional and concentrated in a few firms.
|
| 147 |
+
- **The timeline is a selection.** It keeps the national acts and the main exchanges; many local regulations and local exchanges or trading centres are not listed here.
|
| 148 |
+
- **No independent check.** Figures are taken as published; the dataset does not audit them.
|
| 149 |
+
|
| 150 |
+
## Ethical considerations
|
| 151 |
+
|
| 152 |
+
The dataset contains no personal data. It describes a policy area that bears on privacy, surveillance and state access to data; laws such as the Personal Information Protection Law are listed by date and scope only, without assessing how they are applied. Statements by officials are reported as statements.
|
| 153 |
+
|
| 154 |
+
## Citation
|
| 155 |
+
|
| 156 |
+
```bibtex
|
| 157 |
+
@misc{kukai7_china_data_elements_2026,
|
| 158 |
+
title = {China's Data Elements: Data as a Factor of Production, 2015 to 2026},
|
| 159 |
+
author = {Kukai-7},
|
| 160 |
+
year = {2026},
|
| 161 |
+
url = {https://huggingface.co/datasets/Kukai-7/china-data-elements}
|
| 162 |
+
}
|
| 163 |
+
```
|
| 164 |
+
|
| 165 |
+
## Related
|
| 166 |
+
|
| 167 |
+
- [Greater Bay Area Networks](https://huggingface.co/datasets/Kukai-7/greater-bay-area-networks), logistics, technology and finance around the Pearl River estuary.
|
| 168 |
+
- [China, AI and Manufacturing](https://huggingface.co/datasets/Kukai-7/china-ai-manufacturing), R&D by province and the Hangzhou, Shenzhen and Beijing clusters.
|
| 169 |
+
- [GitHub: Kukai-7](https://github.com/Kukai-7)
|
assets/competition_and_exchanges.png
ADDED
|
Git LFS Details
|
assets/data_in_numbers.png
ADDED
|
Git LFS Details
|
assets/policy_timeline.png
ADDED
|
Git LFS Details
|
code/build.py
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Build the China data elements dataset: data as a factor of production, from policy to market. All tables are curated from official
|
| 2 |
+
releases and press reports of the National Data Administration, the Ministry of Finance, CAICT and others; each row carries its source."""
|
| 3 |
+
import json, os, pandas as pd
|
| 4 |
+
|
| 5 |
+
OUT = 'out/data'
|
| 6 |
+
os.makedirs(OUT, exist_ok=True)
|
| 7 |
+
|
| 8 |
+
# ------------------------------------------------------------------ policy, law and institutions (date of adoption, publication or opening)
|
| 9 |
+
P = [('2015-04-14', 'Exchange', 'Guiyang Global Big Data Exchange opens', '贵阳大数据交易所挂牌', 'Guiyang', 'First data exchange in China'),
|
| 10 |
+
('2015-08-31', 'Policy', 'State Council Action Outline for Promoting Big Data Development', '促进大数据发展行动纲要', 'Beijing', 'First national big-data strategy'),
|
| 11 |
+
('2016-11-07', 'Law', 'Cybersecurity Law adopted', '网络安全法', 'Beijing', 'In force 1 June 2017'),
|
| 12 |
+
('2019-10-31', 'Policy', 'Fourth Plenum names data a factor of production', '十九届四中全会:数据作为生产要素参与分配', 'Beijing', 'Data joins labour, capital, land, knowledge, technology and management'),
|
| 13 |
+
('2020-03-30', 'Policy', 'Opinions on market-based allocation of factors', '关于构建更加完善的要素市场化配置体制机制的意见', 'Beijing', 'Data listed as the fifth factor market'),
|
| 14 |
+
('2021-03-31', 'Exchange', 'Beijing International Big Data Exchange opens', '北京国际大数据交易所', 'Beijing', ''),
|
| 15 |
+
('2021-06-10', 'Law', 'Data Security Law adopted', '数据安全法', 'Beijing', 'In force 1 September 2021'),
|
| 16 |
+
('2021-08-20', 'Law', 'Personal Information Protection Law adopted', '个人信息保护法', 'Beijing', 'In force 1 November 2021'),
|
| 17 |
+
('2021-11-25', 'Exchange', 'Shanghai Data Exchange opens', '上海数据交易所', 'Shanghai', ''),
|
| 18 |
+
('2022-09-30', 'Exchange', 'Guangzhou Data Exchange opens', '广州数据交易所', 'Guangzhou', ''),
|
| 19 |
+
('2022-11-15', 'Exchange', 'Shenzhen Data Exchange opens', '深圳数据交易所', 'Shenzhen', ''),
|
| 20 |
+
('2022-12-19', 'Policy', '"Twenty Measures on Data": opinions on the basic data system', '数据二十条', 'Beijing', 'Separates rights to hold, process and operate data'),
|
| 21 |
+
('2023-08-21', 'Accounting', 'Interim rules on accounting for enterprise data resources', '企业数据资源相关会计处理暂行规定', 'Beijing', 'Data can be booked as intangible assets or inventory from 1 January 2024'),
|
| 22 |
+
('2023-10-25', 'Institution', 'National Data Administration inaugurated', '国家数据局挂牌', 'Beijing', 'Approved in the State Council reform of March 2023'),
|
| 23 |
+
('2024-01-04', 'Policy', '"Data Element ×" three-year action plan 2024 to 2026', '"数据要素×"三年行动计划', 'Beijing', '17 departments, 12 sectors'),
|
| 24 |
+
('2024-03-22', 'Law', 'Rules on promoting and regulating cross-border data flows', '促进和规范数据跨境流动规定', 'Beijing', 'Cyberspace Administration of China'),
|
| 25 |
+
('2024-10-27', 'Competition', 'First "Data Element ×" competition final', '首届"数据要素×"大赛全国总决赛', 'Beijing', 'Over 19,000 teams'),
|
| 26 |
+
('2025-11-25', 'Competition', 'Second "Data Element ×" competition final', '2025年"数据要素×"大赛全国总决赛', 'Shanghai', 'Over 23,000 teams'),
|
| 27 |
+
('2025-12-30', 'Policy', 'National data work conference: 2026 named the year of releasing data value', '全国数据工作会议:"数据要素价值释放年"', 'Beijing', ''),
|
| 28 |
+
('2026-10-23', 'Competition', 'Third "Data Element ×" competition final', '2026年"数据要素×"大赛全国总决赛', 'Qingdao', 'Over 29,000 teams; last year of the three-year plan')]
|
| 29 |
+
pol = pd.DataFrame(P, columns=['date', 'kind', 'event', 'event_zh', 'city', 'note'])
|
| 30 |
+
pol['date'] = pd.to_datetime(pol.date); pol['year'] = pol.date.dt.year
|
| 31 |
+
pol['date_precision'] = 'day'
|
| 32 |
+
pol.loc[pol.event.str.startswith('First "Data'), 'date_precision'] = 'on or shortly before (reported 27 October 2024)'
|
| 33 |
+
pol.to_parquet(f'{OUT}/policy_timeline.parquet', index=False)
|
| 34 |
+
|
| 35 |
+
# ------------------------------------------------------------------ indicators (long format)
|
| 36 |
+
I = [('data_production', 2023, 32.85, 'ZB', 'National data resources survey 2023 (NDA, May 2024)'),
|
| 37 |
+
('data_production', 2024, 41.06, 'ZB', 'National data resources survey 2024 (NDA, 29 April 2025)'),
|
| 38 |
+
('data_production', 2025, 52.26, 'ZB', 'National data resources survey 2025 (NDA, 29 April 2026)'),
|
| 39 |
+
('data_production_growth', 2024, 25.0, '%', 'National data resources survey 2024'),
|
| 40 |
+
('data_production_growth', 2025, 27.28, '%', 'National data resources survey 2025'),
|
| 41 |
+
('data_production_per_person', 2024, 31.31, 'TB', 'National data resources survey 2024'),
|
| 42 |
+
('data_storage', 2025, 2.53, 'ZB', 'National data resources survey 2025'),
|
| 43 |
+
('share_of_world_data_production', 2025, 27.44, '%', 'National data resources survey 2025'),
|
| 44 |
+
('ai_data_training_and_inference', 2025, 199.48, 'EB', 'National data resources survey 2025'),
|
| 45 |
+
('ai_data_inference', 2025, 101.34, 'EB', 'National data resources survey 2025'),
|
| 46 |
+
('high_quality_datasets', 2025, 110000, 'datasets (lower bound)', 'National data resources survey 2025'),
|
| 47 |
+
('high_quality_datasets_volume', 2025, 908, 'PB (lower bound)', 'National data resources survey 2025'),
|
| 48 |
+
('active_data', 2025, 1.67, 'ZB', 'NDA press conference, 11 June 2026'),
|
| 49 |
+
('data_market_trading', 2024, 1600, '100 million yuan (lower bound)', 'NDA director at the national data work conference, January 2025'),
|
| 50 |
+
('data_exchange_trading', 2024, 300, '100 million yuan (lower bound)', 'NDA director at the national data work conference, January 2025'),
|
| 51 |
+
('data_enterprises', 2024, 190000, 'enterprises (lower bound)', 'Xinhua, 28 May 2025'),
|
| 52 |
+
('digital_economy', 2024, 59.2, 'trillion yuan', 'CAICT, China digital economy development report 2025 (March 2026)'),
|
| 53 |
+
('digital_economy_share_of_gdp', 2023, 42.8, '%', 'CAICT, implied by the 2025 report (+1 point in 2024)'),
|
| 54 |
+
('digital_economy_share_of_gdp', 2024, 43.8, '%', 'CAICT, China digital economy development report 2025 (March 2026)'),
|
| 55 |
+
('listed_firms_booking_data_assets', 2025, 136, 'A-share companies', 'SAIF data asset tracking report, via Yicai, 24 June 2026'),
|
| 56 |
+
('data_assets_booked', 2025, 37.86, '100 million yuan', 'SAIF data asset tracking report, via Yicai, 24 June 2026'),
|
| 57 |
+
('data_assets_booked_intangible', 2025, 31.20, '100 million yuan', 'SAIF data asset tracking report, via Yicai, 24 June 2026'),
|
| 58 |
+
('data_assets_booked_development', 2025, 6.44, '100 million yuan', 'SAIF data asset tracking report, via Yicai, 24 June 2026'),
|
| 59 |
+
('data_assets_booked_inventory', 2025, 0.22, '100 million yuan', 'SAIF data asset tracking report, via Yicai, 24 June 2026')]
|
| 60 |
+
ind = pd.DataFrame(I, columns=['indicator', 'year', 'value', 'unit', 'source'])
|
| 61 |
+
ind.to_parquet(f'{OUT}/indicators.parquet', index=False)
|
| 62 |
+
|
| 63 |
+
# ------------------------------------------------------------------ the "Data Element ×" competition
|
| 64 |
+
K = [(2024, 'Beijing', 19000, 100000, None, 12, 119, 'People\'s Daily, 28 October 2024: over 19,000 teams, nearly 100,000 participants'),
|
| 65 |
+
(2025, 'Shanghai', 23000, 117000, 893, 14, 183, 'National Data Administration, 28 November 2025'),
|
| 66 |
+
(2026, 'Qingdao', 29000, 145000, 330, None, None, 'National Data Administration, 23 October 2026')]
|
| 67 |
+
comp = pd.DataFrame(K, columns=['year', 'final_city', 'teams_registered', 'participants', 'projects_in_national_final', 'tracks', 'award_winning_teams', 'source'])
|
| 68 |
+
comp['figures_are'] = 'lower bounds ("over"), except 2024 participants ("nearly")'
|
| 69 |
+
comp.to_parquet(f'{OUT}/competition.parquet', index=False)
|
| 70 |
+
|
| 71 |
+
# ------------------------------------------------------------------ places: exchanges and competition finals
|
| 72 |
+
C = {'Guiyang': (26.647, 106.630), 'Beijing': (39.904, 116.407), 'Shanghai': (31.230, 121.474), 'Guangzhou': (23.129, 113.264),
|
| 73 |
+
'Shenzhen': (22.543, 114.058), 'Qingdao': (36.067, 120.383)}
|
| 74 |
+
pl = pol[pol.kind.isin(['Exchange', 'Competition'])][['city', 'kind', 'event', 'year']].copy()
|
| 75 |
+
pl['latitude'] = pl.city.map(lambda c: C[c][0]); pl['longitude'] = pl.city.map(lambda c: C[c][1])
|
| 76 |
+
pl.to_parquet(f'{OUT}/places.parquet', index=False)
|
| 77 |
+
json.dump(dict(events=len(pol), indicators=len(ind)), open('out/stats.json', 'w'))
|
| 78 |
+
print(pol.kind.value_counts().to_dict(), len(ind))
|
code/viz.py
ADDED
|
@@ -0,0 +1,128 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import json, os, glob, numpy as np, pandas as pd, matplotlib
|
| 2 |
+
matplotlib.use('Agg')
|
| 3 |
+
import matplotlib.pyplot as plt
|
| 4 |
+
from matplotlib import font_manager as fm
|
| 5 |
+
from matplotlib.patches import Polygon as MPoly
|
| 6 |
+
from matplotlib.lines import Line2D
|
| 7 |
+
from shapely.geometry import shape, box
|
| 8 |
+
|
| 9 |
+
for f in glob.glob('fonts/*.otf') + ['/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc', '/usr/share/fonts/opentype/noto/NotoSansCJK-Bold.ttc']:
|
| 10 |
+
if os.path.exists(f): fm.fontManager.addfont(f)
|
| 11 |
+
_n = {f.name for f in fm.fontManager.ttflist}
|
| 12 |
+
FONT = next((n for n in ('Noto Sans CJK SC', 'Noto Sans CJK JP') if n in _n), 'DejaVu Sans')
|
| 13 |
+
plt.rcParams.update({'font.family': FONT, 'font.size': 10, 'axes.edgecolor': '#BBBBBB', 'axes.linewidth': .6,
|
| 14 |
+
'xtick.color': '#666666', 'ytick.color': '#666666', 'axes.labelcolor': '#444444', 'text.color': '#1F1F1F',
|
| 15 |
+
'savefig.dpi': 200})
|
| 16 |
+
INK, MID, LIGHT, LAND, PAPER = '#1F1F1F', '#8A8A8A', '#C9C6BF', '#ECE9E3', '#FFFFFF'
|
| 17 |
+
RED, RED_SOFT, BLUE, BLUE_SOFT = '#B33A2B', '#E3B5AE', '#2F4B7C', '#B9C4D6'
|
| 18 |
+
os.makedirs('out/assets', exist_ok=True)
|
| 19 |
+
|
| 20 |
+
pol = pd.read_parquet('out/data/policy_timeline.parquet')
|
| 21 |
+
ind = pd.read_parquet('out/data/indicators.parquet')
|
| 22 |
+
comp = pd.read_parquet('out/data/competition.parquet')
|
| 23 |
+
pl = pd.read_parquet('out/data/places.parquet')
|
| 24 |
+
KIND = {'Policy': (RED, 'o'), 'Law': (INK, 's'), 'Institution': (RED, 'D'), 'Accounting': (BLUE, 'D'), 'Exchange': (BLUE, 'o'), 'Competition': (MID, '^')}
|
| 25 |
+
v = lambda k, y: ind[(ind.indicator == k) & (ind.year == y)].value.iloc[0]
|
| 26 |
+
|
| 27 |
+
def title(fig, t, s, y=.95):
|
| 28 |
+
fig.text(.05, y, t, fontsize=15, weight='bold')
|
| 29 |
+
fig.text(.05, y - .045, s, fontsize=9, color='#555555')
|
| 30 |
+
def clean(ax, left=True):
|
| 31 |
+
for s_ in ('top', 'right'): ax.spines[s_].set_visible(False)
|
| 32 |
+
if not left: ax.spines['left'].set_visible(False)
|
| 33 |
+
ax.tick_params(length=0)
|
| 34 |
+
|
| 35 |
+
# ------------------------------------------------------------------ 1. from strategy to market: the timeline
|
| 36 |
+
fig, ax = plt.subplots(figsize=(13, 8))
|
| 37 |
+
d = pol.sort_values('date').reset_index(drop=True)
|
| 38 |
+
x = d.date.dt.year + (d.date.dt.dayofyear - 1) / 366
|
| 39 |
+
for i, r in d.iterrows():
|
| 40 |
+
col, mk = KIND[r.kind]
|
| 41 |
+
ax.hlines(i, 2014.6, x[i], color='#F0F0F0', lw=.8, zorder=1)
|
| 42 |
+
ax.scatter(x[i], i, color=col, marker=mk, s=60, zorder=3)
|
| 43 |
+
ax.text(x[i] + .12, i, f"{r.date:%b %Y} {r.event_zh}", va='center', fontsize=7.6, color=MID)
|
| 44 |
+
ax.set_yticks(range(len(d))); ax.set_yticklabels(d.event, fontsize=8.2); ax.set_ylim(len(d) - .4, -1.2)
|
| 45 |
+
ax.axvspan(2024, 2027, color=LAND, lw=0, zorder=0); ax.text(2024.05, -.9, '"Data Element ×" plan, 2024 to 2026', fontsize=7.5, color=MID, va='bottom')
|
| 46 |
+
ax.set_xlim(2014.6, 2033.5); ax.set_xticks(range(2015, 2027)); clean(ax, left=False); ax.xaxis.grid(color='#F2F2F2', lw=.6); ax.set_axisbelow(True)
|
| 47 |
+
fig.legend(handles=[Line2D([], [], marker=m, ls='', color=c, label=k, ms=7) for k, (c, m) in KIND.items()], loc='upper left', bbox_to_anchor=(.045, .9), ncol=6, frameon=False, fontsize=8.5)
|
| 48 |
+
title(fig, 'From big data to data as a factor of production', 'The main laws, policies, institutions and markets for data in China, 2015 to 2026.', y=.965)
|
| 49 |
+
fig.text(.05, .02, 'Curated for this dataset from official releases (State Council, NPC, Ministry of Finance, National Data Administration, CAC) and press reports. '
|
| 50 |
+
'Date of adoption, publication or opening.', fontsize=7, color=MID)
|
| 51 |
+
plt.subplots_adjust(left=.36, right=.98, top=.85, bottom=.07)
|
| 52 |
+
fig.savefig('out/assets/policy_timeline.png', facecolor=PAPER); plt.close(fig)
|
| 53 |
+
|
| 54 |
+
# ------------------------------------------------------------------ 2. the numbers: production, AI data, market, balance sheets
|
| 55 |
+
fig, axs = plt.subplots(1, 4, figsize=(14, 5.2), gridspec_kw=dict(width_ratios=[1.1, .8, 1, 1.1], wspace=.55))
|
| 56 |
+
a = axs[0]; yrs = [2023, 2024, 2025]; vals = [v('data_production', y) for y in yrs]
|
| 57 |
+
a.bar(yrs, vals, color=[LIGHT, LIGHT, RED], width=.6)
|
| 58 |
+
for y, val in zip(yrs, vals): a.text(y, val + 1, f'{val:.1f}', ha='center', fontsize=9)
|
| 59 |
+
for y, g in [(2024, 25.0), (2025, 27.28)]: a.text(y, 3, f'+{g:.0f}%', ha='center', fontsize=8, color=PAPER, weight='bold')
|
| 60 |
+
a.set_xticks(yrs); a.set_ylim(0, 62); a.set_yticks([]); clean(a, left=False)
|
| 61 |
+
a.set_title('Data produced in China\n(zettabytes a year)', fontsize=9.5, loc='left')
|
| 62 |
+
a = axs[1]; tot, inf = v('ai_data_training_and_inference', 2025), v('ai_data_inference', 2025)
|
| 63 |
+
a.bar([0], [tot - inf], color=BLUE, width=.55, label='training'); a.bar([0], [inf], bottom=[tot - inf], color=RED, width=.55, label='inference')
|
| 64 |
+
a.text(0, (tot - inf) / 2, f'training\n{tot - inf:.1f}', ha='center', va='center', color=PAPER, fontsize=8.5)
|
| 65 |
+
a.text(0, tot - inf + inf / 2, f'inference\n{inf:.1f}', ha='center', va='center', color=PAPER, fontsize=8.5)
|
| 66 |
+
a.set_xlim(-.6, .6); a.set_xticks([]); a.set_yticks([]); clean(a, left=False)
|
| 67 |
+
a.set_title('Data used for AI, 2025\n(exabytes)', fontsize=9.5, loc='left')
|
| 68 |
+
a = axs[2]; m, e = v('data_market_trading', 2024), v('data_exchange_trading', 2024)
|
| 69 |
+
a.barh([1, 0], [m, e], color=[INK, BLUE], height=.55)
|
| 70 |
+
a.text(m, 1, f' {m / 10:,.0f} bn+', va='center', fontsize=9); a.text(e, 0, f' {e / 10:,.0f} bn+', va='center', fontsize=9)
|
| 71 |
+
a.set_yticks([1, 0]); a.set_yticklabels(['All data\ntrading', 'On data\nexchanges'], fontsize=8.5); a.set_xlim(0, 2300); a.set_xticks([]); clean(a, left=False)
|
| 72 |
+
a.set_title('Data trading, 2024\n(yuan)', fontsize=9.5, loc='left')
|
| 73 |
+
a = axs[3]; parts = [('Intangible\nassets', v('data_assets_booked_intangible', 2025), BLUE), ('Development\nspending', v('data_assets_booked_development', 2025), MID),
|
| 74 |
+
('Inventory', v('data_assets_booked_inventory', 2025), LIGHT)]
|
| 75 |
+
left = 0
|
| 76 |
+
for nm, val, c in parts:
|
| 77 |
+
a.barh(0, val, left=left, color=c, height=.5)
|
| 78 |
+
if val > 2: a.text(left + val / 2, 0, f'{val:.1f}', ha='center', va='center', color=PAPER, fontsize=8.5)
|
| 79 |
+
if val > 2: a.text(left + val / 2, -.42, nm, ha='center', va='top', fontsize=7.5, color=MID)
|
| 80 |
+
left += val
|
| 81 |
+
a.set_xlim(0, 40); a.set_ylim(-1, .6); a.set_xticks([]); a.set_yticks([]); clean(a, left=False)
|
| 82 |
+
a.set_title(f'Data booked as assets by\n{int(v("listed_firms_booking_data_assets", 2025))} listed firms, 2025 (100m yuan)', fontsize=9.5, loc='left')
|
| 83 |
+
a.text(40, .45, 'inventory 0.2', ha='right', fontsize=7.5, color=MID)
|
| 84 |
+
title(fig, 'Data in numbers', 'Data produced each year, data used for AI, the data market and data on company balance sheets. China produced about 27% of the world\'s data in 2025.', y=.96)
|
| 85 |
+
fig.text(.05, .03, 'Data: National Data Administration (national data resources surveys 2023 to 2025; national data work conference, January 2025); SAIF data asset tracking report via Yicai (June 2026). '
|
| 86 |
+
'"+" marks lower bounds ("over"). Surveys started in 2023, so there is no earlier comparable series.', fontsize=7, color=MID)
|
| 87 |
+
plt.subplots_adjust(left=.05, right=.98, top=.74, bottom=.14)
|
| 88 |
+
fig.savefig('out/assets/data_in_numbers.png', facecolor=PAPER); plt.close(fig)
|
| 89 |
+
|
| 90 |
+
# ------------------------------------------------------------------ 3. the competition and where it happens
|
| 91 |
+
fig = plt.figure(figsize=(13, 6.2))
|
| 92 |
+
a1 = fig.add_axes([.05, .12, .36, .62]); a2 = fig.add_axes([.47, .06, .5, .78])
|
| 93 |
+
yrs = comp.year.values
|
| 94 |
+
a1.bar(yrs - .18, comp.teams_registered / 1e3, width=.34, color=INK, label='Teams registered')
|
| 95 |
+
a1.bar(yrs + .18, comp.participants / 1e3, width=.34, color=RED, label='Participants')
|
| 96 |
+
for y, t, p_, c in zip(yrs, comp.teams_registered, comp.participants, comp.final_city):
|
| 97 |
+
a1.text(y - .18, t / 1e3 + 2, f'{t / 1e3:.0f}k', ha='center', fontsize=8.5)
|
| 98 |
+
a1.text(y + .18, p_ / 1e3 + 2, f'{p_ / 1e3:.0f}k', ha='center', fontsize=8.5, color=RED)
|
| 99 |
+
a1.text(y, -12, f'final: {c}', ha='center', fontsize=8, color=MID)
|
| 100 |
+
a1.set_xticks(yrs); a1.set_ylim(0, 165); a1.set_yticks([]); clean(a1, left=False); a1.legend(frameon=False, fontsize=8.5, loc='upper left')
|
| 101 |
+
a1.set_title('The "Data Element ×" competition', fontsize=10, loc='left')
|
| 102 |
+
EXT = (97, 128, 18, 43)
|
| 103 |
+
clip = box(EXT[0] - 2, EXT[2] - 2, EXT[1] + 2, EXT[3] + 2)
|
| 104 |
+
for f in json.load(open('raw/ne_50m_admin_0_countries.geojson'))['features']:
|
| 105 |
+
if f['properties']['ADM0_A3'] == 'CHN': continue
|
| 106 |
+
g = shape(f['geometry'])
|
| 107 |
+
if not g.intersects(clip): continue
|
| 108 |
+
for p in getattr(g.intersection(clip), 'geoms', [g.intersection(clip)]):
|
| 109 |
+
if p.geom_type == 'Polygon': a2.add_patch(MPoly(np.array(p.exterior.coords), closed=True, fc='#F5F3EF', ec='#E2DFD9', lw=.4))
|
| 110 |
+
for f in json.load(open('raw/china_provinces.geojson'))['features']:
|
| 111 |
+
g = shape(f['geometry'])
|
| 112 |
+
for p in getattr(g, 'geoms', [g]): a2.add_patch(MPoly(np.array(p.exterior.coords), closed=True, fc=LAND, ec=PAPER, lw=.6))
|
| 113 |
+
for c, gp in pl.groupby('city'):
|
| 114 |
+
la, lo = gp.latitude.iloc[0], gp.longitude.iloc[0]
|
| 115 |
+
kinds = gp.kind.tolist()
|
| 116 |
+
if 'Exchange' in kinds: a2.scatter(lo, la, s=90, marker='o', c=BLUE, ec=PAPER, lw=.8, zorder=4)
|
| 117 |
+
if 'Competition' in kinds: a2.scatter(lo, la, s=200, marker='o', fc='none', ec=RED, lw=1.6, zorder=5)
|
| 118 |
+
lab = '\n'.join(f'{r.year} {"exchange" if r.kind == "Exchange" else "competition final"}' for _, r in gp.sort_values('year').iterrows())
|
| 119 |
+
dx, ha = (-10, 'right') if c in ('Guangzhou', 'Guiyang') else (10, 'left')
|
| 120 |
+
a2.annotate(f'{c}\n{lab}', (lo, la), xytext=(dx, 0), textcoords='offset points', ha=ha, va='center', fontsize=7.8, zorder=6, linespacing=1.2)
|
| 121 |
+
a2.set_xlim(EXT[0], EXT[1]); a2.set_ylim(EXT[2], EXT[3]); a2.set_aspect(1 / np.cos(np.radians(30)))
|
| 122 |
+
a2.set_xticks([]); a2.set_yticks([]); [s_.set_visible(False) for s_ in a2.spines.values()]
|
| 123 |
+
a2.legend(handles=[Line2D([], [], marker='o', ls='', color=BLUE, ms=8, label='Data exchange (selection)'),
|
| 124 |
+
Line2D([], [], marker='o', ls='', mfc='none', mec=RED, mew=1.6, ms=11, label='Competition final')], loc='lower right', frameon=False, fontsize=8.5)
|
| 125 |
+
title(fig, 'A national contest and a few exchanges', 'Registrations to the "Data Element ×" competition grew by half in two years; the first data exchanges opened in Guiyang, Beijing, Shanghai and the Greater Bay Area.', y=.96)
|
| 126 |
+
fig.text(.05, .015, 'Data: National Data Administration and People\'s Daily; figures are lower bounds ("over"), except 2024 participants ("nearly 100,000"). Exchanges: the five main ones by opening date, not a complete list. Borders: Natural Earth.', fontsize=7, color=MID)
|
| 127 |
+
fig.savefig('out/assets/competition_and_exchanges.png', facecolor=PAPER); plt.close(fig)
|
| 128 |
+
print(sorted(os.listdir('out/assets')))
|
data/competition.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c2a40067a49477949d004a37069dd231fef7eeb18be085ef68adca7a04db4e3
|
| 3 |
+
size 6514
|
data/indicators.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d2f034a6374e2f48d3da724b4cddbb52d788764108e23401b888c1ed3ffff935
|
| 3 |
+
size 4405
|
data/places.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:966dbf8e6e44214e4d76bb5c270c73a287e626b48f03631c0e38ca08ede7e1bd
|
| 3 |
+
size 4194
|
data/policy_timeline.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2a2377a1c5fed9df5bb74117f45eb9852220599fa40214add08bf55f739c33b3
|
| 3 |
+
size 7277
|