CA-TPU 指数编制方法论

Methodology of the Central Asia Trade Policy Uncertainty Index

新疆大学经济与管理学院 · 张立杰教授团队
Prof. Zhang Lijie's Team, School of Economics and Management, Xinjiang University

中文
English

一、概述

中亚贸易政策不确定性指数（Central Asia Trade Policy Uncertainty Index，简称 CA-TPU）是基于 Baker, Bloom & Davis (2016) 提出的经济政策不确定性指数（EPU）方法论，针对中亚区域贸易政策环境专门构建的量化指标。

该指数覆盖中亚四国：哈萨克斯坦（KZ）、乌兹别克斯坦（UZ）、吉尔吉斯斯坦（KG）和塔吉克斯坦（TJ），通过对中亚主流俄语新闻媒体的系统性文本挖掘，按月度发布。基期设定为 2015–2019 年，基期均值标准化为 100。指数高于 100 表示不确定性高于基期平均水平。

为什么选择中亚？ 中亚地区地处中国"一带一路"倡议与欧亚经济联盟（EAEU）的交汇地带，其贸易政策环境受到大国博弈、地缘政治和区域一体化进程的多重影响，但此前缺乏系统性的量化工具。CA-TPU 旨在填补这一空白。

二、新闻数据采集

2.1 数据来源

本指数从中亚地区 8 家主要俄语新闻媒体网站采集新闻报道：

国家媒体网站类型
哈萨克斯坦www.inform.kz综合新闻
forbes.kz财经新闻
乌兹别克斯坦uza.uz官方通讯社
kun.uz综合新闻
吉尔吉斯斯坦kaktus.media综合新闻
24.kg综合新闻
塔吉克斯坦asiaplustj.info综合新闻
avesta.tj综合新闻

2.2 采集流程

CA-TPU 项目于 2025 年启动，回溯采集上述 8 家媒体自 2009 年以来发布的新闻数据，并实现每日增量更新。截至目前，数据库累计收录新闻报道超过 308 万篇。

关于早期数据：由于部分媒体网站早期存档不完整，2009–2012 年间的新闻数据存在不同程度的缺失。从 2013 年起，各国各媒体的数据覆盖趋于稳定和完整，因此本指数在实际发布中默认从 2013 年 1 月开始展示。
关于土库曼斯坦：由于土库曼斯坦的独立新闻媒体资源极为有限，目前尚无法纳入指数体系。

三、关键词体系

参照 BBD (2016) 三类关键词框架，构建了适用于中亚地区的俄语关键词表。由于俄语为屈折语言，匹配采用词干匹配或包含全部主要变格形式。

3.1 T 类：贸易关键词（Торговля）

俄语原形中文含义主要变格/派生形式
торговля贸易торговли, торговле, торговлю, торговлей, торговый, торговая
экспорт出口экспорта, экспорту, экспортом, экспортный, экспортёр
импорт进口импорта, импорту, импортом, импортный, импортёр
таможня海关таможни, таможне, таможню, таможенный, таможенных
тариф关税тарифа, тарифу, тарифом, тарифы, тарифный
пошлина税пошлины, пошлине, пошлину, пошлинами
санкции制裁санкций, санкциям, санкциями, санкционный
логистика物流логистики, логистике, логистику, логистический
транзит过境运输транзита, транзиту, транзитом, транзитный
товарооборот贸易额товарооборота, товарообороту, товарооборотом
контрабанда走私контрабанды, контрабанде, контрабанду
квота配额квоты, квоте, квоту, квотирование

3.2 P 类：政策关键词（Политика）

俄语原形中文含义主要变格/派生形式
политика政策политики, политике, политику, политический
правительство政府правительства, правительству, правительственный
закон法律закона, закону, законом, законодательство
регулирование监管регулирования, регулированию, регулятор
реформа改革реформы, реформе, реформу, реформирование
президент总统президента, президенту, президентский
парламент议会парламента, парламенту, парламентский
министерство部委министерства, министерству, министр
соглашение协定соглашения, соглашению, соглашений
договор条约договора, договору, договоры, договоров
указ总统令указа, указу, указом, указы
постановление政府决议постановления, постановлению, постановлений

3.3 U 类：不确定性关键词（Неопределённость）

俄语原形中文含义主要变格/派生形式
неопределённость不确定性неопределённости, неопределённостью, неопределённый
риск风险риска, риску, риском, риски, рисков
кризис危机кризиса, кризису, кризисом, кризисный
нестабильность不稳定性нестабильности, нестабильностью, нестабильный
угроза威胁угрозы, угрозе, угрозу, угрожать
волатильность波动性волатильности, волатильностью, волатильный
опасение担忧опасения, опасений, опасениям
непредсказуемость不可预测性непредсказуемости, непредсказуемый
напряжённость紧张局势напряжённости, напряжённостью, напряжённый
дестабилизация不稳定化дестабилизации, дестабилизацию

注：以上为核心关键词示例。实际匹配采用词干匹配技术，完整关键词表定期更新。

3.4 命中规则

标记含义逻辑
OR命中任意一类关键词T ∪ P ∪ U
AND同时命中三类关键词T ∩ P ∩ U

只有 AND 命中的新闻才纳入指数计算。

四、指数计算方法

1 各国月度 TPU 比率

合并同一国家所有站点计算，当 时剔除。

为什么按国家合并？ 部分国家仅 1–2 个新闻源，单站点月新闻量极少时会导致比率极端。合并后更稳健。

2 标准差标准化

为基期（2015-01 ~ 2019-12）内的样本标准差。

3 基期指数化

4 贸易额加权合成

权重基于上一年度各国对外贸易总额占比，采用归一化处理。

国家2025 年贸易额2026 年权重
哈萨克斯坦1,439 亿美元57.18%
乌兹别克斯坦812 亿美元32.27%
吉尔吉斯斯坦158 亿美元6.28%
塔吉克斯坦107 亿美元4.27%

五、平滑处理

为什么使用"向后看"移动平均？ 确保指数一经发布即为终稿，不会追溯修改。

六、指数解读

区间含义
< 80偏低，环境相对稳定
80 – 120正常区间
120 – 150偏高，需关注
> 150显著升高，通常对应重大事件

历史高峰期

时期MA3 峰值驱动事件
2014 年 3–5 月~145克里米亚危机
2015 年 6–10 月~145哈坚戈自由浮动
2020 年 3–8 月~174COVID-19 边境封锁
2022 年 2–5 月~140俄乌冲突制裁溢出
2025 末–2026 初~166美国全球关税战

七、数据更新与发布

  • 更新频率：月度，每月初发布上月指数
  • 数据截止：每月最后一日 23:59（UTC+6）
  • 不可修改性：原始指数与 MA3 一经发布即为终稿

八、参考文献

Baker, S. R., Bloom, N., & Davis, S. J. (2016). Measuring Economic Policy Uncertainty. The Quarterly Journal of Economics, 131(4), 1593–1636. https://doi.org/10.1093/qje/qjw024

技术支持声明
新疆中亚经济与环境研究所为本项目提供了网站运维和数据服务器资源，但未参与研究工作。全部研究成果均由新疆大学经济与管理学院张立杰教授团队完整拥有。

免责声明
本指数仅供学术研究与信息参考之用，不构成投资建议。使用者据此做出的商业决策，责任自行承担。

1. Overview

The Central Asia Trade Policy Uncertainty Index (CA-TPU) is a quantitative indicator specifically designed to measure trade policy uncertainty across Central Asia, adapting the Economic Policy Uncertainty (EPU) methodology proposed by Baker, Bloom & Davis (2016).

The index covers four Central Asian countries — Kazakhstan (KZ), Uzbekistan (UZ), Kyrgyzstan (KG), and Tajikistan (TJ) — based on systematic text mining of major Russian-language news outlets in the region. It is published monthly with a base period of 2015–2019 (mean = 100). Values above 100 indicate above-average trade policy uncertainty.

Why Central Asia? Situated at the crossroads of China's Belt and Road Initiative and the Eurasian Economic Union (EAEU), the region's trade policy environment is shaped by great-power competition, geopolitics, and regional integration — yet no systematic quantitative tool previously existed to track this uncertainty. CA-TPU aims to fill that gap.

2. News Data Collection

2.1 Data Sources

The index collects news articles from 8 major Russian-language news outlets across the four countries:

CountryNews OutletType
Kazakhstanwww.inform.kzGeneral news
forbes.kzBusiness & finance
Uzbekistanuza.uzNational news agency
kun.uzGeneral news
Kyrgyzstankaktus.mediaGeneral news
24.kgGeneral news
Tajikistanasiaplustj.infoGeneral news
avesta.tjGeneral news

2.2 Collection Process

The CA-TPU project was launched in 2025. News data was retrospectively collected from the 8 outlets listed above, covering articles published since 2009, with daily incremental updates maintained thereafter. As of now, the database contains over 3.08 million news articles.

Note on early data: Due to incomplete archiving on some media websites, news data from 2009 to 2012 contains varying degrees of gaps across countries and outlets. Data coverage became stable and comprehensive from approximately 2013 onwards. Consequently, the index is displayed from January 2013 by default in its published form.
Note on Turkmenistan: Due to the extremely limited availability of independent news media, Turkmenistan cannot yet be included in the index framework.

3. Keyword Framework

Following the BBD (2016) three-category keyword framework, a Russian-language keyword dictionary was developed for the Central Asian context. As Russian is a highly inflected language, keyword matching uses stem matching or includes all major declension forms.

3.1 T: Trade Keywords (Торговля)

Keywords identifying news related to international trade: trade (торговля), export (экспорт), import (импорт), customs (таможня), tariff (тариф), duty (пошлина), sanctions (санкции), logistics (логистика), transit (транзит), trade volume (товарооборот), smuggling (контрабанда), quota (квота) — each with all major declension and derivation forms.

3.2 P: Policy Keywords (Политика)

Keywords identifying government policy actions: policy (политика), government (правительство), law (закон), regulation (регулирование), reform (реформа), president (президент), parliament (парламент), ministry (министерство), agreement (соглашение), treaty (договор), presidential decree (указ), government resolution (постановление).

3.3 U: Uncertainty Keywords (Неопределённость)

Keywords identifying uncertainty or risk sentiment: uncertainty (неопределённость), risk (риск), crisis (кризис), instability (нестабильность), threat (угроза), volatility (волатильность), concern (опасение), unpredictability (непредсказуемость), tension (напряжённость), destabilization (дестабилизация).

3.4 Matching Rules

FlagDefinitionLogic
ORMatches any one keyword categoryT ∪ P ∪ U
ANDMatches all three categories simultaneouslyT ∩ P ∩ U

Only AND-flagged articles are classified as TPU-related and included in the index calculation.

4. Index Calculation

1 Country-Level Monthly TPU Ratio

All news sources within a country are merged before computing the ratio. Months where are excluded.

Why merge by country? Some countries have only 1–2 news sources. Per-outlet calculation with very small monthly article counts can produce extreme ratios. Country-level aggregation yields a more robust index.

2 Standard Deviation Normalization

Where is the sample standard deviation of during the base period (Jan 2015 – Dec 2019).

3 Base-Period Indexation

Where is the base-period mean of .

4 Trade-Weighted Composite

Weights are based on each country's share of total foreign trade among the four countries, lagged by one year. Auto-normalization ensures the index can still be computed if one country's data is temporarily unavailable.

Country2025 Trade Volume2026 Weight
Kazakhstan$143.9 bn57.18%
Uzbekistan$81.2 bn32.27%
Kyrgyzstan$15.8 bn6.28%
Tajikistan$10.7 bn4.27%

5. Smoothing

Why a trailing (backward-looking) MA3? A centered MA3 requires future data that does not exist at publication time, causing retroactive revisions. The trailing MA3 uses only historical data, ensuring that published values are final and never revised — a critical property for a regularly released index.

6. Interpretation

RangeInterpretation
< 80Below average — relatively stable trade policy environment
80 – 120Normal range
120 – 150Elevated — warrants attention
> 150Significantly elevated — typically driven by major policy shocks

Historical Peaks

PeriodMA3 PeakDriving Events
Mar–May 2014~145Crimea crisis; first KZT devaluation
Jun–Oct 2015~145KZT free float; regional currency contagion
Mar–Aug 2020~174COVID-19 border closures and supply chain disruptions
Feb–May 2022~140Russia-Ukraine conflict; sanctions spillover
Late 2025–Early 2026~166US global tariff war; secondary sanctions threats

7. Updates & Publication

  • Frequency: Monthly, released in the first week following the reference month
  • Data cutoff: Last day of each month, 23:59 UTC+6
  • Finality: Both the raw index and MA3 are final upon publication — no retroactive revisions

8. References

Baker, S. R., Bloom, N., & Davis, S. J. (2016). Measuring Economic Policy Uncertainty. The Quarterly Journal of Economics, 131(4), 1593–1636. https://doi.org/10.1093/qje/qjw024

Technical Support Statement
Xinjiang – Central Asia Institute for Economic and Environmental Studies provided technical support for this project, including website hosting and data server resources, but did not participate in the research design, data analysis, or index development. All research outputs are solely owned by Prof. Zhang Lijie's team at the School of Economics and Management, Xinjiang University.

Disclaimer
This index is developed for academic research and informational purposes only. It does not constitute investment advice. Users bear full responsibility for any business decisions based on this data.