Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when choosing or defending the data behind a 《世界经济》 (The Journal of World Economy) manuscript — customs micro data (firm-product-destination), cross-country panels, and GVC / input-output decomposition (KWW value-added, ADB / WIOD MRIO). Steers away from generic provincial panels toward internationally-credible sources.
.claude/skills/bilal140202-jwe-data-sources/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 35% | 0% |
> 《世界经济》（中国世界经济学会 + 中国社会科学院世界经济与政治研究所 IWEP 主办，1978 年创刊，月刊）是国内国际经济学/开放宏观重镇。本刊审稿人对数据的"国际属性"尤为敏感：换成《经济研究》/《金融研究》，一套省级或上市公司面板即可成稿；在本刊，数据本身必须带贸易/跨境/GVC 维度，否则"开放经济"立不住。基础事实见 resources/journal-profile.md，来源见 resources/official-source-map.md。
本刊（IWEP + 中国世界经济学会主办的 CSSCI 权威顶级期刊）审稿人对数据的"国际属性"尤为敏感。下表把退修信号显性化。
| 审稿期待 | 达标 | 退修/退稿信号 | |----------|------|----------------| | 数据带贸易/跨境/GVC 维度 | 海关 firm-product-destination、UN Comtrade/BACI、MRIO | 省级或上市公司面板，国际性靠控制变量 | | 海关库匹配可复现 | 写清跨年匹配、剔中间商、清洗规则 | 不交代企业代码逐年变动与代理商处理 | | GVC 分解可追溯 | KWW + 具体 MRIO（WIOD/ADB/OECD ICIO）与年份 | 仅称"基于 GVC"却无任何分解 | | 受限数据合规可复现 | 说明获取渠道、口径，清洗代码可复现 | 受限微观不可公开就不交代构造过程 |
设稿件研究"外资准入放开后下游企业的 GVC 嵌入变化"。按本 skill 配数据：
走查搭建：
核心微观 → 中国海关库（企业-HS-目的国）× 工业企业库，按信用代码+名称地址模糊匹配
GVC 测度 → 用 ADB MRIO（注明年份）做 KWW 分解，构造企业层 DVAR/上游度近似
暴露度 → 负面清单删减环节映射到下游行业（示意：覆盖约 37% 制造业行业）
跨国对照 → 以 WDI/CEPII BACI 校核行业贸易额口径
复现 → 海关原始微观受限，提交清洗脚本与变量构造代码（本刊 2023 年第 10 期起数据/代码公开）> 走查中的行业覆盖比例为示意占位，仅演示数据搭建决策，非真实统计。
本刊贸易/GVC 实证论文多以海关微观、跨国面板（UN Comtrade/BACI/WDI）或 MRIO 投入产出表为主力数据，并交代匹配清洗与分解口径；受限数据强调清洗与构造代码可复现。数据公开的具体执行口径以编辑部最新稿约为准。
【数据类型】海关微观 / 跨国面板 / GVC-MRIO | 一般国内（需升级）
【匹配/清洗】<方法> | 待补
【GVC 分解】KWW + WIOD/ADB/OECD | 仅叙事
【暴露度构造】<进口渗透/出口依赖/GVC联系>
【下一步】jwe-identification| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 57,283 | 22,265 | -61% | 1 | 1 | 0% | 4,231 | 5,086 | +20% | 0 | 0 | — |
case-02 | fail→pass | 29,914 | 21,039 | -30% | 1 | 1 | 0% | 4,671 | 5,021 | +7% | 0 | 0 | — |
case-03 | fail→pass | 25,190 | 19,035 | -24% | 1 | 1 | 0% | 3,909 | 4,614 | +18% | 0 | 0 | — |
case-04 | pass→pass | 24,728 | 15,476 | -37% | 1 | 1 | 0% | 3,694 | 4,211 | +14% | 0 | 0 | — |
case-05 | pass→pass | 23,368 | 24,056 | +3% | 1 | 1 | 0% | 3,467 | 5,312 | +53% | 0 | 0 | — |
case-06 | pass→pass | 15,516 | 15,723 | +1% | 1 | 1 | 0% | 2,574 | 4,317 | +68% | 0 | 0 | — |
case-07 | fail→pass | 15,182 | 15,537 | +2% | 1 | 1 | 0% | 2,226 | 4,059 | +82% | 0 | 0 | — |
case-08 | pass→pass | 27,264 | 23,823 | -13% | 1 | 1 | 0% | 4,498 | 5,758 | +28% | 0 | 0 | — |
case-09 | pass→fail | 25,698 | 23,541 | -8% | 1 | 1 | 0% | 4,090 | 5,532 | +35% | 0 | 0 | — |
case-10 | pass→pass | 20,057 | 18,906 | -6% | 1 | 1 | 0% | 3,251 | 4,645 | +43% | 0 | 0 | — |
case-11 | pass→pass | 30,293 | 27,240 | -10% | 1 | 1 | 0% | 6,008 | 6,907 | +15% | 0 | 0 | — |
case-12 | fail→fail | 25,378 | 18,301 | -28% | 1 | 1 | 0% | 3,642 | 4,711 | +29% | 0 | 0 | — |
case-13 | pass→pass | 28,609 | 26,863 | -6% | 1 | 1 | 0% | 4,521 | 6,109 | +35% | 0 | 0 | — |
case-14 | pass→pass | 26,784 | 23,866 | -11% | 1 | 1 | 0% | 3,606 | 5,301 | +47% | 0 | 0 | — |
case-15 | fail→pass | 26,347 | 25,488 | -3% | 1 | 1 | 0% | 4,398 | 5,920 | +35% | 0 | 0 | — |
case-16 | pass→pass | 21,465 | 21,093 | -2% | 1 | 1 | 0% | 3,380 | 5,199 | +54% | 0 | 0 | — |
case-17 | pass→pass | 24,467 | 23,757 | -3% | 1 | 1 | 0% | 3,750 | 5,414 | +44% | 0 | 0 | — |
case-18 | pass→pass | 19,798 | 41,419 | +109% | 1 | 1 | 0% | 2,990 | 4,442 | +49% | 0 | 0 | — |
case-19 | pass→pass | 17,983 | 17,002 | -5% | 1 | 1 | 0% | 3,015 | 4,501 | +49% | 0 | 0 | — |
case-20 | pass→pass | 23,312 | 19,315 | -17% | 1 | 1 | 0% | 3,157 | 4,530 | +43% | 0 | 0 | — |
case-21 | fail→pass | 18,084 | 22,568 | +25% | 1 | 1 | 0% | 2,676 | 4,856 | +81% | 0 | 0 | — |
case-22 | fail→pass | 23,271 | 19,251 | -17% | 1 | 1 | 0% | 3,846 | 4,665 | +21% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +27 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.