Use when the identification strategy for a 《中国工业经济》 (China Industrial Economics) manuscript is multi-period / staggered DID or an event study. Mandates parallel-trends event-study plots, placebo tests, and modern heterogeneity-robust estimators (Callaway-Sant'Anna, Sun-Abraham, de Chaisemartin-D'Haultfœuille, Goodman-Bacon decomposition) whenever treatment timing varies.
SkillScore82Maintained
Automatic
◆Research pack
Last verified 8/31/2026
What it changes
Passed 18 of 22 test cases with the skill, and 14 of 22 without it.
Without skill64%
With skill82%
Checks it passed that the model alone missed
Contains output block header 【识别策略】
Directs the next step to cie-mechanism
Includes the exact section label 【识别策略】
and 4 more in the full report
4 fixed14 passed either way4 still failing0 broke
Pass-rate lift
+18 pp
Tokens
+57%
Agent turns
0%
Response time
-8%
Model gemini-3.6-flashMeasured 8/31/2026v118 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 4 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 4
Prompt
论文研究地方财政压力对企业污染排放的连续 DID 效应（财政压力指标×Post）。未画平行趋势图，未跑安慰剂置换，未论证压力指标外生性。请给出结构化诊断框。
What was checked, without and with the skill
Checked
Without skill
With skill
Contains output block header 【识别策略】
Missed
Passed
Directs the next step to cie-mechanism
Missed
Passed
case-20 · gemini-3.6-flash · 8/31/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md120 linessingle file
识别策略：多期 DID / 事件研究（cie-did-identification）
触发时机
实证主体是 DID / 多期（交错）DID / event-study
TWFE 用在交错处理上，但没回应异质性处理偏误
平行趋势只口头说"满足"，没画事件研究图
用连续型处理强度（continuous/dose DID）但识别假设没说清
本刊红线（缺一即高危退修）
平行趋势必须画事件研究图（动态效应 + 95% CI，处理前各期系数应不显著）
安慰剂检验必做（随机化处理时点/处理对象 500—1000 次，看真实估计是否落在分布尾部）
交错处理必须用异质性稳健估计——TWFE 在交错+异质性处理效应下有偏（负权重问题）
分支路径
分支 A：标准两期 / 单一时点 DID
平行趋势：事件研究图 + 处理前系数联合检验
安慰剂：随机分配处理组 ≥ 500 次
控制组合理性论证（为什么这些是有效对照）
分支 B：交错（staggered）DID —— 本刊高频
必须做 Goodman-Bacon 分解，报告"坏比较（已处理作对照）"权重
必须改用异质性稳健估计之一并作为主/稳健结果：
Callaway & Sant'Anna（2021）group-time ATT
Sun & Abraham（2021）交互加权事件研究
de Chaisemartin & D'Haultfœuille（2020）did_multiplegt
Use when building or auditing the robustness section of a 《中国工业经济》 (China Industrial Economics) manuscript — the journal's signature "arms race." Drives an exhaustive checklist: alternative measures, dropping competing policies, PSM-DID, changing windows/samples, placebo, winsorizing, and explicitly ruling out alternative explanations. Doing this thoroughly is the norm, not the exception.