Shale Mechanism Data: Oil-Directed Drilling, Associated Gas, and Henry Hub Dynamics (1998–2025)

This dataset supports the manuscript “Shale Mechanism: Oil Incentives, Associated Gas, and Henry Hub Dynamics.”

It contains the full data architecture used to estimate a mechanism-first econometric framework linking oil incentives to natural gas prices through delayed drilling and associated gas production in oil-directed basins.

The dataset includes:

Raw source files from U.S. Energy Information Administration (EIA), rig counts, and price benchmarks
Processed monthly panels used for econometric estimation (1998–2025)
Basin-level datasets distinguishing oil-directed and gas-directed production regions
Derived variables including associated-gas aggregates, storage flows, and system-state controls
Feature matrices used in the machine-learning screening stage
Supporting documentation, including variable definitions and methodological structure

The analytical framework tests a conditional transmission chain:
oil prices → delayed drilling → oil-directed associated gas → Henry Hub

The dataset enables full reproducibility of:

Descriptive timing and correlation analysis
Distributed-lag drilling models
Basin-level associated-gas estimation
Henry Hub econometric models
Cointegration and restricted VECM long-run analysis
Machine-learning screening and translation-back models

Forecast periods present in raw EIA files are excluded from the historical identification sample. The final econometric sample runs from January 1998 to December 2025.

本数据集支持学术论文《页岩机理：石油激励、伴生气与亨利枢纽（Henry Hub）动态》。

其包含完整的数据架构，可用于估算以机理为核心的计量经济框架，该框架通过石油导向型盆地中的延迟钻井与伴生气生产，关联石油激励与天然气价格。

本数据集包含：
1. 来自美国能源信息署（U.S. Energy Information Administration, EIA）的原始源文件、钻机数量数据与价格基准；
2. 用于计量经济估算的经处理月度面板数据（1998年—2025年）；
3. 区分石油导向型与天然气导向型生产区域的盆地级数据集；
4. 各类衍生变量，包括伴生气总量、仓储流量与系统状态控制变量；
5. 机器学习筛选阶段所用的特征矩阵；
6. 配套文档，涵盖变量定义与方法论框架。

本分析框架旨在检验一条条件传导链条：
油价 → 延迟钻井 → 石油导向型伴生气 → 亨利枢纽（Henry Hub）

本数据集可完整复现以下分析：
- 描述性时序与相关性分析
- 分布滞后钻井模型
- 盆地级伴生气估算
- 亨利枢纽计量经济模型
- 协整与受限向量误差修正模型（Vector Error Correction Model, VECM）长期分析
- 机器学习筛选与反向翻译模型

原始EIA文件中包含的预测期已被排除在历史识别样本之外。最终计量经济样本的时间范围为1998年1月至2025年12月。