Supply Chain Transmission of Hot Weather Shocks and Adaptive Behavior of Downstream Firms: An Empirical Study Based on Listed Companies in China

The evidences of temperature, supply chain, and adaptive procurement shift are matched observations from three databases. The weather data are from the National Meteorological Information Center[ The statistics agency branch of China Meteorological Administration. ]; the supply chain data are from China Stock Market & Accounting Research Database (CSMAR); and the trade volume data are from China Customs Database (CCD)[ Published by the General Administration of Customs of the People’s Republic of China. ]. We collect supplier-customer pairs from major suppliers disclosed in the semi-annual reports of China A-share listed companies and use the pairs as the basic observations. Eliminating irregular[ To be specific, we eliminate the pairs with missing values in key explained and explanatory variables; eliminate observations that violate basic accounting principles, such as samples with current assets, fixed assets or net fixed assets greater than total assets; eliminate companies in the financial industries; eliminate companies with ST or PT in the current year; and eliminate companies of negative age. We then match the customer firms in every supplier-customer pair with establishments in the China Customs Database. All pairs with customer firm’s imports data will be kept in the sample. ] data points, we obtain the max sample of 3480 observations spanning 2010 to 2016.

气温、供应链与适应性采购变动的相关证据来自三个数据库的匹配观测样本。其中，气象数据取自国家气象信息中心（National Meteorological Information Center），该机构为中国气象局的统计分支机构；供应链数据取自中国股票市场与会计研究数据库（China Stock Market & Accounting Research Database，CSMAR）；贸易量数据取自中国海关数据库（China Customs Database，CCD），该数据库由中华人民共和国海关总署发布。我们从中国A股上市公司半年度报告披露的主要供应商信息中提取供应商-客户配对数据，并将其作为基础观测样本。针对异常数据点，我们具体执行如下清洗规则：剔除关键被解释变量与解释变量存在缺失值的配对；剔除违反基本会计原则的观测样本，例如流动资产、固定资产或固定资产净额大于总资产的样本；剔除金融行业上市公司；剔除当年被实施ST或PT处理的公司；剔除成立年限为负的公司。随后，我们将每一组供应商-客户配对中的客户企业与中国海关数据库中的企业进行匹配，仅保留拥有进口数据的客户企业所对应的配对样本。经上述数据清洗后，最终得到覆盖2010至2016年的3480个有效观测样本，即最大可用样本量。