企业数据质量对实证研究结论偏差的潜在影响——来自2015年中国企业-员工匹配调查的经验证据  

Impact of Enterprise Data Quality on Empirical Research Conclusions——From 2015 China Employer- Employee Survey

作　　者：程虹[1] 许伟[1] 李唐[1] 

机构地区：[1]武汉大学质量发展战略研究院,湖北武汉430072

出　　处：《华中科技大学学报（社会科学版）》2016年第3期67-80,共14页Journal of Huazhong University of Science and Technology(Social Science Edition)

基　　金：教育部哲学社会科学研究重大课题攻关项目(15JZD023);国家科技支撑计划课题(2015BAH27F01);科技部公益性科研专项(201310202)的研究成果;中央高校基本科研业务费专项资助(2014115010202)

摘　　要：近年来,国内外经济学文献已日益重视对我国企业数据的实证运用,然而,通过对现有主要企业数据来源的分析发现,现有企业数据普遍存在样本信息时效性不强、抽样缺乏随机性和指标多元性不足等质量缺陷,这或对实证研究结论的准确性、科学性和政策指导价值产生一定影响。为此,在广东开展的2015年中国制造业企业-员工匹配调查(CEES),与现有数据相比,本次调查在样本信息时效性、抽样随机性和指标多元性等方面均做出了较大的质量改进。运用上述调查数据,本文选取企业家创新精神与企业经营绩效、出口企业的"生产率悖论"、信贷约束与企业经营绩效等三大研究热点领域为例,首次从实证角度验证了企业数据质量缺陷对研究结论偏差的实际影响。结果发现:现有企业数据样本信息缺乏时效性,影响了研究结论对于当前中国经济真实状况的准确判断;抽样缺乏随机性,影响了研究结论对于中国企业总体状况的科学推断;指标多元性不足,造成了实证研究对企业部分经济行为的测度存在统计定义误差,对研究结论的精度造成一定影响。In recent years, economics literatures at home and abroad have emphasized on using corporate data to do empirical research. However, the analyze of the data source of the literatures on the listed company shows that the existing enterprise data has three defects： lack of timeliness, lack of randomness, and lack of diversity, which possibly have great impact on the accuracy, science and policy guidance of the research con- clusions. Therefore, Wuhan University organized Hong Kong University of Science and Technology, Tsinghua University and the Chinese Academy of Social Sciences to make 2015 Guangdong manufacturing enterprise sur- vey. Compared with the existing data, we have made a big improvement on the sample timeliness, randomness and diversity. This paper chooses three research fields： entrepreneurship and business performance, export enterprises ＂productivity paradox＂, and credit constraints and business performance, to validate the impact of enterprise data quality on research findings. The results show that ： lack of timeliness of enterprise data influ- ences the researchers＇ judgments on Chinese economy; ments on the overall situation; and lacks of diversity of lack of randomues influences the researchers＇ judg- the research findings appear systemic bias.

关 键 词：数据质量 时效性 随机性 多元性 

分 类 号：F019.6[经济管理—政治经济学]