Doctoral Dissertation Research in Economics: The Economic and Behavioral Effects of a Value Added Tax: Evidence from Firm-level Data

增值税（VAT）是一种消费税。它是最年轻但也是最重要的政府收入来源之一。1967年，巴西和丹麦首先引入了增值税，到2013年，已有160多个国家采用了增值税。快速采用增值税的关键原因之一是它被认为是一种有效的税收工具。也就是说，增值税不会对企业的生产决策造成扭曲，这意味着政府可以以最小的成本增加收入，而不会阻碍经济增长。这一预测的一个关键弱点是，它是建立在一个完美的税收合规和完善的税收执法环境的基础上的，而这在实践中经常被违反。因此，增值税是否是一种有效的税收制度，最终是一个实证问题。然而，令人惊讶的是，几乎没有证据可以评估增值税的这些效率特性。这项拟议的研究使用了一个包含数百万企业长达10年信息的商业数据库，以估计增值税对企业增长（如产出、就业、投资支出和生产率）和企业行为反应（如少报收入或多报成本以申请不当税收抵免）的因果影响。随机实验被认为是研究治疗效果的黄金标准。然而，由于伦理或经济原因，大多数社会和经济政策（包括增值税）并不是随机实施的。PI使用回归不连续设计（RDD）通过利用法律规定的注册阈值周围的行为来为局部随机实验创造条件。这项研究有几个贡献。它通过利用为局部随机实验创造条件的增值税规则来分析增值税的影响。其次，之前大多数关于增值税影响的研究都使用了汇总数据，这些数据往往掩盖了大量的可变性。由于许多可能影响增长的遗漏变量，使用汇总数据分离因果关系是困难的。PI使用了丰富的微观数据，可以在增值税和相关结果之间建立更紧密的联系。第三，有大量关于公司的文献。对企业所得税等直接税的行为反应。然而，关于公司的文献？对增值税等间接税的行为反应较小。这项研究为这个虽少但很重要的文献做出了贡献。最后，目前人们对增值税的全面影响知之甚少。本研究的结果将为如何改善增值税制度的管理、执行以及最终的有效性提供见解。

The Value-Added Tax (VAT) is a tax on consumption. It is one of the youngest yet among the most important sources of government revenue. The VAT was first introduced in 1967 in Brazil and Denmark and by 2013 more than 160 countries had adopted it. One of the key reasons for the rapid adoption of the VAT is that it is believed to be an efficient tax instrument. That is, a VAT does not create distortions in firms' production decisions, implying that government can raise revenue with minimum cost and without impeding economic growth. A key weakness of this prediction is that it is based on an environment with perfect tax compliance and perfect tax enforcement, which is often violated in practice. Thus, whether the VAT is an efficient tax system is ultimately an empirical question. There is, however, surprisingly little evidence evaluating these efficiency properties of a VAT. This proposed research uses a commercial database that contains information on millions of firms for up to 10 years, to estimate the causal impact of a VAT on firm' growth (e.g, output, employment, investment spending, and productivity) and firms' behavioral responses (e.g, under-reporting of the revenue or over-reporting of the cost to claim undue tax credit).Randomized experiments are considered the gold standard in studying the effect of a treatment. However, due to ethical or economic reasons, most of the social and economic policies (including a VAT) are not implemented randomly. The PI uses Regression Discontinuity Designs (RDD) to create the conditions for a local random experiment by exploiting behavior around a registration threshold imposed by law. This research makes several contributions. It analyzes the effects of a VAT by exploiting a VAT rule that creates the conditions for a local random experiment. Second, most of the previous studies on the effects of a VAT use aggregate data that often mask great deal of variability. Isolating causality using aggregate data is difficult due to many omitted variables that can affect growth. The PI uses rich micro data that allow establishment of a tighter link between a VAT and relevant outcomes. Third, there is a voluminous literature on firms? behavioral responses to direct taxation such as the corporate income tax. However, the literature on firms? behavioral responses to the indirect taxation such as a VAT is small. This study contributes to this small but important literature. Finally, at present, the full impact of a VAT is poorly understood. The results from this study will provide insights on how to improve the administration, enforcement, and ultimately the effectiveness of a VAT system.