Written by Wei Cui and Xuerui Wang
Posted on March 29, 2020
Last year, two economists in China—Professor FAN Ziying from Shanghai University of Finance and Economics and Professor LIU Yu from Fudan—posted an interesting paper on SSRN. The paper studies the impact of a 2014 tax incentive on investment decisions by Chinese firms. Unlike many other studies of this kind (e.g. the analysis of firm investment response to VAT reform I blogged about here), Fan and Liu explicitly focus on how tax administration and compliance affect firm response to investment incentives. They offer preliminary empirical evidence for an idea that many of us have intuited.
Tax preferences, such as the accelerated depreciation policy Fan and Liu examine, have the potential of helping cash-constrained firms through tax savings. However, Fan and Liu point out that “those firms that face more financial constraints can be more aggressive in their tax avoidance…which reduce firm effective tax rate. A lower effective tax rate reduces the benefit that a firm receives from accelerated depreciation and therefore undermines its effect on firm investment.”
A stronger and blunter version of this reasoning would say: severely cash-constrained firms may have long resorted to tax evasion to survive; tax preferences are meaningful only for firms that are roughly complying with the letters of the law; firms that already flout the law stand to gain little from benefits delivered by modifications of the law.
What empirical evidence do Fan and Liu offer for this hypothesis, which we find intuitively plausible? It is difficult to measure tax non-compliance at the level of each firm. Most tax attributes of business firms—the amount of tax they pay, the ratio of such tax to their revenue and profits—are determined by many factors, and none can be easily pinned to a non-compliance decision. Instead, Fan and Liu suggest that we measure non-compliance at the aggregate level.
They offer two such aggregate measures. The first is a measure of “county-level corporate tax enforcement”. This is derived from observed firm-level tax rates—in particular, the ratio of income tax payment to firm revenue. The idea is that this latter ratio will differ from firm to firm, not only because of firm characteristics such as the industries they operate in, their size in terms of revenue, employees, and fixed assets, or their ownership, but also by virtue of their geographic location. Being located in a given county might in itself predict a higher or lower tax rate. Fan and Liu interpret the ability of county-affiliation to predict a firm’s effective tax rate as reflecting the different levels of enforcement in different counties in China. If a given county predicts a higher effective tax rate, that county must have stronger tax enforcement.
This is an interesting proposal for measuring the level of local tax enforcement. How do we know that it is a valid measure? Logically, since many firm-level characteristics may determine a firm’s effective tax rate (aside from its industry, size and ownership), and since such characteristics may be correlated with geographical location, might the predictive power of county-affiliation simply capture these other firm-level characteristics that have nothing to do with tax enforcement?
We are not sure how Professors Fan and Liu address this question. Sometimes, researchers simply have to work with new empirical measures for a while for those measures to become accepted. Using their somewhat untested measure of enforcement, Fan and Liu show that firms in the “high-enforcement” counties responded to the 2014 investment incentive but firms in “low-enforcement” counties did not. If you already believe that non-compliance would neutralize the effect of tax incentives, you might see this result as a validation of Fan and Liu’s measure of enforcement. The problem is that the hypothesis, that compliance would neutralize the effect of tax incentives, is itself being tested.
The second aggregate measure of tax compliance Fan and Liu use is the “provincial tax fraud rate”, defined as “the number of tax frauds detected by the provincial tax audit office over provincial total number of firms in 2012”. The authors argue that a “higher tax fraud rate implies a higher level of evasion and a lower level of tax enforcement.” And they find that the firms in low-fraud provinces responded to the 2014 investment incentives, whereas firms in high-fraud provinces did not.
The data used by the paper to calculate the provincial tax fraud rate is taken from the China Tax Inspection Yearbook (中国税务稽查年鉴). We have used this data source in ongoing research. We have also reviewed how researchers in China have previously used this data. What is notable is that many scholars seem to interpret the Tax Inspection Yearbook metrics in the opposite way from Fan and Liu.
In particular, the tax fraud rate, as Fan and Liu define it, can be written as a product of two measures: (i) the coverage ratio, i.e. the number of taxpayers inspected relative to the population of firms in a province, and (ii) the target rate, i.e. the number of inspected taxpayers found to owe taxes relative to the number of total inspections in a province.
Most researchers seem to interpret each of these two measures as positive proxies for government enforcement efforts: the higher the coverage ratio, or the higher the target rate, the stronger a province’s tax enforcement is taken to be. For instance, the target rate has been found to be positively correlated with the revenue growth rate of a region (Zhou et al 2011), negatively correlated with some forms of tax planning (Li & Xu 2013), and positively correlated with firm access to credit (Pan et al 2013) and improved disclosure among listed private firms (Wang et al 2019). The coverage ratio, on the other hand, is found to be negatively correlated with firm book-tax difference (Sun 2012). (See references below.)
The puzzle is: How can these two positive measures multiply and produce a negative predictor of tax enforcement?
Intuitively, the detected fraud rate—which is how Fan and Liu’s measure should properly be labeled—is determined by (a) the level of non-compliance in the social environment, (b) the effort made at detecting non-compliance by tax authorities, and (c) the tax authorities’ success in detecting non-compliance. All three of these variables may be dependent on each other (they are “endogenously” determined). Whereas past researchers have taken the detected fraud rate to mainly reflect (b) and (c) (interpreted as the strength of enforcement efforts), Fan and Liu take the detected fraud rate to reflect mainly (a).
In summary, we are not sure that Professors Fan and Liu have arrived at convincing measures of non-compliance. Nonetheless, the hypothesis they lay out is important and deserves further exploration.
Li Wei`an and Xu Yekun (2013) The Effect of Political Identity on Tax Avoidance, Journal of Financial Research, no. 3, p.114, 120. 李维安，徐业坤 ：《政治身份的避税效应》, 《金融研究》 2013年第3期，第114，120页。
Pan et al (2013), Tax Collection and Management, the Government-enterprises Relationship and Corporate Debt Financing, China Industrial Economics, no.8, p. 109, 113. 潘越，王宇光，戴亦一：《税收征管、政企关系与上市公司债务融资》，《中国工业经济 》2013年第8期，第109，113页。
Sun Gang (2012), Family Business, Tax Inspection and Governance and Tax Avoidance，Taxation and Economy, no.3, p. 68-71. 孙刚：《家族企业、税收稽查治理与企业避税行为》，《税务与经济》2012 年第3 期，第68-71页。
Wang et al (2019), Tax Authorities and Enterprises: Zero-sum Game or Win-win Cooperation? Economy and Management, no.2, p.15, 17. 王宇光，赵茜，潘越：《税务机关与企业:零和博弈还是互助共赢?——基于上市公司信息披露质量视角的研究》, 《经济与管理》2019年第2期，第15，17页。
Zhou et al (2011), Tax Effort, Tax Bureaus and the Puzzle of the Abnormal Tax Growth, China Economic Quarterly, no. 1, p. 4-6. 周黎安，刘冲，厉行：《税收努力、征税机构与税收增长之谜》，《经济学(季刊)》2012年第1期，第4-6页。