The acquisition and supervision of tax-related information is an important means to ensuring the standardization of tax enforcement and provides an important guarantee against which the government implements tax policies and promotes tax equity (Gordon & Li, 2009). The Golden Tax Project, which was implemented in 1994, is a representative reform of the tax enforcement system. From 1994 to 1998, the Chinese government implemented “Golden Tax I,” which was designed to focus on the construction of a VAT cross-checking system, and promoted it in more than 50 cities. However, it did not achieve the desired effect due to the frequency of information errors caused by manual entry. To address this problem, China began to formally establish “Golden Tax II” in 1998 and rolled out the system to all provinces, prefecture-level cities and counties in 2001.

The “Golden Tax I” and “Golden Tax II” projects mainly focused on the joint verification of VAT, which could not meet the tax enforcement system’s increasingly complex needs in terms of expanding information processing and supervision, so China officially launched the “Golden Tax III” project in 2013. The “Golden Tax III” project is different from the previous two phases of the project in that it covers all tax types, generally uses big data, cloud computing and other advanced technical means, and is built following the principle of “one platform, two levels of processing, three coverage, and four systems ² .” This project marked a leap forward in the supervision of tax-related information: the unified technical infrastructure platform enables the smooth flow of information between taxation departments and the data can be monitored and cross-checked, the exchange of information between the State Administration of Taxation and local taxation bureaus facilitates the multi-dimensional presentation of taxpayer profiles and prevents tax evasion, and the intradepartmental information exchange compresses the freedom in local tax collection and enforcement while combatting the irregularities in tax enforcement and unfair taxation caused by corruption and collusion.

“Golden Tax III” has made greater breakthroughs than “Golden Tax I” and “Golden Tax II” in terms of influencing reform and expanding the coverage of tax types. Therefore, this paper selects “Golden Tax III” as the research object for tax enforcement system reform and uses its promotion as a quasi-natural experiment to carry out empirical research. The specific rollout of “Golden Tax III” in each province and city is shown in Table 1.

Although social insurance premiums are fees, they have the characteristics of a quasi-taxation. Among more than 170 countries and regions with social insurance premiums, 132 have taxation departments acting as social insurance premium collection and administration agencies. There has long been an ample “free space” for the collection and administration of social insurance premiums in China, and the evasion of social insurance premiums was widespread. Although the “Interim Regulations on the Collection and Payment of Social Insurance Premiums” and the Social Insurance Law of the People’s Republic of China clearly stipulate the scope and norms of collection and payment, enterprises may find ways to evade social insurance premiums by changing their governance structure, compensation plans, and ownership structure as well as by looking for loopholes in the regulatory environment (Desai and Dharmapala, 2006; Chen et al., 2010; Armstrong et al., 2015; DeBacker et al., 2015).

As the most critical factor at the enforcement level, the operation of collection agencies has a significant impact on the standardization of tax and social insurance premium collection as well as the prevention of tax evasion. In 1998, the Ministry of Finance, the Ministry of Labor, the People’s Bank of China, and the State Administration of Taxation jointly issued the “Notice on the Issuance of Interim Provisions on the Implementation of Two Lines of Management of the Basic Pension Insurance Fund for Enterprise Employees,” which clearly stipulates that both social insurance agencies and taxation departments are the legal agencies for collecting and managing social insurance premiums and that localities are free to choose between the two. In 1999, the 13th executive meeting of the State Council considered and adopted the “Provisional Regulations on Collection and Payment of Social Insurance Premiums,” which reconfirmed the collection and administration departments. Although both the social insurance and taxation departments are the legal collection agencies for social insurance premiums, the taxation departments are relatively more regulated and thus more aggressive in their collection efforts. On the one hand, there is a connection between social insurance premiums and income tax and thus the collection of social insurance premiums by the taxation department can allow for more efficient supervision of tax evasion by cross-checking information; on the other hand, the taxation department has stronger enforcement capabilities, a more complete collection and control system, and takes a firmer stance against tax evasion. To restrain tax evasion, more local governments are choosing to implement tax enforcement agency reform and replace third-party agencies with taxation departments as collectors of social insurance premiums. The specific rollout is shown in Table 2.

The key assumption for the above regressions to provide causal inference is that the control group (i.e., firms that did not suffer from tax inequity before the reform) provides an effective counterfactual for the sustainability of firms in the experimental group (i.e., firms that suffered from tax inequity before the reform). The basic premise for this assumption to hold is that the experimental and control groups satisfy the parallel trend condition. Moreover, considering that there may be a lag in the effect of tax enforcement reform and the possibility that the effect of reform may be affected by other related policies, this paper follows Beck et al. (2010) and Wang (2013) by constructing a more flexible DID model to allow the estimated coefficients to vary across years and thus examines the parallel trend before the reform as well as the dynamic effect after it. T F P _ L P i t = α 0 + ∑ k = − 16 7   R O S i t k + C o n t r o l s + Y e a r F E + I n d u s t r y F E + ε i t (3) T F P _ L P i t = α 0 + ∑ k = − 16 7   R O I i t k + C o n t r o l s + Y e a r F E + I n d u s t r y F E + ε i t (4)

Model (3)/(4) replaces ROS _it /ROI _it in Model (1)/(2) with the set of variables ROS _it ^k/ROI _it ^k, where k is the difference between the observation year and the year of reform implementation and takes the value of 0 if it is in the year of reform, n if it is the nth year of reform, and -m if it is the mth year before the reform. In the sample period of this paper (2000–2019), the value of k takes the ranges of [−16,7] in Model (3) and [−19,21] in Model (4). Due to space limitations, this paper only presents the 5-year results before and after the reform implementation.

Table 6 reports the dynamic effects of the tax enforcement reform. Regressions 1–3 show the effects of the reform in the tax enforcement system. Companies are grouped according to whether they suffered from income tax inequity before the reform, with Regression 1 including the experimental group, Regression 2 including the control group and Regression 3 including the full sample. The results show that before the implementation of the tax enforcement system reform, the estimated coefficients are nonsignificant in all three groups. In the years after the implementation of the reform, there is a significant increase in sustainable development in the experimental group, while there is no significant impact in the control group. In the full sample group, ROS _it ^k has a significant effect on corporate sustainability, but the coefficients are smaller than those of the experimental group due to the inclusion of the control group sample. The above results indicate that the reform in the tax enforcement system has a dynamic long-term impact on the sustainable development of enterprises. Regressions 4–6 report the results of the dynamic regressions of the tax enforcement agency reform. The results are similar and indicate that the reform in tax enforcement agencies also has a dynamic long-term impact on the sustainable development of enterprises. Together, the above results verify the findings of the baseline regression.

To further rule out the concern that the increase in enterprise sustainability is a result of the tax enforcement reforms and not random, we conduct a placebo test by setting a pseudo-event time as a placebo during the years of the implementation of the tax enforcement reform. With the placebo replacing the year in which the treatment effect actually occurred, the experimental and control groups are re-selected based on the placebo year. The results reported in Table 7 reveal that the effects of the key variables in both Regressions 1 and 2(UNF_ROS_pb×ROS_pb/UNF_ROI_pb×ROI_pb) on enterprise sustainability are no longer significant, thus indicating that the increase in enterprise sustainability is indeed caused by the tax enforcement reforms and their resulting reduction in tax inequity—that is, the results of the main regressions are robust ³ .

The basic theory of economic growth states that the production function is Y = A * f ( K , L ) , and labor, capital, and technology are the basic factors that affect economic development. In the function, A represents the technology factor, L represents the labor input and K represents the capital input. As a result, there are also three important core measures of sustainable development, namely, total factor productivity (which corresponds to the technology factor), labor intensity (which corresponds to the labor factor), and capital intensity (which corresponds to the capital factor).

In the main regression, the measure of enterprise sustainability is the total factor productivity (TFP) calculated using the LP method. To verify the robustness of the model and more comprehensively measure enterprise sustainability, the total factor productivity calculated using the OP method (TFP_OP), labor intensity (LAB), and capital intensity (CAP) are selected as alternative indicators of enterprise sustainability. Table 8 reports the regression results of the alternative indicators as explanatory variables, respectively, where Regressions 1, 3, and 5 test the effect of reduced tax inequity resulting from the reform of the tax enforcement system on the dependent variable, and Regressions 2, 4, and 6 test the effect of reduced tax inequity resulting from the reform of the tax enforcement agencies on the dependent variable. In all regressions, the main variables UNF_ROS×ROS/UNF_ROI×ROI reveal a significantly positive effect on the dependent variables, which indicates that the effect of tax equity on the sustainable development of enterprises is robust.

Green innovation applications represent the current green innovation output of enterprises, and green invention patent and utility model applications are selected as the metrics used in this paper. Among them, an invention patent refers to a new technical solution for a product, method or improvement, and a utility model refers to a new technical solution for a product shape, structure or combination thereof that is suitable for practical use. Table 9 reports the results of the empirical tests of the reduction in tax inequity brought about by the tax enforcement reform enacted through green invention patent applications.

Running Regressions 1 and 4 is the first step of the mediating mechanism effect test. The interaction terms are significant at the 1% level and the coefficients are positive, which indicates that the reduction in tax inequity brought about by tax enforcement reform significantly contributes to the sustainable development of enterprises. In Regressions 2 and 5, the interaction terms are significant at the 5% level and the coefficients are 0.094 and 0.119, respectively, thus indicating that the reduction in tax inequity brought about by tax enforcement reform increases green invention patent applications by approximately 10%. In Regressions 3 and 6, the coefficients on lnGIPA are significantly positive at the 1% level, which indicates that there is a significantly positive relationship between green invention patent applications and the sustainable development of enterprises. Meanwhile, the coefficients on the interaction term are significant at the 1% level, but less so than their counterparts in Regressions 1 and 3, thus indicating that green invention patent applications partially mediate the relationship between tax equity and the sustainable development of enterprises.

Table 10 reports the results of the empirical tests on how the reduction in tax inequity brought about by tax enforcement reform affects sustainable development through green utility model applications. Similarly, green utility model applications exhibit a partial mediating pattern in the relationship between tax equity and sustainable development.

A patent application indicates that a company has made an invention and applied for authorization from the intellectual property department. However, whether the final authorization is granted depends on whether the patent meets the corresponding requirements. In other words, patent grants are a better indicator of an enterprise’s ability to innovate than patent applications. Based on this, this paper also tests the mediating effects of the number of green invention patents and utility models that have been granted.

Table 11 reports the results of the empirical tests on the mediating effect of green invention model patent grants. In Regressions 1 and 4, the coefficients on the interaction terms UNF_ROS×ROS and UNF_ROI×ROI are both significant and positive at the 1% level, thus indicating that tax equity positively affects enterprise sustainability; in Regressions 2 and 5 the interaction terms remain significantly positive, which indicates that tax equity increases the number of green invention patent grants; in Regressions 3 and 6, lnGIA has a positive coefficient at the 1% level, which indicates that green invention patent grants promote sustainable development. The interaction terms, although significant at the 1% level, have smaller coefficients than their counterparts in Regressions 1 and 4, thus indicating that green invention patent grants partially mediate the effect of tax equity on sustainable development.

Table 12 reports the results of the empirical tests on how the reduction in tax inequity brought about by tax enforcement reform affects sustainable development through green utility model grants. Similarly, green utility model grants exhibit a partial mediating pattern in the relationship between tax equity and sustainable development.

Firms at different life cycle stages have different tax sensitivities and therefore respond differently to changes to the tax regime. Based on this, the paper groups the sample according to firm age. Those above the average age are defined as mature firms and those below the average age are defined as young firms. Table 13 reports the regression results of the heterogeneity tests. The coefficients on UNF_ROS×ROS/UNF_ROI×ROI are significantly positive at the 1% level in all regressions, thus indicating that reducing tax inequities significantly contributes to enterprise sustainability irrespective of age. Furthermore, the coefficient on UNF_ROS×ROS is significantly larger in the group of young firms than it is in the group of mature firms, which indicates that the reduction in tax inequity brought about by tax enforcement reform promotes sustainable development to a greater degree in young firms; the coefficient on UNF_ROI×ROI is larger for mature firms than it is for young firms, which indicates that the reduction in tax inequity brought about by tax enforcement agency reform promotes sustainable development to a greater degree in mature firms. This may be due to the fact that reforming the tax enforcement system breaks down intradepartmental information barriers and is therefore more effective in reducing information asymmetries on the enforcement side and preventing rent-seeking behaviors among enforcement agencies. While young enterprises have fewer social connections and are more likely to suffer from tax inequities resulting from tax enforcement, they benefit more from tax enforcement system reform. Reforming tax enforcement agencies reduces information asymmetries on the payment side by coordinating social security contributions and tax payments and thus discourages tax evasion. Since mature enterprises are generally more disciplined in paying social security contributions, they are more likely to benefit from the tax equity brought about by tax enforcement agency reform.

A favorable market environment is more protective of inventions and more conducive to the transformation of patented technologies into tangible output, so the market environment has an important impact on the sustainable development of enterprises. Therefore, this paper incorporates the marketization index to observe the heterogeneous impact of reduced tax inequities on the sustainable development of enterprises under different marketization conditions.

Table 14 reports regression results. In both the unfavorable and the favorable market environment subgroups, the coefficients on Reform _it ×Inequity _i are all significantly positive. This indicates that the reduction in tax inequity brought about by tax enforcement reform significantly contributes to the sustainable development of enterprises regardless of the market environment. Furthermore, this paper finds that the coefficients on UNF_ROS×ROS/UNF_ROI×ROI are larger in the regressions on the favorable market environment subgroup. This indicates that tax equity can better promote the sustainable development of enterprises in areas with a favorable market environment.