Does Tax Sharing Matter for Export Quality Upgrading? Evidence from China

Abstract

Tax policy is critical for business operations and export decisions. However, the relationship between tax sharing and export has been less frequently discussed. This paper explicitly examines the underexplored tax-sharing system’s effect on manufacturing exporters’ export quality and develops four hypotheses. We use data on Chinese manufacturing exporters and prefecture-level tax-sharing from 2008 to 2013 and employ an instrumental variable approach to alleviate the endogeneity problem. The empirical evidence supports our hypotheses. We find that an increase in the prefecture-level government tax-sharing ratio significantly reduces export product quality of firms. This quality effect can occur through channels, including tax burden effect, production scale effect, and innovation effect. Moreover, more productive firms and those operated in cities with stronger intellectual property protection can face a smaller quality-reducing effect. Our findings offer policy implications for improving China’s modernized tax system and trade upgrading. Policymakers should recalibrate the tax-sharing system to reduce the tax burden on manufacturing exporters, particularly for innovative and high-productivity firms, and bolster intellectual property rights to enhance export quality and support China’s trade and economic modernization.

1. Introduction

Fiscal means are central to the redistribution of economic authority between the central and local governments during the process of China’s market-oriented reforms. China’s 1994 tax reform marked a substantial step in China’s fiscal system marching from the planned economy to the market economy. Featuring China’s central-local fiscal relations, this tax reform resolves the problems originating from the local government’s original fiscal and taxation system and motivates local governments to develop the local economy for higher fiscal revenues [1].

Extensive academic efforts have been put on the economic impact of tax sharing. On one hand, tax-sharing incentives can drive local governments to improve the investment environment, actively attract investment, promote marketization, and stimulate investment-driven economic growth [2,3,4]. On the other hand, tax-sharing incentives may also lead to various social negative impacts, such as regional market segmentation, local debt crises, resource misallocation, and environmental pollution [5,6,7]. Associated economic deficiencies of tax incentives also exist, in the sense that they may distort the structure of the industrial chain [8] and reduce technological upgrading [9], especially in the case that incentives only concentrate on a certain part of the industrial chain. However, some scholars have questioned the effect of tax incentives [10,11]. The critical points are the distortion of investment decisions, the lack of periodic review, and no-exit clauses that are often politically motivated [12]. Therefore, tax incentives are often ineffective because of their inherent defects that can erode the tax base and are prone to abuse and corruption [10].

Tax sharing policy can produce profound impacts on industrial development [13] and the supply of public goods [4] and we call these economic responses the “indirect effect”. Local governments, under the tax sharing system, tend to protect local industries with high profit tax rates on firms, resulting in significant local protection (Bai et al., 2004). Local governments may adjust the degree of tax effort, leading to substantial differences in the actual tax rates faced by different firms [14,15,16]. A higher profit tax rate increases costs disproportionately for large firms compared with small ones. Large firms can argue with local authorities for stronger protection while small firms are likely to exit. This firm turnover redistributes resources across industries and firms. When the tax-sharing ratio increases, local governments enhance tax efforts, leading to an increase in the actual tax rates for firms [17,18]. However, some studies suggest that an increase in the tax-sharing ratio can reduce the tax burden and encourage production and investments for firms [19,20]. We call these firm-level responses the “direct effect”.

In theory, product quality indicates a higher capacity of products to satisfy the needs of consumers and requires more R&D inputs from productive firms. Numerous studies have explored the contributions of up-gradation of exports and export quality [21,22,23], which attracts extensive discussions on the factors and measures of export product quality [24,25,26,27,28]. A few studies have started to uncover the specific impact of certain tax reforms or fiscal policies on firms’ R&D inputs, production factor prices, and credit constraints, thereby influencing the export product quality [13,29,30,31]. For instance, Sui et al. explore the quality effect of counties-managed-directly-by-provinces reform using the DID method [30]. Zhang found a quality-enhancing effect of government subsidies [32]. In contrast, Zhang et al. claim a quality reduction impact of government subsidies because firms can become less independent and have less desire for independent R&D [33]. Nevertheless, the above studies tend to consider the impact of a specific fiscal policy or tax reform as quasi-natural experiments and ignore the role of a verticalized tax-sharing system at the macro level in the export product quality.

In the current paper, we try to explore the relationship between the de facto tax-sharing ratios of the prefecture governments and firms’ exporting performance using intra-national firm data from 2008 to 2013. Specifically, we investigate the impact of value added tax (VAT) and corporate income tax (CIT) sharing on export product quality across Chinese localities since both VAT and CIT mainly derive from the industrial sector and serve as the primary tax sources for local governments. Our main empirical result shows that both the higher local VAT and CIT sharing ratios inhibit the quality of exported products. Moreover, we show that the channels through which tax-sharing takes effect on product quality include increased tax burden, shrinkage in production scale, and expenditure cuts in firms’ innovation activities. Our heterogeneity results indicate that state-owned firms and firms with better production efficiency and situated in cities with better intellectual property protection (IPP) can suffer a smaller quality reduction effect due to a change in the tax-sharing ratio. Furthermore, firms in less competitive industries are significantly influenced by tax-sharing because of their lower market powers and thus less motivation to develop their products.

This study contributes to the existing literature in three aspects. First, our findings extend the existing finance-domestic market economy research to a broader field. A few studies explored the effects of VAT reform on the export product mix; however, we bridge the fiscal policy and export relationship from the export quality perspective. This facilitates the deeper discussion of the new development pattern featuring a “dual cycle” that associates the domestic market economy and the international market economy. Our main finding confirms the importance of fiscal policy for firms’ participation and competitiveness in international markets. Second, we conduct an in-depth analysis of the estimated effects of a specific policy. This discussion departs from other studies of policies of a Procrustean bed using the difference-in-difference approach. Our approach considers cross-region variations in tax-sharing ratios, which enriches our understanding of the effects of one policy, i.e., the tax-sharing reform. Third, we provide detailed empirical evidence of underlying mechanisms linking the tax-sharing system and export quality of firms, including a tax burden effect, production scale effect, and innovation effect. This is equivalent to more insights into possible starting points for fiscal policy intended to enhance competitiveness in international markets for manufacturing firms.

This paper proceeds as follows. Section 2 describes the institutional background of the tax-sharing system and provides research hypotheses; Section 3 presents the data and estimation strategy; Section 4 presents the basic results and robustness checks; Section 5 and Section 6 provide the mechanism analysis and heterogeneity analysis. Section 7 concludes with policy implications.

2. Institutional Background and Hypothesis

2.1. Flexible Tax Sharing System

Since 1994, China has implemented tax-sharing system reforms, primarily involving four major areas. The first step is to redefine the distribution principles of authority and financial power between the central government and local governments; The second is to allocate tax revenues between the central government and local governments taking into account both the combination of authority and financial power; The third is to restructure the tax authorities between national and local governments; The recent step is to re-establish the system of central government tax rebates to local governments.

This tax-sharing system highlights the fundamental principle of aligning government fiscal rights with powers. Based on tax types for the powers and expenditure scopes, governments at different levels thereby develop financially independent and clearly defined responsibilities. In contrast to many other nations, the tax-sharing system rationalizes the transfer of tax authorities through the classification of central taxes, local taxes, and shared taxes. Consumption taxes, tariffs, and other comparable levies, for example, provide the central government with fixed incomes. Local governments, on the other hand, benefit from business taxes, property taxes, and corporate stamp taxes. Taxes such as value-added tax (VAT) and corporate income tax (CIT) are proportionally shared between the central and provincial governments.

Table 1. The adjustment of the tax-sharing ratios in China.

Period	Types of Taxes	Tax-Sharing Ratio (Central-Local)	Notes
1994–2001	Value-added tax (VAT), securities transaction tax (STT), resource tax	VAT (75:25) STT (50:50)	In the resource tax, the marine petroleum resource tax goes to the central government, while all others go to the local government.
2002–2015	Value-added tax (VAT), securities transaction tax (STT), resource tax, corporate income tax, individual income tax	VAT (75:25) STT (97:3) CIT (60:40) IIT (60:40)	Income tax-sharing reform in 2002; Securities transactions tax-sharing reform in 2000.
2016–Present	Value-added tax, securities transaction tax (STT), resource tax, corporate income tax, individual income tax	VAT (50:50) STT (97:3) CIT (60:40) IIT (60:40)	The pilot reform of replacing business tax with VAT in 2012 was fully implemented in 2016.

displays the adjustments of central-local tax-sharing ratios for these shared taxes during several specific tax reforms. Since the tax-sharing reform in 1994, China has adjusted the distribution ratios for different taxes in its fiscal system reforms at various stages of development. This includes reforms such as the 2002 income tax-sharing reform and the 2016 VAT-sharing reform. It is important to note that there are explicit regulations governing the distribution ratios between the central government and local (provincial) governments. For instance, the central-local tax-sharing ratios for VAT and CIT as two major types of tax in China are 75:25 and 60:40 before 2016. Tax-sharing system primarily adjusts the central-to-local ratio of tax distribution for business tax, corporate income tax, and value-added tax and motivates local governments to prioritize local economic production, thereby influencing regional firms’ production behaviors.

The introduction of flexible tax-sharing systems within provinces led to the phenomenon of “flexible sharing contracts” [18,34]. Specifically, the central government oversees solely the tax-sharing ratios only between central and provincial governments. Within provinces, tax revenues collected at city and county levels (below the provincial level) are shared with provincial governments at varying ratios. As a result, the ratios of retained tax revenue below the provincial level vary across provinces, offering an important starting point for assessing the impact of a local government’s tax-collecting behaviors.

We estimate the tax-sharing ratios of prefecture-level governments (including city and subordinate districts and counties) from 2008 to 2013, following the method by Lyu et al. and Mao et al. [18,34,35]. The specific calculation methods and data descriptions will be presented in Section 3.1 and Section 3.2. We identified stylized facts of flexible tax sharing for sub-provincial governments like previous research.

Figure 1 and Figure 2 depict the provincial average of VAT and CIT sharing ratios for prefecture-level governments, respectively. Overall, both tax-sharing ratios are characterized by a relatively stable trend. In Figure 1, the average VAT sharing ratio remains around 20%, with noticeable variations across provinces from 2008 to 2013. Similarly, the trend of CIT sharing ratios, as illustrated in Figure 2, consistently stayed below 40% and also showed distinct differences among provinces. Figure 3 shows the average sharing ratios of VAT and CIT for prefecture-level governments in each province from 2008 to 2013. The disparities in tax sharing primarily stem from the autonomy exercise by the provincial government in the arrangements for tax sharing within the province. Specifically, the highest tax-sharing ratio for VAT is in Yunnan province at 25.0%, while the lowest is in Hebei province at 14.7%. For the CIT share ratio, Zhejiang province has the highest share ratio at 35.8%, and the lowest is in Yunnan province at 14.0%.

2.2. Hypotheses

Although local governments in China are unauthorized to legislate on taxation, they can influence the effective tax rates by adjusting tax collection efforts and providing tax incentives in disguised forms [14,18]. Per the principal-agent model [36], the central government, acting as the principal, establishes the ratio for benefit sharing. While the local governments, functioning as agents, take actions to develop the regional economy, thus resulting in tax revenue increase. Consequently, the tax-sharing ratios significantly influence the tax-collection behaviors of local governments.

Due to the nature of statutory, non-compensatory, and stationarity of taxation, firms are obligated to fully pay taxes in a timely manner, which can reduce their retained earnings and current cash flows [37]. However, the decision to export depends largely on whether firms have sufficient capital to cope with the sunken costs of entering international markets [38,39]. To improve the product quality for exporting, firms need to spend more on the market demand research and innovation. Several studies have shown that inadequate cash flows can lead to financial constraints for firms, purchasing fewer inputs and spending less on quality control [40]. Moreover, some studies also find that cash flow constraints decrease investments in R&D and innovation-related expenditures, which may further affect product quality [41,42]. Hence, we put forward our main hypothesis:

Hypothesis 1.

There is a negative relationship between the tax-sharing ratio of local governments and the export quality of firms.

There are significant differences in the effective tax rates even between adjacent regions because of the varied economic incentives of each sub-provincial government. The effective corporate tax rate is closely associated with government tax collection efforts [18]. Several studies have found that the effective CIT and VAT rates are significantly lower than the nominal tax rates and their distribution among enterprises varies substantially as well [15,16]. This disparity arises not only from tax evasion by enterprises but also from different practices of tax collection adopted by local governments.

The effective tax rates even vary with firm groups, such as profitability, capacity of producing value-added, and years of operation. For instance, firms with higher industrial value-added and a value-added ratio can enjoy lower effective VAT rates. Firms with stronger capacity of profiting pay with higher effective VAT tax rates. Some studies suggest that an increase in the tax-sharing ratio can reduce the tax burden and encourage production and investments for firms [19,20]. However, sub-provincial governments may still utilize the tax rate variations to strengthen their tax collection effect and thus increase tax burdens on firms. Some scholars indeed find that the ratio of tax sharing increases and local governments improve tax collection efforts, leading to higher firms’ effective tax rates and tax burdens [17,18,34]. Hence, we put forward the following hypothesis:

Hypothesis 2.

Higher tax-sharing ratios increase the tax burden of local firms, which is not conducive to the upgrading of export product quality.

Despite a lack of direct evidence that tax sharing affects firm production, several studies mainly focus on the effect of certain tax policies and found positive impacts of policy-induced lower tax rates. For one thing, lower input taxes will reduce the cost of purchasing fixed assets and their financial constraints, which encourage firms to access more advanced machinery and equipment for production [43,44]. For another, tax reforms, such as the VAT reform, in China in 2009 reduced the tax costs paid for investments [45]. One of the primary benefits of VAT is the non-distortion of the choice of production inputs, which can preserve “production efficiency”. In this way, VAT reform can encourage firms to expand reproduction.

Nevertheless, in practice, the input effect depends on how VATs are implemented. The input distortion can sometimes happen [46]. The CIT reform can increase firms’ financial performance by investing more in fixed assets and alleviating external financial constraints [47]. A recent CIT-export research found that CIT reform induces exporters to adjust their product mix and concentrate on their core products [48]. This finding indicates a relationship bridging CIT reform and production scope. Transferring more resources to firms’ core products is conducive to their product quality upgrading. Higher tax-sharing ratios can reduce firms’ disposable funds and leave their production expansion financially constrained. In contrast to previous research, how tax-sharing ratio between the central-local governments affects firms’ production capacity receives less attention and remains unclear. A higher tax-sharing ratio can motivate local governments to retain more tax collected from firms. The above related findings may imply that heavy tax burdens and less funds for production can result in smaller production scales and product scopes. We thus develop the following hypothesis.

Hypothesis 3.

Higher tax-sharing ratios induce firms to downsize their production, which can result in lower export quality.

Among the various factors of global trade competitiveness, the firm’s R&D plays a key role [49]. From the perspective of tax reform, the reduction of VAT for firms can achieve efficiency improvements by enhancing R&D investment or end-of-pipe treatment [50]. Reducing the effective tax rates and subsequently expanding fixed asset investments allow firms to have more abundant funds for R&D and innovation to improve their technology [44,51]. The Chinese government relaxed its R&D tax policy. More newly eligible firms enjoy R&D tax reductions and innovate more [52]. To improve product quality, firms need to develop advanced technology and modify product designs that entail much R&D input. In contrast, higher tax-sharing ratios can inhibit firms’ investment and productivity growth and ultimately decrease firms’ probability of exporting and export performance. Considering this logic, we put forward the following hypothesis:

Hypothesis 4.

Higher tax-sharing ratios can contribute to firms’ lower innovation performance in R&D expenditures, which lowers the export quality of firms.

In Figure 4, we further illustrate the relationship between the four derived hypotheses that higher tax-sharing ratios can add more tax burden on firms, decrease production capacities, and reduce innovation activities, ultimately lowering the quality of exported products.

3. Data and Estimation Strategy

3.1. Data

The data used in this paper are mainly the matched data, including (1) the Annual Surveys of Industrial Enterprises in China (ASIEC) from 2008 to 2013; (2) the General Administration of Customs of China (GACC) as this is the only holistic and disaggregated one for Chinese manufacturing and exporting; (3) the firm-level innovation data obtained from China’s National Bureau of Statistics (NBS); and (4) the prefecture-level tax-sharing ratios data. Problems of ASIEC data pointed out by related literature include outliers and missing values [53].

Following the method developed by Fang et al. [54], the matching of the ASIFC, GACC, and innovation datasets is based on firm names, postcodes, legal persons, and phone numbers. We provide a highly complete firm-level panel data set from 2008 to 2013 that includes data on Chinese manufacturing exporters such as firm export quality, financial, and innovation data. The matching of the tax-sharing data is based on the prefecture names and city codes. Data on prefecture-level tax-sharing ratios are collected from the annual China Statistical Yearbook, China Financial Yearbook, and China Taxation Yearbook.

3.2. Model Specification

To explore the relationship between tax sharing and export quality, we employ a panel data model with multiple fixed effects presented as follows.

$${quality}_{fjct}={\beta }_0+{\beta }_1{Taxshare}_{ct}+{\beta }_3{X}_{ft}{+{\beta }_4{X}_{ct}+{\lambda }_j+{\lambda }_c+{\lambda }_t+\epsilon }_{fjct}$$

(1)

where quality_fjct is the export product quality of firm f in an industry j and located in city c at year t; Taxshare_ct represents the primary explanatory variable which refers to the VAT sharing ratio and CIT sharing ratio of the prefecture-level governments since these two are the most important taxes paid by Chinese firms; X_ft is a set of control variables at the firm level and X_ct refers to a set of prefecture-level macroeconomic fundamentals; λ_j, λ_c, and λ_t represent the industry, city and year fixed effects, respectively; ε_fjct refers to the error term clustered at the 2-digit CIC industry level.

3.3. Variables Description

3.3.1. Firm-Level Export Product Quality

Following the Khandelwal et al. method [25], in this paper we estimate the firm-level export quality, which equals a standardized aggregation of the corresponding product-level estimates [55]. First, we estimate the quality of product n exported by exporter f to destination d in year t using Equation (2).

$$\ln \left(q_{fndt}\right)+{\sigma }_n\ln \left(p_{fndt}\right)={\varphi }_n+{\varphi }_{dt}+{\epsilon }_{fndt}$$

(2)

where q_fndt denotes the demand for product n exported by a particular exporter f in destination country d in year t, p_fndt refers to the price of exported product n charged by the particular export f sold in destination market d in year t; ϕ_n is the product fixed effect; ϕ_dt represents destination-year fixed effect; and σ_n is the elasticity of substitution across 6-digit HS products within the same HS 3-digit products category. The estimation values of elasticity of substitution within HS 3-digit products category σ_n, are taken from Broda et al. who use a sample of data from 73 countries to obtain elasticities of substitution and supply for many sectors in each country [56]. The product-level export product quality is the quality_fndt = ε_fndt/(1 − σ_n). We aggregate the product-level quality to the firm-level one using the export share of each HS 6-digit product within firms as weights.

3.3.2. Measures of Prefecture-Level Tax-Sharing Ratios

The two most important taxes paid by Chinese enterprises are the CIT and VAT. Tax-sharing ratios are differentiated among various prefectures. Prefecture governments include governments of subordinate districts and counties. The tax-sharing ratio of prefecture governments is best measured by the amount of tax revenue received by prefecture governments divided by total local tax revenues of each province. Regretfully, the State Taxation Administration (STA) and State Administration of Local Taxation (SALT) did not reveal a specific amount of taxes for each prefecture from 2008 to 2013 (excluding VAT), therefore tax-sharing ratios for each prefecture cannot be computed directly. To overcome this dilemma, this paper adopts an alternative method following Lyu et al. and Lv et al. [3,18,34], which assumes that all prefectures of a province are regarded as a whole, and the sum of VAT revenues and CIT revenues obtained by prefecture governments of each province is:

$${VATshare}_{cpt}=\frac{{VAT revenue}_{cpt}}{{Total VAT revenue}_{pt}}$$

(3)

$${CITshare}_{cpt}=\frac{{CIT revenue}_{cpt}}{{Total CIT revenue}_{pt}}$$

(4)

where VATshare_cpt is the actual ratio of VAT sharing for prefecture-level government c of province p in year t. VAT revenue_cpt is the sum of VAT revenues obtained by prefecture-level governments of each province, Total VAT revenue_pt is the tax revenue collected by tax authorities of the province p in year t. Similarly, CITshare_cpt is an alternative independent variable in our baseline estimation. CIT share_cpt is calculated as the sum of CIT revenues obtained by prefecture-level governments of each province divided by the tax revenue collected by tax authorities of the provinces. It denotes the CIT tax ratio shared by prefecture-level governments.

The data we use are from the annual China Statistical Yearbook and China Financial Yearbook to obtain the prefecture-level VAT revenues and CIT revenues, and data from the China Taxation Yearbook are to obtain the total tax revenues at the provincial level. The tax-sharing ratios for prefecture-level governments within provinces show fewer variations. We can find significant yearly cross-province variations of VAT and CIT tax ratios. These indicators show similar patterns to those in the existing studies [3,34].

3.3.3. Control Variables

The ASIEC dataset provides a variety of critical financial information on firms and we additionally consider several key firm characteristics following previous literature, such as the firm’s management expense ratio (MER), firm’s leverage (leverage), fixed asset ratio (measured by the proportion of fixed assets held by a firm in its total assets), and age (lnage), as well as prefecture characteristics such as real GDP adjusted for inflation (lnrealgdp) in logarithmic form, the population in log form (lnpop), FDI in log form (lnfdi), and proportion of the secondary industry (prosec).

3.4. Descriptive Statistics

In our sample, each firm operates within a four-digit industry under the Chinese Industry Classification (CIC) system. This classification system changed in 2011 to include more information for certain industries. To guarantee consistency before and after 2011, we unified the CIC code into a 2002 version over the full sample period. We consider the features of different geographic locations and utilize the prefecture-level data from the CEIC database from 2008 to 2013 based on firm location information. Prefecture-level characteristics, such as real GDP, population, FDI, and secondary industry percentage are also incorporated. After matching and cleaning, we obtained an unbalanced panel of 82,822 firms and a total of 235,645 observations between 2008 and 2013.

Table 2. Descriptive statistics.

	N	Mean	Std. Dev.	Min	Max
	(1)	(2)	(3)	(4)	(5)
Panel A: Dependent variables
quality	235,665	0.673	0.141	0.000	1.000
lnquality	235,645	−0.424	0.261	−5.798	0.000
Panel B: Independent variables
VATshare	233,209	0.232	0.024	0.113	0.260
CITshare	233,209	0.299	0.059	0.131	0.403
lnVATshare	233,209	−1.467	0.119	−2.185	−1.345
lnCITshare	233,209	−1.229	0.215	−2.032	−0.909
Panel C: Controls
MER	235,485	0.060	0.059	0.001	0.376
leverage	235,604	0.661	0.597	0.002	5.115
fixed asset	234,937	0.350	0.563	0.000	5.881
lnage	230,166	2.172	0.701	0.000	5.118
lnrealgdp	233,209	5.619	0.914	1.950	7.540
lnpop	233,209	8.547	0.592	5.225	10.422
lnfdi	233,209	7.349	1.420	0.000	9.731
prosec	233,209	0.506	0.078	0.161	0.851

presents the summary statistics for our main variables.

Panel A shows our dependent variable, firms’ export product quality, which indicates that the average export quality of firms is 0.673, with a standard deviation of 0.141, ranging from 0 to 1. In the regressions, the logarithm form of export quality is employed, and the corresponding mean value is −0.424, ranging from −5.978 to 0, with a standard deviation of 0.261. Large variations of export quality across firms occur. Panel B introduces the main independent variables. The main measure of tax sharing, VATshare, ranges from 0.113 to 0.260, with a mean value of 0.238 and a standard deviation of 0.024. Another tax-sharing measure is CITshare, with a mean of 0.229 and a standard deviation of 0.024. We can find small cross-province variations of tax sharing.

4. Empirical Analysis

4.1. Basic Results

The key explanatory variables of interest are these two tax-sharing measures at the prefecture level, lnVATshare, and lnCITshare, as described in Section 3.2. The coefficient, β₁, is the key parameter of interest to be estimated using the fixed effect panel model. It measures the effect of tax sharing incentive at the prefecture level on product quality of what Chinese manufacturers export. We incorporate firm characteristics in our estimating equation such as firm management efficiency (MER), the firm’s leverage (leverage) measured by the ratio of total liability to total assets, the proportion of the firm’s total assets represented by its fixed assets (fixed asset), and establishment duration (lnage), as well as city characteristics, such as inflation-adjusted GDP, population, FDI, and the proportion of secondary industry. Furthermore, we add the year fixed effect to control for the trending impact of non-observable factors affecting all industries and enterprises, and the industry fixed effect, which controls time-invariant non-observable industry-level factors. In addition, since the prefecture-level control variables may absorb the prefecture-fixed effect, we base the regional fixed effect on the province.

Table 3. Baseline regressions.

	Dependent Variable: lnquality
	(1)	(2)	(3)	(4)
lnVATshare	−0.021 *	−0.027 **
	(0.013)	(0.013)
lnCITshare			−0.011	−0.015 *
			(0.009)	(0.009)
MER	−0.291 ***	−0.313 ***	−0.291 ***	−0.313 ***
	(0.010)	(0.010)	(0.010)	(0.010)
leverage	−0.002 **	−0.002 ***	−0.002 **	−0.002 ***
	(0.001)	(0.001)	(0.001)	(0.001)
fixed asset	0.006 ***	0.008 ***	0.006 ***	0.007 ***
	(0.001)	(0.001)	(0.001)	(0.001)
lnage	0.022 ***	0.022 ***	0.022 ***	0.022 ***
	(0.001)	(0.001)	(0.001)	(0.001)
lnrealgdp		−0.005 ***		−0.005 ***
		(0.002)		(0.002)
lnpop		−0.012 ***		−0.012 ***
		(0.002)		(0.001)
lnfdi		0.008 ***		0.008 ***
		(0.001)		(0.001)
prosec		−0.038 ***		−0.038 ***
		(0.010)		(0.010)
YearFE	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes
ProvinceFE	Yes	Yes	Yes	Yes
R2	0.080	0.092	0.092	0.092
Observations	232,375	232,375	232,375	232,375

Notes: Robust standard errors in parentheses. The controls and constant terms are omitted to save space. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

reports the results of our estimating framework according to Equation (1). The key explanatory variables of interest are the VAT tax sharing and CIT tax sharing at the prefecture level. Columns (1) and (2) report the regression results for lnVATshare on export product quality without and with prefecture-level control variables, and Columns (3) and (4) report the regression results for lnCITshare on product quality without and with prefecture-level control variables. Thus, a negative sign of the coefficient implies a negative relationship. In Columns (1) and (3), we include firm controls show that the tax-sharing ratios reduce export product quality. In particular, the coefficient of the VAT sharing is negative and significant at a 10% level. Moreover, in Columns (2) and (4), we further control for prefecture-level variables, and the estimated coefficients of lnVATshare and lnCITshare are negative but significant at the 5% and 10% significance level, respectively. Overall, the coefficients of interest are statistically significant, indicating stronger tax sharing at the prefecture level can suppress exporters from upgrading their export product qualities.

4.2. Robustness Checks

To guarantee the credibility of the baseline results, this paper conducts robustness tests by considering the endogeneity problem, excluding provincial capital cities, winsorizing the extreme values.

4.2.1. Endogeneity Problem

One source of the endogeneity problem could arise from the simultaneity between the prefecture tax sharing and product quality of exporters, i.e., lower profits of exporters due to deteriorated product quality may reduce tax income for local government, making them more eager to lobby to the provincial government to increase the share of taxes retained. Even though the reverse causal relationship could be relatively weak since the tax-sharing ratio is set by the provincial government and cannot be influenced by local firms, we still argue for the existence of this endogeneity source. On the other hand, the unobserved forces absent in our equations may also result in this endogeneity problem. We, therefore, instrument the endogenous variables, VAT sharing ratio, and CIT sharing ratio, with two kinds of instrumental variables.

The first IV we use follows the approach of Bartik [57], by creating a Bartik instrument (the product of one period lagged tax sharing ratio (VATshare_c,t−1 or CITshare_c,t−1), and the first order difference of the tax sharing ratio (ΔVATshare_c,t−1 or ΔCITshare_c,t−1)).We construct Bartik IV = Taxshare_c,t−1 × ΔTaxshare_c,t−1, based on the following considerations: First, the tax sharing ratio between the central and local governments is strictly given and constant over our study period. The tax sharing of each prefecture is often determined by the provincial government and is thus relatively exogenous to a particular prefecture. Second, prefecture demand shocks, in addition to the tax sharing, can also lead to estimation bias, but the Bartik instrument is valid as long as individual prefectures are not so important that their internal demand shocks are significantly correlated with the tax sharing of the province as a whole. We then proceed with the instrumental variable estimation. Columns (1) and (2) in

Table 4. IV approach.

	2nd Stage: Dependent Variable: lnquality
	Bartik IV		Weighted IV
	(1)	(2)	(3)	(4)
lnVATshare	−0.427 **		−0.251 ***
	(0.213)		(0.039)
lnCITshare		−0.709 ***		−0.133 ***
		(0.145)		(0.011)
	1st stage
	lnVATshare	lnCITshare	lnVATshare	lnCITshare
	(1)	(2)	(3)	(4)
Bartik IV	−0.004 ***	−0.009 ***
	(0.000)	(0.000)
mean IV			0.047 ***	0.065 ***
			(0.004)	(0.004)
Controls	Yes	Yes	Yes	Yes
YearFE	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes
ProvinceFE	Yes	Yes	Yes	Yes
Observations	232,375	232,367	232,328	232,306
Kleibergen-Paap rk LM statistic	450.461 (0.000)	1962.639 (0.000)	170.879 (0.000)	239.336 (0.000)
Kleibergen-Paap rk Wald F statistic	470.198	2371.013	170.097	239.275

Notes: Robust standard errors in parentheses. The controls and constant terms are omitted to save space. Mean IV in 1st stage corresponds to the instrumental variables of meanVAT and meanCIT. Similarly, Bartik IV corresponds to the instrumental variables of BartikVAT and BartikCIT, respectively. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

report the regression results where we report the first-stage regression results below the second stage. In the first stage, we find the estimated coefficients of Bartik IV, corresponding to BartikVAT and Bartik CIT, are significant at the 1% level. Controlling for endogeneity, we find that the 2SLS estimates of the coefficient of lnVATshare and lnCITshare in the second stage, are −0.427 and −0.709, being statistically significant. These results indicate that both VAT sharing ratios and CIT sharing ratios still have significant and negative impacts on firms’ export product quality after considering the endogeneity problem.

In addition to using Bartik IV, the second instrumental variable we constructed is the weighted average of the tax-sharing ratios of all other provinces. The weight is the inverse of the geographic distance between provinces. Due to small variations of tax-sharing ratio across prefectures within one province, we construct the provincial IV, which is common to all cities within one province, mean VAT and mean CIT, respectively. Controlling for endogeneity, we find in Columns (3) and (4) of Table 4 that the 2SLS estimates of the coefficient of lnVATshare and lnCITshare, are −0.251 and −0.133, statistically significant at a 1% level. These results indicate that higher tax sharing is not conducive to exporters upgrading their exports which is consistent with the results in Table 3.

Results for the validity of the IV are reported at the bottom of Table 4. The Kleibergen–Paap rk LM statistic reports the results of the under-identification of the chosen IVs. In Columns (1) and (4), all the p-values of Kleibergen–Paap rk LM statistics are 0.000, implying the IVs do not suffer from the under-identification problem. Meanwhile, the Kleibergen–Paap rk Wald F statistics are much greater than the 10% level critical value in the Stock–Yogo weak ID test, implying the IVs are not weak. The two statistics jointly confirm the validity of the chosen IVs.

We find that the IV estimates are consistent with the baseline results in Columns (2) and (4) in Table 3, despite the magnitude of coefficients. Additionally, the reverse causality relationship between export quality and tax sharing ratios may not be serious. Therefore, we still use OLS estimates as the baseline results for subsequent analysis. We also use a one-year lag of tax sharing to serve as an alternative IV, the consistent estimate is obtained, as shown in Columns (1) and (2) in

Table 5. Robustness checks.

	Dependent Variable: lnquality
	One-Year Lag		Exclude Provincial Capitals		Winsorization at 1%
	(1)	(2)	(3)	(4)	(5)	(6)
lnVATshare	−0.035 **		−0.031 **		−0.024 **	−0.013 *
	(0.017)		(0.014)		(0.013)	(0.008)
lnCITshare		−0.034 ***		−0.030 ***
		(0.012)		(0.011)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
YearFE	Yes	Yes	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes	Yes	Yes
ProvinceFE	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.074	0.075	0.092	0.092	0.091	0.074
Observations	140,135	140,135	187,823	187,823	232,375	232,375

Notes: Robust standard errors in parentheses. The controls and constant terms are omitted to save space. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively. The one-year lag denotes that the tax sharing ratios are treated one year lag.

.

4.2.2. Excluding Provincial Capital Cities Sample

Considering the possible statistical bias induced by the difference for provincial capital cities from other prefectures and that firms operating at capital cities on average perform better, we drop firms situated in provincial capitals from our sample. As shown in Columns (3) and (4) in Table 5, both VAT sharing ratios and CIT sharing ratios still have significantly negative impacts on firms’ export product qualities after eliminating samples from provincial capitals, validating the robustness of the baseline regression results. Moreover, the absolute values of the estimated coefficients of lnVATshare and lnCITshare are slightly higher than the baseline results in Table 3, suggesting that tax sharing affects typical prefecture cities more than capital cities.

4.2.3. Sample Extreme Value

Since there may be interference by the sample extreme value, this paper winsorizes the continuous variables in the regression of Columns (5) and (6) in Table 5. The winsoring rate of the whole sample is 1% (0.5% for each side). This result shows that the coefficient of the VAT sharing ratio and CIT sharing ratio are both significantly negative at the level of 5% and 10%, respectively, which is consistent with our baseline results in Table 3.

5. Mechanism

In the previous section, our first hypothesis confirmed that higher VAT sharing and CIT sharing ratios inhibit a firm’s export quality. However, the underlying mechanisms are still to be empirically supported. To examine the mechanism proposed in Hypotheses 2 to 4, which we refer to as the tax burden, production scale, and innovation effect, we set the following estimating equation:

$${mechanism}_{fjct}={\beta }_0+{\beta }_1{Taxshare}_{ct}+{\beta }_2{X}_{ft}+{\beta }_3{X}_{ct}+{\lambda }_j+{\lambda }_c+{\lambda }_t+{\epsilon }_{fjct}$$

(5)

where mechanism_fjct denotes the mechanism variable at the firm level, represented by the firm’s tax burden, production scale and R&D intensity, respectively. Thus, Equation (5) presents three possible channels through which tax sharing affects firms’ export product quality.

5.1. Tax Burden

Considering the fixed statutory tax rates, firms were compelled to underestimate their profits to minimize their income tax liability. Nevertheless, the intricate calculation involving the tax base and profits in corporate income tax presents additional avenues for corporations to engage in tax avoidance. Assuming that the corporate effective income tax rate is influenced by the tax collection initiatives of tax authorities, we employ the ratio of corporate income tax payment to pre-tax profit as a straightforward gauge of the effective corporate tax rate. Robust tax collection efforts reduce opportunities for firms to engage in tax evasion and corporate tax rates are likely to rise, thereby enhancing the tax burden on firms. Columns (1) and (2) in

Table 6. Mechanisms: Firms’ tax burden and production scale.

	Tax Burden		Production Scale		R&D Intensity
	(1)	(2)	(3)	(4)	(5)	(6)
lnVATshare	0.155 *		−0.079 ***		−0.018 ***
	(0.085)		(0.021)		(0.003)
lnCITshare		0.266 ***		−0.099 ***		−0.012 ***
		(0.0550)		(0.010)		(0.002)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
YearFE	Yes	Yes	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes	Yes	Yes
ProvinceFE	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.143	0.143	0.226	0.272	0.134	0.134
Observations	208,087	208,087	209,381	209,381	232,397	232,397

Notes: Robust standard errors in parentheses. The controls and constant terms are omitted to save space. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

show that higher tax sharing encourages local governments to enhance their tax collection efforts, leading to a higher tax burden for firms. This finding is in line with Lyu et al. [18], who suggest that tax efforts constitute an increasing function of the proportion of tax sharing for local governments. Elevating the tax-sharing ratio for local governments can incentivize them to be more proactive in tax administration. Some studies also emphasize an increase in the local proportion of tax sharing implies the strengthened tax administration of local government and higher tax revenue, which leads to an actual increase in the tax burden on enterprises [58]. A higher tax burden on firms is detrimental to their operational performance, thereby affecting their ability to enhance the quality of exported products.

5.2. Production Scale

The quality of firms’ inputs and products produced often positively depends on their production sizes. In other words, the larger the production scale of a firm, the more likely it exports higher-quality products. Columns (3) and (4) in Table 6 present the regression results of the impact of tax sharing on a firm’s production scale. Here, we use the logarithm of production value to reflect the firm’s production scale, and we find that higher VAT and CIT sharing ratios decrease the firm’s production values significantly. Limitations on production scale will hinder firms from investing in factors such as equipment to improve product quality, thus impeding the upgrading of export product quality.

5.3. Innovation

In this case, to explore whether stronger tax-sharing ratios can result in lower export quality, the identification of a causal relationship between tax-sharing and R&D input is necessary. In this part, we use R&D intensity (R&D intensity) to represent the firm’s innovation effects using the ratio of R&D expenses stock to the firm’s total assets. Although the patent is frequently used in innovation literature, the R&D intensity as a measure of innovation may outperform the patent in firm-level studies for two reasons as follows. First, Hall et al. (2005) point out that the ratio of R&D stock/assets is distributed more symmetrically, reflecting a stronger predicting power relative to a variety of knowledge stocks such as patents and citations with extremely skewness [59]. Second, R&D is a time and resource-consuming process that features uncertain time lagging in the eventual transformation into a patent. The R&D intensity data used are from the Chinese firm-level innovation dataset from 2008–2013 which contains relevant firm-level innovation variables. It is assumed that the R&D stock is subject to a 15% depreciation rate which is a common assumption in the literature [59]. The result of this channel is reported in Columns (5) and (6) in Table 6. It suggests that higher tax sharing at the prefecture level obstruct firms from conducting more R&D activity which is conducive to product quality improvement.

6. Heterogeneity Tests

The baseline estimates show a negative effect of local tax sharing on export product quality. However, this quality effect may vary across firms and industries due to their heterogeneity in many aspects such as TFP. We next investigate the potential heterogeneous effects on the firm-level export quality.

6.1. TFP

Firms with higher productivity are less responsive to tax fluctuation but have higher incentives to improve their product quality [60,61]. The estimates are reported in Columns (1) and (2) in

Table 7. Heterogeneity checks: TFP, IPP and the magnitude of tax revenue.

	Dependent Variable: lnquality
	(1)	(2)	(3)	(4)	(5)	(6)
lnVATshare	−0.114 ***		−0.586 **		−0.064
	(0.041)		(0.234)		(0.047)
lnCITshare		−0.029		−0.278 *		−0.203 ***
		(0.023)		(0.146)		(0.023)
lnVATshare×TFP	0.037 ***
	(0.141)
lnCITshare×TFP		0.011
		(0.008)
lnVATshare×lnIPP			0.212 **
			(0.083)
lnCITshare×lnIPP				0.108 ***
				(0.013)
lnVATshare×lnVATinc					0.003
					(0.006)
lnCITshare×lnCITinc						0.033 ***
						(0.004)
TFP	0.139 ***	0.098 ***
	(0.021)	(0.010)
lnIPP			0.351 ***	0.167 ***
			(0.122)	(0.061)
lnVATinc					0.026 ***
					(0.009)
lnCITinc						0.024 ***
						(0.003)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
YearFE	Yes	Yes	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes	Yes	Yes
ProvinceFE	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.103	0.103	0.096	0.096	0.093	0.093
Observations	180,272	180,272	166,080	166,080	232,365	232,369

Notes: Robust standard errors in parentheses. The controls and constant terms are omitted to save space. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

. In Columns (1) and (2), the estimated coefficients of lnVATshare × TFP and lnCITshare × TFP show that firms with higher TFP are less influenced on their product quality by tax-sharing ratios. Specifically, firms appear to be less but still negatively affected by VAT sharing from local government while the effect of the CIT sharing ratio is not significant. In theory, more productive firms can be operated at lower costs and thus they can be less exposed to negative shocks of the economic regulation from local government, thereby promoting the upgrading of export quality.

6.2. City IPP

Intellectual property rights protection (IPP) is an important component of a city’s business environment performance and a crucial part of the local governance agenda. It has a beneficial impact on the city’s capacity to attract innovative enterprises, and the positive impact of IPP on export product quality has also been discussed in literature [28,62,63,64]. To reflect prefectural IPP differences, we follow the method from Dong et al. to exploit the IPP scores of 65 major cities by interacting with the national Ginarte–Park (GP) index [28]. This method can mirror both national IPP and variations in city-level IPP, which only leads to parallel shifts of the IPP at the prefecture level without altering the relative magnitudes across cities. Columns (3) and (4) in Table 7 show the results. We find that the estimated coefficients of interaction terms lnVATshare × lnIPP and lnCITshare × lnIPP are significantly positive, indicating that IPP alleviates the negative quality effect of tax sharing. These results are in line with current literature which maintains that a better IPP helps stimulate firms’ innovation, enhance competitiveness, and provides a stable development environment with less government intervention [65]. A higher quality of products exported is subsequently easier to achieve.

6.3. Magnitude of Tax Revenue

Different regions within a country can have varying tax revenues due to the different industrial composition, economic activities, and population. Large regional differences in tax revenue can lead to disparities in public services, infrastructure development, and overall economic development, thereby influencing local firms’ investments and performances.

In Columns (5) and (6) in Table 7, we further incorporate the impacts of VAT revenue and CIT revenue, respectively, and the estimated coefficients of lnVATshare × lnVATinc and lnCITshare × lnCITinc show that firms located in regions where the tax revenues are large can alleviates the negative quality effect of tax sharing. Specifically, the alleviating effect is more significant by CIT revenue from local government while the alleviating effect of the VAT revenue is not significant.

6.4. Ownership

Due to the different characteristics of firms under different ownerships, the effect of tax policies on SOEs and non-SOEs may vary [66]. The subgroup regression results in

Table 8. Heterogeneity checks: Ownership.

	Dependent Variable: lnquality
	SOEs		Non-SOEs
	(1)	(2)	(3)	(4)
lnVATshare	−0.010		−0.048 ***
	(0.020)		(0.019)
lnCITshare		−0.014		−0.023 ***
		(0.015)		(0.009)
Control Variables	Yes	Yes	Yes	Yes
YearFE	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes
ProvinceFE	Yes	Yes	Yes	Yes
R2	0.076	0.076	0.108	0.108
Observations	103,366	103,366	129,009	129,009

Notes: Robust standard errors in parentheses. The controls and constant terms are omitted to save space. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

show that non-SOEs face significantly estimated coefficients relative to SOEs. This indicates that non-SOEs suffer more quality reduction in response to higher prefecture-level tax sharing ratios. One possible explanation is that non-SOEs on average face more financial constraints and higher tax sharing ratios can add more tax burden on non-SOEs. This could stem from disparities in the relationships between the government and enterprises, as well as the distinct extent of tax efforts, enabling SOEs to access a greater allocation of resources from the government under less management. Previous studies have found that state-owned enterprises possess ample financing channels and economic resources advantages, they may have closer ties to local governments and receive lower tax regulations and higher subsidies [30,67].

6.5. Market Power

The effects of tax sharing on firms’ innovation and export decisions can also be influenced by their market powers. Higher market power means a lower competition level. If a firm is more powerful in the market, it is more likely to transfer the tax burden to consumers by raising product prices, resulting in less motivation to improve their product quality. The Herfindahl index (HHI) is often used to measure the concentration level within industries, and a higher HHI indicates a more concentrated, or less competitive industry. We therefore calculate the Herfindahl index (HHI) for each 4-digit industry. Based on the median value of HHI, we divided the whole sample into two subgroups, high and low market concentration industries.

Table 9. Heterogeneity checks: Market power.

	Dependent Variable: lnquality
	High HHI		Low HHI
	(1)	(2)	(3)	(4)
lnVATshare	−0.032 **		−0.011
	(0.016)		(0.032)
lnCITshare		−0.021 **		−0.002
		(0.011)		(0.022)
Control Variables	Yes	Yes	Yes	Yes
YearFE	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes
ProvinceFE	Yes	Yes	Yes	Yes
R2	0.085	0.085	0.128	0.085
Observations	116,149	116,149	116,226	116,226

Notes: Robust standard errors in parentheses. The controls and constant terms are omitted to save space. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.

presents the subgroup regression results. Firms in high HHI industries have more significant estimated coefficients than low HHI industries. A larger negative effect of tax sharing occurs to firms in less competitive industries. This finding confirms our initial expectation and is consistent with the study that tax policies have smaller impacts on exports in less competitive industries [68].

7. Conclusions and Policy Implications

Higher tax sharing restructures the tax revenue distribution between central and local governments and incentivizes the latter to tax local firms. Using the tax-sharing ratio data at the prefecture level and firm-level data in China from 2008 to 2013, this paper explores the relationship between tax-sharing and firm export quality. To alleviate the endogeneity problem, we employ the IV-2SLS approach by constructing a Bartik IV and a provincial weighted average tax sharing ratio, common to all within-province prefectures, outside the province where firms are located. The empirical analysis confirms all the hypotheses we developed. Our empirical results show that a higher tax-sharing ratio reduces the quality of what firms ship abroad. The mechanism analysis shows that this quality reduction effect is realized because a higher level of tax sharing adds more tax burden on firms, reduces their production scales, and inhibits their innovation intensities. The heterogeneous quality effects exist across firm groups. For instance, firms with higher productivity can face a smaller negative quality effect. Moreover, the quality reduction effect is significant for firms in more concentrated industries.

Our findings extend the existing finance-domestic market economy research to a broader field, which bridges the fiscal policy and export relationship from the export quality perspective. This facilitates the deeper discussion of the new development pattern featuring a “dual cycle” that associates the domestic market economy and the international market economy. Our main finding confirms the importance of fiscal policy for firms’ participation and performance in international markets. Another important contribution is to develop an in-depth analysis of tax-sharing policy effects. This approach offers more insightful information on the effects of tax sharing on exporters’ behaviors. Finally, we provide detailed empirical evidence of underlying mechanisms linking tax-sharing systems and export quality of firms. This is equivalent to more insights into possible starting points for fiscal policy intended to enhance competitiveness in international markets for manufacturing firms. Policymakers should reconsider and adjust the tax-sharing ratios to reduce the negative impact on the quality of corporate exports. For markets with high industry concentration, it is recommended that policymakers take measures to reduce market concentration, thereby lowering the potential negative impact of changes in tax policy on export quality.