Customer and Tax Avoidance: How Does Customer Geographic Proximity Affect a Supplier’s Tax Avoidance?

Abstract

This paper examines the effect of customer geographic proximity on supplier tax avoidance. Based on 5135 Chinese firm–year observations from 2009 to 2020, we find a positive association between customer geographic distance and supplier tax avoidance. Moreover, this association is robust after studying endogeneity concerns. We further find that information asymmetry and detection risk are underlying mechanisms. We also find that this positive relation is more pronounced in suppliers with high financial risk, competitive industrial sectors, and weak marketization environments. Overall, our findings suggest that customer geographic localities within a country are an important factor affecting a supplier’s motivation for tax avoidance. Our research sheds light on how the change in a supplier’s information environment caused by differences in customer geographic proximity impacts its tax strategy.

1. Introduction

This paper investigates whether and how a supplier’s customer geographic proximity affects supplier tax avoidance. Corporate tax burdens firms globally, and firms have strong incentives to engage in tax avoidance activities in their tax planning strategy [1]. Existing studies have found that different stakeholders, such as external auditors [2], analysts [3], and institutional investors [4], may affect the motivations of corporate tax avoidance. There is an increasing awareness that tax avoidance is also noticeably influenced by a firm’s tax planning strategies via the supply chain [5,6]. Recently, implementing tax avoidance strategies via the supply chain has captured the attention of scholars and policymakers. A famous practical example is that of the industrial manufacturer Caterpillar, who, according to a Senate report in 2014, shifted billions of dollars in profits by tax planning related to its supply chain to avoid paying USD 2.4 billion in taxes in the U.S. Existing theoretical studies have mainly focused on firms’ tax strategies in terms of the global supply chain strategy of using the tax rate gap between different countries [6]. Chen and Lin [3] have extended the incentive of tax avoidance to include information asymmetry between firms and stakeholders. However, as an important signal of the corporate information environments, customer geographic proximity has been overlooked in related research. This can directly affect the cost of information exchange in the supply chain and has been ignored in prior tax-related studies (see Hanlon and Heitzman [1] for a review). The purpose of this study is to fill this gap by studying the impact of the geographic proximity of a supplier to its customers on corporate tax avoidance.

Whether customer geographic proximity affects suppliers’ tax avoidance is an empirical question. Intuitively, customers who are proximate to the supplier incur lower costs of information processing and exchange via the supply chain [7,8], which provides opportunities for a supplier firm and its customers to build a closer strategic alliance in tax avoidance [6]. In this way, a supplier with geographically close customers would have an information advantage that might provide more incentive for collusion in tax avoidance than one with geographically remote customers. However, from the supplier’s information environment perspective, customers proximate to the supplier also make information search costs lower for outside stakeholders, such as analysts [9,10], auditors and regulators [11,12], helping them to easily acquire corporate operating and tax-related information This generates more transparent information from the supplier and increases the risk of detection from regulators, thus decreasing their incentives to engage in tax avoidance. Compared to geographically close customers, geographically remote customers are largely opaque to outside stakeholders [13], providing opportunities for firms to engage in tax avoidance free from monitors. Furthermore, customers’ geographic proximity to their suppliers may be optimal for them to monitor their suppliers’ tax avoidance activities. Customers would do this to ensure that their own supply chain remains uninterrupted and thus decrease the extent of the supplier’s tax avoidance. Overall, it is, ex-ante, unclear how customer geographic proximity might affect supplier tax avoidance.

Further, how customer geographical proximity affects supplier tax avoidance is also an interesting and important question to be explored. Existing literature shows that the information environment is an important consideration when firms conduct their tax avoidance strategies [3]. We investigate the potential channels for tax avoidance and provide a further empirical explanation of how a supplier’s information environment affects its tax avoidance. In addition, the motivations for the supplier to engage in tax avoidance may vary across different corporate, industry, and market characteristics, which may also be interesting to explore further.

The context of China helps us to test the research question. First, globally, China has the second-largest land area (https://www.worldometers.info/geography/largest-countries-in-the-world/, accessed on 1 January 2022), which provides suppliers with potential opportunities to choose geographically close or remote customers in a consistent setting that allows us to control for other factors, such as regulation and culture, compared with cross-country analyses. The geographic distance between the suppliers and customers provides sufficient within-country variations to identify the real effect of customer geographic proximity on suppliers’ tax strategies. Second, although research on corporate tax avoidance from a perspective of specific strategies that require the cooperation of other parties has been carried out in a China context (see Tang [14] for a recent review), little is known about the role of the customer. China provides an appropriate setting because listed Chinese firms offer voluntary disclosure of information on their top five customers. Third, The State Council issued the Law of the People’s Republic of China to administer the levying and collection of taxes in 1993 (hereafter, Taxes Collection Act). This law is the first special law for the levying and collection of taxes issued after the founding of the People’s Republic of China, a law that the tax authorities must obey. It requires regulators to carefully verify suppliers’ tax-related information, particularly repeated transactions between suppliers and their major customers. This is achieved by sending letters or conducting on-site supervision. Consequently, China’s unique geographic characteristics and institutional features provide excellent opportunities to investigate our research question.

Using a sample of 5135 firm–year observations of 1430 listed Chinese firms and their top five customers from 2009 to 2020, we found a positive association between major customers’ weighted geographic distance and their supplier’s extent of tax avoidance. Furthermore, we employed a quasi-natural experimental approach, Heckman two-step and entropy balancing methods, to mitigate endogeneity concerns. Finally, we used alternative measures, changed the standard error estimation method, changed the fixed effect, and considered the impacts of the 2008 financial crisis, the value-added tax (VAT) reform in 2016, and the effects of customer concentration to corroborate our main findings. The results are consistent across all additional robust tests.

Furthermore, we uncover two underlying mechanisms. First, the information asymmetry channel supports that the customer geographic proximity can weaken the supplier–customer relationship information advantages through a convenient way for outside stakeholders to simultaneously acquire related information of the supply chain, reducing supplier tax avoidance. Second, the detection risk channel that supports customer geographic proximity can lessen the scrutiny cost of regulators, increasing the risk of detection associated with supplier tax avoidance activities, thus reducing managers’ impetus to engage in tax avoidance. Using the different functions between the supplier office address and the registered address, we further rule out the potential concern exerted by customer monitoring.

Finally, we examine the moderating effects and find that the positive relationship between the major customers’ weighted geographic distance and supplier tax avoidance is more pronounced for suppliers at high financial risk in competitive industrial sectors and weak marketization environments.

Our findings may have several contributions. First, we extend to the literature on the determinants of corporate tax avoidance from the supply chain perspective. Prior studies have examined the shareholders, managers, auditors, and analysts are factors that can explain the variation of corporate tax avoidance [4,15,16,17]. Recent literature extends the factors on tax avoidance to the supply chain and finds that suppliers with more concentrated customers or closer customer relationships are more likely to engage in tax avoidance [5,6]. Our research extends this literature by focusing on the supplier–customers relationship from the geographic localities within-country perspective and provides evidence that customer geographic proximity can reduce supplier tax avoidance. In addition, many prior studies have explored the channel of tax avoidance through the supply chain, mainly focusing on the strategy with customers through subsidiary recognition in low-tax jurisdictions (see Cen, Maydew, Zhang and Zuo [6]). Our study from a within-country perspective on how customers’ geographic location may affect the supplier firms’ tax avoidance strategy provides a clean setting to eliminate other factors such as regulation and tax jurisdiction across countries.

Second, our study contributes to a growing literature on the economic consequences of the firm’s stakeholder geographic proximity on its corporate decisions [18,19,20,21,22], especially the supplier–customer geographic relationships on supplier decisions, such as innovation, stock price crash risk and risk-taking [8,23,24]. We complement the existing research by focusing on the effect of supplier–customer geographic proximity on tax avoidance, which represents a corporate decision that is highly influenced by corporate information transparency and agency problems [3,25,26]. Moreover, evidence on customer geographic proximity shows that customers’ geographical proximity may help customers monitor supplier behaviors and thus increase suppliers’ corporate governance [23]. Our findings give another perspective that customers’ geographical proximity may also help related outside stakeholders monitor supplier’s behaviors through obtaining more supply chain information, which also increases corporate governance and sustainable management. This may also provide some theoretical and practical implications for researchers and policymakers.

Third, our study also contributes to a growing strand of research that links the firm’s information environment to its tax avoidance decisions. Existing scholars find that information environments, such as financial transparency and analyst coverage [3,25], can affect corporate tax avoidance by alleviating information asymmetry between firms and outside stakeholders. However, very few explore the firm’s tax avoidance strategies from the supply chain information environments, reflected by the geographic proximity between the supplier and its customers. Our investigation of the two potential mechanisms pointedly further provides an empirical explanation of how the supplier’s information environment affects its tax avoidance. We find that customer geographic proximity helps the outside stakeholders conveniently and cheaply acquire corporate operating and tax-related information, generating more transparent information from the supplier and detecting risks from regulators. We expect these findings to inform a new viewpoint on monitoring corporate tax avoidance among auditors, tax authorities, and policymakers.

The remainder of this paper proceeds as follows. Section 2 introduces the literature review and hypotheses development. Section 3 describes the data and research design. Section 4 presents the baseline results along with various tests to address the endogeneity problems. Section 5 discusses possible underlying mechanisms. Section 6 offers moderate effects tests, and Section 7 provides additional analyses. Section 8 summarizes and concludes.

2. Literature Review and Hypotheses Development

2.1. Customer Geographic Proximity and Supplier Tax Avoidance

The significant variation in the geographical localities of economic entities has attracted much academic attention (e.g., [11,20,27,28,29,30]). A common assumption from the literature is that geographic distance hurts the linking of information and induces interaction exchange costs across economic entities [31]. Based on this assumption, the existing finance and accounting literature investigates whether and to what extent the geographic proximity between the firm and its stakeholders (e.g., banks, analysts, auditors) could influence the firm. For example, the firms’ geographic distance from their creditors could affect their information asymmetry, impacting their borrowing probability and the magnitude of loan rates [19,20,32]. Firm proximity to its plants increases plant-level investment and productivity because this geographic proximity makes it easier for headquarters to monitor and acquire information about plants [21]. Moreover, the information advantage arising from the geographic proximity of the auditor–client distance may help facilitate the auditors’ supervision and reduce the firms’ real earnings management (REM) [22]. In sum, this body of research argues that geographic proximity plays an important role in corporate decision-making behaviors, mainly by facilitating the ease of collecting and transferring information between the firm and its outside stakeholders and hence creating a more transparent information environment for the firm [33].

In line with these studies on geographic proximity, customer geographic proximity can be crucial because it can directly reflect the level of information asymmetry between the supplier and its customers. Meanwhile, customer geographic proximity is vital because it helps outsiders (e.g., analysts, auditors, regulators, and investors) better understand the supplier’s actual operating situation in the supply chain. However, to our knowledge, few studies explore the economic consequences of customer geographic proximity. Chu, Tian and Wang [8] investigated the economic consequence of customer geographic proximity on one of the suppliers’ investment decisions—innovation. More recently, Cao, Zhang and Yuan [23] found that geographically nearby major customers negatively impact suppliers’ stock price crash risk by using Chinese data. Huang and Fan [24] provide more evidence that customer geographic proximity significantly reduces supplier firms’ risk-taking. As tax planning strategies represent a firm’s important decision that is highly influenced by corporate information transparency and agency problems [3,25,26], we extend this research to investigate whether customer geographic proximity affects supplier tax avoidance.

The customer geographic proximity may affect supplier tax avoidance through the trade-off cost and benefits under the consideration of decreasing information asymmetry and increasing risk detection. On the one hand, customer geographic proximity could enhance suppliers’ tax avoidance because of the decreasing information asymmetry for the supply chain. First, geographic proximity is associated with lower information transformation costs [27,34,35,36]. A lower cost in information transformation helps enhance routine communication and capture timely feedback between the supplier and its customers [8]. Frequent communication and cheap information exchange provide a good opportunity for the supplier to build a closer strategic alliance with its customers. The close supplier–customer relationship could improve the routine coordination between the supplier and its customers, especially in terms of tax avoidance [6]. The alliance based on a close supplier–customer relationship might entail hiding revenue, inflating expenses, managing earnings, and transferring profits intertemporal to engage in more tax avoidance [14,37,38,39,40]. Second, customer geographic proximity implies access to more information sources about customers as Chu, Tian and Wang [8] suggested that geographic proximity between the customer and the supplier can facilitate suppliers to collect more soft information (Soft information refers to the information that is difficult to write down on paper, store electronically, communicate or transfer to others Petersen and Rajan [18]). In this role, more soft information enables the suppliers to have more bargaining power. Geographic proximity to customers could enhance a supplier’s bargaining power because it could provide goods or services at a lower transaction cost. According to the resource dependence theory [41], the supplier’s resource is essential when its customers can acquire raw materials as inputs at a low cost. Customers’ dependence on a supplier for necessary resources further gives the supplier more bargaining power. The higher bargaining power enables the supplier to manage earnings by coercing its customers to do business according to its willingness [42], thus boldly engaging in tax avoidance via the supply chain. Overall, these studies suggest that customer geographic proximity can help the supplier avoid tax aggressively by mitigating information asymmetry in the supply chain and enhancing the supplier’s bargaining power.

On the other hand, however, customer geographic proximity may lessen the supplier’s incentive to avoid tax aggressively because of decreasing information asymmetry for outside stakeholders. Specifically, customers proximate to the supplier make it easier for outsiders to acquire and monitor information about the supplier’s actual operation in the supply chain. Although the farther customer geographic distance would increase asymmetric information between the supplier and its customers [31], it also increases the asymmetric information between the supplier and its outsiders (e.g., analysts, auditors, and tax authorities). Compared to the outsiders who might only limited ways (e.g., from the customers’ disclosures or some unformed channels) to obtain the supplier’s customer information [7], the supplier and its customers would have a relative information advantage due to its strategic relationship with its customers [5,6]. This information advantage thereby potentially imposes excessive opportunities on the managers, forcing them to avoid taxes strategically via the supply chain, such as income shifting among their subsidiaries in low-tax jurisdictions, and to reduce tax payments at a lower financial cost without affecting the supplier’s aggregate reported pretax income [6,39].

Moreover, customer geographic proximity may increase the risk and threat of tax avoidance detection from outside regulators. This risk and threat can increase a firm’s tax avoidance costs, which is one of the main factors that affect the corporate tax strategy reactions [43,44,45], leading to a decline in the supplier tax avoidance incentives and ability. As a result, customer geographic proximity would reduce the supplier’s tax avoidance, while the farther customer geographic distance might increase the supplier’s tax avoidance because of the less risk and threat of tax detection from tax authorities. From the perspective of tax authorities, the amount of tax-related information that suppliers provide in tax declarations is limited, making it hard to fully and truly discern the actual situation of the supplier’s tax obligations. To alleviate information asymmetry and monitor the supplier’s tax obligations [46], tax authorities usually need to make an effort to search for tax-related information in a series of ways [47]. Among these ways, on-site visit supervision is sometimes necessary. The farther customer geographic distance implies a higher cost of traffic, a longer time for on-site monitoring, and a higher information search cost for tax authorities. A higher tax regulation scrutiny cost accompanied by a lower risk and threat of detecting the supplier’s tax avoidance would incent the firm to engage in tax avoidance. In addition, the customer geographic proximity may also make it convenient for customers to monitor their suppliers’ tax avoidance, as customers have the incentive to monitor their suppliers’ decisions in order to maintain safety and stability along the supply chain. The above studies suggest that customer geographic proximity can provide outside stakeholders with more transparent information about the supplier’s actual operation in the supply chain and incur more detection risk from tax authorities to the supplier, thus impeding its tax strategies implementation, lessening its incentives to avoid tax aggressively, and overall reducing supplier tax avoidance.

Collectively, suppliers with geographically close customer proximity might increase or decrease tax avoidance because of lower information asymmetry between the supplier and its customers for the use of outside stakeholders and because of the risk detection, such as higher risk and threat associated with tax authorities. This suggests that the effect of customer geographic proximity on supplier tax avoidance is ultimately an empirical question. Therefore, based on the discussion above, we develop our hypothesis in the null form.

Hypothesis H1:

Customer geographic distance has no significant effect on supplier tax avoidance.

2.2. The Moderating Effect of Financial Leverage, Industry Regulation and Market Environment

We examine three characteristics from firm, industry, and institutional environment levels, namely financial risk, industry regulation, and marketization environment. These characteristics influence a firm’s incentive to engage in tax avoidance, given the cross-sectional variation.

Financial risk indicates that firms face longer financial leverage, a higher possibility of decreasing cash flow, incurring an increased financial burden, running into operating trouble, and even bankruptcy due to financial leverage’s effect on its solvency. This danger leads the firm to save cash and enhance profitability [3]. In addition, the firm is more likely to shore up cash to mitigate future financial distress and maintain its competitiveness because it is more challenging to seek external financing in a high-financial risk situation. Compared to firms with low financial risk, those with high financial risk are more likely to be affected by the impact of customer proximity. Therefore, we conjecture that the effect of customer geographic distance on supplier tax avoidance should be stronger (weaker) in firms with higher (lower) financial risk. This discussion leads to the following hypothesis:

Hypothesis H2a:

The relation between customer geographic distance and supplier tax avoidance is strengthened by high financial risk.

Oh et al. [48] argued that industry characteristics play a crucial role in a firm’s strategic decisions. Compared to regulatory industries, competitive industries face more competition from both factor and product markets [17]. They may not have sufficient credit resources driven by the administrative government’s intervention and do not have super-profits derived from high entry industry barriers, low transaction costs, and strong monopoly power. Under these external situations, firms in competitive industries are, in some ways, more aggressive in saving cash [37]. As a result, tax avoidance might be more needed by firms in competitive industries. In other words, the firms in regulatory industries have little incentive to leverage the relationship between their customers’ geographic proximity and tax avoidance. Therefore, we expect that the effect of customer geographic distance on supplier tax avoidance should be stronger (weaker) in firms in competitive (regulatory) industries. Given the aforementioned discussion, we state our hypothesis as follows:

Hypothesis H2b:

The relation between customer geographic distance and supplier tax avoidance is weakened by regulatory industries.

The process of marketization is vital for enhancing economic growth in China. Chinese marketization denotes institutional reforms, shifting the economic system from a planned economy into a market economy. These reforms refer to a set of institutional reforms of the economy, society, and law rather than some simple daily lifestyle changes [49]. Specifically, the marketization process in China has a significant influence on allocating resources and government policies. Due to this circumstance, corporate decisions are easily affected by the external marketization environment, especially public governance [50]. A high degree of marketization usually means an increased role in public governance. As a result, the supplier–customer relationship may receive more attention from the public. In this role, the supplier–customer relationship might be less impetus to be stable in cooperation.

Meanwhile, firms operating in a high marketization environment have fewer incentives to engage in tax avoidance activities but prefer regular operating activities. In the aspect of business and trade operations, a high level of marketization might also provide firms with an abundance of conveniences, such as lower transaction costs, more workers, and better legal protection.

Therefore, these external factors from higher marketization promote firms to operate normally. On the one hand, firms operating in intense public attention environments face unstable coordination relationships when the firms aggressively avoid tax. On the other hand, firms could easily and immediately grasp an investment opportunity and increase cash inflow when it suddenly appears. Based on the discussion, we conjecture that the relationship between the supplier’s customer geographic proximity and its corporate tax avoidance is stronger in regions with weak marketization environments.

Hypothesis H2c:

The relation between customer geographic distance and supplier tax avoidance is strengthened by weak marketization.

3. Research Design

3.1. Data and Sample Selection

Our initial corporate tax avoidance sample includes Chinese A-share firms publicly listed on the Shanghai Stock Exchange and Shenzhen Stock Exchange during 2009–2020. We start our sample in 2009 to remove the potential effect of the new Corporate Income Tax Law on the consistent measure of corporate tax avoidance, given the legal effect of the new Corporate Income Tax Law on 1 January 2008 [50]. We obtain the statutory tax rate from the WIND financial database and the remainder of the corporate financial data from CSMAR.

We measure the supplier’s customer geographic proximity mainly based on the CSMAR Supply Chain database, a relatively comprehensive database that includes most of the Chinese A-share listed firm supply chain data from 2001 to 2020. Specifically, this database provides the great-circle distance data between a listed firm and its top five customers, calculated by the Haversine formula following Kang and Kim [51] (In this case, the great-circle distance is the shortest distance between the supplier firm and one of its top five customers on the surface of the Earth. After identifying two points on the Earth based on latitudes and longitudes information about the supplier and one of its top five customers, the Haversine formula could generate a great-circle distance as follows:

$$\begin{array}{ll}distanc{e}_{greatcircle}& \\ & =\arccos \{\cos \left(la{t}_s\right)\cos \left(lo{n}_s\right)\cos \left(la{t}_c\right)\cos \left(lo{n}_c\right)\\ & +\cos \left(la{t}_s\right)\sin \left(lo{n}_s\right)\cos \left(la{t}_c\right)\sin \left(lo{n}_c\right)\\ & +\sin \left(la{t}_s\right)\sin \left(la{t}_c\right)\}2\pi r/360\end{array}$$

where $\left(la{t}_s, lo{n}_s\right)$ and $\left(la{t}_c, lo{n}_c\right)$ is the latitude and longitude of the supplier’s registered address and one of its top five customers address, respectively). Other studies using the Haversine formula to obtain the geographic distance between the firm and its stakeholders include Jensen, Kim and Yi [12] and Li, Lin and Luo [22]. To obtain as much customer geographic distance data as possible, we complement the missing data in CSMAR with our manual calculation results produced from a similar method (For the missing data in CSMAR, we collect the top five customers’ information from the fiscal reports of listed firms. We then use Python software to link the customers’ names and information to their corresponding addresses according to a commercial survey website (Sky Eye Search system) or Google Search. Finally, we identify the latitudes and longitudes of the supplier–customers pairs on Google Maps and calculate the circle distance between the supplier firm and its top five customers). Next, we use this dataset to construct the weighted geographic distance as the annual firm-level customer geographic proximity measures following Petersen and Rajan [18].

We conduct the sample selection as shown in

Table 1. Sample selection and description.

Sampling Procedure	Number of Firm–Year Observations	Number of Unique Firms
Initial sample during 2009–2020	43,518	4930
Less: Observations in the A-B or A-H crossing listing	(2262)	(205)
Less: Observations in the financial or real estate industry	(1761)	(39)
Less: Observations with special treatment (ST)	(2590)	(243)
Less: Observations with negative pretax earnings or income tax expense	(3631)	(10)
Less: Observations with missing major customers’ location or financial data on any control variable	(28,139)	(3003)
Final sample	5135	1430

: (1) exclude firms listed on both A-share and B-share (or H-share) stock markets, (2) exclude firms in the financial or real estate industry due to their special regulations, accounting policies and peculiar capital and debt structure [50,52,53,54], (3) exclude firms with special treatment, (4) exclude firm–year observations with negative income tax expenses or pretax earnings, and (5) exclude firm–year observations with missing major customers location and missing key financial variables (e.g., [55,56,57]). Finally, our final sample for the main analysis comprises 5135 firm–year observations from 1430 firms.

3.2. Variable Definitions

3.2.1. Dependent Variable: Tax Avoidance

Following prior literature (e.g., [5,6,50]), we adopt the two major metrics of tax avoidance in year t. The first measure, $ETR\_GAA{P}_{i,t}$, is the total income tax expense divided by the pretax book income in year t. $ETR\_GAA{P}_{i,t}$ use total income tax expense as the numerator and thus capture the extent of tax avoidance caused by permanent and temporary book–tax differences. A higher $ETR\_GAA{P}_{i,t}$ indicates a lower extent of tax avoidance.

The second measure of tax avoidance, $SM{E}_{i,t}$, is the difference between the statutory tax rate and effective tax rate ($ETR\_GAA{P}_{i,t}$) in year t. $SM{E}_{i,t}$ controls for the applicable statutory tax rate and thus could capture the extent of tax avoidance resulting from innate disparity in tax rate and intentional tax avoidance strategy. Tang [14] argues that the effective rate adjusted by the corresponding statutory tax rate is a more appropriate measure of tax avoidance under the Chinese institutional background. A higher $SM{E}_{i,t}$ represents a greater extent of tax avoidance.

We also use other measures of tax avoidance, $ETR\_CU{R}_{i,t}$ (total income tax expense less deferred income tax expense divided by pretax book income), $MSM{E}_{i,t}$ (by the average of the difference of statutory tax rate and effective tax rate (SME) in the last three years to consider the long-run effect), $BT{D}_{i,t}$ (the difference of the book and taxable income in yeat t scaled by total assets) and$DDBT{D}_{i,t}$ (The movements in book–tax differences (BTD) unexplained by total accruals) as alternative metrics of tax avoidance for the robustness test, which also have been used in prior studies (e.g., [25,58,59]). These alternative variables’ definitions are detailed in Table S1.

3.2.2. Test Variable: Customer Geographic Proximity

Following existing studies [23,24], we measure a supplier firm’s customer geographic proximity ($Distanc{e}_{i,t}$) as the weighted geographic distance of the supplier firm i with its major customers in year t. Since a supplier firm i voluntarily discloses its top five customers’ information in year t, the customer-related information theoretically allows us to obtain the corresponding geographic distances with the supplier by the Haversine formula. Accordingly, we estimate the weighted geographic distance of the supplier firm i with its major customers in year t through the following equation:

$$Distanc{e}_{i,t}=\ln (1+{\displaystyle \sum }_{j}DI{S}_{i,j,t}\times Rati{o}_{i,j,t})$$

(1)

where $DI{S}_{i,j,t}$represents the geographic distance between the supplier firm i and one of its top five customers j in year t. $Rati{o}_{i,j,t}$ represents the ratio of the supplier firm i’s sales to one of its top five customers j concerning the total sales of the supplier i in year t. Customer geographic proximity is that the supplier firm is geographically nearness to that major customers’ firm. $Distanc{e}_{i,t}$ demonstrates the supplier firm’s customer geographic proximity (work geographically close), and a larger $Distanc{e}_{i,t}$ indicates that the supplier firm works with customers who are farther away.

Following the method of indicator construction above, we also use $Ratio{1}_{i,j,t}$ (the ratio of supplier firm i’s sales to one of its top five customers j concerning the sum of supplier firm i’s sales to its top five customers in year t) to calculate $Distance{1}_{i,t}$, which is our first alternative test variable. Moreover, we use the simple arithmetic average of the firm’s sales as the weight to calculate $Distance{2}_{i,t}$, which is another alternative test variable. Both alternative measures are employed for the additional robustness tests.

3.2.3. Control Variables

Consistent with the existing literature, we identify a set of control variables firm characteristics estimated in year t that could affect the extent of corporate tax avoidance (e.g., [53,60,61]): $Siz{e}_{i,t}$ is the natural logarithm of total assets; $LE{V}_{i,t}$ is the total liabilities scaled by total assets, which captures the tax shield of debt [33,62]; $RO{A}_{i,t}$ is the net profit scaled by total assets, which affects the firm’s incentives to avoid tax [63]; $Growt{h}_{i,t}$ is the sales of current year less sales of last year divided by the sales of last year; $Tur{n}_{i,t}$ is the book value of total revenue divided by the book value of total assets; $PP{E}_{i,t}$ is the net property, plant, and equipment scaled by total assets; $INTA{N}_{i,t}$ is the net intangibility scaled by total assets; $PI{S}_{i,t}$ is the shares held by institutions scaled by total shares, according to Khan, Srinivasan and Tan [4] and China’s institutional environments; $Los{s}_{i,t}$ is a dummy variable that equals one if net profit is negative in year t-1 and equals zero otherwise, which partly reflects the corporate tax avoidance strategy using intertemporal income shifting [15]; $M{B}_{i,t}$ is the ratio of the market value of total equity to the book value of total assets. $P{C}_{i,t}$is a dummy variable that equals one if the board chairman or CEO has a political connection and zero otherwise [64], reflecting the competitive advantages of firms related to tax avoidance strategies at the institutional level. See Table S1 for the detailed definitions and data resources of all variables.

3.3. Model

To examine our hypothesis, we follow prior studies [5,6] to adopt the ordinary least squares (OLS) method because it can provide a quick benchmark and ensure the precision of our statistical inference by adding a set of control variables and different-level fixed effects in our panel data [65,66]. Therefore, we estimate the following baseline model:

$$\begin{array}{ll}Tax\_Avoidanc{e}_{i,t}\hfill & \hfill \\ \hfill & ={\beta }_0+{\beta }_{1}Distanc{e}_{i,t}+{\beta }_{2}Siz{e}_{i,t}+{\beta }_{3}LE{V}_{i,t}+{\beta }_{4}RO{A}_{i,t}\hfill \\ \hfill & +{\beta }_{5}Growt{h}_{i,t}+{\beta }_{6}Tur{n}_{i,t}+{\beta }_{7}PP{E}_{i,t}+{\beta }_{8}INTA{N}_{i,t}+{\beta }_{9}PI{S}_{i,t}\hfill \\ \hfill & +{\beta }_{10}Los{s}_{i,t}+{\beta }_{11}M{B}_{i,t}+{\beta }_{12}P{C}_{i,t}+{\displaystyle \sum Industry}\hfill \\ \hfill & +{\displaystyle \sum Region+{\displaystyle \sum Year+{\epsilon }_{i,t}}}\hfill \end{array}$$

(2)

Our dependent variable $Tax Avoidanc{e}_{i,t}$ is proxied by $ETR\_GAA{P}_{i,t}$ and $SM{E}_{i,t}$ for the supplier firm i in year t. Our test variable $Distanc{e}_{i,t}$ is the supplier firm i weighted geographic distance with its major customers in year t and represents the supplier’s customer geographic proximity. The control variables of firm factors that could affect a firm’s tax avoidance are described in Section 3.2.3. Following prior studies, we also add industry fixed effects, region fixed effects, and year fixed effects to the model to control the effect of industry, region, and time-series omitted factors (e.g., [67,68]). See Table S1 for the detailed definitions of all variables used in Equation (2). In addition, we cluster standard errors at the supplier firm level.

4. Empirical Results

4.1. Summary Statistics

Table 2. Descriptive statistics.

Variables	N	Mean	Sd	P10	P25	P50	P75	P90
ETR_GAAP	5135	0.203	0.134	0.081	0.130	0.170	0.248	0.337
SME	5135	−0.004	0.131	−0.124	−0.036	0.007	0.060	0.126
Distance	5135	2.863	1.469	0.871	1.841	2.830	3.800	4.828
Size	5135	21.961	1.170	20.575	21.083	21.824	22.678	23.592
LEV	5135	0.427	0.208	0.141	0.260	0.427	0.590	0.706
ROA	5135	0.053	0.043	0.008	0.021	0.043	0.071	0.109
Growth	5135	0.227	0.515	−0.129	0.002	0.128	0.297	0.557
Turn	5135	0.658	0.457	0.226	0.361	0.546	0.815	1.213
PPE	5135	0.244	0.176	0.047	0.106	0.205	0.344	0.509
INTAN	5135	0.051	0.055	0.006	0.018	0.035	0.062	0.109
PIS	5135	0.067	0.075	0.003	0.011	0.041	0.096	0.171
Loss	5135	0.065	0.247	0.000	0.000	0.000	0.000	0.000
MB	5135	2.058	1.340	1.065	1.242	1.606	2.350	3.569
PC	5135	0.382	0.486	0.000	0.000	0.000	1.000	1.000
ETR_CUR	5135	0.221	0.160	0.082	0.139	0.185	0.266	0.373
MSME	4624	0.016	0.121	−0.097	−0.027	0.011	0.072	0.148
BTD	5135	0.000	0.027	−0.027	−0.012	−0.001	0.011	0.030
DDBTD	5135	−0.001	0.034	−0.038	−0.017	−0.001	0.015	0.036
Distance1	5131	4.828	1.820	2.342	3.512	5.080	6.168	6.989
Distance2	5135	5.801	1.502	3.957	4.983	5.971	6.762	7.598

Notes: This table reports sample summary statistics, which shows the descriptive statistics of variables used in our main analysis. All continuous variables are winsorized at the 1% and 99% percentile. See Table S1 for the detailed definitions of variables.

reports the descriptive statistics of the main variables used in our analysis. The average effective tax rate ($ETR\_GAA{P}_{i,t}$) paid by the sample supplier firms is 20.3%, consistent with that for US firms, such as 29.5% (i.e., [6]) and 22.0% (i.e., [69]). The mean value of the difference between the statutory tax rate and effective tax rate ($SM{E}_{i,t}$) is −0.4%, which suggests that the listed Chinese firms have a high actual tax burden [67]. The sample firms have a mean customer geographic proximity ($Distanc{e}_{i,t}$) of around 16.514 km (2.863 reported in Table 2). Other variables are consistent with those in the existing studies (e.g., [50,70,71,72]).

We also conduct the Pearson and Spearman correlation analysis of the key variables. The detailed correlations are presented in Table S2. Not surprisingly, $ETR\_GAA{P}_{i,t}$, and $SM{E}_{i,t}$ are highly correlated with each other, showing the consistency of our dependent variable measures. The correlations between $Distanc{e}_{i,t}$ and the dependent variables are negative (i.e., $ETR\_GAA{P}_{i,t}$) and positive (i.e., $SM{E}_{i,t}$). These correlations are significant at conventional levels, preliminarily suggesting the effect of customer geographic proximity on supplier tax avoidance. Furthermore, the correlations among the independent variables are all below 0.6, which indicates a lower concern of multicollinearity.

4.2. Baseline Findings

In this section, we examine the effect of customer geographic proximity on the supplier firm’s tax avoidance by estimating the OLS regression model. The results presented in

Table 3. Customer geographic proximity and supplier tax avoidance.

	ETR_GAAP		SME
	(1)	(2)	(3)	(1)
Distance	−0.007 ***	−0.004 ***	0.003 *	0.004 **
	(−3.93)	(−2.61)	(1.69)	(2.32)
Size		0.004		0.001
		(1.08)		(0.18)
LEV		0.026		−0.018
		(1.60)		(−1.09)
ROA		−0.889 ***		0.825 ***
		(−12.66)		(11.86)
Growth		−0.006		0.008 **
		(−1.48)		(2.18)
Turn		0.025 ***		−0.019 ***
		(3.98)		(−2.94)
PPE		−0.047 **		0.057 ***
		(−2.46)		(2.93)
INTAN		0.083 *		−0.054
		(1.68)		(−1.09)
PIS		−0.077 **		0.000
		(−2.50)		(0.00)
Loss		0.024 **		−0.014
		(2.06)		(−1.14)
MB		0.003		0.000
		(1.01)		(0.15)
PC		−0.004		0.001
		(−0.88)		(0.29)
Industry FE	No	Yes	No	Yes
Region FE	No	Yes	No	Yes
Year FE	No	Yes	No	Yes
Observations	5135	5135	5135	5135
Adjusted R2	0.006	0.260	0.001	0.199

Notes: This table reports the OLS regressions of the supplier firm’s tax avoidance variables on the customer geographic proximity, including a set of control variables and unreported industry, region, and year fixed effects. Specifically, we choose the city-fixed effect as the region-fixed effect since the observation unit is the firm–year. The estimation window is 2009–2020. All variable definitions are detailed in Table S1. T-statistics shown in parentheses are based on standard errors adjusted for firm-level clustering. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

, based on Equation (2), demonstrate that the supplier firm with more geographically remote customers tends to engage in many tax avoidance activities. Column (1) regresses the effective tax rate ($ETR\_GAA{P}_{i,t}$) on the customer geographic proximity variable ($Distanc{e}_{i,t}$), and column (2) further introduces controls of firm characteristics, industry, region, and time-series omitted factors that could affect the extent of corporate tax avoidance. Columns (3) to (4) repeat the regressors specifications of columns (1) to (2) but substitute the difference in the statutory tax rate ($SM{E}_{i,t}$) for the effective tax rate ($ETR\_GAA{P}_{i,t}$) as the dependent variable for Equation (2).

Table 3 reports the detailed coefficients. Briefly, we find that the adjusted ${\mathrm{R}}^2$of the model has significantly increased after adding a set of control variables and fixed effects based on the same sample, which is consistent with prior studies [73]. Specifically, in columns (1) to (2), the coefficients on $Distanc{e}_{i,t}$ are both significantly negative (t-stat = −3.93 and −2.61). Hence, the supplier firms’ tax burdens are largely attenuated when the supplier firm with greater customer geographic distance. In other words, supplier firms with more geographically remote customers tend to engage in tax avoidance to a similar extent. Economically, the magnitude of this effect is also meaningful. We find that, in terms of the results of column (2), a standard deviation (1.469, i.e., 3.345 km) increase in $Distanc{e}_{i,t}$ leads to a 4.4% decrease in $ETR\_GAA{P}_{i,t}$ at the sample standard deviation (The detailed calculation of economic effect as follows: $ETR\_GAA{P}_{i,t}=\raisebox{1ex}{$\left(0.004\times 1.469\right)$}\!\left/ \!\raisebox{-1ex}{$0.134$}\right.\approx 4.4\%$, where the standard deviation of $Distanc{e}_{i,t}$ is 1.469 and the sample standard deviation of $ETR\_GAA{P}_{i,t}$ is 0.134).

In columns (3) to (4), the coefficients on $Distanc{e}_{i,t}$ are both significantly positive (t-stat = 1.69 and 2.32). These results indicate that the effective tax rate adjusted by the statutory tax rate, proxied by $SM{E}_{i,t}$, is positively associated with customer geographic distance. We find that, in terms of the results of column (4), a standard deviation increases in $Distanc{e}_{i,t}$ leads to a 4.5% increase in $SM{E}_{i,t}$ at the sample standard deviation (The detailed calculation of economic effect as follows:$SM{E}_{i,t}=\raisebox{1ex}{$\left(0.004\times 1.469\right)$}\!\left/ \!\raisebox{-1ex}{$0.131$}\right.\approx 4.5\%$, where the standard deviation of $Distanc{e}_{i,t}$ is 1.469 and the sample standard deviation of $SM{E}_{i,t}$ is 0.131).

The coefficients on the control variables generally are consistent with prior literature. For example, $RO{A}_{i,t}$ has a negative relationship with $ETR\_GAA{P}_{i,t}$ and a positive relationship with $SM{E}_{i,t}$, suggesting that firms with higher profitability have more substantial incentives to engage in tax avoidance (e.g., [6,67]). $Tur{n}_{i,t}$ is positively associated with $ETR\_GAA{P}_{i,t}$ and is negatively associated with $SM{E}_{i,t}$, indicating that firms with a lower turnover of assets also have more substantial incentives to engage in more tax avoidance (i.e., [54]). Moreover, $PP{E}_{i,t}$ has a negative relationship with $ETR\_GAA{P}_{i,t}$ and a positive relation with $SM{E}_{i,t}$, suggesting that firms with a higher number of fixed investments are more likely to engage in tax avoidance through accelerated depreciation (i.e., [68]).

4.3. Endogeneity Issues

In this section, we employ several methods to alleviate the potential endogeneity concerns, which could reasonably corroborate our baseline findings. More specifically, we (1) utilize the operating high-speed railway (HSR) station for the first time in the area near the supplier firm as an exogenous shock in the supplier firm’s customer geographic proximity and attempt to mitigate the concern of omitted variables that are unobservable factors may lead to our baseline findings, (2) perform the Heckman two-step method to mitigate the concern of selective bias that whether our baseline findings are primarily affected by sample selection bias, and (3) employ the entropy balancing method to alleviate non-random treatment assignment bias that whether our baseline findings are affected by selection bias due to observables.

4.3.1. Omitted Variables Bias

Our baseline findings are still affected by unobserved factors, although we include a set of firm characteristics and different-level fixed effects to mitigate the impact of potential factors on tax avoidance. In an attempt to alleviate endogeneity, we follow Zhang, Wu, Zhou and Yuan [36] and consider the operating HSR station for the first time nearby the supplier firm as a plausibly exogenous shock to the supplier’s customer geographic distance to evaluate the effect of customer geographic distance on the supplier tax avoidance. The construction of HSR provides convenience among firms–stakeholders from different regions, especially in China (According to Zhang, Wu, Zhou and Yuan [36], China formally completed the first HSR with 350 km/h in 2008 and formed the largest and busiest railway network globally. The network has covered more than 70% of preference-level cities in the whole of China’s land surface and is expected to cover more cities). Since the construction of HSR can directly decrease the travel time and information exchange cost without changing the geographic localities of entities, operating HSR station near the supplier firm should encourage the supplier firm to seek geographically remote major customers. For example, it is more convenient to communicate between the supplier and its major customers’ corporate executives, thus strengthening the supplier–customer private alliance by exchanging inside information of firm–stakeholder. Furthermore, the decision of the HSR station location is generally not affected by the decision of corporate management. Therefore, operating an HSR station for the first time near the supplier firm could provide an exogenously positive shock in the geographic distance between the supplier firm and its major customers (In descriptive statistics by groups, we find an increase in the cross-section variation of geographic distance between the supplier firm and its major customers after the operating HSR station for the first time near the supplier firm: the standard deviation of $Distanc{e}_{i,t}$ increases from 1.447 (in pre-period) to 1.480 (in post-period) in treated sample. The treated sample refers to the supplier firm sample in which the supplier firm has the operating HSR station for the first time near the firm in the whole sample period. Moreover, the standard deviation of $Distance{1}_{i,t}$ ($Distance{2}_{i,t}$) also increases from 1.720 (1.428) to 1.808 (1.448) in the treated sample).

Hence, we could establish a causal relationship between customer geographic proximity and supplier tax avoidance. Following the prior research, we conduct a time-varying difference-in-differences (DID) specification, as set forth below, to run the regression (e.g., [74]).

$$\begin{array}{ll}Tax\_Avoidanc{e}_{i,t}\hfill & \hfill \\ \hfill & ={\beta }_0+{\beta }_{1}Trea{t}_{i,t}+{\beta }_{2}Siz{e}_{i,t}+{\beta }_{3}LE{V}_{i,t}+{\beta }_{4}RO{A}_{i,t}\hfill \\ \hfill & +{\beta }_{5}Growt{h}_{i,t}+{\beta }_{6}Tur{n}_{i,t}+{\beta }_{7}PP{E}_{i,t}+{\beta }_{8}INTA{N}_{i,t}+{\beta }_{9}PI{S}_{i,t}\hfill \\ \hfill & +{\beta }_{10}Los{s}_{i,t}+{\beta }_{11}M{B}_{i,t}+{\beta }_{12}P{C}_{i,t}+{\displaystyle \sum Industry}\hfill \\ \hfill & +{\displaystyle \sum Region+{\displaystyle \sum Year+{\epsilon }_{i,t}}}\hfill \end{array}$$

(3)

We include a set of control variables and industry, region, and year fixed effects as used in Equation (2). Meanwhile, we also cluster standard errors at the firm level to consider the potential with-firm correlation of the error term. The variable of the test is $Trea{t}_{i,t}$, a dummy variable that equals one in the years after operating the HSR station for the first time in the area that is 5 km from the supplier firm i and zero otherwise. To construct $Trea{t}_{i,t}$, we use the HSR station location information from the official website of the China Railway Administration (See at https://www.12306.cn/index/, accessed on 1 January 2022) or Google search, registered address location information for listed firms from CSMAR. We then identify the latitudes and longitudes of the firm–stations pairs on Google Maps and calculate the circle distance between the firm and each station every year. Finally, the HSR station near firm i is operated for the first time when the circle distance between firm i and one of the stations is less than 5 km in year t.

Table 4. Endogeneity issue derived from omitted variables: Time-varying DID.

	ETR_GAAP	SME
	(1)	(2)
Treat	−0.013 *	0.019 ***
	(−1.95)	(2.79)
Controls	Yes	Yes
Industry FE	Yes	Yes
Region FE	Yes	Yes
Year FE	Yes	Yes
Observations	5135	5135
Adjusted R2	0.259	0.200

Notes: This table presents the time-varying DID estimates of the supplier firm’s tax avoidance variables on the customer geographic proximity, including a set of control variables and unreported industry, region, and year fixed effects. The test variable is Treat, a dummy variable that equals one in the years after the operating HSR station for the first time near the supplier firm and zero otherwise. Specifically, we choose the city-fixed effect as the region-fixed effect since the observation unit is the firm–year. The estimation window is 2009–2020. The control variables included following Equation (3). All variable definitions are detailed in Table S1. T-statistics shown in parentheses are based on standard errors adjusted for firm-level clustering. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

presents the results of the time-varying DID regression. In column (1), we regress $ETR\_GAA{P}_{i,t}$ on $Trea{t}_{i,t}$ by including a set of control variables. As expected, the result shows that the effective tax rate significantly declined after operating the HSR station for the first time in the area that is 5 km from the supplier firm, which suggests that the extent of supplier tax avoidance is significantly increased. Similar results are also found in column (2), in which we change the dependent variable as $SM{E}_{i,t}$. These results provide evidence to support a causal effect of the customer geographic proximity on supplier tax avoidance and further alleviate the concern of the omitted variables.

4.3.2. Selective Bias

Our baseline findings might also face the second potential endogeneity issue because the main results are probably affected by sample self-selection bias (Our main finding is based on the sample of supplier’s voluntary disclosure, and we find that supplier firm’s incentive in tax avoidance decrease with customer geographic proximity. The assumption of this argument is that information disclosure may decrease supplier incentives to engage in tax avoidance. Therefore, for the supplier firm, without voluntarily disclosing supply chain information, they may have more space to engage in tax avoidance, which may strengthen our main results). This concern stems from the supplier firm’s voluntary behavior in disclosing the supplier firm’s top five customers’ information. If the supplier firms do not disclose complete information about their top five customers, we cannot calculate the weighted geographic distance from the dataset. Thus, this sample might lead our baseline findings to face a potential sample self-selection bias.

To mitigate this potential endogeneity issue, we follow Yuan and Wen [75] to conduct a Heckman two-step sample selection analysis as follows:

$$\begin{array}{ll}Disclos{e}_{i,t}\hfill & ={\beta }_0+{\beta }_{1}Siz{e}_{i,t}+{\beta }_{2}LE{V}_{i,t}+{\beta }_{3}RO{A}_{i,t}+{\beta }_{4}Growt{h}_{i,t}+{\beta }_{5}Tur{n}_{i,t}+{\beta }_{6}PP{E}_{i,t}+{\beta }_{7}INTA{N}_{i,t}\hfill \\ \hfill & +{\beta }_{8}PI{S}_{i,t}+{\beta }_{9}Los{s}_{i,t}+{\beta }_{10}M{B}_{i,t}+{\beta }_{11}P{C}_{i,t}+{\beta }_{12}{\mathit{Other}\_\mathit{Distance}}_{i,t}+{\displaystyle \sum Industry}\hfill \\ \hfill & +{\displaystyle \sum Region+{\displaystyle \sum Year+{\epsilon }_{i,t}}}\hfill \end{array}$$

(4)

and

$$\begin{array}{ll}Tax\_Avoidanc{e}_{i,t}\hfill & \hfill \\ \hfill & ={\beta }_0+{\beta }_{1}Distanc{e}_{i,t}+{\beta }_{2}Siz{e}_{i,t}+{\beta }_{3}LE{V}_{i,t}+{\beta }_{4}RO{A}_{i,t}+{\beta }_{5}Growt{h}_{i,t}+{\beta }_{6}Tur{n}_{i,t}+{\beta }_{7}PPE\hfill \\ \hfill & +{\beta }_{8}INTA{N}_{i,t}+{\beta }_{9}PI{S}_{i,t}+{\beta }_{10}Los{s}_{i,t}+{\beta }_{11}M{B}_{i,t}+{\beta }_{12}P{C}_{i,t}+{\beta }_{13}IM{R}_{i,t}+{\displaystyle \sum Industry}\hfill \\ \hfill & +{\displaystyle \sum Region+{\displaystyle \sum Year+{\epsilon }_{i,t}}}\hfill \end{array}$$

(5)

where we include the same control variables and different-level fixed effects as used in Equation (2). We estimate a probit model with a binary dummy ($Disclos{e}_{i,t}$) as the dependent variable in the first step analysis and then estimate the value of inverse Mills ratio ($IM{R}_{i,t}$) and use the $IM{R}_{i,t}$ in the second step analysis.

Specifically, $Disclos{e}_{i,t}$ equals one if a supplier firm voluntarily discloses the detailed name of at least one of its major customers and the percentage of the supplier’s sales to that customer and equals zero otherwise. Following Lennox et al. [76], we need to include an important variable that can affect whether the supplier firm discloses its major customers’ complete information. In the first step analysis, $Other\_Distanc{e}_{i,t}$ is defined as the mean value of $Distanc{e}_{i,t}$ in the same supplier’s city in the same year except for the supplier firm, representing an average geodesic distance from the supplier’s location to its customers’ location. Furthermore, $IM{R}_{i,t}$ is estimated from Equation (4) and is added to Equation (5) in the second step analysis to control for the potential sample self-selection bias.

Table 5. Endogeneity issue derived from selective bias: Heckman two-step analysis.

First-Step Regression			Second-Step Regression
	Disclose		ETR_GAAP	SME
	(1)		(2)	(3)
		Distance	−0.004 **	0.004 **
			(−2.48)	(2.23)
Other_Distance	−0.095 ***	IMR	0.013	−0.020
	(−2.71)		(0.32)	(−0.47)
Controls	Yes	Controls	Yes	Yes
Industry FE	Yes	Industry FE	Yes	Yes
Region FE	Yes	Region FE	Yes	Yes
Year FE	Yes	Year FE	Yes	Yes
Observations	16,891	Observations	4443	4443
Pseudo R2	0.206	Adjusted R2	0.244	0.174

Notes: This table presents the estimates using Heckman two-step sample selection analysis. In the first-step regression, the dependent variable Disclose is regressed on the additional variable Other_Distance in a probit model, including a set of control variables and unreported industry, region, and year fixed effects. In the second-step regression, the measures of supplier tax avoidance are regressed on the test variable Distance, and the inverse Mills ratio (IMR) estimated from the first-step analysis, including a set of control variables and unreported industry, region, and year fixed effects. Specifically, we choose the city-fixed effect as the region-fixed effect since the observation unit is the firm–year. The estimation window is 2009–2020. The number of observations presented in this table is different from those in Table 3 because the calculation of Other_Distance generates the missing data. The control variables in column (1) included following Equation (4). The control variables in columns (2) to (3) included following Equation (5). All variable definitions are detailed in Table S1. T-statistics shown in parentheses are based on standard errors adjusted for firm-level clustering. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

presents the results from estimating the Heckman two-step models. In the first-step regression, column (1) shows that $Other\_Distanc{e}_{i,t}$ have a statistically significant negative relationship with a supplier firm decision to voluntarily disclose at least one of the top five customers’ detailed information. These results preliminarily provide evidence for supporting the validity of the Heckman two-step sample selection analysis.

For the second step regression, columns (2) to (3) show that at the 5% significance level, the coefficients of $Distanc{e}_{i,t}$ is still statistically negative when the dependent variable is $ETR\_GAA{P}_{i,t}$, and is statistically positive when the dependent variable is $SM{E}_{i,t}$. Note that $IM{R}_{i,t}$ in columns (2) and (3) is not statistically significant, implying that the unobserved factors affecting supplier firms to disclose the top five customers’ detailed information do not affect the tax avoidance for supplier firms in our sample. Therefore, taking the two-step analysis results together, we find evidence to support our baseline findings and further mitigate the concern of selective bias.

4.3.3. Non-Random Treatment Assignment

Our baseline findings may face non-random treatment assignment caused by covariate imbalance between supplier firms that are in longer and shorter customer geographic distances [77]. To further alleviate the effect of endogeneity bias on our main findings, we use the entropy matching method that has been widely used in accounting and finance research to address non-random treatment assignment bias due to observables. The entropy matching method helps improve causal inferences by providing reweighted data that covariate distributions could satisfy a set of moment conditions [78].

Following Lee et al. [79], we first conduct our entropy balancing analysis by transforming the continuous test variable ($Distanc{e}_{i,t}$) into a binary measure based on the sample median. We then match 11 different covariates and adjust the differences in the first moment of covariate distributions, which are presented in Panel A of

Table 6. Endogeneity issue derived from non-random treatment assignment: Entropy balancing analysis.

Panel A: Differences in Covariates after Entropy Balancing
	Treated (2506 Units)			Control (2629 Units)
	Mean	Variance	Skewness	Mean	Variance	Skewness
Covariates
Size	21.820	1.303	0.692	21.820	1.243	0.590
LEV	0.413	0.043	0.123	0.413	0.045	0.159
ROA	0.052	0.002	1.619	0.052	0.002	1.529
Growth	0.235	0.235	3.009	0.235	0.203	3.031
Turn	0.614	0.159	1.785	0.614	0.181	1.880
PPE	0.224	0.025	0.938	0.224	0.029	0.958
INTAN	0.051	0.003	2.621	0.051	0.003	2.329
PIS	0.065	0.005	1.587	0.065	0.006	1.670
Loss	0.073	0.068	3.271	0.073	0.068	3.271
MB	2.130	1.919	2.511	2.130	1.992	2.412
PC	0.370	0.233	0.537	0.370	0.233	0.537
Panel B: Customer Geographic Proximity and Supplier Tax Avoidance after Entropy Balancing
	ETR_GAAP			SME
	(1)			(2)
Distance	−0.005 ***			0.004 **
	(−2.63)			(2.17)
Controls	Yes			Yes
Industry FE	Yes			Yes
Region FE	Yes			Yes
Year FE	Yes			Yes
Observations	5135			5135
Adjusted R2	0.261			0.199

Notes: This table presents the results of our entropy balancing analysis. We divide the sample into treated and control groups based on the test variable ($Distanc{e}_{i,t}$) sample median. We then match on 11 different covariates, which are presented in Panel A. Panel B presents the estimates of the supplier firm’s tax avoidance variables on the customer geographic proximity using reweighted sample, including a set of control variables and unreported industry, region, and year-fixed effects. Specifically, we choose the city-fixed effect as the region-fixed effect since the observation unit is the firm–year. The estimation window is 2009–2020. The control variables included following Equation (2). All variable definitions are detailed in Table S1. T-statistics shown in parentheses are based on standard errors adjusted for firm-level clustering. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

. Using the reweighted sample, we re-estimate our baseline regressions as Equation (2).

Panel B of Table 6 presents the estimation results of our entropy balancing analysis. Indeed, the results continue to show that the coefficients of $Distanc{e}_{i,t}$ is still statistically negative when the dependent variable is $ETR\_GAA{P}_{i,t}$, and is statistically positive when the dependent variable is $SM{E}_{i,t}$. Hence, the entropy balancing method confirms the robustness of our baseline findings.

5. Potential Channels

Although the above baseline findings are consistent with the argument that the supplier firm’s customer geographic proximity restrains its tax avoidance activities, the potential channels need to be further explored. In this section, we examine how the geographic distance between a supplier and its major customers works through the two channels to help exacerbate supplier tax avoidance, namely information asymmetry and detection risk.

5.1. The Information Asymmetry Channel

Based on the main findings, we believe that the customer geographic distance plays a vital role by increasing the supplier’s information opacity with its external stakeholders and strengthening the information advantages of the supplier–customers relationship, thus enhancing the information asymmetry between supplier firms and other outside stakeholders. Therefore, the greater customer geographic distance reduces the cost of supplier tax avoidance and promotes managers’ incentives to avoid tax aggressively via the supply chain.

The activities of corporate tax avoidance tend to need opacity that caters to the complexity of tax planning [3,80,81], and the lower information transparency provides the firm more opportunity to evade outside stakeholders’ information search. Meanwhile, the process of tax planning requires good information flow across different internal departments or external related parties and thus executes the firm’s tax planning strategy with efficiency [7]. In particular, Cen, Maydew, Zhang and Zuo [6] point out that compared to other outside stakeholders, information communication between the supplier and its major customers has lower costs in exchanging and implementing related tax planning strategies’ information due to the strategic alliance with relationship-specific investments.

Hence, we predict that if the positive effect of customer geographic distance on supplier tax avoidance is driven by the strategic alliance’s efficient interchange of corporate information, this effect should be weaker (stronger) for firms with lower (higher) information advantages in supplier–customers relationships.

To test whether the empirical evidence might support the information asymmetry channel, with data from CSMAR, we adopt the analysis by adding the interaction term between customer geographic distance and information asymmetry channel-related factor in our baseline regression to test this prediction.

$$\begin{array}{ll}Tax\_Avoidanc{e}_{i,t}\hfill & \hfill \\ \hfill & ={\beta }_0+{\beta }_{1}Distanc{e}_{i,t}+{\beta }_{2}Distanc{e}_{i,t}\times DISP\_D_{i,t}+{\beta }_{3}DISP\_D_{i,t}+{\beta }_{4}Siz{e}_{i,t}+{\beta }_{5}LE{V}_{i,t}\hfill \\ \hfill & +{\beta }_{6}RO{A}_{i,t}+{\beta }_{7}Growt{h}_{i,t}+{\beta }_{8}Tur{n}_{i,t}+{\beta }_{9}PP{E}_{i,t}+{\beta }_{10}INTA{N}_{i,t}+{\beta }_{11}PI{S}_{i,t}+{\beta }_{12}Los{s}_{i,t}\hfill \\ \hfill & +{\beta }_{13}M{B}_{i,t}+{\beta }_{14}P{C}_{i,t}+{\displaystyle \sum Industry}+{\displaystyle \sum Region+{\displaystyle \sum Year+{\epsilon }_{i,t}}}\hfill \end{array}$$

(6)

where $DISP\_D_{i,t}$ represents a channel factor for the supplier firm i with either high or low information asymmetry in year t, and $DISP\_D_{i,t}$ equals one if the dispersion of analysts’ earnings per share forecasts about a supplier firm i is above the sample median in the same year t and zero otherwise. Specifically, we follow Gao and Wang [82] to calculate the dispersion of analysts’ earnings per share forecasts about a supplier firm, measured as the standard deviation of analysts’ forecasts divided by the absolute value of the mean of analysts’ forecasts. The higher the value of analyst forecast dispersion, the larger the information advantages between the supplier and its major customers, which are produced by efficient information dissemination in the strategic alliance. Other outside stakeholders, such as financial analysts, are difficult to obtain more information about the detailed supplier firm’s transaction business by on-site investigation because of the greater customer geographic distance.

Table 7. The information asymmetry channel.

	ETR_GAAP	SME
	(1)	(2)
Distance	−0.001	0.002
	(−0.58)	(1.23)
Distance × DISP_D	−0.006 ***	0.005 **
	(−2.70)	(2.08)
DISP_D	0.030 ***	−0.027 ***
	(3.75)	(−3.38)
Controls	Yes	Yes
Industry FE	Yes	Yes
Region FE	Yes	Yes
Year FE	Yes	Yes
Observations	4414	4414
Adjusted R2	0.276	0.204

Notes: This table presents the regression results for the effect of customer geographic proximity on supplier tax avoidance conditional on information asymmetry. Specifically, we use an indicator variable based on analyst forecast dispersion (DISP_D) to measure the firm-level information asymmetry. We choose the city-fixed effect as the region-fixed effect since the observation unit is the firm–year. The estimation window is 2009–2020. The number of observations presented in this table differs from those with similar model specifications in Table 3 because the firm-level analyst forecast dispersion data exist missing data. The control variables included following Equation (6). All variable definitions are detailed in Table S1. T-statistics shown in parentheses are based on standard errors adjusted for firm-level clustering. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

reports the results of the channel analysis conducted by adding the interaction term. In columns (1)–(2), the coefficient on the interaction of $Distanc{e}_{i,t}$ and $DISP\_D_{i,t}$ is significantly negative (t-stat = −2.70) for $ETR\_GAA{P}_{i,t}$ and significantly positive (t-stat = 2.08) for $SM{E}_{i,t}$. These results provide evidence supporting the information asymmetry channel, which suggests that the positive effect of customer geographic distance on supplier tax avoidance is stronger for supplier firms with higher information asymmetry between the supplier firm and other outside stakeholders.

5.2. The Detection Risk Channel

According to the Taxes Collection Act in China, firms must accept supervision and monitoring from the government department. The firms, which have the incentive to avoid tax aggressively, need to face the detection risk from tax authorities (see Tang [14] for a recent review). In order to meet the annual tax collection budget, tax authorities have both impetus and the ability to detect any suspected tax avoidance behavior by verifying and auditing the related materials about firms’ operations on site. Therefore, the firms could feel threatened in conducting tax avoidance strategies, which harm the government tax revenue [83]. Lin, Mills, Zhang and Li [71] find evidence that the corporate effective tax rate increases with the strength of local tax enforcement, indicating that tax enforcement plays an important external governance role in curbing a firm’s tax avoidance activities.

Because of concerns about detection risk, the firm needs more efficient tax avoidance strategies. Specifically, one of the main corporate tax avoidance mechanisms could be via the supply chain [6]. Moreover, we contend that the greater customer geographic distance could raise on-site detection costs by tax authorities, especially from the costs in detecting actual trade with related customers [13], reducing the detection risk faced by the supplier firm. Supplier firms’ fear of being detected affects the extent to which they seek to profit from tax avoidance by relying on the geographic localities of the supplier–customer relationship. Specifically, the profits associated with a supplier firm’s tax avoidance strategies through the greater customer geographic distance are higher when local tax enforcement power is stronger. The geographically remote customers can provide the supplier with more flexible tax avoidance strategies due to the lower detection risk faced by the supplier. In summary, to the extent that customer geographic distance help, we expect the effect of customer geographic distance on supplier tax avoidance should be stronger (weaker) in firms with higher (lower) detection risk.

To test whether the empirical evidence might support the detection risk channel, with data from the National Bureau of Statistics of China, we also adopt the analysis by adding the interaction term between customer geographic distance and detection risk channel-related factor in our baseline regression to test this prediction.

$$\begin{array}{ll}Tax\_Avoidanc{e}_{i,t}\hfill & \hfill \\ \hfill & ={\beta }_0+{\beta }_{1}Distanc{e}_{i,t}+{\beta }_{2}Distanc{e}_{i,t}\times TE\_D_{i,k,t}+{\beta }_{3}TE\_D_{i,k,t}+{\beta }_{4}Siz{e}_{i,t}+{\beta }_{5}LE{V}_{i,t}\hfill \\ \hfill & +{\beta }_{6}RO{A}_{i,t}+{\beta }_{7}Growt{h}_{i,t}+{\beta }_{8}Tur{n}_{i,t}+{\beta }_{9}PP{E}_{i,t}+{\beta }_{10}INTA{N}_{i,t}+{\beta }_{11}PI{S}_{i,t}+{\beta }_{12}Los{s}_{i,t}\hfill \\ \hfill & +{\beta }_{13}M{B}_{i,t}+{\beta }_{14}P{C}_{i,t}+{\displaystyle \sum Industry}+{\displaystyle \sum Region+{\displaystyle \sum Year+{\epsilon }_{i,t}}}\hfill \end{array}$$

(7)

To measure the detection risk, we construct an indicator variable $TE\_D_{i,k,t}$ proxying for the supplier firm i with either high or low detection risk in province k in year t. $TE\_D_{i,k,t}$ equals one if regional tax enforcement efforts in a supplier firm i’s province k is above the sample median in the same year t and zero otherwise. Specifically, we follow Xu et al. [83] to calculate the tax enforcement efforts in the supplier firm’s region, measured as the actual tax revenue ratio divided by the estimated value of tax revenue ratio (We run a multi-regression model for province-year panel data by regressing the percentage of actual tax revenue to GDP ($AC\_ta{x}_{k,t}$) in province k in year t on the first industry’s proportion to GDP ($IND\_1_{k,t}$) and the second industry’s proportion to GDP ($IND\_2_{k,t}$) and the sum of import and export proportion to GDP ($Opennes{s}_{k,t}$). Moreover, all province-level data exclude the impact of inflation by using GDP deflator or price index based on 2010. Finally, we obtain 31 provinces’ tax enforcement efforts yearly data according to the traditional concept of tax enforcement efforts). The higher the value of tax enforcement efforts, the larger the detection risk faced by the supplier firm.

Table 8. The detection risk channel.

	ETR_GAAP	SME	ETR_GAAP	SME
	(1)	(2)	(3)	(4)
Distance	−0.002	0.001	−0.004 ***	0.004 **
	(−0.95)	(0.64)	(−2.61)	(2.29)
Distance × TE_D	−0.005 *	0.006 **
	(−1.81)	(2.15)
TE_D	0.019 *	−0.021 **
	(1.96)	(−2.15)
Controls	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes
Region FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Observations	5135	5135	4550	4550
Adjusted R2	0.263	0.201	0.262	0.195

Notes: This table presents the regression results for the effect of customer geographic proximity on supplier tax avoidance conditional on external corporate governance. Specifically, we use an indicator variable based on local tax enforcement efforts (TE_D) to measure the province-level governance characteristics in columns (1)–(2). Columns (3)–(4) further report the regression results of model in Equation (2), excluding suppliers whose office addresses relocated closer to customers, but whose registered addresses remain unchanged, which aim at addressing the potential concern that the baseline results are driven by customer monitoring. We choose the city-fixed effect as the region-fixed effect since the observation unit is the firm–year. The estimation window is 2009–2020. The control variables included following Equation (7). All variable definitions are detailed in Table S1. T-statistics shown in parentheses are based on standard errors adjusted for firm-level clustering. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

also reports the results of the channel analysis conducted by adding the interaction term. In columns (1) to (2), the coefficients on the interaction of $Distanc{e}_{i,t}$ and $TE\_D_{i,k,t}$ are significantly negative and positive (t-stat = −1.81 and 2.15), respectively. These results provide some evidence supporting the detection risk channel, which suggests that the positive effect of customer geographic distance on supplier tax avoidance is stronger for supplier firms facing higher detection risk by local tax authorities.

However, major customers may also have the incentive to monitor supplier tax avoidance to maintain a stable and reliable relationship with the supplier [6], which results in a positive association between major customer geographic distance and supplier tax avoidance rather than outside stakeholders monitoring channel. To rule out this alternative interpretation, we adopt an identification strategy to verify our main results using differences in functions based on supplier office address and registered address. Take into account that the supervision of communication between customers and suppliers will only take place at the supplier’s office address, while the concerns of external stakeholders related to the supplier’s tax activities will be more focused on the supplier’s registered address. We exclude the observations in which suppliers whose office addresses relocate closer to customers’ office addresses but whose registered addresses remain unchanged. Columns (3)–(4) in Table 8 show that baseline results still hold when we restrict our analysis to a subsample in which the customer monitoring power can be clearly identified as weaker.

6. Moderating Effects

Thus far, we find that the supplier firm’s weighted geographic distance from the major customers positively affects the supplier tax avoidance, and the relation is robust after controlling for potential endogeneity concerns. Moreover, we investigate the potential channels through which customer geographic proximity affects supplier tax avoidance. In further analysis, to further explore a supplier’s tax avoidance decision when facing geographically proximate customers, we examine how (1) corporate financial risk, (2) industrial regulation, and (3) the marketization environment may influence the relationship between the customer geographic proximity and the supplier tax avoidance.

6.1. Financial Risk

We first test the H2a, which is the moderate effect of corporate financial risk on the relation between the supplier firm’s customer geographic proximity and its corporate tax avoidance. We add the dummy variable $LEV\_D_{i,t}$ and the interaction of $Distanc{e}_{i,t}$ and $LEV\_D_{i,t}$ into Equation (2). $LEV\_D_{i,t}$, equals one if a supplier firm with higher financial leverage in the same industry in the same year and zero otherwise.

As shown in

Table 9. The moderate effect of the corporate financial risk.

	ETR_GAAP	SME
	(1)	(2)
Distance	−0.001	0.000
	(−0.43)	(0.05)
Distance × LEV_D	−0.004	0.005 *
	(−1.56)	(1.82)
LEV_D	0.011	−0.014
	(1.17)	(−1.39)
Controls	Yes	Yes
Industry FE	Yes	Yes
Region FE	Yes	Yes
Year FE	Yes	Yes
Observations	5135	5135
Adjusted R2	0.260	0.199
Marginal effects of financial risk
Low financial risk	−0.001	0.000
	(−0.43)	(0.05)
High financial risk	−0.005 ***	0.005 ***
	(−2.79)	(2.68)

Notes: This table presents a regression of supplier firm tax avoidance on the supplier firm’s geographic distance from its top five customers (Distance), along with its interaction with an indicator for a firm with higher corporate financial risk (LEV_D). We choose the city-fixed effect as the region-fixed effect since the observation unit is the firm–year. The estimation window is 2009–2020. The control variables included following Equation (2). All variable definitions are detailed in Table S1. Given by $\frac{\partial Tax\_avoidanc{e}_{i,t}}{\partial Distanc{e}_{i,t}}={\beta }_1+{\beta }_2\times LEV\_D_{i,t}$, the marginal effects use corresponding coefficient estimates from OLS specifications as Equation (2). T-statistics shown in parentheses are based on standard errors adjusted for firm-level clustering. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

, we generally find supportive evidence for our conjecture. In column (2), the coefficient on the interaction of $Distanc{e}_{i,t}$ and $LEV\_D_{i,t}$ is significantly positive (t-stat = 1.82). This result indicates that the positive effect of the supplier firm’s geographic distance from its major customers on its corporate tax avoidance is greater for supplier firms with a high financial risk. However, the coefficients on the interaction of $Distanc{e}_{i,t}$ and $LEV\_D_{i,t}$ are insignificant (t-stat = −1.56) for $ETR\_GAA{P}_{i,t}$ in column (1). Consistent with the main results above, the bottom panel of Table 9 also presents the marginal effects of customer geographic proximity on supplier tax avoidance when the supplier firm is at low and high financial risk. For example, if the supplier faces a low financial risk, the marginal effect of customer geographic distance on the supplier’s effective tax rate ($ETR\_GAA{P}_{i,t}$) decreases by 0.001 percentage points, but it is not statistically significant; if the supplier faces a high financial risk, the marginal effect of customer geographic distance on the supplier’s effective tax rate ($ETR\_GAA{P}_{i,t}$) significantly decreases by 0.005 percentage points. This result is corroborated by the downward-sloping marginal effect of financial risk shown in Figure S1. Overall, the results provide some evidence to support that the positive effect of customer geographic distance on supplier tax avoidance is more pronounced when the supplier firms with higher financial risk.

6.2. Industry Regulation

We next test H2b, which predicts that the relationship between customer geographic distance and supplier tax avoidance is weakened by regulatory industries. Following Ke et al. [84] regulated industries list, we construct an indicator ($C{I}_{i,l,t}$) that equals one if the supplier firm i is in a competitive industry l in year t and equals zero otherwise. We add the dummy variable $C{I}_{i,l,t}$ and the interaction of $Distanc{e}_{i,t}$ and $C{I}_{i,l,t}$ into Equation (2).

Table 10. The moderate effect of the regulation of the industry.

	ETR_GAAP	SME
	(1)	(2)
Distance	0.001	−0.001
	(0.25)	(−0.43)
Distance × CI	−0.007 **	0.007 **
	(−1.99)	(2.06)
CI	0.011	−0.016
	(0.75)	(−1.07)
Controls	Yes	Yes
Industry FE	Yes	Yes
Region FE	Yes	Yes
Year FE	Yes	Yes
Observations	5135	5135
Adjusted R2	0.261	0.200
Marginal effects of industry regulation
Regulatory industry	0.001	−0.001
	(0.25)	(−0.43)
Competitive industry	−0.006 ***	0.006 ***
	(−3.32)	(3.10)

Notes: This table presents a regression of supplier firm tax avoidance on the supplier firm’s geographic distance from its top five customers (Distance), along with its interaction with an indicator for a firm in the competitive industries (CI). We choose the city-fixed effect as the region-fixed effect since the observation unit is the firm–year. The estimation window is 2009–2020. The control variables included following Equation (2). All variable definitions are detailed in Table S1. Given by $\frac{\partial Tax\_avoidanc{e}_{i,t}}{\partial Distanc{e}_{i,t}}={\beta }_1+{\beta }_2\times C{I}_{i,l,t}$, the marginal effects use corresponding coefficient estimates from OLS specifications as Equation (2). T-statistics shown in parentheses are based on standard errors adjusted for firm-level clustering. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

shows the results of the moderate effect of the regulatory industry. In column (1), the coefficient on the interaction of $Distanc{e}_{i,t}$ and $C{I}_{i,l,t}$ is significantly negative (t-stat = −1.99). Moreover, the coefficients on the interaction of $Distanc{e}_{i,t}$ and $C{I}_{i,l,t}$ are significantly positive (t-stat = 2.06) in column (2). Consistent with the main results above, the bottom panel of Table 10 also presents the marginal effects of customer geographic proximity on supplier tax avoidance when the supplier firm is in a regulatory and competitive industry. For example, if the supplier is in a regulatory industry, the marginal effect of customer geographic distance on the supplier’s effective tax rate ($ETR\_GAA{P}_{i,t}$) increases by 0.001 percentage points, but it is not statistically significant; if the supplier is in a competitive industry, the marginal effect of customer geographic distance on the supplier’s effective tax rate ($ETR\_GAA{P}_{i,t}$) significantly decreases by 0.006 percentage points. This result is corroborated by the downward-sloping marginal effect of industry regulation shown in Figure S2. These results favorably confirm that the relation between the supplier firm’s geographic distance from its major customers and its tax avoidance is more pronounced in competitive industries than in regulatory industries.

6.3. Marketization Environment

We finally examine the H2c, which conjectures that the relationship between customer geographic distance and supplier tax avoidance is strengthened by weak marketization. Following the prior literature (e.g., [50,54,85]), we obtain the provincial-level marketization index from the NERI (More detailed information about Marketization Index could see relevant papers, such as Fan, Wang and Zhu [49], Shen, Gao, Bu, Yan and Chen [50] and so forth) [49]. A higher marketization index means a higher marketization process. We use these data to set an indicator variable $Marke{t}_{i,k,t}$, which equals one if the index value of the province k where the supplier firm i is located is higher than the sample median in the same year t and zero otherwise. We then add the dummy variable $Marke{t}_{i,k,t}$ and the interaction of $Distanc{e}_{i,t}$ and $Marke{t}_{i,k,t}$ into Equation (2).

Table 11. The moderate effect of the marketization environment.

	ETR_GAAP	SME
	(1)	(2)
Distance	−0.010 ***	0.012 ***
	(−3.53)	(4.37)
Distance × Market	0.007 **	−0.011 ***
	(2.27)	(−3.43)
Market	−0.026 **	0.038 ***
	(−2.48)	(3.57)
Controls	Yes	Yes
Industry FE	Yes	Yes
Region FE	Yes	Yes
Year FE	Yes	Yes
Observations	5135	5135
Adjusted R2	0.261	0.202
Marginal effects of marketization environment
Low marketization	−0.010 ***	0.012 ***
	(−3.53)	(4.37)
High marketization	−0.003	0.001
	(−1.40)	(0.76)

Notes: This table presents a regression of supplier firm tax avoidance on the supplier firm’s geographic distance from its top five customers (Distance), along with its interaction with an indicator for a firm in the areas with higher marketization (Market). We choose the city-fixed effect as the region-fixed effect since the observation unit is the firm–year. The estimation window is 2009–2020. The control variables included following Equation (2). All variable definitions are detailed in Table S1. Given by $\frac{\partial Tax\_avoidanc{e}_{i,t}}{\partial Distanc{e}_{i,t}}={\beta }_1+{\beta }_2\times Marke{t}_{i,k,t}$, the marginal effects use corresponding coefficient estimates from OLS specifications as Equation (2). T-statistics shown in parentheses are based on standard errors adjusted for firm-level clustering. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

shows the results of the moderate effect of the marketization environment. In column (1), the coefficient on the interaction of $Distanc{e}_{i,t}$ and $Marke{t}_{i,k,t}$ is significantly positive (t-stat = 2.27) for the effective tax rate ($ETR\_GAA{P}_{i,t}$). In column (2), the coefficients on the interaction of $Distanc{e}_{i,t}$ and $Marke{t}_{i,k,t}$ is significantly negative (t-stat = −3.43) for $SM{E}_{i,t}$. Consistent with the main results above, the bottom panel of Table 11 also presents the marginal effects of customer geographic proximity on supplier tax avoidance when the supplier firm is in low and high marketization environments. For example, if the supplier is in a low marketization environment, the marginal effect of customer geographic distance on the supplier’s effective tax rate ($ETR\_GAA{P}_{i,t}$) significantly decreases by 0.010 percentage points; if the supplier is in a high marketization environment, the marginal effect of customer geographic distance on the supplier’s effective tax rate ($ETR\_GAA{P}_{i,t}$) decreases by 0.003 percentage points, but it is not statistically significant. This result is corroborated by the upward-sloping marginal effect of industry regulation shown in Figure S3. In summary, these results show that the positive effect of the supplier firm’s geographic distance from its major customers on its corporate tax avoidance is greater for supplier firms with a weaker degree of marketization environments.

7. Additional Tests

7.1. Alternative Measures

Following the prior literature, we also use $ETR\_CU{R}_{i,t}$ as an alternative effective tax rate to measure the current level of a supplier firm’s tax avoidance (e.g., [6,71]). Our baseline findings are partly based on annual $SM{E}_{i,t}$, which may not well capture corporate tax avoidance for specific firms, especially those with tax disputes with local tax authorities for a long time [58]. Therefore, we use the adjusted difference between the statutory tax rate and the effective tax rate ($MSM{E}_{i,t}$) as an alternative measure of $SM{E}_{i,t}$, which is calculated by the average of $SM{E}_{i,t}$ in the last three years and thus captures long-run corporate tax avoidance. In addition, we use $BT{D}_{i,t}$, the difference between the book and taxable income in year t scaled by total assets to capture the gap between book accounting income reported to the shareholder and taxable income reported to the tax authorities. Finally, we further use the movements in book–tax differences unexplained by total accruals ($DDBT{D}_{i,t}$), an alternative measure for book–tax differences ($BT{D}_{i,t}$), which captures abnormal book–tax differences after controlling for earning management.

Among the alternative measures of the supplier firm’s customer geographic proximity, $Distance{1}_{i,t}$ captures the supplier firm’s weighted geographic distance from the top five customers based on the weight of the ratio of the supplier firm’s sales to one of the top five customers concerning the sum of these sales to the major customers; $Distance{2}_{i,t}$ further removes the effect of different supplier’s sales to its customers by considering the simple arithmetic average of the supplier firm’s distance from the major customers.

We adopt these tax avoidance or customer geographic proximity alternative measures based on our baseline model for a further robust check. Table S3 reports the results of changing the measures. In column (1), the coefficient on customer-weighted geographic distance ($Distanc{e}_{i,t}$) is negative and statistically significant. We also find that the coefficients on $Distanc{e}_{i,t}$ are both positive and statistically significant from columns (2) to (4). These results indicate that the enhancing effect of customer geographic proximity on supplier tax burden still holds.

In columns (5) to (6), the coefficient on $Distance{1}_{i,t}$ is significantly negative when the dependent variable is $ETR\_GAA{P}_{i,t}$ but is insignificant when the dependent variable is $SM{E}_{i,t}$, providing additional evidence for the robustness of our baseline findings. Similar but strong evidence is presented in columns (7) to (8), in which we regress $ETR\_GAA{P}_{i,t}$ and $SM{E}_{i,t}$ on $Distance{2}_{i,t}$ (coefficients = −0.004 and 0.003; t-stat = −2.43 and 1.58).

7.2. Changing the Standard Error Estimation Method

We then check the robustness of our baseline findings by changing the methods of standard error. To control for heteroskedasticity, we employ the robust supplier firm-level clustering standard error method in our baseline model. Considering the selective bias of standard error estimation methods, we use the cluster standard errors by robust industry level or robust prefecture city level instead of supplier firm level.

We present the results in Table S4. Columns (1) to (2) show the results of clustering standard errors at the industry level, and columns (3) to (4) show the results of clustering standard errors at the prefecture city level. In summary, our baseline findings still hold after changing the methods of standard error estimation.

7.3. Controlling for the Interaction of Industry and Year-Fixed Effect

Although we add industry, region, and year-fixed effects to control the potential omitted factors of tax avoidance, our baseline findings may still be affected by unobserved factors derived from differential industry-specific time effects (e.g., some high-tech industries enjoy tax exemptions beginning in a specific year). To address this concern, we follow Dhaliwal et al. [86] and Wen, Cui and Ke [85] and examine the effect within the same industry and the same year by including industry and year-fixed effect interactions.

The results are shown in Table S5. The supplier firm’s weighted geographic distance from the major customers is still positively associated with supplier corporate tax avoidance. Therefore, our main baseline findings are robust to including industrial time-varying effects.

7.4. The Effect of the Financial Crisis

We still need to check the robustness of our baseline findings by considering the effect of the global financial crisis. This concern is because our sample period starts from 2009 to 2020 and thus includes firms that suffered from the global financial crisis beginning in 2007.

Regardless of the aspects considered, the financial crisis had a catastrophic impact on most firms. The financial crisis might have especially exerted a long-term impact on corporate tax avoidance by reducing the firms’ profits and increasing their incentives to engage in tax avoidance. Based on that, we want to know whether our main findings still exist after controlling for the effect of the financial crisis. Following An and Zhang [87] and Ye et al. [88], we define a dummy variable $Crisi{s}_{i,t}$ as a proxy for the financial crisis, which equals one when the sample firm i existed in the crisis period (2007 or 2008) and zero otherwise.

The results are shown in Table S6 from columns (1) to (2). Consistent with our baseline findings, the results show that the customer geographic proximity attenuates supplier tax avoidance after considering the global financial crisis, indicating the robustness of our baseline findings.

7.5. Tax Reform during the Sample Period

During our sample period, China completed vital tax reform across the board, replacing the business tax with a value-added tax (RBTVT) in 2016. Improving the value-added tax (VAT) deduction chain may change some tax behaviors of the firm. For example, Lin [89] points out that VAT is less likely to be used by a firm to engage in tax avoidance than business tax because of self-policing. On the other hand, RBTVT could reduce the firm’s effective tax rate, leading to less taxation burden [90]. Given this institutional background, we exclude the sample observations from 2016 for a robustness check.

As Table S7 shows, the supplier firm’s weighted geographic distance from the major customers is still positively associated with supplier tax avoidance. Therefore, these results further corroborate that our baseline findings are robust.

7.6. Considering the Effect of Customer Concentration

Prior studies show that customer concentration may play an important role in affecting supplier firms’ corporate behaviors. Customer concentration indicates higher bargaining power from the customer, which increases the supplier’s risk, thus resulting in a higher cost of equity [86], hurting the supplier firm’s profitability [91], and even increasing the supplier’s incentives in tax avoidance [5]. To mitigate the effect of customer concentration on the association between customer geographic proximity and supplier’s tax avoidance, we include customer concentration as a control variable in Equation (2). We define customer concentration as the Herfindahl–Hirschman Index of sales to the top five customers.

As shown in Table S8, the supplier firm’s weighted geographic distance from the major customers is still positively associated with supplier tax avoidance, and the coefficients on $CC\_HH{I}_{i,t}$ are both insignificant in columns (1)–(2), which indicates that our main results are robust to the inclusion of customer concentration in our baseline model.

8. Conclusions

Despite increasing awareness about the effect of the supplier–customer relationship on supplier tax avoidance, geography has mostly been ignored in the literature on tax avoidance via the supply chain. Our paper uses a sample of listed Chinese firms and their top five customers’ information to investigate the impact of customer geographic proximity on the supplier’s tax avoidance. We find that the geographic distance between a supplier and its major customers has a positive and significant effect on its tax avoidance. Our findings are robust after considering endogeneity concerns by examining an exogenous variation in customer geographic distance following the operation of HSR in the near supplier firm for the first time, performing the Heckman two-step method to mitigate the concern of selective bias, and employing the entropy balancing method to alleviate non-random treatment assignment bias due to observables. Our findings are still consistent in additional robust tests.

We also find support for the two potential mechanisms that underline the impact of the customer geographic proximity on supplier tax avoidance: the information asymmetry channel, by which the farther customer geographic distance from the supplier aggravates the suppliers’ information asymmetry between the supplier firm and other outside stakeholders that are positively related to supplier tax avoidance; and the detection risk channel, by which the farther customer geographic distance from the supplier lessens the detection risk for the supplier and is thereby positively related to supplier tax avoidance. We also provide evidence that filters out the potential concern of our baseline results driven by customer monitoring. Furthermore, our results suggest that this effect of customer geographic proximity on supplier tax avoidance is stronger in supplier firms with high financial risk, supplier firms in a competitive industry, and supplier firms in an operating area with weak marketization environments.

Overall, our study sheds light on the role of customer geographic proximity in affecting the supplier’s tax avoidance decisions. From the impact of a firm’s characteristics on corporate tax planning, our findings provide a shred of clear evidence to better understand the debate on the pros and cons of the customer geographic localities within-country for the supplier.

8.1. Theoretical Contributions and Practical Implications

This study has implications for the theoretical literature and management. First, our study adds to the growing literature on related stakeholders and tax avoidance. The existing evidence shows that firms’ relationship with auditors and analysts can affect their tax avoidance [2,3]. We supplement prior studies on tax avoidance by identifying a new factor—supply chain—especially customer geographic distance, which has an incremental effect on tax avoidance. Second, we extend the studies on the effect of the supplier–customer relationship on the corporate decision. Much research has explored the characteristics of customers on a supplier’s corporate governance, such as customer concentration, but few focus on the impact of geographic location [8,23,24]. We complement the existing research by focusing on the effect of supplier–customer geographic proximity on tax avoidance, which is highly influenced by corporate information transparency and agency problems [3,25,26]. Third, we also contribute to the literature about the benefits of information spillover of a firm’s supply chain to the information environment. Specifically, we explored how outside stakeholders use the firm’s supply chain information because the reduced distance to customers affects the supplier’s tax avoidance decisions.

The practical implications of this paper are mainly manifested in two aspects. The first concerns management decision face a trade-off when choosing the remote customer. Our empirical research evidence shows that geographic proximity to customers may improve the information environment as well as higher tax detection risk, which provides implications for management. The second concerns the policymakers, such as tax regulation. The supplier may use information advantage for the customer over tax jurisdiction to engage in tax avoidance. Therefore, policymakers should pay attention to tax avoidance through the supply chain.

8.2. Limitations and Future Research Directions

This study also has some limitations. First, we only investigate the customer characteristics of geographical location, which might only account for part of the reason for tax avoidance. Further studies could consider other customer characteristics to test the influencing factors comprehensively. Second, this study only investigates the relationship between the geographical proximity of major customers and supplier tax avoidance from the perspective of information transfer and tax detection risk. Further study could combine the needs of the customer and the motivation of tax avoidance to explore the complex channels further. Third, we only focus on listed Chinese firms due to the data limitation. We suggest the need for further study to explore the relationship globally if the data allows.