State ownership, performance evaluation and tax avoidance

ABSTRACT

In 2004, the central and local State-owned Assets Supervision and Administration Commission (SASAC) announced how to evaluate the performance of management of state-owned enterprises (SOEs) that are under its control. This article examines how the different performance evaluation systems in different provinces influence the tax avoidance behaviour of SOEs. We find that only four of them – Shanghai, Guangdong, Zhejiang and Fujian – adopt after-tax profit as the evaluation index, while all the other provinces use before-tax profit. We believe that SOEs which use after-tax profit as an evaluation index have higher incentives to reduce the tax burden than those which use before-tax profit as an evaluation index. Following previous literature, two measurements of tax avoidance are used in this paper: effective tax rate and book-tax difference. Based on 8426 observations from 2007 to 2015, we have two main conclusions: (1) compared with SOEs that are evaluated using before-tax profit as the evaluation index, SOEs that are evaluated using after-tax profit have lower effective tax rate (ETR) and higher book-tax difference (BTD); (2) compared with SOEs that are evaluated using before-tax profit as the evaluation index, SOEs that are evaluated using after-tax profit have a higher level of substitution between debt financing and tax avoidance.

KEYWORDS:

• nature of control
• performance evaluation
• tax avoidance

1. Introduction

This article examines how the performance evaluation system of state-owned enterprises (SOEs) in different provinces influences tax avoidance in SOEs. Firms take tax avoidance activities to reduce their actual tax burden, reduce cash flow out and increase firm value. Prior literature has examined various determinants of tax avoidance, for example, the structure of ownership (Liu & Li, 2012; Liu & Ye, 2014; Wang, Wang, & Peng, 2010, 2012; Wu, 2009; Zhang & Lyu, 2011; Zhu & Yu, 2010), the political connection of firms (Li & Xu, 2013; Wu, 2009), internal control (Li & Chen, 2012), local economic development and tax regulations (Fan & Tian, 2013; Liu & Ye, 2014; Zhu & Sun, 2014), and the power of top executives (Xue, Du, & Hong, 2012). However, few papers have focused on how the incentive motivations of top executives, especially the performance evaluation measures, could affect the tax avoidance behaviour. This article contributes to filling that gap.

This paper is based on the performance evaluation systems for top executives in SOEs which are under the control of the central and local State-owned Assets Supervision and Administration Commission (SASAC). Our focus is on how different performance evaluation measures used in different provinces could lead to different tax avoidance results. Although most of the local SASACs choose before-tax profit as the major evaluation index, the SASACs of Shanghai, Guangdong, Fujian and Zhejiang provinces use after-tax profits.¹ We argue that CEOs who are evaluated based on after-tax profit have higher incentives to carry out tax avoidance activities and increase after-tax profits. Consistent with this argument, the empirical tests show that SOEs in Shanghai, Guangdong, Fujian and Zhejiang provinces have a lower effective tax rate (ETR) and higher book-tax difference (BTD).

We make contributions to the literature by first testing the causal relationship between the performance evaluation system of SOEs and tax avoidance decisions. We provide a new explanation for the high tax rate of SOEs: CEOs in most of the SOEs in China are evaluated on before-tax profit, which reduces their personal motivations to carry out tax avoidance activities. As a result, compared with external governance mechanisms, such as reinforcing tax regulations, internal mechanisms through the performance evaluation system can also effectively change firms’ tax behaviour and risk.

This paper is arranged as follows. The next section is the theoretical analysis and hypothesis development. The third section illustrates the sample selection and empirical design. Section 4 shows the empirical results and analysis. The final section is the conclusion.

2. Literature review and hypotheses development

2.1. Literature review

2.1.1. Corporate governance and tax avoidance

Tax is of vital importance in economic studies as it not only influences social wealth distribution, but also affects social value creation. As a result, the determinants and consequences of tax planning are hot topics in both economics and accounting fields (Graham et al., 2012; Hanlon & Heitzman, 2010; Shackelford & Shevlin, 2001). Accounting research especially focuses on the relationship between corporate governance and tax avoidance (Chen & Chu, 2005; Desai & Dharmapala, 2006; Desai, Dyck, & Zingales, 2007; Schön, 2008; Zeng & Zhang, 2009). Within corporate governance, controlling shareholders and top executives with respect to tax avoidance are a major area of concern.

Prior literature offers mixed evidence on how top executives’ motivations influence tax avoidance. Phillips (2003) tested the relationship between CEOs’ performance evaluation and tax results. However, he did not find a reduced ETR for firms that use after-tax profit to evaluate their CEOs. Armstrong, Blouin, & Larcker (2012) support the results of Phillips (2003). However, Gaertner (2014) finds that firms that take after-tax profit as the evaluation index for their CEOs have significantly lower ETR. The sample they used was hand-collected from public firms in the US and is much larger in sample size than those of Phillips (2003) and Armstrong et al. (2012).

In addition to the performance evaluation measures used for CEOs, there are also papers looking at different aspects of top executives. Gallemore, Maydew, & Thornock (2014) show that expected reputation costs influence a CEO’s tax decisions. Wu, Wu, and Rui (2009) and Li and Xu (2013) find that the political status of top executives helps reduce the tax burden.

Compared with US researches, research in China about corporate governance and tax avoidance concentrates on ownership structure. The literature reports similar results. Wu (2009) finds that SOEs have a higher tax burden than non-SOEs. Liu and Li (2012) find that local SOEs have a significantly higher tax rate than non-SOEs and central SOEs. Wang, Wang, & Peng (2010) explain the higher tax burden of SOEs by the lower bankruptcy cost of SOEs. Zhu and Yu (2010) offer the reason of profit and tax integration of SOEs.

This paper is the first one to use Chinese data to test the causal relationship between the performance evaluation system of SOEs and tax avoidance decisions. We provide evidence that personal incentives of CEOs in SOEs influence tax outcomes.

2.1.2. Compensation-performance sensitivity

Hölmstrom (1979) argues that the compensation contract is not optimal when the principal can only observe the outcome of the agent. Principals can improve the effectiveness of the compensation contract by taking advantage of extra information, such as accounting and other signals related to the agents’ internal ability. Jensen and Murphy (1990a) think that the higher the agency problem, the lower the compensation-performance sensitivity.

There is abundant domestic literature on compensation-performance sensitivity and its determinants. Wei (2000) finds no significant relation between top executives’ compensation and firm performance. However, other later papers show that CEO’s cash compensation change is positively related to ROE change (Du & Wang, 2007; Xiao, 2005; Xu, Wang, & Gong, 2007; Yang, Gao, & Yurtoglu, 2009; Zhang, Zhao, & Zhang, 2003; Zhou & Huang, 2006). Furthermore, Fang (2009, 2011) finds the stickiness of executives’ compensation. Xiao (2005) shows different compensation-performance sensitivity in SOEs and non-SOEs. Other factors that are found to influence compensation-performance sensitivity include growth (Zhang & Chen, 2006), stock price volatility (Shao, Liu, & Kong, 2008), internal control quality (Lu, Liu, & Xu, 2011), political connection (Liu, Zhang, Wang, & Wu, 2010), industry competition (Chen & Ding, 2011), stock market reform (Lei, Li, & Jin, 2010), political intervention (Liu, Sun, & Li, 2007) and marketization (Xin & Tan, 2009).

However, these papers use quite different measurements of performance. In the above 17 papers, 11 of them use ROE as the performance measurement, eight of them take stock returns as performance measurement and five of them use ROA as performance measurement. One natural question is whether the performance measurements used in these papers can reflect the actual evaluation process of CEOs and further influence tax behaviour. One advantage of our paper is that we use performance measurement from the compensation contract in the real business world.

2.2. Institutional background

On 1 January 2004, the central SASAC released the Performance Evaluation Method for Executives in Central SOEs, which was further amended in 2006, 2009 and 2012. In this guidance, the compensation of executives in central SOEs is directly linked to annual performance measures. Annual performance measures include basic measures and classified measures. Basic measures include before-tax profit and ROE. Before-tax profit is the integrated total profit after audit. ROE is calculated as after-tax profit divided by net assets. The 2009 modified edition changed the basic measures to before-tax profit and economic value added (EVA). EVA is calculated as audited after-tax profit minus cost of capital. The essence of EVA is similar to ROE. In 2012, the SASAC made a further clarification of how to calculate profit and adjust the relative weight of basic versus classified measures.

After the central SASAC released the Performance Evaluation Method for Executives in Central SOEs, local SASACs made corresponding guidance for the performance evaluation of executives in local SOEs which are under their control. We hand-collect the official guidance and finally get the performance evaluation method in 28 provinces. We find that local SASACs in 17 provinces (Hebei, Liaoning, Heilongjiang, Jiansu, Anhui, Jiangxi, Shandong, Henan, Hubei, Hunan, Sichuan, Guizhou, Yunnan, Shanxi, Gansu, Qinghai and Hainan), three municipalities directly under the Central Government (Beijing, Tianjin and Chongqing) and four municipalities (Guangxi, Neimenggu, Ningxia and Xinjiang) take before-tax profit and ROE as performance measures.

However, local SASACs in Shanghai, Guangdong, Zhejiang and Fujian provinces take after-tax profit and ROE as basic measures. In August 2003, the SASAC of Shanghai released the Performance Evaluation Method for Executives in SOEs of Shanghai.² The following performance indexes are used in the evaluation process: strategic planning and budget implementation, maintenance and appreciation of state-owned assets or ROE, realisable income, and sales growth of key operations. In August 2008, Shanghai released Opinions for Reform of Shanghai SOEs and clearly added ROE and after-tax profit after non-recurrent profit as performance measures. Guangdong, Fujian and Zhejiang released similar performance evaluation methods in 2005, 2006 and 2007, respectively, and used after-tax profit and ROE as major performance evaluation measures.³

Currently, a few papers take a deep look at the actual performance evaluation of Chinese SOEs. Du, Tang and Young (2012) obtain the proprietary data of the final scores of each CEO in SOEs and examine the determinants of their scores. They find that political connections, geographical distance to SASAC and corporate social responsibilities are related to the final scores.

2.3. Hypotheses development

2.3.1. Ownership structure and tax avoidance

In non-SOEs, tax is a burden for the firm. Reducing tax is one important way to increase cash flow and firm value. In China, methods to reduce tax include financing through a debt shield (Modigliani & Miller, 1963) and other mechanisms, for example, transfer pricing (Graham et al., 2012). Different from reducing tax through a debt shield, this paper focuses on tax avoidance through the inconsistent regulation of tax laws and accounting principles. Because of the different identifications of profit between taxation and accounting, firms can perform tax avoidance activities, which lead to temporary difference and permanent difference, reduced ETR, increased BTD, and lower tax burden.⁴

In SOEs, SASAC is the controlling shareholder and SASAC is under the control of local government. For the government, profits from the SOEs and income tax from SOEs are both government receipts. In addition, compared with the profits of the SOEs, income tax is more directly disposable and is one visible index for local economic development. As a result, SOEs have lower incentives to reduce tax because of the controlling shareholders of local government.

Furthermore, local government can directly intervene in the SOEs’ operational decisions and limit the tax saving behaviour. This happens when the local government intends to develop the local economy and mitigate the potential deficit budget caused by various preferential policies (Chen & Li, 2001; Liu & Li, 2012; Zhou & Zhao, 2002).

2.3.2. Performance evaluation and tax avoidance

Compared with controlling shareholders, top executives directly implement operational and financial decisions. Under the separation of ownership and control, a compensation contract is used to align the interests of executives with those of shareholders (Jensen & Meckling, 1976). Jensen and Murphy (1990b) argue that it is how the pay is linked with performance, not the pay level, that can effectively incentivise the management. As a result, the performance evaluation measures are the key in the compensation contract.

The performance measurements of non-SOEs are not required to be public and we get some information on non-SOEs’ compensation contracts through their voluntary disclosure. We find that non-SOEs prefer performance measures based on after-tax profit. For example, in the 2013 Compensation Contract with Directors and Top Executives of Yinxi Technology (Stock Code 300221), after-tax profit is used clearly as one baseline for bonus. This is because the controlling shareholders of the non-SOEs cannot get benefits from tax, and consider tax as a cost. Using after-tax profits to evaluate executives can motive tax-saving behaviour and benefit shareholders.

Although the central SASAC uses before-tax profits to evaluate central SOEs, local SASACs can make their own choice to accommodate local situations. The results we can observe are that local SASACs in different provinces finally use different performance measures, especially the different choice between before-tax and after-tax profits. Central SASACs and most of the local SASACs choose before-tax profit; however, Shanghai, Guangdong, Zhejiang and Fujian choose after-tax profits as the major index.

In financial statements, the difference between before-tax and after-tax profit is tax expense. Keeping before-tax profit constant, then after-tax profits and tax expense are negatively correlated. When CEOs are evaluated on before-tax profit, tax expense will not influence their performance. As a result, the CEO has no incentive to make tax avoidance activities, especially when there are reputation risks and tax inspection risks. The outcome is a higher ETR and lower tax-book difference. In contrast, when CEOs’ compensation is based on after-tax profit, they have higher incentives to make tax saving activities, which will decrease tax expense and increase after-tax profit. The outcome reflected in financial statements is lower ETR and higher tax-book difference.

We find that each province in our sample includes return on equity (ROE) as an evaluation index for local SOEs. Since ROE is calculated as after-tax profit divided by total assets, CEOs of SOEs in each province would care about after-tax profit. We believe that this phenomenon will not influence the argument and results in this paper because the incentive for tax avoidance driven by ROE index is the same for SOEs in each state. Our results show the marginal effect of choice between after-tax profit and before-tax profit on tax avoidance, conditional on common incentives from ROE. Furthermore, the weight on after-tax profit/before-tax profit is large enough to drive the results. Taking Guangdong province as an example, the score on ROE is 20 points while the score on after-tax profit is also 20 points. As a results, the weight on after-tax is large enough to influence the SOEs’ tax avoidance motivation.

Du et al (2012) find that corporate social responsibility can influence the final evaluation score of the CEOs in SOEs. They use two measures for corporate social responsibility: number of employees and industry. In consequence, they do not mention taxation and their paper is not related to our measures.

We propose the first hypothesis in this paper:

H1:

Compared with SOEs that use before-tax profit as the evaluation index, SOEs that use after-tax profit have lower ETR and higher BTD;

2.3.3. Performance evaluation and debt-tax sensitivity

Prior research finds that the cost of tax is one of the most important factors of capital structure (Doidge & Dyck, 2015; Faccio & Xu, 2015; Graham & Tucker, 2006; Richardson, Lanis, & Leung, 2014). Graham and Tucker (2006) provide empirical evidence for the substitution relation between tax avoidance and asset-liability ratio. Richardson et al (2014) find that in firms with a higher percentage of independent directors, there is higher substitution between tax avoidance and tax. They explain this as the advisory function of independent directors.

Compared with independent directors, top executives should play a larger role in tax avoidance behaviour and debt financing. As a result, we argue that the performance evaluation for top executives will exert incentives related to tax financing. The interest expense of tax will reduce both before-tax profit and tax expense. On one hand, when top executives are evaluated on after-tax profit, they would like to reduce the debt financing and interest expense to make sure they can reach the after-tax performance target. The outcome is a high substitution between debt financing and tax avoidance behaviour. On the other hand, when top executives are evaluated on before-tax profit, their tax avoidance activities will not influence before-tax profit. However, tax interest expense will reduce before-tax profit. The outcome is a reduced substitution between debt financing and tax avoidance. We propose our second hypothesis as follows:

H2:

Compared with SOEs that use before-tax profit as the evaluation index, SOEs that use after-tax profit have a higher level of substitution between debt financing and tax avoidance.

3. Research design

3.1. Sample selection

The measurement of income tax expense in annual reports changed in 2007 in China, and our sample consists of listed companies in China from 2007 to 2015. After excluding (1) financial companies; (2) companies located in Tibet, Shanxi, and Jilin;⁵ (3) companies whose before-tax profit or income tax expense is negative, which makes the effective tax rate negative; (4) other financial data required are missing, a total of 8426 firm-year observations are obtained. Table 1 shows the detail of sample selection and number of observations in each year.

Financial data are sourced from the China Stock Market and Accounting Research (CSMAR) database system. Economic data of provinces in China are sourced from National Bureau of Statistics of China. Performance evaluation systems data are hand-collected from the websites of the central and local State-owned Assets Supervision and Administration Commissions. Table 1 shows the detail of sample selection and number of observations in each year.

Table 1. Detail of sample selection and number of observations in each year.

Panel ：Detail of sample selection
Total observations in 2007–2015	12,584
Excluding
-- Financial Companies	–383
-- Companies Located in Tibet, Shanxi, and Jilin	–595
-- Before-tax profit is negative	–629
-- Income tax expense is negative	–1021
-- Other financial data required is missing	–1530
Final observation obtained	8426
Panel B：Number of observations in each year
2007	979
2008	812
2009	877
2010	946
2011	846
2012	926
2013	1045
2014	1018
2015	977
Total	8426

3.2. Variables measurement

3.2.1. Tax avoidance measurement

According to the previous literature, the effective tax rate (ETR) and book-tax difference (BTD) are the most frequently used tax avoidance measurements. In this paper, we use ETR and three different methods to measure BTD, including total BTD (BTD_TOTAL), Desai and Dharmapala (2006) BTD measurement (BTD_DD), and Frank, Lynch, and Rego (2009) BTD measurement (BTD_F).

ETR is the ratio of income tax expense to before-tax profit. Firms will use serval methods to reduce the tax burden. Therefore, lower ETR means a lower level of tax burden or a higher level of tax avoidance. Prior literature frequently uses ETR as a measurement of tax avoidance (Gaertner, 2014; Plesko, 2003). We calculate ETR in this paper as follows.

ETR = Income tax expense/Before-tax profit

The difference between before-tax profit and net profit is income tax expense. SOEs that use after-tax profit as an evaluation index have higher incentives to reduce the tax burden than those that use before-tax profit as an evaluation index. Income tax expense includes current income tax expense and deferred income tax expense. Therefore, the two methods to reduce the income tax expense are reducing both current income tax expense and deferred income tax expense. We use ETR to measure the total level of firm tax avoidance, because ETR cannot show which method is used to reduce the tax burden.

BTD is another measurement of tax avoidance (Manzon & Plesko, 2002), which is the difference between accounting earnings and taxable income. Higher BTD means a higher level of tax avoidance. BTD shows how firms reduce tax burden through adjusting current income tax expense, instead of deferred income tax expense.

We calculate BTD in this paper as follows.

BTD = {before-tax profit – (income tax expense – deferred tax liability increasing + deferred tax assets increasing)/nominal tax rate}/total assets

BTD_TOTAL is an approximate measurement of tax avoidance. Earnings management also influences BTD_TOTAL. A higher level of earnings management through non-taxable items leads to a higher level of discretionary accruals and larger BTD_TOTAL. In order to avoid the above defects, we use another two BTD measurements.

Desai and Dharmapala (2006) used the regression equation to exclude the influence of earnings management. Specifically, ordinary least squares is used to obtain BTD_DD, which is equal to the prediction error ${\epsilon }_{i,t}$.

$\mathrm{B}\mathrm{T}\mathrm{D}\mathrm{\_}\mathrm{T}\mathrm{O}\mathrm{T}\mathrm{A}{\mathrm{L}}_{i,t}={\alpha }_0+{\alpha }_1\mathrm{T}{\mathrm{A}}_{i,t}+u_i+{\epsilon }_{i,t}$

The definition of TA is listed in Table 2. $u_i$ is the firm level fixed effect.

Table 2. Variable definitions.

Variable Name	Abbreviation	Definition
Effective tax rate	ETR	tax expense/before-tax profit
Book-tax difference-total	BTD_TOTAL	{before-tax profit – (income tax expense - deferred tax liability increasing + deferred tax assets increasing)/nominal tax rate}/total assets
Book-tax difference-DD method	BTD_DD	BTD calculated based on DD(2006); see more in the paper;
Book-tax difference-F method	BTD_F	BTD calculated based on Frank et al (2009); see more in the paper;
Performance measure	D	1 for firms registered in Shanghai, Guangdong, Fujian and Zhejiang; 0 for other firms;
Controlling shareholder	LOCAL	1 for controlling shareholders on and under province level; 0 for others;
Total asset	SIZE	log (total asset)
Liability asset ratio	LEV	(long-term debt + short-term debt + non-current liabilities due within one year)/total asset
Profitability	ROA	net profit/total asset
Market to book ratio	MB	(equity market value + book value of debt)/total asset
Nominal tax rate	TAX	nominal tax rate of parent company
Ownership of the largest shareholder	FIRST	ownership of the largest shareholder
Fixed asset	FIXED	net fixed asset/total asset
Inventory	INV	inventory/total asset
Total accrual	TA	(net profit-net cash flow from operational activities)/total asset
Cash flow	OCF	net cash flow from operational activities/total asset
Intangible asset	INTAN
Net profit attributed to parent company’s owner	INCOME	net profit attributed to parent company’s owner/total asset
Net profit of minority shareholders	MI	net profit of minority shareholders/total asset
Lagged BTD_TOTAL	LAGBTD_TOTAL	lagged BTD_TOTAL
Median leverage of the industry	LEV_MEDIAN	median leverage of the industry
GDP	GDP	the GDP of the province
Net budge	NETBUDGET	province level(budget in-budget out)/GDP
Total import and export	IMEXPORT	province level export + import/GDP
Tax from SOE	TAX from SOE	province level tax from SOEs/total tax
Year	YEAR	9 years from 2007 to 2015
Industry	INDUSTRY	Based on <Guidelines for the Industry Classification of Listed Companies>; 12 industries in total without financial industries;

Frank et al (2009) also use the OLS regression to exclude permanent tax difference in BTD_TOTAL. The regression model is as follows:

$\mathrm{B}\mathrm{T}\mathrm{D}\mathrm{\_}\mathrm{T}\mathrm{O}\mathrm{T}\mathrm{A}{\mathrm{L}}_{i,t}={\alpha }_0+{\alpha }_1\mathrm{I}\mathrm{N}\mathrm{T}\mathrm{A}{\mathrm{N}}_{i,t}+{\alpha }_2\mathrm{I}\mathrm{N}\mathrm{C}\mathrm{O}\mathrm{M}{\mathrm{E}}_{i,t}+{\alpha }_3\mathrm{M}{\mathrm{I}}_{i,t}+{\alpha }_4\mathrm{T}\mathrm{A}{\mathrm{X}}_{i,t}\text{break}+{\alpha }_5\mathrm{L}\mathrm{A}\mathrm{G}\mathrm{B}\mathrm{T}\mathrm{D}\mathrm{\_}\mathrm{T}\mathrm{O}\mathrm{T}\mathrm{A}{\mathrm{L}}_{i,t}+{\epsilon }_{i,t}$

The definitions of variables in this model are listed in Table 2. According to Frank et al (2009), OLS is used to obtain BTD_F, which is equal to the prediction error ${\epsilon }_{i,t}$.

3.2.2. Performance evaluation measurement

We use a dummy variable D to measure the performance evaluation system. D is equal to 1 if a firm is located in a province that uses after-tax profit as the evaluation index. D of firms in Shanghai, Guangdong, Zhejiang and Fujian is equal to 1, and D of firms in other provinces is equal to 0. Performance evaluation systems data are sourced from the websites of the central and local SASAC. There are several indexes in the performance evaluation system, including financial indexes or non-financial indexes. Financial indexes contain net profit, ROE, EVA, the rate of value added of state-owned assets. In this paper, we focus on the difference between using after-tax profit and before-tax profit.

3.2.3. Types of firms

We use a dummy variable LOCAL to measure the type of firms: local SOEs and non-SOEs. LOCAL is equal to 1 if the observation is local SOEs, and equal to 0 otherwise. Non-SOEs contain central SOEs and private firms. The central SOEs use before-tax profit as the evaluation index. However, private firms use after-tax profit as the evaluation index, because shareholders of private firms cannot obtain the tax.

3.2.4. Control variables

In H1 testing, ETR and BTD are dependent variables. According to previous literature, we use the following control variables: total assets (SIZE), leverage (LEV), return on total assets (ROA), book to market ratio (MB), nominal tax rate (TAX), proportion of the largest shareholder (FIRST), fixed assets (FIXED), inventory density (INV) and total accruals (TA).

In H2 testing, leverage is the dependent variable. According to previous literature, we use following control variables: total assets (SIZE), return on total assets (ROA), operating cash flow (OCF), fixed assets (FIXED), book-to-market ratio (MB) and median of industry leverage (LEV_MEDIAN).

The definition of all variables is listed in Table 2.

3.3. Model

We use the following models to test H1 and H2.

Model (1)

$\mathrm{T}\mathrm{A}\mathrm{X}\mathrm{A}{\mathrm{V}}_{i,t}={\beta }_0+{\beta }_1{D}_{i,t}+{\beta }_2\mathrm{L}\mathrm{O}\mathrm{C}\mathrm{A}{\mathrm{L}}_{i,t}+{\beta }_{3}D\ast \mathrm{L}\mathrm{O}\mathrm{C}\mathrm{A}{\mathrm{l}}_{i,t}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}{\mathrm{L}}_{i,t}+{ϵ}_{i,t}$

Model (2)

$\begin{array}{rl}& \mathrm{L}\mathrm{E}{\mathrm{V}}_{i,t}={\gamma }_0+{\gamma }_1{D}_{i,t}+{\gamma }_2\mathrm{L}\mathrm{O}\mathrm{C}\mathrm{A}{\mathrm{L}}_{i,t}+{\gamma }_3\mathrm{T}\mathrm{A}\mathrm{X}\mathrm{A}{\mathrm{V}}_{i,t}+{\gamma }_{4}D\ast \mathrm{L}\mathrm{O}\mathrm{C}\mathrm{A}{\mathrm{l}}_{i,t}+{\gamma }_{5}D\ast \mathrm{T}\mathrm{A}\mathrm{X}\mathrm{A}{\mathrm{V}}_{i,t}\\ & +{\gamma }_{6}D\ast \mathrm{L}\mathrm{O}\mathrm{C}\mathrm{A}\mathrm{L}\ast \mathrm{T}\mathrm{A}\mathrm{X}\mathrm{A}{\mathrm{V}}_{i,t}+\mathrm{C}\mathrm{O}\mathrm{N}\mathrm{T}\mathrm{R}\mathrm{O}{\mathrm{L}}_{i,t}+e_{i,t}\end{array}$

Model (1) investigates how the performance evaluation systems of different SOEs in different provinces influence the level of tax avoidance. TAXAV includes four measurements of tax avoidance. We examine whether the SOEs using after-tax profit have a higher level of tax avoidance. Although we can use the sample that only includes SOEs and test the significance of D, the empirical result may be misled due to other province level factors. Therefore, we use the difference between the tax avoidance level of non-SOEs in different provinces as the baseline. In model (1), we focus on ${\beta }_3$, which is a difference-in-difference test model. Namely, ${\beta }_3$ measures the difference as follows:

(the tax avoidance level of local SOEs in four provinces using after-tax profit as a performance evaluation index – the tax avoidance level of local SOEs in other provinces) – (the tax avoidance level of non-SOEs in four provinces using after-tax profit as a performance evaluation index – the tax avoidance level of non-SOEs in other provinces)

We estimate using ETR as TAXAV, ${\beta }_3$ is significant and negative, which means the ETR of local SOEs in four provinces using after-tax profit as a performance evaluation index is lower. Meanwhile, using BTD (BTD_TOTAL, BTD_DD and BTD_F) as TAXAV, ${\beta }_3$ is significant and positive, which means the BTD of local SOEs in four provinces using after-tax profit as a performance evaluation index is higher.

Model (2) investigates how the different SOEs’ performance evaluation systems in different provinces influence the level of substitution between debt financing and tax avoidance. Similarly, we focus on ${\gamma }_6$. We estimate, using ETR as TAXAV, ${\gamma }_6$ is significant and positive. Meanwhile, using BTD (BTD_TOTAL, BTD_DD and BTD_F) as TAXAV, ${\gamma }_6$ is significant and negative.

4. Empirical results and analysis

4.1. Descriptive statistics

Table 3 reports the descriptive statistics of all variables.

Table 3. Descriptive statistics.

Variables	Obs.	Mean	Std. Dev.	Min.	Q1	Median	Q3	Max.
ETR	8426	0.221	0.139	0.00	0.136	0.202	0.273	0.769
BTD_TOTAL	8426	0.0104	0.045	–0.1170	–0.0089	0.0045	0.0229	0.2310
BTD_DD	8426	0.0008	0.043	–0.1191	–0.0186	–0.0039	0.0139	0.2010
BTD_F	7169	–0.0031	0.027	–0.0915	–0.0145	–0.0027	0.0088	0.0829
D	8426	0.313	0.464	0.00	0.00	0.00	1.00	1.00
LOCAL	8426	0.466	0.499	0.00	0.00	0.00	1.00	1.00
SIZE	8426	22.05	1.335	18.93	21.16	21.93	22.83	25.83
LEV	8426	0.29	0.250	0.00	0.10	0.25	0.41	1.39
ROA	8426	0.07	0.085	0.00	0.02	0.05	0.09	0.52
MB	8426	2.05	1.442	0.89	1.22	1.59	2.29	9.45
TAX	8426	0.21	0.056	0.07	0.15	0.25	0.25	0.33
FIRST	8426	36.84	15.56	9.09	24.03	35.02	48.36	75.84
FIXED	8426	0.29	0.229	0.00	0.11	0.24	0.42	1.10
INV	8426	0.23	0.245	0.00	0.06	0.15	0.28	1.23
TA	8426	0.004	0.112	–0.284	–0.052	–0.008	0.042	0.538

Notes: Table 3 reports the descriptive statistics of all variables. All continuous variables are winsorised at 1% and 99%. The definitions of all variables are listed in Table 2.

All continuous variables are winsorised at 1% and 99%. We find the mean of listed companies ETR in our sample is 22.1%, the median of listed companies ETR in our sample is 20.2%, which is less than the nominal corporate income tax rate 33% or 25%. The mean of listed companies BTD in our sample is 0.0104. The mean of D is 31.3%, which means 31.3% listed companies in our sample are located in Shanghai, Guangdong, Zhejiang or Fujian. The mean of LOCAL is 46.6%, which means 46.6% of listed companies in our sample are local SOEs. The mean of ROA is 7%. The mean of FIRST is 36.84% and the maximum value reaches 75.84%, as seen from the fact that the stock rights of listed companies in China are highly centralised. The mean of FIXED and INV are 29% and 23%.

Table 4 reports the mean of all variables in different provinces for our sample. The nominal corporate income tax rates of different provinces cover a wide range. Chongqing, Shaanxi and Ningxia are less than 20%, and other provinces are more than 20%. The ETR of listed companies in Ningxia for our sample is lowest, 16.7% on average. Liaoning is highest, 25.7% on average. The SIZE of listed companies in Beijing for our sample is highest, and The LEV of listed companies in Qinghai is highest.

Table 4. Descriptive statistics by provinces.

Province	Obs	ETR	BTD_TOTAL	BTD_DD	BTD_F	LOCAL	SIZE	LEV	ROA	MB	TAX	FIRST	FIXED	INV	TA
Beijing	656	0.216	0.0079	–0.0010	–0.0031	0.287	22.87	0.25	0.07	1.91	0.21	44.99	0.26	0.21	–0.005
Tianjin	171	0.237	0.0066	–0.0049	–0.0032	0.556	22.39	0.31	0.06	2.08	0.23	40.79	0.24	0.30	0.021
Hebei	201	0.252	0.0062	–0.0026	–0.0028	0.483	22.16	0.34	0.08	2.11	0.23	39.99	0.37	0.20	–0.006
Inner Mongolia	117	0.217	0.0058	–0.0031	–0.0038	0.299	22.43	0.32	0.05	1.70	0.22	38.84	0.32	0.20	–0.005
Liaoning	313	0.257	0.0065	–0.0027	–0.0023	0.451	22.05	0.31	0.06	1.82	0.24	37.00	0.33	0.25	0.002
Heilongjiang	191	0.222	0.0102	0.0006	–0.0051	0.293	21.73	0.25	0.06	2.18	0.22	39.33	0.26	0.26	0.005
Jiangsu	650	0.228	0.0074	–0.0023	–0.0049	0.415	21.95	0.25	0.06	1.91	0.20	36.31	0.24	0.22	0.001
Zhejiang	538	0.226	0.0110	0.0006	–0.0043	0.366	21.94	0.26	0.07	1.96	0.21	35.75	0.23	0.27	0.012
Anhui	306	0.217	0.0026	–0.0054	–0.0019	0.690	22.02	0.26	0.07	1.76	0.20	35.31	0.35	0.19	–0.015
Fujian	340	0.232	0.0116	0.0002	–0.0023	0.521	21.85	0.30	0.08	1.97	0.22	34.66	0.21	0.35	0.025
Jiangxi	170	0.207	0.0121	0.0042	0.0019	0.653	22.16	0.28	0.08	1.87	0.22	43.42	0.35	0.17	–0.022
Shandong	473	0.221	0.0072	–0.0012	–0.0019	0.624	22.21	0.28	0.08	1.96	0.21	34.78	0.35	0.19	–0.011
Henan	198	0.234	0.0133	0.0048	0.0026	0.530	22.18	0.31	0.07	2.00	0.21	38.41	0.37	0.17	–0.014
Hunan	239	0.231	0.0100	0.0011	–0.0015	0.531	21.69	0.31	0.07	2.31	0.21	31.88	0.35	0.19	–0.009
Hubei	424	0.226	0.0112	0.0016	–0.0030	0.351	21.88	0.33	0.07	2.16	0.21	32.37	0.34	0.23	0.003
Guangdong	886	0.212	0.0172	0.0066	–0.0019	0.432	22.11	0.30	0.08	2.11	0.21	35.77	0.26	0.24	0.017
Guangxi	138	0.213	–0.0003	–0.0092	–0.0133	0.630	21.81	0.38	0.06	2.03	0.18	34.38	0.43	0.14	0.002
Hainan	128	0.241	0.0139	0.0028	–0.0027	0.266	21.55	0.26	0.07	2.85	0.22	27.21	0.24	0.18	0.018
Sichuan	397	0.217	0.0114	0.0024	–0.0019	0.363	21.57	0.27	0.08	2.46	0.20	35.13	0.32	0.20	–0.002
Chongqing	186	0.194	0.0164	0.0067	–0.0023	0.500	21.82	0.34	0.07	2.13	0.18	38.68	0.28	0.24	0.020
Guizhou	129	0.205	0.0091	–0.0018	–0.0091	0.465	21.83	0.28	0.09	2.28	0.19	36.67	0.21	0.27	0.018
Yunnan	142	0.227	0.0096	–0.0011	–0.0068	0.627	22.04	0.38	0.07	2.04	0.18	35.85	0.29	0.30	0.016
Shaanxi	167	0.183	0.0126	0.0033	–0.0038	0.557	21.53	0.20	0.05	2.16	0.19	34.76	0.31	0.21	–0.005
Gansu	111	0.210	0.0047	–0.0027	–0.0055	0.496	21.54	0.30	0.07	2.11	0.22	37.85	0.35	0.21	–0.014
Ningxia	62	0.167	0.0157	0.0053	–0.0040	0.258	21.45	0.38	0.05	2.40	0.19	33.59	0.39	0.22	0.018
Qinghai	57	0.239	0.0136	0.0028	–0.0066	0.579	21.75	0.42	0.09	2.59	0.17	34.01	0.32	0.14	0.038
Xinjiang	161	0.216	0.0076	0	–0.0053	0.509	21.93	0.36	0.05	1.91	0.20	35.47	0.39	0.20	–0.021
Shanghai	875	0.211	0.0140	0.0036	–0.0032	0.570	22.23	0.27	0.07	2.06	0.22	38.20	0.23	0.22	0.012

Notes: Table 4 reports the mean of all variables by provinces.

A bivariate correlation matrix of the main variables is presented in Table 5. The correlation between BTD and ETR is significant and negative. The significantly negative relationship between the level of tax avoidance (ETR/BTD) and LOCAL suggests that the level of local SOEs tax avoidance is lower. Local SOEs have higher ETR and lower BTD than local non-SOEs. This means that on average local SOEs have lower tax avoidance than local non-SOEs.

Table 5. Correlation matrix of variables.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)
ETR(1)	1.00
BTD_TOTAL(2)	–0.47*	1.00
BTD_DD(3)	–0.46*	0.98*	1.00
BTD_F(4)	–0.06*	0.63*	0.64*	1.00
D(5)	–0.02	0.05*	0.04*	0.01	1.00
LOCAL(6)	0.04*	–0.03*	–0.03	0.06*	0.01	1.00
SIZE(7)	0.06*	–0.06*	–0.05*	0.03	0.02	0.08*	1.00
LEV(8)	0.06*	0.06*	0.04*	0.11	–0.02	0.02	0.16*	1.00
ROA(9)	–0.17*	0.38*	0.36*	0.02	0.04	–0.06*	–0.17*	0.06*	1.00
MB(10)	–0.11*	0.11*	0.10*	–0.06*	–0.00	–0.16*	–0.48*	–0.24*	0.23*	1.00
TAX(11)	0.23*	0.10*	0.10*	0.40*	0.05*	0.08*	–0.02	0.07*	0.03	0.01	1.00
FIRST(12)	–0.00	0.03	0.04	0.04*	–0.02	0.10*	0.30*	0.07*	0.09*	–0.21*	0.04*	1.00
FIXED(13)	–0.05*	0.04*	0.09*	0.03	–0.14*	0.09*	0.03*	0.32*	0.13*	–0.12*	0.02	0.09*	1.00
INV(14)	0.12*	0.01	–0.04*	0.02	0.07*	–0.05*	0.02	0.26*	0.12*	–0.13*	0.08*	0.09*	–0.32*	1.00
TA(15)	–0.09*	0.21*	0.04*	–0.02	0.07*	–0.08*	–0.10*	0.10*	0.29*	0.09*	0.01	–0.02	–0.24*	0.31*	1.00

Notes: Table 5 reports the Pearson correlation coefficients. * indicate statistical significance at the 5%.

Table 6 reports the univariate analysis results. On average, SOEs in four provinces using after-tax profit as a performance evaluation index have lower ETR and higher BTD compared with the other SOEs, which supports H1. On the other hand, non-SOEs in four provinces using after-tax profit as a performance evaluation index are not significantly different compared with the other non-SOEs, and BTD is similar to the results of SOEs. Based on the univariate analysis results, we have to use non-SOEs in different provinces as a baseline to exclude the province level difference of tax avoidance.

Table 6. Univariate analysis.

	LOCAL=1			LOCAL=0
	(1)	(2)	(3)	(4)	(5)	(6)
	D=0	D=1	T-test: DIFF	D=0	D=1	T-test: DIFF
Obs.	2666	1256	–	3121	1383	–
ETR	0.230	0.219	2.29**	0.213	0.211	0.15
BTD_TOTAL	0.0064	0.0130	–4.72***	0.0107	0.0152	–2.82***
BTD_DD	–0.0021	0.0033	–3.95***	0.0001	0.0038	–1.87*
BTD_F	–0.0018	–0.0000	–1.87*	–0.0045	–0.0056	1.03*
SIZE	22.13	22.26	–3.23***	21.95	21.91	0.91
LEV	0.31	0.27	4.32***	0.28	0.29	–0.98
ROA	0.07	0.07	–0.32	0.08	0.09	–3.97***
MB	1.81	1.79	0.28	2.28	2.28	0.09
TAX	0.22	0.23	4.12***	0.20	0.21	–2.98***
FIRST	38.95	35.53	0.83	35.75	34.13	3.17***
FIXED	0.34	0.25	11.32***	0.29	0.23	8.29***
INV	0.21	0.24	–4.46***	0.23	0.28	–5.86***
TA	–0.01	–0.00	–4.54***	0.01	0.02	–5.12***

Table 6 reports the results of univariate analysis. Column 3 and Column 6 present t-test results of variables between D=1 and D=0. *** denotes a significance level at 1%, ** at 5%, * at 10%. All continuous variables are winsorised at 1% and 99%.

4.2. Empirical results

4.2.1. Empirical results of H1

Table 7 presents the empirical results of H1. We use four proxies (ETR, BTD_TOTAL, BTD_DD and BTD_F) to measure tax avoidance.

All the regression results control for the fixed effect at year and industry level. T values are bracketed. Columns (1), (3), (5) and (7) present the regression results of local SOEs (LOCAL=1). The result is similar to the results of univariate analysis. SOEs in four provinces using after-tax profit as a performance evaluation index have significantly lower ETR and higher BTD_TOTAL/BTD_DD compared with the other SOEs. Columns (2), (4), (6) and (8) present the regression results of DID using non-SOEs as a baseline. The coefficient of LOCAL suggests SOEs have a lower level of tax avoidance (higher ETR and lower BTD_TOTAL/BTD_DD) in general. Furthermore, the coefficient of D_LOCAL means that SOEs in four provinces using after-tax profit as a performance evaluation index have significantly lower ETR and higher BTD_TOTAL/BTD_DD, which excludes other province-level difference of tax avoidance by DID. Accordingly, hypothesis H1 is supported.

Table 7. Empirical results of H1.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	ETR		BTD_TOTAL		BTD_DD		BTD_F
	SOEs	TOTAL	SOEs	TOTAL	SOEs	TOTAL	SOEs	TOTAL
D	–0.019***	–0.005	0.006***	0.002	0.006***	0.002	–0.001	–0.003***
	(–4.17)	(–1.16)	(4.34)	(1.24)	(4.36)	(1.09)	(–0.84)	(–2.78)
LOCAL		0.008**		–0.003***		–0.003***		0.000
		(2.27)		(–2.59)		(–2.64)		(0.70)
D_LOCAL		–0.016***		0.004**		0.005**		0.003**
		(–2.68)		(2.33)		(2.46)		(2.04)
SIZE	0.002	–0.001	0.000	0.001	0.001	0.001	0.000	–0.001*
	(0.91)	(–0.56)	(0.53)	(1.04)	(0.63)	(1.14)	(0.19)	(–1.78)
LEV	0.026**	0.022***	0.003	0.010***	0.002	0.010***	0.012***	0.012***
	(2.53)	(3.24)	(0.55)	(2.83)	(0.48)	(2.90)	(4.87)	(6.60)
ROA	–0.285***	–0.245***	0.174***	0.190***	0.177***	0.195***	0.038***	0.018**
	(–9.67)	(–12.76)	(8.23)	(14.68)	(8.92)	(15.90)	(3.59)	(2.43)
MB	–0.003	–0.005***	–0.002*	0.000	–0.002*	0.000	–0.001*	–0.002***
	(–1.30)	(–3.10)	(–1.76)	(0.32)	(–1.72)	(0.53)	(–1.91)	(–4.53)
TAX	0.478***	0.526***	0.096***	0.088***	0.094***	0.086***	0.192***	0.191***
	(11.52)	(18.85)	(7.33)	(9.15)	(7.28)	(9.29)	(21.42)	(30.23)
FIRST	–0.000*	–0.000**	0.000	0.000	0.000	0.000	–0.000	0.000
	(–1.68)	(–2.23)	(1.24)	(0.57)	(1.09)	(0.31)	(–0.24)	(0.43)
FIXED	–0.006	–0.018**	–0.010**	–0.005	–0.008*	–0.004	–0.015***	–0.010***
	(–0.51)	(–2.24)	(–2.14)	(–1.33)	(–1.93)	(–1.19)	(–6.02)	(–4.62)
INV	0.098***	0.077***	–0.024***	–0.015***	–0.024***	–0.016***	–0.012***	–0.005**
	(7.83)	(9.45)	(–4.06)	(–3.84)	(–4.24)	(–4.31)	(–3.76)	(–2.52)
TA	–0.136***	–0.122***	0.047***	0.052***	–0.026**	–0.018**	–0.017**	–0.009*
	(–6.06)	(–8.47)	(3.43)	(5.82)	(–2.03)	(–2.19)	(–2.44)	(–1.94)
constant	0.087*	0.140***	–0.029	–0.035***	–0.040**	–0.045***	–0.036***	–0.024***
	(1.87)	(4.60)	(–1.50)	(–2.91)	(–2.13)	(–3.90)	(–3.18)	(–3.22)
Industry	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	3922	8426	3922	8426	3922	8426	3392	7169
R^2	0.115	0.126	0.143	0.187	0.141	0.162	0.217	0.182

Notes: Table 7 presents the empirical results of H1. All continuous variables are winsorised at 1% and 99%. The definitions of all variables are listed in Table 2. *** denotes a significance level at 1%, ** at 5%, * at 10%. T values are bracketed.

Although we use DID to exclude other province-level difference of tax avoidance, there still exists endogeneity. The performance evaluation system is not randomly set up by the local SASAC, and the tax avoidance level of firms in provinces may influence the choice of the performance evaluation index. We considered endogeneity in our regression models and adopted an instrumental variable (IV) approach and 2SLS method to resolve the problem of endogeneity. We choose four province level exogenous variables, including fiscal budget net income of the province (NETBUDGET), total tax of local SOEs (TAX from SOE), GDP of the province (GDP) and total export-import volume (IMEXPORT). In this paper, we use the means of these four variables from 2000 to 2003 respectively as IV, which conforms to conditions of IV. On the one hand, the means of these four variables from 2000 to 2003 respectively have a high correlation with the choice of the performance evaluation index. The majority of provinces in our sample have set up performance evaluation systems since 2004, and the economic growth of the provinces will influence the choice of the performance evaluation index directly. For example, fiscal budget net income of the province will affect the choice of the performance evaluation index. On the other hand, these four variables are not correlated with the tax avoidance of firms directly; in particular, especially we use the mean from 2000 to 2003. In second stage of 2SLS, our sample is from 2007 to 2015. It seems impossible that the tax avoidance level of a firm will be influenced by the province level tax income and economic growth in previous years.

Table 8 shows the results of the first stage of 2SLS. Column (1) presents the results of regression with four IVs.

Table 8. H1 first stage of 2SLS.

	(1)	(2)
	D	D
NET BUDGET_AVERAGE	51.206***	67.860***
	(11.33)	(8.88)
TAX from SOE_AVERAGE	–0.000***	–0.000***
	(–9.32)	(–10.09)
GDP_AVERAGE	0.000***	0.000***
	(6.28)	(6.84)
IMEXPORT_AVERAGE	0.000***	0.000***
	(5.40)	(4.11)
SIZE_AVERAGE		–0.011
		(–0.10)
LEV_AVERAGE		1.359***
		(2.98)
ROA_AVERAGE		–2.809
		(–1.39)
MB_AVERAGE		–0.010
		(–0.04)
TAX_AVERAGE		–3.697***
		(–3.32)
FIRST_AVERAGE		–0.015***
		(–3.42)
FIXED_AVERAGE		–0.152
		(–0.32)
INV_AVERAGE		0.260
		(0.48)
TA_AVERAGE		–4.303***
		(–3.26)
constant	2.219***	4.499*
	(6.20)	(1.74)
Industry	Yes	Yes
N	959	959
R^2	0.6312	0.6746

Notes: Table 8 presents the empirical results of H1 First Stage of 2SLS. All continuous variables are winsorised at 1% and 99%. The definitions of all variables are listed in Table 2. *** denotes a significance level at 1%, ** at 5%, * at 10%. T values are bracketed.

Column (2) presents the results of regression with four IVs and the means of control variables from 2000 to 2003 respectively. In the first stage of 2SLS, each observation is a firm, therefore the number of observations is 959. It is more reasonable using a province level sample to estimate the choice of the performance evaluation index; however, the number of provinces in China is only 28, which cannot guarantee the accuracy of estimation. Table 8 shows that the coefficients of four IVs is significance at 1%, suggesting they are highly correlated with the choice of the performance evaluation index. Comparing column (1) with column (2), we find the R² in column (1) is 63%, and it is 67% in column (2), which means the control variables have slight incremental information.

Table 9 shows the results of the second stage of 2SLS.

Table 9. H1 second stage of 2SLS.

	(1)	(2)	(3)	(4)
	ETR	BTD_TOTAL	BTD_DD	BTD_F
D1	0.001	0.000	0.000	–0.000*
	(0.94)	(0.63)	(0.54)	(–1.77)
LOCAL	–0.010***	0.001	0.002	0.002***
	(–2.69)	(1.17)	(1.29)	(2.73)
D1_LOCAL	–0.005***	0.001***	0.001***	0.000*
	(–5.48)	(3.13)	(3.32)	(1.74)
SIZE	–0.001	0.000	0.000	–0.000
	(–0.35)	(0.38)	(0.45)	(–0.37)
LEV	0.024***	0.007*	0.007*	0.009***
	(3.06)	(1.85)	(1.92)	(4.28)
ROA	–0.270***	0.205***	0.211***	0.021**
	(–11.86)	(14.00)	(15.08)	(2.57)
MB	–0.004**	–0.001	–0.001	–0.001***
	(–2.21)	(–0.97)	(–0.75)	(–3.39)
TAX	0.526***	0.091***	0.090***	0.189***
	(16.42)	(8.46)	(8.56)	(26.59)
FIRST	–0.000**	0.000	–0.000	–0.000
	(–2.40)	(0.11)	(–0.04)	(–0.37)
FIXED	–0.020**	–0.003	–0.003	–0.009***
	(–2.18)	(–0.83)	(–0.79)	(–3.98)
INV	0.088***	–0.017***	–0.018***	–0.006***
	(9.28)	(–4.11)	(–4.59)	(–2.76)
TA	–0.135***	0.053***	–0.019**	–0.008
	(–8.08)	(5.49)	(–2.04)	(–1.52)
constant	0.141***	–0.027*	–0.037***	–0.034***
	(3.71)	(–1.85)	(–2.67)	(–3.81)
Industry	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
N	6534	6534	6534	5632
R^2	0.134	0.192	0.170	0.172

Notes: Table 9 presents the empirical results of H1 Second Stage of 2SLS. All continuous variables are winsorised at 1% and 99%. The definitions of all variables are listed in Table 2. *** denotes a significance level at 1%, ** at 5%, * at 10%. T values are bracketed.

In the second stage of 2SLS, we substitute D estimated in the first stage of 2SLS (D1) for D in the main results of H1. The results correspond to the main results of H1. After resolving the problem of endogeneity by IV and 2SLS, we still find that H1 is supported.

4.2.2. Empirical results of H2

Table 10 presents the empirical results of H2.

Table 10. Empirical results of H2.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	TAXAV: ETR		TAXAV: BTD_TOTAL		TAXAV: BTD_DD		TAXAV: BTD_F
TAXAV	0.103***	0.104***	0.231**	0.224	0.247***	0.259*	0.745***	0.830***
	(6.24)	(4.01)	(2.45)	(1.57)	(2.60)	(1.78)	(6.30)	(4.34)
D	–0.002	0.024	–0.004	–0.000	–0.004	0.007	–0.005	0.009
	(–0.44)	(1.56)	(–0.86)	(–0.00)	(–0.86)	(0.92)	(–0.89)	(1.13)
LOCAL	–0.024***	–0.009	–0.023***	–0.014**	–0.023***	–0.015**	–0.026***	–0.017***
	(–5.06)	(–0.82)	(–4.90)	(–2.47)	(–4.91)	(–2.54)	(–5.36)	(–2.85)
D_TAXAV		–0.066		0.571**		0.559**		0.103
		(–1.19)		(2.11)		(2.04)		(0.30)
D_LOCAL		–0.074***		–0.010		–0.023**		–0.028***
		(–3.70)		(–0.99)		(–2.34)		(–2.66)
LOCAL_TAXAV		–0.030		–0.080		–0.069		–0.033
		(–0.78)		(–0.36)		(–0.31)		(–0.11)
D_LOCAL_TAXAV		0.221***		–1.247***		–1.378***		–0.819*
		(3.01)		(–3.28)		(–3.60)		(–1.68)
SIZE	0.012***	0.013***	0.012***	0.012***	0.012***	0.012***	0.017***	0.017***
	(4.79)	(4.88)	(4.63)	(4.79)	(4.63)	(4.80)	(5.96)	(6.04)
ROA	0.435***	0.430***	0.358***	0.349***	0.367***	0.358***	0.329***	0.330***
	(7.60)	(7.52)	(6.07)	(5.91)	(6.38)	(6.23)	(5.02)	(5.00)
OCF	–0.492***	–0.492***	–0.483***	–0.478***	–0.500***	–0.495***	–0.499***	–0.496***
	(–13.20)	(–13.16)	(–12.91)	(–12.73)	(–13.19)	(–13.00)	(–12.28)	(–12.16)
FIXED	0.427***	0.426***	0.427***	0.424***	0.426***	0.424***	0.416***	0.414***
	(24.79)	(24.75)	(24.81)	(24.46)	(24.80)	(24.49)	(22.44)	(22.29)
MB	–0.028***	–0.028***	–0.028***	–0.029***	–0.028***	–0.029***	–0.031***	–0.031***
	(–9.26)	(–9.29)	(–9.58)	(–9.99)	(–9.60)	(–9.99)	(–10.51)	(–10.52)
LEV_MEDIAN	0.405***	0.403***	0.404***	0.400***	0.403***	0.401***	0.389***	0.390***
	(11.04)	(10.99)	(11.03)	(10.96)	(11.03)	(10.98)	(10.24)	(10.26)
Constant	–0.181***	–0.189***	–0.144**	–0.150***	–0.141**	–0.148**	–0.237***	–0.244***
	(–3.11)	(–3.24)	(–2.47)	(–2.60)	(–2.42)	(–2.56)	(–3.87)	(–3.99)
Industry	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	8426	8426	8426	8426	8426	8426	7169	7169
R^2	0.271	0.272	0.269	0.274	0.269	0.275	0.293	0.294

Notes: Table 10 presents the empirical results of H2. All continuous variables are winsorised at 1% and 99%. The definitions of all variables are listed in Table 2. *** denotes a significance level at 1%, ** at 5%, * at 10%. T values are bracketed.

The dependent variable is LEV, the proxy of tax avoidance in columns (1) and (2) is ETR, BTD_TOTAL in columns (3) and (4), BTD_DD in columns (5) and (6), BTD_F in columns (7) and (8). D_TAXAV and LOCAL_TAXAV is not significant in the results, suggesting the level of substitution between debt financing and tax avoidance is not significantly different between SOEs and non-SOEs. Furthermore, firms in different provinces have a similar level of substitution between debt financing and tax avoidance.

In columns (2), (4), (6) and (8), D_LOCAL_TAXAV is significant and consistent with prediction. Compared with other SOEs, SOEs located in the provinces that use after-tax profit have a higher level of substitution between debt financing and tax avoidance.

5. Conclusion

This article examines how the performance evaluation of SOEs influences tax avoidance and capital structure. On one hand, performance evaluation plays a critical role in corporate governance to reduce agency cost. On the other hand, tax is also a key element in operational and financial decisions. As a result, it is an important empirical question of how the tax consideration embedded in a CEO’s compensation contract could affect operational and financial decisions.

We take the unique institutional background in China and analyse the influence of different performance measures used in different local SASACs on the tax avoidance and capital structure. Since 2004, central and local SASACs have released methods to evaluate the CEOs in SOEs under their control. We find that Shanghai, Guangdong, Zhejiang and Fujian choose after-tax profit as the annual performance measures while central SASACs and other provinces use before-tax profits. We have two major findings: (1) compared with SOEs that use before-tax profit as the evaluation index, SOEs that use after-tax profit have lower a ETR and a higher BTD; (2) compared with SOEs that use before-tax profit as the evaluation index, SOEs that use after-tax profit have a higher level of substitution between debt financing and tax avoidance. These two findings conclude that when considering the cost of tax in a performance evaluation process, firms’ operational and financial decisions will be affected.

We have two theoretical contributions. First, we test the relation between performance evaluation systems and corporate tax avoidance behaviour. We explain the high tax burden of SOEs from the compensation contract aspect. Second, our performance evaluation measures are made by local SASACs, not firms and this can mitigate the endogeneity problem to a large extent. Similar research using US data has large endogenous problems because the decision to use before-tax or after-tax profit a as performance measure is made by the firm itself.

This paper also has practical contributions. We reveal the role the compensation contract plays in corporate governance. We show that the performance evaluation system can directly influence CEO’s motivations and outcomes of the firm. In conclusion, it is of vital importance to pick up proper performance evaluation measures to promote the reform of SOEs.

Additional information

Funding

This work was supported by the Key Program of National Natural Science Foundation of China [71632006];National Natural Science Foundation of China [71272008];the Graduate Innovation Program of Shanghai University of Finance and Economics [2016110919];the MOE Project for Key Research Institutes of Humanities and Social Science in Universities [14JJD630005];the Program of the Shanghai Municipal Education Commission [2014111143];the MOE Project for Key Research Institutes of Humanities and Social Science in Universities [11JJD790008];

Notes

¹ We find that each province in our sample includes return on equity (ROE) as an evaluation index for local SOEs. Since ROE is calculated as after-tax profit divided by total assets, CEOs of SOEs in each province would care about after-tax profit. We believe that this phenomenon will not influence the argument and results in this paper because the incentive of tax avoidance driven by ROE index is the same for SOEs in each state. Our results show the marginal effect of choice between after-tax profit and before-tax profit on tax avoidance, conditional on common incentives from ROE. Furthermore, the weight on after-tax profit/before-tax profit is large enough to drive the results. Taking Guangdong province as an example, the score on ROE is 20 points while the score on after-tax profit is also 20 points. As a result, the weight on after-tax is big enough to influence the SOEs’ tax avoidance motivation.

² In August 2003, Shanghai released Opinions of Reform of the Management of SOEs ([2003] No.13). This Opinions includes six attachments, and Performance Evaluation Method for Executives in SOEs of Shanghai is one of them.

³ In September, 2007, SASAC in Zhejiang released Performance Evaluation Method for Executives in SOEs of Zhejiang and clearly makes after-tax profit a performance measure.

⁴ In general, tax saving through debt financing will not influence ETR and BTD.

⁵ Because the data of performance evaluation systems in these three provinces are missing.