Research on the Impact of Mixed Reform of State-Owned Enterprises on Enterprise Performance—Based on PSM-DID Method

Abstract

Based on the data of China Industrial Enterprise Database, this paper uses the propensity score matching double difference method (PSM-DID) to study the impact of mixed ownership reform of state-owned enterprises on enterprise performance. The study found that mixed ownership reform of state-owned enterprises can enhance the performance of enterprises. Further considering marketization, industry competition and regional characteristics, it is found that the effect of reform is heterogeneous. When the degree of marketization is high, the effect of reform on improving productivity is good, and when the degree of marketization is low, the effect of reform on reducing debt is good; the reform effect of industries with low degree of competition is better than that of industries with high degree of competition. The reform of state-owned enterprises in the eastern region has the best effect, and the reform in the central region has a better effect on reducing debt. The effect of mixed ownership reform in the western region is not significant.

1. Introduction

The reform of state-owned enterprises is a commonplace but has always attracted attention. Since the mid-1990s, the reform of state-owned enterprises has been an important part of economic system reform [1]. Ref. [2] found that about 50% of state-owned enterprises have implemented different degrees of reform by analyzing the data of state-owned enterprise reform from 1995 to 2004. The Third Plenary Session of the 18th CPC Central Committee put forward a major task is to promote the reform of mixed ownership of state-owned enterprises, clear ‘the mixed ownership economy with cross-shareholding and mutual integration of state-owned capital, collective capital and non-public capital is an important form of realization of the basic economic system’, mixed ownership reform has become the central link of China’s economic reform.

After the global financial crisis, the average TFP (total factor productivity) growth rate of state-owned enterprises is lower than that of private enterprises. At the same time, state-owned enterprises to acquire or merge private enterprises ‘State Advancement and Private Retreat’ cases increased significantly. This also leads to inconsistent changes in the asset-liability ratio of enterprises of different ownership types. State-owned enterprises increase leverage and private enterprises reduce leverage, but the overall trend of corporate leverage is very obvious. A question arises: Does the obstruction of property rights reform hinder the improvement of corporate performance? Does the mixed ownership reform help to improve the TFP of state-owned enterprises and alleviate the debt problem of state-owned enterprises, thereby improving overall corporate performance and promoting sustainable economic development?

The original intention of the reform of state-owned enterprises is to improve corporate performance, eliminate soft budget constraints, and change the dependence of state-owned enterprises on government subsidies and bank loans, thereby reducing the financial burden and reducing financial risks. If the company’s asset-liability ratio is at a high level for a long time, the financial situation is bound to deteriorate, and maintaining normal operation will be affected. Even if the company intends to carry out innovative activities, it is powerless. Only with a healthy financial situation, sustainable enterprise innovation activities can be implemented. Innovation activities improve the production efficiency of enterprises and enhance the profitability of enterprises, which in turn further improve the financial situation of enterprises, thus entering a virtuous circle. However, the current corporate performance is not optimistic. After the global financial crisis, the average TFP growth rate of state-owned enterprises is lower than that of private enterprises [3]; at the same time, the asset-liability ratio of private enterprises continued to decline from 59.7% in 2001 to 52.1% in 2014. The asset-liability ratio of state-owned enterprises and wholly state-owned companies began to rise from 2008, and climbed to 61.7% in 2012 [4]. This structural corporate debt and TFP changes will affect the sustainable development of the economy from three aspects: First, the inefficiency and waste of resources of state-owned enterprises affect the supply efficiency of public goods and services; second, the financing constraints of private enterprises inhibit the innovation capability of enterprises; third, the enterprise’s debt mismatch, increase the overall economic financing costs, increase the burden of corporate debt [5].

There is no accepted definition and calculation method for enterprise performance. In 1995, the Ministry of Finance issued an evaluation index system for enterprise economic benefits, including 10 indicators such as total asset return rate, capital return rate, asset preservation rate, sales profit rate, asset-liability ratio, current ratio, accounts receivable turnover rate, social contribution rate, and social accumulation rate to evaluate enterprise performance from three aspects: investor, creditor and social contribution.

Table 1. Research on the impact of state-owned enterprise reform on enterprise performance.

Literature	Data	Method	Performance Indicators	Mechanism	Conclusion
[6]	Chinese industrial enterprises from 2001 to 2007	PSM-DID method	TFP (total factor productivity)	The introduction of other capital activates the vitality of enterprises, improves the operating environment of enterprises, and thus has a positive effect on the TFP of enterprises.	Reform helps to improve TFP, and there are differences in the reform of state-owned enterprises under different types of holdings and competition.
[7]	2004–2014 China A-share listed companies	fixed effect model	ROA (return on assets)	The ‘complementarity’ of heterogeneous shareholders enables enterprises to exert the advantages of different ownership capital.	The ownership structure of ‘mixed ownership’ improves corporate performance, but the equity diversification between state-owned capital has no positive impact on corporate performance.
[8]	Chinese state-owned enterprises listed on A shares before 2016	Hybrid panel regression model	Earnings per share, build a comprehensive index.	Imperfect market mechanism	There is no significant correlation between the proportion of state holding and corporate performance.
[9]	China’s two-digit industrial sector classification from 2003 to 2008	linear regression model	Capital appreciation preservation rate.	The mixed reform of state-owned enterprises improves TFP and alleviates the financing constraints of private enterprises (using asset-liability ratio as an indicator), thereby improving corporate performance.	Mixed ownership reform is conducive to the improvement of enterprise performance.
[10]	Competitive state-owned mixed ownership enterprises in 2013–2017	mediating model	Tobin’Q, Value preservation and appreciation rate and market share of state-owned assets.	Equity mixing degree→high source of directors and supervisors→mixed reform performance of state-owned enterprises.	The degree of equity mixing has a significant positive impact on the performance of the mixed reform of state-owned enterprises.
[11]	2008–2015 State-owned listed companies	Tobit model	Ownership structure and top governance structure.	Supervision and restraint of state-owned major shareholders and improve the corporate governance of state-owned enterprises.	The correlation between the proportion of state-owned shares, the proportion of legal person shares and corporate performance is not significant.
[12]	State-owned listed companies in 2013–2017	multiple regression	ROA (return on assets)	equity balance	Inverted ‘U’ relationship between the degree of equity balance and the performance of state-owned enterprises.
[13]	Industrial Enterprise Data 1999–2007	DID model	Operating efficiency (no calculation index is given)	Mixed ownership reform affects policy burden, policy burden affects enterprise performance.	Mixed ownership reform can significantly improve corporate performance.
[14]	Chinese industrial enterprises from 1998 to 2008	fixed effect model	TFP, logarithm of sales output value, product sales profit margin, management cost rate, logarithm of export value.	Mixed private capital with clear property rights can significantly improve the performance of state-owned enterprises.	The entry of non-state-owned capital into state-owned enterprises can significantly improve TFP, sales scale, profit margin and export of state-owned enterprises.
[15]	Shenzhen Stock Exchange after Deducting Financial Companies 1996–1999 All A-share stocks traded.	hybrid panel regression model	Return on net assets, return on main business assets.	The reduction and privatization of state-owned shares are improved through the optimization of corporate governance.	The correlation between the proportion of state-owned shares, the proportion of corporate shares and corporate performance is not significant.
[16]	Companies listed on the GEM before 2011	GMM, simultaneous equations	Return on equity, asset-liability ratio.	mutual effect	There is indeed an interactive relationship between capital structure and corporate performance.
[17]	’Guotai’an 1998–2007 Chinese non-listed company data	DID	TFP, R&D accounted for the proportion of industrial sales revenue.	Improve the supervision and incentive mechanism of the principal to the agent.	State-owned capital is conducive to improving production efficiency rather than innovation efficiency.
[18]	Chinese industrial enterprises from 1998 to 2009	PSM-DID method	ROE	The capital operation efficiency and inventory management ability of an enterprise affect its profitability.	Whether restructuring state-owned enterprises into domestic non-state-owned enterprises or foreign-funded enterprises can promote their profitability.
[19]	listed SOEs in the A-share market in China from 2009 to 2017	Multiple Regression Analyses	earnings management	external monitoring	A significant and positive relationship between state ownership and earnings management.
This paper	Chinese industrial enterprises from 1998 to 2007	PSM-DID method	TTP, LEV, ROA	Market-oriented institutional reform and the complementarity of heterogeneous equity to improve corporate performance.	The mixed ownership reform of state-owned enterprises can improve the performance of enterprises. Further considering marketization, industry competition and regional characteristics, it is found that the effect of the reform is heterogeneous.

summarizes the research on the impact of state-owned enterprise reform on enterprise performance. The indicators of enterprise performance in the literature are not uniform. The factors of total factor productivity change can be essentially classified into two categories: technological progress and organizational management. It is a comprehensive indicator. The asset-liability ratio (LEV) is a comprehensive indicator to measure all aspects of the company (agency cost, financial constraints, creditor power, etc.). These two comprehensive indicators can better measure the performance of an enterprise. Table 1 shows that many articles use the above two indicators. At the same time, it is found that many articles apply the single indicator of ROA (return on assets). This paper uses three indicators of total factor productivity (TFP), asset-liability ratio (LEV) and return on assets (ROA) as the proxy scalar of enterprise performance, and more comprehensively examines the impact of mixed reform of state-owned enterprises on enterprise performance.

The existing literature in Table 1 is mostly analyzed from the perspective of privatization of state-owned enterprises. Due to the lack of data, there are not many research literatures from the perspective of mixed ownership reform of state-owned enterprises. This paper constructs a multi-dimensional index of mixed ownership reform of state-owned enterprises. On the basis of discussing privatization, privatization is subdivided into complete privatization reform of state-owned enterprises, mixed ownership reform of non-state-owned holding type and mixed ownership reform of state-owned holding type. At present, there is a relative lack of research on heterogeneity in the literature. There are huge differences in China’s domestic economic development. Regional marketization, industry competition, and regional differences between the East and the West are very large. These factors will affect the effect of mixed reform of state-owned enterprises. The discussion of subdivision and heterogeneity helps to improve the reliability and accuracy of the estimation results. Most of the research methods and designs have the problem of estimation bias. This paper uses the propensity score matching difference-in-difference method to alleviate the problem of estimation bias to some extent.

Based on the above analysis, the innovation of this paper is reflected in the following three aspects. First, the reform effect is evaluated from the three enterprise performance indicators of total factor productivity, asset-liability ratio and return on assets, making the evaluation more comprehensive and robust. Second, the propensity score matching double difference method is used to empirically test the impact of reform on corporate performance to reduce the estimation bias caused by endogenous problems, and the heterogeneity of different ways of reform and economic environment is considered in the estimation process, which helps to improve the reliability and accuracy of the estimation results. Third, although the asset-liability ratio is used as a performance indicator of the enterprise, few literatures use this enterprise performance indicator for research. In the study of the reform of state-owned enterprises, the effect of the reform of state-owned enterprises on the asset-liability ratio of enterprises is ignored to varying degrees. Under the background of the rising asset-liability ratio of enterprises, it is particularly important to study the deleveraging of enterprises, which enriches the academic research in related fields such as the mixed ownership reform of state-owned enterprises.

This paper is arranged as follows: the Section 2 is the literature review, the Section 3 is the Research Hypothesis and Theoretical Framework, puts forward the research hypothesis, introduces the double difference method model based on propensity score matching; The Section 4 is the database and variable description; the Section 5 is the sample statistical description; the Section 6 is the empirical results and analysis; the Section 7 is a summary, which introduces the research conclusions, related defects and future research.

2. Literature Review

Some preliminary studies on the reform of state-owned enterprises showed that the reform of state-owned enterprises can improve corporate profitability and improve corporate performance [20,21,22]. By summarizing the literature research from Table 1, it can be found that the existing research has the following deficiencies. First, most of the existing literature is analyzed from the perspective of privatization of state-owned enterprises, and the research literature from the perspective of mixed ownership reform of state-owned enterprises is relatively lacking. After the Third Plenary Session of the 18th CPC Central Committee, some scholars have also carried out research on the theme of mixed ownership reform. For example, Ref. [23] used the difference-in-difference method to study the impact of mixed ownership reform on the policy burden of state-owned enterprises, but did not study the impact of enterprise production efficiency. Ref. [6] analyzed the impact of mixed ownership reform of state-owned enterprises on total factor productivity, but did not study the impact of corporate debt. Second, in terms of methods, most of the analysis fails to achieve the matching of reformed enterprises, which will lead to estimation bias. Third, most of the performance indicators use a single indicator, and the representativeness of corporate performance needs to be strengthened. Fourth, the research on the reform of state-owned enterprises ignores the effect of reform on the asset-liability ratio of enterprises to varying degrees. Under the background of the rising asset-liability ratio of enterprises, it is particularly important to study the deleveraging of enterprises.

Refs. [24,25] discusses the relationship between the mixed reform of state-owned enterprises and social responsibility. Ref. [26] discusses the relationship between senior management and the effect of SOE reform. Ref. [27] discusses the relationship between the reform of state-owned enterprises and the efficiency of financial investment. Ref. [28] studied the reform of state-owned enterprises and enterprise innovation.

Further combing the existing research literature on the effect of state-owned enterprise reform, we find that although there are many literatures on the effect of state-owned enterprise reform, these studies have not reached a consistent conclusion [29,30,31,32,33,34,35,36,37,38]. There are three main reasons: (1) the estimation model is endogenous. The endogeneity of the model mainly comes from two aspects: one is the two-way causal relationship. Enterprise reform will affect enterprise performance, and enterprise performance will also affect reform decision-making. The study found that there is a phenomenon of beautiful women marrying first [39]. Second, there are missing variables in the model estimation. For example, the province or industry of the enterprise and the unobservable characteristics of the enterprise may affect the reform decision and performance of the enterprise at the same time [40]. (2) Most of the existing studies only consider the reform of state-owned enterprises into non-state-owned enterprises, and do not further distinguish the specific forms of ownership. In fact, there are three ways of reform, namely, complete privatization reform, non-state-controlled mixed ownership reform, and state-controlled mixed ownership reform. There may be huge differences in corporate culture, corporate governance structure and debt situation among enterprises with different reform modes. Without distinguishing specific reform modes, generalizing reforms as non-state-owned enterprises may cause estimation bias. (3) Improper sample matching results in the inability to effectively identify the effects of corporate reforms [41,42]. However, most of the above three questions have not received enough attention in previous studies, but the answers to the questions are of great significance for further deepening and improving the mechanism design of state-owned enterprise reform. Most of the existing literature has not conducted in-depth research on the above issues. In general, the research on the impact of mixed ownership reform on corporate performance needs to be strengthened.

3. Research Hypothesis and Theoretical Framework

3.1. Research Hypothesis

According to [7], there is no empirical literature systematically supporting the positive effect of ‘mixed ownership’ ownership structure on corporate performance in previous studies on ‘mixed ownership’. Ref. [8] found that in the process of mixed ownership reform of state-controlled enterprises, the property rights reform introduced by private capital did not play a substantial role in improving corporate performance. Simple privatization cannot improve corporate performance. The key factor that can really improve corporate performance comes from competition within the market. Ref. [6] found that the reform of state-owned enterprises can significantly improve total factor productivity (TFP), and different holding types of enterprises have different effects on mixed ownership reform, and will be affected by the strength of industry competition. Compared with monopoly industries, the mixed ownership reform of state-owned enterprises in competitive industries has a greater effect on the improvement of enterprise efficiency. The empirical results of [18] show that the entry of non-state-owned capital into state-owned enterprises can significantly improve the total factor productivity of state-owned enterprises regardless of whether its proportion reaches absolute shareholding. Clear private capital can significantly improve the performance of state-owned enterprises, and capital with advanced technology and management experience can promote the significant improvement of enterprise productivity. Mixed foreign capital can promote a significant increase in exports. Ref. [9] show that the asset-liability ratio of state-owned enterprises is higher than that of joint-stock enterprises and private enterprises, but the capital appreciation and preservation rate of private enterprises is significantly higher than that of state-owned enterprises. Mixed ownership reform can alleviate the financing constraints of private enterprises and improve the total factor productivity of state-owned enterprises, improve enterprise performance. Ref. [13] found that mixed ownership reform can significantly improve enterprise performance by reducing the policy burden of state-owned enterprises, but regional and industry competition factors have heterogeneous effects on the mixed reform of state-owned enterprises. The policy burden of enterprises and monopoly enterprises in the eastern region has the most obvious decline.

The impact of state-owned and private participation on corporate performance is ‘complementary’. State-owned and private shareholders can make up for the shortcomings of heterogeneous controlling shareholders and improve corporate performance. This ‘complementarity’ means that ‘mixed ownership’ is a better ownership structure. Through the balance of non-state-owned shareholders, the mixed reform of state-owned enterprises can improve enterprise management, improve the total factor productivity and asset return rate of state-owned enterprises, and reduce the asset-liability ratio of state-owned enterprises; the existing literature has found that different regional cultures, marketization and industry competition have heterogeneous effects on the mixed reform effect of state-owned enterprises. The following research hypotheses can be made:

H1. 

The mixed reform of state-owned enterprises improves the total factor productivity and asset return rate of reformed enterprises.

H2. 

The mixed reform of state-owned enterprises reduces the asset-liability ratio of reformed enterprises.

H3. 

The mixed reform of state-owned enterprises has heterogeneous effects on marketization, industry competition and region.

3.2. Difference-in-Differences Model Based on Propensity Score Matching

This paper mainly studies the changes of enterprise performance after the reform of state-owned enterprises. T is a dummy variable, if the enterprise has never carried out the reform of state-owned enterprises, then T = 0; if the enterprise has experienced the reform of state-owned enterprises, then T = 1. $y_{it}$ is the performance of enterprise i in year t, which is represented by the productivity ($TF{P}_{it}$) and asset-liability ratio ($LE{V}_{it}$) of enterprise i in year t. $y_i^1$ indicates the potential performance of the enterprise after the restructuring, and accordingly, $y_i^0$ indicates the potential performance of the enterprise if there is no restructuring. Therefore, the average impact of corporate reform on corporate performance can be expressed as: $E\left(y_i^1-y_i^0|T=1\right)$. Where $y_i^0$ and $y_i^1$ cannot be observed simultaneously for the same enterprise. The Difference-in-Difference (DID) method provides a way to identify the average treatment effect of enterprise reform. The identification of the difference-in-difference method requires parallel assumptions that $E\left(\Delta y_i^0|T=0\right),E\left(\Delta y_i^0|T=1\right)$ is comparable. However, the general $E\left(\Delta y_i^0|T=0\right)$ is not equal to $E\left(\Delta y_i^0|T=1\right)$, because whether the reform of state-owned enterprises is non-random, there is a ‘beautiful woman marry first’ phenomenon [22]; differences before the reform of state-owned enterprises will also cause differences in the development trend of reformed enterprises and non-reform enterprises; compared with monopoly industries, enterprises in competitive industries are more motivated to reform. Therefore, if directly using $E\left(\Delta y_i^0|T=0\right)$ to replace $E\left(\Delta y_i^0|T=1\right)$ in Equation (1) will undoubtedly cause estimation errors. Therefore, choosing an effective control group for reform enterprises is the key to identify the effect of state-owned enterprise reform, rather than simply using all non-reform enterprises as a control group. The matching method is to find the unreformed enterprises which are close to the nature of the reformed enterprises from the control group, and the possibility of reform is close to the possibility of reform of state-owned enterprises to eliminate selection bias. Here we use the Propensity Score Matching method to find firms that are most similar to the reformed firms from those that have never been reformed during the sample period to ensure that the control group (non-reformed firms) and the experimental group (reformed firms) have the same trend of change, namely:

$$E\left(\Delta y_i^0|T=0\right)=E\left(\Delta y_i^0|T=1\right)$$

(1)

We match the experimental group samples with the control group samples, as follows.

First, use the logit model to estimate the propensity score for corporate reform:

$$P=P\left(T=1|X\right)$$

(2)

Among them, P is the probability of state-owned enterprise reform. T = 1 represents the experimental group enterprises, X represents the factors affecting enterprise reform. Referring to [43,44], this paper selects the following matching variables: wage level lwage, current assets llas, fixed assets gdas, number of employees labor labor, enterprise age age, exit (export is recorded as 1, otherwise recorded as 0), enterprise size size, agency cost agentcost, paid-in capital lk and the industry in which the enterprise is located. According to this equation, the predicted probability value of each state-owned enterprise reform can be calculated, and then a single nearest neighbor matching method is used to match the samples according to the propensity score.

Based on the matched enterprise samples, this paper constructs the dummy variable $treate{d}_i$ of the reform. The enterprise samples that have undergone restructuring during the sample period take 1, and the enterprise samples that have never undergone restructuring during the sample period take 0. At the same time, the time dummy variable $tim{e}_t$ is defined. For the sample of the experimental group, the period before the mixed ownership reform is 0, and the period after is 1. For the reference group samples, $tim{e}_t$ takes the same value as the matched experimental group samples. On this basis, the following difference-in-difference model can be constructed:

$$TF{P}_{it}={\alpha }_0+{\alpha }_{1}treate{d}_i\times tim{e}_t+{\alpha }_{2}treate{d}_i+{\alpha }_{3}tim{e}_t+{\gamma }_T^{\prime }control{s}_{it}+{\nu }_i+{\mu }_t+{\epsilon }_{it}$$

(3)

$$LE{V}_{it}={\beta }_0+{\beta }_{1}treate{d}_i\times tim{e}_t+{\beta }_{2}treate{d}_i+{\beta }_{3}tim{e}_t+{{\gamma }^{\prime }}_{L}control{s}_{it}+{\nu }_i+{\mu }_t+{\epsilon }_{it}$$

(4)

$$RO{A}_{it}={\gamma }_0+{\gamma }_{1}treate{d}_i\times tim{e}_t+{\gamma }_{2}treate{d}_i+{\gamma }_{3}tim{e}_t+{{\gamma }^{\prime }}_{L}control{s}_{it}+{\nu }_i+{\mu }_t+{\epsilon }_{it}$$

(5)

Among them, $v_i$ is the individual fixed effect, ${\mu }_t$ is the time fixed effect, including size, age, agentcost, labor, lwage, llas, gdas, lk, exit variables. According to the basic idea of difference in difference, the interaction term $treate{d}_i\times tim{e}_t$ coefficient ${\alpha }_1$, ${\beta }_1$, ${\gamma }_1$ respectively reflect the processing effect of total factor productivity, asset-liability ratio and return on assets before and after the reform of state-owned enterprises, which is the main parameter estimated and explained in this paper.

4. Database and Variable Description

4.1. Database Description

This paper combines the micro data of industrial enterprises of the National Bureau of Statistics (1998–2007) and the provincial marketization index of China calculated by [45]. The China Industrial Enterprise Database is the data obtained from the survey of non-state-owned enterprises and all state-owned enterprises with annual sales revenue of more than 5 million yuan by the National Bureau of Statistics every year. The database continuously counts the basic information of each enterprise, such as the province and county code and industry category of the enterprise, the legal person code and registration type of the enterprise, etc., and also includes financial indicators, such as the information of paid-in capital, industrial sales output value, annual average number of employees, total industrial output value, sales volume, fixed assets, annual average balance of net fixed assets, cumulative depreciation, total payable wages, total liabilities, total assets, and current capital. Remove the state capital, collective capital, corporate capital, personal capital, Hong Kong, Macao and Taiwan capital, foreign capital missing enterprises, delete the enterprise’s sales income is less than 0 enterprises. In this paper, the average wage level lwage, current assets llas, fixed assets gdas, employment labor labor, enterprise age age, enterprise asset size size, agent cost of state-owned enterprises agentcost, enterprise paid-in capital lk, total factor productivity TFP, asset-liability ratio LEV were truncated by 1% at both ends, the enterprises with repeated intermediate reforms were deleted, the non-state-owned enterprises were deleted, and the enterprises with only one year were deleted. Finally, 23,740 enterprises were included, with a total of 98,948 samples.

4.2. Total Factor Productivity (TFP) and Asset-Liability Ratio (LEV)

The core variables include total factor productivity (TFP) and asset-liability ratio (LEV). TFP is an indicator of productivity. It is the ‘residual value’ after removing the input of labor, capital and other factors, which can truly reflect the efficiency essence of individual enterprises. The asset-liability ratio reflects how much proportion of total assets is financed by borrowing. It is a comprehensive index to evaluate the debt level of enterprises. It reflects the borrowing ability of different enterprises on the side. Corresponding to the leverage ratio, it essentially reflects the proportion of equity capital and debt. Ref. [1] used the asset-liability ratio as a proxy indicator for the efficiency of capital operations, arguing that an increase in the efficiency of capital operations can reduce the asset-liability ratio, thereby reducing corporate interest expenses and increasing profits to reduce dependence on external financing of enterprises. How to accurately estimate the total factor productivity of enterprises is the key to this article. Ref. [46] used China’s industrial enterprise data to compare and analyze various methods of calculating the total factor productivity of China’s industrial enterprises, and found that the ordinary least squares (OLS) residual estimation of TFP will produce simultaneity bias and sample selection bias. Olley and Pakes (1996)’ s semi-parametric method is usually considered to estimate the production function at the two-digit industry level, and then the consistent estimator of total factor productivity at the enterprise level is obtained [47]. However, the problem of the OP method is that the investment needs to be greater than zero. Ref. [48] proposed to replace investment with intermediate investment products, which solves the situation where more investments are zero. Based on this, this paper estimates the total factor productivity of micro-enterprises by referring to the modified OP method of [49,50]. The modified OP method draws on the LP idea and uses the logarithmic form of actual intermediate input as a proxy variable for productivity shocks that the modeler cannot observe and the entrepreneur knows. This can not only solve the problem of simultaneity bias, but also reduce the data deletion problem caused by the lack of corporate investment data. The modified OP method solves the selectivity bias by calculating the exit probability of the enterprise. TFP_LP represents the TFP estimated by LP method, and TFP_OP represents the TFP estimated by modified OP method.

4.3. Marketization, Industry Competition Variables

In the industrial organization theory, the Herfindahl-Hirschman Index (HHI) is commonly used to describe the degree of competition in the industry in which the enterprise is located. The Herfindahl-Hirschman Index of industry i is $hh{i}_{}={\displaystyle \sum _{j=1}^{n_i}{\left(total\_sal{e}_{ij}/{\displaystyle \sum _{j=1}^{n_i}total\_sal{e}_{ij}}\right)}^2}$, where $n_i$ is the number of enterprises in industry i, enterprise j belongs to industry i, and the total sales income of the enterprise is $total\_sal{e}_{ij}$. The higher the indicator, the higher the degree of industry monopoly, that is, a small number of enterprises in the industry’s market share is relatively high, fully competitive hhi is 0, and fully monopolized hhi close to 1. The index reflects the survival environment of the industry, will have an important impact on production decisions. However, the marketization index has a broader meaning, reflecting the general environment of enterprise survival, and the marketization index measures the relative process of marketization transformation. Ref. [45] constructed a relative index (marketization index) to measure the process of market-oriented transformation from five aspects: the relationship between government and market, the development of non-state-owned economy, the development of product market, the development of factor market, market intermediary organizations and legal system environment. Since the marketization index quantifies the institutional variables, it is possible to quantitatively examine the economic significance of institutional factors. Figure 1 shows that from the time dimension, the overall degree of marketization is rising; from the regional perspective, the developed coastal areas not only have a high degree of marketization, but also the process is fast; central inland areas followed; the western region is slower.

4.4. Reform of Dummy Variables (Treated)

Referring to the treatment methods of [6,19], the entry of non-state-owned capital (personal capital, collective capital, legal person capital, Hong Kong, Macao and Taiwan capital and foreign capital) into state-owned enterprises is defined as the reform of state-owned enterprises. In view of the effect of the reform of state-owned enterprises, we define the enterprises with 100% state-owned capital as state-owned enterprises. Therefore, if the proportion of state-owned capital of state-owned enterprises is less than 100% during the investigation period, it is considered that the enterprise has undergone reform. If the state-owned enterprises have reformed but the proportion of state-owned capital is not 0, it is considered that the state-owned enterprises have undergone mixed ownership reform.

4.5. Matching Variables

Drawing on the research of [34,37], the matching variables and definitions selected in this paper are listed in

Table 2. Definition and description of variables.

Classification	Variable Name	Code	Explanation
kernel variable	total factor productivity	TFP	The part of output exceeding the growth rate of factor input
	return on assets *	ROA	Total pre-tax profit/total assets
	assets-liability ratio	LEV	Total liabilities/total assets
market circumstances	marketization index	market	Fan Gang (2011)’s China Marketization Index, Reflecting the Relative Process of Regional Marketization
	herfindahl-hirschman index	hhi	Sum of squares of the percentage of total assets, reflecting the degree of market competition (in two-digit industry codes)
Reform variables	Reform dummy variables	treated	Reformed enterprise 1, control group 0
	time virtual variables	time	0 before enterprise reform, 1 after reform
key indicators	interaction	treated × time	Reform the interaction term between dummy variable and time dummy variable
Matching variable	Wage level	lwage	The logarithm of the ratio of total wages payable to employment
	Fixed assets	gdas	logarithm of fixed assets
	Current assets	llas	Value of current assets
	Enterprise size	size	The logarithm of total assets
	Age of survival	age	The logarithm of the difference between the establishment time and the reform year
	Agency cost	agentcost	Ratio of management costs to total assets
	Number of persons employed	labor	The logarithm of the number of employees
	Enterprise exports	exit	Export recorded as 1, otherwise recorded as 0
	paid-in capital	lk	The logarithm of paid-in capital
	subsidiary earnings *	sub	log (1 + subsidiary earnings)
	return on equity *	roe	After-tax profit/net assets
	Debt equity ratio *	der	Total liabilities/total owner equity

Note: The variable marked with * is used in the robustness test.

.

5. Statistical Description

Among the sample of state-owned enterprises that reformed during the sample period, there were 751 reformed enterprises in 1999, 869 reformed enterprises in 2000, 583 reformed enterprises in 2001, 445 reformed enterprises in 2002, 465 reformed enterprises in 2003, 254 reformed enterprises in 2004, 259 reformed enterprises in 2005, 223 reformed enterprises in 2006, 509 reformed enterprises in 2007. A total of 4358 enterprises were reformed, accounting for 18.36% of the total sample enterprises. Among them, 1837 enterprises implemented mixed ownership reform, accounting for 42.15% of the number of reformed enterprises.

In

Table 3. Statistical description of variables.

Variable	Sample Size	Mean	Standard Deviation	Median	Minimum	Maximum
TFP_op	56,740	3.690	0.850	3.630	2.040	6.410
TFP_lp	56,740	3.770	0.860	3.720	2.110	6.500
ROA	56,740	0	0.0600	0	−0.180	0.770
LEV	56,740	0.700	0.290	0.710	0.0100	1.470
size	56,740	9.800	1.440	9.760	6.630	13.63
age	56,740	25.25	15.26	27	0	51
agentcost	56,740	0.0700	0.0700	0.0600	0	0.550
labor	56,740	5.030	1.200	5	2.300	7.830
lwage	56,740	1.870	0.720	1.860	0.0100	4.180
llas	56,740	9	1.480	8.980	5.540	13.02
gdas	56,740	8.770	1.640	8.800	4.170	12.69
lk	56,740	8.060	1.520	8.010	4.410	12.31
exit	56,740	0.270	0.450	0	0	1
roe	56,740	2.550	6.830	1.650	−20.79	44.51
der	56,740	0.780	1.990	0	0	8.080
sub	56,740	0.700	0.290	0.710	0.0100	1.470

, the descriptive statistics of each variable under the missing samples of the deleted variables are reported. There are differences in the mean values of TFP_op and TFP_lp. The mean and median values of ROA are zero. From the perspective of return on assets, it can be seen that both good and bad enterprises have, and generally not very ideal. The average LEV is 0.7, and the overall debt of the enterprise is high.

Table 4. Core variable mean test.

Variable	Enterprise	Sample	Mean	Standard Error	Standard Deviation	T Value	Pr (|T| > |t|)
TFP_op	Unreformed	68,828	3.622	0.003	0.899
	Reform	18,062	3.956	0.007	0.900
	diff		−0.334	0.008		−44.376	0.000
TFP_lp	Unreformed	68,761	3.707	0.003	0.906
	Reform	18,056	4.045	0.007	0.907
	diff		−0.338	0.008		−44.683	0.000
ROA	Unreformed	77,145	−0.002	0.000	0.068
	Reform	18,845	0.007	0.000	0.068
	diff		0.001	−0.010		−15.860	0.000
LEV	Unreformed	73,579	0.725	0.001	0.310
	Reform	18,450	0.699	0.002	0.278
	diff			0.002		11.211	0.000

compares the differences between the reform group and the control group for each variable. As shown in Table 4, the total factor productivity of enterprises implementing reform is higher than that of enterprises without reform. The reformed enterprises have higher return on assets than the unreformed enterprises. In terms of asset-liability ratio, enterprises in the reform group are smaller than those in the non-reform group. The preliminary results show that the performance of the reform group is better than that of the non-reform group. This is consistent with previous studies that have found that more efficient state-owned enterprises are more willing to choose privatization [33], and state-owned enterprises with better performance are preferentially privatized [22]. Therefore, whether the performance reflected in Table 4 is just ‘beautiful women marry first’, and whether the reform of state-owned enterprises will promote the improvement of enterprise performance, further analysis is needed.

6. Empirical Results and Analysis of Mixed Reform of State-Owned Enterprises and Enterprise Performance

6.1. Matching Results of Propensity Score

Ref. [51] analyzed and introduced the propensity score matching method in detail. The steps of implementing PSM in STATA are as follows: 1. select the covariates for matching, 2. generate random seeds, 3. generate random numbers, 4. randomize the database, 5. use the psmatch2 command to match, 6. test the balance of covariates between the treatment group and the control group. This paper chooses the non-substitutive one-to-one matching algorithm in the propensity score matching, that is, firstly, based on the observable matching variables, the prediction probability of enterprise reform is calculated, and then the only sample of unreformed state-owned enterprises closest to the sample of state-owned enterprises undergoing reform is found. In the samples examined in this paper, there are 13,698 observations in the experimental group and 27,396 observations in the total sample after matching. Reliable matching should satisfy the balance hypothesis of score matching, so we first need to test whether the distribution of each variable in the treatment group and the control group becomes balanced after matching. If there is no significant difference in the matching variables after matching, it indicates that the matching method is effective. The results of the test in

Table 5. Balance test of sample matching in 2003.

		Mean		Standard Deviation	Standard Deviation Reduction	Statistic	Associated Probability
Variable	Treatment	Treatment Group	Control Group	(%)	(%)	T Statistic	p > |t|
size	Before matching	10.227	9.754	33.2		34.01	0.000
	After matching	10.215	10.217	−0.1	99.6	−0.11	0.912
age	Before matching	22.147	26.265	−26.7		−27.71	0.000
	After matching	22.271	22.27	0.0	100.0	0.01	0.995
agentcost	Before matching	0.075	0.077	−2.7		−2.69	0.007
	After matching	0.075	0.075	0.5	82.5	0.39	0.698
labor	Before matching	5.353	4.997	30.0		30.83	0.000
	After matching	5.345	5.342	0.2	99.3	0.18	0.856
lwage	Before matching	1.968	1.833	17.4		17.60	0.000
	After matching	1.962	1.969	−0.9	94.6	−0.77	0.442
llas	Before matching	9.431	8.951	32.8		33.58	0.000
	After matching	9.418	9.418	0.0	100.0	0.01	0.995
gdas	Before matching	9.155	8.738	25.6		27.07	0.000
	After matching	9.148	9.151	−0.2	99.2	−0.17	0.866
lk	Before matching	8.449	7.976	30.8		32.08	0.000
	After matching	8.435	8.438	−0.2	99.4	−0.16	0.877
exit	Before matching	0.288	0.266	4.9		5.03	0.000
	After matching	0.287	0.283	1.0	78.8	0.84	0.402

show that after propensity score matching, there is no significant difference in each matching variable, indicating that matching effectively controls the potential impact of differences in the nature of enterprises.

6.2. Regression Results and Analysis of State-owned Enterprise Reform

On the basis of propensity score matching estimation, this paper obtains a control group with similar characteristics to the treatment group. The two groups include 26,938 enterprises. Combined with the previously collated and calculated enterprise TFP data, for the matched sample and using the difference-in-difference method to analyze the impact of SOE reform on total factor productivity (TFP) and asset-liability ratio (LEV). TFP_LP represents the TFP estimated by LP method, and TFP_OP represents the TFP estimated by modified OP method. In the empirical analysis, TFP_op is used for analysis, and TFP_lp is used for robustness test. The least squares linear (OLS) regression model ignores the characteristics of enterprises that do not change with time, resulting in the endogeneity of the estimation model and the inability to obtain a consistent estimation. The basic assumption of the mixed regression is that there is no individual effect. Studies have shown that the province or industry of the enterprise and the unobservable characteristics of the enterprise will lead to endogeneity [6,17,34]. F test in fixed effect regression results found that there is a highly significant fixed effect, so we use fixed effect regression analysis in this paper.

Preliminary regression results are reported in

Table 6. Preliminary regression results.

	(1)	(2)	(3)	(4)	(5)	(6)
	TFP_op	TFP_lp	TFP_op	TFP_lp	LEV	LEV
treated × time	0.061 ***	0.062 ***	0.052 ***	0.053 ***	−0.001	−0.017 ***
	(0.018)	(0.018)	(0.018)	(0.018)	(0.006)	(0.006)
time	0.016	0.016	0.001	0.001	−0.005	−0.001
	(0.012)	(0.012)	(0.013)	(0.013)	(0.004)	(0.004)
treated	0.064 ***	0.064 ***	/	/	−0.019 ***	/
	(0.011)	(0.011)	/	/	(0.004)	/
size	0.218 ***	0.220 ***	0.211 ***	0.213 ***	0.051 ***	0.044 ***
	(0.016)	(0.016)	(0.024)	(0.024)	(0.006)	(0.007)
age	−0.005 ***	−0.005 ***	−0.000	−0.000	0.001 ***	0.001 ***
	(0.000)	(0.000)	(0.001)	(0.001)	(0.000)	(0.000)
agentcost	1.704 ***	1.712 ***	1.465 ***	1.470 ***	−0.013	0.170 ***
	(0.078)	(0.078)	(0.106)	(0.106)	(0.028)	(0.032)
labor	0.095 ***	0.098 ***	0.148 ***	0.150 ***	0.007 ***	−0.005
	(0.007)	(0.007)	(0.013)	(0.013)	(0.003)	(0.004)
lwage	0.094***	0.096 ***	0.082 ***	0.083 ***	−0.069 ***	−0.016 ***
	(0.009)	(0.009)	(0.011)	(0.011)	(0.003)	(0.003)
llas	0.150 ***	0.150 ***	0.064 ***	0.063 ***	0.057 ***	0.041 ***
	(0.010)	(0.010)	(0.014)	(0.014)	(0.003)	(0.004)
gdas	−0.120 ***	−0.116 ***	−0.088 ***	−0.085 ***	0.006 **	0.002
	(0.008)	(0.008)	(0.011)	(0.011)	(0.003)	(0.003)
lk	−0.031 ***	−0.030 ***	−0.010	−0.010	−0.111 ***	−0.081 ***
	(0.005)	(0.005)	(0.007)	(0.007)	(0.002)	(0.002)
exit	−0.238 ***	−0.238 ***	−0.112 ***	−0.112 ***	0.035 ***	0.003
	(0.010)	(0.010)	(0.011)	(0.011)	(0.004)	(0.003)
constant term	1.045 ***	1.045 ***	0.820 ***	0.862 ***	0.602 ***	0.571 ***
	(0.049)	(0.049)	(0.167)	(0.167)	(0.017)	(0.051)
Year dummy variable	yes	yes	yes	yes	yes	yes
Region dummy variable	yes	yes	yes	yes	yes	yes
Industry dummy variable	yes	yes	yes	yes	yes	yes
N	26,892	26,892	26,892	26,892	26,892	26,892
r2/r2_w	0.337	0.346	0.073	0.074	0.253	0.111
r2_b			0.219	0.229		0.033
r2_o			0.226	0.235		0.036
F			3.45	3.45		5.12

Note: Standard error in parentheses; **, *** represent the significance level of 5% and 1% respectively.

, (1) is the ordinary least squares regression result of TFP_op, (2) is the ordinary least squares regression result of TFP_lp, (3) is the fixed regression result of TFP_op, (4) is the fixed regression result of TFP_lp, (5) is the ordinary least squares regression result of LEV, (6) is the fixed regression result of LEV. First of all, we examine the impact of state-owned enterprise reform on the total factor productivity (TFP) of enterprises. After controlling other influencing factors, whether it is mixed regression or fixed effect regression, the cross-term is significantly positive at the 1% level, indicating that state-owned enterprise reform will promote the improvement of enterprise TFP. The scale of enterprises is significantly positive at the 1% level. The larger the scale of enterprises, the higher the production efficiency. The coefficient of enterprise survival age is not significant. Ref. [52] pointed out that for enterprises in developing countries, there is no sufficient evidence to prove the relationship between enterprise survival age and enterprise productivity. In addition, the average wage level, current assets, agency costs, employment coefficient is significantly positive, fixed assets, exports coefficient is significantly negative.

Then, we examine the impact of the reform of state-owned enterprises on the asset-liability ratio (LEV) of enterprises. After controlling other influencing factors and the fixed effect of enterprises, the cross term is significantly negative at the 1% level, indicating that the reform of state-owned enterprises will reduce the LEV of enterprises. The size of the enterprise is significantly positive at the 1% level. The larger the size of the enterprise, the higher the credit and the easier the financing. The coefficient of enterprise survival age is significantly positive at the 1% level, which may be due to the increase of LEV caused by the increase of policy burden borne by enterprises with the extension of survival time. In addition, the coefficients of current assets and agency costs are significantly positive, and the coefficients of average wage level and paid-in capital are significantly negative.

Based on the above analysis, the reform of state-owned enterprises can not only improve the total factor productivity (TFP), but also reduce the asset-liability ratio (LEV). The rise and fall brought about by the reform of state-owned enterprises will be conducive to the improvement of production efficiency and reduction of debt, and thus improvement of corporate performance.

The research of [6,34] shows that the reform of state-owned enterprises and the degree of marketization are related to industry competition. Ref. [17] shows that the effect of the reform of state-owned enterprises is sensitive to the characteristics of the industry and the eastern, central and western regions. Therefore, we examined the effects of different degrees of marketization (

Table 7. Effect of state-owned enterprise reform with different marketization degrees.

	(1)	(2)	(3)	(4)	(5)	(6)
Marketization Degree	High	Low	High	Low	High	Low
	TFP_op	TFP_op	TFP_lp	TFP_lp	LEV	LEV
treated × time	0.075 ***	0.040	0.076 ***	0.040	−0.004	−0.019 **
	(0.028)	(0.028)	(0.028)	(0.028)	(0.008)	(0.008)
constant term	0.632 ***	1.753 ***	0.674 ***	1.799 ***	0.630 ***	0.579 ***
	(0.238)	(0.301)	(0.238)	(0.301)	(0.071)	(0.090)
Year dummy variable	yes	yes	yes	yes	yes	yes
Region dummy variable	yes	yes	yes	yes	yes	yes
Industry dummy variable	yes	yes	yes	yes	yes	yes
N	13,513	13,379	13,513	13,379	13,513	13,379
r2_w	0.088	0.046	0.089	0.047	0.098	0.137
r2_b	0.236	0.180	0.247	0.187	0.097	0.030
r2_o	0.251	0.177	0.262	0.183	0.102	0.030
F	3.17	3.51	3.17	3.50	5.05	5.10

Note: Standard error in parentheses; **, *** represent the significance level of 5% and 1% respectively.

), industry competition (

Table 8. The reform effect of state-owned enterprises with different industry competition degree.

	(1)	(2)	(3)	(4)	(5)	(6)
Industry Competition Degree	High	Low	High	Low	High	Low
	TFP_op	TFP_op	TFP_lp	TFP_lp	LEV	LEV
treated × time	0.047 *	0.081 ***	0.048 *	0.082 ***	−0.006	−0.022 ***
	(0.028)	(0.026)	(0.028)	(0.026)	(0.008)	(0.008)
constant term	1.395 ***	0.710 ***	1.443 ***	0.751 ***	0.546 ***	0.492 ***
	(0.252)	(0.273)	(0.252)	(0.273)	(0.075)	(0.084)
Year dummy variable	yes	yes	yes	yes	yes	yes
Region dummy variable	yes	yes	yes	yes	yes	yes
Industry dummy variable	yes	yes	yes	yes	yes	yes
N	13,650	13,242	13,650	13,242	13,650	13,242
r2_w	0.053	0.098	0.054	0.100	0.108	0.123
r2_b	0.151	0.253	0.162	0.264	0.144	0.088
r2_o	0.158	0.261	0.168	0.272	0.152	0.097
F	3.11	3.52	3.10	3.52	4.59	5.06

Note: Standard error in parentheses; *, *** represent the significance level of 10% and 1% respectively.

) and eastern, central and western regions (

Table 9. Effect of SOE reform in different regions.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Region	Eastern	Central	Western	Eastern	Central	Western	Eastern	Central	Western
	TFP_op	TFP_op	TFP_op	TFP_lp	TFP_lp	TFP_lp	LEV	LEV	LEV
treated × time	0.078 ***	−0.009	0.076 **	0.079 ***	−0.008	0.076 **	−0.011	−0.033 ***	−0.004
	(0.027)	(0.036)	(0.034)	(0.027)	(0.036)	(0.034)	(0.008)	(0.010)	(0.011)
constant term	0.806 ***	0.681 **	1.513 ***	0.848 ***	0.730 **	1.549 ***	0.532 ***	0.462 ***	0.625 ***
	(0.231)	(0.307)	(0.403)	(0.231)	(0.307)	(0.403)	(0.068)	(0.090)	(0.136)
Year dummy variable	yes	yes	yes	yes	yes	yes	yes	yes	yes
Region dummy variable	yes	yes	yes	yes	yes	yes	yes	yes	yes
Industry dummy variable	yes	yes	yes	yes	yes	yes	yes	yes	yes
N	11,857	8580	6455	11,857	8580	6455	11,857	8580	6455
r2_w	0.084	0.077	0.097	0.085	0.078	0.098	0.102	0.131	0.134
r2_b	0.225	0.047	0.295	0.236	0.052	0.307	0.131	0.109	0.047
r2_o	0.245	0.068	0.298	0.256	0.074	0.310	0.140	0.110	0.064
F	3.62	3.26	3.26	3.62	3.26	3.26	5.55	4.34	5.39

Note: Standard error in parentheses; **, *** represent the significance level of 5% and 1% respectively.

) on the reform of state-owned enterprises, using a fixed-effect model. First of all, the median of the marketization of the sample as the segmentation point, higher than the median of the sample that the high degree of marketization, on the contrary, that the low degree of marketization. The following analysis controls the following variables: wage level, current assets, fixed assets, employment, firm age, export, firm size, agency costs, firm paid-in capital.

Table 7 are fixed regression results, where (1) is the result of high marketization of TFP_op, (2) is the result of low marketization of TFP_op, (3) is the result of high marketization of TFP_lp, (4) is the result of low marketization of TFP_lp, (5) is the result of high marketization of LEV, (6) is the result of low marketization of LEV. The results of Table 6 show that in terms of productivity, when the level of marketization is high, the cross term is significantly positive at the 1% level, and the reform of state-owned enterprises has effectively improved the total factor productivity of enterprises; when the marketization level is low, the cross-term coefficient is not significant, and the impact of state-owned enterprise reform is not significant. In terms of debt, when the level of marketization is high, the cross-term coefficient is not significant; when the marketization level is low, the cross-term coefficient is significantly negative at the 5% level, and the effect of state-owned enterprise reform on reducing leverage ratio is more obvious in areas with low marketization. The reason may be that in regions with high marketization level, private capital pays more attention to efficiency, while state-owned enterprises have integrated into market-oriented financing means before the reform, so they have not readjusted their financing structure after the reform. Overall, regardless of the degree of marketization, the reform of state-owned enterprises improved the performance of enterprises from different aspects.

The median of the Herfindahl-Hirschman index of the sample industry is a segmentation point. Industries above the median believe that the degree of competition is low and the degree of monopoly is high. On the contrary, it is considered that the industry has a high degree of competition and a low degree of monopoly. The results are reported in Table 8. Table 8 are fixed regression results, where (1) is the result of high degree of industry competition of TFP_op, (2) is the result of low degree of industry competition of TFP_op, (3) is the result of high degree of industry competition of TFP_lp, (4) is the result of low degree of industry competition of TFP_lp, (5) is the result of high degree of industry competition of LEV, (6) is the result of low degree of industry competition of LEV. In terms of productivity, in highly competitive industries, the interaction coefficient is significantly positive at the 10% level; in low-competitive industries, the interaction coefficient is significantly positive at the 1% level, indicating that the reform of state-owned enterprises can significantly improve the TFP of enterprises in low-competitive industries. In terms of debt, in highly competitive industries, the interaction coefficient is not significant. In less competitive industries, the interaction coefficient is significantly negative at the 1% level. In general, the reform of state-owned enterprises has improved the performance of enterprises, but the effect of implementing state-owned enterprise reform in monopoly industries is more obvious. The reason is that by introducing non-state-owned capital, it is beneficial for the government to relax regulations, break administrative monopoly, and promote enterprises to establish a standardized corporate governance structure, thereby improving corporate performance.

Due to the differences in regional development, according to the division method [53], we divide provinces, autonomous regions and municipalities directly under the central government into eastern, central and western regions. The eastern region includes: Beijing, Tianjin, Shanghai, Jiangsu, Zhejiang, Liaoning, Hebei, Fujian, Shandong, Guangdong, Hainan; central region includes: Anhui, Jiangxi, Henan, Hubei, Hunan, Shanxi, Jilin, Heilongjiang, Inner Mongolia; the western region includes: Sichuan, Chongqing, Yunnan, Guizhou, Guangxi, Shaanxi, Gansu, Xinjiang, Qinghai, Ningxia, Tibet. Table 9 are fixed regression results, where (1) is the result of Eastern of TFP_op, (2) is the result of central of TFP_op, (3) is the result of western of TFP_op, (4) is the result of Eastern of TFP_lp, (5) is the result of central of TFP_lp, (6) is the result of western of TFP_lp, (7) is the result of Eastern of LEV, (8) is the result of central of LEV, (9) is the result of western of LEV. The coefficient before the interaction term shows that the impact of reform on TFP is inconsistent in different regions. In the eastern and western regions, the reform of state-owned enterprises has significantly improved the TFP of enterprises. In the central region, the interaction term coefficient is not significant; in terms of debt, in the eastern and western regions, the reform of state-owned enterprises has no significant impact on LEV. In the central region, the coefficient of interaction term is significantly negative. In general, the reform of state-owned enterprises has improved the performance of enterprises in different aspects.

6.3. Performance Comparison of Different Reform Methods

Around 2000, due to the advancement of enterprise reform, there have been a large number of empirical studies on the relationship between the proportion of state-owned shares and corporate performance, but the results of the study are controversial. For example, there is no significant relationship between the proportion of state-owned shares and corporate performance [54]. Ref. [55] found that the proportion of state shares was significantly negatively correlated with corporate performance. Ref. [21] also show that corporate performance is negatively correlated with the proportion of state-owned assets and positively correlated with the proportion of non-state-owned capital. Ref. [56] empirical analysis shows that the proportion of state shares and corporate performance was positively correlated. Ref. [57] found that there is a significant inverted ‘U’ relationship between the proportion of state-owned enterprises and performance. He believes that state-owned property rights are a double-edged sword, which will bring administrative intervention, thus affecting the efficiency of enterprises, and can also bring resources to the development of enterprises. Reform enterprises still rely on the main body of state-owned property rights. Reform enterprises need to make full use of administrative and market resources to maximize efficiency. In order to distinguish and compare the different effects of different reform methods on corporate performance, on the basis of preliminary regression, models (3), (4), (5) are further expanded to examine the heterogeneous effects of different degrees of mixing on corporate performance in the reform of state-owned enterprises.

Firstly, based on the definition of [20], the ratio of state-owned capital to paid-in capital (state-owned ratio) is used as a measure of the degree of mixing. Then, the state-owned ratio is 0 and 0.5 as the critical point, and the state-owned reform enterprises are further divided into three types. dums represents the treatment group with the proportion of state-owned enterprises being 0 after the reform, that is, the complete privatization reform of state-owned enterprises; dumsy represents the treatment group in which the proportion of state-owned enterprises after the reform is less than 0.5 but greater than 0, that is, the non-state-owned mixed ownership reform; dumgy represents the treatment group in which the proportion of state-owned enterprises after the reform is greater than or equal to 0.5, that is, the state-owned mixed ownership reform. By this definition, dumsy and dumgy represent the treatment group of mixed ownership reform of state-owned enterprises. In order to test the impact of mixed ownership reform and full privatization reform on corporate performance, Models (3), (4), (5) are extended to Models (6), (7), (8).

$$\begin{array}{ll}TF{P}_{it}=& {\alpha }_0+{\alpha }_1{}^{\prime }did{s}_{it}+{\alpha }_1{}^{″}dids{y}_{it}+{\alpha }_1{}^{‴}didg{y}_{it}\\ & +{\alpha }_{2}treate{d}_i+{\alpha }_{3}tim{e}_t+{\gamma }_T^{\prime }control{s}_{it}+{\nu }_i+{\mu }_t+{\epsilon }_{it}\end{array}$$

(6)

$$\begin{array}{ll}LE{V}_{it}=& {\beta }_0+{\beta }_1{}^{\prime }did{s}_{it}+{\beta }_1{}^{″}dids{y}_{it}+{\beta }_1{}^{‴}didg{y}_{it}\\ & +{\beta }_{2}treate{d}_i+{\beta }_{3}tim{e}_t+{\gamma }_L^{\prime }control{s}_{it}+{\nu }_i+{\mu }_t+{\epsilon }_{it}\end{array}$$

(7)

$$\begin{array}{ll}RO{A}_{it}=& {\gamma }_0+{\gamma }_1{}^{\prime }did{s}_{it}+{\gamma }_1{}^{″}dids{y}_{it}+{\gamma }_1{}^{‴}didg{y}_{it}\\ & +{\gamma }_{2}treate{d}_i+{\gamma }_{3}tim{e}_t+{\gamma }_L^{\prime }control{s}_{it}+{\nu }_i+{\mu }_t+{\epsilon }_{it}\end{array}$$

(8)

where $\begin{array}{l}did{s}_{it}=treate{d}_i\times tim{e}_t\times dums,\\ dids{y}_{it}=treate{d}_i\times tim{e}_t\times dumsy,\\ didg{y}_{it}=treate{d}_i\times tim{e}_t\times dumgy.\end{array}$

Here by comparing the coefficient ${\alpha }_1{}^{\prime },{\alpha }_1{}^{″},{\alpha }_1{}^{‴}$ to determine the different ways of reform of state-owned enterprises on corporate TFP heterogeneity effects. By comparing the coefficient ${\beta }_1{}^{\prime },{\beta }_1{}^{″},{\beta }_1{}^{‴}$ to determine the different ways of reform of state-owned enterprises on the enterprise LEV heterogeneity effects. By comparing the coefficient ${\gamma }_1{}^{\prime },{\gamma }_1{}^{″},{\gamma }_1{}^{‴}$ to determine the different ways of reform of state-owned enterprises on the enterprise ROA heterogeneity effects.

Table 10. Regression results of different reforms.

	(1)	(2)	(3)
	TFP_op	TFP_lp	LEV
dids	0.069 ***	0.070 ***	−0.033 ***
	(0.023)	(0.023)	(0.007)
didsy	0.027	0.028	−0.038 ***
	(0.032)	(0.032)	(0.010)
didgy	0.030	0.031	0.005
	(0.025)	(0.025)	(0.008)
constant term	0.838 ***	0.881 ***	0.661 ***
	(0.183)	(0.183)	(0.056)
Year dummy variable	yes	yes	yes
Region dummy variable	yes	yes	yes
Industry dummy variable	yes	yes	yes
N	24,548	24,548	24,548
r2_w	0.072	0.073	0.105
r2_b	0.225	0.235	0.028
r2_o	0.228	0.237	0.030
F	3.40	3.39	5.05

Note: Standard error in parentheses; *** represent the significance level of 1% respectively.

uses the fixed effect model, (1) is the result of TFP_op, (2) is the result of TFP_lp, (3) is the result of LEV. According to the regression results of different reforms, the complete privatization reform has a significant effect on the TFP of enterprises, and its coefficient is significantly positive at the 1% level, while the mixed ownership reform has no significant effect on the TFP of enterprises. Complete privatization reform and non-state-controlled mixed ownership reform can significantly reduce the asset-liability ratio (LEV) of enterprises, and the interaction coefficient is significantly negative at the 1% level, but the state-controlled mixed ownership reform has no significant effect on the LEV of enterprises. This shows that for state-owned enterprises, complete privatization not only helps to improve production efficiency, but also helps to improve corporate debt. Although the non-state-controlled mixed ownership reform has no significant effect on improving the production efficiency of enterprises, it can improve corporate debt. The mixed ownership reform of state-owned holding mode has no significant effect on the whole.

Table 11. Regression results of different reforms under different degrees of marketization.

	(1)	(2)	(3)	(4)	(5)	(6)
Marketization Degree	High	Low	High	Low	High	Low
	TFP_op	TFP_op	TFP_lp	TFP_lp	LEV	LEV
dids	0.093 ***	0.029	0.094 ***	0.030	−0.023 **	−0.028 **
	(0.033)	(0.037)	(0.033)	(0.037)	(0.010)	(0.011)
didsy	−0.002	0.041	−0.000	0.042	−0.005	−0.066 ***
	(0.051)	(0.050)	(0.051)	(0.050)	(0.015)	(0.015)
didgy	0.021	0.050	0.021	0.050	0.011	0.004
	(0.039)	(0.036)	(0.039)	(0.036)	(0.012)	(0.011)
constant term	0.766 ***	1.723 ***	0.809 ***	1.770 ***	0.648 ***	0.717 ***
	(0.266)	(0.321)	(0.266)	(0.321)	(0.080)	(0.097)
Year dummy variable	yes	yes	yes	yes	yes	yes
Region dummy variable	yes	yes	yes	yes	yes	yes
Industry dummy variable	yes	yes	yes	yes	yes	yes
N	12,128	12,420	12,128	12,420	12,128	12,420
r2_w	0.088	0.044	0.089	0.045	0.089	0.132
r2_b	0.243	0.177	0.255	0.183	0.093	0.025
r2_o	0.256	0.173	0.267	0.180	0.097	0.026
F	3.11	3.43	3.11	3.43	4.96	4.99

Note: Standard error in parentheses; **, *** represent the significance level of 5% and 1% respectively.

reports the effects of different types of reforms under different degrees of marketization. Table 11 are fixed regression results, where (1) is the result of high marketization of TFP_op, (2) is the result of low marketization of TFP_op, (3) is the result of high marketization of TFP_lp, (4) is the result of low marketization of TFP_lp, (5) is the result of high marketization of LEV, (6) is the result of low marketization of LEV. When the degree of marketization is high, the complete privatization reform method significantly improves the TFP of enterprises, but the mixed ownership reform method is not significant; when the degree of marketization is low, the effect of reform is not significant. When the degree of marketization is high, the complete privatization reform can significantly reduce the LEV of enterprises; when the degree of marketization is low, complete privatization reform and non-state-owned mixed ownership reform can significantly reduce the LEV of enterprises, but the effect of state-owned mixed ownership reform is not significant. Overall, the reform of state-owned enterprises can improve the performance of enterprises, state-controlled mixed ownership reform ineffective.

Table 12. Regression results of different reforms under different degrees of industrial competition.

	(1)	(2)	(3)	(4)	(5)	(6)
Industry Competition Degree	High	Low	High	Low	High	Low
	TFP_op	TFP_op	TFP_lp	TFP_lp	LEV	LEV
dids	0.065 *	0.088 ***	0.065 *	0.089 ***	−0.021 **	−0.034 ***
	(0.034)	(0.033)	(0.034)	(0.033)	(0.010)	(0.010)
didsy	−0.033	0.083 *	−0.031	0.084 *	−0.027 *	−0.045 ***
	(0.052)	(0.044)	(0.052)	(0.044)	(0.016)	(0.014)
didgy	0.050	0.050	0.050	0.051	0.015	0.001
	(0.037)	(0.036)	(0.037)	(0.036)	(0.011)	(0.011)
constant term	1.182 ***	0.735 **	1.229 ***	0.779 ***	0.660 ***	0.479 ***
	(0.273)	(0.299)	(0.273)	(0.299)	(0.082)	(0.093)
Year dummy variable	yes	yes	yes	yes	yes	yes
Region dummy variable	yes	yes	yes	yes	yes	yes
Industry dummy variable	yes	yes	yes	yes	yes	yes
N	12,506	12,042	12,506	12,042	12,506	12,042
r2_w	0.060	0.091	0.061	0.092	0.100	0.117
r2_b	0.158	0.258	0.169	0.269	0.157	0.077
r2_o	0.165	0.264	0.175	0.275	0.167	0.082
F	3.10	3.45	3.10	3.45	4.52	5.01

Note: Standard error in parentheses; *, **, *** represent the significance level of 10%, 5% and 1% respectively.

examines the impact of different types of reforms on the degree of competition in different industries. Table 12 are fixed regression results, where (1) is the result of high degree of industry competition of TFP_op, (2) is the result of low degree of industry competition of TFP_op, (3) is the result of high degree of industry competition of TFP_lp, (4) is the result of low degree of industry competition of TFP_lp, (5) is the result of high degree of industry competition of LEV, (6) is the result of low degree of industry competition of LEV. In terms of productivity, when the degree of industry competition is high, complete privatization reform is significantly positive at 10% level; when the degree of industry competition is low, the total privatization reform and the non-state-owned mixed ownership reform can significantly improve the TFP of enterprises. In terms of debt, no matter how competitive the industry is, complete privatization reform and non-state-owned mixed ownership reform can significantly reduce the LEV of enterprises, but the state-owned mixed ownership reform has no significant effect.

Table 13. Regression results of different reforms in different regions.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Region	Eastern	Central	Western	Eastern	Central	Western	Eastern	Central	Western
	TFP_op	TFP_op	TFP_op	TFP_lp	TFP_lp	TFP_lp	LEV	LEV	LEV
dids	0.066 **	0.051	0.088 **	0.067 **	0.052	0.089 **	−0.030 ***	−0.050 ***	−0.010
	(0.033)	(0.043)	(0.045)	(0.033)	(0.043)	(0.045)	(0.010)	(0.013)	(0.015)
didsy	0.098 **	−0.092	0.082	0.099 **	−0.090	0.084	−0.024 *	−0.064 ***	−0.026
	(0.048)	(0.061)	(0.060)	(0.048)	(0.061)	(0.060)	(0.014)	(0.018)	(0.021)
didgy	0.031	0.016	0.029	0.032	0.017	0.029	0.008	0.005	0.005
	(0.038)	(0.048)	(0.044)	(0.038)	(0.048)	(0.044)	(0.011)	(0.014)	(0.015)
constant term	0.867 ***	0.572	0.879 **	0.911 ***	0.622 *	0.921 **	0.576 ***	0.661 ***	0.737 ***
	(0.261)	(0.349)	(0.440)	(0.261)	(0.349)	(0.439)	(0.079)	(0.103)	(0.147)
Year dummy variable	yes	yes	yes	yes	yes	yes	yes	yes	yes
Region dummy variable	yes	yes	yes	yes	yes	yes	yes	yes	yes
Industry dummy variable	yes	yes	yes	yes	yes	yes	yes	yes	yes
N	10,681	7952	5915	10,681	7952	5915	10,681	7952	5915
r2_w	0.084	0.081	0.091	0.085	0.082	0.092	0.093	0.139	0.114
r2_b	0.210	0.048	0.298	0.222	0.054	0.310	0.114	0.114	0.060
r2_o	0.225	0.068	0.300	0.235	0.074	0.311	0.120	0.116	0.078
F	3.56	3.25	3.15	3.55	3.25	3.14	5.49	4.24	5.31

Note: Standard error in parentheses; *, **, *** represent the significance level of 10%, 5% and 1% respectively.

shows the regression results of different reforms in different regions. Table 13 are fixed regression results, where (1) is the result of Eastern of TFP_op, (2) is the result of central of TFP_op, (3) is the result of western of TFP_op, (4) is the result of Eastern of TFP_lp, (5) is the result of central of TFP_lp, (6) is the result of western of TFP_lp, (7) is the result of Eastern of LEV, (8) is the result of central of LEV, (9) is the result of western of LEV. In the eastern region, the complete privatization reform and the non-state-owned mixed ownership reform have a significant effect on the TFP of enterprises. In the central region, the way of reform to enhance corporate TFP is not significant, in the western region, the way of complete privatization to enhance corporate TFP is significant; in the eastern and central regions, the complete privatization reform and the non-state-owned mixed ownership reform can significantly reduce the LEV of enterprises. In the western region, the reform reduces the LEV of enterprises insignificantly.

6.4. Robustness Test

Considering the estimation deviation caused by the omission of variables, we use the return on assets as the enterprise performance index in the literature. We further introduce the return on assets as the enterprise performance index, and add three variables of subsidy income, return on net assets and debt equity ratio as the matching variables, and then do the robust test. The data set deletes the missing samples of return on assets, subsidy income, return on net assets and debt-to-equity ratio, and passes the balance test of sample matching. According to the research proposal [58], the OLS basic regression is first done, and the regression results are shown in

Table 14. OLS regression.

	(1)	(2)	(3)	(4)
	TFP_op	TFP_lp	ROA	LEV
treated	0.102 ***	0.102 ***	0.004 ***	−0.009 ***
	(0.009)	(0.009)	(0.001)	(0.003)
size	0.236 ***	0.238 ***	−0.007 ***	0.035 ***
	(0.018)	(0.018)	(0.001)	(0.006)
age	−0.005 ***	−0.005 ***	−0.000 ***	0.001 ***
	(0.000)	(0.000)	(0.000)	(0.000)
agentcost	1.830 ***	1.837 ***	−0.012 *	−0.126 ***
	(0.086)	(0.086)	(0.007)	(0.030)
labor	0.055 ***	0.058 ***	0.001	0.005 *
	(0.008)	(0.008)	(0.001)	(0.003)
lwage	−0.006	−0.004	0.014 ***	−0.080 ***
	(0.009)	(0.009)	(0.001)	(0.003)
llas	0.144 ***	0.144 ***	0.002 **	0.069 ***
	(0.011)	(0.011)	(0.001)	(0.004)
gdas	−0.098 ***	−0.093 ***	−0.002 **	0.016 ***
	(0.008)	(0.008)	(0.001)	(0.003)
lk	−0.023 ***	−0.023 ***	0.003 ***	−0.110 ***
	(0.005)	(0.005)	(0.000)	(0.002)
exit	−0.196 ***	−0.196 ***	−0.011 ***	0.027 ***
	(0.011)	(0.011)	(0.001)	(0.004)
roe	0.378 ***	0.378 ***	0.087 ***	0.075 ***
	(0.016)	(0.016)	(0.001)	(0.006)
lma	0.007 ***	0.007 ***	0.001 ***	0.000
	(0.001)	(0.001)	(0.000)	(0.000)
sub	−0.015 ***	−0.015 ***	0.000	−0.001
	(0.002)	(0.002)	(0.000)	(0.001)
_cons	0.799 ***	0.801 ***	0.011 ***	0.581 ***
	(0.044)	(0.044)	(0.003)	(0.015)
N	22,891.000	22,891.000	22,891.000	22,891.000
r2	0.303	0.312	0.239	0.240

Note: Standard error in parentheses; *, **, *** represent the significance level of 10%, 5% and 1% respectively.

, where (1) is the result of TFP_op, (2) is the result of TFP_op, (3) is the result of ROA, (4) is the result of LEV. It can be found that the conclusions are generally consistent with the results obtained by the PSM-DID method, and most of the literature research results are consistent. The reform of state-owned enterprises is conducive to the improvement of ROA. Secondly, more matching variables are added to solve the estimation errors caused by missing variables.

In

Table 15. Preliminary regression results (Robustness Test).

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	TFP_op	TFP_lp	ROA	TFP_op	TFP_lp	ROA	LEV	LEV
treated × time	0.078 ***	0.080 ***	0.004 ***	0.078 ***	0.079 ***	0.009 ***	−0.001	−0.015 ***
	(0.019)	(0.019)	(0.001)	(0.019)	(0.019)	(0.002)	(0.007)	(0.006)
dids	0.094 ***	0.097 ***	0.004 **	0.096 ***	0.097 ***	0.010 ***	−0.016 *	−0.033 ***
	(0.026)	(0.026)	(0.002)	(0.023)	(0.024)	(0.002)	(0.009)	(0.007)
didsy	0.126 ***	0.130 ***	0.009 ***	0.059 *	0.063 *	0.008 ***	−0.029 ***	−0.035 ***
	(0.030)	(0.030)	(0.002)	(0.033)	(0.033)	(0.003)	(0.011)	(0.010)
didgy	0.005	0.007	−0.001	0.050 **	0.050 **	0.008 ***	0.022 ***	0.005
	(0.024)	(0.024)	(0.002)	(0.025)	(0.025)	(0.002)	(0.008)	(0.008)

Note: Standard error in parentheses; *, **, *** represent the significance level of 10%, 5% and 1% respectively.

, where (1) is the ordinary least squares regression result of TFP_op, (2) is the ordinary least squares regression result of TFP_lp, (3) is the ordinary least squares regression result of ROA, (4) is the fixed regression result of TFP_op, (5) is the fixed regression result of TFP_lp, (6) is the fixed regression result of ROA. (7) is the ordinary least squares regression result of LEV, (8) is the fixed regression result of LEV. The overall effect in Table 15 is consistent with the results in Table 6, basically consistent with Table 10, consistent with Table 10 in dids, didsy is consistent on LEV, didgy can increase TFP, consistent on LEV, and state-owned enterprise reform can increase ROA.

Table 16. Effect of state-owned enterprise reform with different marketization degrees (Robustness Test).

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Marketization Degree	High	Low	High	Low	High	Low	High	Low
	TFP_op	TFP_op	TFP_lp	TFP_lp	ROA	ROA	LEV	LEV
treated × time	0.091 ***	0.079 ***	0.093 ***	0.080 ***	0.009 ***	0.006 ***	−0.005	−0.030 ***
	(0.028)	(0.030)	(0.028)	(0.030)	(0.003)	(0.002)	(0.008)	(0.009)
dids	0.105 ***	0.074 *	0.105 ***	0.077 *	0.010 ***	0.006 **	−0.022 **	−0.043 ***
	(0.033)	(0.040)	(0.033)	(0.040)	(0.003)	(0.003)	(0.010)	(0.012)
didsy	−0.002	0.093 *	0.003	0.094 *	0.004	0.011 ***	−0.010	−0.080 ***
	(0.050)	(0.054)	(0.051)	(0.054)	(0.005)	(0.004)	(0.015)	(0.016)
didgy	0.059	0.066 *	0.061	0.066 *	0.008 **	0.004	0.011	−0.009
	(0.039)	(0.038)	(0.039)	(0.038)	(0.004)	(0.003)	(0.012)	(0.011)

Note: Standard error in parentheses; *, **, *** represent the significance level of 10%, 5% and 1% respectively.

are fixed regression results, where (1) is the result of high marketization of TFP_op, (2) is the result of low marketization of TFP_op, (3) is the result of high marketization of TFP_lp, (4) is the result of low marketization of TFP_lp, (5) is the result of high marketization of ROA, (6) is the result of low marketization of ROA, (7) is the result of high marketization of LEV, (8) is the result of low marketization of LEV. The overall effect in Table 16 is consistent with the results in Table 7, which is consistent with Table 11 on LEV, which is consistent with Table 11 on dids, didsy and didgy are consistent on TFP when the degree of marketization is high, and didsy and didgy increase TFP at a 10% significant level when the degree of marketization is low. In terms of ROA, it is generally consistent with the existing literature. Complete privatization increases ROA, and didsy increases ROA when marketization is high, and didgy increases ROA when marketization is low.

Table 17. The reform effect of state-owned enterprises with different industry competition degree (Robustness Test).

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Industry Competition Degree	High	Low	High	Low	High	Low	High	Low
	TFP_op	TFP_op	TFP_lp	TFP_lp	ROA	ROA	LEV	LEV
treated × time	0.088 ***	0.088 ***	0.091 ***	0.088 ***	0.007 ***	0.009 ***	−0.010	−0.023 ***
	(0.029)	(0.027)	(0.029)	(0.027)	(0.002)	(0.002)	(0.009)	(0.008)
dids	0.119 ***	0.078 **	0.121 ***	0.081 **	0.010 ***	0.009 ***	−0.026 **	−0.040 ***
	(0.035)	(0.034)	(0.035)	(0.035)	(0.003)	(0.003)	(0.010)	(0.010)
didsy	0.033	0.091 **	0.038	0.093 **	0.005	0.006	−0.033 **	−0.047 ***
	(0.054)	(0.046)	(0.054)	(0.046)	(0.005)	(0.004)	(0.016)	(0.014)
didgy	0.065 *	0.046	0.069*	0.042	0.005	0.010 ***	0.001	0.003
	(0.039)	(0.037)	(0.039)	(0.037)	(0.003)	(0.003)	(0.011)	(0.011)

Note: Standard error in parentheses; *, **, *** represent the significance level of 10%, 5% and 1% respectively.

are fixed regression results, where (1) is the result of high degree of industry competition of TFP_op, (2) is the result of low degree of industry competition of TFP_op, (3) is the result of high degree of industry competition of TFP_lp, (4) is the result of low degree of industry competition of TFP_lp, (5) is the result of high degree of industry competition of ROA, (6) is the result of low degree of industry competition of ROA, (7) is the result of high degree of industry competition of LEV, (8) is the result of low degree of industry competition of LEV. The overall effect in Table 17 is consistent with the results in Table 8, consistent with Table 12 on LEV, and basically consistent with Table 12 in dids, didsy and didgy, except that didgy is significant at 10% level when the industry is highly competitive, which can improve TFP. In terms of ROA, it is generally consistent with the existing literature. Complete privatization increases ROA, and didgy increases ROA when the degree of industry competition is low.

Table 18. Effect of SOE reform in different regions (Robustness Test).

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
Region	Eastern	Central	Western	Eastern	Central	Western	Eastern	Central	Western	Eastern	Central	Western
	TFP_op	TFP_op	TFP_op	TFP_lp	TFP_lp	TFP_lp	ROA	ROA	ROA	LEV	LEV	LEV
treated × time	0.064 **	0.006	0.191 ***	0.066 **	0.007	0.191 ***	0.008 ***	0.015 ***	−0.000	−0.015 *	−0.030 ***	0.005
	(0.027)	(0.037)	(0.036)	(0.027)	(0.037)	(0.036)	(0.002)	(0.003)	(0.003)	(0.008)	(0.011)	(0.013)
dids	0.049	0.063	0.216 ***	0.049	0.066	0.216 ***	0.008 ***	0.021 ***	−0.004	−0.038 ***	−0.048 ***	0.002
	(0.034)	(0.045)	(0.048)	(0.034)	(0.045)	(0.047)	(0.003)	(0.004)	(0.004)	(0.010)	(0.013)	(0.016)
didsy	0.083 *	−0.088	0.223 ***	0.088 *	−0.084	0.224 ***	0.004	0.015 ***	0.003	−0.029 **	−0.051 ***	−0.023
	(0.048)	(0.064)	(0.065)	(0.048)	(0.064)	(0.065)	(0.004)	(0.005)	(0.006)	(0.014)	(0.019)	(0.022)
didgy	0.020	0.013	0.144 ***	0.023	0.010	0.144 ***	0.007 **	0.011 ***	0.003	0.007	0.002	0.012
	(0.038)	(0.049)	(0.046)	(0.039)	(0.049)	(0.046)	(0.003)	(0.004)	(0.004)	(0.011)	(0.015)	(0.016)

Note: Standard error in parentheses; *, **, *** represent the significance level of 10%, 5% and 1% respectively.

are fixed regression results, where (1) is the result of Eastern of TFP_op, (2) is the result of central of TFP_op, (3) is the result of western of TFP_op, (4) is the result of Eastern of TFP_lp, (5) is the result of central of TFP_lp, (6) is the result of western of TFP_lp, (7) is the result of Eastern of ROA, (8) is the result of central of ROA, (9) is the result of western of ROA, (10) is the result of Eastern of LEV, (11) is the result of central of LEV, (12) is the result of western of LEV. The overall effect in Table 18 is basically consistent with the results in Table 9. The reform of state-owned enterprises reduces LEV at the 10% significant level in the east. It is consistent with Table 13 on LEV, and basically consistent with Table 13 on dids, didsy and didgy. The difference is mainly in the west, where any type of reform can increase TFP. In terms of ROA, it is generally consistent with the existing literature. Except that private holdings have no significant effect in the eastern region, the reform of state-owned enterprises has increased ROA in the eastern and central regions, and the reform of state-owned enterprises in the western region has no significant effect on ROA.

From the above empirical results, it can be concluded that although there are a few differences in measurement estimation results, they are generally consistent, indicating that our results are relatively robust.

7. Conclusions

7.1. Research Conclusions

Based on the data of industrial enterprise database from 1998 to 2007, this paper studies the effect of state-owned enterprise reform on enterprise performance by using the difference-in-difference model with propensity score matching. Overall, the reform of state-owned enterprises can improve the performance of enterprises. After further considering the influence of marketization, industry competition and regional characteristics, it is found that the effect of reform is heterogeneous. When the degree of marketization is high, the effect of reform on improving productivity is good, and when the degree of marketization is low, the effect of reform on reducing debt is good; the reform effect of industries with low degree of competition is better than that of industries with high degree of competition. The reform of state-owned enterprises in the eastern region has the best effect, and the reform in the central region has a better effect on reducing debt. Privatization in the western region has significantly increased productivity. In general, the reform of state-owned enterprises is conducive to im-proving the return on assets of reformed enterprises; complete privatization can in-crease the return on assets. When the marketization is high, private holdings increase the return on assets, and when the marketization is low, state-owned holdings increase the return on assets. When the degree of industry competition is low, state-owned holdings increase asset returns; except that private ownership has no significant effect in the eastern region, the reform of state-owned enterprises has increased the return on assets in the eastern and central regions, and the reform of state-owned enterprises re-turn on assets in the west are not significant.

7.2. Research Policy Implications

The conclusions of this paper show that, first of all, the reform of state-owned enterprises can have a significant positive effect on corporate performance, so we should adhere to and deepen the work of ownership reform; secondly, comparing different reform methods, it is found that different types of reform have different effects. Under different marketization, industry competition and regional characteristics, the reform effects brought by different reform methods are different. Therefore, the reform of state-owned enterprises should not implement a one-size-fits-all policy, and the reform evaluation should not be simplified. Finally, corresponding reform measures should be taken according to the specific indicators of enterprise performance, so as to solve the problems encountered by state-owned enterprises and improve the efficiency of enterprise reform.

7.3. Limitations and Future Perspectives

Limited by data, this paper only analyzes the data of industrial enterprises from 1998 to 2007.During this period, the data quality of industrial enterprises is good, and the calculated total factor productivity is reliable. However, the data is some distance from the present time, and the economic environment is also changing. Whether the discovery law reflects the existing economy and whether it can provide policy guidance to the existing economy needs further analysis. The next step is to collect data on the recent reform of state-owned enterprises, further verify the laws found in this study, and provide theoretical guidance for the reform of state-owned enterprises. At the same time, whether the reform of state-owned enterprises improves the innovation ability of enterprises, and whether there are differences in the reform methods of different state-owned enterprises in different economic environments, which is related to the long-term development of enterprises and the sustainable development of the national economy. Although there have been some studies, there is no consistent conclusion. At present, research in this area is urgently needed to guide the practice of state-owned enterprise reform.