A stochastic frontier analysis of firm efficiency in China

,


INTRODUCTION
Extant research on the corporate governance of transition economies leaves a heated debate on the effectiveness of ownership structure vis-à-vis corporate governance in improving the performance of state-owned enterprises (SOEs).One vein of the literature asserts that privatization is necessary for significant improvements in the performance of SOEs (Boycko et al., 1996;Shleifer, 1998;Dewenter and Malatesta, 2001).Another vein of the literature argues that changes in internal and external corporate governance structure without privatization can be effective in increasing firm performance (Allen and Gale, 2000;Boubakri et al., 2005;Dahya et al., 2008;Ramly and Rashid, 2010).
This study aims to shed additional light on the debate by examining whether and to what extent ownership structure, ownership concentration, board independence and the strength of the external corporate governance mechanism affect firm efficiency, using a large sample of 744 China's privatized former SOEs listed on Shanghai or Shenzhen stock exchanges during 1999 and 2006.
questions: Does the retained state shareholding exert negative impact on firm efficiency?Does ownership concentration allow the controlling shareholder to have more power in monitoring managers or expropriating minority shareholders?Does the identity of the controlling shareholder matter?Is board independence an effective internal corporate governance mechanism?Does the degree of marketization have any impact on firm efficiency?This paper obtains four major findings.First, firm efficiency, as estimated using stochastic frontier analysis (SFA), decreases with state ownership of shares while increases with public and employee share ownership.Second, the relationship between ownership concentration (the percentage shareholding of the largest shareholder) and firm efficiency is U-shaped, providing some evidence that the controlling shareholder expropriates minority shareholders.Third, as the identity of the largest shareholder changes from government, government-controlled legal entity to other types of legal entity, firm efficiency significantly improves.Fourth, board independence, the structure of the supervisory committee and the strength of provincial market development have no impact on firm efficiency.The results suggest that privatization via transfer of ownership and control from the state to the public enhances firm efficiency, while changes in internal and external corporate governance structure without privatization do not affect firm performance.
The rest of paper has the following structure.Subsequently, the study reviews the literature on corporate governance of transition economies.Then, it discusses the SFA methodology and the variables used in empirical analyses.Thereafter, describes the data, characterizes the distribution of firm observations across time and subindustries, and presents summary statistics.Afterwards, it provides empirical results.Finally, concludes with a summary of findings.

LITERATURE REVIEW
Privatization is widely regarded as the major first step to align managerial incentives on increasing profits and improving efficiency (Dennis and McConnell, 2003).A central issue on corporate governance of transition economies is whether or not ultimate ownership changes are necessary for improving the efficiency of SOEs.Boycko et al. (1996), Shleifer (1998) and Dewenter and Malatesta (2001) assert that because governments cannot play an active role in corporate governance, ownership change is necessary for any significant performance improvements of SOEs.Ongore (2011) finds that foreign ownership of firms can lead to more innovations and better firm performance.On the other hand, Allen and Gale (2000) and Dahya et al. (2008) argue that less radical methods such as managerial incentive contracts, market deregulation, and internal and external governance reform can be effective substitutes to outright privatization.Megginson and Netter (2001), Djankov and Murrell (2002) and Megginson (2005) provide comprehensive surveys of over 200 empirical studies on privatization and enterprise restructuring in both developed and transition economies.They find dominant evidence in support of the proposition that private ownership is associated with better firm performance and enterprise efficiency than is state ownership.
Studies of the effectiveness of other governance mechanisms in transition economies have primarily focused on the relationship between a particular aspect of corporate governance and firm performance, for example, board composition (Dalton and Kesner, 1997;Hermalin and Weisbach, 2003), executive compensation and CEO turnover (Gibson, 2003;Bryan, et al., 2010), blockholders (Claessens et al., 2000;Holderness, 2003), takeover market (Allen and Gale, 2000), and investor protection (La Porta et al., 2000, 2002;Durnev and Kim, 2005).Dennis and McConnell (2003) provide a good survey of the literature on international corporate governance.
In recent years, there has been an increasing interest in the effectiveness of corporate governance systems in Su and Dai 11255 transition economies.For example, Khanna and Palepu (2000) examine how the identity of the immediate owners of Indian firms is correlated with the firms' valuation and find that high foreign ownership has beneficial effects either on market valuation or profitability.Johnson et al. (2000) finds that the magnitude of the decline in East Asian stock markets and the degree of depreciation of the exchange rate during the crisis are negatively related to the level of investor protection.Claessens et al. (2000Claessens et al. ( , 2002) ) provide evidence that more than two thirds of East Asian firms are controlled by a single large shareholder, and that the difference between control rights and cashflow rights for the largest shareholder is associated with significant value discounts in 8 East Asian economies.Lemmon and Lins (2003) show that differences in ownership structure at the firm level can explain differences in firm performance in 8 East Asian countries.Klapper and Love (2003) use firm-level data from 14 emerging stock markets and reported that better corporate governance is highly correlated with better operating performance and higher market valuation.Tvaronavičienė and Kalašinskaitė (2005) review various studies on the efficiency and governance of privatized former SOEs and find that privatization is one of the most important economic development tools for a nation.
Using data envelopment analysis, Zelenyuk and Zheka (2006) find that the quality level of corporate governance is positively correlated with firm efficiency in Ukraine.Shah et al. (2011) reports that East Asian firms with higher outside block ownership and more independent board of directors experience smaller decline in stock prices during the Asian financial crisis.In addition, there has been a growing literature on the corporate governance of Chinese firms.Li (1997) provides evidence that improved incentives during the early stage of enterprise reform increase the productivity of SOEs.Aharoni (2000) and Shiu (2002) find that state ownership is negatively related to firm performance and firm value.Hovey et al. (2003) investigate the empirical relationship between governance mechanisms and market valuation of all publicly listed firms in China from 1999 to 2001.They find that ownership concentration, the dual role of general manager/CEO as board chair, and government ownership have negative effects on firm value, as measured by Tobin's q.Lin (2004) shows that corporate governance reform has improved earnings, sales and workers' productivity, but has not increased returns to investors.
In addition, state ownership is associated with poor SOE performance and that legal entity ownership is tied to better firm performance.Fan et al. (2007) find that political connections of CEOs significantly deteriorate the functionality of corporate governance mechanism and lower firm performance in China.Kang et al. (2008) demonstrate that standard corporate governance mechanisms are weak and ineffective for publicly listed firms while alternative governance mechanisms based on reputation and relationship have been remarkably effective in the private sector.Chen et al. (2010) document that China operates a two-tier board system consisting of a main board of directors and a supervisory board.The state has huge influence on the appointment of board and supervisory board members.Because stateappointed bureaucrats are ineffective in monitoring management, the pay-for-performance link is quite weak among listed firms.
Although the mammoth of literature has offered lessons and policy implications of profound importance for research on corporate governance of transition economies, the debate on the relative importance of ownership changes and corporate governance reforms is still inconclusive.Therefore, it is important to continue to carefully examine the subject using a variety of samples and empirical methodologies.

METHODOLOGY
This paper uses productive efficiency in analyzing the relationship between ownership structure, corporate governance and firm performance.Productive efficiency measurement techniques use a production possibility frontier to map a locus of potentially technically efficient output combinations a firm is capable of producing at a point in time, and comprises of parametric or nonparametric methods.Parametric methods rely on pre-specified production or cost functions and utilize econometric techniques such as regression analysis or stochastic frontier analysis (SFA).In contrast, nonparametric methods, for example, data envelopment analysis (DEA), use mathematical programming techniques to evaluate firm performance relative to the efficient frontier. 1This paper uses SFA with the Cobb-Douglas production function to test various hypotheses on the relationship among firm efficiency, ownership structure and corporate governance.

Half-normal and exponential SFA models
SFA models assume a production function of the usual regression form and a distribution type of error item which is equal to the sum of two components, the first part is symmetric and captures statistical noise such as weather, luck, machine breakdown and other events beyond the control of firms, and the second part represents technical inefficiency of firms.
A general SFA model that allows for the existence of firm-year specific unobserved inefficiency term is as follows: is the stochastic production frontier, X i,t is a vector of inputs that typically includes labor, 1 The DEA and SFA methodologies have been widely used in business, economics and management research.See Cooper et al. (2004) for applications of DEA methodology in studies of the productive efficiency.See Kumbhakar and Lovell (2000) for reviews of the basic models, theoretical extensions and recent development of SFA; Shiu (2002), Fu et al. (2008), Hasan et al. (2011) for applications of SFA and DEA methodologies in studies of the productive efficiency of firms in emerging economies.

TE
provides a measure of the shortfall of the observed output from the maximum feasible output in an environment characterized by the stochastic element } exp{ ,t i v .Therefore, the second stage regression can be specified as follows: . Utilizing the log-linear Cobb-Douglas production function, (3) can be rewritten as follows: where output Y is the sales revenue of a firm, LABOR is the number of employees of a firm, Capital represents capital stock measured by lagged one period net fixed assets, and Intermediate represents intermediate inputs including the cost of raw materials, energy and other productive goods.To avoid arbitrary selecting a specific distribution, this paper allows  is a pure random error term.
In addition, t i v , and t i u , are assumed to be distributed independently of each other, and of all explanatory variables.In the context of this paper, (5) can be further specified as follows to include various ownership structure and corporate governance variables:  CEO is also the board chair and 0 otherwise; Supervisorsize is the number of supervisor, Outsupervisor is the number of independent supervisor; Firmage is the number of years after going public for a firm, Firmage2 is the squared term for Firmage, and Marketization is an index that proxies the strengthen of provincial market development.

Conditional mean truncated normal SFA model
The half-normal and exponential SFA models assume that corporate governance mechanisms influence technology and output directly.Hence, regressions ( 6) and ( 7) include all explanatory variables in the production function.Another approach is to assume that outputs are determined by labor, capital stock, intermediate inputs and innovations on technical efficiency ( t i u , ).Corporate governance mechanisms influence output only through their effects on t i u , .In this case, the SFA model can be expressed as the following conditional mean truncated normal distribution model (Kumbhakar and Lovell, 2000):

Data and summary statistics
The sample used in this paper consists of 744 privatized former SOEs in the manufacturing industry, publicly listed on the Shanghai and Shenzhen securities exchanges before 1999.All firms are classified into 9 single-digit and 33 double-digit sub-industries.Ownership structure and corporate governance variables are constructed from the annual and semi-annual financial reports between 1999 and 2006.The final sample is an unbalanced panel consisting of 5218 firm-year observations.All data are from the well-known Chinese Stock Market and Accounting Research (CSMAR) database.
Table 1 contains the frequency distribution of firm observations across 9 single-digit sub-industries.Table 2 presents summary statistics of various ownership structure and corporate governance variables across sub-industries.
As shown in the Table 2, ownership structure does not vary widely across sub-industries.On average, sub-industries 3 (paper and allied products), 6 (metal, mineral and cement) and 7 (machinery, equipment and instrument) have the highest state ownership of shares (around 35% of total shares) while subindustries 2 (textile, apparel, fur and leather) and 9 (other manufacturing) exhibit relatively low state ownership (around 26% of total shares).The average fraction of shares held by legal entities is quite homogenous across the sub-industries.Sub-industry 9 has the highest legal entity share ownership while sub-industry 3 has the lowest.Public ownership of shares is also quite similar across the sub-industries.Sub-industry 8 (medicine and biological products) has the highest (43.2% of total shares) while sub-industry ownership concentration for China's listed firms.Regarding the identity of the controlling shareholder, state is the largest shareholder in about 36.2% of the firms in sub-industry 3 (representing the highest proportion) and 7.8% of the firms in subindustry 5 (representing the lowest proportion).In other subindustries, state is the controlling shareholder in about 13 to 24% of the firms.
State-owned legal entity is the largest shareholder in the majority of firms (greater than 50%).The only exception is sub-industry 2, in which case the state-owned legal entity is the largest shareholder in about 42.7% of the firms.On average, board size, the number of independent directors, the size of supervisory committee and the number of outside supervisors do not vary much across the subindustries.CEO is also the board chair in about 16 to 21% of the firms in sub-industries 1, 2, 4, 6, 7 and 9, and less than 15% of the firms in sub-industries 3, 5 and 8.The average strength of marketization is also quite similar across the sub-industries.
Table 3 illustrates the changes of ownership and governance structure over time.As shown in the table, state ownership of shares decreases from an average of 35% to about 30% between 1999 and 2006, while legal entity and public ownership of shares increases from 25.7 and 31.6% in 1999 to 31.4 and 37% in 2006, respectively.The number of employee shares declines over time, due to the decree issued by the CSRC to cease the issuance of employee shares beginning in 1998.From 1999 to 2006, the average proportion of shares held by the controlling shareholder remains large with very little variations.At the same time, the proportion of firms directly controlled by the state falls from 16.3% in 1999 to 13.4% in 2002, and then rises to 16.6% in 2006.In contrast, the proportion of firms indirectly controlled by the state through ownership of legal entity shares increases from 59.5 to 64.4%, suggesting that the government has tighten its grip of the publicly listed firms through direct and indirect holding of shares.
The frequency of the dual role played by the general manager/CEO as board chair decreases from 27.3% in 1999 to 14.2% in 2006, representing a substantial decline in the power of general manager/CEO.The size of the board and the supervisory committee remain quite stable over time, while the average number of outside directors increases substantially from 0.032 to 2.486, reflecting a move towards more independent board.In contrast, the average number of outside supervisors only increases a little, from

Estimation results from half-normal and exponential SFA models
Tables 4 and 5 contain maximum likelihood estimation results for the half-normal and exponential SFA models, respectively.As shown in the tables, the half-normal and exponential SFA models yield almost identical empirical results.The Cobb-Douglas production function is wellbehaved.Labor, capital stock and intermediate inputs are statistically significant at any level and the sum of their coefficients indicates almost a constant return to scale in Chinese manufacturing industry.Firm efficiency is negatively related to state ownership and positively related to both public and employee share ownership, as the coefficient estimates for State are significantly negative at the 5% level while the coefficient estimates for Public and Employee are significantly positive at least at the 10% level.In particular, a 10% point increase in state shares decreases firm efficiency score by about 0.53 to 0.69% point, after controlling for other variables.
A 10% point increase in public and employee shares increases firm efficiency score by about 1.48 to 2.17 and 2.76 to 3.86% points, respectively.An explanation for the negative effect of state ownership of shares on firm efficiency is that SOEs are not disciplined by the capital market because the state is both the regulator and owner of the firm.The competitively driven monitoring systems and the threats of takeover or bankruptcy are absent, which reduces managerial incentives to improve firm efficiency.In contrast, legal entity shareholders have an important stake in the listed firms and have more incentives to monitor managers.
A larger magnitude of the impact of employee shares than that of public shares shows that employee ownership is more useful in improving firm efficiency, perhaps because employee share ownership can better align the interests of workers (insiders) and outside investors.The previous results are consistent with Li (1997), Shiu (2002) and Hovey et al. (2003), who find that corporate governance changes without more radical ownership reform cannot enhance managerial incentives and improve firm performance.
The coefficient estimates for Largest are significantly negative while the coefficient estimates for Largest2 are significantly positive at the 5% level, indicating that the relationship between ownership concentration and firm efficiency is U-shaped.Therefore, tunneling and expropriation by the controlling shareholder appears to be an important issue to be further addressed in the ongoing corporate governance reform.The identity of the controlling shareholder also affects firm efficiency as the coefficient estimates for Statecrtl and Legalstactl are all significantly negative at least at the 5% level.Among three types of largest shareholders, government is the least efficient of all, followed by government-controlled legal entities.Firms with government and government-controlled legal entities as largest shareholders are around 12 and 5% points less efficient than those with other types of investors as largest shareholders, respectively, indicating that political interferences lead to a large reduction in firm efficiency.
However, the coefficient estimates for Boardsize, Outdirector and Duality are not statistically significant, indicating that board size, the number of independent directors and the dual role of general manager/CEO as board chair have no impact on firm efficiency.The results are similar to Klein (1998) and Hermanlin and Weisbach (2003), who find no significant relation between various characteristics of board composition and firm performance.The structure of the supervisory committee does not affect firm efficiency either, as the coefficient estimates for Supervisorsize and Outsupervisor are all statistically insignificant.The results are consistent with Kang et al. (2008), who find that supervisory committee in China's listed firms are more decorative than functional.Regarding the external corporate governance mechanism, the strength of provincial marketization is unrelated to firm efficiency, as the coefficient estimates for MARKETIZATION are statistically insignificant.These results suggest that internal and external corporate governance mechanisms, such as board of directors, supervisory committee and market freedom, are ineffective in reducing agency problem.
Finally, the relationship between productive efficiency and firm age, the primary control variable in addition to year and sub-industry dummies, is U-shaped.An explanation is that firms only view the stock market as a cheap source of equity capital and do not perform well when they first go public.However, when they find it difficult to raise additional capital later on, they have no choice but to strive to improve firm efficiency.

Estimation results from conditional mean truncated normal SFA model
The SFA regressions ( 8) and ( 9) or ( 8) and ( 10) are estimated jointly by the maximum likelihood method and the results are presented in Table 6.
2 Because expected firm efficiency is negatively related to t i,  , the co- efficients of the ownership and governance structure variables are expected to be in opposite sign with those in the previous models.
As shown in the table, the conditional mean truncated normal SFA models share very similar results with halfnormal and exponential SFA models.Firm inefficiency is positively related to government ownership and negatively related to public and employee ownership.A 10% point increase in state shares is associated with 2 Kumbhakar and Lovell (2000) provide detailed discussions on the maximum likelihood approach in estimating conditional mean truncated normal SFA models.approximately a 0.56 to 0.68 percentage point increase in firm inefficiency and a 10% point increase in public and employee shares is associated with a 1.47 to 2.21 and 2.72 to 3.5% point decrease in firm inefficiency, respectively.The relationship between ownership concentration and firm inefficiency is of inverted "U" shape.Firms with government as the direct controlling shareholders are the most inefficient, followed by those with government-controlled legal entities as the largest shareholders.Firm inefficiency is unrelated to board independence, board size, general manager/CEO duality, the structure of the supervisory committee and provincial marketization.

DISCUSSION
In contrast to the widespread acceptance of econometric and mathematical frontier estimation techniques in many economic, business and management studies, the adoption of these methods in corporate finance research is still in its infancy.In this paper, we examine the relationship between corporate governance and firm efficiency in a sample of publicly listed manufacturing firms in China by estimating parametric half-normal and exponential SFA models and conditional mean truncated normal SFA models.We strive to capture the corporate governance mechanism more fully with a wide spectrum of internal and external governance variables unique to the reform of SOEs in China.We find that our results are robust to the choice of different SFA specifications.Firm efficiency is negatively related to state ownership while positively related to public and employee share ownership.In addition, the relationship between ownership ownership program.Third, firm efficiency is greatly reduced when state or state-owned legal entities serve as the controlling shareholder, in part due to a combination of tunneling activities and government interferences.Hence, corporatization, commercialization and decentralization of SOE via the introduction of more non-government equity block-holders appears to be a useful way to balance the power of control and help to reduce tunneling and expropriation of firm assets.Another way is to transfer state shares to investment holding companies, which will function more like profit-maximizing businesses than administrative branches of the government.
Fourth, although there is some evidence that corporate governance mechanisms (for example, board size and the appointment of independent directors) do not affect firm efficiency, it remains important that the government put in place a set of rules and regulations to ensure that the CEO position is separated from the board chairmanship, board composition is free of influences by the controlling shareholders, and all directors are committed to good corporate governance practices.In addition, the government should foster, or even mandate, better and improved financial disclosures and strengthen external corporate governance mechanisms by further increasing the degree of marketization.Improved financial transparency and law enforcement will help investors strengthen their monitoring roles, and will help firms design more effective incentive mechanisms that lead to efficiency gains.

Conclusion
Using data from 744 publicly listed manufacturing firms in China between 1999 and 2006, we explore the relationship between ownership structure, corporate governance and firm efficiency.Consistent with prior theories of SOE privatization, we find that state ownership exerts negative impact on firm efficiency while public and employee ownerships are beneficial for improving firm performance.The results suggest that strengthening internal incentive via employee share ownership is crucial for the success of SOE reform.
In addition, if the controlling shareholder is government or state-owned legal entities, firm efficiency declines.The relationship between ownership concentration and productive efficiency is of U-shaped.Moreover, internal corporate governance mechanisms, including board and supervisor committee, are ineffective in increasing productive efficiency.The extent of provincial marketization, a proxy for external corporate governance, does not affect firm performance either.Overall, these results indicate that restructuring SOEs via improvements in corporate governance may not be effective in improving firm efficiency.
Although we have obtained some new findings in this paper, we believe that a number of improvements are possible for future research.First, we only adopt two of the most widely used methods in estimating productive efficiency.Future research can apply more sophisticated techniques, such as the method of maximum simulated likelihood estimation, to address potential biases in sample selection and estimation.Second, to ensure that firms are homogeneous in their production, we only use samples from manufacturing industry.It might be interesting to analyze the relationship among ownership structure, corporate governance and firm efficiency using a larger data set with longer time series and more industries.Third, we only consider unique measures of ownership structure and corporate governance with respect to Chinese publicly listed firms.Future research can extend the methodology to accommodate a variety of institutional settings and market characteristics.

Table 1 .
Classification of the nine sub-industries.
where State, Public and Employee are the fractions of shares held by the government, the general public and employees, respectively (ownership structure); Largest is the fraction of shares held by the largest shareholder and Largest2 is the squared term for Largest (ownership concentration); Statectrl and Legalstactl are dummy variables representing the identity of the controlling shareholder, Statectrl takes the value 1 if the government is the largest shareholder and Legalstactl takes the value 1 if an SOE is the largest shareholder; Boardsize, Outdirector and Duality represent board structure and independence, Boardsize is the number of directors, Outdirector is the number of independent directors, Duality is a dummy variable that takes 1 if the general manager or is the mean productive inefficiency for firm i at time t and can be represented by:

Table 2 .
Summary statistics by sub-industries.

Table 3 .
Summary statistics by year.The data consists of 744 publicly listed manufacturing firms during 1999 and 2006 for a total of 5218 observations.Figures in parentheses are standard deviations.

Table 4 .
Maximum likelihood estimates of the relationship between corporate governance and firm efficiency (half-normal SFA model).Labor is the number of employees, Capital is the lagged one period net fixed assets, Intermediate is the sum of the costs of raw materials, energy and other productive factors, State is the fraction of total outstanding shares held by local and central government, Public is the fraction of total outstanding shares held by the investment public, Empolyee is the fraction of total outstanding shares held by workers and managers, Largest is the fraction of total outstanding shares held by the largest shareholder, Largest2 is the squared term for Largest, Statectrl is a dummy variable that takes 1 if government is the largest shareholder and 0 otherwise, Legalstactrl is a dummy variable that takes 1 if government-controlled legal entity is the largest shareholder and 0 otherwise, Boardsize is the number of directors, OUTDIRECTOR is the number of independent directors, Duality is a dummy variable that takes 1 if general manager/CEO is also the board chair and 0 otherwise, Supervisorsize is the number of supervisor, Outsupervisor is the number of independent supervisor, FIRMAGE is the number of years after going public for a firm, Firmage2 is the squared term for Firmage, and Marketization is a proxy for the strengthen of provincial market development.Year, province and sub-industry dummies are omitted for brevity.*, **, and *** denotes 10%, 5% and 1% level of significance, respectively.