Размещено на http://www.allbest.ru/

The Government of the Russian Federation

Federal State Autonomous Institution for Higher Education National Research University Higher School of Economics

St. Petersburg Branch

St. Petersburg School of Economics and Management

CORPORATE GOVERNANCE AND FIRM PERFORMANCE IN CENTRAL ASIA

Master's dissertation

Area of studies 38.04.08 «Finance and Credit»

Master Programme “Finance”

good corporate governance asia

KADAMOVA FARANGIZ ODILBEK KIZI

Research Supervisor

Professor, Dean of St. Petersburg School of Economics and Management,

Department of Finance

Elena Rogova

Saint Petersburg - 2019

• Contents
• Abstract
• Introduction
• I.Literature review
• II.Overview of Corporate Governance in Central Asia
• III.Data Detail, Methodology and Results
• Conclusion
• References


Abstract

Good corporate governance is one of the major drivers of successful economic development and enterprise value creating in both developed and developing countries. However, in the last two decades, there has been very little research on corporate governance and firm performance in Central Asia, whereas the amount of research works on corporate governance influence on companies' performance in West and South Asia has increased significantly. This paper examines the corporate governance impact on firm's performance in Central Asia and suggests algorithms of identifying its main contributing factors. I hypothesize that firm performance in developing countries of Central Asia is likely to benefit more from higher level of corporate governance. Additional hypothesis is that corporate governance is positively associated with firm performance and firm profitability. I use different regression models to test my hypothesis on panel data. The main variables will be various governance features such as size of the board, ownership structure, CEO duality and independence director ratio to better understand the issue of corporate governance. As regards data collection, the main sources of information for my research are national stock exchanges where chosen companies are listed. Furthermore, collected data will be developed by annual reports data of companies. The results will show interpretation of numerical results and recommendations for companies in Central Asia. In conclusion, the topic of my research combines both theoretical and practical aspects of corporate governance and firm performance in Central Asia since studies in this topic are rare. The research will be important for investors, firm managers and directors of companies in Central Asia that aim to develop the corporate governance structure of their company and enlarge its firm value.

Keywords: corporate governance, firm performance, developing countries, Central Asia.



Introduction

Good corporate governance is one of the major drivers of successful economic development and enterprise value creation in both developed and developing countries. However, in the last two decades, there has been very little research on corporate governance and firm performance in Central Asia Central Asia - the region which consists of the former Soviet republics of Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, and Uzbekistan (Encyclopedia Britannica, 2019)., whereas the amount of research works on corporate governance influence on companies' performance in Asia at large has increased significantly. This paper examines the corporate governance impact on firm's performance in Central Asia and suggests algorithms of identifying its main contributing factors. This topic has raised interest of professionals from different countries, and scholars have used different approaches to estimation of corporate governance influence on firm's performance. The majority of previous studies suggest size of the board, board committees, board remuneration and independent directors as corporate governance level indicators; firm size, firm age and leverage ratio as control variables that may have an impact on firm performance; while the parameter of firm performance itself are usually financial performance indicators such as return on asset, return on equity, return on sales and Tobin's Q (Mashayekhi, B. and Bazaz, M., 2008; Iqbal, K. and Kakakhel, S., 2014; Goel, P. and Ramesh, R., 2016 ). Previous studies showed that using several methods raises the accuracy of the estimation of corporate governance influence on firm performance, and suggested algorithms of choice of right methods for different regions and companies that help to improve evaluation process further.

Managers' decision making depends not only on financial characteristics of their companies but also on corporate governance instruments used by shareholders. That is why the analysis of relationship between corporate governance and firm performance is extremely important for understanding of company's business processes. I hypothesize that firm performance in developing countries of Central Asia is likely to benefit more from higher level of corporate governance. Additional hypothesis is that corporate governance is positively associated with firm performance and firm profitability. As regards data collection, the main sources of information for my research are financial statements of the chosen companies. Furthermore, collected data will be developed by annual reports of companies. The results will show interpretation of numerical results and recommendations for companies in Central Asia.

This study aims to investigate relationship of corporate governance and firm performance in terms of financial perspective of Central Asian companies in different industries such as automotive, gas and oil, power supply, constructions, food and beverages and others.

The topic of my research combines both theoretical and practical aspects of corporate governance and firm performance in Central Asia since studies in this topic are rare. Therefore, this study is extremely of current interest and has a lot of practical corporate governance policy implications that are important for firm managers and directors of companies in Central Asia which lead to effective development of the corporate governance structure and enlargement of company's firm value. At first, it gives the evidence on the influence of corporate governance on firm performance, which may be the foundation for further Board structure recommendations for not only directors in Central Asia countries, but also for those from other developing countries. Secondly, it introduces the fundamental factors of corporate governance in Central Asia that influence financial performance of a firm. This will lead to the effective implementation of corporate governance mechanisms specifically in Central Asian companies.

The structure of the thesis is organized as follows: in Section I literature review of previous studies is presented, in Section II the overview corporate governance structure of Central Asia is analyzed. Section III presents the description of data collection, methodology and results.



I. Literature review

The past decades have seen the rapid development in firm-performance of companies in Asia. Most of them advanced their performance by developing their corporate governance structure. Moreover, there is a sustainable evidence on the fact that there is an obvious link between corporate governance instruments and firm performance provided by Boubakri et al. (2004), Bae et al. (2012), Aren et al. (2014). Corporate governance has undergone considerable changes in Central Asia as well. However, in the last two decades, there has been very little research on corporate governance and firm performance in Central Asia, whereas the amount of research works on corporate governance influence on companies' performance in other parts of Asian region, such as Middle East or South Asia, has increased significantly. The purpose of the present study is therefore to ascertain the corporate governance impact on firm's performance in Central Asia and identify its main contributing factors.

To understand how corporate governance affects firm's performance it is first necessary to discuss the corporate governance itself and main corporate governance factors that can have a measurable impact on firm's performance. In consistence with Cadbury Committee (1992), corporate governance is “the system by which firms are directed and controlled”. This definition is one of the first definitions of corporate governance that had been used by the considerable number of researchers in a field of finance. However, it does not specify strategies and instruments by which the company is managed. OECD (2004) introduced a broader definition of corporate governance. It states: "Corporate governance is a system that includes procedures and processes in accordance with which a company is directed and controlled. The corporate governance structure determines the distribution of rights and responsibilities among the different participants in the company - such as the board, managers, shareholders and other parties - and lays down the rules and procedures for decision-making". This research will be based on OECD definition of corporate governance since it covers all main aspects of corporate governance that may play a critical role in evaluation of its effect on firm performance.

Prior studies that have noted the importance of corporate governance as a driving force of firm performance highlighted different corporate governance reforms that had a significant impact on various types of firm performing differently before and after reforms (see, for example, Broadman, H., 2000; Black et al., 2015; Mitchell, A. and Wee, C., 2004; Mickiewicz, T., 2009; OECD, 2011). Moreover, the type of reform appears to be positively related to both labor productivity and firm profitability (Wen M., 2005).

As regards firm performance definition, it is now necessary to consider that firm performance, commonly known as organizational performance, is a notably broad term. As it is stated in (Cameron, K., 1986): “organizational performance indicates the actual output or results of an organization as measured against its intended objectives and consists of three firm outcomes: financial performance, product market performance and shareholders return”.

However, numerous studies have attempted to explain firm performance from different perspectives. For example, Selvam et al. (2016) examined a considerable number of works on firm performance and introduced multidimensional model of firm performance determinants. In their study, researchers draw our attention to two main distinctive categories of firm performance (financial performance and strategic performance) that encompass nine performance determinants. This model is supported by relevant ratios and parameters to each determinant. Notwithstanding their thorough study on determinants of firm performance, Selvam et al. (2016) did not make an attempt to test their model on real companies and present empirical results.

Thus far, several studies demonstrated that not only financial and strategic terms of companies can reveal firm performance. In a comprehensive study of culture and corporate governance, Adnan et al. (2018) examined corporate social responsibility disclosure as one of determinants of firm performance providing a thorough investigation based on more than 200 companies' information in China, Malaysia, India and UK. One unanticipated finding of Adnan and colleagues was that cultural influences and internal corporate governance significantly affect corporate social responsibility reporting. This view is supported in a fresh research by Hernandez and colleagues (2019) who write that firm's information disclosure plays a critical role in evaluation of internal corporate governance level of a company. Liao (2018) found another interesting fact that economic freedom of a firm empowers firm performance in terms of firm value and profitability.

In a recent study, Ciftci et al. (2019) found that ownership concentration considerably affects firm performance. What is surprising is that it is family ownership concentration. Another important finding was that more unique factors such as religion and social heterogeneity of a region affect firm performance (Mertzanis et al., 2019). Since these researches used companies in Turkey and MENA region which are close to companies in emerging markets such as in Central Asia and Middle East, the results are extremely useful for the current research. Despite this, little progress has been made in the evaluation of ways by which corporate governance influence firm performance in this region.

Up to now, a number of studies highlighted corporate governance factors that are associated with firm performance. The wide range of factors such as advertisement policy, labor productivity, or capital structure, play pertinent role in development of firm performance.

The majority of previous studies suggest size of the board, board committees, board remuneration and independent directors as corporate governance level indicators; firm size, firm age and leverage ratio as control variables that may have an impact on firm performance; while the parameter of firm performance itself are usually financial performance indicators such as return on asset, return on equity, return on sales and Tobin's Q (Mashayekhi, B. and Bazaz, M., 2008; Iqbal, K. and Kakakhel, S., 2014; Goel, P. and Ramesh, R., 2016)

A broader perspective has been adopted by Aren et al. (2014), who widened the number of factors that indicate corporate governance level, which include independent audit from Big four companies (PWC, E&Y, KPMG, Deloitte), the independence board, independent board members that have a qualification by Capital market board, with excluded legal regulations. Researchers found that corporate investor ratio, external financial needs and firm size appear to be main determinants of corporate governance. This result seems to be one of the reasons why investors are concerned about corporate governance level while choosing the firm to invest in.

Some writers (e. g. Pillai and Al-Malkawi, 2018) found that corporate governance factors like government shareholdings, type of auditing company, board size, corporate social responsibility and leverage significantly have a considerable impact on firm performance. However, Ciftci et al. (2019) found no link between the size of a board and firm performance.

Other researchers (see Black et al., 2015; Bae et al., 2012) highlighted the relevance of director's expropriation incentives, as one of the key corporate governance indicators, to firm performance. And this finding appears to be valuable for current research due to region-specific reasons.

Another important issue was investigated by numerous authors, whose research was based not only on the standard corporate governance indicators that were listed previously, but also on new developed parameters. Drawing on an extensive range of sources, the authors (see Ararat et al., 2017; Black et al., 2015; Iqbal et al., 2019; Kumar et al., 2017; Hermassi et al., 2017) set out the different ways in which corporate governance index was introduced as a more accurate parameter of firm's corporate governance level.

Notwithstanding that most studies demonstrate a positive impact of corporate governance on firm performance, some studies (e. g. Ofoeda, 2017) generated mixed findings on corporate governance processes and profitability. However, this research was based on Ghananian non-bank financial institutions and is not that important for evaluation of corporate governance influence in Central Asia.

There is a number of instruments available for measuring the corporate governance impact on firm performance, and different authors have measured it in a variety of ways: EVA analysis (Gong et al., 2011), decision tree and mapping (Hernandez et al., 2019), content analysis (Adnan et al., 2018) etc. However, the most widely used method were different variations of OLS and 2SLS regression models (Ofoeda, 2017, Mertzanis et al., 2019, Iqbal et al., 2019 etc.).

The literature review presents insights into the relationship between corporate governance and firm`s performance. However, the synthesis of corporate governance factors affecting firm performance in Central Asia remains a major challenge for companies in these regions.

Existing studies mainly focus on measurement of firm performance in terms of financial performance. Most of them based on data that does not include companies in Central Asia. According to Asian Development Bank (2019) most countries in Central Asia demonstrate extremely weak corporate governance structure. In the light of recent studies, it is becoming difficult to ignore the lack of research on region of Central Asia. Therefore, both qualitative and quantitative methods are needed in this investigation.



II. Overview of Corporate Governance in Central Asia

In the new global economy, corporate governance has become one of the central issues in Central Asia. Over the past century, a wide range of companies in this region has developed processes of property separation rights which led to the management problems of this property. Business development issues have made corporate owners to delegate management functions to other parties. The conflict of interest between managers and shareholders of the company have emerged as one of the main obstacles in company development. This is the reason why in recent years it became extremely difficult to ignore the poor corporate governance not only for market participants, but also for government parties of each country.

Formal investor protection in Central Asia countries was strengthened by regulations of corporate governance in Central Asia that are based on governmental laws and mainly on codes. Thus, there are codes of corporate governance in each country that are proceeding from expert guidelines and standards of international practices of corporate governance contingent on provisions of the current legislation of a country. Today each country in Central Asia has the Code of Corporate governance which follows virtually all recommendations of the OECD Principles of Corporate Governance (OECD, 2018).

Code of Corporate governance in Kazakhstan was inducted in 2005 (Nationalbank.kz., 2019). Republic of Uzbekistan also has the Code, which was established in 2015 (Gkk.uz, 2019). As for Kyrgyz Republic, the Code of Corporate governance enacted in 2012 and last its last updates were in 2018 (Fsa.kg., 2019). Tajikistan does not keep up with neighboring countries since there are only principles of corporate governance for banks that were enacted in 2005 and are successfully used in banks and credit unions (Nbt.tj., 2005). Besides, Ministry of finance of Tajikistan introduced in 2010 the project of National standards of corporate governance, which principles follow all OECD recommendations (Minfin.tj., 2010).

It is important to mention that Codes of Corporate governance of Kazakhstan, Uzbekistan, Kyrgyzstan and Tajikistan have similar structure.

These Codes cover all aspects of corporate governance structure: requirements for General Meeting of Shareholders, Board of Directors functions, Executive agency and Corporate secretary role, Corporate events issues, Information disclosure and Control of financial and economic activities. Furthermore, it includes regulations of dividend policy requirements and regulation of corporate governance conflicts. It is highlighted that the voting process at the General meeting of shareholders should be transparent to shareholders and exclude the possibility of manipulating numbers when tallying votes. Voting is carried out according to the principle “one share - one vote” or by cumulative voting. The Codes of Corporate governance in Central Asia also suggest creating specialized committees such as committee of strategic planning, risk-management, remuneration committee, audit committee etc. According to the Codes, independent director:

a) is not financially or otherwise dependent on the management of the company (the controlling shareholder), large counterparties and companies-competitors;

b) is not a state representative;

c) does not belong simultaneously to the executive management;

d) does not represent consultants that working with the company;

e) receives remuneration only for work on the board of directors;

e) has the necessary qualifications;

g) has a good reputation.

In addition, the Executive body is recommended to initiate a regular analysis of significant corporate events of the company and present the results of such analysis to members of the board of directors, shareholders, investors and other interested parties. However, there is a little number of companies that conduct analysis of changes after material corporate events and, in case of handling such analysis companies usually do not provide the results to investors and other interested parties. Moreover, companies should disclose all necessary information and provide the availability of public information about the company to all interested parties.

In consonance with international standards of corporate governance(OECD, 2018), the internal control system is a process accomplished by the board of directors, the executive body and staff at all levels, aimed at improving the efficiency of firm performance, ensuring the reliability of financial reporting and compliance with current legislation and internal regulations of the company. Internal control system should Control system of financial and economic activity of a company, provide internal audit reports and control the implementation of external auditing.

Traditionally, all codes emphasize the development of a clear dividend policy. It is recommended to publish dividend policy information of a company in a periodical, provided by general meetings of shareholders, and to be posted on the company's website on the Internet, if the company has such an opportunity. However, most companies in Uzbekistan, Kyrgyzstan and Tajikistan do not post dividend policy information on the company's website.

The code of Corporate governance of Uzbekistan apart from principles also provides the model documents (on information, dividend, internal control, conflict of interest policies etc.) in order to provide practical assistance to joint-stock companies, as well as the qualitative implementation of the recommendations of the Corporate Governance Code in their activities. It is interesting to note that in all Codes of Corporate Governance regulations on environmental protection are presented, whereas the Code of CG of Uzbekistan does not provide principles of the most careful and rational attitude to the environment. Nonetheless, Uzbekistan outperforms other countries by monitoring the implementation of Code recommendations and providing detailed instructions on the stages of corporate governance principles implementation.

One unanticipated finding on corporate governance development in Central Asia is that an independent assessment of the corporate governance system in companies is carried out based on a questionnaire approved by ministries of finance (Gkk.uz, 2019). It is probably since the fact that non-compliance with the recommendations of the Codes does not entail the application of measures of responsibility by public authorities and companies have the right not to establish measures of responsibility to company officials for non-compliance with the recommendations of the Code or undisclosed information.

Studies of corporate governance in Central Asia are extremely rare. For example, researches on corporate governance of Uzbekistan, especially with statistical analysis of corporate governance level development are virtually nonexistent. I found only one paper (Ashurov Z.,2015) which analyzes corporate governance from domestic perspective, however, it suggests only qualitative results on the structure of corporate governance in this country compared to other countries. However, there are several theoretical and descriptive studies on corporate governance development. For instance, Khamidulin and Tashmatov (2018) highlighted in their research the significant influence of state ownership on the level of transparency of companies in Uzbekistan. Therefore, this study introduces the idea of state ownership elimination or decreasing government's power in the board of directors. Chinkulov (2016) describes the problems of Code of corporate governance implementation in Uzbekistan. According to this study, the Code of corporate governance have a recommendation character, however, the main aspect of the Code is the key principle «comply or explain», which requires the explanation of reasons why the company could not adhere the recommendation presented in the Code.

The study of corporate governance in Kyrgyzstan was made by IFC, however, only banks were analyzed.

Corporate governance development studies in Kazakhstan, Tajikistan and Kyrgyzstan are mostly undertaken by organizations such as International Finance Corporation under the World Bank (IFC,2011), European Bank for Reconstruction and Development (ERBD, 2015).

The IFC study(Ifc.org., 2011) on Tajikistan corporate governance showed that the companies participating in the study not only do not adhere to the best principles of corporate governance, but also do not comply with certain requirements of the legislation of Tajikistan that regulate the activities of joint stock companies. As the results of the study showed, the internal audit committee exists only in one joint stock company out from all 60 surveyed companies. There is a tendency to perform the functions of external audit by the company's auditor, which itself is a violation of the requirements of legislation and best corporate governance practices and indicates the absence of an independent audit report on the financial statements of a company. The annual report is publicly disclosed by 19 joint-stock companies, of which 13 joint-stock companies are open, representing 32% of the total number of respondents. It was found that the reason for the lack of public disclosure of information, according to most respondents (73%), is “no need to disclose it”. Interested party transactions (with affiliates) are approved by the executive body in 26 joint-stock companies, which accounted for 43% of the total interviewed respondents.

The IFC study (Ifc.org., 2010) on Kazakhstan found that most joint-stock companies are not familiar with the basic international documents in the field of corporate governance, however, the majority of Kazakhstan joint-stock companies consider improving corporate governance practices as an opportunity to improve the company's reputation. Moreover, information about the remuneration of members of the board of directors and the executive body is practically absent (85% of respondents noted this fact in JSCs, 70% in banks). In JSC, the main channel for obtaining information is the investor's request (59%), in banks - a public annual report posted on the bank's website. These studies are informative; however, they are still descriptive in nature since they were based on questionnaires.

III. Data Detail, Methodology and Results

Data collection

This study uses the panel data to investigate the relationship between corporate governance and firm performance in Central Asia. Data for this study was collected from different sources. At first, Thomson Reuters databases was thoroughly observed for data on Central Asia companies. However, there was no data both on financial indicators and corporate governance performance of Central Asia companies. However, there was the list of public companies in Kazakhstan, but data on each company was unavailable. Therefore, for the purpose of this study we focused only on companies which are listed on national stock exchange:

1. The Kazakhstan Stock Exchange (KASE), which was established in 1993, the year of introducing own currency of Kazakhstan.

2. Republican Stock Exchange «Toshkent» (UZSE), formed in 1994 and has more than 100 broker offices all over the Uzbekistan.

3. Kyrgyz Stock Exchange (KSE), that was established in 1994 in the form of non-state and non-profit organization.

4. Central Asian Stock Exchange (CASE) established in Tajikistan in 2015.

5. Ashgabat Stock Exchange was formed in Turkmenistan in 2017 and is the youngest stock exchange among Central Asia countries.

Table 1. Listed companies of Central Asia, 2020. Source: calculated by author.

	Kazakhstan	Uzbekistan	Kyrgyzstan	Tajikistan
Total number of JSC	1082	605	246	325
Listed JSC	201	109	26	1
Listed banks and financial institutions	114	43	24	1
Listed companies	87	66	17	0

The extraction of date from these five sources is considerably problematic. Since companies does not fully disclose the information, we had to search for financial and corporate governance information in annual reports of each company. All banks and financial institutions were not included to the sample due to difference in regulating structure compared to that of listed companies.

Figure 1. Data sample distribution by industry. Source: calculated by author.

The sample ended up with 170 companies from Kazakhstan, Uzbekistan and Kyrgyzstan (Table 1). Unfortunately, Tajikistan companies were removed since there is only one listed company on CASE and it is a bank. And data on Turkmenistan companies was not available at all and their stock exchange still does not list any joint stock companies since it has just started functioning in 2017. We hope that in future the date on companies in Turkmenistan will be available.

Methodology

The sample includes 680 observations: data for 4 years (2015 - 2018) for 170 listed companies in Kazakhstan, Uzbekistan and Kyrgyzstan in different industries such as oil and gas, mining, manufacturing, energy etc. (Figure 1). The interesting finding is that more than 50% of the all listed companies included in our sample have a state ownership of 50% shares of a firm and higher (Figure 1.1). Whereas managerial and foreign ownership makes up just 17% of the sample. This is consistent with studies of Central Asia research community, where authors emphasize the large amount of companies governed by state or state representatives (Ashurov, 2015, Khamidulin et al., 2018, Chinkulov, 2016).

The source variables are data on total assets (Totalassets variable), equity (Equity variable), liabilities (Totalliabilities variable) and net income (Netincome variable). Based on these variables, the quantitative indicators included in the regression models are calculated.

Figure 1.1. Data sample distribution by ownership structure with the share of 50% and higher. Source: calculated by author.

We included in our model variables which are relevant to prior studies on the relationship between corporate governance and firm performance. Table 2 presents the summary of widely accepted variables in previous studies.

Table 2. Popular variables used in previous studies. Source: constructed by author.

Research framework	Determinants	Studies
Corporate governance	size of the board, board committees, board remuneration, independent directors, audit, external financing needs, corporate investor, ownership structure	Mashayekhi, B. and Bazaz, M., 2008; Iqbal, K. and Kakakhel, S., 2014; Goel, P. and Ramesh, R., 2016; Nguyen et al., 2015; Ciftci et al., 2019
	CG index	Ararat et al., 2017; Black et al., 2015; Iqbal et al., 2019; Kumar et al., 2017; Hermassi et al., 2017
Control variables	firm size, firm age and leverage ratio	Pillai and Al-Malkawi, 2018; Mashayekhi, B. and Bazaz, M., 2008; Iqbal, K. and Kakakhel, S., 2014; Goel, P. and Ramesh, R., 2016; Ciftci et al., 2019
Firm performance	return on asset, return on equity, return on sales and Tobin's Q	Mashayekhi, B. and Bazaz, M., 2008; Iqbal, K. and Kakakhel, S., 2014; Goel, P. and Ramesh, R., 2016

Accordingly, we based our study on firm performance as a dependent variable and divided independent variables into two groups (corporate governance variables and control variables).

The dependent variable is a firm performance. To estimate firm performance according to previous studies two indicators were selected from the financial reporting of chosen companies: return on assets (ROA) and return on equity (ROE). Depending on which variable the constructed regression models will show the best results, in conclusion, one or another dependent variable will be selected.

Independent (explanatory) variables - are combined into the following groups:

1) group of control variables:

· the age of the company since listing on the national stock exchange (Age);

· the size of the company (equal to the logarithm of Total assets (Size);

· the Debt-to-Equity ratio, which characterizes the amount of financial leverage (DE);

2) a group of variables indicating corporate governance:

· the number of members of the Board of Directors (Boardsize);

· the proportion of independent directors to the total number of directors on the board (Independent directorratio);

· the CEO duality (that is, CEO also holds a chairperson position on the Board of Directors). This a dummy-variable, which equals to one when the two positions are combined and zero otherwise;

· the proportion of shares held by government (Governmentownership);

· share of foreign co-owners in the company (Foreignownership);

· share of non-state participation in the company (Nongovernmentownership);

· share of managers in the ownership of the company (Managerialownership);

· a dummy variable of Audit company that equals to one if this company is from the BIG4, and 0 otherwise (AuditbyBIG4).

For almost all companies, the amount of the company's participation shares is 100% (only a few have minority shareholders), therefore, when analyzing one of the variables should be excluded.

Hypotheses

The construction of research hypotheses considered the empirical literature on specific factors which tend to affect firm performance (Table 2). Besides, some studies (Boubakri et al., 2004, Bae et al., 2012, Aren et al., 2014) helped to finalize the formulation of research hypotheses in terms of positive and negative effects of explanatory variables.

Thus, several hypotheses consistent with existing theoretical and empirical literature were developed for current research:

Hypothesis 1: The board size positively affects firm performance (effect > 0);

Hypothesis 2: The CEO duality negatively affects firm performance (effect <0);

Hypothesis 3: The BIG4 audit positively affects firm performance (effect> 0);

Hypothesis 4: Age, size of the company and financial leverage positively affect firm performance (effect> 0);

Since this study is based on panel data, several regression models can be constructed to estimate the relationship between corporate governance and firm performance. We studied previous papers on corporate governance and firm performance and discovered that most studies use fixed effects and random effects models (Mertzanis et al., 2019, Ben-Hassoun, A. et al.,2018). Fixed effect model is widely accepted mostly because authors have data samples including factors that vary over time. Moreover, data sample companies usually have their own individual features that may or may not influence the dependent variable of a model. One of the reasons why researchers tend to use random effects model is that it allows time invariant factors to be included in the models. Indeed, we have the variable that indicates industry of a company in the data sample of a current study.

In addition, it is important to highlight that some studies imply generalized method of moments, which usually appears to be the more applicable to analyze panel data since it decreases the probability of autocorrelation and fixed effects issues (Akbar, S. et al., 2016, Nguyen et al., 2015, Wintoki et al., 2012). This study discusses some of the models that were used in previous researches: fixed effects (FE), random effects (RE), first difference (FD) estimator and system generalized method of moments (sGMM).

Descriptive analysis of variables

We performed the descriptive analysis of sample variables using STATA descriptive statistics (Figure 2).

Figure 2. Descriptive statistics. Source: calculated by author in Stata.

According to descriptive statistics, some variables have a considerably high variation in their values compared to the average (this applies to all financial variables, as well as many control variables). We also performed normality test of our variables using the Shapiro-Wilk test.



Figure 3. Shapiro-Wilk test. Source: Calculated by author in Stata.

In this test, the null hypothesis is that the variable is distributed normally. The null hypothesis is not rejected only for the CEODuality and AuditbyBIG4 variables. For the remaining variables, for all this hypothesis is rejected at the 5% significance level and for most - at 1%.

Regression analysis does not require the variables to be normally distributed, but this analysis allows us to see possible issues with the sample. One of the issues to be expected is heteroskedasticity in residual values, as the sample include companies of different size, which is reflected in most of the variables.

We performed correlation analysis to determine how close the relationships between the various variables are to the linear relationship. First, we considered the correlation between possible dependent variables.

Figure 4. Correlation 1. Source: Calculated by author in Stata.

As can be seen from our result, there is a weak connection between return on equity and return on assets, which is explained by the different values of financial leverage. Therefore, the models that will be constructed with both variables will not be equivalent.

We analyzed the correlations between firm performance variables and control variables (Figure 5-6).

Figure 5. Correlation 2. Source: Calculated by author in Stata.

Figure 6. Correlation 3. Source: Calculated by author in Stata.

The correlation matrix shows that the coefficients are low, i.e. those are nonlinear relationships which are mainly observed, and in some cases the relationship may be close to zero.

We also considered correlation of firm performance variables with corporate governance variables (Figure 7-8).

Figure 7. Correlation 4. Source: Calculated by author in Stata

Figure 8. Correlation 5. Source: Calculated by author in Stata

The maximum observed value of the correlation coefficient is 0.5265 which indicates the correlation between non-government and government ownership. Thus, according to the results of the correlation analysis, we can conclude that the relationship between ROA and ROE which is close to linear one does not exist; therefore, the problem of multicollinearity is not expected.

Regression analysis

As a first step to estimation of the relationship between corporate governance and firm performance this study uses fixed effect models on each firm performance variables (ROA, ROE) for all the variables indicated.

The sample includes observations on 170 companies; therefore, the FE-model estimates 169 free variables + constant and regression coefficients.

Model 1

According to the results of the FE-model estimation on ROE variable (Figure 9), it is observed that the following hypothesis is not rejected:

,

where: - is an individual effect for firm i.

Since this hypothesis is not rejected, the model of fixed effects in this case does not make sense, since this model highlights the influence of the observation effect on the overall effect.

In addition, in this model concerning ROE variable 7 out of 10 regression coefficients (not counting the constant) are insignificant. For those coefficients that are significant - the basic hypotheses are confirmed for financial leverage (variable DE) and the board size (variable Boardsize). For the size of the company (variable Size) - the basic hypothesis is not confirmed. The influence of other variables in this regression is close to zero.

Figure 9. Model 1 results. Source: author's calculations in Stata.

Model 2

We constructed the same model for the second firm performance indicator - ROA variable (Figure 10). In this model, the hypothesis is rejected. This means that for the ROA variable, the construction of the FE-model is meaningful, but there are 9 insignificant coefficients out of 10. In terms of the R-squared (intergroup, intragroup and general) values, the regression for ROA is inferior in regression quality in terms of ROE.

Moreover, in this regression, the basic hypothesis is confirmed for the size of the company, which contradicts the result of previous regression.

Figure 10. Model 2 results. Source: author's calculations in Stata.

Thus, models should be adjusted, or other variants of models should be considered in order to obtain a larger number of significant coefficient estimates.

Model adjustments

Since in the initial panel regression models constructed by all factors, many of variables turned out to be insignificant, we changed the approach of models' selection. In order to understand which variables, have a prominent relationship with firm performance this research applies several types of regression models. We have chosen four regression models: fixed effects (FE), random effects (RE), first difference (FD) estimator and system generalized method of moments (sGMM).

The next step is to run different types of simple linear regression models. This instrument gives us the opportunity to identify which type of models generates the best results. Besides, by estimation of different types of regression models we can designate variables that should be added to the regression in the first place so that significant estimates of the coefficients were preserved.

In all discussed regressions, robust standard errors are estimated (using robust function in Stata), since such standard errors consider the heteroskedasticity effect in the models, which is undoubtedly is the case in our research, since the data sample contains firms of different sizes.

Table 3. Predicted signs of the estimates coefficients on independent variables with the dependent variables ROA and ROE in FE and RE models. Source: constructed by author.

Independent variables	Predicted relationship of the estimates coefficients («+» - significant relationship, «-» - no significant relationship), 10% significance level
	FE-model	RE-model
	ROA	ROE	ROA	ROE
Age	-	-	-	-
Size	-	-	-	+ (Н0 is rejected)
DE	-	+ (Н0 is not rejected)	-	+ (Н0 is not rejected)
Boardsize	-	-	-	+ (Н0 is rejected)
Independentdirectorratio	-	-	-	-
CEODuality	-	-	-	+ (Н0 is not rejected)
AuditbyBIG4	-	-	-	-
Governmentownership Foreignownership Managerialownership	-	-	-	-

From comparison of FE and RE models (Table 3) we can see that a model with fixed effects for both ROA and ROE is insignificant even in regression on each of the factors. As for the random effects model, it is significant in regressions on ROE only. Whereas, the null hypothesis in RE model on ROE was rejected for firm size and the size of the Board of directors.

Similarly, we considered the FD ( First Differencies , i.e., the model constructed from the first differences of variables) and sGMM ( System Generalized Method of Moments - methods that estimates dynamic panel regression, in which, in addition to factor variables, the lagged values ??of the dependent variable are added (Table 4). It is important to note that the FD model did not include the Age variable, as the first difference of the Age variable is the same for all observations and equals to 1.

Table 4. Predicted signs of the estimates coefficients on independent variables with the dependent variables ROA and ROE in FD and sGMM models. Source: constructed by author.

Independent variables	Predicted relationship of the estimates coefficients («+» - significant relationship, «-» - no significant relationship), 10% significance level
	FD-model	sGMM-model
	ROA	ROE	ROA	ROE
Age	x	x	-	-
Size	+ (Н0 is not rejected)	+ (Н0 is not rejected)	+ (Н0 is not rejected)	-
DE	-	+ (Н0 is not rejected)	-	-
Boardsize	-	-	+ (Н0 is not rejected)	-
Independentdirectorratio	-	-	+ (Н0 is rejected)	-
CEODuality	-	-	+ (Н0 is not rejected)	-
AuditbyBIG4	+ (Н0 is not rejected)	-	+ (Н0 is not rejected)	-
Governmentownership Foreignownership Managerialownership	-	-	-	-

As a result, in FD model Size, DE, and AuditbyBIG4 variables affect firm performance in terms of ROA and ROE variables. It is noteworthy that for ROA estimation regression the FD model is preferable compared to FE and RE models.

At the same time, the dynamic panel sGMM regression showed good results on ROA estimation - firm size, board size, independent director ratio, CEO duality and AuditbyBIG4 influence firm performance in terms of ROA variable.

Thus, we can conclude that for ROE indicator estimation the regression based on random effects model is more appropriate, and for the ROA indicator, it is preferable to run a regression based on a dynamic panel model estimated using the sGMM method.

Random effect

Therefore, we started from regressions options with random effects for the ROE variable. The selection of the optimal model was constructed as follows: firslty, all variables that were significant in simple linear models (Table 3, 4) were included in the model. Secondly, those variables that turned out to be insignificant were excluded. Then these variables were added to the model one by one. The significance level was assumed to be 0,15.

Table 5. RE regression models on ROE variable at 0,15 significance level. Source: Calculated by author in Stata.

Variables	RE-models
	(1)	(2)	(3)	(4)
Age		-0,4049	-0,4014	-0,4029
Size	-0,4326
DE	0,08297
Boardsize		-1,384	-1,0306	-1,0638
Independentdirectorratio
CEODuality		-6,2538	-6,0191	-5,6286
Governmentownership	1,4414	4,4072
Foreignownership			-4,594
Managerialownership				-3,851
Constant	7,3572	11,1624	13,1095	13,0223
Regression characteristics:
R-sq within	0,3957	0,0124	0,0049	0,007
R-sq between	0,6358	0,0455	0,0386	0,0343
R-sq overall	0,4918	0,0238	0,0184	0,0171
sigma_e	19,2297	24,527	24,6795	24,6553

These models (Table 5) confirm that the variables considered in this study significantly affect the return on equity if of a firm (ROE), except the Independent director ratio variable - this variable was not significant either by itself or in combination with other variables included.

But this influence is not decisive in absolute value; therefore, in models constructed with these variables, a large proportion of unexplained variance remains. This especially appeals to models with control variables, which are mainly qualitative.

We also investigated whether the industry has an impact on these models. We added dummy variables of industry to these models and assumed that if the industry influences the value of ROE, then the coefficient with the added variable will be significant.

The results (Table 6) showed that only regressions with oil and gas and the construction industry dummies added were significant at 0,15 significance level.

Table 6. RE regression models on ROE variable with industry dummies included at 0,15 significance level. Source: Calculated by author in Stata.

Variables	RE models
	(1)	(2)	(3)	(4)
Age		-0,3824	-0,3831	-0,3782
Size	-0,4269
DE	0,0828
Boardsize		-1,3572	-0,9922	-1,0375
CEODuality		-6,5167	-6,2184	-5,8513
Governmentownership	1,5146	8,0343
Foreignownership			-4,1127
Managerialownership				-4,4877
Constant	7,6031	11,9734	13,794	13,968
Industryoil_gas	-1,661	-4,8676	-4,4902	-5,0804
Industryconstruction	-1,799	2,6372	-4,6434	-4,6767
Regression characteristics:
R-sq within	0,3957	0,0129	0,0053	0,0074
R-sq between	0,6377	0,0572	0,0486	0,0462
R-sq overall	0,4924	0,0288	0,0227	0,0223
sigma_e	19,2297	24,527	24,6795	24,6553

Given this, we can conclude that if the company belongs to the oil and gas or construction industry, then this affiliation affects ROE of the company. In all cases, this affiliation affects firm performance negatively, i.e. companies in these sectors have lower return on equity than companies from other sectors. Meanwhile, in other sectors, a significant impact on the value of ROE was non-existent.

System GMM

As a result of models' selection, we also built an extended sGMM model on ROA variable. To build this model, it was necessary to determine which variables were instruments - i.e. variables that are uncorrelated with errors but correlated with regressors.

Since many of the regressors considered are qualitative, i.e. variables change in a narrow range of values, it can be assumed that their correlation with the error will be low and we can use them as instruments for themselves.

We included all the variables in the regression, except the variables characterizing the share of ownership.



Table 7. Results of system GMM model on ROA at 0,15 significance level. Source: Calculated by author in Stata.

From Table 7 it can be shown that in construction of sGMM model all qualitative variables Boardsize, CEODuality, AuditbyBIG4, Independent director ratio, as well as dummy variables corresponding to each year were taken as instruments. Variables DE and ROA (-1) were endogenous variables. The variable Size was considered as exogenous variable. Thus, the result of sGMM regression model obtains significant estimates for all coefficients, except the estimate of the DE variable. 

We used Sargan-Hansen test for exogeneity of instrumental variables. Sargan-Hansen test showed that the hypothesis of their exogeneity is not rejected.

Excluding the variable DE (Table 8), we obtained all estimates of the coefficients that are significant at the 0.15 level.