Government directors

Размещено на http://www.allbest.ru/

FINAL QUALIFYING WORK

Theme: Government directors

Student

Imametdinov Vladislav Tamirovich

Scientific supervisor

PhD Economics, Carsten Sprenger

Moscow, 2016

Contents

Introduction

1. General theory and literature review

2. Data for empirical analysis

2.1 Sample creation

2.2 Problems - solutions

2.3 Dependent variable

2.4 Independent variables

3. Empirical analysis

3.1 Methodology

3.2 Robustness check

Conclusion

References



Introduction

director activity performance company

Quality of corporate governance is a very important and widely discussed issue around the globe. Especially it concerns transition economies where many mechanisms and regulations traditional for market economies are underdeveloped for apparent reasons. In the same time realigning interests of participants of corporate structures is an extremely complicated and multidimensional task due to combination of a fundamental mismatch between incentives of counterparts and the factor of asymmetric information.

In the light of this fact the impact of boards of directors as one of the most significant mechanisms for internal corporate governance control should be critically and thoroughly analyzed. Although the existence of benefits arising from activities of boards is unquestionable, the real weight of services provided by directors in the overall success of a company may vary significantly depending on both internal characteristics of a board and external circumstances.

There are at least two key aspects underlying the question of value added by directors regardless of their specialization. The first is associated with resolving the conflict between owners and managers of a firm, which is addressed via agency theory. Although the monitoring function of boards is not an object of our key interest in terms of this paper, it is absolutely inevitable to touch this issue during the discussion of boards.

The second perspective on boards concerns the advisory, knowledge and other more material resources which directors may provide to enhance the performance of a company. This aspect is closely related to their background and particularly last occupations, from which the specific opportunities originate. It is not a secret that government ties of a certain level may endow a person with a substantial influence and power, so from a view of resource dependence theory politically connected directors are extremely curious subjects. Again it should be noted that Russia is a transition economy, so the influence of state can hardly be overestimated.

The aim of this research is to figure out existing approaches of assessing the activities of directors and their impact on performance of a company to state and statistically evaluate the hypotheses about influence of a certain type of directors, namely, politically connected ones. Existing papers concerning political connections in terms of boards often treat it as a sideline factor and do not pay much attention to their analysis. On the contrary, papers which focus on this issue often address a narrow subgroup of politically connected persons (for instance, only past or current top politicians, or even representatives of a certain political party), omitting the analysis of a general view. For this reason in terms of this research we consider two general types of politically connected directors without specifying the time aspect.

The structure of this paper is as follows: at first we will review the past literature limiting ourselves to papers which help to build up a logical path from general concepts of corporate governance justifying the existence of boards to their main characteristics along with their determinants. Then we will acquaint the general ideas about interrelation between boards and performance of the firm until we finally specify our review to political side of the question. After that we will be ready to state our hypotheses and perform empirical analysis to derive conclusions.



1. General theory and literature review

What is the key problem?

To start a discussion of boards of directors and their features it is essential to figure out what factors determined their existence and their importance in terms of corporate culture.

The core issue here is a conflict between owners and managers of a firm. Not surprisingly, managers are willing to pursue their own interests, which often contradict with the interests of shareholders. This immediately brings us to an agency theory perspective on the existence of boards. In fact, one of the main functions of directors is to monitor the actions of managers, so that they deviate less from strategies derived on behalf of the owners. It is also obvious that managers and shareholders have different perception of risks, as owners bear them to a much greater extent. Thus moral hazard problem arises and the supervision of directors enables to mitigate the impact of asymmetry of information between principles and agent sides.

To sum up, the agency theory perspective enables us to consider boards as entities which reduce agency costs via monitoring actions of managers. Apart from that, many researchers evaluate boards as the key monitoring mechanism (Fama, 1980).

Is the existence of directors a panacea from agency problems?

The fact that boards are present in all joint-stock firms (it is one of obligatory conditions for incorporation), but corporate scandals still take place automatically implies that boards is not an ultimate solution. There are many reasons for that, probably the most obvious of which is that directors are also subject to a human factor and can make mistakes doing their job. In addition they also pursue their own interests, which may lead even to fraud actions.

The existence of other mechanisms of corporate governance dealing with the same problems also indirectly confirms the fact that boards are not almighty in terms of reducing agency costs.

Another generally known alternative to boards of directors is the managerial ownership (Jenses and Meckling, 1976). The idea is that the ownership structure may dramatically affect the amount of risk taken during the decision-making process. Concentration of ownership among top management increases risk aversion of agents, binding their interests with interests of non-managing owners. In contrast with boards system there is no third party involved as a controlling entity, changes occur in the internal intentions of top managers.

Similar idea may be applied to lower level management via binding their payments to the entire performance of a company. This remuneration system enforces interests of managers to partially realign with perspective of owners.

Concentration of ownership as a corporate mechanism does not only concern top management, it is also possible that a big shareholder which is not involved as an employee directly controls the processes in a company, because he has enough power and influence. But as an effective monitoring requires a substantial amount of time and efforts, monitoring via a well-organized board seems to be a more preferable and convenient way for most of shareholders.

Apart from that there are mechanisms such as managerial labor market (beneficial role of competition between managers) and market for corporate control (associated with risks of takeovers), as well as debt financing (Agrawal and Knoeber, 1996).

Is the presence of boards only a pill from agency problems?

From the previous analysis we may conclude that a board of directors exists to act as an overseer and to be an effective solution to a fundamental problem. But if we try to figure out what is required from directors except independence from managers we may conclude that they need special competence and skills. At least it is essential in terms of agency theory to effectively monitor the quality of decisions. On the other hand we may abstract from monitoring and agency theory perspective as a whole and to consider special knowledge of directors as an informational resource. The latter enables boards of directors to act as consulting entities, enhancing performance with their strategic advises. That brings us to a resource dependence perspective, under which the board is considered as an enhancer instead of a purely controlling entity (Hillman and Dalziel, 2003).

In fact knowledge is not the only resource which may come from the representatives of the board. Often they have some special background and consequently a set of useful social, business and government links. Especially it concerns so-called outside directors, which basically stands for directors who are not current employees of the firm. On the contrary, inside directors are directly employed by the company.

What is special about outside directors?

From the agency theory perspective the main feature of outside directors is their independence from CEO, which enables to substantially decrease agency costs. But as we previously discussed outside directors may provide very useful and specific resources depending on their background, which is consistent with the resource dependence theory. There are five significant types of outside directors derived in literature (Hermalin, Weisbach, 2010). The first is associated with banking sphere. Such directors may provide easier access to funds and also to protect interests of banks which they represent, so their presence may be beneficial for all involved counterparties. The second is the venture capitalists. The presence of such director is typically one of the conditions for the corresponding kind of investment to occur. The third is the CEO of other company. Although such type is associated with human resources of a high quality, there is controversial evidence in literature concerning their impact, and the overall result highly depends on the circumstances. The fourth kind is a stakeholder participating in a board, in typical case it is a representative of workers protecting their interests. And finally we come to politically connected directors. As this type is of a special interest in terms of our research, they will be analyzed in more details later.

It is also important to note that the fact of the appointment of an outside director on average positively affects value of the firm (Rosenstein, Wyatt, 1990).

What determines the composition of boards?

Board composition plays a substantial role in terms of effectiveness of board activities, so before moving to analysis of the impact of boards on other factors it is extremely important to figure out what factors affect the composition itself.

The first paper addresses this issue using Russian companies as a foundation (Iwasaki, 2008). Before the empirical analysis some information on Russian special characteristics is given. The main point here is that Russia is a post-communist economy so many features of market economy are underdeveloped. For this reason establishing a well-organized internal corporate control is an issue of a special importance.

Nevertheless, governance mechanisms that we discussed previously in this paper (managerial ownership and ownership concentration) are widely spread in Russia, so the importance of boards of directors tends to be underestimated by individuals. However, minor shareholders are still present in most of Russian companies and in the light of the fact that they have lack of voting power and influence on decision-making processes, and also their coordination with other minor shareholders is often rather weak, it is argued that the presence of boards of directors is extremely important for protection of minority shareholders, which seems to be a hard task in terms of Russia (Lazareva, 2007).

For the empirical analysis of determinants of board composition the dataset of more than seven hundred Russian joint-stock companies is used. Univariate as well as multivariate regression analysis are performed and similar results are obtained in both cases. There are several categories of independent variables. The first group is derived using the idea that CEO has a certain amount of bargaining power to decrease the independence of a board. Such variables as ownership share of top management and ownership share of management group are included in this group. There is a straightforward logic that the more ownership is concentrated in hands of management, the bigger bargaining power the CEO obtains. Also there are several dummies such as newly appointed CEO, here it is obvious that a new CEO is likely to have lack of power and influence starting work in a company from a scratch. The results show that the bargaining variables have substantially greater statistical significance than other two groups (other governance and business-activity variables), so we will not consider them in details. As far as bargaining variables concerns, it is not surprising that empirical evidence confirms that the bigger is bargaining power of CEO, the less outsider directors are present (i.e. the less independent is the board).

The same results might be obtained from another perspective (Enya He and David W. Sommer, 2010). In this paper the relationship between outside directors and managers is addressed not in terms of a struggle, but rather in terms of what are the control requirements under a certain structure of separation of ownership and control. Here we once again address the agency theory. It is argued that the bigger is the extent of separation of ownership and control (i.e. the less is the share of managerial ownership) the bigger are the agency costs and hence the more independent board is needed to limit these costs. This hypothetically implies that the share of outside directors is positively correlated with the extent of separation of ownership and control, so that effective monitoring is performed to prevent agency conflicts.

The empirical analysis via panel regression shows that stock companies closely held by management have the lowest fraction of outside directors than other types. These results are consistent with the proposed theoretical implications.

Ownership concentration as a standalone factor is analyzed in other paper (Jiatao Li, 1994). It is argued that big shareholders may take a role of controlling leaders decreasing the impact of agency costs. This hypothesis is also consistent with the general theory which we discussed above.

Apart from concentration factor, in this paper additional aspects are analyzed, namely, the impact of bank control and state ownership on board composition.

As far as banks are concerned, they usually are major (or at least big shareholders), which seems to be a similar issue to ownership concentration. But there are also specific characteristics of banks which make them special in terms of corporate control. From the theory of financial intermediation we know that banks specialize on monitoring activities, apart from that they have easy access to internal processes of firms via checking their accounts. In result bank seems to be a perfect candidate to control investment activities of a company and as a result the need for additional instruments of decreasing agency costs appears to be less sharp. In terms of outside directors it implies that their share is likely to be negatively correlated with the extent of bank control.

The inverse situation is observed in terms of government control (degree of state ownership). In contrast with bank, state has less monitoring capacity as well as incentives to control the management. The primary objective of state is to realign strategic decisions of company with political interests of state. This can be achieved via adding politically loyal outside directors. This brings up the hypothesis that share of outside directors is positively correlated with state ownership.

Empirical evidence represented by OLS estimators of corresponding variables derived from a sample of almost four hundred large firms from different countries supports all three proposed hypotheses.

How board structure impacts performance?

In the first paper on this topic (Dan R. Dalton, Catherine M. Daily, 1999) the relationship between size of a board and performance of a firm is addressed.

Unsurprisingly, larger boards theoretically may have ambiguous effect on performance, which arises from the payoff between advantages and disadvantages of a big board. The hypothetical benefits straightforwardly come from the resource dependence theory which we already discussed previously. It seems logical, that the greater is amount of directors the bigger is the volume of resources they provide to a firm ceteris paribus.

Disadvantages of large boards seem to be less apparent. But the general idea here is the cohesion between directors, which implies that big boards are less likely to easily come to a consensus. Also the bigger is the board the easier it is (ceteris paribus) for a director to free-ride on monitoring activities.

The meta-analysis of about 20'000 companies (more than hundred samples) provides statistical evidence that there is a positive correlation between board size and performance of the firm (represented by different accounting and market indicators). This result is consistent with the resource dependence theory.

On the contrary, there is a paper in which the inverse conclusion is derived (A. A. Drakos and F. V. Bekiris, 2010). Empirical evidence here shows the negative correlation between board size and performance of the firm. The analysis of performance was made on the basis of Tobin's Q ratio. This inference is also consistent with some earlier studies (David Yermack, 1996).

Apart from board size Drakos and Bekiris consider independence of the board and leadership structure as additional factors in the same paper. As we already know board independence is associated with the fraction of outside directors. As far as leadership structure is concerned, the issue here is whether the CEO is simultaneously a chairman of the board (CEO duality).

However, the empirical analysis shows no statistically significant relationship between these two factors and performance of a firm.

In contrast, there is a paper that approaches the analysis of impact of two other mentioned board characteristics (independence of board and CEO duality) in a specific manner, obtaining statistically significant results (Catherine M. Daily and Dan R. Dalton, 1994). The idea is that there is a relationship between corporate governance and bankruptcy.

Theoretical implications about the impact of CEO duality are based on the concept that bankruptcy is usually a result of steady decline in firm's performance, which means that strategy adhered by management for the corresponding period should be changed or at least corrected to start a reversal of performance trend. The fact that strategic decisions to a great extent reflect the views and personality of a CEO, leads to a great probability of strategic rigidity if CEO has a substantial power and influence, which is implied under CEO duality.

The hypothesis about outside directors is based on similar logic, as higher independence of board increases the probability of deviation of board's view from the previously dominated strategy.

In total theoretically CEO duality should increase the bankruptcy probability, while the share of outside directors vice versa.

Empirical evidence based on logistic regression analysis is consistent with both hypotheses.

Unfortunately, the performance of a company is not always a story about legitimate actions. Corporate governance also has to deal with fraud of different types. In terms of our research, we are interested in a way boards can contribute to resolution of this kind of problem. The following paper enables to figure out this issue (Uzun, Szewczyk and Varma, 2004).

Among the characteristics of boards the independence, CEO duality, board size and committee structure are considered. Logit regression is used to perform tests.

The predictions made by researchers are mostly based on logic similar to previous papers. Share of outside directors and separation of CEO and chairman roles are associated with better controlling activities, which implies lower fraud probability. Higher board size is predicted to facilitate manipulation by CEO due to lack of cohesion. The new issue here is the discussion of committees. The specialization of directors in terms of committee structures is argued to make boards more effective (Klein, 1998). This idea is analyzed in terms of probability of corporate fraud.

Empirical results give no statistically significant evidence on the impact of board size and CEO duality. However, the hypothesis about outside directors is strongly supported. Also the impact of two types of committees is statistically confirmed, but surprisingly the signs are different. Presence of audit committees on average decreases the rate of corporate fraud, while compensation committees vice versa. The latter may be explained by ineffective operation of the corresponding structures.

What is the role of state and affiliated persons in a corporate environment?

Previous papers gave us a general picture of corporate governance and boards of directors as one of its core mechanisms. Now we can focus on a state as a significant participant of corporate activities. From the perspective of common sense the importance of government and its representatives can be hardly overestimated due to amount of resources and influence that a political connection of a high level can provide, especially in country with a transition economy, such as Russia. It is widely known that government plays a key role in macroeconomic processes via monetary and fiscal policies, setting the legal regulations and etc. But what happens when state or an affiliated person enters the corporate structure of a single firm?

In a general way this issue is addressed in the following paper (Frye and Iwasaki, 2011). The core aim of the research is to find out who benefits from this type of interaction. Namely, the model implies three hypothetical types of relations between state and firm. Typology is based on the combinations of benefiting parties, therefore assuming state and frim as counterparts there are three possible scenarios. The political connection of a company is defined by the presence of government directors on a corporate board of corresponding firm.

The first case implies that a company gains a substantial benefit is associated with a strict control by politically connected outside directors to limit agency costs and thus enhance the corporate governance of a firm. Resource dependence perspective also implies that these kinds of directors are predicted to be useful for company. A special feature of this type of relation is the prediction that state would tend to send its representatives primarily to firms with weak performance, which might imply the need for better governance procedures and managerial discipline.

Alternatively, under the second scenario the ruler may use directors for rent extraction from firms to satisfy the interests of powerful social groups to keep solid political positions. Unsurprisingly, under this model better performing firms constitute the focus group for sending representatives to extract more resources. The beneficial asymmetry leads to socially ineffective outcomes.

Under the third scenario firm and state both derive benefits from interaction. The features of the previous types are combined in a straightforward way, namely, ruler still uses strong firms to provide interest groups with rents, but simultaneously state representatives make efforts to monitor and enhance managerial discipline, as well as to provide firm with useful and performance-improving resources.

Empirical evidence supports the third type of relations, associated with collusion. That implies that overall effect for a firm may be ambiguous, depending on the extent of resource extraction.

But not all politically connected directors are directly appointed by the state. Some directors may have prior government involvement and hence have a relevant experience as well as acquaintanceship with influential persons. Thus they can help to deal with business associated with high degree of firm-state interaction and to provide useful resources from state and affiliated entities. But what are exact factors which determine the need for this type of directors. In other words, which variables are highly correlated with the share of directors with government background and experience?

The following paper helps to resolve this puzzle (Agrawal and Knoeber, 2001). Unsurprisingly, it is argued that the share of corresponding directors is higher for firms for which politics is a substantial determinant of performance. The key point here is how to measure the importance of politics for a firm. There are three kinds of such firms derived in the paper. The first is associated with a size of the firm, it is argued that increase in the scale of business activities leads to higher attention and regulation from state. So there is a sharp need in experienced advisors in terms of government procedures. The second type concerns firms for which state directly affects performance via government purchases, trade policy and environmental regulations. However, this list is not exhaustive in terms of reality but is a good proxy for research need. Finally, the third type includes companies with special importance of lobbying activities.

Hypotheses are tested using several methods, including OLS and logit models, under all approaches the results are similar. Summarizing the evidence we can say that for all types the predictions about higher percentage of politically connected directors are statistically significant. The only exception is the subgroup of the second type associated with environmental regulations, which does not have a statistically significant correlation with the number of corresponding directors.

In contrast with the previous paper, in the following research (Faccio, 2006) the effect is analyzed for current politicians (or closely related persons). The political connection of a firm arises from presence of such individual in a corporate structure of a company. The first implication is that the amount and concentration of such connections is not homogeneous round the globe. The second is that the significant effect (positive) is observed only if the previously affiliated person gains a serious political employment, not vice versa. The latter is consistent with previously discussed fact that extraction of resources by state may outweigh benefits (collusion type of relations between firm and state).

There is also empirical evidence that there is a significant relationship between performance and connection of board to a political party winning the elections. There is observed a positive difference in returns (Eitan Goldman, Jцrg Rocholl and Jongil So, 2009) between firms connected to winners and firms connected to losers, as well as increase in receipt of government procurement contracts for the former and decrease for the latter (Eitan Goldman, Jцrg Rocholl and Jongil So, 2013).

Derivation of testable hypotheses

The analysis of past literature also gives us several important implications for introduction of hypotheses and performing empirical tests:

The resource dependence theory implies that politically connected directors could possibly provide a substantial boost to profits using their influential ties.

The business-state relationship is a complicated system with high level of interdependence between participants, so we should control for state presence in the firm.

Apart from controlling for state ownership, we should also control for several fundamental determinants of the profitability of a company and least for one important board characteristic which may have a substantial interdependence with our target variables.

The first idea accompanied with a reasonable prediction that direct government employment implies higher level of access to specific resources leads to the formulation of the following hypotheses:

Politically connected directors associated with direct government employment (i.e. the bureaucrat, no matter current or past) on average have a positive effect in terms of performance of a company.

Indirectly politically connected directors (via employment in a state owned entity) on average have a positive effect in terms of performance of a company

The impact of the bureaucrats is higher than impact of the second type directors.



2. Data for empirical analysis

2.1 Sample creation

For the empirical analysis we need two categories of data. The first is connected to performance of firms, which in terms of this research is represented by accounting figures. The corresponding information was extracted from Ruslana database, prepared by Bureau van Dijk Electronic Publishing. The latter is one of the leaders in the field of producing business information across the globe, so the quality of financial data is high enough to obtain reliable results.

The accounting data is stored in several dimensions, namely, there are figures corresponding to balance sheet and profit and loss statements for a number of Russian publically traded companies for a sufficient number of years. Moreover, data represents consolidated as well as unconsolidated reports, for our purposes it is convenient to use data directly from consolidated reports and to address unconsolidated data only if the former is unavailable.

The pool of publically traded firms includes both associated with MICEX and RTS (and also MOEX, which is a product of the merger of these two entities in 2011). Companies were ranged by market capitalization and hence the pool of 200 biggest were chosen by this criteria (after exclusion of financial entities). The period for which the data is stored constitutes 7 years, namely from 2006 to 2012. Hence the appearance of MOEX lies inside the analyzed period.

The second type of required data is associated with characteristics of boards of directors of corresponding firms. This sort of information was obtained from quarterly reports extracted for the same period. These reports were mainly supplied by SPARK and SKRIN databases. Also the presence of reports enabled to derive the information on state ownership.

To obtain information on directors the section 5.2 of quarterly reports was addressed, where information on participants of controlling bodies of the firm is stored. In some reports the subsection 5.2.1 for the composition of board of directors is specified. In this subsection the data on occupations of each director for the past 5 years was extracted and was consequently used to figure out characteristics of each director and in result the overall composition of a board.

The ownership data is presented in section 6.2 of the quarterly report, where all big shareholders (with more than 5%) are reflected. If the report is properly prepared the information on direct state ownership is also reflected in section 6.3. But for the aims of the research indirect state ownership (for instance, via other state owned firm or non-profit organization) was also traced.

In the Figure 1 (may be found after References) the distribution of observations across industries is presented. The table shows that most observations correspond to electric power industry (30%), than goes engineering industry with 14%, oil and gas industry is the last from top-3 industries with 10%. Chemical and petrochemical industry and mining industry also occupy a substantial share with approximately 8% each.

2.2 Problems - solutions

During the process of obtaining data several problems were faced:

The main problem was the absence of quarterly reports for many companies. For most of them the selective lack of reports was observed (for instance, there was no data for a couple of years or a certain quarters). But in some cases there were no reports at all. The first solution naturally came from the fact that there were two available sources, so although SPARK was the main instrument for obtaining reports, in case of the absence of relevant reports they were checked in SKRIN. If there was still a substantial lack of information (for more than 1-2 years) the company was basically excluded from the sample.

The same problem applied to accounting information. This also led to further reduction of the overall sample. After the exclusion of all problematic companies and some years for certain companies for which the lack on data was present the amount of observations was still sufficient for our purposes.

The data associated with occupations of directors is characterized by a great amount of verbal information, so the combination of hand work and key words analysis was implemented in order to process this data and obtain indicators of different types of directors.

2.3 Dependent variable

It was decided to use accounting indicators as a measure of profitability of the firm, more specifically, the accounting ratios. So the two most apparent choices here were ROA and ROE as commonly used instruments to assess profitability of a firm, as well as to compare firms between each other. In addition they are based on standard and easily available accounting data, which is a great advantage.

The choice between these two options was made in favor of ROA (list of all variables may be found in the Table 1) as it was considered a more convenient candidate. At first it is less subject to problem of outliers and also assets may not be negative. In addition the chosen leverage structure does not affect ROA directly via calculation.

In the Figure 2 there are presented the average values of ROA for each industry and year. As we can see the highest ROA corresponds to oil and gas industry. Before the recent sharp fall of oil prices this result was rather predictable. Also the average ROA above 0.1 is observed in chemical and petrochemical industry. Other industries performed substantially worse, especially trade and engineering, for which average ROA did not hit the 0.04 threshold. The leader in terms of number of observations (electric power industry) along with the transport industry performed slightly better but did not hit the 0.05 score.

In terms of time unsurprisingly the average ROA faced a sharp fall during the crisis in 2008. It is noteworthy that substantial downward trend started in 2007 with a sharp fall of ROA in that year and hit the bottom in 2009. In the following years the consequences of crisis smoothed a bit but the situation was still much less optimistic than in the beginning of the analyzed period.

2.4 Independent variables

The most relevant characteristic of boards in terms of our aims is the share of politically connected directors. As we already stated, we derive two subtypes of such directors, namely, the direct employee of the government and a person affiliated with a state owned entities. It is important to again emphasize that it does not matter whether the involvement is current or occurred in the past for both types.

From Figure 3 we can see that the highest average share of directors of the first type (STN variable stands for share of bureaucrats) corresponds to communications industry (slightly higher than 0.2). Almost the same score but barely lower than 0.2 corresponds to transport industry. Electric power industry also shows rather high figures with almost 0.15 of average share. For other industries figures are significantly lower, especially for food and trade industries, which both show indexes below 0.05. For other industries the average share is the lowest with about 0.02 of bureaucrats in the board of directors.

As far as time concerns, it was observed an apparent increase in share of bureaucrats in 2008, which is potentially also connected to crisis. Particularly, firms might need a support from state and thus the demand for government directors reasonably increased. In the following years there was no particular trend concerning the share of bureaucrats but interestingly the figures for the beginning and the end of analyzed period almost coincide.

As we will see in the further analysis directors of the second type will have a smaller role in terms of empirical research (SON variable stands for their share), so we will not discuss their descriptive statistics in details. However, we can see from Table 2 that the overall mean of their share is almost twice as big as that of bureaucrats.

We are also interested in the sizes of boards (variable N) for two reasons. At first these figures were required to calculate the shares which we discussed above. But in addition as we could see from literature the size is one of the main characteristics of boards and widely discussed issue in the field of corporate governance, particularly in the context of performance of the company. As it potentially may also have a substantial relationship with our target variables, we should account for its impact on performance.

The presence of state in a company definitely plays a great role in our context. It is important to control for state ownership, as this factor is closely related to both presence of politically connected directors and overall performance of a company.

There also should be applied control variables which are abstract from boards and their composition, but have a substantial impact on profitability. Widely used indicators for these purposes are leverage and size of the firm. Different indicators may be used as a proxy of size. Usually it involves taking natural logarithm from total assets, total capital or sales revenues, but other specifications may also be applied (David Yermack, 1996). In our case it was decided to use log of total assets.



3. Empirical analysis

3.1 Methodology

Since we need to find out the relation between ROA and a group of independent variables based on panel data (multidimensional in terms of companies and time), apparently we can use the standard regression model for panel data analysis, which includes unobserved heterogeneity term.

Where is an unobserved heterogeneity term, X stands for observable variables, is the time effect, i and t are unit and time indexes respectively, j is the summation index for observable variables.

The term incorporates firm-specific factors which do not depend on time but have a substantial effect on dependent variable. The estimation by Pooled OLS method implies the assumption that alpha simply equals zero, i.e. there is no unobserved heterogeneity between companies. But if it is not the case in reality we would obtain the omitted variable bias, inconsistent and inefficient estimates.

However, we can start with simple Pooled OLS method to gain some initial estimates.

We should start our analysis by addressing the correlation matrix (Table A1). As we can see there is no apparent relationship between ROA and STN at this point, although it has expected sign (positive), the degree of correlation is very small. For this reason control variables are used, so in further analysis we will analyze this relationship more properly. Correlation between ROA and SON is slightly bigger in absolute terms but has unexpected sign (negative). The sign of correlation between N and ROA is consistent with cohesion perspective on the size of the boards implying that bigger boards are less efficient. Relatively high correlation between N and both SON and STN may be explained by the fact that companies with bigger boards tend to diversify the composition of their boards more than companies with smaller ones and thus attract more directors with specific background. Unsurprisingly, the size of a company(SIZE) have a significant correlation with the size of a board(N), which also explains the fact that correlation patterns of SIZE and shares of both types of politically connected directors are similar to those of N. On the other hand, the same correlation patterns (rather high and positive) apply to share of state ownership (STOWN) and both STN and SON, which was also predictable due to several factors, at first state tends to appoint its representatives to firms with state ownership, and also firms associated with state need directors experienced in this field, in addition inside directors working in state owned companies also counted as a part of SON. The positive correlation between SIZE and STOWN is also consistent with the implication that state tends to deal with big firms. The negative correlation between STOWN and ROA is more consistent with the process of extraction of resources from a firm.

To make further conclusions we should run the regression using the Pooled OLS method (Table A2). Here we can see that all signs except for SON and STOWN are the same as for correlation matrix. Now the effect of SON is consistent with the predictions, but the p-value is big enough to drop the SON from the specification and concentrate on STN for following steps. Although STN is also statistically insignificant at this point, it is rather close to a 10% threshold and may perform better if we obtain more robust estimates. Board size is very significant for this specification, while state ownership is not.

Running Pooled OLS without SON we get the same results except that now STN is closer to become statistically significant at 10 percent level (Table A3).

Up to this moment the hypothesis about direct employees of government seems to be rather promising. But there is still a possibility of existence of unobserved heterogeneity. Fortunately, there are approaches other than Pooled OLS which may resolve this sort of problem.

As we had to selectively exclude some observations for companies which are still present in the sample thus our data is unbalanced (Table A4), but we still can apply specific panel data methods.

The fact that unobservable term may differ from zero does not reject the possibility that it is still constant. This assumption ( = const.) lies in the basis of Fixed effect methods.

There are several commonly used approaches under this framework. The first one implies subtraction of the lagged model from initial one. If we use one period lag the commonly used First difference method is implied. The idea here is that as the unobserved effect is fixed, the subtraction leads to its total removal. But instead the difference model generates the moving average type autocorrelation, which implies decrease in efficiency and invalid tests.

The second is associated with adding dummy variables for individual effect of each firm and running a regression. Consequently the test for joint significance of dummies is performed. However it is not a technically convenient approach in practice for our purposes.

The third approach (within group) is based on deviations of variables from their means, i.e. the steady state equation is subtracted and as a result constant unobserved term crosses out.

The latter two methods give the same estimates, so within group method is usually used due to apparent technical reasons.

The transformations under different fixed effect approaches inevitably lead to loss of constant variables as well as loss in degrees of freedom.

Applying the within groups method (it is set as a default instrument in Stata) to our model we obtain slightly different results (Table A5). The main consequence is that the target variable STN still has a predicted sign and is observed to be statistically significant at 10% level.

But if unobserved heterogeneity term is uncorrelated with regressors the Random effect method is more appropriate (Table A6). The obtained results are very similar to those of Pooled OLS, but before making any inferences we should check whether the Random effect framework is applicable in our case. In order to resolve this issue we should figure out the character of the mentioned correlation.

To check this we should apply the Durbin-Hausmann test to decide which approach will give more robust results in this particular case (Table A7). As we can see from test statistics we need to use the Fixed effect framework. But there is still an issue which may deteriorate the quality of estimators, namely, heteroscedasticity. In case of detection we may obtain more robust estimates correcting for this problem.

At first we use a Modified Wald test for heteroscedasticity (Table A8). The results imply that we are subject to this problem. The corrected version of the model still provides evidence of statistical significance of STN (Table A9).

3.2 Robustness check

Since STOWN and N are slightly out of 10% significance zone, it would be appropriate to check the robustness of obtained results via changing the specification of the model, as the relationship between factors in a true model may be nonlinear.

For the first check we should interpret state ownership as a dummy variable (in this particular case we use Pooled OLS with consideration of heteroscedasticity). Substituting STDUM instead of STOWN we obtain favorable results in terms of STN, while the fact of the presence of state is less statistically significant than the share of the presence. (Table A10)

If we use STOWN^2 instead of STOWN under the fixed effect robust framework the overall inferences do not change. (Table A11)

But what if we abstract from state ownership and analyze the effect of government employees for purely private companies (with zero ownership of state)?

For this aim we take away firms with state ownership from the sample and apply the same specification and method (except including STOWN).

The results concerning STN are still aligned with the previous ones (Table A12). Now we have definitely a serious argument in favor of significance of bureaucrats in terms of profitability of a company.

However, the latest results imply that N is significant and its sign is consistent with the resource dependence theory rather than coherence view. Let us change the specification of N and apply it for a full sample for the last check. Generating BSIZE = log (N) and running a fixed effect robust regression we observe that our finding bears a fruit (Table A13). But the main inference here: STN is again observed to be significant (at 5% level).



Conclusion

The empirical analysis provides robust evidence that share of bureaucrats presented on a board positively affects performance of a company, which supports the first proposed hypothesis. On the other hand share of directors connected to government affiliated entities is observed to be statistically insignificant. Thus the second hypothesis may not be accepted, which supports the third hypothesis that bureaucrats play a more significant role in terms of performance. However there is still a broad field for research, as the assumptions for the second type of directors are rather strict due to complications in tracing indirect state ownership for some companies. Nevertheless the list of such companies composed for this research is sufficient enough to make reasonable inferences. Another cornerstone of all researches in the field is the issue of reverse causation between board characteristics and performance of a firm. Most of papers fail to deal with this problem as reliability of most proposed instruments is fairly questionable, while the short list of good ones is characterized by a hardly accessible data and usually applicable to certain specific cases. But still the statistical evidence on the first type of directors, namely bureaucrats, appears to be rather convincing.



References

A. Drakos and F. V. Bekiris (2010). Endogeneity and the relationship between board structure and firm performance: a simultaneous equation analysis for the Athens Stock Exchange. Managerial and Decision Economics , Vol. 31, No. 6, pp. 387-401

Amy J. Hillman and Thomas Dalziel (2003). Boards of Directors and Firm Performance: Integrating Agency and Resource Dependence Perspectives. The Academy of Management Review. Vol. 28, No. 3, pp. 383-396

Anup Agrawal and Charles R. Knoeber (1996). Firm Performance and Mechanisms to Control Agency Problems between Managers and Shareholders. The Journal of Financial and Quantitative Analysis, Vol. 31, No. 3, pp. 377-397

Anup Agrawal and Charles R. Knoeber (2001). Do Some Outside Directors Play a Political Role? The Journal of Law & Economics. Vol. 44, No. 1, pp. 179-198

April Klein (1998). Firm Performance and Board Committee Structure. Journal of Law and Economics, vol. 41, i. 1, pages 275-303

Catherine M. Daily and Dan R. Dalton (1994). The Academy of Management Journal , Vol. 37, No. 6, pp. 1603-1617

Dan R. Dalton, Catherine M. Daily, Jonathan L. Johnson and Alan E. Ellstrand (1999). Number of Directors and Financial Performance: A Meta-Analysis. The Academy of Management Journal , Vol. 42, No. 6, pp. 674-686

David Yermack (1996). Higher market valuation of companies with a small board of directors. Journal of Financial Economics, 40, 185-211

Eitan Goldman, Jцrg Rocholl and Jongil So, 2009. Do Politically Connected Boards Affect Firm Value? The Review of Financial Studies , Vol. 22, No. 6, pp. 2331-2360

Eitan Goldman, Jцrg Rocholl and Jongil So, 2013. Politically Connected Boards of Directors and The Allocation of Procurement Contracts. Review of Finance, 17, pp. 1617-1648

Enya He and David W. Sommer (2010). Separation of Ownership and Control: Implications for Board Composition. The Journal of Risk and Insurance, Vol. 77, No. 2, pp. 265-295

Eugene F. Fama (1980). Agency Problems and the Theory of the Firm. The Journal of Political Economy, Vol. 88, No. 2, pp. 288-307 Jenses and Meckling, 1976

Hatice Uzun, Samuel H. Szewczyk and Raj Varma (2004). Financial Analysts Journal , Vol. 60, No. 3, pp. 33-43

Ichiro Iwasaki (2008). The determinants of board composition in a transforming economy: Evidence from Russia. Journal of Corporate Finance, 14, 532-549

Jensen and Meckling (1976). Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics, V. 3, No. 4, pp. 305-360

Jiatao Li (1994). Ownership Structure and Board Composition: A Multi-Country Test of Agency Theory. Managerial and Decision Economics , Vol. 15, No. 4, Special Issue: Aspects of Corporate Governance, pp. 359-368