Agency costs and corporate control devices in the Turkish manufacturing industry

The current issue and full text archive of this journal is available at http://www.emerald-library.com Journal of Economic Studies 27,6 566 Agency costs and corporate control devices in the Turkish manufacturing industry Mine UgÏurlu BogÏazicËi University, Turkey Keywords Agencies, Costs, Control systems, Ownership, Manufacturing industry, Turkey Abstract Documents evidence on the interdependence between the mechanisms used to control the agency costs in Turkish manufacturing firms where the external control devices are restricted and most of the firms have concentrated ownership. The ownership concentration, board size and composition, managerial shareholdings, institutional shareholdings, and family shareholdings are the selected devices. Evidence reveals that the proportion of insiders on the board is positively related to the percentage of family shareholdings and negatively related to the percentage of foreign institutional shareholdings and ownership concentration. Board size shows a significant negative relation with all the control mechanisms except the debt ratio. The finding that the managerially controlled firms have lower debt ratio than the institutionally controlled firms and the family controlled firms supports the entrenchment hypothesis. The capital market seems to complement the institutional shareholdings, family shareholdings, and ownership concentration in monitoring the CEOs. Journal of Economic Studies, Vol. 27 No. 6, 2000, pp. 566-599. # MCB University Press, 0144-3585 Introduction Agency costs arise within a firm whenever managers pursue their own interests at the expense of shareholders. When ownership and control are separated, the conflicts of interest between managers and shareholders constitute a source of agency costs. Several mechanisms can be used to reduce these costs. The primary disciplining of managers comes through external sources such as the capital market, the managerial labor market and the market for takeovers. Other control devices used to align conflicting interests of managers and shareholders are institutional shareholders, blockholders, managerial shareholdings, outside directors on the board and leverage used by the company. The primary disciplining of the managers may come through managerial labor markets both within and outside the firm (Fama, 1980), or may be performed by the risk bearers (Alchian and Demsetz, 1972; Jensen and Meckling, 1976) who depend on the signals provided by an efficient capital market and rely on control devices such as the board of directors and incentive structures. Efficiency of the capital market as a disciplinary device varies across firms depending on the level of stock concentration. Some evidence documents that the existence of blockholders in publicly owned firms results in long-term views of strategy, encourages investments in R&D and increases productivity (Hill and Snell, 1989; Bethel and Liebeskind, 1993; Barclay and Holderness, 1991; Holderness and Sheehan, 1988; Mikkelson and Ruback, 1991; Shleifer and Vishny, 1986). Greater concentration of ownership may produce

behavior closer to profit maximization. Concentrated ownership means that the market for the firm's shares is less developed and the capital market discipline effect is weak (Leech and Leahy, 1991). Concentrated ownership is one of the components of a good corporate governance system which should combine large investors with legal protection of both their rights and those of small investors. While large investors rely on the legal system, they do not need as many rights as the small investors do to protect their rights. Corporations in successful market economies such as the USA, Germany and Japan are governed through somewhat different combinations of legal protection and concentrated ownership (Shleifer and Vishny, 1997). This paper attempts to reveal the factors that affect the use of control devices in corporations most of which have concentrated ownership. The agency problems, that may be encountered in concentrated firms that have outside CEOs and in the dispersed firms where less control is exerted on the CEOs, are expected to lead to differences in managerial behavior with respect to the use of leverage, the proportion of insiders on the board and board size. The firms included in this study differ with respect to the type of shareholders, degree of ownership concentration and location of control. The ownership typology developed in this paper is used to examine the relation between ownership structure and monitoring mechanisms. The ownermanaged firms, called owner controlled, serve as the control group where no agency problems exist. The dispersed firms, managerially controlled, are included to detect the differences in managerial behavior when compared with the externally controlled firms where the dominant shareholders closely monitor the CEOs. The managers in the latter group are assumed to be subject to the closest monitoring as documented by McEachern (1976). The focus of this paper is to search for interrelations between the control devices used to reduce agency costs. Recently, the interaction between the monitoring mechanisms has been the focus of extensive research (Jensen et al., 1992; Agrawal and Knoeber, 1996; Bathala et al., 1994; Chaganti and Damanpour, 1991). Evidence reveals that increasing institutional ownership can offset the need for debt and managerial ownership to reduce agency costs (Bathala et al., 1994) and this relation is moderated by the existence of family and inside institutional shareholdings (Chaganti and Damanpour, 1991). Furthermore, research that documents the interaction between market for takeovers and board of directors reveals that the market for takeovers and outside directors on the board are not perfect substitutes implying that the function of outsiders may vary depending on the nature of the external market for corporate control (Brickley and James, 1987). This study documents evidence on the interdependence between control mechanisms used in Turkish manufacturing firms to align interests of managers and shareholders. The market for takeovers does not exist as an external mechanism to control the corporate managers. The capital market has limited disciplinary effect on the managers because of the highly concentrated ownership structure of Turkish firms. The majority of the firms in the Istanbul Agency costs and corporate control devices 567

Journal of Economic Studies 27,6 568 Stock Exchange have only a small portion of publicly owned equity which leads to information asymmetries between investors in the market and the managers. The average shareholdings of the largest shareholder is 55 per cent and the average shareholdings of the three largest shareholders is 71 per cent for the firms included in the sample. The average proportion of equity that is publicly traded is 25 per cent for the 160 manufacturing firms in the Istanbul Stock Exchange. Another environmental constraint is the high rate of inflation in Turkey which increases the need for external financing for companies with insufficient equity. The high rates of interest on Treasury bills and bonds makes bond financing very costly for firms in need of capital. Most of the corporations rely heavily on short-term bank loans which increase the cost of capital and the risk of bankruptcy. Although existence of large creditors in corporate ownership is detected, bank governance is less effective in Turkey. Only 20 per cent of the firms in the sample have banks or financial institutions as investors. Firms remain family controlled and profitable corporations tend to prefer internal financing. Consequently, the institutional shareholders are not further classified as banks or financial institutions in this research. This paper focuses on managerial shareholdings, leverage, board size and composition, institutional shareholdings, ownership concentration and family shareholdings as control devices used to monitor the managers. Institutional shareholdings are further decomposed into foreign institutional shareholdings and domestic outside institutional shareholdings. The interdependence between these control devices are examined using an OLS simultaneous regression equations approach. The empirical evidence reveals that the negative relation detected between institutional shareholdings and the proportion of inside directors, ignoring the interdependence of the mechanisms, remains when all the devices are examined together. The evidence shows that the foreign institutional shareholdings and outside directors complement each other as managerial control devices. The significant negative relation between ownership concentration and proportion of insiders on the board implies that they are substitutes. Family shareholdings are positively related to the proportion of insiders revealing that family blockholders monitor best when accompanied with insiders on the board. Furthermore, the significant relation between the type of the dominant shareholder and leverage detected in the bivariate analysis disappears in the multivariate framework. The regression analysis reveals that leverage is negatively related only with domestic institutional shareholdings. Leverage seems to be associated with foreign institutional shareholdings revealing a positive beta coefficient which is significant at only 10 per cent confidence level. The negative relation between profitability and leverage implies that profitable firms tend to prefer internal financing to escape from lender monitoring. Managerial shareholdings do not have any significant relation with other control devices. The foreign institutional shareholdings and the ownership concentration reveal a negative and significant relation with the size of the board.

Consequently, as the board size decreases, foreign institutional shareholdings and ownership concentration increases. Since small boards serve the control function, board size seems to complement other control devices except leverage. The next section of the paper covers an overview of the related literature. This is followed by the empirical framework section that covers the hypotheses tested in this paper. The research design and methodology sections cover the sample specifications, the variables and measures, and the description of the techniques used in the analyses. The findings are presented and interpreted in the results section. The concluding remarks are documented in the last section of the paper. Agency costs and corporate control devices 569 Theoretical overview The separation of ownership and control gives rise to conflicts of interest between managers and owners. In contrast to the models that rely mainly on shareholders for the control of management (Jensen and Meckling, 1976; Alchian and Demsetz, 1972), Fama (1980) concludes that dependence on the risk bearers for disciplining managers is not sufficient since they can spread their wealth across many firms and are not interested in directly controlling the management of any specific firm. The efficiency of the capital market depends on the quality of information at its disposal (Fama, 1970) and the ability of the stockholders to remove managers that fail to maximize shareholder wealth by arranging proxy battles or takeover bids (Hill and Snell, 1989). If information asymmetries exist between managers and shareholders, managers may pursue strategies in their own interests. When ownership is concentrated, it is relatively easy for individual stockholders to demand information from management and overcome information asymmetries. Significant information asymmetries are likely to exist when stockholdings are diffused. Stockholders may lack resources to arrange a proxy battle or takeover even if they realize that management is not acting in their interests. Similarly, significant information asymmetries may exist between corporate managers and investors in the market when corporations have a small portion of equity publicly owned. The argument on the primary source of discipline of the managers is inconclusive and depends on the extent to which the external markets perform their function. The selection of the internal devices of control vary with respect to the degree of concentration and type of control within the firm which in turn affects the board composition, selection and compensation of the CEOs and the firm's debt policy. Each of the internal control devices has its costs. Excessive managerial ownership can result in entrenchment problems and high leverage may increase bankruptcy risk substantially. So, there should be an optimum combination of agency cost reducing mechanisms. The evidence on the effects of institutional shareholders on corporate policies is mixed. The literature recognizes that institutional investors have a significant role as external monitors in the stock market (Agrawal and Mandelker, 1990; Shleifer and Vishny, 1986; Coffee, 1991). Monitoring role of

Journal of Economic Studies 27,6 570 institutional owners is found to have higher significance for OTC firms compared to those listed in the NYSE (Bathala et al., 1994). The controversy on the monitoring role of institutional investors rests on two different streams of arguments. One of these arguments rests on the premise that since institutions manage diversified portfolios, they can't effectively monitor the firm and tend to focus on short-term performance (Lowenstein, 1991; Drucker, 1986; Mitroff 1987). The other stream of argument supports the view that increased monitoring by institutional investors is reflected in changes of corporate policies such as increases in research and development expenses (Hansen and Hill, 1991; Pound, 1992; Jarell and Lehn, 1985) and diversification (Hill and Snell, 1989; Dennis et al., 1997). The evidence reflecting the negative association of institutional ownership with the level of leverage and managerial shareholdings supports the argument that institutional investors serve as effective monitoring agents and help in mitigating agency costs (Bathala et al., 1994; Chaganti and Damanpour, 1991). Higher institutional ownership constrains the managerial shareholdings and substitutes for lender monitoring that stems from the use of leverage. Some researchers provide evidence supporting the myopic view of institutional owners when making investment decisions. Research based on computer manufacturers reveals that there is a significant negative relation between institutional ownership and R&D spending (Graves, 1988). Institutional ownership is found to have a significant positive relation to dividend payouts indicating that institutional shareholders don't serve to discipline managers (Bethel and Liebeskind, 1993; Lowenstein, 1991). The functions of institutional shareholders as monitors may depend on the ownership concentration. In dispersed firms, institutions holding minority shareholdings may tend to focus on short-term performance and act like portfolio managers. As the shareholdings of institutions increase, they tend to be more aggressive monitors of managers. Demsetz and Lehn (1985) reveal evidence relating to the significant impact of firm size, instability of profits and systematic regulation on ownership structure. These authors recognize the transaction and information costs associated with maintenance of corporate control and suggest that alterations in the structure of corporate ownership is partly related to these costs. Their evidence documents that there is no significant relation between ownership concentration and accounting rates of return which contrasts with the Berle- Means thesis (Berle and Means, 1932). Several attempts have been made to study the effect of ownership structure on the behavior and performance of firms. While some authors find no significant differences between diffused and concentrated firms with respect to performance (Holderness and Sheehan, 1988) an extensive amount of literature provides contrary results. Some studies reveal that the existence of blockholders in publicly owned firms results in efficiency increasing restructuring and downsizing (Hill and Snell, 1989; Bethel and Liebeskind, 1993). The observed decreases in diversification are strongly linked with

market disciplinary forces such as block purchases, takeovers, and management turnover (Denis et al., 1997). These results are consistent with the evidence provided by Barclay and Holderness (1991), Holderness and Sheehan (1988) and Mikkelson and Ruback (1991) who suggest that blockholder ownership increases firm value by precipitating changes in managerial policy. Consequently, the evidence confirms that blockholders exert a disciplinary effect on managers. Executive compensation is a device used to align conflicts of interest between managers and shareholders. Managers are not involved in profit maximizing behavior within a certain range of managerial ownership. Evidence reveals that there is a convergence of interests between managers and owners for the upper and lower levels of managerial ownership and a divergence of interests over the middle range (Carter and Stover, 1991; Morck et al., 1988). Over the middle segment of managerial ownership, managers have enough power and influence to guarantee their employment and are less concerned about the results of their behavior. The managers of public firms tend to expand and diversify firms without increasing their value (Grant et al., 1988; Lubatkin and Chatterjee, 1991; Rumelt, 1974; Wernerfelt and Montgomery, 1988). Extensive diversification into unrelated businesses is the best way for managers to maximize the size of the firm to satisfy desires for high renumeration, power, security and status. Compensation policy based on growth in sales encourages this trend. Conflicts between managers and shareholders are severe when corporations have substantial free cash flow. Leverage reduces the agency costs of free cash flow by reducing cash at the discretion of managers (Jensen, 1986). The control function of debt is more important in organizations that generate large cash flows but have low growth prospects. The free cash flow argument suggests that the entrenched managers may reduce leverage to avoid having to meet necessary performance levels associated with required debt payments. Evidence documents that leverage is positively related to the proportion of CEO stock ownership, presence of a blockholder and the proportion of outside directors (Berger et al., 1997). Leverage is found to be significantly and negatively related to CEO tenure and size of the board of directors. These results present evidence that as CEO entrenchment increases or monitoring decreases, managers adopt an underlevered suboptimal capital structure. Therefore, managerial entrenchment affects capital structure decisions. The effect of the structure and composition of the board of directors on firm performance is the subject of extensive research which does not reach a consensus. The degree of aligment between board and shareholder incentives varies with the composition of the board. Some evidence supports a view that outside directors act in the interests of the shareholders and serve to monitor the managers (Weisbach, 1988; Rosenstein and Wyatt, 1990; Hermalin and Weisbach, 1988; Mayers et al., 1997; Fama and Jensen, 1983; Lee et al., 1992). Outside directors are appointed in the interests of the shareholders and are more likely to remove CEOs following poor performance. Moreover, when Agency costs and corporate control devices 571

Journal of Economic Studies 27,6 572 boards are outsider dominated, fired CEOs tend to be replaced with executives from outside the firm (Borokhovic et al., 1996). Some research provides contrary evidence on the monitoring role of outsiders. The view that inside ownership aligns managerial incentives with the interests of all shareholders rests on the finding that dual class announcement returns are associated with insider shareholdings but not with the increase in outside directors (Bacon et al., 1997). The evidence that outside directors have a negative impact on performance is common in some researches and explained within different contexts (Agrawal and Knoeber, 1996; Mace, 1986; Lorsch and McIver, 1989). When boards are extended by political reasons, new outside directors tend to reduce firm performance. Furthermore, outside directors may feel uncomfortable in challenging the CEO on decisions outside their area of expertise. When CEOs dominate the director nomination process and their interests are not aligned with shareholders, they may nominate outside directors who tend to support their decisions. The monitoring role of outside directors is investigated together with other control mechanisms such as the market for takeovers and ownership structure (Brickley and James, 1987; Mayers et al., 1997). Although market for takeovers and outside directors are alternative mechanisms for monitoring the managers, the finding by Brickley and James (1987), who make a comparative study among states that prohibit bank acquisitions and states that allow bank acquisitions, reveals that more outside directors are present on the boards of banks in acquisition states compared to those in nonacquisition states. This result suggests that these mechanisms are not perfect substitutes. When the market functions freely, outside directors may assume the role of strategy makers rather than monitors. Similarly, ownership structure and function of outside directors seem to be related. In firms without publicly traded equity, the importance of outside directors in the control process increases. When a stock company's manager has controlling shares, the firm has significantly fewer outside directors (Mayers et al., 1997). This evidence may reflect the desire to avoid outside directors by the entrenched managers or a reduced ownermanager conflict. The finding that more diversified firms have a larger fraction of outside directors may point to the need for greater outside expertise. Consequently, ownership structure and board composition seem to be complements in monitoring the CEOs. The size of the board is another device used to align interests of managers and board members. Small boards serve to control managers whereas a larger board may not function effectively as a controlling body and leave management free. However, a larger board may be more valuable for the breadth of its services (Chaganti et al., 1985). Since most of the control mechanisms exist together, several empirical studies focus on the interdependence between these devices (Agrawal and Knoeber, 1996; Brickley and James, 1987; Chaganti and Damanpour, 1991; Mayers et al., 1997). Findings reveal that shareholdings by blockholders and

institutions are positively related to the market for corporate control revealing a complementarity between large outside shareholders and the market for corporate control. None of the other mechanisms affects insider shareholdings or outside direction but an increase in each of these leads to greater use of debt. This leads to the conclusion that the discipline implied by lender monitoring is most effective when coupled with internal monitoring either by inside shareholders or outside members of the board (Agrawal and Knoeber, 1996). The market for takeovers, outside directors and ownership concentration are not perfect substitutes. As institutional ownership increases, leverage and managerial shareholdings decrease, implying that they are substitute devices. However, the coalitions between executives, family owners and insider institutions can influence or alter the degree of outside institutions' influence. Agency costs and corporate control devices 573 Empirical framework In this section of the paper, an empirical framework is presented to assess the factors that affect the use of control devices and the interdependence between these mechanisms. The institutional shareholdings, family shareholdings, managerial shareholdings, board size, the proportion of insiders on the board of directors, ownership concentration and leverage are the selected devices used to control agency costs within a firm. The capital market is considered as an external monitoring device that constrains managerial behavior and is employed in the analyses. The hypotheses that are tested in this paper are grouped with respect to the type of control mechanism. The analyses rest on a set of multiple regression equations where each control device is used as a dependent variable in one of the equations and as an independent variable in the rest of the equations. The factors that may affect each control mechanism are included as regressors in the regression analysis where the specific monitoring device serves as a dependent variable. The simultaneous regression analysis is explained in detail in the methodology section. Some hypotheses relying on the ownership typologies are tested using the ANOVA technique. The symbols used for the variables in the regression equations are clearly defined in Table VIII (on p. 594). Debt The high rate of inflation prevailing in the country, coupled with the growth rate of the firms, raises the working capital requirements and increases the cost of debt. The average debt ratio for the 500 largest firms in the Istanbul Chamber of Industry (ICI) is 61 per cent while it is 55 per cent for the firms in the sample. Most of the corporations in the sample rely on the use of short-term bank loans to finance their working capital needs. The average proportion of short-term debt in total debt is 75 per cent for the 500 largest firms in the Istanbul Chamber of Industry. There are only a few corporate bond issues in the Istanbul Stock Exchange (ISE). The corporations face the burden of high interest rates and environmental uncertainty in relation to economic and political changes. Therefore, they are reluctant to issue long-term bonds at high rates of interest.

Journal of Economic Studies 27,6 574 Other financial instruments such as convertibles, preferred stocks and derivative securities do not exist. Consequently, Turkish corporations mainly rely on retained earnings and short-term bank loans for financing. Since leverage is risky due to the high costs and short maturity of the available funds, increased monitoring by lenders will be accompanied by leverage. Consequently, as ownership gets more concentrated and outside directors on the board increase, there will be more reliance on leverage. Similarly, owner-managers will be willing to take risk since they will also reap the high returns associated with it. So, as managerial shareholdings increase, the conflicts of interests between managers and stockholders will disappear and leverage will be used more extensively. There are four hypotheses related to the use of leverage in corporations. H1: Debt is positively related to managerial shareholdings, ownership concentration and outside directors on the board. H2: Debt is expected to be lower for managerially controlled firms relative to institutionally controlled and family controlled firms. Since borrowing under high interest rates increases bankruptcy costs significantly, entrenched managers will tend to adopt a conservative approach to avoid unfavorable effects of high risk or low performance. H3: CEO tenure is lowest in externally controlled firms relative to owner controlled and manager controlled firms. CEO tenure is a measure of manager entrenchment and the managers of externally controlled firms are in the most sensitive position (McEachern, 1976). In the externally controlled firms, corporate managers are closely monitored by the major shareholders who are outside the company. Owner controlled firms are lead by the major shareholder(s) and these managers are the least constrained and don't face agency problems. The shareholders of the dispersed firms do not control the managers closely since no shareholder is in a position to exert pressure alone. So, these managers are restrained by external devices such as the capital market and the market for managerial labor. H4: Leverage increases as firm size increases since the expected bankruptcy costs of debt are lower for large firms. The regression equation for debt is given below: (1) DEBT = a 0 + a 1 FINST + a 2 OCONCEN + a 3 PIBOARD + a 4 FAMSHARE + a 5 SBOARD + a 6 LSALES + a 7 ISE + a 8 MSHARE + a 9 DINST + a 10 ROA + e For the debt equation the regressors include foreign institutional shareholdings (FINST), ownership concentration (OCONCEN), proportion of insiders on the board (PIBOARD), the percentage of family shareholdings (FAMSHARE), board size (SBOARD), log sales (LSALES), existence in stock exchange (ISE), managerial shareholdings (MSHARE), domestic institutional shareholdings (DINST) and return on assets (ROA).

Board composition and size The board size and outside directors on the board are the mechanisms used to control the agency costs. In this research, the number and proportion of inside directors are used to measure the board composition. In environments where takeovers are restricted or costly, there will be more outsiders as board size increases (Brickley and James, 1987). Since the market for takeovers does not exist in Turkey, we expect more use of the other control devices. However, the extent to which outside directors are used depends on the type of shareholder and owner concentration. Although there are conflicting views on the function of outside directors as monitors, we expect that performance is positively related to outside directors implying that interests of outsiders are aligned with shareholders. H5: There is a negative relation between institutional shareholdings and proportion of insiders on the board of directors. Ownership concentration affects the role of outsiders on the board of directors. In firms with no publicly traded equity, the importance of outside directors as monitors increases (Mayers et al., 1997). Since major shareholders play an active role in selecting the members of the board (Holderness and Sheehan, 1988), family blockholders are expected to have a higher proportion of insiders on the board. The inside directors are less likely to challenge the CEO and managerial decisions. H6: Ownership concentration and proportion of inside directors are negatively related. The blockholders have a disciplinary effect on managers which is complemented by more outsiders on the board. H7: The proportion of insiders is positively related with family shareholdings. The regression equation for the proportion of insiders is as follows: (2) PIBOARD = b 0 + b 1 DEBT + b 2 FINST + b 3 OCONCEN + b 4 FAMSHARE + b 5 LSALES + b 6 DINST + b 7 ACEO + b 8 ROA + b 9 RD + b 10 AFIRM + v The existence of insiders on the board may be linked with the success of the CEO as measured by ROA. The size and age of the firm may affect the board composition as well. Newly established firms may need closer monitoring for the strategic decisions and R&D investments that are needed. Listing in the ISE may lead the firms to employ more outsiders on the board for a more representative board. So age of the CEO (ACEO), profitability (ROA), proportion of R&D to sales (RD), and age of the firm (AFIRM) are the variables which are included as regressors in addition to the control devices displayed in the equation for proportion of insiders on the board (PIBOARD). Sales variable is included as a control for size. The regression equation for board size is as follows: (3) SBOARD = c 0 + c 1 DEBT + c 2 FINST + c 3 OCONCEN + c 4 PIBOARD + c 5 FAMSHARE + c 6 LSALES + c 7 ISE + c 8 ROA + k Agency costs and corporate control devices 575

Journal of Economic Studies 27,6 576 Size of the board is expected to be affected by firm size, listing in the ISE and profitability in addition to some of the control devices. We expect board size to be positively related to firm size, profitability and being listed in the ISE. Larger boards tend to be more representative and the breadth of services provided tends to increase profitability. Foreign institutional shareholders The institutional shareholders are grouped as foreign institutional shareholders and domestic outside institutional shareholders. The domestic outside institutional shareholders are Turkish firms that have no relation with the shareholders of the firms that they have invested in. The foreign institutional shareholders are multinationals that have invested in Turkish firms or have formed joint ventures with domestic firms. Institutional shareholders serve as the control mechanisms in reducing agency costs. The ownership concentration and type of institutional shareholders may affect the function of the board, debt policy and managerial ownership as devices of control. H8: Debt is negatively related to foreign institutional shareholdings. The monitoring role of institutional investors tends to substitute the need for leverage to eliminate agency conflicts (Bathala et al., 1994). H9: Foreign institutional shareholdings are negatively related to managerial ownership. H10: There is a positive relation between research and development investments and foreign institutional blockholders. Most of the coalitions of Turkish and foreign firms are in the form of joint ventures where the firms benefit from economies of scale in operations and share risks related to research and development operations. The foreign firms assume the role of strategy makers as well as monitors in these coalitions and can't be regarded as portfolio managers like institutional investors in the capital market. Therefore, it is more likely that as foreign institutional shareholdings increase, firms will adopt a long-term perspective in managerial decisions rather than focusing on short-term performance. The regression equation that aims to estimate the factors that affect the percentage of foreign institutional shareholdings is as follows: (4) FINST = d 0 + d 1 DEBT + d 2 OCONCEN + d 3 PIBOARD + d 4 FAMSHARE + d 5 SBOARD + d 6 LSALES + d 7 SINST + d 8 ISE + d 9 MSHARE + d 10 DINST + d 11 ROA + d 12 RD + t In addition to the five control devices (DEBT, OCONCEN, PIBOARD, FAMSHARE, SBOARD) and the control variable sales (LSALES), this equation aims to detect the effect of subsidiary shareholdings (SINST) domestic outside institutional shareholdings (DINST), profitability (ROA) and R&D expenses (RD) on the foreign institutional shareholdings.

Ownership concentration The ownership concentration in a firm may alter the degree to which some of the devices of control are used. The regression equation for ownership concentration is as follows: (5) OCONCEN = f 0 + f 1 DEBT + f 2 FINST + f 3 PIBOARD + f 4 FAMSHARE + f 5 SBOARD + f 6 LSALES + f 7 ISE + f 8 DINST + f 9 ROA + f 10 AFIRM + g We expect a negative relation between ownership concentration and firm size (LSALES), listing in the ISE and age of the firm (AFIRM). A positive relation between ownership concentration and domestic institutional shareholdings (DINST) is expected. Furthermore, the existence of blockholders may result in higher profitability (ROA) because of a greater emphasis on efficiency increasing investments as documented by Hill and Snell (1989), and Bethel and Liebeskind (1993). So, (DINST), (ROA) and (AFIRM) are included as regressors in this equation in addition to the monitoring devices and the control variable size (LSALES). Agency costs and corporate control devices 577 Family shareholdings Most of the firms in the sample are family controlled corporations. For this reason, the interdependence of family shareholdings with other devices of control is examined using the following regression equation: (6) FAMSHARE = h 0 + h 1 DEBT + h 2 FINST + h 3 OCONCEN + h 4 PIBOARD + h 5 SBOARD + h 6 LSALES + h 7 ISE + h 8 MSHARE + h 9 DINST + h 10 ROA + h 11 RD + j Family blockholders are expected to be positively related to R&D expenses and profitability (Holderness and Sheehan, 1988). Firms with family blockholders tend to be smaller than firms that have corporate majority shareholders. Furthermore, we expect that family shareholdings are positively related to managerial shareholdings and proportion of insiders on the board. In the sample, 22 of the 73 family controlled firms are lead by their owners, whereas 51 are controlled by family members who are on the board of directors. Managerial shareholdings The managerial shareholdings equation is presented below: (7) MSHARE = s 0 + s 1 DEBT + s 2 FINST + s 3 IBOARD + s 4 FAMSHARE + s 5 SBOARD + s 6 DINST + s 7 ROA + s 8 LSALES + s 9 R&D + s 10 ACEO + s 11 CTENURE + n If managerial shareholdings serve as incentives to align conflicts of interest between managers and shareholders, then firms will reward their managers who have shown long years of successful performance. So managerial shareholdings are expected to be positively related to CEO tenure and firm profitability (ROA). Higher managerial shareholdings are associated with increased levels of R&D which suggest high agency costs (Bathala et al., 1994).

Journal of Economic Studies 27,6 578 Therefore a higher level of managerial shareholdings is expected to lead to higher leverage which serves as a medium to constrain the high agency costs. H11: Managerial shareholdings are expected to be positively related to domestic outside institutional shareholdings and family shareholdings. H12: Managerial shareholdings are positively related to the number of insiders on the board. H13: Managerial shareholdings are positively related to age and tenure of the CEO. Domestic outside institutional shareholdings The domestic outside institutional shareholdings equation is presented below: (8) DINST = p 0 + p 1 DEBT + p 2 FINST + p 3 OCONCEN + p 4 PIBOARD + p 5 FAMSHARE + p 6 SBOARD + p 7 MSHARE + p 8 LSALES + p 9 ISE + p 10 ROA + r H14: A positive relation is expected between family shareholdings and domestic outside institutional ownership. H15: Domestic outside institutional ownership is negatively related to debt and the proportion of insiders on the board. Since institutional investors tend to focus on short-term performance of the firms, we expect a positive relation between the domestic outside institutional shareholdings and ROA in the DINST equation. Research design The sample The sample of this study is confined to 134 manufacturing corporations. These corporations are selected from the 500 largest manufacturing firms of the Istanbul Chamber of Industry (ICI) and from firms that are listed in the Istanbul Stock Exchange (ISE). Although the research rests on a survey conducted through a questionnaire directed to the CEO of the firms, secondary data is needed for the multivariate analyses. Some selected ratios are provided through questions in the questionnaire and most of the data required for multivariate analyses rely on the published financial statements of the firms. While the firms listed in the ISE have published financial statements, the firms from the 500 largest corporations publish only part of their financial data; sales, net income, total net assets and total equity figures. Furthermore, the declaration of the names of the firms and their partial financial data are on a voluntary basis. Therefore, 20 firms out of the 500 largest don't declare their names and data at all, while some of the remaining corporations choose not to expose some of the selected figures such as the net income and equity figures. Consequently, the number of observations included in each of the analyses varies depending on the number of missing cases related to each variable that is included. This paper is based on a cross-sectional research that rests on year 1997 data.

The firms in the sample are selected so as to satisfy the three criteria stated below: (1) There should be sufficient number of firms from each industry that is selected to allow for comparative studies. (2) Different levels of ownership concentration should be represented. Since most of the firms have a highly concentrated ownership structure, whether in the ISE or not, some firms are from those listed in the Istanbul Stock Exchange to represent diffused ownership structure as well. (3) Firms should have different types of shareholders. Firms in the sample are selected from those that have foreign institutional ownership, family major shareholders, and domestic outside institutional shareholders to obtain adequate representation of different compositions of ownership structure. Consequently, the firms in the sample are selected from two different groups, namely the 500 largest firms from the ICI and ISE lists. The two groups are not mutually exclusive. Some of the firms that satisfy the above constraints exist in both groups and some exist only in the 500 largest group or only in the ISE. The breakdown of the sample with respect to each group representing the population is shown in Table I. The breakdown of the sample of 134 manufacturing firms is as follows: Number of firms in the 500 largest of the ICI 102 Number of firms in the ISE 77 Those that exist in both groups (45) Total 134 The t tests that are carried out reveal that the sample significantly represents both populations at 5 per cent confidence level. The summary statistics for the ownership concentration, the ownership distribution by type of shareholders, industry allocation of the sample and the financial variables are omitted here but can be provided by the author upon request. In some of the statistical analyses where the data such as firm sales, debt ratio and profitability (ROA) are used, the total number of responses in the sample drop to 98 due to the constraints on the availability of these figures. The corporations are from seven industries which are initially classified on the basis of three digit standard industrial classification (SIC) codes. Although Agency costs and corporate control devices 579 500 largest ISE Total number of manufacturing firms in each group for 480 160 which partial or full financial data can be obtained Total number of firms representing each group in the sample 102 77 Table I. Population and the sample

Journal of Economic Studies 27,6 580 a balance is sought in allocation of firms among industries, insufficient response rates in some sectors led to reclassification into two digit SIC codes. The number of firms in basic metal and paper products industries remains insufficient for some of the statistical tests and they were eliminated where necessary. The number of public firms is limited and does not allow for comparative tests. Therefore, state ownership is not included in the analysis. Previous studies reveal that firm size is a moderating variable in determining the relation between ownership structure and performance. In this study, total sales is used as a proxy of size. Firm sales has a highly skewed distribution. Thus, log sales is used as size variable in the multiple regression analyses. There are wide sales discrepancies among firms included in the 500 largest group. Half of the firms included in the sample (52 per cent) fall within the first quartile of the sales distribution of the 500 largest firms and 25 per cent in the second quartile. The remaining 23 per cent is grouped within the third quartile of the sales distribution. Consequently, log sales is used as a measure of size control. The preliminary tests reveal that ownership structure is not significantly related to firm sales (F = 1.8531, p = 0.16) and a weak relation is detected between ownership structure and log sales (F = 3.13; p = 0.048). Data collection This paper rests on a questionnaire directed to the top executives of the manufacturing firms. The questionnaire consists of seven parts that are designed independently. The first part includes information about the industry of the firm reflecting the rate of technological change, intensity of competition and fluctuation of sales. The second part includes questions related to the firm's performance with respect to liquidity, return on equity, rate of sales growth, breakeven sales, capital intensity and the market share using likert scales. The third part covers questions related to corporate and investment policies that rely on the use of likert scales. In addition, the executives are asked to rank their preferences of financing sources. Part four covers questions relating to the firm's hedging policies. Part five includes several measures of ownership that are explained in the following section. Part six covers questions related to the executive performance evaluation and compensation. In the last part, ratios regarding R&D investments, capital investments and diversification are provided by the firms. This paper relies on the data generated in the last three parts of the questionnaire. The questionnaire was delivered with prior notice, through appointments with top managers and collected by the interviewers personally. The questionnaire was directly given to the CEOs. The CEOs explained that they completed the questionnaire leaving the questions that ask for some ratios of the firm to the finance department. Initially, 200 firms were selected using the criteria explained in the research design section of this paper. Since only 134 firms returned the questionnaire, the response rate was 67.5 per cent.

In addition to the primary data provided by the questionnaire, the empirical framework of the paper rests on some figures that are obtained from the financial data published by the 500 largest firms of the ICI and the financial statements published by the Istanbul Stock Exchange for the listed companies. The variables derived from secondary data include sales, debt ratio and return on assets for the firms. However, only 98 firms out of 134 in the sample have the mentioned data, and the statistical tests including these variables are confined to 98 observations. Agency costs and corporate control devices 581 Variables from the survey and measurement The variables related to ownership structure such as ownership concentration, percentage shareholdings of the investors and ownership typology are derived from the questionnaire. Furthermore, CEO tenure, age of the firm, ratio of R&D to sales, proportion of insiders on the board (PIBOARD) and board size (SBOARD) are the primary data obtained from the survey. The abbreviations used for the variables included in the regression analyses are provided in Table VIII. Ownership structure Ownership structure is measured using three different groups of variables. Ownership concentration. In the first group the degree of ownership concentration is measured using percentage shareholdings of the largest shareholder and the three largest shareholders. These two variables reflect a highly concentrated ownership structure in Turkish manufacturing firms. The number of observations related to these variables are confined to 127 and 106 firms which have responded to these questions. The firms in the sample have highly concentrated ownership. The average shareholdings of the largest investor is 55 per cent and the average of the shareholdings of the three largest shareholders is 71 per cent. When these figures are compared with 456 Fortune 500 firms, where mean percentage holdings of the largest shareholders is 15.4 per cent and mean percentage holdings of three largest shareholders is 28.8 per cent (Shleifer and Vishny, 1986), the differences reflect the contrasting features of the two settings. Shareholdings of investors. The second group of variables related to ownership reflects the percentage shareholdings for each type of shareholder. The shareholders are grouped as family, insider institutions, domestic outside institutions, foreign institutions and managers. The managerial shareholdings have a highly skewed distribution which has not obtained normality after several transformations. The log transformation of this variable is used in the regression analyses. Ownership typology. The typology that relies on a dichotomy of ownership called ``owner-controlled firms and managerially controlled firms'', developed by Berle and Means (1932), is found to be insufficient in explaining the relation between ownership structure and performance. McEachern (1976)

Journal of Economic Studies 27,6 582 distinguished between concentrated firms managed by the owners and those that have ``hired managers''. He called the latter category of firms ``externally controlled'' firms. Thus McEachern's (1976) ownership typology consists of three types of firm ownership: (1) Firms in which the managers are the dominant shareholders (ownercontrolled firms). (2) Firms in which the dominant shareholder is not the manager (externally controlled firms). (3) Firms in which no dominant shareholder exists (managerially controlled firms). This typology is similar to the one developed by Mintzberg (1983) but uses three categories instead of the four groups determined by Mintzberg. A comparative breakdown of the sample with respect to these typologies is presented in Table II. This paper relies on McEachern's typology of ownership. The ownership variable includes the type of dominant shareholders of the externally controlled firms as well. Data is generated by defining the ownership categories of firms and asking the executive to select the one that best describes the firm's ownership structure. The number of responses is limited to 130 firms since four CEOs omitted this question. The ownership structure is reclassified by type of dominant shareholders in some of the analyses. The ownership typology rests on two dimensions used to separate ownership and control. These are the location of control and the degree of control. The location of control is decomposed into internal and external control depending on whether the control is internal or external to the manager. These are similar to the involved and detached categories defined by Mintzberg (1983). The degree of control is a continuous variable which measures the discretion which the controlling group has in pursuing its objectives and is related to voting power (Cubbin and Leech, 1983). The breakdown of the Typology I (McEachern, 1976) Number of firms (%) Typology II (Mintzberg, 1983) Number of firms (%) Table II. Sample ownership classification typologies Externally controlled by Concentrated detached firms Family 51 39.2 Family 51 Institutions 37 28.5 Institutions 37 Foreign institutionsminority 9 6.9 Total 88 67.7 Total 97 74.6 Dispersed-involved firms 9 6.9 Owner-controlled 22 16.9 Concentrated-involved firms 22 16.9 Managerially controlled 11 8.5 Dispersed-detached firms 11 8.5 Total 130 a 100.0 Total 130 a 100.0 Note: a Four cases are missing