Master Thesis
Leverage, Ownership, and Firm Performance:
A Study of Banks in the Euro Area
Jiayuan Shi
January 24, 2013
University of Groningen: S2086808 Uppsala University: 890205-P211
Email: jacintashi89@gmail.com
MSc International Financial Management Faculty of Economics and Business
MSc Business and Economics Faculty of Social Sciences
Supervisor: Prof. Dr. C.L.M. Niels Hermes
1
Abstract
This paper studies leverage, ownership and firm performance of listed banks in the Euro area. I mainly explain how leverage and ownership structure influence firm performance from the agency-theory perspective. Therefore I follow prior research and choose a firm performance measurement which indirectly reflects the degree of agency costs – profit efficiency. In this paper, higher profit efficiency represents lower level of agency costs and better firm performance. The panel regression analysis shows that leverage has a positive influence on firm performance and the relationship is non-monotonic. Furthermore, managerial insider ownership influences firm performance negatively at low levels and then positively. The influence is entrenched and becomes negative again when the proportion of shareholding is very high. Outside institutional holding can lower agency costs and influence firm performance positively but the effect is also non-monotonic and becomes negative at high levels.
JEL classification: G32
2
1. Introduction
Banks play an important role in the financial system in modern economy. They act as financial intermediaries and allocate funds among investors. Besides providing credit to nonfinancial firms, the banking industry also has other important functions: it transmits the effect of monetary policy and provides stability to the whole economy (Berger and Bonaccorsi di Patti, 2006). However, corporate governance problem is usually severe in banks. The bankruptcies of some international banks (e.g. Lehman Brothers) reveal large losses in this industry and the corporate governance issue has been emphasized more and more. Some recent research also points out the seriousness of the flaws in banks’ corporate governance (e.g. Diamond and Rajan, 2009).
Corporate governance is a system under which companies are directed and controlled; it includes “regulatory and market mechanisms, the roles and relationships between a company’s management, the board, shareholders and other stakeholders, and the goals for which the corporation is governed” (Tricker and Adrian, 2009). It is related with the mitigation of interest conflicts between stakeholders1, among which the principal-agent problem caused by the separation of ownership and management is one major thread of discussions2 in corporate governance literature (Berle and Means, 1932; Jensen and Meckling, 1976; Shleifer and Vishny, 1986). The principal-agent issue (also referred to as the agency problem) is thoroughly investigated in the seminal paper of Jensen and Meckling (1976), where they develop the agency cost theory. Following Jensen and Meckling’s (1976) work, scholars have done many relevant researches. Some of them suggest that the choice of capital structure can mitigate agency conflicts (e.g. Grossman and Hart, 1982; Williams, 1987; Jensen, 1986; Harris and Raviv, 1990a; Stulz 1990), and some suggest that certain ownership structure can also mitigate agency conflicts (e.g. Shleifer and Vishny, 1986; Stulz, 1988; Pound, 1988).
Corporate governance problems and agency cost issues put forward particularly important research and policy questions in regard to the banking industry as banks hold opaque
1
Stakeholders include external stakeholders (e.g. shareholders, debtholders, suppliers, customers and communities affected by a firm's activities) and internal stakeholders (e.g. board of directors, executives, employees). (Haidar and Ibrahim, 2009)
2
3
information (e.g. personal information of loan customers, other private information of credit counterparties) by their nature (Berger and Bonaccorsi di Patti, 2006). However, the existing studies in the banking literature have one major limitation. There is relatively little research which studies agency costs and how capital structure and ownership mitigate the problem in the banking literature (Berger and Bonaccorsi di Patti, 2006) as compared with the studies done in other industries. Many empirical studies investigate capital structure, ownership and agency costs by using different firm performance measures, but the samples are non-financial institutions (e.g. Jensen, 1986; Wruck, 1994; Ang et al., 2000; Zhou, 2001; Cai and Zhang, 2006; Margaritis and Psillaki, 2010). A lot of studies on financial institutions, especially banks, focus mainly on the differences in firm performance as measured by either cost efficiency or profit efficiency, but few of them investigate possible factors that cause the efficiency differences of banks. Most of these studies are centered on U.S. banks (e.g. Berger and Mester, 2003; Rogers, 1998; Berger and DeYoung, 2001; Färe et al., 2004). There are studies which investigate efficiency difference of banks in other countries (Miller and Noulas, 1996; Maudos et al., 2002; Bonin et al., 2005a), but they are also only based on the comparison of efficiencies.
A study which fills the gap of the above limitation is done by Berger and Bonaccorsi di Patti (2006). They study capital structure and firm performance of U.S. banks. Firm performance is measured by profit efficiency to reflect the degree of agency cost3. Three problems in prior research are pointed out, namely the choice of firm performance measurement, the possible reverse causality problem, and ignoring the ownership structure influence. They prove that an increase in leverage yields a predicted increase in firm performance as measured by profit efficiency. They also include ownership as control variable and find that the relationship between managerial insiders and firm performance is non-monotonic while the relationship between institutional shareholders and firm performance is strongly positive. However, there are two limitations in their work: the whole study is based only on the U.S. and the study only covers the period from 1990 to 1995. Moreover, they predict that the relationship between leverage and firm performance may be non-monotonic – it will reverse when debt increases to a very high level, but their results are not consistent with this prediction.
3
4
In this paper, I also conduct a study of leverage, ownership and firm performance. Since many prior studies are about U.S. banks, I base my study on European banks4. Because I mainly explain how leverage and ownership structure influence firm performance from the agency-theory perspective, I follow Berger and Bonaccorsi di Patti’s (2006) study by choosing profit efficiency, which is regarded as a good reflection of the level of agency costs, as the indicator of firm performance. To be more specific, in this paper, I will study how leverage and ownership structure influence agency costs of banks in the Euro area, and I will test whether the influences are non-monotonic5 or not. By conducting a research covering a longer period and using a sample of banks from different countries, the study aims to add to the work done by Berger and Bonaccorsi di Patti (2006) and also make a contribution to the banking literature. The study aims to answer the following research questions:
Does higher leverage lead to better firm performance, and will the relationship reverse when leverage is very high?
Does ownership structure have a significant influence on firm performance?
Something to be mentioned in advance is that I choose banks in the Euro area - the largest economic and monetary union (EMU) in the world. Although there may be some country-level differences, the focus in my study is the differences across banks as caused by different corporate governances, and the effect of regulation and country-level difference are not within the scope of the research in this paper. Moreover, this paper assumes that leverage is exogenous as with prior studies (e.g. Mehran, 1995; Mc Connel and Servaes, 1995) and does not consider the possible reverse causality between leverage and firm performance. I also do not distinguish long-term and short-term debt but consider debt as a whole (leverage) influences firm performance. Similar to Berger and Bonaccorsi di Patti (2006), I assume the choice of capital ratio is not affected by ownership structure, so ownership structure mainly influences firm performance directly. And because it is difficult to empirically distinguish between the agency costs of debt and the agency costs of equity, firm performance is viewed as an indirect
4 For Continental Europe, firms have higher ownership concentration and more dominant control, so the nature of ownership
and control is different from that in the U.S. (Faccio and Lang, 2002). La Porta et. al (1998) also argue that due to different institutional settings, the findings for firms in the U.S. may not be representative for firms in other countries.
5
5
reflection of total degree of agency costs as is the same in Berger and Bonaccorsi di Patti’s (2006) study.
Thus, I set out my study in the Euro area and collect data from the period 2003 to 2011. ‘BankScope’ serves as the main database. The panel regression with time-period fixed effects shows that leverage influences firm performance both significantly and positively, but the relationship reverses when leverage is very high6. Besides, the results also show that managerial insider ownership influences firm performance negatively at low levels and then positively. The influence is entrenched and becomes negative again when the proportion of shareholding is very high. Outside institutional holding lowers agency costs and influences firm performance positively but the effect is also non-monotonic and becomes negative at high levels. All of the results are statistically significant and hold when controlling for the possible influence of different years. The results for the non-monotonic relationship between leverage and firm performance and the results for the effect of managerial ownership also hold either with the whole sample or the subsample.
My study has several contributions: 1) it is the first paper which studies capital structure, ownership and agency costs particularly on banks in the Euro area; 2) the time span is from 2003 to 2011 which is longer with up-to-date data and covers the recent financial crisis; 3) the study also considers how ownership structure influences firm performance which is a problem ignored in most prior researches as pointed out by Berger and Bonaccorsi di Patti (2006); 4) results in this paper confirm the predictions - the non-monotonic relationship between leverage and firm performance, untested in Berger and Bonaccorsi di Patti (2006)’s work; 5) the relationship between institutional holding and firm performance is also found to be non-monotonic7.
The structure of the paper is organized in the following order: a list of relevant literature is provided first and hypotheses are developed as well, then the measure of firm performance and the technique used are discussed next. In the methodology part, the main model will be built after which the data selection criteria and a relevant description are presented.
6
In my study, I mainly test whether the relationship is non-monotonic or not and does not specify the turning points.
7
6
The results are provided thereafter followed by a discussion, and finally, a conclusion is made to end this paper.
2. Literature Review and Hypotheses
This section gives a review of the relevant literature on agency costs, capital structure and ownership structure. Meanwhile, I develop all the hypotheses to be tested based on existing theories, empirical evidence and my own arguments. I will first introduce the agency cost theory, and then discuss why the choice of capital structure can be a tool to mitigate agency costs. Finally, I will consider how ownership structure may also influence agency costs.
2.1 Agency costs
Agency costs are essential problems in corporate governance. Berle and Means (1932) first indicate the possible conflict of interests between the management and various shareholders within a corporation when the management has no equity in the firm. The problem is also emphasized by many other scholars (e.g. Baumol, 1959; Williamson, 1964). Jensen and Meckling (1976) later formalize the problem in their agency cost theory. The theory states that if both parties are utility maximizers, the agent will not keep acting in the best interest of the principal due to the conflict of interests. Because of the separation of ownership and control of firms, the agency problem is common in modern corporations and agency costs are suffered almost by all firms. Jensen and Meckling have identified two types of agency costs in a firm.
7
later.
Another type of agency cost within a firm is the agency costs of debt. Debt contracts will stimulate the suboptimal investment problem in a firm. When an investment produces large returns, equity holders obtain most of the gain after paying the face value of the debt. However, when an investment fails, debt holders suffer the loss. As a result, firms financed with external debt may engage in risky-project investments. To protect debt holders, there are provisions to limit a firm’s participation in risky projects. The opportunity wealth loss on a firm’s investment decisions caused by the covenants of debt, the monitoring and bonding costs by debt holders, and the bankruptcy and reorganization costs8 constitute the agency costs of debt9 in a firm. (Jensen and Meckling, 1976)
Both agency costs exist in firms while the reduction of either one has the possibility to improve firm performance. Following Jensen and Meckling (1976), scholars have done much relevant research and suggest that the choice of capital structure and the existence of some ownership types can mitigate agency conflicts10.
2.2 Capital structure, agency costs and firm performance
There are many theories on capital structure. Before the agency cost theory, Modigliani and Miller (1958) state that, under perfect market conditions (e.g. no tax and transaction costs, no asymmetric information), the debt-equity ratio does not influence a firm’s market value and owners of a firm are indifferent about the financial structure. Their theory is based on many assumptions, and one of them is that firms’ production plan is independent from the capital structure. Following Modigliani and Miller (1958), scholars try to relax the assumptions and explain the function of debt. After introducing tax, Modigliani and Miller (1963) revise their previous conclusions and suggest that firm value is positively related to debt as interest payments caused by debt lower a firm’s tax liability11. Further relaxing the assumptions (e.g.
8
In case of bankruptcy and reorganization, there is a transfer of ownership, a loss value of debt, and an increase of rate of return required by investors etc. Contracts representing claims on the firm need to be written. The reorganization also includes the adjustments of claims among various parties and the business. (Jensen and Meckling, 1976)
9
Some scholars also pointed out other agency costs of debt: e.g. Meyer (1977) points out that when a firm is likely to go bankruptcy, equity holders have no incentive to contribute capital even for value-increasing projects because the return may be mainly obtained by debt holders in the liquidation. Therefore, at higher leverage level, firms may face more objections on value-increasing projects.
10
There are also other ways to mitigate agency costs but is not within the scope of this study. e.g. Green (1984) points out that the use of convertible bonds and warrants can also mitigate agency costs.
11
8
with taxes and bankruptcy costs), Myers (1977) implies in his study that tax advantage of debt can be offset by personal taxes, and other costs like bankruptcy costs also make borrowing less desirable.
The agency cost theory provides another theoretical perspective to explain the role of debt and the choice of capital structure. Jensen and Meckling (1976) contend that firms use debt commonly before the tax subsidies on interest payments, so there must be other reasons to explain firms’ capital structure. Relaxing Modigliani and Miller’s (1958) assumption that firms’ production plan is independent from the capital structure, Jensen and Meckling (1976) consider a situation where the entrepreneur can raise funds either by issuing equity or issuing debt for an investment project. Because the entrepreneur owns the firm, if the investment project is financed with external equity, the entrepreneur’s share on the firm will be less than 100 percent. As a result, the entrepreneur will not manage it carefully and will take many perquisites. However, if the project is financed with debt, the entrepreneur’s incentive is improved because he can get the benefit of all increase in profits except in bankruptcy. Thus, debt provides a way for firms’ expansion without sacrificing incentives. There is a trade-off for the entrepreneur: financing with equity will share risks, while financing with debt will give rise to a high value of the project since the entrepreneur’s incentive is improved.
Following Jensen and Meckling (1976), scholars have explained how capital structure has the effect of mitigating agency costs from different point of views (e.g. Grossman and Hart 1982, Williams 1987, Jensen 1986; Harris and Raviv 1990a, Stulz 1990).
One of the perspectives is that leverage reduces agency costs as the use of debt creates a bonding effect on managers through the threat of liquidation. Grossman and Hart (1982) develop the theory and explain the use of debt as a financial instrument. They argue that the possibility of bankruptcy can be an important factor to solve managers’ incentive problem while bankruptcy as a discipline for managers depend on firm’s financial structure. Because managers will lose the benefits they can receive from the firm in the event of bankruptcy, they may prefer to seek high profits instead of behaving differently to risk “sacrificing their perquisites”. If there is no debt, managers face no threat of bankruptcy. The market would recognize that managers
9
are in an unconstrained position with less incentive to maximize profit and thereby put a low valuation on the firm. By issuing debt, managers pre-commit themselves that in order not to lose their positions they can only be more productive. In this case, shareholders know that not maximizing profit is personally costly for managers. As recognizing that profit will be higher, the market will put a high valuation on the firm. (Grossman and Hart, 1982) The incentive effect in Grossman and Hart’s (1982) model comes from the desire to avoid bankruptcy. One issue that needs to be pointed out is that their theory is developed on the hypothesis that management does not have a shareholding inside the firm, so a change from equity finance to debt finance does not change management’s managerial benefits. The influence of insider ownership will be discussed later in this part.
Another way to explain how capital structure mitigates agency problem is from the free cash flow perspective. Jensen (1986) first explains the role of debt in his “free cash flow theory”. In his line of reasoning, Jensen contends that one major problem is how to motivate managers to use cash in a proper way and not disgorge it (e.g. using cash for investment projects for which the return rate is below cost of capital or spending it on other inefficient projects). The issue of debt bonds managers to pay out future cash flows (e.g. interest payments) and in this way, the available cash flow for spending at the discretion of mangers is reduced. Thus debt can reduce the agency costs of free cash flow and the pressure of making interest and principal payments of debt also effectively motivates firms to be more efficient (Jensen, 1986).
10
Stulz (1990) explains how capital structure mitigates agency problem from the perspective of managers’ investment decisions. He investigates how managers’ ability to obtain their own objectives is restricted by financial policies. Stulz (1990) supposes that shareholders are hard to force managers to pay out cash flow when it accrues because to act collectively is costly for them. Managers prefer to use all the available funds for investments, so they can derive more perquisites. Stulz (1990) denotes this as an overinvestment problem. The model developed by him shows the benefit of debt in limiting managers’ overinvestment activities because it requires managers to pay out funds when cash flows accrue. This is similar with Jensen’s (1986) arguments. Therefore, in the view of Stulz (1990), debt lowers agency costs by limiting the possible overinvestment problem in a firm and has a positive influence on firm performance.
Next, I will review empirical evidence that have tested the positive effect of debt (leverage) on firm performance which is measured by different methods.
11
individuals to make value maximizing decisions – the company then has a better firm performance on the short and long term. Wruck’s (1994) study in addition shows that the findings by previous scholars also hold for a particular firm.
With the attempts to measure the magnitude of agency costs, some scholars use accounting figures to reflect firm performance. One of the studies is done by Ang et al. (2000). The accounting figures used by them are the ratio of operating expenses to annual sales and the ratio of annual sales to total assets. Ang et al. (2000) use the first ratio to measure how effectively managers in a firm control operating costs which include excessive consumption of perquisites and other direct agency costs, so high ratios represents high agency costs and poor firm performance. They use the second ratio to measure how effectively managers deploy the firm’s assets, and low ratio represents poor utilization of assets and high agency costs. Ang et al. (2000) find that as leverage increases, external monitoring by debt holders will lower agency costs and improve firm performance. The results are based on a study of a large sample of small U.S. firms. Following the study of Ang et al. (2000), Flemming et al. (2004) study agency costs and ownership structure of Australia firms. With the attempt to measure the magnitude of agency costs as well, they use accounting ratios similar to Ang et al. (2000) - operating expenses to sales and asset utilization ratio. One of their findings is that higher leverage is related with more efficient asset usage, indicating the existence of debt creates a bonding effect on managers and agency costs therefore become lower. Another method which tries to reflect the degree of agency costs is using efficiency figures12 (e.g. profit efficiency, productive efficiency). Berger and Mester (1997) conduct a study on bank efficiency and the variables that affect agency costs. They use profit efficiency as a measurement of firm performance - higher profit efficiency represents lower level of agency costs and better firm performance. One of their findings is that banks with higher leverage have higher profit efficiency. In Margaritis and Psillaki’s (2010) study, they use productive efficiency - higher productive efficiency figures represent lower level of agency costs and better firm performance. They use a sample of French manufacturing firms and show a positive relationship between leverage and productive efficiency. They also explain that the principal-agent conflict is one of the reasons that firms engage in inefficient activities. To sum up, empirical evidence has suggested the existence of a positive relationship between
12
12
leverage and firm performance. Scholars have done researches among both non-financial and financial institutions and for either a large sample or a single firm. They try to indirectly measure the degree of agency costs, and firm performance is measured by different methods.
In the study of Berger and Bonaccorsi di Patti (2006), they also analyze the positive effect of debt on firm value from the agency-theory perspective and state that one testable hypothesis is “increasing the leverage ratio should result in lower agency costs and improved firm performance, all else held equal”. Similar with Berger and Bonaccorsi di Patti’s study, I will first test the “agency costs hypothesis”:
H1: Higher leverage is expected to lower agency costs, enhance efficiency and improve a firm’s performance.
13
Therefore, debt payment influences shareholders’ wealth positively by limiting the available funds for managers and thus mitigating the overinvestment problem, but it can also influence shareholders’ wealth negatively through restraining potential advantageous investments. In Stulz’s review, the costs and benefits of debt indicate the existence of an optimal capital structure which maximizes firm value. To summarize, the above statements imply that debt does not always have a positive effect on firm value, and there exists an optimal level of capital structure. Once leverage is above the optimal level, the cost of debt will outweigh the benefit of debt and leverage’s influence becomes negative.
In Berger and Bonaccorsi di Patti’s (2006) study, they allow for the relationship between firm performance and leverage to be non-monotonic, but the result is not consistent with the prediction that a very high leverage ratio may have a reverse effect on firm performance. To test the non-monotonic relationship between leverage and firm performance, I include a supplement hypothesis for H1:
H1a: The relationship between leverage and firm performance is expected to be non-monotonic.
2.3 Ownership structure, agency costs and firm performance
There is also much literature on how ownership structure influences agency costs and
firm performance (e.g. Shleifer and Vishny, 1986; Stulz, 1988; Pound, 1988). One problem as pointed out by Berger and Bonaccorsi di Patti (2006) is that most prior studies do not take ownership structure into account when studying capital structure and agency costs (e.g. Titman and Wessels 1988); while some evaluate the effects of ownership structure on firm performance, but fail to include the influence of capital structure (e.g. Mester 1993, Pi and Timme 1993, DeYoung, Spong and Sullivan 2001). Berger and Bonaccorsi di Patti (2006) contend that it is the separation of ownership and control that creates agency cost, so ownership structure and capital structure should be included together when studying agency costs.
Two types of ownership13 are discussed most in previous research: managerial insider holding which means a portion of the shares is owned by corporate officers and members of the board of directors; large external institutional holding which means a large proportion of the
13
14
shares is owned by an outside institution (e.g. a corporate group, a financial institution etc.). (McConell and Servaes, 1990) I will discuss the influence of the two types of shareholdings one by one.
Scholars have different conclusions14 on how ownership structure influences agency costs and firm performance. When discussing the agency problem, Jensen and Meckling (1976) advocate that managerial insider shares reduce managerial incentives to consume perquisites or to engage in other sub-optimal activities and thus will lower agency costs. Managerial insider shares represent manager’s ownership claim on a firm’s profit. A reduction of managerial insider shares means a less ownership claim for managers on the firm’s profit. It leads to a further divergence of interest between managers and outside equity holders as a result of which manager’s incentive to spend enough efforts in finding “new profitable ventures” for the firm falls (Jensen and Meckling, 1976). So in the view of Jensen and Meckling (1976), the increase of managerial insider shares will lower agency costs and improve firm performance.
On the other hand, Demsetz (1983) and Fama and Jensen (1983) contend that when managerial ownership is low, market discipline will force managers to remain committed to the firm’s value maximization objectives. Then when the insider holding reaches a certain level, although managers’ consumption of perquisites will lead to a reduced value of the firm, the benefit of the engagement in perquisite consumption may outweigh the loss caused by the value decrease of the firm. Morck et al. (1988) argue that at high levels, managerial insider ownership will be “entrenched” because it is difficult for external shareholders to discipline the actions of such managers – managers have enough control to follow their own objectives. Therefore, the interest convergence and the entrenchment effect together predict a non-monotonic relation between managerial insider holding and firm performance.Stulz (1988) develops a model and confirms the non-monotonic relationship between the fraction of managerial insider shares and firm performance. In his study, firm value first increases and then decreases to a minimum when managerial insider ownership reaches 50%. There is a point after which managers are
14
15
entrenched to pursue private benefits, especially when managers have the ability to block value-enhancing takeovers and the discipline caused by the threat of takeover weakens (Stulz, 1988).
In empirical studies, the earlier discussed study of Ang et al. (2000) is also about ownership and agency costs. Besides the finding concerning leverage, their major findings are that agency costs are negatively related with managerial ownership and are significantly higher when the firm is managed by an outsider rather than an insider. Therefore, the study of Ang et al. (2000) mainly supports Jensen and Meckling’s (1976) arguments. The results in the paper of Fleming et al. (2005) also show that agency costs decrease as managerial equity holdings increase. Because of the difficulty to measure the level of agency costs, McConnell and Servaes (1995) use Tobin’s Q to reflect firm performance, and they use managerial ownership squared to test the non-monotonic relationship. The results indicate that the relationship between managerial ownership and firm performance is first positive and then becomes negative when ownership is concentrated by managers, so the non-monotonic relationship is confirmed. The difference between McConell et al. (1990) and McConell and Servaes (1995) is that the former study finds the entrenchment effect appears when the managerial ownership is 5% to 25% while the latter study finds entrenchment occurs when the managerial ownership is above 50%. Both studies are based on U.S. firms. Short and Keasey (1999) extend the research to U.K. firms. They find that managerial ownership is entrenched at high levels and also confirm the relationship to be non-monotonic. In the study of Berger and Bonaccorsi di Patti (2006), as they use profit efficiency to reflect agency costs, high profit efficiency means lower level of agency costs and better firm performance. Their results show that the effect of managerial insider ownership on firm performance is slight negative when the shareholding is below 16%, then it becomes positive from 16% to 60%. And above 60%, it is entrenched and shows negative influence again.
16
will test the possible influence of both effects. The hypothesis is developed as follows:
H2: Managerial insider holding lowers agency costs and influences firm performance positively due to the effect of interest convergence, but at high levels of ownership, managerial insider holding influences firm performance negatively due to the entrenchment effect.
As for institutional ownership, Shleifer and Vishny (1986) have a study about the influence of large outside institutional shareholders. They indicate that the existence of large external institutional holding will mitigate agency conflicts because they have strong incentives to monitor and discipline managers. Besides, large institutional holders can facilitate takeovers and divide their large share gains with the bidder. Shleifer and Vishny (1986) contend that it is a strategy to improve the firm’s operation when the large institutional shareholders are not pleased with the incumbent management and want to replace them. It is an effective tool to monitor managers. Shleifer and Vishny (1986) find that as the large shareholder’s share proportion increases, the possibility of a takeover also increases. For small shareholders, they only obtain on shares owned by them. If the monitoring and takeover costs overwhelm what they can gain, small shareholders do not have strong incentives to monitor managers. While for large shareholders, the return of their large shares is usually large enough to cover the monitoring and takeover costs (Shleifer and Vishny, 1986). So according to Shleifer and Vishny’s (1986) arguments, large external institutional holding will lower agency costs and improve firm performance.
17
and has a positive relationship with firm performance. Under the influence of the other two forces, institutional holding will cause an increase of agency costs and has a negative relationship on firm performance. And in Pound’s (1988) empirical study, the result mainly shows the negative effect of institutional holdings on firm performance.
Brickley, Lease and Smith (1988) also conduct an empirical study and provide evidence showing the negative relationship between institutional holding and firm performance. They find institutional shareholders participate more actively to support antitakeover amendments but oppose proposals which are harmful to their interests. In the study of McConell and Servaes (1990), the results mainly report a positive influence of institutional holdings on firm performance which is consistent with the first force – the efficient monitoring prediction as proposed by Pound (1988). Claessens et al. (2002) conduct a study of both incentive and entrenchment effects of large shareholdings and they find concentrated ownership is associated with better firm performance regardless of the ownership type. Besides, the study of Berger and Bonaccorsi di Patti (2006) reports a strong positive effect of institutional holdings on profit efficiency which indicates the increase in institutional holding will lower agency costs and improve firm performance. They argue that the positive influence may be caused by the favorable monitoring effects of institutional holders.
It can be summarized that empirical studies have both found positive and negative effects of institutional holdings, but one limitation is that few have tested whether the relationship is non-monotonic or not. It is also possible that at certain level the monitoring force dominates, but at high levels, it is difficult to discipline institutional holders and they can co-operate with managers. To test both the positive and the negative effects, I develop the following hypothesis:
H3: Institutional holders can provide more monitoring on managers and therefore influence firm performance positively, but they can co-operate with mangers due to conflict of interest and therefore influence firm performance negatively at high levels of ownership.
18
As is mentioned earlier, previous studies have employed different ways to measure firm performance. The common measures include: accounting ratios derived from financial statements (e.g., Demstetz and Lehn 1985, Mehran 1995, Ang et al. 2000), market returns and the volatility of relevant stocks (e.g., Jensen 1989, Saunders et al. 1990, Cole and Mehran 1998, Cai and Zhang 2006), a combination of market data and accounting data - Tobin’s q (e.g., Tobin and Brainard 1977, Morck et al. 1988, Himmelberg et al. 1999, Zhou, 2001).
Another measure developed to measure firm performance is firm efficiency. It is first brought up by Leibenstein (1966) who measures the discrepancy between firm’s real output and potential maximum output as X-inefficiency. Leibenstein contends that the inefficiency is caused by different objectives between managers and shareholders, inadequate motivation and managerial slack. Stigler (1976) later links productivity efficiency with agency costs in his studies. One advantage of using efficiency to measure firm performance is that it does not consider the difference caused by market factors which influence firm value but are out of the management’s control (factors that are not caused by managers’ activities), and can better reflect the degree of agency costs (Berger and Bonaccorsi di Patti, 2006). Therefore, efficiency is a good way to measure firm performance when studying agency costs.
19
Akhavein et al. 1994, DeYoung and Nolle 1996) also find that revenue inefficiency constitutes the major part of total inefficiency, so profit efficiency which contains both the influence of revenue inefficiency and cost inefficiency can better reflect the overall efficiency level of a firm. By using the distribution free approach (DFA) method and studying a sample of 6000 U.S. banks from 1990 to 1995, Berger and Mester (1997) show that the profit-inefficiency standard deviation is five times of that of cost inefficiency. The study done by them further demonstrates the importance of profit efficiency and the under-estimation of cost efficiency in evaluating a firm’s inefficient activities. Berger and Bonaccorsi di Patti (2006) point out the use of cost efficiency and other measurements as one problem in prior studies. They use profit efficiency in their studies as a reflection of total agency costs. To better reflect total inefficiency and the overall level of agency costs within a firm, I also choose profit efficiency to measure firm performance.
3.2 The technique to measure firm performance
20
which assumes that efficiency is persistent over time and random errors tend to cancel each other out in the course of time.
This paper chooses the non-parametric DEA15 approach to calculate profit efficiency as a measure of firm performance. It is less complex using the DEA-Solver developed by Zhu (2002). Besides, while a large sample size is needed for parametric techniques to make reliable estimates (Isik & Hassan, 2002), the DEA approach can do well with a small sample size and does not require the necessary knowledge of any functional form of the frontier (Ariff and Can, 2008). The empirical efficient frontier of used inputs and outputs represents the “best practice” frontier (Zhu, 2002) – the best-practice banks in the sample constitute the frontier, and the profit efficiency scores of other banks are relative figures to the efficiency scores of best-practice banks.
Because of viewing banks as production or intermediation units, different choice of inputs and outputs has been used in efficiency models in previous studies, producing different results (Berger and Humphrey, 1997). The intermediation approach is used widely in prior studies which considers banks’ primary role as intermediating funds between savers and investors (Chen et al., 2005). Similar to Berger and Bonaccorsi di Patti (2006) and other prior studies (e.g. Maudos et al., 2002; Ariff and Can, 2008), this paper will use the intermediation approach.
4. Methodology
4.1 The model to calculate profit efficiency
Following prior studies (Ariff and Can, 2008), the paper uses three inputs: total loanable funds, number of employees, physical capital. The two outputs used are: total loans, total investments. Table 1 contains a description of all the variables used for profit efficiency calculation. Considering banks’ primary role as intermediating funds between savers and investors, “loanable funds” is therefore one major input of banks. I use total deposits and other funds which serve as the main source of banks’ funds for borrowers. The deposits create interest
15
21
Table 1: Description of variables required for efficiency calculation
This table shows the variables used for calculating profit efficiency and explains their meanings. To be specific, x1, x2,and x3 are the three inputs;w1, w2, and w3 are the price of inputs. And y1 and y2 are the two
outputs; while r1 and r2 are the price of outputs. The intermediation approach is used, considering banks as
intermediaries that mediating funds between savers and investors.
Variables Description
x1 Loanable funds The total deposits and other loanable funds of a bank in a year
x2 Employee number The number of employees of a bank in a year
x3 Fixed assets Book value of property, plant and equipment
y1 Loans Loans net and interbank lending
y2 Total investments The total investments
w1 Price of loanable funds Total interest expenses/ x1
w2 Price of employee number Salaries and benefits of employees/ x2
w3 Price of fixed assets Other operating expenses/ x3
r1 Price of loans Total interest income from loans/ y1
r2 Price of total investments Total investment return
expenses that banks have to pay to individual savers, therefore the price for the input “loanable
funds” is the ratio of total interest expenses and loanable funds. “Employee number” represents
the input of the labor source, and theprice of labor is obtained by using total expenses (salaries and benefits of employees) divided by the total employee number. “Fixed assets” is a measure of physical capital input and is measured as the book value of property, plant and equipment while the price is the ratio of other operating expenses and fixed assets (Maudos et al., 2002; Ariff and Can, 2008). Two major outputs are considered, namely loans and total investments. Banks use the loanable funds from savers and other sources and loan them to meet borrowers’ demands. Furthermore, they also use loanable funds to engage in some investment activities and gain a profit. The total loans here include the net loans and the interbank lending. As banks receive interest income from their loans, the price for the output “loans” is calculated by using total interest income from loans dividing total loans. “Total investments” is the second output and the price for it is the ratio of total investment income and total investments which is the total investment return.
22
assumes an imperfect competition and firms have market power to set output prices (Berger and Mester, 1997). Therefore, in standard profit efficiency function, firms maximize their profit by adjusting inputs and outputs amount as the input price and output price is given. In alternative profit efficiency function, firms maximize their profit by adjusting the price of output and the quantity of inputs (Maudos et al., 2002). Berger and Mester (1997) contend that alternative profit efficiency can better represent reality because there are differences of production quality among banks and the assumption of perfect competition in pricing is questionable. For the above reasons, this paper uses alternative profit efficiency to measure firm efficiency.
This paper follows the alternative profit efficiency model developed by Maudos and Pastor (2003). Suppose there are N firms (i=1,…,N), they use a vector of p inputs xi=(xi1,…, xip)
∈Rp++ and pay prices wi=(wi1,…,wip)∈Rp++ to produce a vector of q outputs yi=(yi1,…,yiq)∈
Rq++ that are sold at prices ri=(ri1,…,riq)∈Rq++. The alternative profit efficiency model assumes
that banks take the quantity of outputs (yi ) and the price of inputs (wi) as given, so they
maximize profits by adjusting the price of outputs ri (e.g. the investment return rate) and the
quantity of inputs xi (e.g. the amount of employees). The APE for firm j can be expressed in a
linear programming:
Max ∑rjq yjq –∑wjp xjp
s.t. ∑λiriq yiq ≥ rjq yjq ∀ q
∑λi xip ≤ xjp ∀ p (1)
The solution corresponds to the output price rj* = (rj1*,…, rjq* ) and the input demand vector xj*
=(xj1*,….., xjp*) that maximize profits with the given prices of inputs (wj) and quantity of outputs
(yj ). The solution is obtained from a linear combination of firms that uses the same or less
amount of inputs and obtains at least as much revenues as firm j. The APE for firm j is calculated as follows:
APEj = Pj/Pj* = (∑rjq yjq –∑wjp xjp)/(∑rjq* yjq –∑wjp xjp*) (2)
where APEj≤1 indicates the ratio between the observed profits Pj and the maximum profits Pj* in
association with the maximum revenue and the demand for inputs xj* that maximize profits for
firm j.
23
Table 2 Summary of all variables
This table shows all the variables that will be employed in the regression model. Independent variables and control variables are one-year lagged.
Variables Meaning Description
effi,t Profit efficiency of a firm To measure firm performance
ecapi,t-1 Equity capital/ gross total assets An inverse measure of leverage
ecap2i,t-1 Square of ecapi,t-1 A quadratic form to test non-monotonicity
owneri,t-1 Largest share percentage Measure the largest proportion of shareholding
institi,t-1 1 if the largest shareholder is an institution Dummy variable to distinguish institutional shareholding
insidei,t-1 1 if the largest shareholder is an insider Dummy variable to distinguish managerial insider holding
otheri,t-1 1 if the largest shareholder is other type Dummy variable to distinguish with institi,t-1 and insidei,t-1
profiti,t-1 The ratio of return on equity Control variable for profit
sizei,t-1 Natural log of total assets Control variable for firm size
4.2 Variables
As is discussed in section 4.1, to measure how leverage and ownership influence firm
performance, I use profit efficiency score as a measure of firm performance, so the profit efficiency of firm i in year t is denoted as effi,t. Similar to Berger and Bonaccorsi di Patti (2006),
higher effi,t. means lower agency costs and better firm performance. All independent variables
and control variables are one-year lagged to allow for lagged effects as is in Margaritis and Psillaki’s (2010) study. They assume that the influence of leverage on firm performance is not instantaneous and time lags also prevail when considering the effect of other variables on firm efficiency. The ratio of equity capital to gross total assets is used as an indicator of capital structure which is denoted as ecapi,t-1, the capital structure of firm i in year t-1. It is an inverse
measure of leverage which is standard in banking research as is contended by Berger and Bonaccorsi di Patti (2006). The first hypothesis predicts that higher leverage will lower agency costs and improve firm performance, so a negative relationship between effi,t and ecapi,t-1 is
expected – the estimated coefficient is therefore interpreted as how an increase in leverage of firm i in year t-1 causes a predicted increase in profit efficiency of firm i in year t. To test H1a, a quadratic form ecap2i,t-1 is included in the model and is expected to be positively significant as to
24
managerial insiders and outside institutions, I divide ownership type of the top-tier holding company into three groups: managerial insider holdings where the largest shareholding are corporate officers or board members; institutional holdings where the largest shareholding is an external institution (e.g. financial institution, other corporate groups); and other types where the largest shareholding is neither of the above two. Therefore, as in Berger and Bonaccorsi di Patti (2006), I use the largest proportion of shares to measure ownership and to distinguish managerial insiders and outside institutions, I include three dummies: insidei,t-1 which equals 1 if
the largest shareholding of firm i in year t-1 are managerial insidersand 0 if otherwise, institi,t-1
which equals 1 if the largest shareholding of firm i in year t-1 is an outside institutionand 0 if otherwise, and otheri,t-1 which equals 1 when the largest shareholding of firm i in year t-1
belongs to other type. I use owneri,t-1 to represents the largest share proportion among all
holdings of firm i in year t-1, so insidei,t-1* owneri,t-1 measures the proportion of shares held by
managerial insiders, institi,t-1*owneri,t-1, measures the proportion of shares held by outside
institutions, and otheri,t-1*owneri,t-1 measures the proportion of shares held by other ownership
type when both insidei,t-1 and institi,t-1 equal 0. To test whether the relationship between
ownership and firm performance is non-monotonic or not, I also include the quadratic forms of the three types of ownership structure: (insidei,t-1*owneri,t-1)2, (institi,t-1*owneri,t-1)2, and (otheri,t-1*owneri,t-1)2. In the study of Berger and Bonaccorsi di Patti (2006), they also include a
third-order term of managerial ownership because it may be entrenched at different levels. To better reflect the relationship between managerial ownership and firm performance, I will also include the third-order term - (insidei,t-1*owneri,t-1)3.
Following prior research (e.g.Berger and Bonaccorsi di Patti, 2006; Ariff and Can, 2008; Margaritis and Psillaki, 2010), profit and size are used to control for firm characteristics. Profit is measured as the ratio of return on equity, so profiti,t-1 represents firm i’s profit in year t-1, and the estimated coefficient measure how a change of firm i’s profit in year t-1 influence
25
In these cases, larger firms are expected to have better performance. On the other hand, hierarchical managerial inefficiencies may also exist in larger firms (Williamson, 1967) and thus have negative influence on firm performance. Some scholars have reported a negative relationship between bank size and efficiency (e.g., DeYoung and Nolle, 1996; Isik and Hassan, 2002), while other studies did not find significant advantage of large banks in efficiency (e.g., Berger and Mester, 1997). In this paper, size is measured by the natural log of a firm’s assets, so
sizei,t-1 refers to the natural log of firm i’s assets in year t-1.
A summary of all the variables and their meanings is available in Table 2.
4.3 Regression model
In this part, I introduce all the models that will be used and the steps of my analysis.
Because my research contains data with cross-section and time series characteristics, panel regression will be used. To investigate the effect that leverage and ownership have on firm performance, I use multivariate panel regression which allows multiple variables into the model for explanation. More precisely, I will use panel regression with time-period fixed effects, assuming each year has a unique intercept. The choice of time-period fixed effects is described in Appendix 1.
The analysis will be done in the following steps: I will first include leverage indicator variables ecapi,t-1 and ecap2i,t-1 in the model to test how leverage influences firm performance as
is described in hypotheses H1 and H1a. Then I will add ownership variables insidei,t-1*owneri,t-1, (insidei,t-1*owneri,t-1)2, institi,t-1*owneri,t-1, (institi,t-1*owneri,t-1)2, otheri,t-1*owneri,t-1, and (otheri,t-1*owneri,t-1)2 into the model and in this step only the quadratic form is included to test
the non-monotonic relationship. Because empirical studies find managerial ownership will be entrenched at different levels – either around 5%-25% (e.g. McConell et al., 1988) or at even higher level above 50%-60% (e.g. McConell and Servaes, 1995; Berger and Bonaccorsi di Patti, 2006), I will add a third-order term (insidei,t-1*owneri,t-1)3 in the model in the third step.
For robustness test, I will first use a subsample16 and redo the regression analysis to see whether the findings still hold. To ensure there are enough data to conduct the regression analysis, I use banks from the two countries - Italy and France which contain both managerial
16
26
and institutional holding as the subsample. For the second robustness check, I will test whether the findings also hold for different time periods. As the time period in this study covers the recent financial crisis, I will add a year dummy ydumi,t to distinguish period before the financial
crisis and period during the financial crisis and see whether the results hold for both periods. The regression model in the first step is as follows:
effi,t=α0+α1*ecapi,t-1+α2*ecap2i,t-1+α3*profiti,t-1+α4*sizei,t-1+εi,t,
(3)
where all variables are designed as is discussed in Section 4.2 and also summarized in Table 2; ε is the mean-zero disturbance term. The influence of leverage is estimated by α1, and α1 is
estimated to be negative for Hypothesis H1 to hold while α2 is estimated to be positive for
Hypothesis H1a to hold.
In the second step, I add ownership variables and only include the quadratic forms to test the non-monotonic relationship between ownership and firm performance. The model is as follows:
effi,t=α0+α1*ecapi,t-1+α2*ecap2i,t-1+α3*profiti,t-1+α4*sizei,t-1+α5*insidei,t-1*owneri,t-1
+α6*(insidei,t-1*owneri,t-1)2+α7*institi,t-1*owneri,t-1+α8*(institi,t-1*owneri,t-1)2
+α9*otheri,t-1*owneri,t-1+α10*(otheri,t-1*owneri,t-1)2+εi,t,
(4)
where all the variables are the same as is designed in Equation (3) except the ownership variables. The influence of managerial ownership is estimated by α5, and α6 will test whether the
relationship is non-monotonic. For the non-monotonic relationship hypothesis to hold, α5 and α6
are expected to have opposite signs and both should be significant. The influence of institutional ownership is estimated by α7, and similarly for the non-monotonic relationship hypothesis
between institutional ownership and firm performance to hold, α7 and α8 are expected to have
opposite signs and both should be significant.
In the third step, the third-order term (insidei,t-1*owneri,t-1)3 is added in case managerial
ownership will be entrenched at different levels. The model is written in the following way:
effi,t=α0+α1*ecapi,t-1+α2*ecap2i,t-1+α3*profiti,t-1+α4*sizei,t-1+α5*insidei,t-1*owneri,t-1
+α6*(insidei,t-1*owneri,t-1)2+α7*institi,t-1*owneri,t-1+α8*(institi,t-1*owneri,t-1)2
+α9*otheri,t-1*owneri,t-1+α10*(otheri,t-1*owneri,t-1)2+α11*(insidei,t-1*owneri,t-1)3
+εi,t, (5)
27
Table 3: Final Sample (Euro area listed banks, 2003-2011)
Banks with missing data from 2004 to 2011 for calculating profit efficiency are excluded which left the sample with 64 banks. Therefore, each bank in each year has a profit efficiency score, and there are 512 efficiency scores. Because one-year lagged value is required as is designed in the model, only banks with consistent data for independent and control variables from 2003 to 2010 are finally selected. The final sample consists of 469 firm-year observations – there are missing data on ownership for the same 5 banks from 2003 to 2010 and for another two banks in year 2005 and another one bank in year 2006 (one-year lagged).
Panel A: Number of firm-year observations for each country
Country Number of observations Country Number of observations Austria 40 Ireland 16 Belgium 16 Italy 104 Finland 16 Luxembourg 0 France 111 Netherlands 8 Germany 39 Portugal 23 Greece 40 Spain 56 Total 469
Panel B: Number of firm-year observations in each year
Year Number of observations Year Number of observations
2004 59 2008 59
2005 59 2009 59
2006 57 2010 59
2007 58 2011 59
Total 469
managerial ownership. Therefore, the influence of managerial ownership on firm performance is together estimated byα5, α6 and α11.
Then as the robustness check, I will first re-estimate Equation (5) with time-period fixed effects using a subsample to see whether all the findings still hold. As the second robustness check17, I will add a year dummy to Equation (5) to see whether the results hold for different time periods if controlled for the influence of the financial crisis.
5. Sample and Data
5.1 Sample selectionAs the calculation of profit efficiency and the use of ownership structure variable
17
Scholars haveargued that in the time of a financial crisis, firms face higher default probabilities and other conditions which may influence the financial decisions (e.g. Hackbarth et. al, 2006; Bharma et. al, 2008). It is not the focus of my paper, but to distinguish the possible differences caused by financial crisis, I include a year dummy to test the robustness of my results. The model is as follows: effi,t=α0+α1*ecapi,t-1+α2*ecap2i,t-1+α3*profiti,t-1+α4*sizei,t-1+α5*insidei,t-1*owneri,t-1+α6*(insidei,t-1*owneri,t-1)2
28
require detailed information disclosure, this paper chooses listed banks18. The time span considered in this study is from 2003 to 2011 which includes the period of the recent financial crisis. The geographical coverage is the countries which joined the euro area before 2003, namely Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, and Spain. Only banks which are listed before 2003 and still in operation till 2012 are selected. After applying these criteria, a primary sample of 83 banks are drawn from the London-based International Bank Credit Analysis LTD’s ‘BankScope’ database.
Because the DEA-Solver recognizes missing data as 0 which will bias the final efficiency score, further work is done on the primary sample. First, for banks that lack information on items that are need for profit efficiency calculation, additional data and double checks are done from the other two available sources: the Datastream database and banks’ annual reports. Since data from either BankScope or Datastream initially come from banks’ annual reports, data after the combination from the three sources are consistent as well. Second, for banks which still have missing data on items required for profit efficiency calculation after additional data collection work is done, they are excluded from the primary sample.
Furthermore, as is explained earlier, this paper allows for lagged effects in the specification of the empirical model, so one year lagged data are needed. To sum up, data from 2004 to 2011 are used for calculating efficiency (the dependent variable), and data from 2003 to 2010 are used for independent and control variables in the regression model.
The final sample is an unbalanced panel which consists of 64 banks in every year from 2004 to 2011 and 469 firm-year observations without missing data. A full description of the final sample is in Table 3.
5.2 Sample description
In this paper, data required for efficiency calculation are collected first. Table 4
presents the descriptive statistics of the variables for profit efficiency calculation sorted by year. As is shown in Table 4, from 2004 to 2011, banks are growing since all the inputs x and the outputs y are gradually increasing. There are more and more loanable funds (deposits) and the price of loanable funds first increases and then decreases in the financial crisis from 2008 to
18
29
Table 4: Descriptive statistics (variables in efficiency calculation) according to year
The table shows the descriptive statistics of variables used for the calculation of profit efficiency adjusted for inflation in the first stage, and they are sorted by different years. Data for loanable funds, fixed assets, loans and total investments are adjusted to the nearest 1000. It contains 469 firm-year observations and the mean and median are reported respectively. The data are collected from BankScope and Datastream.
x1 (Loanable funds) w1 (Price of x1) x2(Employee number) w2 (Price of x2) x3 (Fixed assets)
Mean Median Mean Median Mean Median Mean Median Mean Median 2004 41,648 7,978 0.0714 0.0416 15,232 3,718 60.0364 55.1746 862 250 2005 46,150 8,544 0.0933 0.0457 15,431 3,688 64.7272 59.1868 977 288 2006 52,427 9,651 0.0946 0.0580 17,009 3,769 67.0885 62.8985 1,045 274 2007 60,430 10,635 0.1179 0.0741 18,209 3,737 70.9043 64.0570 1,185 306 2008 62,208 11,872 0.1207 0.0760 20,684 3,965 66.6067 62.6534 1,195 304 2009 66,380 12,985 0.0726 0.0452 22,095 4,194 66.5300 64.4333 1,189 292 2010 70,996 12,975 0.0574 0.0351 21,736 4,017 68.3009 65.3650 1,243 279 2011 69,898 13,274 0.0694 0.0449 23,139 5,173 67.0245 65.3767 1,286 302 Total 58,764 11,202 0.0872 0.0533 19,192 3,992 66.4023 63.0772 1,123 283
y1 (Loans) r1 (Price of y1) y2(Total investment) r2 (Price of y2) w3 (Price of x3)
Mean Median Mean Median Mean Median Mean Median Mean Median 2004 59,151 16,867 0.0547 0.0507 36,273 2,743 1.2176 1.1699 1.2990 0.8626 2005 74,694 19,516 0.0575 0.0504 56,272 5,137 1.2495 1.2276 1.8066 1.0329 2006 85,145 24,347 0.0608 0.0512 65,263 6,181 1.2651 1.2355 1.7195 1.1412 2007 99,902 27,830 0.0697 0.0579 82,090 6,184 0.9776 0.9651 1.6365 1.0768 2008 112,634 30,287 0.0707 0.0650 83,823 6,042 0.5098 0.4662 1.6551 0.9714 2009 109,696 30,024 0.0520 0.0483 72,376 7,736 1.2601 1.1929 1.8562 1.2953 2010 113,614 27,487 0.0458 0.0431 75,073 8,207 0.8499 0.8706 1.8445 1.2703 2011 110,675 29,412 0.0505 0.0442 78,133 8,428 0.6969 0.7862 1.6900 1.2535 Total 95,679 25,250 0.0578 0.0514 68,663 5,691 1.0033 1.0401 1.6885 1.0838
2009. As is explained earlier in Section 4.1, x1*w1 represents the interest expenses of a bank;
although the price of loanable funds w1 decreased in 2009, loanable funds x1 in 2011 are nearly
twice of the loanable funds in 2004, banks’ overall interest expenses are still increasing. The trend is similar with banks’ loans. y1*r1 represents the interest income of a bank, though the
price of loans r1 decreased in 2009, the amount of loans y1 keeps increasing and banks’ interest
30
Table 5: Descriptive statistics (variables in efficiency calculation) according to country
The table shows the descriptive statistics of variables used for the calculation of profit efficiency adjusted for inflation in the first stage, and they are sorted by different countries that are included in this paper. Data for loanable funds, fixed assets, loans and total investments are adjusted to the nearest 1000. It contains 469 firm-year observations and the mean and median are reported respectively. The data are collected from BankScope and Datastream.
x1 (Loanable funds) w1 (Price of x1) x2 (Employee number) w2 (Price of x2) x3 (Fixed assets)
Mean Median Mean Median Mean Median Mean Median Mean Median
Austria 20,153 5,110 0.0769 0.0411 9,468 1,372 57.8110 63.8485 469 110 Belgium 140,900 144,814 0.2533 0.0696 38,368 37,371 61.0368 60.9047 2,038 1,851 Finland 2,712 2,062 0.2302 0.0622 1,525 908 64.3873 60.1565 85 57 France 62,504 2,662 0.0865 0.0889 16,431 1,816 69.7752 64.2201 1,162 74 Germany 108,161 17,420 0.1743 0.0845 21,403 4,479 84.3142 77.4289 1,032 296 Greece 28,102 27,068 0.0516 0.0452 13,814 12,186 42.7725 43.6933 880 843 Ireland 66,744 65,367 0.0531 0.0486 20,262 20,496 58.5899 59.2398 629 586 Italy 4,364 9,103 0.0528 0.0326 19,492 3,388 75.6813 69.4134 1,152 285 Netherlands 12,735 12,993 0.0557 0.0562 2,158 2,170 96.5310 94.0529 175 179 Portugal 28,555 27,100 0.0604 0.0580 12,893 9,459 48.3894 45.5021 518 565 Spain 102,339 35,988 0.0542 0.0486 39,581 10,277 56.7806 54.5234 2,351 815 Total 58,767 11,202 0.0872 0.0533 19,192 3,992 66.4023 63.0772 1,123 283
y1 (Loans) r1 (Price of y1) y2 (Total investments) r2 (Price of y2) w3 (Price of x3)
Mean Median Mean Median Mean Median Mean Median Mean Median
Austria 32,297 8,325 0.0504 0.0478 10,974 2,075 1.0366 0.0504 0.5622 0.4480 Belgium 271,037 248,532 0.1150 0.0684 153,620 136,986 0.9353 1.0951 2.7005 2.7626 Finland 12,142 9,080 0.0538 0.0413 3,350 1,907 1.0535 1.0714 3.9252 1.0792 France 100,294 8,720 0.0507 0.0493 96,886 1,031 1.0669 1.0681 2.3496 1.4122 Germany 144,542 57,198 0.0749 0.0512 238,938 28,529 1.0588 1.0061 2.7421 1.3340 Greece 37,687 38,828 0.0742 0.0717 8,818 7,687 0.9022 1.1011 0.4724 0.4122 Ireland 114,705 119,607 0.0520 0.0509 34,606 34,763 0.7468 0.7124 2.5444 1.7188 Italy 81,405 18,019 0.0520 0.0499 30,946 3,756 0.9852 0.9920 1.1149 0.8823 Netherlands 17,631 17,519 0.0573 0.0541 1,604 1,562 0.9777 0.9486 0.8739 0.8158 Portugal 51,049 55,122 0.0559 0.0558 10,087 8,829 0.9135 1.0561 1.9143 0.9439 Spain 167,641 73,758 0.0521 0.0493 66,147 16,841 1.0108 1.0370 1.1016 1.1586 Total 95,679 25,250 0.0578 0.0514 68,663 5,691 1.0033 1.0401 1.6885 1.0838
overall expenses on salaries have increased. Banks have more fixed assets and also invest more;
r2 represents the investment return (investment income divided by total investments), it is clearly
