Capital Struture Equity Ownership and Firm Performence

Journal of Banking & Finance 34 (2010) 621632
Contents lists available at ScienceDirect
Journal of Banking & Finance

journal homepage: www.elsevier.com/locate/jbf
Capital structure, equity ownership and rm performance

Dimitris Margaritis a,*, Maria Psillaki b
a b
Department of Finance, Faculty of Business, AUT, Auckland 1020, New Zealand Department of Economics, University of Piraeus, Piraeus 18534, Greece
a r t i c l e
i n f o
a b s t r a c t
This paper investigates the relationship between capital structure, ownership structure and rm performance using a sample of French manufacturing rms. We employ non-parametric data envelopment analysis (DEA) methods to empirically construct the industrys best practice frontier and measure rm efciency as the distance from that frontier. Using these performance measures we examine if more efcient rms choose more or less debt in their capital structure. We summarize the contrasting effects of efciency on capital structure in terms of two competing hypotheses: the efciency-risk and franchisevalue hypotheses. Using quantile regressions we test the effect of efciency on leverage and thus the empirical validity of the two competing hypotheses across different capital structure choices. We also test the direct relationship from leverage to efciency stipulated by the Jensen and Meckling (1976) agency cost model. Throughout this analysis we consider the role of ownership structure and type on capital structure and rm performance. 2009 Elsevier B.V. All rights reserved.
Article history: Received 24 April 2008 Accepted 28 August 2009 Available online 1 September 2009 JEL classication: D24 G32 Keywords: Capital structure Agency costs Firm efciency Ownership structure DEA
1. Introduction This paper uses an estimate of a rms production efciency to assess the role of leverage in addressing agency conicts within a rm. More specically, we rst assess the direct effect of leverage on rm performance as stipulated by the Jensen and Meckling (1976) agency cost model. Second, we investigate if rm efciency has an effect on capital structure and whether this effect is similar or not across different capital structure choices. Throughout these analyses we consider explicitly the role of equity ownership on both capital structure and rm performance. Corporate nancing decisions are quite complex processes and existing theories can at best explain only certain facets of the diversity and complexity of nancing choices. By demonstrating how competing hypotheses may dominate each other at different segments of the relevant data distribution we reconcile some of the empirical irregularities reported in prior studies thereby cautioning the standard practice of drawing inferences on capital structure choices based on conditional mean estimates. By using productive efciency as opposed to nancial performance indicators as our measure of (inverse) agency costs we are able to carry
out tests of the agency theory that are not confounded by factors that may not be related to agency costs. Our methodological approach is underpinned by Leibenstein (1966) who showed how different principal-agent objectives, inadequate motivation and incomplete contracts become sources of (technical) inefciency measured by the discrepancy between maximum potential output and the rms actual output. He termed this failure to attain the production or technological frontier as Xinefciency. Based on this we model technology and measure performance by employing a directional distance function approach and interpret the technological frontier as a benchmark for each rms performance that would be realized if agency costs were minimized.1 We then proceed to assess the extent to which leverage acts as a disciplinary device in mitigating the agency costs of outside ownership and thereby contributes to an improvement on rm performance. To properly assess the disciplinary role of leverage in agency conicts we control for the effect of ownership structure and ownership type on rm performance. We also allow for the possibility that at high levels of leverage the agency costs of outside
* Corresponding author. Tel.: +64 9 921 9999; fax: +64 9 921 9940. E-mail addresses: dmargaritis@aut.ac.nz (D. Margaritis), psillaki@unipi.gr (M. Psillaki). 0378-4266/$ - see front matter 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jbankn.2009.08.023
1 As we explain in Section 3, the directional distance function gives the maximum proportional expansion of output(s) and contraction of inputs that is feasible for a given technology thereby yielding a measure of rm efciency relative to best practice. The directional distance function has a dual association with the prot function and thus it provides a useful performance companion when protability is the overall goal of the rm.
622
D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632
debt may overcome those of outside equity whereby further increases in debt can lead to an increase in total agency costs. We turn next to analyze the effects of efciency on capital structure using two competing hypotheses. Under the efciencyrisk hypothesis, more efcient rms may choose higher debt to equity ratios because higher efciency reduces the expected costs of bankruptcy and nancial distress. On the other hand, under the franchise-value hypothesis, more efcient rms may choose lower debt to equity ratios to protect the economic rents derived from higher efciency from the possibility of liquidation (Demsetz, 1973; Berger and Bonaccorsi di Patti, 2006). Thus our paper contributes to the literature in four directions: (1) using X-inefciency as opposed to nancial indicators as a measure of rm performance to test the predictions of the agency cost hypothesis; (2) showing that X-inefciency as a proxy for (inverse) agency costs is an important determinant of capital structure choices; (3) demonstrating how competing hypotheses may dominate each other at different segments of the leverage distribution; and (4) providing new empirical evidence on the relationship between ownership structure, capital structure and rm efciency.2 This is to our knowledge one of the rst studies to consider the association between productive efciency, ownership structure and leverage. In a recent study Berger and Bonaccorsi di Patti (2006) examined the bi-directional relationship between capital structure and rm performance for the US banking industry using a parametric measure of prot efciency as an indicator of (inverse) agency costs while Margaritis and Psillaki (2007) investigated a similar relationship for a sample of New Zealand small and medium sized enterprises using a technical efciency measure derived from a non-parametric Shephard (1970) distance function. In this paper we use a directional distance function approach on a sample of French rms from three different manufacturing industries to address the following questions3: Does higher leverage lead to better rm performance? Would different ownership structures have an effect on rm performance? Does efciency exert a signicant effect on leverage over and above that of traditional nancial measures? Are the effects of efciency and the other determinants of corporate nancing decisions similar across different capital structures? To what extent our results are driven by certain types of owners e.g., family vs. non-family rms? The reminder of the paper is organized as follows. The next section discusses the relationship between rm performance, capital and ownership structure. Section 3 outlines the methodology used in this study to construct measures of rms efciency. Section 4 describes the empirical model used to analyze the relationship between efciency, leverage and ownership. Section 5 describes the data and reports the empirical results. Section 6 concludes the paper.
shareholders lie at the heart of the corporate governance literature (Berle and Means, 1932; Jensen and Meckling, 1976; Shleifer and Vishny, 1986). While there is a relatively large literature on the effects of ownership on rm performance (see for example, Morck et al., 1988; McConnell and Servaes, 1990; Himmelberg et al., 1999), the relationship between ownership structure and capital structure remains largely unexplored.4 On the other hand, a voluminous literature is devoted to capital structure and its effects on corporate performance see the surveys by Harris and Raviv (1991) and Myers (2001). An emerging consensus that comes out of the corporate governance literature (see Mahrt-Smith, 2005) is that the interactions between capital structure and ownership structure impact on rm values. Yet theoretical arguments alone cannot unequivocally predict these relationships (see Morck et al., 1988) and the empirical evidence that we have often appears to be contradictory. In part these conicting results arise from difculties empirical researchers face in obtaining direct measures of the magnitude of agency costs that are not confounded by factors that are beyond the control of management (Berger and Bonaccorsi di Patti, 2006). In the remainder of this section we briey review the literature in this area focusing on the main hypotheses of interest for this study. 2.1. Firm performance and capital structure The agency cost theory is premised on the idea that the interests of the companys managers and its shareholders are not perfectly aligned. In their seminal paper Jensen and Meckling (1976) emphasized the importance of the agency costs of equity arising from the separation of ownership and control of rms whereby managers tend to maximize their own utility rather than the value of the rm. These conicts may occur in situations where managers have incentives to take excessive risks as part of risk shifting investment strategies. This leads us to Jensens (1986) free cash ow theory where as stated by Jensen (1986, p. 323) the problem is how to motivate managers to disgorge the cash rather than investing it below the cost of capital or wasting it on organizational inefciencies. Thus high debt ratios may be used as a disciplinary device to reduce managerial cash ow waste through the threat of liquidation (Grossman and Hart, 1982) or through pressure to generate cash ows to service debt (Jensen, 1986). In these situations, debt will have a positive effect on the value of the rm. Agency costs can also exist from conicts between debt and equity investors. These conicts arise when there is a risk of default. The risk of default may create what Myers (1977) referred to as an underinvestment or debt overhang problem. In this case, debt will have a negative effect on the value of the rm. Building on Myers (1977) and Jensen (1986), Stulz (1990) develops a model in which debt nancing is shown to mitigate overinvestment problems but aggravate the underinvestment problem. The model predicts that debt can have both a positive and a negative effect on rm performance and presumably both effects are present in all rms.5 We allow for the presence of both effects in the empirical specication of the agency cost model. However we expect the impact of leverage to be negative overall. We summarize this in terms of our rst testable hypothesis.
2. Firm performance, capital structure and ownership Conicts of interest between owners-managers and outside shareholders as well as those between controlling and minority
2 Most studies focus on analyzing the nancial structure-performance relationship for large rms in the US and UK. These ndings may not be representative for countries with different legal and institutional settings (see La Porta et al., 1998). There is relatively little evidence for Continental Europe where the legal environment and the nature of ownership and control are different, ownership concentration is higher and family ownership and control are more dominant compared to US/UK (see Faccio and Lang, 2002). 3 Civil law systems provide less investor and creditor protection than common law systems and among the civil-law systems the French system provides the least protection (see La Porta et al., 1998). In addition, de Jong et al. (2008) report that creditor right protection has a signicant effect on capital structure. As legal structures with little investor and creditor protection tend to exacerbate information asymmetries and contracting costs, a study focusing on French rms presents some interesting features for the purposes of our investigation.
4 Recent international studies in this area include Brailsford et al. (2002) for Australian rms, Short et al. (2002) for UK rms, and King and Santor (2008) for Canadian rms. 5 A common element in the models of Myers, Jensen and Stulz is the focus on the link between the rms investment opportunity set and the effects of debt on the value of the rm. Thus a reasonable conjecture will be that for rms with few growth opportunities the positive effect of debt on rm performance will be more dominant whereas the opposite effect will apply for rms with high growth opportunities (see McConnell and Servaes, 1995).
623
According to the agency cost hypothesis (H1) higher leverage is expected to lower agency costs, reduce inefciency and thereby lead to an improvement in rms performance. 2.2. Reverse causality from rm performance to capital structure But rm performance may also affect the choice of capital structure. Berger and Bonaccorsi di Patti (2006) stipulate that more efcient rms are more likely to earn a higher return for a given capital structure, and that higher returns can act as a buffer against portfolio risk so that more efcient rms are in a better position to substitute equity for debt in their capital structure. Hence under the efciency-risk hypothesis (H2), more efcient rms choose higher leverage ratios because higher efciency is expected to lower the costs of bankruptcy and nancial distress. In essence, the efciency-risk hypothesis is a spin-off of the trade-off theory of capital structure whereby differences in efciency, all else equal, enable rms to ne tune their optimal capital structure. It is also possible that rms which expect to sustain high efciency rates into the future will choose lower debt to equity ratios in an attempt to guard the economic rents or franchise value generated by these efciencies from the threat of liquidation (see Demsetz, 1973; Berger and Bonaccorsi di Patti, 2006). Thus in addition to a equity for debt substitution effect, the relationship between efciency and capital structure may also be characterized by the presence of an income effect. Under the franchise-value hypothesis (H2a) more efcient rms tend to hold extra equity capital and therefore, all else equal, choose lower leverage ratios to protect their future income or franchise value. Thus the efciency-risk hypothesis (H2) and the franchise-value hypothesis (H2a) yield opposite predictions regarding the likely effects of rm efciency on the choice of capital structure. Although we cannot identify the separate substitution and income effects our empirical analysis is able to determine which effect dominates the other across the spectrum of different capital structure choices. 2.3. Ownership structure and the agency costs of debt and equity The relationship between ownership structure and rm performance dates back to Berle and Means (1932) who argued that widely held corporations in the US, in which ownership of capital is dispersed among small shareholders and control is concentrated in the hands of insiders tend to underperform. Following from this, Jensen and Meckling (1976) develop more formally the classical owner-manager agency problem. They advocate that managerial share-ownership may reduce managerial incentives to consume perquisites, expropriate shareholders wealth or to engage in other sub-optimal activities and thus helps in aligning the interests of managers and shareholders which in turn lowers agency costs. Along similar lines, Shleifer and Vishny (1986) show that large external equity holders can mitigate agency conicts because of their strong incentives to monitor and discipline management. In contrast Demsetz (1983) and Fama and Jensen (1983) point out that a rise in insider share-ownership stakes may also be associated with adverse entrenchment effects that can lead to an increase in managerial opportunism at the expense of outside investors. Whether rm value would be maximized in the presence of large controlling shareholders depends on the entrenchment effect (Claessens et al., 2002; Villalonga and Amit, 2006; Dow and McGuire, 2009). Several studies document either a direct (e.g., Shleifer and Vishny, 1986; Claessens et al., 2002; Hu and Zhou, 2008) or a non-monotonic (e.g., Morck et al., 1988; McConnell and Servaes, 1995; Davies et al., 2005) relationship between ownership structure and rm performance while others (e.g., Demsetz and Lehn, 1985; Himmelberg et al., 1999; Demsetz and Villalonga,
2001) nd no relation between ownership concentration and rm performance.6 Family rms are a special class of large shareholders with unique incentive structures. For example, concerns over family and business reputation and rm survival would tend to mitigate the agency costs of outside debt and outside equity (Demsetz and Lehn, 1985; Anderson et al., 2003) although controlling family shareholders may still expropriate minority shareholders (Claessens et al., 2002; Villalonga and Amit, 2006). Several studies (e.g., Anderson and Reeb, 2003a; Villalonga and Amit, 2006; Maury, 2006; King and Santor, 2008) report that family rms especially those with large personal owners tend to outperform non-family rms. In addition, the empirical ndings of Maury (2006) suggest that large controlling family ownership in Western Europe appears to benet rather than harm minority shareholders. Thus we expect that the net effect of family ownership on rm performance will be positive. Large institutional investors may not, on the other hand, have incentives to monitor management (Villalonga and Amit, 2006) and they may even coerce with management (McConnell and Servaes, 1990; Claessens et al., 2002; Cornett et al., 2007). In addition, Shleifer and Vishny (1986) and La Porta et al. (2002) argue that equity concentration is more likely to have a positive effect on rm performance in situations where control by large equity holders may act as a substitute for legal protection in countries with weak investor protection and less developed capital markets where they also classify Continental Europe. We summarize the contrasting ownership effects of incentive alignment and entrenchment on rm performance in terms of two competing hypotheses. Under the convergence-of interest hypothesis (H3) more concentrated ownership should have a positive effect on rm performance. And under the ownership entrenchment hypothesis (H3a) the effect of ownership concentration on rm performance is expected to be negative. The presence of ownership entrenchment and incentive alignment effects also has implications for the rms capital structure choice. We assess these effects empirically. As external blockholders have strong incentives to reduce managerial opportunism they may prefer to use debt as a governance mechanism to control managements consumption of perquisites (Grossman and Hart, 1982). In that case rms with large external blockholdings are likely to have higher debt ratios at least up to the point where the risk of bankruptcy may induce them to lower debt. Family rms may also use higher debt levels to the extent that they are perceived to be less risky by debtholders (Anderson et al., 2003). On the other hand the relation between leverage and insider share-ownership may be negative in situations where managerial blockholders choose lower debt to protect their non-diversiable human capital and wealth invested in the rm (Friend and Lang, 1988). Brailsford et al. (2002) report a non-linear relationship between managerial share-ownership and leverage. At low levels of managerial ownership, agency conicts necessitate the use of more debt but as managers become entrenched at high levels of managerial ownership they seek to reduce their risks and they use less debt. Anderson and Reeb (2003b) nd that insider ownership by managers or families has no effect on leverage while King and Santor (2008) report that both family rms and rms controlled by nancial institutions carry more debt in their capital structure.
6 McConnell and Servaes (1995) indicate that the relation between ownership structure and rm performance may differ between low- and high-growth rms. Their conjecture is that ownership is likely to be more important for low-growth than for high-growth rms.
624
3. Benchmarking rm performance In this section we explain how we benchmark rm performance. More technical details are given in the appendix. We rely on duality theory and the use of distance functions. Directional distance functions are alternative representations of production technology which readily model multiple input and multiple output technological relationships. They measure the maximum proportional expansion in outputs and contraction in inputs that rms would be able to achieve by eliminating all technical inefciency relative to the performance of their best performing peers. They are the primal measures; their dual measures are the more familiar value functions such as prot, cost and revenue. We interpret these inefciencies to be the result of contracting costs, managerial slack or oversight. They differ from allocative inefciencies which are due to the choice of a non-optimal mix of inputs and outputs.7 We estimate the production technology and derive rm efciency measures using data envelopment analysis (DEA) which is a non-parametric technique that employs linear programming methods to construct a piecewise linear representation of the frontier technology. DEA is a deterministic enveloping technique and hence it is not subject to standard econometric problems. There are no regression parameters to estimate, there is no functional form, no error term and no distinction between dependent and independent variables. Hence there are no endogeneity problems to handle. In addition, by modelling technology using a directional distance function we allow rms to optimize in both input and output directions rather than in a single direction as for example is the case with the more traditional Shephard type distance functions. DEA is purely a data driven technique. It operates by tting the tightest cone enveloping the data. At each segment of the piecewise frontier it identies a peer group of efcient rms for each individual rm being evaluated. As it makes no accommodation for noise it envelopes the data set fully rather than nearly the way for example a stochastic frontier does (see Fried et al., 2008). While it may be more sensitive to the presence of outliers, DEA has several advantages over parametric stochastic frontier methods; e.g., it avoids functional form misspecication problems and the effects of endogeneity or incorrect error term distribution assumptions that can impact signicantly on efciency measures. More importantly, in our context DEA provides efciency measures (performance index numbers) that are not subject to the endogeneity bias problems that arise when stochastic frontier estimates of rm inefciency are used as explanatory variables in second stage regressions. A rms ability to achieve best practice relative to its peers will be compromised in situations where it is forced to forego valuable investment opportunities, participate in uneconomic activities that sustain growth at the expense of protability or being subject to other organizational inefciencies. Following Leibenstein (1966) we use technical or X-inefciency as a proxy for the (inverse) agency costs arising from conicts between debt holders and equity holders or from different principal-agent objectives. These conicts will give rise to resource misallocations and potential output will be sacriced. The magnitude of agency costs will vary from rm to rm (see Jensen and Meckling, 1976) and thus individual rms with similar technologies can be benchmarked against their best performing peers. As in Berger and Bonaccorsi di Patti (2006) we view these best practice rms as those which minimize the agency costs of outside equity and outside debt.
7 Duality theory provides the basis for the decomposition of prot efciency into allocative efciency and the technical efciency component captured by the directional distance function and described in terms of input and output quantities (see Fre et al., 2008). However, we cannot measure prot efciency since we do not have data on input and output prices.
4. The empirical model We use a two equation cross-section model to test the agency cost hypotheses (H1) and (H3/H3a) and the reverse causality hypotheses (H2 and H2a). 4.1. Firm performance The regression equation for the rm performance model is given by:
EFF i;t a0 a1 LEV i;tl a2 LEV 2 a3 Z 1i;tl ui;t ; i;tl
where EFF is the rms efciency measure8; LEV is the debt to total assets ratio; Z1 is a vector of control variables; u is a stochastic error term; and as we explain in Section 5 we allow for lagged effects in the specication of the empirical model. According to the agency cost hypothesis the effect of leverage (LEV) on efciency should be positive. However, the possibility exists that at sufciently high leverage levels, the effect of leverage on efciency may be negative.9 The quadratic specication in (1) is consistent with the possibility that the relationship between leverage and efciency may not be monotonic, viz. it may switch from positive to negative at higher leverage. Leverage will have a negative effect on efciency for values of LEV < a1/2a2. A sufcient condition for the inverse U-shaped relationship between leverage and efciency to hold is that a2 < 0. The variables included in Z1 control for rm characteristics. More specically, we assume that protability, size, asset structure, growth opportunities, ownership structure and ownership type are likely to inuence rm efciency. Protability (PR) is measured by the ratio of prots (EBIT) to total assets. In general we expect a positive effect of (past) protability on efciency. More protable rms are generally better managed and thus are expected to be more efcient. Firm size (SIZE) is measured by the natural log of the rms sales. The effect of this variable on efciency is likely to be positive as larger rms are expected to use better technology, be more diversied and better managed. Larger rms may also enjoy economies of scale in monitoring top management (Himmelberg et al., 1999). But larger rms may suffer from hierarchical managerial inefciencies and also incur larger monitoring costs (see Williamson, 1967). Thus as in Himmelberg et al. (1999) we allow for nonlinearities in the effect of rm size on performance by including the square of the natural log of sales in the efciency equation. Tangibility (TANG) is measured as the ratio of xed tangible assets divided by the total assets of the rm. Tangibles are easily monitored and provide good collateral and thus they tend to mitigate agency conicts (Himmelberg et al., 1999). As in Himmelberg et al. (1999) we allow for nonlinearities in the effect of asset structure on rm performance by including the square of the tangibles to assets ratio. Intangibility (INTG) is measured by the ratio of intangible assets to the rms equity. This variable may be considered as an indicator of future growth opportunities (see Titman and Wessels, 1988) but its effect on rm performance is generally ambiguous
8 In the empirical part of the paper efciency is measured as 1=1 DT where DT is the value of the directional distance function (see the Appendix). This has the advantage of restricting the efciency measures in the 01 range and hence facilitates comparisons with more conventional e.g., Shephard type efciency measures where a fully efcient rm has a score of 1. 9 Highly levered rms may face nancing constraints that will prevent them from adjusting their capital to labor ratios and thus attain more efcient production allocations (see Spaliara, 2009). Debt nancing may also have a negative effect on rm performance for rms with plentiful growth opportunities (see Myers, 1977; Jensen, 1986; McConnell and Servaes, 1995).
625
especially if these opportunities are the result of excessive risktaking behavior given the size of equity (see Myers, 1977). Sales growth (GROWTH) this variable can also serve as a proxy for growth prospects and investment opportunities. It is likely to have a positive effect on rm performance (see Claessens et al., 2002; Maury, 2006; King and Santor, 2008). We consider both the effects of ownership concentration and ownership type on rm performance. We measure ownership concentration (OWNC) by the percentage of shares held by those classied as large shareholders. We allow the effect of ownership to vary in a piecewise linear form (see Morck et al., 1988) across different segments of ownership concentration by introducing dummy variables (Independence Indicators) dened over three different ranges of ownership holdings: low concentration (OWN1) with no owner holding more than a 25% stake in the company; intermediate concentration (OWN2) with the largest shareholder(s) holding between 25% and 50%; and high concentration (OWN3) representing equity holdings in excess of 50%. Hence the piecewise ownership structure variable (OWNER) used in the empirical model is the product of OWNC times OWN. To the extent that large or block owners are more capable of monitoring and aligning management to their objectives ownership concentration would be expected to have a positive effect of rm performance (see Jensen and Meckling, 1976; Shleifer and Vishny, 1986; Jiraporn and Gleason, 2007). But increased ownership share may adversely affect performance because it raises the rms cost of capital due to decreased market liquidity or decreased diversication opportunities (Fama and Jensen, 1983). Morck et al. (1988) argue that concentrated ownership may be associated with a negative (entrenchment) effect on rm performance where the overall effect on rm value may be positive at low concentration but negative at high concentration levels. They also suggest that the relationship between ownership structure and rm performance is likely to vary across industries. These predictions are corroborated by McConnell and Servaes (1995) who report that ownership has a positive effect on performance for low growth rms but an insignicant albeit positive effect for high growth rms. Demsetz (1983) on the other hand argues that although different types of ownership may intensify agency problems, they also generate compensating advantages so that overall ownership structure should not have any signicant effect on rm performance. This view is supported by the empirical ndings of Demsetz and Lehn (1985) and Demsetz and Villalonga (2001). We control for the effect of ownership type by dividing ultimate owners into three groups: (1) rms owned by families or related individuals (Family); (2) rms owned by nancial institutions banks, mutual funds, investment and insurance companies (Financial); and rms with other types of ownership (Other). The latter is a generic category that includes non-nancial companies holding shares in other companies (cross-holding structures). Small rms are more likely to be family controlled and their owners are likely to be involved in the management of the company while nancial companies are more likely to be widely held with no owner involvement in the companys management. Since family ownership reduces the classic owner-manager conict, agency theory would predict a positive effect of family ownership on rm performance (Morck et al., 1988; Anderson and Reeb, 2003a; Villalonga and Amit, 2006). This effect may be offset in situations where family managed rms forego the opportunity to hire professional managers that may be able to run the business more efciently. Similarly, while there may be positive effects of cross-ownership ties on rm performance (e.g., technology transfers, vertical ties), such structures also entail offsetting entrenchment effects, increasing hierarchical group structure monitoring costs (see Williamson, 1967) or other costs associated with inefcient business diversication.
4.2. The leverage model The capital structure equation relates the debt to assets ratio to our measure of efciency as well as to a number of other factors that have commonly been identied in the literature to be correlated with leverage (see Harris and Raviv, 1991; Myers, 2001). The leverage equation is given by:
LEV i;t b0 b1 EFF i;tl b2 Z 2i;tl v i;t ;
where Z2 is a vector of factors other than efciency (EFF) that correlate with leverage; v is a stochastic error term and again we allow for lagged effects in the empirical specication of the model. Under the efciency-risk hypothesis, efciency has a positive effect on leverage, i.e. b1 > 0; whereas under the franchise-value hypothesis, the effect of efciency on leverage is negative, i.e. b1 < 0. We use quantile regression analysis to examine the capital structure choices of different subsets of rms in terms of these two conditional hypotheses. This is in line with Myers (2001) who emphasized that there is no universal theory but several useful conditional theories describing the rms debtequity choice. These different theories will depend on which economic aspect and rm characteristic we focus on. The variables included in Z2 control for rm characteristics that are likely to inuence the choice of capital structure (see Harris and Raviv, 1991; Rajan and Zingales, 1995). They are the same variables used in the agency cost model such as protability, size, asset structure, growth opportunities, and ownership structure and type. There are conicting theoretical predictions on the effects of protability on leverage (see Harris and Raviv, 1991; Rajan and Zingales, 1995; Booth et al., 2001). Myers (1984) predicts a negative relationship because he argues rms will prefer to nance new investments with internal funds rather than debt. According to his pecking order theory rms nancing choices follow a hierarchy in which internal cash ows (retained earnings) are preferred over external funds, and debt is preferred over equity nancing. Thus more protable rms are more likely to nance their growth by retained earnings whereas less protable rms will use more debt nancing. Most empirical studies report a negative relationship between protability and leverage although this association may be complicated by the presence of strong investment opportunities (see Booth et al., 2001). The effect of size on leverage is expected to be positive. As larger rms are more diversied and tend to fail less often than smaller ones, we would expect that they have better access to credit and are able to sustain more debt (Friend and Lang, 1988). The tangibility of the rms assets can serve as a proxy for the agency costs of debt and the costs of nancial distress (Myers, 1977; Booth et al., 2001). Firms with more tangible assets have in general greater ability to secure debt as these assets can be used as collateral (Jensen and Meckling, 1976). Thus asset tangibility is expected to have a positive effect on leverage (Titman and Wessels, 1988). Arguably the use of collateral as a device to lower agency costs associated with debt may play an even more important role in countries like France where creditor protection is relatively weak in comparison to other developed countries (see La Porta et al., 1998). The degree of asset intangibility measured by the ratio of intangible assets to total assets can serve as both a proxy for growth opportunities and as a source of collateral. Growth opportunities are generally associated with an increase in the agency costs of debt and are thus expected to have a negative effect on leverage (Myers, 1977). To the extent that intangibles may be perceived by lenders as providing some form of security they will have a positive effect on leverage. The overall effect of intangibles on leverage is likely to be negative; especially for rms that have greater opportunities to expropriate bondholder wealth by substituting
626
safer assets for riskier assets (see Booth et al., 2001; Anderson et al., 2003). Sales growth may be considered as another indicator of future growth opportunities. Low growth rms will have less opportunities to substitute low risk for high risk (high return) investments; hence they should incur lower agency costs of debt and should be able to carry more debt in their capital structure. High growth rms on the other hand may face a more intense debt overhang problem of the type described by Jensen and Meckling (1976) and Myers (1977). But if recontracting costs are kept low the underinvestment incentives are much smaller (see Booth et al., 2001). And if growth opportunities are viewed as an indicator of a successful business the effect of growth on leverage may be positive. This effect may be reinforced by owners of smaller private especially closely held companies who are fearful to lose control or are unable to issue new equity and thus opt to fund growth opportunities with leverage (see Giannetti, 2003). Ownership structure may have a positive or a negative effect on the amount of debt held in the rms capital structure. Firms where shareholders rights are weak are expected to carry more debt in their capital structure as these rms are expected to incur higher agency costs (Jiraporn and Gleason, 2007). Because of their longterm commitments to the rm, family owned rms have stronger incentives to mitigate agency conicts with debt claimants and as a result they may face lower costs of debt nancing (Anderson et al., 2003). Thus family rms may carry more debt in their capital structure. On the other hand diversied shareholders have incentives to expropriate debtholder wealth by investing in risky projects (Jensen and Meckling, 1976) in which case we would expect debt holders to require a higher return. Similarly, when leverage is high the risk of bankruptcy increases which may then induce rms to lower debt. For example, an increase in insider ownership may push rms to reduce leverage for fear of bankruptcy or loosing control to banks (Friend and Lang, 1988). Firms owned by nancial institutions may have better access to nance but nancial institutions also tend to be risk averse in their investment strategy.
5. Empirical results In this section we provide answers to the questions of Section 1. As stated in the introduction we are interested in examining how capital structure choices affect rm performance as well as the reverse relationship between efciency and leverage. More precisely, we want to examine if leverage has a positive effect on efciency and whether the reverse effect of efciency on leverage is similar across the spectrum of different capital structures. We are also interested in assessing empirically the effects of ownership structure on capital structure and on rm performance. As explained in Section 3, we measure rm efciency using the directional distance function. We choose to estimate the directional distance function using deterministic non-parametric frontier methods (DEA). The DEA model is constructed using a single output (value-added) and two inputs (capital and labor) technology. The labor input is measured by the total number of full-time equivalent employees and working proprietors whereas capital is measured by the rms xed tangible assets. We set the elements of the directional vector (g) equal to the sample averages of the input and output variables (see the Appendix). Table 1 (Panel A) gives the descriptive statistics of the rms in the sample for 2005. The data comprises samples of French rms from two traditional manufacturing industries (textiles and chemicals) and a growth industry (computers and related activities and R&D). We collect data from 2002 to 2005 to allow for sufcient lagged dynamic structure to resolve the identication and endogeneity problems in the empirical specication of the cross-section
model. We limit the data to all rms that report non-negative values for all inputs and output. The source of the data is the Diane database compiled by the Bureau van Dijk.10 On average rms in the chemicals industry are much larger and more capital intensive than rms in the computers and textiles industries. Firms in the computers and R&D industry have higher intangibles to assets ratios and carry on average more debt in their capital structure. Protability appears to be much higher on average in the textiles industry but its distribution is highly skewed note that the median chemicals rm is more protable than the median textiles rm. Firms in the computers industry appear to be closer on average to the technological frontier compared to those in the chemicals and textiles industries. We do not nd any signicant differences in efciency performance over time (i.e. from 2002 to 2005). There appears to be a slight improvement in performance for rms in the chemicals industry and a slight decline on average for rms in the textiles industry. Panel B of Table 1 shows that family ownership is highest in the textiles industry and comparatively low in the chemicals industry. While ownership is quite concentrated across all industries, rms in the computers and R&D industry appear to have the least concentrated ownership structure. This observation is consistent with the predictions of the Mahrt-Smith (2005) model. For growth rms where long-term project discovery and development investments are more important than short-term projects, ownership is likely to be more dispersed as managers are motivated to protect these long-term rents. Table 2 shows how efciency, protability and leverage vary across family vs. non-family rms; and across rms with dispersed (<25%) vs. more concentrated (>25%) ownership.11 We nd that family rms are much more efcient and more protable than non-family rms. These differences are statistically signicant on average across all industries. We also nd that family rms in the computers and R&D and textiles industries carry less debt in their capital structure than non-family rms and these differences are statistically signicant on average. This is an interesting nding since family rms in both industries are also shown to be more efcient. As Anderson et al. (2003) point out family monitoring and control may result in better operating performance for family rms and this may also mitigate the owner-manager agency conicts and hence the need to use debt as a disciplinary device. There is no evidence of statistically different debt ratios between family and non-family rms in chemicals. Table 2 shows that the net effect of ownership concentration on efciency varies across industries. We nd that efciency is higher on average for rms with more concentrated ownership in computers and R&D and for rms with more dispersed ownership in chemicals. These differences in efciency between higher and lower concentration are statistically signicant on average. On the
10 We collect data for rms with at least ve employees and an annual turnover greater than EUR 200,000. The majority of these rms are small (dened as those with 550 employees), followed by medium-sized rms (those with 51500 employees). In particular, 60% of the rms in chemicals are small, 32% are medium-sized and 8% are large rms (more than 500 employees). In computers and R&D, 85% of the rms are small, 13% are medium and 2% are large. In textiles, 75% of the rms are small, 23% are medium and 2% are large. 11 As in Claessens et al. (2002) and Anderson et al. (2003) we do not separate family ownership from family management and we do not expect this distinction to be important in smaller unlisted rms. Faccio and Lang (2002) report that about twothirds of French family controlled rms have top managers from the controlling family and this ratio is expected to be much higher for unlisted rms. Similarly, we do not expect a signicant wedge between ownership and control for the type of rms we consider and hence we do not control for mechanisms that may be used to enhance control. Such mechanisms (e.g., cross-holdings, dual class shares) are rare in France (see Faccio and Lang, 2002) or are used infrequently (e.g., pyramidal structures) especially by smaller family rms.
627
Table 1 Descriptive statistics. The table reports the 2005 sample statistics for inputs, output, efciency measures and rm characteristics for the three industries (Chemicals, Computers and R&D, and Textiles). In Panel A, output (Y) is value-added; labor (L) is the number of employees; tangibles (K) is xed tangible assets; Y/L is labor productivity; K/L is capital intensity; efciency is measured as 1/(1 + distance function value) and is reported for 20032005; prot is earnings before interest and taxes; PR is the prot to assets ratio; LEV is the debt to assets ratio; INTG is the intangibles to total assets ratio; TANG is the tangibles to equity ratio; growth is the percentage change in sales revenue; OWNC is the proportion of largest shareholder(s) equity ownership. In Panel B all variables are 0/1 dummy variables, OWN1 denotes low, less than 25% ownership concentration; OWN2 denotes intermediate, greater than 25% but less than 50% ownership concentration; OWN3 denotes high, greater than 50% ownership concentration; family, nancial and other are ownership type variables. Chemicals Mean Panel A: summary statistics Output (Y) 17855.81 Labor (L) 166.24 Tangibles (K) 10101.60 Y/L 94.89 K/L 47.76 Efciency_05 0.82 Efciency_04 0.81 Efciency_03 0.81 Revenue 69714.61 Prot 5973.86 Total assets 66863.22 Total debt 33239.00 Intangibles 3771.61 PR 0.07 LEV 0.58 INTG 0.04 TANG 0.70 Growth 0.07 OWNC 0.76 OWN1 Std. Dev. 63789.21 467.06 37225.20 209.82 123.25 0.24 0.24 0.23 258236.50 51290.61 416953.10 191354.70 38017.68 0.13 0.26 0.10 3.58 0.30 0.27 OWN2 Median 2363.00 37.00 719.00 65.05 20.07 0.94 0.92 0.91 8338.50 213.00 5621.00 2928.50 38.50 0.06 0.56 0.01 0.32 0.05 0.90 OWN3 Computers and R&D Mean 4413.77 55.95 341.01 78.78 7.11 0.87 0.86 0.87 8245.26 395.53 7134.45 4175.51 643.29 0.08 0.69 0.08 0.23 0.14 0.65 OWN1 11.83 4.22 2.65 4.96 3253 Std. Dev. 21226.04 235.44 2888.03 130.61 83.32 0.17 0.16 0.16 35689.20 5338.94 44318.08 20492.68 5595.82 0.26 0.42 0.14 3.88 0.34 0.30 OWN2 23.63 15.48 2.27 5.88 Median 997.00 15.00 35.00 64.14 2.06 0.93 0.92 0.93 1816.00 75.00 1092.00 679.00 11.00 0.05 0.65 0.01 0.08 0.09 0.52 OWN3 64.54 35.58 4.92 24.04 Textiles Mean 2568.00 51.09 711.32 51.43 11.95 0.78 0.79 0.80 9283.36 432.97 6407.74 3271.51 252.55 0.15 0.56 0.05 0.37 0.02 0.68 OWN1 4.92 2.80 1.14 0.98 1705 Std. Dev. 6491.76 94.16 2186.38 36.71 25.70 0.22 0.20 0.19 21864.07 3192.63 19825.20 8747.22 1299.80 0.05 0.30 0.10 2.36 0.30 0.27 OWN2 21.90 17.74 0.59 3.57 Median 976.00 23.00 137.50 41.95 5.42 0.86 0.86 0.87 3129.00 71.00 1879.50 867.50 15.00 0.05 0.54 0.01 0.15 0.00 0.51 OWN3 72.21 40.21 9.18 22.82
Panel B: ownership concentration tabulation by ownership type (%) 3.87 24.16 71.97 Of which: Family Financial Other Obs. 1.94 0.42 1.51 1188 18.86 0.76 4.50 13.89 8.00 50.18
other hand there is no statistical evidence of such differences in efciency for rms in the textiles industry. We turn next to empirically assess the relationship between leverage and efciency controlling for the effect of ownership and other rm characteristics. The simultaneous equation system given by (1) and (2) above requires adequate structure to be properly identied. An obvious way to deal with the identication problem is by imposing relevant restrictions on the structural system. Undoubtedly the task of both properly identifying the system of equations for efciency and leverage and ensuring that the conditioning variables entering these two equations are indeed exogenous is fraught with difculty. We have dealt with the identication and endogeneity issues in the following way. Arguably both the effect of leverage on efciency and the reverse effect from efciency on leverage are not expected to be instantaneous. Time lags are also likely to prevail when considering the effect of other conditioning variables on efciency and leverage. For example, the pecking order theory states that it is past not current protability that is envisaged to have an effect on leverage.12 An explicitly account of the dynamics in the relationship between efciency and leverage would thus help solve the identica-
tion problem while rendering a structure that is more robust to simultaneity bias problems. Based on this we have proceeded to estimate the agency cost and leverage equations using both static and dynamic model specications.13 We have estimated structural forms of these equations using instrumental variables techniques and their dynamic or reduced form specications using OLS and quantile regressions. The results we obtained from different models or estimation techniques appear to be quite robust, particularly in relation to assessing the predictions of the agency cost and efciency hypotheses. We only report the results obtained from estimating dynamic models for both the efciency and leverage equations. The regressors in these equations are predetermined (lagged endogenous or exogenous) variables thereby circumventing simultaneity problems. Parsimonious forms of these equations were obtained by applying a standard general to specic methodology starting with models that used variables with up to three year lags. We use two-year average to mitigate the effects of year to year uctuations in some of these variables. Table 3 reports the estimates of the rm performance equation. We report both cross-section results (Panel A) and panel estimates with random-effects (Panel B). The panel estimates use larger samples of rms-years. The results show that leverage has a signicant
12 Given the stability of ownership patterns that we observe in our sample we treat ownership as an exogenous variable rather than the endogenous outcome of competitive selection as advocated by Demsetz (1983). This is contrary to the results of Demsetz and Lehn (1985), Himmelberg et al. (1999) and Demsetz and Villalonga (2001) for large publicly traded US rms but consistent with the stable ownership structures of Continental European rms and in particular smaller unlisted companies that comprise the bulk of the rms in our sample.
13 Given the limited number of time periods for which data is available we focus on cross-section rather than panel model estimates. This ensures sufcient dynamic conditioning of the rm performance and leverage equations. In addition, it would have been difcult to apply quantile regression methods to panel data as quantiles of convolutions of random variables are highly intractable objects (see Koenker and Hallock, 2001). For completeness, we present panel estimates for the rm performance model.
628
Table 2 Distribution of efciency, protability and leverage by ownership type and ownership concentration. The table reports the 25th (p25), 50th (p50) and 75th (p75) percentiles of the efciency, prot to assets ratio (protability) and debt to assets ratio (leverage) distributions by ownership type (family vs. non-family) and by ownership concentration (less than 25% vs. greater than 25%). Also reported are the means and standard deviations by ownership type and concentration and the F-statistics with p-values in brackets for testing the hypotheses of equal means across ownership types and concentration, respectively. p25 Panel A: chemicals Efciency Family Non-family Low (<25%) concentration High (>25%) concentration Protability Family Non-family Low (<25%) concentration High (>25%) concentration Leverage Family Non-family Low (<25%) concentration High (>25%) concentration Panel B: computers and R&D Efciency Family Non-family Low (<25%) concentration High (>25%) concentration Protability Family Non-family Low (<25%) concentration High (>25%) concentration Leverage Family Non-family Low (<25%) concentration High (>25%) concentration Panel C: textiles Efciency Family Non-family Low (<25%) concentration High (>25%) concentration Protability Family Non-family Low (<25%) concentration High (>25%) concentration Leverage Family Non-family Low (<25%) concentration High (>25%) concentration
* **
p50
p75
Mean
Std. Dev.
F-statistics
Obs.
0.952 0.603 0.934 0.720 0.012 0.002 0.012 0.004 0.387 0.412 0.370 0.402
0.984 0.850 0.975 0.935 0.072 0.052 0.063 0.056 0.560 0.565 0.539 0.565
0.994 0.963 0.990 0.985 0.137 0.127 0.142 0.129 0.745 0.736 0.655 0.736
0.949 0.751 0.927 0.815 0.083 0.057 0.079 0.065 0.571 0.579 0.541 0.579
0.097 0.257 0.257 0.237 0.124 0.132 0.106 0.131 0.229 0.266 0.259 0.258
227.74* (0.000) 10.05* (0.002) 9.900** (0.017) 0.43 (0.514) 0.19 (0.666) 0.95 (0.330)
412 776 46 1142 412 776 46 1142 412 776 46 1142
0.892 0.726 0.790 0.842 0.020 0.018 0.047 0.008 0.484 0.515 0.428 0.500
0.946 0.880 0.886 0.929 0.083 0.048 0.034 0.073 0.629 0.681 0.608 0.655
0.978 0.953 0.953 0.971 0.173 0.135 0.110 0.156 0.768 0.843 0.751 0.814
0.915 0.803 0.840 0.868 0.092 0.019 0.004 0.064 0.648 0.739 0.596 0.697
0.098 0.209 0.161 0.167 0.184 0.267 0.231 0.227 0.312 0.521 0.276 0.430
405.59* (0.000) 6.65* (0.010) 75.30* (0.000) 14.95* (0.000) 36.98* (0.000) 13.87* (0.000)
1806 1447 264 2989 1806 1447 264 2989 1806 1447 264 2989
0.825 0.531 0.734 0.705 0.014 0.002 0.007 0.008 0.351 0.380 0.354 0.364
0.919 0.745 0.892 0.875 0.064 0.045 0.040 0.056 0.533 0.546 0.556 0.536
0.967 0.898 0.952 0.954 0.132 0.113 0.121 0.123 0.690 0.726 0.698 0.702
0.870 0.688 0.817 0.799 0.078 0.049 0.061 0.067 0.542 0.580 0.548 0.557
0.139 0.246 0.186 0.208 0.210 0.129 0.145 0.185 0.246 0.347 0.267 0.302
379.23* (0.000) 0.49 (0.483) 9.45* (0.002) 0.07 (0.796) 6.63* (0.010) 0.06 (0.810)
1046 659 66 1639 1046 659 66 1639 1046 659 66 1639
Indicates signicance at the 1% level. Signicant at the 5% level.
effect on efciency. This effect is positive at the mean of leverage for each industry and it remains positive over the entire relevant range of leverage values. Thus we nd support for the agency cost hypothesis (H1) that higher leverage is associated with improved rm performance. Based on the magnitude of estimated coefcients, we observe that the effect of debt on efciency appears to be stronger for rms in the traditional (chemicals and textiles) industries. This nding corroborates the conjecture of McConnell and Servaes (1995), namely that debt has a fundamentally different role on performance between rms with few and those with many growth opportunities (see footnote 5 above). The nding that debt is more important for rm performance for industries
with less growth opportunities is also consistent with the theoretical predictions of Jensen (1986) and Stulz (1990) (see also Booth et al., 2001). Protability has a positive and signicant effect on efciency for all industries. The effect of size is mostly signicant in chemicals and is non-monotonic, i.e. positive for smaller rms but negative for larger rms. Similarly, asset tangibility has a non-monotonic effect; negative at low xed tangibles to total assets ratios and positive at high tangibles to asset ratios. An explanation for this nding is that a high proportion of hard tangible assets would reduce the extent of the rms growth opportunities and as a result diminish the agency costs of managerial discretion (see Booth et al., 2001).
629
Table 3 The rm performance model. Panel A reports least squares estimates with heteroskedasticity-consistent standard errors (SEs). The dependent variable is the rms efciency score in 2005 computed as 1/(1 + distance function value). LEV_AVG(1) is the debt to assets average ratio for 2003 and 2004; PR_AVG(1) is the prot to assets average ratio for 2003 and 2004; LOGR(1) is log sales revenue in 2004; TANG(1) is the tangibles to assets ratio in 2004; INTG(1) is the intangibles to equity ratio in 2004; GROWTH_AVG(1) is average sales growth for 2003 and 2004; Owner1 = OWN1 * OWNC, Owner2 = OWN2 * OWNC, and Ownwr3 = OWN3 * OWNC capture the piecewise linear effect of ownership structure where OWN1 is a dummy variable that denotes low <25% ownership concentration; OWN2 denotes intermediate >25% but <50% ownership concentration; OWN3 denotes high >50% ownership concentration; and OWNC is the percentage of largest shareholder equity ownership; family and nancial are ownership type dummy variables. Panel B reports random-effects GLS panel regression estimates with Huber-White robust standard errors (SEs) for the period 20022005. The dependent variable is the rms efciency score in 2005 computed as 1/(1 + distance function value). LEV_AVG(1) is a two-year average debt to assets ratio with one period lag; PR_AVG(1) is a two-year average prot to assets ratio with one period lag; LOGR(1) is log sales revenue with one period lag; TANG(1) is the tangibles to assets ratio with one period lag; INTG(1) is the intangibles to equity ratio with one period lag; GROWTH_AVG(1) is a two-year average sales growth rate with one period lag. Dependent variable: efciency (EFF) Variable Chemicals Coefcient Panel A: cross-section estimates LEV_AVG(1) LEV_AVG(1)_Square PR_AVG(1) LOGR(1) LOGR(1)_Square TANG(1) TANG(1)_Square INTG(1) Growth_AVG(1) Owner1 Owner2 Owner3 Family Financial Constant R-squared F-statistic Obs. 0.0716* 0.0002 0.0910** 0.1469* 0.0128* 0.9508* 0.8516* 0.0015* 0.0089 0.0009** 0.0004*** 0.0004* 0.0148** 0.0190 0.6393* 0.792 261.9* 1188 SE 0.0180 0.0002 0.0386 0.0387 0.0021 0.0689 0.1037 0.0004 0.0116 0.0004 0.0002 0.0001 0.0076 0.0164 0.1669 Computers and R&D Coefcient 0.0423* 0.0036* 0.0345* 0.0140 0.0071** 1.5487* 2.2679* 0.0004 0.0018 0.0007** 0.00003 0.00004 0.0099** 0.0008 1.2221* 0.653 252.0* 3253 0.0330* 0.0026* 0.0216* 0.0208 0.0072* 1.5829* 2.2904* 0.0001 0.0179* 0.0004*** 0.0001 0.00001 0.0079** 0.0011 1.1822* 0.694 2655.6* 5530 SE 0.0100 0.0012 0.0112 0.0462 0.0029 0.1298 0.3914 0.0004 0.0093 0.0003 0.0002 0.0001 0.0049 0.0101 0.1815 Textiles Coefcient 0.0769* 0.0175* 0.1101** 0.0654 0.0109** 1.4275* 1.3937* 0.0003 0.0022 0.0002 0.0001 0.0001 0.0261* 0.0087 1.0706* 0.738 226.9* 1705 0.1144* 0.0488* 0.0637** 0.0443 0.0091** 0.3271* 0.1148* 0.0003 0.0499* 0.0002 0.0001 0.0002 0.0438* 0.0039 1.0280* 0.688 2210.19* 3069 SE 0.0226 0.0048 0.0478 0.0795 0.0048 0.0654 0.1160 0.0007 0.0145 0.0003 0.0002 0.0001 0.0072 0.0137 0.3288
Panel B: panel estimates 20022005 LEV_AVG(1) 0.0898* LEV_AVG(1)_Square 0.0203** PR_AVG(1) 0.0443** LOGR(1) 0.1578* LOGR(1)_Square 0.0131* TANG(1) 0.8150* TANG(1)_Square 0.6847* INTG(1) 0.0003 Growth_AVG(1) 0.0159* Owner1 0.0008* Owner2 0.0005** Owner3 0.0003* Family 0.0177** Financial 0.0256*** Constant 0.5487* R-squared overall 0.802 Wald Chi-square (14) 2573.48* Obs. 2138
* ** ***
0.0218 0.0096 0.0199 0.0287 0.0015 0.0565 0.0799 0.0002 0.0060 0.0003 0.0002 0.0001 0.0088 0.0142 0.1287
0.0069 0.0090 0.0063 0.0420 0.0026 0.1051 0.3157 0.0002 0.0045 0.0002 0.0001 0.0001 0.0037 0.0082 0.1663
0.0192 0.0126 0.0284 0.0618 0.0038 0.0206 0.0124 0.0004 0.0088 0.0003 0.0002 0.0001 0.0081 0.0135 0.2488
Signicant at the 1% level. Signicant at the 5% level. Signicant at the 10% level.
The effect of intangibles is negative and signicant only for rms in the chemicals industry while the effect of growth is insignicant across all industries in the cross-section regressions. Turning now to the ownership results we nd that the contrasting effects of convergence of interest (H3) and ownership entrenchment (H3a) vary across industries. Ownership concentration has a positive and signicant effect on rm performance across different concentration ratios in the chemicals industry but its strength diminishes at higher concentration segments suggesting the presence of entrenchment effects (note that differences in estimated coefcients are only statistically signicant between low and high concentration with an F-statistic value of 2.96). On the other hand for rms in the computers industry we nd that low ownership concentration has a negative effect on rm performance. Higher ownership concentration has an insignicant effect in this industry
suggesting the presence of offsetting entrenchment and incentive alignment effects.14 For rms in the textiles industry we nd no evidence that ownership concentration has a signicant effect on performance across the three concentration segments. The absence of a statistically signicant relationship between ownership concentration and efciency for textiles and in part for rms in computers and R&D appears to support the views expressed by Demsetz (1983). However we nd that ownership type has a signicant effect on rm performance. The results of Table 3 show quite consistently that family rms perform better on average in comparison to non-family rms.
14 This nding provides partial support for the conjecture of McConnell and Servaes (1995), namely that the effect of ownership on performance should be more important for low-growth rather than high-growth rms (see also footnote 6 above).
630
Table 4 The leverage (LEV) model. The table reports least squares (OLS) estimates with heteroskedasticity-consistent standard errors and simultaneous quantile regression estimates at leverage quantiles .20 (q20), .50 (q50) and .80 (q80) with bootstrap standard errors (SEs). The dependent variable (LEV) is the debt to assets ratio in 2005. Efciency(1) is the rms efciency score in 2004 computed as 1/(1 + distance function value); PR_avg(1) is the prot to assets average ratio for 2003 and 2004; LogR(1) is log sales revenue in 2004; tang(1) is the tangibles to equity ratio in 2004; intg(1) is the intangibles to assets ratio in 2004; GR_avg(1) is average sales growth for 2003 and 2004; Owner1 = OWN1 * OWNC, Owner2 = OWN2 * OWNC, and Ownwr3 = OWN3 * OWNC capture the piecewise linear effect of ownership structure where OWN1 is a dummy variable that denotes low <25% ownership concentration; OWN2 denotes intermediate >25% but <50% ownership concentration; OWN3 denotes high >50% ownership concentration; and OWNC is the percentage of largest shareholder equity ownership; family and nancial are ownership type dummy variables. Coefcient OLS Panel A: chemicals OLS and quantile regression estimates 0.0569 Efciency(1) 0.1840* PR_avg(1) 0.848* 0.0847 ** 0.0084 LogR(1) 0.0169 * 0.0027 tang(1) 0.0071 0.0659 intg(1) 0.187* 0.0289 GR_avg(1) 0.0870* owner1 0.0004 0.0012 owner2 0.0003 0.0006 ** 0.0003 owner3 0.0007 Family 0.0060 0.0225 Financial 0.0054 0.0248 0.1243 Const. 0.2637** R-squared 0.23 SE Coefcient q20 0.1675** 0.593* 0.0255** 0.0067 0.181** 0.1597* 0.0003 0.0008 0.0008*** 0.0065 0.0149 0.0115 0.09 0.0746 0.1035 0.0114 0.0178 0.0845 0.0521 0.0022 0.0009 0.0005 0.0394 0.0418 0.1753 SE Coefcient q50 0.1833* 0.7727* 0.0179 0.0143 0.1867** 0.0753** 0.0012 0.0004 0.0008*** 0.0164 0.0302 0.2287 0.14 0.2088* 0.6829* 0.0200* 0.0189 0.0586 0.0746* 0.0018** 0.0004 0.0008* 0.0237 0.0415** 0.3195* 0.11 0.2506* 0.546* 0.0315* 0.0659* 0.2280** 0.1094*** 0.0007 0.0009 0.0007* 0.0127 0.0398 0.0327 0.11 0.0738 0.0893 0.0130 0.0166 0.0947 0.0313 0.0013 0.0007 0.0004 0.0316 0.0286 0.1830 SE Coefcient q80 0.2518* 0.866* 0.0212 0.0189** 0.231*** 0.0491 0.0008 0.0006 0.0009* 0.0057 0.0064 0.3113 0.15 0.1679* 0.986* 0.0027 0.0059 0.0767*** 0.0351 0.0019*** 0.0010* 0.0011* 0.042** 0.0270 0.6488* 0.17 0.3018* 0.773* 0.0284 0.0513* 0.0946 0.0334 0.0018 0.0008 0.0008** 0.0293 0.0255 0.2103 0.16 0.0879 0.1061 0.0135 0.0093 0.1219 0.0428 0.0023 0.0007 0.0004 0.0289 0.0348 0.2031 SE
Panel B: computers and R&D OLS and quantile regression estimates 0.0793 0.1527*** Efciency(1) 0.2826* 0.1581 0.312* PR_avg(1) 0.897* LogR(1) 0.0076 0.0105 0.0148 tang(1) 0.0060 0.0043 0.0184 intg(1) 0.0607 0.0938 0.0010 GR_avg(1) 0.0436 0.0494 0.1326* owner1 0.0002 0.0011 0.0014 0.0005 0.0002 owner2 0.0016* 0.0004 0.0008** owner3 0.0025* Family 0.0154 0.0193 0.0025 Financial 0.0060 0.0581 0.0330 ** 0.1358 0.2022 Const. 0.3081 R-squared 0.22 0.05 Panel C: textiles OLS and quantile regression Efciency(1) 0.1769* PR_avg(1) 0.868 * LogR(1) 0.0225** tang(1) 0.0350* intg(1) 0.1812** GR_avg(1) 0.0366 owner1 0.0001 owner2 0.0008 owner3 0.0006*** Family 0.0296 Financial 0.0440 Const. 0.2203*** R-squared 0.21
* ** ***
0.0803 0.0767 0.0096 0.0232 0.0771 0.0302 0.0010 0.0009 0.0004 0.0212 0.0358 0.1251
0.0692 0.0690 0.0079 0.0196 0.0575 0.0230 0.0008 0.0005 0.0002 0.0168 0.0110 0.1252
0.0414 0.0758 0.0076 0.0080 0.0451 0.0278 0.0011 0.0003 0.0002 0.0176 0.0347 0.1105
estimates 0.0675 0.1270 0.0092 0.0074 0.0908 0.0443 0.0009 0.0006 0.0003 0.0194 0.0286 0.1229
0.0799 0.310* 0.0242 0.0393** 0.1590 0.1082*** 0.0003 0.0007 0.0003 0.0081 0.0416 0.0541 0.06
0.0883 0.0849 0.0154 0.0185 0.1603 0.0593 0.0009 0.0008 0.0005 0.0286 0.0299 0.1816
0.0614 0.1169 0.0124 0.0190 0.1129 0.0584 0.0015 0.0007 0.0004 0.0233 0.0265 0.1538
0.0850 0.1329 0.0123 0.0142 0.1460 0.0407 0.0015 0.0006 0.0003 0.0248 0.0286 0.1579
Signicant at the 1% level. Signicant at the 5% level. Signicant at the 10% level.
Table 4 reports the estimates of the leverage model. We report OLS results and estimates for three leverage quantiles (0.20, 0.50 and 0.80); estimates for other quantiles are available from the authors. The results from the OLS and quantile regressions show that the effect of efciency on leverage is positive and signicant in the low to high range of the leverage distribution supporting the efciency-risk hypothesis (H2): more efcient rms with relatively low levels of debt tend to choose higher debt ratios because higher efciency lowers the expected costs of bankruptcy and nancial distress. However there is no evidence to suggest that the franchise-value effect (H2a) outweighs the efciency-risk effect even for the most highly levered rms. Consistent with pecking order theory, protability has a negative effect on leverage for all industries on average and also across different capital structures. The effect of protability appears to be
stronger for rms with higher debt. For example, the F-statistic for the null hypothesis of no difference in protability coefcients in the chemicals industry is 4.74 at leverage quantiles (.20 and .80) which is signicant at the 5% level. We also nd that a higher proportion of tangible assets are more important in increasing debt capacity for the smaller typically riskier rms in the textiles industry. The effect of intangible assets is negative in chemicals, positive in textiles and generally not signicant for rms in computers and R&D. Growth has a positive effect on leverage on average (OLS estimates) for rms in chemicals as well as for low to medium leveraged rms across all industries. The owners of these rms appear to opt for debt nance for reasons we have discussed earlier on. We nd that in general rms with more concentrated ownership carry more debt in their capital structure. For mid- to highleveraged rms in the computers and R&D industry we nd that
631
low ownership concentration has a negative effect on leverage. Finally, our results show that the effect of ownership type on leverage is generally non-signicant across all industries with the exception of some rms in computers and R&D. In particular, we nd that among the highly levered rms in this industry, family rms opt to carry less debt in their capital structure. 6. Conclusion This paper investigates the relationship between efciency, leverage and ownership structure. This analysis is conducted using directional distance functions to model the technology and obtain X-inefciency measures as the distance from the efcient frontier. We interpret these measures as a proxy for the (inverse) agency costs arising from conicts between debt holders and equity holders or from different principal-agent objectives. Using a sample of French rms from low- and high-growth industries, we consider both the effect of leverage on rm performance as well as the reverse causality relationship while controlling for the effects of ownership structure and ownership type. We nd support for the core prediction of the Jensen and Meckling (1976) agency cost hypothesis in that higher leverage is associated with improved efciency over the entire range of observed data. The contrasting alignment and entrenchment agency effects of ownership concentration vary across industries and across concentration ratios. We nd that while more dispersed rms face higher agency costs in computers and R&D; the reverse is true for rms in the chemicals industry. Moreover, we nd that family rms outperform non-family rms. We have also investigated the reverse causality relationship from efciency to leverage in terms of two competing hypotheses: the efciency-risk hypothesis and the franchise-value hypothesis. Using quantile regression analysis we show that the effect of efciency on leverage is positive in the low to high ranges of the leverage distribution supporting the efciency-risk hypothesis. We also nd that more concentrated ownership is generally associated with more debt in the capital structure. However, we generally nd no evidence that ownership type has an effect on leverage choices. Our methodology has gone some way in reconciling some of the empirical irregularities reported in prior studies. In particular, we have shown how competing hypotheses may dominate each other at different segments of the relevant data distribution thereby cautioning the standard practice of drawing inferences on capital structure choices using conditional mean (OLS) estimates. By using productive efciency as opposed to nancial performance indicators as our measure of (inverse) agency costs we have been able to carry out tests of the agency theory without the confounding problems that may be associated with the more traditional nancial measures of rm performance. In future research it will be of interest to extend this analysis across different countries and across different industries as well as examine further aspects of ownership and governance characteristics. Acknowledgements We are grateful to an anonymous referee and the Editor (Ike Mathur) for helpful comments and suggestions. We also thank the participants at the 2008 EFMA Conference in Athens and at a Massey University seminar for helpful comments and Chris Yung for research assistance. Appendix A This section outlines the specication of the efciency measures. Following Fre and Grosskopf (2004) and Fre et al.
(2008) we assume that rms employ N inputs denoted by x x1 ; . . . ; xN 2 RN to produce M outputs denoted by y y1 ; . . . ; yM 2 RM .15 Technology may be characterised by a technol ogy set T, which is the set of all feasible inputoutput combinations, i.e.
T fx; y : x can produce yg;
x 2 RN :
A:1
The technology set is assumed to satisfy a set of reasonable axioms. Here we assume that T is a closed, convex, nonempty set with inputs and outputs which are either freely or weakly disposable.16 To provide a measure of efciency we use a directional technology distance function approach. This function completely characterizes technology (i.e., it is equivalent to T), it is dual to the prot function and allows for adjustment of inputs and outputs simultaneously. Thus the directional distance function entails an extremely exible description of technology without restricting rms to optimize by either increasing outputs proportionately without changing inputs or by decreasing inputs proportionally for given outputs. To dene it we need to specify a directional vector, denoted by g = (gx, gy) where g x 2 RN and g y 2 RM . This vector determines the direction in which technical efciency is assessed, i.e. the path of the projection of the observed data to the frontier of technology. The directional technology distance function is dened as:
~T x; y; g ; g supfb : x bg ; y bg 2 Tg: D x y x y
A:2
The directional distance function expands outputs in the direction gy and contracts inputs simultaneously in the direction gx to the frontier T. If the observed input output bundle is technically efcient, the value of the directional distance function would be zero. If the observed input output bundle is interior to technology T, the distance function is greater than zero and the rm is technically inefcient. In this paper we choose g ; y which implies x x that output(s) may be increased by ~T x; y; ; y y and inputs deD x x creased by ~T x; y; ; y for a rm to eliminate all technical inefD ciency relative to its best performing peers. The directional distance function can be estimated non-parametrically using DEA a mathematical programming enveloping technique under a VRS (variable returns to scale) technology as follows:
~T x; y; g ; g max b D x y
subject to:
A:3
XK XK XK
kx k1 k kn ky k1 k km k k1 k
xkn bg x ; ! ykm bg y ;
n 1; . . . ; N; m 1; . . . ; M;
1;
kk ! 0; k 1; . . . ; K:
The intensity variables kk form combinations of inputs and outputs from the observed set of inputs and outputs of the rms in the sample, one for each activity or observation (k) of data. These are non-negative variables whose solution value may be interpreted as the extent to which an activity is involved in frontier production. Thus at each segment of the piecewise frontier DEA identies a peer group of best practice reference rms (i.e. those with non-zero kk ) for each rm being evaluated. Each rm (k) can produce no more output using no less input than a linear combination of all the rms inputs and outputs in the sample. Therefore the technology is constructed from the data xk ; yk by forming
15 In the empirical section of the paper technology is restricted to a specication that uses two inputs (capital and labor) to produce a single output. 16 Input weak disposability means that if all inputs increase proportionally then output will not decrease. Strong or free disposability on the other hand requires that output does not decrease if any or all feasible inputs are increased. Disposable outputs are similarly dened.
632
D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632 Grossman, S.J., Hart, O., 1982. Corporate nancial structure and managerial incentives. In: McCall, J. (Ed.), The Economics of Information and Uncertainty. University of Chicago Press, Chicago. Harris, M., Raviv, A., 1991. The theory of capital structure. Journal of Finance 46, 297355. Himmelberg, C., Hubbard, G., Palia, D., 1999. Understanding the determinants of managerial ownership and the link between ownership and performance. Journal of Financial Economics 53, 353384. Hu, Y., Zhou, X., 2008. The performance effect of managerial ownership: evidence from China. Journal of Banking and Finance 32, 20992110. Jensen, M., 1986. Agency costs of free cash ow, corporate nance, and takeovers. American Economic Review 76, 323329. Jensen, M., Meckling, W., 1976. Theory of the rm: Managerial behavior, agency costs and capital structure. Journal of Financial Economics 3, 305360. Jiraporn, P., Gleason, K.C., 2007. Capital structure, shareholding rights, and corporate governance. Journal of Financial Research 30 (1), 2133. Koenker, R., Hallock, K.F., 2001. Quantile regression: an introduction. Journal of Economic Perspectives 15, 143156. King, M.R., Santor, E., 2008. Family values: ownership structure, performance and capital structure of Canadian rms. Journal of Banking and Finance 32, 2423 2432. La Porta, R., Lopez de Silanes, F., Shleifer, A., Vishny, R.W., 1998. Law and nance. Journal of Political Economy 106, 11131155. La Porta, R., Lopez de Silanes, F., Shleifer, A., Vishny, R.W., 2002. Investor protection and corporate valuation. Journal of Finance 57, 11471170. Leibenstein, H., 1966. Allocative efciency vs. X-efciency. American Economic Review 56, 392415. Mahrt-Smith, J., 2005. The interaction of capital structure and ownership structure. Journal of Business 78, 787816. Margaritis, D., Psillaki, M., 2007. Capital structure and rm efciency. Journal of Business Finance and Accounting 34 (910), 14471469. Maury, B., 2006. Family ownership and rm performance: empirical evidence from Western European corporations. Journal of Corporate Finance 12, 321341. McConnell, J., Servaes, H., 1990. Additional evidence on equity ownership and corporate value. Journal of Financial Economics 27, 595612. McConnell, J., Servaes, H., 1995. Equity ownership and the two faces of debt. Journal of Financial Economics 39, 131157. Morck, R., Shleifer, A., Vishny, R.W., 1988. Management ownership and market valuation: an empirical analysis. Journal of Financial Economics 20, 293315. Myers, S., 1977. Determinants of corporate borrowing. Journal of Financial Economics 5, 147175. Myers, S., 1984. The capital structure puzzle. Journal of Finance 39 (3), 575592. Myers, S., 2001. Capital structure. Journal of Economic Perspectives 15 (2), 81102. Rajan, R.G., Zingales, L., 1995. What do we know about capital structure? Some evidence from international data. Journal of Finance 50, 14211460. Shephard, R., 1970. Theory of Cost and Production Functions. Princeton University Press, Princeton. Shleifer, A., Vishny, R.W., 1986. Large shareholders and corporate control. Journal of Political Economy 94, 461488. Short, H., Keasey, K., Duxbury, D., 2002. Capital structure, management ownership and large external shareholders: a UK analysis. International Journal of the Economics of Business 9 (3), 375399. Spaliara, M., 2009. Do nancial factors affect the capitallabour ratio? Evidence from UK rm-level data. Journal of Banking and Finance 33, 19321947. Stulz, R., 1990. Managerial discretion and optimal nancing policies. Journal of Financial Economics 26, 327. Titman, S., Wessels, R., 1988. The determinants of capital structure choice. Journal of Finance 43, 119. Villalonga, B., Amit, R., 2006. How do family ownership, control and management affect rm value? Journal of Financial Economics 80, 385417. Williamson, O.E., 1967. Hierarchical control and optimum rm size. Journal of Political Economy 75, 123138.
the tightest convex cone that includes all data hence the descriptive title data envelopment analysis. Constraining the intensity variables to add up to one imposes the VRS technology.
References
Anderson, R.C., Mansi, S.A., Reeb, D.M., 2003. Founding family ownership and the agency cost of debt. Journal of Financial Economics 68, 263285. Anderson, R.C., Reeb, D.M., 2003a. Founding family ownership and rm performance: Evidence from the S&P 500. Journal of Finance 58 (3), 13011328. Anderson, R.C., Reeb, D.M., 2003b. Founding family ownership, corporate diversication, and rm leverage. Journal of Law and Economics 46, 653684. Berger, A.N., Bonaccorsi di Patti, E., 2006. Capital structure and rm performance: a new approach to testing agency theory and an application to the banking industry. Journal of Banking and Finance 30, 10651102. Berle, A., Means, G., 1932. The Modern Corporation and Private Property. Harcourt, Brace and World, New York. Booth, L., Aivazian, V., Demirguc-Kunt, A., Maksimovic, V., 2001. Capital structure in developing countries. Journal of Finance 56, 87130. Brailsford, T.J., Oliver, B.R., Pua, S.L.H., 2002. On the relation between ownership structure and capital structure. Accounting and Finance 42, 126. Claessens, S., Djankov, S., Fan, J., Lang, L., 2002. Disentangling the incentive and entrenchment effects of large shareholdings. Journal of Finance 57 (6), 2741 2771. Cornett, M.M., Marcus, A.J., Saunders, A., Tehranian, H., 2007. The impact of institutional ownership on corporate operating performance. Journal of Banking and Finance 31, 17711794. Davies, D., Hillier, D., McColgan, P.M., 2005. Ownership structure, managerial behavior and corporate value. Journal of Corporate Finance 11, 645660. De Jong, A., Kabir, R., Nguyen, T.T., 2008. Capital structure around the world: the roles of rm- and country-specic determinants. Journal of Banking and Finance 32, 19541969. Demsetz, H., 1973. Industry structure, market rivalry, and public policy. Journal of Law and Economics 16 (1), 19. Demsetz, H., 1983. The structure of ownership and the theory of the rm. Journal of Law and Economics 26, 375390. Demsetz, H., Lehn, K., 1985. The structure of corporate ownership: causes and consequences. Journal of Political Economy 93 (6), 11551177. Demsetz, H., Villalonga, B., 2001. Ownership structure and corporate performance. Journal of Corporate Finance 7, 209233. Dow, S., McGuire, J., 2009. Propping and tunneling: empirical evidence from Japanese keiretsu. Journal of Banking and Finance 33, 18171828. Faccio, M., Lang, L., 2002. The ultimate ownership of Western European corporations. Journal of Financial Economics 65, 365395. Fama, E., Jensen, M., 1983. Separation of ownership and control. Journal of Law and Economics 26 (2), 301325. Fre, R., Grosskopf, S., 2004. New Directions: Efciency and Productivity. Kluwer Academic Publishers, Boston. Fre, R., Grosskopf, S., Margaritis, D., 2008. Efciency and productivity: Malmquist and more. In: Fried, H.O., Lovell, C.A.K., Schmidt, S.S. (Eds.), The Measurement of Productive Efciency and Productivity Growth. Oxford University Press, New York. Fried, H.O., Lovell, C.A.K., Schmidt, S.S., 2008. Efciency and productivity. In: Fried, H.O., Lovell, C.A.K., Schmidt, S.S. (Eds.), The Measurement of Productive Efciency and Productivity Growth. Oxford University Press, New York. Friend, I., Lang, L., 1988. An empirical test of the impact of managerial self-interest on corporate capital structure. Journal of Finance 43 (2), 271281. Giannetti, M., 2003. Do better institutions mitigate agency problems? Evidence from corporate nance choices. Journal of Financial and Quantitative Analysis 38 (1), 185212.

Capital Struture Equity Ownership and Firm Performence

Загружено:

Сведения о документе

Исходное описание:

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Capital Struture Equity Ownership and Firm Performence

Загружено:

Авторское право:

Доступные форматы

Journal of Banking & Finance 34 (2010) 621632

Contents lists available at ScienceDirect

Journal of Banking & Finance

Capital structure, equity ownership and rm performance

D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632

D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632

D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632

EFF i;t a0 a1 LEV i;tl a2 LEV 2 a3 Z 1i;tl ui;t ; i;tl

D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632

LEV i;t b0 b1 EFF i;tl b2 Z 2i;tl v i;t ;

D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632

D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632

D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632

412 776 46 1142 412 776 46 1142 412 776 46 1142

1046 659 66 1639 1046 659 66 1639 1046 659 66 1639

Indicates signicance at the 1% level. Signicant at the 5% level.

D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632

D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632

D. Margaritis, M. Psillaki / Journal of Banking & Finance 34 (2010) 621632

T fx; y : x can produce yg;

Вам также может понравиться