Ownership structure and rm performance: evidence from the UK nancial services industry

Transcription

1 Applied Financial Economics, 1998, 8, 175Ð 180 Ownership strctre and rm performance: evidence from the UK nancial services indstry RAM MUD AMB I* and C ARMELA NI COS IA *ISMA Centre, University of Reading, Reading, UK and School of Bsiness, University of Bckingham, Bckingham MK 18 1EG, UK Theory tells s that managerial ownership of shares in a rm generates two con icting e ects on management behavior, i.e. the convergence e ect whereby increased managerial ownership can improve corporate performance, and the entrenchment e ect which conters it. A nmber of stdies have soght to evalate these e ects empirically. The reslts in the literatre are not niformly in agreement. In this paper, we distingish between measres of ownership and measres of control implied by this ownership. Frthermore, we provide evidence spporting the entrenchment and convergence e ects sing UK data. I. INTRO DU CTIO N Economists have been interested in the e ects of the separation of ownership and control in the modern corporation at least since the classic works of Berle and Means (193) and Coase (1937). This interest contines, as evidenced by major stdies in the last decade (Cosh and Hghes, 1987; Jensen and Mrphy, 1990). The major focs of concern has been the potential con icts of interest between managers (who rn corporations) and shareholders (who own them). One method sggested to redce this potential con ict is to increase the identity between the two grops, typically throgh indcing managers to own shares in the company. Modern work on this isse has shown that managerial ownership of shares in a rm generates two con icting forces on management behavior, namely the convergence of interest e ect and the entrenchment e ect (Jensen and Meckling, 1976; Fama and Jensen, 1983; Hart, 1983; Jensen and Rback, 1983). Under the former, the relationship between managerial stake and market vale of the rm is positive, as management and shareholder interests converge. Under the latter, this relationship is negative, as a larger managerial stake entrenches and inslates management from the market for corporate control. A nmber of stdies have soght to evalate empirically the link between managerial share ownership and rm performance. However, the reslts are not niformly in agreement. Demsetz (1983) arges that there shold be no relationship between ownership strctre and rm performance. Prsing this argment empirically, Demsetz and Lehn (1985) nd no signi cant correlation between pro t rates and varios measres of ownership concentration in a sample of 511 US companies sing 1980 data. Tsetsekos and DeFsco (1990) constrct portfolios arranged according to managerial share ownership and report no signi cant di erences in the retrns on the varios portfolios. In contrast, Mehran (1995) provides evidence of a positive relation between managerial eqity ownership and rm performance. Similarly, Wrck (1988) nds a strong and, on average, positive link between the change in ownership concentration and rm performance. Several other stdies nd evidence for a non-monotonic relationship between ownership concentration and the market vale of the rm. Using a sample of 371 Fortne 500 rms, Mork et al. (1988) nd that the relationship is positive for managerial ownership between 0 and 5%, negative between 5 and 5% and positive thereafter. Hermalin and Weisback (1991) stdy 134 NYSE rms and se CEO stock ownership to represent managerial share holding. They nd the relationship to be positive for CEO stock ownership between 0 and 1%, negative between 1 and 5%, positive between 5 and 0% and negative thereafter. McConnell and Servaes (1990), analysing 1976 data for a sample of 1173 rms and 1986 data for a sample of 1093 rms, nd the relationship to be positive between 0 and somewhere in the range 35Ð 50% and negative thereafter. Similar 0960Ð 3107 Ó 1998 Rotledge 175

2 176 R. Mdambi and C. Nicosia ndings are reported by Jarrell and Polsen (1988) and Stlz (1988). Most of the work in the literatre assmes that managerial ownership of shares is representative of managerial control. It is ths sed as an explanatory factor in rm performance withot considering the ownership pattern of the rest of the rm s shares. This means that the concentration of managerial share ownership is considered, bt the degree of control that this ownership confers is not. Or rst objective in this paper is to assess whether the concentration of share ownership and the degree of control exerted by the large share holding grops go hand in hand in terms of their e ects on company performance. Using a measre developed by Cbbin and Leech (1983), we nd that the e ects of concentration and control on rm performance are radically di erent. Ths, we sggest that it is managerial control, rather than the level of managerial share holding that is the relevant explanatory variable for assessing rm performance. Or second objective is to attempt to estimate ranges of managerial share ownership over which the entrenchment and convergence e ects are predominant, sing UK data. Mch of the research ths far has sed US data and it is of interest to examine whether data from other contries spport similar reslts. We set p and test a non-monotonic relationship and nd it to be well-spported by the data. Frther, the estimated ranges for the entrenchment and convergence e ects are very similar to those estimated in some US stdies. In Section II we present or methodology. In Section III we describe or data and estimation, and discss or reslts. We o er some conclding remarks in Section IV. II. METH O DOL OGY We se data for the UK nancial services indstry. The nancial services indstry was chosen for a nmber of reasons. The growing importance of nancial services in the economic activities of advanced contries and the recent developments in nancial markets, sch as dereglation, credit expansion and the introdction of new nancial instrments have pt a sharp focs on the indstry. In addition, problems relating to information asymmetry and investor protection are of particlar relevance for nancial services. This is becase nancial instittions are particlarly ssceptible to information problems and externalities. Measres of the rm s performance are the dependent variables in or stdy. We measre the rm s performance by its rate of retrn on the stock market. We se both the actal rate of retrn as well as the abnormal rate of retrn. We provide precise de nitions of the variables sed in Section II below. In approaching the rst of or objectives, we mst se measres of share ownership concentration and measres of control implied by the pattern of share holding as independent variables. It is well established that of all the available measres of concentration, the Her ndahl index (H) is particlarly strong in terms of satisfying the reqirements of a good index (Hannah and Kay, 1977; Schmalensee, 1977; Encaoa and Jacqemin, 1980). Ths, we se the `H index as or measre of ownership concentration. In order to measre the `degree of shareholder control, we se the CbbinÐ Leech index (a ). This measre depends positively on the proportion of shares held by a particlar grop of investors and inversely on the degree of concentration of the remaining shares. This expresses the notion that control depends not only on the nmber of shares owned, bt on the dispersal of the nowned shares. We address the rst qestion by contrasting the estimates based on the `H index with those based on the à index. The dependent variable is the rm s nancial rate of retrn, which can be a ected by the riskiness of its investment strategy and by the sensitivity of its share price to movements in the overall share market. Hence, we normalize the estimation by incorporating rm-speci c nancial measres. In particlar, we introdce the coe cient of systematic risk (beta) and the coe cient of speci c risk as additional regressors. We se the percentage of directors eqity ownership as or proxy for managerial share ownership. We recognize the possibility that the relationship between this variable and the rm s performance may be non-monotonic. We introdce this possibility by setting p a standard cbic for estimation, allowing the data to spport or reject the hypothesis of non-monotonicity. We note that there is a sbstantial literatre docmenting the relationship between rm size and nancial retrns. Roll (1981), Banz (1981) and Dimson and Marsh (1989) are jst a few examples of work estimating this relationship. We therefore normalize for rm size in or estimation procedre. III. DAT A, ESTIMATIO N AN D RES ULTS Data The primary sorce of data is Datastream, spplemented by the London Bsiness School Risk Measrement Service. Data relate to the UK nancial services sector. The data set comprised of a total of 111 rms and inclded banks, merchant banks, insrance companies and insrance brokers. The data relate to the period 199Ð 94. Smmary statistics describing the data are presented in Table 1. The actal rate of retrn is denoted by ACROR and is the percentage capital appreciation pls the dividend yield over the year, where gross dividends are assmed to be

3 Ownership strctre and rm performance 177 Table 1. Smmary statistics Variable Mean Std. Dev. Maximm Minimm - ACROR H ALPHA BETA SPRISK MCAP DIROWN re-invested in the rm s shares at the end of the month in which they are paid. The abnormal rate of retrn is denoted by ABROR and is the performance of the rm s shares over the year, relative to the market as a whole. It is eqal to the di erence between the actal retrn on the share and percentage retrn available over the same period from an investment in a diversi ed portfolio with the same risk characteristics, as measred by beta. The Her ndahl index is well known and de ned as H = + j (m j ), where m j is the percentage holding of the jth largest share holding grop. We compte the CbbinÐ Leech index associated with the largest share holding grop. 1 This is de ned to be a 1 = F[(m 1 )/(H - m 1 ) 1 / ], where F[.] is the standard normal distribtion fnction and m 1 is the percentage holding of the largest share holding grop. We denote this vale by ALPHA. The rm s systematic risk is measred by BETA. This is the standard measre derived from the Capital Asset Pricing Model (CAPM). The risk of non-market related ctations in share price is measred by the rm s coe cient of speci c risk. This is denoted by SRISK. The measre of directors eqity is eqal to the sm of the percentage stake owned by bene ciary and non-bene ciary directors of each company. This variable is denoted by DIROWN. Given that or sample consists of rms in the nancial services sector, the interpretation of measres of trnover is not straightforward. We therefore measre the size of the rm by sing its capital base. We se the rm s market capitalization, denoted by MCAP, as or measre of size. Fig. 1. OL S residals and rm size Estimation and reslts In order to address or two qestions, we set p the following eqation for estimation ACROR = f [DIROWN, (DIROWN), (DIROWN) 3, H, ALPHA, MCAP, BETA, SRISK] + 1 (1) We also estimate Eqation 1 sing ABROR as the dependent variable. However, the reslts are all in extremely close agreement with those for ACROR. Hence, in the interests of brevity, only the reslts relating to ACROR are presented. (Fll reslts are available from the athors.) We initially estimate the eqations sing ordinary least sqares (OLS). However, examining the residal plots, we nd considerable evidence of heteroscedasticity relating to rm size (Fig. 1). This is con rmed by noting that the OLS estimates conclsively fail the BreschÐ Pagan test (Table ). 1 Ideally, we wold like to know whether the largest share-holding grop consists of directors of the company or not. Unfortnately, this information is not available in the data set. This methodology is open to the criticism that rms may di er in their technology, i.e. rms may se di ering levels of capital intensity, so that two rms of the same size may have di erent levels of market capitalization. However, we nd that when we compare market capitalization with another inpt-based measre of rm size, namely employment, the two are in close agreement. This sggests that the technology, as represented by capitalð labor intensity, is roghly the same for all rms in the indstry. Conseqently, market capitalization is a good measre of rm size.

4 x 178 R. Mdambi and C. Nicosia Table. L east sqares estimates of the rate of retrn sing ownership concentration and shareholder control (Her ndahl Index and CbbinÐ L eech Index) Regressor OLS Estimates WLS Estimates Constant (1.38) (1.78) (1.94)* ALPHA 1.15 (1.93)* (.01)* (.03)* H (1.5) (1.39) (3.70) * BETA (.50)* (3.07) * (4.51)* SPRISK (.86) * (1.83)* (0.01) MCAP (0.3) (.47)* (3.14)* DIROWN.9934 (.34) * (.0)* (.48)* (DIROWN) (.00)* (.69)* (.7) * (DIROWN) (.83)* (.45)* (.5) * R Diagnostics Adj `F Stat; (d.f ) 8.44; (8,89) 17.39; (8,80) Joint exclsion restriction on ; (3,86) ; (3,8) DIROWN coe cients F Stat; (d.f.) Log-likelihood BreschÐ Pagan Test: (7); (p vale) ; (0.000) 5.446; (0.374) Akaike IC Dependent variable: ACROR Notes: (1) t-statistics in brackets. Those calclated sing White s heteroscedasticity-consistent varianceð covariance matrix are shown in italics. () Estimates signi cant at the 5% level are marked with a *. We adopt three alternative means of correcting for this problem. First, we adjst the OLS standard errors, sing the White heteroscedasticity-consisten t varianceð covariance matrix. Second, we generate weighted least sqares (WLS) estimates sing rm size measred by employment as the weighting variable. It may be seen in Table that sing WLS ensres that the BreschÐ Pagan test is passed and that the overall t of the eqation improves. Third, we estimate the model with assmed mltiplicative heteroscedasticity sing the method of maximm likelihood. The t of these estimates is also an improvement on the OLS estimates (Table 3). OLS and WLS estimates of Eqation 1 are presented in Table, while maximm likelihood estimates are presented in Table 3. In all cases, the coe cient of the H index appears with a negative sign. In the case of the WLS estimates, this negative e ect is highly signi cant. In the case of the maximm likelihood estimates, the t-statistic is qite near the critical vale. By contrast, in all cases, the coe cient of ALPHA appears with a positive sign. Frther, this positive coe cient is statistically signi cant, with the lowest signi cance level observed in the case of the OLS estimates. Ths, we nd that we are able to spport the nding of Leech and Leahy (1991), that share ownership concentration does `not necessarily ¼ (have) ¼ any implications for the degree of control. Or estimates of the e ects of ownership concentration and the degree of control are diametrically opposed. We now proceed to examine or second qestion. We have allowed for the possibility of a non-monotonic relationship between managerial share ownership and performance by estimating an eqation that is a standard cbic in DIROWN, i.e. it is of the form ACROR = 1 (DIROWN) 3 + (DIROWN) + 3 (DIROWN) + Q (.) () where Q (.) is the linear fnction containing all the remaining regressors and the constant term. Eqation will exhibit trning points if its derivative has real roots. These roots are the trning points of the cbic. Frther, the trning points will be a local maximm followed by a local minimm if the derivative has a minimm which is negative. The derivative of Eqation is ACROR/ DIROWN = 3 1 (DIROWN) + (DIROWN) + 3 (3) Examining the estimates of 1, and 3 in Tables and 3, we nd that they are consistent in their retention of sign, and the t-tests indicate that they are always individally

5 Ownership strctre and rm performance 179 Table 3. Maximm likelihood estimates of the rate of retrn sing ownership concentration and shareholder control (Her ndahl Index and CbbinÐ L eech Index) Regressor Linear model with mltiplicative heteroscedasticity Constant (1.91) H (1.76) ALPHA (.80)* BETA (3.51)* SPRISK (.56)* MCAP (1.36)* DIROWN (3.5)* (DIROWN) (4.38)* (DIROWN) (4.93)* Estimates of the variance process (S ) (6.64)* SIZE (5.01)* Diagnostics Log-likelihood Restricted log-likelihood LR Test x (1); (p vale) 4.989; (0.015) Joint exclsion restriction on DIROWN coe cients LR test x (3); (p vale) ; (0.0000) Iterations 8 Dependent variable: ACROR Notes: (1) The estimated model sets the conditional variance as a linear fnction of rm size. OLS estimates are sed as starting vales. () Estimates signi cant at the 5% level are marked with a *. signi cant (even in the case of the sspect OLS estimates). Frther, the joint test of inclsion for all three polynomial DIROWN terms is easily passed in all cases. An `F test is sed for the OLS and WLS estimates and a likelihood ratio test is employed in the case of the maximm likelihood estimates. All these estimates are collected in Table 4. The soltions for the roots of Eqation 3 and its minimm vale are also presented here. 3 The minimm vale is always negative as reqired. However, we need to test whether these negative vales are statistically signi cant. This is done by noting that a s cient condition for the minimm to be negative is > 0 (4) This is a non-linear restriction on the parameters. We test this restriction sing a non-linear Wald test, which generates a x statistic with one degree of freedom. The reslts of this test are presented in Table 4 and indicate that the restriction is always statistically signi cant. Table 4. Estimation of trning points in the marginal e ect of managerial shareholding on the rate of retrn Using the CbbinÐ Leech Index (a ) Estimating method OLS WLS Max.L x Negative root Positive root Derivative minimm Wald Test (1); (p) (0.0001) (0.0400) (0.0068) Notes: The estimated derivative is: ACROR DIROWN = 3 1 DIROWN + DIROWN + 3 The roots and minimm vale of this derivative are presented in the table > 0 garantees that the derivative has two real roots, i.e. that its minimm vale is negative. This is a non-linear restriction on the parameters and is tested sing a non-linear Wald test. Examining the soltions for the roots of Eqation in Table 4, it may be seen that the estimated relation between managerial share holding and rm performance is nonmonotonic. This is the case even thogh the H index and the CbbinÐ Leech index are both sed in estimation. Based on or above argments, we emphasize the estimates sing the CbbinÐ Leech index. Amongst these, the WLS and maximm likelihood estimates, which make an explicit correction for heteroscedasticity, are to be preferred to the OLS estimates. Hence, it appears that the relation is positive for a managerial share holding between 0 and approximately 11%. It is then negative between approximately 11% and 5% (the entrenchment e ect) and positive thereafter (the convergence e ect). IV. CON CLUD I NG RE MARK S We condcted this stdy to examine two qestions, sing data from the UK nancial services indstry. The rst was to assess the sitability of ownership concentration as an explanatory factor for rm performance. The second was to estimate the relationship between managerial share holding and rm performance, and test whether it is non-monotonic, as has been reported in several US stdies. 3 S cient conditions for the relationship in Eqation to have two trning points, with a local maximm followed by a local minimm are: (a) 1 > 0, < 0 and 3 > 0; and (b) the minimm vale of the derivative of Eqation 3 is negative. It may be seen from Table 4, that all of or estimates satisfy these conditions.

6 180 R. Mdambi and C. Nicosia We nd that ownership concentration and the degree of shareholder control have radically di erent e ects on rm performance. In particlar, increased control vested with the large share holding grops appears to be positively related to performance. However, increased concentration seems to be inversely related to the same performance measre. This sggests that ownership concentration per se may be an incomplete measre in this context. Secondly, we nd strong evidence pointing to a nonmonotonic relationship between managerial share holding and rm performance. This appears to spport the theory sggesting the con icting entrenchment and convergence e ects of managerial share holding. ACKN O WL ED G EMEN TS We shold like to thank Chris Pal and an anonymos referee for comments which have sbstantially improved this paper. The sal disclaimer applies. REFEREN CES Banz, R. (1981) The relationship between retrn and market vale of common stocks, Jornal of Financial Economics, 9, 3Ð 18. Berle, A. A. and Means, G. C. (193) The Modern Corporation and Private Property (revised edition, 1967) Harcort Brace and World, Inc, New York. Coase, R. H. (1937) The natre of the rm, Economica NS, 4, 386Ð 405. Cosh, A. D. and Hghes, A. (1987) The anatomy of corporate control: directors, shareholders and exective remneration in giant US and UK corporations, Cambridge Jornal of Economics, 11, 85Ð 313. Cbbin, J. and Leech, D. (1983) The e ect of shareholding dispersion on the degree on control in British companies: theory and measrement, Economic Jornal, 93 (), 351Ð 69. Demsetz, H. (1983) The strctre of ownership and the theory of the rm, Jornal of L aw and Economics, 6, 375Ð 90. Demsetz, H. and Lehn, K. (1985) The strctre of corporate ownership: cases and conseqences, Jornal of Political Economy, 93, 1155Ð 77. Dimson, E. and Marsh, P. (1989) The smaller companies pzzle, The Investment Analyst, 91, Janary. Encaoa, D. and Jacqemin, A. (1980) Degree of monopoly, indices of concentration and threat of entry, International Economic Review, 1, 87Ð 105. Fama, E. F. and Jensen, M. C. (1983) Separation of ownership and control, Jornal of L aw and Economics, 6, 301Ð 5. Hannah, L. and Kay, J. A. (1977) Concentration in Modern Indstry: Theory, Measrement and the UK Experience, Macmillan, London. Hart, O.D. (1983) The market mechanism as an incentive scheme, Bell Jornal of Economics, 14, 366Ð 8. Hermalin, B. and Weisback, M. S. (1991) The e ect of board composition and direct incentives on corporate performance, Financial Management, 0, 101Ð 1. Jarrell, G. and Polsen, A. (1988) Dal-class recapitalization as an anti-takeover mechanism: the recent evidence, Jornal of Financial Economics, 0, 19Ð 5. Jensen, M. C. and Meckling, W. H. (1976) Theory of the rm: managerial behavior, agency costs and ownership strctre, Jornal of Financial Economics, 3, 305Ð 60. Jensen, M. C. and Mrphy, K. J. (1990) Performance pay and top management incentives, Jornal of Political Economy, 98, 5Ð 64. Jensen, M. C. and Rback, R. (1983) The market for corporate control: the scienti c evidence, Jornal of Financial Economics, 11, 5Ð 50. Leech, D. and Leahy, J. (1991) Ownership strctre, control-type classi cations and the performance of large British companies, Economic Jornal, 101, 1418Ð 37. McConnell, J. J. and Servaes, H. (1990) Additional evidence on eqity ownership and corporate vale, Jornal of Financial Economics, 7, 595Ð 61. Mehran, H. (1995) Exective compensation strctre, ownership and rm performance, Jornal of Financial Economics, 38, 163Ð 84. Morck, R., Shleifer, A. and Vishny, R. W. (1988) Management ownership and market valation: an empirical analysis, Jornal of Financial Economics, 0, 93Ð 315. Roll, R. (1981) A possible explanation for the small rm e ect, Jornal of Finance, 36, 879Ð 88. Schmalensee, R. (1977) Using the H index of concentration with pblished data, Review of Economics and Statistics, 59, 186Ð 93. Stlz, R. (1988) Managerial control of voting rights: nancing policies and the market for corporate control, Jornal of Financial Economics, 0, 5Ð 54. Tsetsekos, G. P. and DeFsco, R. A. (1990) Portfolio performance, managerial ownership and the size e ect, Jornal of Portfolio Management, 16, 33Ð 9. Wrck, K. H. (1988) Eqity ownership concentration and rm vale, Jornal of Financial Economics, 3, 3Ð 8.