Printer Friendly

Management ownership and firm compensation policy: evidence from converting savings and loan associations.

The distinction between mutual and stock savings and loan associations has permitted consideration of various facets of the owner-manager conflict. One such facet, expense preference, refers to the tendency of managers to spend more on prequisities than profit-maximizing would dictate. Both theoretical (Deshmukh, Greenbaum, and Thakor [5], Fama and Jensen [8], and Rasmussen [24]) and empirical (Verbrugge and Jahera [27], and Akella and Greenbaum [1]) research suggest the existence of expense preference behavior among mutual savings and loan association (SLA) management when compared with that of stock firms. However, Morck, Shleifer, and Vishny [19] find that the managers of nonfinancial stock firms are not necessarily value-maximizing within a certain range of managerial ownership of the firm. Therefore, the presumed lack of expense preference behavior among stock SLAs may be incorrect for all levels of management ownership.

This study examines a select sample of stock savings and loan associations that have recently converted from mutual ownership as a test of a relatively uncontaminated change in ownership structure on managerial behavior. Such behavior is quantified in terms of employee compensation because of the broad use of that measure in previous studies of financial institutions. (1) The exclusive examination of the compensation of SLAs avoids any interindustry influences. Both Krueger and Summers [14] and Kole [13] have raised questions regarding interindustry effects on traditional measures of managerial behavior, particularly the findings of Morck, Shleifer, and Vishny [19]. To assess differential changes in expense preference behavior for the stock SLAs, employee compensation and certain control variables are examined relative to a matched sample of SLAs remaining mutual in the same market. Further, external factors potentially affecting the level of expense preference such as market concentration (Edwards [7], Hannan and Mavinga [10], James [11], and Verbrugge and Jahera [27]), board of directors' composition (Brickley and James [3], and Morck, Shleifer, and Vishny [19]), and the market for corporate control (James [11]) are explicitly considered.

Our results relating employee compensation, as an example of expense preference behavior, and management ownership are consistent with the Morck, Shleifer and Vishny [19] conclusions concerning overall firm value. We employ a switching regression model which avoids the arbitrary switching points of Morck, et al[19]. Our findings suggest a negative relationship between compensation and management ownership at both low and high levels of ownership with a positive relationship occurring in the middle range. None of the external factors examined altered these conclusions. The conclusions also provide potentially useful information for the regulation of the conversion process. In particular, the results suggest that it may be advantageous to control management's purchase of a converting firm's common stock to avoid the tendency toward expense preference behavior.

The paper progresses as follows. Section I provides a background of the previous research in expense preference behavior and management ownership, with particular reference to stock versus mutual SLAs. The sample and empirical methodology are presented in Section II. Section III contains the results, and Section IV the summary and conclusions.

I. Organizational Form, Expense

Preference Behavior and Ownership

Previous research relating to the mutual form of organization has documented the existence of expense preference behavior on the part of management. (2) In their review of earlier studies of mutual SLAs, Verbrugge and Jahera [27] found that these organizations tended to exhibit higher operating expenses, higher personnel expenditures, lower profitability, and less risky portfolios than comparable stock institutions. Further, they found the expense preference tendency in their study of California SLAs and, separately, in the SLAs in other states.

The difference between mutual and stock forms of organization is in the locus of profit control. The management of mutual banks face a unique set of conditions within which to make decisions (Deshmukh, Greenbaum, and Thakor [5], and Rasmussen [24]). The mutual manager's incentives are different from those of a comparable stock firm. Lacking a direct reward for value maximization, the mutual SLA manager receives value from perquisite consumption. The inability of mutual shareholders to effectively monitor managers results in the agency costs referred to by Alchian and Demsetz [2] and Jensen and Meckling [12].

Despite the evidence of expense preference behavior within mutual organizations, research on stock firms has led to incomplete conclusions. Verbrugge and Jahera [27] concluded that conversion to stock among SLAs should lead to improved managerial efficiency. However, they did not examine the proportionate ownership of management in the stock firm. Their results may not be necessarily congruous at all levels of management ownership. Morck, Shleifer, and Vishny [19] found that the market value of the firm increases with board and management ownership (convergence of interests) up to a 5% level of ownership, but then declines through the 25% level. Several studies of financial institutions (Brickley and JAmes [3], Hannan and Mavinga [10], James [11], and Saunders, Strock, and Travlos [25]) have directly or indirectly examined the effect of management or director equity holdings on bank performance. Hannan and Mavinga [10] find that manager-controlled banks exhibit expense preference behavior when unconstrained by other exogenous factors. In his study of converting SLAs, Masulis [16] concludes that "management incentives are altered so as to increase their (managers') interest in the association's long-term profitability."

In this paper, we examine the effect of management ownership on the level of employee compensation of SLAs that have recently converted from the mutual to the stock organizational form. By examining the SLAs shortly after the change in ownership structure, we minimize any contaminating effects that may result from factors other than conversion. We expand upon earlier expense preference research by using an empirical model that is similar in structure, employing compensation as the measure of expense preference behavior. Consistent with Jensen and Meckling [12], we predict an overall negative relation between the stock ownership of managers and directors and the level of compensation of the SLA. Additionally, if the nature of management behavior changes at different levels of ownership, as argued by Morck, Shleifer, and Vishny [19], we expect a negative relation between compensation and ownership at relatively low and relatively high ownership levels but a positive relation over the middle range.

II. Sample and Empirical Model

A. Sample

The objective of the sample selection was to develop a set of matched SLAs in the same market to control for the many exigencies identified in previous studies. The initial sample consists of all approved conversions that were completed during the 1985-1986 period. Earlier conversions were not examined principally because the Federal Home Loan Bank Board (FHLBB) changed its regulations regarding management stock subscription rights during the 1982-1983 period. There was also difficulty in collecting the relevant financial data from the Federal Home Loan Bank tapes prior to 1984. (3) This initial sample consisted of 174 SLAs that were given approval to convert by the FHLBB. Converting SLAs which were excluded were those that were acquired immediately after conversion, those that were part of the Federal Home Loan Bank Board (FHLBB) Management Consignment Program, those not included in the U.S. Savings and Loan Directory, and those with incomplete data on the FHLBB tapes. In addition, since one of the objectives was to replicate previous studies in terms of exogenous variables, the sample was reduced further because of incomplete data on these variables.

A control sample of comparably sized mutual SLAs in the same geographic market was selected. (4) The control sample diminishes the effect of other factors that may affect the performance of the converting SLA. These factors include concentration and competition within banking markets, demographic considerations and regulation. For example, James [11] explicitly examined the competition factor by comparing a sample of banks in states without an acquisition market to a sample in states with an active acquisition market. Similarly, Hannan and Mavinga [10], Mester [18], and Verbrugge and Jahera [27] limited their samples to institutions within a single state. Given that our sample of converted SLAs is dispersed throughout the U.S., control of these exogenous factors is paramount. The resulting sample consists of 56 converting SLAs along with their matched mutual SLAs.

Descriptive statistics for both samples are presented in Exhibit 1. The variables are measured four quarters before (Panel A) and after (Panel B) conversion. To assure comparability between the test and control samples prior to conversion, we first examined the differences in means both on a univariate and multivariate level for seven pertinent variables. (5) Only return on assets yielded significant results, with the converting firms exhibiting a higher return on assets in the pre-conversion period. Mester [18] had concluded that the general tests of expense preference are only applicable if the production functions are assumed to be comparable. Using a cross-section of both mutual and stock SLAs, she found that such comparability of product functions could not be assumed. The variables nonfinancial income (Fama and Jensen [8], Masulis [16], and James [11]), total assets (Akella and Greenbaum [1]). and loan-deposit ratio are included in Exhibit 1 to examine whether the cost structures between those SLAs intending to convert and the comparable mutual SLAs are different. A lack of significant differences in these variables prior to conversion indicates no likely difference in the production technology between these two samples of mutual SLAs. The F-statistic for the joint test of differences in means between the two samples in the pre-conversion period was 1.47, which is also insignificant.

A comparison of the post-conversion samples also suggests no significant differences in firm characteristics with the exception of the net worth/assets ratio for the stock sample. This is not surprising with the immediate influx of equity capital due to the sale of stock. While the joint F-test for these variables is significant at the one percent


level (F = 4.00), a joint test without the net worth/assets ratio is not (F = 1.23).

B. Empirical Model

Since the objective of this study is to expand upon earlier research, we examined the converting SLAs employing an empirical model that has been widely used in the study of firms in both nonfinancial and financial industries (Edwards [7], Hannan and Mavinga [10], James [11], and Verbrugge and Jahera [27]). (6) The model is as follows:



COMP = officers' and employees' compensation,

L/D = loan/deposit ratio,

SIZE = natural logarithm of assets,

NONFIN = nonfinancial income,

CONC = three-firm concentration ratio,

INCOME = market total personal income,

WAGE = market manufacturing wage, and

OWN = percentage of converted SLA's stock controlled by management and directors.

The first four variables in Equation (1) are measured as the difference between the stock and the control sample mutual SLA four quarters following the conversion. This interval was selected to permit management time to alter the compensation level within the new organizational structure without the contamination of external market changes and, as Exhibit 1 shows, without changes in production technology. (7)

The independent variables were selected from previous expense preference literature to assure that we have controlled for the appropriate factors which may effect a change in personnel compensation. The L/D variable provides a measure of output heterogeneity and its impact on compensation. Additional loan activity requires increasing labor, suggesting a positive coefficient for L/D. While the level of significance varied, empirical studies (Brickley and James [3], Edwards [7], Hannan and Mavinga [10], James [11], and Verbrugge and Jahera [27]) have validated this relationship.

Although our matching process appears to have selected SLAs of comparable asset size, a SIZE variable has been included to control for any unforeseen differences. While Masulis [16] and McNulty [22] suggest that organizational efficiencies improve for larger firms, Murphy [21] indicates that their managements may increase firm size to enhance compensation. Smirlock and Marshall [26] concluded that because large bank management is more difficult to monitor, they exhibit expense preference behavior. Hence, the addition of SIZE minimizes these potentially obscuring effects albeit with an uncertain sign.

In their discussion of mutual versus stock forms of organization, Fama and Jensen [8] found that the principal difference between stock and mutual financial firms is that the former have relatively more nonfinancial business receipts (revenues other than interest, dividends and capital gains) than mutual organizations. Masulis [16] found a similar result when comparing SLA conversion applicants and nonapplicant mutuals. NONFIN, the difference in nonfinancial business income between the stock and mutual SLAs, is included as an exogenous variable to permit the examination of the ownership structure of the SLA given this potential difference in activity. It is expected that NONFIN exhibits a positive coefficient as increasing nonfinancial business activity will require more, specialized labor.

Unlike the other variables in the model, the competitive market considerations (CONC, INCOME, WAGE) are measured as stock variables because, in most cases, both the sample and mutual SLAs were drawn from the same geographic market. For 16 of the stock SLAs in nonmetropolitan statistical areas (MSAs), however, a mutual SLA could not be found in the same community. For these firms, the observations for CONC, INCOME and WAGE were those of the community for the stock firm. Use of matched pairs of SLAs in the same market tends to minimize the role of market variables. Nevertheless, CONC, INCOME, and WAGE have been included to assure that any identified relation between management ownership and differential wage levels is a function of ownership and not strictly due to differences in competition, nonprice demand and demand for labor across geographic regions.

While alternative concentration measures were available, we employed the three-firm concentration ratio because of its common use in earlier research. (8) The observed coefficient sign for the concentration variable in other empirical studies by Brickley and James [3], Edwards, [7], Hannan and Mavinga [10], and Verbrugge and Jahera [27] has been inconsistent as well as of varying significance. Conventional wisdom suggests a positive coefficient between market concentration and compensation, as less competitive markets should display increasing expense preference behavior. Verbrugge and Jahera [27] employed an interaction term between SLA organizational form and market concentration. They concluded that mutual SLAs exhibited greater expense preference behavior in more concentrated markets.

Total personal income in each market, INCOME, represents nonprice demand, which is expected to be positively related to the demand for labor. Earlier studies (Brickley and James [3], Edwards [7], Hannan and Mavinga [10], James [11], and Verbrugge and Jahera [27]) yielded significantly positive coefficients. WAGE, the hourly manufacturing wage rate in each market, is included to control for the market's demand for labor. The above studies, which included an income variable, found a negative (albeit with varying significance) relationship between market wages and compensation.

Because of the 16 converting SLAs where a matched mutual firm was not available in the same community, use of the stock firm's observations for CONC, INCOME and WAGE may have introduced a small measurement error. To limit the effect of such an error, three additional independent variables were included in the model described in Equation (1). These variable are DIFCONC, DIFINC, and DIFWAGE, the differences between the stock and mutual MSA observations for CONC, INCOME, and WAGE, respectively. For 40 cases, these observations are zero.

Finally, OWN is the percentage of newly issued shares subscribed to by converting SLA management. Management was defined as those officers and directors who


purchased stock on the conversion according to the respective SLA's conversion offering circular. (9) Conversion regulations generally limited the amount of conversion stock that management and their associates could purchase as a group. This amount varied between 25% and 35% depending on the asset size of the converting institution. As suggested by Jensen and Meckling [12], managers are less likely to engage in expense preference behavior as their ownership stake rises. With increasing ownership, management's share of the costs of non-value maximizing behavior also increases. This suggests a negative coefficient for OWN.

III. Empirical Results

Exhibit 2 presents the results of the empirical model developed in Equation (1). The F-statistic, which is significant at better than the 1% level, and the adjusted [R.sup.2] of 0.94 suggest that the model does well in explaining the cross-sectional differences in SLA wages. The variable employed to measure ownership in the newly converted SLAs, OWN, has the anticipated sign but is not significant. Thus, using the proportion of equity controlled by management and directors as a single-equation, continuous variable, we cannot reject the notion that management ownership has no effect on expense preference behavior. (10)

In an alternative examination, we consider the possibility that the relation between compensation and the ownership variable cannot be expressed in a single linear equation. Specifically, we test the hypothesis of Morck, Schleifer, and Vishny [19] that the relation between ownership levels and COMP is more accurately represented by three continuous but separate linear regression equations. Over the first equation segment, representing low levels of management ownership, Morck, et al [19] suggest that market discipline, the marginal labor market and the market for corporate control, force the manager toward value maximization. Hence,


we expect the coefficient for low levels of ownership to be negative, suggesting a convergence of interest of management and ownership. Over the middle segment, Morck, Shleifer and Vishny [19] contend that managers have enough power and influence to guarantee their employment and are, therefore, less concerned about the results of their behavior. We expect a positive coefficient for this segment, reflecting entrenchment. Finally, over the upper levels of ownership, Morck, at el [19] predict a return to convergence-of-interests behavior as larger ownership levels have a more direct impact on managers' wealth. A negative coefficient at the upper levels of ownership would reflect this convergence of interest.

Morck, Shleifer, and Vishny [19] argue that an alternative explanation of a negative coefficient at the lower levels of ownership reflects increasing remuneration in the form of stock for better performing firms. In this study, we are able to differentiate between the conflicting hypotheses for lower levels of management as stock for remuneration purposes had not previously been available for the converting SLAs.

To avoid the subjective element of arbitrarily set ownership levels in the examination of possible equation segments, we employed a switching regression technique which allows for a maximum of two switching points. (11) This approach is particularly appropriate because of the lack of a prior relationship between ownership levels and compensation. Using a derivation developed by Fuller [9] of the procedure first employed by Quandt [23], we searched for the most likely switching points. In all, we tested 77 possible switching point combinations. By minimizing the mean squared error of the overall model, we were able to identify the most likely points. As an additional examination, we tested the hypothesis that the coefficients for each equation segment were equal at these points. The results are represented in Exhibit 3.

The F-statistic for the switching model is significant at better than 1%. The coefficient for the independent variables, other than OWN, are qualitatively similar to the comparable coefficients in the single equation model (see comparable coefficients in the single equation model (see Exhibit 2). (12) For the OWN variable, the model identified the 15% and 27% ownership levels as the most likely switching points. (13) The upper switching point is very similar to that of Morck, Shleifer and Vishny [19], but the lower switching point is substantially higher (15% versus 5% for Morck, et al [19]). Our observed ranged is similar to that used by Hanna and Mavinga [10], who differentiated between owner and management control by eliminating the middle range of ownership levels defined as between 10% and 25%. However, after a review of pertinent literature, we were able unable to discover additional reasons for the specific ownership switching points identified in our study.

Using the levels to differentiate the three equation segments minimized the model's mean squared errors. As anticipated, the coefficient for the first segment is negative and significant. The coefficient for the middle segment is positive and significant at the 10% level. While the coefficient for the third segment is significant at the 12% level, the sign suggests that above ownership levels of 27% some decline in wages is present. The tests of the difference in coefficient values for the OWN variable for the lower and middle segments and the middle and upper segments are significantly different, albeit at the 10% level for the latter. (14) Alternative estimations of the model in Equation (1) yield similar results. (15)

To examine whether the preceding results are robust to the presence of additional variables, we introduce a series of independent variables in the switching regression which quantify additional labor and capital market control considerations. Each of the following variables (or similar variables proxying for the same effect) have been developed by previous authors in the context of expense preference behavior examinations.

The first of these variables is the interaction effect between management ownership and the concentration ratio (OWN x CONC). Verbrugge and Jahera [27] suggest that if management is more likely to exhibit expense preference behavior at low levels of ownership, this propensity should be exacerbated in less competitive markets.

Second, James [11] found that the likelihood of expense preference behavior for banks in states where takeovers are permitted is lower than in those states restricting acquisitions. For the SLA industry, interstate takeovers were only permitted for ailing SLAs during this period. To estimate the financial health of the SLAs, we used the ranking provided by S & L-Savings Bank Financial Quarterly (RANK). The rank (RANK) is developed from financial ratios and ranges from 1 to 300.S & L-Savings Bank Financial Quarterly reports that all but nine of the 143 thrift reorganizations and closings in 1985 and 1986 had a ranking of 50 or less. The mean and standard deviation of the ranks of the 56 converting SLAs in this sample is 160.56 and 65.1163, respectively. Only three of the SLA's have ranks below 50.

If management is more likely to exhibit expense preference behavior at low levels of ownership, this propensity should increase when takeovers are unlikely. (16) Thus, the this independent variable included is the interaction between ownership and the quality ranking (OWN x RANK).

Finally, we examine the interaction effect between the fraction of the board of directors from "outside" the SLA (DIR) and the quality ranking (DIR x RANK). Brickley and Jame [3] argue that the market for takeovers and the independence of corporate boards (outside versus management membership) are substitutes for one another in the discipline of management. The percentage of outside directors was obtained from the converting SLA's offering circular. In the markets where a takeover is difficult or impossible, more independent boards will emerge.

To measure the independent effects of OWN x CONC, RANK, OWN x RANK and DIR x RANK, we reestimated the model including each variable in subsequent regressions. A general linear models test was used to examine the contribution of each of the variables. The F-statistic generated by the test measures the relative change in the model's sum of squared errors with and without the variable. [17] In this manner, we determine the marginal contribution of each variable, if any, and their effect on the OWN coefficients.

In general, the switching results for OWN were unchanged, and the F-statistic for the general linear models test was not significant for any of the variables. [18] Moreover, none of the coefficients of these additional variables was significant. [19] These results imply that the relationship between expense preference behavior and managerial ownership in savings and loan associations is unaffected by both labor and capital market controls.

IV. Summary and Conclusions

The finding of M[phi]rck, Shleifer, and Vishny [19], that the relationship between managerial ownership and firm value is nonlinear, has potentially important implications for the expense preference literature. Previous research has suggested that mutual savings and loan associations are more likely to exhibit expense preference behavior than comparable stock associations. However, if the non-linear relationship between managerial ownership and firm value also applies to particular forms of firm behavior, the presumed lack of expense preference behavior among all stock associations may be inaccurate.

Converting savings and loan associations provide a unique form of organizational change within which to examine this issue. Management of the converting firm is able to purchase stock within conversion regulatory constraints. We compare the behavior of the converting firm with a similar mutual savings and loan association in the same market to allow for potential contaminating factors such as the banking market, demographic and regulatory considerations. Consistent with earlier studies, employee compensation is used as the measure of expense preference behavior.

As recent research has suggested that the relation between ownership and management behavior may switch at one or more levels, our empirical model is designed to estimate these points. We find that for SLAs where management ownership is less than 15% of the firm's value the relation between compensation and ownership is negative. For those SLAs where management ownership is greater than 15% but less than 27%, the relation is positive. Beyond the 27% level, the evidence, though not conclusive, suggests a negative relation between ownership and compensation. These results provide support for the convergence of interests hypothesis (management acts in the interest of the owners) between management and ownership for the lower and upper levels of management ownership and the management entrenchment hypothesis (management acts in its own interest) over the middle range.

Finally, these conclusions raise important regulatory concerns. Particularly during the 1980's, thrift regulatory agencies were preoccupied with the need for savings and loans to raise additional capital. In so doing, management was encouraged to purchase up to the maximum amount of stock prescribed by regulatory guidelines. Such action was deemed an essential signal to the market that management was sufficiently optimistic about the institution's prospects. However, by sanctioning such levels of equity participation by management, the regulatory agencies appear to also have implicitly endorsed some degree of expense preference behavior which may limit future capital growth. A normative conclusion from this research may be that additional considerations be given to the level of initial equity ownership in converting institutions.

(1) Studies which have included compensation as a form of expense preference behavior include: Demsetz [4], Edwards [7], Hannan and Mavinga [10], James [11], Verbrugge and Jahera [27], and Williamson [28]. Alternative measures include occupancy expense, number of employees and various forms of financial risk.

(2) To our knowledge, only Mayers and Smith [17], in their study of life insurance companies, found that conversion from the stock to mutual form of organization improved operating efficiency.

(3) Prior to 1984, the Federal Home Loan Bank Board tapes were reported on a quarterly basis. Further, the data was noncumulative.

(4) The geographic market was determined by examining the U.S. Savings and Loan Directory. When the converting SLA existed in a metropolitan statistical area (MSA), the mutual SLA was identified in that same MSA. If the converting SLA was not in an MSA, the control SLA was determined first by examining the same city, and, if not successful, comparable-sized cities in the same county.

(5) Since individual differences in means tests ignore the correlations among variables, they fail to use the total available information in assessing group differences. Alternatively, multivariate analysis, employing Wilks' lambda, considers these interactions by implicitly testing the linear combinations of the variables that provide the strongest evidence of overall group differences. For a further description of overall group differences with Wilks' lambda, see Morrison [20, Ch. 5]. Wilcoxon rank sum tests, for both pre- and post-conversion data sets, were also conducted for each of the variables. Except for the net worth/assets ratio in the post-conversion data, none of these differences were significant.

(6) Akella and Greenbaum [1] examined SLA output instead of inputs to quantify possible existence of expense preference behavior. Comparing samples of mutual and stock SLAs, the authors concluded (i) mutual SLAs expanded deposits and loans beyond profit-maximizing levels, and (ii) greater diffusion of ownership implies higher agency costs.

(7) Mester [18] argues that the commonly employed tests of expense preference behavior which compare mutual SLAs to stock SLAs are incorrect because of differences in production functions between these two types of institutions. Our comparison of mutual SLAs with recently converting stock SLAs should reveal marginal changes in expense preference behavior without allowing sufficient time for the converting SLAs to substantially alter their production function. See Exhibit 1, Panel B.

(8) Several previous studies have included a threshold concentration level based on Quandt's [23] linear switching regression technique. In replicating this technique for our sample, however, we were unable to find a switching point. Consequently, the continuous concentration measure was retained.

(9) Subscription data was obtained from the offering circulars and subscription prospectuses of the converting SLAs. The authors obtained these documents from the Freedom of Information Act Office of the Federal Home Loan Bank Board.

(10) Edwards [7] and James [11] examined the number of employees as an alternative dependent variable. We also analyzed Equation (1) with the number of employees as the dependent variable. The results were qualitatively similar to those reported for COMP in Exhibit 2.

(11) See the Appendix for a description of the switching regression model.

(12) The mode' was tested for the problems of heteroskedasticity and multicollinearity using techniques described by White [29] and Marquardt [15], respectively. The tests failed to identify the existence of either problem.

(13) The switching regression model was also estimated without the competitive market variables (CONC, INCOME, and WAGE) and without DIFCONC, DIFINC, and DIFWAGE. The results were qualitatively similar.

(14) The difference between ownership coefficient values is as follows: For the first versus the second segment, the F-statistic is 5.67** for the second versus the third, the F-statistic is 3.18*. Other variables which were included in other studies were also examined. These were growth in assets and number of offices. Neither provided substantial additional information to our heuristic model nor statistically significant coefficients and were not reported in the tables.

(15) In Edwards [7], an empirically testable model is derived assuming a firm characterized by a two-factor Cobb-Douglas production function. The model is presented in the following equation (ln = natural logs):

In(SMSA Comp) = [b.sub.0] + [b.sub.1]ln(WAGE) + [b.sub.2]ln(INCOME) + [b.sub.3](CONC)

where SMSA Comp is the total wage for banks in the SMSA and the remaining variable abbreviations are generally the same as those described for Equation (1). For our estimation of Edward's equations, the dependent variable is the natural log of compensation for the converting SLA. The natural log of compensation for a matched sample of mutual SLAs is included as a control variable and the natural log of ownership as the test variable. The adjusted [R.sup.2] for the switching regression is 0.62 with an F-statistic of 13.9. The coefficients for the lower and upper ownership levels are negative and significant at the 1% and 5% levels, respectively. The middle level is positive, but not significant.

(16) Attempts were made to find a switching point for the ranking variable. No significant level was found, however.

(17) For a further explanation of the general linear models test, see Draper and Smith [6, pp. 75, 76].

(18) The results for the general linear models test for the inclusion of each of the aforementioned variables is as follows:
 Variable F-statistic
OWN x CONC 1.440
RANK 0.005
OWN x RANK 0.125
DIR x RANK 2.700

None of the F-statistics are significant at traditional levels.

(19) The measure of board independence (DIR) was also used as an independent variable. The F-statistic for the general linear models test was not significant (2.06) and the variable had little effect on the model.


[1] S. Akella and S. Greenbaum, "Savings and Loan Ownership Structure and Expense Preference," Journal of Banking and Finance (July 1988), pp. 419-437.

[2] A. Alchian and J. Demsetz, "Production, Information Costs and Economic Organization," American Economic Review (December 1972), pp. 777-795.

[3] J. Brickley and C.M. James, "The Takeover Market, Corporate Board Composition, and Ownership Structure: The Case of Banking," Journal of Law and Economics (April 1987), pp. 161-180.

[4] H. Demsetz, "The Structure of Ownership and the Theory of the Firm," Journal of Law and Economics (June 1983), pp. 375-390.

[5] S.D. Deshmukh, S. Greenbaum, and A. Thakor, "Capital Accumulation and Deposit Pricing in Mutual Financial Institutions," Journal of Financial and Quantitative Analysis (December 1982), pp. 706-725.

[6] N. Draper and H. Smith, Applied Regression Analysis, New York, John Wiley and sons, 1966.

[7] F. Edwards, "Managerial Objectives in Regulated Industries: Expense-Preference Behavior in Banking," Journal of Political Economy (February 1977), pp. 147-162.

[8] E. Fama and M. Jensen, "Agency Problems and Residual Claims," Journal of Law and Economics (June 1983), pp. 327-349.

[9] F. Fuller, Introduction to Statistical Time Series, New York, John Wiley and Sons, 1976.

[10] T. Hannan and F. Mavinga, "Expense Preference and Managerial control: The Case of the Banking Firm," The Bell Journal of Economics (Autumn 1980), pp. 671-682.

[11] C. James, "An Analysis of the Effect of State Acquisition Laws on Managerial Efficiency: The Case of the Bank Holding Company Acquisitions," Journal of Law and Economics (April 1984), pp. 211-226.

[12] M. Jensen and W. Meckling, "Theory of the Firm: Managerial Behavior Agency Costs and Ownership Structure," Journal of Financial Economics (October 1976), pp. 305-360.

[13] S. Kole, "A Reexamination of the Interaction Between Board Ownership of Equity and Firm Performance," Working Paper, University of Chicago, Chicago, IL, 1990.

[14] A. Krueger and L. Summers, "Efficiency Wages and the Inter-Industry Wage Structure," Econometrica (March 1988), pp. 259-293.

[15] D. Marquardt, "Generalized Inverse Ridge Regression Biased Linear Estimation and Nonlinear Estimation," Technometrics (August 1970), pp. 591-612.

[16] R. Masulis, "Changes in Ownership Structure: Conversions of Mutual Savings and Loans to Stock Charter," Journal of Financial Economics (March 1987), pp. 29-59.

[17] D. Mayers and C. Smith, Jr., "Ownership Structure and Control," Journal of Financial Economics (May 1986), pp. 73-98.

[18] L. Mester, "Testing for Expense Preference Behavior: Mutual Versus Stock Savings and Loans," Rand Journal of Economics (Winter 1989), pp. 483-498.

[19] R. M[phi]rck, A. Shleifer, and R. Vishny, "Management Ownership and Market Valuation," Journal of Financial Economics (January/March 1988), pp. 293-315.

[20] D. Morrison, Multivariate Statistical Methods, New York, McGraw-Hill, 1967.

[21] K. Murphy, "Corporate Performance and Managerial Remuneration," Journal of Accounting and Economics (April 1985), pp. 11-42.

[22] J. McNulty, Economies of Scale in the Savings and Loan Industry, Atlanta, GA, Federal Home Loan Bank, January 1981.

[23] R. Quandt, "The Estimation of the Parameters of a Linear Regression System Obeying Two Separate Regimes," Journal of the American Statistical Association (January 1958), pp. 873-880.

[24] E. Rasmussen, "Mutual Banks and Stock Banks," Journal of Law and Economics (October 1988), pp. 395-421.

[25] A. Saunders, E. Strock, and N. Travlos, "Ownership Structure, Deregulation and Bank Risk Taking," Journal of Finance (June 1990), pp. 643-654.

[26] M. Smirlock and W. Marshall, "Monopoly Power and Expense-Preference Behavior. Theory and Evidence to the Contrary," Bell Journal of Economics (Spring 1983), pp. 166-178.

[27] J. Verbrugge and J. Jahera, Jr., "Expense-Preference Behavior in the Savings and Loan Industry," Journal of Money, Credit and Banking (November 1981), pp. 465-476.

[28] O. Williamson, "Managerial Discretion and Business Behavior," American Economic Review (December 1963), pp. 1032-1057.

[29] H. White, "A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity," Econometrica (May 1980), pp. 817-838.

Appendix. Switching Regression

Quandt [23] considered a model that examines the relationship between two variables Y and X, where the structure of this relationship changes at one or more points (S[unkeyable]). The model estimates regression equations for each of these structures or switching regimes.

The Quandt model can be extended to include additional control variables through a modification described by Fuller [9]. If we assume that the relationship is continuous and that the regressions have equal variances, a two-switching-point model can be expressed in the following set of equations:

[Mathematical Expressions Omitted]

where Z represents exogenous, control variables, [[mu].sub.i] is the regression error and [[alpha].sub.i] and [[beta].sub.i] are the intercept and slope coefficient for regression i, respectively. Using the continuity assumption, it follows that:

[Y.sub.1] = [Y.sub.2], if X = [S.sub.1] (A4)


[Y.sub.2] = [Y.sub.3], if X = [S.sub.2]. (A5)

Substituting and solving for [[alpha].sub.2] and [[alpha].sub.3] provides the following equations:

[[alpha].sub.2] = [[alpha].sub.1] + [[beta].sub.1][S.sub.1] - [[beta].sub.2][S.sub.1] (A6)

[[alpha].sub.3] = [[alpha].sub.1] + [[beta].sub.1][S.sub.1] + [[beta].sub.2]([S.sub.2] - [S.sub.1]) - [[beta].sub.3][S.sub.2]. (A7)

Substituting (A6) and (A7) into (A2) and (A3), respectively, provides the following empirical model:

[Y.sub.1] = [[alpha].sub.1] + [[beta].sub.1][X[mu].sub.1], if X [is less than or equal to] [S.sub.1] (A8)

[Y.sub.2] = [[alpha].sub.1] + [[beta].sub.1][S.sub.1] + [[beta].sub.2](X - [S.sub.1]), if [S.sub.1] < X [is less than or equal to] [S.sub.2] (A9)

[Y.sub.3] = [[alpha].sub.1] + [[beta].sub.1][S.sub.1] + [[beta].sub.2]([S.sub.2] - [S.sub.1]) + [[beta].sub.3](X - [S.sub.2]), if X > [S.sub.2] (A10)

where [[beta].sub.1], [[beta].sub.2] and [[beta].sub.3] are the coefficients of interest and represent the slopes of each regime.

The regression model (Equations (A8) - (A10)) can be estimated with computations of possible switching points to find those that minimize the mean squared error. The resulting slope coefficients for each regime can then be tested for equality.

Richard B. Carter is currently an Assistant Professor of Finance and Roger D. Stover is a Professor of Finance at the College of Business of Iowa State University, Ames, Iowa.
COPYRIGHT 1991 Financial Management Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1991 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Corporate Compensation Policy Special Issue; includes appendix
Author:Carter, Richard B.; Stover, Roger D.
Publication:Financial Management
Date:Dec 22, 1991
Previous Article:Factors affecting price earnings ratios and market values of Japanes firms.
Next Article:ESOPs and profit-sharing plans: do they link employee pay to company performance?

Related Articles
Compensation policy and the investment opportunity set.
Monitoring versus bonding: shareholder rights and management compensation.
Best practices in director pay.
The positive effects of stock ownership.
The director as employee of management.
A reply from the author.
Uncertainty in Executive Compensation and Capital Investment: A Panel Study.
Building Long-Term COMPANY VALUE.
Why compensation committees need your help: CPAs can play an important role in developing effective pay policies.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters