«by Charles P. Himmelberg, R. Glenn Hubbard, and Darius Palia* This Draft: June 22, 1999 * Respectively: Columbia University, Columbia University and ...»
UNDERSTANDING THE DETERMINANTS OF MANAGERIAL OWNERSHIP AND
THE LINK BETWEEN OWNERSHIP AND PERFORMANCE
Charles P. Himmelberg, R. Glenn Hubbard, and Darius Palia*
This Draft: June 22, 1999
* Respectively: Columbia University, Columbia University and the National Bureau of Economic
Research, and Columbia University and the University of Chicago. We are grateful for helpful
comments and suggestions from two anonymous referees and from Anup Agrawal, George Baker, Sudipto Bhattacharya, Steve Bond, Charles Calomiris, Harold Demsetz, Rob Hansen, Laurie Hodrick, Randy Kroszner, Mark Mitchell, Andrew Samwick, Bill Schwert (the editor), Scott Stern, Rob Vishny, and Karen Wruck, as well as participants in seminars at Boston College, Columbia, University of Florida, Harvard, London School of Economics, M.I.T., Oxford, Virginia Tech, the 1998 Western Finance Association meetings, and the National Bureau of Economic Research.
Charles P. Himmelberg R. Glenn Hubbard Darius N. Palia Graduate School of Business Graduate School of Business Graduate School of Business Columbia University Columbia University Columbia University 606 Uris Hall 609 Uris Hall 413 Uris Hall 3022 Broadway 3022 Broadway 3022 Broadway New York, NY 10027 New York, NY 10027 New York, NY 10027 (212) 854-2622 (212) 854-3493 (212) 854-4402 email@example.com firstname.lastname@example.org email@example.com
1. Introduction Since Berle and Means (1932), the conflict between managers and shareholders has been studied extensively by researchers seeking to understand the nature of the firm. When shareholders are too diffuse to monitor managers, corporate assets can be used for the benefit of managers rather than for maximizing shareholder wealth. It is well known that a solution to this problem is to give managers an equity stake in the firm. Doing so helps to resolve the moral hazard problem by aligning managerial interests with shareholders’ interests. Therefore, Jensen and Meckling (1976) suggest that managers with small levels of ownership fail to maximize shareholder wealth because they have an incentive to consume perquisites. In a similar fashion, some commentators have decried low levels of managerial ownership in U.S. corporations, and the theme has even appeared in discussions by compensation specialists and boards of directors.
In this paper, we propose an equilibrium interpretation of the observed differences in ownership structures across firms. Rather than interpret low ownership levels as per se evidence of suboptimal compensation design, we argue that the compensation contracts observed in the data are endogenously determined by the contracting environment, which differs across firms in both observable and unobservable ways. In particular, low levels of managerial ownership might well be the optimal incentive arrangement for the firm if the scope for perquisite consumption (or more generally, the severity of the moral hazard problem for managers) happens to be low for that firm. We do not deny the importance of agency problems between stockholders and managers, but rather emphasize the importance of unobserved heterogeneity in the contracting environment across firms.
We begin by examining the observable determinants of managerial ownership. This
the level of managerial ownership is determined by the riskiness of the firm, measured by the volatility of the stock price. They argue that the scope for moral hazard is greater for managers of riskier firms, which therefore means that those managers must have greater ownership stakes to align incentives. They also point out that riskiness makes it costlier for managers to hold nondiversified portfolios (assuming that equity holdings in the firm are not easily hedged), so the relation between managerial ownership and nondiversifiable stock price risk is not necessarily monotonic.
To document the extent to which managerial ownership is endogenously determined by the contracting environment, we extend the empirical specification used by Demsetz and Lehn by including a number of additional explanatory variables other than stock price variability (see also Kole, 1996). Most important, we include variables (such as firm size, capital intensity, R&D intensity, advertising intensity, cash flow, and investment rate) designed to control for the scope for moral hazard. To the extent that our additional explanatory variables proxy for moral hazard, our specification clarifies the role of stock price variance as an explanatory variable for managerial ownership. We also use panel data that allow us to estimate the importance of unobserved (time-invariant) firm effects. These results show that a large fraction of the crosssectional variation in managerial ownership is “explained” by unobserved firm heterogeneity. In our subsequent analysis of the determinants of firm value, we argue that this unobserved heterogeneity generates a spurious correlation between ownership and performance.
The second goal of this paper is to reexamine theoretical explanations of the empirical link between managerial ownership and firm performance. Mørck, Shleifer, and Vishny (1988)
Tobin’s Q increases and then decreases with managerial ownership. McConnell and Servaes (1990) examine a larger data set than the Fortune 500 firms examined by Mørck, Shleifer, and Vishny and find an inverted U- shaped relation between Q and managerial ownership, with an inflection point between 40% and 50% ownership. Hermalin and Weisbach (1991) analyze 142 NYSE firms and find that Q rises with ownership up to a stake of 1%; the relation is negative in the ownership range of 1-5%, becomes positive again in the ownership range of 5-20%, and turns negative for ownership levels exceeding 20%. The pattern identified by Mørck, Shleifer, and Vishny has been corroborated for a cross-section of U.S. firms from 1935 by Holderness, Kroszner, and Sheehan (1999). Kole (1995) examines the differences in data sources used in several recent studies and concludes that differences in firm size can account for the reported differences between those studies. These studies generally interpret the positive relation at low levels of managerial ownership as evidence of incentive alignment, and the negative relation at high levels of managerial ownership as evidence that managers become “entrenched” and can indulge in non-value-maximizing activities without being disciplined by shareholders. However, these studies do not address the endogeneity problem that confronts the use of managerial ownership as an explanatory variable, a problem noted early by Jensen and Warner (1988, p. 13).
We investigate the degree to which this heterogeneity makes managerial ownership an endogenous variable in models of firm performance. Following in the tradition of Demsetz and Lehn, we describe the contracting problem faced by the firm and develop a simple empirical model to illustrate the econometric issues that are encountered when estimating the relation among managerial ownership, its determinants, and its effect on firm performance. Distinct from
managerial ownership on firm performance, we use panel data to test for the endogeneity of managerial ownership in models linking ownership to performance (measured by Tobin’s Q). In particular, we use panel data to investigate the hypothesis that managerial ownership is related to observable and unobservable (to the econometrician) firm characteristics influencing contracts.
If the unobserved sources of firm heterogeneity are relatively constant over time, we can treat these unobserved variables as fixed effects, and use panel data techniques to obtain consistent estimates of the parameter coefficients. This approach provides consistent estimates of the residuals in the Q regression, which we use to construct a test for correlation between managerial ownership and unobserved firm heterogeneity.
Our principal findings are threefold. First, proxies for the contracting environment faced by the firm (i.e., observable firm characteristics) strongly predict the structure of managerial ownership. We substantially extend the set of explanatory variables examined by Demsetz and Lehn, and we show that many of our results are robust to the inclusion of observed determinants of managerial ownership, industry fixed effects, or firm fixed effects. Second, we show that the coefficient on managerial ownership is not robust to the inclusion of fixed effects in the regression for Tobin’s Q. Our formal statistical test rejects the null hypothesis of a zero correlation between managerial ownership and the unobserved determinants of Tobin’s Q, thus supporting our conjecture that managerial ownership is endogenous in Q regressions. That is, managerial ownership and firm performance are determined by common characteristics, some of which are unobservable to the econometrician. Third, we explore the use of instrumental variables as an alternative to fixed effects to control for the endogeneity of managerial ownership
performance, but this evidence is tentative because of the weakness of our instruments. We argue that future progress will require a more structural approach to the model.
Kole (1996) also argues that managerial ownership is endogenous; she further argues that causality operates in the opposite direction, from performance to ownership. Using a panel-data vector autoregression, we corroborate Kole’s reverse causality evidence (results available upon request). Our research, however, supports the idea that both ownership and performance are determined by similar (observed and unobserved) variables in the firm’s contracting environment. Thus, our interpretation is different from Kole’s interpretation. That is, we find evidence of endogeneity caused by unobserved heterogeneity, as opposed to reverse causality.
The paper is organized as follows. In Section 2, we outline a simple model of managerial ownership and explain why it is difficult to estimate the relation between managerial ownership levels and firm performance, particularly in the context of cross-sectional data.
Section 3 describes the sample selection criteria and the data we use in our empirical analysis of managerial ownership and firm performance. In Sections 4 and 5, respectively, we present empirical evidence on the determinants of managerial ownership and on the relation between managerial ownership and firm performance. Section 6 concludes.
2. An empirical framework for analyzing executive contracts A common approach for estimating the impact of managerial ownership on firm value is to regress Tobin’s Q on such variables as the percentage of equity held by managers. In this section, we argue that this regression is potentially misspecified because of the presence of
also determinants of managerial ownership, then managerial ownership might spuriously appear to be a determinant of firm performance. To motivate our focus on the endogeneity of managerial ownership, we provide three examples of likely sources of unobservable heterogeneity, and in each case, we discuss their econometric consequences for cross-sectional regressions. We follow this discussion with a more formal exposition, in which we assume that the unobserved heterogeneity is a “firm fixed effect,” and we show how, under this assumption, panel data can be used to mitigate the endogeneity problem. In Section 5, we return to this model to describe a test for the endogeneity of ownership in regressions for Tobin’s Q.
For our first example of unobserved heterogeneity, consider two firms that are identical except that the owner of one of the firms has access to a superior monitoring technology. Under the optimal contracting regime, the owners with access to the superior monitoring technology will choose a lower level of managerial ownership to align incentives, and this firm will have a higher valuation because fewer resources will be diverted to managerial perquisites. If measures of the quality of the monitoring technology are omitted from the specification, a regression of firm value on managerial ownership will spuriously (and falsely) indicate a negative relation, because ownership is a negative proxy for the quality of monitoring technology.
Intangible assets provide a second example of unobserved firm heterogeneity. Suppose two firms are identical except that one of the firms operates with a higher fraction of its assets in the form of intangibles. Under the optimal contracting regime, the owners of this firm will require a higher level of managerial ownership to align incentives because the intangible assets are harder to monitor and therefore subject to managerial discretion. This firm will also have a
book value of assets in the denominator will understate the value of intangibles (because Tobin’s Q is measured as the ratio of the market value of the firm’s outstanding debt and equity divided by the book value of assets). In this example, the unobserved level of intangibles induces a positive correlation between managerial ownership and Tobin’s Q, but this relation is spurious, not causal.
A third example of unobserved heterogeneity is variation in the degree of market power.