Home | Business News | Browse by Publication | A | American Economist

Executive compensation incentives in a volatile market.

Article Excerpt
I. Introduction

In recent years, advances in principal-agent theory have allowed a more thorough analysis of pay-for-performance incentives in executive compensation contracts. By taking into account the risk-averse attitudes of executives, Rajesh Aggarwal and Andrew Samwick (1999) concluded that executive compensation contains a large pay-for-performance element that is a much more significant incentive than previously estimated using simpler models. However, this study used data from a unique period in U.S. financial history: the bull market of the 1990s (specifically 1993-1996). More recently, in the bear market of the early twenty-first century, reports in the popular press and works by academics have questioned whether executive compensation contracts are actually constructed as incentive mechanisms. (1) The criticism has been most forcefully made by Lucian Benchuk and Jesse Fried (2003, 2004) and Benchuk, Fried and David Walker (2002), who argue that manager contracts are not written for the benefit of shareholders, as discussed below.

This paper seeks to determine if pay-for-performance incentives are significant when estimated in the context of Aggarwal and Samwick (1999) type models during the recent period of turbulent stock market returns. The sample period, January 1999 through December 2001, includes periods of both sharply rising and falling stock prices. The differences between the sample period in this paper and that of Aggarwal and Samwick (1999) are dramatically depicted in Figure 1, which shows the Standard and Poor's index of stock prices (adjusted for dividends) for the largest 500 firms (S&P 500). In 1999-2001, stock prices are much more volatile, and contain extended periods of falling prices. (2)

[FIGURE 1 OMITTED]

The results of the analysis indicates that executive compensation does contain a pay-for-performance element in the 1999-2001 period. However, the magnitude of the incentive mechanism seems to be diminished by at least one-half. For 2001, preliminary results suggests that incentive schemes may have disappeared or even reversed. Caution is needed in interpreting these results as differences in the sample and methods make the results not fully comparable to previous studies.

II. Executive compensation models

The problem of how to best compensate executives is a classic application of the principal-agent model. (3) In the agency framework, the collection of owner-shareholders (the principal), desires the executive (the agent) to maximize shareholder value, but cannot observe or evaluate the executive sufficiently. Central to the problem is that the goals of the executive may conflict with those of the shareholders. For example, a manager may be more interested in amassing and defending personal power rather than pursuing profit-maximizing strategies (see e.g. Bebchuk and Fried (2003) for a review).

Michael Jensen and Kevin Murphy (1990a) conducted an early empirical examination of the pay-performance sensitivity with the goal of testing the predictions of agency theory. This paper estimated the total effect of incentive mechanisms, including performance-based salary and bonuses (salary and bonus revised annually based on company performance), stock ownership, stock option grants, and the threat of dismissal. A simple representation of the Jensen and Murphy (1990a) model is expressed in equation (1):

w = a + b[pi] (1)

where w is total executive compensation, a is the guaranteed, or safe component of compensation, b is the sensitivity of compensation to performance, and [pi] is the measure of performance (e.g. change in shareholder wealth). Jensen and Murphy (1990a) found that while a positive statistical relationship did in fact exist between firm performance and executive pay (i.e., b > 0), the total pay-for-performance sensitivity was at most $3.25 for $1,000 change in shareholder wealth, which was claimed to be too low to be consistent with agency theory predictions. (4)

John Garen (1994) took issue with this conjecture, noting that the optimal value of b varies by firm and an aggregate estimation of b will yield downward biased results. (5) Specifically, Garen (1994) considered a standard example when the executive has the utility function:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)

where [rho] is the coefficient of absolute risk aversion, and [mu] is the mean of the random normally-distributed [pi], and k is a constant. In this case, the optimal b for a competitive firm is:

b = 1/1 + k[rho][[sigma].sup.2] (3)

where [[sigma].sup.2] is the variance of [pi]. By substituting plausible values for the parameters in this model, Garen (1994) illustrated that Jensen and Murphy's (1990) estimates may indeed be consistent with agency theory.

Because optimal values for b vary greatly with functional form and other standard assumptions, Garen (1994) suggested that a more appropriate test of agency theory is to construct an optimal contract model and derive the comparative static predictions for b. The model developed in Garen (1994) showed the characteristics of contracts that meet the individual rationality and incentive compatibility constraints for a chief executive officer (CEO) of a representative corporation. In a key extension of the Jensen and Murphy (1990a) approach, this optimal contract model took into account the effect of the risk associated with such contracts when executives are risk averse. Simple models like equation (1) fail to take into account the inherent tradeoff between the insurance and incentive portions of a contract, and the Garen (1994) model showed that the element of risk must be explicitly accounted for when estimating pay-for-performance. (6) This is a clear result of equation (3), where b is an inverse function of both the level of risk aversion and the risk of the performance measure. Garen's (1994) empirical results were consistent with principal-agent theory, albeit with weak statistical significance. (7)

Aggarwal and Samwick (1999) used these theoretical results to develop a linear approximation of the optimal contract (essentially a linearization of (1) and (3) that is...

View this article FREE - Now for a Limited Time, try Goliath Business News
Free for 3 Days!