The myth of executive compensation: do shareholders get what they pay for?

Abstract

We use compensation data for a sample of 701 US public firms and document a significant positive relation between the level of executive compensation and the subsequent realized stock returns. In present value terms, shareholders in firms incurring a total compensation cost of $78 million above the median experience subsequent incremental wealth gains of $385 million above the median over the 5-year period following compensation awards. The results are robust to the use of size and industry-adjusted measures of compensation and returns, and hold after we segment the sample by firm size and industry. Our evidence is consistent with rational executive labour markets where more talented executives command higher compensation and produce superior returns.

Notes

¹ Recently, Core et al. (1999) find that managerial pay is also influenced by agency conflicts of interest under weak corporate governance structures.

² Coughlan and Schmidt (1985) argue that it is asymmetric information resulting from the delegation of authority rather than the separation of ownership and control that is responsible for the agency problems which execute compensation attempts to mitigate.

³ Because a significant portion of compensation is in the form of new grants of stock options, we examine the characteristics of the options granted. We find that about 80% of firms in our sample granted options, and nearly all options are granted at the money (see also, Murphy, 1998). There is a high degree of uniformity in option maturity for firms in our sample. The median maturity is 10 years, and <5% of firms grant options with maturity <5 years. Thus, the types of options granted to top executives as part of their compensation package are similar across cross sections of our sample.

⁴ See, for example, Rosen (1982), Kostiuk (1990), Smith and Watts (1992) or Murphy (1998).

⁵ See e.g. Fama and French (1992), Loughran and Ritter (1995, 1997), Lee (1997) and Daniel and Titman (1998).

⁶ Studies of relative performance include Coughlan and Schmidt (1985), Murphy (1985), Antle and Smith (1986), Gibbons and Murphy (1990), Janakiraman et al. (1992) and Sloan (1993).

⁷ The nonuniform distribution of sample firms across size quintiles is because quintile breakpoints are defined using the population of NYSE/AMEX firms (e.g. see Lee, 1997), and because ExecuComp focuses on relatively large firms. The limited number of observations and the nonuniform distribution of firms across size quintiles prohibit us from conducting analyses for each quintile separately. Unreported analyses reveal that our definition of ‘large’ and ‘small’ most effectively captures size-based differences in the return performance of firms in our sample. Our results are not affected by the use of the median market capitalization of NYSE/AMEX firms to classify sample firms.

⁸ This is similar to the compensation measures used by, for example, Coughlan and Schmidt (1985), Murphy (1985) and Jensen and Murphy (1990).

⁹ In unreported regressions for 2-, 4- and 5-year BHRs, the coefficient of ΔTC remains insignificant. We also find that the coefficient of TC and ΔTC are similar to those in Table 3 for sub-samples of large and small firms and firms in each of the four industrial classifications.

¹⁰ In unreported regressions, we find that the coefficient of ΔTC is small and insignificant in the models used in Table 4.

¹¹ See Murphy (1998) for a review of some of these studies.

¹² From Table 1, median compensation for the full sample is $3.34 million. From Table 2 the median 1-year return for the full is 13%. This means that $1 invested on 1 January 1993 will have grown to $1.13 by year's end. In estimating Equation 2 returns are entered as the terminal wealth at the end of the period, e.g. 1.13, but for expositional purposes we report returns in Table 2 using convention notation, e.g. 0.13.

¹³ Masson (1971) discounts wealth gains at the bond rate of 6%. We feel that a discount rate based on equity returns more adequately reflects the riskiness of equity cash flows.

¹⁴ Since we require data for the year prior to and following a CEO change, we restrict the test to CEO changes that occurred between 1 January 1993 and 31 December 1995 (the June 1997 ExecuComp database contains data beginning in 1992 and ending in 1996). We restrict our analysis to CEOs because ExecuComp does not provide reliable information on turnover for the other top executives.