Tracking pay for performance.

Are CEOs paid to run up the size of their companies and not to enhance performance and shareholder value? Media pundits say yes, but their research methods may be flawed.

The level of CEO compensation--and whether executives are paid for performance--has recently become an issue with political overtones. A number of observers, including consultant Graef Crystal, contend their analyses provide strong evidence that compensation is highly related to firm size, but only weakly related to firm performance. We disagree with that assertion, as does CE contributor David Meredith of Personnel Corp. of America (see article "Exploding Myths Of CEO Pay" CE: September 1992). As our analysis demonstrates, changes in firm performance are directly related to changes in chief executive compensation.

But we also challenge another compensation myth advanced by various media pundits, namely, that many CEOs are rewarded for simply running up the size of their companies. Our analysis indicates that changes in firm size exhibit only a trivial association with changes in chief executive compensation.

TAINTED METHODOLOGY

If arguments in the popular and business press are correct, the structure of compensation arrangements has disturbing consequences. Generally speaking, the absence of pay-for-performance incentives can prompt poor corporate investment choices and erode shareholder value.

But many observers base their criticisms of CEO pay on skewed data and tainted methodology. Most analyses are based on a "cross-sectional" association between the level of compensation, on one hand, and firm performance and size on the other. To a somewhat lesser extent, other factors deemed to have an impact on chief executive pay are CEO tenure, firm risk and geographical location.

These analyses, using a technique known as multiple regression, are usually conducted at a single point in time. Most indicate a company's size (which is usually measured by sales or total assets) is strongly related to CEO compensation, while its performance (which is usually measured by accounting or stock market rates of return on total assets) is not.

Using such an approach, we analyzed cash compensation--defined as base salary plus annual bonus--for CEOs of "Forbes 500" companies between 1970 and 1988. More specifically, we calculated the proportion of cash compensation that is explained by firm sales and return on assets. Higher figures indicate a stronger relationship between pay and the explanatory variables.

CONTROVERSIAL INTERPRETATION

In a cross-sectional analysis for each year, sales (averaged across all years and all industrial groupings) explained 17.65 percent of the variation in CEO cash compensation. But the incremental explanatory power of ROA was only 3 percent. We found comparable results for six different industrial groups, ranging from primary industrial to financial firms.

At first glance, these results indicate CEOs are paid to increase the size of their firms. But this interpretation is more controversial than it might seem, because it assumes the relationship between compensation, and firm size and performance is the same across a diverse set of firms.

When using cross-industry data, it is necessary to assume the relationship between pay and performance is the same for a chemical company, food company, electric utility, and money center bank. It is clear, however, that considerable differences exist in the risk, complexity and strategy of corporations across industries--and even for those in roughly the same industrial sector. Similarly, no uniform pattern exists for equity held by executives. Thus, caution must be used in drawing conclusions from cross-sectional analyses about the compensation incentives of CEOs.

ALTERNATIVE MODEL

A more appropriate way to examine such incentives is to analyze the relationship between compensation, and firm size and performance over time for the same firm. In contrast to other methods, this approach assumes the relationship between pay and other variables is the same over time for a single firm--but not the same across all firms or types of business. Thus, it yields a firm-specific assessment regarding the determinants of executive compensation.

Using this method, we analyzed the same CEO cash compensation data over the same time period. We conducted a separate regression analysis for each of 568 firms in our sample that had at least 10 consecutive yearly observations, and then computed the proportion of the change in compensation that is explained by changes in firm size and/or performance.

These results shed a different light on the determining factors of CEO pay packages. When data for all industries is used in the analysis, on average, changes in ROA explain 12.49 percent of changes in CEO cash compensation, while the incremental explanatory power contributed by sales is approximately 3 percent. These figures are practically mirror images of those yielded by the cross-sectional analysis.

Our conclusion: With the possible exception of regulated service firms, CEO cash compensation is strongly tied to firm performance, but has an inconsequential relationship with firm sales. In corporate America, pay-for-performance is a reality.

MITIGATING FACTORS

We do not argue that ROA is the most appropriate measure for evaluating CEO performance. In many situations, an increase in sales is highly desirable; in others, the rise or fall of a company's stock price is the most accurate measure.

The choice of performance measures for a compensation contract--as well as the selection of measures for statistical analysis--is an extremely complex task. Detailed knowledge of corporate strategy and the characteristics of executives are necessary to make an informed choice. Our primary point is merely to illustrate the problems inherent in cross-sectional analysis.

One limitation to our method is that we focus only on the cash compensation (salary and annual bonus) component of CEO pay. Obviously, executive pay also consists of stock options and other forms of remuneration contingent on firm performance. We omitted these components largely because of the variation in disclosure requirements over our time period.

Nonetheless, it is clear that options and other "long-term" compensation arrangements only provide a payoff if a company performs well. In other words, an executive will only pocket profit from a stock option if the stock price increases over time. Moreover, the value of the shares owned by the CEO is perfectly correlated with changes in the stock price. As a result, it is likely that our analysis substantially understates the extent to which changes in firm performance translate into changes in CEO wealth.

It is also important to realize that these statistical analyses can only capture the contemporaneous portion of pay-for-performance incentives (e.g., the relationship between change in pay and corporate performance this year). But CEOs are also paid for accomplishing various nonfinancial corporate and individual objectives, including the implementation of executive training programs, quality improvement, and succession planning. Each of these nonfinancial objectives can affect a firm's future performance. However, if the CEO is paid now for completing these objectives, the explanatory power of our regression analysis will be quite low.

The upshot: Without further inquiry into the terms and conditions of the compensation arrangement, it is difficult to conclude that the low explanatory power in the regression implies the absence of pay-for-performance.

CAUTIONARY NOTE

Of course, PCA's David Meredith aims to calculate what an executive's stock options will be worth five years from now--and then to gauge the current-year effect of the equity kicker on compensation. But that isn't the only methodological difference between our analysis and that of Meredith. In particular, he compares pay and performance across several hundred firms at a single point in time (in the current CE article, compensation and firm performance data for 1991). In contrast, we examine the association between pay and performance individually for hundreds of firms over 19 years.

We believe our approach more directly measures how firm performance--presumably affected by the decisions of the chief executive--translates into changes in compensation.

Finally, we also sound a cautionary note about the so-called leverage index created by Meredith of PCA and similar measures constructed by others. To be sure, such a benchmark is potentially helpful to verify and gauge the relationship between pay and performance. Nonetheless, such an index also helps to foster the impression that a strong relationship between pay and firm performance (e.g., a "high" score of more than "1.00" in Meredith's calculations) is desirable in all cases. However, under certain circumstances, a highly leveraged pay package may actually motivate executives to take actions that are contrary to the interest of shareholders.

At one extreme, CEO wealth might fluctuate dollar-for-dollar with changes in shareholder wealth. One significant problem with this type of compensation contract is that CEOs tend to hold a relatively undiversified portfolio, certainly much less diversified than that of their shareholders. For example, CEO compensation, reputation in the managerial labor market, and equity holdings are highly correlated with changes in the performance of the firm. Increasing pay-for-performance in the compensation contract beyond a certain level may result in executives becoming excessively conservative. That is, they may decide to pursue "safe" strategies that offer a low expected return to protect the portion of their pay at risk. In addition, strengthening the link between pay and performance could prompt executives to devote resources to costly diversification programs that are not beneficial to shareholders.

HOW MUCH IS TOO MUCH?

One final point is that neither cross-sectional nor firm-specific analyses provide much information about the appropriate level of executive pay. Without a well-defined compensation benchmark that assigns a value to CEO output, it is difficult--perhaps impossible--to make across-the-board statements about CEO pay.

Cross-sectional analysis may be helpful in predicting the level of a CEO's pay relative to other CEOs. However, it can be that all CEOs are substantially overpaid--or conversely, underpaid. Similarly, the analysis of a single firm over a certain period of time can help to gauge the relationship between pay and performance but does not help to resolve broader pay questions.

There are likely to be instances where the level of executive pay is difficult to justify. Even so, the recent findings of many compensation analysts are open to question.

We conclude there is a strong pay-for-performance component in CEO compensation and that it would be inappropriate to regulate or legislate the level or structure of pay packages.

At the very least, we should endeavor to understand more completely any flaws in existing executive pay systems before we adopt compensation arrangements with a questionable impact on shareholder wealth.

Directed by Edward H. Bowman, the Reginald H. Jones Center of the Wharton School focuses on "CEO concerns, "drawing on diverse faculty of the University of Pennsylvania to study leadership and strategic issues.

David F. Larcker is Ernst & Young Professor of Accounting at The Wharton School, University of Pennsylvania. His research examines various issues in executive compensation and corporate strategy.

Richard G. Sloan is Goldman Sachs Assistant Professor of Accounting at The Wharton School. His current research focuses on compensation and financial statement analysis.