Incentive Contracting and Value-Relevance of

Earnings and Cash Flows

Rajiv Banker*

Rong Huang†

Ram Natarajan**

*TempleUniversity

†CityUniversity of New York-BaruchCollege

** The University of Texas at Dallas

June 11, 2008

We thank seminar participants at Temple University, City University of New York—Baruch College, The University of Texas at Dallas, and the American Accounting Association 2006 Annual meeting for helpful comments and suggestions.

Abstract

Accounting performance measures such as earnings and cash flows are useful for both valuation and performance evaluation purposes. However, little evidence exists on whether there is any association between these two roles. In this study, we provide large sample empirical evidence that the value-relevance of earnings explains a significant amount of the cross-sectional variation in the pay-sensitivity of earnings and the incremental value-relevance of cash flows explains variation in the marginal pay-sensitivity of cash flows. We document that while both value-relevance and compensation weight on earnings decline from the sub period of 1993 to 1997 to the sub period of 1998 to 2003, both value-relevance and compensation weight on cash flows increase from the earlier sub period to the later sub period. Overall, our results provide additional evidence that value-relevance of a performance measure plays a significant role in its use for performance evaluation.

Keywords: Incentive contracting, value-relevance, earnings, cash flows, executive compensation

I.Introduction

This study examines the association between pay-sensitivity and value-relevance of earnings and cash flows. Accounting performance measures such as earnings and cash flows serve a variety of purposes in organizations and markets, including valuation and performance evaluation. While a number of prior studies have examined the valuation roles and incentive-contracting roles of accounting performance measures separately, very little evidence exists on whether there is any association between these two roles. Using CEO compensation and accounting data for a large number of U.S. firms, we examine the empiricalassociation between value and incentive relevance of earnings and cashflows over an eleven year period from 1993 to 2003. The graph in Figure I shows that the relative compensation weight on cash flows versus earnings tracks remarkably closely to their relative valuation relevance for our sample period from 1993 to 2003. The correlation between the relative compensation weight and the relative valuation relevance is as high as 0.87. When we analyze the data more formally to control for other influential factors at the firm level, the results confirm that compensation weight on each of earnings and cash flows is higher for firms that exhibit high levels of value-relevance for the performance measure. The evidence also indicates that both value-relevance and compensation weight for earnings decline from the sub period of 1993-1997 to the sub period of 1998-2003, while both the value-relevance and the compensation weight for cash flows increase from the earlier sub period to the later sub period.

Gjesdal (1981) puts forth the basic premise that the incentive-informativeness of performance measures that determines their compensation weights may be different from their valuation-informativeness. He considers differential uses of accounting information in organizations and shows that the ranking of information systems for valuation purposes need not align with the ranking of those information systems for control purposes. Lambert (1993) expands on this issue by remarking that valuing the firm is not the same as evaluating the manager’s contribution to the value of the firm. The observations made by Gjesdal (1981) and Lambert (1993) are based on single-action, single-period settings where optimal compensation contracts assign lower (higher) weights on performance measures that have lower (higher) sensitivity-to-noise ratio (Banker and Datar 1989).

The implication of the above observations for empirical accounting researchers is that the association between value-relevance and incentive-informativeness of performance measures is context-specific and varies across agencies. However, it appears that these implications may have been misunderstood as indicating instead that there is no association between value-relevance and incentive-informativeness. A notable exception is Bushman, Engel and Smith (2006) who examine linkages between the weight placed on earnings in compensation contracts and the weight placed on earnings in stock price formation. They show that valuation earnings coefficients and compensation earnings coefficients are positively associated and point out that further theoretical and empirical investigation of this association remains an interesting challenge for future research.[1] Our main motivation behind this study stems from a similar desire to provide evidence on the positive association between the valuation roles and incentive contracting roles in settings where multiple accounting performance measures are used to evaluate top managers.

We consider two primary accounting performance measures, earnings and cash flows, for the purposes of this study. We focus on these performance measures for several reasons. First, since cash flows is a component of earnings, the research setting lends itself to examining incremental value-relevance and marginal pay-performance sensitivity of cash flows when cash flows information is available in addition to earnings for valuation and performance evaluation purposes. Second, a number of studies have established that earnings and cash flows have differential implications for firm value (Rayburn 1986, Bowen, Burgstahler and Daley 1986, Ali 1994, Sloan 1996). Prior studies have also shown that the incremental value of cash flows over earnings varies cross-sectionally depending on factors such as the persistence of earnings and cash flows (Sloan 1996, Xie 2001, Richardson, Sloan, Soliman and Tuna 2005), the time interval over which performance is measured, the volatility of the firm’s working capital requirements, and the length of the firm’s operating cycle (Dechow 1994, Dechow, Kothari and Watts 1998). Third, studies examining the stewardship value of components of earnings have found that there is significant cross-sectional variation in the way cash flows and earnings are used in determining top management compensation (Natarajan 1996, Nwaeze, Yang and Yin 2006). Fourth, very limited empirical evidence exists on whether the incremental value-relevance and the marginal pay-performance sensitivity of cash flows have changed over time since cash flow information was first made available to shareholders in 1987 through SFAS 95.

To understand the structural factors influencing the context-specific nature of the above-mentioned association, we first derive pay-sensitivities and value-relevance measures using a highly stylized principal-agent setting characterized by two performance measures.[2] The two performance measures are each modeled as consisting of a distinct managerial effort component, a common pay-off relevant noise term and a specific non value-relevant noise component. We show that the context-specific nature of the associationbetween valuation weights and compensation weights is critically dependent on the cross-sectional differences in the variances of the pay-off relevant noise and idiosyncratic noise of the performance measures under consideration. The insights provided by our stylized model are used to generate empirical proxies of the variances of the relevant noise terms from the firm-specific variance-covariance matrix of earnings and cash flows and to explicitly quantify the co-movement of the theoretical, endogenously determined, valuation and compensation weights at various deciles of the cross-section of a large sample of Compustat firms. The analysis based on estimated values of the variances of the relevant noise terms suggests an expected positive association between compensation weights and valuation weights when earnings and cash flows are the performance measures under consideration.

We formally test this prediction using actual compensation, valuation and performance measure data. We use a sample of 7,076 CEO-years spanning the period 1993 to 2003 in our empirical analysis. In the first stage of our analysis, we estimate value-relevance of earnings and incremental value-relevance of cash flows for each firm-year by using a time-series of 8 to 10 years of past data on earnings, cash flows, stock prices and adapting the metrics suggested in prior literature (Ohlson 1995, Collins, Maydew and Weiss 1997, Barth, Beaver, Hand and Landsman 1999, Engel, Hayes and Wang 2003, Bushman, Chen, Engel and Smith 2003) to our context. In the second stage, we estimate cross-sectional yearly regressions that use CEO and firm level data on cash compensation, earnings, cash flows, value-relevance of earnings and cash flows, as well as a variety of control variables that have been identified in prior literature as determinants of cross-sectional variation in pay-performance sensitivity of earnings and cash flows. The regression coefficients from the second stage regression enable us to estimate the average magnitude of the association between value-relevance and pay-performance sensitivity during our sample period for both earnings and cash flows.

The empirical results support our predictions. The estimated association between pay-earnings sensitivity and the value-relevance is significantly positive for both earnings and cash flows. We evaluate the robustness of our findings by considering cash flows as the primary performance measure and earnings as the supplementary performance measure, by using total compensation instead of cash compensation and by employing changes rather than levels in earnings and cash flows as performance measures. These results also support our predictions.

Our study contributes to a stream of research that has examined the association between the valuation and performance evaluation roles of accounting performance measures. Two notable studies that belong to this stream are Bushman, Engel and Smith (2006) and Engel, Hayes and Wang (2003). In contrast to Bushman, Engel and Smith (2006) who empirically examine the association between the valuation and incentive contracting role of accounting earnings, we focus on a pair of correlated accounting performance measures namely, accounting earnings and cash flows. Our empirical results confirm Bushman et al.’s (2006) findings that higher the value-relevance of earnings, higher the pay-sensitivity of earnings. More importantly, we alsoprovide evidence that higher the incremental value-relevance of cash flows, higher the incremental pay-sensitivity of cash flows.

Engel, Hayes and Wang (2003) employ a research design somewhat similar to ours to examine the association between CEO turnover probability and accounting earnings. They predict that CEO turnover probability is decreasing in the timelines of earnings and find empirical evidence consistent with their predictions. “Timeliness” of earnings is measured through its association with contemporaneous stock returns.

There are however, some significant differences between Engel, Hayes and Wang (2003) and our study. Our study focuses on the compensation decision while Engel, Hayes and Wang (2003) examine turnover decisions. The empirical model in Engel, Hayes and Wang (2003) is similar to Sloan (1993) and Lambert and Larcker (1987) in that the primary focus is on the relative weights on accounting earnings and stock returns for performance evaluation. In contrast, we focus on the association between valuation and performance evaluation roles of two accounting performance measures e.g., earnings and cash flows. The empirical analysis in Engel, Hayes and Wang (2003) is based on the predictions of a two-period model where current earnings, by construction, reflect only a portion of current period managerial effort, and which predicts that relative weight on earnings increases with timeliness of current period earnings. In contrast, our analytical characterization underlines the influence of cross-sectional differences in the variances of pay-off relevant noise and idiosyncratic noise of performance measures on the association between compensation and valuation weights of performance measures and does not make any directional predictions for this association.

To summarize, our study contributes in three ways to the existing literature on the use of accounting performance measures in valuation and performance evaluation. First, it confirms and quantifies the positive association between value-relevance and pay-performance sensitivity for earnings and cash flows. Second, it quantifies the decline in value-relevance and pay-performance sensitivity of earnings and corresponding increase in value-relevance and pay-performance sensitivity of cash flows over the past decade. Third, it provides evidence that value–relevance of performance measures plays a significant role in the choice of accounting performance measures for performance evaluation in organizations.

The remainder of this paper proceeds as follows. Section II develops the main hypotheses. Section III discusses the research design and sample selection. Section IV presents the empirical results. Finally, section V concludes the paper.

II.Theory Development

Lambert’s (1993) remark that valuing the firm is not the same as evaluating the manager’s contribution to the value of the firm is intuitively appealing but provides limited guidance to empirical researchers on the similarities and differences in the way value-relevance measures and pay-sensitivities are influenced by the underlying agency and performance measure characteristics. Since both the pay-sensitivities and value-relevance measures are endogenously determined and are functions of the characteristics that differ across the agencies in the cross-section, it is important to understand the impact of the variation of the various characteristics on the association between these endogenous variables in the cross-section. Changes in some of the characteristics result in changes in valuation and contracting weights that are of the same sign while changes in other characteristics may have opposing effects on these weights. The stylized model that we develop in this section and the analysis of performance measure data to quantify the insights from the model are oriented towards understanding the dominant characteristics that drive the cross-sectional distribution of the set of agencies (firms) under consideration.

We formally consider a simple, single period, two-action, two-signal principal-agent setting based on the LEN (Linear contract, Negative exponential utility and Normally distributed random variables) framework to better understand the factors influencing the association between value-relevance measures and pay-sensitivities of performance measures. Complete details of the model are provided in the Appendix I and we briefly describe the setting here.

The setting we consider involves a risk-averse manager who is in charge of two distinct productive activities. These productive activities along with factors beyond the control of the manager determine the unobservable outcome and generate two observable, contractible, performance measures. Each performance measure is driven by one of the productive activities while the outcome or value to the principal is influenced by both activities. The two performance measures also contain a common as well as a specific random component. From a valuation point of view, the performance measures are useful because the common component they contain is a key component of the outcome. Market participants, therefore, use these signals to update their beliefs about the outcome. From a contracting point of view, the performance measures are useful because they are informative about the agent’s unobservable activities. Without loss of generality, we normalize the risk-aversion coefficient and the sensitivities of the performance measures to managerial actions to unity to focus on a minimal set of parameters that vary across agencies.[3]

While the setting we consider is highly stylized, it captures some necessary elements of a contracting environment where accounting performance measures such as earnings and cash flows are key determinants of executive compensation. The two performance measures have a positive correlation by construction which is representative of the empirically documented positive correlation between earnings and cash flows (Dechow 1994). Productive managerial activities which may differentially impact cash flows and earnings ultimately contribute to increases in firm value and this essential aspect is also captured in the way the unobservable outcome is modeled. Investor beliefs about firm value get updated on the release of information about realizations of earnings and cash flows (Rayburn 1986, Bowen, Burgstahler, and Daley 1986) and we operationalize this by modeling the random components of the performance measures as garbled versions of the random component of the outcome. Our simplified set up, however, does not capture the effect of prior period productive actions on current period earnings and cash flows as well as the effect of earnings management practices on reported accounting numbers.

In this model, we can completely characterize the pay-sensitivities and value-relevance metrics of the performance measures in terms of the elements of the variance-covariance matrix of the performance measures. The elements of the variance-covariance matrix are functions of two specific variances and a common variance.[4] In our model, different agencies (firms) are characterized by different values for the three-component, positive-valued, variance vector.