Corporate Performance and Compensation Objectives

1. Introduction

Furthermore, our analysis delves into the behavioral implications of performance goals, revealing potential managerial myopia around “jump points” in the pay-performance relationship [1]. Managers may be incentivized to take short-term actions to push reported performance just beyond these goals, potentially sacrificing long-term value creation. Additionally, we uncover nuances in the relationship between performance goals and reported firm performance, considering factors such as the type of metrics employed and the nature of the payout structure [2–4].

By examining the distribution of reported performance around these goals, we identify significant discontinuities, indicating a clustering of performance just above the target level [3]. This clustering effect is particularly pronounced for grants tied to single metrics, suggesting a strategic focus on narrowly meeting specific targets. Moreover, we find evidence suggesting that managers may adjust their behavior to avoid falling short of performance targets, potentially due to concerns about forced turnover or target ratcheting effects [4].

Our analysis also extends to the examination of firm behavior in meeting accounting performance goals. We find that firms narrowly exceeding EPS goals tend to exhibit higher abnormal accruals, indicating possible earnings management practices [5]. Similarly, firms exceeding profit goals by a small margin demonstrate lower SG&A expenses, suggestive of cost-cutting measures to meet targets [6].

Overall, our findings underscore the complexities inherent in linking executive compensation to explicit performance goals and highlight the need for careful consideration of the behavioral and reporting implications associated with such practices.

2. Literature Review

Our study aligns closely with previous research examining how executives navigate performance management to either meet or exceed predetermined performance targets. Notable studies in this realm include [7], which explores the zero EPS goal, and [3, 8], which delves into consensus analyst estimates. While existing literature demonstrates executives’ manipulation of earnings to sidestep losses or satisfy analyst predictions, our contribution lies in unveiling their manipulation to fulfill compensation plan objectives. Our findings underscore the drawbacks of explicit performance targets when executives wield influence over reported performance metrics.

Moreover, we distinguish ourselves by not only acknowledging the financial repercussions managers face for falling short of performance benchmarks, thus enabling more precise cross-sectional analyses, but also by shedding light on the pivotal role performance goals play in forecasting actual firm performance. Our research also intersects with the body of work investigating optimal contracts in settings where agents can manipulate observable performance metrics. For instance, [9] demonstrate that compensation contracts based on reported earnings fail to incentivize managers to maximize profits while accurately reporting them.

Similarly, [10] explore a principal-agent scenario where information distortion emerges as an equilibrium behavior due to the agent’s private knowledge of production costs. Additionally, [11, 12] reveal equilibria wherein irregularities in reported earnings distributions arise endogenously. Within the accounting and finance domain, numerous studies document the correlation between performance and performance goals. [12] illustrate how firms resort to share repurchases to manipulate EPS and meet bonus targets, while [13] highlight cost-cutting strategies to enhance reported margins and avert losses.

Furthermore, [14] expose CEOs’ willingness to sacrifice long-term value for earnings stability, while Bergstresser and Philippon (2006) establish a link between discretionary accrual usage and stock-based pay. In contrast, our study reveals how firms bolster accruals and reduce discretionary spending to achieve meticulously outlined performance goals embedded in compensation contracts.

Additionally, our work contributes to the discourse on the utilization of performance provisions in executive compensation. [15, 16] examines short-term accounting-based performance goals, emphasizing the connection between accruals and managerial income incentives. [16] distinguishes between internally and externally determined performance standards, while explore the implications of performance-vesting provisions in executive stock and option grants, showcasing their positive impact on subsequent operating performance.

Moreover, investigate performance grants tied to relative performance, albeit finding a weak correlation with peer group performance. Unlike these studies, our focus lies on the role of performance provisions incentivizing managed reported performance. Additionally, our research contributes to the evaluation of executive performance metrics, echoing advocacy for metrics reflective of CEO effort and hypothesis regarding metric selection in response to past performance.

Furthermore, our study holds implications for the ongoing policy-oriented discourse surrounding executive compensation. With large investors and proxy advisory firms advocating for explicit performance goals, our findings underscore the necessity for meticulous design and stringent board oversight to curb executive manipulation of reported performance to meet these goals.

3. Methodological Approach

In this section, we outline the procedures employed to detect potential manipulation of firm performance in order to meet predefined goals. All tests conducted aim to identify discrepancies within the reported performance distribution.

The initial test we apply follows the methodology outlined, designed specifically to uncover discontinuities within a density at a given point. To execute this test, we create variables measuring the disparity between actual performance and stated goals and examine for discontinuities at the goal threshold. The process entails two steps. Firstly, we generate a finely gridded histogram of the underlying variable, meticulously defining bins to ensure they encapsulate either side of zero distinctly. Subsequently, we smooth the histogram by conducting weighted regressions separately on each side of zero. Here, the midpoint of each bin serves as the regressor, while the normalized counts of observations within each bin constitute the outcome variable. A triangular kernel weighting function is utilized, assigning greater weight to bins proximate to the point of interest. The discontinuity test is then executed as a Wald test, assessing the null hypothesis of zero discontinuity. This test is implemented utilizing the “DCdensity” function in STATA, yielding both the initial histogram and the smoothed histogram, accompanied by 95% confidence intervals (CI) of the smoothed density. Crucial parameters in this test include the bin size for the initial histogram and the bandwidth employed in the subsequent estimation. For our analysis, default values for these parameters are adopted as suggested by the “DCdensity” function.

The second test, adapted from [5], serves not only as a robustness check on McCrary’s [13] test but also facilitates the detection of discontinuities across the entire density. Similar to McCrary’s approach, this test involves data binning, smooth density estimation, and comparison of observed versus expected counts. The optimal bin size for the initial histogram is determined to minimize mean-square error and is calculated as \(0.7764 \times 1.364 \times min [\sigma, Q 1.34] \times n^(-1/5)\), where \(\sigma\) represents the standard deviation, Q denotes the interquartile range, and n signifies the total number of observations. In the subsequent stage, a Gaussian kernel is employed for density estimation with bandwidth set equal to the bin size from the first stage. This test utilizes an estimate of sampling variation in the histogram to ascertain whether observed counts significantly deviate from those expected under the null hypothesis of a smooth distribution.

An inherent limitation of the aforementioned tests is the inability to compare the magnitudes of discontinuities at different points or across densities. To address this, we complement these tests with a bootstrapping exercise and regression-based analyses. In the bootstrapping exercise, we repeatedly draw random samples from the variable of interest and compare the means of observations lying in bins adjacent to zero. Cross-sectional tests are conducted separately for single and multiple metric-based grants to compare differences in magnitude. Regression-based tests are described in detail below.

4. Hypotheses

In this section, we delineate hypotheses pertinent to our context, outlining anticipated predictions. If there is a surge in managerial compensation upon achieving a performance goal, managers may endeavor to manipulate reported performance slightly beyond the goal threshold, especially if they perceive actual performance to fall near, yet slightly short of the target. However, in scenarios where the Pay-Performance Relationship (PPR) exhibits concavity at the performance goal, managers lack incentives to surpass the goal. Both scenarios are likely to yield a disproportionate number of firms reporting performance either at or marginally above the goal [5].

Firm performance may converge around the target value specified in grants for reasons unrelated to direct payout considerations. Managers may refrain from significantly exceeding the target to avoid subsequent period target escalation (termed “target ratcheting effect”). Additionally, board scrutiny may focus on target performance, penalizing underperformance through means other than reduced bonuses. Consequently, we anticipate a higher proportion of firms reporting performance slightly above the goal compared to those falling slightly short, forming our first prediction.

An important aspect of our analysis involves grants contingent on either a single metric or multiple metrics. Given the positive correlation among metrics, managers face challenges in narrowly surpassing all metrics simultaneously. Hence, we expect greater clustering in performance for grants tied to a single metric, leading to a larger discontinuity in the density of underlying performance compared to grants contingent on multiple metrics, our second prediction.

The slope of the PPR at the performance goal also influences performance clustering. We anticipate clustering around concave kinks rather than convex ones, as slower pay increases beyond concave kinks incentivize performance clustering. However, a concave kink may not necessarily result in clustering just above the target, as agents may aim slightly above to ensure goal attainment amidst uncertainty, rather than precisely hitting the target, forming our third prediction.

Furthermore, grants denominated in cash versus equity introduce convexity in the PPR due to the positive relationship between stock price and performance metric. Consequently, we expect a greater discontinuity at the performance goal for cash payouts compared to equity payouts.

Additionally, we anticipate targets in one period to positively correlate with performance in the previous period, consistent with the “target ratcheting effect.” Firms nearing the target are expected to have a higher likelihood of surpassing subsequent targets, akin to findings by [6].

Finally, depending on the metric involved, managers may employ various means to meet goals, such as manipulating accruals, cutting discretionary expenditures, or repurchasing shares. By comparing the levels of various financial indicators between firms slightly exceeding versus missing the goal, we aim to assess the extent of performance manipulation and elucidate potential adverse implications of such pay contracts.

5. Empirical Analysis

A. Full Sample Analysis

Panel A illustrates the histogram of the deviation between actual and target performance alongside a smoothed density. The histogram, with a bin width of 0.029 as recommended by STATA, reveals clustering around zero, skewed towards the left. Notably, the smoothed density exhibits a mode to the left of zero. Panel B presents the output from the “DCdensity” function, indicating the density of the actual minus target performance. The x-axis delineates the performance difference, with the vertical line representing the goal. The solid line represents the fitted density function, while thinner lines denote a 95% confidence interval. The presence of a discontinuity at zero (p-value = 0.01) suggests a significant number of firms reporting performance slightly exceeding the target, compared to those falling short.

In Panel C, a test for discontinuity at points other than zero is conducted, revealing significant t-values primarily to the right of zero, consistent with a higher number of observations exceeding the goal. Further, a bootstrapping exercise confirms this observation, indicating a statistically significant difference between observations just to the right and left of zero.

Separate tests for discontinuities at zero are performed for various metrics, with statistically significant results obtained only for EPS goals. However, bootstrapping demonstrates a larger discontinuity for EPS goals compared to profit and sales goals, indicating greater clustering around EPS goals.

To address potential selective reporting of targets, additional tests are conducted on firms reporting targets consistently across contiguous years, yielding robust results.

Extends the analysis to test for discontinuities at zero for “Actual less threshold,” revealing a significant discontinuity (p-value = 0.018) akin to the target level results.

B. Subsample Analysis

Focuses on “Actual less target,” dividing the sample into subsamples based on single versus multiple metrics. A larger discontinuity is observed for single metric grants, consistent with the hypothesis of greater performance clustering.

Conducts cross-sectional tests based on the concavity or convexity of the Pay-Performance Relationship (PPR) at the target. A significant discontinuity at zero is identified only for concave PPRs, indicating more pronounced clustering.

The sample is segmented into non-equity and equity-based grants, revealing a discontinuity only for non-equity grants, consistent with the graphical results.

C. Regression Analysis

A regression analysis is performed to statistically compare discontinuity sizes and accommodate multiple points of density discontinuity. The model estimates the expected number of firms in each bin, incorporating a fourth-order polynomial of the bin midpoint and controlling for metric groups and year fixed effects. Results confirm the presence of discontinuities, with additional advantages of metric-specific distributions and cross-sectional tests.

D. Relative Performance-Based Awards

The focus shifts to relative performance-based awards to examine whether firms with such awards tend to meet these targets precisely or fall short. These tests serve as both indicators of firms’ tendencies to beat relative performance goals and falsification tests for the main hypothesis. If firms manage reported performance to beat goals, this tendency should be less prevalent for grants tied to relative performance due to the unpredictability of peer group outcomes until after the performance period ends.

Panel A illustrates the histogram of the difference between actual and target ranks, showing no clear tendency for firms to meet or exceed their performance target. [12] test for a discontinuity at zero, finding no statistically significant evidence of such a discontinuity. These results contrast with the previous evidence of firms’ tendencies to meet absolute performance targets precisely, suggesting that firms do not exhibit a similar tendency when targets are based on relative performance. Together, these findings support the notion of firms managing reported performance to meet absolute performance targets but not relative ones.

E. Strategies to Exceed Performance Goals

In the following tests, firms that just exceed a manager’s compensation goal are compared to those that just miss a goal on various dimensions to understand how firms ultimately exceed performance goals in practice. These comparisons shed light on the extent to which firms manage accruals and discretionary expenditures to manipulate reported performance.

Panel A focuses on EPS goals, revealing similarities between firms that exceed and miss their EPS goals on most observable characteristics. Surprisingly, firms that exceed their EPS goals engage in less share repurchase and experience smaller changes in R&D expenditures compared to firms that miss their EPS goals.

Panel B compares firms that just exceed and just miss their sales goals, with the former exhibiting a higher sales growth rate but no other significant differences observed.

Panel C examines profit goals, revealing that firms exceeding their profit goals are larger, have lower market-to-book ratios, lower sales growth, and smaller changes in SG&A compared to firms that miss their profit goals. The smaller change in SG&A for firms that exceed their profit goals aligns with prior research indicating that firms often decrease discretionary spending to boost short-term earnings.

Study conducts multivariate tests to further compare firms that exceed and miss their performance goals, offering deeper insights into the strategies employed to achieve performance targets.

6. Conclusion

Our study reveals compelling evidence that executives manipulate reported performance to meet compensation goals, with a disproportionate number of firms exceeding targets rather than falling short. This behavior is particularly pronounced for earnings-based goals, single metric grants, and those with concave pay-performance relationships. However, relative performance goals show no such trend, suggesting nuanced responses to different metrics. These findings underscore the need for cautious design of performance-based compensation, favoring relative metrics and smoother performance-pay relationships to minimize distortions.