Testing for Distortions in Performance Measures: An Application to Residual Income‐Based Measures like Economic Value Added

• Date: 01 March 2015
• DOI: http://doi.org/10.1111/jems.12080
• Author: Mirjam Praag,Randolph Sloof
• Published date: 01 March 2015

Testing for Distortions in Performance Measures:

An Application to Residual Income-Based Measures

like Economic Value Added

RANDOLPH SLOOF

Amsterdam School of Economics

University of Amsterdam

Amsterdam, The Netherlands

r.sloof@uva.nl

MIRJAM VAN PRAAG

Department of Innovation and Organizational Economics

Copenhagen Business School

Copenhagen, Denmark

mvp.ino@cbs.dk

Distorted performance measures in compensation contracts elicit suboptimal behavioral responses

that may even prove to be dysfunctional (gaming). This paper applies the empirical test developed

by Courty and Marschke (Review of Economics and Statistics, 90, 428–441) to detect whether

the widely used class of residual income-based performance measures—such as economic value

added (EVA)—is distorted, leading to unintended agent behavior. The paper uses a difference-in-

differences approach to account for changes in economic circumstances and the self-selection of

firms using EVA. Our findings indicate that EVA is a distorted performance measure that elicits

the gaming response.

1. Introduction

Understanding the incentive implications of performance measures used in remuner-

ation contracts has become vital since the rapid introduction of performance-based

compensation schemes in the last three decades. The classic principal-agent model ac-

knowledges that the noise of a given performance measure determines its suitability

for use in compensation contracts. The “noisier” the measure, the lower is the optimal

incentive intensity. However, as Baker (2002) points out, the critical issue in most in-

centive contracts may not be the noisiness of the performance measure, but rather its

“distortion.” Baker defines distortion inversely as the extent to which the effect of effort

on measured performance is aligned with the effect of effort on the firm’s objective

function.1

If a performance measure is “distorted,” then agents can increase their performance

outcomes in two ways: by engaging in productive activities that are intended by the

principal or by engaging in unintended actions that are easier or cheaper from the

We gratefully acknowledge the researchassistance of Dennis Schoenmakers and comments by George Baker

and Luis Garicano. The usual disclaimer applies. We thank the (associate) editor and an anonymous referee

for their excellent suggestions.

1. Theoreticalmultitasking models show an inverse relationship between the distortion of the performance

measure and the efficient incentive intensity (see Holmstromand Milgrom, 1991; Feltham and Xie, 1994; Datar

et al., 2001; Baker, 2002).

C2015 Wiley Periodicals, Inc.

Journal of Economics & Management Strategy, Volume24, Number 1, Spring 2015, 74–91

Testing for Distortions in Performance Measures 75

agent’s perspective. Performance pay based on a distorted performance measure will

thus motivate agents to put costly effort into “cheap” activities (increasing measured

performance) to the possible detriment of organizational value. Such unproductive

efforts are typically referred to as “gaming efforts” (see Courty and Marschke, 2008).

Performance measures are often assessed and chosen based on their correlation

with firm value (see Baker, 2002; Courty and Marschke, 2008). However, some measures

that seem to be informative ex ante about the performance of the company become less

informative when used for incentive purposes, revealing their distortion. Implementing

a compensation scheme based on such distorted measures undermines the association

between the measure and the value of the company. Put differently, the mere use of a

distorted performance measure in compensation degrades its quality as a measure of

intended performance. The degree of (potential) degradation depends on the extent to

which the measure is distorted. Thus, the suitability of a performance measure cannot be

measured by the ex ante correlation between the measure and the value of the company,

as this correlation disregards the distortion not yet revealed.

An example involves the distortion of the performance measure “client satisfac-

tion” when it is used in compensation schemes for sales employees. Client satisfac-

tion, used as a performance measure, may provide valuable information about a sales

employee’s contribution to company performance. Using the measure as a basis for

performance pay, however, creates an incentive for a sales employee to increase client

satisfaction through “cheap” effort: selling at low prices, for example, or even giving

products away for free. The performance measure “client satisfaction” thus becomes

degraded: it is less useful than originally expected as a basis for performance pay.

Courty and Marschke (2008) developed an empirical test to detect distortion. Their

test, based on a simple theoretical framework, assesses how a performance measure

degrades when it is first used in a compensation contract or when its use in such a

contract is intensified. The focus is on measuring the change in the association between

the performance measure and organizational value due to an increase in the weight

of the measure in the compensation scheme. A negative change indicates degradation

of the quality of the measure, revealing a distortion. Courty and Marschke (CM) apply

their test to performance measures introduced in the course of a natural experiment

among agencies managing the US Governmental Job Training Partnership Act. They

find weak support for the distortion of these measures.

This paper applies the Courty and Marschke (2008) approach to test for distor-

tions in the performance measure residual income (RI), a key indicator of corporate

performance. More specifically, we collected a sample of firms that have introduced the

RI-based performance measure economic value added (EVA) in the remuneration con-

tract of executive board members. EVA was developed and copyrighted by Stern Stewart

& Co. Unlike RI,EVA cannot be measured solely on the basis of firms’ accounting data.

It differs from RI due to some—for researchers opaque—discretionary and standard

adjustments. In our empirical analysis we measure RI for the firms in our sample and

refertothisastheaccounting-based value of EVA.2

Unlike CM, the application in this paper does not concern a natural experiment,

but reviews the experience of listed (US-based) firms that introduced this popular

performance measure to reward management performance, mostly in the mid-1990s.

2. Wewill discuss the differences between RI and EVA,and the possible consequences of these differences

for our analysis, in Section 3. Using EVA data obtained from Stern Stewart is not possible in our application

(see Section 3).