Rater Bias and Incentive Provision

• Published date: 01 October 2015
• Author: Philipp Weinschenk,Daniel Müller
• DOI: http://doi.org/10.1111/jems.12118
• Date: 01 October 2015

Rater Bias and Incentive Provision∗

DANIEL M¨

ULLER

University of Bonn

Adenauerallee 24 D-53113 Bonn, Germany

daniel.mueller@uni-bonn.de

PHILIPP WEINSCHENK

Technical Universityof Kaiserslautern

Gottlieb-Daimler-Straße D-67663

Kaiserslautern Germany; and

Max Planck Institute for Research on Collective Goods

Kurt-Schumacher-Str. 10 D-53113 Bonn, Germany

weinschenk@coll.mpg.de

The tendency of supervisors to judge an employee as either good or bad and then to seek out

evidence supporting that earlier established opinion is regarded as one of the major problems

of performance appraisal. We investigate the implications of this rater bias in a dynamic moral

hazard model with a wealth-constrained agent. Although rater bias weakens the agent’sincentives

to exert effort in late periods, at the same time it strengthens implicit incentives in early periods.

Under the optimal contract, as long as rater bias is not overly strong, its adverse effect on late-

period incentives is fully offset by exploitation of stronger early-period incentives and thereby

leaves the principal’s profits unchanged.

1. Introduction

A widespread instrument for incentive provision within firms and organizations is the

implementation of a merit pay plan in combination with some kind of performance ap-

praisal process. Contingent pay is seen as a fair and equitable way to reward employees

differentially for their different contributions and—by linking pay to performance—to

motivate employees to perform better.1Many firms, however, rely (at least to some ex-

tent) on performance measures which are subjective rather than objective in nature, that

is, where an employee’s pay is at the discretion of the impressions of a supervisor.2This

makes performance appraisal a process by which humans judge other humans, thereby

opening the door for behavioral biases and inferential errors to enter and to distort

this process. Rater bias in general is considered a severe problem because “[b]iased and

∗In preparing this paper, we have benefited from comments made by Oliver Hart, Fabian Herweg, Matthias

Kr¨

akel, Patrick Schmitz, and Fritz Strack. We also thank the conference audiences at the Nordic Conference

on Behavioral and Experimental Economics (Helsinki, 2010) and at the IMEBE (Barcelona, 2011) as well as the

seminar participants at the University of Marburg and the University of W¨

urzburg for helpful comments.

1. For textbook expositions on the basic functions of and problems with performance appraisal, see, for

example, Porter et al. (2008) or Dessler (2008).

2. For example, Murphy and Oyer (2003) find that 67% of 280 U.S. firms display discretion in the allocation

of a bonus pool across participants. Likewise, as reported in Baker et al. (1994), half a worker’s pay at Lincoln

Electric comes froma bonus based on subjective aspects of performance, for example, the worker’s cooperation,

innovation, or dependability. For a comprehensive review of the theoretical and empirical literature on the

rationales for and the problems associated with the use of subjective performance measures, see Prendergast

(1999).

C2015 Wiley Periodicals, Inc.

Journal of Economics & Management Strategy, Volume24, Number 4, Winter 2015, 833–862

834 Journal of Economics & Management Strategy

inaccurate performance evaluation reduces productivity by reducing the effectiveness

of incentives in the organization” (Baker et al., 1988, p. 608). 3

In this paper, we formally inquire into the implications for incentive provision

and optimal contract design when the evaluation of an employee’s performance is di-

rectionally influenced by that employee’s past appraisal records.4More specifically, we

model the tendency of supervisors to judge an employee’s performance as either good

or bad and then to reaffirm that previously established opinion in subsequent ratings of

later performance. The cognitive processes that have been hypothesized to drive such

spillover or assimilation effects in appraisal patterns are manifold. For example, accord-

ing to the priming hypothesis, the use of cognitive categories that reflect the quality of

work behavior in organizing one’s perception of previous performance increases the

probability of using the same categories again when evaluating present performance.

Likewise, though operating at a different level of information processing, hypothesis

confirmation bias is conjectured to lead individuals to seek out information that con-

firms their early impressions, and to discount disconfirmatory information.5

These spillover effects—by whatever cognitive process they are brought about—

are commonly acknowledged to lead to a distortion of the appraisal process which

constitutes a source of employee frustration, in particular for those employees who

are discriminated against, and in turn deteriorates incentive provision and firm perfor-

mance.

Toallow for past performance records to prejudice current performance evaluation,

we set up a moral hazard framework in which a risk-neutral agent, who has no resources

of his own, decides in two consecutive periods how much effort to exert. 6In each period,

the agent’s unobservable effort decision influences the realization of a performance

measure. Although the performance measure itself is (at least partially) subjective and

therefore cannot be contracted upon, the supervisor’s reports are verifiable and can be

used for contracting purposes. The supervisor’s perception of the performance measure,

however, does not necessarily coincide with its actual realization: whenever the agent’s

second-period performance does not match his first-period performance record, the

supervisor tends to misperceive the agent’s performance in the second period as identical

to his first-period performance. In what follows, we refer to this stylized notion of current

appraisal being prejudiced by past performance records plainly as rater bias.

In contrast to the prevalent opinion in human resource management, we find that

the implications of rater bias for incentive provision and overall performance are by

far not as dire as common wisdom suggests. Admittedly, rater bias makes provision of

second-period incentives more difficult: the value of high effort for an employee who

has received an unfavorable first evaluation is diminished by rater bias because a second

poor evaluation becomes more likely even when he puts much devotion into his task;

likewise, an employee who was rated favorably during the first period does not consider

exertion of high effort to be very necessary,because due to rater bias a second favorable

3. Likewise, according to Murphy (1992), “[t]he success of companywide incentive compensation pro-

grams therefore depends on managers’ effectiveness in monitoring and appraising performance.” (p. 37)

4. Our paper is thus closely related to the literature on biased ratings and incentive provision. See, for

example, Prendergast and Topel(1996), Prendergast (2002), and Giebe and G ¨

urtler (2012) on the implications

of favoritism and the leniency bias for incentive design.

5. See, for example, Murphy et al. (1985) for an overview of these and alternative cognitive processes

resulting in assimilation effects in performance evaluation.

6. For other papers considering limited liability of a risk-neutral agent in a repeated moral hazard frame-

work, see, for example, Cr´

emer (1995), Che and Yoo (2001), Schmitz (2005a, 2005b, 2013), Ohlendorf and

Schmitz (2012), and Kr¨

akel and Sch¨

ottner (2010).