Vous êtes sur la page 1sur 19

APJHR_45_2_Kavanagh.

qxd

132

29/05/2007

9:58 PM

Asia Pacific Journal of Human Resources

Page 132

2007 45(2)

Understanding performance appraisal fairness


Peter Kavanagh*
University of Melbourne, Australia
John Benson
University of South Australia, Adelaide, Australia
Michelle Brown
University of Melbourne, Australia

Employee reactions to the performance appraisal (PA) process have been


identified as a potentially important influence on employee acceptance of the
performance appraisal process. One such reaction is the perceived fairness of the
performance appraisal experience. Previous studies have tended to focus on single
aspects of the PA process that impact on PA fairness. In this paper, we combine
findings from previous research on PA fairness into a single model and then
investigate their impact using data from 2377 public sector employees. We find
that participation in PA, attitudes towards the supervisor, and knowledge of the
performance appraisal process are all positively and significantly associated with
employee perceptions of performance appraisal fairness.
Keywords: employee reactions, fairness, performance appraisal

Performance appraisal (PA) involves measuring job performance (Murphy


and Cleveland 1995, 1). This view captures an essential element of the PA
process without specifying the actual techniques used for measurement. In
addition, the reference to measurement points to the preoccupation of much of
Correspondence to: Associate Professor Michelle Brown, Department of Management and
Marketing, Faculty of Economics and Commerce, University of Melbourne, Parkville, Vic.
3010, Australia; fax: +613 9348 4293; e-mail: brownm@unimelb.edu.au
* This research was funded by grants received from the Faculty of Economics and
Commerce, University of Melbourne. We would like to thank management of PSR and
the officials of the Community and Public Sector Union for their ongoing support of the
project. We would also like to thank PSR staff for their willingness to participate in the
research.
Asia Pacific Journal of Human Resources. Published by SAGE Publications (Los Angeles, London, New Delhi and
Singapore; www.sagepublications.com) on behalf of the Australian Human Resources Institute. Copyright 2007
Australian Human Resources Institute. Volume 45(2): 132150. [1038-4111] DOI: 10.1177/1038411107079108.

APJHR_45_2_Kavanagh.qxd

29/05/2007

9:58 PM

Page 133

Understanding performance appraisal fairness

the research in PA. Evans and McShane (1988, 177) argued that the focus of PA
research has been on identifying the conditions which affect the validity,
reliability and other characteristics of employee performance measurement.
This focus has resulted in a lack of attention to what have been described as
reaction criteria (Murphy and Cleveland 1995), though, as Cawley, Keeping
and Levy (1998, 615) point out, in order for appraisal systems to be effective
they need to be accepted and supported by its employees. The value of such
an approach is reflected by recent findings that PA reactions impact on
employee attitudes such as job satisfaction and organisational commitment
(Blau 1999; Levy and Williams 1998; Pettijohn, Pettijohn and dAmico 2001).
One employee reaction to PA that has been the focus of a number of
studies is that of perceived fairness of the PA process (Dobbins, Cardy and
Platz-Vieno 1990; Evans and McShane 1988; Fulk, Brief and Barr 1985;
Greenberg 1986; Landy, Barnes and Murphy 1978; Landy, Barnes-Farrell and
Cleveland 1980; Taylor et al. 1995). Research in this area was encouraged by
the finding of Lawler (1967) that employee beliefs about the fairness of a PA
system were an important influence on the ultimate success of any PA system
because perceived fairness was linked to confidence in and hence, acceptance
of the PA system.
Despite the importance of PA fairness and the attention given to it in the
literature, there have been few attempts to develop and test a range of PA
process factors that affect an employees perception of performance appraisal
fairness. This omission is significant as understanding the correlates of
performance appraisal fairness will assist in the design of effective performance appraisal systems. There are many dimensions of a performance appraisal
process and, using data from 2377 Australian public sector employees, we
investigate the role of employee participation in PA, attitudes towards the
supervisor and knowledge of the PA system in promoting employees perceptions of performance appraisal fairness.
The paper commences with a review of the relevant literature and the
presentation of hypotheses. The next section outlines the methods adopted and
operationalises the theoretical model. This is followed by a section detailing
the findings of the research, and the theoretical and practical implications of
the results. A short conclusion completes the paper.
Performance appraisal fairness: Background and hypotheses
A range of factors might be identified to explain employee perceptions of
performance appraisal fairness. From a policy and practice perspective it is
useful to focus on the processes of performance appraisal as these can be
managed by organisations. An organisation can change the rules and affect the
operation of a performance appraisal system, through for example, the training
they provide to supervisors who conduct appraisals.

133

APJHR_45_2_Kavanagh.qxd

134

29/05/2007

9:58 PM

Asia Pacific Journal of Human Resources

Page 134

2007 45(2)

Theoretical support for a focus on PA processes can be found in process


control theory and social exchange theory. Process control theory suggests that
fairness perceptions are driven by the level of control that individuals are able
to exercise over processes that determine outcomes (Thibaut and Walker 1975).
Control plays a key role in shaping peoples views about the fairness of the
procedures and individuals view procedures as most fair when control is
vested in the participants (Konovsky 2000, 493). In other words, the more
control employees have of the PA process the more they will consider the
processes to be fair. Process control theory states that control of the process is
important to people for instrumental reasons: that control of the process will
result in outcomes that are more favourable to the individual. This provides
the justification for two aspects of our PA processes model: participation in PA
and knowledge of the PA system. Participation provides an opportunity to
influence performance targets while knowledge is important in exercising
control over processes.
Social exchange theory refers to relationships that entail unspecified
future obligations. Like economic exchange, social exchange generates an
expectation of some future return for contributions though the exact nature
of that return is unspecified (Konovsky 2000, 493). In these circumstances fair
treatment of employees by management is particularly important. Fair
treatment by management demonstrates respect for the dignity of employees
and removes fear of exploitation. This provides the rationale for another aspect
of our PA processes model: attitudes towards the supervisor. PA fairness not
only is a function of the control employees have in the process but will also be
influenced by their need to be treated fairly by a key player in the performance appraisal process, namely their supervisor. These three aspects of PA
processes (participation in objective setting, attitudes towards the supervisor
and knowledge of the PA process) are presented in figure 1. We now turn to
an evaluation of the empirical research on each of these three aspects and the
development of hypotheses.
The first aspect of the PA process model outlined in figure 1 is participation in the PA process. This was identified as a positive influence on
PA fairness by Greenberg (1986), Evans and McShane (1988), Dobbins, Cardy
and Platz-Vieno (1990) and Cawley, Keeping and Levy (1998). Dobbins,
Cardy and Platz-Vieno (1990) reported that appraisal fairness was found to
have strong positive correlations with the level of two-way communication.
Fairness perceptions are also enhanced by two-way communications as they
are useful in plotting employees progress towards their performance objectives and provide the employees with an opportunity to raise issues that
influence their ability to achieve their performance objectives. Moreover, twoway communications provide an opportunity to clarify the rules of the game:
knowing the rules of the game allows employees to make choices about how
to operate within that system (Mani 2002).
Participation in the setting of performance objectives provides a means

APJHR_45_2_Kavanagh.qxd

29/05/2007

9:58 PM

Page 135

Understanding performance appraisal fairness

Figure 1

Model of performance appraisal fairness

PA process
Participation

Level of two-way communication

Involvement in setting of
objectives
Supervisor

Supervisor neutrality
Knowledge of PA system

Clarity of PA

Understanding of objectives

Acceptance of objectives

Situational characteristics
Role conflict
Role overload
Resource inadequacy
Work overload
PA rating

Personal characteristics
Age
Gender
Salary level
Negative affectivity

Performance appraisal
fairness

through which employees are able to exercise some control over the process.
Participation can ensure that the right number and type of objectives are set for
the employee. In the absence of employee participation supervisors may impose
objectives on their subordinates without regard to the complexities of the job
or the span of control of the employees (Lewis 1998). Further, an excessive
number of performance objectives can result in conflict between those objectives (Lewis 1998) and make it difficult for the employee to achieve a good
performance rating. This suggests that higher levels of participation in the
appraisal process should result in higher levels of perceived PA fairness.
Accordingly it is hypothesised that:

135

APJHR_45_2_Kavanagh.qxd

136

29/05/2007

9:58 PM

Asia Pacific Journal of Human Resources

Page 136

2007 45(2)

Hypothesis 1: Higher levels of two-way communication in the PA process


are related to higher levels of PA fairness.
Hypothesis 2: Higher levels of involvement in the setting of objectives as
part of the PA process are related to higher levels of PA fairness.

The second aspect of the PA process (figure 1) involves the supervisor.


Supervisors are pivotal in PA as they are generally responsible for setting
performance objectives, providing formal and informal feedback, and the
overall rating of the employees performance. Uses of the performance rating
vary but include determining the extent of any subsequent pay rise, promotion
or dismissal. So an employees current and future employment prospects can
be heavily influenced by their supervisors evaluation of their performance.
A number of studies have demonstrated a link between attitudes towards
the supervisor and perceptions of PA fairness. The findings of Landy, Barnes
and Murphy (1978) and Greenberg (1986) identified the ability of a supervisor
to make an accurate evaluation of subordinates performance as an important
influence on fairness perceptions. Fulk, Brief and Barr (1985) identified a belief
that open communication with the supervisor would not result in negative
repercussions as important in promoting perceptions of PA fairness (Fulk,
Brief and Barr 1985, 302). Greenberg (1986, 340) used an open-ended questionnaire to discover what individuals considered as the most decisive factor in
a particularly fair or unfair performance evaluation. The most important
factor, out of a final list of five, was supervisory consistency in the application
of standards. Cumulatively, these studies suggest that fairness perceptions are
more likely to be enhanced when the supervisor is seen to be neutral: that is
when an individual employee is treated without bias (Konovsky 2000, 494). It
is therefore hypothesised that:
Hypothesis 3: Higher levels of supervisor neutrality are related to higher
levels of PA fairness.

The third aspect of the PA process (figure 1) involves employee


knowledge of the PA process. Levy and Williams (1998) found that knowledge
of the PA system was a significant and positive influence on fairness perceptions. Levy and Williams (1998) examined the relationship between perceived
system knowledge (PSK) and PA fairness in two separate studies conducted
with bank employees. In both studies PSK referred to an understanding of the
objectives and operation of the appraisal system as well as the overall goal of
the PA process. The first study found that where appraisees believed they
understood the appraisal system they were more likely to judge the system as
fair (Levy and Williams 1998, 62). This finding was supported by the second
study, which used a longitudinal design. Evans and McShane (1988) provided
confirmation of the importance of agreement on the duties to be evaluated

APJHR_45_2_Kavanagh.qxd

29/05/2007

9:58 PM

Page 137

Understanding performance appraisal fairness

(Landy, Barnes and Murphy 1978) as influences on PA fairness. Higher levels


of agreement on each of these factors were associated with higher levels of PA
fairness.
Knowledge of the PA system can be seen as consisting of a number of
elements: clarity about the role of appraisals, understanding of performance
objectives and acceptance of those objectives. Each of these three dimensions of
knowledge add to an employees feelings of process control: employees are
aware of why the appraisal is taking place, what they are required to do in order
to be successful in the appraisal, and the consequences of the appraisal. There
will be no surprises for the employee during the appraisal cycle, which is likely
to contribute to perceptions of PA fairness. It is therefore hypothesised that:
Hypothesis 4: Higher levels of clarity of the PA system are related to higher
levels of PA fairness.
Hypothesis 5: Higher levels of understanding of the PA system are related
to higher levels of PA fairness.
Hypothesis 6: Higher levels of acceptance of objectives under the PA system
are related to higher levels of PA fairness.

Method
Setting and subjects

The setting for this study was an Australian public sector, scientific research
organisation (PSR). At the time of the research, PSR received about 70 per cent
of its funding from government with the balance of funds derived from
commercial activities. PSRs primary roles are to carry out scientific research,
to assist Australian industry and further the interests of the Australian
community, and to encourage and facilitate the application and use of scientific
research. The organisational structure revolves around three principal areas:
agribusiness; information, manufacturing and minerals; and sustainable energy
and the environment.
At the time of the research, PSR employed 6957 employees at over 50 sites
in a range of occupations including tradespersons, technicians, clerical, managerial and scientific research workers. The use of a single organisation such
as PSR, as the site for research, meant that all respondents were subject to the
same organisational environment and worked under a standard set of employment conditions and human resource policies.
All PSR employees are covered by an enterprise agreement, which provides
for a salary scale of nine levels. The levels relate to particular job classifications.
Levels 7 to 9 cover senior management, levels 5 and 6 are research scientists and
level 4 and below cover clerical and research support staff. Each of the pay levels
has a salary range attached to it, which is generally the midpoint plus or minus
12 per cent. This range is broken into several increments.

137

APJHR_45_2_Kavanagh.qxd

138

29/05/2007

9:58 PM

Asia Pacific Journal of Human Resources

Page 138

2007 45(2)

Employees are usually appointed to the bottom of the range for a particular salary level. Progression through the increments depends on the
outcome of the performance, planning and evaluation system (PPE). The
outcome assessment is based on the achievement of objectives, which are
jointly decided by the supervisor and the employee. The employees performance in respect of these objectives is then monitored throughout the year. At
the conclusion of the 12 months evaluation period the supervisor makes an
assessment of the employee using a 5-point scale. A rating of three or better
will result in the employee being granted a salary increment. The pay system
does not provide for downward movement through the increments.
The data were collected by a survey of the employees of PSR conducted
between December 1998 and March 1999. With the support of management
and the union the researchers were able to forward the survey to all 6957
employees using PSRs internal mail system. Completed questionnaires were
returned directly to the researchers at their university using the reply-paid mail
system. This approach ensured the confidentiality of the respondents and the
independence of the researchers. Completed questionnaires were received
from 3335 employees resulting in an overall response rate of 47.9 per cent. The
mailing lists used for the survey were based on demographic information
concerning the workforce, which had been supplied by PSR. As such it
was possible to compare the sample and the population by gender and
geographical location. Using t-tests, no significant differences between the
sample and the population were found on these criteria. After allowing for
missing data the effective sample size for this research was 2377.
Measures

All the measures used in this study were based on employee self-report, fivepoint scales unless otherwise indicated. Definitions, items and descriptive
statistics for all variables used in this study are set out in table 1. A reliability
analysis was carried out for all multi-item scales (Cronbach 1951) and this
figure is also reported in table 1. Given the relatively high reliability scores of
these scales they were all included in the model.1 With the exception of the
measures for Involvement, Neutrality and Accept all scales were taken
from previous studies where their psychometric properties had been established. Confirmatory factor analysis for the existing scales confirmed that the
items making up these scales loaded on one factor only (eigenvalue > 1) with
the minimum factor loading for each scale ranging from 0.63 to 0.91.

1 The alpha coefficient for three scales (Roleover, Understand and Involvement) was below
the 0.7 level which is usually considered the lower limit for inclusion of a scale (Nunnally
1978). Some further investigations were carried out to understand their impact on the
model. The results indicated that the inclusion of these variables did not have a significant
impact on the regression analyses reported in table 3.

APJHR_45_2_Kavanagh.qxd

Table 1

29/05/2007

9:58 PM

Page 139

Variable definitions

Label

Definition

Mean

SD

3.58

0.92

42.20

9.55

Dependent variable
PA fairness

Fairness of performance, planning and evaluation system as measured by


six items from Tang and Sarsfield-Baldwin (1996), alpha = .95. A five-point
scale where 5 = very fair. Items were My last performance rating truly
represented how well I performed in my job; My last performance appraisal
was fair; My performance was accurately evaluated; My supervisor was
justified in his/her rating of my performance; My rating was free from bias
and If I had been evaluating my own performance I would have given the
same rating as my supervisor. Five-point scale where 5 = very fair.

Personal characteristics
Age

Age in years.

Gender

Female = 1, male = 0.

0.33

0.47

Salary

Level of salary classification where 1 = lowest level, 9 = highest level.

4.82

1.75

Negative
affect

The extent to which an individual experiences aversive emotional states as


measured by three items from Watson, Pennebaker and Folger (1987),
alpha = .86. A five-point scale with 5 = high negative affectivity.

2.81

0.89

Work setting characteristics


Role conflict The degree to which an employees role expectations are incompatible as
measured by three items from Kahn et al. (1964), alpha = .74. A five-point
scale where 5 = high role conflict.

2.78

0.79

Role
overload

The extent of role overload as measured by three items from Iverson,


Deery and Erwin (1995), alpha = .62. A five-point scale where 1 = high
role overload.

3.72

0.61

Resource
inadequacy

The degree to which an employee perceives resources are adequate to


perform their jobs as measured by three items from Iverson (1992),
alpha = .80. A five-point scale where 5 = high resource inadequacy.

2.76

0.88

Work
overload

The extent of job overload as measured by four items from Iverson (1992),
alpha = .81. A five-point scale where 5 = high level of work overload.

3.30

0.81

Rating

Performance appraisal rating in the previous assessment cycle


(1 to 5, with 5 as top).

3.61

0.75

Two way

The degree of two-way communication under performance, planning and


evaluation system as measured by six items from Tang and Sarsfield Baldwin
(1996), alpha = .83. A five-point scale with 5 = high two way communication.

2.49

0.85

Involvement

Involvement in the setting of objectives as measured by four items from a


self-devised scale. Alpha = .61. A five-point scale with 5 = high levels of
involvement.

3.65

0.69

Neutrality

The degree of supervisor neutrality as measured by a two item scale


covering supervisor favouritism and employee equality. Self devised scale,
alpha = .81. A five-point scale with 5 = high levels of neutrality.

3.45

0.97

Clarity

The degree of clarity of the performance, planning and evaluation system as


measured by three items from Tang and Sarsfield Baldwin (1996), alpha = .86.
A five-point scale with 5 = high knowledge.
2.82

1.20

Understand

The degree of approachability of the supervisor as measured by three items


derived from Tang and Sarsfield Baldwin (1996), alpha = .68. A five-point
scale where 5 = high level of understanding.

3.71

0.80

Accept

The degree to which employees accept the objectives set in the performance
appraisal process. A three item self-devised scale, alpha = .73. A five-point
scale where 5 = a high level of acceptance of the objectives.
3.61

0.66

PA process

N = 2377

APJHR_45_2_Kavanagh.qxd

140

29/05/2007

9:58 PM

Asia Pacific Journal of Human Resources

Page 140

2007 45(2)

Three scales (Involvement,Neutrality and Accept) were specifically


developed for this research. In this case the scales were pre-tested using data
from a pilot survey of 20 retired PSR employees. This process was followed
by confirmatory factor analysis using the full data set. In all cases the analysis
generated only one factor when the minimum eigenvalue was set at 1.
Moreover, in all cases the eigenvalue would have to be lowered to under 0.25
for a second factor to be generated. For the scale Involvement the factor
loadings on the four items all exceeded 0.40. Similarly, for Neutrality the two
items loaded on one factor with all values exceeding 0.75. Again a similar
situation appeared with Accept with the three items loading only one factor
being generated by the analysis and all factor loadings exceeding 0.62.
Dependent variable

The dependent variable PA fairness was measured by a scale created by Tang


and Sarsfield-Baldwin (1996). The scale consisted of six items focusing on the
perceived degree of fairness of the process of evaluating performance under
the PPE system. A sample item is If I had been evaluating my own performance I would have given the same rating as my supervisor (see table 1 for
items). Factor analysis was performed on this scale. Each of the items loaded
on one factor only (eigenvalue > 1) and all values were in excess of 0.82. The
scale therefore possesses both convergent and discriminant validity. The alpha
coefficient of this scale was 0.95.2
Independent variables

Six PA process variables were included in the explanatory model as outlined


in the previous section. Two way is a six-item scale developed by Tang and
Sarsfield-Baldwin (1996) that measures the level of two-way communication
that took place during the PPE process. A sample item is How much input

2 This finding is consistent with that of Dobbins, Cardy and Platz-Vieno (1990). These
writers used a similar scale to measure what they described as PA satisfaction. Concerned
at the possibility that the scale was measuring two constructs: fairness and satisfaction,
Dobbins, Cardy and Platz-Vieno (1990, 625) undertook further analysis of the scale.
They found strong evidence of its unidimensionality and elected to retain the term PA
satisfaction for the reaction they examined. This finding suggests, however, that the
distinction between the reactions of fairness and satisfaction is artificial. The term PA
fairness would have been equally justified for this reaction. Indeed, Organ (1988) argued
that satisfaction judgements are in essence fairness judgements. Fairness would also
appear to have been more suitable because it reflected the focus of the scale and avoided
confusion with those studies of PA satisfaction, which have focused exclusively on
satisfaction without reference to fairness and/or accuracy (Giles and Mossholder 1990;
Dipboye and de Pontbriand 1981).

APJHR_45_2_Kavanagh.qxd

29/05/2007

9:58 PM

Page 141

Understanding performance appraisal fairness

does your supervisor ask for during the evaluation cycle? Involvement is a
self-devised, four-item scale that measures the level of involvement in the
performance objectives set as part of the PPE process. A sample item is I was
able to participate effectively in the setting of objectives. Neutrality is a selfdevised, two-item scale that measures the level of neutrality displayed by the
manager. A sample item is My manager is serious about treating all employees
equally. Clarity, is a three-item scale taken from Tang and Sarsfield-Baldwin
(1996) that measures the clarity of the various uses of the PPE system. A sample
item is When you took up your current position how clear was it that your
performance would be periodically evaluated? Understand is a three-item
scale derived from the work of Tang and Sarsfield-Baldwin (1996) that
measures the employees level of understanding of the PPE system. A sample
item is I understand the performance appraisal process at PSR. Accept is a
three-item self-devised scale that measures the employees acceptance of the
performance objectives used in the PPE process. A sample item is My objectives are achievable.
Control variables

The nine control variables were divided into two groups: personal characteristics and situational factors. Four personal characteristics were included in the
model: age, gender, salary and negative affect. Age has been linked to
attitudes to PA with younger workers tending to take a more positive view of
the appraisal process (Cleveland and Landy 1981; Ferris et al. 1985). In the
analysis, age is a continuous variable. The influence of gender on attitudes
toward the PA process is unclear. Dobbins, Cardy and Truxillo (1986) found
that gender was an influence in the ratings students gave to professors. Two
studies conducted in work settings, however, did not find gender effects (Shore
and Thornton 1986; Thompson and Thompson 1985).
Salary indicates the position of the employee in the nine-level salary
structure used at PSR. As noted earlier, the pay levels are based around job
classifications within the organisation. The inclusion of this variable in the
model thus acts as a control for the potential influence of different job characteristics on the reaction to the appraisal process. As employees in different
pay classifications may experience the PA process differently due to variations
in the skills and knowledge of both the appraiser and appraisee, it was considered appropriate to control for this possibility. Negative affectivity (Negative
affect) refers to a broad range of aversive mood states of employees (Watson
and Pennebaker 1989, 2345). As a consequence, individuals with high
negative affect tend to overemphasise the negative side of life experiences. In
the PA context, Lam, Yik and Schaubroeck (2002, 196) found that negative
affectivity moderated the attitudinal effects of positive performance feedback
on employees reactions. Negative affect was measured by a three-item scale
developed by Watson, Pennebarker and Folger (1987).

141

APJHR_45_2_Kavanagh.qxd

142

29/05/2007

9:58 PM

Asia Pacific Journal of Human Resources

Page 142

2007 45(2)

Five situational characteristics were selected for inclusion in the model.


These are role conflict (conflict), role overload (role overload), resource inadequacy (resource inadequacy), and work overload (work overload). Landy,
Barnes and Murphy (1978, 753) found that agreement between the supervisor
and the subordinate on job duties was a significant influence on judgements
regarding the fairness of PA evaluation. This finding is supported by the work
of Folger, Konovsky and Cropanzano (1992) who argued that agreement
between the appraiser and appraisee over the standards to be applied in judging
performance is critical to enhancing beliefs about the accuracy and fairness of
the PA system. Each of the situational characteristics pertains to the likelihood
of agreement between the parties to the appraisal over the duties of the
appraised. It is argued that a lack of agreement is likely to exist when an
employee believes that their role is not clearly defined, their duties are in conflict
with one another or the role or tasks required of them is too demanding. The
inclusion of these variables is designed to control for this effect.
The final situational control variable is the last performance rating for the
employee. The performance rating an employee receives could potentially
affect their view of the fairness of PA process so we have included it as a
control: a higher rating being more likely to generate perceptions of PA
fairness. The influence of PA outcome on judgements of fairness has been the
subject of debate in the literature. Landy, Barnes-Farrell and Cleveland (1980)
examined the potential influence of the PA rating on the formation of fairness
judgments. They found that variance in the perception of the fairness of evaluation could not be explained by rating level (Landy, Barnes-Farrell and
Cleveland 1980, 356). Fulk, Brief and Barr (1985) provided confirmation of
these findings. Alternatively, Dobbins, Cardy and Platz-Vieno (1990), Evans
and McShane (1988), and Greenberg (1986) found that the PA outcome is an
important influence on fairness perceptions. PA outcome consists of a single
item representing the performance rating achieved at the last PA with a 5
representing the highest available rating and 1 the lowest. This data was
provided by our respondents. Previous research has identified self-report as an
acceptable method of collecting this information because employees are able
to accurately report their most recent appraisal (Levy and Williams 1998, 57).
Method of analysis

The measure of fairness is composed of interval scale data and so the most
appropriate estimation method is the ordinary least squares (OLS) regression
technique.3 The size and sign of the co-efficient of each of the PA process

3 Fairness scores ranged from 1.00 to 5.00 with 25 different values. As these scores are
approximating interval data, OLS was selected because this technique is robust enough
to handle this type of data (Berry and Feldman 1985).

APJHR_45_2_Kavanagh.qxd

29/05/2007

9:58 PM

Page 143

Understanding performance appraisal fairness

variables was the major focus of the analysis. To identify the influence of each
of the groups of variables on fairness judgements, hierarchical regression
analysis was undertaken for each of the models. Model 1 included personal
characteristics and situational factors; model 2 included all the model 1
variables with the addition of the participation variables, while model 3 built
on model 2 with the addition of the supervisor variable. Finally, model 4
included all of the personal, situational and PA process variables.
The presentation of the results includes the regression co-efficient and the
standard error for each of the independent variables. Significance levels are
reported at the .10, the .05 and .01 levels. This latter level is particularly
important as with a large sample size the adoption of a more rigorous requirement for statistical significance will ensure a higher likelihood of practical
significance. The adjusted R2, change in adjusted R2, and the overall F-test for
each of the models are also reported.
Regression analysis may encounter two problems that undermine the reliability of the estimates. The first of these problems is the degrees of freedom
deficit problem. This problem is caused by a high number of independent
variables relative to the sample size. In this study this is not an issue as the
number of variables is 15 and the sample size is 2377. The second problem that
may arise is that of multicollinearity which can be caused by common method
variance or by strong correlations between predictor variables in a model.
Common method variance may occur in survey work when the dependent
variable and some or all of the predictor variables are composed of self-report
attitude measures (Campbell and Fiske 1959; Fiske 1982).
To reduce the likelihood of common method bias, positively and negatively worded items were used in each of the multi-item scales (Gordon and
Ladd 1990). Method variance marker variable analysis (Lindell and Whitney
2001) was also undertaken. Although of a post-questionnaire design, the
marker variable (financial referent when determining satisfaction with pay)
had high reliability (alpha coefficient = .85), was placed between fairness and
the majority of independent variables, and had a non-significant correlation
coefficient of .07 (p > .05) which supported the theoretical proposition that
there is no relationship between these variables. Moreover, to the extent that
common method variance did influence the results it will have inflated all
variables to a similar extent (Campbell and Fiske 1959).
The second cause of multicollinearity is the relationship between the independent and control variables. Studenmund and Cassidy (1987) suggested that
multicollinearity is unlikely to be a problem if the correlations between
predictor variables are less than .8. This is clearly the case in this study (see
table 2). However, given that many of the correlations were statistically significant, further examination using the methods suggested by Belsley, Kuh and
Wulsch (1980) was undertaken. The Variance Inflation Factor for all variables
in the model was less than 1.5. This is well below the considered acceptable
threshold level of 10 (Neter, Wasserman and Kutner 1985). These results

143

.40

.38

.46

.29

.49

.38

Two-way

Involvement

Neutrality

Clarity

Understanding

Accept

.01

.07

.06

.01

.08

.11

.07

.09

.02

.04

.06

.06

.03

.02

.17

.13

.05

.08

.01

.14

.06

.01

.33

.07

.07

.34

.06

.01

.02

.03

1.00

.25

.23

.10

.25

.16

.14

.07

.32

.28

.26

.29

1.00

p < .05 (1tailed) for correlations greater than or equal to .04

.39

Rating

.40
.03

.35

.29

.09

.33

.26

.16

.09

.26

.27

.24

1.00

.32

.22

.06

.17

.20

.09

.08

.19

.23

1.00

.35

.27

.19

.26

.20

.17

.10

.29

1.00

.33

.08

.06

.12

.03

.08

.07

1.00

.14

.18

.10

.14

.17

.12

1.00

10

.24

.47

.34

.38

.19

1.00

11

.45

.45

.20

.30

1.00

12

.30

.48

.21

1.00

13

.20

.32

1.00

14

.40

1.00

15

1.00

16

Asia Pacific Journal of Human Resources

.08

.00

.01

.01

.04

1.00

9:58 PM

N = 2377

.26

.12

Role overload

.14

.30

Role conflict

Work overload

.19

Negative affect

.48

.21

1.00

29/05/2007

Resource
inadequacy

.05

Salary

.40

Age

.05

1.00

PA fairness

Gender

Variables

Correlation matrix

144

Table 2

APJHR_45_2_Kavanagh.qxd
Page 144

2007 45(2)

APJHR_45_2_Kavanagh.qxd

29/05/2007

9:58 PM

Page 145

Understanding performance appraisal fairness

suggest that multicollinearity is not a threat to the substantive conclusions


drawn from the parameter estimates (Hair et al. 1999).
Results
This study confirmed that the way in which performance appraisals are
conducted have a significant impact on the perceived fairness of the appraisal
system, as demonstrated by the results in table 3.
The PA process variables were grouped under three headings: participation in the appraisal process, attitude towards the supervisor, and knowledge
of the appraisal system. The first aspect of the PA process focused on identifying the influence of participation in the PA process on fairness perceptions.
These variables were the extent of two-way communication and the level of
employee involvement in the setting of objectives. Each of these variables was
found to have a significant and positive effect on PA fairness across all specifications of the model (2, 3 and 4), so hypotheses 1 and 2 were supported.
The second aspect of PA process focused on the employees supervisor.
The analysis indicated that the perceived neutrality in the conduct of the PA
by the supervisor had, as predicted (hypothesis 3), a significant positive effect
on PA fairness. The final aspect of the PA process examined was the extent of
knowledge of the appraisal system. These variables included the clarity of the
PA system (hypothesis 4), the level of understanding of the PA system (hypothesis 5) and the degree of acceptance of the objectives to be pursued as part of
the PA system (hypothesis 6). Each of these variables was found to have a
positive and significant impact on perceived fairness.
We were concerned the significant results were being driven by the large
sample size, so we took a random sample of 20 per cent and then ran the
analyses again. The results were consistent with the full sample. We therefore
can have some confidence that the results presented in the paper reflect the
underlying relationships in the data and are not merely a function of the
response size.4

Discussion and conclusions


This study has investigated the role of a range of PA processes in promoting
employee perceptions of PA fairness. Earlier research has demonstrated the
importance of PA fairness to a range of outcomes and evaluated the impact of
individual PA process variables in promoting PA fairness. We integrated the
findings from these previous studies by combining them into a single model to
test the individual and combined effect of a range of PA process variables. This
4 The results are available from the authors upon request.

145

APJHR_45_2_Kavanagh.qxd

29/05/2007

9:58 PM

Page 146

Asia Pacific Journal of Human Resources

146

Table 3

2007 45(2)

Performance appraisal fairness

Variable

Model 1

Model 2

Model 3

Model 4

Control variables: Personal characteristics


Age

.0037*
(.00)

.0013
(.00)

.0007
(.00)

.0001
(.00)

Gender

.0854**
(.04)

.0796**
(.03)

.0652*
(.03)

.0575*
(.03)

Salary

.0505***
(.01)

.0087
(.01)

.0120
(.01)

.0059
(.00)

Negative affect

.0606***
(.02)

.0248
(.02)

.0024
(.02)

.0049
(.02)

Control variables: Situational characteristics


Role conflict

.2303***
(.02)

.1460***
(.02)

.0957***
(.02)

.0782***
(.02)

Role overload

.0409
(.03)

.0752***
(.02)

.0847***
(.02)

.1146***
(.02)

Resource inadequacy

.1403***
(.02)

.0771***
(.02)

.0540***
(.02)

.0223
(.02)

Work overload

.0852***
(.02)

.0982***
(.02)

.1025***
(.02)

.0781***
(.02)

Performance rating

.4385***
(.02)

.3794***
(.02)

.3657***
(.02)

.3502***
(.02)

Two-way communications

.3068***
(.02)

.2349***
(.02)

.1520***
(.02)

Involvement

.3094***
(.03)

.2568***
(.03)

.1507***

PA process: participation

PA process: Supervisor

(.03)

Supervisory neutrality

.2236***
(.02)

.1815***
(.02)

PA process: Knowledge of performance appraisal


Clarity

.0541***
(.01)

Understanding

.1844***
(.02)

Acceptance

.1166***
(.02)

Constant
Adjusted R2

3.5123***

1.5004***

.2547

.3835

.4241

.4510

.1288

.0406

.0269

adjusted R2
Mean vif

1.27

F test
N = 2377

146.92
*p < .10

**p < .05

***p < .01.

.9467***

.4131**

1.30

1.32

1.43

135.37***

146.80***

131.12***

Note: standard errors in parentheses.

APJHR_45_2_Kavanagh.qxd

29/05/2007

9:58 PM

Page 147

Understanding performance appraisal fairness

study provides strong evidence that the use of particular practices within a PA
system will enhance fairness perceptions. Participation in the appraisal process,
a belief in the neutrality of the supervisor, and knowledge of the appraisal
system all had a positive influence on the PA fairness. Also, the addition of
further PA process variables added to the overall perceptions of PA fairness,
suggesting that these practices complement rather than work against each
other. The findings provide strong evidence that the goal of designing PA
systems that are perceived as fair is achievable as the practices included in the
present study are within the control of management.
These findings have important implications for the many contemporary
organisations which have formal appraisal systems. By incorporating the identified practices into their PA system, organisations are likely to have employees
who view the PA system as fair. Two consequences flow from this. First, as
PA fairness has been linked to the acceptance of decisions based on the PA
process (Lawler 1967), it is possible that decisions such as pay adjustments,
promotions, terminations and training requirements will be met with higher
levels of acceptance of these decisions thereby enhancing the effectiveness of
the PA system. Second, the higher level of PA fairness resulting from such
practices is likely to result in more positive organisational attitudes such as job
satisfaction and organisational commitment (Blau 1999; Levy and Williams
1998; Pettijohn, Pettijohn and dAmico 2001). Both of these outcomes would
help to justify the investment required to adequately train both appraisers and
appraisees in the procedures required to enhance the perceived fairness of the
performance appraisal system.
Future researchers should consider extending the model to include the
interpersonal aspects of performance appraisal, such as the treatment of individuals with dignity and respect (Bies and Moag 1986). Research by Cobb, Vest
and Hills (1997) found that the way in which procedures are enacted is as
important in influencing fairness judgments as the procedures themselves.
This suggests that efforts to improve the perceived fairness of the PA system
need to be broadened to include practices consistent with fair interpersonal
treatment. Such practices might include the punctuality, tone of voice and
attentiveness of the appraiser.
Despite the contributions of the present study, it is subject to limitations.
First, the cross-sectional character of the study leaves the direction of causality
open to interpretation. The results provide strong support for the view that
the identified aspects of the PA process cause PA fairness. Nevertheless, alternative readings of the relationship are possible. For example, it is plausible that
PA fairness generates higher levels of communication and information. Thus,
those employees who display confidence in the fairness of the PA system may,
for example, be given greater opportunity to obtain information and opportunities to participate in the process by their supervisor. As has been demonstrated, however, there is strong theoretical and empirical evidence to support
the approach taken in this paper.

147

APJHR_45_2_Kavanagh.qxd

148

29/05/2007

9:58 PM

Asia Pacific Journal of Human Resources

Page 148

2007 45(2)

Second, care must be taken in generalising these findings to other settings.


The sample consisted of employees primarily involved in white-collar tasks in
a public sector organisation. In addition, the sample contained a significant
proportion of tertiary educated employees; some 30 per cent of employees
possessed PhDs. On a broader level, these results come from an Australian
organisation and while broadly consistent with findings from US studies may
not be applicable to a different cultural setting. Further, as already indicated,
contextual factors seem to play an important part in influencing fairness
perceptions. Nevertheless, the large sample size (N = 2377) and the strength of
the findings suggest that although the specific influences on perceptions of
fairness may be different elsewhere, the influences on PA fairness identified
in this study should not be ignored.

Peter Kavanagh (MCom (Hons), Melb.) has a special interest in the application of HR practices in
educational settings with a particular focus on the use of performance appraisals to manage the
performance of school teachers.
John Benson is professor and head of the School of Management at the University of South Australia.
Prior to taking up his present appointment John was chair of the MBA Programme in International
Business at the University of Tsukuba, Japan after having spend many years as a reader in the
Department of Management, University of Melbourne, Australia. His major research interests are
Japanese management and unions, the restructuring of Chinese industry, outsourcing, and knowledge
work. John has published numerous papers and monographs and his most recent work is
Unemployment in Asia with Dr Ying Zhu.
Michelle Brown is an associate professor in the Department of Management and Marketing at the
University of Melbourne. Previous appointments have been with the University of Wisconsin, Milwaukee
and Victoria University. She has research interests in labour market adjustments, employee participation
and performance management systems. Recent work in these areas has appeared in various
international journals. She is co-author, with John S. Heywood, of Paying for performance: An
international comparison.

References
Belsley, D., E. Kuh, and R. Wulsch. 1980. Regression diagnostics. New York: John Wiley.
Berry, W., and S. Feldman. 1985. Multiple regression in practice. Beverly Hills, CA: Sage.
Bies, R., and J. Moag. 1986. Interactional Justice; communication criteria of fairness. In
Research on negotiation in organisations, eds R. Lewicki, B. Sheppard, and B. Bazerman,
4355. Greenwich, CT: JAI Press.
Blau, G. 1999. Testing the longitudinal impact of work variables and performance appraisal
satisfaction on subsequent overall job satisfaction. Human Relations 52(8): 109913.
Campbell, D., and D. Fiske. 1959. Convergent and discriminant validation by the multitraitmultimethod matrix. Psychological Bulletin 56(2): 81105.

APJHR_45_2_Kavanagh.qxd

29/05/2007

9:58 PM

Page 149

Understanding performance appraisal fairness

Cawley, B., L. Keeping, and P. Levy. 1998. Participation in the performance appraisal process
and employee reactions: a meta analytic review of field investigations. Journal of Applied
Psychology 83(4): 61533.
Cleveland, J., and F. Landy. 1981. The Influence of rater age on two performance
judgements. Personnel Psychology 34(1): 1929.
Cobb, A., M. Vest, and F. Hills. 1997. Who Delivers Justice? source perceptions of procedural
fairness. Journal of Applied Social Psychology 27(12): 102140.
Cronbach, L. 1951. Coefficient alpha and the internal structure of tests. Psychometrika 16(3):
297334.
Dipboye, R., and R. de Pontbriand. 1981. Correlates of employee reactions to performance
appraisals and appraisal systems. Journal of Applied Psychology 66(2): 24851.
Dobbins, G., R. Cardy, and S. Platz-Vieno. 1990. A contingency approach to appraisal
satisfaction: An initial investigation of the joint effects of organisational variables and
appraisal characteristics. Journal of Management 16(3): 61932.
Dobbins, G., R. Cardy, and D. Truxillo. 1986. Effect of ratee sex and purpose of appraisal on
the accuracy of performance evaluations. Basic and Applied Social Psychology 7(3):
22541.
Evans, E., and S. McShane. 1988. Employee perceptions of performance appraisal fairness in
two organizations. Canadian Journal of Behavioural Science 20(2): 17791.
Ferris, G., V. Yates, D. Gilmore, and K. Rowland. 1985. The influence of subordinate age on
performance ratings and causal attributions. Personnel Psychology 38(3): 54557.
Fiske, D. 1982. Convergent-discriminant validation in measurements and research strategies.
In New directions for methodology of social and behavioral science: Forms of validity in
research, eds D. Brinberg and L. Kidder, 7792. San Francisco: Jossey-Bass.
Folger, R., M. Konovsky, and R. Cropanzano. 1992. A due process model for performance
appraisal. In Research in organisational behaviour, eds B. Staw and L. Cummings, vol. 14,
12748. Greenwich, CT: JAI Press.
Fulk, J., A. Brief, and S. Barr. 1985. Trust-in supervisor and perceived fairness and accuracy
of performance evaluations. Journal of Business Research 13: 299313.
Giles, W., and K. Mossholder. 1990. Employee reactions to contextual and session components
of performance appraisal. Journal of Applied Psychology 75(4): 37177.
Gordon, M., and R. Ladd. 1990. Dual allegiance: renewal, reconsideration, and recantation.
Personnel Psychology 43(1): 3769.
Greenberg, J. 1986. Determinants of perceived fairness of performance evaluations. Journal of
Applied Psychology 71(2): 3402.
Hair, J., R. Anderson, R. Tatham, and W. Black. 1999. Multivariate data analysis, 5th edition.
Upper Saddle River, NJ: Prentice-Hall.
Iverson, R. 1992. Employee intent to stay: An empirical test of a revision of the Price and
Mueller model. Doctoral dissertation, University of Iowa.
Iverson, R., S. Deery, and P. Erwin. 1995. Absenteeism in the health services sector: a causal
model and intervention strategy. In Managing absenteeism: Analysing and preventing
labour absence, eds P. Reidel and A. Preston, 25794. Canberra: AGPS.
Kahn, R., D. Wolfe, R. Quinn, J. Snoek, and R. Rosenthal. 1964. Organisational stress: Studies
in role conflict and ambiguity. New York: John Wiley.
Konovsky, M. 2000. Understanding procedural justice and its impact on business
organizations. Journal of Management 26(3): 489511.
Lam, S., M. Yik, and J. Schaubroeck. 2002. Responses to formal performance appraisal
feedback: The role of negative affectivity. Journal of Applied Psychology 87(1): 192201.
Landy, F., J. Barnes, and K. Murphy. 1978. Correlates of perceived accuracy and fairness of
performance evaluation. Journal of Applied Psychology 63(6): 7514.
Landy, F., J. Barnes-Farrell, and J. Cleveland. 1980. Perceived accuracy and fairness of
performance evaluation: A follow up. Journal of Applied Psychology 65(3): 3556.

149

APJHR_45_2_Kavanagh.qxd

150

29/05/2007

9:58 PM

Asia Pacific Journal of Human Resources

Page 150

2007 45(2)

Lawler, E. 1967. The multitrait-multirater approach to measuring managerial job


performance. Journal of Applied Psychology 51(5): 36981.
Lewis, P. 1998. Managing performance related pay based on evidence from the financial
services sector. Human Resource Management Journal 8(2): 6677.
Levy, P., and J. Williams. 1998. The role of perceived system knowledge in predicting
appraisal reactions, job satisfaction and organisational commitment. Journal of
Organisational Behavior 19: 5365.
Lindell, M., and D. Whitney. 2001. Accounting for common method variance in crosssectional research designs. Journal of Applied Psychology 86(1): 11421.
Mani, B.G. 2002. Performance appraisal systems, productivity, and motivation: A case study.
Public Personnel Management 31(2): 14159.
Murphy, K., and J. Cleveland. 1995. Understanding performance appraisal: Social, organisational,
and goal based perspectives. Thousand Oaks, CA: Sage.
Neter, J., W. Wasserman, and M. Kutner. 1985. Applied linear statistic models: Regression,
analysis of variance, and experimental design. Homewood, IL: Richard D. Irwin Inc.
Nunnally, J. 1978. Psychometric theory. New York: McGraw Hill.
Organ, D. 1988. A re-statement of the satisfaction-performance hypothesis. Journal of
Management 14(4): 54757.
Pettijohn, C., L. Pettijohn and M. dAmico. 2001. Characteristics of performance appraisals
and their impact on sales force satisfaction. Human Resource Development Quarterly
12(2): 12745.
Shore, L., and G. Thornton. 1986. Effects of gender on self and supervisory ratings. Academy
of Management Journal 29(1): 11530.
Studenmund, A., and H. Cassidy. 1987. Using econometrics: A practical guide. Boston: Little,
Brown and Company.
Tang, T., and L. Sarsfield-Baldwin. 1996. Distributive and procedural justice as related to
satisfaction and commitment. SAM Advanced Management Journal Summer: 2530.
Taylor, M., K. Tracy, M. Renard, J. Harrison, and S. Carroll. 1995. Due process in
performance appraisal: a quasi-experiment in procedural justice. Administrative Science
Quarterly 40: 495523.
Thibaut, J., and L. Walker. 1975. Procedural justice: A psychological examination. Hillsdale, NJ:
Erlbaum.
Thompson, D., and T. Thompson. 1985. Task-based performance appraisal for blue-collar
jobs: evaluation of race and sex effects. Journal of Applied Psychology 70(4): 74753.
Watson, D., and J. Pennebaker. 1989. Health complaints, stress and distress: Exploring the
central role of negative affectivity. Psychological Review 96: 2345.
Watson, D., J. Pennebaker, and R. Folger. 1987. Beyond negative affectivity: Measuring stress
and satisfaction in the workplace. Journal of Organisational Behavior Management 8(2):
14157.

Vous aimerez peut-être aussi