This chapter starts by identifying the focal independent and criterion variables of interest in
the study. It then goes on to link the research questions to the gaps identified in the literature
review. This is accomplished via presentation of the models o f the hypothesized relationship
between the variables of interest, which are then summarized into testable hypotheses. The models o f relationships investigated by most o f the control systems studies are represented in
Figures 3.1 - 3.3 below; while that employed in this study is represented in Figure 3.4 with construct mappings in Figure 3.5.
3.1
Identifying Focal and Criterion Variables
Identifying the focal independent variable o f interest in a study is important as it brings focus
and clarity to the study. It also determines the choice o f criterion variable(s) and has a direct
bearing on theory development. Briers and Hirst (1990) note that the choice o f dependent
variables is constrained by the focal independent variable and that theory development involves not only providing explanations for the relations among the variables o f interest but
also choosing the dependent variables. Similarly, Hartmann (2000, p474) note that
“ ...understanding o f the multiple roles of accounting and o f RAPM (reliance on accounting
performance measures) may be enhanced if researchers explicitly consider which aspect of
RAPM they address...” To this end therefore the focal variable o f interest to this study is
performance evaluative style (PES) conceptualised as encompassing the nature o f information
used in performance evaluation, the way this information is used in evaluating performance
and how it is linked to rewards. A detailed discussion o f the conceptualisation o f PES, its
measurement and empirical application is provided in Chapter 5.
In terms o f the criterion variables, behaviour and performance are the focal variables. In this
Chapter 3 Theory D evelopm ent and H ypothesis Formulation 39
impact on behaviour and performance, thus they are not criterion variables as such. Also, both
functional and dysfunctional behavioural effects of PES are examined. Functional behaviour
investigated is work effort and dysfunctional behaviour is represented by quality threatening
behaviour and intention to turnover, as a surrogate o f real turnover. Performance on the other
hand is gauged by actual job performance ratings and managers self rating o f performance.
Other variables o f interest in the present study are introduced in the discussion o f the models
o f relationships investigated starting from the simple direct or overall effects o f PES on to the
more complex relationships combining mediating and moderating influences.
3.2
Theoretical Models of Relationship Examined
Over time, MCS researchers have employed three main models o f relationships in studying
control systems and its consequences. In historical sequence, researchers have explored this
relationship using a direct or overall effect model, a moderated effect model, and a mediated
or indirect effect model.
a) The direct/overall effect model, graphically depicted in Figure 3.1 below, represents the
generic model o f relationships investigated by early control systems studies as highlighted
in the literature review that focus on PES (see also Covaleski et al., 2003). With this
model researchers explored the direct relationships between control systems and its
consequences. In relation to PES, researchers investigated the direct consequences o f PES
on mental states, behaviour and performance (e.g. Argyris, 1952; Hopwood, 1972; Otley,
1978).
b) The moderated effect model, graphically depicted in Figure 3.3 below, represents the
relationships studied by the second stream o f control systems studies. With this model,
researchers explored the factors that moderate the direction and/or strength o f the
relationship between control systems and their consequences. In relation to PES,
Chapter 3 Theory D evelopm ent and H ypothesis Formulation 40
behaviour and performance (e.g. Brownell, 1982a; 1985, 1987; Hirst, 1983; 1987;
Govindarajan, 1984; 1985; Brownell & Hirst, 1986). These studies were motivated by the
conflicting findings from the first stream o f studies and tried to explain these in terms of
possible moderating influences that could account for the conflicting findings,
c) The mediated or indirect effect model, graphically depicted in Figure 3.2 , represents the
form o f relationships investigated by later control systems studies (e.g. Nouri & Parker,
1998; Shields et al., 2000; Chong & Chong, 2002; Wentzel, 2002). These studies
generally hypothesize and investigate an indirect effect of control systems features,
predominantly participation, on job performance via other variables - predominantly goal
commitment and goal difficulty. However, in relation to PES it does not appear to the best
o f my knowledge that any o f the studies focusing on PES investigated this model of
relationships perhaps because of the prime focus on explaining conflicting findings of
earlier studies in terms o f contextual or moderating factors. This model entails identifying and examining the factors that intervene in the relationship between PES and the criterion
variables. To this end, the mental state variables are modelled as the mediating factors and
behaviour and performance as the criterion variables.
In addition to the above forms o f relationships, this study also examines a fourth and more
comprehensive form o f relationship which, to the best o f the author’s knowledge, has not been
studied in the performance management and control systems literature. This study integrates
into one comprehensive model all the 3 models previously examined separately in what is
termed a mediated moderation model.9 With this model both the moderating and mediating
influences on the overall effect o f PES on behaviour and performance are examined
simultaneously. Thus the reality' o f the complexity o f the mechanism by which PES affects
behaviour and performance is explored enhancing our understanding of control systems and
its effects. The mediated moderation model is graphically depicted in Figure 3.4 below.
9 This form o f relations is currently discussed in the social psychology literature as mediated moderation and moderated mediation. For a more detailed discussion and distinction between the two see Muller, Judd, & Yzerbyt, 2005; Simons et al., 1993; and Baron & Kenny, 1986.
Chapter 3 Theory D evelopm ent and H ypothesis Formulation 41
The approach adopted in this study in using this model is to first develop hypotheses on the
overall effects o f PES on performance and behaviour in line with the direct effects model.
Then hypotheses are developed incorporating the moderating influences on the overall effects
o f PES on performance and behaviour in line with the moderated effects model. Furthermore,
hypotheses are also developed incorporating the mediating influences on the overall effects of
PES in line with the mediated effects model. Finally, hypotheses are developed incorporating
both the moderating and mediating influences in the mediated moderation model. This models
the mediating influences as the mechanisms by which the moderation effects are produced. To
this end, possible moderating influences considered include organisational hierarchical level,
the interactive use o f control systems, trust in supervisor or supervisory trust, organisational
commitment, and goal difficulty. And the mediating variables considered include equity and
Chapter 3 Theory D evelopm ent and H ypothesis Formulation 43
3.3 M od el D evelop m en t & Form ulation o f H ypothesis - O verall E ffects M od el
We start the model development and formulation of hypothesis by looking at the direct effects
o f PES on both performance and behaviour as did the initial control systems studies (links A
& B in the model, Figure 3.4). PES is hypothesized to affect two types o f behaviour -
functional and dysfunctional behaviour. Dysfunctional behaviour refers to those behaviours
exhibited by employees which are ordinarily considered to threaten the best interests o f the
firm and thus its long term success, whereas functional behaviour is considered to enhance the
firm’s interests.