The budgetary slack has been studied extensively in the management and accounting literature, but results are inconclusive. This could be because the research has focused on economic factors rather than on psychological variables, such as trust. This paper tries to contrast psychological and economic causes in the creation of budgetary slack. Particularly, we examine whether a higher amount of subordinates' trust in their superiors with an economic incentive helps to reduce the tendency of subordinates to create budgetary slack. This study is based on a laboratory experiment conducted with 240 managers in order to investigate how trust, understood as a psychological and moral factor, contributes toward the reduction of slack in the absence and presence of pecuniary incentives. Subjects were divided into three groups: managers, executives and controller. Results support partially our hypotheses. This paper shows that non-monetary incentives could help managers to reduce budgetary slack in organizations.
The existence of slack leads harmful consequences for companies like lost business
opportunities and/or inflated costs. The word “slack” is used to describe a
circumstance in which the resources and effort employed in the development of an activity
no longer contribute to the achievement of organizational objectives
The goal of this paper is to develop new theories that integrate behavioral and economic factors, and thus we treat together a psychological factor (trust) and an economic factor (economic incentives).
The current dominant economic view of slack is based on agency models. Agency
models theorize that providing agents (subordinates) with more information than
principals (supervisors) need not result in greater efficiency; the reason for this is that
agents may use this information to shirk. Assuming an agency perspective, numerous
experimental studies have studied the effects of risk aversion, information asymmetry,
and pay schemes on budgetary slack; the goal of this research is to understand the
incentives that promote honesty in agents
In summary, results in prior experiments suggest that subjects with no economic incentives to cooperate (because they are paid with slack-inducing schemes) nonetheless reduce the amount of slack in the laboratory, and as a consequence their wealth. Not only do subjects create less budgetary slack than expected, but in addition honesty can prevail in the absence of pecuniary incentives. In other words, the introduction of explicit monetary incentives may weaken non-pecuniary incentives. The experiment conducted by Rankin et al. (2005) disentangles the preference for honesty from other non-pecuniary preferences, demonstrating that subjects who have final budget authority significantly prefer honesty. In addition, the slack generated in this experiment was less than the theory predicted.
Trust can be defined as the willingness of one party (trustor) to be vulnerable to the
actions of another party based on the expectation that the other will perform in the
way that trustor expect
In summary, we expect that in the laboratory: a) subjects who distrust or don’t trust but are economically encouraged, are prone to decrease slack; b) subjects who distrust or don’t trust but are not economically encouraged, are prone to ever-increasing slack; c) subjects who trust but are economically encouraged, submit budgets with higher slack; and d) subjects who trust but are not economically encouraged, submit budgets with low slack. Thus, we formulate the following hypothesis:
Hypothesis 1: Subjects who trust in their superiors and are not economically encouraged choose budgets with less slack than subjects who evidence distrust or “no trust” and are economically encouraged.
The manipulation of the level of trust in the laboratory should have consequences
for subjects’ choice of budgetary slack. When suspicion about superiors arises,
budgetary slack should increase. This results from the fact that trust is formed over time
It is always feasible to move managers from their initial positions along the
continuum of trust-distrust because trust is an induced mind-set. Trust is a non-personality
factor, susceptible to change when individuals interact in laboratory experiments.
Thus, an individual can change his or her level of trust (or mistrust) while attempting
to solve a problem
Having established a level of trust with another person, a perception that trust is
one-sided leads to some diminution. When individuals begin to doubt that another
person is operating in good faith, they manifest suspicion. Suspicion, in turn, results
in a loss of trust. Similarly, individuals begin to distrust when they anticipate
violations of trust in the future. The thought that unfulfilled expectations in one
interpersonal exchange are likely to manifest in all other exchanges leads to distrust. Distrust
emerges through negative expectations, assumptions, or beliefs about others’ motives.
Recurring abuses further increase distrust
Hypothesis 2: A reduction in trust generates an increase in slack, independently of the presence and absence of economic incentives. 3
The laboratory experiment employed a 5 (trust-slack levels) x 3 (information asymmetries on trust and slack) factorial design. We randomly assigned 240 participants to the roles of 30 executives, 90 managers, and 120 controllers, and in 30 groups (see Figure 1). In order to design our experiment, we based on a previous one. That experiment tried to check if subjects who trust in their superiors choose budgets with less slack than subjects who evidence no trust or distrust without any monetary incentives. It was recruited a total of 240 businesspersons enrolled in postgraduate business courses to participate in an experiment about the effect of trust on budgetary slack. Subjects were pseudo-volunteers, as the experiment was part of a class assignment. The subjects did not receive payment for their participation in this experiment. Businesspersons were invited to participate in the experiment as a means of improving their knowledge of the budgetary process, consistent with the notion that classroom experiments have pedagogical value (Friedman and Sunder, 1988). The average managerial experience of a participant was 3.57 years. The percentages of males and females in the sample were 68% and 32%, respectively. 48% of the participants were currently dealing with budgets in their professional activities, while all of the subjects had experience dealing with budgets at some time.
The experiment consisted of a simulation study of a business game, where
participants were assigned simulation tasks
It was verbally informed participants regarding the general purpose of the experiment, the resource and information endowment, the set of actions available to them, and the moral and economic consequences of each action (Friedman and Sunder, 1988). Participants also received private written instructions, which they were not allowed to reveal at any time during the experiment. It was also provided all participants with written information about the nature of the budgets under discussion. In particular, they knew the global profitability underlying each budget: a) 5.35% (budget 1), b) 5.78% (budget 2), c) 6.31% (budget 3), d) 6.68% (budget 4), and, e) 7.09% (budget 5). Nonetheless, only subjects in the roles of managers knew the amount of budgetary slack, as they were told privately that 7.10% was the maximum attainable global profitability. The amounts of slack were 0.01% (budget 5), 0.42% (budget 2), 0.79% (budget 3), 1.32% (budget 2) and, 1.75% (budget 1). Thus, they were aware of the slack associated with each budget.
Variables Measurement
The endogenous variables are: the first budget proposed, which represents the earliest
manifestation of slack (V1), and the final budget, which is the last manifestation of
slack (V2)
Both executives and controllers were uninformed about slack conditions and the amount of slack. Hence, to identify how aware executives and controllers were of slack during discussion of the budget, they were asked about the effort that managers invested in their last budget proposal. A variable based on effort was built, which varies from 1 (very easy to attain) to 5 (required their maximum effort) (V5). It was also checked if executives and controllers were conscious of: a) managers’ success in submitting budgets easily attainable (V6); b) if budgetary targets induce high managerial productivity (reverse code) (V7); c) if it was costly to manage budgets carefully (reverse code) (V8); and d) if they thought that budgets had motivated managers to be concerned with improving efficiency (V9). Executives’ and controllers’ responses were on a scale from 1 (definitely true) to 7 (definitely false). With regard to trust, executives gleaned some indirect information through the level of cooperation, whereas controllers knew nothing. To differentiate between these two situations, a binary control variable that we denote as the absence of information on trust was defined (V10); this provides a value of one for controllers and zero for executives. We also control for gender differences (V11), professional experience (years in the workplace as a manager) (V12), and previous knowledge of budgets (V13). 4
To test our hypotheses, a multinomial logit model was specified
P (yk 1 / ßk xexp ßkTx)/ exp ßkTx) , (1)
Where y is the class indicator for the kth budget; x is the predictor vector extended by one to be paired with the intercept parameter. Each k is a vector of parameters, one for each class (the letter T means total). The initial budgets diverge. Subjects in the role of managers start the budgetary meeting with budget 3 (amount of slack: 0.79%) 28.9% of the time, followed by budget 1 (slack: 1.75%) 24.4% of the time, and budget 2 (slack: 1.32%) 21.1% of the time. Budgets 4 (slack: 0.42%) and 5 (slack: 0.01%) are chosen less frequently, 12.2% and 13.3% of the time, respectively. A Wald test permitted appraising the significance of the individual logistic regression coefficients for the variable group (V4) and the insignificance of the control variables (Table 3). Using the Wald statistic, group is significant with the exception of Budget 2. Likelihood ratio tests also corroborate the significance of group and the insignificance of the control variables (see Table 1).
The odds ratio, Exp (B), in Table 1 shows that as group increases by one unit, the odds ratios of budget 3 (slack= 0.79%), budget 4 (slack= 6.68%), and budget 5 (slack= 0.01%) increase by multiples of 4.05, 5.14, and 2.43, respectively, once the variables for sex (V11), years at work (V12), and budget experience (V13) were controlled. Thus, the parameter estimates confirm that when one-time prior conditions of subjects move from distrust-high slack to trust-low slack, the probability of a subject submitting initial budgets with low slack (0.42%), medium slack (0.79%), and very low slack (0.01%) increases. This result confirms, to some extent, Hypothesis 1: Subjects who previously trust create less slack than managers who distrust, i.e., they intend to invest more effort. We cannot show, however, that subjects who evidence low distrust in their superiors produce more (or less) sack than he ones who evidence high distrust. V1 V1 V3 V1
A different multinomial logit model for closing budgets was constructed. Final budget is the dependent variable with five categories generating four equations. Each of the four equations is a binary logistic regression that contrasts other budgets with Budget 1 (very high slack). Multinomial logistic regression simultaneously estimates the four logits.
Final budgets show some discrepancy. A greater number of subjects (32.2%) finish the budgetary meeting agreeing to budget 3 (slack=0.79%). Smaller numbers of 0.106 3.996 managers decide on other budgets: 14.4% are inclined to close the meeting with budget 1 (slack=0.79%), 20.0% with budget 2 (Slack=1.32%), 18.9% with budget 4 (slack=1.32%), and 14.4% with budget 5 (slack=0.01%). It was found that the amount of slack in final budgets is less than in initial budgets. Therefore, disagreement appears to reduce slack on average. 57 subjects adhere to their opening budget proposals, however, while 33 subjects change their final proposal from the opening offer. Using the Wald statistic, group (V3) is significant with the exception of budget 2 (Table 2), and as well as in the likelihood ratio tests (Table2). The odds ratio, Exp (B), bears out the preceding outcome. A one unit increase in group, i.e., subjects moving towards early high-trust and low-slack, brings about an increase of 1.998 in the odds ratio of budget 3 (slack=0.79%), and 2.152 in the odds ratio of budget 4 (slack=0.42%). The odds of budget 2 (slack=1.32%) and budget 5 (slack=0.01%) as final proposals by subjects in meetings, however, are not significantly explained by the initial group.
Sig. V13 –0.177
The Wald test also indicates that final trust (V4) explains the odds ratios of final budgets 2 and 4 (Table 2). The likelihood ratio tests, however, do not strongly support the significance of final trust (Table 2). If the ending trust increases by one unit (towards high trust), the odds ratios of budget 2 (slack=1.32%) and budget 4 (slack=0.42%) are less than one. Further units of final trust generate a reduction of 0.399 in the odds ratio of Budget 2, and 0.386 in the odds ratio of Budget 3 (Table 2). Accordingly, once final trust increases, the probability that subjects propose low and medium slack final budgets, instead of budgets with the maximum slack, is less. The exception is the odds ratio for budget 5, which is found to be insignificant. These findings validate to some extent Hypothesis 2: By taking into consideration the fact that final trust produces consequences for subjects’ slack choices, follow-on slack is greater than before as one introduces suspicion. The initial trust and slack conditions, however, are determinants for most managers. For example, 57 managers do not adjust their original budget suggestions. 5
The experiment shows that trust, understood as a moral and psychological factor,
ameliorates the problem of slack in the absence of any explicit link between trust
preferences and the distribution of wealth
The budgets initially and finally proposed by subjects in the role of managers
contain less slack than expected, which is consistent with previous findings that indicate
that subjects tend to produce less budgetary slack than agency theory predicts
Several subjects in the role of managers adjusted their budgetary choice. As soon as their final trust moved from distrust towards trust, subjects modified their budgets and thus their slack. In response, most subjects decided to reduce, rather than increase, slack. When suspicion appeared in the budgetary setting, and the managers’ final trust shifted from a position of distrust to trust, the possibility that managers chose budgets with high, medium, and low slack, instead of very high slack, diminished as the final conditions depended more on trust. This is a key finding: Budgetary slack levels decrease in trust settings, even in the absence of any direct pecuniary incentive. This finding is relevant for management accounting researchers as trust, understood as a psychological and moral factor, has a positive effect on the amount of slack. But we ask: budgetary slack levels decrease more (in trust settings) in the presence of monetary incentive? Or conversely budgetary slack levels decrease less? We try to design an experiment that answers these questions. 6