Sabotage in Capital Budgeting: The Effects of Control and Honesty on Investment Decisions

Abstract

This study investigates the sabotage of investments in response to hurdle contracts as a means of formal control in capital budgeting. We conduct a laboratory experiment to examine factors that drive or inhibit sabotage. Sabotage occurs when the manager provides false information to prevent the firm from realizing a profitable investment, which is costly both for the manager and the firm. Our results show that managers sabotage investments to reciprocate for distrustful formal control. Conversely, budget communication that requires factual assertions and thus activates managers' preference for honesty inhibits sabotage. Moreover, honesty suppresses negative reciprocity and thus reduces sabotage not only directly but also indirectly. Our findings warn firms to consider the sabotage of investments as a hidden cost of control in budgeting. They show that honesty has a spill-over effect, as it absorbs negative reciprocity.

Acknowledgments

We thank Victor Maas (associate editor) and two anonymous reviewers for many helpful comments. We are particularly indebted to Jeremy Douthit, whose suggestions have greatly improved this study. We also thank Gunther Friedl, Steve Kachelmeier, Alwine Mohnen, Steven Schwartz, Bill Tayler, Michael Williamson, and Rick Young for their comments and support, along with participants of the 2017 Annual Meeting of the American Accounting Association, the 2017 AAA Management Accounting Section Midyear Meeting, the 2016 GEABA annual meeting, the 2016 Annual Conference of the Verein für Socialpolitik, the 2016 Joint International Conference of JIAR and AOS, the 2016 Annual Congress of the European Accounting Association, the 2015 ENEAR summer school, and the Workshop on Research in Ethics and Accounting at the University of Halle. We appreciate the advice and help of Simon Bierbaum in setting up the experiment.

ORCID

Markus Brunner  http://orcid.org/0000-0002-1687-279X

Andreas Ostermaier  http://orcid.org/0000-0001-9143-5103

Notes

1 Hurdle rates are used in more than half of US and Canadian firms (Graham & Campbell, 2001). When hurdles are used, they are usually above the cost of capital (Poterba & Summers, 1995). We refer to a hurdle contract as the combination of participative budgeting and the hurdle as a formal control, which is commonly found in firms. Hence, the term hurdle contract is used to describe Evans et al.'s (2001) modified trust contract rather than Antle and Eppen's (1985) truth-inducing hurdle contract.

2 Sabotage is technically possible in Rankin et al. (2003), albeit not the focus of their research. We provide their outcomes for comparison with our results below.

3 We refer to the manager with male pronouns and to the owner with female pronouns throughout the article.

4 In their survey among managers and planning executives with more than 1300 responses, Collins et al. (1987) ask participants about their attitude toward budgeting. The responses suggest that it is not uncommon for managers to withhold their support of the budget and have erroneous or incompatible data enter and hamper the budgeting process.

5 Let c denote the actual cost, $\hat{c}$ the reported cost, and $\tilde{c}$ the maximal reported cost that the owner funds (i.e. the hurdle). A self-interested manager who maximizes slack will always report $\hat{c}=\tilde{c}$ as long as the actual cost is below the hurdle. With $c\sim \mathcal{U}(0,100)$, $\tilde{c}=50$ maximizes the owner's expected payoff, which is $\tilde{c}/100\cdot (100-\tilde{c})$.

6 Irrespective of the owner's motive in choosing the hurdle contract, it affects the distribution of payoffs. If the investment is made, it leaves the owner with a larger payoff than the manager, which the manager likely considers unfair. Inequity creates a cost that the manager charges against his payoff (Fehr & Schmidt, 1999). However, we keep inequity constant across all treatments of our experiment to disentangle the effects of negative reciprocity and honesty from the effect of fairness preferences.

7 Let c and $\hat{c}$ be the actual and the reported cost. Suppose that the manager misreports the cost to maximize slack ($\hat{c}=50$ if $c\le 50$ and $\hat{c}=100$ otherwise). With $c\sim \mathcal{U}(0,100)$, the owner's expected payoff is $Pr(c\le 50)\cdot (100-50)=0.5\cdot 50=25$ with the hurdle compared to 0 without. If the manager reports the cost honestly ($\hat{c}=c$), the owner's expected payoff is $Pr(c\le 50)\cdot (100-\mathrm{E}(c\mid c\le 50\left)\right)=0.5\cdot 75=37.5$ with the hurdle compared to $100-\mathrm{E}\left(c\right)=50$ without. To generalize, let $\pi \in [0,1]$ denote the manager's expected percentage of honesty (the slack he does not create, although he could; see Evans et al., 2001), where the owner expects the manager to maximize slack for $\pi =0$. The owner implements the hurdle contract if $Pr(c\le 50)\cdot (50+\pi (50-\mathrm{E}(c\mid c\le 50)\left)\right)>\pi (100-\mathrm{E}(c\left)\right)\iff 0.5\cdot (50+\pi (50-25\left)\right)>\pi (100-50)\iff \pi <\frac{2}{3}$. Hence, the hurdle contract suggests serious doubts about the manager's honesty.

8 Rankin et al. (2003) have owners set the hurdle but do not report how often managers sabotage investments. Moreover, they do not include conditions where formal control is chosen exogenously to disentangle inequity aversion and negative reciprocity, which related research does (Christ, 2013; Falk & Kosfeld, 2006; Houser et al., 2008; Kuang & Moser, 2009).

9 In addition to the magnitude of lies, people have been found to care about consequences and condition their decision to lie on these (Gneezy, 2005; Maas & van Rinsum, 2013). The manager might thus refrain from sabotage because of the severe consequences for the owner. Slack also hurts the owner, but far less so than sabotage.

10 The fixed payment made sure that no one left the experiment without compensation. We follow similar experiments, which all included fixed payments (Evans et al., 2001; Hannan et al., 2006; Rankin et al., 2008; Stevens, 2002). In particular, the fixed payment avoids excessive tension for managers who want to be honest (Rankin et al., 2008).

11 The manager cannot understate the cost or allocate a lower payoff to himself than what he needs to cover his cost. Hence, he cannot leave the experiment with a net loss.

12 The cost in the rounds that allowed sabotage averages ECU 24.59 in the Factual Assertion condition and ECU 24.42 in the No Factual Assertion condition. It does not differ between the Reciprocity and No Reciprocity conditions, as the latter replicates the former. Likewise, the cost in the rounds with the hurdle contract averages ECU 51.16 in the Factual Assertion condition and ECU 49.63 in the No Factual Assertion condition.

13 The binding budgetary announcement by the owner in Rankin et al. (2003) allowed the manager to sabotage the investment. As the announcement was the owner's choice and the manager was required to submit a cost report, Rankin et al.'s setting resembles our Reciprocity–Factual Assertion condition. We thank the authors for providing us their data for comparison. Unreported results show that managers sabotaged 13.14% of the investments that they could sabotage, which is similar to the 11.12% in our Reciprocity–Factual Assertion condition.

14 To control for the lack of independence caused by each manager making multiple decisions, sabotage is averaged over all rounds. The manager's averaged sabotage is treated as a single observation in the ANOVA. Note that averaging turns the dummy variable sabotage into a continuous variable ranging from 0 to 1, which allows the use of the ANOVA.We also estimate a logistic regression, which confirms the results of the ANOVA, to account for the effect of the round. We find neither a significant main effect of the round nor an interaction effect with our between-subject variables. More detailed results are reported below (Section 4.3).

15 Conducting these analyses with slack instead of sabotage as the dependent variable, we find the same effects of negative reciprocity and factual assertion. However, the effects are only significant when we include data regardless of whether investments were realized or failed, as prior literature does (e.g. Douthit and Stevens, 2015; Rankin et al., 2008). Table 1 reports the slack that managers actually created rather than the hypothetical slack from rounds where investments were sabotaged or failed because of the hurdle.Interestingly, Panel C of Table 1 shows that slack is higher in the Factual Assertion than in the No Factual Assertion condition under the trust contract, which seems inconsistent with Rankin et al. and Douthit & Stevens. Slack creation, however, should be seen within the context of sabotage. Honesty prevents sabotage in the Factual Assertion condition, resulting in frequent unequal payoffs. Given the frequent inequity under the hurdle contract, managers appear to push for equity under the trust contract in turn.

16 Note that the question was phrased in general terms to measure Intent to Be Honest regardless of the condition rather than to check whether managers were more aware of the honesty norm in the Factual Assertion than in the No Factual Assertion condition. It is obvious, however, that post-treatment responses are contingent on the condition, as described in the text.

17 In addition to the significant paths depicted in the figure, the model includes insignificant paths (1) from Reciprocity to Intent to Punish (z=1.49, p=.137); (2) from Reciprocity to the path from Distrust to Intent to Punish in order to interact Reciprocity and Distrust (z=0.95, p=.343); (3) from Factual Assertion to Sabotage (z=−0.09, p=.926). The first two paths are intended to rule out that Distrust results in Intent to Punish because hurdles are set by owners only. The third path rules out an effect of Factual Assertion on Sabotage other than developed in our argument for Hypotheses 2 and 3.

18 Managers indicated their agreement with the following statement on a 7-point Likert scale, ranging from ‘fully disagree’ (1) to ‘fully agree’ (7): ‘I wanted the profit to be fairly distributed between the owner and me’.

19 The hurdles and payoffs from the three non-paying rounds where the managers assumed their future role as managers were considered because they might affect expectations. Expectations about hurdles and payoffs therefore already exist for the first round.

20 More precisely, the manager reports $50+\epsilon$ in order to sabotage the investment, where $\epsilon \to 0$.

21 Participants do not learn whether they will be owners or managers until this information is displayed to them on the screen after the end of the trial stage.