Audits, Reputation, and Repeated Interaction in a Capital Budgeting Setting

Abstract

This experimental study investigates the use of audits as a control instrument in capital budgeting processes and factors that contribute to audit efficiency (or inefficiency) in a repeated relationship. We disentangle reputational aspects of superiors and subordinates from an increased social content in repeated relationships. In settings where subjects face new counterparts every round, we find that reputational aspects strongly affect the superiors' and subordinates' behavior. This leads to a decrease in slack and an increase in the superior payoff. However, in a repeated relationship with an increased level of social content, these benefits are no longer present. In fact, the subordinates' ability to retaliate against their superior for prior punishments in the repeated setting by reporting more slack in later rounds has negative effects on the use of audits and decreases audit efficiency. These findings imply that, in repeated relationships, social factors play an important role and might impair the disciplining effects of audits.

Acknowledgments

We thank participants at the 2008 AAA annual meeting, the 2009 MAS Research and Case Conference, the 2008 EAA annual meeting, the Operations Research Conference 2008, the annual meeting of the German Academic Association for Business Research, the accounting section meeting of the German Academic Association for Business Research, and seminar participants at the University of Maastricht, the University of Lausanne, the University of Graz, the University of Passau, the University of Göttingen and the European Business School, for many helpful comments and suggestions. We are particularly grateful to Ann Gabriel, Jason Kuang and two anonymous reviewers, and to Alexander Brüggen, Stefan Dierkes, Robert M. Gillenkirch, Jari Huikku, Robert Knechel, Olaf Korn, Frank Moers, Dirk Simons and Alfred Wagenhofer for their helpful comments. We also thank Achim Hendriks for his excellent research assistance.

Data availability: From the authors upon request.

Notes

See, for example, Fehr and Gächter (2000) and Camerer and Fehr (2004) on social norms and their economic importance.

These contributions often recommend to use audits as a feedback instrument, i.e., as a learning device for managers in order to improve future decision making (Chenhall and Morris, 1993; Pierce and Tsay, 1992).

This is necessary in order to disentangle monetary and non-monetary motivations to conduct audits. We will explain this in more detail in Section 3.

See Hannan et al. (2010) for its use in a capital budgeting setting. Their results with respect to the superiors' decisions are not affected by the two methods they use. Schwartz et al. (2009) find that superiors are less affected by emotional factors in a capital budgeting setting when the strategy method is used.

Schwartz and Young (2002) investigate reputation and repeated interactions in a joint investment setting involving two division managers, but they do not consider the possibility that one of the managers can use audits or punish the other for not cooperating. Stevens (2002) and Webb (2002) also examine reputation formation in a budgeting environment but consider only reputation concerns of the acting managers.

Consistent with this, Bohnet and Huck (2004) find that in a trust game, trust is more likely to emerge in a repeated partner setting than in a setting with random matching and full information, and Bolton et al. (2004) find that buyers in a market context react more strongly to information about sellers' history in a repeated setting than in a setting with random matching and full information.

We conducted the treatments T2–T4 in order to exclude that the pure information about the cost reports made by the subordinates or seen by the superiors could influence the results. This is necessary as the information provided in these treatments corresponds to the information that would be shown in the audit treatments T6–T8 if the superiors never conducted any audits. Recall that this would be the case if everybody had only monetary preferences and this was common knowledge.

In the experiment, we avoided the term ‘fine’ and simply spoke of ‘deducting points’. We also used the term ‘inspection’ instead of ‘audit’, and we referred to superiors and subordinates as ‘A-players’ and ‘B-players’.

The average fine amounts to 89.6% of the slack. In 75.6% of the cases, the maximum fine is imposed. This is further evidence for the notion that superiors consider excess slack creation to be a norm violation. The differences in fines across the audit treatments are small. The average fines vary between 83.1% and 93.7% of the slack, and a Kruskal-Wallis test reveals that the differences are not significant ( , p > 0.5).

For SLACK ^rel , it is not necessary to calculate a mean value as SLACK ^rel already aggregates the slack over all rounds and sets it into relation to the maximum slack.

T-tests are used throughout the analysis as, for each measure, we cannot reject the hypothesis that it is normally distributed in every treatment.

We also conducted three ANOVAs with all three factors manipulated in the random matching treatments as independent variables (ability to audit Y/N, information about the superior Y/N and information about the subordinate Y/N). SLACK ^rel and both payoffs are the dependent variables, respectively. For the ability to audit, we find a highly significant main effect in every case but no two-way or three-way interaction effects with the other two independent variables.

In every OLS regression, standard errors are clustered at participant level in order to control for multiple observations within subjects, and we include individual fixed effects. Additionally, we include period fixed effects whenever ROUND is not used as an independent variable. However, the fixed effects' coefficients are not reported in the corresponding tables.

If we repeat the regression analysis separately for T6 in which only information about the subordinates' history exist, we find qualitatively the same results.

This is also the case if we examine T7 separately.

Although this is inconsistent with a purely economic perspective, prior experimental research has shown that subjects do not always correctly predict the consequences of contract choices (Fehr and Rockenbach, 2003; Irlenbusch and Sliwka, 2005).

For example, there might be spillover effects from audits conducted with a ‘learning’ purpose for performance evaluation, or the threat of audits might be strong enough to deter subordinates from misreporting costs to the point that punishing subordinates is never actually necessary.

Anderson and Putterman (2006) find that participants do not completely refrain from punishing even if their punishment cost exceeds the fine for the other player. Moreover, the effects of increased punishment costs are smaller under the strategy method used in our experiment (Brosig et al., 2003).