Audit schedules in the presence of concealing effort in supply chains
Final Report Abstract
Firms often discover unexpected issues in their products or supply chains that, if not addressed promptly, are harmful to other parties. Rather than fixing the problem, however, these organizations sometimes prefer to conceal it and exert effort to evade detection. For instance, when technology executives at the auto manufacturer Volkswagen discovered that their new engine would not comply with the US Clean Air Act, they chose to develop a sophisticated software that evaded emission tests rather than investing in the creation of more expensive but effective emission equipment. These concealment efforts greatly hinder the ability of the affected parties to address potentially costly issues in a timely manner. In this project, we advance the operations research theory on how organizations and auditors can uncover and remedy such adverse issues that may occur in firms with concealment capabilities. To that end, we first formalize the problem of designing efficient auditing and remediation strategies as a dynamic principal-agent problem in continuous time. The generality of this framework allows us to explore the efficacy of a large variety of auditing policies, which include random audit times following any (well behaved) distribution, deterministic schedules, random audits at deterministic times, or any combination of the above. We then show that a generalized version of the so-called revelation principle holds in our context. As a result, our formulation reduces to the optimal stochastic control of a piecewise deterministic Markov process (PDP), a class of stochastic processes that generalizes semi-Markov decision processes. Optimal control of PDPs are, in general, not analytically tractable. Nonetheless, we fully characterize the principal’s optimal policy, which yields new insights concerning auditing in the presence of evasion capabilities. This optimal policy describes a family of novel auditing strategies that are i) efficient at detecting adverse issues and ii) relatively easy to implement. Further, our analytical characterization enables us to prove results concerning the impacts of the evasion capability on efficient remediation strategies, without resorting to numerical studies. Overall, this analysis suggests that the auditor should incentivize the firm to always disclose the issue as soon as it occurs. For this purpose, the auditor should audit the firm periodically, but also offer to cover part of the agent’s remediation cost if the agent voluntarily reports the issue. These monetary incentives for self-reporting should decrease as the time of the next audit approaches. This corresponds to a dynamic cost sharing contract, in which the firm pays a time-dependent penalty upon self-disclosing an issue, while the auditor covers the remaining remediation cost. The firm’s payment towards the remediation is always strictly positive and the auditor should never cover all costs. Perhaps more importantly, our results reveal that the auditor should audit the firm according to a pre-determined schedule if the agent’s evasion capability is limited, but should conduct random audits otherwise. In this sense, the firm’s evasion capability affects the very nature of the auditing policy. Further, as the firm’s evasion capability becomes more limited, the auditor needs to audit more frequently and thus incurs higher audit costs. In other words, the firm’s evasion capability actually reduces the overall auditing costs, which may be counter-intuitive. However, and as expected, this evasion capability also increases the auditor’s contribution to the remediation costs, which is required to induce the agent to voluntarily disclose the issue. Besides the problem of evading detection, the optimal control problem of PDP we propose in this project provides a fruitful framework to address other types of issues related to auditing.
Publications
- “Audit and Remediation Strategies in the Presence of Evasion”, INFORMS Conference 2020
Francis de Véricourt
