Auditing Section Research Summaries Space

A Database of Auditing Research - Building Bridges with Practice

This is a public Custom Hive  public

research summary

    Causal Inference in Auditing: A Framework.
    research summary posted October 14, 2015 by Jennifer M Mueller-Phillips, last edited October 14, 2015, tagged 09.0 Auditor Judgment 
    Causal Inference in Auditing: A Framework.
    Practical Implications:

    These findings have implications for both research and practice. Most importantly, given that both practitioners and experimental researchers often consider a fairly small initial set of hypotheses it is critically important to know whether the set is considered by the auditor or by the experimental participants to be exhaustive or non-exhaustive, and to know the expected interrelationships (positive, negative, or independent) among the hypotheses. Differences in these characteristics imply different interpretations of audit evidence or research results. The findings highlight the importance of explicitly measuring or controlling interdependencies among hypotheses evaluated in future experiments. While the focus has been on causal inference in auditing research, similar ideas apply to causal inference in other areas of accounting. Researchers pursuing their questions involving diagnostic reasoning would also benefit from consideration of the framework.


    Srivastava, R. P., T. J. Mock, K. V. Pincus, and A. M. Wright. 2012. Causal Inference in Auditing: A Framework. Auditing: A Journal of Practice & Theory 31 (3): 177-201.

    applied probability, audit analytical procedures, audit judgment, causal inference, causal schema, discounting, multiple hypotheses, uncertain reasoning
    Purpose of the Study:

    Causal inferencethat is, determining the root cause(s) of an observed eventis a pervasive and crucial component of many audit tasks, such as analytical procedures. Auditing standards stress the need to approach audit tasks with professional skepticism, encouraging the consideration of multiple hypotheses or causes, and the careful evaluation of evidence.

    The authors decide to develop an analytical framework that could address the complexities present in many audit judgment tasks. This paper describes the resulting framework, including appendices containing the key proofs and derivations related to the debate. The framework provides valuable insights for conducting experimental audit judgment research on causal inference and for developing the skills of novice auditors. The authors examine four realistic conditions that are present in audit practice.

    Causal inference tasks in auditing share four important characteristics:

    • First, there are typically multiple potential causes for an observed effect; that is, the decision-maker inherits or generates multiple hypotheses.
    • Second, hypotheses may be non-exclusive; that is, more than one cause may contribute to an observed effect.
    • Third, the hypotheses set under consideration may be non-exhaustive; that is, the auditor acknowledges not all potential causes may have been identified.
    • Finally, causal inference tasks usually entail relying on evidence that is not perfectly diagnostic; that is, evidence that may be convincing, but does not establish the cause(s) with certainty.
    Design/Method/ Approach:

    The authors develop a general framework for auditing causal inference and explore the implications for various cases an auditor may encounter.


    The key analytical findings include:

    • The assessment of the potential causes of an unexpected fluctuation (anomaly) and the related discounting or inflating of causes is contingent on the number of potential causes and their interrelationships.
    • The type (positive, negative, or mixed), quantity (single or multiple items), and strength of audit evidence have a dramatic impact on the magnitude of probability revisions.
    • In a common audit setting with non-exhaustive and non-exclusive hypotheses, inflating should occur in the posterior probability of a cause, say A, if there is positive evidence that another cause, say B, is present, as long as A and B are positively correlated. In some situations, this will lead to superadditive probabilities.
    • In the audit setting with non-exhaustive and non-exclusive hypotheses, the posterior probabilities of all the hypotheses may logically add to more than 1 (superadditive) when there are strong positive items of evidence in support of several of the positively correlated hypotheses or when some of the hypotheses are independent.
    Auditor Judgment