The Causal Inference Problem: When Can Managers Use Data to Inform Decisions and How Can Organization Design Help Them?

Abstract

Most research on organizations presumes that leaders can direct their organizations toward a set of goals, but that ability requires leaders to anticipate the consequences of their actions, a problem of causal inference. To explore this problem, we develop a formal model of the organization as a system of causal relationships. Managers observe some elements of this system but not others. Hidden (unseen) elements can bias managers’ assessments of the expected consequences of their actions, but they only do so under a specific set of conditions. Absent those conditions, simple and even incomplete theories can often allow for accurate assessments. Interestingly, the specificity of the problematic condition also suggests several ways in which organization design could mitigate it, either by eliminating the influence of the problematic hidden factor or by revealing it to the manager.