Summarization with Errors: A Perspective on Empirical Investigations of Agency Relationships

Abstract

Empirical research in the social sciences often proceeds in the following manner. First, the researcher formulates an hypothesis regarding a phenomenon of interest. Second, the researcher identifies the empirical predictions of competing hypotheses. Third, the researcher gathers relevant data, and conducts statistical tests to determine whether the data support the predictions of the maintained hypothesis.

Typically, the researcher does not attempt to model formally every aspect of the phenomenon under study. Analytic and computational complexity necessitate a focus on only the most important and the most fundamental forces. Indeed, the real art to much of social science research lies in distinguishing between the forces that are central to the phenomenon under consideration and the forces that, while of some potential relevance are less central. Any such differentiation between central and less-central forces raises the possibility that important elements of the phenomenon under consideration might be omitted from the analysis. The differentiation also implies that empirical research in the social science must be viewed as a form of summarization with errors. At best, the empirical analysis summarizes imperfectly the central forces at play. At worst, empirical research can overlook critical forces and thereby misinterpret the observed statistical relationships among key variables.

The purpose of this research is to examine some of the key summarization errors that are prone to arise in empirical investigations of agency relationships. Broadly speaking, agency relationships entail interactions where incentives must be employed to induce desired behaviour.

JEL Classification: M40, C12, C51, D82