THE THEORY AND EMPIRICAL MEASUREMENT OF CONSERVATISM

Abstract

Conservatism is important in achieving efficient contracting and protecting the interests of debt holders. However, the lack of a standard way to define and measure conservatism results in various conservatism measurement models, leading to potential confusion or even loss of credibility in conservatism research. This paper addresses this issue by analyzing the consistency and the robustness of five major conservatism measurement models. For the consistency assessments, industries and companies are classified and ranked by the five measurement models. The correlation among the conservatism scores derived from each of the measurement models is also looked into. We find that the different conservatism measurement models classify and rank industries and companies differently. The correlation among the conservatism scores is low or even negative, with statistical significance. These results show that the five conservatism measurement models are not consistent. For the robustness assessments, after benchmarking against Penman and Zhang (2002), all five conservatism measurement models are used in the same experimental settings as in Penman and Zhang (2002). Industry analysis is introduced as well. We find that results across the five measurement models are not robust but the results within the same type of conservatism measurement models are robust, for both conditional conservatism models and unconditional conservatism models. However, at the industry level, diverse conclusions are reached across different industries with regards to robustness. Intertemporal stability remains in most cases. This paper highlights the inconsistencies and some degree of robustness among the five major conservatism measurement models, with an attempt to provide some insights into conservatism measurement and attract future research in this area.