A Dynamic Ordinal Item Response Theory Model of Political Repression and Accountability Standards

I argue that the use of repressive policy tools appears to have remained flat over time because of a changing standard of accountability that reporting agencies use to hold states responsible for abuse. The standard of accountability is a set of expectations or norms that state behaviors are measured against. Though scholars have discussed the possibility of a changing standard of accountability at length, this concept has not been systematically incorporated into measurement models of repression. The quality of inferences made about repression levels in different countries and over time depends on choosing a theoretically informed model that best approximates reality. This paper is the first to test for a changing standard of accountability in a model of repressive outcomes by extending a dynamic Bayesian measurement model to incorporate multiple sources of events-based and standards-based information on repression. The measurement model extends existing latent variable models by (1) accounting for the fact that human rights indicators can be more or less informative about the latent level of repression, (2) providing a theoretical motivation for the modeling of temporal dependence in human rights levels within countries over time, and (3) and providing a theoretical motivation for the changing baseline probability of being coded at a given level of repression on the repression data included in the model. I also present several techniques that each demonstrate that the model outperforms alternatives. The analysis finds substantial evidence for an increasing standard of accountability, especially for the violation of the right not to be tortured. By allowing this standard to vary with time, a new picture emerges of improving conditions over the period of study (1949-2010).