the construction of human rights: accounting for systematic bias in common human rights measures

Empirical human rights researchers frequently rely on indexes of physical integrity rights created by the Cingranelli-Richards (CIRI) or the Political Terror Scale (PTS) data projects. Any systematic bias contained within a component used to create CIRI and PTS carries over to the final index. We investigate potential bias in these indexes by comparing differences between PTS scores constructed from different sources, the United States State Department (SD) and Amnesty International (AI). In order to establish best practices, we offer two solutions for addressing bias. First, we recommend excluding data before 1980. The data prior to 1980 are truncated because the SD only created reports for current and potential foreign aid recipients. Including these data with the more systematically included post-1980 data is a key source of bias. Our second solution employs a two-stage instrumented variable technique to estimate and then correct for SD bias. We demonstrate how following these best practices can affect results and inferences drawn from quantitative work by replicating a study of interstate conflict and repression.