Footnote 48, Chapter 3, Page 67:

Fearon and Laitin (2003) point out additional drawbacks of the MAR dataset which are less likely to impact this project’s findings. The aspects of the MAR dataset that have been used here are its identification of cases containing MARs (for compilation of cases of plural societies) and data for two dependent variables, “Protest Index” and “Rebellion Index.” They argue that, “despite careful coding procedures,” the dataset still has “substantial coding errors” (Fearon & Laitin, 2003, p.5). They then discuss one variable represented by data that has been affected substantially by such errors. This variable measures linguistic difference between minority and dominant groups but it is not used here. I am not aware of any substantial coding errors affecting data representing the dependent protest and rebellion variables. Fearon and Laitin (2003, p.7) also observe that a number of variables they might have expected to have been included, such as GDP, population growth, and “basic aspects of state structure,” were not included in the dataset. Those exclusions do not impact this project. They argue that a “considerable number of the MAR variables seems to be endogenous to ethnic conflict or cannot be reliably coded independent of observation of the value of the dependent variable” (Fearon & Laitin, 2003, p.7). Domestic conflict and antagonism are not represented by independent variables for this project so that problem does not pose problems here.