I don't find it very interesting what a schoolboy can infer from those stats. It takes an expert in statistics to interpret stats, and even experts can be fooled because statistics is not always intuitive.
The estimate behind the Pan Arab Games comes from 965 athletes participating in 27 different sports, only 1 of which is the sport of Athletics. The Pan Arab Games athletes are also not quite the same caliber as World Championship athletes. So any comparison between "PAG" and "WCA" estimates is apples to oranges. We could, like the fictitious schoolboy Coevett invented, infer Arabs are "dirtier". But we could equally infer Athletics is "cleaner", on average, than 26 other sports, or that World Championship caliber athletes are "cleaner" than Pan Arab Games caliber athletes. But we shouldn't place too much emphasis on fictitious schoolboy inferences, or any assumed or made up stats from Coevett.
Although Coevett simply made up the "low 20% range" figure, for non-Russian, non-African athletes, there is a data-backed reason to arrive at this estimate, for all 2011 WAC athletes, including Russians and Africans and non-Africans. Table 6 from an Appendix the same paper publishes "SSC method" data -- an alternate RRT technique with the additional ability to estimate non-compliance. The results of plugging this raw data into an SSC calculator gives an estimate of 21% doping prevalence, while estimating 30% non-compliance to the survey.
Regarding who dopes, no one actually claims "EVERYBODY dopes". WADA publishes annual figures telling us how many athletes from each country are busted for doping. These tables should serve as a minimum indication of the breadth of doping.