This tells us nothing of who is actually traveling to those countries. For instance, is it mainly Pakistani Americans traveling to Pakistan? How does that play out with where they are going and who they are interacting with? Additionally, the state department may be making an attempt at predicting violence in a specific country or region of a country as opposed to looking at historical data. If they feel a conflict is brewing, perhaps they advise to stay away. They have many potential reasons to issue a warning. Additionally, those not heeding the warning may be safer in the regions they travel to depending on who they are or their experiences there. An uninitiated Caucasian would do well to avoid the Philippines alone, but someone who grew up there would generally know how and where to stay safe. This is well illustrated when you consider what happened to German tourists visiting Miami for nearly a decade (it didn't go well). Simply put, the data doesn't tell us much. It lacks context.
“On the whole, there is a significant relationship between the number of American deaths abroad per capita and the number of travel warnings a country receives (r = 0.56, p = <.001)” the graph associated with his claim does not look anything like a linear relationship.
I feel like the comparison of deaths abroad to total population of the US is a bit disingenuous. Surely, not all 316 millions travelled abroad, or travelled abroad for a significant amount of time, and so on. So comparing this arbitrary ratio to the homicides occurring in the US seems to me like comparing apple to oranges.
I wonder whether something like homicides per total days spent abroad by Americans might be a more interesting measure.
I thought that was weird too (several other things on the site seem that way at first, until they explain something like "oh by the way, we filtered only for murders, terrorist acts, etc., not naively by number of deaths"), but it's the definition of "per capita" that is weird. If you look at the table, per capita is number of murders per travellers (to that country) times 100000, which seems fairly reasonable, though indeed the amount of time there also should factor into it (maybe I still didn't read far enough :).
