After spending maybe more time than I meant to digging into one of the most popular sources for the econometric history of Africa, I found some time to write up one very small problem with George Murdock’s Ethnographic Atlas and its uses by economists: the ‘focal year’ variable, often used in regressions as a control for the ‘year of observation’ of the data, is not actually a year of observation, and if that’s what you’re using it for, then one of your controls may be measured with error (which, depending on the correlation structure of your RHS variables, may mean that other coefficients are contaminated. Attenuation bias is not the only kind of bias you need to worry about with an iffy dataset!). Here is the draft: comments welcome.