Small issue with the studywriteup: the emigration pool from each country was not homogenous. Unless the authors addressed that within the study I'm hesitant to believe the outcomes. The outcome is wages-from-occupation, which occupational skills may have been the impetus for moving from a home country.
> There is also the problem of reverse causality. It is quite possible that richer migrants were more inclined to Americanise their name. (They might, for example, live in a richer area with fewer migrants, which might incentivise them to change their name to fit in). So, to establish a causal link from name-changing to wage-boosting—and here’s my favourite part of the paper—the authors turn to Scrabble. They calculate the Scrabble score for the name of each arriving migrant (the score of all the letters in the name) and show that individuals who decided to Americanise had higher Scrabble points. That finding soothes worries about reverse causality: it suggests that it was not the wealthier people that changed their names, but those with unusual names.
When I saw the headline, that's not even the reverse causality story that jumped into my mind; I thought instead, 'adopting an American name proxies for willingness to assimilate and abandon the Old Country and hence naturally predicts increased success' (similar to the 'black' name studies). Their Scrabble/unusualness metric fits this story just as well the obvious discrimination story.
I hate it when I miss valid points in my sklm :). That's a perfectly reasonable concern and an innovative bias-correction strategy.
It still misses the homogeneous point though: Russia might have a lot of doctors emigrating for higher wages, whereas Italy might have had people fleeing as refugees. We'd expect higher skilled workers with hard names to adjust their hard-to-pronounce-and-therefore-hard-to-sell-services names.
It might be innovative, but I don't think it's valid. What if having a longer name makes your 30% more likely to change your name and being richer makes you 30% more likely to change your name?
The Scrabble metric would still show that name length is a factor, but it doesn't necessarily serve as evidence that other factors weren't in play as well. Selection bias seems like an obvious suspect here and this Scrabble discussion just seems like hand-waving to me.
My assumption is that they ran a correlation test, which checks how much of the variation in name-changing is explained just by Scrabble score. If that's high enough, there just isn't much room for other factors that aren't themselves correlated with Scrabble score.
