The paper does not show foreigners commit crimes at the same rate as Germans
Instead, it claims to explain why foreigners commit crimes at higher rates than Germans.
So the title is the opposite of the truth.
Academics, please point out this error candidly, openly.
I hope @tylercowen & @ATabarrok will note the misinterpretation of multivariate regression in this paper.
And @mattyglesias could help the cause of improved empirical debates by pointing the difference between claiming something is false versus explaining why it's true.
This Reuters post about the study gets at part of the issue, what economists call overcontrol bias, what I've called an Everest Regression:
"Controlling for barometric pressue, Everest is the same altitude as Death Valley."
In Ek's excellent new @JPolEcon piece on migrants to Sweden, he reports a "significant dispersion in human capital across countries [of origin] with a 90/10 percentile ratio of 3.2."
So, which are at the 90th percentile?
The ethics review board wouldn't let him tell us.
🧵:
The screenshot above is from his ReadMe file here, part of a ZIP that includes all of the replication data.
These words should be enough to find the files via a Google search:
"Replication Data for: Cultural Values and Productivity"
You can see the replication folder includes the cross-country data in an Excel file-- with countries listed by number, not name.
"FodelseLandnamn" translates to "Birth country name," so that might have been the one column Ek was required to change by the Lund ethics board.
"A great example of Jones’s descriptive abilities lies in a description of task diversity.. by Adam Smith.. The positive channels of diversity are limited to specific settings, while the negative channels of diversity are broader & more likely to be realized."
"Jones’s greatest strength is in his ability to distill novel research into easy-to-understand concepts... His examples are excellent and easy to follow, and he frames the problems in a way that promotes further curiosity and exploration."
"Cardinal Richelieu [favored] talent importation, encouraging... the establishment of small, industrious communities of Spanish conversos... and discreetly shielding these economically productive ‘peregrinos’, or wanderers, from persecution."
Saint-Simon on how banning Protestantism hurt French productivity:
"The revocation of the Edict of Nantes... depopulated a quarter of the realm, ruined its commerce, weakened it in every direction... banished our manufactures to foreign lands, made those lands flourish and… twitter.com/i/web/status/1…
Bryan Caplan:
"[N]one of the.. papers on [immigrant assimilation] I’ve read seriously worried that self-reports.. severely understate assimilation."
From the paper I cite on Page 20 of TCT:
"[S]avings rates might suffer from self-reporting bias. To address such a concern..."
If Bryan Caplan is concerned that self-reporting bias of subjective attitudes might be driving cultural transplant findings...
....Then I hope he chooses to focus on the transplant of savings behavior I report in Chapter 1 of TCT.
Page 25 of The Culture Transplant:
Caplan has never AFAIK mentioned the Thrift Transplant.
Notably, he also hasn't discussed the evidence in TCT that across Southeast Asia, the Chinese diaspora has not fully assimilated --and that lack of assimilation has very likely been a boon to prosperity in the region.
The private return to AGI will plausibly be as low as the private return to HGI:
About 1% more output per IQ point.
So an IQ 400 AGI can do about 20X of an average person.
If you're looking for god-tier AGI, that ain't it.
The paradox of IQ is that the private return to IQ (a measure of human general intelligence) is pretty low, but the payoff to IQ at the national level is pretty high-- enough to explain half or more of the percentage differences across countries.
So smarter groups get a lot more done than smart individuals-- on a per-person basis!
Same will plausibly be true for AGI: individual payoffs to AGI will be modest--like the 1 IQ point = 1% more output-- so a single superintelligence won't be unrecognizably more productive IRL.