Can one get their hands on this data? 
Brazil has ~5500 municipalities. Everybody takes a scholastic test every year, sample size was 8.6M in 2016. So we calculate White% population and average test score. Looks like this.pic.twitter.com/W2eLtv0MrF
-
-
-
It's all public. Massive amounts of it. This one is from Enem 2016. http://portal.inep.gov.br/microdados Spatial data here ftp://geoftp.ibge.gov.br/organizacao_do_territorio/malhas_territoriais/malhas_municipais/municipio_2018/Brasil/BR/
-
You are great!
- End of conversation
New conversation -
-
-
Seems pretty misleading since there are many omitted variables. It also assumes white% exogenous. Did you try controlling for income, education expenditures, poverty levels?
-
Main causal pathway is white% -> intelligence -> social outcomes. Controlling for social outcomes is the sociologist fallacy.
-
How did you get a causal pathway out of a simple regression?
-
Never said I got it out of this regression.
-
That's true. I just don't see how this regression is meaningful at all with only one explanatory variable
-
Weird comment. I hope you keep your standards and say the same thing for every other scatterplot you see. ;) Joking aside, this plot is highly informative because many people don't know of this pattern, but it's present all over New World. Ancestry is root cause of inequality.
End of conversation
New conversation -
-
-
"Everybody takes a scholastic test every year" Very interesting. Any other countries do this?
-
A lot of countries do, they just don't publish the data. E.g., review in http://www.nefmi.gov.hu/europai-unio-oktatas/eurydice/national-testing-of …
End of conversation
New conversation -
-
-
Nice data set and interesting analysis. The endogeneity problem is obvious, though. Did you control for average income/wealth in the municipalities?
New conversation -
-
-
Th variance of the municipalty means after controling for european % is interestingly quite large. Larger than I'd have naively expected. Some of it self sorting, no doubt, others just SE. What accounts for the rest? Also too many data points at 0,50,100% due to rounding.
-
It's not European%, it's self-reported White (Branco) status. The amount of European ancestry this amounts to varies by region, so it's not that great a proxy. I think values at 0.5, 1.0 due to having 2 cases in that unit.
End of conversation
New conversation -
-
-
Worth noting that Enem became a heavily politicized exam in the last few years, especially the essay. They became left-wing-ideological tests that students then learned to game in order to score high. Data on older tests (2008 or prior) and excluding the essay might be the best.
-
I did not use the essays on my scoring.
End of conversation
New conversation -
-
-
Could you do the same with a loess fit line?
-
Thanks.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.