In my experience, there is a huge difference between blacks/Africans and "everyone else". If no or very few blacks are present in a given school, town/city, or workplace etc, diversity or composition of ethnic background per se makes very little difference. Culture and class do matter a lot, but the mapping between these and ethnic background is not straightforward and also changes over time (e.g. to take a US example, as more Hispanics are second- or third-generation rather than recent immigrants).
On the other hand, going from zero to say 15-20% black is a massive change in the energy and socio-political dynamics of the space. The great migration out of Africa is going to have much more serious consequences for Europe than the Hispanicization of the USA will for that country.
Emil are there any good statistical resources for ethnicity/race and migrant crime in Germany? I got into a discussion not too long ago with someone who pointed out the unrigorous nature of their crime stats. For example, this person stated that a gang of neo-Nazis committed a string of murders (over a period of 10 years), and that the suspects of the case included migrants (when the real perps were ethnic Germans, but that hadn’t been established yet), thus, “suspect” rates are not the most accurate measurement. I imagine the best way to calculate migrant/non-German crime is to look at conviction rates, prison population, and crimes which involve witnesses/victims, such as sexual assault or rape. That, and the German statical agencies don’t seem to have ethnic/racial breakdowns of either “non-Germans” or “Germans” (the latter could be a Kurd with German citizenship/nationality, rather than an ethnic German or West European). Their system compared to the US or even the UK, is horrible and myopic to say the least.
"Racially homogeneous counties in the USA are typically rural White counties."
How about throwing in a control for population density then? Is there a way to also control for population of neighboring counties? For example, a suburb of a big city might have the same population density as a small town, but might systematically differ in important ways.
In these homogeneity investigations, I wish that people would focus more on the more social outcomes like atomization, number of friends, marital satisfaction etc instead of wealth and crime. Bet these things would have positive homogeneity effects which survive controls for g and %white.
checked it with g only instead of the direct effects of various races, and the semipartial between homogeneity and s when g is partialed from homogeneity is -0.0012916674.
for crime, the corresponding semipartial is -0.2210903977, so it seems like the weird non-g idiosyncracies of various groups are more important for crime outcomes. I wish there was some kind of general r/k variable which is easily measurable widely used/popular.
Partial correlation is equal to assuming causality is A<C>B, then you relate A~~B after controlling both for C. This is not generally what one wants to do.
I can just send you the data, you can use the proper ways.
In my experience, there is a huge difference between blacks/Africans and "everyone else". If no or very few blacks are present in a given school, town/city, or workplace etc, diversity or composition of ethnic background per se makes very little difference. Culture and class do matter a lot, but the mapping between these and ethnic background is not straightforward and also changes over time (e.g. to take a US example, as more Hispanics are second- or third-generation rather than recent immigrants).
On the other hand, going from zero to say 15-20% black is a massive change in the energy and socio-political dynamics of the space. The great migration out of Africa is going to have much more serious consequences for Europe than the Hispanicization of the USA will for that country.
Emil are there any good statistical resources for ethnicity/race and migrant crime in Germany? I got into a discussion not too long ago with someone who pointed out the unrigorous nature of their crime stats. For example, this person stated that a gang of neo-Nazis committed a string of murders (over a period of 10 years), and that the suspects of the case included migrants (when the real perps were ethnic Germans, but that hadn’t been established yet), thus, “suspect” rates are not the most accurate measurement. I imagine the best way to calculate migrant/non-German crime is to look at conviction rates, prison population, and crimes which involve witnesses/victims, such as sexual assault or rape. That, and the German statical agencies don’t seem to have ethnic/racial breakdowns of either “non-Germans” or “Germans” (the latter could be a Kurd with German citizenship/nationality, rather than an ethnic German or West European). Their system compared to the US or even the UK, is horrible and myopic to say the least.
"Racially homogeneous counties in the USA are typically rural White counties."
How about throwing in a control for population density then? Is there a way to also control for population of neighboring counties? For example, a suburb of a big city might have the same population density as a small town, but might systematically differ in important ways.
In these homogeneity investigations, I wish that people would focus more on the more social outcomes like atomization, number of friends, marital satisfaction etc instead of wealth and crime. Bet these things would have positive homogeneity effects which survive controls for g and %white.
Don't have such measures for counties usually.
ik. I'm just saying it sucks.
checked it with g only instead of the direct effects of various races, and the semipartial between homogeneity and s when g is partialed from homogeneity is -0.0012916674.
for crime, the corresponding semipartial is -0.2210903977, so it seems like the weird non-g idiosyncracies of various groups are more important for crime outcomes. I wish there was some kind of general r/k variable which is easily measurable widely used/popular.
Why would you use this method over straight up regression? Partial and semi-partial correlations are hard to interpret.
Did that because it's quick/possible to do purely with the summary stats as posted here.
>hard to interpret
A) i think otherwise; B) in what sense?
Partial correlation is equal to assuming causality is A<C>B, then you relate A~~B after controlling both for C. This is not generally what one wants to do.
I can just send you the data, you can use the proper ways.
the semipartials I did only controlled racial homogeneity for g while letting g still be associated with SES and crime.