Fact that US still have to break down so much into racial cohorts in polls is weird. And offer nothing of value other than the generalizations like "who tend to be poorer" that doesn't help the discussion nor analysis. What do you do with this information in this context?
How about breaking down by religion or something else that at least points to what kind of media/message the group is more exposed to? or something that hints at extraneous interests, like homeowners or not. There are so much better ways to break down this data.
It would be nice if the actual racial distribution of the population of the county is added for context. For a random visitor from the internets the lack of context hinders the ability to understand what implications the data has.
When 90% of the population is black the conclusions are different from when the population is 90% white.
The irony of any statistics regarding race, etc. is that the very act of analysis is itself a form of unnecessary segregation.
For example, during the election I saw CNN blabbing about each candidate's popularity among white women voters or black voters or this or that, and I just turned it off. The whole problem is that people treat these as relevant and appropriate categories. The statistics should be completely uninteresting, in a color-blind world.
Yes, if you slice the stats fine enough you can eventually find some statement of apparent interest, but it comes from your stat slicing, not anything useful. You name the race and I'll find a location in the world where they have poverty rates in excess of some other race. It's not useful information.
This is why it's so important to first formulate a question, then consult the stats; examining the stats for questions is actually very dangerous, you'll always find something, but not necessarily something important. Particularly when you start moving the goalposts when the stats didn't say what you wanted them to say.
That's fair. I think my point is that a statistical representation allows you to see if one race is being favored over another one. But of course it's not a straight forward analysis as you correctly pointed with the WA black population example.
Its extremely common in public opinion polling in the US, and I've never seen an explanation for it. If the subsamples of, e.g., various non-White Catholic, evangelical, etc., groups were too small to report, you'd expect that they'd just not do a race breakdown at all, and report the overall Catholic, evangelical, etc.
I suspect that the explanation might be that not only are the non-White subsamples too small, but the racial composition of the various religious groups in the polls is (because of sample size issues) often not representative of the racial composition of the groups in the population and the sample size of the non-White subgroups in the religious groups is too small, so that its quite likely that the overall numbers for the religious groups are distorted as a result of the racial demographics being non-representative, and they don't have meaningful numbers for most of the racial subgroups, but they do have meaningful numbers for the White subgroup.
OTOH, this often results in even more misleading presentation, where the White subgroup of each religious group is all that is reported, but news articles based on the poll still treat that as if it was a result for the religious group as a whole. And it doesn't explain why the non-religious group doesn't get the same treatment, because it almost certainly would have at least as much of the same problem.
It'd probably be better -- for media polls, and media coverage of polls at least -- to either do polls with a big enough sample that your sub-samples in various non-White race/religious subgroups are big enough to report on or just not report religious breakdowns at all where the only way you feel the numbers can be meaningful is just to report the White subgroup for religious groups.
The most disingenuous use of stats I see regularly published, are those bemoaning wealth inequality between ethnic and racial groups that don't control for family makeup. There is a 30+% difference between some groups on one vs. two parent households, but I guess that doesn't fit the right narrative an would presume people have some agency in their circumstances.
I think that to make any real sense of this number, you have to understand the demographics and other characteristics of their sample set much better than the article allows you to.
all of that should reflect in economic indicators of parents and their say marital status/history etc... - you shouldn't need an imprecise proxy like race to level the playing field.
I agree - interpretation is super important. The above is a good example; someone who isn’t paying attention might not take into account that there are 5-6x the number of white vs black people in the US, which puts a very different spin on the raw figures.
All I'm saying is pooling the entire United States into a single data point is fairly useless. We need to know which demographics are the ones bringing down the average the most in order to make moves that will improve this metric the most.
Having a look at how people with various ethnic background fare is interesting and can be essential to optimise the effectiveness of health policies by targetting the right groups.
However, they are still all US Citizen, there is nothing misleading about the article or their analysis unless you want them to start splitting the stats by 'race', which would be a strange starting point (why would ethnic background be a better way to group people than socio-economic brackets anyway?).
Another issue I see based on your comment is that segmenting based on locale (diverse mix in SF, white majority in Kansas) is that it can just take what knowledge and norms exist now and harden them.
For example, according to actual data I've seen, poor people really are disproportionately likely to a) have missed payments before, b) not repay your loan, and c) belong to certain ethnic groups. Some people may fit into one of these categories and not the others, but the correlation exists, so you can't clamp down on one without also measurably reducing the others unless you go out of your way to do so.
(Don't know who's downvoting, but I'm just stating pretty well-known facts. Don't take this to indicate approval — I'm just saying that, statistically, some groups do get the short end of the stick. What you do with this information is a different question entirely.)
A lot of the things you list are not "cold quantitative analysis." I don't have strong feelings about considering racial/ethnic data but I expect a lot of qualitative factors would end up being racial/ethnic proxies.
I think your assumption here is that if you were to filter based on those things you'd get an outcome that would be representative of the total population. But I don't see why you think that. Look at relatively poor Asian kids in the NYC public schools system. Even if you were to apply your filter there, they would still likely end up massively over-represented. Similar (but less dramatic) outcomes are likely even with Whites. Even factors like social class and family income don't carve things up in such obvious ways.
How about breaking down by religion or something else that at least points to what kind of media/message the group is more exposed to? or something that hints at extraneous interests, like homeowners or not. There are so much better ways to break down this data.
reply