Apophis

2300 people isn't a lot. I'd like to see this re-performed with a larger sample size and a more diverse population. I still found the results interesting and don't feel that it in itself is a headline-grabbing survey. I think the opinion article was headline-grabbing and meant to stir up controversy, but the survey itself doesn't have that feel to me.

__________________
Diplomacy is the art of saying "good doggie" while looking for a bigger stick.

leejo

I guess they got lucky and picked the 276 non-racist republicans and the biggest scumbag racist 2024 democrats.

Whew!

Kerostasis

Statistically speaking, anything over 1000 gives you a pretty good picture of the population, most of the time. When they do polls on the Presidents approval ratings, they only poll 1 to 2 thousand people.

Of course, with only 270-odd republicans responding, I would feel much less confidant about the accuracy of that portion of the study.

Edit: Anecdotal evidence time! My dinner crowd is mostly full of people without strong political beliefs, but there is 1 far-left liberal in the group. So tonight, she complained that we shouldn't be spending money on building hospitals in Iraq, fearing that there might not be enough money left over to provide her with FREE health care here in the US. The potential to save hundreds of lives of non-whites somewhere else was meaningless, if it meant she might have to pay for her OWN health care.

So, does that count as racist?

__________________

Karkianman101

Only if you try to deconstruct it into being about race. You could really pick just about any kind of general characteristic about the people over there (muslim, non-white, poor, non-english speaking, non-american, non-north american), and blame the choice on one of those factors. I also don't know whether you are saying that she actually got upset about not getting her own healthcare, or if you yourself deduced that as the motivation, and even if she did say so, it may have been more for example than to actually complain about her not getting her healthcare.

Kerostasis

No, she didnt complain about not getting health care. She complained about not getting FREE health care, using those exact words. There wasn't really much interpretation required.

__________________

Wulfyn

The survey is flawed. The collection of data is not the issue here, nor the sample of the population (2300 is a decent number, most biological science research (as opposed to medical testing) is conducted using only a few hundred samples in the US, and often less than 100 in the UK). The problem is completely with the presentation of the analysis.


The key headline which I was looking for from the first paragraph, and which I was not disappointed in finding was the sample of the population that was used was predominantly white. To get an investigation such as this and then to produce a report that stipulates that one part of the results was dependent upon another is not good science, nor statistics. It is propoganda. Remember stats is all about how you present the results.


To highlight this issue I'll pull a key point:

This shows that we have another very strong demographic which the report does not highlight (possibly as it is far less controversial). You could conclude that as more money was given to white people and the sample was largely white that people were just more likely to give money to those of the same colour of skin as they were, and thus that on the whole racism exists. Controversial I know

The report goes on to speculate that the sample that falls into the white educated bracket are also more likely to be liberal, but I did not quote that as I did not see any raw numbers to back it up (the article seemed to imply this rather than state it so I was a touch skeptical).


As an example of this sort of thing I'll give you a real world example that I face in my job. I work as an underwriter for an insurance company and a large part of my job is building models on risks and premiums and propensities. Each question you answer has an associated rating factor. 2 common ones are driver age and how long you have been driving for. The stats agree with common sense - young drivers have more accidents and drivers with less experience tend to have more accidents (or at least more expensive accidents). Thus the rating factor for under 20's is much higher than for the over 30's, and the rating factor for people who have been driving for under 2 years is much higher than those that have been driving for more than 10.

There is a real problem tho. A 19 year old driver can at most have been driving for 2 years, and thus accidents caused by 19 year olds also fall into the accidents caused by people with little driving experience. This would be unfair to people who learn to drive later as you could be applying a disproportionate amount of accidents based on associated but uncorrelated factors (in this case thinking that drivers with little experience are more dangerous than they are because of accidents caused by being young rather than just inexperience). The same applies here, if the majority of the liberals are white (and the majority of white are liberal) then whites favouring whites would make it appear that liberals favour whites.


There are two ways that you can overcome this. Either you knock out one of the associated factors to get a base line (science) or you take a sample which accurately reflects the larger population (sociology). The first gives you an accurate reflection of what an individual is likely to do in order to create models from the ground up, whereas the latter tells you what is likely to happen (which very often is a lot more useful). Top down or bottom up, take your choice but try not to cross to far into the other side.


Also:

I don't have the raw numbers to do any sort of stat analysis, but a $100 difference per month over a sample of 2300 people doesn't sound like a huge amount, and maybe not enough to be statistically significant. Simply stating a $1000 difference makes it sound like a lot more than it is when you consider that this difference is spread over a year. I'm not trying to say it is inconsequencial, but it is a case of presenting in the best light again. It's akin to a debate I gave back in college where one of my classmates did not feel that there was much of a difference between 80% and 97% (being only 17%). However when presented as the difference between 1 in 5 and 3 in 100 it sounds a lot worse.



To summarise I'm not trying to say that liberals are not racist, just saying that the presentation of the results is deeply flawed. My advice is to make sure you do enough math to be able to work out your own conclusions and not have to rely on second hand evidence.

__________________

xTYBALTx

If the conservatives in this study were indeed Stanford university students and facutly, as it appears at first they are, then it stands to reason that they have unusually strongly held beliefs and are of strong character and intellectual disposition - they've managed to sit through countless hours of liberal blowhard professor's presentations and still hold strong.

The liberals in this study, however, have largely bent to carry the majority view. That means that they are not as "dominant" - that they will bend a bit to agree with whatever they're hearing; including, it just might happen, a few news reports which make black perpetrators seem more violent or generally worse than their white counterparts.

Might this study expose biases in reporting moreso than biases in political groups?

And, in reaching back to my faded PSTAT5 memories in 2002, I recall needing a sample size of only a few hundred to extrapolate one's findings to the 300MM Americans.


Edit:

The survey size required for a 95% confident survey with a 5% confidence interval (margin of error) to be extrapolated upon a population of 300,000,000 people is 384. For a 99% confident level, one needs a sample size of only 666 people.

Linkage.

__________________
A policy of freedom for the individual is the only truly progressive policy. -F.A. Hayek