Question Authority -

[kyburg]

Remember the whiny, insecure kid in nursery school, the one who always thought everyone was out to get him, and was always running to the teacher with complaints? Chances are he grew up to be a conservative.

At least, he did if he was one of 95 kids from the Berkeley area that social scientists have been tracking for the last 20 years. The confident, resilient, self-reliant kids mostly grew up to be liberals.

OH RLY.

Nah, the folks in Berkeley don't have a axe to grind or anything.

It's an interesting study on 95 subjects, hardly double-blind and where was this done again? *snaps fingers* OH YEAH.

Confident, resilient folks comes in every stripe known to man - the rigid, uncomfortable, insecure ones? Extremists of every caliber.

Conservative or Liberal. Bah. What a crock. Like you're completely one or the other - and your behavior dictates the decision, not the content.

Dumb as dead cats.

But that's just my opinion - and I encourage you to question it.

Comments

[http://users.livejournal.com/little_e_/]

Nah, the folks in Berkeley don't have a axe to grind or anything.

It's an interesting study on 95 subjects, hardly double-blind and where was this done again? *snaps fingers* OH YEAH.

Um, yeah...

You can't do this kind of study as a double-blind study. Double blind is where you take a population of subjects, do a real experiment on half of them and a fake experiment on the other half, and don't tell anyone involved which half is which. For example, out of 2,000 people, 1,000 received medicine XYZ to deal with their arthritis and 1,000 received placebos. You then write down everything that happens to everyone, and at the end, after all of the medicine's been taken and the funding's almost out, you look up who had which and see if anything correllates. If the people on placebos report just as much pain relief as the people with the medicines, then we know the medicine isn't causing pain relief.


This, by contrast, is an observational study, to see if there's any correllation between nursery school behavior and later political behavior. How should we do this double-blind? Put half the kids in nursery school and half not? You can't. The children are not being experimented on. The fact that this study was not double-blind in no way discredits it because it did not attempt to measure anything which was done to the group of children.

Now, if the people were selected based on political affiliation, IE 100 liberal and 100 conservative adults, and then the researchers then went and interviewed their old preschool teachers about their behavior all those years ago, then we could say that there's bias. But since the subjects were chosen and observed long before the manifestation of anything even remotely resembling political opinions had observed, we can't claim bias on those counts.


The fact that the study only followed 95 is troubling. It is too small a sample to really tell us much, and the fact that they all lived in the same environment also means the results can't be extrapolated very far, exactly as it says in the article. We could claim that conservatives are more insecure, or we could claim that insecure people who grow up in an extremely liberal environment will react to that by becoming more conservative. It may be that insecure people who grow up in extremely conservative environments become liberals.

Finally, the fact that the folks at Berkley would prefer a certain result doesn't make them incorrect. Almost all scientific research is done with the intention of producing a certain result. While I personally think this leads to a lot of bad research, in a simple observational study done with an eye towards issues of size and applicability (on which points the study really fails) the results can be quite accurate. If we say the Berkley folks dislike conservatives and want to prove that conservatives are bad/icky/wrong, we could just as easily say that evolutionary biologists have an axe to grind and want to prove creationism wrong. Almost all scientists have something they want to prove/disprove when they do their research.


At any rate, the article only reports a 7% correllation. So, you know, there's a 7% chance the whiny kids will turn out more conservative than they would have by chance, while there's a 7% chance the self-reliant kids will turn out liberal. (excuse me if I'm misinterpreting this point, I haven't slept in about 24 hours.) So that's not exactly a huge correllation. Enough that it might be statistically robust, but small enough that other factors (such as family life, genetics, getting dumped in hs, etc.) may play a much bigger and more important role.

Which, you know, makes the title of the whole thing rather deceptive, but '7% chance whiny kids grow up to be conservative' just isn't that catchy.

[foogod.livejournal.com]

Actually, I note that the reporting of the results in this particular article appears to be rather inconsistent. They actually state that there was a .27 correlation (27%), but then in the next sentence say this amounts to a 7% contribution to conservative/liberal, which makes no sense to me. A 7% correlation would not be anywhere near statistically significant; however if r=0.27, then this is actually fairly significant for that sample size by most scientific standards (alpha < 0.01).

There are serious questions about sample selection, however, particularly considering they were reportedly using a sample set which was originally selected for completely unrelated research. If the sample was chosen largely randomly (which would likely be the best we could hope for under the circumstances), from my own personal experience I don't believe that the "whiny, dependent kids" would account for fully half of such a sample. Likewise, it seems unlikely that of 95 randomly-chosen individuals in Berkeley, 50% would end up being conservative, so it seems likely that both the dependent group and the conservative group were substantially smaller than their alternates, which would reduce the significance by an appropriate degree, possibly even casting a .27 correlation into the range of a borderline result.

As mentioned, there are also some obvious issues with external factors (environment, parenting, etc), and although I can't really tell without reading the actual study, the characterization makes me somewhat suspicious of the precision of the chosen hypothesis criteria as well.

It should also be noted that we're talking about political orientation of people in their early 20s, basically at the beginning of the political part of most people's lives, so whether this has any bearing on longer-term political affiliation is debatable as well (it's been my experience that many people get more conservative as they get older, for a variety of reasons).

I don't see anything in this to suggest that there was undue bias or other "axe-grinding" involved, however. There are undoubtedly some issues which make the significance of these results debatable, but I think automatically assuming they're not legitimate just because of where (not even who, but just where) they came from is a bit of a biased conclusion itself.

[http://users.livejournal.com/little_e_/]

Yeah.

I am basically in agreement.