Humanities & Social ScienceThere are several questions regarding speech which which have huge sample sizes in the GSS:
SPKRAC (Allow Racist to Speak), SPKHOMO (Allow Homosexual to Speak), SPKCOM (Allow Communist to Speak) and SPKATH (Allow Anti-Religionist to Speak) all have sample sizes of 53,000. This means that one can look for trends at a relatively granular scale. I decided to check how people lined up as a function of Age, belief in God, political views and intelligence (vocabulary). Lots of charts below.
A key:
1) Age goes from 18-89, left to right.
2) God goes from (left to right) atheist, to agnostic, to higher power, to some belief, to mostly belief and some doubt and finally those who know god exists.
3) Politics goes from extremely liberal on the left to extremely conservative on the right (with liberal, slightly liberal, moderate, slightly conservative and conservative in the middle).
4) Vocabulary scores go from 0 out of 10 on the left to 10 out of 10 on the right. Dumb to smart.
Observations:
1) The young support free speech more robustly than the old...but, the very young are waning in their support for free speech compared to the older genration. The strength of the second trend varies by question; Gen Y is far less absolutist on free speech for racists than Gen X than on the other questions.
2) There's a pretty clear monotonic relationship between stupidity and support for censorship. I won't bother to report the data by degree attainment because it's what you'd expect.
3) It looks like extremism in position might have some relationship for intolerance of socially marginalized stances. Look at peaks in support for free speech among the agnostic. Remember that the sample sizes here are large, so the small but consistent difference between atheists & agnostics are likely real and perhaps reflect the difference in mental outlook between the two groups. Additionally, there is little gain in support as one becomes more liberal.
4) Political moderates are relatively stupid. I suspect that explains some of their aversion to offensive speech...but even when I filter for intelligence it looks like moderates have an inordinate aversion to absolutist free speech. I guess one can chalk that up to the fact that as moderates they're averse to extremism in general. The political charts exhibit an interesting dynamic, as moderation, extremism and Left-Right variables are all at work.
Comments
You've got plenty of data there for a good multiple regression. How about computing regressions correcting for the other factors? Things like age and liberalism and religiousness are all correlated, but you might be able to partial out the strongest factors... And then plot the graphs based on the marginals rather than the raw data...
Posted by: Harlan | March 10, 2009 10:42 AM
Did you produce plots for "Allow Militarist to Speak"?
Thats the one I'd naively expect to break the trends all the others seem to share, although perhaps only as a function of political orientation.
Posted by: Dave S | March 10, 2009 11:01 AM
It's good to know that the statistical evidence supports what I had long suspected just based on personal experience.
Posted by: The Science Pundit | March 10, 2009 11:10 AM
So as a moderate on pretty much everything except free speech issues I must be an outlier.
Joking aside, I'm not sure that the questions as a whole are that great for measuring attitudes towards free speech. In particular, given that agnostic liberals are more likely to have sympathy for communist, anti-religious or pro-homosexual views one would expect them to support free speech for those groups more. Now, it looks like the general pattern still occurs when one looks at the data for the question about allowing racists to speak. So the pattern may be real beyond just the general questions asked. But given that data set, I'm not sure a strong conclusion can be drawn in that regard.
Also, questions of vocabulary might say more about education than intelligence intelligence. It might be interesting to correct for education levels and see if the strong correlation with vocab still held then.
Posted by: Joshua Zelinsky | March 10, 2009 11:58 AM
Thats the one I'd naively expect to break the trends all the others seem to share, although perhaps only as a function of political orientation.
same.
Posted by: razib | March 10, 2009 3:05 PM
How about allow human biodiversity-ist to speak? :-P
.... about human differences as clinically detached and as without value judgement (therefore as non-racistly) much as a BBC documentary about say, different toad species.
;-P
Posted by: deadpost | March 10, 2009 3:29 PM
You might look into PNG for your graphs instead of JPEG. JPEG is suitable for photographs, but PNG is far superior for graphics with sharp lines, and would make them look far more professional.
Posted by: Ryan | March 10, 2009 3:56 PM
Your assumption that vocabulary correlates directly to intelligence is pretty problematic. Non-native speakers, lower-class individuals and people of color might all have less developed vocabularies than upper-class white Americans, but that doesn't make them stupider.
Posted by: Andrea | March 10, 2009 3:58 PM
Your assumption that vocabulary correlates directly to intelligence is pretty problematic.
no it's not.
Posted by: razib | March 10, 2009 4:03 PM
So, you would categorically argue that non-native speakers of English, and speakers of creole dialects (for example, Jamaican creole: http://www.jumieka.com/) are less intelligent than well-educated upper-class American liberals?
Posted by: Andrea | March 10, 2009 4:18 PM
@Harlan: Do it yourself; the data are all in the GSS link. You could probably even write a script to fetch the data for you.
@Joshua: Of course the vocabulary test says more about education than intelligence. So do standardized undergrad/graduate admissions tests. So do IQ tests, for that matter. What's "intelligence intelligence?"
Also, I ran SPKMIL on the linked database, and it follows the same trends as the other SPK's, though there are probably some differences that I couldn't pick out off the cuff.
Posted by: Luke | March 10, 2009 4:24 PM
So, you would categorically argue that non-native speakers of English, and speakers of creole dialects (for example, Jamaican creole: http://www.jumieka.com/) are less intelligent than well-educated upper-class American liberals?
no, i'm not retarded
1) the sample is 95% native born english speaking
2) there are no colored people in it (look at the text)
so your objections are irrelevant. the question is how well vocabulary tests correlate with general IQ. last i checked, pretty well. there's no "perfect" measure of intelligence. but people with smaller vocabs tend to be less intelligent.
Posted by: razib | March 10, 2009 4:48 PM
Out of interest, is the Wordsum a better proxy for intelligence than the reasoning tests (ie the 'Alike' questions) on the GSS?
Posted by: Ross | March 10, 2009 5:11 PM
I would be interested to see a few more questions concerning for example:
1)those who don't believe in anthropogenic global warming
2)religion, from the opposite standpoint
Posted by: Tom Bri | March 10, 2009 5:21 PM
amazing that every discussion of wordsum turns into an argument about the validity of testing -- it's not 1970, these things have been studied.
little known/appreciated conclusion from psychometrics -- the quality of an IQ test is surprisingly difficult to estimate from a superficial examination of content of the items themselves. this is why IQ test construction is such a data intensive business and why test makers guard their tests so jealously.
other vocabulary-based IQ tests are really good so long as the tested individuals come from the general English-speaking US population -- such as the GSS sample. wordsum is short, relatively reducing its reliability, but in practice its not bad for this kind of aggregated analysis.
also, multivariant genetically-informed analyses do not support the contention that education (currently) plays a substantial *independent* role in test performance. that is, if you forced everyone to get the same amount of education, you won't change their performance on tests very much relative to the current circumstances. note that's not saying that education isn't an important causal factor -- so too are a myriad of factors such as say nutrition (or healthcare or whatever) -- just that you can't increase education to achieve improvements in test performance outcomes all that much (much more than increasing nutrition or healthcare would).
Posted by: richard | March 10, 2009 5:34 PM
Very interesting study.
Stark charts.
Thanks for posting.
I have never understood why people are against free speech. If the people are promoting something stupid, letting them speak is the fastest way to reveal it.
Posted by: sg | March 10, 2009 5:40 PM
It would be interesting to pose the question "should women be allowed to speak?"
Posted by: James | March 10, 2009 9:35 PM
Eighteen percent of atheists wouldn't let an anti-religionist speak? What?
Posted by: Z. M. Davis | March 11, 2009 12:08 AM
I would of liked an "Allow child-sex advocate to speak" variable to be added, because as far as socially unacceptable positions go I would venture to say that it is the most unacceptable of them all. I doubt that even most young irreligious liberals would be particularly supportive of the NAMBLA crowds right to voice their opinion.
Posted by: B.B. | March 11, 2009 5:24 AM
B.B. that would effectively be a purely a measure of free speech absolutism, since apart from the a few pedophiles there is essentially no sympathy for that position. Sounds like a good idea, and one could correlate it with an explicit question about free speech absolutism.
And then I'd be interested in substracting that percentage from the various groups here and seeing perhaps a measure of their degree of sympathy for the various positions/groups represented. I.e. are liberals really more sympathetic of racist speech or are they more into free speech absolutism?
Posted by: Coriolis | March 11, 2009 11:36 AM
The "Moderate Pinch" is very interesting, I wonder if your view of them is accurate. What's the data on vocab scores, IQ or other intelligence measures for people claiming moderate political views?
Posted by: mnuez | March 12, 2009 8:46 AM
This was interesting. Kind of funny how many negative comments ended up on this page, but I found it amusing anyway.
Posted by: David | March 12, 2009 10:00 PM