<p>Again, this may have skew. Working class will likely perform worse on those tests than college educated. </p>
<p>I'd be more interested in seeing the data normalized with equal numbers of people in specific educational groups. Make it a good sample and then we'll see. Otherwise, you can't control for the inevitable difference in educational levels' performance on verbal IQ tests.</p>
<p>
[quote]
However, when the sample size is large enough, and the GSS is. Interesting data can be gleaned.
[/quote]
</p>
<p>There's more to building a good sample than just numbers. I can sample 1000s of people and still have a bad representation of the population...</p>