From the FAQ on that site : "Most of the musical artists are popular across wide SAT ranges"
Since the data is not consistent, that makes it a lot harder to say listening to that band makes it more likely that you are dumb or smart. I expect if the sample size was bigger and the selection of participants more representative, then the differences here between bands would smoothen out more.
And of course music doesn't make you dumb .. It's just a hook.
edit :
Just for fun and to prove a little point, I used my own
die-roller script to generate 10 numbers between 1 and 400 which is roughly the range of the SAT scores. Then I divided each value by 10 to get the averages.
These are the numbers I got :
160,6
174,0
194,7
220,9
210,4
241,7
181,0
162,7
209,3
221,9
Most of them are relatively uniformly distributed around the mean, but I've bolded the highest and lowest scores. What you can see here is that even though we
know the real average here is ~200, we have one value at 160 and one at 240.
(To check if the average is ~200, I can pick 1000 numbers between 1 and 400 and see if that approximates 200000 .. then I get 196325. 196325/1000 = 196,325 ~ 200)
If you say that 240 is U2 and MC Hammer is 160, then you can draw the conclusion that people listening to MC Hammer are more stupid than people listening to U2. However, in this example, it doesn't matter what they listen to since they are all random numbers. The difference is explained by stochastics, random chance. I don't know what kind of numbers of observations he has behind each group here, but I assume they are very small numbers and that makes it almost certain that random chance will mess with his results. F.ex, how many people are behind Beethoven's ratings on the chart?
Abuse of statistics, I think.