[hammock]

Something like this, any sample size 300 or above (assuming it was drawn in a representative way) is generally regarded as legit.

Source: I run and analyze surveys for a living

[buntress]

300 is kind of a paltry number. One high school’s population, for a medium sized town.

Look at one high school, then another. Take each high school from opposite ends of the country. Same results? Not sure about that. But hey, magic number says we’re all good. Run the article, right?

[sakarisson]

You've just illustrated in your own comment that it really isn't as much about the number of participants as it is about the random distribution of participants.

Would you think it was any better if it was 3000 people from the same area rather than 300? Probably not a lot.

If you pick 300 or more random people throughout the country, you've going to get a pretty accurate representation of the average person.

[saagarjha]

I'm sure that the results hold up reasonably well if they're taken randomly (whether the survey actually is random is another story, though).

[misterbowfinger]

Thanks for correcting! Good to know!

AmeriSpeak seems to have national representation, so I can only assume it was drawn randomly throughout the nation.