[citation][nom]drutort[/nom]thats a very small sample to draw a conclusion for whole America LOL really? only 2.2k adults? and its done over phone and cell phone... ok why not like public survey or other none tech surveying... how about asking others who have Internet if they have adults who never connect on the net.. that might work for an online study lol though not very accurate[/citation]
Actually a survey size of 2,200 people allows for a very manageable survey with a high certainty. For example, the Gallup Polls that correctly predicted the 2008 election with a sample size of 3,050 had only 2.1% error. The reason why the survey was conducted over cellphones and phones is that pseudorandomness is easily attained. A public survey has regional bias and asking internet users if they know people who do not use the internet has a socioeconomic bias. While using mail would have been suitable, it has a much higher overhead with not noticeable benefits over a phone survey. From the information given in this article, the survey looks as if it could be very accurate; however, I do not know any other details about it.