How a poll of Best answer on the web

January 9, 2009 on 7:33 am | In mybachcars.com |
How a poll of Best answer on the web
  • There have been many polls of who's in the lead in the presidential race, but I was surprised that a national poll of less than 800 was used to determine who's in the lead. Can someone point me to critiques or support of the accuracy of polling with such a small sample?


  • Firstly the key point here is that the sample taken is "representative".Just like a doctor who takes a sample of blood ( doesnt pump out the whole bloods out of you) for diagnosis.
    The margin of error due to sampling decreases as sample size increases, to a point. For most purposes, samples of between 1,000 and 2,000 respondents have a sufficiently small margin of error that larger samples are not cost-effective.
    If there was even a sub grouping required within these (say you wanted to find out what percent of the people who say they vote for democrats also are from a particular minority) then you would definitely need a larger sample size.


  • In general, the margin of error in a poll is inversely proportional to the square root of the sample size. A poll of 10 people would have a standard error of 1/sqrt(10) =~ 31.6%. A poll of 100 people would have an error of 1/sqrt(100) =~ 10%, a poll of 1000 people would have an error of 1/sqrt(1000) =~ 3.16%, and a poll of 10,000 people would have an error of 1/sqrt(10,000) =~ 1%.
    As you can see, one quickly reaches a point of diminishing returns in terms of reducing the standard error once the sample size reaches several hundred people (assuming the people chosen are a truely random sampling).
    See http://www.publicagenda.org/polling/polling_error.cfm
    http://www.studyworksonline.com/cda/content/article/0,,EXP545_NAV2-76_SAR542,00.shtml http://whyfiles.org/009poll/math_primer.html