Lesson 40 of 1524
The Politics of Public Opinion
A poll is only as good as its sample and its question.
Practice this chapterA representative sample looks like the population on the traits that matter for the question. A large sample can still be wrong if the people in it were chosen badly.
The margin of error is the range the method expects around the result. A lead smaller than that range is not a lead the poll can honestly claim.
A usable poll
Representative sample + honest question + stated margin of error
Drop any one of the three and the number is a story, not a measure.
Worked example
A website asks readers to vote on whether a bill should pass. 40,000 people click yes. The site calls the result a national poll with no margin of error. What is wrong with the claim?
- 1The respondents chose themselves by visiting the site and clicking. That is not a random draw from the national adult population.
- 2People who visit that site and care enough to click are not a representative sample of the country.
- 3A large count does not repair a biased selection. Forty thousand volunteers can all share the same slant.
- 4Without a probability sample there is no honest margin of error to report. The figure is a straw tally of that audience.
Result: It is a straw poll of site visitors, not a scientific national poll.
Why. Scientific polling depends on how people were chosen, not on how many happened to answer. A self-selected crowd measures the crowd.
A bigger sample reduces random error only when the sample is random. It does not cure a biased list of respondents.
Practice margin
This chapter
A fresh set from this chapter only. Choose 10 or 20. Multiple choice and fill-in, with no repeat inside the set.