Leaders and reporters want to know what the public thinks, and the views you studied in political socialization add up to public opinion. Asking everyone is impossible, so polls ask a sample and estimate.
| A good poll | Why it matters |
|---|---|
| A random sample: everyone in the group has a known chance of being chosen | So the sample looks like the whole population |
| A large enough sample (often about 1,000) | Larger samples have smaller margins of error |
| Neutral wording and order | Loaded questions push answers one way |
| A reported margin of error | It tells how far the result may be from the truth, from sampling alone |
Why it works
The most famous failure shows why. In 1936 the Literary Digest mailed millions of ballots, mostly to people on lists of car owners and telephone users, and predicted that Alf Landon would beat Franklin Roosevelt. Roosevelt won in a landslide. Size could not fix a sample that was not random. The margin of error comes from probability: for a random sample of about 1,000 people it is roughly plus or minus 3 percentage points (a handy rule: about 1 divided by the square root of the sample size). It covers only sampling chance, not bad wording or people who refuse to answer. And when comparing two candidates, the uncertainty in the gap between them is larger, about twice the stated margin.
Key idea
Polls estimate opinion from a random sample. Good polls use neutral wording and report a margin of error, about ±3 points for 1,000 people. A lead smaller than the margin of error is too close to call.
Worked example: Too close to call?
A poll of 1,000 voters (margin of error ±3) shows Candidate A at 48 percent and Candidate B at 46 percent. Who is ahead?
- The gap is 2 points.
- 2 is less than the 3-point margin, and the margin on the gap is even larger.
Answer: The poll cannot tell; the race is too close to call.
Worked example: Sample size
About what margin of error goes with a random sample of 2,500 people?
- Rule of thumb: 1 ÷ √n.
- √2,500 = 50, and 1 ÷ 50 = 0.02.
Answer: About ±2 percentage points.
Common mistake
Thinking a candidate ahead by 2 points with a 3-point margin of error is clearly winning. When the gap is within the margin of error, that poll cannot say who leads. Check: compare the gap with the margin (and remember the margin on a gap is bigger still).