Confidence Intervals
Level 5 · Testing ideas
How a range around a guess shows how sure you can be. What the margin of error in a poll means, and why a narrow range needs a bigger sample.
Course contentsLesson 10 of 10
Level 1 · Parts and chances
Level 2 · Chance over many tries
Level 3 · Describing data
Level 4 · Samples and links
Level 5 · Testing ideas
After this lesson you can
- Read a confidence interval
- Explain what 95 percent confident means
- Say why two close poll numbers may be a tie
11 mincovers 1 glossary entries
Worth reading first: Hypothesis Test and P-Value
What does a confidence interval tell you?
A confidence interval is a range around a guess that likely holds the true value. A poll that finds 52 percent support, plus or minus 3 points, gives a range of 49 to 55 percent. The plus or minus is the margin of error. A narrow range means you are fairly sure, and a wide one means you are not.
Most intervals are given at 95 percent confidence. That means the method catches the true value about 19 times in 20. If you ran the same poll 20 times with new random samples, about 19 of the ranges would contain the real figure, and about 1 would miss.
It does not mean there is a 95 percent chance the true value is in this one range. The true value is fixed. Either this range caught it or it did not. The 95 percent describes how often the method works.
Bigger samples give narrower ranges, but slowly. To cut the margin of error in half, you need about four times as many people. A poll of 1,000 gives about 3 points. A poll of 4,000 gives about 1.5.
Intervals help you read close results. If one party polls 48 percent and another 46, each with a margin of 3 points, their ranges overlap heavily. The race is too close to call, whatever the headline says.
Intervals and tests are linked. If a range for a difference between two groups does not include zero, the difference would also clear the usual bar in a test. But the range tells you more, because it shows how big the effect might be.
The common mistake is to report a single number as if it were exact. Every estimate from a sample has a range. A number with no margin of error hides how much it could be off.
Worked example
A close election poll
A poll of 1,000 voters puts Party A on 48 percent and Party B on 45, each with a margin of 3 points. A news site says A is ahead.
Ivy works out the ranges: 45 to 51 for A, 42 to 48 for B. They overlap by 3 points. She calls it too close to call.
The words for this lesson1
Check yourself
To halve a poll's margin of error, about how many more people do you need?
Finished this lesson?
Marking it complete also marks its 1 glossary entries learned, which updates your roadmap and your saved page. 0 are already marked.