A confidence interval gives you a range of plausible values for something you can’t measure directly about a whole population — the true average, say, or the true percentage — estimated from a single sample. When you see a “95% confidence interval,” here’s the part to hold onto: the 95% describes the method, not the one interval in front of you. If you took sample after sample and built an interval each time, about 95% of those intervals would contain the true value. The one you’re holding either contains it or it doesn’t — you just don’t get to know which.
That distinction sounds like splitting hairs the first time you hear it. It isn’t, and it’s the single most misunderstood idea in an intro statistics course. So let me walk through it properly — what a confidence interval is, where it comes from, and what it genuinely tells you versus what students think it tells them.
What a confidence interval actually is
You almost never get to measure a whole population. You can’t ask every voter, weigh every box, or test every patient. So you take a sample, calculate something from it — a mean or a proportion — and use that to make an educated estimate about the whole group.
But a single sample number, all by itself, is a little dishonest. If you say “the average is 71,” you’re implying a precision you don’t actually have, because a different sample would have given you a slightly different number. A confidence interval is the honest version of that statement. Instead of one number, it gives you a range, and it attaches a confidence level to that range. “I’m 95% confident the true average is between 68 and 74” is a far more truthful thing to say than “the average is 71.”
Where the interval comes from
Almost every confidence interval you’ll meet in an intro course has the same shape:
point estimate ± margin of error
The point estimate is your best single guess from the sample — the sample mean or sample proportion. The margin of error is the cushion you build around it to account for the fact that your sample isn’t the whole population. Add the margin to the estimate and you have the top of the interval; subtract it and you have the bottom.
So when a news poll says “52% support, margin of error 3 points,” the confidence interval is 49% to 55%. The 52% is the point estimate; the 3 points is the margin of error.
And the width is a real trade-off. Say I want the average shoe size of a group and I announce a margin of error of 5 — “it’s somewhere between a 4½ and a 14½.” I’m 100% confidentin that, and it’s completely useless. Tighten the margin to 0.2 and now I’ve pinned it to about one shoe size — far more useful, but I’m far less sure I’ve actually captured the truth. Narrower tells you more but promises less; wider promises more but tells you less. The confidence level is just where you decide to sit on that seesaw.
What “95% confidence” really means
Here’s the careful version, and it’s worth reading twice.
The confidence level is a statement about the procedure, made over the long run. A 95% confidence level means: if you repeated this whole process many times — new sample, new interval, again and again — about 95% of the intervals you’d build would capture the true population value.
It does notmean there’s a 95% probability that the true value lies inside the particularinterval you calculated. Once you’ve built a specific interval, the true value is already either in it or out of it. There’s no probability left to talk about for that one interval — the randomness was all in the sampling, and the sampling is done.
Picture it concretely: take a hundred different samples and build a 95% interval from each. About 95 of them land across the true value; a handful miss it entirely. You’re holding one of the hundred, and you have no way to know whether yours is one of the 95 or one of the few. The one comfort: even when an interval misses, it usually doesn’t miss by much.
I know that feels like a technicality. But it’s the difference between understanding what the tool does and just operating it, and exam questions love to test exactly this point.
What makes an interval wider or narrower
Three things move the width, and seeing how they trade off tells you almost everything:
- Sample size. Bigger samples give narrower intervals. More information, more precision.
- Confidence level. Wanting to be more confident — 99% instead of 95% — makes the interval wider, not narrower. To be more sure you’ve captured the true value, you have to cast a wider net.
- Variability in the data. More spread in the underlying data means a wider interval.
That second one surprises people, so sit with it: there’s a tension between confidence and precision. A 99.9% interval might be so wide it’s useless (“somewhere between 40% and 90%”), while a 50% interval is tight but you can barely trust it. The usual 95% is a compromise everyone has agreed is reasonable.
Let me show you how I’d read one
Say a study reports that a new medication lowers blood pressure by an average of 8 points, with a 95% confidence interval of 5 to 11 points.
Here’s what I’d take from that. My best single estimate of the true effect is 8 points. The plausible range — given the honesty the interval forces on me — runs from 5 to 11. And critically, because the whole interval sits above zero, I have good reason to believe the medication does something: even the low end of the range is a real drop. If that interval had been −2 to 18, I couldn’t say that, because zero (no effect) would be a plausible value. That’s often the real question a confidence interval is built to answer: is some meaningful value — usually zero — inside the range or outside it?
The mistake I see most often
Two slips, and they’re cousins.
The first is the probability one I keep coming back to: saying “there’s a 95% chance the true value is in this interval.” Tempting, natural-sounding, and not what the interval claims.
The second is confusing the confidence interval with the spread of the data itself. A confidence interval is a statement about a population parameter— the average, the proportion — not about where individual data points fall. It does not mean “95% of people scored between 68 and 74.” It means “I’m 95% confident the true averageis between 68 and 74.” Different claim entirely, and mixing them up is one of those small gaps that quietly makes everything afterward feel impossible.
Want to work through it together?
If confidence intervals still feel slippery, that’s normal — and it’s usually one small idea from a few weeks back doing the damage, not the whole topic. Finding that idea is the first thing I do with every student, and I’ll never make you feel foolish for having a gap. That’s just how this subject is learned.
