Confidence level vs confidence interval

By Jude Wallis · Published July 28, 2026

The confidence level is a percentage, like 95%, saying how often the method captures the parameter in repeated sampling. The confidence interval is the range of plausible values, like 0.504 to 0.696, from one sample. The level is the method's success rate; the interval is this sample's result.

AP Statistics: Unit 3 (topics 3.3, 3.4). Constructing a confidence interval for a proportion is topic 3.3 and interpreting the interval and its confidence level is topic 3.4, both in Unit 3 of the Fall 2026 AP Statistics course. The same level-versus-interval distinction carries into the mean intervals of Unit 4.

Confidence level vs confidence interval: the short answer

These two travel together, which is why they blur, but they are different objects. The confidence level is a percentage you choose before looking at the data, such as 90%, 95%, or 99%. The confidence interval is the range of numbers you get out afterward, such as from 0.504 to 0.696.

Think of the level as a property of the method and the interval as the result of running it once. The level describes how the procedure behaves over many samples: at a 95% level, about 95% of the intervals built this way would capture the true parameter. The interval is the single answer your one sample produced. One is a long-run success rate; the other is a concrete range of plausible values.

What the confidence level is

The confidence level, written $C\%$ , is set before you collect data and controls how sure the method is. Its meaning is a long-run capture rate: in repeated random sampling with the same sample size, approximately $C\%$ of the intervals you build will contain the true population parameter.

The level enters the arithmetic through the critical value. For a 95% level the critical value from the standard normal distribution is $z^* = 1.96$ , the number of standard errors that brackets the middle 95% of the curve; for 90% it is $1.645$ and for 99% it is $2.576$ . A higher level uses a larger critical value, which is why raising your confidence makes the interval wider. The what 95% confidence means guide unpacks this long-run idea in full.

What the confidence interval is

The confidence interval is the range of plausible values your sample produces for the parameter. It is built from the formula on the AP sheet,

\text{statistic} \pm (\text{critical value})(\text{standard error})

where the statistic is your point estimate, such as a sample proportion $\hat{p}$ (read "p-hat"), and the standard error measures how much that estimate varies from sample to sample. The critical value comes from the confidence level, and the two pieces after the $\pm$ sign together form the margin of error.

Once computed, the interval is a fixed pair of numbers, like from 0.504 to 0.696. That single interval either contains the true parameter or it does not; you just cannot tell which. What you can say is that you are $C\%$ confident the interval contains the parameter, meaning it came from a method that captures the truth $C\%$ of the time. You can compute one with the confidence interval calculator.

The differences side by side

Feature	Confidence level	Confidence interval
What it is	A percentage, like 95%	A range of values, like 0.504 to 0.696
Chosen or computed	Chosen before collecting data	Computed from the sample
Describes	The method's long-run capture rate	This sample's plausible values
Changes with the sample	No, it is fixed by you	Yes, a new sample gives a new interval
Role in the formula	Sets the critical value	Is the final result
Effect of raising it	Not applicable	Makes the interval wider

The level is an input you set; the interval is the output you read. Raising the level widens the interval, because a larger critical value stretches the margin of error.

How the level and the interval fit together

You choose the confidence level to set how often the method should succeed, and that choice flows into the width of the interval it returns. A 95% level is the usual default: high enough to be trustworthy, tight enough to be useful. Push the level to 99% and the interval widens, because you are demanding the method capture the parameter more often, which forces a bigger margin of error.

Sample size is the other lever. A larger sample shrinks the standard error, so at a fixed level the interval gets narrower, which is how you buy precision without giving up confidence. So the level fixes the capture rate, the interval reports the plausible values, and sample size trades off against width. To turn the interval into a claim, check whether a value of interest lies inside it, the idea behind how to calculate a confidence interval.

The classic mix-up and how to avoid it

The signature error is attaching the level to the single interval you computed, saying "there is a 95% probability the parameter is between 0.504 and 0.696." Once the interval is a fixed pair of numbers, the parameter is either inside it or not; there is no probability left to assign. The 95% describes the method across many samples, not the one interval in front of you.

Two fixes keep it straight. First, phrase the interval correctly: "we are 95% confident the interval from 0.504 to 0.696 contains the true proportion," which points the confidence at the procedure, not at a random parameter. Second, keep the words attached to the right object: level is the percentage and always a property of the method, while interval is the range and always the result from your sample. Mixing the two is the most common wording mistake on inference free-response questions.

Same sample, two confidence levels

In a random sample of 100 voters, 60 support a measure, so the sample proportion is 0.60. Build the 95% confidence interval, then the 99% interval, and identify which piece is the level and which is the interval.

Find the standard error of the sample proportion: $SE = \sqrt{\frac{\hat{p}(1-\hat{p})}{n}} = \sqrt{\frac{0.60 \times 0.40}{100}}$ .
Compute inside the root: $0.60 \times 0.40 = 0.24$ , and $\frac{0.24}{100} = 0.0024$ , so $SE = \sqrt{0.0024} \approx 0.04899$ .
For the 95% level, the critical value is $z^* = 1.96$ . Margin of error $= 1.96 \times 0.04899 \approx 0.0960$ .
Build the 95% interval: $0.60 \pm 0.0960$ , which is $0.60 - 0.0960 = 0.504$ to $0.60 + 0.0960 = 0.696$ .
For the 99% level, the critical value is $z^* = 2.576$ . Margin of error $= 2.576 \times 0.04899 \approx 0.1262$ .
Build the 99% interval: $0.60 \pm 0.1262$ , which is $0.60 - 0.1262 = 0.474$ to $0.60 + 0.1262 = 0.726$ .
Name the parts: 95% and 99% are confidence levels, chosen in advance; the ranges 0.504 to 0.696 and 0.474 to 0.726 are the confidence intervals, computed from the same sample.

At 95% confidence the interval is 0.504 to 0.696; at 99% confidence it widens to 0.474 to 0.726. The level (95% or 99%) is the method's long-run capture rate you chose beforehand, and the interval is the range your sample produced. Raising the level from 95% to 99% raised the critical value from 1.96 to 2.576, which widened the interval.

Frequently asked questions

Is there a 95% probability the parameter is in my confidence interval?

No. Once the interval is computed it is a fixed range, and the parameter is either in it or not, so no probability applies to that one interval. The 95% is the confidence level, the long-run rate at which intervals built this way capture the parameter.

How does the confidence level change the interval's width?

A higher level uses a larger critical value, which enlarges the margin of error and widens the interval. Going from 95% to 99% raises the critical value from 1.96 to 2.576, so the 99% interval is wider than the 95% interval built from the same sample.

Can I make an interval narrower without lowering the confidence level?

Yes. A larger sample size lowers the standard error, so at a fixed confidence level the interval gets narrower. That is how you gain precision without giving up any confidence, at the cost of collecting more data.