Confidence level explained: what 95% means

By Jude Wallis · Published

A confidence level is the long-run success rate of the method used to build a confidence interval. At a 95% confidence level, about 95% of the intervals built from repeated random samples of the same size would capture the true parameter. You choose the level before collecting data.

AP Statistics: Unit 3 (topics 3.3 Constructing a Confidence Interval for a Population Proportion, 3.4 Justifying a Claim Based on a Confidence Interval for a Population Proportion). Confidence levels are first used in Unit 3, topics 3.3 and 3.4 of the Fall 2026 AP Statistics course, where a confidence interval for a population proportion is constructed and interpreted in context.

What a confidence level means

A confidence interval is a range of plausible values for a parameter, such as a population proportion pp or a population mean μ\mu. The confidence level describes how reliable the method behind that range is.

Picture taking a random sample, building an interval, and then repeating that process many times with new random samples of the same size. The intervals differ, because each sample differs. Some capture the true parameter and some miss. The confidence level is the percentage that capture it in the long run.

So 95% confidence is a statement about the procedure. It is not a statement about one finished interval. Once an interval is computed, the parameter is either inside it or not. You do not know which, but there is no longer any chance involved.

You can watch this happen in the confidence interval coverage interactive, which draws many intervals and counts how many capture the parameter.

Where the confidence level enters the formula

Most confidence intervals in an introductory course have the same structure:

statistic±(critical value)(standard error)\text{statistic} \pm (\text{critical value})(\text{standard error})
  • The statistic is the point estimate from the sample, such as p^\hat{p} ("p-hat") or xˉ\bar{x} ("x-bar").
  • The standard error estimates how much that statistic varies from sample to sample.
  • The critical value, written z∗z^* or t∗t^*, is set by the confidence level.
  • The product of the critical value and the standard error is the margin of error.

The confidence level only changes the critical value. For intervals based on the normal distribution:

Confidence levelCritical value z∗z^*
90%1.645
95%1.960
99%2.576

z∗z^* is the z-score that leaves the middle CC% of the standard normal distribution between −z∗-z^* and z∗z^*. For 95%, that leaves 2.5% in each tail. For a mean, t∗t^* plays the same role and also depends on the degrees of freedom.

Higher confidence means a wider interval

Raising the confidence level raises the critical value, which raises the margin of error. The interval gets wider.

That is the trade-off. To capture the parameter more often, the method has to cast a wider net, and a wider interval is a less precise estimate. In the worked example below, moving from 95% to 99% confidence increases the margin of error from about 0.049 to about 0.064, roughly 31% wider.

The usual way to keep a high confidence level and a narrow interval is a larger sample. A larger nn shrinks the standard error. See why is my confidence interval so wide and the margin of error calculator.

How to interpret a confidence level in context

There are two different sentences, and they answer different questions.

Interpreting the interval: "We are 95% confident that the interval from 0.511 to 0.609 captures the true proportion of all students at the school who prefer morning classes."

Interpreting the confidence level: "If we took many random samples of 400 students from this school and built a 95% confidence interval from each, about 95% of those intervals would capture the true proportion of all students who prefer morning classes."

The first sentence names the specific interval and the parameter. The second describes repeated sampling and never mentions the specific endpoints. If a question asks what the confidence level means, give the second.

Common mistakes

  1. "There is a 95% probability the parameter is in this interval." After the interval is computed, nothing is random. The 95% describes the method across many samples.
  2. "95% of the data fall in the interval." A confidence interval estimates a parameter. It does not describe where individual values lie.
  3. "95% of sample proportions will fall in this interval." Other samples produce their own intervals centered on their own estimates.
  4. Describing the sample. The interval is about the population parameter. The sample proportion is known exactly and needs no interval.
  5. Thinking higher confidence is always better. Higher confidence costs precision. A 99.9% interval may be too wide to be useful.
  6. Mixing up confidence level and confidence interval. The level is a percentage. The interval is a pair of endpoints. See confidence level vs confidence interval.
  7. Expecting the level to fix a bad sample. The stated success rate holds only when the conditions for the procedure are met, starting with random sampling.

Choosing a confidence level

The level is chosen by the person doing the study, before looking at the data. 95% is the most common convention. 90% is used when a narrower interval matters more than the extra assurance. 99% is used when missing the parameter would be costly.

The confidence level is tied to the significance level of a two-sided test. A confidence level CC, written as a decimal, pairs with a two-sided test at α=1−C\alpha = 1 - C: a 95% interval goes with α=0.05\alpha = 0.05, and a 99% interval goes with α=0.01\alpha = 0.01. The pairing is exact for tt procedures about means. For proportions the test builds its standard error from p0p_0 and the interval builds it from p^\hat{p}, so the two can disagree in borderline cases. See significance level vs confidence level and the p-value explainer.

Confidence level in AP Statistics

Confidence intervals first appear in Unit 3 of the Fall 2026 AP Statistics course, in topic 3.3, Constructing a Confidence Interval for a Population Proportion, and topic 3.4, which uses an interval to justify a claim. The same ideas carry into two-proportion intervals and into Unit 4 for means. Free-response question 3 on the exam is an inference question, either a hypothesis test or a confidence interval.

Be ready to find the critical value for a given level, interpret an interval and a confidence level as two separate things, and explain how changing the level or the sample size changes the width. Build intervals with the confidence interval calculator, then try the interpreting confidence intervals practice set.

One sample, three confidence levels

A random sample of 400 students at a large high school finds that 224 prefer morning classes. Construct 90%, 95%, and 99% confidence intervals for the proportion of all students at the school who prefer morning classes, and compare them. Assume the school has more than 4,000 students.

  1. Point estimate: p^=224400=0.56\hat{p} = \frac{224}{400} = 0.56.

  2. Check conditions: the sample is random, 400 is less than 10% of the population, and the counts of successes and failures, 224 and 176, are both at least 10.

  3. Standard error: p^(1−p^)n=(0.56)(0.44)400≈0.0248\sqrt{\frac{\hat{p}(1 - \hat{p})}{n}} = \sqrt{\frac{(0.56)(0.44)}{400}} \approx 0.0248.

  4. 90% level, z∗=1.645z^* = 1.645: margin of error ≈1.645(0.0248)≈0.041\approx 1.645(0.0248) \approx 0.041. Interval: 0.56±0.0410.56 \pm 0.041, or about (0.519,0.601)(0.519, 0.601).

  5. 95% level, z∗=1.960z^* = 1.960: margin of error ≈0.049\approx 0.049. Interval: about (0.511,0.609)(0.511, 0.609).

  6. 99% level, z∗=2.576z^* = 2.576: margin of error ≈0.064\approx 0.064. Interval: about (0.496,0.624)(0.496, 0.624).

  7. Compare: the estimate 0.56 and the standard error are the same in all three. Only the critical value changed, and the interval widened each time the level went up. The 99% interval is the only one of the three that contains 0.50.

90%: about (0.519,0.601)(0.519, 0.601). 95%: about (0.511,0.609)(0.511, 0.609). 99%: about (0.496,0.624)(0.496, 0.624). We are 95% confident that the interval from 0.511 to 0.609 captures the true proportion of all students at the school who prefer morning classes.

Frequently asked questions

What does a 95% confidence level mean?

It means the method used to build the interval captures the true parameter in about 95% of repeated random samples of the same size. It does not mean there is a 95% probability that one particular computed interval contains the parameter.

Is a 99% confidence level better than 95%?

Not automatically. A 99% level captures the parameter more often, but the interval is wider and so less precise. In the worked example the margin of error grows from about 0.049 at 95% to about 0.064 at 99%. The right level depends on how costly a miss is compared with how much precision you need.

What is the difference between confidence level and significance level?

The confidence level CC is the long-run capture rate of a confidence interval method. The significance level α\alpha is the cutoff for a p-value in a hypothesis test and the probability of a Type I error when the null hypothesis is true. For a two-sided test they are linked by α=1−C\alpha = 1 - C.

What are the z critical values for 90%, 95%, and 99% confidence?

They are 1.645, 1.960, and 2.576. Each is the z-score that leaves the stated percentage in the middle of the standard normal distribution. For intervals about a mean, use a t∗t^* critical value with the right degrees of freedom instead.