One Proportion Z Interval Calculator: Formula, Conditions, Examples & Software Steps

Number of Successes

How many in your sample had the outcome you're testing for.

Sample Size

Total number of people or items in your sample.

Confidence Level (%)

Common choices are 90, 95, or 99.

About This Test

What Is a One-Proportion Confidence Interval?

A confidence interval for a single proportion gives you a range of plausible values for the true population proportion, based on what you observed in your sample — rather than a single point estimate you have to take on faith.

When to Use It

Estimating the true conversion rate of a website from a sample of visitors
Estimating the true defect rate of a production line from a sample batch
Estimating the true approval rating in a population from a survey sample
Any situation with one sample and a yes/no outcome

The Formula

\[ \hat{p} \pm z_{\alpha/2}\sqrt{\frac{\hat{p}(1-\hat{p})}{n}} \]

Worked Example

Out of 150 customers surveyed, 45 said they'd recommend the product. What's the 95% confidence interval for the true recommendation rate?

\( \hat{p} = 45/150 = 0.30 \)
Critical value for 95%: \( z_{\alpha/2} = 1.96 \)
Margin of error: \( 1.96 \times \sqrt{0.30 \times 0.70 / 150} \approx 0.0733 \)
95% confidence interval: (0.2267, 0.3733)
We're 95% confident the true recommendation rate is between about 22.7% and 37.3%.

Your Results

Quick answer:

A one proportion z interval estimates a plausible range for the true population proportion from your sample data.
Formula: p̂ ± (critical value from the z-distribution) × √(p̂(1−p̂)/n)
Check three conditions first: random sample, independence (the 10% condition), and both np₀ and n(1−p₀) at least 10.
Use the calculator above for an instant result, or keep reading for the full formula, a worked example, software code, and common mistakes to avoid.

I have been checking dissertation stats chapters and thesis appendices for twelve years now, and the one proportion z interval is one of those topics students get half right almost every time. They either mix it up with the hypothesis test, or they skip the conditions check completely and jump straight to plugging numbers into a formula.

So let me clear up the one proportion z test definition properly, the way I explain it to my own clients.

A one proportion z interval is a range of values, calculated from your sample, that you can be reasonably confident contains the true population proportion. It is built using the sample proportion, a critical z value based on your confidence level, and the standard error of that proportion. The one proportion z interval equation is simply p̂ plus or minus that margin of error.

Use the calculator above to get your answer in seconds. Below, I will walk you through everything the calculator does not show you on the surface, including the hypothesis test version of this same procedure, because most people searching for this actually need both.

One Proportion Z Interval vs One Proportion Z Test: What’s the Difference?

This confuses almost everyone I work with, so let me be direct about it.

A one proportion z interval gives you a range. It answers “what is the true proportion likely to be, between what two numbers.” A one proportion z test gives you a decision. It answers “is the true proportion different from some specific claimed value, yes or no.”

Both use the same sample data. Both rely on the same normal approximation. But they answer completely different research questions, and I have seen dissertation students use the wrong one because their supervisor asked for “a significance test” and they submitted a confidence interval instead.

	One Proportion Z Interval	One Proportion Z Test
Question it answers	What range is the true proportion likely in?	Is the true proportion equal to a claimed value?
Output	Lower bound, upper bound	Test statistic (z), p-value
Needs a hypothesised value?	No	Yes, p₀
Typical use case	Estimating an unknown rate	Testing a claim against a benchmark

If your research question sounds like “what percentage of customers actually prefer X,” you want the interval. If it sounds like “the company claims 65% prefer X, is that true,” you want the test. Many dissertation methodology chapters actually need both, one to test the claim and one to report the precision around it.

Related tool: if you’re comparing two groups instead of testing against a fixed value, use the confidence interval two proportion z-test calculator instead.

Why Proportions Use a Z-Interval, Not a T-Interval

If you have already studied confidence intervals for a mean, you might expect a t-distribution to show up here too. Proportions work differently.

A t-interval exists to correct for extra uncertainty when you are estimating the population standard deviation from a small sample, which is exactly the situation with sample means. A proportion does not have that problem, since its standard error, √(p̂(1−p̂)/n), is calculated entirely from p̂ and n, with no separate standard deviation to estimate.

That is why a one proportion z interval uses the z-distribution regardless of sample size, as long as the np₀ and n(1−p₀) conditions below are satisfied. There is no standard “t-test for one proportion,” even though I get asked about this constantly.

Calculator Notation: 1-PropZTest and 1-PropZInt

If you have used a TI-84 or TI-83 for an AP Statistics or intro stats class, you have probably seen this same procedure under a shorter name. Press STAT, then TESTS, and “1-PropZTest” and “1-PropZInt” sit right next to each other, “1” simply being calculator shorthand for “one.”

1-PropZTest runs the one proportion z test formula covered below and asks for x (successes), n (sample size), and p₀ (your hypothesised value). 1-PropZInt runs the interval formula from this page and asks for x, n, and your confidence level, no p₀ needed, matching the “no hypothesised value” row in the table above.

So if your assignment says “run a 1 prop z test” or “find the 1 proportion z interval,” your professor means the exact same procedure this article walks through. The conditions and assumptions, the hypothesis setup, and the equation underneath stay identical, no matter which method you use: by hand, in Excel, or straight off your calculator screen.

What Are the Conditions for a One Proportion Z Test?

I will not sugarcoat this: these one proportion z test assumptions, or a 1 prop z test assumptions check as some professors call it, are not optional. If you skip this step, your interval or your test result is not trustworthy, no matter how correctly you punch numbers into a calculator.

This whole procedure rests on the central limit theorem. Your raw data follows a binomial distribution, just success or failure, yes or no, but as n grows, the sampling distribution of p̂ starts behaving like a normal distribution. That is the only reason a z critical value is valid here at all.

Three things need to be true before you run a one proportion z interval or z test:

Random sample. Your data has to come from a random sample or a properly randomised experiment. Convenience sampling technically breaks this assumption, though in practice a lot of published research quietly ignores it.
Independence, or the 10% condition. If you are sampling without replacement from a finite population, your sample size should be no more than 10% of that population.
Large enough sample for the normal approximation. Both np₀ and n(1−p₀) need to be at least 10 (some textbooks say 5, but 10 is the safer standard most journals expect). Penn State’s STAT 200 course notes walk through this exact threshold check using real survey data, if you want to see it applied outside a dissertation context.

That third condition is the one I see violated most often in dissertation drafts. Students collect 40 responses, get 3 “yes” answers, and try to build a z interval anyway. The maths will still spit out a number, but that number is not reliable at that sample size.

The Plus-Four Method: What to Do With Small Samples

Here is something almost none of the beginner guides mention, and it genuinely surprises me every time I bring it up with a student.

When your sample is too small for the standard normal approximation, you do not have to abandon the z interval entirely. This one proportion plus four z interval procedure, also called the plus-four method or Agresti-Coull interval, adjusts your numbers before you calculate: add 2 successes and 2 failures to your data (4 observations total, hence “plus four”), then calculate the proportion and interval as usual using this adjusted count.

It is a small trick, but it noticeably improves accuracy for small or moderate samples, and it is something I recommend to any client whose np₀ falls under 10. It sits in the same family as the Wilson score interval, another small-sample correction you will see in more advanced statistical software, though plus-four is simpler to calculate by hand and is what most intro-level courses expect. Most calculator tools online will not walk you through this, which is exactly why I am including it here.

How Sample Size Affects Your Margin of Error

Bigger samples always produce narrower, more precise intervals, because n sits inside a square root in the denominator of your standard error. Quadruple your sample size and your margin of error roughly halves. Double it, and the margin of error shrinks by about 30%, not 50%, which surprises a lot of students expecting a straight-line relationship.

This matters when you are planning data collection, not just analysing it afterwards. If a pilot survey gives you a margin of error too wide for your research question, the fix is almost always a bigger sample, not a different formula. The sample size and power calculator works out how many responses you actually need before you go back into the field.

The One Proportion Z-Test Formula (Hypothesis Testing Side)

If your research question involves testing a specific claimed value rather than just estimating a range, here is the full setup.

Setting Up the Null and Alternative Hypothesis

Every one proportion z test starts with a claimed value, p₀, and three possible ways to frame your hypotheses:

Two-tailed: H₀: p = p₀ vs H₁: p ≠ p₀ (you just want to know if it’s different, in either direction)
Left-tailed: H₀: p ≥ p₀ vs H₁: p < p₀ (you’re testing if it’s lower)
Right-tailed: H₀: p ≤ p₀ vs H₁: p > p₀ (you’re testing if it’s higher)

Most dissertation research questions I encounter use the two-tailed version by default, unless the hypothesis specifically predicts a direction.

What Is the One Proportion Z Test Statistic Formula?

The one proportion z test statistic formula, sometimes written as the one prop z test equation, is:

[ z = \frac{\hat{p} – p_0}{\sqrt{\dfrac{p_0(1-p_0)}{n}}} ]

Notice the denominator uses p₀, the claimed value, not p̂, your observed sample proportion. This is the single most common calculation error I see in submitted work. Students plug in their own sample proportion into the standard error formula instead of the hypothesised one, and it throws the whole test statistic off.

How Do You Find the P-Value for a One Proportion Z Test?

Once you have your z statistic, you compare it against the standard normal distribution to get a p-value.

Two-tailed test: p-value = 2 × P(Z > |z|)
Left-tailed test: p-value = P(Z < z)
Right-tailed test: p-value = P(Z > z)

If that p-value comes in under your chosen significance level, usually 0.05, you reject the null hypothesis. If it’s above, you fail to reject it. I always tell my students, “fail to reject” is not the same as “prove the null is true,” and examiners do check for that exact wording.

If you need this same conversion for a different test entirely, chi-square or t or F, the p-value calculator for all distributions handles all of them in one place instead of you rebuilding this lookup by hand each time.

Worked Example: Hypothesis Test and Confidence Interval Together

Let me walk you through a full example, start to finish, using both procedures.

A university career services office claims that 40% of graduating students accept a job offer within one month of graduation. A researcher surveys 200 recent graduates and finds that 92 of them accepted an offer within a month. Test the claim at α = 0.05, and also build a 95% confidence interval for the true proportion.

Step 1: Check conditions. np₀ = 200 × 0.40 = 80, and n(1−p₀) = 200 × 0.60 = 120. Both well above 10, so the normal approximation holds.

Step 2: State the hypotheses. H₀: p = 0.40, H₁: p ≠ 0.40 (two-tailed, since the researcher just wants to know if the claim is accurate, not specifically higher or lower)

Step 3: Calculate the sample proportion. p̂ = 92/200 = 0.46

Step 4: Calculate the test statistic. z = (0.46 − 0.40) / √(0.40 × 0.60 / 200) = 0.06 / 0.0346 ≈ 1.73

Step 5: Find the p-value. For a two-tailed test with z = 1.73, p-value ≈ 0.0836

Step 6: Make a decision. Since 0.0836 > 0.05, we fail to reject H₀. There is not enough evidence at the 5% level to say the true acceptance rate differs from 40%.

Step 7: Now build the confidence interval separately. This time we use p̂, not p₀, in the standard error: SE = √(0.46 × 0.54 / 200) ≈ 0.0353 Margin of error = 1.96 × 0.0353 ≈ 0.0692 95% CI = (0.46 − 0.0692, 0.46 + 0.0692) = (0.3908, 0.5292)

In plain English: we are 95% confident that the true proportion of graduates accepting a job offer within one month falls somewhere between about 39% and 53%. That is the sentence you actually write in a results chapter, not just the bare numbers.

So even though the hypothesis test failed to reject the 40% claim, the confidence interval shows the true rate could plausibly be anywhere from about 39% to 53%. This is exactly why I tell clients to report both numbers together. A test alone hides how wide the uncertainty actually is.

A short case from my own practice: a marketing dissertation student I worked with last year had almost this exact scenario, testing a brand’s claimed customer satisfaction rate against her own survey data. Her test came back non-significant, and she wanted to write “the claim is confirmed true.” I had to explain that failing to reject a null hypothesis is not confirmation, it just means her sample didn’t have enough evidence to disprove it. That one sentence saved her an awkward comment from her external examiner.

How to Calculate a One Proportion Z Interval in Excel, R, Python, and SPSS

If you need to reproduce this in your actual analysis software, here is the exact command or formula for each.

Excel

Excel has no single built-in function for this, so you build it manually:

=phat - NORM.S.INV(1-(1-confidence)/2) * SQRT(phat*(1-phat)/n)
=phat + NORM.S.INV(1-(1-confidence)/2) * SQRT(phat*(1-phat)/n)

Replace phat, confidence, and n with your cell references for the lower and upper bound respectively.

R

R’s base function for this is prop.test():

prop.test(x = 92, n = 200, p = 0.40, correct = FALSE)

Set correct = FALSE if you want the uncorrected z-test result rather than R’s default Yates continuity correction, which some professors specifically ask students to disable for this comparison. CRAN’s own walkthrough of the one-sample z-test for a proportion shows the correction’s effect on the same formula if you want to see it side by side.

Python

Running a one proportion z test in Python is cleanest with statsmodels:

from statsmodels.stats.proportion import proportions_ztest, proportion_confint

stat, pval = proportions_ztest(count=92, nobs=200, value=0.40)
ci_low, ci_upp = proportion_confint(count=92, nobs=200, alpha=0.05, method='normal')

The full parameter list, including alternative methods like Wilson and Agresti-Coull, is in the official statsmodels documentation.

SPSS

SPSS does not have a dedicated one-proportion z-test menu option in most versions. The common workaround is to use the Binomial Test under Analyze > Nonparametric Tests > One Sample, which uses the exact binomial method rather than the normal approximation. For the confidence interval specifically, you will usually need to compute it manually using the Compute Variable function with the same formula shown above.

I get asked about this SPSS gap constantly, and honestly, it is one of the reasons I built out these calculators in the first place. Software menus do not always match what a stats course actually teaches.

Practice Problems (Test Yourself)

These one proportion z test practice problems cover both the interval and the hypothesis test, so try them before checking the answers.

A survey of 250 voters finds 140 in favour of a proposal. Build a 90% confidence interval for the true proportion in favour.
A factory claims its defect rate is 5%. A quality check on 300 units finds 21 defective. Test this claim at α = 0.05.
Out of 80 patients given a new treatment, 52 showed improvement. Is this sample large enough to use the normal approximation, and if not, what should you do instead?

Answer key:

p̂ = 0.56, z* = 1.645, margin of error ≈ 0.0516, so 90% CI ≈ (0.5084, 0.6116)
np₀ = 15, n(1−p₀) = 285, both above 10 so conditions hold. p̂ = 0.07, z ≈ 1.59, two-tailed p-value ≈ 0.112, fail to reject the 5% claim at α = 0.05
np₀ depends on what you set p₀ to, but with only 80 patients and no stated benchmark, always check np₀ and n(1−p₀) before assuming the approximation holds. If either falls under 10, use the plus-four adjusted interval instead of the standard formula.

Common Mistakes to Avoid

After correcting hundreds of these calculations in student drafts, a few errors show up again and again.

Using p̂ in the test statistic denominator instead of p₀. This is the single biggest error I correct. The hypothesis test standard error always uses the claimed value, never your observed sample proportion.
Skipping the condition check entirely. I still see submitted chapters with no mention of np₀ or n(1−p₀), which examiners increasingly flag.
Reporting only the p-value, without the interval. A p-value tells you significance, not precision. Report both.
Confusing “fail to reject” with “proven true.” These are not the same statement, and careless wording here is an easy mark deduction.
Forgetting the 10% condition when sampling without replacement from a small, finite population.

If your dissertation methodology chapter needs this level of precision checked properly before submission, this is exactly the kind of review I do for clients. Need expert help? Get professional statistical consulting.

For related calculations you’ll likely need in the same chapter, check the one proportion z test calculator for hypothesis-test-only output, the critical z score calculator if you need z* values for other confidence levels, the margin of error calculator to isolate just that piece of the formula, and the sample size calculator if you are planning your data collection before running the test at all.

Frequently Asked Questions

Can I run a one proportion z test in Excel? Yes, though Excel has no built-in function for it specifically. You calculate the test statistic manually using NORM.S.DIST() for the p-value, working from the standard formula with your sample proportion, hypothesised value, and sample size.

What sample size do I need for a one proportion z test? There is no fixed number. What matters is that np₀ and n(1−p₀) both come out to at least 10. Depending on your hypothesised proportion, that could mean a sample of 50 or a sample of 500.

Is a one proportion z test the same as prop.test() in R? Mostly yes, with one difference. R’s prop.test() applies a Yates continuity correction by default, which makes the result slightly more conservative than the plain z-test formula taught in most statistics courses. Add correct = FALSE in R to match the textbook version exactly.

What is the plus-four (Agresti-Coull) interval and when should I use it? It is an adjustment for small samples where the standard normal approximation is not reliable. You add 2 successes and 2 failures to your observed counts before calculating the proportion and interval. Use it whenever np₀ or n(1−p₀) falls below 10.

Does SPSS give a p-value or a confidence interval for this test? SPSS’s Binomial Test option gives you a p-value using the exact binomial method, not the normal approximation. It does not automatically output a confidence interval for a one-sample proportion, so that part typically needs to be calculated manually.

What’s the difference between a one proportion z test and a one proportion z interval? The test gives you a yes-or-no decision about whether the true proportion equals some claimed value. The interval gives you a range of plausible values for the true proportion, without needing a claimed value to test against. They use the same underlying data but answer different questions.

What is a 1-PropZTest on a calculator? It is the exact same one proportion z test covered in this article, just under its TI-84/TI-83 menu name. “1-PropZInt” is the confidence interval version of the same procedure, sitting right next to it in the STAT > TESTS menu.

How does sample size affect a one proportion z interval? A larger sample narrows your interval, because n sits inside the square root in the standard error formula. Quadrupling your sample size roughly halves your margin of error, so if your interval comes out too wide to be useful, collecting more data is usually the fix, not a different formula.

Siddharth Gupta has spent 12 years helping graduate students and researchers across the USA, UK, and UAE with dissertation statistics, from proposal to final defence. Connect on LinkedIn.