This two proportion z test calculator tests whether two independent population proportions are equal. Enter successes and sample size for each group, choose the direction of the alternative hypothesis, and set α. The result shows both sample proportions, p̂₁ − p̂₂, the pooled proportion, pooled standard error, z statistic, p-value, decision, conditions, and calculation steps.

Use this test when the outcome is binary and the two groups are independent. For paired responses, repeated measurements, matched subjects, or more than two groups, this pooled two-sample procedure is not the correct analysis.

02Input

How to Use the Two Proportion Z Test Calculator

Define the same success outcome for both groups. Enter successes x₁ and total observations n₁ for group 1.
Enter successes x₂ and total observations n₂ for group 2. Each success count must be a whole number from zero through its sample size.
Choose the alternative. The calculator always reports the observed difference as p̂₁ − p̂₂, so group order controls the meaning of greater and less.
Enter the preselected significance level α as a decimal between zero and one.
Compare the p-value with alpha, then review the effect difference, sampling design, and large-count condition before drawing a conclusion.

A two-sided selection tests for any difference. “Greater” asks whether the population proportion for group 1 exceeds group 2; “less” asks whether group 1 is lower. Reversing the group order reverses the sign of z and the observed difference, but does not change a two-sided p-value.

03Calculation

Two Proportion Z Test Calculator Formula

Calculate each sample proportion and their signed difference:

p̂₁ = x₁ / n₁

p̂₂ = x₂ / n₂

difference = p̂₁ − p̂₂

The null hypothesis says the population proportions are equal. Under that model, both samples estimate one common proportion, so the test pools their successes:

p̂ = (x₁ + x₂) / (n₁ + n₂)

For a compact standard-error expression, let K = 1/n₁ + 1/n₂. Then:

K = 1/n₁ + 1/n₂

SE₀ = √[p̂(1 − p̂)K]

z = (p̂₁ − p̂₂) / SE₀

The selected alternative determines the normal-tail area used for the p-value. The calculator applies no continuity correction.

Null modelp₁ − p₂ = 0

Pooling is appropriate here because the equality null treats both population proportions as one value. A confidence interval for the difference answers a different estimation question and normally uses an unpooled standard error.

04Hypotheses

Two Proportion Test Hypotheses and Tail Direction

The null hypothesis is H₀: p₁ = p₂, equivalently p₁ − p₂ = 0. Choose one alternative before inspecting the sample results.

ATwo-sidedp₁ ≠ p₂

Are the Two Population Proportions Different?

Use a two-sided test when either direction would be meaningful. Both tails of the standard normal distribution count toward the p-value.

BGroup 1 higherp₁ > p₂

Is the Group 1 Proportion Greater?

Use the greater alternative only when a higher group 1 proportion was the predefined claim. Positive z values support this direction.

CGroup 1 lowerp₁ < p₂

Is the Group 1 Proportion Less?

Use the less alternative for a predefined decrease in group 1 relative to group 2. Negative z values support this direction.

05Worked example

Two Proportion Z Test Example with Pooled Estimate

Suppose two independent, preassigned user groups see different onboarding flows. In group 1, 72 of 120 users complete onboarding. In group 2, 51 of 120 complete it. Test for any population difference at α = 0.05.

The hypotheses are H₀: p₁ = p₂ and Hₐ: p₁ ≠ p₂.

p̂₁ = 72 / 120 = 0.600 and p̂₂ = 51 / 120 = 0.425.
The observed difference is 0.600 − 0.425 = 0.175, or 17.5 percentage points in the group 1 minus group 2 direction.
The pooled estimate is p̂ = (72 + 51) / (120 + 120) = 0.5125.
SE₀ = √[0.5125(1 − 0.5125)(1/120 + 1/120)] ≈ 0.064530.
z = 0.175 / 0.064530 ≈ 2.711936.
The two-sided p-value is approximately 0.006689.

Because 0.006689 ≤ 0.05, the samples provide enough evidence that the two population completion proportions differ. That conclusion assumes the groups and observations are independent and representative of the populations named in the question. The 17.5-point difference should be reported with the p-value; it communicates effect direction and size in the sample.

The pooled test evaluates whether the 17.5-point sample difference is compatible with equal population proportions.

06Validity

Conditions for a Two Sample Proportion Z Test

Check the design and counts before relying on the normal approximation:

Same binary outcome: success and failure must have the same operational definition in both groups.
Independent groups: no person or unit may appear in both groups. Matched pairs and before/after responses require a paired categorical method.
Independent observations within groups: clustering by classroom, household, clinic, or repeated exposure can make the ordinary pooled standard error too small.
Representative sampling or random assignment: random sampling supports generalization to a population; random assignment supports a treatment comparison under the study design. Neither property can be verified from aggregate counts.
Large counts: this calculator checks for at least 10 observed successes and 10 observed failures in each group. If a count is smaller, an exact, randomization, or other categorical-data method may be preferable.
Sampling fraction: when sampling without replacement, each sample is commonly kept below about 10 percent of its population unless the dependence is handled explicitly.

Passing the displayed count check does not certify the study. It only indicates that one numerical guideline for the normal approximation is met.

07Interpretation

Interpreting Difference in Proportions, Z, and P-Value

What does a positive or negative z score mean?

The sign follows p̂₁ − p̂₂. A positive z means group 1’s observed proportion is higher; a negative z means it is lower. For a two-sided test, equally large positive and negative z scores have the same p-value.

Is the pooled proportion the estimated effect?

No. The pooled proportion estimates a common rate only for the equality null model. The observed effect is p̂₁ − p̂₂. Keep these quantities separate when reporting the result.

Does a nonsignificant result prove the proportions are equal?

No. A p-value above alpha says the data do not provide enough evidence for the selected difference claim. Equality or equivalence requires a prespecified equivalence margin and a suitable analysis; it is not established by failing to cross a significance threshold.

How is this different from a two-proportion confidence interval?

The z test evaluates the null difference of zero with a pooled standard error. A two proportion confidence interval calculator estimates a range for p₁ − p₂ and generally uses an unpooled standard error. The two results complement one another but should not share the pooled formula automatically.

Can this calculator analyze paired or repeated data?

No. If the same subjects are measured twice, or observations are matched, the four entered totals omit the pairing information needed for a valid paired analysis.

08Method

Method, Limitations, and Two-Proportion Test Sources

This calculator implements the pooled large-sample z test for H₀: p₁ = p₂. It does not apply a continuity correction, compute Fisher’s exact test, accept a nonzero null difference, adjust for multiple comparisons, model clusters or survey weights, or establish causation. All-success or all-failure combined data produce a zero pooled standard error, so the calculator stops instead of reporting an undefined z statistic.

The method was checked against the NIST Engineering Statistics Handbook comparison of two proportions and Penn State STAT 200 lesson on two independent proportions. The visible threshold of 10 successes and 10 failures per group follows the Penn State large-count guideline. For high-stakes scientific, medical, policy, or financial decisions, use a prespecified analysis plan and qualified review.