What this Sample Size Calculator for a Proportion calculator does
The Sample Size Calculator for a Proportion estimates how many independent observations a study needs to estimate a population proportion, such as the share of respondents supporting a proposal or the share of reviewed records meeting a criterion. It uses three inputs: a confidence level represented by a z-score, a margin of error in percentage points, and an expected proportion. Its result is an estimated number of people or observations.
The calculation uses a normal approximation for a proportion. It is a planning estimate under stated assumptions, rather than a guarantee that a particular study will achieve a given precision. For example, a survey estimate can be affected by how participants were selected, whether people respond, how questions are worded, and how the data are analyzed. This tool does not account for those design features, finite population corrections, clustering, or adjustments for multiple outcomes.
The formula is n = z² × p × (1 − p) ÷ e². Here, n is the estimated sample size, z is the selected confidence z-score, p is the expected proportion expressed from 0 to 1, and e is the margin of error expressed from 0 to 1. Because a study cannot enroll a fraction of a person, interpret the displayed estimate as a planning target and round upward to the next whole observation.
How to use the Sample Size Calculator for a Proportion calculator
Choose a confidence level, enter a margin of error in percentage points, and enter the expected proportion as a percentage. The default expected proportion is 50%. Use that conservative planning value when there is no reasonable prior estimate, because it produces the largest sample size under this formula for a fixed confidence level and margin of error.
- Confidence level: This choice sets the z-score used in the calculation. Higher confidence levels use larger z-scores and therefore require larger samples.
- Margin of error: Enter the absolute precision you want, in percentage points. A five-point margin means an estimate near 40% has an approximate interval width of five percentage points on either side under the model.
- Expected proportion: Enter a plausible percentage from 1% through 99%. A prior study or pilot may inform it, provided the population and measurement are sufficiently similar.
For a worked example, choose 95% confidence, enter a 5 percentage-point margin of error, and use an expected proportion of 50%. The formula is 1.96² × 0.50 × (1 − 0.50) ÷ 0.05². That equals 3.8416 × 0.25 ÷ 0.0025, or 384.16 observations. Round upward, so the simple estimate is 385 observations. This is the mathematical result before any allowance for nonresponse or design effects.
If the expected proportion is known to be near 20%, with the same confidence and margin settings, the calculation is 1.96² × 0.20 × 0.80 ÷ 0.05², or 245.86. That result rounds upward to 246 observations. These examples use proportions as decimal fractions inside the formula, even though the inputs are entered as percentages.
How the Sample Size Calculator for a Proportion formula works
The formula comes from the approximate standard error of a sample proportion. That standard error depends on p × (1 − p) divided by the sample size. Solving the margin-of-error relationship for sample size gives the equation used here. The confidence z-score reflects the selected normal-approximation confidence level, while the requested margin of error controls how narrow the estimate should be.
The relationship is not linear in every input. The z-score is squared, and the margin of error is squared in the denominator. Halving the margin of error therefore multiplies the sample estimate by four when the other inputs stay fixed. Moving from 95% to 99% confidence also increases the estimate because the selected z-score is larger.
The proportion term is largest at 50%, where p × (1 − p) equals 0.25. Values closer to zero or one produce a smaller estimate in this formula. If the true proportion is uncertain, using 50% is a cautious choice for this specific calculation. It does not cover other sources of uncertainty, such as poor measurement or biased sampling.
Use cases for the Sample Size Calculator for a Proportion
This estimate can support early planning in decision analysis when a team needs to gather evidence about a percentage before comparing options. A service organization might estimate the share of users who complete a process, or a policy team might measure the share of surveyed residents who favor an option. In each case, the estimate helps make the desired precision and data-collection effort explicit.
It can also help compare design choices. A decision maker can calculate estimates for several margins of error and confidence levels, then discuss the resource implications alongside the value of more precise information. The result is one input to that discussion. It does not identify the best decision, determine whether a study is worthwhile, or measure the consequences of acting on an inaccurate estimate.
For a small pilot, the tool can show how an initial estimate of the proportion affects a later plan. Pilot findings should be treated carefully when the pilot sample is small or differs from the intended study population. A team should document the assumptions used so that others can understand why a particular target was chosen.
Other factors to consider when calculating sample size
This calculator models a simple random sample of independent observations and a single proportion. Real studies may need a larger sample. Nonresponse reduces the number of usable observations, and cluster sampling can make observations more alike than the simple formula assumes. If data are collected in groups, a statistician may need to apply a design effect. If the population is small relative to the planned sample, a finite population correction may also be relevant.
The calculation also does not target statistical power for comparing groups, detecting a difference, or testing a hypothesis. Those questions require additional inputs, such as the smallest effect that matters, variability, allocation between groups, and analysis method. A study measuring a mean rather than a proportion also needs a different formula. Match the calculation to the study question before using its estimate in a budget or research plan.
Finally, distinguish the number of completed usable observations from the number of people invited. If a study expects some participants not to respond or some records to be unusable, the invitation target must be higher. Estimate that adjustment from relevant prior experience and state it separately, since this calculator does not include a response-rate input.
FAQ
What does the confidence level mean?
The confidence level selects the z-score used by the normal approximation. In repeated comparable samples, the method is designed so that intervals built this way cover the fixed population proportion at the stated rate, subject to the model and sampling assumptions. It does not mean there is that probability that this particular computed interval contains the proportion.
Why does a 50% expected proportion give a larger estimate?
The variance term p × (1 − p) reaches its maximum at 50%. That makes the required sample largest for a fixed confidence level and margin of error. When there is no dependable prior estimate, 50% is commonly useful as a conservative planning assumption within this formula.
Should I round the result?
Yes. Round up to the next whole observation because a fraction of an observation cannot be collected. Then consider whether nonresponse, exclusions, or a more complex design requires an additional adjustment.
Does the estimate tell me how many people to contact?
No. It estimates completed independent observations under the stated model. The number to contact depends on expected participation and the share of data that will remain usable. Plan that separately using evidence suited to the study population.
Can I use this for a comparison between two groups?
Not directly. A comparison needs a sample size method based on the outcome, expected difference, variability, confidence level, power, and group allocation. This calculator estimates one population proportion and does not model a test or comparison.