331+ free tools. No sign-up. No hidden fees. Explore the toolkit

Margin of Error Calculator

Every poll result comes with a margin of error — the “plus or minus” that says how far the true figure could be from the survey result. A poll of 1,000 people saying 52% support something, with a margin of error of ±3.1 points, really means the true support is likely somewhere between about 49% and 55%. The margin of error is the honest half of any survey number.

Enter your sample size, confidence level and proportion, and this calculator gives the margin of error, the standard error, and the interval the true value likely falls in.

How many people were surveyed. Must be a whole number.

95% is the standard choice for polls and surveys.

The share who gave the answer, as a decimal: 0.5 means 50%. Leave it at 0.5 if you don't know — it gives the largest (safest) margin of error.

Margin of error (95% confidence)

Save to Pinterest
Note: the margin of error covers sampling error only — the uncertainty from asking a sample instead of everyone. It does not cover biased questions, unrepresentative samples, or people who refuse to answer. A poll can have a tiny margin of error and still be wrong. The proportion p = 0.5 gives the largest margin of error, so it is the safe default when you don't know the true share.

How It Works

The formula is:

Margin of error = z × √(p × (1 − p) / n)

The square-root part is the standard error: how much a sample proportion typically wobbles around the true proportion. It shrinks as the sample grows — but only with the square root of the sample size, so doubling the sample only shrinks the error by about 30%. The z-value stretches the standard error to the confidence level you want: 1.645 for 90% confidence, 1.96 for 95%, 2.576 for 99%. Higher confidence means a wider interval — you pay for certainty with precision. Margin of error is about sampling uncertainty; percent error measures how far a single measurement is from the accepted value.

The proportion p matters too: p × (1 − p) is largest when p = 0.5 (a perfect 50/50 split) and shrinks towards 0 as p approaches 0 or 1. That is why pollsters use p = 0.5 when they don't know the true share — it gives the most conservative (largest) margin of error, so the real error can only be smaller. A poll where 90% say yes has a much smaller margin of error than one where 50% say yes, with the same sample size.

The result is a confidence interval: your sample proportion, plus or minus the margin of error. If 52% of a 1,000-person sample said yes at 95% confidence, the interval is 52% ± 3.1%, or roughly 48.9% to 55.1% — and "95% confident" means that if you ran the same poll 100 times, about 95 of the intervals would contain the true value.

Worked Example

Take a poll of n = 1,000 people at 95% confidence with p = 0.5. The standard error is √(0.5 × 0.5 / 1,000) = √0.00025 = 0.01581, and the margin of error is 1.96 × 0.01581 = 3.10%. So if 52% said yes, the true value is likely between 48.9% and 55.1%.

At 99% confidence with the same sample, the z-value rises to 2.576 and the margin of error becomes 2.576 × 0.01581 = 4.07% — the price of extra confidence is a wider interval. And with a smaller sample of n = 385 at 95% confidence, the standard error is √(0.25 / 385) = 0.02548, giving a margin of error of 1.96 × 0.02548 = 4.99% — which is why 385 is the classic sample size for a ±5% poll.

Assumptions

Sources: No external data is used on this page — every figure is calculated from the values you enter, using the standard margin-of-error formula MOE = z × √(p(1−p)/n) with z-values 1.645 (90%), 1.96 (95%) and 2.576 (99%). Last reviewed: 15 November 2026.

Built and tested by SumKitTool — see how we check our numbers.

How to Use This Calculator

Enter the sample size (a whole number), pick a confidence level — 95% is the standard — and enter the proportion as a decimal (0.5 means 50%). If you don't know the proportion, leave it at 0.5: that gives the largest, safest margin of error. Pick "Custom z-value" if you need a confidence level that isn't listed. The result updates as you type; Reset restores the worked example.

Try the worked examples: the defaults give ±3.10%. Switch to 99% confidence and it becomes ±4.07%; change the sample size to 385 (at 95%) and it becomes ±4.99%. Then change the proportion to 0.9 and watch the margin of error shrink — a 90/10 split is much less uncertain than a 50/50 one.

Tips and Practical Advice

To halve your margin of error, you need four times the sample — the error shrinks with the square root of n, not with n itself. Going from 1,000 to 4,000 respondents takes ±3.1% to about ±1.5%. That square-root law is why most serious polls stop at 1,000–1,500 people: beyond that, each extra point of precision gets very expensive.

A lead smaller than the margin of error is not a lead. If candidate A is at 52% and candidate B at 48% with a ±3.1% margin of error, the race is a statistical tie — the true gap could be anywhere from A ahead by 10 points to B ahead by 2. Always compare the gap with twice the margin of error, not with the margin itself.

Need to work backwards — "how many people must I survey to get a ±3% margin of error?" — use the Sample Size Calculator, which answers exactly that question.

Frequently Asked Questions

What is a margin of error?

The margin of error is the 'plus or minus' on a survey result — how far the true figure in the whole population could be from what the sample found. A poll of 1,000 people with a margin of error of ±3.10% that finds 52% support really means the true support is likely between 48.9% and 55.1%. It covers sampling error only, not biased questions or unrepresentative samples.

Why is 0.5 the default proportion?

Because p × (1 − p) is largest when p = 0.5 — a perfect 50/50 split gives the most sampling uncertainty, and therefore the largest margin of error. Using 0.5 when you don't know the true share is the safe, conservative choice: the real margin of error can only be smaller.

How do I reduce the margin of error?

Survey more people — but the error shrinks with the square root of the sample size, so halving it takes four times the sample. Going from 1,000 to 4,000 respondents takes ±3.10% to about ±1.55%. Lowering the confidence level from 95% to 90% also shrinks it, but at the cost of less certainty.

What does 95% confidence actually mean?

It means that if you ran the same survey 100 times, about 95 of the 100 confidence intervals would contain the true value. It does not mean there is a 95% chance this particular interval contains the truth — the true value either is in it or it isn't. It is a statement about the method, repeated many times, not about one poll.

Last updated: