On this concept: Formula

Not sure which lesson to do first? Find your level and get a plan →

Statistics · Grade 9-12

Margin of Error

The margin of error is the distance that the gap between a sample statistic and the true parameter stays within in a stated share of samples, usually 95% — the ± value in a result like '52% ± 3%.

margin of error=z∗×standard error

Quick Answer

The margin of error is the distance that the gap between a sample statistic and the true parameter stays within in a stated share of samples, usually 95% — the ± value in a result like '52% ± 3%.' You compute it as critical value times standard error, E=z∗⋅SE, and use it to judge how far a poll or survey result could be from the truth by chance. It shrinks as the sample grows (four times the people halves it) and grows when you demand more confidence. It covers random sampling error only, not a biased sample or a badly worded question.

Why This Matters

The margin of error decides whether a headline is news: "52% support, ±3 points" and "52% support, ±8 points" say very different things. Knowing how it depends on n lets you plan a survey: to get within 3 points on a proportion near 50% at 95% confidence you need about 1,070 people, and halving the margin takes four times as many. Ignore it and you will read noise as a trend.

Intuitive Explanation

A poll of 1,000 random voters finds 52% support. The standard error of that proportion is 0.52×0.48/1000≈0.0158, about 1.6 points. Multiply by 1.96 and you get the 95% margin of error: about 3.1 points. In 95% of samples of 1,000, the poll's percentage lands within 3.1 points of the true support, so the report reads "52%, ±3 points".

Three dials set the size. Confidence: 90% uses z∗=1.645 (a margin of 2.6 points here), 99% uses 2.576 (4.1 points). Spread: p^(1−p^) is largest at 50%, so close races have the widest margins. Sample size: the margin shrinks with n, so 4,000 voters give about 1.5 points. Population size barely matters; 1,000 random voters measure a city or a whole country about equally well.

The word "error" suggests mistakes, but the margin measures only the luck of the draw. If a poll reaches only people who answer landline phones, or asks a leading question, its result can be off by far more than its margin, and a larger sample does not help: 10,000 biased answers are still biased. When a sample was not random, the stated margin of error describes a precision the poll does not have.

Think of it like this. A phone map shows your location as a dot inside a shaded circle: you are probably somewhere in the circle. More satellite signals shrink the circle, as a bigger sample shrinks the margin, and asking for more certainty would widen it. Like the margin, the circle covers random jitter only; if the map itself is wrong, the circle cannot warn you.

Core idea. The margin of error is critical value times standard error: in, say, 95% of samples the estimate lands within that distance of the truth. It shrinks with n, grows with the confidence level, and covers random sampling error only, not bias.


When to Use

Use the margin of error when a result comes with a plus-or-minus, when you must decide whether a lead or a change is bigger than sampling noise, or when you plan the sample size a survey needs. Compute it as z∗×SE. For the full range, add and subtract it to get the confidence interval; for raw precision with no confidence level attached, stop at the standard error. When the worry is who was sampled, the fix is a better design, not a margin.

Tip. Ask: Do you need just the plus-or-minus amount that says how far a reported estimate could be from the truth by chance?

How to Recognize It

Before using a reported plus-or-minus figure, check what it is attached to, what confidence it carries, and what it leaves out.

  1. Does the problem give or ask for one plus-or-minus number next to an estimate, like 52% plus or minus 3 points?

    Yes, that number is the margin of error. If both endpoints are wanted, like 49% to 55%, the answer is the Confidence Interval.

  2. Is a confidence level, usually 95%, attached or implied?

    It must be, since the margin is a critical value times the standard error. A spread with no confidence level is the Standard Error alone.

  3. Is the worry about chance error rather than a slanted sample or a loaded question?

    Yes is what the margin covers. A poll of volunteers or a leading question carries bias, such as Sampling Bias, that no margin of error reflects.

  4. Are you comparing two candidates or answers from the same poll?

    Then the margin on the gap between them is roughly twice the reported margin, so a lead must clearly beat that doubled figure before it means much.

Margin of Error vs Confidence Interval vs Standard Error vs Sampling Bias

A one-line poll report packs several of these together. Ask whether you need the plus-or-minus amount, the full range, the raw spread before any confidence level, or an error that no plus-or-minus can show.

Margin of Error

Meaning
The margin of error is the plus-or-minus amount that covers chance error in a stated share of samples, usually 95%.
Key test
One plus-or-minus number tied to a confidence level.
Formula
E=z∗p^(1−p^)n
Example
A poll of 1,000 finds 60% support with a margin of about 3 points at 95% confidence.

Confidence Interval

Meaning
A confidence interval is the range running from the estimate minus the margin of error to the estimate plus it.
Key test
Both endpoints of the range.
Formula
p^±E
Example
57% to 63% support among all voters.

Standard Error

Meaning
The standard error is the typical chance error of an estimate, before any confidence level is applied.
Key test
The raw spread, about half the 95% margin.
Formula
SE=p^(1−p^)n
Example
0.6×0.41000≈0.015, or about 1.5 points.

Sampling Bias

Meaning
Sampling bias is error from a selection method that favors some people, which the margin of error leaves out.
Key test
Error from who was asked, which no plus-or-minus can show.
Example
An online poll of 1,000 volunteers still reports plus or minus 3 points but leans toward heavy internet users.

Formula & Notation

margin of error=z∗×standard error
The margin of error for a mean is E=zα/2⋅σn. For a proportion, E=zα/2⋅p^(1−p^)n. Doubling n reduces E by a factor of 2.

How to read it: E is the margin of error. The confidence interval is θ^±E, where θ^ is the sample estimate.

Worked Examples

Example 1 — Margin of error for a poll

Easy

A random sample of 900 voters finds that 54% plan to vote yes on a school budget. Find the 95% margin of error and the confidence interval. Does the poll suggest that a majority supports the budget?
Show the steps and the answer
  1. Find the standard error: 0.54×0.46/900=0.000276≈0.0166.

    The margin is built from how much p^ varies across samples of 900 voters.

  2. Multiply by the critical value: E=1.96×0.0166≈0.033, about 3.3 points.

    1.96 standard errors cover the middle 95% of the normal curve.

  3. Add and subtract: 54%±3.3% gives 50.7% to 57.3%.

    The interval is the estimate plus and minus the margin.

  4. Compare with 50%: the whole interval lies above it.

    Every plausible value for true support is a majority, though only just at the low end.

Answer: E≈3.3 points; the interval is 50.7% to 57.3%, so the poll suggests a narrow majority in favor.

Takeaway: Margin of error =z∗×SE. Before calling a result, check where key values such as 50% fall relative to estimate ±E.

Example 2 — A precise-looking margin on a self-selected sample

Standard

A news website posts a poll on its homepage, and 5,000 readers choose to answer it; 70% say a new law is bad. The site reports "70%, margin of error ±1.3 points". Is that a fair description of what all adults in the country think?
Show the steps and the answer
  1. Check the arithmetic: 1.960.7×0.3/5000≈1.96×0.0065≈0.013.

    The 1.3 points is what a random sample of 5,000 would give, so the formula itself was applied correctly.

  2. Ask how the sample was chosen: readers decided for themselves whether to answer.

    People with strong feelings, and people who read this particular site, are much more likely to respond, so the sample is not random.

  3. Recall what the margin covers: only the chance variation between random samples.

    It says nothing about bias from who answered, and 5,000 self-selected answers can sit far from the population's view.

Answer: No. The ±1.3 points assumes a random sample; this one chose itself, so the 70% could be off by much more and the margin of error does not apply.

Takeaway: A margin of error is only as good as the randomness of the sample. Ask how people were chosen before trusting a small ±.

Example 3 — Find the mistake: a lead inside the noise

Application

A random poll of 1,000 voters in a two-candidate race gives Lee 52% and Diaz 48%, with a margin of error of ±3.1 points. Maya writes: "Lee leads by 4 points, which is more than the 3.1-point margin of error, so Lee is clearly ahead." Find the mistake and correct it.
Show the steps and the answer
  1. Build each candidate's interval: Lee 48.9% to 55.1%, Diaz 44.9% to 51.1%.

    The two ranges overlap, so the poll is consistent with, for example, Diaz at 50.5% and Lee at 49.5%.

  2. Find the error: 3.1 points is the margin for one candidate's share, not for the lead.

    In a two-person race every point Lee gains is a point Diaz loses, so the lead 52−48 moves twice as fast as either share.

  3. Compute the lead's margin: about 2×3.1=6.2 points, so the lead is plausibly 4±6.2, from −2.2 to 10.2 points.

    That range includes 0 and negative values, and a negative lead means Diaz is ahead.

Answer: The race is too close to call: the 4-point lead has its own margin of about ±6.2 points, so a Diaz lead is still plausible.

Takeaway: In a two-way race, compare the lead with about twice the reported margin of error, or check whether the candidates' intervals overlap.

Where You'll See It

  • News poll '52% ± 3%'

    The ±3% is the margin of error: the reported 52% could reasonably be anywhere from 49% to 55%, so a 51% vs 49% result is really 'too close to call.'

  • Class election exit survey

    Asking 80 of 400 students which candidate they voted for gives an estimate with a wide margin of error; surveying 300 students shrinks it, because margin of error falls as sample size rises.

  • Two candidates within the margin

    If Candidate A polls 48% ± 4% and Candidate B polls 45% ± 4%, the overlap means the 3-point lead is inside the margin of error, so the data cannot confidently say who is ahead.

Common Mistakes

  1. The slip: A bigger margin of error means the survey was done badly.

    Why it's tempting: 'Error' sounds like a mistake, so a large value feels like poor work.

    Instead: Margin of error mainly reflects sample size and confidence level, not quality; a small honest survey simply has a larger margin than a huge one.

  2. The slip: A 3-point lead in a poll means the leader is really ahead, regardless of the margin of error.

    Why it's tempting: A clear numeric gap looks like a clear winner.

    Instead: The lead between two candidates in the same poll has its own margin of error, about twice the reported one, so a 3-point lead with a 3-point margin is well within chance; compare the lead with about twice the margin before claiming anyone is ahead.

  3. The slip: Doubling the sample size cuts the margin of error in half.

    Why it's tempting: Twice as many people feels like it should give twice the precision.

    Instead: The margin shrinks with n, so halving it takes four times the sample. For a 52% result, 1,000 voters give about 3.1 points, 2,000 give about 2.2 points, and 4,000 give about 1.5 points.


Mini Practice

Try these on your own. Tap Reveal when you want to check.

  1. A survey reports that 62% of residents favor a new library, with a margin of error of 3 percentage points. What are the lowest and highest plausible values for the true percent?

    Hint: Go the margin below and above the estimate.

  2. A poll of 1,000 voters finds 47% support. Compute the 95% margin of error using z∗=1.96 and SE=p^(1−p^)/n.

    Hint: Find the standard error with p^=0.47 first.

  3. A poll keeps its 95% confidence level but increases its sample from 250 to 1,000 people. What happens to the margin of error?

    Hint: The margin is proportional to 1/n. How does n change?

  4. How large a sample is needed for a 95% margin of error of 3 percentage points, using the safe guess p^=0.5?

    Hint: Solve 1.960.25/n=0.03 for n and round up.

  5. A poll write-up lists a standard error of 1.5 points and a 95% margin of error of 2.9 points. A reader wants the plus-or-minus range around the estimate. Which number should they use, and how are the two linked?

    Hint: One of these is multiplied by a critical value to make the other.

  6. A news anchor says: "Candidate A leads Candidate B 51% to 49%, with a margin of error of 3 points, so A is winning." Correct the claim.

    Hint: Apply the margin to both candidates' numbers and compare the ranges.

Want the full set?

50 practice questions for this concept — free to try, every one with a complete worked solution showing the why, not just the answer.

Frequently Asked Questions

Is there a quick way to estimate a poll's margin of error?

Yes. For a result near 50%, the 95% margin is 1.960.25n=0.98n1.96\sqrt{\frac{0.25}{n}} = \frac{0.98}{\sqrt{n}}, which is about 1n\frac{1}{\sqrt{n}}. So 100 people give about 10 points, 400 give 5, and 1,000 give about 3. The margin is largest near 50%; for a 90% result from the same sample it is only 60% as large, since 0.9×0.1=0.3\sqrt{0.9 \times 0.1} = 0.3 while 0.25=0.5\sqrt{0.25} = 0.5.

Why do polls with a 3-point margin sometimes miss an election by more?

The margin covers only random sampling error, and only in 95% of samples. Real polls also face people who refuse to answer, wrong guesses about who will actually vote, and minds that change before election day. Studies of past elections have found that actual poll errors are often about twice the stated margin. Averaging many polls reduces chance error but cannot fix a mistake they all share.

Why is the margin of error for a subgroup so much bigger?

Because the subgroup is a smaller sample. If a poll of 1,000 people has a margin of about 3 points overall, the 200 people aged 18 to 29 inside it have a margin of about 7 points, since 1200≈0.07\frac{1}{\sqrt{200}} \approx 0.07. News stories about a big swing among young voters in a single poll are often just reading that larger wobble.

Learning Path

Margin of Error

You are here

You need Standard Error (how much an estimate wobbles) and Confidence Interval (estimate ± a half-width). The margin of error isolates that half-width, E=z∗⋅SE, so you can see how confidence level and sample size control it and solve for the n a survey needs. It leads to Statistical Significance, which asks whether a difference is larger than chance alone would produce.

See Also