Margin of Error and Its Meaning

What a Poll’s Margin of Error Tells You

A news poll may report a result such as “47% support the proposal, with a margin of error of plus or minus 3 percentage points.” The margin of error (often abbreviated \(ME\)) describes the distance from the sample estimate to either endpoint of the associated confidence interval. It gives a scale for the sampling uncertainty in the estimate—not a guarantee that the estimate is within that distance of the truth.

For a one-proportion \(z\)-interval, the sample proportion \(\hat{p}\) is the point estimate of the population proportion \(p\). The margin of error is the critical value multiplied by the estimated standard error. As in Structure of a One-Proportion z-Interval, adding and subtracting this margin of error gives the interval’s endpoints.

Definition: The margin of error for a one-proportion confidence interval is the distance from the sample proportion \(\hat{p}\) to either endpoint. It measures the interval’s sampling uncertainty at the stated confidence level.

ME=z^*SE_{\hat{p}} =z^*\sqrt{\frac{\hat{p}(1-\hat{p})}{n}}

Here, \(n\) is the sample size, \(z^*\) is the critical value for the chosen confidence level, and \(SE_{\hat{p}}=\sqrt{\hat{p}(1-\hat{p})/n}\) is the estimated standard error. For a 95% confidence interval, \(z^*=1.96\). The interval is \(\hat{p}\pm ME\).

A margin of error is often reported in percentage points. For example, an estimate of 47% with a margin of error of 3 percentage points gives endpoints of 44% and 50%. The margin is 3 percentage points, not 3% of 47%. In proportion notation, \(47\%=0.47\) and 3 percentage points is \(0.03\), so the endpoints are \(0.47-0.03=0.44\) and \(0.47+0.03=0.50\).

Interpret a Reported Margin of Error in Context

To interpret a poll’s stated margin of error, identify the population proportion the poll is estimating, the sample estimate, and the confidence level if one is reported. Then use the margin to find the interval around the estimate. The confidence-level interpretation follows the repeated-sampling idea from Interpreting a 95% Confidence Level Correctly: the level describes the long-run performance of the method, not a probability that one particular interval contains the fixed population proportion.

For instance, “47%, with a 3-percentage-point margin of error at the 95% confidence level” means the poll’s 95% confidence interval extends 3 percentage points below and above 47%. The interval is a range of plausible values for the population proportion, given this sample and method. It does not mean that 95% of people support the proposal, nor does it guarantee that the true proportion lies between 44% and 50%.

Interpretation pattern: If a poll reports a sample proportion \(\hat{p}\) and margin of error \(ME\), its interval is \(\hat{p}-ME\) to \(\hat{p}+ME\). Describe those endpoints as a confidence interval for the true proportion in the named population, at the stated confidence level.

The margin of error describes sampling uncertainty under the poll’s sampling method. It does not, by itself, measure every source of error. For example, the margin does not account for a sample that systematically leaves out part of the population, inaccurate responses, or confusing question wording. As discussed in earlier tutorials on sampling variability and bias, a small margin of error does not eliminate bias. Read a reported margin in light of how the poll was conducted.

Worked Example: Interpreting a News Poll’s Stated Margin

A hypothetical news poll reports that 47% of eligible voters in a city support a proposed transit plan, with a margin of error of 3 percentage points at the 95% confidence level. The poll used a random sample of 1,000 voters from a population of 60,000 eligible voters. Interpret the reported margin and find the interval.

State.
Let \(p\) be the true proportion of eligible voters in this city who support the transit plan. The poll’s sample estimate is \(\hat{p}=0.47\), and the reported margin of error is \(0.03\).

Plan and check conditions.
The sample is stated to be random. For sampling without replacement, the 10% limit is \(0.10(60{,}000)=6{,}000\), and \(1{,}000\leq6{,}000\). The estimated counts are \(1{,}000(0.47)=470\) supporters and \(1{,}000(0.53)=530\) non-supporters, both at least 10. These checks support using a one-proportion \(z\)-interval.

Do.
Subtract and add the stated margin: \(0.47-0.03=0.44\) and \(0.47+0.03=0.50\). The interval is \((0.44,0.50)\), or 44% to 50%.

Conclude.
We are 95% confident that the true proportion of eligible voters in this city who support the transit plan is between 44% and 50%. The 95% describes the long-run capture rate of the interval method under its conditions; it is not a guarantee for this one interval.

The “plus or minus 3 percentage points” is another way to report the interval centered at the sample estimate. It does not say that the estimate is known to be no more than 3 percentage points away from the truth. The true proportion is fixed, and this particular interval either contains it or does not.

Calculate a Margin of Error from \(n\) and \(\hat{p}\)

When the margin of error is not provided, calculate it using the one-proportion \(z\)-interval formula. Use the critical value for the stated confidence level and the estimated standard error based on the observed sample proportion. In this interval, \(\hat{p}\)—not the unknown population proportion \(p\)—is used in the standard error calculation.

Before applying the formula, check the conditions for the interval. As in Constructing a One-Proportion z-Interval by Hand, verify that the sample is random or comes from a suitable random process, check the 10% condition when sampling without replacement from a finite population, and check the Large Counts condition using the observed success and failure counts. These conditions support the interval calculation; they do not prove that the poll has no bias.

Worked Example: Calculating a 95% Margin of Error

A hypothetical community poll randomly samples 625 adults from a community of 50,000 adults. Of those sampled, 250 say they would use a proposed public garden. Calculate the margin of error for a 95% one-proportion \(z\)-interval and interpret it.

Let \(p\) be the true proportion of adults in the community who would use the public garden. The sample proportion is:

\hat{p}=\frac{x}{n} =\frac{250}{625} =0.40

The sample is stated to be random. The 10% limit is \(0.10(50{,}000)=5{,}000\), and \(625\leq5{,}000\). There are 250 successes and \(625-250=375\) failures, both at least 10. The random, 10%, and Large Counts conditions support using a one-proportion \(z\)-interval.

For a 95% interval, \(z^*=1.96\). First calculate the estimated standard error, then multiply by the critical value:

SE_{\hat{p}} =\sqrt{\frac{(0.40)(1-0.40)}{625}} =\sqrt{\frac{0.24}{625}} =\sqrt{0.000384} \approx0.01960

ME=1.96(0.01960) \approx0.0384 =3.84\text{ percentage points}

The 95% confidence interval is \(0.40\pm0.0384\), or approximately \((0.3616,0.4384)\). We are 95% confident that the true proportion of adults in this community who would use the proposed public garden is between about 36.16% and 43.84%. The margin of error is about 3.84 percentage points from the estimate to either endpoint.

Why the Sample Proportion Matters

For a fixed sample size and confidence level, the margin of error can still vary with \(\hat{p}\), because the estimated standard error includes \(\hat{p}(1-\hat{p})\). This product is largest when \(\hat{p}=0.50\). Consequently, for a given \(n\), a sample proportion near 0.50 produces a larger estimated standard error—and margin of error—than a proportion closer to 0 or 1.

This does not mean that one particular result is more accurate than another. It describes how the formula’s estimate of sampling variability depends on the observed proportion. Keep the confidence level fixed when comparing margins this way; the next tutorial examines how changing the confidence level affects interval width.

Worked Example: Comparing Margins for Two Possible Poll Results

Imagine a random sample of 400 residents from a city population of 30,000. Compare the 95% margins of error that would result if the sample proportion were 0.25 and if it were 0.50. Treat these as two possible outcomes from samples of the same size, not as two estimates from one sample.

For either sample, the sample is random and the 10% limit is \(0.10(30{,}000)=3{,}000\), which is greater than 400. If \(\hat{p}=0.25\), the estimated counts are \(400(0.25)=100\) successes and \(400(0.75)=300\) failures. If \(\hat{p}=0.50\), the estimated counts are 200 successes and 200 failures. All these counts are at least 10, so the Large Counts condition is met for both possibilities.

For \(\hat{p}=0.25\), the estimated standard error and margin of error are:

SE_{\hat{p}} =\sqrt{\frac{(0.25)(0.75)}{400}} =\sqrt{0.00046875} \approx0.02165

ME=1.96(0.02165) \approx0.0424 =4.24\text{ percentage points}

For \(\hat{p}=0.50\), the calculations are:

SE_{\hat{p}} =\sqrt{\frac{(0.50)(0.50)}{400}} =\sqrt{0.000625} =0.025

ME=1.96(0.025) =0.049 =4.90\text{ percentage points}

At the same sample size and 95% confidence level, the margin is larger for \(\hat{p}=0.50\) than for \(\hat{p}=0.25\). The formula gives the largest margin when the estimate is 0.50 because \(\hat{p}(1-\hat{p})\) is largest there. In context, these are margins for estimating the proportion of city residents with the characteristic measured by the poll; they do not describe other sources of survey error.

Common Mistakes and What a Full-Credit Answer Says

Confusing percentage points with percent. A 3-percentage-point margin around 47% gives 44% to 50%. It does not mean taking 3% of 47%.
Calling the margin the actual error. The actual difference between \(\hat{p}\) and the fixed population proportion is not known just from the margin. Say that the margin describes the interval’s sampling uncertainty under the method.
Treating the margin as a guarantee. A reported margin does not guarantee that the population proportion is within that distance of the estimate. The interval method can miss the true proportion.
Leaving out the population and characteristic. “The result is between 44% and 50%” is incomplete on its own. A full interpretation names whose proportion is being estimated and what characteristic is being counted.
Using \(p\) instead of \(\hat{p}\) in the interval’s estimated standard error. For a one-proportion \(z\)-interval, calculate \(SE_{\hat{p}}\) using the sample proportion.
Reporting only a calculator result. A margin of error is meaningful only when the stated confidence level and conditions support the interval method. Check the random, 10%, and Large Counts conditions.
Assuming the margin covers every kind of poll error. The formula addresses sampling variability under the method. It does not automatically measure nonresponse, response bias, or other problems in how data were collected.

AP Exam Tip: Show the margin formula, identify the confidence-level critical value, and substitute the sample proportion and sample size. Then express the result in percentage points and, if asked to interpret it, name the population and characteristic. A strong interpretation describes a confidence interval and the method’s confidence level without claiming that the margin guarantees the true proportion is inside it.

Key Takeaway

A poll’s margin of error is the distance from its sample proportion to either endpoint of the corresponding confidence interval. For a one-proportion \(z\)-interval, calculate it as \(z^*\sqrt{\hat{p}(1-\hat{p})/n}\). Interpret it as sampling uncertainty at the stated confidence level, not as a guarantee or a measure of every possible poll error.

Key takeaway: Read “estimate plus or minus margin of error” as the two endpoints of an interval for the population proportion. To calculate the margin, use the stated confidence level, \(n\), and \(\hat{p}\), and check the conditions for the interval method.

Check Your Understanding

Use the poll context, confidence level, and interval formula to answer each question.

A poll reports 52% with a margin of error of 2 percentage points at the 95% confidence level. What are the interval endpoints, and what population quantity do they estimate?
A random poll samples 500 people, with \(\hat{p}=0.30\). Write the formula for the 95% margin of error and identify the critical value to use.
Why is a margin of error of 4 percentage points not a guarantee that the sample estimate is within 4 percentage points of the true population proportion?
For a fixed sample size and confidence level, which value gives the larger margin of error: \(\hat{p}=0.20\) or \(\hat{p}=0.50\)? Explain using the formula.
Name one source of poll error that a one-proportion margin of error does not automatically measure.