Point Estimates and Margins of Error for a Proportion

From a Sampling Distribution to an Estimate

In Free Response Practice on Sampling Distributions for Proportions, you used a population proportion \(p\) to describe how sample proportions might vary across repeated random samples. In a real survey, however, \(p\) is usually unknown. We observe one sample and use its proportion, \(\hat{p}\), to estimate the population proportion.

That estimate is useful, but it is not an exact measurement of the population. A different random sample of the same size could produce a different \(\hat{p}\), even if the population and survey method stayed the same. The sampling distribution ideas from earlier in this course help us describe that ordinary sample-to-sample variation.

Definition: A point estimate is a single statistic used to estimate a population parameter. For a population proportion \(p\), the sample proportion \(\hat{p}=x/n\), where \(x\) is the number of sampled individuals with the characteristic and \(n\) is the sample size, is the point estimate of \(p\).

Keep the statistic and parameter distinct: \(\hat{p}\) describes the sample, while \(p\) describes the population. As discussed in Parameters Versus Statistics for Proportions, the sample result is known after collecting data; the population proportion is typically what we want to learn about. Earlier, in Mean of the Sampling Distribution of \(\hat{p}\), you learned that \(\hat{p}\) is unbiased under an appropriate random sampling process. That means its sampling distribution is centered at \(p\)—not that every particular sample’s \(\hat{p}\) equals \(p\).

Why One Point Estimate Is Not Exact

Suppose a town’s true proportion of residents who support a proposal is \(p\), but it is not known. A random sample provides one value of \(\hat{p}\). If another random sample were selected, its proportion might be slightly higher or lower. This change from sample to sample is sampling variability, as introduced in Sampling Variability Versus Bias in Proportions.

For example, a sample proportion of 0.58 does not establish that exactly 58% of the entire population has the characteristic. It is the observed share in that sample, used as an estimate of the population share. How much confidence to place in the estimate depends partly on the sample size and the sampling process. A larger random sample generally has less sampling variability, as you saw in How Sample Size Changes the Spread of \(\hat{p}\).

A point estimate alone does not tell us how far it is likely to be from \(p\). A margin of error gives a measure of the estimate’s sampling uncertainty at a stated confidence level. It is reported in the same units as the proportion—often as percentage points—and describes a typical allowance around the point estimate under the sampling model. The next tutorial develops how this margin fits into a one-proportion \(z\)-interval.

Definition: The margin of error is the distance from a point estimate to either endpoint of an estimate interval. For a one-proportion estimate, it depends on the confidence level, the sample size, and the estimated standard deviation of \(\hat{p}\). A margin of error describes sampling uncertainty; it is not a guarantee that the estimate is within that distance of the true population proportion.

Estimating a Margin of Error

For a one-proportion estimate, the estimated standard error uses \(\hat{p}\) in place of the unknown \(p\) in the standard deviation formula for a sample proportion. A confidence level determines a multiplier, commonly called a critical value. For a 95% confidence level, the standard Normal critical value is approximately 1.96.

\text{estimated standard error} =\sqrt{\frac{\hat{p}(1-\hat{p})}{n}} \qquad\text{and}\qquad \text{margin of error} =z^*\sqrt{\frac{\hat{p}(1-\hat{p})}{n}}

Here, \(z^*\) is the critical value for the stated confidence level. The formula estimates the amount of sample-to-sample variation in \(\hat{p}\); it does not use the unknown population proportion \(p\). The margin of error is often reported as a percentage-point amount. For instance, a margin of error of 0.04 is 4 percentage points, not 4 percent of the estimate.

The formula relies on a suitable sampling process and a Normal approximation. For a survey sampled without replacement from a finite population, check the 10% condition. Also check the Large Counts condition using the observed counts: there should be at least 10 sampled individuals with the characteristic and at least 10 without it. These checks support the calculation, but they cannot repair a biased sampling method.

Worked Example: Estimating a Population Proportion

Worked Example: Support for Extended Trail Hours

A parks department selects a random sample of 150 residents from a town of 5,000. In the sample, 87 residents support keeping a local trail open later in the evening. Find the point estimate of the proportion of all town residents who support extended hours, and explain what it means.

State. Let \(p\) be the proportion of all town residents who support extended trail hours. Let \(\hat{p}\) be the proportion of sampled residents who support them. The sample has \(x=87\) supporters and \(n=150\) residents.

Do. Calculate the sample proportion:

\hat{p}=\frac{x}{n} =\frac{87}{150} =0.58

Interpret. The point estimate of the proportion of all town residents who support extended trail hours is 0.58, or 58%. This means 58% of the sampled residents supported the change; it does not show that exactly 58% of all town residents support it. Another random sample could produce a different estimate.

The sample was selected randomly, which supports using it to estimate the town-wide proportion. Because the sample was taken without replacement, the 10% condition also applies: \(0.10(5{,}000)=500\), and \(150\leq500\). These details support treating the sample as informative, but the calculated point estimate remains a sample statistic, not a census of the town.

Worked Example: A Survey Margin of Error

Worked Example: Interest in a Community Compost Program

A community group takes a random sample of 600 residents from a town of 40,000. Of those surveyed, 342 say they would use a proposed compost program. Calculate the point estimate and an approximate 95% margin of error. Interpret the result and check the conditions relevant to the calculation.

State. Let \(p\) be the proportion of all town residents who would use the program. The point estimate is the sample proportion, based on \(x=342\) and \(n=600\):

\hat{p}=\frac{342}{600}=0.57

Plan and check conditions. The survey uses a random sample. Since it samples without replacement, check the 10% condition:

0.10N=0.10(40{,}000)=4{,}000 \qquad\text{and}\qquad 600\leq4{,}000

The 10% condition is met. The observed counts of supporters and non-supporters are \(342\) and \(600-342=258\), both at least 10. Thus, the Large Counts condition for a Normal approximation is met.

Do. For 95% confidence, use \(z^*=1.96\). Substitute the sample proportion and sample size into the margin-of-error formula:

\text{ME} =1.96\sqrt{\frac{0.57(1-0.57)}{600}} =1.96\sqrt{\frac{0.2451}{600}} =1.96\sqrt{0.0004085} \approx1.96(0.02021) \approx0.0396

The estimated margin of error is about 0.0396, or 3.96 percentage points, which rounds to 4.0 percentage points.

Conclude in context. The sample estimates that 57% of town residents would use the compost program, with an approximate 95% margin of error of 4.0 percentage points under the stated sampling model. The margin describes sampling uncertainty; it does not account for possible problems such as residents misunderstanding the question or people who do not respond differing from those who do.

Worked Example: How Sample Size Affects Margin of Error

Worked Example: Comparing Two Library Surveys

Two independent random samples are taken from a large population of library members to estimate the proportion who borrow digital books. One sample has 400 members, of whom 192 borrow digital books. The other has 1,600 members, of whom 768 do. Find the point estimate and approximate 95% margin of error for each sample, then compare the margins.

State. In both samples, the point estimate is 0.48:

\hat{p}_1=\frac{192}{400}=0.48 \qquad\text{and}\qquad \hat{p}_2=\frac{768}{1{,}600}=0.48

Plan and check conditions. Both samples are stated to be random. Assume the population has at least 16,000 members, so even the larger sample is at most 10% of the population; this supports treating observations within each sample as approximately independent. In the first sample there are 192 digital-book borrowers and \(400-192=208\) others. In the second there are 768 borrowers and \(1{,}600-768=832\) others. Both samples meet the Large Counts condition.

Do. Use the same 95% critical value, 1.96, for each calculation:

\text{ME}_1 =1.96\sqrt{\frac{0.48(0.52)}{400}} =1.96\sqrt{0.000624} \approx1.96(0.02498) \approx0.0490

\text{ME}_2 =1.96\sqrt{\frac{0.48(0.52)}{1{,}600}} =1.96\sqrt{0.000156} \approx1.96(0.01249) \approx0.0245

The approximate margins of error are 4.90 and 2.45 percentage points. The second sample is four times as large, and its margin of error is about half as large. This matches the square-root relationship: multiplying the sample size by four divides the estimated standard error, and therefore the margin of error at the same confidence level, by two.

Conclude in context. Although both samples estimate that 48% of members borrow digital books, the larger sample gives a more precise estimate under the random sampling model. A larger sample reduces sampling variability, but it does not eliminate bias from a flawed survey method.

What a Margin of Error Does Not Measure

A margin of error is about sampling variability under a model. It does not measure every way a survey could go wrong. A poorly worded question, a sample that leaves out part of the population, or nonresponse related to people’s opinions can make the estimate systematically too high or too low. Increasing \(n\) can reduce sampling variability, but it does not automatically fix these sources of bias.

The confidence level matters too. Holding the sample and \(\hat{p}\) fixed, a higher confidence level uses a larger critical value and produces a larger margin of error. This is a trade-off: asking for more confidence in the method generally means accepting a wider range of estimates. When comparing two survey margins, check that they use the same confidence level and are based on comparable sampling methods.

Common Mistakes and AP Exam Tips

Calling \(\hat{p}\) the population proportion. Write that \(\hat{p}\) is the sample proportion and point estimate of \(p\). State what the sample and target population are.
Claiming the estimate is exact. A sample statistic varies across random samples. Say it estimates the population proportion; do not claim it proves the population has exactly that proportion.
Confusing percent with percentage points. A margin of error of 0.04 is 4 percentage points. It is not necessarily a 4% relative change in the estimate.
Using \(p\) when it is unknown. For an estimated margin of error, substitute \(\hat{p}\) into the standard-error expression. The true population proportion is not available in a typical survey.
Ignoring the sampling method. A small margin of error does not guarantee a representative survey. Mention randomness and relevant conditions, and recognize that bias and nonresponse are not summarized by the margin of error.
Comparing margins without checking the confidence level. A higher confidence level tends to produce a larger margin of error. Compare sample sizes directly only when confidence levels and methods are comparable.

AP Exam Tip: A clear response distinguishes the statistic from the parameter: “In this random sample, \(\hat{p}=\dots\), so the point estimate of the population proportion is \(\dots\).” When asked about a margin of error, name its confidence level, give the value in percentage points, and describe it as sampling uncertainty—not as a guarantee about the true proportion or a measure of all survey error.

Key Takeaway

A sample proportion gives one data-based estimate of a population proportion. Because different random samples can produce different values, a margin of error helps describe the estimate’s sampling uncertainty at a stated confidence level. The quality of that description still depends on the sampling process and the conditions supporting the model.

Key takeaway: Use \(\hat{p}=x/n\) as the point estimate of \(p\). A margin of error summarizes sampling uncertainty around that estimate at a stated confidence level; it is not a guarantee of accuracy and does not account for every source of survey error.

Check Your Understanding

For each situation, distinguish the observed sample result from the population quantity it estimates.

A random sample of 250 students includes 145 who favor later school start times. Calculate the point estimate of the proportion of students who favor later starts, and interpret it in context.
Explain why the point estimate in Question 1 does not establish the exact proportion of all students who favor later starts.
A random survey of 500 households finds that 200 use a rain barrel. For a 95% margin of error, identify \(\hat{p}\), the two observed counts to check for Large Counts, and the critical value to use.
Without calculating, explain what happens to the 95% margin of error if the sample size is quadrupled while \(\hat{p}\) stays the same.
Name one source of survey error that a calculated margin of error does not measure, and explain why a larger sample does not necessarily remove it.