Same Percentage, Different Amount of Information
A percentage can make a result easy to compare, but it leaves out how many people or items were counted. If 3 of 5 people select an option, the sample proportion is \(3/5=0.60\), or 60%. If 300 of 500 select it, the sample proportion is \(300/500=0.60\), also 60%. The percentages match, but the amount of information behind them does not.
In Finding a Sample Proportion From Counts, you learned to calculate a sample proportion by dividing the count in the category of interest by the sample size. In Converting Proportions to Percentages, you learned to express that proportion as a percent. Here, we add an important habit: when interpreting a percentage, report or check the count and the total that produced it. A percentage from a small count can change substantially when just one response changes.
Consider the effect of one more response in the category. With 5 responses total, changing the category count from 3 to 4 changes the percentage from 60% to 80%: a 20-percentage-point increase. With 500 responses total, changing the count from 300 to 301 changes the percentage from 60% to 60.2%: a 0.2-percentage-point increase. One response has a much larger effect on the percentage when the total is small.
This does not mean that a small sample’s percentage is automatically wrong, or that a large sample’s percentage is automatically right. It means that the small sample’s percentage can be more sensitive to the particular individuals or cases included. To decide what a result says beyond the observed group, also consider how the data were collected and whether the sample represents the population of interest.
Keep the Count and the Total in View
A percentage without its denominator can hide useful context. “60% chose the option” could describe 3 of 5 people, 30 of 50 people, or 300 of 500 people. The percentages are identical, but the counts reveal different sample sizes. When space permits, give both the percentage and the count, such as “60% (3 of 5)” or “60% (300 of 500).”
For a sample proportion, \(x\) is the count in the category of interest and \(n\) is the total sample size. As in Choosing the Correct Denominator for a Percentage, make sure that \(n\) is the total for the group the percentage describes. Then keep the count and total attached to the result:
In this notation, \(\hat{p}\) (“p-hat”) is the sample proportion. For 3 successes out of 5 observations, \(\hat{p}=3/5=0.60\). For 300 successes out of 500 observations, \(\hat{p}=300/500=0.60\). Each sample has the same observed proportion, but the second includes many more observations.
A larger sample often gives a more stable estimate of a population proportion when the data come from a suitable sampling process. The reason is intuitive: if the sample includes many observations, an unusual response or a few different cases usually have less effect on the overall percentage. This is a tendency, not a guarantee that every large sample percentage is closer to the population value.
Worked Examples
Worked Example: Matching Percentages With Different Totals
Two fictional groups of volunteers are asked whether they would use a proposed community tool-sharing service. In Group A, 3 of 5 volunteers say yes. In Group B, 300 of 500 volunteers say yes. Compare the observed percentages and explain why the results do not carry the same amount of information.
For Group A, the sample proportion is:
For Group B, the sample proportion is:
The observed percentages are equal: both groups have a 60% yes response. But Group A’s percentage is based on only five volunteers, while Group B’s is based on 500. In Group A, one volunteer represents \(1/5=20\%\) of the group. In Group B, one volunteer represents \(1/500=0.002=0.2\%\) of the group.
Conclusion: Both groups have an observed yes percentage of 60%, but Group B’s result is less sensitive to the response of any one volunteer. If the groups were selected using the same suitable method to represent a population, the result based on 500 observations would generally be more stable evidence about that population’s yes-response percentage. The matching percentages alone do not establish that the groups represent the population equally well.
Worked Example: One Response Changes a Small-Sample Percentage Sharply
A fictional school club asks five students whether they would attend a weekend event. Three say yes. A sixth student is then asked, and that student also says yes. Find the percentage before and after the additional response, and explain the effect.
Before the additional response, 3 of 5 students say yes:
After one more student says yes, the count is 4 out of 6, not 4 out of 5, because the total also increases:
The percentage rises by about \(66.7\%-60\%=6.7\) percentage points. This particular change is smaller than 20 percentage points because the added yes response increases both the numerator and denominator. To see the effect of changing one response while holding the total at five, compare 3 of 5 with 4 of 5:
Conclusion: A single response can have a large effect on a percentage from a very small group. Here, adding an observation changed the observed percentage from 60% to about 66.7%; if one of the original five responses had instead changed from no to yes, the percentage would have risen by 20 percentage points. With a small total, describe the count as well as the percentage so the reader can see how many responses the summary represents.
Worked Example: A Large Sample That Still Has a Limitation
A fictional town wants to estimate the percentage of all residents who favor a new park. A volunteer collects responses from 500 people attending a weekend gardening event; 300 say they favor the park. Calculate the observed percentage, then explain what the sample size does and does not tell us.
The sample proportion is \(300/500\). Converting it to a percentage gives:
The observed percentage is 60% among the 500 event attendees who responded. Because the sample is large, any one response has a small effect: one person is \(1/500\times100\%=0.2\) percentage points of the sample total.
However, the respondents were gathered at a gardening event. Residents who attend such an event may differ in their interest in parks from residents who do not attend. The sample could therefore overrepresent people with a particular viewpoint, even though it contains 500 responses. The number of responses describes the sample’s size; it does not show that the sample represents all town residents.
Conclusion: In this fictional group of 500 event attendees, 60% favor the park. The large sample makes the percentage less sensitive to any one response, but the way the respondents were recruited limits how confidently the result can describe all town residents. A large sample is not a substitute for a sound sampling method.
What “Stronger Evidence” Means Here
When comparing 3 of 5 with 300 of 500, “stronger evidence” does not mean that the larger percentage is more impressive or that its value must be closer to the truth. It means that, under comparable and appropriate sampling conditions, a result based on more observations is generally less affected by random differences in which individuals happened to be selected. The larger sample usually gives a more stable basis for estimating a population proportion.
That qualification matters. Suppose the smaller group is a properly selected random sample from the population of interest, while the larger group consists only of volunteers recruited from one special-interest event. The larger group may be less useful for describing the whole population despite containing far more responses. Sample size and sample selection answer different questions: size concerns how many observations there are; selection concerns whom those observations represent.
Also keep separate the observed percentage and a claim about the population. “60% of the 5 surveyed people said yes” is a statement about those five people. “About 60% of all residents would say yes” extends the result to a larger group and needs support from the way the sample was obtained. A percentage alone cannot supply that support.
Common Mistakes and AP Exam Tips
- Reporting only the percentage. “60% said yes” hides whether the result was 3 of 5 or 300 of 500. Include the count and denominator when they are available.
- Claiming the percentages are equally informative because they match. The observed proportions are equal, but the small sample is more sensitive to individual responses. State both facts.
- Saying a large sample guarantees an accurate result. A large sample can still be biased or unrepresentative. Discuss how people or cases entered the sample.
- Confusing evidence about the sample with evidence about a population. The sample percentage describes the observed cases directly. A population claim depends on how the sample was selected.
- Forgetting to update the denominator. When new observations are added, both the category count and the total may change. For example, adding one yes to 3 of 5 gives 4 of 6, not 4 of 5.
- Using “significant” without support. In an introductory comparison like this, do not call a difference statistically significant based only on the percentages or sample sizes. Describe the observed results and the limits of the sampling information given.
A clear AP-style explanation names the category, gives the count and total, and distinguishes the observed result from a population conclusion. For example: “In the sample of 5 volunteers, 3, or 60%, said yes. The sample of 500 also had a 60% yes response, but one response represents a much smaller share of the larger sample. If the samples were selected comparably and representatively, the larger sample would generally provide a more stable estimate; sample size alone does not establish representativeness.”
Check Your Understanding
Use the counts and sampling descriptions to explain what each percentage does—and does not—tell you.
- Two fictional samples have 4 of 10 and 40 of 100 responses in a category. Find both percentages. Which sample percentage is more affected by one response, and why?
- A fictional poll reports that 8 of 10 people prefer a new bus route. State the sample percentage and explain why reporting “80%” alone may give an incomplete impression.
- In a group of 200, 120 people choose an option. If one more person in the group had chosen it instead of not choosing it, how many out of 200 would choose it, and what would the percentage be?
- A survey has 1,000 responses, all collected from visitors to a hobby convention. Give one reason the large sample size does not guarantee a good estimate for the whole city.
- Write one careful sentence comparing 3 of 5 with 300 of 500 when both give 60%. Include what is the same and what differs.