# Margin of error on the SAT: what that plus-or-minus really says > A statistic plus or minus a margin of error claims a plausible range for the population value. What shrinks it, what it never covers, and the phrasings. Published 2026-05-29 | Math | StudyHall Canonical: https://trystudyhall.com/blog/sat-margin-of-error At some point the Math section stops asking you to compute and asks you to interpret: a random sample, a mean of 4.2 hours, a margin of error of 0.3 hours, and four choices offering to tell you what that means. Three are wrong in ways the SAT recycles constantly. The margin of error is one idea, and once you can say it in a sentence, these are fast points. **Key takeaways:** - A statistic plus or minus its margin of error gives a **plausible range for the population value**, from a random sample. - It’s a claim about the population’s **mean or proportion**, never about any individual. - A **larger random sample** shrinks the margin. Re-surveying the same people does not. - It does **not** cover bias: a huge non-random sample is just precisely wrong. ## What the interval actually claims Say researchers sample 400 students at random from a large high school and find a mean of 4.2 hours of daily screen time, margin of error 0.3 hours. The claim is exactly this: it is plausible that the mean for **all** students at the school is between 3.9 and 4.5 hours. Three parts carry everything. *Plausible*, because a sample can never guarantee. *The mean*, because the interval describes the average, not any one student. *At the school*, because that’s the population sampled, not the district or the state. Why an interval at all? A different random sample would land on a slightly different mean; the [margin of error](https://trystudyhall.com/learn/inference-and-margin-of-error) is the honest width of that wobble. ## What shrinks the margin of error One lever matters on the SAT: **sample size**. A larger random sample pins the estimate down, and every correct answer to a "which change would decrease the margin of error" question is some version of "select a larger random sample." Things that don’t shrink it: surveying the same respondents twice, or swapping in a bigger but non-random group, which trades wobble for bias. No fresh randomly selected data, no smaller margin. ## What it never covers > **Watch out: "The margin of error accounts for bad sampling"** It doesn’t. The margin quantifies random sampling variation only, and the math *assumes* the sample was random. Volunteers or one lunch table break that assumption. Bias isn’t inside the plus-or-minus; it’s outside the whole calculation. ## Try one: interpret the interval **Example: Inference and margin of error.** > Researchers surveyed a random sample of 400 students at a large high school and found a mean daily screen time of 4.2 hours, with a margin of error of 0.3 hours. Which is the most appropriate interpretation of these results? - A) Every student at the school has a daily screen time between 3.9 and 4.5 hours. - B) It is plausible that the mean daily screen time of all students at the school is between 3.9 and 4.5 hours. - C) The mean daily screen time of all students in the state is between 3.9 and 4.5 hours. - D) The mean daily screen time of all students at the school is exactly 4.2 hours. **Answer:** B. The interval runs from 3.9 to 4.5 and describes one thing: the plausible range for the **mean** among **all students at this school**. The first choice applies it to individuals, whose times can sit far outside it. The state version stretches to a population never sampled. And "exactly 4.2" treats the statistic as the truth, which is what the margin of error exists to deny. ## Try one: shrink the margin **Example: Inference and margin of error.** > A city planner surveys a random sample of 200 residents to estimate the mean commute time of all city residents, reporting the estimate with a margin of error. Which change to the survey design would most likely decrease the margin of error? - A) Selecting a random sample of 800 residents instead - B) Selecting a random sample of 50 residents instead - C) Asking each of the 200 residents to report their commute twice and averaging the reports - D) Replacing the random sample with 500 residents of the neighborhood nearest the planner’s office **Answer:** A. Margin of error shrinks as the random sample grows, so 800 beats 200. Fifty residents grows the margin. Asking the same 200 people twice adds no new information about the city. And the 500 neighbors are a bigger but non-random group: that introduces bias, the one problem the margin cannot absorb. ## The phrasings to expect - **"Which is the most appropriate interpretation..."** Pick *plausible range*, *population mean or proportion*, *the sampled population*. Eliminate individuals, certainty, and wider groups. - **"Which change would decrease the margin of error?"** Larger random sample. Everything else is decoration. - **"Can this estimate be applied to [another population]?"** Only to the one randomly sampled. - **"Which conclusion is justified?"** [Evaluating statistical claims](https://trystudyhall.com/blog/sat-evaluating-statistical-claims) covers which study designs earn which conclusions. **[Practice margin of error questions](https://trystudyhall.com/learn/inference-and-margin-of-error)**: Drill these questions with a tutor that makes you say what the interval claims before you see the choices. ## FAQ ### What does margin of error mean on the SAT? The plus-or-minus attached to a sample statistic: together they give the plausible range for the population’s mean or proportion, based on a random sample. It measures sampling wobble, not certainty and not individuals. ### Does the margin of error account for a biased sample? No. It measures random sampling variation and assumes the sample was random. A sample of volunteers or a convenient group can be badly off in a way no margin of error captures. ### Does the interval apply to individual people in the population? No. The interval describes the population’s mean or proportion only. Individual values routinely fall far outside it, so a plausible range for the mean says nothing about any particular person. [Read the HTML version](https://trystudyhall.com/blog/sat-margin-of-error) | [Browse the StudyHall blog](https://trystudyhall.com/blog)