Inference questions ask you to reason from a sample back to a whole population — and to understand what margin of error actually tells you about that estimate. This guide covers the vocabulary, the logic, and two full worked examples so these questions stop feeling like guesswork.
Why Samples, Not Whole Populations
It's often impossible or impractical to survey an entire population — every student in a country, every product off a factory line — so researchers survey a smaller, randomly selected sample and use it to estimate something about the full population. The SAT tests whether you understand what makes this estimate trustworthy: random selection, adequate sample size, and an honest accounting of uncertainty through margin of error.
Margin of Error, Explained Plainly
A margin of error describes a range around a sample estimate within which the true population value is likely to fall. If a poll reports that 42% of respondents prefer a product, with a margin of error of 3 percentage points, the true population percentage is estimated to fall somewhere between 39% and 45%. A margin of error is typically reported alongside a confidence level (commonly 95%), meaning the described interval would contain the true population value in about 95 out of 100 samples drawn the same way — though the SAT rarely requires you to compute a confidence level, only to reason about what the interval implies. Two things affect margin of error in predictable directions:
- Larger sample size → smaller margin of error. More data generally narrows the range of plausible values.
- More variability in the population → larger margin of error. If responses are wildly inconsistent, a sample gives you less certainty.
You won't be asked to calculate a margin of error from a formula involving standard deviation and sample size — the SAT tests the conceptual relationship, not the computation.
Worked Example: Interpreting a Confidence Interval
A school district wants to estimate the average number of hours per week that its students spend on homework. Researchers randomly sample 150 students from the district and find:
| Statistic | Value |
|---|---|
| Sample size | 150 students |
| Sample mean | 6.2 hours/week |
| Margin of error | 0.5 hours |
Question 1: Based on this sample, what is a reasonable interval estimate for the true mean homework time of all students in the district?
Step 1 — Subtract and add the margin of error to the sample mean.
\(6.2 - 0.5 = 5.7\) and \(6.2 + 0.5 = 6.7\)
So the estimated interval is 5.7 to 6.7 hours per week.
Question 2: If the district had instead sampled 600 students (using the same random method), what would you expect to happen to the margin of error?
Answer: It would likely decrease. A larger, still-random sample tends to produce a more precise estimate of the population mean, narrowing the margin of error — the interval might shrink to something like 5.9 to 6.5, for example, rather than the wider 5.7-to-6.7 range from the smaller sample.
This is the exact reasoning skill tested repeatedly: connect sample size to precision, without needing to compute an actual formula.
Worked Example: Comparing Two Polls with Different Sample Sizes
Two independent polling organizations survey voters ahead of an election, both using proper random sampling:
| Poll | Sample size | Reported support for Candidate X | Margin of error |
|---|---|---|---|
| Poll A | 400 voters | 52% | ±4.9 percentage points |
| Poll B | 1,600 voters | 52% | ±2.45 percentage points |
Question 1: What is the interval estimate for Candidate X's true support according to each poll?
Poll A: \(52 - 4.9 = 47.1\%\) to \(52 + 4.9 = 56.9\%\).
Poll B: \(52 - 2.45 = 49.55\%\) to \(52 + 2.45 = 54.45\%\).
Question 2: Based only on Poll A's interval, can you conclude Candidate X is ahead of a rival polling at 48%?
Answer: No. Poll A's interval (47.1% to 56.9%) overlaps with 48%, so the data doesn't rule out the two candidates being roughly tied — the reported 52% could reflect a true value close to 48% within the margin of error. Poll B's tighter interval (49.55% to 54.45%) does not overlap with 48%, so it provides stronger evidence Candidate X is genuinely ahead.
Notice both polls reported the identical 52%, yet Poll B's quadrupled sample size (400 to 1,600) roughly halved the margin of error — a pattern worth recognizing: you need a dramatically larger sample for a modest gain in precision, so a poll's sample size matters as much as its headline number.
What Margin of Error Does NOT Tell You
A small margin of error only means the estimate is precise — it says nothing about whether the sample was selected fairly. If the sample of 150 students was drawn only from one grade level, or only from students who volunteered, the estimate could be precise and still biased. Precision (small margin of error) and accuracy/lack of bias (good sampling method) are separate ideas, and the SAT likes to test whether you can tell them apart.
Population vs. Sample Language
Watch for precise wording: a parameter describes the whole population (often unknown), while a statistic describes a sample (what you actually calculate). SAT answer choices sometimes swap these terms deliberately — a correct interpretation talks about what the sample statistic estimates about the population parameter, not what it definitively proves.
Frequently asked questions
Do I need to memorize a margin of error formula for the SAT?
No. You need to understand the relationship between sample size, variability, and margin of error conceptually, and be able to compute simple interval bounds by adding and subtracting, as shown above.
Can margin of error ever be zero?
In practice, no — unless you survey the entire population (a census), there's always some uncertainty in a sample-based estimate, though a very large, well-chosen sample can make that margin quite small.
Can two studies have the same margin of error but different levels of trustworthiness?
Yes. Margin of error only measures the precision of an estimate based on sample size and variability — it says nothing about whether the sample itself was randomly and fairly selected. A precise-looking result from a biased sample is still a biased result.
How does this connect to evaluating statistical claims?
Margin of error is one piece of judging whether a claim is well-supported; the broader question of study design is covered in our guide to evaluating statistical claims, including observational studies versus experiments.
To build comfort with this reasoning under timed conditions, work through instant-scored SAT Math sets for this subtopic, and if scoring feels uncertain, our article on setting a realistic SAT Math score goal can help you plan your prep timeline.