When can you claim cause and effect on SAT study design questions?

Only when subjects were randomly assigned to groups. An observational study, where groups chose their own behavior, can show an association but not causation, because the groups might differ in many other ways. Random assignment is what makes an experiment an experiment, and the trap answers on the SAT claim causation from an observational study.

What is margin of error on the SAT?

Margin of error is the wiggle room around a survey estimate. If 62 percent of a random sample supports something with a margin of error of 4 percent, the plausible range for the population is 58 to 66 percent. The margin of error belongs to the population estimate, not to any individual, and a larger random sample shrinks it. The SAT asks what a margin of error means, not how to calculate it from a formula.

SAT Inference and Study Design

Q: What do SAT inference and study design questions test?

Whether you can match a conclusion to the evidence a study provides — and refuse to go past it. These questions cover two of the seven official testing points in the Problem-Solving and Data Analysis domain, which supplies 5 to 7 of the 44 math questions: inference from sample statistics and margin of error, and evaluating statistical claims from observational studies and experiments. There is almost nothing to compute; the answer choices are four conclusions, and three of them claim more than the study can back up.

Last updated: July 19, 2026

Inference and study design questions on the digital SAT hand you a study — a survey, a sample, an experiment — and ask which conclusion the design can support. There is almost nothing to compute. Two facts decide every answer: whether the sample was selected at random, which decides who the results apply to, and whether subjects were randomly assigned to groups, which decides whether you can claim cause.

What does this skill actually test?

Whether you can match a conclusion to the evidence — and refuse to go one inch past it.

This post covers two of the seven official testing points in the Problem-Solving and Data Analysis domain, which supplies 5–7 of the 44 math questions: inference from sample statistics and margin of error, and evaluating statistical claims from observational studies and experiments. On paper that sounds heavy. In practice these are reading questions in math clothing. The answer choices are four conclusions, and three of them claim more than the study can back up.

What can a sample tell you?

If a sample is selected at random, whatever you measure in it is a reasonable estimate for the population it was drawn from. And only that population.

A random sample of 200 students from one high school tells you about that high school. Not the district, not the state, not "teenagers." The moment an answer choice stretches the conclusion to a bigger group than the one sampled, it's wrong — no matter how sensible it sounds.

And if the sample wasn't random — volunteers who chose to respond, or people surveyed outside a gym about exercise — it doesn't generalize at all. "Random" isn't decoration in these problems. It's the load-bearing word.

When can you claim cause?

Only when subjects were randomly assigned to groups.

That's the entire test. A study that watches two groups who chose their own behavior — an observational study — can show an association: people who do X tend to have Y. It cannot show that X causes Y, because the groups might differ in a hundred other ways. Random assignment is what scrambles those differences away, and it's what makes an experiment an experiment.

So when a question asks for the best conclusion, check two things in order:

Random selection? If yes, results generalize to the population sampled. If no, they stay with the participants.
Random assignment? If yes, cause-and-effect is on the table. If no, the strongest available word is "associated."

The trap answers claim causation from an observational study. Every time.

What is margin of error?

A survey estimate is a best guess with wiggle room, and margin of error is the wiggle room. If 62% of a random sample supports something with a margin of error of 4%, the plausible range for the population is 58% to 66%. That's the whole move: estimate minus margin, estimate plus margin.

Two things the SAT checks:

The margin of error belongs to the population estimate, not to any individual person in it.
A larger random sample shrinks the margin of error. If a question asks how to make the estimate more precise, the answer is more people, sampled the same random way.

You will not be asked to calculate a margin of error from a formula. You'll be asked what it means. That's a different skill, and it's the one students don't practice.

Practice routine

Work 8–10 of these and make the design call before reading the choices:

Underline who was sampled and how, because the sampling method fixes the largest group any conclusion can mention
Label the study "assigned" or "observed" before judging any answer, so a causal claim from an observed study is dead on arrival
For margin of error questions, write the interval — estimate ± margin — before looking at the options, so the right range is in your hand first

These questions live in the same Problem-Solving and Data Analysis group as percentages and probability, but they fail differently: not from bad arithmetic, from overclaiming. HIROSCORE tracks this skill separately, so you know whether your losses come from reading study design wrong or from somewhere else entirely. The GPS for your SAT score.

If you'd rather stop guessing which skill is leaking points, join the HIROSCORE beta and help us build it.

References