Learn to inspect samples, distinguish association from causation, and evaluate the choices behind a political score.
3 lessons · approximately 14 minutes · Sources checked 2026-11-18
A percentage describes a numerator divided by a denominator. Before sharing it, ask: what exactly was counted, who could participate, who actually participated, and who was left out? “Eighty percent of people who completed this quiz selected option A” is a statement about those recorded completions. It is not automatically a statement about all Americans, all voters, or everyone who saw the invitation.
An open online quiz is usually a self-selected, nonprobability sample. People’s interest in the topic, where the link appeared, the invitation’s wording, and willingness to finish can affect who appears in the results. A large number of responses does not, on its own, remove this selection problem. Weighting can help under some assumptions, but cannot be assumed to fix every unmeasured difference.
Pew Research Center’s comparisons of online samples illustrate why recruitment method matters and why accuracy must be checked against credible benchmarks. Those studies do not establish the error of a particular OppScore dataset; that requires inspecting its own collection process and data.
Try it: a fictional quiz has 1,000 starters, 400 completions, and 300 selections of A among completers. That is 75 percent of completers, not 75 percent of starters. It also tells you nothing yet about the opinions of the people who never started. Write the denominator into the headline.
Two things can change together without one causing the other. A third factor may influence both, the direction may be reversed, or the association may result from how observations were selected. Correlation describes a pattern. Causation is a stronger claim about what produces a change.
Imagine that people who finish an educational quiz score higher afterward than people who never open it. Perhaps the lesson helped. Perhaps people with more prior knowledge were more likely to finish. Perhaps the groups differed in available time or interest. The comparison alone does not isolate the effect of the lesson.
A more informative learning study can compare equivalent knowledge assessments before and after instruction, include a suitable comparison group, and disclose who dropped out. Random assignment, when appropriate and ethically conducted, can help separate an intervention’s effect from pre-existing group differences. Measurement quality and follow-up still matter. This is a way to study learning, not a method for tailoring political persuasion.
Repeating the same quiz and finding a similar percentage can show stability in that process. It does not prove that the sample represents the public or that the quiz caused someone to change a political belief. Try it: rewrite “the lesson caused everyone to understand” as a narrower statement naming the participants, the test, the comparison, and the uncertainty.
A political score is constructed. Its designers choose which issues to include, which sources to accept, how to code positions, how much weight each item receives, and how to handle missing or conflicting evidence. A documented vote can be a factual input; calling one direction “better” or “pro-opportunity” adds an evaluative framework. The resulting number should be read together with that framework.
Consider a fictional two-question score. Person A agrees with one item and disagrees with the other. If the first item receives nine times the weight of the second, the overall result will look very different from equal weighting. Neither calculation tells you whether the weighting reflects your priorities. A reproducible method lets a reader see exactly what changed.
For a survey result, look for the sponsor, recruitment method, field dates, sample sizes, question wording, processing rules, and limitations. AAPOR’s disclosure standards explain why those details belong beside published findings. Disclosure helps evaluation; it does not certify that a method is unbiased or correct.
For a score, add the source dates, issue definitions, weights, evidence links, version history, and correction process. This checklist is an editorial recommendation, not a claim that OppScore already publishes every item. Try it: label one statement as an observed fact, one as an interpretation, and one as a value judgment. Good information makes those distinctions visible.
Two-question knowledge checks accompany each lesson with JavaScript enabled. Guest practice stays on this device. Connect with BOLD One to save server-graded progress in BOLD Academy. Learning results do not affect political ratings.