A two-percentage-point poll lead means the survey's estimate favors one candidate by two points. It does not establish that the candidate will win or even that the population lead is clearly different from zero.
Interpretation depends on the survey design, uncertainty, electorate, and interview dates.
Read the two estimates
Imagine a hypothetical poll with Candidate A at 48% and Candidate B at 46%. The difference is two percentage points.
That is not the same as saying A has a 48% chance of winning. It is also not a count showing that A has received more valid ballots.
If the release reports an uncertainty measure for each candidate, the uncertainty for their difference needs separate consideration. Source: Pew's margin-of-error explainer.
Why small leads can be fragile
A narrow estimate can be affected by the sample drawn, weighting choices, who participates, and how likely voters are defined.
Future voter decisions also remain open. A survey measures a period before voting is completed, even if many respondents already intend to participate.
The phrase “statistical tie” is sometimes used informally. A clearer description is that the poll does not establish a clear lead under its stated statistical assumptions.
What would strengthen the evidence?
Look for comparable independent surveys, a sustained pattern, transparent methods, and credible information about the electorate.
Even then, avoid converting a recurring small lead into a guarantee. Several pollsters can share errors, and a close contest is especially sensitive to modest mismeasurement.
What a responsible headline says
“Candidate A holds a small estimated lead in this poll” identifies the scope. “Candidate A will win” goes beyond the evidence.
If a forecast supplies a probability, explain that it comes from a model rather than from subtracting the two vote estimates.
The practical lesson
Use narrow leads to identify races worth watching, not to close the question. Keep the poll's source and dates visible and revisit the evidence as new releases arrive.
Related reading: margin of error and poll averages.