Margin of Error vs Confidence Interval

One Statement, Two Formats

Because the two are interconvertible, neither is “more correct.” Intervals shine when the endpoints matter (does the range include 50%? does it include zero effect?); margins shine for quick precision comparisons between studies.

Anatomy of the Margin

The two factors separate cleanly. The standard error measures how noisy the estimate is — it belongs to the data. The critical value encodes how confident you asked to be — it belongs to you: 1.645 for 90%, 1.96 for 95%, 2.576 for 99% (z case). Demanding more confidence at the same sample size widens the margin; collecting more data at the same confidence narrows it, at the square-root rate covered in the standard error guide.

Worked Example: A Poll of 1,000 Voters

A poll finds 52% support for a proposal among n = 1,000 respondents. At 95% confidence:

1. Standard error: √(0.52 × 0.48 / 1000) ≈ 0.0158.
2. Margin of error: 1.96 × 0.0158 ≈ 0.031 → ±3.1 percentage points.
3. Confidence interval: 52% ± 3.1 → (48.9%, 55.1%).

The interval format immediately shows the headline that matters: the range dips below 50%, so majority support is not established at 95% confidence, despite the 52% estimate. The margin format shows precision at a glance. Same math, different emphasis — and both reproduce in the margin of error calculator and the confidence interval calculator.

Upgrading to 99% confidence multiplies the same standard error by 2.576 instead: MoE ≈ 4.1 points, interval (47.9%, 56.1%) — more certainty about a vaguer claim.

What the ±3 Points Does and Does Not Cover

• It covers random sampling error only. Question wording, undercoverage, non-response, and late opinion shifts are all outside the margin — real polls miss by more than their margins more often than the pure math predicts.
• It fails at its stated rate by design. At 95% confidence, about 1 poll in 20 misses by more than its margin even with perfect methodology — the long-run reading explained in the confidence intervals guide.
• It applies to each share, not to gaps. A 52–48 lead with a 3-point margin is not “outside the margin”: the lead of 4 points has an effective margin near 6, because the gap moves twice as fast as either share.
• Subgroups have bigger margins. The 200 under-30 respondents inside a 1,000-person poll carry a margin near ±6.9 points, not 3.1 — precision belongs to the subgroup's n, not the survey's.

Planning Backwards From a Target Margin

Study design usually runs the formula in reverse: choose the margin you need, then solve for n. For a proportion at 95% confidence, n = (1.96/MoE)² × p(1 − p), computed at the worst case p = 0.5 when nothing is known. A ±3-point target needs about 1,067 respondents; ±2 points needs about 2,401; ±1 point needs about 9,604 — the square-root economics in action. The sample size calculator performs this inversion, including the finite population correction for small populations.