Margin of Error: Formula, Calculator & How to Calculate It

What Is the Margin of Error?

When a news organization reports a poll result with "±3 percentage points," that figure is the margin of error. It does not mean the poll is wrong by 3 points. It means that if you repeated the same poll many times under identical conditions, 95% of the resulting confidence intervals would contain the true population value. The margin of error is the half-width of that interval.

One constraint worth stating early: the margin of error only accounts for random sampling error. It says nothing about non-sampling errors — question wording bias, non-response bias, undercoverage, or measurement error. A poorly designed survey with n = 10,000 can have a tiny MOE and still produce a deeply misleading estimate. This limitation is frequently overlooked when poll numbers are reported in the press.

Sampling and Polls: A Brief History

The modern polling margin of error traces directly to the work of Jerzy Neyman, whose 1934 paper on stratified sampling established the mathematical framework for confidence intervals that underpins survey reporting today. George Gallup's successful prediction of the 1936 U.S. presidential election — using a sample of roughly 50,000 against the Literary Digest's flawed poll of 2.4 million — made it clear that sample design mattered far more than raw sample size. The margin of error became the standard summary of that design's precision.

Today the same principles apply to study design in clinical trials, quality control in manufacturing, and A/B testing in software products. The underlying math is described in detail in the sampling distributions section on Statistics Fundamentals.

The Three Variables That Control MOE

These three relationships are not independent. Increasing the confidence level from 95% to 99% requires a larger critical value (from z* = 1.960 to z* = 2.576), which directly increases the MOE. Recovering that precision requires a larger sample. This trade-off is exactly what sample size calculators are built to solve.

How to Calculate Margin of Error: Step-by-Step Formulas

There are two standard versions of the margin of error formula, one for estimating a population mean and one for estimating a population proportion. The choice depends on what you measured in your sample.

Formula for a Population Mean (μ)

When the population standard deviation σ is unknown — which is the usual case — substitute the sample standard deviation s and use the t-distribution critical value t* instead of z*. The degrees of freedom are df = n − 1. This matters most when n is small (under 30). For large samples, the t and z critical values converge and the distinction becomes negligible. See the full confidence interval for a mean guide for the t-distribution version.

Formula for a Population Proportion (p)

When p̂ is unknown in advance — for example, before a poll is conducted — use p̂ = 0.5. This produces the most conservative (largest) possible margin of error, because the expression p̂(1 − p̂) reaches its maximum at p̂ = 0.5. Sample size tables in research proposals almost always assume p̂ = 0.5 for this reason.

Live Margin of Error Calculator

Enter your study parameters below. The calculator handles both the proportion formula (most common for surveys) and the mean formula. Select the calculation type first to see the relevant inputs.

Worked Examples: Three Real-World Scenarios

Case Study 1 — Political Polling at 95% Confidence

Case Study 2 — Clinical Trial Efficacy at 99% Confidence

Case Study 3 — Manufacturing Quality Control at 90% Confidence

Quick-Reference Calculation Matrices

Critical Value Cheat Sheet

Sample Size Impact on MOE (95% CL, p̂ = 0.5)

This table demonstrates the diminishing returns of increasing sample size. Going from n = 100 to n = 1,000 (ten times larger) reduces MOE from 9.8% to 3.1% — not by a factor of ten, but by roughly two-thirds. That square-root relationship means you need 100× the sample to get 10× the precision. The sample size calculator makes it straightforward to solve for n given a target MOE.

Statistical Comparisons: MOE vs. Related Concepts

Margin of Error vs. Standard Error

These two terms are frequently confused because both measure variability in sample statistics. The distinction is precise and worth getting right.

The standard error is a property of the statistic and the sample design. The margin of error adds the decision about how confident you want to be, converting SE into the language of interval estimation. For more on SE and its role in inference, see the guide to sampling distributions of sample proportions.

Margin of Error vs. Confidence Interval

The confidence interval tells you where the truth probably lives. The margin of error tells you how wide that region is. You need both to interpret a survey result fully. See the confidence interval for a proportion and confidence interval for a mean pages for the full derivation of each.

Where Margin of Error Appears in Practice

Python and R Implementation

Python (scipy.stats)

import scipy.stats as stats
import math

# Margin of error for a proportion
def moe_proportion(p_hat, n, confidence=0.95):
    z = stats.norm.ppf(1 - (1 - confidence) / 2)
    se = math.sqrt((p_hat * (1 - p_hat)) / n)
    moe = z * se
    return {
        "moe": round(moe, 4),
        "lower": round(p_hat - moe, 4),
        "upper": round(p_hat + moe, 4),
        "z_star": round(z, 3)
    }

# Margin of error for a mean (known σ)
def moe_mean(sigma, n, x_bar=None, confidence=0.95):
    z = stats.norm.ppf(1 - (1 - confidence) / 2)
    se = sigma / math.sqrt(n)
    moe = z * se
    result = {"moe": round(moe, 4), "z_star": round(z, 3)}
    if x_bar is not None:
        result["lower"] = round(x_bar - moe, 4)
        result["upper"] = round(x_bar + moe, 4)
    return result

# Examples
print(moe_proportion(0.54, 1200, confidence=0.95))
# {'moe': 0.0282, 'lower': 0.5118, 'upper': 0.5682, 'z_star': 1.96}

print(moe_mean(6.5, 80, x_bar=12.4, confidence=0.99))
# {'moe': 1.8727, 'z_star': 2.576, 'lower': 10.5273, 'upper': 14.2727}

R Script

# Margin of error for a proportion
moe_proportion <- function(p_hat, n, confidence = 0.95) {
  z <- qnorm(1 - (1 - confidence) / 2)
  se <- sqrt((p_hat * (1 - p_hat)) / n)
  moe <- z * se
  list(moe = round(moe, 4),
       lower = round(p_hat - moe, 4),
       upper = round(p_hat + moe, 4),
       z_star = round(z, 3))
}

# Margin of error for a mean (known sigma)
moe_mean <- function(sigma, n, x_bar = NULL, confidence = 0.95) {
  z <- qnorm(1 - (1 - confidence) / 2)
  se <- sigma / sqrt(n)
  moe <- z * se
  result <- list(moe = round(moe, 4), z_star = round(z, 3))
  if (!is.null(x_bar)) {
    result$lower <- round(x_bar - moe, 4)
    result$upper <- round(x_bar + moe, 4)
  }
  result
}

# Examples
moe_proportion(0.54, 1200, confidence = 0.95)
moe_mean(6.5, 80, x_bar = 12.4, confidence = 0.99)

Entity and Formula Glossary

Frequently Asked Questions