A sample estimate usually differs from the unknown population value. A confidence interval expresses that sampling uncertainty by giving a range of plausible values, often written as an estimate plus or minus a margin of error.
Read the center and the margin separately
A survey reports that 54% of adults support a proposal, with a margin of error of 4 percentage points. The interval is 50% to 58%:
The margin describes the distance from the estimate to either endpoint, not the total interval width. The width here is 8 percentage points.
Percentage points differ from percent change. Adding 4 percentage points to 54% gives 58%. Increasing 54% by 4% would multiply by 1.04 and give 56.16%, which is a different operation.
What a confidence level means
A 95% confidence procedure is designed so that, across many repeated random samples using the same method, about 95% of the resulting intervals contain the fixed population parameter.
Once an interval is calculated, the parameter is still fixed. The confidence level describes the reliability of the method, not the percentage of individual people inside the interval. Nor does it mean that 95% of sample answers were accurate.
In ordinary reporting, “We are 95% confident that the population proportion is between 50% and 58%” is an appropriate summary when the sampling design and interval conditions are justified.
Example: compare claims with the interval
If a candidate's estimated support is 52% with a margin of error of 3 percentage points, the interval is 49% to 55%. It includes values below and above 50%, so the survey does not establish a clear majority at that confidence level.
If two groups have reported intervals, simple overlap is not a complete formal test of their difference. Use a method designed for the difference when a formal conclusion is required. At the basic interpretation level, avoid claiming a certain lead from point estimates alone.
The interval from 49% to 55% contains 50%, so the estimate does not establish a clear majority.
Describe this figure
- 49%: 49
- Estimate: 52%: 52
- 55%: 55
How sample size changes precision
For comparable random-sample designs at a fixed confidence level, margin of error is approximately proportional to Quadrupling the sample size approximately halves the margin. Multiplying the size by 9 approximately divides the margin by 3.
Thus a margin of 6 percentage points from 400 responses would be about 3 points from 1,600 comparable responses. Doubling the sample size does not halve the margin; the square root controls the improvement.
Higher confidence requires a wider interval for the same data. You gain a more reliable capture method by accepting a less precise range.
Uncertainty is not the same as bias
A stated margin of error usually addresses sampling variability under a particular method. It does not automatically cover misleading wording, undercoverage, nonresponse, or a voluntary-response sample. A narrow interval from a biased design can still miss the intended population value badly.
A confidence interval also does not predict the range of individual observations. An interval for a mean estimates one population average. Individual values can vary much more widely around that average, so a narrow mean interval is compatible with a highly variable population.
Try it yourself
A report gives an interval from 37% to 45%. Find the point estimate and margin of error. Approximately how must the sample size change to reduce that margin from 4 points to 2 points?
Answer: The midpoint is 41% and the margin is 4 percentage points. Halving the margin requires about four times the sample size, assuming comparable conditions and confidence.
What to remember
Interpret the interval, confidence level, and target population together. More data can reduce random uncertainty, but they do not automatically remove systematic bias.