Calculate the relative standard error, coefficient of variation, and quality rating for survey estimates. Includes confidence intervals and sample size planning.
The Relative Standard Error (RSE) Calculator computes the RSE — the standard error expressed as a percentage of the estimate — along with confidence intervals, quality ratings, and sample size requirements for target precision levels.
RSE is the primary measure of estimate reliability in survey statistics. It normalizes the standard error against the estimate's magnitude, making it easy to compare precision across estimates of different sizes. Government agencies like the Census Bureau and Bureau of Labor Statistics use RSE thresholds to classify data quality: estimates with RSE under 10% are considered reliable, while those above 25% should be used with extreme caution.
This calculator goes beyond the basic RSE formula by providing confidence intervals at multiple levels, a visual quality gauge with color-coded bands, and a sample size planning table showing how many observations you'd need to achieve different RSE targets. It is useful when you need to judge whether an estimate is publishable and what sample-size increase would materially improve precision.
Survey statisticians and researchers need a standardized way to assess and communicate estimate quality. RSE provides that standard — it's used by the U.S. Census Bureau, Bureau of Labor Statistics, and statistical agencies worldwide to flag unreliable estimates.
This calculator is essential for survey analysts, government statisticians, market researchers, and anyone who publishes estimates with associated standard errors. The sample size planning feature helps design future surveys with adequate precision.
RSE = (Standard Error / |Estimate|) × 100. CV = SE / |Estimate|. 95% CI = Estimate ± 1.96 × SE. Required n ≈ n_current × (RSE_current / RSE_target)².
Result: RSE = 2.67%, CV = 0.0267, Quality = Excellent
RSE = (120 / 4500) × 100 = 2.67%. This is under 5%, so the estimate is rated "Excellent" quality. The 95% CI is 4500 ± 235.2 = (4264.8, 4735.2).
Government statistical agencies publish thousands of estimates annually, each with associated standard errors. The RSE provides a quick, standardized quality metric. The U.S. Census Bureau suppresses estimates with RSE above 50% and flags those above 25% with warnings. The Bureau of Labor Statistics uses similar thresholds for employment data.
One of the most practical applications of RSE is planning future surveys. If your current survey of n=500 yields RSE=12%, and you need RSE≤5%, you'd need approximately n=500×(12/5)²=2,880 observations. This calculation assumes the same design effect and similar variability in the new survey.
In complex survey designs (cluster sampling, stratification), the standard error differs from what simple random sampling would give. The design effect (DEFF) measures this ratio. RSE calculations should use the actual SE accounting for the survey design, not the simple random sampling SE. Most survey software (SAS, Stata, R survey package) computes design-adjusted standard errors automatically.
Standard error is in the same units as the estimate (e.g., dollars). RSE divides by the estimate to create a percentage, making it comparable across different-sized estimates. An SE of $100 on a $1,000 estimate (RSE=10%) is worse than $100 on a $100,000 estimate (RSE=0.1%).
Common standards: RSE ≤ 5% = Excellent, 5-10% = Good, 10-15% = Acceptable, 15-25% = Use with caution, >25% = Do not use for analysis. These are guidelines; specific agencies may use different thresholds.
The primary way is to increase sample size. RSE decreases proportionally to 1/√n, so quadrupling the sample halves the RSE. Better survey design (stratification, optimal allocation) can also reduce RSE without increasing sample size.
RSE is the CV multiplied by 100 (expressed as a percentage). CV = SE/|Estimate| is a unitless ratio; RSE = CV × 100 is a percentage. They convey the same information in different scales.
Halving the RSE requires quadrupling the sample size (because RSE ∝ 1/√n). Achieving very low RSE targets (like 5%) from a current high RSE may require dramatically larger samples, which may not be feasible.
Yes, when the estimate is close to zero. A mean of 0.01 with SE of 0.005 has RSE = 50%, but the absolute precision (±0.01) might be perfectly adequate for the application. Always consider RSE alongside the absolute margin of error.