Relative standard error

Indian Economy glossary

Also called: RSE · Topic: Economic Data: Census, NSS, Surveys and Statistical Tools · NCERT: Beyond NCERT

Meaning

Relative standard error (RSE) tells you how reliable a survey estimate is. It starts from the standard error, which measures how much an estimate would change if the survey were repeated with a different random sample. RSE then shows that standard error as a percentage of the estimate itself:

RSE = (Standard error of the estimate ÷ Estimate) × 100

A low RSE means the estimate is precise. A high RSE means it could be far from the true value, so it should be used with care. RSE only covers sampling error. It cannot detect non-sampling problems such as refusals, recall lapses or wrong recording.

Example

MoSPI publishes RSE tables in the "reliability of the estimates" section of the PLFS quarterly bulletins. Suppose a monthly bulletin estimates the unemployment rate at 5.0% with a standard error of 0.2. The RSE is (0.2 ÷ 5.0) × 100 = 4%. Each monthly estimate uses only that month's interviews, so its RSE is usually higher than a quarterly one. A month-on-month change smaller than this margin may not be a real change at all.

Don't confuse with

  • Standard error: This is stated in the units of the estimate, for example percentage points. RSE is a percentage of the estimate, so it has no units. That lets you compare reliability across estimates of different sizes.
  • Sampling error: This is the actual gap between the population parameter and the sample estimate, and it is usually unknown. RSE is a measure of how big that gap is likely to be.

Read more