Errors and data agencies

Data can contain two kinds of error.

A sampling error is the difference between the true value for the whole population (the parameter) and the estimate from a sample. For example, five workers earn ₹300, ₹400, ₹500, ₹600 and ₹700; their average is ₹500. If a sample of two (₹300 and ₹500) is taken, its average is ₹400, so the sampling error is ₹500 – ₹400 = ₹100. A larger sample comes closer to the whole population, so it reduces the sampling error.

Non-sampling errors are more serious, because a bigger sample does not remove them. They include:

• Sampling bias: the sample does not include some parts of the population.
• Non-response: some people chosen do not reply.
• Errors in data acquisition: mistakes in recording or copying, such as writing 56 instead of 65, or wrong answers given by respondents.

Even a full census can contain non-sampling errors.

India has several agencies that collect data nationally:

• The Census of India, held every ten years since 1881. The first Census after Independence was in 1951.
• The National Sample Survey (NSS), which carries out surveys in continuous rounds on different topics, such as health care (the 60th round, 2004) and consumer spending (the 68th round, 2011-12). It publishes its results in its journal, Sarvekshana.
• The Central Statistical Organisation (CSO), the Registrar General of India (RGI), the Directorate General of Commercial Intelligence and Statistics (DGCIS) and the Labour Bureau.

(In 2019, the CSO and the NSS Office were merged into the National Statistical Office.)