Rethinking Census 2021

Activities for the next decennial Census of India, falling due in 2021, will be spooled in no time. While it is likely to be another traditional census, there has been a remarkable transformation in census methods elsewhere in the world, mostly in Europe, within the last two decades or so. Instead of implementing the questionnaire-based method, attempts are being made to use data from various administrative data registers, mostly from government sources, to produce useful statistics.

In 2000, the Austrian government decided that the 2001 census would be the last traditional one involving a high burden for respondents, and a huge cost (€72 million). Consequently a register-based “test census” was conducted in 2006, where the methods, data procedures and use of registers were successfully tested. The 2011 census was the first complete register-based one having no burden for respondents, and the cost declined to €10 million.

Most of the data were already available in several registers like the Integrated Data Bases for persons, families, households, buildings and dwellings, and locations of work. Data was also available from municipalities, geo-information statistical databases and interactive maps. The difficult task of combining all large registers was done by using a special identification number for persons. Certainly, there were occasions where same variables featured in many registers. Also, some variables such as “language mostly spoken” and “religion” were not in any register; these were then collected by suitable sample surveys.

While the Austrian population is less than 10 million, Germany also conducted a nationwide census in 2011 after a 20-year gap. This first register-based census was a multiple-source, mixed-mode method to collect data from administrative registers such as population registers, full enumerations and a sample survey. The structure of the Swiss Census also changed completely in 2010. The information is now primarily drawn from population registers and supplemented by sample surveys of about 5 per cent of the population.

The Nordic countries have a long tradition of using administrative registers for producing official statistics. Countries such as Norway, Finland, Sweden and Denmark now conduct their population censuses using administrative data registers rather than through a nationwide survey of households.

The Netherlands has held virtual censuses since 1981, using the Population Register and surveys. For example, in the 2011 census, they used the registers for population, jobs, self-employment, fiscal administration, social security administration, pensions and life insurance benefits, and housing, supplemented by a labour force survey.

Estonia, perhaps the most advanced digital nation in the world, used a combined census methodology using several registers along with an e-census in 2011, the preparation for which was initiated in 2006. However, Estonia is now moving towards a completely register-based census for 2021. A trial census of the register-based population and housing was conducted in 2016 to test the quality of the registers to form the census characteristics, the functioning of the methodology, and readiness of the support software. The result confirmed that Statistics Estonia, the country’s official statistical agency, is ready for the exercise in 2021. However, a second trial census in 2019 will test the effectiveness of the measures used to address weaknesses detected in the first one.

In 2014, the UK government announced that the country will replace the decennial census beyond 2021 by statistics produced by more regular and timely administrative data, such as council tax records, school admission statistics, tax records or NHS data — at a lower cost and on a more frequent basis. The executive director of the Royal Statistical Society has said the government had “made the right call”.

Greenland in North America and Singapore and Bahrain in Asia are also making effective use of registers for their censuses. Singapore had its first register-based census as early as 2000, and the second one in 2010, where administrative data for basic population count and profile like age, sex, ethnic group, place of birth and type of dwelling were supplemented by sample enumeration for additional topics like marital status and fertility, education, language and literacy.

Is it time to think beyond the traditional questionnaire-based approach in India as well? While the country’s 2011 Census cost about Rs 22 billion, the 2021 Census could cost about Rs 46 billion, assuming $0.50 per person. Billions of rupees can be saved by making use of the administrative data of several available registers instead. 

However, what a country of 1.3 million (Estonia, for example) can easily do is a daunting task for a country of 1.3 billion. But, I believe that we have the statistical and technological expertise to combine several government registers along with tax, hospital and educational records to produce statistics similar to the census, saving lots of money in the process. The loss of information, if any, should be minimal, and can easily be compensated by suitable sample surveys. Let the 2021 exercise be the last traditional census in the history of India.

If India aspires to step forward towards a register-based census system now, it might also be an act of great homage to P C Mahalanobis, India’s “Plan Man”, whose 125th birth anniversary it was earlier this year.

The writer is Professor of Statistics, Indian Statistical Institute, Kolkata.