The Covid-19 pandemic has delayed the houselisting phase of the 2021 Census that should have been completed before the onset of monsoon in most states. Even if houselisting is somehow completed this year, there will be hardly any time to plan the household phase scheduled for February 2021.
In all likelihood the February 2021 Census will have to be rescheduled for February 2022 to ensure comparability with earlier censuses. This will also affect the National Sample Surveys and others that use the census as the sampling frame. The delay can, however, be used to introduce much-needed reforms to this gigantic exercise whose roots go back to the late 19th century.
The need for reforms
While the Indian census is unparalleled in the developing world, in terms of its regularity and coverage over a century and a half, the past four decades have seen a decline in the quality of data and growing delays in its release despite technological innovations. The use of census data in delimitation and federal redistribution has been questioned on grounds of poor quality, while the Covid-19 pandemic revealed the obsolete and poor quality of data on internal migration.
To enable the census to provide timely and good quality inputs to policymakers in a fast-changing world, reforms need to go beyond the introduction of digital technologies and attend to problems such as lack of clarity about the purpose of census, lacunae in training of enumerators, and growing politicisation of statistics. Despite sustained problems, the census has not seen any major reform after 1994 when both the Census Act, 1948 and Census Rules, 1990 were amended. Concern about the quality of census is invariably linked to other issues and fades away with interest in those issues.
The legal foundation of the census has remained largely unchanged since newly independent India enacted a permanent census legislation in 1948. The methodological core – extended de facto (synchronous) canvasser-based enumeration – too has remained intact even though the length and layout of schedules changed quite a bit. The Household Schedule, for instance, grew with the footprint of the state, from 14 questions in 1951 to 29 questions in 2011.
Understand the ‘purpose’ of census
Reforms should begin with the design of schedules based on a clear understanding of two essential functions of the census. First, census facilitates rule-based distribution of power and resources through constitutionally mandated redistribution of taxes, delimitation of electoral constituencies and affirmative action policies. It is also used in routine policy-making across tiers of government. Second, census serves as the sampling frame for surveys and is also the basis of population projections.
Redistribution and delimitation require only the geographical distribution of headcount. Affirmative action requires the distribution of headcount by caste, which is conditional upon religious identity, and tribe. Other routine policies require distribution of the headcount by households, marital status, age, sex, literacy, migrant status, and mother tongue. Put together, these variables are sufficient for choosing representative samples for surveys. Population projections additionally require fertility estimates. Information collected beyond these is superfluous for the census’ core functions.
Further, data collection has not kept pace with improvements in data processing technology due to the lack of motivated and adequately trained enumerators. Given the high salaries of school teachers, the modest honorarium paid for census work does not cover their opportunity cost of conducting door-to-door enumeration. So, both from the perspective of the essential purpose of census as well as workforce constraints, the schedules need to be revisited.
Cut the questions
Nearly half of the ‘Houselisting and Housing Schedule’ of the census is devoted to questions on household amenities and assets. This information is not needed to conduct the household phase of census or sample surveys. These questions can be dropped because the information can be more appropriately collected through sample surveys and administrative statistics. Likewise, a quarter of the questions in the ‘Household Schedule’ probe characteristics of workers, which can be dropped except for, say, the question on occupation. It is unfair to expect inadequately trained enumerators to assess employment when even full-time, trained surveyors armed with more nuanced questionnaires find this challenging. For similar reasons, the issue of disability is better studied using specialised surveys.
Cutting down the length of unwieldy schedules has several advantages. First, it will improve data quality by reducing the workload of enumerators. Second, it will also free up senior census officials and help revive the earlier tradition of producing detailed administrative and other reports crucial for understanding the context of data. Third, shorter schedules will seem less invasive and assure respondents uncomfortable with sharing too many details. Fourth, it will cut down processing time and help in reducing delays in the release of data.
It, however, will not curb political interference at the Union level that has been a key factor behind delays over the past two decades. This problem can be addressed if the Census Rules, 1990 are revised to introduce desirable timelines for the release of data and the Registrar General commits to a calendar of release in advance.
Deal with data manipulation
The proposed changes will streamline census operations and ensure timely release of data but not address problems such as the poor accounting of migrants that distorts estimates of urbanisation as well as inter-state distribution of population, and grassroots manipulation of data driven by political and economic considerations.
Better accounting of seasonal migrants is possible if the instructions for the question on the duration of migration allow recording durations less than one year. Further, enumerators need to be properly trained on how to account for different types of migration under extended de facto (synchronous) enumeration.
Addressing local manipulation will require community outreach along the lines of the 2011 Census of Nagaland to demystify census operations and build trust in the impartiality of the exercise, better scrutiny of electoral records and welfare schemes to weed out bogus beneficiaries, and amendment of the Census Act, 1948 to empower the Registrar General to formally reject manipulated data.
These reforms are essential to ensure that the census exercise is able to fulfil its constitutional, policy and statistical obligations and also clear the ground for debates on the future of census in the digital era.
Vikas Kumar teaches at Azim Premji University, Bengaluru, and is co-author of Numbers in India’s Periphery: The Political Economy of Government Statistics, Cambridge University Press (2020).