46 | Appendix - Overview of Census Operations 1981-2011

Curated by Shivakumar Jolad and Gaurav Kalyani

Published on: 09 November 2026

This appendix documents the exhaustive summary of how the four decennial censuses of 1981, 1991, 2001 and 2011 were planned, conducted, processed and evaluated. It draws on A Treatise on Indian Censuses Since 1981, published in 2023 by the Office of the Registrar General & Census Commissioner, India (ORG&CCI).

The Process of Census-Taking, 1981–2011

Population characteristics

Conducting a population census in India is an exceptionally demanding administrative and statistical task because of the country's geographical spread, the size of its population, uneven literacy and the scattered pattern of urbanisation. Between 1981 and 2011 the population almost doubled, from 683 million to 1,210 million. Literacy rose from 43.6 per cent to 73.0 per cent and the urban share from 23.1 per cent to 31.1 per cent (Table 1).

Table 1: Population characteristics of India, 1981–2011

¹ Literacy rate = (literate persons aged 7 years and above ÷ population aged 7 years and above) × 100. The 1981 and 1991 totals include estimates for Assam (1981) and Jammu & Kashmir (1991), where the census could not be held; the enumerated populations were about 665 million and 839 million respectively. Source: Treatise, Table 1.1.

Geographies covered

Census data must be compiled and released for a fixed hierarchy of units: State, District, Sub-District, Village and Town. To organise fieldwork across this hierarchy, ORG&CCI uses two operational units. Charges are formed at the sub-district level in both rural and urban areas, while Enumeration Blocks (EBs) are compact clusters of households within villages and towns, each sized to be covered by one enumerator. Table 2 shows how the number of jurisdictional units grew over the period.

Table 2: Number of jurisdictional units in India, 1981–2011

*Excluding Assam. # Excluding Jammu & Kashmir. NA: not available. Source: Treatise, Table 1.2.

Census functionaries and their honorarium

The fieldwork is carried out by enumerators and supervisors drawn mainly from State education departments, most of them local school teachers. Besides visiting every household, the enumeration task involves training these functionaries, distributing schedules and materials to them, and collecting the completed records. Their number rose from about 12.5 lakh in 1981 to 27 lakh in 2011; 29.9 lakh were approved for Census 2021 (Table 3).

Although enumerators and supervisors are salaried State government employees, they receive an honorarium for census work done in addition to their regular duties. The amount has varied considerably from census to census (Table 4).

Table 3: Number of enumerators and supervisors, 1981–2021

Source: Treatise, Table 1.3. The census-specific chapters of the Treatise give different approximate figures for 1981 (about 17 lakh) and 1991 (16 lakh); the series above is used for comparison across years.

Table 4: Honorarium paid to enumerators and supervisors (Rs.), 1981–2021

² The 2011 amount includes the honorarium for collecting data for the National Population Register. The 2021 rates were approved for data collection on mobile phones; for paper-mode collection they were Rs. 5,500 (HHC) and Rs. 8,250 (PE), with a separate provision for NPR work. Source: Treatise, Table 1.4.

Questionnaires and census circulars

The number of questions asked of each household affects both the quality of the data collected and their later use. A Data Users' Conference, at which ministries, planners, academics and research bodies comment on the draft schedules, has been a hallmark of every Indian census. The census organisation avoids abrupt changes so that data remain comparable over time, but questions have been added and dropped to meet changing needs (Table 5). Instructions for each stage of the operation, from freezing jurisdictions to conducting post-enumeration surveys, are issued to State governments and Directorates of Census Operations (DCOs) as numbered Census Circulars (Table 6). Our Questions Explorer delves deeper into the evolution of census questionnaire

Table 5: Number of questions canvassed, 1981–2021

³ The 2021 HHC questions were notified in the Gazette of India; the PE questions were approved but not notified before the census was halted by the Covid-19 pandemic. Source: Treatise, Table 1.5.

Table 6: Number of circulars issued by ORG&CCI, 1981–2021

Source: Treatise, Table 1.6.

Census reference date and method of enumeration

A country's population is declared as of a fixed reference point, the census moment. Up to 1971 different dates were used; since 1981 the reference date has been 1 March. The reference time changed later. Until 1991 the census moment was the sunrise of 1 March, and from 2001 it has been 00:00 hours of 1 March.

Two concepts describe how people were counted.

Under the de facto method, everyone physically present in an area at the reference point is counted, whatever their legal residence or citizenship.

Under the de jure method, only usual or legal residents are counted.

India also uses a variant, the extended de facto method, in which people are enumerated where they are found over the enumeration period, and the count is then brought up to the census moment through a revisional round that records births, deaths and new arrivals.

A second distinction is between synchronous and non-synchronous censuses. A synchronous census is taken across the whole country within the same short period; a non-synchronous one is taken at different times in different areas. The 1872 census was de facto but non-synchronous. Every census since has been synchronous, and since 1941 the method has been both extended de facto and synchronous (Table 7). Areas that are snow-bound in February are the standing exception, and are enumerated earlier on a non-synchronous basis.

Table 7: Reference date and method of census-taking, 1872–2011

Source: Treatise, Table 1.7.

Technology and data processing

Changes in processing technology determined how much of the collected data could be used and how quickly. Data were once digitised by hand tabulation, punching and keyed entry, and until 1991 no more than 45 per cent of the data collected could be digitised. The use of Intelligent Character Recognition (ICR) in 2001 and 2011 made it possible to digitise all of it. Census 2021 was planned to use a mobile application on enumerators' own phones, with a paper option, and the approach was tested in the 2019 pre-test (Table 8).

Table 8: Technology used and time taken for data processing, 1961–2021

⁴ Mobile applications were developed for Census 2021 alongside the paper mode. Source: Treatise, Table 1.8.

Mapping, training and publicity

Mapping. Maps are used before, during and after the census. At the outset, Tahsil, District and State maps are redrawn from the previous census's jurisdictions and reconciled with lists obtained from State governments, which can only be done once complete lists of administrative units have been received. During the census, every enumerator must draw a layout map of the Enumeration Block before collecting data; these EB maps are digitised afterwards. Once the data are final, a range of thematic maps is published, and the Census of India is regarded as the largest producer of thematic maps in the country.

Training. Training is at the core of census-taking. Enumerators and supervisors assemble at training venues in their own jurisdictions, and trainers from the census organisation, State governments and non-government organisations have been used to give uniform instruction across India. A hierarchy of trainers operates at national, state, district and charge levels. Training aids have changed over time, from blackboards, flip-charts and reading from manuals to projectors, PowerPoint presentations and audio-visual material, and classes have been held under trees, in cinema halls and in the open.

Publicity. Publicity through traditional media and street performances has built public awareness of why the census matters, and the trust and cooperation that complete coverage depends on.

Data tables published

Once fieldwork is complete and the data have been scrutinised, tables are published for data users. Their type, range and level of detail are decided by a Technical Advisory Committee of leading data users, academics and former Registrars General. Census 2011 produced 293 tables, the largest number in the period (Table 9).

Table 9: Number of data tables published, 1981–2011

Source: Treatise, Table 1.9.

Post Enumeration Surveys

A Post Enumeration Survey (PES), known as the Post Enumeration Check (PEC) until 1991, is an independent sample survey conducted after enumeration to measure the accuracy of the census. It estimates coverage error, the omission or duplication of households and persons, and content error, the misreporting of characteristics. The net omission rate it produces is a key indicator of field quality. PES results have never been used to adjust the census count (Table 10).

Table 10: Net omission rate, India (per 1,000 persons), 1981–2011

Source: Treatise, Table 1.10.

Census of India 1981

Scale and coverage

The 1981 Census was the 12th census in India since 1872 and the fourth after Independence. It was conducted in two phases: houselisting, held at different times during 1980 (largely in April–May), and population enumeration, from 9 to 28 February 1981. Some 665 million persons were enumerated across 30 States and Union Territories and 402 districts, in more than 5,60,000 inhabited villages and towns. About 23 per cent of the population was urban and the literacy rate was about 43 per cent. The census could not be conducted in Assam because of the disturbed conditions prevailing in the state at the time.

Preparation

Preparations began at the national level around 1976. At the State level, work started in January 1978 with two letters from the Registrar General: one on planning, jurisdictional changes and the updating of maps and village lists, and a demi-official letter asking Chief Secretaries to support the census. Most Directorates set up a Census Cell to coordinate with State departments. Village lists supplied by Tahsildars were reconciled with the 1971 lists, and the corrected tahsil maps were returned to the Tahsildars for certification. The Registrar General issued 35 circulars in all to guide the DCOs.

The questionnaires were developed through two Data Users' Conferences and three rounds of field testing:

• First Data Users' Conference (February 1978). Representatives of Central and State governments, the Planning Commission, the Central Statistical Organisation (CSO) and research bodies discussed the concepts and questionnaires.

• Pilot study (12–18 June 1978). Held in nine States (Andhra Pradesh, Gujarat, Haryana, Jammu & Kashmir, Karnataka, Kerala, Rajasthan, Uttar Pradesh and West Bengal), mainly to test whether the new economic questions on the draft Individual Slip could be canvassed, and how comparable their results were with 1971.

• First pre-test (September–October 1978). A full-drill exercise testing five schedules: the Houselist, the Household Establishment Schedule, the Universal Individual Slip, the Sample Individual Slip and the Degree Holders and Technical Personnel Survey Schedule.

• Second Data Users' Conference (21–23 December 1978, New Delhi). Finalised the questionnaires in the light of the two field surveys.

• Second pre-test (June–July 1979). Unlike the first, it used State government staff as enumerators, introduced a Household Schedule formed by moving items from the Houselist and merging the population record into it, canvassed schedules in regional languages, and used the enumeration block rather than the household as the sample unit. Three conferences of Directors of Census Operations then settled the operational plan:

• August 1979, New Delhi (India International Centre, 24–29 August). Fixed the hierarchy of census officers, retained the 1971 urban criteria, adopted the two-phase design, set the reference date and enumeration period, and decided to integrate the Economic Census with houselisting.

• January 1980, Mysore (16–19 January). Reviewed arrangements for houselisting and finalised instructions for the Abridged Houselist, Household Schedule and Individual Slip.

• November 1980, New Delhi (Vigyan Bhavan). Fixed workload norms for enumeration blocks (750 persons in rural areas and 600–650 in urban areas, with a 10 per cent reserve of staff), the reporting of provisional totals, the plan for the Primary Census Abstract, the Post Enumeration Check and the tabulation plan.

Changes in questionnaire and method

Compared with 1971, the 1981 Census introduced several changes intended to widen the data collected and to speed up processing:

Integration with the Economic Census. For the first time, houselisting was combined with the Economic Census of the CSO. Enumerators canvassed an Enterprise List alongside the Houselist in May 1980.

Physically handicapped persons. Data on three categories (the totally blind, the totally crippled and the totally dumb) were collected during houselisting at the request of the Social Welfare Department, the first such enquiry since 1931.

Housing questions moved to enumeration. The Houselist was reduced to 14 columns. Questions on housing materials, amenities and the SC/ST status of the head of household were shifted to a separate Household Schedule canvassed during enumeration, so that amenities could be cross-classified by population characteristics.

Abridged Houselist (AHL). Because of the long interval between the two phases, an Abridged Houselist was prepared for each enumeration block to update the frame before enumeration and to serve as a frame for sample surveys.

Main and marginal workers. The economic questions distinguished main work (work for the major part of the year) from marginal work, and asked non-workers and marginal workers whether they were seeking or available for work.

Universal and sample slips. A Universal Individual Slip was canvassed for everyone. A Sample Individual Slip, which carried the migration and fertility questions, was canvassed in a 20 per cent sample of enumeration blocks in 12 larger States, and for the whole population in the smaller States and all Union Territories. The two were printed in different colours.

Direct data entry and Statistical Quality Control. Individual Slip pads were not broken up: data were keyed from the bound pads directly onto magnetic tape. Part I of the Household Schedule was processed fully by computer. Statistical Quality Control was introduced for the first time in census history, limited to the coding of the economic questions and carried out by independent verification at the headquarters of the major States.

Field operations

Table 11: Calendar of the 1981 Census

Source: Treatise, Chapter 4.

The 'sea population', meaning those aboard Indian naval ships, Indian merchant vessels and ocean liners arriving from abroad, was enumerated on 28 February and 1 March 1981. Strictly military areas were enumerated by Military Census Officers, who were trained alongside civilian officers and then trained the enumerators for their areas.

Training of enumerators and supervisors followed a cascade from district officers to charge officers to field staff, in four rounds between the second half of November 1980 and January 1981, ending with practical field exercises. Publicity used posters, articles in local newspapers, messages on All India Radio, Doordarshan broadcasts and a documentary film, To Meet Your Needs, shown in cinemas from late November 1980 in English and regional languages.

Printing was itself a large undertaking. At least 71 crore Individual Slips were needed (one for every person, plus reserves) in addition to houselists, abstracts, household schedules and instruction books. The Registrar General's office procured paper and printed the English forms centrally; the DCOs translated and printed the regional-language versions.

Provisional results, processing and evaluation

• Provisional totals. Charge officers compiled totals from enumerators' abstracts and sent them to the District Census Officers and DCOs by express telegram, telephone or police wireless. Directors were to report to the Registrar General by 12 March 1981. National figures were to be released by 17 March (Paper 1 of 1981) and State and UT figures by 23 March (Supplement to Paper 1)

  
  

• Tabulation. The Primary Census Abstract (PCA) was to be compiled within ten months, starting in January 1982, from the population record in Part II of the Household Schedule. National estimates were first produced from a 5 per cent sample of blocks, followed by detailed tables from the 20 per cent sample. The General Population Tables and the PCA were tabulated manually; all other series were processed by computer.

  
  

• Post Enumeration Check and Census Evaluation Study. Both were conducted in 15 States and the Union Territory of Delhi by staff independent of the census fieldwork, with fieldwork to be completed by 15 April 1981. The PEC measured coverage, while the Census Evaluation Study used Sample Registration System blocks to examine the omission of children under five and the accuracy of their reported ages.

  
  

Challenges and learnings

House numbering. Despite detailed instructions, numbering was poor in many areas and especially confused in towns. Methods were not uniform; government and public-sector colonies used their own systems unrelated to municipal numbers; and rapid urban growth with ad hoc numbering of new houses had disrupted existing systems.

Layout maps. Many enumerators found it hard to prepare the second notional map showing landmarks and the location of every census house, and could not distinguish a building from a census house.

Enterprise List. Most enumerators found the Enterprise List and its abstract difficult because of its many codes, and totals were often struck wrongly.

Abridged Houselist. The honorarium for the AHL was too low. Copying about 1,000 names from voters' lists earned Rs. 20, while writing an AHL covering about 6,000 persons paid only half that. Charge officers struggled to find willing writers, and AHLs in large cities were often defective because of bad house numbering.

Supply of schedules. Late, heavy demands from charge officers had to be met from State reserves by special messenger, yet large quantities of unused forms were returned after enumeration, showing over-indenting. The Treatise traces this, as in 1971, to instructions not being followed at charge level, where the Tahsildar's many other duties left little time for census work.

Lessons from the pre-tests. Enumerators struggled to separate main from marginal work. Many women who did agricultural work alongside household chores were returned as engaged in 'household duties', raising concern that female work participation would be under-recorded. Enumerating one household took about 30 minutes, and the large schedules and 100-leaf pads were cumbersome to carry from door to door.

Census of India 1991

Scale and coverage

The 1991 Census was the 13th census since 1872 and the fifth after Independence. Houselisting was conducted between 1 April and 30 September 1990, and population enumeration from 9 to 28 February 1991. It counted over 839 million persons, excluding Jammu & Kashmir, where the census could not be held. About 25.5 per cent of the population was urban and the literacy rate was about 52 per cent. Assam, left out in 1981, was enumerated again. The 1991 Census was the first to use computers on a large scale for data processing.

Preparation

Planning began nearly three years before enumeration and followed a month-by-month Calendar of Operations. The main preparatory steps were:

• First Data Users' Conference (18–20 April 1988). Attended by Central ministries, the Planning Commission, State governments, universities, demographic institutes and individual experts.

  
  

• Advisory Committee on Technical Issues (appointed 31 August 1988). Vetted the questionnaires after each round of testing.

  
  

• Pre-tests. The first pre-test, in November–December 1988, used census staff. The second, in June–July 1989, used local school teachers and government officials of the kind who would do the actual census work. The questionnaires (the Houselist, the Household Schedule and the Individual Slip) were finalised in November 1989.

  
  

• Freezing of boundaries. State governments were asked to make any jurisdictional changes before October 1989, and then directed that no administrative boundaries be changed between 1 January 1990 and 30 June 1991.

  
  

• First DCO Conference (Delhi, 6–9 November 1989). Reviewed preparations across the agenda, from preparatory steps to the Post Enumeration Check.

Changes in questionnaire and method

Expanded Houselist. The Houselist grew from 14 columns in 1981 to 24 in 1991. The questions that had been moved to the Household Schedule in 1981 (materials of wall and roof, SC/ST status of the head, number of living rooms, tenure, and whether the household cultivated land) were returned to the Houselist. Further items were added on the predominant material of the floor and the source of drinking water. For the first time, columns were added for the sex of the head of household and the type of fuel used for cooking, and the availability of toilet facilities in rural households was also asked for the first time. Primary census data were also generated at the Community Development Block level, which had replaced the tahsil as the planning unit.

Women's work. The under-counting of women's economic activity had long been a concern. With support from UNIFEM (the United Nations Development Fund for Women) and the Research Centre for Women's Studies at SNDT University, training stressed probing for unpaid family labour, and a UNIFEM poster depicting the range of women's work was distributed across the country before enumeration. The Treatise credits these efforts with improving the recording of women's unpaid work.

Audio-visual training. For the first time, training used an overhead projector, with the instructions condensed into a set of 69 A4-size transparencies so that complex concepts could be explained clearly and uniformly at the divisional, district and charge levels.

Charge Registers. In 1981 a single Charge Register had been updated across both phases. In 1991 a Preliminary Charge Register was used for houselisting, and a Final Charge Register was prepared afresh from the houselisting data for enumeration, giving a cleaner frame of blocks.

Urban frame. The 1981 urban criteria were retained: statutory towns, together with census towns having a population of at least 5,000, at least 75 per cent of the male working population in non-agricultural work, and a density of at least 400 persons per sq. km. One particular problem arose in Uttar Pradesh over the Special Area Development Authorities (SADAs) created under the Nagar Tatha Gram Nivesh Adhiniyam, 1973. These were statutory bodies with municipal powers, but most were rural in character. Only those SADA areas that met the census-town criteria were treated as urban.

Field operations and logistics

Table 12: Calendar of the 1991 Census

Source: Treatise, Chapters 5 and 6.

Requirements for each schedule were worked out from population projections. For the Houselist, the projected population was divided by an assumed household size of five. Allowance was then made for vacant and non-residential houses (using 1981 figures inflated by 25 per cent) and for multiple households and enterprises (a further 50 per cent), with 10 per cent for wastage, and the result was doubled because the list was prepared in duplicate. Individual Slips were estimated by inflating the projected population for March 1991 by 15 per cent (5 per cent for wastage and 10 per cent as reserve). Household Schedules were taken as one-fifth of the Individual Slips, since one form covered five or six persons.

The Enterprise List was again canvassed with the Houselist on behalf of the CSO. Instruction booklets for it, printed at the Government of India Press, Bhubaneswar, were not supplied in time because of what the Treatise describes as the CSO's poor planning and nonchalant attitude. In States such as Orissa, material had to be dispatched without the booklets, and timely distribution was achieved only after the Registrar General and the DCO, Orissa pursued the matter vigorously.

Table 13: Honorarium for field staff, 1991 Census

Source: Treatise, Chapter 5. The Treatise notes a general view that these rates were inadequate for the work involved.

Data processing and dissemination

A mainframe computer was procured for census processing, supported by microcomputers for decentralised work. The General PCA, the separate PCAs for Scheduled Castes and Scheduled Tribes, and a few other tables were produced manually from the Household Schedules. All other tables were generated by computer from the Individual Slips in three stages:

•Stage I. A 10 per cent sample of Individual Slips in the larger States (those with populations of 10 million or more) and all slips in other States and UTs, producing the fertility tables, most migration and social-cultural tables, and broad economic tables.

•Stage II. Full processing of slips for main workers other than cultivators and agricultural labourers, marginal workers, and non-workers seeking or available for work, giving detailed economic tables.

•Stage III. Full processing of all Scheduled Caste and Scheduled Tribe slips for special tables.

Magnetic tapes had first been offered to users in 1981; in 1991 tables were also planned for release on floppy disks and CD-ROMs alongside printed volumes. Provisional totals were transmitted from the field for the first time over the National Informatics Centre's satellite link (NICNET). Because the link was still experimental, telegram, wireless and telephone were also used. Control rooms ran round the clock from 1 March 1991.

Evaluation

A Post Enumeration Check was held immediately after enumeration in States and UTs with a population of two million or more, on a sample of 4,200 enumeration blocks for coverage error and 840 blocks for content error. The blocks were selected centrally and communicated to the Directorates on 1 January 1991.

Challenges and learnings

The main lesson of 1991 was legal. The Census Act, 1948 (Act No. 37 of 1948) did not cover houselisting as a statutory operation. Offences under Section 11 were non-cognizable, since under the Code of Criminal Procedure only offences punishable with three or more years' imprisonment are cognizable. Prosecution required the prior sanction of the State government (and, for gazetted Central government employees, of the Government of India), and because Section 468 of the Code sets a one-year limit, cases often became time-barred before sanction was obtained. The Treatise concludes that the Act gave defaulting respondents and functionaries no effective deterrent. It recommended that offences be made cognizable, that the requirement of prior sanction be dropped except in corruption cases (on the lines of the Representation of the People Act, 1951), and that houselisting be expressly covered by the Act.

Census of India 2001

Scale and coverage

The 2001 Census, the 14th since 1872 and the first of the twenty-first century, was called the Millennium Census. Houselisting and the Housing Census were conducted between April and September 2000, and population enumeration from 9 to 28 February 2001, with a revisional round from 1 to 5 March. The census counted 1.03 billion people across 35 States and UTs, 593 districts and 5,463 sub-districts, in about 5,98,900 villages and towns. About 27.8 per cent of the population was urban and the literacy rate was 64.8 per cent. (The overview paragraph of the Treatise's chapter on 2001 gives the literacy rate as about 51.6 per cent, which repeats the 1991 figure; 64.8 per cent is the value in the Treatise's Table 1.1.) More than 20 lakh enumerators and supervisors took part. It was the first census to fix the census moment at 00:00 hours of 1 March, in place of sunrise.

Preparation and the geographic frame

The first circular, issued in September 1997, asked States to freeze administrative boundaries as on 31 December 1999. The Home Ministry then advised that no changes be made between 1 January 2000 and 31 March 2001. Even so, three new States (Uttaranchal, Jharkhand and Chhattisgarh) were created in November 2000, only three months before enumeration.

Permanent Location Code Numbers (PLCN). Until 1991, villages were numbered serially within each tahsil, so codes changed whenever boundaries were redrawn, and inter-censal comparison was difficult. In 2001 every village in a State or UT was given a code in a single continuous series from 1 to N, independent of tahsil or district. The code had up to eight digits: the first six identified the village and the last two were set to '00' to allow for later division of villages or the addition of new ones.

Urban frame. The standard demographic criteria were retained. Following the Nagar Palika Act of 1994, all town panchayats became statutory towns and were therefore classed as urban whether or not they met the demographic criteria. For an area to be treated as an Urban Agglomeration, two conditions were added: the core town or at least one constituent town had to be a statutory town, and the total population of the constituents had to be at least 20,000 as per the 1991 Census.

Changes in questionnaire and method

The questionnaires were shaped by a Data Users' Conference in April 1998 (24–25 April, Vigyan Bhavan) and the Technical Advisory Committee. A single pre-test was held in all States and UTs in July 1999. The main changes were:

•A single Household Schedule. From 1961 to 1991, enumeration had used an Individual Slip for each person with a summary Household Schedule. In 2001 the Individual Slip was replaced by one detailed Household Schedule of 23 questions in 39 columns, printed on both sides of the form. Most questions had pre-coded answers.

•Houselisting and housing. New items included the condition of the house, drainage, availability of a bathroom and a kitchen, type of latrine, use of banking services, and household assets (radio/transistor, television, telephone, bicycle, scooter/motorcycle/moped and car/jeep/van).

•Disability. Disability questions, dropped in 1991, were restored, covering five types: seeing, speech, hearing, movement and mental.

•Other population items. Age at marriage was asked of males as well as females. A new question recorded the distance to the place of work and the mode of travel. The definition of economic activity was widened to include the production of milk and of crops for household consumption.

•Respondent authentication. The signature or thumb impression of the respondent was taken on the Household Schedule.

Field operations

Table 14: Calendar of the 2001 Census

Source: Treatise, Chapter 6.

In rural areas an enumerator covered about 150 households (some 750 persons) and in urban areas about 130 households (about 650 persons). Several areas had to be enumerated on different dates:

• Snow-bound areas. In Jammu & Kashmir and the snow-bound parts of Himachal Pradesh and Uttaranchal, enumeration was held from 11 to 30 September 2000, with a revisional round from 1 to 5 October 2000.

  
  

• Kashmir Valley and Doda. In six districts of the Kashmir Valley the enumeration period was extended from 1 October to 15 November 2000 (reference date 16 November). In four blocks of Doda district it ran until 15 December 2000 (reference date 16 December).

  
  

• Kinnaur. Flash floods in August 2000 forced the postponement of enumeration in Kinnaur district of Himachal Pradesh to 12–31 May 2001, with 1 June 2001 as the reference date.

  
  

• Gujarat earthquake. The earthquake of 26 January 2001, two weeks before enumeration, led to the suspension of census work in the whole of Kachchh district, the Morvi, Maliya-Miana and Wankaner talukas of Rajkot district, and Jodiya taluka of Jamnagar district. A new calendar was later drawn up for these areas.

Training. Because charge officers (Tahsildars, Block Development Officers and municipal executive officers) were too busy with routine administration to give training the time it needed, a separate cadre of Master Trainers was created for the first time. They were drawn from experienced lecturers, principals and other officials, trained intensively at district (and in some States, state) level, and then trained enumerators and supervisors in classes generally limited to 70–80 participants.

Publicity and monitoring. A census logo, designed by Shri Prashant Roy of the Directorate of Census Operations, West Bengal, carried the motto 'People Oriented'. It showed a house with four figures, two adults and two children ('ham do hamare do'), standing for the national goal of population stabilisation, and a commemorative postage stamp carrying it was issued. In cities with more than 50,000 people, red stamp-sized stickers were fixed to the door of every enumerated house. These helped monitor progress and prompted households that had been missed to come forward. Census Help Lines provided technical support to enumerators and, from February to 5 March 2001, took complaints from people who had not been enumerated.

Data processing: Intelligent Character Recognition

Before 2001, data were keyed in manually, only part of the collected information was digitised, and full results took five to ten years. In 2001 the Office of the Registrar General adopted image-based Intelligent Character Recognition (ICR):

• Software. The eFlow 2.0 ICR software of M/s Top Image Systems (TIS), Israel, was customised in-house by ORGI with TIS and the selected vendor, with CMC Ltd as system integrator.

  
  

• Infrastructure. Fifteen scanning centres were equipped with 25 high-speed Kodak 7230 scanners, 45 Windows NT servers, 1,060 Pentium III PCs and ZIP, SLR and DLT backup drives.

  
  

• Workflow. All 228 million filled-in schedules, in 16 languages, were scanned. Numeric entries were read automatically by ICR, while handwritten text (Scheduled Castes and Tribes, mother tongue, religion, education, industry and occupation) was captured through image-based Computer Assisted Coding.

  
  

• Outcome. All the data collected were processed for the first time (Table A.8). Editing was done in phases, starting with the items needed for the Primary Census Abstract.

Evaluation

The Post Enumeration Check, part of every census since 1951, was renamed the Post Enumeration Survey in 2001. Whereas the 1991 PEC covered only States and UTs with populations of two million or more, the 2001 PES covered all States and UTs. It measured coverage error, divided into the omission or duplication of households (Type I) and of persons within enumerated households (Type II), as well as content error in selected characteristics.

Challenges and learnings

• New States. With enumeration barely three months away when Uttaranchal, Jharkhand and Chhattisgarh were formed, separate Directorates could not be set up. The DCOs of Uttar Pradesh, Bihar and Madhya Pradesh also served as DCOs for the new States.

  
  

• Vacant Director posts. Several Directors joined late or were available only ex officio. Assam had three Directors in the two months before enumeration, with the DCO Sikkim holding additional charge for a time. Meghalaya was looked after first by the Assam Director and then by the DCO Arunachal Pradesh.

  
  

• Lesson. The Treatise concludes that all Directors of Census Operations must be in position at least two years before enumeration begins.

Census of India 2011

Scale and coverage

The 2011 Census was the 15th since 1872 and was conducted under the tagline 'Our Census, Our Future'. Houselisting and the Housing Census took place between 1 April and 30 September 2010, and population enumeration from 9 to 28 February 2011. It counted over 1.21 billion people across 35 States and UTs, 640 districts and 6,05,541 villages and towns (5,97,608 villages and 7,933 towns). About 31.1 per cent of the population was urban and the literacy rate was 73 per cent. More than 27 lakh functionaries took part. A separate schedule was canvassed alongside houselisting to collect the details needed to create the National Population Register (NPR).

Table 15: Scale of the 2001 and 2011 Censuses

Source: Treatise, Tables 1.1–1.3.

Preparation

Work began in 2007 with a circular of 22 October 2007 on recording jurisdictional changes since 2001 and compiling lists of administrative units, each with a unique code. Questions were tested twice: in a small field trial in six States in 2007, and in a full pre-test in all States and UTs between June and 17 August 2009, which covered 1,26,326 households and followed every procedure of the actual census.

Changes in questionnaire and method

The houselisting schedule was similar to that of 2001. Ownership of a computer or laptop (with or without internet) and of a mobile phone was added, and the question on the number of married couples with separate sleeping rooms was dropped. The Population Enumeration schedule had 29 questions. The main changes approved by the Technical Advisory Committee were:

• Gender. In a radical departure from the past, respondents could report a gender other than male or female.

  
  

• Date of birth. Both age in completed years and date of birth were recorded. Reporting of a complete date of birth rose from about 40 per cent in the 2007 field trial and 65 per cent in the 2009 pre-test to nearly 78 per cent in the census; the Treatise suggests that the NPR exercise may have contributed.

  
  

• Disability. The five disability codes of 2001 were increased to eight by adding mental retardation, mental illness, any other, and multiple disability. For the first time, respondents could report up to three disabilities.

  
  

• Marital status. Separate codes were provided for divorced and separated persons.

  
  

• Work status. Main workers were defined as those who worked for six months or more. Marginal workers were split into those who worked for three to six months and those who worked for less than three months. A new 'others' category of non-economic activity covered commercial sex workers and persons in illegal activities, and a code for rentiers was introduced.

  
  

• Migration. Because many district names had been left blank in 2001, the names of the village or town of birth and of last residence were recorded so that districts could be identified during processing.

  
  

• Questions dropped. Questions on land under cultivation, irrigated land and land tenure, asked in 2001 but not tabulated because of widespread non-response, were dropped.

Field operations

Each State and UT chose a 45-day window between April and September 2010 for houselisting, instead of the usual month, because the NPR schedule was canvassed at the same time. In Jammu & Kashmir and Himachal Pradesh, houselisting was done in two phases because of snow, and in Delhi for administrative convenience. About 25 lakh enumerators and 2 lakh supervisors visited some 25 crore households.

Population enumeration followed the extended de facto canvasser method: everyone was enumerated over a period of about three weeks, from 9 to 28 February 2011, and the count was brought up to the census moment of 00:00 hours on 1 March 2011 through a revisional round from 1 to 5 March. In the snow-bound parts of Jammu & Kashmir, Himachal Pradesh and Uttarakhand, enumeration was held from 11 to 30 September 2010, with a revisional round from 1 to 5 October 2010. The houseless population was counted on the night of 28 February 2011.

Training followed a four-tier cascade. Some 90 National Trainers trained 725 Master Trainer Facilitators. The facilitators trained about 54,000 Master Trainers, drawn from universities, colleges, research institutes and government departments, and these in turn trained the 27 lakh enumerators and supervisors as well as charge officers.

Printing and logistics: the partnership with India Post

About 220 million scannable schedules had to be printed, delivered to some 17,000 locations and returned to 17 scanning centres. ORGI entered into a partnership with the Department of Posts (DoP) to handle this.

• Paper and printing. In all 8,577 tonnes of paper were procured, including 4,636 tonnes of 90 GSM ICR-scannable paper conforming to BIS standards. Scannable schedules were printed at four presses in the houselisting phase and six in the enumeration phase. These included Manipal Press (Udupi), Gopsons Paper (Gautam Budh Nagar) and KL Hi-Tech Secure Print (Medak).

  
  

• Pre-printed codes and barcodes. State, district, sub-district and town codes were pre-printed on the schedules, and 3D barcodes were used to track material.

  
  

• Movement of material. DoP moved about 11,500 tonnes of material in the houselisting phase and about 10,500 tonnes in the enumeration phase, mainly through Speed Post. The logistics cost Rs. 52,27,91,546.

Mapping, data processing and evaluation

Digital mapping of towns was completed during the period. A digital database of 937 statutory towns was created under the Tenth Five Year Plan and the remaining 3,104 were digitised under the Eleventh Plan, covering 4,041 statutory towns in all. Data were captured by ICR using eFlow software. The Treatise records several constraints: entries written outside the boxes or touching the gridlines, overwriting and the use of whitener caused misrecognition, with close to 2 per cent 'false positives'. There was also no built-in quality check and only a limited management information system in the production environment.

The Post Enumeration Survey was conducted between April and July 2011 in all States and UTs on a sample of 4,000 enumeration blocks, with at least eight blocks in each smaller State or UT, grouped into six zones. Fieldwork was done by agencies independent of the enumeration. Coverage error was measured as before through Type I (households) and Type II (persons within households) errors, and content error was measured in a 20 per cent sub-sample of the PES blocks.

Challenges and learnings

Areas affected by left-wing extremism. Houselisting and the Housing Census could not be conducted in 103 villages of Chhattisgarh and 260 villages of West Bengal.

Manipur. Houselisting covered the whole State, but data for the Mao-Maram, Paomata and Purul sub-divisions of Senapati district were initially withheld and published only after joint verification with the State government.

Supply of boxes. DoP had proposed to supply 75,000 boxes a week to the presses but managed only 20,000, 50,000 and 52,000 in the first three weeks. Sub-contracted transport also caused delays.

Lost and misdirected material. In Jammu & Kashmir, 22 boxes of filled schedules went missing and could not be traced despite special efforts. In Uttar Pradesh, boxes delivered to wrong addresses delayed training until the district administration recovered them.

Comparative Summary, 1981–2011

Table 16: Key parameters of the four censuses

⁵ Includes estimates for areas not enumerated. Sources: Treatise, Tables 1.1–1.10 and Chapters 4–7.