Bangladesh Population Census Post Enumeration Check 1981
This report of post enumeration check (PEC) survey programme of the 1981 population census discusses elaborately the procedures and also explains in detail all findings of the survey programme. Such comprehensive documentation of the PEC study has been done in order to present complete analytical basis for understanding correctly the level of precision of the data collected in the 1981 Population Census of Bangladesh.
Achieving complete success for large-scale statistical operation like census of population is not an easy task in Bangladesh. The country has high density of population and low literacy rate as well as temporal and spatial barriers imposed by weaknesses of communication and transport facilities. Therefore, a well designed plan, complete in every detail, was worked out and adopted in carrying out the census of population in 1981.
From the beginning, attention was paid to adoption of stategies which could contain and substantially reduce the indercoverage error and provide reliable data for the census. With this end in view, the field enumeration was done nation-wide in a sfort period of three days, 6 through 8 March 1981. A short optical mark reading (OMR) questionnaire was utilized in the nation-wide count. All logistics support procedures were organised and followed through by use of updated maps with control system established around the computer. Training of census field workers was imparted using a verbatim training manual. Census campigning was done through radio, television, newspapers, posters, pamphlets, handbills, brochures, mike-announcements, drumb-beatings, screening a census flim in
cinema houses and television net-work and other media. Strict field supervision was ensured by placing trained staff members of the BBS as supervisors at thana, subdivision, district and division levels. Also, a core of siior officers of the Bureau constituted headqiiarters supervisory team which frequently paid field visits during census field work. All these management and control logistics made it possible to carry out the census enumeration successfully.
A rigourous and independent evaluation of coverage of the census was done by conducting the PEC survey about two weeks later. The PEC was designed to provide information for estimating coverage errors at the national level and separately for urban and rural areas as well as for deriving indicators of data quality. The PEC universe consisted of all census households from which the sample was selected.
The results of the post Enumeration Check :(PEC) Survey showed that the 1981 Population Census undercounted the population of Bangladesh by 3.1%. When this factor is applied to the census count of 87.1 million the adjusted count becomes 89.9 millio This net undercount of 3.1% can be broken down into a "missed" rate of 4,2% and an "erroneously enumerated (over count)" rate of 1.1% providing a net of 3.1%..
Marked differences may be observed when the census coverage rates are calculated for different population groups. Rural areas showed a 2.5% net undercount while urban areas showed a 7.7% net undercount or about three times the rural rate. When the 7.7% urban rate is broken down by sex we find that males had a 8.0% net undercount while females had a 7.1%. Within age groups one finds that for urban males the highest rate is for babies under one year of age (11.5%) followed by males 15 to 49 years (9.3%). Among urban females the highest rate is again for babies under one year of age (13.6%) followed by females over 55 years of age (9.9%) and females 15-19 years of age (8.0%).
In the rural areas males showed a net undercount of 2.3% while females had a 2.8%. Again males under one year of age had the highest rate of 5..4% while males 10-19 year had the next highest rate of 3.4% , The highest rate for rural females was also babies under one year of age (6.3%) followed by females over 50 years of age (4.5%) and females 15-19 year (45%).
The response variance associated with different questions included in the census also showed some marked differences. Response variance here is defined as the degree of agreement in reporting between the original census enumerator and the independent reinterview by the PEC enumerator for a matched set of persons/households. The item "sex" showed the highest level of agreement (99,9%) followed by marital status (97%). The other items in descending levels of agreement were "tenure of the dwelling unit" (95%), "literacy" (94%),."roof material of household structure" (89%), "highest class passed" (88%), "wall material of household structure" (85%) and "age classified into five-year age groups" (63%). As age increases the level of agreement decreases from 89% for age group Cr4 years down to 34% for persons over 55 years of age.
It was found in the operation of the PEC that the reliability of coverage estimates was very sensitive to sources of non-sampling error -- in particular (1) the degree of operational independence that is maintained between the main census and the PEC survey and (2) in the quality of matching the PEC households/persons with census household/persons. A great deal of emphasis was placed on developing effective procedures which would provide PEC estimates of acceptable quality.
Kind of Data
Census/enumeration data [cen]
Producers and sponsors
Authoring entity/Primary investigators
BANGLADESH BUREAU OF STATISTICS
The sample comprised a tot,250enumeration areas (EAs)- 150 rural and 100 urban EAs. For selection of the urban sample the 79 pourashavas or municipalities were definec$ban areas. All sample EAs were selected by adopting systematic random sampling scheme and two-way stratification design. The primary stratification was the grouping of rural and urban census EAs and the secondary stratification was done by arranging these EAs according to their geo-code sequence. The sample was self-weighting with constant sampling fraction of 1/1128 in rural areas and 1/167 in urban areas.
Deviations from the Sample Design
independent reenumeration of 250 sample EAs was done at the first stage by specially trained employees of the Bureau. The reenumeration quality was ensured by use of maps and enforcing strict supervision. Sixty fouL higher level headquaters staff personnel performed the supervision job. The fiLst stage reenumeration aim was to measure the number of persons missed in the census enumeration. At next stage, matching work was carried out in the headquarters. Each BA was independently matched twice by two different matchers and finally scrutinised by a supervisor for determinjng the ctha1.matchstatu of each person. Finally, intensive field revisits were made in order to estimate the number of "erroneausly enumerated" persons and also verify if persons classified "non-match-" were actually missed.
Census coverage estimates were signifiCantly liigher for for rural areas. The completion rate of rural areas was 96.4% compared to 91.1%/urban localities. The same was true of the missed rate-the rural missed rate was 3.6% against 8.9% of the urban. However, the errorneouslyenumerated rates of rural and urban areas were not significantly different. The net error rate was found to 2.5% for rural areas while the same for urban areas was 7.7%. At the national level the net error rate was calculated 3.1% with standard error 0.24. This was a record accomplishment if one compares the coverage errors of all earlier censuses which were found to range between 8 to 15%. The FEC programme was worked through with all the rigidity and care for correctly assessing the coverage and the data quality of the census. i believe that the FEC report presents all the details which are helpful for precise appraisal of success achieved in realising objectives of the 1981 Population Census. I express my sincere thanks to my collegues who have participated in bringing out this comprehensive report of the FEC survey programme.