Benefits of Sample long Form to Enlarge the scope of Census Data Analysis: The Experience Of Bangladesh

Transcription

1 yed S. Hossain, University of Dhaka A K M Mahabubur Rahman Joarder, Statistics Division, GOB Md. Abdur Rahim, BBS, GOB eeds Assessment Conference On Census Analysis III Benefits of Sample long Form to Enlarge the scope of Census Data Analysis: The Experience Of Bangladesh Syed S. Hossain, University of Dhaka A K M Mahabubur Rahman Joarder, Statistics Division, GOB Md. Abdur Rahim, BBS, GOB November 24, 2011

2 Outline Resume of 5 th Census Why Sample Census (Long Form)? Long Form Questionnaire Sampling Method Sample Size Limitation of Long Form

3 Resume of Census in Bangladesh Bangladesh Bureau of Statistics (BBS) was mandated to conduct the population and housing census decennially. Census history in this territory: Pre-independence: 1872, 1881,1891, 1901, 1911, 1921, 1931, 1941,1951, Post-independence: 1974, 1981, 1991, Present: Fifth census, 2011.

4 Main Objectives of the 5 th Census To develop benchmark database for all areas of the country for policy implication when necessary and consistent. To develop sampling frame for all socio-economic and demographic surveys. To use in area delimitation and updating of electoral demarcation of all national and local level elections.

5 5 th Census Phases Population and Housing Census, 2011 has three phases: Phase-1: Main Count: Collection of basic information for all households and individuals using an ICR format questionnaire; Duration March 15-19, Phase-2: Post Enumeration Survey (PES): A special survey to evaluate the quality and coverage the Phase-1 enumeration. Duration April 10-14, 2011 (conducted by independent organization Bangladesh Institute of Development Studies). Phase-3: Sample Census: A Sample Census with a longer questionnaire to collect detailed socio-economic and demographic information from selected households and individuals in the selected areas. Duration September 15-25, 2011.

6 Census Enumeration House to house interview method. Floating Population were enumerated at census night. Modified De-Facto method of enumeration was adopted for collection of data for the main census and the PES. Modification: People on duty and on board were counted at their usual residences. De-Jure method of enumeration was followed in the sample census. Enumeration Area (EA) were costructed using GIS mapping. On average, each Enumeration Area (EA) consisted of about 120 households.

7 Census Enumeration (Cont.) A total of 296 thousand EAs were constructed. Villages having more than 120 households were divided into two or more enumeration areas. Villages having less than 80 households were merged with other adjacent smaller village(s) to define the enumeration areas. A supervisor was appointed over groups of 6 to 7 enumerators. A total of 48,531 supervisor areas were created for the country and these areas were delineated on maps. In total 2,612 Zones were created. 130 Census District Coordinators were appointed from high officials of BBS.

8 Census Questionnaire There are 25 questions in the census questionnaire: 11 on household and housing (HH) module. 14 on individual module. Questions on type of household, number of houses, tenancy of the house, source of drinking water, toilet facilities, electricity connection were asked in the HH module. Information on age, sex, marital status, religion, educational status, field of education, literacy, activity status, field of employment were collected in the individual module. Ethnicity and Disability are the new variables (compared to 2001 census) included in the questionnaire.

9 Census Form/Questionnaire 2011

10 Technology Used Data is being Captured by using Intelligent Character Recognition (ICR) machine. Integrated Computer Assisted Data Entry (icade) software is used for data capturing. CSPro will be used for data analysis. Data capturing process and icade Software is supported by US Bureau of Census.

11 Main anticipated outputs of the Census Population count. Household count. Desegregation by household and individual characteristics. Cross classification by above variables. Any other analysis required and possible with the available indicators.

12 Preliminary Results

13 yed S. Hossain, University of Dhaka A K M Mahabubur Rahman Joarder, Statistics Division, GOB Md. Abdur Rahim, BBS, GOB eeds Assessment Conference On Census Analysis III Preliminary Results Table: Population in thousands in Diferent Census Bangladesh: Population 71,479 87, , , ,319 Change in Pop. - 15,641 19,195 18,040 17,964 Pop. increase % Av. annual increase - 1,931 1,920 1,804 1,772 Av. annual growth rate % Note: 2011 PES adjustment yet to be made.

14 Why Long Questionnaire Survey? The 1st phase census questionnaire was designed for quick counting and did not accommodate many of the stake-holders demand regarding Demographic characteristics like: Fertility, Mortality, Nuptiality, Migration etc. Socio-economic characteristics like: Asset, access to mass media, access to electricity, water and sanitation issues etc. Inclusion of all these indicators in Census questionnaire would make it lengthy, hard to manage and in short time frame and incur a huge cost. A long questionnaire survey was thus planned to follow the 1st phase census.

15 Expected benefits Information on the variables Mortality, Fertility, Migration and other topics could be gathered at a contemporary time of the total count. The estimated indicators obtained by long questionnaire may be used as compliment of the census indicators. The set up of the census (sampling frame, trained enumerators, administrative body and equipment) can be used right away.

16 Sample Census(Long Form) BBS has conducted the sample census for estimating the socioeconomic and demographic condition of the country following the Population and Housing Census It is the last phase of completion the census process, which was continued for twelve days.

17 Long Form Questionnaire The Long Form questionnaire included 81 questions in 10 modules. The modules are Module 1: Identification and Household Summary Module 2: Housing related Module 3: Household Related Module 4: Individual Module 5: Economic Activities (For members of age 10 years and above) Module 7: Fertility (For currently married/divorced/separated/widowed women of age 10 years and above) Module 8: Members returned from abroad Module 9: International Migration Module 10: Death (During last 12 months) LongFormEnumeration

18 Sample Design Stratified two-stage cluster sampling. First stage units EA (Enumeration Area). Second stage units households. EAs consisting of about HH were created all over the country and GIS Map were developed for each EA prior to the census. HHs were listed prior to 2011 census.

19 Stratification Estimation for the district delineation will be required, each of the administrative districts were treated as 64 strata. At the national and divisional level the estimation will be required to be segregated for rural, urban and City Corporation region. The district sample will therefore be consisted of both rural and urban region. In addition, for estimation at the city corporation level, six additional strata are defined as the 6 city corporations. Total Strata= 64+6=70

20 Resume of Long Form Survey The total country territory has been divided into 70 domains including 64 districts and 6 City Corporations. 96 Enumeration Areas (EAs) were chosen from each domain. 25 households were selected per EA general households were thus chosen from each domain. A total number of 6720 Primary Sampling Units (PSUs) all over the country were covered. 168,000 households was covered under 70 domains in the sample census.

21 Determination of Sample Size The sample size is determined by the usual formula. Where the main issues taken into consideration were: Previous estimate of proportions for the variables of interest vary from p=0.005 (for MMR) to p=0.97 (for access to drinking water). Proportions for the rare variables needed to be kept in mind. The value of Design effect is taken to be 2. A moderately small margin of error is needed to be chosen for estimating rare variables. Attempting to obtain district estimate of rare variables like MMR with a formidable confidence would require a huge sample size.

22 Determination of Sample Size (Cont.) The sample size for the study was determined to allow a maximum error of ±25% (of the proportion) in estimating a characteristic as rare as p =.05. The minimum required sample size was calculated to be 2335 HH per stratum (approximately 2400). With this sample size a calculation is done to see how much margin of error could be allowed for what type of estimation.

23 Margin of error to be achieved for district and division estimate

24 Achievable margin of error for different Variable for District Estimate

25 Achievable margin of error for different Variable for Dhaka Division

26 Limitations of the long form Sample size / statistical validity / cost Duration of interview (About 45 minutes) Long/short timing: Simultaneous, Deferred, When? Long form enumerators: Same? Others? Supervisors? Long/short common variables

27 Some Recommendation/thoughts The analysis plan should be carefully handled. If augmentation/adjustment of the short form results is required to be done using Long form findings, the associated theories are needed to be researched. Suggestions/advices from Demographer/Statisticians/Experts from home and abroad should be sought during and before analysis as required.

28