Evaluation of Census Data for the General Population Census Data of Cambodia 2008 BY: Mr. Meng Kimhor Deputy Director General of NIS/MoP Presented at Workshop on Census Data Evaluation for the South East Asian Countries, Phnom Penh, Cambodia, 14-17 November 2011 1 Overview (1) Pop. Censuses in Cambodia: First Pop. Census 1962, Second Pop. Census 1998, Third Pop. Census 2008, There was a gap of 36 years between 1962 and 1998 because of conflict and civil war. The majority of the 1998 General Population Census was supported fully by UNFPA. UNDP provided vehicles, but took them back. 2008 Pop.Census was supported by UNFPA, JICA, Japanese Gov’t, Germany Gov’t, and RGC The de facto was used for 1998 and 2008 population census. 2 Overview (2) Post-enumeration Survey (PES) was conducted immediately after a census for the purpose of evaluating the census quality (from 25 March to 04 April, 2008). The Provincial Planning Director and Deputy Director in each province coordinated and supervised the PES operations in the respective province. About 200 enumerators and supervisors carried out the field task. The net coverage error of missed persons (2.77 percent) and the level of response variance of selected characteristics as estimated by PES show that in the general 2008 General Population Census of Cambodia was conducted satisfactorily. 3 Objective • The main objective of PES is to quantify the omission and duplication and ultimately estimate in the net error in the census count, and also to measure the response errors in respect of a few selected characteristics canvassed in the census. • Apart from evaluation, the PES provides a feedback on operational issues which will be useful while organizing future censuses. 4 Method Used To Evaluate 1998 and 2008 Census Data Recruit enumerators and supervisors for field task (200 Enu. and Sup for PES of 2008 census). Sample Design: the sample plan for conducting the PES was very simple. It is a probability sample selected in single stage, using systematic random sampling. Sample Frame: The sample frame for the PES was the final list of Enumeration Area (EAs) in Cambodia they were to be covered by the 2008 General Population Census. There were about 28,000 EAs listed for the census. The sample selected consisted of 100 EAs. The selected EAs were reenumerated for PES. The overall size of the PES sample that was selected about 9,600 households with 46,000 persons for 2008 and in 1998 census PES sample was selected 8,703 households or 45,021 persons. Sample selection: After sorting the EAs in the sequence indicated the sample was selected at the rate of 1 in 282.6 using a random start for 2008 census and 1998 census, the sample was selected at the rate of 1 in 249.2, using random start of 78.9. The sample was systematic, equal probability selection design as geographic ordering, urban and rural, EAs arranged in order of their estimated sizes within each urban or rural sector. 5 Method Used To Evaluate 1998 and 2008 Census Data (2) Limitation of PES: many countries undertake post-census studies by re-checking a sample of the census returns, in one way or anther, to try to measure the degree of undercounting. Training and Fieldwork: There were three distinct stages of operation in the PES such as: i) Listing and enumeration of persons in all households in buildings situated within the allotted Enumeration Area (EA), ii) Desk matching of particulars collected during listing with those contained in the corresponding census schedule, Form B, and iii) Field recollection of particulars, where necessary. Some Important Aspect of PES Procedure: In order to minimize the non-sampling error in conducting PES field operation the following situation that occurred during the PES re-enumeration were reconciled with the census in order to determine true census misses: i) Births, ii) Deaths, and iii) In-movers and out-movers 6 Main results of evaluation of Census Data 7 Estimates of coverage error Estimate Completion rate Missed rate Over-count rate Net missed rate Percent 2008 1998 98.22 2.15 0.37 1.78 97.23 3.39 0.62 2.77 8 Net missed rates by age group and sex for 1998 census (percent) Age group Both Sexes Males Females All ages 1.78 1.81 1.76 Under 1 0-9 19-Oct 20-29 30-39 40-49 50-59 60+ 4.64 1.78 1.49 3.14 1.84 1.18 0.91 1.56 4.04 1.63 1.34 3.48 2.44 1.41 1.04 1.02 5.29 1.94 1.63 2.85 1.34 1.01 0.82 1.96 9 Data Evaluation 10 Standard Error of Census Coverage Estimates Item statistics Gross under-count (Missed persons) Gross over-count (Over counted persons) Net under-count (Net missed persons) Net missed rate Coefficient of variation % 2008 1998 1998 2008 1998 2008 250,626 466,597 23,095 62,619 9.21 13 42,994 84,795 7,402 14,468 17.22 17 207,632 381,802 24,192 62,450 11.65 16 0.178 0.0277 0.002 0.0045 11.24 16 Estimate Standard Error 11 Net Undercount Rates in some countries Countries Australia Bangladesh Canada Cambodia China India New Zealand United Kingdom Year 1996 1991 1991 1996 1998 2008 2000 1991 2001 2001 2001 Net undercount rate (%) 1.6 4.6 2.9 2.4 1.8 2.8 1.8 1.7 2.3 2.2 2.2 12 Level of Agreement 13 Rate of Agreement and Aggregate Inconsistency Index for selected characteristics Characteristics Index of Inconsistency (IAG ) 1998 Age 2008 Rate of Agreement 1998 2008 12.08 5.6 89.1 94.8 Mother tongue 4.97 13.3 99.4 99.4 Marital Status NA 7.1 NA 95.9 16.3 7.7 91.0 97 NA 14.1 NA 98.9 Main activity 17.47 19.0 88.0 89.5 Employment period 44.34 43.2 84.0 75.1 Children ever born 15.60 16.0 86.0 90.3 Children surviving 10.18 19.0 91.4 86.5 Literacy Physical/Mental Disability 14 Net Difference Rate and Inconsistency Index of enumerated by age Age Group 0 1-4 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70+ Net Difference Rate 1998 -0.02 0.16 NA -0.02 -0.05 -0.09 0.10 0.01 0.07 -0.08 0.00 -0.13 0.02 0.02 0.02 -0.05 0.03 2008 NA NA -0.2 0.2 -0.1 -0.1 0.1 0.1 0.1 N -0.2 0.1 N -0.1 N N N Index of Inconsistency 1998 11.25 9.47 NA 11.19 11.50 9.95 13.86 12.55 13.49 13.44 13.89 14.70 15.10 14.70 15.96 17.16 7.40 2008 NA NA 7.4 7.8 5.9 2.1 3.2 3.2 8.4 5.3 4.6 4.9 9.5 6.3 6.8 15.6 5.1 15 Net Difference Rate And Inconsistency Index for Person Enumerated By Marital Status- Cambodia Marital Status Category Net Difference Rate Index of Inconsistency Never married 0.6 3.2 Married (i.e. currently married) 0.3 5.3 Widowed -0.7 14.3 Divorced -0.3 24.5 0.1 14.3 Separated 16 Net Difference Rate And Inconsistency Index for Person Enumerated By Type of Physical/Mental disability-Cambodia Type of Disability None In seeing In speech In hearing In movement Mental Net Difference Rate% 1.3 -0.4 0.1 N -0.5 -0.3 Index of Inconsistency 16.3 2 11.2 N 10 5.8 17 Net Difference Rate And Inconsistency Index for Mothers classified by children ever born Category (Number of Children Ever Born) 0 1 2 3 4 5 5+ 6 7 8 9 10+ Not Stated Net Difference Rate (%) 1998 0.91 0.44 0.39 -0.11 0.19 0.02 NA -0.05 -0.39 -0.22 -0.5 -0.66 -0.02 2008 Index of Inconsistency 1998 1.3 0.2 0.6 0.6 -0.3 NA -2.3 NA NA NA NA NA NA 4.28 11.78 14.8 16.11 19.83 21.13 NA 22.44 22.24 22.65 25.64 29.13 12.51 2008 7.3 16.7 16.7 14.8 15.4 NA 9.6 NA NA NA NA NA NA 18 Net Difference Rate And Inconsistency Index for Mothers classified by children surviving Category (Number of Children Ever Born) 0 1 2 3 4 5 5+ 6 7 8 9 10+ Not Stated Net Difference Rate (%) 1998 2008 0.77 0.14 0.02 -0.22 -0.07 0.08 NA -0.17 -0.14 -0.25 -0.12 -0.03 -0.01 Index of Inconsistency 1998 2.3 -0.9 -0.9 1.6 -0.3 NA -1.8 NA NA NA NA NA NA 4.27 9.51 10.77 11.51 11.2 13.52 NA 14.21 15.76 16.24 17.29 20.14 12.51 2008 13.2 18.4 16.9 22.5 21.8 NA 12.2 NA NA NA NA NA NA 19 Problems faced (1) • There were some problems at the household listing as households increased. In some EAs, the households increased to three or four times as compared with the results of the census mapping. Therefore the NIS faced the problem to print and send additional forms to the relevant EAs. • Printing a large volume of census forms and distributing them to the field on time was very crucial. 20 Problems faced (2) • Difficulties in enumeration in urban areas: – Multiple store building; – Locked houses; – Vacant houses; • During PES operation, some respondent forget some information that they answered during census time (hour 00 in 3 March 2008) 21 Conclusion • The PES is the one method to evaluate the census data in 1998 and 2008 • Through the PES, the 1998 and 2008 population census can be estimated over-count (duplicate plus erroneously counted) is only 0.37 in 1998 and 0.62 in 2008. • The quality of enumeration was generally good as revealed by the level of response variance is selected characteristics, age, martial status, mother tongue, literacy physical/mental disability, main activity, the aggregate level of inconsistency is moderate. • The net coverage error of missed persons is 1.78 percent in 1998 population census and 2.77 percent in 2008 population census. So, the generally the 1998 and 2008 was conducted satisfactorily. 22