POST ENUMERATION SURVEY Experiences from India What is PES? Post Enumeration Survey (PES) is a sample survey conducted shortly after census for the primary purpose of evaluating the census. Helps in identifying areas of deficiencies that would need improvement for subsequent censuses, such as: concepts and definitions employed, procedures of enumeration and related instructions to the field staff etc, areas of under Enumeration and misreporting. Objectives of PES It provides an assessment of the accuracy of the census figures. It estimates the magnitude of Coverage error Omission (under count) Duplication (over count) of individuals. Content error Response errors in respect of certain selected characteristics PES in India No precise estimate of the extent of error in the census count was ever made till 1941 census (prior to independence). Indian Census Organization, since 1951, has traditionally been evaluating the results of the Population Census through PES, also called Post Enumeration Check (PEC). No attempts, however, were ever made to adjust the census results based on the PES results. Types of studies in PES The PES combines two types of matching studies A post-censal matching survey for the purpose of measuring census coverage error A re-interview survey for the purpose of evaluating the extent of content error in the recorded census characteristics of the population. Types of studies in PES (contd..) In simpler terms, the PES envisages answers to the following two questions: Coverage Error: How accurately has every individual been enumerated in the Census? (Omission or duplication) Content Error: How precisely have the particulars of individuals been recorded in the Census? What is Coverage Error For any study of coverage error, the most important item is the instructions issued on ‘whom to enumerate’ in census. The coverage error investigated in PES consists of two components: i. Type I error: Omission or duplication of persons due to omission or duplication of households. ii. Type II error: Omission or duplication of individuals in enumerated households. Persons enumerated in Indian Census The Indian census defines 4 types of persons as eligible for enumeration in a particular household: i. Persons, who normally reside and are present in that household during the entire period of enumeration i.e. , from 9th February to 28th February (both days inclusive) ii. Persons, who are known to be normally residing and had actually stayed during a part of the enumeration period (9th February to 28th February) in the household but are not present at the time of enumeration Persons enumerated in Indian Census (contd..) iii. Persons, who are known to be normally residing in the household and are not present at the time of visit of the census enumerator but are expected to return by 28th February iv. Visitors, who are present in the household at the time of population enumeration and are expected to be away from the place of their usual residence during the entire enumeration period. For the purpose of census enumeration, such visitors are enumerated in the household where they are actually found during the enumeration period provided they have not been enumerated elsewhere Persons enumerated in PES Initial Listing to make the list exhaustive : All the persons in the household at the time of PES enumeration (as on the date of PES) and in addition, any other person during the reference period of the population enumeration List considered for PES: persons covered under the four categories in Census are only considered for the PES Possible reasons for coverage error Omission Unconventional Houses (e.g. garages, temples, gurdwaras, school buildings) where people, though not expected to live, do live. Locked Houses at the time of census: A situation observed mostly in urban areas. Visitors and domestic servants omitted due to ignorance of respondents and limited probing by enumerators. Under-enumeration of rented households: mainly in urban areas Under-enumeration of new born babies Possible reasons for coverage error (contd..) Duplication: There may be cases where persons, who usually live at one place have come to visit their families at weekend, may also be included in the household unless probed by the enumerator. In fact he/she should be enumerated at his/her normal place of residence. Time of PES data collection PES data collection should take place as soon as possible after the Census This is to minimize matching difficulties related to changes in the composition of households between the two dates. At the same time, there must be enough time to retrieve all the Census materials from the field to avoid any contamination between the two operations. Keeping this in view, PES for the 2001 census was conducted in all the states and union territories between April to July 2001. Preparation of PES Step 1: Determination of the sample size, sample design and the Schedules for canvassing. Sample size depends on level of dis-aggregation (Level of data presentation) level of reliability required for analysis resource constraints Sample Size in 2001 PES Based on 1991 PES omission rates and the PRSE, sample size at the national level was fixed at 3,000 EBs For estimating coverage error, all the 3000 EBs were considered For estimating content error, 600 EBs (20% of EBs used to estimate coverage error) were considered. These EBs were distributed among all the 35 States and Union territories in proportion to their projected population of 2001 Linear Systematic Sampling within each stratum (Rural, City & Non-City) was used for selection of EBs Organization of PES Conducted in all the States through Dte. of Census Operations (DCO) or Dte. of Economics & Statistics (DES) Uniform one-day training on concept and methodology etc. was imparted to one Nodal Officer from each DCO/DES at ORGI Nodal officers in turn imparted 2 days training to their officials (Enumerators/Supervisors) for field/desk work and supervision Officials drafted for PES work were not involved in Census data collection Organization of PES PES work starting from re-listing to field reconciliation: 10 days per Enumeration Block were considered necessary. In addition to PES enumerators, PES supervisors were appointed to supervise the work of enumerators. contd. Each supervisor was allotted 5 Enumeration Blocks. Overall monitoring was done by senior officers of the Census organisation and State Govts (wherever they were involved in PES work) Data Collection Methodology The unit of analysis in the PES is person (member of household). The target universe included the residential and partly-residential houses. Only normal households were covered; institutional and houseless households were excluded in view of its insignificant proportion to total population. Schedules Canvassed in PES Three main schedules Related to Coverage Error: Schedule on listing of Building/Census House/ Households (Schedule-I): To identify omitted or duplicated households (Type I error) Schedule on listing of members of the Household (Schedule-IV): To find out persons omitted or duplicated in households enumerated in the census (Total error) Related to Content Error: Schedule on individual particulars (Schedule-VI): To determine content error in respect of selected questions like age, sex, marital status, literacy, educational level, disability, work and fertility status Three more Schedules - used for desk matching & field reconciliation PES: Stages of work Three distinct stages of operation : i. Field Operation: Independent listing and independent canvassing of the PES Schedules I,IV and VI (wherever applicable) ii. Desk Job: Desk matching of particulars collected in the PES and the corresponding Census records of an EB iii. Field Reconciliation: a field operation Criteria for matching households A household was considered as matched between PES and Census only when the following were same in both PES Schedule I and Census: house number use of the census house name of the head of the household Partially matched case, established at the desk match stage, were probed during field reconciliation (repeat visit in that household) for arriving at a decision regarding its status as matched or not. Criteria for matching persons In case of persons listed in Schedule IV, a match was established at desk match stage when all the following were same in both the records Name Sex relationship to head marital status of the person The partial match cases were further investigated during field reconciliation. Similarly, the entries in PES Schedule VI were matched with the corresponding Census records and the partial matched cases were investigated during field reconciliation. Independency of PES from Census For realistic evaluation of Census, independence between the Census and the PES operations is essential To maintain independence The persons other than Census field functionaries were engaged for PES work One group of persons for re-listing / re-enumeration during PES Another group of persons for desk matching and field reconciliation Independency of PES from Census contd. Census enumeration records were brought to the concerned DCO before commencement of PES. The list of the selected EBs was not disclosed till the population enumeration was over. The Census field staff did not have any prior knowledge of the selected EBs of PES. The Abridged House List (AHL) and filled –in Census Household Schedule 2001 were not shown to the PES enumerator till completion of field work. Results of PES 2001 Coverage Error: Omission rate is 24 per 1000 persons at National level Duplication rate is 1 per 1000 persons at National level Net omission rate (NOR) at National level: 23 per 1000 persons NOR higher in urban areas (40 per 1000 persons) than that of rural areas (17 per 1000 persons) NOR higher in the age-group of 0-4 and 20-24 years No significant difference in NOR by sex Results of PES 2001 Content Error: Content error is estimated only for matched persons and selected variables Indicators of Content Error: Gross Difference Rate: Percentage of persons for whom the classification differs both in Census and PES to total persons with regard to a variable Net Difference Rate: (Value of Census for a characteristic – Value of PES) x 100 Sample population size Index of inconsistency: It estimates the ratio of response variance to total variance for a question answer. It is a relative measure of response variance. <20 (low), 20 to 50(moderate) and >50(high) Results of PES 2001 contd. Content Error Characteristic Gross Difference Rate Net Difference Rate Index of Inconsistency Persons by marital status 0.3 to 0.9 -0.3 to 0.3 Low ( < 20 ) Persons by Literacy 0.5 to 6.3 -0.8 to 0.1 Low ( < 20 ) Persons by disability 0.3 to 1.6 -1.6 to 1.6 Low ( < 20 ) Persons by work status 0.3 to 1.2 -3.6 to 3.6 Low ( < 20 ) Persons by category of economic activity 0.2 to 0.4 -1.4 to 1.2 Low ( < 20 ) Values shown in ranges are for different sub-groups within each characteristic