UN Banner
HomeAboutExperiences and LessonsCase StudiesRecommendationsResourcesChecklist
Home
About▶
BackgroundPurposeRationaleStructureTarget Audience
Experiences and Lessons▶
1. Project Planning and Management2. Geospatial Mapping and EA Database3. Enumeration Instruments and Tools4. Data Capture and Management5. Census Testing and Pilots6. Recruitment and Training7. Deployment and Supervision8. Enumeration and Logistics9. Quality Assurance and Risk Management10. Census Analysis and Products11. Partnerships and Collaboration12. Advocacy and Publicity13. Procurement and Financial Management14. Post-Enumeration Survey15. Alternative Approaches16. Planning Checklist
Case Studies▶
1. Project Planning and Management2. Geospatial Mapping3. Enumeration Instruments4. Data Capture5. Census Testing6. Recruitment and Training7. Deployment and Supervision8. Enumeration and Logistics9. Quality Assurance10. Census Analysis11. Partnerships12. Advocacy13. Procurement14. Post-Enumeration Survey15. Alternative Approaches16. Planning Checklist
Recommendations
Resources
Checklist
Skip to main content

Challenges lessons learnt​

  • If person and household information (demographic and metadata) is processed as two separate datasets, ensure that there is an easy and unambiguous way to join this information back together again as automatic matching will rely on using information from both datasets. For example, assign a unique identifier to each household and make sure that each person within a household is also labelled with that identifier. This applies to both census and PES data.

  • If there is no addressing system that can be used to identify buildings, then stickers can be used to label building enumerated during census, and the information written on the sticker can be collected during PES enumeration. If householders agree then labelling building with a permanent marker pen is also effective. Hence the same building in census and PES can be easily matched together prior to matching the people within the buildings and this will improve the automatic matchrate and the accuracy of the matching. (See case study from Rwanda). However, this method only works if:

    1. the stickers/marker pens are good quality and remain in place in-between census and PES enumeration

    2. both census and PES enumerators are trained well so that they understand the importance of the census enumerator writing the unique household ID clearly on the sticker and sticking it in a place where is it easy to find and read, and the PES enumerator understands the importance of correctly transcribing the unique household ID from the sticker.

There have been examples of countries that used chalk to mark households enumerated during the census, but these markings washed off in heavy rain so that they couldn’t be read during PES enumeration. In other countries stickers proved to be useless a lot of the time because the writing was not clearly readable, or the stickers had fallen off or been removed by householders.

  • A more high-tech method to identify matching households in the data is to collect latitude and longitude at the entrance to each structure. However, whilst useful, this can be challenging if the co-ordinates are not accurate, or where many households live within the same structure.

  • Carry out the PES as soon as possible after census enumeration has completed (within three months). Countries where there was a long time e.g., seven or eight months between census night and PES enumeration tend to have a much worse matchrate due to people moving, people not remembering who was in the household at the time of the census and stickers/household enumeration labels becoming illegible. The results of the subsequent dual system estimation may no longer be valid due to the difficulty of knowing whether people are correctly not matched due to an omission in either the census or the PES or whether they are people who moved in-between census and PES.

  • Give clear instructions to enumerators (on census and PES) of what to do in the case of missing information. For example, if day and/or month of birth is unknown then enter an ‘X’ for the unknown fields. In cases where enumerators have not been given clear instructions, they have sometimes used a ‘1’ causing a big spike of birthdays on 1st January. It is not possible to distinguish between people who genuinely have a birthday on 1st January and people who do not know their date of birth, which makes matching harder. The same applies for missing names – give enumerators clear instructions of what they should enter e.g., ‘UNKNOWN’. These values can then easily be found in the data and replaced with a code for missing. Otherwise, time will have to be spent finding all the different ways that enumerators have used to record an unknown value.

  • Although the census and PES should be independent, it is important that the teams running the census and the PES do interact with each other. The census team must understand the purpose of the PES and how the data the census team collect will be used for the purposes of matching. This understanding will feed into the census and PES questionnaire design and the training for the enumerators.

  • There have been examples where quality issues have been caused by lots of enumerators recording ‘99’ for a question because they were not sure how to complete the question. This could be prevented by improved questionnaire design, and the issue should be picked up during analysis of data collected during the census and/or PES pilot.