80 likes | 187 Views
Centre for Actuarial Research (CARe) A Research Unit of the University of Cape Town. Lessons from Southern Africa Census and Survey Data. Outline. Lessons learned … … from CSOs … from censuses … from surveys Suggestions and questions for the future. Lessons learned … from CSOs.
E N D
Centre for Actuarial Research (CARe) A Research Unit of the University of Cape Town Lessons from Southern Africa Census and Survey Data
Outline • Lessons learned … • … from CSOs • … from censuses • … from surveys • Suggestions and questions for the future
Lessons learned … from CSOs • Dangers associated with the loss of institutional memory • Paradoxically, strengthens desire to perpetuate analytical approaches and processes of the past • The need to properly archive data, and maintain strict version control • Or, what happens when your server crashes? • The need to preserve documentation • Or, what happens when you have handed out the last existing copy of your report?
Lessons learned … from CSOs • Need to ensure reasonable access to data • Not handing it over to outsiders who promise to act as a gatekeeper, but then deny access later • Not abusing the confidentiality of unit-record data to deny access to bona fide researchers • Political, social or developmental goals can lead to pressure on CSOs to ensure that demographic indicators are ‘moving in the right direction’ • Estimation methods chosen to give the ‘right’ results or which conform to other expectations • E.g. Millennium Development Goals
Lessons learned … from censuses • Need to distrust data put there by editing rules; and particularly those dreamed up by foreign consultants • Enumerators matter • Not using enumeration as a (temporary) job-creation exercise • Not using unskilled enumerators • Proper training of enumerators • Not using untried technology for data capture or processing
Lessons learned … from censuses • Need for greater checks on the data, and a willingness to admit to errors • Need to validate results against those from other sources (e.g. school enrolments by age give a minimum population at that age) • Need for long(er)-term planning cycles • Rushed censuses invariably produce bad data • Longer-term planning horizon will lead to less of a boom-to-bust cycle in terms of staffing and management
Lessons learned … from surveys • Not all DHSs are of good quality • Depends on the partner organisation used to do the fieldwork • Not all DHSs are in the public domain • The same points as the two above apply, but to an even greater extent, to other national surveys
Things that might be challenged • How do we view Post-Enumeration Surveys? • Do we adjust the data for estimated undercount? • What are the implications of not doing so? • Subnational estimates • What are the implications of doing so? • Would a short-form census be better? • A poor census enumeration has knock on effects onto surveys using that census to determine a sampling frame • Transparency in editing of censuses data