Postediting a source of inconsistencies We are not

Post-editing: a source of inconsistencies We are not alone! Ton de Waal, UN/ECE Work Session on Statistical Data Editing, Bonn, September 2006

Results of sociological and psychological research n Over 40 years of research has shown that basically two kinds of persons exist: ¨ Us ¨ Them n Main characteristic of us: ¨ Always n right, unless misinformed by them Main characteristic of them: ¨ Always wrong, unless they listen to advice given by us

Example of one of us n One of us could be responsible for monthly business statistics for certain branch of industry ¨ focuses on specific domains for that branch ¨ uses population frame available at that moment ¨ uses background information available at that moment ¨ aims for stable time series without much fluctuations over time n n reducing fluctuations (variation) is more important than bias many outliers are identified

Example of one of them n One of them could be responsible for National Accounts ¨ may focus on other domains as us ¨ uses updated population frame ¨ uses updated and more background information ¨ aims for overview of entire economy n n reducing bias is more important than reducing variance less outliers need to be identified

Problem of NSI’s need to produce information on related topics, but on different moments and from different points of view ¨ Short term business statistics ¨ Annual business statistics ¨ Price indexes ¨ National Accounts n Who are “us” and who are “them”?

Conclusion n Need to aim for coherence and consistency across chain of related subprocesses ¨ Agreements need to be made between various subprocesses ¨ Methods of various subprocesses need to be geared one another n We are not alone!