Regional Course on Statistical Business Registers: Data sources, maintenance and quality assurance Perak, Malaysia 21-25 May, 2018 4 .1 Survey fram e methodology REVIEW For sampling purposes, a snapshot of the live register at a particular point in t im e is needed. The collect ion of active statistical units in the snapshot is referred to as a frozen frame. REVIEW A sampling frame for a survey is a subset of t he frozen fram e t hat includes units and characteristics needed for t he survey. A single frozen fram e should be used for all surveys in a given reference period Creat ing sam pling fram es SPECIFICATIONS Three m ain t hings need t o be specified t o draw appropriate sampling frames: ▸ Target population (which units?) ▸ Variables of interest ▸ Reference period CHOICE OF STATISTICAL UNIT Financial data Production data Regional data Ent erpri ses are Establishments or Establishments or typically the most kind-of-activity local units should be appropriate units to units are typically used if regional use for financial data. the most appropriate disaggregation is for production data. necessary. Typically a single t ype of unit is used for each survey, but t here are except ions where t arget populat ions include m ult iple unit t ypes. CHOICE OF STATISTICAL UNIT Enterprise groups are useful for financial analyses and for studying company strategies, but they are not normally the target populations for surveys because t hey are t oo diverse and unstable. SURVEYS OF EMPLOYMENT The sam pling fram es for t hese include all active units that are em ployers. SECTOR-SPECIFIC SURVEYS E.g., manufacturing surveys. The sam pling fram es for t hese include all active units with an ISIC code in the specified group. CAPITAL EXPENDITURE SURVEYS The sam pling fram es for t hese m ay include all active units above a cert ain size. BASIC STEPS St ep 1 St ep 2 At specified t im es A frozen frame is (m ont hly, quart erly, creat ed, t ypically yearly), snapshots are containing all ext ract ed from t he live statistical units that register containing all have been act ive in t he units and their past 12 months. characteristics. BASIC STEPS St ep 3 St ep 4 Frame populations are A sampling frame is t he com plet e set s of t ypically select ed from each statistical unit a single fram e t ype from t he frozen population. fram e. FRAME GENERATION FREQUENCY Annual Quarterly or monthly Som e count ries generat e If SBR frozen fram es are frozen frames annually, generat ed quart erly or and the annual frozen monthly, a relationship will fram e is used for all need t o be est ablished bet ween monthly, quarterly and survey fram es for different annual survey frames. periods (keeping in mind characteristics will likely not be addit ive over t im e). FRAME ERRORS: UNDERCOVERAGE Undercoverage, when t he fram e populat ion does not include all units in the target population for the survey, can occur for many reasons, including: ▸ A size threshold for inclusion in administrative s ources ▸ Active units that are erroneously marked inactive ▸ Lag time before an active unit is included in the SBR FRAME ERRORS: OVERCOVERAGE The opposit e, overcoverage, is when t he fram e population includes units not in the target populat ion for t he survey. Reasons include: ▸ Mistakenly keeping inactive units in the SBR ▸ Double counting when combining multiple data s ources FRAME DIAGRAM REDUCING FRAME ERRORS While fram e errors cannot be ent irely avoided, t here are ways t o reduce t hem . One m et hod is t o ext ract t ables from t he live regist er and compare them with the current frozen frame in order t o check for errors before t he next frozen fram e is generat ed. CASE STUD Y: THE D UTCH SBR In t he Net herlands, a frozen fram e is creat ed from t he live regist er each m ont h, consist ing of enterprise groups, enterprises and local units and their main characteristics. In addition the statistical units are linked to the legal and the administrative units valid for that month. CASE STUD Y: THE D UTCH SBR A list is kept with frame errors referring to activity codes or size classes of statistical units in order t o be able t o generat e several inst ances of each frozen fram e wit h im proved qualit y each t im e. REFERENCES This presentation was primarily based on UNECE’s Guidelines on Statistical Business Registers, in particular Chapter 8. Other m at erial used: ▸ The African Development Bank’s Guidelines for Building Statistical Business Registers in Africa ▸ Eurostat’s Business registers: Recommendations manual Country maps based on Natural Earth data. Presentation template by SlidesCarnival.