Use of administrative data in short term economic indicators Statistics NZ Rochelle Barrow
Overview Statistics New Zealand – environment current use of admin data in short term economic indicators future use of admin data
Statistics NZ – current environment large emphasis on respondent load Minister of Statistics is also Minister of Small Business efficiency
Current use of admin data Quarterly Manufacturing Survey KAU based Operating income, purchases, raw material stocks, finished goods stocks, additions and disposals of assets Quarterly Wholesale Trade Survey KAU based Operating income, raw material stocks and finished goods stocks Monthly Retail Trade Survey GEO (geographic location/establishment) based Monthly sales, quarterly stocks
Type of admin data used data sourced from Inland Revenue Department registrations and deregistrations goods and services tax returns employer monthly schedules data received monthly approximately 6 weeks after the end of the reference period
Data manipulation admin data not equal to survey collected data match data to statistical model manipulation of GST data Group apportionment Estimation and apportionment of non monthly data Estimation of missing data result = a monthly series of GST sales and purchases by enterprise
How the data is used update population details stratify enterprises replace direct surveying of small business Result = 25 percent decrease in respondent load
Updating population details registrations and deregistrations update the dynamic business frame business rules used to determine which enterprises can be updated directly from admin data
Stratify businesses stratification by industry (ANZSIC) stratification variables: Annualised GST Rolling mean employment count use of two stratification variables has improved the efficiency of the sample design – decreasing the required sample size tax sample Full coverage Rolling mean EC Annualised GST
Manufacturing Survey (Mar 04 quarter) StratumNumber of KAU % of population Contribution to total ($m) % of total Full coverage ,56975 Sampled61132,94818 Non sampled 2,85413 Tax17,752801,2818 TOTAL22,12516,798
Retail Trade Survey (April 04 month) StratumNumber of GEO %Number of EN %Contributi on to total ($m) % of total Full coverage 5,86991,17322,29352 Sampled2,42242,13641,67438 Non sampled 25, ,69639 Tax32, , TOTAL66,59158,5934,415
Replace direct surveying of small business desire to reduce respondent load problems encountered Lack of timelinessForecastingQMS, RTS, WTS Incomplete coverage ModellingQMS, RTS, WTS Missing dataImputationQMS, RTS, WTS Definitions differModellingQMS, WTS – not deemed necessary RTS – modelling for shocks
Potential use of admin data - background information Electronic Funds Transfer at Point of Sale (EFTPOS) NZ has a relatively high level of debit and credit card useage businesses with electronic terminals deal directly with banks banks then engage one of two switching houses to process EFTPOS transactions the switching houses are owned (jointly) by the banks both switching houses have provided SNZ with data currently confirming scope and classifications e.g. regional and arranging ongoing supply of data
Data requested from switching houses MC = industry code RG = region includes internet transactions excludes overseas transactions
Analysis EFTPOS data for recent periods closely track movements in the Retail Trade Survey
Advantages and potential uses timeliness – available days after the end of the reference period more robust small domain estimates e.g. regional trading day adjustment analysis possible reduction in respondent load (if Retail Trade Survey moves to quarterly survey) other analysis and validation e.g. money spent by overseas visitors
Issues relating to EFTPOS data penetration of EFTPOS use differing degrees of card useage by storetype e.g supermarkets vs motor vehicle retailing one bank already produces estimates based on their EFTPOS data
Conclusion SNZ makes extensive use of admin data in short term indicators benefits include significant reductions in respondent load considerable opportunities exist with respect to EFTPOS data