- 1 - index number: 060701 title: proceedings author: date ... · oracle with sql will be utilised...

49
- 1 - Index Number: 060701 Title: Proceedings Author: Date: October, 1991 Country: Australia Round Table: 6th Round table Coolangatta (AUSTRALIA) October 14 - 18, 1991

Upload: nguyentuyen

Post on 10-May-2018

217 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 1 -

Index Number: 060701

Title: Proceedings

Author:

Date: October, 1991

Country: Australia

Round Table: 6th Round table

Coolangatta (AUSTRALIA)

October 14 - 18, 1991

Page 2: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 2 -

Session No. 1OPENING SESSIONSession Leader : Australia

OPENING SESSION Session Leader: AUSTRALIAMr Sarossy welcomed delegates to the 6th Roundtable. He tendered apologies for the AustralianStatistician, Ian Castles, who had Electoral Commission responsibilities, Deputy Statistician, BillMcLennan who is in China and John Carrol who is acting Australian Statistician in Canberra; Johnwould attend the Roundtable from Tuesday and Mr Castles will attend from Thursday.

Administrative matters: Mr Sarossy introduced organisers Linda Crooks, Michael Toole andthen, the real "movers and shakers" of the Australian register; Graham Boxsell, Ian Marshall,Tony Carr, Dennis Robson, Brian Finegan and Lance Giles.

Mr Sarossy noted the program for accompanying persons and referred to the free day onWednesday. He promised that the organised activities would be enjoyable. After discussing someother organisational and social matters, Mr Sarossy explained the use of session presenters.Rather than being chairpersons as in the past , They had the role of developing the theme for thesession as well as chairing the proceedings.

Mr Sarossy asked delegates to think of agenda items for the next Roundtable as the conferenceproceeded.He then referred to the papers for each agenda item and asked for omissions and missing papersto the identified. A list of papers was prepared and circulated to each delegate so that they couldidentify papers they had missed.

Agenda Item 1 was then introduced.

Page 3: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 3 -

.Session No. 1PROGRESS REPORTSSession Leader : Australia

Chair: George Sarossy, Australian Bureau of Statistics

George opened proceedings by reminding participants that at the Williamsburg meeting it wasagreed that reports would be restricted to 5 to 10 minutes and would emphasise issues that willnot be dealt with in other sessions.

Reports would be presented alphabetically with the host country (Australia) last.

Progress Report 1: Statistics CanadaPresenter: Jacob Ryten

Jacob commented that the Statistics Canada register was meeting, or exceeding, its budgetary andquality goals that had been set. This assessment was based on a number of traditional measures ofquality - coverage, classification and costs. Also, updates from regular surveys (wholesale, retailand employment) are now incorporated into the register on a continuous basis without significantbacklogs.

As a result legal entities as recorded on the register will be used in the 1991 Industry Surveys.

Some of the objectives which have been temporarily dropped to allow this work to proceed wereoutlined by Jacob

Some simplification was introduced into the profiling activity by assuming, for all but the top 500companies, only one statistical company in each entity. The register as a source of small area datawill have shortcomings as individual locations will not be recorded within each province.

Matching of survey data to the register at the establishment level has been suspended at this stage.

Plans for the future include the planned profiling of the largest 300 businesses and a longer terminitiative to utilise the data generated by the administration of the GST to provide enhancedregister update information.

Page 4: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 4 -

In summary, Jacob advised that in terms of achieving a register to allow surveys to be conductedwithout duplication and with acceptable classifications some considerable success has beenachieved. The register however did not as yet provide an exhaustive record of the geography orlongitudinal record of businesses.

Progress report 2: Statistics DenmarkPresenter: Bente Dyrberg

Bente reported that the project to build a new register which was started two years ago isexpected to be completed by the end of 1992. This register is being established using data basetechnology using ORACLE and will be improved further by the integration of the Tax and theVAT registers (as described in a later session).

Substantial resources have been utilised in the development of a new Danish activity classificationwhich will be based on the NACE Rev 1. In 1992 it is planned to industry recode the 45000 unitson the register. The Yellow Pages will be used to reduce the number of units which will need tobe surveyed later.

Progress Report 3: Central Statistical Office of FinlandPresenter: Kaija Hovi

Kaija advised that the enterprise register unit of Statistics Finland, from the beginning of 1991,provides under a Finnish law, information services on the identification number of enterprises. Theaim is to have units identified by the same organisation number in the information systems in thepublic sector and much of the private sector.

Plans are also in place for the register coverage to be improved by adding own account workers(non-employing units) and local government establishments. This will add approx 60000 and38000 units respectively.

Use of the register has been extended by its use in activity and place of work identification in the1990 Population Census.

Comment was also provided on the reorganisation of Statistics Finland which incorporatedmanagement by results principles, customer orientation and user pays services. As part of thismove, Kaija outlined the introduction of a flat rate charge for telephone inquiries for register data.

Page 5: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 5 -

Progress Report 4: FrancePresenter: Phillippe Pommier

The major developments since the Williamsburg meeting identified by Phillippe were the use ofstatistical survey results to update the register for activity details, size (number of workers) andwhether the unit is still alive, a move to measure the quality of the register annually instead ofperiodically as in the past and the start of the process to link enterprises within the one enterprisegroup.

Some problems have been encountered with the electronic provision of data to enable these linksto be established, so completion of this project is still 1 or 2 years away.

The register will adopt a new activity classification harmonised with NACE and CITI on 1January 1993. Various techniques (annual survey data, automatic procedures and an additionalsurvey utilising optical reading of bar codes for data entry) will be used to move units to the newclassification.

Progress Report 5: JapanPresenter: Isao Mitake

Isao reported that for the first time the Establishment Census conducted by the Management andCoordination Agency and the Census of Commerce conducted by the Ministry of InternationalTrade and Industry were conducted on a joint, coordinated basis as at July 1991 to avoidduplicated enumeration and to reduce the burden on establishments surveyed.

The results of the censuses will be published by May 1992. The establishment data prepared forthe conduct of the survey will be used to establish a new directory covering all establishmentsincluding small individual proprietorship establishments previously excluded from the currentdirectory.

Future use of the new directory is intended to include a business survey frame system.

Progress report 6: Netherlands Central Bureau of StatisticsPresenters: Jean Ritzen and Ad Willeboordse

Jean reported that the CBS was undertaking a revision of the model of the business register. Thefirst stage, the inputs model, was completed in 1990. Research was now proceeding on an outputmodel. This revision is being done in conjunction with a review of the software to be utilised.Oracle with SQL will be utilised instead of the presently used ADABAS. An incrementalapproach utilising prototypes will be adopted with the aim of having a new operational system bythe beginning of 1993.

Page 6: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 6 -

A new approach to the definition of units is being introduced in the Netherlands emphasising legalunits. Local units will probably not be maintained.

Investigations are also proceeding on alternative sources of register data. One area of particularinterest here is the future requirement, by law, of firms to provide information on intra-EuropeanCommunity trade and the usefulness of this information for register purposes.

Jean advised that the possibility of a fourth cohort survey is being investigated in response todemands.

Also a survey carried out in 1990 indicated a gap of about $US 3 billion in interest paid andreceived by units. The relevance of particular units such as non-employers in some parts of theSIC is an important consideration when enumerating the economic process. This has particularimportance for National Accounts.

Ad reported that the Dutch SIC has been revised to be compatible with NACE Rev 1 and will beintroduced when special surveys to convert units have been completed. This may not be before 1January 1993.

Ad reported that a "top 100" data base, comprising data from a number of economic statistics, hasbeen completed for the 100 largest businesses in the Netherlands.

As a result of work undertaken in the Netherlands on unit definitions, Ad invited participants toprovide information on units definitions utilised in their systems and offered to utilise anyinformation provided by countries on the definitions to prepare for discussions at the nextroundtable.

Progress report 7: New Zealand Department of StatisticsPresenter: Keith Sykes

Keith reported on a change in the philosophy of updating the register from a system where usersupdate the register online to a system where all updating is undertaken by the Directory Sectionon contract to survey users. The contract is a formalised written agreement committing bothparties to specific criteria including content, quality and delivery.

The commercial environment faced by the Department has seen the development of key accountcontracts and the invitation to list brokers, direct marketing and market research firms to establishcommission agency arrangements for the sale of register products.

A review of National Accounts requirements have pointed to a need to increase the coverage ofunits on the Directory. The GST registrations will be a source of this expansion which is plannedto be completed in April 1992.

Page 7: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 7 -

The Agriculture Directory and the Business Directory have been merged to create one integrateddirectory to ease expansion of coverage plans and to simplify ongoing maintenance tasks. Also theintegration will result in a move from a land based system for agricultural units to an economicbasis.

A significant development reported by Keith is a review of the Directory with the objective ofexamining the existing operations of the Business Directory and to prepare a long term (at least 5years) development plan. The review is planned for completion in the second quarter 1992.

Progress report 8: Statistics SwedenPresenter: Marianne Lager

Marianne reported that changes to the VAT rules has consequently reduced, in January 1991, theregister by about 100000 units (from 570000).

A new source of information , the National Telecommunications Administration, is now beingutilised monthly for telephone numbers, mailing addresses and an indication whether the legal unithas more than one local unit.

About one third of the cost of the register is now financed by sales (approx $US 2 million).Resellers now number ten and cooperative arrangements with Finland and Denmark for sales ofextracts from their registers are in place. Two CD-ROM producers are producing informationutilising the Business Register.

Quality improvements reported by Marianne include the provision of updating data by theNational Tax Board every two weeks. At the end of October results from a quality study of threedifferent registers, including the Statistics Sweden register, will be released.

In 1991 Statistics Sweden commenced a project to amend the activity code utilised to NACE Rev1. The work is planned to be completed at the end of 1993, with 1994 statistics being the firstpublished on the new basis.

Progress report 9: EurostatPresenter: Keith Perry

Keith provided information on the current status of discussions within the European Communityregarding registers. The principle areas of discussion are units definition and harmonisation ofmember registers including the establishment of common core data. Papers outlining finalproposals are expected to be available in November.

Discussions have also canvassed a move away from the concept of a cutoff based on number ofemployees to an emphasis on the coverage of most economic activity.

Page 8: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 8 -

Register maintenance in the member countries will also be enhanced by a law providing forstatistical offices to have access to administrative data for register updating purposes.

Progress report 10: UK Employment DepartmentPresenter: Piers Grey-Wilson

Piers reported that the UK would move to the NACE Rev 1. As the UK National Accounts willnot be moved to the new classification by 1993 it will be necessary for the register to maintain 2industry codes on the register for some time after 1993.

The Employment Department will conduct a full Census of Employment in 1993, the firstcomplete collection since 1981, to establish a new register following a change to the InlandRevenue's PAYE system. Electronic capture methods such as imaging will be assessed and testedon a smaller survey in 1992.

In 1991 a study of a number of options for the new register was conducted by the Department.This study incorporated a wide ranging survey of user needs and consideration of the use of anInter Departmental Business Register (IDBR) being investigated jointly with the UK CSO.

Progress Report 11: UK Central Statistical OfficePresenter: John Perry

John commented that the past year has seen rapid change driven largely by the "Chancellor'sInitiative" to improve the quality of the register. The main points to note were a move to increasestaff and other resources inputs, the conduct of a number of new inquiries and the move of themonthly retail sales from a voluntary to a mandatory basis. Profiling of the top 2000 enterprise hasbeen required to conduct new surveys in particular a quarterly profits inquiry which is conductedat the enterprise group level.

Also the Office completed the integration of a separate register used for the Property ServicesInquiry, with the main VAT based register.

John advised that a revised SIC will be introduced in stages from 1993 to 1998 to provide forNational Accounts purposes, optical scanning and CD storage devices have been installed to assistwith form handling and enforcement procedures have been introduced, although no prosecutionshave been undertaken as yet.

A technical study of the IDBR has been completed with prototyping using a SEQUENT computerrunning UNIX and the INGRES relational database manager. A decision has now been made todevelop the register that will serve the needs of both the CSO and the Employment department. Itis planned to have the new register in place by 1994.

Page 9: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 9 -

Also the CSO is to investigate the form filling burden on businesses.

John also commented that the CSO is to become an agency, in line with a general move offunctions away from central government departments where possible. The agency would have achief executive and a management board.

Progress report 12: US Bureau of Labour StatisticsPresenter: Brian McDonald

Brian reported that the BLS has continued the process started in 1986 to improve the quality ofthe business register. A particular achievement noted was the addition of a considerable number ofunits (eg 1 million physical location addresses), the increase in the number of comment codes andthe enhancement of editing of incoming data files. The long term trend in number of errors wasnoted as being down.

Technical enhancements reported by Brian included the completion of the design of an extractmodule for the register system which will allow the business register users to interact directly withthe data base to extract records they require. Also a sample data base has been established forresearch and testing purposes to avoid the need to access the main data base.

Brian reported that the Business Establishment List (BEL) improvement project is changing thefocus of reporting by multi unit employers from a reporting unit level (often county orcounty/industry) to an worksite level. The first quarter of 1991 saw the introduction of a MultipleWorksite Report to facilitate the change.

It was also noted that the BLS and the Census Bureau have signed a Memorandum ofUnderstanding in April 1991 which will provide for the sharing of data, in particular SIC codesheld by BLS. Results of the matching process are being evaluated with a view to establishingappropriate procedures for the future.

Progress report 13: US Census BureauPresenter: Roger Bugenhagen

Roger outlined the Standard Statistical Establishment List (SSEL) maintained by the CensusBureau. It contains approx 6.6 million businesses with employees and during Economic Censusyears approx 11.5 million non employer businesses are added.

Activities in the period since the last Roundtable noted by Roger included the use of the 1987Economic Censuses data to update the register and conversion of the processing system to DECminicomputers to provide for interactive processing.

Page 10: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 10 -

Planning has also begun for the 1992 Economic Censuses which will be mailed in December 1992.Coverage of the Register will be expanded to cover communications and public utilities andfinance, insurance and real estate sectors. This will raise the coverage from about 75% to 98% ofGNP.

As mentioned above the Census Bureau and the BLS signed a Memorandum of Understanding inApril 1991 facilitating the sharing of data. Under the arrangement BLS will access certain Taxinformation and the Census Bureau will access BLS SIC data. Considerable analysis is required toassess the quality of information provided by the matching processes utilised.

Roger informed the meeting that the Census Bureau is sponsoring an International Conference onthe Classification of Economic Activity on November 6 to 8 at Williamsburg, Virginia. He alsomentioned that a concordance between the Canadian and US SIC systems has been published andwork is proceeding on a joint concordance between the US SIC, the Canadian SIC and NACE forcompletion in 1993.

Progress report 14: US National Agricultural Statistics Service.Presenter: Ron Bosecker

Ron reported that demands for NASS list frames have continued to escalate and that additionalfunds have been provided for approx 100 additional statisticians to cope with expanding programrequirements such as environmental statistics requirements and to improve list frame maintenance.

It was noted that a new Policy and Standards Memorandum was issued in 1991 to establishstandards for coverage, duplication checks, out of scope records,etc. A copy of the memorandumwas provided to participants for information.

The process of bringing all state offices on line to the mainframe data base has now beencompleted and by 1992 local area networks will be installed in all state offices. Also a new surveyprocessing system (SPS) involving interactive editing and based on BLAISE software is to bedeveloped.

In response to the need to keep abreast of changes the NASS headquarters is being restructuredto re-establish a separate Research Division, a new sampling frame research Section and units toaddress the expanding environment statistics program.

Ron mentioned to participants plans for an international conference on Business and Agriculturalprocessing methods to be conducted in 1993 or 1994.

Page 11: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 11 -

Progress report 15: Australian Bureau of StatisticsPresenter: John Struik

John provided to the meeting an update of the status of the major activities being undertaken inAustralia.

Profiling of large businesses is into its second round and is regarded as a success, although anexpensive operation requiring organisational commitment. Also public sector units have beenupdated and will now be maintained.

Considerable effort has been expended in order to match Australian Taxation Office (ATO)records to overcome undercoverage and to upgrade the system of identifying births to theregister. In addition, a survey of units to identify units no longer alive has been conducted.Lessons to be learnt from these exercises are that early advice to all affected areas of the likelychanges to their frame is essential to allow for appropriate adjustment to surveys for suchinfluences as undercoverage of new businesses.

The ATO modernisation project, due to be completed in 1993, was outlined by John with themajor point being the potential for a considerably enhanced source of all register updatinginformation including non employer data which have had to be excluded from the register at themoment. This project has led to preliminary thinking within the ABS on a new processing systemwhich will be developed after 1993. Also the enhanced data will allow the register to be used as asmall area data base.

Closing remarks

Mr Sarossy commented throughout the session on the common threads raised during discussion.These were that a number of agencies are adopting a user pays policy, are addressing qualityissues, moving towards using relational data bases, and that National Accounts requirements werea greater focus for the use of registers than in the past.

The session closed with an opportunity for participants to clarify issues with the presenters of theindividual progress reports.

Page 12: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 12 -

Session No. 2QUALITY ASSURANCE PRACTICES IN REGISTER MANAGEMENTSession Leader : Canada

Chair: Michael Colledge, Statistics Canada

Introduction by Chair:

Michael introduced the main themes as being the measurement and the management of registerquality, where "quality" was to be interpreted in a broad sense to include relevance, accuracy,timeliness and cost effectiveness.

Four of the papers to be presented addressed the theme of quality measurement. The first, fromthe US Bureau of the Census, looked at register quality on the input side, while the second, fromthe UK Central Statistical Office, compared register content with other data files. The third,Australian paper discussed aspects of the quality of the input processes and then introduced theidea of adjusting survey estimates in accordance with known register deficiencies. The theme ofcompensation for register errors was also the subject of the fourth paper from France whichoutlined the adjustments made for inactive units.

The fifth paper summarized and analysed the results of the questionnaire on quality assurancecirculated by Canada. The sixth paper was not specific to registers, but a description of theapproach of a whole agency, the US Bureau of the Census, to the general issue of qualitymanagement.

Paper 1:Quality Assurance of Administrative Record Data

Presenter: C. Harvey Monk, United States Bureau of the Census

Presentation Summary:Quality assurance is particularly important to the Bureau of the Census (USBC) in the context ofupdating the Standard Statistical Establishment List (the business register) using administrativedata from the Internal Revenue Service (IRS).

Some of the disadvantages in using administrative records were experienced during the 1982Economic Censuses. The data which were collected primarily for the internal use of IRScontained a number of serious data errors. Although the IRS was willing to adjust its proceduresfor future processing it was not prepared to reprocess data that had already passed through thesystem.

Page 13: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 13 -

Based on this experience USBC decided that additional quality control measures were needed toensure the early detection of errors contained in IRS data for the 1987 Economic Censuses. Thevolume of data concerned was very large: over 65 million records from ten separate IRS centres,with a variety of data items.

A quality assurance system was developed to suit the situation. The major element wasmechanised data verification. The USBC established standards that reflected the desiredcharacteristics of the incoming IRS data. The standards were discussed with IRS prior toimplementation. They related to both timing and quality.

The reporting patterns of the actual data are compared to these standards based on the automatedanalysis of a 2 percent sample of records. Deviations from the standards were signals indicatingpotential problems in the data. For example, during the course of processing data for the 1987Censuses, a major deviation in the proportions of two form types flowing into the system wasdetected. However, this proved to be a consequence not of an error but of a genuine shift in thepopulation caused by changed tax laws.

The system was successful in demonstrating that quality standards were being met. Similartechniques have been and are being introduced for other SSEL operations.

Discussion:In response to a query about the source of industry codes it was reported that the IRS respondentis asked to self-code, and that these codes are accepted without question when they are supplied(which occurs 80 to 85 per cent of the time). An industry description is also requested which willbe coded by the IRS if required. The IRS take this responsibility for coding very seriously ascoding is viewed as part of the service purchased by the Census Bureau.

It was also noted that the quality standards are essentially measures of expectation based onhistorical experience. At this stage there is little else available.

Paper 2:Measuring Quality: Business Register Data for Northern Ireland

Presenter: John Perry, UK Central Statistical Office (CSO)

Presentation Summary:Northern Ireland has been used as a prototype area for testing the establishment of an inter-departmental business register (IDBR). Several aspects of the quality of administrative data havebeen investigated.

Page 14: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 14 -

Records on the CSO's VAT register were matched against the definitive Companies Houseregister. Matching on an 18 character name key produced a good degree of matching with a 2 percent error rate. The quality of VAT data was relatively high. However, names were not alwayscorrect, reflecting the fact that having precision of names is not essential to VAT operation.

Records from the 1989 Census of Employment were matched against records on the CSO's VATregister, using VAT numbers supplied on the Census form. After automated matching only halfthe Census records were matched. VAT numbers were often missing or invalid. Also many deadunits had remained in the database although they had been subject to a census.

Units were missing from the VAT register because they were below the VAT size threshold orbelonged to the public sector,and hence were VAT-exempt. Special arrangements would have tobe made to record public sector units on the register.

Discussion:There was some discussion of the implications of extending the exercise to the whole of the UK.While considerable clerical checking was required in the Northern Ireland exercise, only one staffyear would be allocated to the full UK job. Duplicates arising from failed matching would beaccepted on the assumption that they would constitute no more than 5 per cent of the population.Matching between the Census of Employment records and the VAT register could be improvedby putting more emphasis on Census collection of the VAT number. Previously VAT numberquality had not been a concern as the number had only been collected to resolve occasionalqueries.

The question was raised whether the error rates in reporting identification numbers (to be used forcross matching) was similar across all data sources and all countries.

It was suggested that it was not necessary to worry unduly about devising a units structure for therecording of general government and other units which behaved differently from other economicunits and for which any pragmatic structure would be satisfactory. It was noted that such unitswere included in regular economic collections, for example of employment and of finalconsumption expenditures.

Page 15: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 15 -

Paper 3:Undercoverage of the ABS's Business Register over the 1980s

Presenter: Geoff Lee, Australian Bureau of Statistics

Presentation Summary:Until the end of 1989, the ABS had accessed paper copies of group employer records from theAustralian Taxation Office to update the business register. Thus, if the units were accidentlyomitted or deleted from the register, it had not been possible to re-establish them by checkingagainst the existing stock of group employers. Since then a system had been established to obtainan electronic file of records of all group employers. An analysis of the units missing from theregister indicated that the 'leakage' accounted for about 4.5 per cent of all units. Furthermore, ithad been fairly constant over time, to some extent affecting all industries and employment sizes.

The change from the paper to electronic system of accessing the administrative records wasaccompanied by a delay of several quarters in updating the register. It was also recognised thatthe normal time lag between the birth of a business and its placement on the register had beenchanged by the new system. Therefore modelling had been carried out to impute the birth ratesthat would have been expected on the register had there been no changes in the system. Theinvestigation indicated that an exponential decay model fitted well, and suggested that businesseshave a half-life of 5 years, and that undercoverage due to leakage had remained fairly constantover the years.

The model was used as the basis for determining the appropriate adjustments to be made tosurvey estimates to allow for leakage and delays in the birthing process.

Geoff concluded by noting that to minimise future risks, a quality control program should be putin place and there should be regular measurement of timeliness, duplication, undercoverage, nilsand defuncts (inactive units), and miscoding.

Discussion:In response to a suggestion that constant undercoverage is not a major problem and thatresources could be better spent fixing problems than in making quality measurements, Geoffstated that a complete quality improvement program should ensure that resources were spent inthe most effective way possible. He also pointed out that it was not known before the analysisthat undercoverage was in fact constant.

In response to a question Geoff said that timing changes in the birthing of businesses would resultin artificial lumpiness in estimates.

Page 16: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 16 -

Paper 4:Quality Control and Demographic Analysis

Presenter: Philippe Pommier, INSEE, France

Presentation Summary:The French SIRENE business directory is updated with information on legal and legal local unitsFor demographic analysis, good estimates of active enterprises and establishments are required.To get these estimates, a 'counting base' is derived from the original register information. Thecounting base now spans 1986 to 1991.

Firstly, legal units without economic activity are omitted. Secondly, adjustments are made forthose cases where more than one legal entity is associated with a single enterprise. Legal entitieswith only occasional or subordinate activity (such as house renting) are also omitted.

A more difficult problem is to allow for the closure of businesses, for which there is no automaticnotification. Estimates of business closures are made from the annual business survey. Instead ofimputing for non-response as in the past, non-respondents are carefully checked to see which areactually business closures. The information derived in this way is used to adjust the counting base.Different closure rates are derived for different sectors and size classifications.

The overall effect of the adjustment results is that an initial count of 3.5 million units on theregister is reduced to an estimate of 2.2 million live enterprises. However, the adjustment is ofthe same order as movements over time in the counting base itself, emphasising the need forregister quality control.

Discussion:Philippe confirmed that information to define a legal unit as non-productive is available on theregister itself.

Jean Ritzen asked which population is used for survey frames. Philippe responded that it dependson the survey. Usually it is the list of units that have all non-productive units excluded. But asurvey of investment may wish to include extra units such as those who are capital owners but notoperators.

In response to a question by Geoff Lee, Philippe said that the correction factor can move by aboutone percentage point. The total correction figure is about 10 per cent.

Page 17: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 17 -

Paper 5:Business Register Quality: Measurements and Management: Summary of Survey Results

Presenter: Michael College, Statistics Canada

Presentation Summary and Discussion:The survey was designed to see what quality measures are available within agencies, to providebenchmark measures for year to year tracking, to provide a means of comparison betweenagencies, and as a stimulus to thinking about what developments are required in this area.

In discussing the results of the questionnaire, it was clear that some questions are particularly hardto answer, for a variety of reasons. Insufficient quality monitoring procedures may have beenestablished. The institutional environment of an agency may not be reflected in the way that aquestion is framed. Clarification may be needed as to whether some measures relate to idealobjectives or current accepted practices (for example, the definition of certain types of unit beingout of scope). Therefore some questions are not particularly suitable for interagency comparison,but may be of considerable interest for tracking within an agency over time. In other cases, it isuseful to agencies to see which quality measures are considered important elsewhere.

Overall, quality measurement does not seem to play a large role in register improvement practices.This raises the question as to whether more quality measurement would be an efficient use ofscarce resources, and whether there should be a core of measurements that all agencies shouldstrive for.

It was noted that quality measurements can be used in negotiating with other parts of statisticalorganisations with respect to resources, performance, etc.

Paper 6:Total Quality Management at the United States Bureau of the Census

Presenter: Thomas L. Mesenbourg, US Bureau of the Census

Presentation Summary:The Census Bureau has introduced a program based on the concepts of Total QualityManagement.

In planning this program it was seen that a number of barriers would have to be faced. Previousattempts to introduce some Deming principles had failed because the approach had been too 'topdown' and there had not been enough follow up on initiatives taken. Staff were aware of problemsrelating to lack of teamwork, protection of their own 'turf', general resistance to change, and poorcommunication.

Page 18: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 18 -

To emphasise that the program was relevant to the organisation, the name Census QualityManagement (CQM) was adopted. The basics of the program were: everyone has customers;quality data is needed from suppliers; it is the customer's right to define quality, and continuousmeasurement is required; continuous process improvement is needed; and the process involvespeople.

The implementation of the program had several components. A vision statement and qualitypolicy have been developed. Changes in management behaviour are expected - managers shouldsee themselves as leaders. Organisational changes have been made, such as the appointment of aQuality Coordinator. Everyone has been trained in the principles of CQM. The organisation'sstrategic plans have been integrated with the quality plans.

Lessons learned include the importance of involving managers in training, and the importance ofgetting some action started.

Discussion:Ron Bosecker asked about the measurement of success and progress. Tom said that measurementwas difficult in some respects, but was not of great concern. It has been explained to staff thatsome failures could be expected and that the teams concerned would not be treated harshly.However, it is necessary to find ways of recognising or promoting the successes.

David Dodds asked whether the suggestions coming forward were old ideas finding an avenue forthe first time, or whether they were new ideas. Tom said that they mostly seem to be the former atthis stage, which is consistent to the recurring statement that there have been communicationblocks in the past.

George Sarossy referred to ABS experience, noting that the hardest part is to get people torenegotiate their respective roles and change hierarchical structures. Tom said that there was notyet a clear idea of the changes needed in the organisational culture in the Census Bureau.

Page 19: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 19 -

Session No. 3SPECIAL UNITS IN THE BUSINESS REGISTER ORGANISATION TO HANDLELARGE BUSINESSESSession Leader : Australia

Session Leader: John Struik, Australian Bureau of StatisticsSummary of Introductory Remarks:

The session was to examine the experience of established large business units and the innovationsbeing considered by agencies who have recently set out on this path.

The three papers are from the Netherlands, Canada and New Zealand and the emphasis in thepapers is clearly on innovations. At previous Roundtables Australia and New Zealand havereported on the establishment of large business units (LBU's) and their procedures and progress inprofiling large businesses.

Although Australia was not presenting a paper this time, John considered it appropriate to statethe Australian position which was that:-

. LBU's were set up to profile the top 600 businesses.That had been completed and the second round of profilingwas well underway with a number of businesses having beenprofiled more than twice.

. LBU's act as the central contact within the ABS for thesebusinesses and undertake the despatch and collectioncontrol collection, but not editing activities for only theannual production collections and

. The ABS Classification and Standards area has theresponsibility for standardising data item definitions andfor overseeing the reporting arrangements for these largebusinesses.

The objectives of the LBU operations were to achieve an increase in the quality of statistics, areduction in respondent burden and an increase in the standing of the ABS in the large businesscommunity.

Page 20: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 20 -

To further set the scene, John cited extracts from two of the papers:-

"Experience during a number of years has shown that despitecoordination principles and directives, in practice asatisfying degree of coherence between data from differentsurveys was not obtained and that large and complexbusinesses accounted for a major part of the problem"(Netherlands), and"The objective of this project is toimprove the quality of our industrial and financialstatistics through a better and more integrated method ofreporting by large businesses" (Canada).

The three papers all cover to some extent the following:-. Profiling. Integrated reporting arrangements for large businesses,and. Other activities beyond profiling to improve the accuracy and consistency of statistical reporting.

It is the latter where the innovations lie and each agency proposes something different and veryinteresting.

PAPER 1:LARGE AND COMPLEX BUSINESSES IN THE DUTCH SYSTEM OF ECONOMICSTATISTICSPRESENTER: Ad Willeboordse, Central Bureau of Statistics of the Netherlands.

Presentation Summary:Ad commenced his presentation by stating that he felt progress could be summarised by the titlewhich referred to "large and complex systems in the Dutch System of Economic Statistics"whereas in the past we probably would have referred to "in the Business Register." This typifiedthe wider scope of the process being undertaken.

While describing the units (enterprise and kind of activity (KAU)) in place in the Dutch system,Ad put in a plea for greater uniformity and coordination of units terminology.

Statistical coordination is a distinct issue in the Netherlands where some 300 monthly, quarterlyand annual surveys are undertaken by fourteen separate departments.

There was recognition as far back as 1975 that there was a need to treat large and complexbusinesses (LCB's) separately, so that data could be consistent. At that time, a special officer wasappointed for LCB's but this was seen to be a somewhat passive response in that it sought to onlyput some order into an otherwise uncoordinated system.

Page 21: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 21 -

In 1989 a special unit (5 staff) was formed with the principal function of determination of theappropriate statistical structure for some 170 conglomerates.

The experience has been that the determination of appropriate structures and the attendantspecially negotiated collection arrangements are not altogether sufficient to guarantee that correctand consistent data will be obtained for all economic statistical collections involving the largebusinesses. It has been found that businesses do not always advise all relevant changes and thatthere are not always uniform interpretations of the arrangements within the CBS as well as withinformants.

An ideal solution would be to centralise all surveys of LCB's but this has not been achievable.The measures taken to address the problems have been the introduction of a "TOP100 Data Base"and the extension of the program of production statistics with an annual survey on holding units.

TOP 100 DATA BASEWithin the TOP100 data base, data from a variety of economic surveys for large businesses andtheir KAUs are stored. It is now in the last stage of design and will come into full use by the endof 1991. The data base contains the legal and KAU structure for the 116 enterprises, 1350 KAUsand 3750 legal units designated as large businesses. For data recorded in the TOP100, priority isgiven to those data that play a central role in linking production and financial accounts and also todata that are particularly useful in consistency analyses. Longitudinal relationships and causes ofchange are also registered. The data from various sources are analysed for consistency andplausibility - number of employees is the key variable for checking.

In summary, the positive effects of the TOP100 are:- more accurate and up to date unit structures in the CBS,- the opportunity for resolution of misinterpretations by respondents and surveying staff, and- the opportunity to sharpen arrangements for future reporting.

HOLDING UNITS

The detailed experience of the TOP100 database will be reported on at the next Roundtable.The "holding" unit industry is being surveyed for the first time in 1991, with a view to sorting outproblems anticipated with internal service payments and overhead costs. There are perceivedbenefits for arrangements with businesses on statistical and reporting structures.

Page 22: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 22 -

PAPER 2:LARGE ENTERPRISE STATISTICS PROJECT (LESP)

Presenters: Jacob Ryten and Dick Collins, Statistics Canada

Jacob introduced the session by commenting on the rationale for why particular attention shouldbe paid to large businesses. This included:-

. as a group large complex businesses (LCB's) while small innumber contribute significantly to aggregates and in certaineconomies (Canada included) for an inordinate "chunk",

. they tend to be multi activity oriented, and if we can'tidentify that we run the risk of distorting industrystatistics,

. they are typically geographically dispersed with consequenteffects on regional statistics,

. there are undesirable (lack of) response consequences forthem,

. they are recipients of many questionnaires and management ofthat workload is required

. not having the composite picture can lead to respondentincoherence and difficulties in resolution,

. without a detailed understanding, it is not easy for us orthem to understand the overall picture (we need some tools toassist them to provide coherent responses.)

Jacob also commented that the earlier point made by John Struik on increasing the standing of thestatistical agency in the large business community was a particularly important one.

20 staff years have been allocated over the next three years to advance the LESP, with theobjective of improving quality of industrial and financial statistics through a better and moreintegrated method of reporting by large businesses, initially confined to surveys of annualproduction, capital formation and financial statistics and their quarterly derivatives.

Page 23: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 23 -

Although there is nothing sacrosanct about it, a target number of 300 large businesses has beeninitially settled on, although experience and an assessment of benefits will ultimately dictate thefinal number.The planned innovations are:-

. improving methods of liaison with large businesses,

. special reporting arrangements for all annual surveys forthem, and

. testing a Consolidation Questionnaire (CQ).

Dick Collins spoke on the development of the CQ which aims to ensure the complete andunduplicated coverage of large businesses for production statistics data which are consistent withthe consolidated financial statements by obtaining control units for each unit in the accountingstructure.

There is great variability in how large businesses manage themselves. Consolidation is thecommon point. Like the Netherlands experience, the point was made that correct unitsidentification does not in itself ensure that correct information is reported.

The CQ is designed to extract information from the consolidation worksheets of business andincludes:-

. Operating Revenue

. Operating Expenses

. Capital Expenditure

. Operating Profit

. Amortisation & Provision for Future Expenditure

. Depreciation & Depletion and

. Closing Inventoriesfor all the recognised units in the organisational structure of a business (including such units asCorporate Head Offices and Research Divisions which most surveys don't cover).Other perceived benefits of the CQ are that it will provide:-

. a meaningful method of defining operating and hencestatistical units, and

. a means to link financial and production statistics.

Page 24: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 24 -

PAPER 3:COMPLEX BUSINESS UNITS IN THE NEW ZEALAND SYSTEM OF ECONOMICSSTATISTICSPresenter: David Archer

David made the point that the Complex Business Unit (CBU) about to set up has the sameobjective as that of the Netherlands, but with some differences because of concepts and thelimitations of current New Zealand financing unit statistics.

New Zealand has been profiling for some years and had decided it was not effective to continuewith the original high resource levels because of resource considerations and a tendency for "overprofiling". There are currently 2 staff in the Large Business Unit (LBU) who update structures ofall large groups twice a year primarily by a postal structural collection but also utilising surveyfeedback and the annual (location) postal update.

David commented that Business Directories don't detect all the problems. You can have all thestructures statistically pure, but there are intricacies that are not apparent from the structuresthemselves.

The New Zealand CBU will centralise efforts, not at the "Enterprise Group" level like Canada andNetherlands but at the Enterprise level. In the longer term, they expect to move to considerationof the enterprise group and involving the Balance of Payments collections which currently sufferfrom a lack of "integration".

The CBU will be involved, for 400 enterprises, in the centralised dispatch, follow up and editingfunctions of all economic surveys data that are integratable and that use the Business Directory asa population frame. The 400 enterprises encompass all multi accounting unit enterprises, localauthorities with trading enterprises, Government Departments, and those GovernmentDepartments changing into State Owned Enterprises. The CBU will also reconcile structures andoutputs on a quarterly basis and maintain the appropriate structure on the Business Directory.

An aspect of the setting up of the CBU was that resources are being funded by transfer fromexisting subject matter collections, and not by the creation of new positions.

DISCUSSION

Before opening discussion, John Struik commented that although large business unit equivalentscontinue to address issues identified previously such as scope and resourcing of profiling, theimportant point emanating from the three very interesting papers was that the determination ofcorrect structures was not sufficient to guarantee appropriate linking of financial and productionstatistics.

Page 25: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 25 -

Following a response to a question on the Consolidation Questionnaire content, George Sarossypointed out that if we thought the esoteric arguments that had been conducted on productionunits were conceptually demanding, then issues affecting financial units were going to besignificantly more so.

He pointed out:-

. taxation/security commission/corporate regulating bodies eachof which have different requirements across and withincountries,

. creative accounting mechanisms particularly relevant tofinancial units, and

. the degree of off balance sheet, off shore etc problems tobe addressed.

John Perry in introducing a question reiterated Ad's earlier plea for uniformity in unitsterminology. Dick Collins' response to John's question as to whether the ConsolidationQuestionnaire was building in a delay in reporting was that although businesses in Canada have 45days to provide consolidated annual data this was about the same delay as that in providingproduction data.

Jacob Ryten offered the suggestion that we perhaps should be looking at multiple classificationsrather than a unique one. This was in response to a discussion on the treatment of CorporateHead Offices (CHO's). Jacob posited that we may want to view a CHO as being part of itsprincipal industry under one set of circumstances, (when looking at overall industry effects) yetnot under another, ("eg. you don't want miners in Sydney").

John Perry noted that the papers didn't substantively address respondent burden issues. GeorgeSarossy commented that burden was indeed an issue and the emphasis had moved recently fromthe burden on small businesses to that on the large ones. Looking at numbers of forms is onething but burden is also related to the closeness of the form to record keeping practices, thecloseness of the statistical unit to the accounting structure, and sending forms at the appropriatetime. Details of time taken to complete questionnaires are being maintained in Australia and a subsample of these will be followed up to identify further savings options. Roger Bugenhagenreferred to a record keeping practices survey undertaken in the US and offered to circulate thereport to those requesting a copy.

Jacob recognized the difficult task that confronted decisions that need to be taken in trade offsbetween respondent burden and statistical purity. It really was an agency to agency decision.

Page 26: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 26 -

Session No. 4SESSION 4: STATISTICAL DATA FROM BUSINESS REGISTERS

SESSION 4.1 GEOGRAPHYSession Leader : New Zealand

SESSION LEADER: David Archer, Department of Statistics, New Zealand

PAPER 1: GEOGRAPHIC CODING OF BUSINESS REGISTERS: A REPORT ON INTERNATIONAL PRACTICES

Presenter: David ArcherPresentation Summary:David introduced Session 4.1 by presenting his paper which summarised the results of a survey ofgeocoding practices of business registers by countries participating in the Roundtable. Hereported that the main findings were that all countries used geocodes but at different levels ofcommitment. The business unit which was coded was in all cases the single physical location andthat the spatial units to which locations were coded ranged from city/town/district level to ones asfine as streets.

David said that the conclusions presented in the paper had prompted him to consider whether thelevel of commitment to geocoding was appropriate in the context of other priorities which had tobe handled in maintaining a business register. He proposed to delegates that this was an issuewhich they may wish to consider in the context of the objectives for their own business registers.

Jean Ritzen clarified the geocoding practices which were in use in the Netherlands. These werethat the business register itself was not used as a source for small area statistics. Geocoding ofbusiness units on the register was undertaken only to ensure that survey samples selected from theregister were geographically diverse. Respondents selected in surveys were required to providedata which was disaggregated geographically.

Delegates from the United Kingdom cautioned that the concept of geocode was one that was notentirely clear because it had different meanings depending on the requirements of respective users.For this reason coordinates of latitude and longitude were the best geocodes but in practice,mainly because of cost considerations, something less precise had to be used.

Karsten Stetkaer reemphasised the importance in Denmark of the used of geocodes to the level ofstreet or road. This enabled compilation of statistics on important topics such as night and daypopulations of localities.

Page 27: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 27 -

George Sarossy advised the meeting that in Australia some work had been done to assess thefeasibility of geocoding with coordinates of latitude and longitude. Coordinates could be obtainedeasily from a hand-held decoder which interpreted satellite signals which pinpointed the locationof the decoder.

Presentation 1: TOPOGRAPHIC INTEGRATED GEOGRAPHY ENCODING REFERENCE

Presenter: Thomas L Mesenbourg, Bureau of the Census, USA

Thomas reported to the meeting on the Bureau of the Census', Topographic IntegratedGeography Encoding Reference (TIGER). This was a computerised geocoding aid which hadbeen constructed by the Bureau to automate mapping of results of the 1990 Population Censusand to link results of the 1980 and 1990 censuses. TIGER would be used to code some 4.3million establishments to street block level. This would be undertaken in the context of the 1992Economic Census and would, subject to disclosure considerations, enable release of results atuser-specified small areas. A Congressional District was an example of such an area. Anotheruse of TIGER was that it was providing geocodes for all collections and this enabled the linkingof all datasets.

Thomas illustrated his presentation by screening a video which showed the methodology used todevelop and build TIGER.

Presentation 2: MAPPING FROM THE BUSINESS REGISTER BY THE AUSTRALIAN BUREAU OF STATISTICS

Presenter: Ian Marshall, Australian Bureau of Statistics (Queensland Office)

Ian demonstrated a range of computer-generated maps being derived by the ABS from itsbusiness register. This work used MAPINFO software and was at a relatively early stage. TheABS intended to produce an industrial atlas compiled from maps from the business register.

Presentation 3:MAPPING FROM THE BUSINESS REGISTER BY THECENTRAL STATISTICS OFFICE

Presenter: John Perry, Central Statistical Office, UKJohn demonstrated a range of computer-generated maps being produced by the CSO using AtlasGraphics software. John emphasised that the most important part of a mapping facility was itsboundary file; the better the quality of this file the better the quality of the maps produced. TheCSO was using the maps as an editing tool for the business register by gauging whether, whenmapped, businesses were clustering where they should. Maps were also being used in discussionswith users to clarify their requirements for data for small areas.

Page 28: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 28 -

Presentation 4:PROCEEDINGS OF THE INTERNATIONAL STATISTICALINSTITUTE MEETING HELD IN CAIRO, 9-17 SEPTEMBER, 1991

Presenter: John Perry

John reported briefly to the Roundtable on the deliberations of the session on statistical geographyat the recent meeting of the International Statistical Institute. The centrepiece was a paper "TheUse of GIS in Spatial and Space-time Statistical Surveys" by Messrs G Arbia and M Coli, twoItalian academics. The paper discussed the concept of Geographic Information Systems. Johnobserved that the concept of GIS needed further development. The presentation of the paper wasrather long and left insufficient time for any meaningful discussion. John said that notice wasgiven of a forthcoming US paper on statistical geography and he thought this could proveinteresting.

Page 29: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 29 -

SESSION 4.2SESSION 4.2 LONGITUDINAL AND DEMOGRAPHIC ANALYSISSession Leader : CanadaSESSION LEADER: David Dodds, Statistics Canada

Summary of Chair Remarks:David opened the session by advising delegates that there were four papers and threepresentations in this session. The two Canadian papers would be presented jointly.

PAPER 1

LONGITUDINAL ANALYSIS USING THE BLS BUSINESS REGISTRY

Presenter: Brian MacDonald, Bureau of Labor Statistics, USA

Presentation Summary:Brian said that the principal use of data from the BLS business register had been to present point-in-time statistics in publications and press releases. Not much work had been done onlongitudinal statistics but a recent redesign of the business register now facilitated longitudinalstudies of micro units.

Brian said his paper described progress to date on a research project on longitudinal analysis ofsome 400,000 establishments involving about 6 million employees. These units were drawn from6 small- to medium-sized states.

David Birch, a US researcher, had published a number of studies which argued that employmentgrowth in the US over many years could mainly be attributed to growth in the number of smallbusinesses. Mr Birch's conclusions had been contentious and the BLS was hoping ultimately toshed some light on the subject.

Brian described the record linkage procedures being used and the criteria used for classifying unitsas births, deaths or continuous.

Ad Willeboordse questioned what weight was assigned to the factors taken into account inclassifying units which were taken over or merged. Brian undertook to let Ad have a copy andremarked that even if there were a change of ownership of the establishment but most of its othercharacteristics remained the same, the unit could be classified as continuous.

Brian referred delegates to the tables of results included in his paper and highlighted that the deathrate of units appeared to be 3% per quarter, the birth rate about 4% and net employment growthwas 118,000. He cautioned delegates that drawing conclusions from the data in respect ofagricultural industries was not easy because the BLS business register did not cover allagricultural units.

Page 30: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 30 -

PAPER 2:

DEVELOPMENT OF LONGITUDINAL PANEL DATA FROM BUSINESSREGISTERS:CANADIAN EXPERIENCE by John Baldwin, Richard Duprey and William Penner.PAPER 3:PROPOSED DEFINITIONS AND CLASSIFICATIONS OF LONGITUDINALCONTINUITY AND CHANGE FOR GLOBAL ENTERPRISES AND ENTERPRISES

Presenter: Richard Collins, Statistics CanadaPresentation Summary:Richard commenced his presentation with the observation that he supported the conclusions in thePenner paper. The work being done in this field in Canada was capitalising on the longitudinalelements of survey editing which was already undertaken. No additional data had been collectedfor the purpose of these studies.

What had become clear during early work was a need for stable concepts of units, ownership ofunits and changes in units.

The Penner study had shown that there was a need for flexibility in definition of businesses. Theevents which longitudinal studies were trying to measure had an impact on which view, and hencedefinition, of a business was most relevant. Penner had also shown that the greater the range ofdata which was linkable from one period to the next, the greater the value which could be derivedfrom longitudinal studies. The ability to track at several levels, units in a business, was importantin longitudinal studies, but can significantly complicate maintenance of records to support theanalyses.

Turning to his own paper Richard showed that for his purposes the enterprise and globalenterprise were being used for longitudinal tracking. He highlighted the definitions of continuityand structural change which he had adopted for use in his study and described how these could bedetermined in the cases of enterprise and global enterprise units.

Phillipe Pommier commented that choices of the points-in-time across which longitudinal studieswere conducted would have a major bearing on the results of studies. He said the outcome of astudy across a quarter could be very different from one across a year. Richard noted that hisdefinition of continuity was based on a unit existing at the ends of both periods.

Page 31: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 31 -

PAPER 3:NEW ZEALAND BUSINESS DEMOGRAPHY DATABASEPresenter: David Archer, Department of Statistics, New Zealand

In opening his presentation David contrasted the situation in New Zealand with that in the USBureau of Labor Statistics. David said that New Zealand's work had now progressed to thedevelopmental stage. The Department of Statistics had made some limited releases of data aboutactivity units from its Business Demography Database. These releases had carried caveats andwere aimed at prompting discussion with and comment from users.

Four key observations could be made from the New Zealand experience. There was a need for anacross-the-board consistent source of business births. This was obtained from VAT registrations.There was a need for a source to identify deaths and this was obtained from the February surveyof the business directory. There was a need for information about changes in businesses and therewas a need for clear definitions of what constituted a birth and a death of a business.

David referred delegates to the tables in his paper and said that notwithstanding the caveatsapplying to this data at this stage users were still prepared to purchase it. David's expectation wasthat this data would raise significant revenue. David also said that tabulating the data in thismanner had indicated some faults in the business register and in this regard there was a qualityimprovement dimension to the results.

On a final note David observed that the higher up the hierarchy of business units one chose tomake ones longitudinal studies the greater the complexity of the concepts and definitions neededto support the analysis. He noted too that duplicates in the units studied was a serious problem.

General Discussion:Barbara Slater commented on the importance of the integrity of the lowest level of units inbusinesses which were subject to longitudinal analysis. These were the building blocks uponwhich the studies depended.

John Perry advised that in the UK data availability meant these studies could only be made at theenterprise level. This was a slow process in any case because the studies had to be based onanalyses of the published accounts of the largest 2000 businesses.

Keith Perry enquired whether presenters had encountered any fundamental problems in theirstudies such as changes in methods in compiling of administrative sources. Brian MacDonaldreported that the US Bureau of Labor Statistics had checked for this. He emphasised the need forvigilance in this regard.

Page 32: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 32 -

Jacob Ryten observed that there was a certain arbitrariness in choosing 50% as a measure forownership and control of businesses. He said he thought it was important to make an assessmentin whether small changes in this ratio would have a significant impact on the results of longitudinalstudies. He also enquired what affect would there be from businesses which restructured over thecourse of a long time after a merger or takeover. Phillipe Pommier reported that in France suchbusinesses were flagged on the business register. While this flag was in place the business couldnot be accessed until all the changes arising from the restructuring were reflected on the businessregister. This could take up to two years and in this period special reporting arrangements weremade to collect data.

Chairs summary of session discussion:David concluded the session by noting that in many countries there was some capacity forlongitudinal analysis of business life cycles as a by-product of maintaining a business register.Discussion at this Roundtable had shown that the issues surrounding longitudinal studies werecomplex ones. The selection of level of unit for study was of fundamental importance and thepapers had discussed both establishment and enterprise units. Another crucial aspect was theneed for precise rules and definitions of births, deaths and continuity to underpin longitudinalstudies.

Page 33: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 33 -

Session No. 5ADMINISTRATIVE DATA FOR REGISTERSSession Leader : USBC

ADMINISTRATIVE DATA FOR REGISTERS

Chair: Roger Bugenhagen, US Bureau of the Census

Roger opened this session by stating that the major aims of this topic were to learn how countriesuse administrative data for registers, whether this usage has changed over time, what evaluationstudies have been undertaken and the relationship that we now have and should be developingwith the agencies responsible for this administrative data.

The US use administrative data for compiling mailing lists and for the provision of data for smallestablishments. This enables the Bureau of the Census to address issues such as response burdenand cost while at the same time the use of taxation data gives better quality information.Administrative data is also used to check the completeness of large businesses. Roger commentedthat the US were trying to foster better relations with the tax agencies and were even trying tohave these relationships established legally.

The US was also using administrative data to assist in producing reports on minority businessese.g. information on businesses owned by females. The general tenor or philosophy in the US is touse administrative products more and to mail a little less.

Paper 1Registers, Administrative Records, Statistical Agencies &Revenue Departments

Presenter: Jacob Ryten, Statistics Canada

Presentation Summary:Statistical agencies are increasingly using revenue departments to obtain information aboutbusinesses and are spending more effort in analysing this data. It is therefore prudent to developmeasures which ensure access to these records is retained.

The relationship that exists between statistical and tax agencies will differ from country to countryand there are no easy or universal answers on how agencies can maintain and improve thisrelationship. .

Page 34: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 34 -

A major finding in the Wilk report on the functions and efficiency of the statistical unit in the RCTwas that the tax office could no longer claim that its only function was to collect tax. Theyprovide information that is vital to other agencies. The tax office does not consider itself liable forthe quality of this information and are unaware of how it is used. Therefore, by convincing thetaxation department to adopt statistical concepts and standards, the dependence of the tax officeon statistial agencies would increase.

Martin Wilk's report recommended that the taxation department use more statistical techniques,that staff should be exchanged between the two agencies and an independent Register Unit shouldbe set up under legislation. This Unit would be the only gate through which businesses couldobtain a registration number. The Unit would be self-financing. The report has been shelved butthe issue may be resurrected in the medium term.

In conclusion Jacob stressed a strong need for co-operation between statistical and taxationdepartments and that the grounds for co-operation should be extended. Whilst it would beimpossible to completely redress the imbalance between taxation and statistical departments, it isimportant to find ways to reduce it.

Discussion:The objections to the report came from the civil service. The Registration Unit was a replacementnot an addition to other registers. There were concerns that if Canada had a single register,application for remittance may go to the wrong agency. The register would serve the Tax andStatistical agencies. The sale of the lists would depend upon the administrative agreements but itwould be preferable for the Registration Agency to sell the data. Jacob said that there would beno feedback to the tax agencies on information gained by Statistics Canada.

Karsten advised that in Denmark since 1988 the VAT and PAYE systems have a uniformcompany number. Jacob said that company securities legislation was Province based in Canadaand would be too difficult to co-ordinate with the Registration Unit.

John Struik said that the Tax authorities in Australia used statistical techniques and standards intheir auditing processes. John Perry commented that by setting up a separate register unit theauthorities could be concerned with risks to revenue collection. Jacob said that therecommendations came at a favourable time in Canada (the introduction of VAT). Thesefavourable circumstances have not disappeared; VAT is in need of change while PAYE will needto be redeveloped. If Statistics Canada can promote a co-ordinated development of these systems,then the establishment of the Register Unit could be examined at the same time.

Page 35: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 35 -

Paper 2Use of Administrative Records to Maintain Business Survey Frames

Presenter: Keith Sykes, Department of Statistics, New Zealand

Presentation Summary:The most important source of administrative records to maintain the Business Directory in N.Z. isGST data obtained from the Inland Revenue Department. These records are used to birth newbusinesses, and to assist with directory maintenance. The Statistics Department has on-line readonly access to the GST registration data base.

Prior to introduction of the GST, the Statistical Department used a number of sources to updatethe Business Directory (PAYE, licensing lists etc.). Information on balance of payments wasobtained from the Reserve Bank. With deregulation, surveys now need to be undertaken to obtainbalance of payments information.

The major benefits of moving to a GST source were that coverage increased by 30 per cent, itallowed the building of a common business directory, improved systems for monitoring responseburden, reduced directory resources and established links with the taxation office by usingrecognition codes.

New developments included the merging of the Business and Agricultural directories, the use ofnew classification codes and the allocation of IRD number to each enterprise. There is also a planto drop the GST as it is too restrictive and replace it with the IRD's new client registrationsystem. There are plans to have each business survey use the business directory. The ability of theIRD to provide financial information is also being investigated with a view to reducing respondentload.

DiscussionSome groups registered for GST voluntarily to get tax credits. New Zealand collected informationfrom the IRD under the authority of the Statistics Act. Michael College asked whether financialinformation would be available from individual taxpayers. Keith said that financial informationfrom administrative sources was still being investigated but it may not be timely enough forstatistical purposes. Information on private individuals would not be sought.

A proving questionnaire was sent to businesses as information from the GST form was incompleteand the tax office would not include any other information on the form. In addition, New Zealandneeded to sell the information on the Register and the Statistics Department does not own thedata until it is confirmed or updated by a Statistics Department collection. John Perry askedwhether any analysis had been undertaken on the number of businesses below the threshold leveland what effect these businesses had on survey information and longitudinal studies.

Page 36: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 36 -

Keith said that grossing up could be undertaken once quality was known. Jacob commented thatthe threshold level would define the scope of the collections. However, it was noted that majorchanges to the threshold level could cause problems for surveys.

Paper 3Australian Use of Administrative Data

Presenter: Leon Pietch, Australian Bureau of Statistics.

Presentation Summary:The Australian Taxation Office (ATO) is the major external source of coverage to update theBusiness Register and this practice will continue.

The Australian system is dependant upon information obtained from the PAYE "business line" ofthe ATO. The Business Register therefore only records employers (except for the Agriculturalsector). There is an undercoverage of about 7 per cent in the non-agricultural, non-governmentsector. This undercoverage is not evenly spread across industries because non-employers are moreimportant in industries such as construction and retail.

Leon mentioned some of the major problems experienced in Australia in maintaining the Register.These include the high costs of proving surveys and recent policy changes in the ATO to converttax stamp payers to group employers. This has highlighted problems of duplication and leakagefrom the system.

CATO (Clients of the Australian Taxation Office), a new client data base, is being developed bythe ATO. This system will provide coverage for all "business lines" for the first time> It shouldprovide a suitable source for non-employer units, and it will record information such as industrycode, employment and principal location.

The Australian Securities Commission is a new Federal body which subsumes the roles that oncebelonged to the State Corporate Affairs Commissions. The ABS hopes to make use of the newAustralian Company Number (ACN) which is issued by the Commission.

Discussion:Keith Perry asked whether information on individuals could be obtained from the CATO systemand placed on the Business Register. Leon said that there would be a distinction betweenindividuals and businesses, and personal information would not be available. On the coding ofinformation by the ATO, John Struik advised that he expected the ATO to move to the newindustry classification. John was optimistic that the quality of this information would be good asthe ABS would train ATO staff and the classification is a benefit to the ATO. John said that theATO were using statistical techniques in their audit procedures.

Page 37: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 37 -

John pointed out that the ABS did not have to convince the ATO that it was beneficial to useABS classifications and was therefore different from the Canadian experience. Roger closed thediscussion by stating that it was an encouraging admission by the tax agency that they had"clients" and hoped other agencies had a similar outlook.

Papers 4 and 5Use of Administrative Records for Business RegistersEvaluation of Coverage of Business Activity by Administrative Data

Presenter: John Perry. U.K. Central Statistical Office

Presentation Summary:The VAT system has been in place in the U.K. since 1973 and has been used as a statistical sourcesince 1975. John said that VAT operated similarly to New Zealand's GST. It has a cut off of35,000 pounds with some exempt traders (wider than in NZ). Coverage had not been a problem inthe past but was becoming more so, particularly as there has been increasing demand forinformation on the service trades.

Relations with H.M. Customs are extremely good. VAT data would be in the statistical systemwithin two weeks and penalties ensure that coverage problems are minimal. Deregistration cantake 6 to 9 months as traders are included on the register if they have a tax liability. Quality of theinformation would depend on the data items e.g. address is excellent, the industry classification isbased on an old SIC and has limitations, while turnover is relatively good.

John said that a major concern for them is that authorities have increased VAT threshold levelsabove inflation. A recent decision could take as many as 80,000 traders off the Register. Thiswould cause problems for longitudinal analysis but other effects would be small.

John discussed the future developments in the area including feedback of information to VAToffices and the potential to obtain information on imports and exports. The VAT register is theone used for Intrastat, the EEC register for monitoring trade. John closed the discussion on thefirst paper by discussing the use of the other major administrative systems, PAYE and CODA.PAYE has good coverage but the names and addresses listed would not be as reliable as VATaddresses. CODA would be an excellent system for recording the self-employed on the Register,but legislation restricts this use.

John's second paper dealt principally with the data to create an ideal register which could besourced from VAT, PAYE and CODA registrations. This register would service the needs of theEmployment Department and the Business Statistics Office. John discussed three groups ofoptions which were considered. In discussing the various options John pointed out that the optionof using PAYE and CODA data is the ideal for coverage.

Page 38: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 38 -

Other options will by definition omit some businesses. John then discussed the problem with VATexemption level and how it would effect coverage problems.

Discussion:In general discussion Jacob asked whether companies register with more than one VAT office andif this would create a problem of duplication. John advised that it is normal only to register withone VAT office but companies which have separate accounts at different locations may register atmore than one VAT office. These duplications can be picked up through statistical enquiries.

Paper 6:The Use of Administrative Records for the CBS Business Register

Presenter: Jean H.G. Ritzen, Netherlands Central Bureau of Statistics

Presentation Summary:In the Netherlands, the use of administrative records started in the 1960s, when a central registerwas started on the basis of the records held by the Industrial Insurance Boards. These covered allfirms with paid employees. From 1978, the registers of the Chambers of Commerce and Industrywere also accessed because they gave coverage of non-employing businesses. Matching has to becarried out to avoid duplication. The two sources still miss some government units, which aresourced from annual reports, and some exempt organisations.

The Trade Register of the Chambers of Commerce has a legal basis. It provides publicinformation about firms and is an important tool of the Chambers in fulfilling their role ofproviding information about the regional structure and development of trade and industry.

The Register datamodel contains the legal unit, the firm (or economic unit), and the local unit.Units are recorded with a description of their activities, address, date of establishment and sizeclass. If a local unit is designated as a 'head office', it is supposed to carry the industry code of thewhole firm. But this rule is not always followed in practice, creating difficulties in interpreting SICcoded data of firms.

Businesses report changes in the information registered in the Trade Register. However, theclassifications used in recording the changes are administrative rather than statistical, e.g. in thedefinition of a birth or death of a unit. Additional information sometimes needs to be gathereddirectly from businesses. Overall, the data quality of the Chambers of Commerce Register is nothigh.

There are regular meetings between the CBS and the Chambers to discuss issues of concern.These include differences between apparently similar statistical series published by the twoorganisations.

Page 39: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 39 -

CBS has started a revision of the model of the business register in connection with a conversionto Oracle and new computer hardware. Possible new sources are also being examined, includingthe VAT register. Consideration is being given to joining with the Chambers of Commerce in theirquality improvement project. Uniformity in definitions and classifications and their implementationhave a major role to play in quality improvement.

Discussion:Jean confirmed that the matching of records is done on the basis of name and address. There areduplicates within the Chambers of Commerce records which have to be identified, and additionalinformation from the Industrial Insurance Boards has to be incorporated.

John Perry inquired about the low likelihood of using VAT information. In the Netherlands,confidentiality has been a problem. The possibility of integrating all registers is being examined,however, as is the use of information to be provided to Eurostat.

Paper 7:Administrative Data for Registers - The Danish Case

Presenter: Karsten Stetkaer, Danmarks Statistik

Presentation Summary:The Danish Central Register of Enterprises and Establishments (CER) was set up under an Act ofParliament. It has both an administrative and statistical role, and is administered by DanmarksStatistik (because they have the expertise). These two roles are legally clearly separated, butphysically the two registers are combined. Moreover, in setting up the statistical business register,no additional contact with respondents was permitted - all information had to be obtained fromadministrative data.

Data is primarily obtained from the Tax and Customs Directorate. Until April this year, Tax andCustoms were separate directorates with PAYE data coming from Tax and VAT data comingfrom Customs. Each had their own register, and there were difficulties in combining them. Thetwo directorates have now been combined and these problems will lessen with the establishmentof a Common Basic Register. However, a staff cut of 10 per cent was imposed at the time of theamalgamation, and the consequent pressures have resulted in some difficulties in liaising with thetax authorities.

Danmarks Statistik currently receives a computer file of changes to the Common BusinessRegister, and copies of registration, deregistration and change of kind of activity forms. Hardcopy forms are required to check kind of activity coding done by Customs and to determinewhether a new registration has more than one establishment.

Page 40: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 40 -

Registers of businesses are also maintained by the Danish Commerce and Companies Agency andfor the labour market supplememntary pension fund, but these are not used directly to update thebusiness register.

Discussion:Barbara Slater asked whether the Common Business Register would replace the two registers onwhich it was based. Karsten said that there would still be separate registers to carry items that areof specific interest only to VAT or PAYE.

Kaija Hovi explained that Finland had gone through a similar process of combining two separatetax based registers when two taxing authorities were combined.

Michael Colledge observed that there is a spectrum of experiences between countries. At one end,countries such as Canada and Australia have no linking key between their various administrativesources, while France, at the other end, has all authorities using a single registration system.George Sarossy noted that many countries are moving towards the French system, even thoughthis seemed impossible only four years ago.

Paper 8:French Experience in Statistical Processing of Administrations Files

Presenter: Philippe Pommier, INSEE

Presentation Summary:French experience so far suggests that two factors are important in using administrative data forstatistical purposes. Access to accounting records is very desirable. It is also very useful to havestatisticians within the administrative authorities to have input to the development ofadministrative procedures.

There are some typical shortcomings of using administrative records. The scope is often deficientbecause, for example, late declations are not included or the absence of a declaration does notresult in sanctions. Statistical classification of units, to industry for example, is likely to be poor.Information not directly connected with the applicaiton of the regulations is generally not reliable.File sizes may be so big that sampling of records is required.

If the administrators are not interested in helping to fix these problems, the statistical authoritiesare likely to have to undertake complementary work on their own to address the issues. Forexample, VAT thresholds mean that administrative data is not available for some small businesses.

Page 41: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 41 -

Discussion:There was some surprise at the idea that it is relatively cheap to survey units that are below aVAT threshold, because there are a large number of small units. Even if a small fraction issatisfactory for sampling, all units have to be included on the frame. In some cases, area samplingis required because there is no source to identify the small businesses to create a sampling frame.

Session summary

In summary, Roger noted that there have been definite and positive changes in this area since itwas last discussed.

Jacob Ryten questioned whether a greater dependence on administrative records was necessarily apositive change. For example, if tax authorities decide that less data is required for auditing, theymay stop collecting information that is currently also used for statistical purposes. Also,developments in EDI and funds transfer with banks is likely to make it more difficult to getadministrative data.

Roger suggested that the trends to date had been for improved cooperation and recognition of thestatisticians' needs.

George Sarossy said that the ABS had to lobby the Australian Taxation Office to retain thestatistical content of the taxation forms when Australia moved to self assessment of tax. It may bea good idea to outpost a high level statistician to watch out for emerging problems.

John Perry noted that the statistical office is more vulnerable when dependent on only oneupdating source. In the U.K., VAT thresholds had been increased without any consultation withthe statisticians. It would be good to have legislative support for data colleciton fromadministrative sources. But it would be prohibitively costly to return to depending on economiccensuses for all statistics.

Page 42: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 42 -

Session No. 6COMPUTER TECHNOLOGIES IN REGISTER OPERATIONSSession Leader : US Bureau of Labor Statistics

Computer Technologies in Register Operations

Session Leader: Mr Carl Lowe

Carl opened the session with a summary of its aims; ie, to share information on computingtechnologies used or intended to be used in agencies and to learn how users access directoriesthrough computer technology.

Paper 1. Electronic Data Collection in the US Economic Censuses, Roger H. Bugenhagen,US Bureau of the Census.

The paper presented by Roger outlined the history of the economic programs of the US Bureau ofthe Census. These comprise the 5-yearly Economic Censuses, and the current (monthly, quarterlyand annual) surveys which are generally less detailed but more frequent and timely than thecensuses.

Reporting of data on media other than questionnaires began in 1972 when a few large companiesprovided data on computer printout. In 1987 the US Census Bureau accepted reporting onmagnetic tape when ten retail companies, accounting for over 25,000 establishments, reported inthis way.

The Census Bureau has now embarked on a major program to provide for electronic filing of datain the 1992 Economic Censuses following expressions of interest from a sufficient number oflarge companies. The criteria set for this exercise (industry usage of EDI, homogeneity of thecompanies operations and the number of establishments operated by the company) are met by theRetail Trade Sector in the US and the large businesses in this sector will be targeted to commencethe introduction of EDI reporting in the 1992 Economic Censuses.

A standard transaction set has been approved and the Census Bureau are now proceeding topromote the facility to the Retail Sector, to establish a standalone computing facility (to protectconfidential data held on other machines) and to establish appropriate software. A small scalepilot test is expected to be undertaken early in 1992.

Page 43: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 43 -

It is expected that a combination of magnetic tape input and EDI will be utilised in the 1992Economic Censuses.

Paper 2. Changing Technologies Utilized in Register Operations, Ron Bosecker, USNational Agricultural Statistics Service.

Ron Bosecker outlined the activities in the US National Agricultural Statistics Service (NASS) toaddress issues associated with the development of a new list management system. Severalcommittees have been established to evaluate a suitable database for local area networkprocessing, identify the appropriate functions for the system and to investigate historical dataneeds for use in sampling, editing and analysis.

At this stage analysis of the cost comparisons between centralised mainframe and LAN PC basedprocessing for list management has indicated that distributed processing by State Offices isfeasible and can provide cost savings.

During the session Ron quickly surveyed participants to obtain information of the extent ofcentralised or distributed processing in their agencies. While current usage of systems varied itwas apparent that there was a move towards distributed processing in a number of agencies.

Paper 3. Name and Address Matching, John Perry, UK Central Statistical Office.

The paper presented by John described the current situation regarding name and address matchingin the UK Central Statistical office and the Employment Department. This process has beenunder review as a result of the move to develop a new business register and the probablerequirement to match large numbers of source records to the business register.

The review has included a study of data bases and database management systems. Also researchinto a range of available software tools able to be utilised for name and address matching has beenundertaken. Currently it appears that only one product satisfies the criteria set for the proposedregister. This product, SSANAME from the company SPL, provides tools that can be fitted to anumber of hardware platforms and offers a flexible solution to the name and address matchingproblem. One of the particular criteria considered was that the software needed to be compatiblewith the likely environment being considered for the UK register; i.e. the INGRES RDBMSrunning under UNIX.

Page 44: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 44 -

Paper 4. Current status of Computerisation of the Establishment Directory, Isao Mitake,Statistics Bureau, Japan.

Isao described the current status of the computerisation of the Japanese Establishment Directory(ED). The ED is produced and updated as a result of the processing of the Establishment Census.This processing is undertaken by contracted data processing companies with errors being referredto the prefectural office of the Statistics Bureau for correction. A paper based register of allestablishments is prepared from details available from enumerators in the Establishment Census.However, a magnetic tape based register is also compiled containing details (name, address,industry, number of employees) of only establishments with five or more employees. While thestatistical data from the Establishment Census is stored on a database, the ED is only maintainedon magnetic tape. This tape-based register is used for less than half the request for registerrecords; the other requests being for small numbers of records and being satisfied from the paperbased system.

Paper 5. Computing Technology Used in Register Operations at Statistics Canada: AReview, Michael Colledge, Statistics Canada.

Michael Colledge provided an outline of the current situation regarding the computing technologyused in register operations in Statistics Canada.

The Central Frame Data Base (CFDB) had been rebuilt in the period 1984 to 1989 and is basedon ADABAS with use of PL/I and NATURAL. Three databases are available; developmental,acceptance testing and the production database.

A distinctive feature of the database is its temporal capacity. It can provide a view of the contentof the database at any specified date since its inception.

Features of the CFDB system described by Michael included separate treatment by the system oflarge and small businesses to allow for the simple requirements in regard to small businesses, taskscheduling for clerks via a CFDB "workbench", the chronological updating of businesses on theregister and some of the restrictions this places on operations, quality control via acceptancechecking of updates prior to finally updating the database, and the availability of only "committed"or updated units to users.

Michael also outlined the automatic system which records businesses contacted in surveys whichdraw their list from the register. Future use of the system for response burden control is to beconsidered.

Page 45: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 45 -

The method of record linkage via the Revenue Canada payroll deduction (PD) account numberand business name was summarised. Names are encoded using a Statistics Canada routine,NSKGEN7. It is planned to strengthen this linkage procedure by the use of postcode, andpossibly by the use of size and industry attributes.

It was also mentioned that industry codes are automatically allocated to incoming sources recordsby matching industry descriptions to a file of recognised descriptions and codes. Where an exactmatch is found the industry code is automatically assigned to the register record. Two othersystems in use in Statistics Canada (the Automated Coding by Test Recognition (ACTR) and thestandards Automated Coding System) were outlined and it was noted that while they may provideenhanced industry coding in the future, resources and time prevented developments at this stage.

Paper 6. Summary of responses to a questionnaire on computing technologies used inmember countries, Carl Lowe, US Bureau of Labour Statistics.

Carl Lowe completed the session by providing a summary of the written and graphicalpresentation of information gathered by questionnaire from various agencies.

Points to note were:• The typical mainframe is an IBM or IBM compatible and there has been a move towards IBM

compatible machines since 1988.• While many agencies have moved to larger mainframes, a number have moved, or are

considering moving, to mini or micro based systems for register operations. Given theincreasing power of these latter machines and the connectivity of hardware, this may be anincreasing trend in the future.

• There is a significant move towards the use of relational databases, with the most common inuse at the moment being ADABAS.

• Applications programming languages used by agencies are predominantly COBOL or PL/I.NATURAL has grown in usage; while the language C is seen by some as being a considerableinfluence in the future.

• Structured Query Language (SQL) is seen as a major factor in the portability of softwareacross different processing environments by providing a common interface for databases.Updating information for registers is most commonly obtained from taxation records and mostagencies match records utilising standardised names and addresses.

• There is considerable variety in the range of users able to access data from agencies' registers -ranging from external government and public users to restricted access by government agenciesfor statistical purposes only.

Page 46: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 46 -

Session No. 7CLOSING SESSIONSession Leader : Australia

A ITEM 7 - CLOSING SESSION

Initial discussion for the session centred around how best to organise and make available thepapers presented at this and all previous Roundtables, given that a substantial amount ofpotentially valuable information had been forthcoming. While the proceedings were useful, theywere not seen as sufficient by themselves.

Mr Sarossy proposed that for this Roundtable, the papers should be compiled into a volumewhich would be readily available to any other interested parties. There was no objection to thisproposal. However, it was decided not to do this for previous Roundtables; instead, author, titleand keyword indexes would be compiled for all papers, excluding progress reports, of eachRoundtable. The indexes would be compiled by the relevant host agencies who would alsomaintain a master set of complete papers for their Roundtable. The ABS would produce the firstsuch indexes, for the 6th Roundtable, and then circulate these to other host agencies for use as aguide to their index creation. The ABS would subsequently consolidate the indexes for allRoundtables and circulate this to participants.

Discussion then moved to arrangements for next year's Roundtable, which will take place inCopenhagen, from 12 to 16 October.

It was generally agreed that the session leader idea proposed at Williamsburg had worked well,providing better focus and discussion, and there was consensus that session organisation shouldcontinue in this way.

Mr Ryten reiterated the importance of getting papers out 1 or 2 months before the Roundtable, itbeing very detrimental to discussion if this did not happen; it also enabled session leaders toprovide more focus to their session. Mr Sarossy said that the host agency should be prepared tochase session leaders to ensure this happened. Mr Lee raised the idea of "contributed" papers; itwas generally considered that we should move slowly in introducing this concept. Mr Perryproposed that a final agenda including titled papers should have a firm target date of some 4 to 6weeks before the Roundtable; papers arriving after that date ought to be excluded frompresentation, but allowed as room documents. There was consensus to this proposal.

Agenda items for 1992 were then considered. It was agreed that the shortened progress reportdeliveries were more stimulating than previously since their focus was sharpened and it wasagreed to continue with the 5 minute presentations.

Page 47: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 47 -

Agenda items were called for, and the following were proposed:* . implementation of revised SIC's, experience and pitfalls;. marketing matters;. where the register section fits into the statistical organisation;. demands on the register for enterprise group surveys;* . technology follow-up;. relationship of business register to survey users;. response burden control;. review of automatic and semi-automatic coding procedures;* . statistical units;* . quality assurance;. complex business units;. longitudinal analysis;* . case study of a register development in a specific

agency.

The statistical units proposal (Mr Willeboordse) and case study proposal (Mr Dodds) generatedsubstantial discussion.

Regarding units, Mr Willeboordse noted we would soon have a reference frame in the ISICintroduction and Eurostat introduction, however, the trouble with these is that they leave toomuch to interpretation; at least as a first step we could assist in firming up the definitions. Weshould also pursue uniformity in the use of terminology. Mr Perry noted that a contribution fromEurostat would be useful, especially given the time and place of the meeting.

Mr Dodds raised the idea of circulation of several examples of business organisation toparticipants for them to classify in order to assist understanding and comparison of agencystatistical unit rules. This idea was seen to be very worthwhile.

Then followed debate on the worth of having a session on units. Since the topic covered othermatters of general interest such as the handling of general government, religious organisations etc,it was agreed that the units agenda item should stand.

The idea of a case study of one agency's progress was suggested in the context of the 1stRoundtable which focussed on the Statistics Canada register redevelopment. It was suggestedthat it would be better to select an agency which was in the early stages of a development ratherthan one in the final stages. After some debate on potential agencies, the UK CSO was selected.

Page 48: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 48 -

Initial agreement was reached about which of the proposed agenda items would be proceededwith (marked with an * in the list above). Subsequently there was discussion on which oftechnology or response burden control should be included, with the latter being selected. Theinnovations planned by several agencies in the treatment of complex business units were ofsubstantial interest and the agencies concerned were asked to pay special attention to progress inthis area in their progress reports.

Session leaders were then selected, and some further development of themes took place, asfollows.

1. Quality Performance MeasuresMr Archer noted that unless specific quality performance measures were identified, we

risked covering previous ground. On this basis it was agreed that the theme for the session wouldrelate to:

. the quality of industrial classification;

. timeliness of births.Session Leader: UK Department of Employment

2. Case StudySession Leader: UK CSO

3. Statistical UnitsSession Leader: US BLS

4. Response Burden ControlSession Leader: Australia

5. SIC ImplementationSession Leader: Canada

6. MarketingSession Leader: New Zealand

Page 49: - 1 - Index Number: 060701 Title: Proceedings Author: Date ... · Oracle with SQL will be utilised instead of the presently used ADABAS. ... survey is being investigated in response

- 49 -

Papers were then called for, with session leaders requested to lobby for more if sufficient were notforthcoming at this session. The following agencies volunteered:1. SIC : Australia, Denmark, Netherlands,

UK CSO, US BC, US BLS2. Response Burden : Netherlands, New Zealand, UK CSO,

US Department of Agriculture3. Statistical Units : Canada, Netherlands, US BC4. Case Study : UK CSO5. Quality: UK DE to undertake a survey6. Marketing : Denmark, Finland, France, Sweden

Mr Sarossy then noted that the ABS, in preparing for this Roundtable, had drawn up a manual ofprocedures and would now pass this on to Denmark for their use and further development.Certificates were then distributed to all participants acknowledging them as "Knights of theRoundtable". Mr Sarossy then formally closed the 6th Roundtable.