2nd Working Group Meeting

Report of Meeting

Introduction

Document 23: Agenda for 2^nd meeting

Document 24 : List of participants for 2^nd meeting

The meeting was opened by the Director-General of the Estonian Statistical Office. He stressed the importance of the project and the meeting. He warned participants to pay more attention to quality than to technological issues.

The agenda (document 23) was agreed. Document 24 (list of participants) was circulated for updating and correction. Where papers were updated they would be re-circulated and marked as final version. In addition to the papers circulated specifically for the meeting, participants were referred to document 9 (report of the first meeting), 10 (B3 questionnaire) and 11 (progress report to 1 August 1998). All papers except 14, 16 and 25 had been sent by email prior to the meeting.

1. Organisational issues

Document 22 : Place in the national statistical office of the pilot project (John Perry)

Document 12 : Summary of register situation in CECs (John Perry)

The Slovak republic had raised the issue of local unit information prior to the meeting. In addition prior to the meeting the Czech Republic had raised a question regarding a European business register. The latest information available to the project manager was that a group (“Best”) had reported to the European Commission in May proposing a standard for business registers in the EU. The Commission had produced a response and discussion was due to take place in the Industry Committee on 16 November.

Document 22 was essentially circulated only for information. In relation to document 12, the headings of the summary table on gross domestic product (GDP) per head had been reversed. a corrected version incorporating other changes notified by participants would be circulated following the meeting.

Eurostat mentioned that GDP at purchasing power parity (PPP) had been published in September and could be quoted in the document. Further information was also available on responses to the (Andree Bombeeck) Eurostat questionnaire: Estonia and Albania had in fact replied but Eurostat had not been able to read the files; Latvia and FYROM had also replied; Bulgaria agreed to check the position. Subject to the Bulgarian check, Andree Bombeeck would be asked by Eurostat to produce and circulate a summary. Eurostat advised participants that this questionnaire was put to EU member states annually and that the questions differed each year.

There was considerable interest about local units on the business register. These were present only for some countries. Latvia was planning to include local kind of activity units (LKAU) and to record their employment. Most were small shops or service activities. The information was difficult to update, as there was no administrative source. They felt it was worthwhile persevering as there was a demand from the national accounts division for the regional accounts. Hungary also had experience of adding local units for the hotel, retailing and catering (HORECA) trades from a census in 1995 and for the whole economy from a census in 1997. They were updating the local units through the registration system. This was helped in early 1997 by the requirement for businesses to renew permits. Poland collects local unit information through structural surveys of enterprises with more than 5 employed. Local units would in future be incorporated explicitly on the business register. One additional feature of the situation in Poland was that up until now natural persons would register for each separate activity. In future they will register only once for the enterprise. The Czech Republic does not currently have local units but was planning to add them for large enterprises through profiling in 1999.

2. Seminar on uses of business registers

Document 15 : Characteristics, maintenance and uses of the business register (Jean Ritzen)

Document 20 : Items and questions on the review of the use of business registers in CECs (Hans Velt)

Jean Ritzen presented his paper (document 15) through slides (attached). The paper relates to 1995 and there have been developments since then that were covered in the presentation. The position for the Netherlands is similar to other countries but solutions differ by country. Two uses: economic statistics and business demography. Moving from using mainly surveys to electronic collection and direct use of administrative data. [attach copies of slides]. Register has a key role in relating administrative to statistical units. Objectives of the business register are: construction of statistical units; preparation and co-ordination of surveys; demography; mobilising administrative data. Main variables are for identification ; stratification; demography. Units: source and legal; enterprise; local unit; enterprise group. Sample frame is only at the enterprise and enterprise group level.

Uses: administrative (optional); statistical (surveys including co-ordination and control of burden; tool for linking data sources). Single business register will be an important tool with a unique business identifier. Survey users obliged to use the register. Register quality is a problem, especially for small enterprises. Register should be accurate and up to date. There are gaps in sources and inaccuracies in recording (e.g. of NACE code). Because the Netherlands does not sell lists, address errors are not too important. How to measure quality : surveys; checks against other sources; area samples. Plan to use more sources. This can create more differences and requires prioritisation of sources to minimise the cost to the office of maintenance.

The business register is becoming more and more a system of information, used as a tool for estimation of populations. Can apply correction factors to deal with under-coverage and errors in classification. Register needs to be integrated into the statistical process with good access.

Each month a file is produced for monthly statistics; similarly for quarterly. Annually two files are created: before and after changes in register variables. Two versions of the register are kept: one for processing (continuously updated); and fixed files for surveys and analysis.

Hungary asked about the many to many relationship between legal units and enterprises. There are some problems identified by profiling, whereby one legal unit belongs to more one legal unit. Also there are some problems with defining enterprises in the public sector. All enterprise groups with less than 10 employed are now treated as enterprises which reflects reality and the tax administrations.

Eurostat recognises that most EU states equate legal unit to enterprise. Eurostat has a draft on recommendations for large and complex businesses that can be circulated to participants. Will also be development of definitions for the public sector, education, health and non-profit institutions. Enterprise groups are a future work area.

The Netherlands was planning to introduce a single business register maintained outside the statistical office. France was known to be redeveloping its SIRENE system as a sngle business register (a paper is on the business survey frames roundtable internet site - http:\\www.stat.fi\roundtable\

Hans Velt presented document 20 (planning for the seminar on uses) through a series of slides (attached). A number of topics for papers were considered:

- Practical issues in delineating local units - possibly by the Netherlands (Jean Ritzen would check)

- Uses of business registers - Latvia

- Registers as a tool for co-ordination - Philippe Pommier reporting on a study by CESIA for Eurostat

- Co-ordination of feed-back to the business register - Romania

- Business registers from the viewpoint of the users - Poland

- Estimation of populations from imperfect frames - Hans Velt with the Netherlands Economic Institute

- Quality - possibly Eva Elvers (Statistics Sweden)

Other topics for which there were no specific offers included: rules for stability and practical issues of industrial classification; the role of local units; and use of tax data in business registers

This gave enough guidance for Hans Velt to produce an agenda for the seminar. The agenda would include the names of authors. Papers would be needed in mid-January.

In addition, Eurostat would provide an opening session on the Regulation and recommendations for business registers in the EU.

Participants agreed to provide answers to the questions raised by Hans Velt in document 20 by the end of November, drawing on other responses (document 12, the national accounts study of business registers and the annual Eurostat questionnaire). One change to the questions in document 20 was agreed: other uses (e.g. administrative or commercial) would be added to section 3. Hans Velt would then summarise as a paper of the seminar.

Eurostat would send an invitation letter to director-generals. This would link the seminar closely to the DOSME project. It would ask for national accounts interests to be represented, either directly or indirectly (through the business surveys). It would also suggest representation from enterprise statistics and from methodology areas. It would mention the languages for the meeting (English, German and French) and the fact that up to four participants would be funded (and how the funding was organised). Arto Luhtio would check procedures prior to finalising the letter. The letter would confirm the dates (8-10 February 1999) and Luxembourg as the location. The building and room were not yet agreed.

3. Update on technological matters

Document 5 : Summary of computer systems in CECs (Infostat)

Document 13 : Update on technological issues (Infostat)

Infostat presented the documents and the web site demonstration in this session.

There had been a number of changes since the 1st working group meeting. The PECO Help had been completed. The CDB file for Bulgaria was finalised, together with all of the C survey files and the tables required to make the final payments to countries.

For the DOSME project, the help file had been supplied for the B3 survey and the sample allocations finalised. Modifications had been made to the DOSMENU and to the sampling and data entry packages. Programs now had more checks built in to make them more robust. With the increasing number of files as the surveys proceeded, Infostat had introduced a more consistent identification of files.

The year 2000 problem has not been addressed.

Countries could now use Blaise 2.5.

Communication was now generally by email, although this had not been without problems, mainly to do with encoding and size of attachments. The WWW site had been developed on htttp:\\www.infostat.sk\DOSME\. This site was not secure but could be accessed only by those who know this precise address. A demonstration of the site showed up problems in access when the internet service was very slow. Improvements in speed might be expected from a text only version. Some screen changes were pending. Only the database manager could action changes and only after a delay because of other pressures. This was a problem particularly for activity close to meetings.

Participants were requested to check policy on internet access within their offices (all national statistical offices had some web presence), to gain access and to check the system. Comments, both positive and negative, were to be sent to Infostat through the web site email by the end of October.

Participants were also requested to check document 5 and to update for recent changes and planned changes. Hans Velt would send an email with a MS Word 97 attachment immediately after the meeting to each member of the group. All would be asked to detach the file and attempt to read it. confirmation of receipt and readability was required from all.

4. Outstanding activities on PECO Panel project

Detailed A publication (John Perry/ Hans Velt)

B2 publication (Paul Smith/ Arto Luhtio)

Document 21 : Current longitudinal database and future outlook (Hans Velt)

Document 19 : Enterprises in the CECs: initial longitudinal analysis 1995-7 - draft results (Hans Velt)

Document 14 : Recommendations for further analysis of the data collected in the C survey (Hans Velt)

Each country, Infostat and Eurostat received a copy of the detailed A and the B2 publication (pre-publication version) from the project manager. The detailed A publication was referenced in the B2 publication but would not appear other than electronically. Latvia queried the figures in table 23 (from the B1 survey publication originally). They stated the death dates were too high for creations in 1995. This was because a block of 6,000 natural persons had been omitted from the A survey frame and added to the B1 frame. This would also affect the A survey results as the sampling from was deficient because of this action. The project manager and Eurostat would consider the course of action to take, prior to the actual publication by Eurostat, which was now scheduled for some time in November. Other than this, Infostat could release the databases to countries who could proceed with their own publications. B2 would be published in the same series and format as B1.

Poland was planning to publish data within about one month. Slovakia was also planning a short publication towards the end of October, to be followed later by a more detailed publication. No other countries stated that they planned their own publications.

Hans Velt presented his three papers that had a common theme of dealing with the longitudinal aspects of the project. The first paper (document 21) covered the creation of a longitudinal database. The data quality for the surveys is generally good but some meta-data were incomplete, eg the definitions of some variables. There are also file name inconsistencies. Some questions in different questionnaires have different answers and coding is not always consistent between surveys.

The document recommends a relation database design to permit flexible data holding, unique identifiers for units, clear and logical variable names and consistent coding. While this was an objective, transition from the existing linked datasets would not be simple. In addition, it would not be easy to copy a relational database, so communication through flat files would still be required.

What was also missing from the existing datasets were the tables that formed the publications. They should be stored with their underlying individual data.

Document 19 covered the planned firs C publication. There was a need to publish both estimates of populations and analyses of change. The first C publication would deal only with the second of these requirements. Three types of analysis were planned: panel; cross-panel; and trend. Comparisons would be between A and C and B1 and C.

There was a problem with missing death dates. To permit early publication, death dates would be ignored, with death codes being used. In the longer terms complex rules to deal with inconsistencies and imputation of missing death dates would be necessary. The definition of survivors was important to get right. The publication would focus on this rather than active businesses.

The structure of the first C publication would be: methodology; text, key figures and graphs; copies of questionnaire for A, B1, B2 and C and an explanation of differences between them. In theory, five tables were possible for each B2 table.

A selection of draft tables was then presented to highlight the methods and data quality issues for countries to consider (table numbers are those used for B1 and B2 publications):

3a - High average employment for Albania

3b - Large differences in average employment for Bulgaria

4 - A question included in C but not in A and B1 - Estonia shows extreme data

7c - Bulgaria and Estonia show outliers. Hungary has zero entries.

8a - Hungary and Estonia have some zero entries

9a - Albania is out of line because of a coding error in survey A - Albania is clear that this was corrected for the publication - if so, it throws doubt on the data for all other countries, which may be wrongly coded.

12c- Latvia, Lithuania and Romania appear to be miscoded

14 - This is a new table - the question was modified, with codes for yes and no being effectively reversed - For Lithuania, it is possible that the wording of the question was not change, resulting in the coding being reversed.

16c - One of the options “secondary payment insolvency” had been modified to “ non or late paying customers” - this had affected consistency of response.

The plan was to finalise the publication in December. Provisionally publication would be through a press notice and electronic publication only. Eurostat will need to confirm.

The actions required were: tables to be sent to countries; responses from countries; Infostat database corrections; final tables to Hans Velt; final report to countries; comments from countries; databases to countries; and final publication.

Document 14 outlined more detailed analysis of the C survey database. Various options were considered. Eventually countries agreed that the publication should produce population estimates and some other tables. It should not be like the detailed A publication. Poland offered to produce this publication in co-operation with a suitable other contractor. Poland would need to be paid by Eurostat for the work. The plan for the publication would be presented as a paper for the 3rd working group meeting.

5. Progress on survey B3

Document 16 : B3 summary tables (Infostat)

For this session each country provided an oral situation report:

Albania - They had experienced many problems with the sampling programs but these had all been resolved by the end of September. Fieldwork had started on 1 October. Unusually, more than half of the sample was in Tirana. Interviews would be complete by the end of October.

Bulgaria - Fieldwork would be by interviewing using the 28 regional offices. Interviewing would be complete by 5 December.

Czech Republic - Mailing of questionnaires had started on 15 September. Final non-response contact will be by interview by early November.

Estonia - Some problems had been experienced with the sampling programs but all had been resolved. Mailing of questionnaires had started on 1 October and already 8% had been returned.

Hungary - There had been a delay in starting the survey but the survey forms had probably been despatched to regional offices on 12 October. There had been some local technical problems in producing the questionnaires. Replies to the postal questionnaires were required by 26 October and final take-on of data was due for completion by the end of November. Some problems had been encountered in running the programs but all had been resolved.

Latvia - The questionnaire had been printed early in September and fieldwork started on 15 September. There was already more than 50% response of the live units on the sample and five regional offices had actually finished the fieldwork. A lot of data checking was being done.

Lithuania - Surveying was b interviewers from the regional offices, with 35% of fieldwork completed.

Poland - Around 300 units in the original sample were found to be dead through checking the register. Questionnaires wer emailed out at the end of September.

Romania - Questionnaires had been sent to regional offices on 25 September, with a deadline for data collection set at beginning of November.

Slovakia - Fieldwork was due to finish early in December. No serious problems had been encountered.

Slovenia - Staff shortages through illness had caused local problems but the survey had started on 7 October with mail out of questionnaires. Reminders were planned for mid-October followed by telephone interview of non-responders. Of the 1400 units in the sample, 95 had been found dead on the register.

FYROM - They were included only on a pilot basis. The scheme had been agreed with the project manager in Skopje following the 1st working group meeting. The survey manager and programmer had been given training in Bratislava on 1 and 2 October but no feedback was possible as neither was at this 2nd working group meeting. A panel of 200 units (75 legal units in 1997 and 75 before then with 50 natural persons registered in 1997) had been selected from two of the eight regional office areas. The forms had been mailed out following the training session. Only dead letters had so far been returned.

Overall the B3 survey was proceeding with t actions by countries being on or ahead of schedule and only minor problems being encountered. On the finance side, several countries complained once again of delays in payment. In particular, Poland had still not received payment, although CESD had stated that this was made on 22 September. Countries found the delays and lack of adequate information to be unacceptable.

Infostat reported on progress on the central processing. For this working group they had supplied data (for document 16) on diskette. all of the problems with programs had been resolved bilaterally with the countries affected and there had been no resulting delay to the project. All sampling tables had been received from all countries prior to the working group meeting. In general, the April 1998 business registers had been used as the sampling base and this had been a good choice as it dealt with the normal technical delays in getting registrations on to the registers. The number of late registrations (Y07 period) had been lower than previously indicating an improvement in register procedures. The quality of the registers also seemed to be improving, with smaller numbers of units with incorrect legal status or invalid classification codes. There remained however problems with coding of employment, with confusion in some instances between zero and undefined employment codes. High inactive rates had been noted especially in countries with re-registration processes. Infostat requested all countries to check the data on the diskettes very carefully.

6. Imputation issues

Document 25 : Imputation examples (Paul Smith)

Paul Smith presented document 25 and a detailed example of the imputation variables selection process. Excel was used as it was a program available to all countries, although a statistical package such as SAS would be preferable. In particular, Excel offered only Chi-squared as a test statistic. He expected that following the session, each country would work through the examples to be provided on floppy diskette and discuss any problems with him. The eventual outcome would be that all countries (other than probably FYROM) would be expected to run the imputation for B3 locally without assistance from Infostat or experts.

One important point about the process of choosing good variables (through tabulation and examination of test statistics) was that improvements in imputation would be expected whatever the choice of variable. A wrong selection would do no damage and would simply not give improvements to quality that a good choice would give. If a chosen set of variable fits well, there is no need to search for the best fit. For the sample, employment size was a good starting point as an imputation variable. Paul Smith would check the impact of different versions of Excel. The diskette work contain 97 and 5.0 versions.

7. Preparation of survey B4

Document 17 : B4 survey sample and questionnaire (John Perry)

Discussion of survey B4 was based on document 17. The sampling scheme was agreed. For the 4th meeting of the working group, an analysis of the longitudinal database would be undertaken to determine if the sample sizes and stratification were optimal. A final opportunity would be available to review the questionnaire at the 3rd working group meeting. On the basis of the work so far it would not be advisable to change the questions. Countries would however check the national wording in view of the queries raised from the C survey analysis.

Eurostat would consider plans for publication of B4 survey data prior to the 3rd meeting.

1. Proposals for project 2

Document 18 : Proposals for project 2 (John Perry)

Eurostat was unable to provide information on the process of getting approval to project 2 (document 18). However, with approval of project 1 consideration had been given to the possible extension to project 2. Eurostat would check procedures within its organisation. In discussion, some important improvements were suggested to the project proposal. The proposal should stress the importance to the national economic statistics (in particular the national accounts and the structural surveys) and refer to the pre-accession partnership. Although the balance needed to be correct, it was also important to stress the need to improve quality of business registers and to align them with the EU Regulations. Particular stress should be placed on the differences from the previous PECO Panel survey. This was a difficult task because the major element of the project cost was for the conduct of the B5 and C2 surveys.

Hungary informed the group that the Vilnius unit would soon disappear. A new policy group will make decisions about projects.

Guidance to the content of the proposal could come from examination of the future work programme of the business registers committee within the EU. This was being presented to the Statistical Programme Committee (SPC) in November and then to Council early in 1999. Eurostat would keep the group informed. Eurostat would also need to advise all countries how to lobby effectively for support of the proposals.

Poland mentioned its planned seminar on structural business surveys scheduled for June 1999. The plan was close to approval by Eurostat. It would provide a good opportunity to underline the importance of the project to all countries. Poland would ask the group to provide a paper.

A mid-term evaluation of project 1 was agreed for the 3rd working group meeting.

10. Preparation of the next meeting

A number of actions prior to the next meeting were agreed:

a. General

b. Seminar

c. Technology

d. B2 survey

e. First C survey

f. Later C publication

g. B3 survey

h. Imputation

i. Other

The actions under f) were discussed in some detail. Countries agreed to try to meet a timetable for preparation of the publication prior to mid-December. If they could not meet this timetable, they themselves would not be able exercise the right to publish immediately.

Romania confirmed an offer made prior to the meeting to host the 3rd working group meeting. They were advised to consider transportation and facilities before choosing the location. The meeting would require access to email and to the internet. An overhead projection for computer screens was required to the standard offered in Tallinn. The PC should have MS Office 97 for Word, Excel and Powerpoint. The date agreed was during the week of 20 April 1999. Photocopying and computer printing was required on-site.

The agenda for the meeting was provisionally agreed as:

1. Introduction and agenda

2. Progress

- B3 processing

- B4 final planning

- Report on second C survey publication

3. Evaluation

- Mid-project review

- Review of February seminar

- Review of imputation work for B3 survey

4. Project 2

5. Next (4th) meeting

Countries were requested prior to the 3rd meeting to consider if they could run the 4th meeting and to advise the project manager.

The meeting closed on schedule at 12:00, with thanks to the Estonian statistical office for their efficient and friendly organisation of the meeting.

John Perry

Project manager