3nd Working Group Meeting

Report of Meeting

Monday 26 April 1999: 0930 to 17:15

1. Introduction

Document 27: List of participants for 3^rd meeting

Document 28: Agenda for 3^rd meeting

The meeting was opened by the Director of Statistics of the NSC of Romania (Victor Dinculescu). He welcomed participants to Romania and, in particular, to Brasov. He commented that the DOSME project had provided knowledge on modern statistical methods for business surveys, specifically on imputation, estimation and grossing-up. The funding by Eurostat had been important in the process of achieving comparability with EU member states.

The list of participants had undergone several changes but the version circulated the previous week was final as far as names were concerned. A contact point had been asterisked for each organisation. This contact was different from the usual contact in recent months for three countries as that person was not able to participate this time. Participants were asked to check details so that a final version could be circulated after the meeting.

Jean Ritzen, who had been on the list until a week previously could also not attend but would still be associated with the project as an expert and expected to attend the next meeting. A paper he had intended to talk to (produced for the recent International Roundtable on Business Survey Frames had been circulated to contribute to the discussion under item 4).

The agenda was agreed. Participants were asked to check that they had all of the papers. Document 29 was to be produced after the meeting. Three contributions had been received prior to the meeting: from Lithuania, Poland and Estonia. Other written contributions were presented to the organisers for copying: from Albania, Czech Republic, Hungary, Romania, Slovakia and Slovenia. Written contributions from Bulgaria and Latvia would follow the meeting

Papers had been circulated by email. This had been largely successful, although neither Albania not Bulgaria had email access in recent days. Papers were also posted to the web site:

http:///www.infostat.sk/DOSME/

The hosts provided access to the site so that the DOME screens could be demonstrated. Access was reasonably fast with no breaks in connection. Albania, the FYR of Macedonia, Hungary, Poland and Slovenia reported problems with web site access in recent weeks. They were asked to check again following the meeting and to report problems to Infostat. Bulgaria had only a 9.6k connection. Infostat would check whether this could access the site successfully.

Once these issues are resolved, it is hoped that papers can be circulated on the Internet site instead of by email. Initially, Infostat proposed to send sampling programs to CECs using this method.

Eurostat asked Infostat to provide a link to their methodology site from DOSME. CECs were asked to provide site addresses for their own statistical offices that could also be added as links from the DOSME site.

Progress on the web would be monitored.

2. Progress on B3

Document 29: B3 report by CECs (Document to be produced following meeting)

Document 30: Infostat report on B3

This item was based round the written contributions from ten CECs and two oral presentations (from Bulgaria and Latvia) to be written up following the meeting. The presentations were designed to help drafting of the B3 publication.

Lithuania

The decrease in the number of active enterprises was related to the events in the Russian economy of mid-1998. Similarly the limited growth in GDP between 1997 and 1998 and an increase in unemployment were caused by the same factor.

Poland

GDP shows negative growth, but there is no crisis in the economy; inflation (measured with the CPI) is much reduced, and for the first time is <10%, coinciding with falls in interest rates. Price rises for services (notably gas, electricity and rents) are still a problem. The Russian crisis had a small effect (noticeable particularly in the food and textiles industries), but recently unemployment has begun to rise, and there is a noticeable drop in the “profitability rate” at this stage. There has been a large influence of non-economic factors, especially the new statistical law, and this impact is expected to be bigger for B4, with changes of regional classification, and reform of the health, education and social security systems. There has also been a development where businesses have divided to take advantage of new regulations, even though the new businesses are basically subcontractors, so we expect an artificial increase in the number of businesses in 1999.

Macedonia

The position is essentially as in the paper; GDP figures for 1998 will be available soon. Particular points to draw out are that the rate of registration of new businesses seems to be fairly stable. There will be a new registration law in 1999, and this will be coupled with a move to classification based on NACE. If there is no re-registration of a business, under the new law it will be assumed to be dead. It is also the case that the law makes it easier for businesses to stay the same, and more work for them to change.

Slovenia

GDP had risen by 4.1% in 1998 (estimated from 3 quarters), and inflation continued its downward trend with the CPI rising 6% annually. There was also a slowdown in manufacturing production, with rise at approximately 4%. This growth was concentrated amongst exporters, largely exports to the EU (up 11.8%). Unemployment is 7.8%. The number of businesses is increasing by 0.5% quarter on quarter, mostly in the distributive trades and mostly sole proprietors. Most of start-ups are by individual entrepreneurs, since for them no capital is needed.

Slovakia

The proportion of dead units had gone down in 1997 (because of an improved register), and fewer of these units had definitely ceased trading. GDP in 1997 had changed by - 2.1% in constant prices, and other indicators also show a general contraction in the economy.

Latvia

The B3 survey showed a higher number of active enterprises, but this is before the crisis in the Russian economy had a major effect. Two large banks went bankrupt in this crisis, taking many small businesses with them and resulting in very high levels of bankruptcy and insolvency. These, however, occurred after the fieldwork for B3. As in Lithuania, the main sectors to be affected were transport, food and textiles. Unemployment is now rising suddenly (>10% for the first time), and some unemployed people are trying to generate income from their own enterprises, which typically start and stop very quickly. Obtaining reasons for cessation for these businesses is difficult.

Hungary

From 1 July 1998 a new system of registration began in Hungary: the “one window” system. Previously registration had been the responsibility of the statistical office only, but this is now split between the registry courts, chambers of commerce and the statistical office. Care is being taken to ensure that the correct number of units is identified.

Estonia

Obtaining credit had become difficult in 1998 as three banks had gone into bankruptcy. Natural persons should not register on the business register. They typically operate for only a short time and many sole proprietors have stopped their activity.

Czech Republic

There are a large number of legal/natural persons from foreign registrations, in particular some new businesses from Russia and Vietnam. Domestic output is going down, as are exports, and the unemployment rate is climbing. In many enterprises there is too much administration, and not enough time actually working. There is a large change in both active and inactive businesses for the B3 survey, mainly resulting from fewer creations.

Bulgaria

There was a new register and a new registration law in Bulgaria, which had given rise to B2 information with unusual characteristics. The law meant that temporarily inactive businesses that did not reregister by a certain time were declared to be dead. In practice some of these have restarted activity, but must now reregister, and so count as new registrations even though their starting date is old. This accounts for the high proportion of period 07 registrations in B3. In addition for B3 there are three large recently privatised businesses in the sample, which cause distortion in the results. In this case we should probably make sure they are not used as donors in imputation, and re-weight by size so that they do not have a disproportionate effect. Additional work to identify “false new registrations” is promised, and this will be used to provide estimates (probably in footnotes to tables) of the true creations.

Albania

Albania asked about imputing the NACE code from the register; this was not a good idea as part of the survey procedure is to collect this information to see how well it corresponds with the register data (and in some cases the correspondence is not good). It was also noted that the border regions of Albania were very much affected by the crisis in Kosovo, and that this might make businesses in these regions both non-responders, and different from other businesses in the country. This is a non-ignorable non-response problem, and the solution will need to be considered for B4 imputation.

Romania

The rate of creations seems to be stable (increasing from year to year) even though the economy is in recession. The definition of a new enterprise from the register is not clear, as parts of the population started trading two years ago. The registration forms have been redesigned recently to catch this information.

These comments are referenced again in the B3 publication section.

Document 29 would be finalised two weeks after the meeting to allow countries to make changes to their presentations. The document would be simply a set of the individual presentations.

This was followed by a summary of the position on payments for the B3 survey, based on information supplied by CESD up to 23 April

Albania Final invoice not submitted

Bulgaria Final invoice not submitted

Czech republic Final invoice paid 20 April

Estonia Final invoice paid 20 April

FRY of Macedonia Final invoice paid week of 26 April

Hungary Final invoice not submitted

Latvia Final invoice not submitted (actually submitted in last few days)

Lithuania Final invoice paid 9 April

Poland Final invoice paid 20 April

Romania Final invoice paid 20 April

Slovakia Final invoice paid 16 April

Slovenia Invoice to be re-submitted before final payment can be made (the figure in the final invoice covering both second and final payments was incorrect)

Infostat then presented document 30. This raised a number of issues:

3. Publications

Situation report on B2 and C publications

The B2 publication had come out at the end of February. Three copies had been sent to each CEC and two to John Perry. In addition, the Eurostat publication office has a separate circulation list. Some spare copies exist and can be requested. Eurostat would like copies of any national publications based on the survey.

The C survey publication had been delayed by some data problems regarding the survey in Bulgaria. The final version had been sent to Brussels at the end of March. Publication had now been approved and the document had been sent to the publication office two weeks ago. Actual publication is expected within one month.

Poland had plans to publish C and also to produce a comparative publication for B1, B2 and B3 data. Latvia has advertised the Eurostat (English language) version but has no plans to publish its own version. Latvia quoted the limited usefulness of a publication based on old data. Slovenia plans to publish a brief report in the next two months. No other plans exist for publication.

Document 31: B3 publication

Paul Smith raised a number of data issues from the document and from the presentations earlier in the day. The issues were left “on the table” overnight.

Tuesday 27 April 1999: 0900 to 1715

Document 31: B3 publication (continued)

A number of specific queries were discussed to ensure that the production of the B3 publication could proceed smoothly.

Legal form in Lithuania

Joint stock companies are a superset of joint ventures in the Lithuanian register, with the result that they cannot easily be separately identified. In this case all businesses have been treated as code 5 (not 6). Joint stock companies are essentially the same as limited liability companies in the Lithuanian system

Large proportion of units in 07 registration period

The table for registration period (RY) was based only on live units. Bulgaria confirmed that the late registrations were as a result of the new statistical law. The Czech Republic stated that a change in the administrative processes from April 1997 (when an electronic form was introduced) had probably affected the registration of small businesses. Slovakia stated that a check had been made for the first time in 1997 between the administrative and statistical registers. This would have affected the figures. A further check was planned for 1999.

In the Slovak Republic, the statistical register was compared with the register of entrepreneurs compiled by the Ministry of the Interior for the first time in 1997. Units missing from the register were added, and units dead on the register of entrepreneurs were removed, resulting in some additional delayed registrations coming through. This exercise was repeated in 1998 (which may affect B4).

Slovenia had two registers previously (of companies and of craft businesses) The Act passed in 1995 on business registers has resulted in these being combined and data being harmonised in March 1997. This had affected registrations for prior to 1997 and in the first trimester of 1997. The new units in B3 are from the introduction of this single register in March 1997.

Lithuanian high proportion of active units

The registration law has been changed from 1999, but this should not affect B3. The high rate of inactive units in B2 may be wrong. The Lithuanians confirmed that there was no exclusion of dead enterprises from the sampling base.

Poland, natural/legal persons

The sample had mostly natural persons, and Poland could offer no explanation other than that it is the structure of the economy. The sample distribution was very different to the sampling base, where the proportions were roughly half and half, which might have an implications for the sample allocation for the B4 survey..

Active units with zero employment

Infostat confirmed that this was a valid response (if an owner spent <50% of his/her time in the business) and would not be auto-corrected. In general discussion it seemed that this rule was not being strictly applied: Poland would in theory have many cases, but this is not the case as there is no part-time/full-time split on the questionnaire. Estonia asked how part-year working fitted into this framework. Slovakia and Hungary would check their units. The Czech Republic thought there was a mistake in the data for their one unit. The conclusion was that for analysis of B3, the 0 and 1 categories should be combined and treated as active, and that an additional check should be made on the treatment of this for B4.

Active and ceased trading units

The three countries involved would check the corresponding questionnaires. Infostat thought that this condition was not automatically checked. Infostat would add one more test. The leading response in checking and imputation is “active/inactive”, so in the absence of any other information we would assume that “active/inactive” was correct and impute any missing data in the normal way.

Lumpy distribution of units over trimesters

For Slovenia this was also caused by the introduction of the new register. In Albania 1997 was a period of considerable political instability, so there were very few new business registrations until things had settled down.

Split of legal/natural persons in Bulgaria

There had been increasing proportions of legal persons, but there was a discrepancy between register figures for B3 (54% reported) and the provisional table figures (78%). Bulgaria would check this information.

Capital area/other coding in Albania

The coding was not reversed even though the figures were substantially different from B2. The B3 information is consistent with the political-economic effects in Albania.

Increased “other” activity in Poland

Coding would be checked, but this is thought to be part of the economic trend in Poland. The increased percentage of “other industrial classification” is explained by the significant increase of newly created businesses in financial intermediation.

Stability of the level of registrations in Poland

Entrepreneurs are trying to minimise costs, and small businesses do not have to pay social security, so there is a tendency for businesses to split into smaller pieces, each of which registers. Hence there are more businesses but no change in activity. Due to heavy competition many businesses have ceased trading, which has an opposite, but smaller, effect. The B4 survey is expected to show a further decline in new registrations.

All agreed that the B4 survey should include GDP for 1997 for all countries. These were in the C survey publication. Eurostat would check the current figures for 1997 for all countries. Albania and FYR of Macedonia would check they had figures for their own countries.

The B3 publication should be treated (as B2) as an historical publication.

Bulgaria and Slovenia needed to check further on the impact of the legal changes in their countries on the survey results.

Any economic statements needed to be checked with the relevant parts of the national statistical institutes.

Document 32 : C survey – exploitation of data

The contact immediately prior to the meeting by Charles Monck and Associates (stated to be working for DG 1A) was relayed to the working group. A full list of his contacts would be circulated for comment after the meeting.

Offers were requested from CECs for work on a further C publication. This would be under a DOSME project.

Exploitation of data meant adequate publicity. Eurostat had regular meetings with professional organisations. They also had contacts with other DGs. They would use these links to publicise the project.

The Warsaw meeting in June and the UNECE meeting also in June provided further immediate opportunities.

The issue of confidentiality was raised. Any contractor of Eurostat within the DOSME project could have free access to the full data set. Nobody else was allowed access. CESD would be asked to include a specific clause for contractors to cover data access specifically.

4. Evaluation

The project had seen more and more processing (e.g. imputation and the longitudinal study) moving towards the individual countries. This was demonstrating that the project was “making things happen”.

Processing was becoming quicker and with less errors. Countries however should look more carefully at their own data to achieved further quality improvements. Changes to programs were not good news for users. They should be kept to a minimum.

Review, and actions arising from the Seminar on Uses of Business Registers

Eurostat had collated comments from participants that demonstrated that the seminar was a success (rated good to excellent). Costs could be reduced by not having interpretation facilities (only English being used)

Latvia felt that the inclusion of participants from areas other than the DOSME project had been very helpful. The separate Eurostat Internet presentation (by Steve Vale) was also thought to be excellent.

Latvia also felt that relating theory to practice for statistical units was becoming a pressing issue. Arto Luhtio, for Eurostat, confirmed that this was a difficult and important subject. He mentioned the recent project led by the Netherlands with the UK and Denmark. He would check that the report was available on the Eurostat web site. Eurostat was also doing a comparative study of the units mentioned in EU legislation. This would be discussed at the EU registers working group in Luxembourg in June and a report was expected in the autumn. Further work was also proposed on case studies.

The “Salomon” sample selection product had been demonstrated at the Luxembourg seminar. Countries had now had a chance to examine it further. Romania was enthusiastic, although they were not yet in a position to implement it as the selection tool for their regular business surveys. Lithuania was also not yet in a position to use it, partly because the business register did not itself satisfy the coverage requirements of those surveys. Also random numbers are not held on the register. (Infostat pointed out that it worked on a separate file and not directly on the register.) Salomon also produced only simple stratified samples and more complicated sampling methods may be required.

Latvia showed the product to a professor of mathematics, who was positive about it. Latvia also stated that it lacked rotation. (Infostat corrected misconception.) Estonia was currently testing Salomon but currently used stratified random sampling based on the Swedish system. (Infostat mentioned that the Swedish system had been examined before Salomon was developed.) They might be interested in using some parts of the product.

Slovakia had asked a mathematician for a view and the response was positive. Use in the Slovak statistical office was limited, however, because the business register was not itself used for selecting samples. Surveys were selected by merging different registers to create a sampling frame. They also had many full-cover surveys and some large businesses that had to be forced inclusions for more than one branch of industry. The Czech Republic had a need to co-ordinate samples. They have tested Salomon but cannot yet agree to use it. They had not decided where to switch resources from an existing sample selection development.

Poland now had requirements to consider costs and co-ordinate samples. They used SAS and felt that it was a better approach to selection of regular surveys. However, there could be some use directly by survey statisticians when selecting ad hoc surveys but Salomon would have to be developed further.

Documentation had been sent to the UK, Norway, Luxembourg and Italy. John Perry gave a copy (as received from Philippe Pommier but converted to an MS Word 6 file) to Romania during the meeting. Some further development could be done on Salomon, although as CESIA had ceased to function in this area of work, a developer would need to be found.

Only a couple of comments had been received prior to circulation of the revised seminar document 7. The document was also on the DOSME web site. Following this Slovakia had made a further comment. The further revised version would be put on the web site and countries would be able to download it. The full document was too large and not sufficiently well-structured to develop, although the first part (a) could be used. This would be presented to the UNECE meeting on business registers in Luxembourg in June.

Document 33: Eurostat questionnaire summary

This document comprised three parts: a summary with the UNEC questionnaire; a table of estimates of numbers of enterprises; and the 1999 EU member states questionnaire.

The questionnaire used By Andree Bombeeck had been too complicated and the data extracted from the responses was of dubious quality. The same questionnaire had been used for EU member states but had now been dropped in favour of much less complicated software.

There was considerable debate over the numbers. John Perry pointed out that he had produced some comparative figure in 1997 with a Eurostat national accounts project. This compared regular register counts with data from the PECO survey. What was clear was that countries were not providing consistent information on different questionnaires and they were generally not using the PECO/ DOSME results. This problem needed urgent action by CECs.

The UNECE questionnaire was criticised specifically for its concentration on total numbers on the register, ignoring activity rates.

All countries were urged however to complete the questionnaire. This would provide a full set of data for the DOSME group.

There were some questions about the Eurostat questionnaire. Definitions of public enterprises (businesses controlled by central or local government) and of NACE P And Q were unclear. Slovakia mentioned the need to hold ESA sector codes.

Document 37: Review of imputation work for B3 survey

Countries had generally not used the Windows version of the imputation. Poland stated that they had not used the Windows version, even for testing. This would be the only one supported for the B4 survey, so countries should ensure immediately that they are happy with it. Infostat agreed to check that it was « year 2000 » compliant. Updated versions of the imputer would be placed on the web site for downloading. All CECs were asked to report to Infostat the success or otherwise of their attempts to download.

The topic of non-ignorable non-response was touched on briefly in the presentation. An example had been noted in the B3 survey data : parts of Albania affected towards the end of 1998 by the Kosovo situation had businesses that were impossible to contact. The issue would need to be considered for the processing of the B4 survey.

Document 39: Quality of coding of industrial classification

This item was covered on Wednesday. All CECs agreed to examine 50 questionnaires from the A survey and their equivalent questionnaires from the C survey. They were to be coded again directly from the questionnaires, without reference to the previous coding and the NACE codes were to be sent in a file to John Perry within three months of the meeting.

5. B4 survey

Document 34: Planning for B4 survey

Lithuania and Latvia commented that the costs were increasing but the unit payment was still 8 Euros. Latvia commented on the delays in payments from the B3 survey that made it difficult to pay for survey services.

The group agreed that the three payments should be of 25%, 50% and the balance of the total sum (up to 25%).

The project manager would write to CESD about the continuing delays in paying invoices.

So as not to penalise those CECs that finished imputation by the due date, those completing the work and having their imputation approved would be able to send an invoice for final payment immediately following the closure date.

Some detailed issues needed to be covered. The weighting for Hungary would be checked by Infostat, in the light of the problem found in the B3 survey. Special treatment might be necessary for Albania and FYR of Macedonia data. DOSME help would be guaranteed by Infostat to operate under Windows 95, 98 and NT (but not 3.1/3.11). Infostat would check the “year 2000” compliance of the updating tool (Forefront Author). Infostat would also check the compliance of other programs and data files, as the survey would operate over the millennium period. As “year 2000” clause would be included in contracts

The timetable was revised outside the formal meeting to reflect discussion and the Infostat document 30. The revised timetable is attached. This was agreed formally on Wednesday morning.

John Perry, Infostat and FYR of Macedonia participants also discussed the special treatment of FYR of Macedonia. Additional training would bee required, although the timing was not yet certain. The sample of 2000 would include 600 previously registered units. FYR of Macedonia will produce % active rates for legal and natural persons, for 1997 and previously registered units and for capital and other areas to allow the sampling programs to work. The sample size also remained provisional until a register extract had been taken to give counts of numbers of units registered in 1998 and for earlier by legal and natural persons. Infostat would make all programs and necessary additional documentation available immediately.

All other countries would also now immediately take a register extract, unless they had already done so.

Wednesday 28 April 1999: 0900 to 1230

6. Making population estimates

Document 35: Longitudinal database logical design issues

This paper was a detailed exploration of the issues. Several points were made in discussion that indicated that the issues raised were extremely important and warranted further examination prior to the next project group meeting.

Infostat was requested to produce a further paper for that meeting describing briefly the definitions for unique sequence numbers, consistent file types, unambiguous coding for RY and other dates reflecting the “year 2000” problem and standard country codes. This should also include a programme for revising all files and documentation where they were still relevant to the project. This was particularly important, if data were to be explored by researchers.

The work by Hans Velt and Poland on longitudinal estimates also indicated that the size and complexity of files needed to be given further consideration.

Paul Smith suggested that the confusion between survey and questionnaire codes within the file type variable could best be resolved by separating into two variables: on describing the survey; the other relating to the questionnaire. As examples he quoted:

Infostat was requested to explore this option.

Other comments on the paper should also be sent to Infostat

Document 36: Longitudinal estimates

The work was originally to be part of the C survey publication but was excluded from the contract. A new contract was set up but was delayed and progress has been limited as a result. The outcome of the work will be estimate of active, dormant and dead units for 1/1 of 1995, 1996 and 1997 for Poland, external validation for three countries and similar estimates for all CECs.

The key task is to impute cessation dates using analysis of distribution of company lifetimes. From May to July, population estimates will be made for all CECs. Comments will be requested from CECs and a final report will be presented at the fourth working group meeting.

Imputation is required for active units from A and B1 but not in C and non-active units not dead in A and B1 and not in B2. Four variables were used for imputation: unit type, PD2, PD6 and RY (actually a revised version of this called CATLRY). Eventually location (PD6) was shown not to have a significant relationship on the data for Poland. This may not be the case for other countries. More generally, the model may differ from country to country. For Poland, a lognormal distribution was then chosen as the model. The model might be improved by changing the explanatory variables, dealing with outliers and changing assumptions about RY (in particular the 05 and 06 variables).

In addition, it is necessary to impute missing creation dates. This is a complex issue, especially taking into account the limited scope of the study and the fact that communication with Poland takes place primarily by e-mail.

The contracts required checks to be made on two other countries. Those selected were Slovenia and the Czech Republic from the first wave of accession countries and Slovakia. The statistical offices will be contacted formally for their co-operation. Romania queried the selection criteria. Latvia queried the treatment of real start dates that could be before registration dates. Hans Velt responded that register data would be used as little as possible.

5. Project 2

Document 38: Updated proposals for the second DOSME project

Arto Luhtio stated that although the total PHARE funding had been reduced, allocation between projects could be altered, e.g. with business statistics regulation funding. Arto Luhtio will have a meeting next Friday.

We need to argue that there has been a substantial investment so far that could be wasted with reduced funding. We could involve the Chambers of Commerce. There was a strong links to the business structural statistics regulation. Support could be sought at the Warsaw seminar in June.

Two further options were suggested (the others being: proceed with full survey, drop the C survey, drop the second follow-up cases from C survey). The two options were: drop the B5 survey; and run only 50% of the C survey. The “drop B5” option seemed to have some support as it might be the least harmful to the overall objectives.


6. Next (4^th) meeting

The Lithuanian Statistical Office agreed to host the 4^th meeting in Vilnius from 25 to 27 October 1999, with the following provisional agenda:

Depending on the work to be done in preparing the B5/C survey, the meeting would be 2.5 or 3 full days.

The offer from Lithuania was accepted with thanks. Participants were thanked for their enthusiastic contributions. The Romanian NSI was thanked for its organisation of the meeting and hospitality and helpfulness throughout the meeting.

Overview of Documents

Document 1 List of participants (1^st meeting)

Document 2 Agenda (1^st meeting)

Document 3 Project description (this is a copy of Document 78 of the previous CEC panel project)

Document 4 Contents and planning of the project

Document 5 Summary of computer systems in CECs

Document 6 Survey B3: sampling and questionnaire

Document 7 Draft contracts for the B3 survey

Document 8 The seminar on the use of business registers

Document 9 Report of 1^st meeting

Document 10 B3 questionnaire

Document 11 Progress report to 1 August 1998

Document 12 Summary of register situation in CECs

Document 13 Update on technological issues

Document 14 Recommendations for further analysis of the data collected in the C survey

Document 15 Characteristics, maintenance and uses of the business register

Document 16 B3 summary tables

Document 17  B4 survey sample and questionnaire

Document 18 Proposals for project 2

Document 19 Enterprises in the CECs: initial longitudinal analysis 1995-7 - draft results

Document 20 Items and questions on the review of the use of business registers in CECs

Document 21 Current longitudinal database and future outlook

Document 22 Place in the national statistical office of the pilot project

Document 23 Agenda for 2^nd meeting

Document 24  List of participants for 2^nd meeting

Document 25 Imputation examples

Document 26 Report of 2^nd meeting (Tallinn, Estonia)

Document 27 List of participants 3^rd meeting

Document 28 Agenda 3^rd meeting

Document 29 :   B3 report by CECs

Document 30 :   Infostat report on B3

Document 31 :   B3 publication

Document 32 :   C survey – exploitation of data

Document 33 :   Eurostat questionnaire summary

Document 34 : Planning for B4 survey

Document 35 :   Longitudinal database logical design issues

Document 36 :   Longitudinal estimates

Document 37 : Imputation issues – B3 and B4 surveys

Document 38 :   Updated proposals for the second DOSME project

Document 39 : Quality of coding of industrial classification

Document 40: Report of the 3^rd working group meeting