Division for Energy and Industrial Production Statistics

Division for Energy and Industrial Production Statistics

Anne Sofie Abrahamsen, Morten Qvenild Andersen, Britt-Inger Sande and Nils Petter Skirstad

Project: Development and Analysis of the Norwegian PRODCOM Survey

Final report

Preface

This report is based on contract No 2002 44501 002 between Eurostat and Statistics Norway, and is the final report on the project. The title of the project is Development and Analysis of the Norwegian PRODCOM survey. Project leader has been Mr. Morten Qvenild Andersen.

The report is written by Mr. Morten Qvenild Andersen, Ms. Anne Sofie Abrahamsen, Ms. Britt-Inger Sande and Mr. Nils Petter Skirstad at Statistics Norway. The steering group for the project has been Mr. Bjørn Bleskestad, Head of Division, Division for Energy and Industrial Production Statistics, and Mr Frank Foyn, Statistical Adviser, Division for Energy and Industrial Production Statistics. Reference group for the project has been Ms. Randi Hallen, Division for National Accounts and Mr. Trond Sandmo, Division for Environmental statistics.

Statistics Norway, Division for Energy and

Industrial Production Statistics

Oslo 29 September 2003

Contents

1Background and introduction

2Summary

3Development and analysis of the PRODCOM survey

3.1Background for the Norwegian PRODCOM survey

3.1.1Main results 2001, comparisons between PRODCOM and NACE

3.2The unit for the statistics

3.2.1Data collection unit

3.2.2Distribution of subclasses in enterprises

3.3Coverage and sampling methods

3.3.1Coverage and sampling

3.4Linking External Trade, PRODCOM and the National accounts

3.4.1New national PRODCOM codes

3.4.2External trade

3.4.3Comparisons between PRODCOM and External Trade

3.5Volume imputation in PRODCOM

3.5.1Method of imputing missing values

3.5.2Outlier diagnostic

3.5.3Results from simulating price estimation.

3.6Better distribution on PRODCOM codes

3.6.1Background

3.6.2Analysis of the effect of using groups

3.7Methods for more effective editing

3.7.1Editing tools and remedies:

4Annexes

4.1More examples of outliers

4.2Questionnaires groups of PRODCOM codes

4.3Cut off limits lower than 20 where the limits are unchanged

4.4Example of information on kind-of-activity unit sent to the enterprises

4.5Editing database

1Background and introduction

This is the final report for the project Development and Analysis of the Norwegian PRODCOM survey, based on contract No 2002 44501 002 between Eurostat and Statistics Norway.

The work on the project started in 2002 and have focused on four main topics. According to the contract these topics are:

• The unit for the statistics
• The sample coverage and the sampling methods
• Linking the External Trade statistics with the PRODCOM Survey
• Methods for editing and calculation of missing values for quantities

The unit for the statistics is discussed and analysed in chapter 3.2. In chapter 3.3 we present information and analyse the coverage and sampling methods for the statistics. The possibilities of linking External Trade statistics to PRODCOM are discussed in chapter 3.4. Methods for calculation of missing values for quantities, and methods for more effective editing are presented in chapter 3.5 and 3.7. In chapter 3.6 we present background and some results from efforts in the 2002 survey to get better specifications from enterprises for the distribution of sold production on PRODCOM codes. This can be seen as a part of the last topic of the project.

Our aim with the project has been to improve the data from the PRODCOM survey for use by international organisations (Eurostat) and for national users and finding ways of improving the quality of the final figures. We have for national users focussed on the needs from the National Account and the Environmental statistics. We have based our analysis mostly on the 2000 PRODCOM survey and have implemented some changes, based on the results of the project, in the 2002 PRODCOM survey. The rest we plan to implement in the 2003 PRODCOM survey.

2Summary

In chapter 3.1 we present the background for the Norwegian PRODCOM survey as it is today and we present some main results from the 2001 PRODCOM statistics. This gives some information on the structure of the manufacturing sector in Norway.

We have in chapter 3.2 examined how the structure of the enterprise included in the survey is and the consequences to alter the data collection unit for the PRODCOM survey. Some of the reasons to change the data collection unit are

• A better link to the structural business statistics
• Better information basis for estimation of units outside the sample
• Better information on production within a NACE group
• Demand for kind-of-activity figures from large users as the National Accounts and the Environmental statistics

An advantage by changing the data collection unit is that it is possible to produce PRODCOM figures both on the enterprise and the kind-of-activity level.

The decision was made to choose local kind-of-activity unit as the basis for the sampling for enterprises with production within more than one NACE subclass for production. However, for some enterprises using the local kind-of-activity unit instead of kind-of-activity unit would mean a much larger burden on the enterprises. For these enterprises we have used the kind-of-activity unit. Only small changes in the sampling were necessary to be able to get information needed to make figures for the kind-of-activity units. I t was only needed to change the data collection unit for 122 enterprises. The changes were implemented in the 2002 PRODCOM survey

In chapter 3.3 we presents the coverage of the Norwegian PRODCOM from two angles, the coverage compared with the EU-15 and the coverage compared with the total production in Norway measured by the Norwegian Structural statistics. We wanted to find out if the sample did give us a good coverage of the Norwegian Production. As a result of the investigation we found that an adjustment in the sample should be made to improve the coverage for some NACE codes and we also decided to increase the total coverage by increasing the number of enterprises included in the sample.

The linking between External Trade and PRODCOM is important to be able to produce good figures for the National Accounts. In chapter 3.4 we have looked on this topic from two angles, improving the PRODCOM list by introducing new national PRODCOM codes and checking the coherence between the codes used in the two surveys for some larger enterprises. From 2003 the organisation number of the enterprise has been included on the customs declaration forms. This makes it possible to get a good link between the PRODCOM survey and External trade for the enterprises included in the PRODCOM survey. By linking the commodities used in the External trade to PRODCOM codes and comparing them with the codes used in the PRODCOM survey, we have started the work to get a better coherence between the figures from the PRODCOM survey and the External Trade survey in the 2002 statistics, and will continue the work in the 2003 statistics.

Chapter 3.5 focuses on volume imputation. We have found that there are two improvements that can be made to increase the number of volume figures. Outliers is always a problem for the imputation of missing figures. We find that by using an outlier model instead of a common ratio model for imputation for the codes where the imputation failed we were able to impute some more volume figures. For many of the volume units there are only missing figures from one or just a few enterprises necessary to be able to produce volume figures. By focusing more on this during the editing process it might be possible to get this information from the enterprises.

In chapter 3.6 we present the results from trying to increase the number of PRODCOM codes used by the enterprises. The background for this is that enterprises tend to distribute their production figures only on the pre-printed PRODCOM codes on the questionnaire, and thereby not specifying all commodities correctly, We have tried to pre-print more PRODCOM codes on the questionnaires to see if this makes the enterprises distribute the sold production better. We found a little increase in the use of PRODCOM codes but also found an increase in the use of "other" codes which might be a problem for the statistics.

In chapter 3.7 we examine the editing tools and remedies used in the editing process to find ways of improving the process, both to get better timeliness and better quality of the Norwegian PRODCOM figures. We have made some improvements in the database, made a new checking lists, changed some of the procedures and made other tools and sources for checking the PRODCOM data accessible for the editing personnel.

3Development and analysis of the PRODCOM survey

3.1Background for the Norwegian PRODCOM survey

PRODCOM is the title of the EU production statistics for Mining and Quarrying, Manufacturing, and Electricity, Gas and Water Supply, i.e. Sections C, D and E of the Statistical Classification of Economic Activities in the European Community (NACE Rev. 1). The first four digits of the PRODCOM code correspond to the classes of NACE Rev. 1.

As the EEC-treaty came into force 1 January 1994, Statistics Norway choose to alter its production statistics from a HS-based survey with local kind-of-activity unit as data collection unit to a survey based on PRODCOM with enterprise as the data collection unit. So the first reference year of the Norwegian PRODCOM survey was 1995. The survey has since been based on the PRODCOM list for the prevailing year, and only annual data have been collected. One of the changes in the survey has been that the previous production statistics consisted of actual production, while the PRODCOM survey mainly consider sold production, even though total production is gathered for some of the codes.

The main users of the Norwegian PRODCOM statistics are Eurostat and the National Accounts of Statistics Norway.

3.1.1Main results 2001, comparisons between PRODCOM and NACE

The Norwegian PRODCOM survey for 2001 was based on information from the 2,507 largest enterprises with production within manufacturing. The sample consist of all enterprises with at least one local kind-of-activity unit with 20 persons or more employed in mining quarrying and manufacturing. In some divisions the limit has been set to 15 or 10 persons employed. The sold production for these 2.507 enterprises was NOK 375 billion in 2001.

In Table 3.1.11 we have made aggregates of 2 digit PRODCOM for sold production. As we can see from the table the largest division is food products, beverages and tobacco. The large figures for other transport equipment is due to the large oil industry in Norway. The production of oil platforms and modules has a large share of this production. Norway has also a long history of building of ships. A large share of the Norwegian production are in the raw material and semi finished goods industries. Norway is a large producer of aluminium, chemicals, wood and wood products and fish. The refining of oil products is also large but there are only two producers and the figures are confidential. The largest industry within sections C, D and E of NACE in Norway is Extraction of crude petroleum and natural gas. For this industry we have an own survey with its own publication and the results are not included here. In the later years there has been a tendency of outsourcing the production units to foreign countries. This is something which is and will be a challenge for the production of industrial statistics.

Table 3.1.11 Sold production distributed by the first two digits in the PRODCOM code

2 digit PRODCOM / Sold production
Mill. NOK
10,12-37Mining, quarrying and manufacturing / 374,620
10, 12-14Mining and quarrying / 4,260
15-16Food products, beverages and tobacco / 84,396
17-19Textile and textile products, leather and leather products / 3,729
20Wood and wood products / 12,687
21Pulp, paper and paper products / 17,297
22Publishing and printing etc / 16,839
23Coal and refined petroleum products / :
24Chemicals and chemical products / 27,509
25Rubber and plastic products / 5,205
26Other non-metallic mineral products / 10,954
27Basic metals / 42,577
28Metal products, except machinery and equipment / 13,954
29Machinery and equipment n.e.c. / 22,330
30-33Electrical and optical equipment / 26,026
34Motor vehicles, trailers and semitrailers / 4,574
35Other transport equipment / 45,104
36-37Manufacturing n.e.c. / :

An enterprise may include several local kind-of-activity units (LKAUs) with activity within different sectors. Since an enterprise classified outside manufacturing may have LKAUs within manufacturing, some of these enterprises are included in the sample for the PRODCOM survey.

By aggregating sold production on a two digit level for the NACE code of the enterprise we get a somewhat different distribution than in Table 3.1.11. As we can see from Table 3.1.12 (next page) there is mostly in the construction and the wholesale and trade industries where we find enterprises with LKAUs within manufacturing outside the sections C, D and E.

Table 3.1.12 Sold production distributed by the NACE code of the enterprise

NACE Industry division of the enterprise / Number of enterprises / Sold Production
Mill. NOK
Total / 2,507 / 374,620
5Fishing / 3 / 610
11 Extraction of crude petroleum and natural gas / 2 / :
10,12-14 Mining and quarrying / 68 / 4112
15-16Food products; beverages and tobacco / 397 / 80,962
17-19Textile and textile products, leather and leather products / 102 / 3,712
20Wood and wood products / 243 / 12,516
21Pulp, paper and paper products / 37 / 16,087
22Publishing and printing etc / 269 / 17,013
23Coal and refined petroleum products / 0 / 0
24Chemicals and chemical products / 62 / 30,734
25Rubber and plastic products / 78 / 5,070
26Other non-metallic mineral products / 96 / 10,246
27Basic metals / 52 / 42,641
28Metal products, except machinery and equipment / 327 / 12,641
29Machinery and equipment n.e.c. / 177 / 23,315
30-33Electrical and optical equipment / 132 / 21,904
34Motor vehicles, trailers and semitrailers / 37 / 4,908
35Other transport equipment / 221 / 47,174
36-37Manufacturing n.e.c. / 166 / 9,527
40-41 Electricity, gas, steam and hot water supply / 1 / :
45 Construction / 8 / 476
50-52 Wholesale and retail trade / 22 / 12,706
55 Hotels and restaurants / 1 / :
60-64 Transport, storage and communication / 1 / :
70-74 Real estate, renting and business activities / 2 / :
85 Health and social work / 1 / :
90-93 Other community, social and personal service activities / 2 / :

3.2The unit for the statistics

3.2.1Data collection unit

We have had a dialog with the reference group for the project regarding changing from enterprise to local kind-of activity unit as the data collection unit for the Norwegian PRODCOM survey. Through this discussion we found that neither in the Norwegian National Accounts nor in the Environmental statistics the regional dimension is used for the production statistics, but they have a demand for figures on the kind-of-activity unit. The National Accounts and the Environmental statistics are also the most important users, besides Eurostat, of the Norwegian PRODCOM figures. Some of the reasons for change the in data collection unit are

• A better link to the structural business statistics
• Better information basis for estimation of units outside the sample by getting information on more homogeneous units.
• Better information on production within a NACE group
• Demand for kind-of-activity figures from large users as the National accounts and Environmental statistics

National Accounts estimate aggregates based on the PRODCOM survey for all NACE codes on a 3 digit level including figures for enterprises not included in the sample. This will be much easier to do when they get information on the kind-of-activity level.

An advantage by changing the data collection unit is that it is possible to produce figures both on the kind-of activity unit and the enterprise. On this basis we decided to look into the consequences of changing the data collection unit.

One problem by choosing either local kind-of-activity unit or kind-of-activity as data collection unit has to do with the handling of internal deliveries. We decided not to alter the questionnaires used for enterprises, but underlined in the guidance that it was sold production we are asking for. We also sent out the questionnaires for kind-of activity unit together with the questionnaires for the structural business statistics and used the same deadline for all the questionnaires. This means that we will for many of the enterprises be able to use the information from the structural business statistics in the editing process of the PRODCOM statistics. The structural business statistics has local kind-of-activity unit as the data collection unit, and we ask for internal deliveries. In this way we are able to check for the differences between the sold production adjusted for internal deliveries from the structural business statistics with the figures from the PRODCOM survey.

In chapter 3.2.2 we show how we investigated the enterprises included in the 2002 survey to find out how to adjust the sampling to be able to get information on the kind-of-activity unit and also having the opportunity to make good figures for the enterprise. For enterprises with only one kind-of-activity unit, or with more than one kind-of-activity unit but only one within production, the PRODCOM figures for the kind-of-activity unit will be the same as the figures for the enterprise. So for these enterprises no changes are needed. When it concerns the other enterprises there is a need for a change in the data collection unit. For these enterprises we have to use either local kind-of-activity unit or kind-of-activity unit as the data collection unit. The enterprises included in the sample of the PRODCOM survey all get questionnaires for the local kind-of-activity unit in the Structural Business Statistics survey. This means that they are used to reporting on this unit. We have also already a production system with enterprises and local kind-of-activity units in the PRODCOM survey. This is due to the fact that we have earlier collected the production statistics for 10-20 enterprises on the local kind-of-activity level, due to the special needs of the environmental statistics. An advantage by using the same data collection unit in the two surveys, is that it makes it easier to check the coherence between the data. We decided that in the cases that the local kind-of-activity unit corresponds with the kind-of-activity unit, we would use the local kind-of-activity unit as the data collection unit. The same decision was made for enterprises with one or two kind-of-activity units less than the number of local kind-of-activity. For enterprises with more than 3 local kind-of-activity units more than the number of kind-of-activity units we have used kind-of-activity unit as the data collection unit