EUROSTAT
Directorate G: Business Statistics
Unit G-2: Structural Business Statistics /
Report on Eurostat Quality Visit
to Statistics France
Version: / 1.00
Authors: / Mihaela Agafitei
Revised by: / Brian Williams
Approved by: / Merja Hult
TABLE OF CONTENTS
1. Purpose of the visit 3
2. Institution 3
3. Survey 3
4. Definition of the target population 3
5. Sampling Frame 3
6. Product nomenclature 3
7. Sampling design 3
8. Method of data collection 3
8.1. Questionnaire 3
9. Data processing 3
10. Measurement errors 3
1. Purpose of the visit
The Eurostat Prodcom team is conducting a number of "Quality Visits" to the Prodcom teams in the NSIs, in order to discuss the methodology used, detect potential problems and identify best practice that can be shared with other countries. The visit to Statistics France is the 6th of these visits.
After each visit, Eurostat will prepare a report on the discussions and, after checking it with the team visited, will load it to the Prodcom CIRCA website, together with any material such as Powerpoint slides provided by the team in the NSI.
2. Institution
Official statistics in France are produced by statistical services under different bodies of the French government and is coordinated by the National Institute of Statistics and Economic Studies (INSEE) which is a general directorate of the French Ministry of Economy, Industry and Employment.
Ensuring the designing and coordination of national statistical surveys, the Directorate General of INSEE includes:
- a general secretariat which manages all the resources of the Institute
- a general service which conducts audits and assessments,
- six units that contribute to the missions of INSEE:
The Unit of Statistical Coordination and International Relations
The Unit of Business Statistics
The Unit of Demographic and Social Statistics
The Unit of Economical Analysis and Summaries
The Unit of Dissemination and Regional Action
The Unit of Dissemination of Regional Statistics
3. Survey
“The Annual Survey of Production” (EAP) has been carried out since 2009 (2008 as observed year) and replaces two previous business surveys: the Annual Enterprise Survey (EAE) and “the Annual Survey of Industry (EAB)[1]”. It was designed so that to minimize the information requested by limiting them to those INSEE does not have elsewhere.
The EAP has two major objectives:
- Identify the different activities performed by enterprises through a very detailed breakdown of their turnover in industry, which form the basis for identifying the main activity of enterprises.
- Provide data on industrial production in order to meet the requirements of the EU Regulation on Prodcom and to meet the demands of domestic users, including the demands of professional organizations.
In addition, the EAP contributes to a better quality of short term statistics being a valuable auxiliary source of information (i.e. to apply calibration method).
The major advantage of this new survey design is that it ensures a high coherence between industrial statistics (Prodcom), SBS and STS by a single interrogation of the surveyed enterprises. The INSEE has estimated that, in doing so, the burden on business is reduced by about 60% while the level of detail is increased considerably.
4. Definition of the target population
The target population consists of all enterprises which carry out an activity listed in section B (“Mining and quarrying”), C (“Manufacturing”), D(“Electricity, gas, steam and Air”) and E (“Water supply, sewerage, waste management and remediation activities”) of NACE Rev 2.
5. Sampling Frame
The sampling frame named FUL is drawn from the French Business Register – SIRENE – and contains all active enterprises which, on French territory and during the reference period, carry out an activity listed in section B (“Mining and quarrying”), C (“Manufacturing”), D(“Electricity, gas, steam and Air”) and E (“Water supply, sewerage, waste management and remediation activities”) of NACE Rev 2. The size of sampling frame (FUL) is about 170,000 enterprises.
The sampling frame is drawn from FUL in November of the reference year and the information contained is related to the year before the reference period.
The average coverage error contained in FUL is estimated at about 4%. As the highest incidence of coverage errors is observed to very small enterprises, they do not have significant impact on Prodcom data.
6. Product nomenclature
The national product nomenclature (PRODFRA)[2] is based on the Prodcom list (8 digits) supplemented by some more detailed headings (9th and 10th digit) used for national purposes
The French nomenclatures of activity and product are coherent with the European nomenclatures as follows:
Sometimes, a different volume unit is used in order to facilitate the response but the data transmitted to Eurostat use the volume unit requested by the Prodcom list. For those cases where a direct correspondence between volume unit used by the PROFRA nomenclature and the volume unit requested by Prodcom does not exist, a coefficient of equivalence is set based on consultation with professionals.
The 2009 PRODFRA contains 4695 headings and 4540 headings are linked to the PRODCOM list. So about 3.3% of national codes are outside the scope of Prodcom. For 359 Prodcom headings (9.3%), PRODFRA uses more detailed national codes.
Data at PRODFRA heading level is converted into PRODCOM headings before transmission to Eurostat.
In general, the PRODFRA is updated yearly, the rate of change is close to the PRODCOM rate (national changes are not frequent).
7. Sampling design
The INSEE uses a cut-off sample supplemented by a small random sample. So all information related to the units over a threshold is collected via the survey. The EAP overall sample (about 40,800 units) includes the following types of units:
- all enterprises with 20 or more employees or with a yearly turnover of more than a threshold defined by activity (regardless of numbers of employees) and having their main activity within the scope of the survey. There are about 30,000 units.
- for those NACE classes where the total turnover of sampled units represents less then 85% of the total turnover by NACE class, a simple random sample is drawn from the enterprises under these thresholds and having their main activity within the scope of the survey. In order to minimise the burden on businesses, a rotation schema of 5 years is used for the sample. There are about 10,000 units.
- all enterprises having their main activity outside the scope of the survey but having a secondary activity within it, over a threshold depending on the turnover of the branch in the same way as described previously. There are about 1000 units.
8. Method of data collection
On-line (Internet) and paper questionnaires. The enterprises have to report by Internet, exceptionally a paper questionnaire is sent to them. Currently, over 85% of the respondents use the Internet, the others still prefer the paper questionnaire. The questionnaire is customised. The questionnaire design does not allow the enterprises to see and correct their previous data.
The data collection is carried out by another statistical service (located in Caen – Normandy) in charge of data collection of enterprises in all the industrial sectors. Data collection is organised by sector; it starts a month after the end of the survey period and lasts until late June. In between there are two reminders, a threat of a fine and finally, if the enterprise fails to respond by the deadline, it receives a fine.
The overall average response rate is over 90% of the total turnover.
8.1. Questionnaire
The annual survey of production (EAP) focuses on the analysis of the turnover of enterprises. This analysis is very detailed for the sales of goods or industrial services and it is more general for the rest.
The information requested through the EAP questionnaire is organised as following:
Global indicators:
- The number of employees (all persons having a work contract and
paid directly by the enterprise) - The total turnover which is split into five categories:
The sale of industrial products, which covers all products of the manufacturing and
extractive activities as defined in statistical classifications, except division 33.
Installation of industrial products (33.2)
Repair and maintenance of industrial products (33.1)
Other sales of industrial products, covering all industry products that are not classified
in the manufacturing and extractive industries (NACE sections D and E).
The sale of non-industrial products/services.
For each category of turnover the questionnaire requests more detailed information. The description below covers the aspects relevant to the first category (mainly Prodcom statistics)
Turnover arising from the sale of industrial products
In this part, the turnover coming from the sale of industrial products is split into five economic models, by showing the percentage for each model by product (as defined by the PRODFRA product nomenclature):
Model 1 – the product sold was purchased from the market and not manufactured by the enterprise, or the manufacturing was entirely outsourced to a subcontractor without providing any input.
Model 2 – the manufacture of product was completely outsourced to a subcontractor and all necessary inputs were provided to subcontractor free of charge.
Model 3 – the product was entirely manufactured by the enterprise itself (including
assembly of purchased components).
Model 4 – the product was manufactured for another enterprise which designed the product or owns copyright or patent on the product the product, without receiving any input from that enterprise free of charge.
Model 5 – the product was completely manufactured for another enterprise and the main part some or all of the inputs were provided by that enterprise free of charge or partial processing of the principal’s products
In addition the following indicators are requested by product:
- The quantity of sold production which must correspond to the value of sold production
- The quantity of total production
- The breakdown of the turnover coming from the sale of the industrial product by destination nature: wholesale or retail
A general picture of the five economic models is shown in the following chart:
Industrial branch
9. Data processing
A lot of micro and macro plausibility checks are performed at different aggregation levels (e.g. local KAU, KAU, enterprise, NACE class, NACE division, CN or Prodcom headings) and by comparison with historical data. To do this, the following information is used in the data processing:
- current/previous Prodcom data
- VAT data
- external trade data
- SBS data
Based on these checks the following decisions and actions are taken during data processing:
- Questionnaire assessment: different types of error are assigned by questionnaire and a decision is taken whether or not to validate the questionnaire. If the questionnaire is not validated, based on the priority rank, the statistician in charge calls the enterprise for clarifications and corrections.
- The imputation and estimation phase: based on different data analyses, the missing data is imputed and erroneous data that are not corrected by the enterprise (see the questionnaire assessment) are replaced by estimations. The imputation/estimation applies for both the current and previous reference years. So two datasets, corresponding to the current and previous reference year, are analysed and, depending on the situation, an amount either for the current reference year or for the previous year could be imputed or estimated.
The Oracle system (in-house applications) is used to perform data processing.
10. Measurement errors
The main problems encountered by enterprises? during data collection are
- to report in the requested measurement unit;
- to classify products;
- to provide complete data.
The following actions are taken in order to minimise measurement errors:
- During the whole data collection process, the staff of the regional service offices maintain direct and close contact with the surveyed units. This is the case especially when a unit encounters problems in providing an estimation for the volume data or in classifying a product to a PRODFRA heading.
- A lot of effort is devoted to these contacts and it is sometimes difficult to find the right person in the unit to provide the information required;
· Ensuring easy and user-friendly access to the product nomenclature, including a search engine, indexes and correspondence tables.
- Providing respondents with a free electronic listing of the national product classification.
Report on Eurostat Quality Visit to Statistics Germany - Page 7 / 8
Document Version 1.00 dated ………….
[1] The EAB was one of the oldest business surveys carried out in France being set up in 1938. A part of theses surveys, formely a large part, were performed by professional organisations under INSEE supervision, this collaboration with INSEE ended in 2008 for quality reasons The EAB surveyed production by recording the quantities of sold production, value of sold production as a sum of invoices, total production and orders for industrial products, no result by activity branches.
[2] PRODFRA + technical lines only used by EAP is callded the PRODEAP list or simply PRODEAP (5507 headings)