The target population of the survey is made up of all enterprises contributing to French industrial production. The sampling frame is taken from the statistical business register (SIRUS) which determines the scope of business statistics based on the information available in the national register of legal units and establishments (SIRENE). The sampling frame thus composed of industrial enterprises, or non-industrial enterprises but with an important industrial branch, comprises about 150 000 legal units. Of these, about 35 000 are surveyed, covering 90% of French production for each industry sector.

Annual

Businesses can respond by internet or by paper. Each enterprise surveyed receives a letter with its connection parameters to the INSEE collection site. Paper questionnaires are only sent to companies that request them.

Whether on the internet or on paper, the questionnaire is personalised for each business, based on the products manufactured in previous years, or on those manufactured for businesses in the same sector of activity, for businesses that have never responded to the survey. However, businesses can add products not included in their questionnaire by accessing the exhaustive list of products available on the collection site.

A large majority of respondents (97%) use the online questionnaire.

The collection is annual and takes place from January N+1 to September N+1 with, however, a concentration on the first half of the year in order to provide a response to PRODCOM by 30 June N+1.

By Internet

Legal unit

Businesses with more than 20 employees or more than 5 millions turnover are surveyed exhaustively. For the smallest enterprises, a random draw by stratum (NAF subclass * number of employees) is made.

35,000 enterprises are surveyed

Data validation involves two steps :

• validation of individual data : each questionnaire is checked by an automatic chain of checks, which also adjusts for partial non-response. There are consistency checks with other sources (e.g. comparison of turnover with VAT data), as well as temporal checks and internal coherence checks. Some checks are blocking and others only warnings. Questionnaires judged to be of poor quality at the end of the checks, or containing blocking errors, are processed manually ;

• validation of aggregates: aggregates are subject to a score calculation taking into account the N/N-1 evolution of the aggregate, the largest contribution to the evolution of the aggregate and the extent of the contributions of enterprises to the evolution of the aggregate. If this score exceeds a certain threshold, the aggregate is considered suspect and is analysed manually.

The disseminated results are obtained by summing the output of all in-scope legal units. No weighting is therefore used, but the imputations of non-surveyed and non-respondents.

Every unit/product couple is then estimated at the beginning of the campaign (adjustment for total non-response). These estimates are then replaced by the respondents' answers, once the questionnaires received have been validated.

For total estimates (total non-response), the value of the sold production is estimated with the value of the previous year, multiplied with the evolution of the unit's turnover from the VAT source, if this information is available, otherwise with the rate of evolution in the unit's sector, again from VAT. The same methodology is applied to the sold quantity and total quantity. It is then assumed that the product structure remains unchanged. If the unit never responded before, there is no estimation.

For responding units, the questionnaires may be adjusted for partial non-response :

• breakdown of billings by economic mode : either the N-1 data are used if they exist, or 100% of billings are allocated to a particular model, depending on the type of product ;

• sold quantity : either the unit price N-1 is applied if it can be calculated, or a median unit price is applied to the billings ;

• total quantity: either the sold quantity is used if it exists, or the total quantity N-1 is used, to which the evolution of billings is applied if it can be calculated. The adjustment is only made if the sold quantity and the total quantity are expressed in the same unit.

No particular adjustment is made

Not applicable