French ministry of agriculture and food  (Ministère de l'Agriculture et de l'Alimentation):

http://agriculture.gouv.fr/

Statistical and Foresight Department (Service de la Statistique et de la Prospective):

http://agreste.agriculture.gouv.fr/

Department of agricultural, forest and agrifood statistics (Sous-direction des statistiques agricoles, forestières et agroalimentaires: SDSAFA)

Unit of structural, environmental and forestry statistics (Bureau des Statistiques Structurelles, Environnementales et Forestières: BSSEF)

MAA- SG

SSP-BSSEF
Complexe agricole d'Auzeville
BP 32688
31326 CASTANET TOLOSAN

The first farm structure survey was conducted in France in 1967 for European needs. Between 2000 and 2010 (census), there were 3 FSS : 2003, 2005 and 2007.

Since 2010 census, there were 2 FSS : FSS2013 and FSS2016.

FSS surveys include Eurostat's needs, and some national questions.

For the first time, the 2016 survey included the orchard survey for Eurostat needs (Regulation(EU) No 1337/2011). We carried out in 2013 an orchard census for these Eurostat needs.

The legal framework for French FSS survey is the Eurostat legal framework : Regulations 1166/2008 and 715/2014. 

In 2016, it also includes orchard statistics legal framework: Regulation 1337/2011.

The statistical service of the French Ministry of Agriculture , which is an integral part of the public statistics system, acted as contracting owner and project manager.

Our national statistical programme includes surveys for Eurostat needs and  specific national surveys, as for agricultural practices (for vineyards; fruits; vegatables; cereals; animals...).

The law obliges farmers to give precise answers to the questions asked.
Failure to reply or a deliberately inaccurate response gives rise to court proceedings, and an administrative fine is imposed as the final penalty.

The definition of agricultural holding used in France is that set out in Regulation (EC) No 1166/2008. 
The agricultural holding is defined, for the purpose of agricultural statistics, as an economic and production unit which meets all three of the following conditions:
- its activity is agricultural in nature,
- it is subject to independent everyday management,
- it is of a certain size or larger (area, number of animals, production, etc.) according to the thresholds defined in the tables presented in 2.6 Statistical population

The FSS 2016 covers the entire scope of the EU Regulation.

Holdings with land maintained in good agricultural and environmental conditions, even if they do not have other agricultural activity, are in the scope of the survey.

See 2.6 for thresholds.

• In 2010, according to our census results, there were 491 000 agricultural holdings in mainland France and around 25 000 agricultural holdings in overseas departments (Guadeloupe, Martinique, La Réunion, Guyane).

• FSS 2016 universe is our national agricultural register, which is our sampling database called BALSA. Balsa is constructed with 2010 census farms, plus administrative data and surveys results in order to update 2010 census and to add new farms. The total number of farms of this universe for FSS sample was 546 285.

National Thresholds

The general threshold is 1 hectare of UAA, except for certain types of specialised production, where it is lower:

- The threshold is 20 ares for the following crops :

• hop (B_1_6_2),
• tobacco (B_1_6_1),
• aromatic, medicinal and culinary plants (B_1_6_12)
• vegetable, floral, fodder or industrial seeds
• fresh vegetables, melons and strawberries outdoor - open field (B_1_7_1)
• floral and ornamental crops (B_1_8)
• permanent crops (vineyards, orchards, berry plantations) (B_4)
• tree nurseries (B_4_5) : wine-producing, ornamental, orchard, forestry.

- If the holding does not own at least 1 ha of UAA or is not over the previous thresholds, the survey covers units that have :

The survey uses a threshold of 1 hectare of utilised agricultural area, thus art 3.2 is not applicable.

The thresholds used for farms are lower than the ones proposed by the European regulation.

With these thresholds, holdings with at least 1000 poultry or 10 bovine animals are covered.

Geographical coverage

The geographical scope is mainland France and four overseas departments: Guadeloupe, Martinique, Guyana and Réunion.

In Guyana, the FSS only concerns coastal zone or  reachable areas ; which represents 50 % of total farms : based on 2010 census results, 3000 farms are excluded from the geographical scope since, there are less representative of Guyana agriculture, and collecting data in this area is very expensive (5 983 farms in total in Guyane in 2010).

In the French overseas departments, the territories of Saint-Martin, Saint-Barthélemy and Mayotte are excluded.

There are 55 units (without weights) with a standard output equal to 0.

- 27 are with permanent grassland no longer used for production purposes and eligible for the payment of subsidies ; kept in good agricultural and environmental conditions.

- 4 units are with animals which are not economically valued based on standard output coefficients :

     1 unit fattening rabbits (1927932)

     1 unit breeding pheasants (1584266)

     1 unit breeding wildboars (1853926)

     1 unit breeding Angora rabbits (2103625)

- 24 units with 'chicons' (endive).

The thresholds used for farms are lower than the ones proposed by the European regulation:

•hops (B_1_6_2),

•tobacco (B_1_6_1),

•aromatic, medicinal and culinary plants (B_1_6_12)

•vegetable, floral, fodder or industrial seeds

•fresh vegetables, melons and strawberries outdoor – open field (B_1_7_1)

•floral and ornamental crops (B_1_8)

•permanent crops (vineyards, orchards, berry plantations)(B_4)

•tree nurseries (B_4_5): wine-producing, ornamental, orchard, forestry.

The criteria used to determine the location of the holding is the location of its main agricultural building, or the location of the biggest parcel (physical size) if there are no buildings.

The reference period for data on crops, production methods and labour is the 2015-2016 agricultural production year (1 November 2015 to 31 October 2016). 
For livestock, the date is 1 November 2016 or, in case of cull, the day before the cull.

For rural development measures : measures approved for 2014, 2015 or 2016.

National organisation

The Statistical and Foresight Department (Service de la Statistique et de la Prospective) of French Ministry of Agriculture is in charge of FSS surveys  :

- a project manager is working with other people in the unit of structural, environmental and forestry statistics (BSSEF) ; 

- the Methods and Information technology Unit (BMIS) of the Statistical and Foresight Department provides support on sample selection, CAPI development, weights calculation, quality indicators.

- 1 person responsible for IT systems

- 1 person responsible for methods

Regional Organisation (NUTS 2 or merged NUTS 2)

In each French region (NUTS 2 or merged NUTS 2 areas: 13 in regions  + 5 in overseas areas), there is a regional statistic and economical service for data collection, data validation and data dissemination. They are in charge of recruiting, training interviewers, planning and monitoring the work of the interviewers, data collection and first level validation.

For FSS 2016, 881 interviewers took part in data collection.

Weight calculation were longer than expected.

We did different weight calculation regarding first aggregated results.

See 3.5 for weight calculations and 6.3.1 for coverage issues.

Our data source is BALSA. The scope of 2016 Farm Structure Survey is a total of 546 285.

BALSA was constructed with 2010 census farms, and has been updated since 2010 with:

• administrative data : French business register, IACS (CAP area declarations database), BDNI data (Bovine animals register)
• national survey results.

The 2016 data is collected for a sample of holdings for which some characteristics are taken from administrative sources.

We used our agricultural statistical register (BALSA) for the FSS 2016 sample.  It is a list frame of agricultural holdings with several characteristics useful to realize samples:

- farm type based on 2010 census results which may have been updated. For a few farms, we used administrative information from business register  (NACE code);

- Agricultural area characteristics if any, from 2010 census results or from IACS;

- Information on animals if any, from 2010 census results or from our bovine register (BDNI);

Our Statistical register is updated with administrative data once a year. It was updated in July 2016 for the FSS 2016 sample.

FSS sample is a stratified sample. We used a single stage sampling. Sampling units are agricultural holdings.

The sample is a balanced sample by NUTS 3 on standard output,  legal status and type of farming for organic farms and non organic farms.

The stratification is based on:

- NUTS3,

- Farm type :

• 1) field crops
• 2) Horticulture, vineyards
• 3) Fruits and Other permanent crops
• 4) Grazing livestock
• 5) Granivores (Pigs and Poultry Distinguished in FR52)

- standard output :

• Small farms : SO <= 25 000 €
• Medium farms : 25 000 € < SO <= 100 000 €
• Big farms : 100 000 € < SO up to 1 500 000 €

- organic production : YES / NO

Specific strata

Specific strata were defined for : 

- collective units

- vacant units

- FADN units

- full coverage strata

In total, there are 3 925 strata in FSS 2016 sample.

There are full coverage strata for :

- units with standard output above 1 500 000 € and / or

- with 50 permanent salaried people or more and / or

- producing 500 tons of mushrooms or more and / or

- producing 300 tons of endives (or chicons) or more in some regions - FR21 / FR22 / FR30 and / or

- unusual farms regarding their economical size (the size criterion depends on the region and the production considered)

- FADN farms (which were not selected in the sample: 5 701 units).

The sample size has been defined in order to meet several requirements: 

- the cost of the survey: maximum 70 000 units in total

- RSE by NUTS 2 under 5 % for productions in Annex IV of 1166/2008 Regulation;

- RSE by NUTS 2 under 5 % for other productions:

non dairy cows / breeding sows / vegetables - FR52

poultry - FR26

potatoes - FR22, FR30

sugar beet - FR30

other bovine - FR71,  FR43

sheep - FR61, FR63

- RSE by NUTS3 under 5 % for SO of organic farms; 

- RSE by NUTS3 under 2.5 % for SO of non organic farms;

- integration of the 7 500 FADN farms in the sample;

- RSE at the national level under 3 % for orchard characteristics.

In other NUTS 2 areas (FR 61, FR 81, FR 82, FR 71); common lands are selected with a precision requirement of 10 % regarding the utilised agricultural area.

Vacant farms: in a full coverage strata in NUTS 2 regions with less than 50 vacant farms. In other NUTS 2 regions, we set a limit of 50 farms. 

The official statistical service (SSP) of the Ministry in charge for Agriculture (MAAF) is entitled by law to access administrative registers, such as SIRENE, data from CAP aid declarations (IACS), BDNI (national identification database), etc. 

Please access the information in the file at the link: (link available as soon as possible)

CAP area declarations:
We thus only prefill with 2015 data for some crops with a good matching in FSS and CAP declaration.

CAP area declarations: ~= 80 % of farmers realise a CAP area declaration

BDNI:  2 % of units with bovine animals declared in FSS but not in BDNI (478)

CAP data only used for farms with cap payments, for the others, we only have data from 2010 census or FSS 2013 if in the sample.

BDNI:We considered that administrative data is true. We put 0 for bovine in FSS. 

Our management of metadata depends on the sources, as different units from our services are involved.

Between 2010 and 2020, FSS data is collected in 2013 and in 2016.

The FSS 2016 was conducted in face-to-face interviews. An interviewer visits holdings to get the required data.

We used computer-assisted personal interviewing (CAPI) on touch screen table PCs (Stylus).

When an interviewer was confronted with a first refusal, he/she had to inform the competent regional service, which tried to obtain a response. This procedure made it possible to solve most cases of refusal.

Regional services sent written reminder letters.

Number of holdings in the survey frame plus possible (new) holdings added afterwards

In case of a census 1=3+4+5

546 285

No holding introduced afterwards

Number of holdings in the gross sample plus possible (new) holdings added to the sample

Only for sample survey, in which case 2=3+4+5

71 341

No holding introduced afterwards

Number of ineligible holdings with ceased activities

This item is a subset of 3.

Number of holdings with unknown eligibility status

4>4.1+4.2

Number of eligible holdings

5=5.1+5.2

Number of eligible non-responding holdings

5.1>=5.1.1+5.1.2

Number of the records in the dataset 

6=5.2+5.1.2+4.2

The questionnaire contains specific questions for orchard Eurostat statistics needs.

We applied different levels of validation:

1) Validation rules integrated in the application (CAPI) 

There are different kinds of checks: minor or major errors, with or without compulsory comments.

There are data format checks, completeness checks, relational checks, consistency checks.

See annex for consistency checks.

Questionnaires involving major controls cannot be transmitted to the central server by the interviewer unless this is specifically requested.

In such cases, the regional service must correct the ongoing anomaly, possibly by asking the farmer to provide additional information.
Comments can be written by interviewer or statistical services in order to justify the persistence of an error.

2) Additional checks, during data collection, underlying some errors, for regional statistical services: Daily automatically updated with a specific online application related to CAPI.

• checks on administrative identification numbers (EDE / PACAGE / SIRET) for multiple identifiers,

• examination of maximum values at the NUTS 2 or merged NUTS 2 level for some characteristics,
• other checks: if there seems to be no production; checks on collective units; specific checks for some erros on CAPI integrated validation rules etc.

3) Validation rules applied on CAPI extraction: central staff (the same as integrated in the CAPI application and other ones) (with SPSS); including Eurostat validation rules.

4) Eurostat validation rules applied on the FSS 2016 file created: central staff; after national validation (with SPSS)

5) Control of aggregated data:

central staff  and regional staff: comparisons with 2010 census results on main indicators at the NUTS 2 or 3 level. 

central staff only: comparisons with annual agricultural statistics results for main categories of UAA or some results on animals (heads). See 6.3.1 for weight calculation.

Our steps to produce datasets are:

Export from CAPI application (csv files)

Set of corrections; calculations; validations; format adjustment

Validated dataset

Used to produce 2 datasets: FSS 2016 file for EUROSTAT (csv file)/ FSS 2016 national files (sav files)

In each stratum, design weights are: the size of the population in the stratum by the size of the sample in the stratum:

                N_h/n_h in stratum h

                N number of units in the population

                n number of units in the sample

We did several weight adjustments, particularly because we had coverage problems, see 6.3.1.

• Extreme weights were trimmed for units with vineyards; with fresh vegetables, melons and strawberries, UAA>0; ewes; and breeding sows within each stratum. 
• We adjusted the weights  for some ceased units for which the buyer was interviewed while he was not in the sample but in the population: weight = 1 for 2 680 units.
• 11 units were classified in the exhaustive strata regarding their size after data collection.

Some questions were added for national purposes :

     - administrative identification numbers,

     - farm management: tax system; details on rented UAA,

     - details on land categories (in particular for fruits),

     - questions on pasture management; livestock breeding capacities; poultry annual production,

     - quality certification schemes and marketing via low food-mile systems. These relate to production and marketing systems which are encouraged.

     - diversification via a body which is legally separate from the farm.

     - detailed questions on labour force; for exampe non-salaried people working on societies or partnerships.

Please access the information in the file at the link: (link available as soon as possible)

The main source of errors is coverage error. See 6.3.1.

There are some measurement erros for some specific questions related to soil and manure management practices.

The estimation of RSEs, expressed as a percentage, is equivalent to the coefficient of variation. It is estimated by the formulae:

Where   is the estimated variance of the estimator, and    is the estimator of the total.

The stratification accounts for the calculation of the double inclusion probabilities that will be used for the calculation of the variance.

A calibration on margins is carried out. We proceed as follows:

• in each calibration post-stratum, the variable Y is regressed on the variables used for margin calibration, with weighting as the weight before calibration, after which the residuals  of the regression are retrieved;

• the variable  is created : , where  is the weight at the output of the calibration on margins;

• this variable is used as input for classical variance calculation formulae, instead of Y (i.e. replace  with  ).

To estimate the double inclusion probabilities of units, the following formula is used:

The variance is then calculated as follows :

Tool

SAS macro developed within our agricultural statistical service ("macro CALVA")

- NUTS2 - permanent grasslands excluding rough grazing ; FR82 Estimated RSE = 10.5 %

- NUTS 2 - other pigs ; FR51 Estimated RSE = 7.1 %

- NUTS 2 - poultry ;

• FR51 Estimated RSE = 5.6 %;
• FR52 Estimated RSE = 6.6 %
• FR53 Estimated RSE = 12.3 %                          
• FR61 Estimated RSE = 5.7 %

- NUTS1 - poultry ; FR2 Estimated RSE = 9.78 %.

Concerning the area of pasture and meadow in FR82, we knew before conducting FSS that the precision could be above the thresholds: farms in this region are very scattered. To improve lightly the RSE, we should have added too many questionnaires, for an expensive final cost.

Concerning poultry and other pigs, the deterioration of precision can be due to the fact that we do not well know these farms in our farm register (BALSA). The information necessary when we had to build the sample came mainly from the 2010 census; but there may have been changes for these productions since the census.

In order to avoid disturbing farmers for several agricultural surveys (see 10. Cost and burden), we have excluded 23 000 farms from our frame. Two kinds of units were excluded (negative coordination) :

- 18 000 units which were in the sample of our national survey of agricultural practices for fruit production (surveyed in 2015).  62 % of these units were big farms (Standard Output >= 100 000 euros).

- 5 000 units which were in the sample of our national survey of breeding practices (surveyed in 2015). 65 % ot these units were big farms (Standard Output >= 100 000 euros).

We consequently had an undercoverage in our FSS sample for big fruit farms and big farms with animals. We tried to made corrections after data collection with weight calibration. 

See 3.5 weight calculation.

Some units were interviewed although their administrative number was ceased, because they still had an agricultural activity with another administrative number (SIRET).

We should have stopped the survey but we pursued the survey although this new unit may be yet in our farm register (BALSA) with an identification number different from the selected one.

In order to limit the effect of this kind of error, we set weights to 1 for 2 680 units (when we saw that the respondant was yet in our database with another identification number).

We also may have some units that were below the thresholds but we do not know how many (counted with those who have ceased their activity).

They can be ceased or they can have obtained, since 2010, an administrative number and thus they can be twice in our farm register, with and without SIRET.

Furthermore, 1 % of the units of our farm register have a part of the same identification number (SIREN) : one of the SIRET can be ceased but not the other one.

--> We may not correctly notice ceased units in our register.

We had misclassification for some units : wrong size or wrong  farm type classification at the moment of the sample extraction compared with size or farm type calculated after data collection (~25 % of the units).

We maintained design weights.

See 6.3.1 /  2. Over-coverage errors

Proportion of out-of-scope units in the initial population identified with the sampling frame:

NOT AVAILABLE.

If any, very low (under threshold maybe), as we controlled the entered units in our sampling database (BALSA).

We asked respondent units if they are in production or not.    It not, we considered the units as ceased units.

1. Fresh vegetables, melons, strawberris - OPEN FIELD or MARKET GARDENING : B_1_7_1_1 / B_1_7_1_2

The source of measurement errors is the complexity to explain the difference between these characteristics in our instructions' document.

We had an unclear formulation of the questions  although we aimed to give clarification compared to FSS2010 or FSS2013. 

We detected errors when we analysed first weighted results compared to FSS2010 and FSS2013.

Actions taken: individual data corrections based on FSS2013 or FSS2010 individual data collected.

2. Permanent Pastures

The source of measurement errors is the complexity of the distinction between B_1_9_1 / B_3_1 / B_3_2 as the categories used for CAP area declaration (IACS) are quite different with a lot of changes in 2015. 

We detected errors when we analysed first weighted results compared to FSS2010 and FSS2013.

Actions taken: individual data corrections based on FSS2013 or FSS2010 individual data collected, for units with > 100 ha of B_3_1 in 2016 but with B_3_2 in 2010.

3. Soil cover  / Share of arable land with crop rotation

The sources of measurement errors are the complexity of the characteristic and the unclear formulation of the question.

Actions taken:

- checks after data collection and individual correction for soil cover (in order to be coherent with arable land area)

- % of arable land with crop rotation: we did not have the same codification on our national questionnaire as in the 715/2014 Regulation.

There are two categories of non-respondent units:

   - those which could not be contacted for different reasons: 3 156.

   - and those which were contacted but refused to respond to the survey: 1 185.

We applied re-weighting: new weights for respondents (in production or ceased) with the same NUTS2 and farm type.

To treat the issue of non-response, we first used a logistic regression to determine the characteristics of respondents. We used the Akaike Information Criterion (AIC) and a forward method to conclude that both localization and types of farming were the most significant variables to explain the response. We calculated the probability to answer in new strata built with those two variables. These new strata are what we generally call the homogeneous response groups. We then multiplied the weights of respondent units with the inverse of the probability to answer.

The Computer - Assisted Personal Interviewing (CAPI) interface checked that all the questions had been answered. There was no partial non-response.

6.5 % (4 341 / 67 000)

  -->  UNIT NON RESPONSE :  3 156 units who could not be contacted +  1 185 who refused to respond to the survey = 4 341

  -->  RESPONDING UNITS : 62 377 in production + 4 623 non producing = 67 000 units

4 341 non responding units, for which we do not know if they were eligible or not (ceased or in production, above or under our thresholds).

According to Eurostat guidelines, the unit non-response rate is calculated as 4341/(4341+62377) giving also 6.5%.

There were no partial non-responses.

No imputation in 2016. 

% of arable land with crop rotation: we did not have the same codification on our national questionnaire as in the 715/2014 Regulation.

We made calculations in order to have FSS 2016 categories.

- checks in our CAPI survey : direct corrections by surveyers OR permanent agents in regional statistics services;

- validations rules at the national level after data collection to verify consistency; some corrections;

- validation rules in a specific CAPI tool for regional statistical service to check some specific points as maximum values of some characteristics.

See 3. Statistical processing / 3.4 Data validation

No imputation in 2016.

We only publish FINAL estimates, no preliminary results are published for FSS. 

In 2016, in order to avoid delays, we have strengthened individual data validation (individual checks) after data collection in order to prepare:

- first aggregated results for regional statistical services,

- individual national structure survey data open access for our statistical service (national + regions),

- Eurostat file.

Errors were detected after the first transmission of aggregated results to regional statistical services (sample and weights issue). We prepared thus new version of our files.

With our first publication (called "PRIMEUR"), data is considered final. No more revisions.

- First file transmitted to Eurostat: July 2017: 7 months after the end of the year of the survey.

- First aggregated table only for agricultural statistical staff: October 2017: 9 months after the end of the year of the survey.

First national results: June 2018

Final national file: June 2018.

--> From 31 December 2016 to June 2018: 18 months

Final file transmitted to Eurostat: July 2018

This delay was mainly due to sample and weighting issues.

The national and the European definitions of the holding are the same. The definition of an agricultural holding used is France is the one described in Regulation (EC) No 1166/2008.  See item 2.5.

Handbook FSS WG/2015/1/11 revision 1 was used for the farm structure survey.

In FSS 2016, the "mushrooms" variable is measured in terms of production and not in terms of area. A correspondence key has therefore been used to meet the Community requirements. One hectare is considered to produce 270 tons (changes in 2016: for FSS 2013 and FSS 2010: 180 tons of mushrooms per year.)

The farm manager (chef d'exploitation) is defined as the person in charge of day-to-day management. We therefore look at the family of the farm manager and not at that of the holder (exploitant). This divergence does not pose any problems in the case of individual farms, i.e. the majority of farms. In FSS 2016, when considering extrapolated results, only 2 276 farms (0.5 %) were found with a sole holder who is not the manager.

% of arable land in crop rotation: we asked for  <=10 % /  10< - =<50 % / 50 <- <=75 % / > 75 % rather than 0 / 0<-<25 / 25<=-<50 / 50<=-<75 / >=75.

In FSS 2016, an AWU equals to 225 days. Considering that a day equals to 7 hours, an AWU equals to 1575 hours. We asked in 2016 for % AWU and then we constructed percentage bands.

The concept of common land exists in France. 

Methodology used for collecting information on common land units is the same in FSS 2016 as in FSS 2013 and FSS 2010.

Target population: the common land is maintained and managed by special holdings (i.e. common land holdings). They all may file the application for subsidy themselves (less favored area subsidy; called in France ICHN Indemnité Compensatrice de Handicap Naturel, which covers now what was previously called PHAE Prime Herbagère Agroenvironnementale (grass subsidy), which they then divide between the farmers on a pro-rata basis according to their use of the grassland. These units have been registered as agricultural holdings for which areas and labour data have been collected. 
The general questionnaire was used.

The land type of the common land is only permanent grassland and meadow (B_3); with 87 % of rough grazings (B_3_2).

Common lands covers 606 847 ha of utilised agricultural area (2.2 % of the total) in 2016.

In 2016, when considering extrapolated results, there were 1128 farms of common land (0.2 %).

As for FSS 2010, the territories of Saint-Martin, Saint-Barthelemy and Mayotte (in Mayotte, knowledge on farm holdings was significantly incomplete and of low quality, thus it was not possible to draw a stratified sample in this department) are excluded.

In 2016, a restricted territory of Guyane was surveyed compared to 2013 in order to focus to accessible holdings. In 2013, there yet were restrictions compared to FSS 2010 :  it was limited to 3 052 holdings, instead of 5 983 units surveyed by FSS 2010; on professional and more accessible holdings, on the coast side).

Impact of changes in Guyana scope :

2010 census results :

         Guyana, total = 5 983 farms

         Guyana, restricted scope as defined for FSS 2013  = 3 052 farms

         Guyana, restricted scope as defined for FSS 2016  = 2 785 farms

2010 census results for Guyana

The definition of the threshold used for farms has remained unchanged since 1955.

The only change introduced in 2016 is the removal of the threshold of fruit trees standing alone for production purposes in our manual for surveyors, but it didn't impact our results.

During FSS 2010, in France, the farmer (or holder) was called the responsable économique et financier (REF). For FSS 2013, we introduced the word of exploitant, which means "farmer", in order to reconcile our national terminology with European concepts. The REF benefits from the operating profits and suffers any losses, and may be a natural or legal person. This change of terminology allows to be more consistent with European definitions, and has not any impact on the results.

Farm labour:

In FSS 2010, the recording of farm labour, family or otherwise, was identical regardless of the farm status. In FSS2013, for the first time, we adopted Eurostat definition of the family: in the case of farms managed under legal person status (EARL, SCEA, SA, SARL or other person), co-farmers and their families who work on the farm are considered as non-family labour. This change of definitions may have had an impact of the family labour collected in 2013.

In FSS 2016, we asked for % of AWU, instead of AWU band; and we made calculation after data collection in order to have AWU bands. This may have a marginal impact on our result.

Dairy / non-dairy cows:

In FSS 2013, for the first time, in order to increase the use of administrative sources, bovine animals were not collected during the survey but were provided, according to the race of cows, by the national identification database (BDNI), thanks to farmers’ identification numbers (EDE). The eligible bovine animals are the ones detained by the farmer on 1 November 2013.

In FSS 2010, the part of dairy cows in each farm had been asked to the farmer, and this part had been applied to the number of cows known in the BDNI for this farm on 1 November 2010, to determine the number of dairy cows and other cows for every farm.

This method used for FSS 2013 is consistent with the one we are using in the framework of the animal Regulation No 1165/2008.

For FSS 2013, in certain mixed races, used as well for the dairy production as for the production of meat, the breakdown between dairy cows and other ones, based on races in the BDNI, can differ from that operated by the farmer during the survey of 2010.

For 2016, we came back to 2010 way of collection: the part of dairy cows in each farm had been asked to the farmer, and this part had been applied to the number of cows known in the BDNI for this farm on 1 November 2016, to determine the number of dairy cows and other cows for every farm. The number of other bovine animals directly comes from BDNI.

A part of the evolutions observed on the cattle (number of dairy cows and ranking within the farmtype - characteristic A06) is thus connected to the change operated in the mode of collection.

Fresh vegetables:

We tried for FSS 2016 to clarify the distinction between open field and market gardening but it introduced confusions for investigators.

Mushrooms:

One hectare is considered to produce 270 tons in 2016.  For FSS 2013 and FSS 2010: 180 tons of mushrooms per year.

We observe a decrease of common land in 2016: -11 % of common land units; - 10 % of common land areas,  if we compare to 2013.

437 416 

472 207

-7%

27 703 018

27 739 432

0%

18 190 695

18 466 196

-1%

9 631 087

9 623 200

0%

2 408 946

2 434 578

-1% 

4 508 674

4 887 400

-8%

452 389

494 235

-8%

8 565 856

8 242 242

4%

942 350

1 024 473

-8%

21 950 540

21 871 298

0%

18 935 003

18 905 862

0%

6 733 094

7 379 858

989 926

1 423 334

- 30 %

13 498 278

13 467 846

0%

304 776 416

297 077 263

3%

425 540

491 054

-13%

This decrease is due to the decrease of individual farms in which sole holder is the manager, for which we ask for family labour force.

269 602

296 676

-9%

397 066

416 025

-5%

328 693

343 805

-4%

In order to check data collected by FSS 2016, comparisons were made at the questionnaire level with administrative data:

• with areas declared by farmers for CAP payments (IACS) in 2015, for some crops for which data were available, in order to detect errors (either in crops areas or in administrative numbers).  Differences can be explained by changes between 2015 and 2016.
• with BDNI data for cows (total: C_2_6 + C_2_99)

These checks were conducted during data collection. It made it possible to detect some errors made by investigators while recording data in CAPI (unit errors, etc...).

We also introduced in the survey data from our orchard census for orchard specific questions (not for FSS).

After data validation at micro level, provisional results of FSS 2016 (especially main characteristics) were compared with several sources :

• data from FSS 2010;
• data from the annual agricultural statistics (AAS).These are drawn up by each statistical service of the regional directorates for agriculture and forestry.Their data relate to the statistics on land use and agricultural production in several fields, at department level: areas, yield, quantities harvested for plants, numbers of workers, weight or average quantities and totals for animal production. Each departmental piece of data is the result of comparing and balancing data from surveys, administrative sources (subsidies, taxation, etc.), and estimations from expers : questions addressed to the Chambers of Agriculture, trade associations, producer groups, etc.

Explanations for discrepancies between FSS and AAS :

- differences between definitions and concepts,

- sources used are not the same : AAS is the results of annual estimations from surveys results + administrative data + expert estimation,

• data from external sources, mainly the areas declared by farmers in the context of agricultural aid (IACS).

We then applied adjustments to FSS weights (See 3.5 Data compilation).

First results are available, at the national level, in a publication called "Primeur": 6 pages, text, graphs and tabulations.

We will only publish FSS results at national level.

However, some regions can publish their own results. 

National first results, Primeur, 6 pages: June 2018

Tabulations on our website: autumn 2018

Our PRIMEUR is available here: http://agreste.agriculture.gouv.fr/IMG/pdf/primeur350.pdf

Example of regional publication: Normandie results : http://agreste.agriculture.gouv.fr/IMG/pdf/R2818A09.pdf

All the results will be available on our website: http://www.agreste.agriculture.gouv.fr/ 

Basic tables will be available on the website of the Ministry of Agriculture:
https://stats.agriculture.gouv.fr/disar-web/disaron/!searchurl/searchUiid/search.disar

FSS 2016 - micro-data files can be accessed for research purposes.

Researchers or institutes can ask for this access. If the request is approved, the data may be accessed only via a safe access centre and the resulting tables are checked for the application of statistical confidentiality.

There is a fee to have an access.

Published information:

See on http://agreste.agriculture.gouv.fr/enquetes/structure-des-exploitations-964/enquete-structure-2016/

Methodological documents:

• Questionnaire:  see item 3.3;
• Instructions manual designed for regional staff and investigators, with definitions and explanations about the concepts used in the questionnaire; see the annex;
• A forum was also daily updated with questions from regional staff, and answers from national FSS team;
• Methodological letters transmitted to regional staff before and during the survey concerning the questionnaire, data collection, data checks, weight calculation.

No references on quality. 

Co-ordination between surveys:

In order to avoid the situation that some farms have to answer multiple agricultural questionnaires, we made a "negative co-oordination" beetween FSS 2016 and two other national surveys:

- Breeding practices, 2015 (questionnaires for farmers with animals);

- Orchards management - practices 2015.

See item 6.3.1 for more information.

Questionnaire:

A great expertise was conducted before the survey 2013 to simplify FSS questionnaire, in order to focus on collecting data requested by Eurostat.

The same Decree specifies that the answers collected come under the Statistical Confidentiality Act (No 51-711 of 7 June 1951) and the "Information Technology and Freedoms" Act (No 78-17 of 6 January 1978). In accordance with the Statistical Confidentiality Act (No 51-711 of 7 June 1951), the data collected are professional, and the interviewers and statisticians are bound by professional secrecy. 

The data circulates after encryption.
In the tables, data concerning fewer than three units and data where one unit represents more than 85% of the information are replaced by "s".