FARM STRUCTURE SURVEY 2007

FARM STRUCTURE SURVEY 2007 NATIONAL METHODOLOGICAL REPORT November 2008 Member State: FRANCE

Résumé En France, l enquête "structure 2007" s'est déroulée d octobre 2006 à novembre 2008 pour l ensemble des étapes de réalisation, de la préparation à la publication des résultats. Toutes les exploitations agricoles, au sens du recensement 2000, sont dans le camp de l'enquête. C'est le même écantillon qu'en 2005 qui a été enquêté en 2007, soit au total, 77 774 exploitations interrogées par enquêteur en France Métropolitaine et dans les départements d outre-mer, soit 13.3% des exploitations agricoles françaises. La livraison des données individuelles à EUROSTAT porte sur 72 908 dossiers, c'est à dire les exploitations ayant eu une production en 2007. La réalisation de l enquête "structure" a été confiée au Service de la Statistique et de la Prospective (SSP) qui est le département statistique central du Ministère en carge de l'agriculture (MAP). Le service central est en carge du tirage de l'écantillon, de la rédaction du questionnaire et des instructions, de la formation des services régionaux, du contrôle final de la qualité des données recueillies et de la première publication des résultats. Le SSP s'est appuyé sur ses écelons déconcentrés spécialisés, les services statistiques régionaux (NUTS 2) d'information statistique et économique (SRISE). Les agents permanents locaux n'étant pas assez nombreux pour supporter la carge de travail que représente l enquête, il a fallu recruter et former quelque 1 100 personnes non permanentes (enquêteurs). Les SRISE, outre la formation et l'encadrement de la collecte sont cargés de l'acquisition des données et du premier niveau de contrôle. L enquête s'est déroulée en 6 grandes étapes : la préparation, le recrutement de personnel temporaire, la formation, la collecte, la validation des données et leur diffusion. La collecte et les retours sur le terrain pour vérifications se sont déroulés entre le mois d octobre 2007 et le mois d avril 2008. La pase de contrôlevalidation a été réalisée au niveau régional et national jusqu'au mois de juillet 2008. La diffusion des premiers résultats de l enquête a débuté au mois d'octobre 2008 par la publication d un quatre pages Agreste Primeur numéro 215. En parallèle, les ciffres sont stockés dans la base de données DATAGRESTE du SSP et un certain nombre de résultats sous forme de tableaux sont mis à disposition du public sur le site Internet du Ministère en carge de l'agriculture ttp://www.agreste.agriculture.gouv.fr/ A partir des résultats définitifs, le département statistique central a pris en carge la réalisation des produits standard de ciffres bruts pour l ensemble des écelons géograpiques (Ciffres et données Série Agriculture avec CD- ROM à publier en décembre 2008). Les services régionaux publient également sous des formes diverses les résultats de l enquête dans des documents Agreste (tableaux, caiers, mémento ). Octobre 2008 Page 3 sur 23

L ensemble du travail de diffusion sera acevé en fin d année avec la livraison des données individuelles à EUROSTAT en octobre ou novembre 2008. Octobre 2008 Page 4 sur 23

Summary (as drafted by te Frenc autors) In France, te farm structure survey 2007 (FSS2007) as been carried out from October, 2006 till November, 2008 including all te steps from preparation to results dissemination. All te farms, witin te census 2000 definition, are in te field of te survey. It is te same sample, as in 2005, wic was surveyed during 2007 survey. A total of 77 774 farms as been questioned by investigator in Metropolitan France and in te overseas departments (DOM), i.e. 13.3% of Frenc farms. Te delivery of te individual data at EUROSTAT relates to 72 908 files, only farms aving ad a production into 2007. FSS2007 realization was entrusted to te "Service de la Statistique et de la Prospective" (SSP) wic is te central statistical department of te Ministry of agriculture (MAP). Te central service is in carge of sample drawing, drafting of te questionnaire and te surveyors' instructions, training of te regional services, final quality control of te data collected and te first publication of te results. Te SSP was assisted by its specialized decentralized levels, te regional statistical services (NUTS 2) of statistical and economic data (SRISE). Te local permanent agents not being enoug numerous to support te workload wic te investigation represents, it was necessary to recruit and train some 1 100 non permanent persons (investigators). Te SRISE, in addition to te training and management of te data collection are in carge of te data acquisition and te first level of control. Te investigation proceeded in 6 great steps: preparation, recruitment of te temporary staff, training, data collection, data validation and data dissemination. Main data collection and te complementary surveys in order to proceed to field cecks were carried out between October 2007 and April 2008. Te step of control-validation was carried out at te regional and national level until July 2008. Te diffusion of te first survey results began in October 2008 by te publication from one four pages "Agreste Primeur" number 215. In parallel, te figures are stored in te data base "DATAGRESTE" of te SSP and main results are stored as tables on SSP website (free of carge) ttp://www.agreste.agriculture.gouv.fr/ From te final results, te central statistical department dealt wit te realization of te standard products of roug figures for te wole of te geograpical levels ("Ciffres et Données Série Agriculture" wit CD-Rom to be publised in December 2008). Te regional services also publis in various forms te survey results in "Agreste" documents (tables, books, memorandum...). Te wole dissemination will be completed at te end of te year wit te delivery of te individual data at EUROSTAT in October or November 2008. Octobre 2008 Page 5 sur 23

Octobre 2008 Page 6 sur 23

SOMMAIRE 1 Introduction 8 1.1. Background, context 8 1.2. National law 8 2 General caracteristics 9 2.1. National caracteristics and reference period 10 2.2. Questionnaires 10 3 Survey metodology 11 3.1. Organisation of te survey 11 3.1.1 National organisation 11 3.1.2 Regional organisation 12 3.2. Survey timetable 12 3.3. Preparation of te survey 13 3.3.1 Context and population 13 3.3.1.1 Definition of an agricultural olding 13 3.3.1.2 Divergences from te European Union's variables 13 3.3.2 Sampling plan 14 3.3.2.1 Scope of te survey 14 3.3.2.2 Stratification 14 3.3.2.3 Determination of an exaustive stratum 15 3.3.2.4 Determination of te number of questionnaires and te variables of interest 16 3.3.2.5 Expected accuracy 17 3.3.3 Testing of te questionnaire 17 3.3.4 Information and communication wit te respondents 17 3.4. Sample, data collection 17 3.4.1 Creating te sample 17 3.4.2 Organising collection 17 3.4.3 Use of administrative data 18 3.4.4 Cecking data 18 3.4.4.1 Field cecks 18 3.4.4.2 Capture/control in te region (NUTS 2) 18 3.4.4.3 Cecks on aggregated data 19 3.4.4.4 Transmission of information to te central level 19 3.4.5 Total non response 20 3.5. Processing of te data, analysis and estimation 20 3.5.1 Metods for correcting and extrapolating missing data 20 3.5.2 Estimation 20 3.5.3 Oter types of errors 21 3.5.4 Evaluation of te results 22 4 Publication and dissemination 22 5 Annexes 23 Octobre 2008 Page 7 sur 23

1 Introduction Te 2007 Structure Survey is te tird suc intercensal survey, following tose of 2003 and 2005. 1.1. Background, context Exaustive knowledge of te structure of agricultural oldings first became available after te Agricultural Census carried out in 2000. Te goal of te Structure Survey is to update te data on te structure of agricultural oldings - size, legal status, type of farming, population and production, and teir development in between censuses, at bot national and regional level. Cooperation was significant for te 2000 Agricultural Census, wic was te first of te series. Te directorates of te Ministry of Agriculture, Food, Fiseries and Rural Affairs (MAAPAR) and te survey's users monitor te work and are consulted before te survey takes place, via a users committee. 1.2. National law Te 2007 Structure Survey was entrusted to te "Service de la statistique et de la prospective" 1 (SSP), wic is te central statistical department of te Agriculture Ministry, wit te support of its regional statistical services (NUTS 2). Te 2007 Structure Survey meets a Community requirement: Council Regulation (EEC) No 571/88; Commission Decision No 98/377/EC of 18 May 1998 amending Annex I to Council Regulation (EEC) No 571/88; Commission Regulation (EC) No 143/2002 of 24 January 2002 amending Annex I to Council Regulation (EEC) No 571/88 in view of te organisation of te Community surveys on te structure of agricultural oldings in 2003, 2005 and 2007 Regulation (EEC) No 1444/2002 on te definitions of te variables; Regulation (EEC) No 204/2006 on te non-existent and/or nonsignificant caracteristics. Te survey is included in te Frenc public services' programme of statistical surveys for 2007. Te opinions of te National Committee on Computerisation and Freedoms (CNIL) and of te National Statistical Information Council (CNIS) are essential to tis operation. Te CNIL's favourable opinion relating to te automated processing of te 2007 Farm Structure Survey was delivered in an Order of 9 August 2007, publised in te Official Journal of te Frenc Republic of 29 August 2007. 1 Te "Service de la statistique et de la prospective" (SSP) was created on 1 July 2008, as part of te general overaul of public policies. It as taken over all te activities of te "Service central des enquêtes et études statistiques" (SCEES) and as suc remains te central statistical department of te Agriculture Ministry and a member of te Frenc public statistics system. Octobre 2008 Page 8 sur 23

Te CNIS is an advisory body in wic te different social partners are represented. It gives opinions on te advisability of statistical operations and certifies tat te conditions of implementation guarantee tat te rules are being properly complied wit. Te committee's compliance notice in respect of te label no. 216 / D131 of 3 February 2005, notified on 7 February 2005, covered bot te 2005 survey and its renewal in 2007. As te 2007 Structure Survey is acknowledged to be of general interest, it is compulsory. Te Minister for Agriculture and Fiseries and te Minister for Economic Affairs and Industry ave issued an approval number (2007 A 107 AG) for tis operation. It is valid for 2007. In accordance wit te Act on Statistical Confidentiality (51-711 of 7 June 1951), te data collected are confidential. Interviewers and statisticians are bound by professional confidentiality. Te law requires te farmers to reply accurately to te questions. Te data collected are confidential, but farmers ave te rigt to access and correct teir declarations (Information Tecnology and Freedom Act (78/17 of 6 January 1978)) Proceedings are initiated against farmers wo fail to respond or provide deliberately misleading information, te final penalty being an administrative fine. 2 General caracteristics Surveys on te structure of agricultural oldings are intended to meet tree objectives: to learn ow oldings are structured and monitor developments; to monitor te development of agricultural production; to keep track of te agricultural population. Te main types of information collected concern: Knowledge of te structure of oldings and measurement of its development: te legal form of oldings; type of tenure; area of oldings; economic size of oldings measured by standard gross margin (estimated from data on te area and livestock); type of farming (estimated from data on te area and livestock). Knowledge of agricultural production and measurement of its development measuring oldings' individual fields; measuring te number of livestock, poultry production. Keeping track of te agricultural population: caracteristics of te family working on te oldings (training, working time, etc.); measuring te amount of work needed to run te olding: family labour, permanent, seasonal or temporary employees. Tese data, wen compared to tose collected for previous agricultural censuses or structure surveys make it possible to, for example: measure te concentration of agricultural activity; follow te development of te legal form of oldings (GAECs (agricultural groups for joint farming), companies, etc.); follow te evolution of speculation (crops and livestock); evaluate structural policy, for example: monitor te future of agricultural oldings: Octobre 2008 Page 9 sur 23

estimate te number of oldings wic disappear, are vacant or are taken over, and te appearance of new oldings originating from existing oldings. 2.1. National caracteristics and reference period Te survey covered 74 220 oldings in continental France, and 3 554 in te overseas départements, a total of 77 774. Tis is a representative sample of te oldings interviewed for te 2000 Agricultural Census. Te oldings were drawn at random and may be quite small, provided tat tey produce agricultural goods, reac a certain tresold (in terms of area, number of livestock, volume of production, etc.), and are independently run. Tey are oldings wic were already questioned in 2005. Te reference period for te survey is te agricultural year from 01/11/2006 to 31/10/2007 for all questions, except for numbers of livestock present, wic are recorded on te day of te interviewer's visit. 2.2. Questionnaires Te Structure Survey questionnaire makes it possible to determine te agricultural population working on te farm (te farmer, is family, paid family and non-family labour) and te means of production (breakdown of areas, plant and livestock products, buildings and equipment, etc.). To a lesser extent, te Structure Survey also makes it possible to bring togeter associated activities, namely te processing of products on te farm, reception and tourism in a rural environment, contract services, etc. Te questionnaire does not contain any questions on te values of financial results or income since tese subjects are reserved for te Agricultural Accounts and te Agricultural Accounting Information Network (Rica) operations. Te questionnaire complies wit all European obligations. However, additional national questions ave been added: inventory of sprays used (page 6), signs of quality for products (page 10), proportion of turnover accounted for by diversification activities (page 10), management for te system of taxation, VAT, sickness insurance, aid facilitating establisment and modernisation, long-term agricultural contracts, Internet use, te year te type of farming was cosen, succession (page 11) and management of occupational waste (page 12). Tere are 2 versions of te survey structure questionnaire. One version is intended for continental France, te oter for te overseas départements, i.e. Guadeloupe and Martinique, Frenc Guyana and, Reunion (cf. versions of te questionnaire in Annex 1). Te questionnaires also ave 20 questions applying to regions on pages 5, 6 and 7. Tese are left to te discretion of local units, wic may coose to use tem or not, depending on teir needs. Te questionnaire's lengt as been restricted to 12 paper pages maximum to avoid over-long interviews. Te 2007 Farm Structure Survey is structured as follows : Te first tree pages contain a set of questions (common to all te surveys) designed to elicit information on any structural canges in te olding in question since 2005. Tis information makes it possible to decide weter te olding sould be kept in te sample and wat extrapolation coefficient to assign to it; Octobre 2008 Page 10 sur 23

te following pages are allocated to te questions wic are relevant to te survey in question, broken down by subject: main crops (page 4); oter information on te land (page 5); irrigation and sprays (page 6); livestock; rearing (page 7); population and family labour (page 8); non-family labour (page 9); valorisation and diversification activities (page 10); management (page 11); management of occupational waste (page 12). Page 12 also contains te list of PACAGE identification numbers used by te olding in its dealings wit te bodies responsible for collecting teir landarea declarations and paying te CAP subsidies. Te aim is to be able to link te structural information collected in te survey and te administrative information (area declarations, amounts of premiums, etc.). Only te first and last pages of te questionnaires are pre-printed, on te basis of te information existing in te SSP's Updated Sampling Frame, wic results from a partial updating of te units on te basis of postcensus surveys. Te following are pre-printed: name and address of te eadquarters of te olding to be surveyed; name and address of te ead of te olding to be surveyed; name and address of te farmer (REF - economic and financial manager) of te olding to be surveyed; te name, legal form and SIRET number of te various structures connected wit te olding to be surveyed (pre-printed if available); te list of known PACAGE numbers. 3 Survey metodology Te metodology employed for te 2007 survey was very similar to tat employed in 2005. Te cief innovation was an experiment in partially precompleting te questionnaires on te basis of te area declarations (subsidy application dossiers). 3.1. Organisation of te survey Te survey is designed and run by te SPP, wit data collection and firstlevel validation being entrusted to te regional delegations. 3.1.1 National organisation Witin te SSP, national responsibility for te survey lies wit te Subdirectorate for Agricultural, Forestry and Agri-food Statistics (SDSAFA), and more specifically wit its Office of Statistics on Agricultural Holdings and te Environment (BSSAE), te national unit responsible for te operation. Te preparation and running of te survey in te overseas départements is conducted by te unit coordinating te agricultural statistics for te overseas départements, in conjunction wit te BSSAE. In addition to preparing te survey, and auditing and cecking te survey at national level, tese two bodies store te surveys and transmit te individual data to Eurostat. Oter units or persons from te Ministry of Agriculture are also asked to elp. Tus, te Administrative Bureau of te Statistics Department is responsible for budgetary aspects, suc as te delegations of appropriations Octobre 2008 Page 11 sur 23

to te regional services, and regulatory aspects (employment contracts, for example). Te publications office is in carge, in particular, of page makeup and printing te documents, duplication, sending out te materials, relations wit te printers, etc. IT personnel from te Ministry of Agriculture's computer centre are eavily involved as managers of te network for te transmission of data linking te two geograpical levels. Tese units work in constant collaboration wit te statistics and IT teams. Moreover, a users committee specific to te 2007 Structure Survey as been set up. It includes te main Directorates of te Ministry of Agriculture, Food, Fiseries and Rural Affairs, te National Institute for Agronomic Researc (INRA) and te Permanent Assembly of te Cambers of Agriculture (APCA), and it gave its opinion on te definitive questionnaire before preparation of te survey began. In order to carry out tis operation, a tecnical group comprising national and regional statisticians was set up for te purpose of: finalising te questionnaires and instructions after consulting users; conducting test interviews; creating teacing tools for te regional services in order to train interviewers; creating tools for cecking responses bot at regional and national level. 3.1.2 Regional organisation Te regional statistical and economic information services (SRISE) are responsible for recruiting and training interviewers. Tese services plan and monitor te work of te interviewers, ensure collection and first-level validation of te data and publis regional results once te survey is declared validated. 3.2. Survey timetable Te survey took place in six major stages: preparation, recruitment of temporary staff for te collection pase, training, data collection, data validation and data dissemination. Operation Date Preparation of te survey October 2006 to October 2007 Recruitment - Training September - October 2007 Data collection Feedback October 2007 - February 2008 Cecking and validation of data February 2008-July 2008 Definitive arciving in te DATAGRESTE database Dissemination: First publication, 4 pages, Agreste Primeur Dissemination: Delivery of file containing individual data to Eurostat Metodological report August 2008 October 2008 October - November 2008 Preparation began in October 2006 wit te creation of te various working groups. It was completed in February 2008 wit te delivery of te tools for cecking responses. Octobre 2008 Page 12 sur 23

Interviewers were trained wit teacing tools made available to te wole of te network. Depending on teir prior experience, it took from 1 to 3 days to train te interviewers for tis survey. Collection in te field took place from October 2007 to February 2008. Data capture took place as collection proceeded witin te regional services. Te process is one of capture/control caracterised by te activation of a set of controls at te time of capture. Tis stage was accompanied at regional level by a final validation of te data using a set of data control procedures, from February 2008 to July 2008. After te data were validated at national level, tey were arcived in te DATAGRESTE database in August 2008. Te first publications were completed by October 2008. Dissemination of te data from te Structure Survey will extend, essentially, until te end of 2008. 3.3. Preparation of te survey Preparation of te survey was andled exclusively by te statistical service (SPP) of te Agriculture Ministry. 3.3.1 Context and population Te population comprises all Frenc agricultural oldings as recorded in te 2000 Agricultural Census and subsequently updated by surveys and/or oneoff operations for te preparation of temed surveys. 3.3.1.1 Definition of an agricultural olding Te statistical definition of an agricultural olding as not canged since 1955. It now co-exists wit oter administrative or statistical definitions, namely te legal definition of farming (farming by nature or by extension), social approac (sickness insurance and pension scemes), aid files and te definition of te general statistical list. An agricultural olding is defined by Decree 2000-60 and by te Order of 24 January 2000 prescribing te census. In terms of agricultural statistics, it is an economic and production unit wic simultaneously meets tree preconditions. It produces agricultural products. It is at least of a certain size in terms of area, livestock numbers, production, etc.. Finally, it is subject to independent everyday management. Te definition of an agricultural olding used for te Structure Survey is te same as te one used for te 2000 Agricultural Census. 3.3.1.2 Divergences from te European Union's variables Certain caracteristics diverge from te European Union's standard variables. Some relate to aspects for wic France as requested and obtained an exception. In te case of oter points, wic are limited in number, te Community requirement is complied wit by reprocessing te information collected. France as been allowed two exceptions. Cristmas trees are not considered to be agricultural products. Tey are classed wit te olding's woodlands and forests. Breeding cocks are not sown wit layer en numbers but wit broilers. Octobre 2008 Page 13 sur 23

As regards te oter points, France's treatment does diverge in places, ranging from differences of vocabulary to te problem of content. In France, te farmer is called te economic and financial manager (REF), te party wo enjoys te operating profits and suffers te operating losses. Tis may be eiter a natural or a legal person. Tis farmer, or economic and financial manager, is not te reference person. Te reference person is te farm manager, defined as te person in carge of te day-to-day management. Tus we are interested in te family of te farm manager and not tat of te farmer. Tis divergence does not pose any problem in te case of individual oldings, wic represent 73 % of all oldings. In te case of te remaining 17 %, complex treatments are necessary to retranscribe tese data in accordance wit Community rules wen te file is relayed to Eurostat. Te same applies were te olding is farmed on bealf of anoter natural person. Te recording of te olding's labour force, weter family or not, is identical watever te status of te olding. As far as Eurostat is concerned, in te case of oldings wit company status (EARL, SCEA, SA, SARL or oter legal personality, or statuses 3, 5, 6 and 7), te co-farmers and teir families appear as non-family labour. Te musrooms variable is measured in terms of production and not in terms of area. A correspondence key as terefore been used to meet te Community constraint. France as no areas wic are subject to set-aside incentive scemes and wic ave been converted to permanent pasture, meadow, wooded areas, etc. 3.3.2 Sampling plan Let us reiterate tat te Structure Survey works by following a permanent sample of oldings taken from te latest Agricultural Census. To cover te time between te last census in 2000 and te next one in 2010, tis sample will be interviewed in 2003, 2005 and 2007. Te permanence of te sample is guaranteed by specific rules governing wic oldings are to be interviewed if an event occurs wic affects a unit of te sample (absorption, partial break-up, etc.). At te request of certain local units, te sample was reinforced for te 2005 Structure Survey. In 2007 te sample is strictly identical to te 2005 sample, apart from tose oldings no longer operating. 3.3.2.1 Scope of te survey All te oldings recorded in te 2000 Agricultural Census are part of te universe from wic te sample is drawn. An effort was made wit te local units to identify te oldings to be avoided during te interview pase, i.e. particular oldings wic work in extremely different ways from oter oldings (referred to as "atypical" oldings). 3.3.2.2 Stratification Stratification is basically determined from two variables: OTEX 1896 (an 18- digit number describing te type of farming calculated from MBS 1996 coefficients) and te CDEX 1096 variable (a 10-digit number indicating te economic size class). Octobre 2008 Page 14 sur 23

Before te sample is drawn from eac stratum, te oldings are sorted using OTEX 1896 as a major criterion, and ten by MBS as a minor criterion. 3.3.2.3 Determination of an exaustive stratum In order to improve te precision of te estimators, an exaustiveness tresold was defined for eac département (NUTS 3) to ensure tat all units wit a very eavy weigting witin a given département are interviewed. As a general rule, tis tresold was set at 250 ESU. However, for certain départements and certain OTEX values, tis tresold was raised because of te number of oldings concerned. Octobre 2008 Page 15 sur 23

3.3.2.4 Determination of te number of questionnaires and te variables of interest Te number of questionnaires and te variables of interest were determined by multi-stage iteration taking te following constraints into account: te number of questionnaires ad to be at least 300 but no more tan 1 200 per geograpic unit (NUTS 3); te total number of questionnaires for all of te metropolitan départements was to be approximately 84 000 (in Annex 3, page 3, tis limit is set at 75 000, wic corresponded to te domestic budget ceiling but wic, ultimately, was insufficient to meet te accuracy requirements for te tree intercensal surveys; te budgetary ceiling was terefore revised upwards); accuracy ad to be good for a certain number of variables representative of eac département's production, and bearing in mind te anticipated decline in te number of oldings between 2000 and 2007. Te number of questionnaires per stratum for eac département was determined using a Neyman allocation. Te number of questionnaires was ten increased in te strata were te extrapolation coefficient was too ig. Of te 84 294 oldings selected for te first survey, 77 774 remain today, i.e. a drop of 6 520, altoug te total UAA (utilised agricultural area) and livestock numbers ave not fallen accordingly. Tis reflects te trend towards concentration of means of production. Most of tese canges are noted by te interviewer in te course of is survey, at individual olding level, and are recorded in te first tree pages of te questionnaire. Strict rules are ten applied to determine te oldings remaining in, joining or leaving te sample and to determine te extrapolation coefficient. In addition, interviewers are informed of olding movements ascertained from oter sources (updating of te sample survey basis by oter surveys or administrative sources): oldings wic existed wen te sample was selected and wic no longer exist (as discovered by te interviewer) represent teir weigting on cessation of teir activity oldings wic ave canged farmer (as discovered by te interviewer) but wic are oterwise uncanged are questioned wit te same extrapolation coefficient oldings wic ave created anoter olding by contributing land and/or livestock: bot oldings are questioned wit te original extrapolation coefficient oldings wic ave merged: it is te result of te merger tat is questioned. Te olding wic contributed te most in terms of means of production is identified, and tat olding's extrapolation coefficient is applied to te wole new oldings, weter wit livestock or wit cleared land, are identified tanks to administrative data (mandatory declarations), but tey are few in number. For most départements, te variable of interest for allocating questionnaires is te standard gross margin (SGM). However, for a number of départements (91, 96 2, 24, 33, 09, 12, 38, 73, 74, 11, 30, 34, 66, 04, 05 and 84), te variable of interest is te utilised agricultural area (UAA). 2 Département 96 does not exist as suc. It is an amalgamation of tree départements (92, 93 and 94). Octobre 2008 Page 16 sur 23

In response to local requests te sample was reinforced in 2005, being increased to 84 294 oldings out of a total of 698 535 recorded in te 2000 Agricultural Census, or 12.1%. Tese oldings existing at te time of te census ave given birt to new oldings, wic explains ow te total number of oldings in te 2005 Structure Survey came to be 86 126. In 2007, 77 091 oldings active in 2005 were interviewed, togeter wit 683 oters created by te former, tus giving a total of 77 774 oldings interviewed. 3.3.2.5 Expected accuracy Te table below contains te variation coefficients for certain variables at te level of metropolitan France. Metropolitan France: variation coefficients for certain variables (%) (%) Utilised Agricultural Area 0.26 Total bovines 0.49 Cereals 0.34 Dairy cows 0.62 Industrial crops 0.52 Nurse cows 0.77 Permanent grassland 0.52 Total pigs 1.75 Fodder 0.55 Total seep 1.65 Vineyards 0.68 Goats 4.07 Perennial crops 1,71 Total poultry 2,62 3.3.3 Testing of te questionnaire As te questionnaire for te 2007 Structure Survey is very similar to tat used in te last census, no additional testing was carried out in te field on te traditional parts. Only te new questions were tested in te field. 3.3.4 Information and communication wit te respondents As it depends on te local context, te coice of te means of communication wit respondents was left to te local units. Te tecnical working group, working at national level, prepared a form letter for announcing te survey (cf Annex 3 to tis report), a leaflet describing te survey, wic was made available to te local units, and a folder on statistical confidentiality and te obligation to respond. 3.4. Sample, data collection 3.4.1 Creating te sample In eac stratum, oldings are sorted by type of farming and economic size. Holdings are selected randomly until te number of questionnaires expected for te stratum is reaced. 3.4.2 Organising collection Te 2007 Structure Survey consisted exclusively of face-to-face interviews. To tis end, in te field te survey is supported by te Ministry of Agriculture's specialist units, namely te 26 regional (NUTS 2) Statistical and Economic Information Services. Tey conduct te survey in te field via te network of interviewers wic tey use for te entire programme of agricultural statistical surveys. Octobre 2008 Page 17 sur 23

As te local permanent officials were too few to cope wit te extra workload generated by te Structure Survey, additional non-permanent staff placed under te autority of te ead of te statistical service ad to be recruited. Tus, over 1 100 interviewers were recruited and given te task of collecting information from te farmers during a direct interview and transcribing te information onto questionnaires. Once completed, te questionnaires are sent to te regional statistical department. Te interview lasts from 20 minutes to 1 1/2 ours. 3.4.3 Use of administrative data Te SSP used te data provided in te area declarations made by te farmers wen applying for CAP subsidies. Tese declared areas were communicated to te interviewers via an information seet for eac farm included in te sample. Te interviewer was required to get tese data validated by te farmer. Ultimately, as in all statistical surveys, it is te reply given in te interview tat is used. Tis experience revealed serious difficulties in te use of te administrative data: Less tan 50% of te data was available at te time of launcing of te survey in September 2007. Te crop nomenclatures are not identical and tere are difficulties in switcing from one to anoter, Te areas recorded in te CAP declarations are net, wereas in te Structure Survey tey are gross. 3.4.4 Cecking data Data cecking, te last step before dissemination, began during te collection stage on te farmers' premises. It ten continued trougout te processing cain. 3.4.4.1 Field cecks Field cecks are performed at two levels: by te interviewer and by te ead of te regional statistical service responsible for te survey. Te interviewer's role is to verify te totals and consistency directly at te respondent's premises. After te interviews, te interviewer must ceck te additions and carryovers and make sure tat te questionnaires are complete, consistent and legible. Once te questionnaires ave been cecked, te interviewer sends tem to te regional service. 3.4.4.2 Capture/control in te region (NUTS 2) An initial, visual ceck is carried out by te staff of te regional units, particularly at te beginning of te collection process. Te questionnaires are ten keyed in using capture/control software developed centrally by te SSP and made available to te regional Statistical and Economic Information Services (NUTS 2). Tis computer application allows responses to be input wile, at te same time, activating a series of cecks. Octobre 2008 Page 18 sur 23

Tere are tree types of cecks: minor, major or blocking. A minor ceck is a warning wic does not necessarily require any correction. Example: an area wic is too small for te numbers of bovine animals. A major ceck requires correction, but not necessarily immediately. Te questionnaire can continue to be keyed in, but te dossier contains a specific indicator wic excludes it from furter processing operations. Example: an erroneous total. Finally, a blocking test requires immediate correction. One instance of tis is wen a capture code does not belong to te list of accepted codes. Wen te capture/control programme is written, a proper balance needs to be found between te efficiency of te cecks and teir sensitivity, wic, if excessive, would increase capture times. Tis task is all te more tricky since tere is no indicator wic measures, for eac minor ceck, te number of corrections made compared to te number of times wen te keyboard operator as carried on regardless. Te best possible design is sougt wen te software is tested. For example, for te 2007 Structure Survey, 666 cecks were defined in te capture-control tool for te national part of te questionnaire. Of tese, 366 were minor, 265 were major and 35 were blocking cecks. 3.4.4.3 Cecks on aggregated data Cecks on aggregated data are a means of improving te quality of results. Tey supplement te system set up for collecting te data, namely training, field cecking and capture/control. Tey spot aberrant values, ensure tat te interviewer applies rules systematically, detect errors missed by te capture/control programme and flag wen instructions ave not been complied wit. A tool for cecking survey data, usable bot nationally and regionally, was developed in te same language as te capture-control programme. Wit it, inventories, ratios and pre-defined or user-programmed lists can be obtained. In tis way, eac region was able to define te most suitable cecks for te context of local agriculture. All tese cecks were used at NUTS 3 level. Main functions used: verification of te "forced" cecks during input in order to spot potential systematic errors; list of te X largest values for eac quantitative variable, enabling aberrant values to be spotted; comparison, for eac farm, of te data common to te 2005 and 2007 surveys, and identification of suspect trends; comparison between 2005 and 2007 of te aggregated results at "département" level, and identification of suspect trends. 3.4.4.4 Transmission of information to te central level Subsequent to completion of te cecks conducted at NUTS 3 level by te regional services, temselves operating at NUTS 2 level, all of te questionnaires are sent to te central level for analysis, via te Ministry of Agriculture's secure transmission network. IT personnel from te Ministry of Agriculture's computer centre are tus also eavily involved in tis procedure as managers of te network for te transmission of data linking te two geograpical levels. Octobre 2008 Page 19 sur 23

If anomalies or inconsistencies are still detected at central level during te final pase of cecking, te regional services are asked to confirm or, if necessary, modify te information wic is considered suspect. Once corrected, te individual questionnaires are ten retransmitted to te central level. 3.4.5 Total non response Out of te 77 754 oldings surveyed in te original sample (74 220 in metropolitan France, and 3 554 in te overseas départements), plus 20 newly created oldings, some refusals were met wit during collection. Generally, wen an interviewer meets wit a refusal te ead of te service or a permanent member of staff intervenes. In general, a telepone discussion is enoug to cange te farmer's mind about is refusal and te questionnaire is completed normally. To ensure tat samples can be followed up, replacements of oldings are proibited. Tis is because maintaining a permanent sample from one survey to te next is crucial for an accurate measurement of ow structures are evolving. Ultimately, te file does not include any total non-responses. All te missing variables were eiter obtained after an additional survey conducted by te permanent staff of te services or were obtained from te administrative declarations. Tese latter cases represent less tan 0.1% of te oldings questioned. 3.5. Processing of te data, analysis and estimation 3.5.1 Metods for correcting and extrapolating missing data If information is missing, te interviewer or local statistician may contact te respondent by telepone. If tis fails to obtain te information, te local statistician uses te administrative sources available witin te competent Departmental Directorate for Agriculture and Forestry (CAP dossiers, etc.). 3.5.2 Estimation Once te strata ave been defined and te number of oldings per stratum to be surveyed ave been determined, te following information per stratum is available: N te number of oldings from te universe wic belong to stratum n te number of oldings to be surveyed in stratum. Te weigting of eac olding to be surveyed in stratum is equal to wic is te olding's extrapolation coefficient. N, n It sould be kept in mind tat subsidiary oldings may be discovered (and terefore surveyed) during te course of te survey. As a result, te number of oldings wic are actually surveyed ( n ) will be equal to or greater tan te number of oldings wic were to be surveyed ( n ). Octobre 2008 Page 20 sur 23

For stratum ( varies from 1 to H in so far as we ave defined H strata), if Y is te value of variable Y for olding i (i being a olding on te list to be i sampled) for stratum, estimators of te average and total for variable Y in stratum are constructed as follows: te estimator for te average of variable Y in stratum is Yˆ = n i=1 Y n i te estimator for te total of variable Y is Tˆ = N Yˆ For te strata taken togeter, estimators of te average and total for variable Y in stratum are constructed as follows: H N te estimator for te average of variable Y is Yˆ = Yˆ = 1 N te estimator for te total of variable Y is Tˆ = H T ˆ = 1 For te strata taken togeter, te estimated variance of te estimators of te average and total for variable Y is defined as follows: H 2 2 N = n s Vˆ ( Yˆ) = 1 1 N N n wit s 2 = n i=1 ( Y Yˆ ) i n 2 Te estimated variance of te estimator for te total of variable Y is derived from te formula: V ˆ ( Tˆ) = N 2 Vˆ( Yˆ) Te estimated relative precision of te estimator for te average of variable Y (expressed as a percentage) is: ˆ( ˆ) ( ˆ) V Y pr Y = 100 Yˆ Te estimated relative precision of te estimator for te total of variable Y is equal to te estimated relative precision of te estimator for te average of variable Y. 3.5.3 Oter types of errors Errors are detected tanks to te various cecking tools wic are provided at regional level. Te regional units are responsible for ensuring tat errors in te questionnaires are corrected. Once te data are submitted to te central level, tey are cecked again. If furter errors are detected at Octobre 2008 Page 21 sur 23

central level, te regional unit is asked to carry out te additional corrections. 3.5.4 Evaluation of te results At central level, an initial assessment is carried out tanks to te expertise of te Office for Plant and Forestry Statistics and te Office for Animal Statistics. Tis assessment is carried out by comparing: certain data from te annual agricultural statistics. Tese are collated by te statistical services of eac Departmental Directorate for Agriculture and Forestry. Tis is for data wic concern statistics regarding land use and agricultural production in te départements: areas, yields, quantities arvested for plants, numbers of workers, average and total weigts or quantities for animal production. Eac département piece of data is te result of comparing and balancing data from surveys, administrative sources (subsidies, taxation, etc.), questions addressed to Cambers of Agriculture, trade associations, groups of producers, etc.; te results of Community livestock surveys or specialised surveys, suc as tose for orcards, vegetables, etc. If variations of more tan 8% are detected for certain common variables suc as tose regarding land area and livestock numbers, a more in-dept analysis is carried out, including referring te matter back to regional level if necessary. 4 Publication and dissemination Te first step was to make te data available to te wole agricultural statistics network by loading tem into DATAGRESTE, SSP's database. Tey can ten be accessed according to structural, geograpical, agronomic and oter criteria. A series of pre-programmed tables was prepared centrally. Tis provides te agricultural statistics network wit some uniform output at national, regional, or even département level. te standard dissemination of te 2007 Structure Survey consists of: a 4-page "Agreste Primeur" (early results), No. 215, at national level, in October 2008; a 4-page "Agreste régionaux" (regional results); a national edition of "Ciffres et Données" (planned for December 2008), presenting te main results in table form; posting of national, regional and département tables on te website of te Ministry of Agriculture and Fiseries, wit access free of carge: ttp://www.agreste.agriculture.gouv.fr/ Octobre 2008 Page 22 sur 23

5 Annexes Questionnaire métropole Questionnaire used in metropolitan France Questionnaire DOM Questionnaire used in te four overseas départements (DOM) Secret statistique Leaflet given to farmers by interviewers Lettre aux exploitants Specimen letter notifying farmers of te visit of an interviewer. Plan de sondage Note on te selection of te sample. Tis is identical to te 2005 note, as te same sample was questioned in te two surveys. Agreste_primeur_21 5 4-page publication Double-click on te icons to open te documents Octobre 2008 Page 23 sur 23