Selecting the sample of farms from the FADN's field of observation
This chapter describes how farms are selected and the techniques that are used to achieve as high a degree of representativity as possible.
Member States conduct their own selection
Before the creation of FADN, several Member States were already conducting agricultural surveys based on farm accounts. Some of these surveys were based on a selective sample of farms - as opposed to the entire population of farms. To select a sample of farms, these Member States had established their own selection plans.
Most Liaison Agencies of the Member States continue to conduct national surveys and have thus retained their own selection plans. Current practice is for Liaison Agencies to design their own selection plans for the European Union survey. The plans are submitted to the FADN European Union Committee for approval. They vary in technical sophistication from one Member State to another.
The use of stratification
Within FADN's field of observation, there is a great diversity of farming. Some farms are very large (in terms of their economic size) while others may be very small. Some farms concentrate on crop production, others specialise in livestock rearing while yet others practice mixed farming, that is, both crop and livestock production. On these two criteria alone i.e. - economic size and type of farming -, the field of observation of European Union farms is highly heterogeneous.
To ensure that the sample of farms adequately reflects this heterogeneity, Liaison Agencies stratify the field of observation before the sample of farms is selected. If this were not done, there would be a greater risk that particular categories of farm (say, large dairy farms in one region, or small fruit farms in another region) would not be represented adequately (or at all) by the sample.
Stratification is a statistical technique which is used to increase sampling efficiency (i.e. to minimise the number of farms required to represent the variety of farms in the field of observation). The Commission makes extensive use of this technique and uses three criteria for stratification: region, economic size and type of farming, as described in the following sections.
For FADN purposes the European Union is divided into FADN regions. All farms in FADN's field of observation are classified into economic size classes and type of farming.
A detailed typology has been created for use by various bodies at European Union level. It is sufficiently broad to encompass the many different types of farming that are found in the European Union. This typology is described in Commission Regulation (EC) No 1242/2008 of 8 December 2008.
Typology identifies the principal types of farming, which are then further broken down. How are farms allocated to a specific type? In other words, what are the definitions of different types of farming?
Types of farming are defined in terms of the relative importance of the different enterprises on the farm. Relative importance is itself measured quantitatively as a proportion of each enterprise's SO to the farms' total SO. (see example of classification of a farm).
For the purpose of computing Standard Results, the Commission uses groups of farming types. For more information on how these Standard Results are calculated and published, refer to Diffusion chapter.
Example of classification of a farm according to European Union typology
Assume: a farm with 50 dairy cows and 10 breeding sows and 5 dairy cows
two principal types of farming would appear to be suitable descriptions of this farm:
|74||mixed livestock, mainly granivores|
To which type of farming does this farm belong?
|Enterprise||SO||Size of enterprise||Enterprise SO||Enterprise SO as proportion of& farm's total SO|
|Breeding sows||1000||50 breeding sows||50.000||83.3|
|Dairying||2000||5 dairy cows||10.000||16.7%|
| || ||Farm's total SO||60.000||100%|
The definition of the two principal types of farming are as follows:
|51||specialist pigs||pigs > 2/3 of farm's total SO|
|74||mixed livestock, mainly granivores||grazing livestock and forage ≤ 1/3 granivores in farm's total SO|
Since breeding sows > 2/3 of farm's total SO, this farm is classified as "specialist pigs" for principal type of farming.
The definition of the 3 particular types of farming is as follows:
|511||specialist pig rearing||breeding sows contribute > 2/3 of farm's total SO|
|512||specialist pig fattening||piglets and othe pigs contribute > 2/3 of farm's total SO|
|513||pig rearing and fattening combined||holdings in class 51 excluding those in classes 511 and 512|
Since breeding sows contribute > 2/3 of farm's total SO, this farm is classified as "specialist pig rearing" for particular type of farming.
The universe and field of observation represented as a matrix of cells
The 3-way stratification of the universe allows it to be represented as a 3-dimensional matrix of cells. The number of farms in each cell is derived from the Farm Structure Survey (FSS) organised by Eurostat. This survey employs the same typology as that used for FADN.
Each cell corresponds to a specific category of farms. Some cells represent a large number of farms: for instance, in Ireland there are approximately 18 000 farms in the economic size class from 4 000 to less than 8 000 euro in the cell that practise farming type 46 - specialist cattle (rearing and fattening). Other cells represent very few farms: in Denmark for example, there are only about 40 farms of size from 15 000 to less than 25 000 which practise farming type 2 - specialist horticulture. Needless to say, some cells are empty - such as those representing vineyards, big or small, in Finland.
The Commission and the Liaison Agencies select the sample of farms not from the field of observation as a whole but from the cells which make up the field of observation. Sample farms are thus selected from each cell - in this way all the cells are, in principle, represented in the sample. Thus the FADN sample of farms reflects the heterogeneity in the field of observation.
Determining the optimal size of the sample
Sampling fractions vary from cell to cell. In some Member States, the Liaison Agencies have sufficient data on the variability of farms within the field of observation to compute optimal sampling fractions. In other cases, this is not possible and sampling fractions are set according to the number of farms in the cell. After the selection plan is drawn up, farms can shift from one cell to another if there is a change in their economic size or type of farming. This and other similar factors influence the sampling fraction as described below.
The extent to which the sample is random
Ideally, farms are selected at random from the field of observation. However, various factors prevent full randomisation:
- Availability of farm accounts. To complete the European Union FADN Farm Return, a suitable set of farm accounts (or similar financial information such as receipts, invoices, etc.) must be readily available. Some farmers do not have the necessary information at hand, and in these instances it is impractical to attempt to complete the Farm Return. In some countries, the Liaison Agencies assist farmers to keep accounts if these would not otherwise be kept.
Overall, the number of farmers keeping accounts is gradually increasing.
- Voluntary participation. The participation of farmers is on a voluntary basis. Some of those farms initially selected for the sample may not want to participate. In this case, the farm will be replaced by drawing another farm from the same cell in the field of observation.
As a result; the sample is; effectively, drawn at random from the subset of farms within the field of observation which fulfil the above two conditions.
Before the beginning of each accounting period, Liaison Agencies are obliged to draw up a selection plan in accordance with Commission Regulation (EU) No 1291/2009 of 18 December 2009 and subsequent amendments. A variety of selection procedures are used in Member States. During implementation of the selection plan several problems can occur, for example there are not enough farmers who are willing to participate from particular cells and that the actual sample may fall short of the intended size and distribution.
The size of the FADN sample
The reasons why in practice the intended sample size may not be attained
There are several reasons why the intended sample size may not be attained or, indeed, may be surpassed. For example, it may be difficult to find sufficient farmers in a particular cell who are both willing to participate and who have the necessary information. Another explanation may be that a participating farmer may give up farming before the completion of the accounting year. A further reason may be that the European Union FADN Farm Return is incorrectly completed and unable to be corrected, thus failing at the control stage. (See Data Quality Chapter).
Actual sample size in recent years
The FADN sample size and average weights of a sample farm differ between Member States, as shown below.