The need for a weighting system
A special weighting system is used in the calculation of EU-FADN results.
It is based on the principle of "free expansion": a weight calculated for the sample applies to each holding of the sample (extrapolating factor). In order to calculate this individual weight, holdings in the sample and in the field of survey are stratified according to the same three criteria: FADN region, type of farming and economic size class. The individual weight is equal to the ratio between the numbers of holdings, of the same classification cell (FADN region x type of farming x economic size class), in the population and in the sample.
Consider, for example, very large specialist dairy farms in Brittany. If there are 20 farms belonging to this group in the FADN sample and if there are 1000 in the population, then each individual farm in the sample for that group will have a weight of 1000 / 20 = 50.
The purpose of the weighting system is to take account of different sampling fractions for different cells. In the production of FADN results, weighted averages are calculated using these weights applied to each individual farm recorded in the sample.
Information on the field of observation
To calculate weighting factors it is necessary to have an accurate and up-to-date field of survey. The FADN field of survey is a subset of the EUROSTAT Farm Structure Survey (FSS). As this survey is not annual (but every 2-3 years), the Commission services use the most recent information available.
The aggregation of cells
The theoretical grid used for stratification comprises numerous classification cells (140 FADN regions x 62 types of farming x 14 economic size classes = 121 520 cells).
It can be represented as a 3-dimensional matrix of cells:
Certain cells (type of farming x economic size classes) do not exist or are very infrequent in some Member States. Through the application of selection plans, Member States ensure that all significant categories of holdings - that is, all classification cells that contain holdings in the field of survey - are represented in the farm samples. In practice, however, the intended sample may not be attained so that some cells in the field of survey are not represented in the sample.
From its knowledge of the field of survey and selection method, the Commission services, assisted by the relevant Member State Liaison Agency, are able to judge for which farm types cells in the sample may be empty. Cells with similar characteristics are then clustered (aggregated) and treated as a single cell when calculating the weights.
In the European Union typology of agricultural holdings, two similar types of mixed livestock farms are distinguished: 73 (mixed livestock, mainly grazing livestock) and 74 (mixed livestock, mainly granivores). In most Member States there are a large number of both types of such farms in the field of survey and it can be expected that the cells representing these types of farming at the level of the sample will have a sufficient number of farms. In some other Member States there are comparatively few farms of each of these types.
Thus, there is a risk that one of these types will not be represented in the sample. To counter this risk, the two principal farm types are aggregated to a single broader farm type at a higher level in the typology: in this case, principal type of farming 73 and principal type of farming 74 would be aggregated to form one cell representing general type of farming 7 (Mixed livestock).
Aggregation of cells increases the coverage of the field of survey. However, it may also involve a loss of precision and representativity for certain farm types and/or sizes of farms.
Tables in annex show the clustering schemes that the Commission and the Member States have agreed to apply for weighting EU-FADN results, as from the accounting year 1995.