Farm structure survey – survey coverage

This article presents the set of thresholds defined by the legislation applicable to the farm structure survey (FSS) and the various sets of thresholds applied by countries while respecting the minimum coverage requirements laid down in the legislation, as well as the impact of the application of these thresholds on the comparability of the FSS results over time. The article is part of an online publication on methodological articles dealing with the farm structure survey. The analysis and findings are based on the datasets and the national methodological reports (NMRs).

Full article

Thresholds described in legislation

In accordance with the legislation applicable to the FSS, the target population is the universe of agricultural holdings. Countries have, however, been allowed to apply various survey thresholds to limit the survey population and thereby the survey sample and costs, as long as the minimum survey coverage set up in the FSS legislative framework was guaranteed.

As illustrated in Table 1, until the FSS 2007, in accordance with Article 6 of Regulation (EEC) No 571/88, the countries which applied a threshold of above one hectare of utilised agricultural area (UAA) committed themselves to fix this threshold at a level excluding only the smallest holdings, which together contributed 1 % or less to the total standard gross margin (SGM).

Table 1: Thresholds used in FSS legislation

Following the entry into force of Regulation (EC) No 1166/2008, the minimum requirements for survey coverage from the 2009/2010 FSS onwards have been modified. Countries which used a survey threshold above one hectare of UAA were allowed to fix this threshold at a level that excludes only the smallest agricultural holdings which together contribute 2 % or less to the total UAA excluding common land, and 2 % or less to the total number of farm livestock units (LSU). In addition, in all cases, countries have to include in the survey populations all holdings which comply with at least one of the following set of physical thresholds, defined in Annex II of Regulation (EC) No 1166/2008:

Utilised agricultural area (arable land, kitchen gardens, permanent grassland, permanent crops): 5 hectares
Permanent outdoor crops (fruit, berry, citrus and olive plantations, vineyards and nurseries): 1 hectare
Other intensive production:
- Fresh vegetables, melons and strawberries, which are outdoors or under low (not accessible) protective cover: 0.5 hectares
- Tobacco: 0.5 hectares
- Hops: 0.5 hectares
- Cotton: 0.5 hectares
Crops under glass or other (accessible) protective cover:
- Fresh vegetables, melons and strawberries: 0.1 hectares
- Flowers and ornamental plants (excluding nurseries): 0.1 hectares
Bovine animals (all): 10 heads
Pigs (all): 50 heads
Breeding sows: 10 heads
Sheep (all): 20 heads
Goats (all): 20 heads
Poultry (all): 1 000 heads.

Overview of survey coverage changes between 2007 and 2013

Table 2 provides an overview of thresholds used by countries since 2007 and presents the modifications to these thresholds applied in the 2010 and 2013 surveys .

Table 2: Overview of the survey coverage in countries, for FSS 2007-2013

Regulation (EC) No 1166/2008 brought changes in the thresholds and therefore in the coverage of the 2009/2010 agricultural census in some countries. A similar situation occurred also in the FSS 2013, when an additional adjustment of the country-specific thresholds took place in some countries (Belgium, France, Croatia, Italy, Luxembourg, Malta, Poland, Finland and Iceland). These modifications limit the comparability of the FSS results over time.

The survey coverage changes can be grouped into 3 different categories, see Table 3:

1. Decrease in coverage

As compared to the FSS 2007, in the 2010 agricultural census, the Czech Republic, Germany, Poland, Slovakia and the United Kingdom increased the thresholds to determine the survey population, decreasing thus the population covered with significant impact on some key characteristics.

Furthermore, as compared to the 2010 census, in the FSS 2013, Italy and Finland increased their thresholds, decreasing to some extent the population covered. The increase of UAA and LSU thresholds was possible as the smallest holdings, although representing a large share of the total agricultural holdings in the country, are insignificant with regards to their share in the total UAA, livestock population and standard output (SO) of the country. Additionally, Malta excluded from the FSS 2013 those holdings which had type 9 in 2010 (holdings with only fallow land or kitchen gardens and with no standard output in 2010), also decreasing the covered population.

2. Increase in coverage

In Denmark, Croatia, Italy and Sweden, the entry into force of Regulation (EC) No 1166/2008 led to an increase of the population covered in the FSS 2010 compared to 2007.

3. Change of coverage with no significant impact on comparability over time at least for key characteristics

The change of coverage between 2007 and 2010 does not significantly affect comparability of data over time, at least for the key characteristics in Bulgaria, Estonia, Ireland, Spain, France, Cyprus, Latvia, Luxembourg, Hungary, the Netherlands, Austria, Portugal and Norway^[1] . Furthermore, the change of coverage between 2010 and 2013 does not significantly affect comparability of data over time at least for the key characteristics in Belgium, France, Croatia, Luxembourg and Poland^[2].

Table 3: Impact of the modifications of survey coverage on national populations

In Iceland, the 2013 survey coverage is not perfectly comparable with 2010. The new holdings added in 2013 used a different coverage principle than in 2010, see Table 2. However, Eurostat does not have the necessary information to analyse the impact over time.

Lithuania, Romania and Slovenia have not changed the survey coverage since their first implementation of the FSS.

Greece did not change thresholds over the period 2007-2013 (in fact over the whole period 2000-2013). However, the lack of update of the Farm Register led to an overestimation of the FSS characteristics in 2007 and trends across years should, therefore, be carefully interpreted (see the 2009/2010 NMR).

Changes of survey coverage in each country between 2007 and 2013

This section presents the countries which have modified their survey coverage over time. Although the presentation focuses on the changes since the FSS 2007, the data compare the populations fulfilling the set of coverage criteria common over the entire period. These comparable populations are obtained by applying the sets of thresholds for all years to the data of each year and taking the holdings above all sets of thresholds, in the geographical territory common to all years. The analysis considers the following key characteristics: number of holdings, total UAA, total LSU, arable land, kitchen gardens, permanent crops, permanent grassland, cattle, sheep, goats, pigs, laying hens, labour force directly employed by holdings in annual work units (AWU) and economic size of holdings (standard output (SO) between 2007 and 2013 and standard gross margin (SGM) between 2017 and 2013).

Even in the presence of coverage changes, the comparability of data on the total UAA and the total LSU is in some cases guaranteed as Regulation (EC) No 1166/2008 requires that at least 98% of the total UAA (excluding common land) and 98% of the total LSU should be covered in case the UAA threshold exceeds 1 hectare ^[3]. When changes have a significant impact on data, they usually affect the comparability over time of characteristics related to the holdings, such as number of holdings, legal type of holdings, characteristics related to the holders and managers, labour force, other gainful activities and, in some cases, specific land use and livestock categories, typically kitchen gardens. In this case, trends cannot be directly interpreted from disseminated data over survey populations.

Belgium

The survey coverage did not change between 2000 and 2010 surveys. Compared to the 2010 census and the previous structural surveys (which used the Farm Register established by Statistics Belgium as a sampling frame), the 2013 FSS was based on the Integrated Administration and Control System (IACS) register. In general, the IACS does not include those holdings that are not subject to IACS declarations, for example small holdings or holdings belonging to retired farmers. However, Belgium explained that this additionally covered all holdings which did not apply for subsidies as long as they were above the thresholds of at least 2 hectares of arable land or 50 ares of land under greenhouse or produced animal manure equivalent of (at least) 300 kg of phosphate (P2O5). As before 2013, Belgium filtered out all holdings which did not produce for sale in the FSS 2013. All in all, the change of the population frame between 2010 and 2013 caused the loss of very few farmers, usually those who are retired and who did not submit IACS declarations. As it is not possible to distinguish between holdings which apply and holdings which do not apply for subsidies in Eurostat data, more detailed analyses are not available.

For more information, see the national methodological reports.

Czech Republic

The Czech Republic increased the survey thresholds in the 2010 census, which led to the decrease of survey coverage between 2007 and 2010. The 2010 thresholds remained the same in the FSS 2013.

Table 4 shows the impact of the coverage changes on key characteristics of all survey years, by presenting first the relative changes of key characteristics in the survey populations and then in the comparable populations obtained after applying all years' thresholds to each year's data. While apparently, between 2007 and 2010, the number of holdings decreased by 42 %, the kitchen gardens by 70 %, and the labour force by 21 %, and the population of sheep increased by 6 % and that of goats by 7 %, in reality, the number of holdings decreased only by 4 %, the kitchen gardens by 43 %, and the labour force by 14 %, and sheep increased by 12 % and goats by 25 % over comparable populations. If the same thresholds as in 2010 and 2013 are applied to the previous surveys (FSSs 2003, 2005 and 2007), on average 43.2 % fewer holdings would be covered, as well as 8.5 % less labour force and 1 % less SO. The same thresholds would decrease the number (in LSU) of goats counted in the FSS between 2003 and 2007 on average by 21.6 %, sheep by 5.6 %, kitchen gardens by 42.9 % and permanent crops by 3.0 %, while the impact on total UAA, the total LSU, total standard output, cattle, pigs, laying hens, arable land and permanent grassland would remain negligible. Therefore, the change of thresholds in 2010 mainly affects trends in indicators on number of holdings and labour force, indicators related to holding characteristics such as the legal personality of the holding, characteristics related to the holders and farm managers of the holdings, as well as other gainful activities. Also, though the impact on the total LSU, the total UAA and the total SO is negligible, the impact is significant for specific livestock and land use categories (kitchen gardens, permanent crops, sheep and goats), while not significant for others (cattle, pigs, laying hens, arable land and permanent grassland). Trends of the affected characteristics cannot be directly interpreted from survey data disseminated on Eurostat's website. To do that, trends have to be interpreted over comparable populations.

Table 4: Key characteristics, relative changes in the surveyed and comparable populations (above all years' thresholds), 2003-2013, Czech Republic

For more information, see the national methodological reports.

Denmark

In compliance with Regulation (EC) No 1166/2008, Denmark surveyed all agricultural holdings having at least 5 hectares UAA or a standard output of at least EUR 7500. Moreover, compared to the previous survey years, all farms falling below these thresholds, but complying with a set of different physical thresholds (see Table 2), were also included in the target population from the FSS 2010 onwards. This modification led to a general increase of survey coverage between 2007 and 2010. In particular, Denmark mentioned that the new thresholds introduced with Regulation (EC) No 1166/2008 led to the inclusion of 431 more farms in the Danish farm structure census 2010. Denmark also chose to include farms with fur animals from 2010 onwards, which resulted in a further 735 and 545 farms added to the 2010 and 2013 populations, respectively. However, as these farms are not eligible under the above Regulation, they were removed from the final datasets in Eurostat. The survey coverage between 2010 and 2013 did not change.

Table 5 shows the impact of the coverage change on key characteristics of survey years by presenting the relative changes of key characteristics, first in the survey populations and then in the comparable populations obtained after applying all years' thresholds to each year's data. The total number of holdings in the survey populations decreased by 7 % in 2010 compared to 2007, while in the comparable populations (above all years' thresholds) by 12 %. As Table 5 illustrates, the effect on the total UAA and UAA main categories is negligible. However, the addition of specific thresholds relative mainly to different crop and animal categories increased some FSS indicators. In particular, the standard output increased by 22 %, while in the comparable populations it would be only by 10 %, the total livestock increased by 7 %, while in the comparable populations it would drop by 6 % and the labour force decreased by 9 %, while it would be by 17 % in the comparable populations over 2010/2007. As Table 5 shows, for specific livestock categories, instead of a small decrease in laying hens and goats (by 6 % and 10 %), or even an increase in pigs and sheep LSU (by 9 % and 2 %) in 2010 compared to 2007, there would be a (much larger) decrease in all these categories over the comparable populations 2007/2010 (i.e. decrease in LSU of laying hens by 25 %, goats by 20 %, pigs by 8 % and sheep by 7 %). Similarly, if we apply all sets of thresholds to all years, in 2010 and 2013 on average 5.5 % fewer holdings, 12.6 % fewer LSU, 9.2 % less labour force and 9.4 % less SO would be covered. While the impact of the change of coverage is negligible on the total UAA, UAA main categories (arable land, kitchen gardens, permanent crops and permanent grassland) and cattle, it is significant for the number of holdings, the labour force, the standard output, the total LSU and specific livestock categories (sheep, goats, pigs and laying hens), see Table 5.

Table 5. Key characteristics, relative changes in the surveyed and comparable populations (above all years' thresholds), 2000-2013, Denmark

For more information, see the national methodological reports.

Germany

Germany increased the survey thresholds in the 2010 agricultural census which led to the decrease of survey coverage between 2007 and 2010. The survey coverage between 2010 and 2013 did not change.

Table 6 shows the impact of the coverage change on key characteristics of survey years by presenting the relative changes of key characteristics, first in the survey populations and then in the comparable populations obtained by applying all years' thresholds to each year's data. While apparently, between 2007 and 2010, the number of holdings decreased by 19 %, the kitchen gardens by 11 %, the labour force by 10 % and the permanent grassland by 4 %, in reality, the number of holdings decreased only by 6 %, the kitchen gardens increased by 1 %, the labour force decreased only by 6 % and the permanent grassland decreased only by 2 % over comparable populations. The impact of the survey coverage change on the total UAA, the total LSU and the standard output, as well as on arable land, permanent crops, permanent grassland and the analysed livestock categories (cattle, sheep, goats, pigs and laying hens) is negligible. On the other hand, the impact is (very) significant for number of holdings, kitchen gardens and labour force. For example, if all thresholds are applied to all survey years, the comparable populations would cover on average 9.5 % fewer holdings, 9.6 % fewer kitchen gardens and 3.4 % less labour force, see Table 6. The change of survey coverage, therefore, mainly affects comparability of indicators on labour force, kitchen gardens and number of holdings, and indicators related to holding characteristics, such as legal personality of the holdings, characteristics related to the holders and farm managers of the holdings and other gainful activities. Trends of these enumerated characteristics cannot, therefore, be directly interpreted from survey data disseminated on Eurostat's website. To do that, trends have to be interpreted over comparable populations.

Table 6: Key characteristics, relative changes in the surveyed and comparable populations (above all years' thresholds), 2000-2013, Germany

For more information, see the national methodological reports.

Ireland

Before 2010, Ireland included holdings which had at least 1 hectare of UAA, at least one LSU or a specialist activity. In 2010, Ireland did not apply any threshold and included in the survey population all holdings with agricultural activities, i.e. holdings in the agricultural register (based on the IACS and Bovine Registers). The survey coverage between 2010 and 2013 did not change. According to information received from Ireland, the application of thresholds used in 2007 has a negligible impact on the number of holdings in 2010. The change of thresholds in 2010 on the population covered can, therefore, be considered negligible with regards to the key characteristics and analysis of trends can be made on survey data available in Eurostat's online database without corrections.

For more information, see the national methodological reports.

Greece

In Greece, the Farm Register was updated until 2009 by the agricultural census, which was carried out once every decade, and to a certain extent by sample agricultural surveys and administrative sources. Greece specified in its 2009/2010 NMR that it was not possible to record and depict all the changes in the structure of the agricultural holdings for a number of reasons. For example, a number of holdings that were not operational any more were shown in the Farm Register as active. Thus the number of holdings included in the Farm Register, which was the sampling frame for the FSS 2007, is potentially greater than the real one in that particular case. Greece considers that this led to an overestimation of the FSS 2007 variables (and possibly also 2003 and 2005), explaining to a significant degree the apparent large decrease in the number of holdings (and utilised land) between the 2007 FSS and the 2009 Census, see Table 7. ELSTAT has already taken actions to improve the methods and the techniques of updating the Farm Register. In the Framework of the Joint Overall Statistical Greek Action Plan (JOSGAP), an external consultancy has already taken place and ways for improving the register have been indicated. ELSTAT envisages implementing a project with the aim of linking its Farm Register with the registers kept by other Greek Authorities, which will enable ELSTAT to update the Farm Register continuously and more entirely. Greece did not modify the FSS thresholds over the period 2000-2013 and the application of all years' thresholds to each year's data did not affect the trends, see Table 7. However, because of the above-mentioned problem, trends in the data on Eurostat's website between years 2007 and 2009 should be carefully interpreted as changes in characteristics may not only reflect real changes, but results are likely to have been affected by methodological issues.

Table 7: Key characteristics, relative changes in the surveyed and comparable populations (above all years' thresholds), 2000-2013, Greece

For more information, see the national methodological reports.

Spain

Before 2009, Spain applied a threshold of 0.1 hectare UAA together with some other physical thresholds, but sent data to Eurostat only on holdings which passed the threshold of 1 hectare of UAA and some additional physical thresholds, see Table 2. In the 2009 agricultural census, Spain increased the threshold of 0.1 hectare to 1 hectare of UAA and introduced some additional physical thresholds, in particular on tobacco, hops and cotton. These additional thresholds increased the population covered in 2010 compared to the population covered in previous survey years. According to information received from Spain, applying the thresholds used in 2010 on data collected in 2000 would increase the number of holdings in 2000 by 0.2 %. The change of thresholds in 2009 on the population covered can, therefore, be considered negligible with regards to the key characteristics and analysis of trends can be made on survey data available in Eurostat's online database without corrections. The survey coverage between 2010 and 2013 did not change.

For more information, see the national methodological reports.

France

The thresholds applied by France to agricultural holdings have remained unchanged since 1955. However, there are differences in geographical coverage related to the overseas territories over survey years. Table 8 shows the impact of this coverage change on key characteristics over years, by presenting the key characteristics first in the survey populations and then in the comparable populations. Before 2003, the FSS did not cover the French overseas territories. Table 8 shows that including the overseas territories in 2003 had a negligible impact on the total UAA, the total LSU, the standard gross margin, as well as on arable land, kitchen gardens, permanent grassland, permanent crops, cattle, sheep, pigs and laying hens. However, the number of holdings and the labour force directly employed on the holdings increased by about 3 % after including the overseas territories. While apparently, between 2000 and 2003, the number of holdings decreased by 8 %, the labour force decreased by 4 % and the goats increased by 9 %, if the overseas territories had not been included, the number of holdings would have decreased by 11 %, and the labour force by 7 %, while the goats would have increased by only 4 %. Therefore, the change of coverage in 2003 mainly affects trends of indicators on labour force, goats, number of holdings and indicators related to holding characteristics such as the legal personality of the holding, characteristics related to the holder and farm manager of the holdings and other gainful activities. Trends of these enumerated characteristics cannot, therefore, be directly interpreted from survey data disseminated on Eurostat's website. To do that, trends have to be interpreted over comparable populations, see Table 8.

Table 8: Key characteristics, relative changes in the surveyed and comparable populations (without the overseas territories), 2000-2013, France

In the FSS 2010 and 2013, the covered geographical area was mainland France and four overseas departments: Guadeloupe, Martinique, Guyana and Réunion, while the territories of Saint-Martin, Saint-Barthélemy and Mayotte were excluded. The territories of Saint-Martin and Saint-Barthélemy were excluded because their status changed and they are no longer part of the national territory. Mayotte became a department only in 2011 (after 2010). For the 2013 survey, 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. On the other hand, in 2013, as compared to 2010, the coverage of Guyana territory was reduced to professional and more accessible holdings on the coast side. In 2010, in Guyana, these kind of holdings amounted to 3 052 holdings out of the total of 5 983 Guyanese holdings (a difference of 2 931 holdings). While the number of holdings in Eurostat's online database decreased in 2013 from 516 104 holdings to 472 207 holdings, i.e.by 8.5 % compared to 2010, the correct trend is obtained after removing the 2 931 holdings from 2010. This means that a comparable number of holdings decreased from 513 173 to 472 207 holdings (by 8 %) over 2010-2013. The impact of the change of coverage in France between 2010 and 2013 is negligible for number of holdings, UAA, LSU, labour force directly employed by holdings, standard output, arable land, kitchen gardens, permanent grassland, permanent crops, cattle, sheep, goats, pigs and laying hens and these indicators can be compared over 2010-2013 without problems.

For more information, see the national methodological reports.

Croatia

Before becoming a Member State in July 2013, Croatia organised two FSSs, in 2007 and 2010, each carried out as a sample survey with different survey thresholds. While in the FSS 2007 Croatia applied the thresholds of 1 hectare of UAA and 2 LSU to obtain the target population, in the FSS 2010 Croatia modified these thresholds to 0.4 hectare UAA and 0.5 LSU, thus increasing the survey coverage. According to the 2010 NMR and the analysis of the 2010 data, the change of thresholds between 2007 and 2010 affected mainly the number of holdings and kitchen gardens and somewhat the labour force directly employed by holdings, goats, pigs and laying hens. Trends of these characteristics cannot be directly interpreted from survey data disseminated on Eurostat's website. To interpret trends for these indicators, a correction needs to be made to both survey years, to exclude from the analysis holdings falling under both sets of survey thresholds. On the other hand, the change of thresholds between 2007 and 2010 did not significantly affect the total UAA, the total LSU, the standard output, arable land, permanent grassland, permanent crops, cattle and sheep.

As concerns the FSS 2013, the thresholds applied did not change compared to FSS 2010, except for some minor modifications, see Table 2. Furthermore, unlike in 2013, in 2010 Croatia did not apply the thresholds again after the data collection in order to remove the holdings below all thresholds during the reference period. However, Croatia confirmed that these differences have a non-significant impact on the comparability between 2010 and 2013 datasets. As Table 9 shows, the impact of the coverage change is negligible for all key FSS indicators, including the number of holdings, and as Croatia mentioned, the comparability between these years is not affected.

Table 9 shows the impact of all coverage changes on key characteristics over years by presenting the key characteristics, first in the survey populations and then in the comparable populations, obtained after applying all years' thresholds to each year's data. The key indicators mostly affected by the coverage changes as a whole are number of holdings, kitchen gardens, permanent crops, livestock, goats, pigs, laying hens and labour force. Table 9 further illustrates that if we keep only the comparable populations for the analysis over the years, on average 19.2 % fewer holdings, 2.2 % fewer LSU and 8.2 % less labour force would be covered. Furthermore, the impact is indeed (very) significant for specific crop and livestock categories: kitchen gardens, permanent crops, goats, pigs and laying hens. On the other hand, the impact on the total UAA, standard output, arable land, permanent grassland, cattle and sheep is not significant.

Table 9: Key characteristics, relative changes in the surveyed and comparable populations (above all years' thresholds), 2007-2013, Croatia

For more information, see the national methodological reports.

Italy

Before the FSS 2010, Italy included all holdings with at least 1 hectare of UAA, or with the production reaching a certain proportion for sale, or exceeding certain physical thresholds. In 2010, Italy lowered the thresholds on UAA and fixed them at a regional level, see Table 2. This modification led to the increase of survey coverage between 2007 and 2010. About 10 % of the holdings in 2010 are under the 2007 thresholds, while about 1 % of holdings in 2007 is under the 2010 thresholds.

In the FSS 2013, the survey thresholds were again modified and set to 1 hectare of UAA, resulting this time in the decrease of the survey coverage. However, the reduction in the number of holdings was caused by the change of the survey coverage of holdings sent to Eurostat rather than by the modification of the survey thresholds in Italy. While in 2010, the records sent to Eurostat covered both:

holdings over 1 hectare of UAA and over the thresholds of physical characteristics other than UAA defined in Annex II of Regulation (EC) No 1166/2008, and
holdings under the above-mentioned thresholds but above regional UAA thresholds, or with livestock totally or partially destined for market, or with any surface of flowers and ornamental plants, vegetables, vineyards and fruit trees because of their economic value,

the records sent to Eurostat in 2013 covered only the holdings in the first category above. Thus, 29.7 % of the 2010 holdings were no longer covered by the data sent to Eurostat in 2013. This coverage is in fact enough to fulfil the minimum requirements concerning UAA and LSU set in the Regulation (EC) No 1166/2008. Apparently, the number of holdings decreased from 1 620 884 to 1 010 328 holdings, i.e. by 38 % between 2010 and 2013. However, taking into consideration only the part of the 2010 population which corresponds to the 2013 population, this decrease is reduced to only 11 %. Such a decrease can be further explained by the natural tendency of decrease in the number of holdings and family labour force and increase in the non-family labour force (especially contractors and employees on a non-regular basis). Table 10 shows that the change of thresholds over 2010-2013 significantly affected the comparability of number of holdings, area of kitchen gardens, permanent crops, labour force, and as emphasised by Italy in the 2013 NMR, of olive plantations. On the other hand, the comparability of UAA, LSU, standard output, arable land, permanent grassland, cattle, sheep, goats, pigs, laying hens was not (significantly) affected.

Table 10 also shows the impact of all coverage changes on key characteristics of all survey years by comparing the key characteristics, first in the survey populations and then in the comparable populations obtained after applying the most constrained 2013 thresholds to the survey data from all years. Data collected in the previous surveys and in the FSS 2013 are not directly comparable for number of holdings, kitchen gardens, permanent crops and labour force. Trends for these characteristics cannot be directly interpreted from the survey results disseminated on Eurostat's website. To interpret trends for these indicators, a correction needs to be made to the previous survey years, to exclude from the analysis holdings falling under the 2013 thresholds. On the other hand, data collected in the previous surveys and in the FSS 2013 are comparable concerning UAA, LSU and, to a certain extent, standard output, arable land, permanent grassland, cattle, sheep, goats, pigs and laying hens. Finally, as Table 10 illustrates, if we retain only comparable populations over the 2013 thresholds in all years, between 2000 and 2010 on average 30.1 % fewer holdings, 13.5 % less labour force directly employed by the holdings and 2.4 % less standard output would be covered. Though the impact on the total UAA and the total LSU is negligible, it can be very significant for specific crop categories, e.g. on average 7.7 % fewer permanent crops would be covered between 2000 and 2010, see Table 10.

Table 10. Key characteristics, relative changes in the surveyed and comparable populations (above 2013 thresholds), 2000-2013, Italy

For more information, see the national methodological reports.

Cyprus

Unlike in the surveys carried out in 2003, 2010 and 2013, when the FSS threshold was set to 0.1 hectare of UAA, in the FSSs 2005 and 2007 the UAA threshold was reduced to 0.05 hectare, see Table 2. However, as Table 11 shows, these modifications had only a negligible impact on the FSS key indicators of which trends can be rather attributed to the normal evolution rather than to the methodological changes.

Table 11: Key characteristics, relative changes in the surveyed and comparable populations (above all years' thresholds), 2003-2013, Cyprus

For more information, see the national methodological reports.

Latvia

The first FSS in accordance with Regulation (EEC) No 571/88 was carried out in 2001 in the form of a census. Since then, the FSS threshold has been modified several times. While in 2001 and 2003, the threshold of 1 hectare of agricultural area or certain economic threshold applied, in 2005 and 2007, an economic threshold of 0.1 European size unit (ESU) (EUR 120) and for non-active farms an agricultural area related threshold were used. Finally, in the FSS 2010 and 2013, the thresholds of 1 hectare of agricultural area or a certain level of standard output (SO) applied. Table 12 shows the impact of threshold modifications on key characteristics of all survey years by comparing the key characteristics, first in the survey populations and then in the comparable populations obtained after applying the 2001 thresholds to data from all years. The population covered in 2010 was 23 % less than in 2007 and 41 % less than in 2001, while the comparable population (above 2001 thresholds) in 2010 would be 21 % less than in 2007 and 43 % less than in 2001. The NMR 2010 explains that the definition of the survey thresholds was based on the ongoing trend of the progressive concentration of the agricultural production in large holdings.

Table 12: Key characteristics, relative changes in the surveyed and comparable populations (above 2001 thresholds), 2001-2013, Latvia

Table 12 further illustrates that while the total UAA increased by 1 % and the total LSU decreased by 3 % between 2007 and 2010, the changes in some land and livestock related characteristics were quite significant. In particular, the area of kitchen gardens increased by 173 %, permanent crops decreased by 52 %, goats decreased by 35 %, laying hens increased by 21 % and sheep increased by 19 % between 2007 and 2010. Additionally, the standard output increased by 33 % and the labour force decreased by 19 % in the same period. However, having compared the above data with the same data in comparable populations (above 2001 thresholds), it can be concluded that the modifications of thresholds between 2007 and 2010 did not have any significant impact on the FSS key indicators or even on the main specific livestock and land use categories. Latvia also informed that the change does not have any impact on FSS characteristics.

If the same thresholds as in 2001 are used in the subsequent surveys (FSSs 2003, 2005, 2007, 2010 and 2013), on average 4.7 % fewer holdings would be covered, 5.1 % fewer kitchen gardens, 2.1% fewer goats and 2.0% less labour force, while the loss in the other key characteristics would be negligible.

For more information, see the national methodological reports.

Luxembourg

In the 2010 agricultural census, Luxembourg raised the survey thresholds which led to the decrease in survey coverage compared to the previous FSSs. Furthermore, in 2013, Luxembourg added a threshold of beehives to the set of thresholds used in 2010, this time increasing in principle the survey coverage.

Table 13 presents the impact of threshold modifications on key characteristics of all survey years by comparing the key characteristics, first in the survey populations and then in the comparable populations obtained after applying all years' thresholds to the data from all years. Between 2007 and 2010, the changes in some key characteristics were small or absent, e.g. the number of holdings decreased by 4 %, the total UAA remained stable, the livestock increased by 4 %, the labour force decreased by 1 %, while the changes in other key characteristics were quite significant, e.g. LSU of goats increased by 81 %, standard output by 18 % and the kitchen gardens decreased by 17 %. Having compared the above data with the data obtained from comparable populations (above all years' thresholds), in reality, the key characteristics decreased/ increased by nearly the same percentages. Therefore, the change of thresholds in 2010 on the population covered can be considered negligible with regards to the key characteristics analysed. However, the impact on other characteristics may be significant and caution should be taken when interpreting trends across time for these characteristics.

Table 13: Key characteristics, relative changes in the surveyed and comparable population (above all years' thresholds), 2000-2013, Luxembourg

With respect to 2010 and 2013, Table 13 shows the same relative changes in the key characteristics over the survey populations and over the comparable populations, thus the impact of adding the threshold on beehives in 2013 can be considered negligible on comparability of the key characteristics analysed over these two years.

Finally, as Table 13 illustrates, if all years' thresholds are applied, between 2000 and 2013 on average 2.3 % fewer holdings, 2.7 % fewer kitchen gardens and 2.6 % fewer goats would be covered. The impact is negligible on the other analysed key characteristics: UAA, arable land, permanent grassland, permanent crops, LSU, cattle, sheep, pigs, laying hens, labour force and standard output, see Table 13.

For more information, see the national methodological reports.

Malta

In Malta, no physical or economic threshold applies to FSS surveys. However, it should be noted that the coverage in 2013 is different from the coverage in 2010 and earlier years. Compared to 2010 and earlier years, the 2013 coverage excluded those holdings which were type 9 holdings in 2010 (holdings with only fallow land or kitchen gardens and with no standard output in 2010). Still, Malta covered these holdings in the national FSS 2013 dissemination, with the values of the characteristics from FSS 2010. These holdings are not covered and disseminated by Eurostat in the FSS 2013.

Apparently, the number of holdings decreased by 25 % between 2010 and 2013. However:

If we remove the type 9 holdings from both 2010 and 2013 populations, the actual decrease is only by 12 %. However, this approach is not optimal, as unlike in 2010, the 2013 population is under-covered (holdings which were type 9 in 2010 and type different from 9 in 2013 are not, and cannot possibly be included in 2013, as they were not surveyed in 2013). Table 14 shows the impact of the coverage change on key characteristics of all survey years by comparing them first in the survey populations, and then in the populations obtained after removing type 9 holdings from all years. This approach, as mentioned above, has the disadvantage of under-coverage in 2013. Table 14 shows that, if we compare only the populations without type 9 holdings over the years 2003-2013, on average there would be a loss of 18.0 % holdings, 5.0 % UAA, 6.7 % labour force, 2.6 % arable land and 32.6 % kitchen gardens. There would be no loss of permanent crops, livestock and standard output. This is obvious as it is expected that type 9 holdings have fallow land (part of arable land, which is part of UAA), kitchen gardens, labour force, but do not have permanent crops, livestock and standard output.
If we remove the type 9 holdings only from the 2010 population (which are no longer covered in 2013), the actual decrease is only by 1 %. However, this approach is also not optimal, for example because the structure of holdings by typology over time is affected: while the 2010 data will not include any type 9 holdings, the 2013 data will include those type 9 holdings which acquired this status between 2010 and 2013. Moreover, this approach cannot be sustained long term as it would mean following a panel in future surveys (always removing the holdings which had type 9 in 2010). Table 14 also illustrates the impact of the coverage change on key characteristics over 2010-2013 by comparing them first in the survey populations, and then in the populations obtained after removing the type 9 holdings only from 2010. The impact is negligible on permanent crops, livestock, all livestock main categories and standard output and (very) significant on number of holdings, UAA, arable land, kitchen gardens and labour force.

Table 14: Key characteristics, relative changes in the surveyed and comparable populations (without type 9 holdings), 2003-2013, Malta

We can conclude that there is no optimal approach to show actual trends over time. There is always under-coverage of 2013 population concerning holdings that were of type 9 in 2010 and of a type different from 9 in 2013, which cannot be dealt with.

For more information, see the national methodological reports.

Poland

Poland increased the survey thresholds in the 2010 agricultural census, which led to the decrease of survey coverage between 2007 and 2010. Table 15 shows the impact of this change on key characteristics of survey years by comparing the key characteristics, first in the survey populations and then in the comparable populations obtained after applying all years' thresholds to the data of each year.

Table 15: Key characteristics, relative changes in the surveyed and comparable populations (above all years' thresholds), 2003-2013, Poland

While apparently, between 2007 and 2010, the number of holdings decreased by 37 %, the total UAA by 7 %, of which the biggest drop was for the area of kitchen gardens (by 59 %), the total LSU decreased by 7 %, of which the biggest drop has been recorded for the goats (by 26 %), the labour force directly employed by the holdings decreased by 16 % and the SO increased by 11 %, in reality, the number of holdings decreased only by 16 %, the kitchen gardens by 31 %, the goats by 7 %, the labour force by 10 % and the SO increased by 13 % over comparable populations. If all thresholds (basically 2010 thresholds) are applied to the previous surveys (FSSs 2003, 2005 and 2007), on average over 2003-2007, 26.5 % fewer holdings and 7.7 % less labour force directly employed by the holdings would be covered. Although the impact on the total LSU, the total UAA and the standard gross margin is negligible, the impact can be very significant for specific livestock and land use categories. The thresholds would decrease the number of goats counted in the FSS between 2003 and 2007 on average by 20.2 %, sheep by 3.6 %, laying hens by 4.9 %, kitchen gardens by 28.8 % and permanent crops by 4.4 %. The impact on cattle, pigs, arable land and permanent grassland would remain negligible. The change of thresholds in 2010 as compared to 2007 therefore does not affect UAA, LSU and standard output/standard gross margin key characteristics, but mainly affects trends of number of holdings, labour force, indicators related to holding characteristics such as legal personality of the holdings, characteristics related to the holders and farm managers of the holdings and other gainful activities, and specific livestock and crop categories (goats, sheep, laying hens, kitchen gardens and permanent crops). Trends of these characteristics cannot, therefore, be directly interpreted from survey data disseminated on Eurostat's website. To interpret trends for these characteristics, a correction should be made to the survey data, to exclude from the analysis holdings falling under all years' thresholds (basically 2010 thresholds being enough), see Table 15.

Poland introduced in the FSS 2013 new physical thresholds for natural persons (mushrooms area, number of rabbits, wild animals kept for slaughter and beehives, as well as covering all organic farming holdings regardless of thresholds). Furthermore, concerning the legal persons, all such holdings were included in 2010, while in 2013 only the ones with specific agricultural activities were included and, in addition, regardless of the agricultural activities, the ones with agricultural land of at least 1 hectare and the ones which run livestock production. Although the thresholds in 2013 include wild animals, holdings with only fur animals (other than rabbits) were not sent to Eurostat. Moreover, fur animals are not counted in 'other livestock' in the dataset, but the rearing of fur animals is counted among the other gainful activities. It is not possible to analyse the change of coverage completely accurately based on the information available in Eurostat, because not all the thresholds applied used characteristics as defined in Eurostat's database. Using some approximations, Eurostat made however an analysis. As Table 15 shows, the change of thresholds between 2010 and 2013 can be considered negligible with regards to all analysed key characteristics. The 2013 NMR also mentions that 2010 and 2013 data are comparable.

For more information, see the national methodological reports.

Slovakia

Slovakia increased the survey thresholds in the 2010 agricultural census which led to the decrease of survey coverage between 2007 and 2010. The survey coverage did not change between 2010 and 2013.

Table 16 shows the impact of the coverage changes on key characteristics of all survey years by comparing the key characteristics, first in the survey populations and then in the comparable populations obtained after applying all years' thresholds to the data of each year. While apparently, between 2007 and 2010, the number of holdings decreased by 65 %, kitchen gardens by 71 %, permanent crops by 18 %, goats by 22 %, labour force directly employed by the holdings by 39 % and laying hens increased by 30 %, in reality, the number of holdings decreased only by 37 %, kitchen gardens by 46 %, permanent crops by 12 %, goats by 3 %, labour force by 29 % and the laying hens increased by 37 %. The total UAA, the total LSU, the standard output, as well as arable land, permanent grassland, cattle, sheep and pigs remained rather unchanged over comparable populations.

Table 16: Key characteristics, relative changes in the surveyed and comparable populations (above all years' thresholds), 2000-2013, Slovakia

Table 16 also illustrates that if all thresholds are applied to all surveys, the 2013 and 2010 data are not affected, but previous years' data are affected as follows: on average over 2000-2007, there would be 43.8 % fewer holdings, 48.3 % fewer kitchen gardens, 7.1 % fewer permanent crops, 22.7 % fewer goats, 4.9 % fewer laying hens and 13.4 % less labour force. The other key characteristics are not affected. Therefore, the change of thresholds in 2010 mainly affects trends in indicators using data on the number of holdings and labour force, indicators related to holding characteristics, such as legal personality of the holdings, characteristics related to the holders and farm managers of the holdings and other gainful activities, as well as kitchen gardens, permanent crops, goats and laying hens. Trends of these characteristics cannot, therefore, be directly interpreted from survey data disseminated on Eurostat's website. To interpret trends for these indicators, a correction needs to be made to all years, to include only those holdings situated above the thresholds from all years. However, the key characteristics concerning the total UAA, the total LSU, the standard output, arable land, permanent grassland, cattle, sheep and, to a certain extent, pigs can be compared over time without population corrections.

For more information, see the national methodological reports.

Finland

Until the FSS 2005, Finland included holdings exceeding the threshold of 1 hectare of UAA or with an economic farm size of at least one European size unit (ESU). These thresholds were, however, modified in 2007, when the survey coverage included all farms with at least 1 hectare of arable land or at least one LSU, as well as those horticultural enterprises with less than 1 hectare of arable land but which are engaged in horticultural production intended for sale (e.g. greenhouse enterprises). According to the information provided in the NMR 2007, the impact of this modification was negligible on all key FSS characteristics. The same thresholds, as applied in 2007, were applied to the FSS in 2010. In the FSS 2013, Finland decided to replace these physical thresholds with an economic threshold, i.e. with the standard output of at least EUR 2 000, decreasing thus to some extent the population covered. Table 17 shows the impact of the coverage changes on key characteristics of all survey years by comparing the key characteristics, first in the survey populations and then in the comparable populations obtained after applying all years' thresholds to all years' data. In fact, as Table 17 illustrates, the modification of the thresholds affected the number of holdings, the kitchen gardens and the permanent grassland. The impact of the changes in thresholds on the population covered can, therefore, be considered negligible with regards to the other characteristics analysed (UAA, LSU, labour force directly employed by the holdings, standard output, arable land, permanent crops, cattle, sheep, goats, pigs and laying hens). The impact on specific characteristics may, however, be significant and caution should be taken when interpreting trends across time for these characteristics. Finally, as Table 17 illustrates, if all years' thresholds are applied, on average 2.6 % fewer holdings would be covered between 2000 and 2013.

Table 17: Key characteristics, relative changes in the surveyed and comparable populations (above all years' thresholds), 2000-2013, Finland

Table 17 also illustrates that the specific changes between 2010 and 2013 significantly affect only the number of holdings out of all analysed characteristics, the comparability of the other characteristics being maintained.

For more information, see the national methodological reports.

Sweden

Until 2010, thresholds applied to FSSs differed from the thresholds established by the relevant legal framework. According to the information provided in the Swedish NMR in 2005, the comparability with data from previous surveys, especially for the number of holdings and areas with temporary grasses, is limited. This is because of the implementation of the Common Agricultural Policy (CAP) reform from 2005 and the change from area based subsidies and animal subsidies to single farm payments. Between 2005 and 2007, the data comparability is good.

However, in the FSS 2010, a combination of the old national thresholds and the new thresholds that are in line with EU Regulation was introduced to prevent discontinuity in national time series. This modification led to the increase of survey coverage compared to the previous surveys. The survey coverage between 2010 and 2013 did not change.

Table 18 shows the impact of the coverage changes on the characteristics of all survey years, by comparing the key characteristics first in the survey populations and then in the comparable populations obtained after applying all years' thresholds to each year's data. The total number of holdings decreased by 2 % in 2010 compared to 2007, while in the comparable populations (above all years' thresholds) it would be by 6 %. The effect on the total UAA, the LSU and the standard output is not significant. However, the addition of specific thresholds, related mainly to different crop and animal categories, increased some FSS indicators. In particular, the area of permanent grassland decreased between 2007 and 2010 by 7 %, while it would be by more than 13 % in comparable populations, sheep increased by 11 %, while in the comparable populations it would be only by 7 % and the labour force decreased by 13 %, while it would be by 15 % in the comparable populations over 2010/2007. Similarly, if all thresholds are applied to all data, on average 3.7 % fewer holdings, 9.8 % fewer permanent crops and 2.5 % less permanent grassland would be covered, though the impact on the other FSS indicators is negligible, see Table 18.

Table 18: Key characteristics, relative changes in the surveyed and comparable populations (above all years' thresholds), 2000-2013, Sweden

For more information, see the national methodological reports.

United Kingdom

The FSS thresholds, and hence the survey coverage, have been modified several times in the United Kingdom since the 2000 survey. Moreover, the various UK countries used slightly different thresholds. For example, compared to the threshold of 6 hectares of total area and other physical thresholds applied in 2000, between 2003 and 2007 England and Wales covered all active farms, i.e. all holdings thought to have some form of agricultural activity at the time of the survey according to the Farm Register. However, this threshold has again been modified following the entry into force of Regulation (EC) No 1166/2008 and from 2010 onwards the threshold of 5 hectares of UAA, together with additional physical thresholds, applies. The increased 2010 threshold led to the decrease in survey coverage in 2010 compared to previous surveys. Between 2010 and 2013 the survey coverage did not change.

Table 19 shows the impact of the survey coverage changes on key characteristics of all survey years, at first by comparing the characteristics in the survey populations and then in the comparable populations obtained after applying the most constrained 2010 or 2013 thresholds to the data of all years. While apparently, between 2007 and 2010, the number of holdings decreased by 18 %, the labour force directly employed by the holdings by 13 % and the number of goats (in LSU) increased by 8 %, in reality, the number of holdings decreased only by 7 %, the labour force by 10 % and goats increased by 12 % over comparable populations. If the 2013/2010 thresholds are applied to the previous surveys (FSSs 2000-2007), on average 22.9 % fewer holdings, 8.1 % less labour force and 13.6 % fewer goats would be covered. The impact on the total LSU, UAA, SO and other analysed characteristics is negligible.

The change of thresholds in 2010, therefore, affects trends of number of holdings, goats, labour force, holding characteristics such as legal personality of the holdings, characteristics related to the holders and farm managers of the holdings and other gainful activities. Trends of these characteristics cannot, therefore, be directly interpreted from survey data disseminated on Eurostat's website. To interpret trends, a correction has to be made to previous survey years, to include in the analysis only holdings above 2013/2010 thresholds, see Table 19. On the other hand, the time series data is comparable on the total UAA, LSU, SO as well as arable land, permanent grassland, permanent crops, cattle, sheep, pigs and laying hens.

Table 19: Key characteristics, percentage in the surveyed and comparable population (above 2013 thresholds), 2000-2013, United Kingdom

It is also important to note that there are differences between the coverage of data disseminated by Eurostat and by the United Kingdom (though not by much in the majority of cases) because Scotland, Wales and Northern Ireland use slightly different thresholds in their national surveys:

for Northern Ireland, the thresholds include all active farm businesses with at least 1 hectare of farmed land, or with any cattle, sheep, goats, pigs, significant poultry or horticultural activity.
for Wales, the thresholds include all holdings with a SGM greater than zero.
for Scotland, the thresholds include all holdings with at least 0.5 hectares of farmed land or with any crops (including temporary grass) or with more than 1 livestock unit or at least 1 worker, with deer or at least 20 poultry.

For more information, see the national methodological reports.

Conclusions

Where national survey coverage has been modified following EU legislation and national decisions, some trends in disseminated data likely mislead users. Users could get information on survey coverage changes from Eurostat metadata and the national methodological reports, but such action is less probable for data extracted for a large number of countries. Besides, the metadata information on the impact of the coverage change on the main indicators does not clarify the impact on specific indicators that users might use. Therefore, countries and Eurostat are now discussing approaches to disseminate more comparable FSS results on a specific place on Eurostat's website.

Source data for tables and graphs

Farm structure survey - survey coverage: tables

Direct access to

Notes

↑ According to the 2010 NMRs from Bulgaria, Estonia, Spain, Hungary, Austria, Portugal and Norway, and according to information from Ireland, France, Cyprus, Latvia, Luxembourg and the Netherlands.
↑ According to the 2013 NMR from Belgium and the analysis in this article for the other countries.
↑ For FSS 2007 and previously, the minimum coverage was required in terms of standard gross margin by Regulation (EEC) No 571/88. Standard gross margin is directly related to UAA and LSU.

[1] According to the 2010 NMRs from Bulgaria, Estonia, Spain, Hungary, Austria, Portugal and Norway, and according to information from Ireland, France, Cyprus, Latvia, Luxembourg and the Netherlands.

[2] According to the 2013 NMR from Belgium and the analysis in this article for the other countries.

[3] For FSS 2007 and previously, the minimum coverage was required in terms of standard gross margin by Regulation (EEC) No 571/88. Standard gross margin is directly related to UAA and LSU.

[1]

[2]

[3]