Department for Environment Food and Rural Affairs (Defra)

Supplementary contacts:

• Scottish Government (SG)
• Welsh Government (WG)
• Department of Agriculture, Environment and Rural Affairs (DAERA-NI) 

In England, the Agricultural Statistics Act 1979 as amended by the Agriculture (Amendment) Act 1984, provides the power to obtain statistical information relating to agriculture from any owners or occupiers of land used for agriculture.  

In Scotland, the Agriculture Act 1947 provides the Scottish Government with the power to obtain statistical information relating to agriculture from any owners or occupiers of land used for agriculture.

In Wales, the Agricultural Statistics Act 1979 as amended by the Agriculture (Amendment) Act 1984, provides the power to obtain statistical information relating to agriculture from any owners or occupiers of land used for agriculture. The powers enshrined in this Act were devolved to the National Assembly for Wales following the Government of Wales Act 2006.

In Northern Ireland, the Agricultural Statistics (Northern Ireland) Order 2004 provides the Department of Agriculture and Rural Development (Northern Ireland) with the power to obtain agricultural information in relation to any land comprised in an agricultural unit.

GAEC = good agricultural and environmental conditions

England use the FSS thresholds (Annex II of Regulation 1166/2008) plus additional records with > 1 ha of hardy nursery stock, any area of mushrooms and holdings with temporarily reduced levels of activity e.g. empty pig/poultry sheds or short term rented out land (land > 5 ha that has been let on a short term basis at the last survey).

Northern Ireland include all active farm businesses with >=1 ha of farmed land, or with any cattle, sheep, goats, pigs, significant poultry or horticultural activity.

Wales include all holdings with a Standard Gross Margin (SGM) >0.

Scotland include all holdings with >=0.5 ha farmed land or have crops (including temporary grass) or have >1 livestock unit or >=1 worker, have deer or >=20 poultry. 

For all the UK, the FSS thresholds are implicitly covered in the national survey definition.

The thresholds mentioned above are as follows:

Notes:

• Tobacco, cotton, olive plantations, citrus plantations are NE characteristics.
• Hops is NS characteristic.

For England, the scope of the records sent to Eurostat is the same as the scope of the national survey.

For the rest of the UK, the scope of the records sent to Eurostat is smaller than their national survey (see above in item 2. for national threshold definitions).

They have areas included in “fallow land” (B_1_12) and permanent grassland and meadow no longer for production purposes (B_3_3) which are all eligible for subsidies, therefore the land is kept in GAEC conditions.

If a holding only has fallow or permanent grassland which is no longer being kept in GAEC conditions, it is removed from our holding register.

1. Firstly, where the holding has returned a Single Payment Scheme (SPS) subsidy claim, we use the grid reference provided on SPS. This relates to the grid reference of the land parcels covered by the claim.

2. If no grid reference can be found, we use grid references provided on the Cattle Tracing System (CTS). This relates to the location of the cattle.

3. For the very small number of holdings where no grid reference can be found, the postcode of the farm address is used to estimate a grid reference.

Whilst we have a location for every holding, caution needs to be exercised when using them. Ultimately they are our best point estimate for the centre of a spatial entity which can often cover a large land area. In the UK around 80% of holdings are smaller than 100 ha (1 km²) but there are almost 3 500 whose area is at least 500 ha (and this includes just over 1 000 farms whose area exceeds 1 000 ha or 10 km 2). These estimates provide a very useful aid to interpretation but ultimately should be considered as an indicative location and not a precise one.

Exceptions are:

(a) livestock items which were obtained from the bovine register as at 1 June 2016, and 

(b) rural development payments, which were approved between 1 June 2013 and 1 June 2016 and sourced from administrative data. 

Each of the statistical teams have their own well established timetable and organisation for the survey. Typically, the process is

• Develop the paper questionnaire and online data collection system (where application). Finalise the questionnaire. Test online data collection system. Online system goes live. (April 2015 - December 2015)
• Determine sample strategy and finalise numbers in selection. Complete selection. (February 2016 - April 2016)
• Complete validation specification and coding. (February 2016 - April 2016)
• Survey forms and email invitations to complete survey are sent to respondents. (May 2016)
• Reminders are sent (email and paper). (June 2016 - August 2016)
• Data received is validated, queried and corrected. (June 2016 - October 2016)
• Online collection and paper survey is closed. (September 2016 - October 2016)
• Administrative data is obtained, principally cattle data from the bovine register. (August 2016 - December 2016)
• Statistical analysis is carried out. (August 2016 - December 2016)
• Early results for England, Wales, Scotland and Northern Ireland are compiled and published. This is not a FSS publication but a routine statistics release of the June survey of agriculture results. (August 2016 - October 2016)
• Final results for England, Wales, Scotland and Northern Ireland are compiled and published. This is not a FSS publication but a routine statistics release of the June survey of agriculture results. (September 2016 - December 2016)
• Further work is carried out to prepare the FSS dataset for transmission to Eurostat. (January 2017 - December 2017)

These domestic arrangements mean that agricultural policy, data collection methods and data availability can vary slightly between the four countries. As data supplier, Defra tries to harmonise the data collection methodologies where possible but ultimately, as long as they supply the data specified in Regulation (EC) No 1166/2008, Defra has limited powers to influence the methodologies chosen in these countries.

Each administration has its own statistical staff who are responsible for cleaning and analysing their data. A lead statistical analyst in each administration undertakes the bulk of the statistical work with support from additional analysts and a survey support team. When the data is returned to the statistical teams, they run a series of validation checks on the data and the survey support teams follow up these with phone calls where appropriate to ensure data quality.

The following characteristics have been collected only from the sample: labour force, other gainful activities, irrigation, soil management, rural development, legal type, A_3_3_1 (more than 50% of production self-consumed by the holder) and A_3_3_2 (more than 50% of sales are direct sales).

A list frame approach is used in each country, i.e. a list of agricultural holdings.

In the UK any farmer will need to register a holding number for the holding in question if they intend to buy, sell or move livestock, sell crops for human consumption or claim any agricultural subsidies thus there is generally a very strong incentive to register. All of the statistical teams in the UK receive regular updates to their register that capture new holdings (and changes to contact details for an existing holding number). These registers are maintained continuously as they are used for a number of farm surveys.  

We select one sample but for England only, we apply different extrapolation factors depending on the characteristics:

- For some characteristics such as crops and livestock, we produce a full holding level imputed dataset using response data to estimate for the non-sampled and non-responding holdings. Thus we obtain land and livestock characteristics for all holdings in England. These characteristics have an extrapolation factor of 1 (and we use extrapolation factor A09).

- For characteristics where we are unable to impute, we use response data only and use the extrapolation factor A10 to enable grossing up to national totals.

Separate strata with higher sampling rates for horticultural holdings are used to ensure adequate precision.

Neyman Allocation was used to determine sample size allocation between strata. The allocation was calculated for all land and livestock variables separately using June 2005 data, then the mean allocation for each strata was used in each stratum. 

Rural development payments: There are a large number of clauses within the rural development payments regulation and it is up to member states to choose how to divide up their pot and which parts of the regulation they will fund.  Within the UK each devolved administration has opted to fund slightly different parts of the regulation, but overall the only area where there are significant numbers of beneficiaries is for agri-environment scheme payments. The exact source of the administrative data varies from country to country but basically it comes from the unit who makes the payments.  In some cases this is a single central source (as occurs in Wales and NI) or a number of sources (England and Scotland).

Cattle: In Great Britain, the headcounts of the numbers of cattle were obtained from the Cattle Tracing System (CTS). In Northern Ireland, the Animal and Public Health Administration (APHIS) system, which is also an EU audited cattle tracing system, is used. These provide a continuous record of the births, deaths and all movements of individual cattle in Great Britain thus it is possible to obtain a snapshot of all cattle on any given date. This date can be chosen to coincide with the reference date of the FSS.

Crop areas: In Scotland the Single Application Form (SAF) that farmers complete annually to claim farm subsidies contains detailed crop information. Any holdings that were in-scope for FSS2016 who submitted a Single Farm Payment (SFP) claim had their land data from the SAF dataset automatically integrated into the 2016 FSS dataset. 

Poultry and pigs: In Northern Ireland, data on poultry and pigs were derived from administrative sources, the Northern Ireland Annual Inventory of Pigs and the Northern Ireland Bird Register Update.

When the UK was considering switching to an administrative system in 2005, we opted to collect data by both methods in 2006 to check the feasibility of switching methods. A considerable amount of background analysis was undertaken at the time¹. The change of data source did introduce a step change into the series by identifying a higher number of cattle (6% or 360 thousand more in England in June 2006), but the trends were very similar to those of the June Survey.

When the admin data source was first used, the metadata was seen and examined for consistency with the Regulation 1166/2008. The admin data source is responsible for storing and recording the metadata but the metadata can be requested by Defra when needed.

The slight problem with agri-environment schemes is that a farmer need not (or possibly cannot) enter their entire farm into the scheme under one agreement.  The current generation of agri-environment agreements in the UK tend to run for 5 or 10 years.  When the latest generation of schemes were introduced a number of farmers had land within “classic schemes⁷” and could not therefore include all of their land within these new schemes otherwise there would be issues of double funding.  Such farmers would make a claim under the new scheme for land that was not covered by any existing schemes and then once their classic agreements come to an end make a second application.  Thus there can be situations where a given holding has a number of different agri-environment agreements logged within the database.  

For the remaining measures, the data are downloaded from the systems of the appropriate payment body.  For example, in England the data are downloaded from the CMEF on-line system⁸.  On this English system a registered user can download a list of all of the projects funded under a given rural development measure between point X and Y – in our case 1 June 2013 and 1 June 2016.  The slight problem with this system is that it does not include the CPH reference⁹. The data was matched to our farm register on the basis of names, addresses and postcodes.  Whilst this matching was quite a manual process it was possible because of the small number of projects claimed upon over the reference period.  The devolved administrations used a similar approach to England although often they had the benefit of a CPH or business identifier to aid the matching and weed out the non-agricultural claimants.

Crop area: In Scotland, any holdings that were in-scope for FSS2016 who submitted a Single Farm Payment (SFP) claim had their land data from the SAF dataset automatically integrated into the 2016 FSS dataset. These farmers could then receive a simplified survey form.  

6 CPH numbers are of the form cc/ppp/hhhh and it is common for leading zeros on the parish or holding part to be omitted.

7 These are the suite of older schemes, like Countryside Stewardship, that have been replaced by the current generation of Environmental Stewardship Schemes.

8 A system for monitoring rural development programme projects.

9 The reason is that some of the recipients of funding under options such as the encouragement of tourism do not actually need to be farmers.

In England farmers were also offered the chance to complete the survey online.  These farmers received an email inviting them to take part instead of a survey form¹¹.  The farmers were then able to log into the website and enter their data.  The system uses shaping questions to ensure that the farmer only see questions relevant to them. Links to the online system are included on the paper form and reminder cards so even farmers who received a paper survey form could opt to log in and complete the form online.  Similarly farmers who received the email invite but did not want to complete an online return could contact us and request a paper form.  

10 Scotland has a significant quantity of land-use data available from administrative systems.

11 The numbers targeted in this way were limited by the number of holdings for which we had captured an email address from previous correspondence.

The UK does not use face to face interviews for any component of the FSS due to the costs for government and the burden placed upon the farmer involved in collecting such data.  Although a number of respondents answered full survey questions over the telephone we would not formally constitute this as an “interview”.

One of our key tools to improve response rates is through sending reminders to farmers.  The exact nature and timing of these varies slightly between the four countries but the overall process is similar.  For the main survey form there were typically two reminders sent in the form of a reminder card/email.  The first is sent two or three weeks after the survey day and the second in mid July.  In some instances these were followed up with a third reminder.  In Northern Ireland, these included a duplicate survey form and were targeted towards larger holdings and/or minority sectors (e.g., horticulture) to maximise their impact.  Wales targeted larger non-responders through telephone calls instead of a third reminder whilst Scotland did not need one because of the quantity of data that they can access through administrative systems.

Larger farms that are part of composite units (primarily pig and poultry farms) are issued with separate special survey forms to ensure that maximum coverage is achieved.  This is a routine operational procedure that we run each year for our regular June Survey and it generates good data.  The strategy used is to send the forms to the head office for the company rather than the individual farms.  They then collate the information and send it back to us.  Some of the larger farming businesses, particularly those involved in horticulture, have multiple holding numbers and often a significant number of staff who work across all of their units.  This can make it hard for them to complete the labour data for a 12 month period.  To encourage responses we allow them to record all of their farm labour data under a single holding number and not complete the labour section on their other forms.  The farm then indicates to us all of the holdings within their group and on which one they have recorded their labour data.  We then use our point in time estimates of worker numbers and the activity on the holding to apportion these workers between the holdings that the farmers have not supplied data for.      

Our mailing lists and contact details are derived from the live agricultural holdings database in each of the four countries.  These are updated on a daily basis.¹² All staff, and in particular those recruited on a temporary basis specifically to deal with the survey receive a significant amount of training in advance of the survey.  This includes techniques on how to obtain information from difficult farmers and how they can use other information that we already hold about the farm to resolve issues with the supplied data.

12 The updates come from both other government registers / administrative systems, direct contact from the farmer (either comments on our survey forms or telephone calls) and undelivered mail that is returned to us.

In case of a census 1=3+4+5

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

This item is a subset of 3.

4>4.1+4.2

5=5.1+5.2

5.1>=5.1.1+5.1.2

21 863 (England only and excluding common land units -
we also imputed an additional 50 314 holdings in England which were not included in the gross sample)

6=5.2+5.1.2+4.2

When loaded onto their systems the statisticians run a series of validation checks against this incoming data.  These will be things like checking the components sum to the total and subtotals, comparing the data against the return from the previous year to flag up large changes and highlighting inconsistent responses in categorical questions.  

All of these basic checks were run whilst the survey was still operational, additional checks were subsequently run.

A validation database was built that mimics all of the Eurostat rules and this highlighted other records needing checking, mainly comparing the organic administrative data against the conventional land/livestock data collected in the FSS.      

Once a complete dataset was achieved with suitable source data for every FSS item the data were transferred from our data collection format to that required for FSS.  In many cases this was converting numeric data to the Eurostat categorical variables or summing data items together.

The tools used for data validation were IT system, charts for erroneous values at analysis stage, calculated influential statistics to assess influence of individual data points.

There is validation built into the online data collection system which prevents farmers from moving on to the next section if the response they have given is not appropriate (for example they have specified a greater area of crops than their UAA). The validation is carried out separately by each UK country, then the overall validation is carried out at the UK level by Defra.

• Estimates for farm labour as at 1 June (not those who carried out farm work on the holding during the 12 months ending on the reference day of the survey) are collected each year.
• In the livestock categories there are a number of instances where we collect more detail than required by the Regulation. For example, domestically sheep are quite important and therefore we have additional data categories such as lambs and three different categories for ewes.
• In the vegetable and soft fruit categories, we collect more detail than required by the Regulation.  For example, we collect raspberries, blackcurrants and other small fruit as three separate categories when the regulation only calls for “berry species”.

These characteristics are requested by the UK Government to meet its need for evidence to support the making of Government policy on agriculture.

For the main FSS items probably the biggest error results from non-response. 

Coverage of the poultry sector is difficult because the rapid turnover on the industry and its dominance by a few very large companies. The survey questions are designed to be a snapshot of the industry at 1 June. It is possible that on the day of the survey a number of producers may have no birds on their farm and be disinfecting their premises. To try and help with coverage, data is collected directly from the head office of the larger companies in a “special poultry exercise”.

Our validation rate of checks is around 30% on land and livestock characteristics and around 50% on labour characteristics as farmers find it difficult to understand the complex requirements. The main source of data error is farmers incorrectly completing the questionnaire. We carry out validation checks for these.

A number of data items were collected from administrative data rather than the farmer. Whilst this improves the coverage there are some issues in matching this data to the FSS register and fitting the data to the FSS categories.

Thus far none of the UK administrations have pursued legal action in relation to non-response on FSS because legal advice indicated that a successful prosecution would be unlikely and it was considered that such actions are more likely to lower response to subsequent surveys rather than improve them and would lead to a very significant increase in the costs of the survey.  The administrations would prefer to work together with farmers and farming industry to improve survey response. Non-responses are imputed into the results.

For the main FSS items probably the biggest error results from non-response. Non-response affects all farm types and whilst this varies slightly with farm type, UK level records of non-response by farm type are not compiled.

Any records that are flagged with issues (or are partially complete) are contacted via telephone using our own bespoke Computer Aided Telephone Interview (CATI) technology.

For data items collected from administrative data they are considered to offer complete coverage and the data are used in the supplied form.  The only changes applied to this data are those to map the data to the correct categories or in the case of the organic data to constrain the organic data to the total organic and inorganic data supplied from the survey.

For the surveyed items when the data is returned a series of validation checks are run against the data, for example, checking the components sum to the total and subtotals, comparing the data against the return from the previous year to flag up large changes and highlighting inconsistent responses in categorical questions. Any records that are flagged with issues (or are partially complete) are investigated by our survey support team. They use additional information such as comments farmers have written on the form, data from other surveys/admin data systems to try to resolve these issues – including data from the previous year (base data). Where they cannot use existing data the farmers are contacted via telephone to resolve the issues. Where a common problem arises, global updates can be applied to streamline the process, for example many farmers omit the section totals so we can automatically set the section total to be the sum of the components. In general less information is available to resolve issues with labour and diversification data than the land area and livestock data hence the former group usually require more telephone calls to farmers to clarify. 

Despite our best efforts we do not get responses from every farmer.  In addition, despite our call back strategy not all of the responses are fully complete and correct¹. The methods used to complete missing or incorrect sections and deal with non-response changes depending on the data group (labour / diversification and land-use / livestock). For the land-use and livestock categories, there is a well established method of dealing with the non-response and the returned forms that are partially complete or in error following the actions of our survey support team. The data for each holding is broken down into chunks that represent a group of data items such as sheep, grassland, arable crops². If there was an error within the group which could not be resolved by the survey support team, the data items for that group are imputed. 

1 In some cases we make numerous unsuccessful attempts to contact the farmer or the information that the farmer supplies will not resolve the issue.

2 Note no imputation is required for cattle as these come from the Cattle Tracing System (CTS) administrative system.  The Cattle Tracing System is a bovine register that records births, movements and deaths of cattle.

Data revision policy for the 4 institutions producing statistics that form the FSS dataset.

• Defra: https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/423291/statspolicy-revisions-22apr15.pdf
• Scottish Government: http://www.gov.scot/Topics/Statistics/About/CPSonRevisionsCorrections
• Welsh Government: http://gov.wales/statistics-and-research/about/statement-of-compliance/revisions-errors-postponements/?lang=en
• DAERA-NI: https://www.daera-ni.gov.uk/publications/daeras-statistics-charter

There are no differences between national and EU definitions of characteristics and classifications of characteristics.

1800 hours are taken to be the minimum annual working hours for a full-time employee, equivalent to 225 working days of eight hours each.

All the UK common land holdings are tenure classified as common land.

The data compiled for the FSS are not collected elsewhere in a suitable form with sufficient coverage or a large enough sample to permit widespread comparisons. Generally the best method of comparison is against previously compiled data. Land, livestock and point in time estimates for labour are collected annually on 1 June, thus land use and livestock data compiled for FSS 2016 can be compared to data compiled in June 2015 and earlier years to monitor year-on-year changes.

Confidentiality policy for the four institutions producing statistics that form the FSS dataset:

• Defra: https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/401606/statspolicy-dataprotection-05feb15.pdf
• Scottish Government: http://www.gov.scot/Topics/Statistics/About/CPSconfidentiality
• Welsh Government: http://gov.wales/statistics-and-research/about/statement-of-compliance/confidentiality-data-access/?lang=en
• DAERA-NI: https://www.daera-ni.gov.uk/publications/daeras-statistics-charter

An additional level of protection is applied if the tables are for spatial scales of NUTS4 or finer. This additional level involves calculating the proportion of the cell total contributed by the highest contributing farm. Where this value exceeds 85%, the cell value is suppressed to protect the identity of this dominant contributor.