Eurostat, the statistical office of the European Union.

E1: Agriculture and fisheries

2920 Luxembourg LUXEMBOURG

Animal production statistics cover three main sub-domains based on three pieces of relevant legislation and related gentlemen’s agreements.

• Livestock and meat statistics are collected under Regulation (EC) No 1165/2008. They cover meat production, as activity of slaughterhouses (monthly) and as other slaughtering (annual), meat production (gross indigenous production) forecast (semi-annual or quarterly), livestock statistics, including regional statistics. A quality report is also collected every third year.
• Milk and milk product statistics are collected under Decision 97/80/EC implementing Directive 96/16/EC. They cover farm production and utilisation of milk (annual), collection (monthly for cows’ milk) and production activity by dairies (annual) and statistics on the structure of dairies (every third year). An annual methodological report is also collected.
• Statistics on eggs for hatching and farmyard poultry chicks are collected under Regulation (EC) No 617/2008, implementing Regulation (EC) No 1234/2007 (Single CMO Regulation). They cover statistics on the structure (annual) and the activity (monthly) of hatcheries as well as reports on the external trade of chicks.

European Economic Area countries (EEA - Iceland, Liechtenstein and Norway) are requested to provide milk statistics, with the exception of those related to home consumption, as stated in Annex XXI of the EEA Agreement. Liechtenstein is exempted in the Agreement, only Norway is concerned.

The Agreement between the European Community and the Swiss Confederation on cooperation in the field of statistics states that Switzerland must provide Eurostat with national milk statistics and, after 2013, livestock and meat statistics.

The same statistics are requested from the candidate and potential candidate countries as EU acquis.

The statistical tables disseminated by Eurostat are organised, under Animal production (apro_anip), into three groups of tables on Milk and milk products (apro_mk), Livestock and meat (apro_mt) and Poultry farming (apro_ec). This later label covers statistics on hatcheries and trade in chicks and on production of eggs for consumption. The regional animal production statistics collected on livestock (agr_r_animal) and on cows’ milk production on farms (agr_r_milk_pr) are disseminated separately.

Due to the change in the legal basis or in the methodology, the time series may be broken. This is indicated by a flag in the tables.

Further data about the same topics refer to repealed legal acts or agreements. The tables on statistics on the structure of rearing (apro_mt_str) and the number of laying hens (apro_ec_lshen) or of equidae (apro_mt_lsequi) are therefore no longer updated. The same applies to some variables (external trade of animals and meat), periods (surveys in April or August) or items (number of horses) included in other tables.

The detailed content of each table and the reference to its legal definition is provided in the table below.

Table 3.1: Data tables disseminated regarding animal production statistics

Some standard code lists cover the concepts over the domains. The list of items and their definition is in any case derived from the legislation, but the coding integrates different approaches. The following concepts have been integrated in a single list agriprod:

• Live animals
• Meat and meat products
• Dairy and other animal products (except meat)

A handbook on concepts and definitions used for animal production statistics is provided in Annex 1.

Explicit size classes are used to describe the dairy entreprises by yearly quantities (t) collected, processed or produced and the hatcheries (egg incubation capacity).

Regional data

The territorial classification of regional data (tables agr_r_milkpr and agr_r_animal) is broken down according to the NUTS classification for Member States and to Eurostat’s list of Statistical Regions for Candidate countries and EFTA countries.

Statistics on livestock, on farm production, on other slaughtering and utilisation of milk cover agricultural holdings in Member States.
The minimal coverage for livestock sample surveys is of at least 95 % of the national population with reference to the last survey on the structure of agricultural holdings (FSS).

Livestock surveys may be conducted independently by livestock category or as a sub-set of items surveyed with a wider scope (livestock survey as a whole, farm production survey, annual census) or recorded with a wider objective in the case of registers (every animal owner). Depending on the design, some over-coverage can be observed.

Statistics on external trade in chicks are designed to reflect foreign trade in chicks from hatcheries with more than 1,000 incubation places.

Other animal production statistics cover EU plants whose activity is slaughtering, milk collection, milk product production or hatching fertilised eggs. The surveys must cover exhaustively dairies representing 95 % of the cows’ milk collected by Member States, the balance being represented by sampling or other sources. The methodological questionnaire collected from Member States reports whether a correction is applied for cross-border collection of milk from dairy farms in a country by dairy enterprise in a neighbour country.

See also Annex 1 for further explanations.

Among concepts used in animal production statistics (see Annex 1), some can be reported because of their specificity.

Gross indigenous production (GIP) is the number of animals slaughtered plus the balance of intra-Community and external trade for the same kind of live animals. GIP is thus the number of animals from a Member State (indigenous) apparently (gross) slaughtered or exported alive.

Slaughtering is measured through activity of slaughterhouses from 1 January 2009 (application of Regulation (EC) No 1165/2009), i.e. production of marketable meat for human consumption. Estimates of ‘other slaughtering’ can be added for a more accurate picture of meat production.

Livestock population is accounted by categories that capture their rearing, either for fattening then slaughter, or for herd renewal, i.e. for breeding and/or milking.

Milk statistics are led by the concept of ‘national dairy’, i.e. the dairy sector is considered as a single process, which internal flows are not (intended to be) taken into account.

Use of raw milk is followed through production of its two main components, fat and protein content. Milk processed is thus accounted for as an aggregate of UWM (utilised whole milk, with the full content of fat and proteins) and USM (utilised skimmed milk, with the full content of proteins,without fat). As a process can produce skimmed milk further to the main (fat) product and, in such a case, USM can be negative. This is especially the case for cream and butter production.

Bovine animals are domestic animals of the species Bos taurus and Bubalus bubalis, including hybrids like Beefalo. Clarification on the implementation of this definition led to the integration of buffaloes and hybrids into the category 'bovine animals'.

Chickens means all animals of species Gallus gallus, including broilers and boiling hens. This concept was applied with Regulation (EC) No 1165/2008 on 1 January 2009.

Regional data

Region means a sub-division of a Member State territory. Depending on the statistics, 'region' refers to:

• NUTS 2 for milk production (table agr_r_milkpr), the NUTS reference being the version applicable on the date of data transmission.
• NUTS 2, except for DE and UK (NUTS1) for livestock statistics (table agr_r_animal).

For data on the structure of hatcheries, a particular region (the most important) can be considered as representative of the national data in BG, EL, LV and AT.

Agricultural holdings are the statistical units for livestock surveys and animal production statistics at farm level (milk production and meat). Depending on the statistics collected, a more precise definition can be used, based on their activity or their structure, e.g. dairy farms producing raw milk, or farms with livestock or with sheep or goats.

Dairy enterprises -- undertakings of two types:

• Collection centres collect milk or cream and transfer it in whole or in part to other enterprises without any processing. They are often defined as referred to in Article 2(2) of Council Directive 96/16/EC.
• Dairies and agricultural holdings purchase milk or milk products from agricultural holdings or collection centres with a view to transforming them into milk products. They are often defined as referred to in Article 2(1) of Council Directive 96/16/EC.

Some enterprises process milk products obtained from a dairy as defined above, e.g. skimmed milk into milk powder or yogurt, and may appear to be excluded from the definition of dairy enterprises. Nevertheless, non-packed intermediate products are considered as raw products and such dairies are therefore covered as statistical units for the purpose of some statistics, whatever the enterprise supplying them.

Slaughterhouses are registered and approved establishments used for slaughtering and dressing animals whose meat is intended for human consumption. In countries in which ‘hygiene package’ is not fully implemented (slaughterhouses not registered or not approved by the EU can nevertheless produce for the local market) all slaughterhouses are covered.

Hatcheries are establishments for incubating eggs, hatching and supplying chicks (exceptionally almost hatched eggs).

For some animal production statistics, the statistical units are not explicitly defined, i.e. they refer to all enterprises. This is the case for the statistics on trade of chicks and this used to be the case for slaughtering statistics. Indirectly, the statistical units are the reporting enterprises dealing with one of parameters to be measured by the statistics.

The statistical population is the framework of the statistical units in the reference Member State or country for the reference period.

Nevertheless, data collection may be organised in a different way by a respondent other than the statistical unit. For instance, milk delivered by farms to dairies is accounted for by both units, and can be obtained more easily from dairies, of which there are fewer.

For monthly milk statistics, the population covers the dairies collecting cows’ milk. The quantity of milk products processed may be therefore underestimated compared to national production.

The reference area is the territory of the Member States as defined by Decision 91/450/EEC. For non-EU countries, territory follows the definition agreed bi-laterally with Eurostat.

Data are presented in chronological series for each country and for the EU. The covered period varies according to the country, depending on the date of accession to the EU or of application of the legal requirements for statistics. They are available at least for the EU Member States, depending on the date they became members of the EEC or EU, from legislation to repeal of the legal text and insofar as Member States comply with legislation.

Statistics are available, at least partly, since reference years 1960 (milk statistics), 1964 (meat statistics) and 1970 (livestock statistics and statistics on eggs for hatching and chicks) for the countries that were Member States at this time.

Further time periods may be available if collected on a volunteer basis, especially:

• Data collected under a gentlemen’s agreement
• Data collected for accessing and acceding countries
• Data collected under other specific agreements

The production forecast cover four or six quarters for pigs, three or four semesters for bovine animals, and two or zero semester for sheep and goat, depending on the size of the national livestock. At the end of September, year N, the forecast must be available up to the end of: - year N+1 for bovine animals and - first half of year N+1 for pigs. At the end of February, the forecast regarding sheep and goats must be available up to the end of the year for the countries with significant production.

Not applicable.

The data sources may be sample surveys or censuses for milk and livestock statistics. Nevertheless administrative sources may be used for obtaining these results in order to limit burden on the respondents. This is especially the case for bovine livestock. The milk quota registers have been used up to the end of the milk quota regime (April 2015).

See also Annex 3 for livestock and meat statistics.

For most of the data collected, collection frequency is the time granularity of the data, i.e. monthly data are collected monthly and annual data annually.

Annual totals provided in tables apro_mk_cola, apro_mt_pann and apro_ec_poula are disseminated at the same time as the data from December.

GIP forecast for meat are collected twice a year or, for the countries with a limited livestock population, annually.

Data on the structure of dairy enterprises are collected every third year.

Data and metadata transmission is executed in the same manner in all three sub-domains. The input consists of the national data which are ready for transmission and the output comprises the XML data files in eDAMIS servers.

CNAs transmit regional (where required) and national statistics to Eurostat exclusively via eDAMIS. The data comprise all variables listed in Chapter 2 heading 5 except those for which it is indicated that they are computed by Eurostat. The data comprise nine datasets on milk statistics, 16 datasets on livestock and meat production statistics and three datasets on poultry statistics, displayed in Table 2. Special notes about the transmission of some livestock and meat datasets are given in Table 3.

The transmission means used is eDAMIS Web Forms (eWF). A specific form exists for each dataset, presented as a spreadsheet-like table on screen. According to usual eWF functionality, the form automatically retrieves from eDAMIS servers and displays the data transmitted for earlier reference periods. CNA staff type or import the statistics for the current reference period, and also for earlier periods if they wish. The form implements pre-validation, i.e. “real-time” validation.

If there are no critical errors and no basic errors without justification the data can be transferred to eDAMIS, where they are converted into XML files and are stored in eDAMIS’ servers.

When the data are transmitted the domain manager is automatically notified by eDAMIS.

Confidential data

Each data item is accompanied by a flag which shows whether the data is confidential or not.

Regarding milk statistics in particular, 14 countries flagged certain items as confidential since three or less dairy enterprises had a significant contribution to the production of milk or of milk products such as powered dairy products.

Monitoring of data transmission

The monitoring of data transmission starts at the time the collection takes place. Countries, which have not provided data one week after the end of the deadline, are notified with a reminder sent by the eDAMIS collection system automatically. Eurostat domain managers contact the data provider three weeks later. In case of non-response Eurostat contacts the data provider again before sending the formal reminder letter, which is the last step before starting an infringement procedure.

1. Data validation in Member States

Data validation in Member States is known depending on their reporting.

• For milk statistics, Member States validate the data collected from farms and dairies. Comparisons with data obtained from the same dairy/farm in previous collection rounds of the survey and with data from other surveys are usually carried out. Moreover, variables collected from dairies are compared with similar ones collected from farms. No more information about other validation performed by the Member States before data submission to Eurostat is available.
• For livestock and meat statistics, the information about data validation is scarce. It covers all MSs with the exception of Austria since at the time of writing the present document no quality report was received from the country.
  Eleven CNAs report that they implement “real-time” validation of the data they collect, i.e. validation at the moment of collection. 22 CNAs mention explicitly that they validate the data: 18 compare the data with data for the same statistical units from earlier data collections, while 14 CNAs do other comparisons, which are not described. One CNA uses the results of a sample survey to check the data about slaughtering collected monthly from administrative sources.
  Reporting of the use of imputation, re-calibration or other correction / adjustment methods is connected explicitly with the treatment of non-response. It cannot therefore be known whether they are also used to correct other errors identified during validation. Re-calibration is used by eight CNAs, to treat non-response in livestock surveys. Imputation is employed by four CNAs in the case of monthly slaughtering statistics and by 14 CNAs for livestock statistics. The most usual imputation method is to take data for the same statistical unit from the previous data collection. A small number of CNAs use donor imputation or take data from other surveys or registers.
• On eggs for hatching and chicks, no information is available about data validation performed by the Member States before data submission to Eurostat.

2. Validation rules agreed with Member States

The validation rules agreed with Member States are documented in a Handbook on data validation in animal production statistics.

The validation implemented by eWF in “real time”, as the user types the data, is known as “pre-validation” in the domain. There are two severity classes of validation rules:

- Basic rule: a validation rule, which can be broken. However the CNA is required to provide a justification within the form for the violation of the rule;

- Critical rule: a validation rule, which must be satisfied. As long as the form contains critical errors the functionality for transmitting the data to Eurostat is disabled.

Pre-validation of animal production statistics
Validation, level 0
It is checked that all mandatory cells contain information (data or flags) and that data types are valid. The only non-numerical value allowed in data cell in this domain is “NA” for cells for which data are not available. Flag “C” indicates confidentiality of the value. The flags that can be inserted in flag cells are pre-defined and available to the user as a drop-down list.
All data cells of all transmitted datasets are mandatory with few exceptions:

• All data cells for regional data of countries with a single region,
• Livestock statistics for the regions having fewer than 75 000 bovines, 150 000 pigs, 100 000 sheep or 25 000 goats, if these regions represent 5% or less of the national population of the relevant animals,
• Data cells regarding optional items (Milk table B).

Validation, level 1
Validation of level 1 includes numerous rules specifying the allowed range of values for data cells and also certain rules for consistency between cells (mainly totals and sums of components).

3. Data validation - detection (Eurostat)

The first processing step is to validate the data and load them in the production system. The input to this process are the datasets received and the output are either the data loaded or validation reports and requests to the CNAs for feedback or revised data.

It is reminded to the reader that during data transmission eWF has performed a type of validation, the so-called “pre-validation” (see Data collection).

Data files are loaded and users can review the available information and the comments made by the CNAs and they can combine data from the available input files and create ad hoc data tables. Automated and non-automated validation is carried out with the tool.

Validation, level 2 - detection of outliers

The automated validation is the detection of outliers in the received data. Time series are created by retrieving data from earlier reference periods. Each current value is considered an outlier when it is outside the interval (Q1 – 1.5*(Q3-Q1), Q3 + 1.5* (Q3-Q1)), where Q1 and Q3 denote the first and third quartiles respectively of the corresponding time series. This validation is applied to all cells of the received datasets. Outliers are not necessarily wrong data.

Validation, level 1 - consistency in the data file

Eurostat may implement additional validation rules to check the consistency of the outliers with current data from the same data file. The following types of additional checks are used:

- Aggregation of items: data items that represent category totals (e.g. total bovine animals) must be equal to the sum of the items representing the different categories, allowing for certain tolerance limits.

- Aggregation of regions: data items that represent regional totals (e.g. totals at country level) must be equal to the sum of items representing the corresponding sub-regions.

- Consistency of totals and partial components: sometimes not all categories corresponding to a particular total are collected (e.g. cows’ milk part of which is obtained from dairy cows’ milk, in Table C). In these cases the data items that represent category totals must be greater than or equal to the sum of the items representing the different categories.

- Sign of the values: all variables should be non-negative. Some variables however regarding milk production in Tables B and C are allowed to receive negative values (e.g. in Table B-Milk production, the variable ‘Input of skimmed milk (USM)’ can be by definition negative in case that the process produces skimmed milk (e.g. cream processing)).

- Consistency by size class: the average size class of milk collected and volume of milk treated or milk production in dairies enterprises is computed by size category and is checked whether it falls in specified ranges.

Validation, level 3 - comparison with other data sets

Eurostat may implement additional validation rules to check the consistency with related datasets. The following types of additional checks are used:

- Correspondence between variables from different data sets: same or similar variables from different datasets are compared (e.g. the quantities of products obtained in Table B-Milk Production should be equal to the ones reported in Table H-Milk Protein Contents).

- Consistency between variables from different data sets: related variables in different tables must be consistent. For example if Quantity of utilized product in Table B-Milk Production is zero (plus tolerance limits) then the Protein content of the same product in Table H-Milk Protein content must be less than 1000 (plus tolerance).

These checks are additional to the consistency checks already implemented in eWF. Since they are not applied to all data items they are not systematic and moreover they are not automated; the user chooses on which items to apply them and which rules to apply, based on intuition.

At the end of this step, validated data are forwarded for subsequent processing and production of EU statistics.

Validation of data compiled by Eurostat, after dissemination

Eurostat presents every year to the ‘Animal Production Statistics’ Working Partyp (WP) tables which summarize the data received from the CNAs since the previous WG meeting (Figure 10 Annex 1). More specifically, the tables present statistics produced on the basis of the received. The CNAs are asked to examine the results for their country. Whereas such validation is not powerful or fast, it is the only exhaustive validation covering efficiently every collected variable. Moreover, it provides Eurostat with the CNAs’ explicit approval of the results, which is stronger than an implicit non-rejection.

4. Data validation - correction (Eurostat)

Depending on the results of validation Eurostat may decide to reject the dataset (which is quite rare) or to request feedback from the CNA postponing loading in the production system. In either case it prepares an email to the registered data provider with a report of the validation findings.  

Validation of data compiled by Eurostat, after dissemination

The CNAs are asked to examine the disseminated results for their country and either to confirm that they are correct or to provide remarks and / or revised data if they identify errors.

• Calculation of results for country aggregates (e.g. EU-28, EU-27, etc.).
• Calculation of annual totals based on monthly values
• Calculation of item aggregates
• In table apro_ec_poulm, derived series are calculated from the primary data; from these it is possible to estimate numbers of birds and production (eggs) by applying coefficients.
• Statistics on trade of chicks are drawn from Intrastat/Extrastat (Comext) for six Member States and under their responsibility
• On regional statistics, imputation of values to equivalent regions, collected only once (e.g. LU, LU0, LU00)

None.

Animal production statistics are subject to the general quality assurance framework of Eurostat, where domain-specific quality assurance activities (the use of best practices, quality reviews, self-assessments, compliance monitoring) are carried out systematically.

The latest quality review should provide findings and recommendations in 2019, for the whole process of all Eurostat animal production statistics.

In 2012, the main strengths and weaknesses of the process were the following.

Principal strengths of statistical output:

• Long history and experience;
• Legal basis;
• Data collection based on common methodology and harmonised questionnaires;
• Good knowledge of key users and their needs;
• Sound monitoring of compliance with the requirements of data collection;
• Close cooperation with data providers;
• Share of best practices on data validation;
• Good completeness of the data received;
• Validation processes in place;
• Detailed methodological information is available to the public;
• Data is normally disseminated a few hours after being received;
• Coherence analysis with other statistics is undertaken;
• The statistical process is revised and improved according to emerging needs;
• Innovative IT applications are used for data collection, validation and dissemination.

Principal weaknesses of the statistical output:

• More information on profiles and specific needs of users would be useful;
• Coordination with other international organisations could be improved with a view to reducing the burden on respondents and further improve data coherence;
• Timeliness of data delivered from certain countries could be improved;
• More information on metadata and certain validation procedures at country level is needed.

Meanwhile some work has been conducted to improve the situation, with

• Documentation of the user needs in the framework of the Strategy on agricultural statistics for 2020 and beyond and preparatory work on the future EU legislation;
• Systematic monitoring of punctuality and report to the Working Group and the the Director's Group (DGAS);
• Use of the ESS metadata-handler for metadata collection, with dissemination of the national reports (access them through the annexes of the present metadata).

Coordination with other international organisations is now managed at the level of agricultural statistics as a hole.

The main users are other Directorates General of the European Commission (e.g. DG Agriculture and rural development and DG Health). However, there are other major users such as other European institutions, national administration services, national statistical institutes, other international organisations, agro-industry, producer groups, research institutes, journalists, third countries and the public in general. The objectives of these users vary, but animal production statistics are especially useful for market management/monitoring, production forecasts and policy-making in agriculture and food.

Key users are well known and their needs are met. In addition, specific questions from individual users are answered.
Eurostat conducts regular user surveys with a wider scope.

Since 2010, completeness has been measured regularly and discussed during meetings of the Working Party on Annual Production Statistics. Meanwhile it has been regularly improved.

To reduce the burden on respondents, some statistics are collected less often in countries with limited importance in the EU-28 totals. Statistics on bovine population are due only once a year for the Member States where it is below 1.5 million head. Statistics on pig population are due only once a year for the Member States where it is below 3 million head. Statistics on sheep population are not expected from the Member States where it is below 500 000 head. Statistics on sheep population are not expected fromthe Member States where it is below 500 000 head.

Also some further data are not due and may appear to be missing, due to the ‘cube’ approach of the slection tools in the Eurostat dissemination database (each selected code of a dimension is crossed with each selected code of each other dimension).

Regional livestock population is not mandatory when it is under 75 000 bovine animals, 150 000 pigs, 100 000 sheep and 25 000 goats.

Achieved completeness, as measured by the percentage of mandatory values transmitted for reference year 2019, was the following:

• 96.1 % on average for the Member States resulting in 91.9 % for the EU totals on livestock populations;
• 96.2 % on average for the Member States resulting in 94.7 % for the EU totals on slaughtering;
• 100 % on average for the Member States resulting in 100 % for the EU totals on GIP forecast.

One should notice that the EU totals are not expected for the values mandatory for only some Member States, and that only one EU value results from many regional data. On the other hand, one missing national value out of 27 cancels one EU total.

Accuracy is normally accuracy displayed and reflects accuracy of computation. Nevertheless threshold for significance is half of the displayed unit. It means that '0.000' with flag 'n' is lower than 0.5 whereas 0.001 is lower than 0.0005. '0.000' without flag 'n' is a true zero, i.e. with full accuracy.

Regulation (EC) No 1165/2008 states that sampling errors for the results of each Member State shall not exceed (with a confidence interval of 68 %):

• 1 % of the total number of bovine animals (5 % where the bovine animal population is below 1 000 000 head);
• 1.5 % of the total number of cows (5 % where the cow population is below 500 000 head);
• 2 % of the total number of pigs (5 % where the pig population is below 1 000 000 head); and
• 2 % of the total number of sheep and goats (5 % where the sheep and goat population is below 1 000 000 head).

The Member States report to Eurostat on the actual accuracy but these national figures are not publishable.

Directive 96/16/EC states that the Member States may carry out the milk monthly surveys in the form of sample surveys. In that case, the sampling error must not exceed 1 % of the total national [milk] collection (with a confidence interval of 68 %).

No quality criterion is defined in Regulation (EC) No 617/2008.

The Member States conducting sample surveys must take all necessary steps to ensure that the extrapolated national survey results on livestock statistics meet the precision requirements set out in Annex III of the Regulation, as follows:

Sampling errors for the results of each Member State shall not exceed (with a confidence interval of 68 %):

• 1 % of the total number of bovine animals (5 % where the bovine animal population is below 1 000 000 head);
• 1.5 % of the total number of cows (5 % where the cow population is below 500 000 head);
• 2 % of the total number of pigs (5 % where the pig population is below 1 000 000 head); and
• 2 % of the total number of sheep and goats (5 % where the sheep and goat population is below 1 000 000 head).

However, Regulation (EC) 1165/2008 does not define any precision requirements for meat statistics.

Regulation (EC) 1165/2008 also requires estimates of the extend of other slaughtering in order to cover all slaughtered animals. The CPSA agreement states that the expected quality of monthly estimates cannot be higher than that of annual estimates. The burden on the respondents should not be increased by the collection of monthly estimates of other slaughtering, unless is not possible otherwise, or if such estimates are clearly not of a sufficient quality.

The quality must be such that the values to be provided must reflect the monthly changes in other slaughtering.

Meat production forecasts are not subject to precision requirements.

6.2.1.1. Coefficient of variation achieved for the main variables

In the frame of the 2019 quality reporting exercise, the Member States were requested to provide the coefficient of variation (CV) of the variables "Number of animals" for the livestock and slaughtering statistics (bovine animals, pigs, sheep, goats and poultry), where the data were drawn from a sample survey. Hence, the value reported in the cells and the comments enable one to identify the type of source of these statistics. If the value of CV was greater than 0, the data were supposed to be drawn from a sample survey. Otherwise, if the value of CV was equal to 0 or not reported, data were supposed to be drawn from another source, such as administrative source, register, or exhaustive (full) survey.

Based on the information provided in the quality reports, the majority of Member States (19/26) use a sample survey to collect data on at least one livestock category. 18 Member States used sample survey for pigs and 13 out of the 19 involved Member State for sheep, whereas the exhaustive sources were often used for bovine animals (18 Member States, of which 16 used administrative register). For goat livestock statistics, 9 Member States used sample surveys and 10 other ones used exhaustive sources (register or census).Only 7 Member States used exclusively exhaustive sources to collect data on all livestock categories. However, it should be noted that some countries consider a sample survey carried out by another institution to be an administrative source. For such a sample survey, no coefficient of variation has been provided.

On the contrary, all the Member States used sources other than sample survey to collect data on slaughtering statistics, i.e. administrative data source or exhaustive survey.

For more detailed information please see Tables 6.2.1.1.1., 6.2.1.1.2. and 6.2.1.1.1.3.

Livestock statistics

Table 6.2.1.1.1.: Coefficient of variation of the main variables (livestock statistics)

Cells in grey: Non-mandatory data.

Red values are over the legal requirements (Annex III – Regulation (EC) no 1165/2008).

*ES: 2019 quality report is not available.

Table 6.2.1.1.2.: Type of data sources used for livestock statistics (EU Member States) 

*NL: April data collection on bovine animals and pigs- use of administrative data sources (an exhaustive survey carried out by other institution.

**IE carries out a sample survey on sheep also in June, but this is not included here as in accordance with Regulation (EC) 1165/2008 only Nov/Dec data collection on sheep has been required.

***ES: 2019 quality report is not available.

Slaughtering statistics (in slaughterhouses)

Table 6.2.1.1.3.: Type of data sources used for slaughtering statistics in slaughterhouses (EU Member States)

* According to the information provided in the quality reports 2019 no Member State has used a sample survey to collect data on slaughtering in slaughterhouses. However a sample survey has been used by some Member States to collect data on slaughtering other than in slaughterhouses (see Table 3.3.2. above). Two Member States (LV and LT) have provided in the quality report 2019 the CVs of the main variables for the sample surveys organised  to collect the information on the animals slaughtered on the farm or sold for slaughter and the results of these surveys are submitted to Eurostat as other slaughtering.

**ES: 2019 quality report is not available.

SAMPLE DESIGN

Sample design refers to the way a sample is organised and to the statistical methods used to check whether the sample complies with predefined criteria (e.g. representativeness).

Usually, a simple stratified sample is used (if any) but the way the strata are defined (stratification criteria) may be complex. A sample could also be a purely random sample or be designed with several levels.

A stratified sample uses stratification criteria, usually the herd size (number of animals) and the region (large countries). Some specific strata may also be defined for a particular purpose, e.g. farms with direct sales activity, organic farms, new farms, type of farming (dairy, meat on grass, bull fattening, selection, etc.). In general, the largest farms are surveyed exhaustively because they provide information on a large proportion of the livestock and at low cost (few statistical units or mandatory administrative information, sound bookkeeping).

A multi-level sample may be based for instance on geographical units (sample of municipalities or districts) within which some or all the strata are represented.

Various sample designs are possible. However, data providers are more likely to use a census or an administrative source in the following cases:

• Number of bovine animals (bovine registers);
• Limited contribution to the national total (administrative sources);
• Small numbers of statistical units (small countries) or high diversity of units (required sampling rate will be high anyway);
• Availability of administrative sources (legal entities, sanitary reports by slaughterhouses) whose updating frequency and quality make them suitable for statistical use.

Therefore, the pig population (efficient sampling), surveys of (numerous) small units and monthly statistics are more often sampled. The stratified samples are mostly based on the size of units but the specialisation of units may also be a factor for efficient sample design. Location is a stratification criterion, except for small countries and regional surveys (federal administrative organisation). Special strata may cover the largest units, new units and/or legal units (or with other specific status) which regularly provide mandatory reports on their activity.

Finally additional sample surveys may be combined with administrative information to cover:

• Units beyond the scope of administrative sources;
• Detailed variables not recorded by administrative sources;
• Monthly figures to be combined with annual administrative information;
• Specific information at farm level collected through a specific survey, e.g. on farm production.

The highest sampling rates reflect a combination of sampling and an exhaustive source. In contrast, really low sampling rates reflect considerable numbers of small units.

With only few exceptions, the samples are stratified in strata which the sampled units seek to represent. Of the few surveys reported as non-stratified, most turn out to be composite data collections from different sources. Few samples are stratified afterwards or are multi-level samples.

In pig and poultry statistics, exhaustive strata often include more than 90 % of animals in just a few per cent of the frame units. In rare cases, exhaustive strata are at the margin of the frame, i.e. they represent units beyond the scope of a main source.

In 2019, the Member States that used stratified sampling applied most often the stratification criteria of location and size of unit.

6.2.1.2. Sampling rate

The sampling rate is the ratio between the size of the sample and the size of the statistical frame, expressed as a percentage. It is a simple indicator for describing a sampling scheme because it is easy to calculate. It also gives an indication of the cost of a survey but provides less information on the expected statistical efficiency than the coefficient of variation.

Depending on the way the survey is organised, sampling may refer to farms with the various species or to the whole frame of farms with animals (Table 6.2.1.2.).

Sampling rates provided in the 2019 national quality reports ranged from 1.34 % to 17.7 % for bovine farms, from 0.58 % to 61.0 % for pig farms, from 1.56 % to 43.95 %.for sheep farms and from 1.21 % to 20.4 % for goat farms. For the animal farms, where the survey was designed for all livestock together, the sampling rates ranged from 2.35 % to 15.0 %. The sampling rate was not relevant for slaughterhouses as their data were collected everywhere from exhaustive surveys or administrative data sources. Seven Member States did not use sample surveys to collect data on livestock and thus the sampling rate was irrelevant for them.

Table 6.2.1.2.: Indicative sampling rate (%)

*ES: 2019 quality report is not available.

Measurement errors are due mainly to lack of harmonisation in statistical methods. For instance, when EU concepts do not fit with national concepts, there may be significant measurement errors. As well over-coverage (surveying units out of the statistical population) may be due to the same, when the harmonised definition of the statistical unit is not implemented. Data providers are expected to correct these deviations.

See item 6.3.1.1.

Coverage error occurs due to divergences between the target population and the frame population. A good sampling frame covers all the units in the target population, excludes all units not in the target population and has accurate information on the unit (e.g. information allowing contacting the unit). Ideally each unit should have a unique identifier.

Coverage error may be deliberate (usually on cost grounds) or identified afterwards. Overcoverage in livestock statistics is due especially to farms which ceased their activities and no longer have livestock. Moreover, non-agricultural animal owners are outside the statistical frame but this is not necessarily known before information is collected. It should be noted that the coverage error is not specific to surveys.

In the quality reports 2019, the countries provided relevant information on the register or other frame source to assist in understanding coverage errors and their effects. This is summarised below.

• Regarding the geographical coverage - only three Member States have indicated some exceptions in the quality report 2019, in FR the overseas departments are included in the survey with the exception of Mayotte and in EL Mount Athos is excluded, and DE (has not specified the indicated exception yet). Geographical coverage is sensitive, because it has a direct impact on comparability.
• Concerning the other possible sources of coverage error, most of the Member States indicated for data collection on livestock that the small farms, as well as farms that have just started/ceased their activities were covered. Around half of the Member States declared that they cover also empty farms/buildings in the data collection on livestock.
• Similarly, a majority of the Member States which indicated the source of coverage error, referred to slaughterhouses also including the small ones or those having just started or ceased the activity. Approximately half of Member States (13) declared that they cover emergency slaughtering.
• The provided rate of over-coverage varied, for those countries that provided the required information, from 0 to 20 %.

Not applicable.

Measurement error results from deviation in the accuracy of measurement during data collection. For the surveys, it covers both incorrect recording of an accurate response and correct recording of an inaccurate response.

Deviation from standard definitions

When EU concepts do not fit with national concept and thus data are not recorded in line with the EU definitions, there may be significant measurement errors. However, the data providers are expected to correct these deviations using for example the technical coefficients. In 2019, the Member States were asked to report on the Checklist for possible measurement errors in data collection on slaughtering. In particular, the questions related to possible deviation from variables covered and from standard definitions of carcass weight of animals, including poultry. The definition of poultry carcass weight refers to the animal plucked and drawn, without head and feet and without neck, heart, liver and gizzard, known as ‘65 % chicken’, or otherwise presented. The latter wording allows some deviations that should be reported. The deviations reported in the 2019 are provided in the table below.

Table 6.3.2.1.: Measurement error - Slaughtering

Survey questionnaire

The main tool used for statistical measurement is the questionnaire, either an actual questionnaire filled in by a surveyor or a form completed directly by the respondent. The key to limiting errors at this stage is to ensure the questionnaire is clear, with the items well explained and well understood. Because field work cannot be easily repeated, possible weaknesses are assumed to be remedied with experience. Therefore, the age of the questionnaire and the experience of the surveyors are factors limiting this error.

In 2019, the Member States provided information onsurvey questionnairefor each identified specific process.

Experience in using the same questionnaire improves comparability over time but revisions are needed to adapt the questionnaire to changing requirements.

In 2019, the number of surveys already performed with the same questionnaire, as reported by the Member States, confirmed that the stable requirements of Regulation (EC) No 1165/2008 since 2009, preserved most of the questionnaires from recent significant revisions.

The surveyor experience was not seen as a quality factor or was not relevant, the data being collected on-line (without surveyors) or drawn from administrative data source and/or obtained from other entities (e.g. other authority, regional administration, etc.). Most of the Member States replied that it is not known whether the surveyors had already performed the survey or that this was not relevant. In counterpart, most of the Member States answers indicated that a hot-line support or an on-line FAQ, and even often both, were accessible to the surveyors/respondents. Furthermore, in most Member States, the questionnaire are based on usual concepts for the respondents and the results are cross-checked. Really few Member States answered for instance on field testing of the new questionnaires and , therefore, no conclusion could be drawn. Training new surveyors, where known, takes on average usually no more than one day. 

Non-response is a failure in data collection. The difference between the statistics calculated from the data actually collected and those that would be calculated obtained if all values were available is the non-response error.

Unit non-response

Unit non-response occurs when no data are collected at all about a given statistical unit intended to data collection. It can be managed either by providing substitution values (imputation) or by correcting the weight of units in the strata (re-calibration). For detailed information on unit non-response rate, see 6.3.3.1 below.

Item non-response

Item non-response occurs when data about a given statistical unit are collected on only some, but not all, of the variables. It may also be corrected by imputation, especially when other valuable information has been collected for the relevant unit. If the non-responses are randomly distributed, the statistics remain nevertheless representative. Otherwise, further corrections are required. For detailed information on item non-response rate, see 6.3.3.2 below.

Unit non-response analysis

Unit non-responses have been analysed for at least one 2019 data collection in more than half of the Member States. These countries indicated that the risk of bias due to non-response has been assessed as insignificant or proved null, but in some cases it was unknown. No assessment was conducted in some cases, where the response rate was 100 %, but also in some other cases. Nevertheless, the non-response rate and the way to correct non-response errors should draw attention everywhere.

Imputation procedure

Imputation is explicitly connected with the treatment of non-response. Imputation has been used by majority of Member. The most usual imputation method consists in using previously collected data for the same statistical unit. A small number of countries use donor imputation or take data from other surveys or registers.

In 2019, the response rate was higher than 95 % in a majority of Member States, or even close to 100 % for livestock data collection and, in almost all cases, it reached 100 % for the data collection from slaughterhouses (based on census or administrative sources).

Unit non-responses should be resolved by imputation where this is more efficient than re-calibration (recent data, other source, strata with few units, etc.). Approximately half of the responding Member States used imputation while the other half use ‘other methods‘. Seven Member States used re-calibration to resolve unit non-response.

Known item non-response rates are quite rare. Nevertheless, those Member States which provided a response, indicated that item response rate has been either 100 % or close to it. Incomplete questionnaires have been treated mainly by imputation where relevant.

Processing errors occur during data entry, data editing, coding, imputation and transmission.

Internal processing errors

Internal processing errors reflect weaknesses in process organisation. They cannot easily be measured and, when detected, they are usually directly corrected and the process is amended to avoid any repetition. Changing IT systems has been identified as a representative factor in processing errors.

No change in the IT system has been reported in 2019 and, therefore, the possible impact of IT changes on statistical results was not considered.

Transmission processing errors

Transmission processing errors impact data where responsibility is shared and correction requires coordination between sender and receiver. Eurostat focuses particularly on transmission between the Member States and the Commission.

Data transmission to Eurostat may be centralised in a specialised service or managed in the statistical production service. In the first case, there is internal data transmission between national services (with a risk of data transmission error) and in the second case a risk of insufficient knowhow in terms of transmission procedures and tools.

According to the quality reports 2019, data transmission has been in most Member States conducted by the department in charge of animal production statistics, especially where livestock statistics and meat statistics are produced in the same service. However, in a few other cases the data were centralised by the NSI when some or even all statistics are produced elsewhere. Such centralised data transmission can be efficiently coordinated with other national activities through the ESS.

Control procedure

Control procedures guarantee a level of quality by detecting errors. Conducted at the relevant stage, their automation avoids slowing down the process excessively.

In 2019, for livestock statistics like for meat statistics, at most, the data have been checked at four stages before transmission to Eurostat.

In more than half of the Member States, the data on livestock and meat have been checked interactively by the surveyor or the respondent upon data entry. Where the collected data have been cross validated, it was in most cases against previous results, because they are the main source of available and comparable individual information. Sometimes other sources have been used in addition to these previous results.

The imputation rate was not requested from the Member States for reference year 2019.

Model assumption error was not described in the 2019 quality reports.

Seasonal adjustment is not applicable to the livestock and meat statistics.

Data are revised if figures of higher quality become available.

In order to provide fresh statistics, Regulation (EC) No 1165/2008 foresees that provisional livestock statistics are delivered first (in February and September) and definitive results are provided later (in May and October).

Results can be corrected by the countries concerned during the months after data has been sent.

For annual milk statistics, after a first validation round, the values which does not return error are published as provisional. Advanced validation is then run, leading either to confirmation or to revision of these values.

Revisions of historical data or of long time series are exceptional and the most recent have been documented (see 15.2. Comparability over time).

See item 6.6 

The transmission deadlines for the various animal production statistical tables are set out in Regulation (EC)1165/2008, given relative to the end of the reference period (see Table 7.1 below).

The actual timeliness (length of time between the event and availability of the statistical output) can be shorter than the legal timeliness if data are provided earlier. The time lag between the actual release date and the planned (agreed or legal) date is called punctuality.

The actual timeliness for EU-28 results depends on timeliness achieved among Member States. The time taken for data validation and dissemination by Eurostat is also taken into account.

Table 7.1.: Transmission deadlines for livestock and meat statistics

Table 3.1. (see 3.1. data description) shows the deadlines for the various animal production statistical tables, given relative to the end of the reference period.
The actual timeliness (length of time between the event and availability of the statistical output) can be shorter than the legal timeliness if data are provided earlier. The time lag between the actual release date and the planned (agreed or legal) date is called punctuality.
The actual timeliness for EU-28 results depends on timeliness achieved among Member States. Time taken for data validation and dissemination by Eurostat is also taken into account.

This indicator represents the number of days (or weeks or months) from the last day of the reference period to the day of publication of first results.

In 2019, 24 out of 26 reporting Member States (ES has not provided the 2019 quality report) provided this indicator for the livestock statistics and 20 Member States for the meat statistics. This covers, altogether, 63 statistical processes on livestock and 48 statistical processes on meat statistics (slaughtering). The number of processes for each Member State as well as the data source for each process differs from Member State to Member State, with from one process on livestock in some of them to four processes in another one. For livestock statistics, timeliness in days from a Member State point of view varies between 30 and 106 days depending, especially, on the data source type (sample survey, register, structural surveys, etc.). Regarding meat statistics, the time lag on provision of preliminary results varies between 22 and 112 days.

Table 7.1.1.1.: Timeliness - first/preliminary results

The definitive results are published later and all the Member States except ES reported the corresponding timeliness. The minimum timeliness is higher than the average time lag of preliminary results The range of time lags is wider, (from 20 to 230 days for livestock statistics and 25 to 180 days for meat statistics) because the definitive results published directly (without preliminary publication) are the earliest and the others may be published later as an improvement to the already available (preliminary) results.

Table 7.1.2.1.: Timeliness - final results

Data are normally received within the legal deadlines. However, some countries may experience delays in sending their data to Eurostat due to exceptional circumstances.

The below attached document on punctuality intends to summarise punctuality of the Member States without disclosing the bilateral issues between Eurostat and the Member States. Punctuality of the EU results is, for each value, punctuality of the latest Member State. 

Data are normally received within the legal deadlines. However, some countries may experience delays in sending their data to Eurostat due to exceptional circumstances.

The table below on punctuality for the 2019 data reflects the Eurostat point of view and intends to summarise punctuality of the Member States without disclosing the bilateral issues between Eurostat and them. Punctuality of the EU results is, for each data set, punctuality of the latest Member State.

Table 7.2.1.: Punctuality of the 2019 data

From a Member state perspective, the statistical results are usually available to the national users’ in the Member States for livestock and meat statistics at about the same time as transmitted to Eurostat.

The table below gives an overview of punctuality for all the processes described in the national quality reports.

Table 7.2.1.1.:Availability of data to national users

In general, animal production statistics are mostly comparable over countries and regions with the following exceptions:

• The legislation itself provides for differentiated approaches for countries depending on their livestock population;
• Specific agreements with EEA countries and with the Swiss Confederation envisage limited application of the legislation and exceptions to the definitions.

Other non-EU countries may partly implement the EU legislation and some concepts may be irrelevant for them.

Comparability of regional data over time has been and will be affected by changes in the NUTS classification.

Animal production statistics are mostly comparable over countries and regions with the following exceptions:

• The ESS and CPSA agreements may be granted with only some countries
• A derogation to EU statistical legislation may have been granted
• The legislation itself provides for differentiated approaches for countries depending on their livestock population (see 12.3 Completeness)
• Specific agreements with EEA countries and with the Swiss Confederation envisage limit application of the legislation and exceptions to the definitions

Other non-EU countries may partly implement EU legislation and some concepts may be irrelevant for them. Quality of the statistics is usually not at the same level as for statistics under EU legislation.

Comparability of regional data over time has been and will be affected by changes in the NUTS classification.

This concept is not applicable to the livestock and meat statistics.

Animal production statistics are largely comparable over time, with the following exceptions:

• Implementation of Regulation (EC) No 1165/2008 (see ASA-TE-696.Inventory_of_changes.rev1.doc )
• Change in NUTS for regional statistics
• Other changes in the methodology flagged with "b", indicating breaks in the time series

Major revision may downgrade comparability over time or, when they intend to be corrective, improve it (below attached document)

Length of comparable time series was not requested in 2019 quality reports.

Most livestock categories are comparable with certain characteristics from the farm structure survey, but differences in the design of each survey (especially reference date and calibration method) may produce discrepancies in estimates.

Milk products are comparable with the Prodcom products, but the recorded flows are different (added value for Prodcom, black box for milk statistics).

Not applicable.

Not applicable.

Internal coherence is insured through pre-validation checks during the data collection and validation on the data received in Eurostat.

Accessibility and clarity represent the conditions and modalities by which users can obtain, use and interpret data.

News releases on-line.

Publications

Publications available in pdf format are accessible on Eurostat's website (Statistics / Agriculture and fisheries / Agriculture / Publications). Eurostat has published in 2019 the Eurostat regional yearbook 2019.

The Directorate General on Agriculture and Rural Development (DG-AGRI) publishes regularly a Short-term outlook.

Statistics Explained

Statistics Explained is an official Eurostat website presenting many statistical topics in an easily understandable way. Together, the articles make up an encyclopaedia of European statistics, completed by a statistical glossary clarifying all terms used and by numerous links to further information and the very latest data and metadata, a portal for occasional and regular users alike.

Animal production articles:

• Milk and milk product statistics
• Agricultural production - livestock and meat

Articles referring to animal production statistics

• Agriculture statistics atregional level: Livestock and milk

Re-dissemination

The Eurostat statistics are widely re-disseminated. For improving the market transparency, the EU Milk Market Observatory provides for instance statistics on milk production and the EU Meat Market Observatory statistics on meat production or livestock.

The data on animal production statistics and statistics for regional production may be found on the Eurostat’s dissemination database under the link: https://ec.europa.eu/eurostat/data/database and under the following paths: Database / (+) Data navigation tree / (+) Database by themes / (+) Agriculture, forestry and fisheries / (+) Agriculture (agr) / (+) Agricultural production (apro) / (+) Animal production (apro_anip) / (+) Poultry farming (apro_ec), or (+) Milk and milk products (apro_mk), or (+) Livestock and meat (apro_mt).

This channel displays all the public Animal Production results.

Public access to data 

The data on animal production statistics and statistics for regional production are available on the Eurostat’s dissemination database under the link: https://ec.europa.eu/eurostat/data/database and under the following paths: Database / (+) Data navigation tree / (+) Database by themes / (+) Agriculture, forestry and fisheries / (+) Agriculture (agr) / (+) Agricultural production (apro) / (+) Animal production (apro_anip) /  (+) Livestock and meat (apro_mt).

In 2019, a public on-line database provided the livestock statistics for all Member States and the slaughtering statistics for almost all of them. However, the level of detail published online depends on the Member State and type of statistics (Table 9.3.1.1.). Only 5 Member States reported that at least some data were not accessible on-line to the public. In four of these cases these were slaughtering other than in slaughterhouses.

Table 9.3.1.1.: Public access to data

Not applicable.

Confidentiality is especially an issue for milk statistics due to the dominant contribution of the largest enterprises in the national totals, and especially for some products (e.g. dairy powders). Further investigations are currently conducted in order to improve the dissemination of the EU results without threatening statistical confidentiality.

Public documentation on methodology and legislation is available on-line.

Member States provide Eurostat with a report on the quality of the data transmitted every third year under the current legislation on livestock and meat, and for the first time for reference year 2010.

Member States send Eurostat an annual methodological questionnaire on milk statistics providing some quality indicators.

This information may be disseminated to the public once this is agreed with Member States.

The specific EU metadata files on livestock and meat statistics and on statistics of milk and milk products (see the links at the bottom of the present file) summarise the national quality reports and provide the direct links to them.

Not applicable.

In 2019, the Member States should check whether certain metadata were available as a paper document, as electronic information, or not available. Table 9.7.2.1. below displays detailed illustration of the replies.

 Table 9.7.2.1.: Available metadata

Permanent efforts are done to limit the burden on the respondents and, if relevant on the competent national authorities.

• The design of Regulation (EC) No 1165/2008 makes optional the delivery of some data sets for the countries under a certain threshold. This approach by differentiate intensity of burden relatively to the policy interest is favoured in the new data collection (gentlemen's agreements). This might make difficult using the tables with national values, due to the gaps.
• Use of administrative sources is favoured (livestock statistics) when it is not the prefered data source (slaughtering).
• Involvement of the respondents as data users (milk statistics) contributes to a balanced burden.
• The supply balance sheets, requested previously at national level are now compiled , with few exceptions, at EU level
• The Strategy on Agricultural Statistics for 2020 and beyond provides an opportunity to review the statistical needs.

See 11.1. and 11.2.

Regulation (EC) No 223/2009 on European statistics (recital 24 and Article 20(4)) of 11 March 2009 (OJ L 87, p. 164), stipulates the need to establish common principles and guidelines ensuring the confidentiality of data used for the production of European statistics and the access to those confidential data with due account for technical developments and the requirements of users in a democratic society.

A confidentiality charter for animal production statistics has been adopted in 2016 (below) and should lead to disseminate more EU totals from 2020 on.

Regulation (EC) No 223/2009 on European statistics (recital 24 and Article 20(4)) of 11 March 2009 (OJ L 87, p. 164), stipulates the need to establish common principles and guidelines ensuring the confidentiality of data used for the production of European statistics and the access to those confidential data with due account for technical developments and the requirements of users in a democratic society.

A confidentiality flag is 'computed' for each aggregate. By default, if one or more components are confidential the aggregate is also flagged as confidential. Implementation of the confidentiality charter is based on the set of methods agreed and refer to the number of cells, the possible dominance of some of them and the logical links between the values with the same table or in different tables.

The current description covers decades of methodology on statistics. It reflects especially the current design of animal production statistics. Usability of the data promotes continuity of the time series, which may downgrade interpretability of the statistics, especially considering the permanent improvement in the statistical production process. Eurostat expects having reached a satisfying trade-off. The data user is nevertheless invited referring to the relevant legislation applicable at the reference date for more accurate analysis involving long time series.