COMPARABILITY OF RESULTS, BREAKS IN SERIES AND COHERENCE WITH OTHER STATISTICS

IMPORTANT REMARKS ON THE EU TOTALS:

Last update: 02/10/2009

From 2005 data, EU totals are based on the available country data for a given quarter.

Until 2004 data, EU totals are located in quarter 2.

(based on quarter 2 for most of countries, on quarter 1 data for Austria and France, quarter 4 for IT in 1992 and quarter 1 for PL in 1999).

Between 2003 and 2006, the survey in Luxembourg provides data for the whole reference year only. In absence of quarterly results, the same yearly figures are repeated each quarter.

1. Comparability of results between countries

Perfect comparability among countries is difficult to achieve, even were it to be by means of a single direct survey, i.e. a survey carried out at the same time, using the same questionnaire and a single method of recording. Nevertheless, the degree of comparability of the EU Labour Force Survey results is considerably higher than that of any other existing set of statistics on employment or unemployment available for Member States. This is due to:

(a) the recording of the same set of characteristics in each country;

(b) a close correspondence between the EU list of questions and the national questionnaires;

(c) the use of the same definitions for all countries;

(d) the use of common classifications (e.g. NACE for economic activity);

(e) the data being centrally processed by Eurostat.

The EU Labour Force Survey, although subject to the constraints of the EU's statistical requirements, is a joint effort by Member States to coordinate their national employment surveys, which must serve their own national requirements. Therefore, in spite of the close coordination between the national statistical institutes and Eurostat, there inevitably remain some differences in the survey from country to country. This is especially the case for:

- Population coverage (restricted to age: 16-74 in IS; 15-74 in NO*; CH, 15 and over in PL before 2006, before 1999 in EE, before 2002 in LT and before 2001 in BG; 15-74 in SE (1995)),

- Labour status coverage (restricted to age: 16 and over in ES, IT (from 2008 onwards) and UK; 16-74 in IS and SE (1995 - 2000); 15-74 in DK, EE, LV, HU, FI, NO^* and SE (from 2001 onwards). In 1999, the division between unemployment and inactivity was not possible for Cyprus),

- Main economic activity definition (BG: number of persons whose main activity is in the agriculture sector lacks reliability. Due to the very high proportion of persons having agricultural activity in addition to another main occupation the LFS does not provide a precise estimate of total employment in this sector),

- Participation in education and training (see the notes for the indicator ‘Life-long learning’),

- Highest educational attainment level (see the notes for the indicator ‘Youth education attainment level’) ,

but also due to the transition to a quarterly continuous survey as explained below.

Notes:
* NO: before 2006, only people who completed 16 years at the end of the year where included

2. Comparability of results between successive surveys

(N.B. this section applies only to LFS series – Detailed survey results since the LFS Adjusted Series includes corrections for this type of breaks)

Since 1983, improved comparability between results of successive surveys has been achieved, mainly due to the greater stability of content and the higher frequency of surveys. However, the following factors may somewhat detract from perfect comparability:

(a) the population figures used for the population adjustment are revised at intervals on the basis of new population censuses;

(b) the reference period may not remain the same for a given country due to the transition to a quarterly continuous survey;

(c) in order to improve the quality of results, some countries may change the content or order of their questionnaire;

(d) countries may modify their sample designs;

(e) the manner in which certain questions are answered may be influenced by the political or social circumstances at the time of interview.

The main factors affecting the comparability of the data for successive surveys are detailed in the table below.

From 1983 to 1997, the EU Labour Force Survey was conducted only in spring (quarter 1 or 2 depending on the country). The data for remaining quarters started to become progressively available from 1998 onwards. Since 1998, the transition to a continuous quarterly survey (where the reference weeks are spread uniformly throughout the year) has been gradually conducted by Member States. Some countries first introduced a continuous annual survey (meaning the reference weeks were uniformly distributed throughout the reference spring quarter) and then switched to a quarterly collection, whereas the others moved directly to a continuous quarterly survey.

Regulation 1991/2002 of the European Parliament and of the Council introduced a deadline for the period of transition given to the Member States to introduce a continuous quarterly survey. In 2003, all countries conducted a quarterly continuous survey except IT, CY and AT (starting in 2004) and DE (starting in 2005). For this reason, EU results are still provided on the basis of spring results (quarter 2 for all countries except France and Austria for which quarter 1 is used) [1]. Croatia started with a quarterly continuous survey in 2007 and Switzerland plans to start with a quarterly survey in 2010.

MAIN BREAKS IN SERIES
Country	Due to transition to a quarterly continuous survey	Due to census revisions and implementation of new concepts
BE	1999
BG		2001: unemployment data lack comparability with previous years due to changes in certain survey characteristics (questionnaire and sample design)
DK		2007: new survey structure (including significant increase of sample size) with impact e.g. on education data
DE	2005
EL		2004: Re-design of the survey. 1998: Post-stratification taking into account the changes in regional, age and sex distribution based on the 2001 census was implemented.
ES		2005: significant changes in the questionnaire with impact on employment and unemployment data 2001: revised unemployment definition
FR	2003	2003: revised unemployment definition
IT	2004
LT	2002
LU	2003	2009: re-organisation of fieldwork with impact on survey results
AT	2004
PL	2000
PT		1998: re-design of the survey
RO		2002: results lack comparability with previous years due to significant changes in employment and unemployment definitions
FI	2000
SE	2001	2005: revised unemployment definition
UK		1999Q2: break due to census revisions
IS	2003

For the complete listing of the revisions since 2005, please follow this link LFS_data_revisions.

Please consult the annexes for the more detailed information on EU aggregates and breaks in series

Annex 1. Breaks due to the transition to the continuous survey

Annex 2. Breaks due to census revision and implementation of new concepts

3. Coherence of LFS with employment from other statistics

LFS and National Accounts are the two main sources of employment data. Users regularly compare them and ask about their differences. These sources are not independent; indeed LFS is frequently an input to National Accounts employment estimates. Although using common ILO-based definitions, LFS and National Accounts have their own aims and measurement approaches, which may lead to different results. In addition, other statistics based on business surveys also provide estimates of employment which may differ.

The LFS is a sample survey of households. A fraction of the population living in private households is interviewed and the results grossed up to country totals. The survey focuses on employment and unemployment, but some other 100 variables are also collected, including: gender, age, educational level attained, features of each job held, occupation, hours worked etc. This wealth of information makes the LFS unique in providing social breakdowns and interrelations of employment/unemployment with other variables, mostly demographic and social. LFS is also used for non-labour market studies, for instance demographics, household composition, education, etc.

The LFS is an EU harmonised survey, meaning that the concepts, definitions and variables are harmonised across the EU. LFS is one of the most important household surveys in the EU. It has good quality and provides reliable results. However, like in any survey of households, the identification of economic activity lacks the quality that employers report (e.g. in business surveys or employment registers). Being a sample survey, LFS is also exposed to the risk of response bias that is more difficult to identify and correct than in National Accounts.

Business surveys, like structural business statistics (SBS) or short-term business statistics (STS), are focused on production-related variables like output, turnover or value added, but they also produce some estimates of employment. These estimates may and frequently are different from LFS. There are two broad groups of reasons:

• Different scope and different units: business surveys gather information on production units operating in the territory whereas LFS gathers information on people living in the country. Cross-border workers, or seasonal workers, are correspondingly recorded in different countries. LFS does not cover people living in collective households. Business surveys typically do not gather information on certain economic activities, like agriculture or some services. Business surveys estimate number of jobs whereas LFS counts jobholders. Business surveys rarely have access to jobholders' features like age, gender, etc. for which LFS is the only source.

• Different measurement strengths: business surveys are based on a business register that may not include small enterprises below a certain threshold. Update and comprehensive coverage in the register of small production units and self-employed persons might be an issue. As business surveys inquire employment simultaneously to other variables like turnover or profits, they are more exposed to underreporting of employment than household surveys. In addition, employment not included in the payroll or in the accounting books, like trainees or family aids, could be left out. On their side, household surveys like LFS do not suffer from those downsides. LFS is more adequate to measure the total employment levels. However the identification of economic activity (industry) is much better from business surveys than from household surveys, because the information is directly obtained from the production units. As comparisons between business surveys and LFS can only be done at industry level (because some activities are out of business surveys scope, as explained above), weaknesses in both sides undermine comparisons.

National Accounts is a conceptual framework comprising definitions, classifications, variables and presentational arrangements. National Accounts are compiled by comparing and combining all the relevant data sources available in the country. This is a key feature of National Accounts: it allows taking the best from each source, increasing coherence and obtaining a more comprehensive result. For the variable employment, this means more robust estimates and improved consistency with other key national accounts variables like salaries and output. The National Accounts integration is however done at macro level, meaning that the results are produced for the whole economy plus a few standard industry breakdowns. Certain breakdowns like gender and age are not available from National Accounts. The macro-level adjustments and the absence of certain breakdowns do not make it possible to cross National Accounts employment with other variables in the way LFS allows.

The integration of data sources into the National Accounts is done differently in each country. In general, the LFS is the most important single source used for National Accounts employment. Other sources are business surveys, employment registers, social security registers, population census, etc. Some countries use LFS as the only source for National Accounts employment, many others complement the LFS with other sources and a few countries do not use LFS. Whenever LFS is used in National Accounts, some scope alignments are needed prior to any integration, the main ones being:

• Different geographical coverage. ESA95 acknowledges two employment concepts: resident persons employed (i.e. the so-called national employment concept) and employment in resident production units irrespective of the place of residence of the employee (i.e. domestic concept). The domestic concept is more frequently used as it allows putting employment in relation to GDP. On its side, LFS gives information on the national concept (i.e. resident workers). Adjustments for cross-border workers are needed to transform one concept into the other.

• Other coverage issues: LFS excludes persons below 16 years old from the definition of employment. On its side, National Accounts do not exclude individuals from employment because of age. LFS leaves out of scope the following: persons living in institutional households, staff of national embassies working abroad and crews of national fishing boats.

• Other small borderline differences (in some countries and circumstances): regarding recording of conscripts, unpaid apprentices and trainees, work in agriculture for own-consumption, etc.

Those scope alignments plus the integration of LFS with other sources (in countries where done) leads National Accounts employment to be different from LFS. The revision policies of the two statistics may also give prominence to the differences. All in all, National Accounts is judged more suitable to measure employment levels, employment growth and industry breakdowns. LFS is more adequate to measure participation in the labour market (i.e. employment rates, activity rates, flows between employment and unemployment, etc.), demographic or social breakdowns (e.g. by age, gender or educational level) and it is more suitable for socio-demographic studies.

Annex 1. Breaks due to the transition to the continuous quarterly survey^*

Country	Spring	Quarterly	Continuous survey from	Remarks
	results from
BE	1983	1999	1999
BG	-	2000	(2003)	2000-2002: One week per quarter 2003+: Uniformly spread over the first 12 weeks of each quarter
CZ	-	1998	1997	1997: Seasonal quarters 2 and 4
DK	1983	1999	1994	1992-1993: More than one week spread unequally over 1st and 2nd quarter
DE	1983	2005	2005	1983-2004: One week per quarter 2005+: Quarterly continuous survey
EE	1997	2000	2000	1997-1999: All weeks in 2nd quarter not uniformly spread
IE	1983	1999q2	1998	1992-1997: More than one week, but not uniformly spread in one quarter 1998-2008: Seasonal quarters
EL	1983	1998	1996	1992-1995: All weeks in 2nd quarter not uniformly spread
ES	1986	1996	1999	1996-1998: Evenly spread with the exception of 4 weeks in August (not surveyed due to interviewers' holidays)
FR	1983	2003	2003	1992-2002: More than one week, but not uniformly spread in one quarter
IT	1983	1997q2	2004	1983-2003: One week per quarter
CY	1999	2004q2	1999
LV	1998	2002	2002	1998-2001: All weeks in 2nd and 4th quarter not uniformly spread (semi-annual results)
LT	1998	2002	2002q3	1998-2001: One week in 2nd and 4th quarter each (semi-annual results) 2002q1-q2: One week per quarter
LU	1983	2007	(2003)	1983-2002: One week per quarter 2003-2006: All weeks of the year, but not uniformly spread and no quarterly results
HU	1996	1999	2003	1999-2002: One week per month 2003-2005: 3 weeks per month not uniformly spread 2006+: All weeks surveyed, nearly uniformly spread
MT	2000	2002	2004
NL	1987	2000	2000	1992-1999: 1st to 22nd/23rd week surveyed, not uniformly spread
AT	1995	1999	2004	1995-2003: More than one week at the end of the quarter, not uniformly spread
PL	1997	2000	2000
PT	1986	1996q2	1998	1996-1997: More than one week per quarter, not uniformly spread
RO	1997	1999	(1998)	1998+: Most or all weeks of the quarter, not uniformly spread
SI	1996	1999	(2002)	1996-2001: One week per quarter 2002-2005: All or most weeks surveyed, not uniformly spread 2006+: All weeks surveyed, nearly uniformly spread
SK	-	1998	1998	1998-1999: Seasonal quarters
FI	1995	1998	2000	1995-1999: One week per month 2000+: Monthly survey. Uniformly spread over the weeks of the month, months of each quarter have 4-4-5 weeks
SE	1995	2001	1999	1995-1998: Uniformly spread over 4 weeks of one month
UK	1983	1999q2	1992	1992-2006: Seasonal quarters
HR	2002	2007	2007	2002-2006: Half-year results, one reference week per month uniformly spread over the months
MK	-	2006	2006
TR	-	2006	-	Only one week per month covered
IS	1995	2003	2003	1995-2002: One week per quarter
NO	1995	2000	1996
CH	1996	(2010)	(1995)	1995+: All or most weeks surveyed, not uniformly spread. 2010+: Quarterly, continuous planned

Annex 2. Breaks due to census revisions and implementation of new concepts ^*

Additionally, the publication of the final results from the recent population census round makes it possible to revise LFS series. This implies breaks with past series in the following years/quarters:

Country	Breaks due to census revisions	Further planned census revisions	Other breaks
BG	2001		2001: unemployment data lack comparability with previous years due to changes in certain survey characteristics (questionnaire and sample design)
DK			2007: new survey structure (including significant increase of sample size) with impact e.g. on education data
IE	1998
EL	1998		2004: re-design of the survey 1998: Post-stratification taking into account the changes in regional, age and sex distribution based on the 2001 census was implemented
ES			2005: significant changes in the questionnaire with impact on employment and unemployment data 2001: revised unemployment definition
FR	2003	2000-2002	2003: revised unemployment definition
IT	1992
CY	1999
LV	1998
LT	1998
LU			2009: re-organisation of fieldwork with impact on survey results
HU	2001	1998 – 2000
NL	2003
AT	2003
PT	1998		1998: re-design of the survey
RO	2003	2002	2002: results lack comparability with previous years due to significant changes in employment and unemployment definitions
SE			2005: revised unemployment definition
UK	1999Q2	1992 – 1998
NO			2006: some definitions and parts of the questionnaire were changed. These led to some lower estimates for unemployment (- 3 000) and some higher estimates for employment (+ 9 000) in the 1st quarter. A break in series also occurred for the estimates on actual working hours per week.
CH	2002

PL and MT revised unemployment definition in 2004, HR plans to revise it in the near future.

^* The main breaks are highlighted in bold font

---

⁽¹⁾ Complete quarterly series as well as corresponding annual averages are available for the main indicators in the domain Employment and Unemployment in which missing LFS quarters are estimated to present EU aggregates.