Labour Force Survey (LFS)
SUBJECTS ON THIS PAGE:
LFS is the main government survey for analysis of the workforce, in terms of both the jobs people do and the characteristics of the people themselves. So, for example, major uses include analyses by work status, qualifications and training received.
LFS also includes data about all the adults in a household and can therefore be used to analyse the overall work status of households (e.g. numbers of children who are in workless households).
Although LFS includes extensive data on earnings, it should only be used to analyse low pay when the desired analyses are not available from the Annual Survey of Hours and Earnings.
- Available from: UK data archive.
- Registration required: yes.
- First survey available: 1975.
- Frequency: quarterly.
- Updated: Feb, May, Aug, Nov.
- Scope: UK-wide.
- Format: SPSS, STATA or TAB.
- Files: a single individual-level file per quarter.
- Documentation: comprehensive.
- Weighted or unweighted: weighted.
- Household income data: no.
The LFS datasets comes in two flavours:
- An individual dataset, published quarterly.
- A household dataset, published six-monthly. This dataset is the same as the individual dataset for the relevant quarter but includes additional variables relating to the combined economic status of the adults in the household.
From 2005 (but not before), each individual LFS dataset comes in two flavours:
- With specific ages. This (special licence) version has an approval process which is granted at the behest of ONS based on various criteria and which, even if granted, requires the user to make certain undertakings.
- With age bands only.
For most purposes, the version with the age bands only will be sufficient, the obvious exception being those where a specific age (e.g. 19-year-olds) is required.
In addition to the dataset itself, ONS publishes various statistics from LFS on its website, for example, various time trends.
Which software to use
Any individual LFS dataset comprises around 120,000 records. As such, the summations should usually be done using SPSS or equivalent rather than by exporting the base data into Excel.
When to use the individual and household datasets
Most analyses will use the individual datasets, the exceptions being those concerned with workless households and/or lone parents (lone parenthood being a household variable).
For each quarter, the LFS household dataset is the same set of records as the individual dataset, with additional household variables on the end. This means that if a household contains X people then there will be X records in the household dataset (rather than just one). Clearly, any simple analyses on this dataset will therefore result in a count of individuals rather than a count of households. Most times that one uses the household dataset, however, what one wants is a count of households rather than individuals. This can be achieved by filtering the dataset to only pick up the ‘Household Reference People’, for whom there is one per household.
Which individual datasets to use
There a two complications here, one relating to which quarters to use and the other relating to age-based analyses.
Taking the issue of which quarters to use. The four quarters of LFS from 2006 onwards are the calendar quarters, namely January to March, April to June, July to September and October to December. However, for each year prior to 2006, the survey was actually undertaken using seasonal quarters, namely Winter (December to February), Spring (March-May), Summer (June-August) and Autumn (September to November). These seasonal quarters have subsequently been re-grouped into standard calendar quarters by ONS, the net result being that there are actually eight datasets in the archive for each year (the four calendar ones and the four seasonal ones). Clearly any research should be based on only one of these two sets, and not both.
The obvious set to use is the calendar quarters and this is what is done in most parts of the website. However, in their re-grouping of the quarters, ONS have deleted some of the data which makes use of these quarters problematic for certain types of analysis. More specifically, where a variable changed definition (and name) from one seasonal quarter to another, the re-grouped calendar quarter excludes either one or both of these definitions, meaning that either one-third or all of the records for the re-grouped quarter do not have any data for that variable. Particular variables and quarters where this occurs include:
- Highest qualifications: quarter 1 2005 (no data for one-third of the records) and quarter 1 2004 (no data for any of the records).
- ‘O levels’: quarter 1 2005 (no data for any of the records) and quarter 1 2004 (no data for any of the records).
- Apprenticeships: quarter 1 2004 (no data for any of the records).
- Ethnic group: quarter 1 2001 (no data for any of the records).
Moving on to age-based analyses. The standard datasets from 2007 onwards only record which five-year age band the respondent falls into rather than their age in single years. This is because ONS has decided that age is a ‘disclosive’ variable which should not typically be made available to researchers. As part of this change, they have also introduced ‘special license’ versions of the datasets, which are currently identical to the standard license versions except with the addition of the age variable. These special license versions have to be applied for and this is a lengthy process with no guarantee of a successful conclusion.
Clearly, most research should use the standard license versions. There are, however, some types of analysis where the specific age of the respondents is important and where it might therefore be worth while applying for the special license versions.
Which quarters to use
LFS is a ‘semi-panel’ survey, with every household being interviewed for five consecutive quarters before being dropped. In reaction, some researchers, when producing annual results, merge the four quarters together and remove the duplicates before doing their analysis. This is, however, a time-consuming process which makes little difference to the answers and a more practical approach is to estimate the annual figures as the average for the four relevant quarters.
It is important to average across the four quarters because there are seasonal variations across these quarters.
How to avoid making errors when using the household dataset
The first issue is the point made earlier that, although it is a household dataset, each record actually relates to an individual so any simple analyses of the dataset will result in a count of individuals rather than of households. If a household count is what is wanted, this can be achieved by filtering to only pick up the ‘Household Reference People’.
A second issue is that there is a difference between ‘households’ and ‘families’ (for a discussion of the differences between these two terms, see the page on households, families and benefit units). So, for example, each record in the dataset is given both a ‘household type’ and a ‘family type’. For some purposes (e.g. analysis of workless households), the right variables to use are those that relate to the household. For most purposes, however, the right variables to use are those that relate to the family and the two sets of variables can lead to very different results. For example, consider a lone parent living with her parents: at a household level, the household type will be something like ‘couple living with non-dependent child’ and it is only at the family level that the type will be something which includes the words ‘lone parent’.
A third issue is that, for some reason, LFS uses slightly different definitions than are standard for both ‘families’ and ‘child’:
- Families: in the standard definition used by most surveys, all adults living with their parent(s) would be considered to be a different family than their parent(s). In LFS, however, they are considered to be part of the same family as their parent(s) so long they have never been married children and have no children of their own living with them
- Child: the standard definition used by most surveys is anyone who is either aged under 16 or aged 16-18, unmarried and on a course up to and including A level standard (or Highers in Scotland). In LFS, however, the term ‘child’ is limited to those aged under 16.
These differences are important. Consider a lone parent living with two of her children, one a dependent 11-year-old and the other a non-dependent 20-year-old: a straightforward analysis of the dataset would then result in the 20-year-old being erroneously considered to be a lone parent even though they have no children(!). The only practical (non-ideal) way round this to to exclude such individuals from any analysis by filtering out those adults who are neither the ‘head of the family’ nor their spouse/partner.
Some important terms in LFS
|Household||A group of people living at the same address who share common housekeeping or a living room. In other words, everyone who lives behind the same ‘front door’.|
Either a single person, or a married/cohabiting couple, or a married/cohabiting couple and their never-married children who have no children of their own living with them, or a lone parent with such children.
Everyone aged between 16 and retirement age (65 for men and 60 for women).
|Household reference person||The householder, who is the household member who owns the accommodation; or is legally responsible for the rent; or occupies the accommodation as reward of their employment, or through some relationship to its owner who is not a member of the household.
If there are joint householders, then one with the highest income. If their income is the same, then the eldest one.
Note that there is precisely one household reference person per household, so a count of household reference people is the same as a count of households.
|Unemployed||‘Unemployment’ is the ILO definition, which is used for the official unemployment numbers.
It comprises all those with no paid work in the survey week who were available to start work in the next fortnight and who either looked for work in the last month or were waiting to start a job already obtained.
|Economically active||Those in paid work + the ILO unemployed.|
|Economically inactive||All those with no paid work in the survey week who do not meet the criteria for ILO unemployment because either they are not available to start work for some time or they are not actively seeking work.
Many lone parents and disabled people fit into this category rather than ILO unemployment.
The economically inactive + the economically active = the whole population.
|Economically inactive but wanting paid work||People who are economically inactive but say that they would like to be in paid work.|
|Lacking but wanting paid work||The ILO unemployed + those who are economically inactive but want paid work.|
What variables to use and what do they mean
Clearly, the variables to be used will vary according to the purpose at hand. There are, however, a small number of variables which will be used much more than any of the others. The table below lists some of these variables.
|Variable name||Variable description||Comment|
|ages||Age bands||Mostly 5-year age bands|
|govtof||Government office regions||The standard regional boundaries|
|govtor||Government office regions||Somewhat more disaggregated than ‘govtof’|
|hiqual5||Highest qualifications/Trade apprenticeship|
|ilodefr||Economic activity (reported)||Broad categories only|
|inecac05||Economic activity (reported)||Much more disaggregated than ‘ilodefr’|
|pwt03||Integer weight||The individual-level weights|
In the LFS dataset, each variable has both a name and a description plus a description of what each of its values mean. This material is then repeated in the user manual (in a form which is easier to read) together with more detail explanatory text where needed. The table below lists those used in the analyses of this website, together with whether their use is major or minor.
Note that some of the detailed variable definitions change from year-to-year. Such changes are always accompanied by a change in the variable name. In other words changes in variable definitions always go hand in hand with changes in variable name and so, for example, one knows that if the variable name is unchanged then the variable definition is also unchanged.
|Primary dataset||Variable name||Official variable description||Extent of use on this website|
|appr4||Recognised trade apprenticeship||minor|
|cured||Current Education received||minor|
|ed13wk||J R Edu & Train in last 13 weeks (in wrk)||minor|
|ed4wk||J R Edu & Train in last 4 weeks (in wrk)||minor|
|ethas||Asian ethnic group||minor|
|ethbl||Black ethnic group||minor|
|ethmx||Mixed ethnic group||minor|
|ethwh||White ethnic group||minor|
|ftptwk||Full-time or part-time in main job||minor|
|futur4||J R Edu & Train in last 4 weeks (unemp)||minor|
|govtof||Government office regions||major|
|govtor||Government office regions||major|
|hiqual5||Highest qualif/Trade apprenticeship||major|
|hourpay||Gross hourly pay||major|
|ilodefr||Economic activity (reported)||major|
|inds92m||Industry SECTION (main job)||minor|
|inecac05||Economic activity (reported)||major|
|jobtmp||Types of temporary job||minor|
|llord||Landlord of accommodation||minor|
|numas||Num (A/S level passes) already held||minor|
|numol5||Num (O-level, GCSE etc pass) already held||minor|
|piwt03||Integer weight – Income||major|
|qulnow||Currently working / study towards qual||minor|
|union||Trade union or staff assoc member (GB)||minor|
|whytmp6||Reason for temporary job||minor|
|yptjob||Reason for part-time job||minor|
|Household||futype6||Type of family unit||minor|
|hdc515||No of children in household between 5 and 15||minor|
|hdpch18||No dependent children in household aged 16-18||minor|
|hdpch4||No of children in household age 4 year or less||minor|
|heacomb||Household economic activity variable||minor|
|hhtype6||Type of household||minor|
|hnftstud||No in household who are full-time students||minor|
|hnwkage||No of people in Household of working age||minor|
|relhfu||Relationship to head of family unit||minor|
|relhrp6||Relationship to HRP||major|
Analysis by region
The ‘North West’ government region is split into two in LFS, namely ‘North West’ and ‘Merseyside’. The results for these two ‘regions’ therefore need to be added together to calculate the true ‘North West’ results.
Note that, as well as recording where someone lives, LFS also records where they work. It can therefore be used to analyse cross-boundary flows to work.
Analysis by work status
There are around 35 possible values for work status. Any particular analysis will need to aggregate these into a smaller number of groups, for example: working, unemployed, economically inactive but wanting work; and economically inactive and not wanting work. The table below provides such a mapping for the 2006 datasets.
|3||Government emp & training programmes||Working|
|4||Unpaid family worker||Working|
|6||Inact- seeking, unavailable, student||Economically inactive – wanting work|
|7||Inact- sking,unav,lking after fam,home||Economically inactive – wanting work|
|8||Inact- sking,unav,temp sick or injured||Economically inactive – wanting work|
|9||Inact-sking,unav,long-term sick,disabled||Economically inactive – wanting work|
|10||Inact- sking, unavail, other reason||Economically inactive – wanting work|
|11||Inact- sking, unavail, no reason given||Economically inactive – wanting work|
|12||Inact- not sk,wld like,wait res job app||Economically inactive – wanting work|
|13||Inact- not sking, wld like, student||Economically inactive – wanting work|
|14||Inact- not sk,like,lking after fam,home||Economically inactive – wanting work|
|15||Inact- not sk,like, temp sick,injured||Economically inactive – wanting work|
|16||Inact-not sk,like,lng trm sick,disabled||Economically inactive – wanting work|
|17||Inact-not sk,like, believes no job avail||Economically inactive – wanting work|
|18||Inact- not sk,like, not yet looking||Economically inactive – wanting work|
|19||Inact- not sk,wld like,doesn’t need job||Economically inactive – wanting work|
|20||Inact- not sk, like, retired||Economically inactive – wanting work|
|21||Inact- not sk,like, other reason||Economically inactive – wanting work|
|22||Inact- not sk,like, no reason given||Economically inactive – wanting work|
|23||Inact- not sk,not like,wait results app||Economically inactive – not wanting work|
|24||Inact- not sk,not like, student||Economically inactive – not wanting work|
|25||Inact-not sk,not like,lk after fam,home||Economically inactive – not wanting work|
|26||Inact- not sk,not like, temp sick,injur||Economically inactive – not wanting work|
|27||Inact-not sk,not like,long-term sick,dis||Economically inactive – not wanting work|
|28||Inact-not sk,not like,belvs no job avail||Economically inactive – not wanting work|
|29||Inact- not sk, not like, not yet looking||Economically inactive – not wanting work|
|30||Inact- not sk,not like,doesn’t need job||Economically inactive – not wanting work|
|31||Inact- not sk,not like, retired||Economically inactive – not wanting work|
|32||Inact- not sk,not like, other reason||Economically inactive – not wanting work|
|33||Inact- not sk,not like, no reason given||Economically inactive – not wanting work|
Analysis by highest qualification
There are around 40 possible values for highest qualification, comprising a mixture of academic (e.g. GCSEs) and vocational (e.g. NVQs) qualifications. Any particular analysis will need to aggregate these into a smaller number of groups, and these groups can either be academic groups or vocational groups. The Department for Education (DfE) publish ‘equivalence tables’ to do this grouping, a version of which is provided in the table below as it applied in 2006.
Regarding the vocational grouping, note that:
- The ‘A level or equivalent’, ‘SCE higher or equivalent’, ‘A,S level or equivalent’, and ‘O level, GCSE grade A-C or equivalent’ academic qualifications cross NVQ boundaries depending on the numbers obtained. So, for example, ‘O level, GCSE grade A-C or equivalent’, is considered to be NVQ2 if the person has 5+ but only NVQ1 if the person has 1-4. To analyse this, the numbers of these qualifications that a person has needs to be analysed.
- There is no obvious vocational mapping for either ‘trade apprenticeship’ or ‘other qualifications’. DfE recommends that such people be allocated to particular NVQ levels in specific proportions. So, for example, 10% of those with ‘other qualifications’ should be allocated to NVQ3, 35% to NVQ2 and 55% to NVQ1.
- DfE sometimes publishes reports which refer to ‘Level 2’ and ‘Level 3’ qualifications. These are similar, but not identical, to NVQ2 and NVQ3 respectively.
|Value||Highest qualification||Academic grouping||Vocational grouping|
|1||Higher degree||Higher degree||NVQ5|
|2||NVQ level 5||Higher degree||NVQ5|
|3||First/Foundation degree||First degree||NVQ4|
|4||Other degree||First degree||NVQ4|
|5||NVQ level 4||Other higher education||NVQ4|
|6||Diploma in higher education||Other higher education||NVQ4|
|7||HNC, HND, BTEC etc higher||Other higher education||NVQ4|
|8||Teaching, further education||Other higher education||NVQ4|
|9||Teaching, secondary education||Other higher education||NVQ4|
|10||Teaching, primary education||Other higher education||NVQ4|
|11||Teaching foundation stage||Other higher education||NVQ4|
|12||Teaching, level not stated||Other higher education||NVQ4|
|13||Nursing etc||Other higher education||NVQ4|
|14||RSA higher diploma||Other higher education||NVQ4|
|15||Other higher education below degree||Other higher education||NVQ4|
|16||NVQ level 3||A-level or equivalent||NVQ3|
|17||Advanced Welsh Bacheloriate||A-level or equivalent||NVQ3|
|18||International Bacheloriate||A-level or equivalent||NVQ3|
|19||GNVQ/GSVQ advanced||A-level or equivalent||NVQ3|
|20||A level or equivalent||A-level or equivalent||NVQ3 if 2+, otherwise NVQ2|
|21||RSA advanced diploma||A-level or equivalent||NVQ3|
|22||OND, ONC, BTEC etc, national||A-level or equivalent||NVQ3|
|23||City & Guilds advanced craft||A-level or equivalent||NVQ3|
|24||Scottish CSYS||A-level or equivalent||NVQ3|
|25||SCE higher or equivalent||A-level or equivalent||NVQ3 if 3+, otherwise NVQ2|
|26||Access qualification to higher education||A-level or equivalent||NVQ3|
|27||A,S level or equivalent||A-level or equivalent||NVQ3 if 4+, NVQ2 if 2-3 or NVQ1 if 1 only|
|28||Trade apprenticeship||A-level or equivalent||Assume 50% NVQ3 and 50% NVQ2|
|29||NVQ level 2||GSCEs A*-C||NVQ2|
|30||Intermediate Welsh Bacheloriate||GSCEs A*-C||NVQ2|
|31||GNVQ/GSVQ intermediate||GSCEs A*-C||NVQ2|
|32||RSA diploma||GSCEs A*-C||NVQ2|
|33||City & Guilds craft||GSCEs A*-C||NVQ2|
|34||BTEC, SCOTVEC first or general diploma||GSCEs A*-C||NVQ2|
|35||O level, GCSE grade A-C or equivalent||GSCEs A*-C||NVQ2 if 5+, otherwise NVQ1|
|36||NVQ level 1||Other qualification||NVQ1|
|37||GNVQ, GSVQ foundation level||Other qualification||NVQ1|
|38||CSE below grade1,GCSE below grade C||Other qualification||NVQ1|
|39||BTEC, SCOTVEC first or general certificate||Other qualification||NVQ1|
|40||SCOTVEC modules||Other qualification||NVQ1|
|41||RSA other||Other qualification||NVQ1|
|42||City & Guilds other||Other qualification||NVQ1|
|43||YT, YTP certificate||Other qualification||NVQ1|
|44||Key skills qualification||Other qualification||NVQ1|
|45||Basic skills qualification||Other qualification||NVQ1|
|46||Entry level qualification||Other qualification||NVQ1|
|47||Other qualifications||Other qualification||Assume 10% NVQ3, 35% NVQ2 and 55% NVQ1|
Analysis by disability
LFS records two types of disability, namely ‘DDA disability’ and ‘work-limiting disability’. For work-related analyses, the obvious type to use is ‘work-limiting disability’.
Analysis by ethnicity
LFS uses multiple fields to code ethnicity, the first field dividing people into broad groups (White, Black, Asian, etc) and the other fields subdividing each broad group into narrower groups (Pakistani, Bangladeshi, Indian, etc).
Analysis by lone parenthood
Unlike most other datasets (e.g. HBAI), adults living with their parents are classified in LFS as being in the same family as their parents. This means that, when their parent is a lone parent, they themselves get classified as being in a lone parent family. To ensure that they are excluded from the lone parent counts, the syntax needs to check the relationship of every adult to the head of the family and only consider them as a lone parent if they are both in a lone parent family and are the head of that family.
Analysis by trade union membership
This can only be done using the 4th quarter datasets as the other quarters do not ask this question.
Analysis of NEETs
NEETs (not in education, employment or training) are analysed in LFS by a process of exclusion rather than inclusion. In other words, someone is not in education, employment or training if they are ‘not in education’ and ‘not in employment’ and ‘not in training’. The syntax for analysing this is complicated and should only be undertaken after discussions with DfE.
Relevant graphs on this website
|Children in workless households||household||first three||Filter on HRP to ensure that each household is only picked up once.
In line with ONS methods, children comprise all those under the age of 16 (i.e. not including dependent children aged 16 to 18).
Needs a lookup table to group household types into couples and lone parents.
The information published by ONS is only for selected quarters.
|Young adults without a basic qualification||individual||first three|
|Impact of qualifications on work||individual||all|
|Not in education, employment or training||individual||first two||Syntax developed in consultation with DfE.
The information published by ONS is only for selected quarters.
|Young adult unemployment||individual||all|
|Young adult low pay||individual||fourth and fifth|
|Wanting paid work||individual||all||Use the variable ‘durun’ to distinguish between short-term and long-term unemployed.|
|Work and disability||individual||first||Uses work-limiting disability rather than DDA disability.|
|household||second, third and fourth||Uses work-limiting disability rather than DDA disability.
Define someone as a lone parent only if both family type = lone parent and relationship to family unit head = head (the latter condition being to exclude adults living with their lone parent).
|individual||fifth and sixth||Uses work-limiting disability rather than DDA disability.|
|Work and lone parents||household||all||Filter on family type = lone parent and relationship to family unit head = head to ensure that only lone parents are picked up (the latter condition being to exclude adults living with their lone parent).|
|Work and ethnicity||individual||first and second||Exclude Northern Ireland because ethnic group not collected.|
|household||fourth||Filter on HRP to ensure that each household is only picked up once.|
|Work and gender||n/a||all||From the ONS website.|
|Blue collar jobs||individual||third and fourth||Include self-employed workers.
Only includes the first job for those who have multiple jobs.
|In workless households||household||first two||Filter on HRP to ensure that each household is only picked up once.
Exclude households which are entirely composed of full-time students plus households where their economic status is not known.
Exclude full-time students from the calculations to decide whether the household has one or more than one adult.
In line with ONS methods, children comprise all those under the age of 16 (i.e. not including dependent children aged 16 to 18).
|Low pay by industry||individual||all||Uses LFS rather than ASHE because of the age range.|
|Low pay and disability||individual||all||Uses LFS because analysis not possible from ASHE.
Uses work-limiting disability rather than DDA disability
|Low pay by ethnicity||individual||all||Uses LFS because analysis not possible from ASHE.
Omit chinese because of small sample sizes.
|Insecure at work||individual||second and third|
|individual||fourth||Data available for the fourth quarter of each year only.|
|Access to training||individual||all|
|Without educational qualifications||individual||first three|
|Polarisation by housing tenure||individual||second, third and fourth|
- Sometimes the options for a particular question (e.g. highest qualification) change from year to year. This is reflected in a change of name for the variable. In such cases, the SPSS syntax needs to be amended, both to pick up the right variable and, within this, to pick up the right values. When the variable in question is ‘highest qualification’ then the new values need to be added to the lookup tables, together with their equivalences (where the latter can usually be judged from the position of the new values as the highest qualifications are in descending order from ordered from highest first).
- Even if the options for a particular question remain the same, their precise wording can change from year to year. In such cases, the lookup tables need the new wording added.
- To get results for the North West region, add the results for North West and Merseyside together.
Scotland, Wales and Northern Ireland graphs
These are effectively a subset of the UK graphs using government region as a filter.