Labour force survey

The definitions in the Norwegian LFS are in accordance with the definitions in other European countries. Norway complies with the EU regulations and participates in the European statistical cooperation.

Employed persons are persons who performed work for pay or profit for at least one hour in the reference week, or who were temporarily absent from work because of illness, holidays etc. Persons engaged by government measures to promote employment are also included if they receive wages. Persons laid off 100 per cent with a continuous duration of until three months are defined as employed.

The employment rate is calculated as a percentage of the whole population in the same age bracket.

Unemployed persons are persons who were not employed in the reference week, but who had been actively seeking work during the preceding four weeks, and were available for work in the reference week or within the next two weeks. Persons laid off 100 per cent are defined as unemployed after three continuous months of leave if they also fulfil the criterions of active searching and being available to take a job.

The unemployment rate is calculated as a percentage of the labour force.

The labour force is the sum of persons in employment and the unemployed, that is, those who activel yoffer their labour on the labour market. They are also referred to as economically active.

Persons outside the labour force are those who are neither in employment nor unemployed. The group inclueds persons laid off 100% and who do not fulfil the criterions of active search and availability.

Status in employment groups the employed on being employees, self-employed, or unpaid family workers. In the LFS, persons in employment who during the interview confirm that their registered job in the A-scheme is in fact their job will be classified as employees. The exception is those who are contractors or free lance workers. They are asked about the number of clients / customers and their payment forms. Those who worked for only one client / customer in the reference week and who received a wage from them are classified as employees. The remaining are classified as self-employed. Persons in employment who do not confirm the register information in the A-scheme are asked if they are employees, self-employed, familiy workers, contractors or freelance, and coded accordingly.

In addition to the measurement of employment and unemployment according to the international recommendations and definitions, it is also asked a single question in the LFS to all non-employed persons, and to the part-time employed persons, about their main activity. This variable gives the persons' self-perception regarding their activity or status. The purpose is to estimate how many people are in education, homemakers, pensioners etc., and how many have a part-time job besides.

The potential labour force consist of persons who were not employed in the reference week, but had either been seeking work during the preceding four weeks or were available for work in the reference week or within the next two weeks.

The extended labour force consist of the (ordinary) labour force and the potential labour force.

The labour market slack consist of the unemployed, persons working part-time involuntary and persons in the potential labour force.

Man-hours worked include all actual working hours, i.e. including overtime and excluding absence from work. This is published as man-weeks per week, and can therefore also be read as the number of employed full time equivalents in the month under consideration.

Contractual/usual working hours refer to the weekly number of working hours determined by the working contract. Absence from work because of illness, holidays etc. is not subtracted, and overtime is not included. Employees whose contractual working hours vary from week to week give information on the average of their contractual working hours per week over the last three months. This also applies to self-employed and family workers.

Full-time/part-time distinction is based on usual working hours per week. Usual working hours of 37 and more is full-time, in addition to varying working hours or usual working hours between 32 and 36 where the respondent classify this as full-time work. The rest is classified as part-time, i.e. usual working hours less than 32 hours and varying hours or usual working hours between 32 and 36 where the respondent classify this as part-time work. For persons with more than one job, only the usual working hours in the main job is used in the classification of full-time/part-time work.

Overtime is defined as working hours which exceed the contractual working hours. The overtime may be compensated by payment or by time off, or be without any compensation.

Underemployed persons consist of part-time employed persons wishing and seeking longer usual working hours and who were able to start with increased working hours within a month. This definition corresponds to that of the unemployed in the meaning that they must be both seeking and available.

Persons who work involuntary part-time consist of employed persons wishing longer usual working hours and who are able to start with increased working hours within a month. This group differs from the underemployed since they don't need to be seeking more increased working hours, it suffices to wishing it and beeing available for it.

The employees are asked whether they have a permanent job (a work contract of unlimited duration) or a temporary job (a work contract of limited duration). If the person has more than one job, only the main job is classified by permanence.

There are two kinds of working time arrangements outside ordinary hours (Monday to Friday from 6 am to 6 pm).

• Shift work is working time outside normally working hours in pre specified periods.
• Work outside ordinary hours, not shift work. This refers to work on evenings, nights, Saturdays and Sundays which is not shift work. Evening work is defined as work between 6 pm and 10 pm. Night work is between 10 pm and 6 am.

Respondents who report that they work on one or more of these working time schedules outside ordinary hours are also asked about the frequency of this kind of work during a four week period. This is done for each of the working time schedule separately. Based on this, the work outside regularly hours is divided into the categories "regularly" or "sometimes".

For evening and night work, the respondent must have this kind of work on at least half of their working days in the four week period to be defined as having regularly evening work and/or regularly night work.

For work on weekends, they must work 2-4 Saturdays and/or 2-4 Sundays during the four week period to be defined as having regularly Saturday and/or Sunday work. The respondents who have these kinds of working schedules, but more seldom than indicated above, are defined as sometimes having this kind of work.

Age means completed years in the reference week.

Immigrants are persons born outside Norway to parents who were also born outside Norway. This information is collected from registers.

The industrial classification is based on the EU-standard of NACE Rev. 2.

The occupational classification is based on the International Standard Classification of Occupations 2008 (ISCO-08).

The educational classification is in accordance with the Norwegian Standard Classification of Education.

Most 15 year old and some 16 year old have not yet completed primary education, but in the LFS they are still coded as completed.

Classification of regions follow the standard for regions

Name: Labour force survey (LFS)

Topic: Labour market and earnings

Division for labour market and wage statistics

Most of the results are only shown for the whole country.

The main indicators are available by region.

We recommend using the register based employment statistics for results by municipality and county.

Monthly, quarterly and annual. Monthly disseminating of trend and seasonally adjusted figures 3-4 weeks after the end of the month. Quarterly and yearly figures are normally disseminated 5-6 weeks after the end of the quarter or year.

Quarterly and annual micro data files are sent to Eurostat. Tables are sent each month to Eurostat. Tables are sent annually to OECD, ILO and IMF. A selection of variables are sendt to Sikt.

Collected and revised data are stored securely by Statistics Norway in compliance with applicable legislation on data processing.

Statistics Norway can grant access to the source data (de-identified or anonymised microdata) on which the statistics are based, for researchers and public authorities for the purposes of preparing statistical results and analyses. Access can be granted upon application and subject to conditions. Refer to the details about this at Access to data from Statistics Norway

The main purpose of the survey is to provide data on employment and unemployment.

The Norwegian LFS started in 1972.

The main changes in the survey are:

• 1988 (second quarter): The number of persons in employment increased by about 10 000 as a consequence of starting to collect data monthly.
• 1996 (first quarter): The number of unemployed increased by 11 000, or 0.5 percentage points, as a consequence of the new production system. The change was from one reference week per month to continous data collection, the unemployment definition was changed to follow the ILO recommendations, and the data collection mode was changed from paper to electronic.
• 2006 (first quarter): The number of persons in employment increased by 8 000 and the number of unemployed decreased by 1 000 as a consequence of the new production system. The lower age limit was changed from 16 to 15 years of age.
• 2021 (first quarter): The number of persons in employment increased by 22 000 and the number of unemployed increased by 5 000, as a consequence of the new EU LFS regulation. The production system was revised in full, with changes in sampling, weighting, use of register data, and a new questionnaire.

Development tasks and additional modules are partially financed by the EU.

The survey give information to the labour market authorities and other users about the situation on the labour market, and provide data for labour force research and forecasts, as well as for international organizations and mass media. Some of the main users are the Ministry of Finance, the Ministry of Work and Social Affairs, and the Work and Welfare Administration.

No external users have access to statistics before they are released at 8 a.m. on ssb.no after at least three months’ advance notice in the release calendar. This is one of the most important principles in Statistics Norway for ensuring the equal treatment of users.

Differences between the LFS and the National Accounts

The main reason for the discrepancies between the LFS and the NA is that the LFS measures employment among persons who live in Norway, whereas the NA measures employment in Norwegian owned enterprises.

Differences between the LFS and registered unemployed with the Work and Welfare Administration (NAV)

The figures on unemployment based on the LFS differ from the figures on unemployed persons registered at the Employment Offices. For more information, see the article Hvorfor ulike arbeidsledighetstall? (only in Norwegian)

Information about immigrants in the LFS

The main source to describe the situation for the immigrants on the labour market is the register based statistics on employment and unemployment. By using data from the LFS as a supplement we will achieve still more information regarding this group, for example on actual and desired working hours, temporary employment and patterns of working time. Moreover the LFS has more data than the registers on job seeking and desire for work.

The main problem using the LFS for statistics on immigrants is the size of the sample and the statistical uncertainty. In order to minimise the uncertainty we have subdivided the immigrant group in only two: Persons from EU/EFTA, North America, Australia and New Zealand. Persons from Eastern Europe outside the EU, Asia including Turkey, Africa, South and Central America, and Oceania excluding Australia and New Zealand.

The statistics are developed, produced and disseminated pursuant to Act no. 32 of 21 June 2019 relating to official statistics and Statistics Norway (the Statistics Act).

Council Regulation (EC) 2019/1700. Commission implementing regulations 2019/2240 and 2019/2241.

The goal of the regulations is to establish a common framework for European social statistics.

The LFS covers everyone who lives in Norway. The survey contains most information of the age group 15-74 years, but it also has a substantial data amount about persons 74 - 89 years. Those under 15 and over 89 are only covered by register information.

The observation units are persons and households.

The main source for the LFS is quarterly, representative samples. The data is collected by interview by telephone.

Inhabitants in all municipalities are randomly selected, on the basis of the population register. The sample consists of about 21 000 persons each quarter. Each respondent participates in the survey 8 times during a period of 8 quarters, and are asked about their connection to the labour market.

Additionally, all members of the main respondent's household are interviewed once during these two years. They get a shorter interview. They make up aproximately 3 000 persons, so the total sample consist of about 24 000 persons.

Editing is defined here as checking, examining and amending data.

Coding of industry is done from information from registers.

Demographic data are collected from the Central Population Register, and data on education are based on a register of individual data collected by Statistics Norway from the educational institutions.

All weeks of the year are covered with data collection.

The LFS collects data by telephone interviews.

Some information from previous interviews are re-used. For instance, if the respondent confirms that they have the same job as the last time we talked to them, we do not ask about the respondent's occupation.

The respondent is the same person as the observation unit.

As the data collection is made by use of computer-assisted interviews, some procedures for electronic control of the registration of answers are included in the questionnaire, for example concerning the number of working hours during the reference week. In some cases the interviewers become a "warning" by recording an answer, in other cases maximum or minimum values have been set beforehand.

The most common analysis unit is person. The absolute numbers from the LFS are presented in the form of estimated total for the entire population aged 15-74. The weights or inflation factors vary, but have an average of about 195 for quarterly figures.

The estimation method uses more demographic data and register information relevant to the connection to the labour market in order to minimize standard errors and correct more for bias in the response sample in LFS since the nonresponse is not random. New method was launched in April 2018, is also used on LFS data back to 2006 to get the most comparable time series.

The estimation method in LFS is done in several stages, and are called multiple model calibration. Initially, the main labor market status of LFS, which is employed, unemployed, outside the workforce, are modeled consistent with a multinomial logit model, explained with a number of register variables known to all in the population. The register information are registered completely unemployed, on measures and persons with disabilities at NAV, register information on disability pensioners, education level, marital status, family size and immigrant category, country of origin, gender and age, residence, and information from the A scheme and the Tax Recovery Register. The model provides predictive probabilities every month for each main labor market status in LFS for everybody in the population.

The monthly weights in LFS are calibrated using these predicted probabilities and some register variables directly[1]. This means that the weights also become consistent with the population for the number in the population register by gender, different age groups and region, as well as consistent weights for the number of full / part-time wage earners by gender and registered employed (yes/no) cross classified by immigrants in 2 groups.

Multiple model calibration provides some variance reduction, utilizing more efficiently that we have available good help information about everyone in the population from various registries that Statistics Norway has linked.

The initial weights before calibration are the ratio of the number of people in the population to the gross sample per. county (NUTS3), and takes into account that people in different counties have different probabilities of being selected to LFS sample.

For more detailed technical information about the new estimation method, please see Documents 2018/16 [https://www.ssb.no/en/arbeid-og-lonn/artikler-og-publikasjoner/new-estimation-methodology-for-the-norwegian-labour-force-survey]

Week-proportional weighting of the months in quarterly averages

LFS have continuous data collection. In order for all reference weeks to weighted evenly in quarterly averages, we now make week-proportional adjustment of monthly weights in our quarterly averages. That is, the monthly weights are multiplied by 4/13 or 5/13 depending on whether the months in the LFS contain respectively 4 or 5 whole weeks. Weekly proportional weighting of quarterly average has been made on all quarterly figures in our StatBank back to 2006. Before that is the even adjustment off all monthly weights was used, i.e. multiplied by 1/3 for all months.

[1]That the weights are consistent for a register variable, such as gender, means that the sum of the weights in the responses in the LFS equals the number of the population for each category of the register variable, such as the number of men and the number of women in the population

The monthly figures are seasonally adjusted using X13-ARIMA-SEATS.

The seasonal adjustment of employed and unemployed are done indirectly; we seasonally adjust the series for women and men respectively over and under 24 years of age separately. For man-weeks workes, we seasonally adjust the following 3 series separately: persons aged 24 and below, men above 24 years and womenabove 24 years. We get the totals by summing the seasonally adjusted partial series afterwards.

We construct sex distributed seasonally adjusted figures for employed, unemployed and weekly hours worked for people aged 24 and below by using seasonally adjusted series by age and breaking them down by gender using monthly distribution keys. The distribution keys are made of trend numbers from extra runs of X-12ARIMA based on the LFS. We leave it to the seasonal adjustment program to decide whether additive or multiplicative decomposition of the series gives the best results, in addition to the choice of ARIMA model. These optimal picks are usually locked in for one year at a time. We allow X-12ARIMA to recalculate the seasonal components and parameters in the precorrection regressions each month. This is best practice for seasonal adjustment program settings according to Eurostat, and is done to have little revision of the seasonally adjusted time series throughout the year.

We allow the seasonal adjustment program to pre-correct the time series using regression analysis. We have specified right-hand side variables to take into account extreme values ​​and effects of holidays that do not fall on weekdays in the same month in the LFS each year. The seasonally adjusted series for employment and unemployment for those over 24 and the figures for man-hours worked are precorrected if Easter is in March. The figures for man-hours worked are also sensitive to more individual holidays, and are therefore additionally pre-corrected for the effects of Easter Monday in March, May 1, May 17, Ascension Day and Whitsunday 2. In addition, we correct for the number of public holidays on weekdays in December and the number of normal working days in Christmas and the number of weekdays in June which in the LFS end up in July, because in the LFS you never divide reference weeks (Monday-Sunday) between months as the calendar does. We also take into account that two of the above-mentioned holidays can end up on the same date (for example, Ascension Day on 17 May in 2007 and 2012 and on 1 May 2008). All the right-hand side variables counted as deviations are given respective monthly averages. In addition, the variables have been adjusted with a week multiplier to better adjust for the fact that some months in the LFS have 5 reference weeks, while others have 4.

Seasonal adjustment during the corona crisis (March 2020 to March 2022) is done in such a way that the figures from the crisis are not included in the basis for the calculation of the seasonal pattern. This is in line with the recommendations from Eurostat. This handling means that trend figures will initially follow seasonally adjusted figures in the period. In the corona period March 2020 to March 2021, the trend figures are therefore smoothed with a three-month moving centered average.

The corona crisis is modeled with consistent level shift (LS) specification for each month from March 2020 through March 2022.

The trend numbers represent the longer-term tendency in the data, including the business cycle. The trend can often be slightly revised when new observations are incorporated, especially towards the end of the time series, called revision uncertainty. The final trend is calculated by smoothing the seasonally adjusted figures. The program selects a moving trend average based on statistical properties of the data. For monthly series, either a 9-, 13- or 23-term Henderson moving average will be selected, and associated asymmetric variants towards the end. For detailed information on trend filters in X-12ARIMA, see for example the Australian Bureau of Statistics website or chapter 12.6 of the Handbook on Seasonal Adjustment.

See the "About seasonal adjustment" section below for more information about the seasonal adjustment.

Employees of Statistics Norway have a duty of confidentiality.

Statistics Norway does not publish figures if there is a risk of the respondent’s contribution being identified. This means that, as a general rule, figures are not published if fewer than three units form the basis of a cell in a table or if the contribution of one or two respondents constitutes a very large part of the cell total.

Statistics Norway can make exceptions to the general rule if deemed necessary to meet the requirements of the EEA agreement, if the respondent is a public authority, if the respondent has consented to this, or when the information disclosed is openly accessible to the public.

More information can be found on Statistics Norway’s website under Methods in official statistics, in the ‘Confidentiality’ section.

We use international standards for definitions, which means that the results of the Norwegian LFS can be compared to results in other countries.

There are several times series breaks in the survey:

• 1975 (first quarter): 13 000 more persons in employment, 22 000 more persons in the total population because of a new estimation method.
• 1976 (first quarter): 30 000 more persons in employment due to new questions about absences from work and about family workers.
• 1980 (first quarter): 26 000 fewer persons in employment for the annual averages of 1980-1986 (new estimation method in use from 1987, with effect backwards to 1980). The reduction was however balanced out because conscripts (about 25 000 persons per year) were now counted as in employment.
• 1986 (first quarter): Persons in employment increased by 15 000 because the limit for being counted as a family worker was reduced from 10 hours per week to one hour per week.
• 1988 (second quarter): 10 000 more persons in employment, as a result of starting to collect data every month.
• 1996 (first quarter): Unemployment up 11 000. Substantial changes in the production system: continous data collection, updated the unemployment definition to follow the ILO definition, change from paper based to electronic data collection.
• 2006 (first quarter): Change of the lower age limit from 16 to 15 years. Employment increased 8 000, unemployment down 1 000.
• New weighting method in 2018, with back calculations to 2006. All published tables in Statbank are revised back to 2006. See https://www.ssb.no/arbeid-og-lonn/artikler-og-publikasjoner/_attachment/346996?_ts=162d85820c8 for more details.
• 2015 (first quarter): The data base at the Work and Welfare administration was upgraded to the A-scheme. This lead to better and faster register information about jobs. See https://www.ssb.no/arbeid-og-lonn/naermere-om-forholdet-mellom-gammel-og-ny-statistikk for more information.
• 2021 (first quarter): Employment up 22 000 and unemployment up 5 000. New EU regulation, updated sampling and weighting methods, more use of register data, updated questionnaire.

The estimation method which we started to use in 2018 uses more registers that will reduce non-response bias and sample uncertainty. This leads to a slightly lower level of employed and almost correspondingly higher levels for people outside the workforce than the previous method. The total number of unemployed changes only a little. To get the most comparable figures, the time series are revised back to 2006 in our StatBank. However, the adjustments vary slightly between groups and over time.

There is a time series break between 2020 and 2021 because of a restructuring of the LFS in 2021. This led to a break in the number of employed people estimated to 21 900 and a break in the number of unemployed people to 5 400 (both estimates, the latter not significant). The method used in the break estimation is documented in a separate paper.

The unemployment rate in the new LFS is 0.1 percentage points higher than in the old LFS.

The main changes were:

• The sample was changed from being persons 15-74 years living in Norway to persons 15-89 years living in private households in Norway. This means that the age bracket is wider, but also that persons who live in communal households, like those serving in the national service, are excluded. The changes are implemented gradually throughout 2021 and 2022, since only 1/8 of the sample is renewed each quarter.
• The sampling unit was changed from family to person, and the sampling plan is changed accordingly.
• Some central variables have new defintions, following the new LFS regulation. Some of them are employment, full time, under-employment.
• New questionnaire. Several variables are changed, the language has been updated to reflect the modernisation over the last 20 years, increased coordination with other countries, and questions which are designed to work in a multi mode setting.
• Proxy interviews are no longer allowed.
• Information is collected not only about the reference person, but also about the household. This takes place only in one interview wave.

The table below show break estimates for employed and unemployed (4th quarter 2020) for more age groups than the one mentioned above.

Tables 13618 and 13619 contain comparable time series across the break between 2020 and 2021. Theses tables contain figures for the labour force, employed persons and unemployed, where the figures for the period 2009-2020 are adjusted according to the new LFS level using the same method as described above.

Since the LFS population was changed from 2021 to excluding residents in collective households, there are also breaks in the shares of employed persons and persons in/outside the labour force. Therefore, the tables 13618 and 13619 also contain population figures for the period 2009-2020 in accordance with the new LFS population from 2021 on, e.g. persons living in private households. Consequently, they contain labour force and employed persons in per cent of the population back to 2009.

Table 13760 also contain break adjusted figures. These figures are in addition seasonally adjusted.

Possible breaks in other variables

Other time series breaks are analysed furtherin a separate article. This is regarding the number of employees and self-employed, and the number of part time employees. There is also a possible time series break in actual working hours and man-hours per week, but this is not possible to pinpoint because the Corona crisis took place at the same time as the break in the LFS.

Additionally there looks to be a time series break in the number of persons 15-19 years of age who are neither in employment nor in education.

Coverage error in sample surveys happen if the sampling frame does not correspond to the target population. This can lead both to undercoverage and overcoverage. We sample a predetermined number of persons from the population register to the LFS each quarter. The sampling takes place two weeks before the start of the quarter.

Overcoverage is not likely to be a problem, because the target population of the LFS is persons who live in private households in Norway and the sample frame only contains persons who live in Norway. It can however happen, for instance when persons have moved from Norway without reporting it to the population register. Since there is a time lag of a few weeks between the sampling and the interview it is also possible to have some cases each quarter where the respondent has moved from Norway, moved from a private to a collective household, or died, before the interview. Such cases would in most instances be noted by the interviewers, and the interview would just be registered as non-response. Persons who have completed an interview but who are no longer in the population register at the end of the quarter will not be assigned a weight in the LFS.

Undercoverage, the situation where one should have been sampled for the LFS but wasn't, is not a large problem either, but it is nevertheless larger than the overcoverage. The population register is updated continously and has a good overview of immigration. The time lag between sampling and the interview means that a few immigrants are not included, but this is not a large problem. The main issue here is that the LFS has eight waves over two years. So if large and homogenous groups immigrate over a short time span they will not be captured by the LFS, as it takes two years to rotate inn a full new sample.

Measurement and processing errors

Measurement errors happen because of the data collection method. In a survey of persons, as the LFS, this will typically be because of unclear questions. The result will be that the respondent gave the wrong answer because they misunderstood what we meant.

The LFS questionnaire is based on the recommendations from Eurostat, which again are based on thousands of hours of testing and analysis from a large range of statistics offices in Europe. Additionally, the Norwegian questionnaire has been tested extensively. We are therefore reasonably sure that the measurement errors are small. The LFS also uses large amounts of register data, as for instance from the A-scheme. These data sets are in constant use by Statistics Norway, the Tax Administration, and the Work and Welfare Administration, are under constant surveilance and analysis. If there are errors they are spotted and fixed fast.

Processing errors are errors which appear after the data collection, that is, they are created by the post processing of data in Statistics Norway. Traditional examples are errors which appear in the transfer of data from one medium (e.g paper) to another (e.g. electronic). But as the LFS is fully electronic this is not a possible source of errors, but there could be errors in merging one file with another, or in the computer coding. Both Statistics Norway and Eurostat keep a close eye on the data quality and check the reasons for unexpected results or invalid combinations of outcomes.

Nonresponse errors happen when respondents refuse to answer. There are two types of nonresponse error: item nonresponse when the respondent has refused to participate at all, and partial nonresponse when the respondent has refused to answer to some of the questions.

The size of the item nonresponse has varied a lot since the start of the LFS in 1972. The first 20 years it varied between 10 - 12 percent. In 1992-1997 it was particularly low, at 6 - 8 percent, but with an increase up to 21 percent in 2013. This lead us to work systematically with the response rate, and it was reduced to 14 percent i 2018. In 2021 the sample unit was changed from family to person, which meant that we stopped offering proxy interviews. The nonresponse increased to about 20 percent in 2021 and further up to just below 25 percent in 2022.

The partial nonresponse varies a lot between the variables, and goes from 0 to about 40 percent. The most important variables, like underemployment, working time, occupation and industry have a nonresponse of less than 1 percent.

Sampling errors are a consequence of the results coming from a sample of persons and not the whole population. The sampling error is therefore the expected deviation of the result from a sample survey from what we would have gotten if the entire population was interviewed.

The standard error is substantially lower for the annual averages than for the quarterly averages. This is because the statistical uncertainty is diminished when we have more observations. One should be careful when assessing changes over time from one survey to another. For groups where the gross flow in and out is relatively small compared to the size of the group we can safely assume that the standard error for the change from one point in time to another is approximately the same as the standard error for the total group. Where the gross flows are larger we must multiply the standard error with a factor of 1.3. The confidence intervals for the changes can be so large that we cannot say for certain if a change has taken place.

Total nonresponse

We adjust for the total nonresponse in the estimation procedure, and we make sure that the sample matches the population each quarter for age, sex, and region. The margin of error for the main indicators in the LFS are publised in table 13618 (annual) and table 13619 (quarterly).

Totalt frafall

Vi justerer for totalfrafallet i estimeringen, der vi sikrer at utvalget samlet sett er riktig hvert kvartal i forhold til alder, kjønn og landsdel.

Feilmarginene for årsgjennomsnitt for hovedindikatorene i AKU er publisert i tabell 13618 og for kvartalsresultater i tabell 13619 i statistikkbanken.

A revision is a planned change to figures that have already been published, for example when releasing final figures as a follow-up to published preliminary figures. See also Statistics Norway’s principles for revisions.

Revisions in previously published seasonally adjusted figures can take place when new observations (or revised previous observations) are included in the basis of calculation. The scope of the revision is usually greatest in the most relevant part (last 1–2 years) of seasonally adjusted time series. A corresponding revision in trends is also typical, particularly at the end of the time series. The extent of the revision of trends and seasonally adjusted figures is partly determined by the revision policy, see Section 4 of the European Statistical System (ESS) Guidelines on Seasonal Adjustment on the Eurostat website. For more information on the revision of seasonally adjusted figures, see the ‘About seasonal adjustment’ section in the relevant statistics.

Monthly and quarterly time series are often characterised by considerable seasonal variations, which might complicate their interpretation. Such time series are therefore subjected to a process of seasonal adjustment in order to remove the effects of these seasonal fluctuations. Once data have been adjusted for seasonal effects by X-13-ARIMA-SEATS or some other seasonal adjustment tool, a clearer picture of the time series emerges.

For more information on seasonal adjustment: metadata on methods: seasonal adjustment

As a consequence of the Corona outbreak im March 2020 we did some changes to the seasonal adjustments. Additive outliers and level shifts are incorporated into the precorrections. The monthly trend for the last observation is replaced with the seasonally adjusted result, since we are not able to specify the level shift in the last observation and point extreme value. This adjustment form was ended in March 2022.

Due to fixed and movable holydays and holidays in July and December the figures of man-weeks worked vary across the year. Also the number of employed persons and unemployed persons vary across the year, especially for the young people. Among other matters this is because of summer substitute and young people searching for summer jobs. This makes it difficult to compare one month to another. We therefore seasonally adjust the figures so that we can better analyse the underlying development, i.e. the business cycle.

Persons in employment and unemployed are seasonally adjusted by age over / under 24. Man-hours are adjusted in three separate series: persons 24 years and younger; men above 24 years; women above 24 years

Pre-treatment is an adjustment for variations caused by calendar effects and outliers.

Running a detailed pre-treatment. This means using models which are specially adapted for the pre-treatment of the raw data for a given series.

Calendar adjustment

Calendar adjustment involves adjusting for the effects of working days/trading days and for moving holidays. Working days/trading days are adjustment for both the number of working days/trading days and for that the composition of days can vary from one month to another.

We perform calendar adjustments on all series showing significant and plausible calendar effects within a statistically robust approach, with RegARIMA.

Methods for trading/working day adjustment

No correction.

Comments : LFS is surveying persons attachment to the labour market in whole reference weeks. The reference weeks are not split to follow the exact months of the calendar. Therefore, all the monthly data files include as many Mondays as Tuesdays and so on. The monthly data files include either 4 or 5 reference weeks, and the inflation factors adjust fore that. Due to this the number of working days in a month is only affected by holidays and national days off.

Correction for moving holidays

Correction based on an estimation of the duration of the moving holidays effects, specifically adjusted to Norwegian circumstances.

Comments :

The seasonally adjusted time series for man-weeks worked are pre-adjusted if Easter is in March.

Due to sensitivity to single holidays or common days off, the man-weeks worked series in addition are pre-adjusted for 2nd Easter day, May 1st, May 17th, Whit Monday and Ascension Day. (We also take account of the effect when two of these days occur on the same date, for instance Ascension Day on May 17th in 2007 and 2012 and on May 1st. in 2008).

In addition we pre-adjust man-weeks worked series for the number of common days off that falls on week days in December, and for the number of normal days in the week between Christmas and New Year's Eve that falls in December/January in the LFS. All regression variables are deseasonalized by subtracting the long term monthly average. Also the regression variables are fine tuned by a week multiplier to take account of that some months represent 4 reference weeks and other 5 reference weeks.

National and EU/euro area calendars

Use of the Norwegian calendar

Comments : We also take account of the fact that LFS don’t divide reference weeks (Monday – Sunday) between months, like in the ordinary calendar. For instance the holiday May 1 st may in the LFS be in April. Correspondingly may New Year's Day be in December in the LFS.

Treatment of outliers

Outliers, or extreme values, are abnormal values of the series.

Outliers are detected automatically by the seasonal adjustment tool. The outliers are removed before seasonal adjustment is carried out, and then reintroduced into the seasonally adjusted data.

Comments : Only additive outliers are detected automatically.

Handling of the Corona period (March 2020 to March 2022)

The Corona period is modelled for all time series, with specification of the extreme value for the last observation (after March 2020), and forwards with level shifts for each month, starting in March 2020, except from the most recent observation. For instance, results as of March are precorrected wit the specification AO2020.3. Results as of May are precorrected with LS2020.3 LS2020.4 and AO2020.5. Precorrected effects are returned to the seasonally adjusted numbers after the calculation of the seasonal component. Usual trends for the last observation should not be used, since they are not affected by the last extreme point estimate, only by the level shift. We therefore replace the trend for the last month with seasonally adjusted results for the last month. This is our best estimate for the last observation. In the statbank we present time series which are smoothed three month moving averages, both for seasonally adjusted and for trend.

Model selection

Pre-treatment requires choosing an ARIMA model, as well as deciding whether the data should be log-transformed or not.

Automatic model selection by established routines in the seasonal adjustment tool.

Comments : X-13-ARIMA-SEATS (latest Linux version)

Model structures and filters are kept unchanged between annual reviews (partial concurrent adjustment).

The ARIMA-selection procedure pickmdl{} is preferably used, with the TRAMO-like procedure automdl{} only as an alternative. RegARIMA-models without calender effects and automatic outlier detection for the model-span from April 2022 with high critical value are used.

The supplementary ARIMA-modelse in pickmdl{} are:

(0,1,[1,3])(0,1,1)12

(0,1,3)(0,1,1) 12

(0,1,1)1(0,0,1)3(0,1,1)12

Decomposition scheme

The decomposition scheme specifies how the various components – basically trend-cycle, seasonal and irregular – combine to form the original series. The most frequently used decomposition schemes are the multiplicative, additive or log additive.

Automatic decomposition scheme selection.

Choice of seasonal adjustment approach

X-13-ARIMA-SEATS

Consistency between raw and seasonally adjusted data

Impose the equality over the year of seasonally adjusted data to original data (e.g. sum or average). Since the annual averages only are ready together with quarter four we use preliminary adjustment factors from the previous year before they are ready. The whole year is therefore level adjusted when quarter four is published.

Consistency between aggregate/definition of seasonally adjusted data

In some series, consistency between seasonally adjusted totals and the original series is imposed. For some series there is also a special relationship between the different series, e.g. GDP which equals production minus intermediate consumption.

Impose the equality between aggregated series and the component series.

The labour force is not seasonally adjusted. The labour force, seasonally adjusted is defined as the sum of employed persons, seasonally adjusted and unemployed persons, seasonally adjusted.

The equality is imposed by indirect approach.

Direct versus indirect approach

Direct seasonal adjustment is performed if all time series, including aggregates, are seasonally adjusted on an individual basis. Indirect seasonal adjustment is performed if the seasonally adjusted estimate for a time series is derived by combining the estimates for two or more directly adjusted series.

Indirect approach where the seasonal adjustment of components occurs using the same approach and software, and then totals are derived by aggregation of the seasonally adjusted components.

Horizon for estimating the model and the correction factors

When performing seasonal adjustment of a time series, it is possible to choose the period to be used in estimating the model and the correction factors. Correction factors are the factors used in the pre-treatment and seasonal adjustment of the series.

The whole time series is used to estimate the model and the correction factors

Trend

The trend represents the long term tendency in the data, including the business cycle. For the monthly series we use either 9-, 13- or 23-term Henderson gliding averages, and the corresponding asymetrical variants towards the end. For detailed information about trend filters i X-12ARIMA, see for instance chapter 12.6 in the Handbook on Seasonal Adjustment.

General revision policy

Seasonally adjusted data may change due to a revision of the unadjusted (raw) data or the addition of new data. Such changes are called revisions, and there are several ways to deal with the problem of revisions when publishing the seasonally adjusted statistics.

Seasonally adjusted data are revised in accordance with a well-defined and publicly available revision policy and release calendar.

Revisions second months in the quarter .
The figures from the first months in the quarter are preliminary. They are re-estimated, at the same time as the data from the second months in the quarter are estimated, supplemented with the extra interviewees that we got after the preliminary deadline and with more up to date or better auxiliary register variables. For the re-estimation, also more up to date auxiliary information about the whole population from the Register of Employee and the Central Population Register are utilized.

Quarterly revisions . The figures from the first two months in the quarter are preliminary. They are re-estimated at the end of the quarter supplemented with the extra interviewees that we get after the preliminary deadlines and with better auxiliary register variables. For the re-estimation, also more up to date auxiliary information about the whole population from the Register of Employee are utilized.

Yearly revisions . After seasonal adjustment, the levels are corrected to make the annual averages of the seasonally adjusted figures equal to the corresponding unadjusted figures from the LFS. The annual averages of a year are not ready until we publish the 4 th quarter figures. At that point we use new factors of level correction. Until the 4 th quarter publication is available, preliminary level correction factors from last year have to be used. For the estimation of the 4 th quarter also more up to date auxiliary information about the whole population from the annual Tax Register are utilized.

Concurrent versus current adjustment

Partial concurrent adjustment
The model, filters and calendar regressors are re-identified once a year and the respective parameters and factors re-estimated every time a new or revised data becomes available.

Additive outliers are detected automatically and parameters re-estimated every time new data becomes available.

Horizon for published revisions

The entire time series is revised in the event of a re-estimation of the seasonal factors.

Evaluation of seasonally adjustment data

Continuous/periodical evaluation using standard measures proposed by different seasonal adjustment tools.

Quality measures for seasonal adjustment

For most of the series, a selected set of diagnostics and graphical facilities for bulk treatment of data is used.

For more information on the quality indicators in the table, see: metadata on methods: seasonal adjustment

Seasonal adjustment of short time series

All series are sufficiently long to perform an optimal seasonal adjustment.

As from 2006 the LFS was revised, among other things the lower age limit covered by the survey was lowered from 16 to 15. In order to avoid breaks in the series, we start all the time series in 2006.

Treatment of problematic series

Problematic series are treated in a special way only when they are relevant. The remaining series are treated according to normal procedures.

Due to large random variation (sample uncertainty) and small seasonal variation, we do not seasonally adjust all the main variables of the LFS gender divided. This is the case for unemployment, employment and for Employed persons man-weeks worked for persons 15-24 years. Afterwards, gender divided figures are constructed from seasonally adjusted figures by utilizing monthly gender distributions calculated from trend-cycle figures from additional unofficial seasonal adjustments of the LFS.

Data availability

Three months moving averages of seasonally adjusted data and trend are available.

All metadata information associated with an individual time series is available.

Historical data are available to enable revision analysis.

Easily retrieve and integrate Statistics Norway’s seasonally adjusted LFS data with your own systems with API.

In order to reduce the uncertainty, the presented series are three months moving averages of the seasonally adjusted figures. For instance the figure from September is the average of the estimates from August - October.

See previous releases for historical data.

Press releases

Previously we only published three month moving averages as the monthly results. They do not comply with the international demands for timeliness, but nevertheless are useful for several important national users. We therefore continue to publish these figues, in addition to seasonally adjusted monthly results, trends, and unadjusted monthly results.

The sesonally adjusted and the unadjusted monthly results are aimed at expert users. Both of the series are highly volatile, and are usually revised rather heavily at each new publication. Statistics Norway therefore advises to use the trends and the three month gliding averages. The trends show the long term development and the business cycle, whereas the three month moving averages are smoothed montly results adjusted for seasonal variations. The trends are more timely than the three month moving averages.

• The Norwegian Labour Force Survey sampling design
• Weighting methodology for the Norwegian Labour Force Survey from 2021 onwards
• Break estimation in the Norwegian LFS due to the 2021 redesign