Statistics Finland

Working life and wages and salaries

See below.

The labour cost survey data (LCS) for 2020 was collected from existing data sources and was without its own survey questionnaire. This production model was first time implemented in 2012 and similar production process was used in 2016 and 2020 with some modifications to the earlier production. 

The calculation of the LCS is based on the General Industrial Classification of Economic Activies (NACE). In the national publication data is classified also by sectors (private, central government and local sector).

LCS covers NACE rev. 2 sections B to S.

Concepts and definitions of the statistics are available on the website.

https://stat.fi/en/statistics/documentation/tvtutk#Concepts

The statistical unit is the enterprise.

LCS applies to all activities in sections B to S of NACE Rev.2 and represents all statistical units occupying 10 and more employees.

The concepts of Finnish LCS correspond to those of European. Instead of local units, enterprises are used as statistical units. In the national publications, data is provided by sectors (private, central government and local government sector).

The LCS is released every fourth year. The transmission to Eurostat take place within 18 months after the reference year.

The LCS was published in 1996 for the first time covering only some of the NACE sectors. Since 2008 it has covered all sectors - private with more 10 employees,  local and central government.

Not applicable.

See below.

 The main source of the LCS 2020 are:

• Labour cost index survey (LCI) – For information on wages and salaries and either paid or worked hours for enterprises with more than 20/30 employees
• Structure of earnings statistics (SES) – For information on wages and salaries and paid hours for enterprises with less than 20/30 employees
• Labour force survey (LFS) – For information on structure of working hours and different types of absences

   

The above mentioned surveys contribute to most part of the variables in the LCS:

• Number of employees (A.1-A.131)
• Hours actually worked (B.1-B.13)
• Paid hours (C.1-C.13)
• Wages and salaries (D.11)

Most of the LCS data is derived from LCI survey data, which is summed up to yearly data. In the data 46% of units and 85% of the cost sum comes from LCI data. 50% of units and 7% of cost sum comes from SES data. The rest (data for central government sector), contributes to 8% of the cost sum.

The LCI survey includes enterprises that has 30 employees or more, and within some NACE main classes the survey concerns those with 20 employees or more (in NACE classes construction F, wholesale and retail trade G, accommodation and food service activities I, real estate activities L, professional, scientific and technical activities M, and administrative and support service activities N). For enterprises with 10-20 or 10-30 employees the data from SES is used in LCS.

In addition to data on small enterprises, SES is used to evaluate structure of working time on all the enterprises. This is done by modelling the use of working time (hours worked, different types of absences, hours paid) from LFS data, and imputing that structure into individual level SES data. The imputed individual data is then aggregated into enterprises. The data for small enterprises is used as such and the SES data for enterprises with 20/30 employees or more is used to evaluate missing information on hours in the LCI data.

The local government sector survey inquiry of LCI differs from that of the private sector such that both the number of hours worked and hours paid along with absent hours total and hours of sickness and maternity/paternity leave are asked. Thus no modelling for structure of hours is needed for local government sector.

In addition to these surveys, a number of other statistics and register data was used. These data give information on social costs, vocational training costs and subsidies received by the employer as well as supplementary information for wages, salaries and working hours.

All of the variables concerning social costs (D.12-D.123) are based on register data, except for the variable “guaranteed remuneration in the event of sickness” (D.1221). The variable D.1221 is based on a combination of register data on compensation for expenses of sickness and maternity leave and LFS structure data on hours of sickness and maternity leave.

The vocational training costs for enterprises and local government sector an additional variable was added to the LCI questionnaire. The register data for government sector included vocational training costs.

The shares of items that fall under variable “other expenditure paid by the employer” (D.3) were estimated to be of the same share ratio to wages and salaries as in 2008 since these data were not available from existing sources. Except for central government where information was available in 2020. The share of other expenditure is (according to 2008 survey) less than 1 per cent of total costs.

The cost item “Taxes” (D.4) is irrelevant in Finland, since there are no employer taxes in use. The cost item “Subsidies received by the employer” (D.5) is based on register data.

The government sector data is based on administrative register “TAHTI” which is maintained by the State Treasury. The local government sector data is mainly drawn from LCI except for social costs.

The sources of data of LCS are summarized in Table 1 by sector and by size class of enterprises.

The source data is collected every fourth year for the product of the Labour Cost Survey.

The main sources of the LCS are:

• Labour cost index survey (LCI) – For information on wages and salaries and either paid or worked hours for enterprises with more than 20/30 employees
• Structure of earnings statistics (SES) – For information on wages and salaries and paid hours for enterprises with less than 20/30 employees
• Labour force survey (LFS) – For information on structure of working hours and different types of absences

In addition a number of other statistics and register data is used.  These data fie information on social costs, vocational training costs and subsidies received by the employer as well as supplementary infomation for wages, salaries and working hours.

The source data has been validated in the production process of each statistics. Therefore no systematic checks have been made for the unit data in LCS. Instead there have been made checks for relating to hours worked and paid and for elements for costs and validations for aggregated data in different phases of the production. 

Not available.

Not applicable.

See below.

The concepts and classifications of the LCS are defined in the Comission's implementing regualation.

The quality report is deliverd to Eurostat and evaluated by Eurostat.

The labour cost survey data (LCS) for 2020 was collected from existing data sources and was without its own survey questionnaire. This production model was first time implemented in 2012 and similar production process was used in 2016 with some modifications to the earlier production. Also 2020 the same production model was used with some modifications to the earlier production 

Main users of the statistics

The main users of statistics on labour costs in Finland are Eurostat and European commission. National users are employer and employee organizations and unions, their economists, public administration, organizations and enterprises purchasing or selling labour or financing diverse projects, enterprises and organisations practising cost monitoring, scientific researchers and the media.

When assessing labour costs, the users need information on the costs and incomes of their members or customers. In collective bargaining negotiations this information is used to estimate the real costs employers will incur from increases in earnings. The information can also be utilised in competitive tendering or in considering financial support for projects that have applied for it.

To their users the statistics provide information about the average level of labour costs. Information on the structure of labour costs can be used to assess the impact of e.g. diverse tariff changes on total costs and any possible effects on employment. The relatively long time it takes to obtain results after the reference year is a disadvantage of the LCS. Thus the more timely data of yearly labour costs, based on LCS and changes in LCI is needed.

National users require sector based data which is nationally published (in contrast to Eurostat's publication). The relatively long time it takes to obtain results after the reference year is a disadvantage of the LCS. Thus the more timely data of yearly labour costs, based on LCS and changes in LCI is needed. Nationally no yearly data is published.

Finland cannot deliver results with NUTS classification because data does not cover enough of units from Åland Island.

All except NUTS classified data.

See below.

The population of LCS 2020 is derived from business register. The sample is derived from respondents of labour cost index and structure of earnings statistics. There was no separate survey for LCS 2020. Being based on other statistics and registers, the number of enterprises included in the data was increased from LCS 2012 (there were over 4 500 answers from enterprises in 2012 and under 4 800 answers form enterprises in 2016 whereas in 2020 there is data on over 5 800 enterprises).

Data for LCS 2020 is derived from LCI survey data and SES data.

The coefficients of variation have been calculated by treating the main source of information (LCI or SES) as stratum. The coefficients of variation are calculated with breakdowns by NACE sections and size band. They are expressed in percentages.

The following tables include the coefficients of variation of annual labour costs and hourly labour costs, in total and by NACE section classes and size classes. The coefficient of variation in total is 0.6% for hourly labour cost and 4.1% for annual labour cost (Table 2).

The coefficient of variation of annual labour cost by NACE section (Table 3) varies from 5.7% in Manufacturing (C) to 29.3% in public administration and defence; compulsory social security (O). The coefficient of variation by size class (Table 4) varies from 2.8% to 9.5%.

The coefficients of variation of hourly labour costs (Table 5 and Table 6) are much smaller to those of total costs. They vary from 1.1% in Wholesale and Trade (G) to 4.4% in real estate activities (L) when compared by NACE sections (Table 5). The coefficients of variation of hourly cost broken down by size classes vary from 0.7% to 1.9% (Table 6).

The sources of non-sampling errors are listed in the subsections.

The study population of LCS 2020 is drawn from the Business Register. It includes enterprises with 10 or more employees in the private sector and all municipalities and joint municipal authorities in the local government sector.

According to the Business Register, about 24 per cent of employees and 17 per cent of wage sum is within the enterprises with less than 10 employees within NACE Rev 2 sections B-S. Thus the study population covers more than 75 per cent of all employees and more than 80 per cent of wage sum in the private sector.

The data of an enterprise in LCI was only used if there was data on at least two quarters of the reference year. Thus if there were enterprises that closed down operations after the first quarter of the reference year they were excluded from the LCS data.

Enterprises that were in the sample but were not included in the study population were given a weight of 1 in the estimation. These enterprises had either less than 10 employees but had nevertheless answered the inquiry or were missing from the study population since they had a lower number of employees in the Business Register (below 10) than in the LCI or SES inquiry.

The reference population and the sample population in the local government sector differs such that the municipalities with less than 5 000 inhabitants are left out from the LCI survey questionnaire and thus are not included in LCS either.

The central government data is derived from TAHTI register and includes data on all government sector employer organizations.

11% of the enterprises in the sample had less than 10 employees. Thus they are not a part of the target population which is enterprises with 10 employees or more.

Not applicable.

All of the surveys that are utilized in the production of LCS go through validations and quality checks. In the production of LCS only some variables were validated separately for the study. These were mainly the variables concerning the numbers of employees reported in LCI.

In the LCI questionnaire the numbers of employees are asked four times a year and the answer is based on the number at the end of each reference period. In enterprises with a lot of fixed-term employees or with notable changes in the numbers of employees the yearly average of these figures might not give good enough an estimate for number of employees. Therefore for all of the enterprises where both LCI and SES data existed, the number of employees was corrected to correspond the paid hours per employee in SES.

Each survey contributing to LCS 2020 is subject to non-response error. In LCS, it is assumed that the non-response in the main contributing surveys, in LCI and SES, is randomly distributed such that the non-respondents do not differ from respondents in terms of survey variables. The correction coefficients for LCS are calculated by taking an inverse of the ratio of wages and salaries sum of the respondents and wages and salaries sum in the population by industry and size classes.

The number of enterprises in the study population is about 20 700. There was 4 448 enterprises in the sample. The numbers of enterprises in the sample and in the population are presented in Table 7. The number of employees and ratios of wages and salaries sums are also presented in the Table 7. The number of units, employees and wages and salaries sum in local government sector are presented in Table 8.

Not applicable.

Not applicable.

The production model where multiple sources are combined together is subject to processing errors. The LCS 2020 production included combining individual level data as well as enterprise level data.

The most relevant variables subject to processing errors are the amount of hours worked, paid and absent. In the production process of Finnish LCS, the hours are estimated through a combination of responses in labour cost index, data in structure of earning statistics and labour force survey. This method of estimating the structure of working hours was tested against the 2008 survey data and was found to be reliable enough to be used in the production process of the so called register-based LCS.

In almost all of the cases where individuals and organizations were combined, there were unique identity codes for both individuals and enterprises used in combining the data together. Thus the information on unique combinations of employments could be reliably combined. This was important in order to link the individual level data into the right enterprise. If there was for example an individual working in two employments in 2020, the individual tax register data of wages could be reliably combined with SES data by using both individual and enterprise identity codes.

A possible source of processing error comes from combining individual level information of apprenticeship and SES. The data on apprentices includes the unique identity codes, employer and the time of apprenticeship. There is some uncertainty whether all the wages paid during the time of apprenticeship are in fact apprenticeship wages or whether these employees work already as full time workers and get paid accordingly. In the data there were about 10 000 apprentices of 1 000 000 employees.

In combining the register data with LCS data, a proportional share of the cost item to wage bill was used instead of nominal values. The obtained percentage was used to calculate the shares of the corresponding contributions of the wage sum reported in LCI or SES. Some of the units of an enterprise or some of the employees in the enterprise could be left out from the inquiries and therefore using the percentage ensured that the obtained result was correct.

Most important variable that was imputed was hours actually worked. The worked hours were imputed for about 80 per cent of the enterprises. The hours paid were imputed for about 20 per cent of the enterprises.

About 5 per cent of units could not be combined with register data on social security costs. These units were given industry specific mean values of social cost components.

The main data source, LCI, contains survey data of either paid or worked hours reported by enterprise. For hourly-paid employees the data of labour input is always bound to concern hours worked. In the case of monthly-paid workers the labour input is either hours worked or hours paid, depending on respondents’ choice. 

The missing data of hour of absence, sick leave, paid holiday and other paid leave is imputed for enterprises. The imputation method involves a regression imputation that uses LFS data to model working time. The estimates of working time are imputed on individual level data of SES and the individuals are then aggregated to enterprise level.

Modelling the working time from LFS has some challenges, mostly concerning how to restrict the LFS data when modelling and the question of how well the population of LFS corresponds to the population in SES. The working time estimates were tested with 2008 data and they were found to be very close to those gained from questionnaire in LCS 2008.

The regression model for the usage of working time in LFS is as follows:

where  states for hours worked per day for employee i in stratum j in LFS (T indicates LFS). Explanatory variables in the model are variables from LFS data, e.g. occupation, age, sex, working experience, overtime, part-time and different types of holidays. Variable  is a random error term with expected value of zero and a finite variance.

The working time structure is imputed into SES data for every employment, taking advantage of background information such as employees’ occupation, age, sex and working experience. After imputation of the working time, the SES data is aggregated into enterprise level and the amount of paid hours, holidays, sickness and maternity leave and other paid holidays are calculated for each enterprise.

There are three types of enterprises in the data:

1) enterprises of which there is information in both LCI and SES,

2) enterprises of which there is information only in LCI and

3) enterprises of which there is information only in SES.

In the first case, the structure of working time that is imputed in SES is used to calculate hours of absence and hours paid or hours worked on the basis of the reported hours in LCI. In the second case, the average structure of working time (on NACE 2-digit level) is used to impute the missing hours into LCI data. In the third case, the imputed absent hours and worked hours are used.  

For the enterprises where there exists both LCI and SES data the hours that an enterprise has reported in LCI is taken as the basis for hours. If an enterprise has reported paid hours in LCI, then the hours worked is being calculated by using the ratio a_i/w_i, where a_i equals all the absent hours in enterprise i in SES and w_i equals all the paid hours in enterprise i in SES. The paid hours in LCI are multiplied by this ratio to gain the number of hours absent and the number of hour worked.

For the enterprises where there is only LCI data available, the industry averages for the ratio a/w is used. When there is only SES data available for an enterprise, the paid hours in SES along with the imputed hours of absences are used as such.

Not applicable.

Not applicable.

Not applicable.

Not applicable.

See below.

The preliminary data was submitted to Eurostat on 29th of June 2022. The final data was submitted in December 2022. Revisions of the final data was sent to Eurostat in October and November. The revisions to final data was made because local government social security costs was corrected and there were inconsistencies relationships between variables.

The final results were published nationally on 19th December 2022. This was about 24 months after the end of the reference period of 2022.

Time lag was 18 months.

The final results were published nationally on 19th December 2022. This was about 24 months after the end of the reference period of 2022.

As there was no survey implemented for LCS 2020 the question of deadlines, recalls and follow-ups for respondents is irrelevant.

The main data source, LCI data was ready for use in the spring of 2021 when the index figures of the 4th quarter and thus the whole year of 2020 were collected and published. Another important source, the LFS data for 2020 was also ready in early 2021. However, as other sources, especially the SES data for 2020, were still unavailable, the processing of the data started only in early 2022. These three sources together lay the basis for all of the variables concerning numbers of employees, number of hours worked and paid, and wages and salaries. Other sources than these contributed to social costs and details of the wages and salaries data as well as vocational training costs and information on apprentices.

Some of the data are based on registers from other institutes than Statistics Finland, and these institutes were contacted with orders of the data in fall 2021 and in spring 2022. Some register data was available of 28th March 2022.

The main period for processing of the data was timed from January 2022 to June 2022. This processing included handling the data from different sources as well as combining them, testing and carrying out quality assurance.

the first delivery was sent on time.

See below.

The concepts of Finnish LCS correspond to those of European. Instead of local units, enterprises are used as statistical units. In the national publications, data is provided by sectors (private, central government and local government sector).

Not applicable.

Differing from previous years, the LCS 2012 was implemented on the basis of existing sources and not by a survey questionnaire as before. This could cause difficulties in comparison of the LCS 2012 with the previous years. The LCS 2020 was implemented with similar production model, so comparison between 2016 and 2012 should be less uncertain.

Before the year 2012, the labour cost surveys were conducted by survey questionnaires where the number of employees, hours worked and paid, and wage costs were surveyed. These data was complemented with register data. From 2012 onwards the LCS is compiled of over 20 different sources without a survey of its own.

The most important source of LCS 2020 is labour cost index data, which is the source for number of employees, hours (either worked or paid) and wage costs for enterprises over 20 or 30 employees, depending on branch. The wage costs and paid hours for enterprises with less than 20 or 30 employees is gained from structure of earnings data. An important part of gaining different absences and full profile of the working hours is using an imputation model where labour force survey data is modelled and values for different types of absences imputed in the structure of earnings data.

For about half of the enterprises with 20/30 employees, the number of hours worked is based on the number of paid hours reported by enterprise and imputation of structure of working time from the LFS model. For the other half of enterprises with 20/30 employees, the number of paid hours is based on the number of worked hours reported by enterprise. For all of the enterprises with 10-20/30 employees, the number of hours worked is based on paid hours in SES where the structure of working time is imputed using the modelled LFS working time.

In the surveys before 2012 the values of benefits in-kind were asked from the respondents. In the LCS 2012 and onwards the benefits in-kind are taken from tax register and are thus taxable values. Untaxable benefits in-kind where asked in extra questions at LCI inquiry 2020 to get all the costs of benefits in-kind.

Some cost items from 2012 onwards differ somewhat from previous labour cost surveys. Differing from the earlier years, the taxable values of stock options are included in labour costs from 2012. The holiday remunerations in LCS 2012 and 2020 are included in one-off items, whereas previously they were classified as wages and salaries for days off. Severance payments are also classified as one-off items whereas before LCS 2012 they were classified as social costs. These changes are due to classification of cost items in LCI. In LCI questionnaire these cost items are one-off items.

Incomes register has been used in LCI since 2020 and used in the production of LCS 2020. Enteprises' wage and salary costs have been calculated from the Incomes Register and pre-filled on the online forms. Working hours and employer's social insurance contributions are not reported to the Incomes Register. Enterprises must report this data in the data collection. Incomes register was used in LCS 2020 to replace a material previously obtained from the Finnish tax administration. Companies which were included in LCI and Stucture of earnings was picked first. Then personal income data was extracted for these companies. Change in the procedure makes comparisons with previous Labour Cost Surveys difficult.

As a difference to the previous LCS, social contributions are calculated from net amounts rather than gross amounts due to received data. In previous surveys wage earner's share of social contributions has been included in data. In previous surveys employers' share was calculated by substracting the average share of wage earners's from the total share (gross amount).

Not applicable.

Comparison of Labour Cost Survey and Labour Force Survey

Table 9 shows the hours actually worked according to LCS and LFS. The amount of hours actually worked is systematically greater in LFS than in the LCS except for Transportation and storage (H), Accommodation and food service activities (I), Arts, entertainment and recreation (R) and Other service activities (S). LFS figures are based on individual level data and include employees of all enterprises, whereas the small enterprises are excluded in LCS. The hours actually worked in LFS include unpaid overtime whereas in the LCS the worked hours do not include unpaid overtime. These differences contribute to differing values for hours actually worked in the two surveys – even though the new production model of LCS takes use of the LFS working time structure.

In total, the difference in hours worked is 4.5 per cent. The largest discrepancies are in Administrative and support service activities (N). The number of part-time employees in these industries and their different amount in the two surveys could have an effect on the differing figures along with the fact that there are a lot of small enterprises in these branches that are included in LFS but not in LCS.

Comparison of Labour Cost Survey and Structural Business Statistics

The second comparison concerns wages and salaries per employee in private sector according to Labour Cost Survey and Structural Business Statistics (SBS). For better comparison, the wages and salaries in LCS were calculated only for private sector enterprises.

SBS includes small enterprises and therefore it is natural that the wages and salaries are higher according to LCS where there are no enterprises with less than 10 employees. On the other hand, pay for sick leave is included in wages and salaries in SBS whereas in LCS it is classified as social cost. In total, the wages and salaries of SBS are about 93 per cent of those in LCS (Table 10).

The wages and salaries calculated per full-time equivalent employees (Table 11) differ somewhat from figures calculated per employee. The differences of LCS and SBS are smaller in the comparison of full-time equivalent employees within most of the NACE Rev 2 sections compared to the figures per employee (indicated by the ratios of SBS to LCS closer to 1 in table 11 in comparison to table 10). This indicates that there are differing amounts of part-time employees or differing processing of part-time employees in these two statistics.

Comparison of Labour Cost Survey and Labour Cost Index

The third comparison shows the differences of the average annual growth rates of Labour Cost Index and Labour Cost Survey for time period between 2016 and 2020 (see table 12). In total, the difference of growth rates is 0.7 percentage points which is bigger difference than survey's comparisons between 2012 and 2016. According to the LCS the hourly labour costs have risen slower than in the LCI.

The largest discrepancies between average annual growth rates in LCS and LCI are in Mining and quarrying (B), Electricity, gas, steam and air conditioning supply (D) and Administrative and support service activities (N). Of these, NACE classes B and N have a lot of enterprises with less than 20 employees. Part of these enterprises (those with 10-20 employees) are included in the LCS but not in the LCI.

Not applicable.

The table 13 compares the compensation of employees per employee in LCS and National Accounts (NA). In total the compensations per employee in the these two statistics are close to each other. On the other hand, there are large discrepancies, difference of over ten percentages, in some NACE sections: Mining and quarrying (B), Manufacturing (C), Accommodation and food service activities (I), Administrative and support service activities (N) and Public administration and defence (O). Differences are large especially in section B.

Generally the compensation per employee is larger in the LCS statistics, which is expected since LCS does not include small enterprises whereas NA does.

Not applicable.

See below.

None.

The most important tables of the publication have been made available in Statistics Finland’s online StatFin service on the Internet in Finnish, Swedish and English. The dates of the publications were published in the end of 2021 in the release calendar of Statistics Finland. The links to the publications are as follows:

The publications:

https://stat.fi/en/statistics/tvtutk

The online StatFin service:

https://stat.fi/en/database-tables?topic=all&statistic=tvtutk

Not available.

Not applicable.

No information on the results of LCS 2020 was sent to enterprises or other units that were included in the sample.

A description of the statistics is provided in three languages on the Statistics Finland’s website:

https://stat.fi/en/statistics/documentation/tvtutk#Basic%20data%20of%20the%20statistics

Concepts and definitions of the statistics are available on the website.

https://stat.fi/en/statistics/documentation/tvtutk#Concepts

A quality report of the LCS 2020 is included in the Statistics Finland publication on the Statistics Finland website in Finnish, Swedish and English.

https://stat.fi/en/statistics/documentation/tvtutk

Not provided.

Not available.

Because there was no direct survey involved cost and burden was not assessed.

See below.

Data is not published if there is small number of units .

If units (E2) <= 5 then data is not published.

None