Statistics Denmark

Earnings and absence

Statistics Denmark

Sejroegade 11

DK-2100

Copenhagen OE

The Structure of Earnings Survey (SES) 2010 results is based upon a comprehensive data material with more than 1.5 million records covering the public sector and private enterprises with at least 10 fulltime employ-ees. This quality report aims to contain a thorough review of the data in question, the statistical process and its relevance. Therefore, in accordance with the Commission Regulation (EC) No 698/2006 of 5 May 2006 implementing Council Regulation (EC) No 530/1999 regarding quality evaluation, this report contains the following subjects

• Relevance
• Accuracy
• Timeliness and Punctuality
• Accessibility and Clarity
• Comparability
• Coherence

Furthermore the results presented follow the Eurostat’s arrangements for implementing the Council Regulation 530/1999, the Commission Regulations 1916/200 and 1738/2005 on 24. November 2010 in order to fulfil the requirements for the SES 2010.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

-

The statistical process used by Statistics Denmark to collect earnings data is best described as a census. In accordance with the Act on Statistics Denmark the population covered is thus imposed the reporting of Earnings data. The population consist of the public sector and private enterprises with at least 10 employ-ees.

[Not requested]

[Not requested]

[Not requested]

[Not requested]

[Not requested]

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

[Not requested]

-

Statistics Denmark collects Earnings data from the public sector as well as from all private enterprises with at least 10 fulltime employees.

Potential user needs varies a lot especially regarding detailed information on earnings for a variety of DIS-CO-08 groups. DISCO-08 (Danish International Standard Classification of Occupation (only published in Danish) is the revised Danish version of the International Standard Classification of Occupation (ISCO-08). It contains a detailed 6-digit classification of occupation. Nevertheless it may still be impossible to extract specific occupations from the dataset, due to the fact that the DISCO- classification in spite of its detail includes some occupations which are closely related, into the same DISCO-classification.

For the purpose of the SES 2010 the DISCO-08 classifications is converted into ISCO-08 so as to meet the SES 2010 requirements regarding the classification of occupation.

The main users of the Danish Earnings Statistics are researchers, private business enterprises, ministries, counties, municipalities, national organisations (employer- as well as employee organisations) and interna-tional organisations.

The scale of coverage is at a level that is regarded as satisfying when it comes to user needs.

The 2010 Public image survey of Statistics Denmark showed that Statistics Denmark in general has a high degree of user confidence. 83 per cent thus trusted Statistics Denmark, while 4 per cent distrusted it. 94 per cent of the people who knew the institution well trusted it.

Read more about the Public Image Survey of Statistics Denmark on the following website: http://www.dst.dk/en/OmDS/maal-og-resultater 

[Not requested]

[Not requested]

The accuracy and reliability of the survey are both very good as the survey is a full-scale census covering all enterprises with more than 9 employees.

[Not requested]

The statistical process being a census leads to Statistics Denmark not calculating the coefficient of variation for the SES 2010.

The statistical process being a census leads to Statistics Denmark not calculating the coefficient of variation for the SES 2010.

Despite of being a census the accuracy is affected by different types and degrees of errors such as coverage-, measurement- and non-response errors.

The margins of statistical errors are especially linked to hours of work. Data reported on paid absence can thus be subject to inaccuracies. In addition to this, there may be errors in the periodic delimitation, which are essential to the compilation of hours worked as well as the agreed working time. However, efforts are continuously made to improve the data quality through feedback to enterprises and through update and improvement of the production systems.

The framework used in selecting the target population is Statistics Denmark’s Central Business Register. The Central Business Register contains information on all enterprises and local units in Denmark (covering both the private and public sector). Each enterprise is identified by means of an 8-digit registration number (CVR number), which is the same number used in the administrative registers operated by the tax authorities, etc. The local units are identified by a 10-digit local unit code number. The Central Business Register is continuously updated, which implies that there are only minor problems in achieving close to complete coverage. Information on firms and economic activity of the local units, geographical location and legal owner-ship is collected from the Central Business Register. Furthermore, the register contains information on total employment measured in full time units.

When analysing the range of principal economic activities used in the SES 2010 the results shows about 21,000 private enterprises with at least 10 full time employees in the target population. The amount of employees in the study population private enterprises covers 87 per cent of the amount in the target population private enterprises. A measurement of the public sector coverage has not been made.

There is no measurement of over-coverage of the survey. Only units with more than 9 employees are included, all other units are sorted out.

[Not requested]

The data collection system is designed to explore the potentials of the latest computer technology for efficient and rational collection and reporting of data. In this respect, a Standard for Electronic Statistical Reporting has been developed, which is used by computer agencies and respondents utilising standard systems and by business enterprises which have set up a wage system. In addition to this, Statistics Denmark has prepared an electronic questionnaire as an alternative method of reporting data on paper questionnaires. The use of paper questionnaires is, at the present time, only used to very limited extends.

The data reported to Statistics Denmark, which form the basis for the SES, have been collected annually since 1994, and since then Statistics Denmark has systematically contacted business enterprises in cases where they have reported inadequate or erroneous data to Statistics Denmark. This continuous contact with business enterprises has improved data reports.
The discrepancy between the target and study population can be attributed erroneous data and a small amount of non-response error. The erroneous data is filtered through the Structure of Earnings process flow illustrated in the figure below.

Process flow leading to Structure of Earnings Survey data

Validation process 1 ensures that the data of the study population fulfill some of the most basic data quality and coding requirements. After validation process 1, the amount of data has been reduced with 12.6 percent. During validation process 1, the most frequent errors causing deletion/elimination of data are the following:

• the number of employees in the enterprise/organization not exceeding 9 employees (3.9 percent)
• invalid or erroneous isco-coding (1.8 percent)
• invalid or erroneous coding of highest completed level of education (2.9 percent)
• negative and invalid value of overtime pay (2.6 percent)

The study population used for the SES 2010 is still erroneous despite validation process 1. This is due to the various limitations set by plausibility checks and conversions between formats which are implemented through the following SES 2010 production process. Of the study population after the first process of validation, 3 percent is deleted as a consequence of validation process 2. The table below outlines the six most common measurement and processing errors during validation process 2 as a percentage of the study population. It is important to keep in mind that a single record can contain more than one error and thus add to the percentage count of more than one type of measurement error. Furthermore table 3.3 only contain the most common errors. Due to these reasons it is not possible to summarize the percentages in order to reach the 3 per cent erroneous data.

Measurement and processing errors

 The most common type of error, B27=FT and (B32-B321)<130, mainly affects part-time employees in the public sector working in the municipalities and regions.

The Act on Statistics Denmark makes it statutory for enterprises with at least 10 employees to report earnings statistics to Statistics Denmark. If Statistics Denmark doesn’t receive data from a target population enterprise they are legally justified to impose a fine on the enterprise not reporting data. Nevertheless not all enterprises report their earnings statistics either consciously or do to reasons that exempt them from reporting. Almost all the units to which statistics Denmark has made contact, for the purpose of making use of the enterprise or local unit’s earnings and absence data, submit information at some point.

[Not requested]

[Not requested]

Although processing errors may occur, especially in relation to the transfer of data between systems, there has not been an overall measurement of the extent of processing errors.

There is no imputation done on observations, only on some of the variables included in the survey. The imputation rate of these variables vary.

Statistics Denmark collects earnings data for a full year when considering the private sector and on a monthly basis when considering the public sector. In order to combine the two data sources, the earnings data are put together in the Common Earnings register which primarily contain yearly figures.

As a consequence Statistics Denmark is not directly able to differentiate between different months. In order to provide EUROSTAT with the variables measured in relation to the reference month some assumptions must be made.

One employee – multiple records

Statistics Denmark collects data from salary systems used by the target population enterprises. If there is a change in the employee information (fx B23 (ISCO) or (B28)) a new collection of data considering that employee will be initiated. One employee may thus appear in the dataset multiple times within a given year.

As an example let’s say that an employee halfway through the year decides to change to another local unit within the same enterprise and work there for the rest of the year. Because of this change of local unit the employee will be counted twice in the dataset. In this example the two records will have the same grossing up factor and the number of weeks to which the gross annual earnings relate (B31) is 26.07 weeks in both cases.

When measuring the amount of people working less than 30 weeks in the reference year the person in the above mentioned example will count twice. The data from Denmark should therefore not be used in such a comparison between countries.

Reference month

A common benchmark is the reference month which is constructed on basis of the employees working period.

The working period of the employee to which the gross annual earnings relate are given in variable B31 (number of weeks to which the gross annual earnings relate). This variable is constructed on the basis of information in the earnings data. B31 is then used when calculating the number of months in which the gross annual earnings relate by applying the following formula

[3.1]       

where 4,345 is the average number of weeks per month in a full year (52,14/12).

By applying this formula some observations in the dataset have a [3.1] which is less than 1. This can cause problems when calculating various variables. This will be outlined below.

Number of hours paid during the reference month (B32)

In the Common Earnings Register (target population) data the number of hours paid are given in yearly amounts. In order to provide the number of hours paid during the reference month, the yearly amounts are split equally throughout the year of employment, hence

[3.2]       

When the earnings data and number of paid hours per year is related to a working period of less than a month (3.1<1), then B32 may turn out to be greater than “paid hours per year”. In these cases B32 reflects the number of paid hours during the reference month as if the person has worked a full year.

Earnings related to overtime in the reference month (B421)

The earnings related to overtime are given in yearly amounts in the Common Earnings Register. In order to calculate the earnings related to overtime in the reference month (B421) the following method has been used

[3.3]        

The same implication which may arise when calculating B32 by dividing paid hours per year with [3.1] may arise when calculating earnings related to overtime. Hence, monthly earnings related to overtime may in some cases exceed the yearly earnings related to overtime (only if 3.1<1).

Special payments for shift work in the reference month (B422)

The earnings related to special payments for shift work are given in yearly amounts in the Common Earnings Register. In order to calculate the earnings related to overtime in the reference month (B422) the following method has been used

[3.4]         

The same implication which may arise when calculating B32 by dividing paid hours per year with [3.1] may arise when calculating earnings related to special payments for shift work. Hence, monthly earnings related to special payments for shift work may in some cases exceed the yearly earnings related to overtime (only if 3.1<1).

Full-time or part-time employee (B27)

There exists no information about the degree of full-time or part-time employment ship in the Common Earnings Register. In order to give a meaningful indication of this and in order to calculate the percentage share of a full-timer’s normal hours (variable B271) Denmark has adopted the following assumption:

[3.5] A person is:

Full-time employed if hours worked per week ≥ 29, 91

Part-time employed if hours worked per week <29, 91

Collective pay agreement (A15)

Statistics Denmark collects data for a sample called Other Labour Costs. The data in this sample are being used to estimate the degree of collective pay agreement for the local unit. This is possible because the sample is randomly drawn from a representative population representing the enterprise structure in Denmark.

Firstly the following conditions apply. Local units and enterprises which are members of either DA (The Confederation of Danish Employers) of FA (The Danish Employers’ Association for the Financial Sector) are all covered by a collective agreement at an industry level (code B). This is also the case with local units and enterprises in the public sector.

The remaining earnings data primarily collected by Statistic Denmark has to be subject to estimation of variable A15 on the basis of information from the Other Labour Cost survey.

In order to estimate the degree of collective pay agreement for the remaining data the following assumptions are made

1. The sample records are grouped according to local unit size and economic activity (NACE rev. 2).
2. This information is then matched with the data in the study population which holds no information regarding collective pay agreement, based on local unit size and economic activity (NACE rev. 2).
3. The information in the sample is then transferred to the study population in order to give a representative picture of the degree of collective payment agreement.

[Not requested]

[Not requested]

[Not requested]

[Not requested]

[Not requested]

-

The public sector earnings data for 2010 was collected on a monthly basis beginning January 2010 ending December 2010. There is a 1 month time-lag when collecting and receiving public sector data. Hence, data concerning January 2010 is available February 2010.

[Not requested]

[Not requested]

Public Sector (Local and central government)

Since the earnings data only contain information concerning earnings, the time span for collecting the absence data in the public sector is a bit broader. The absence data for 2010 was received on June 2011. The total time for public sector data collection is thus 18 months.

The first results based on the public sector earnings data was published on December 2011.

Processing of the earnings data for the public sector including quality checks, plausibility checks etc. amounted to 6 months.

The private sector data collection process consists of 2 modules. The first relates to the earnings and absence data for 2010 and the other relates to the “Other Labour Costs” data which contain information re-garding the degree of collective pay agreements.

The 2010 earnings and absence data collection started December 2010 stretching 6 months. In cases where a respondent failed to report the data, extra time was given to transmit the data. If a respondent failed to supply Statistic Denmark, DA or FA with earnings data within this additional time, the enterprise was reported to the police.

The first publication based on the private earnings data was published on December 2011.

Private Sector

Processing of the earnings data for the private sector including quality checks, plausibility checks etc. amounted to 6 months.

Regarding the 2010 “Other Labour Costs” data the time span for data collection was February 2011 to August 2011. The information gathered here is not used in the national statistics on earnings in the public or private sector.

When the earnings, absence and Other Labour Cost data was all gathered the Common earnings register could be constructed. The data collection process for this register amounted to 20 months (January 2010 to August 2011).

Common Earnings Register and SES

The processing of the 3 sources for the Common earnings register 2010 including quality checks, plausibility checks etc. amounted to 3 months. This included the validation process 1 (see figure Process flow leading to Structure of Earnings Survey data in chapter 5.3.2).

Since the Common earnings register 2010 forms the basis for the Structure of Earnings survey 2010 the total processing time of data amounted to 6 months, since the SES 2010 data was delivered to EUROSTAT on time (before the end of June 2012).

[Not requested]

-

-

The regulation demands information on grossing-up factors for the employees and for the local unit. Statistics Denmark has compiled the grossing-up factors with respect to the number of employees in the company/enterprise as a whole and not the local unit. Each local unit is still supplied with a unique identification key, but it is supplied with the same grossing-up factor as the rest of the local units belonging to the respec-tive company/enterprise.

The reason for the above is that we are not in control of the single local unit to the same extent, as if we collected all the information directly from the company or the local unit. When we get information from a company on their employees and their local units, we have to match the information with our central business register to get information on NACE-section, region and the number of persons they actually employ. We need the number of employees to be able to gross up to the entire population. However, in the case of Denmark two sources of errors exist.

Firstly the information from the companies in relation to the local unit identification number and secondly the information from the business register can be erroneous, which can cause problems in compiling the grossing-up factor.

The relation (number of employees in the population)/(number of employees in the sample) will in a few cases be high and in other cases below 1. These cases are not an expression of how it really appears in the practice but just an expression of internal technical match problems with the business register.

In addition to the above there is a small discrepancy between the ISCO-08 and DISCO-08 classification. The ISCO-08 codes 631, 632, 633 and 634 do not have a matching DISCO-08 classification. The DISCO-08 codes 342, 622 and 962 has not been included in the SES 2010 because they are not a part of the list showing which ISCO-08 codes that’s to be a part of the SES 2010 according to Eurostat’s arrangements for implementing the Council regulation 530/1999, the Commission Regulations 1916/2000 and 1738/2005.

[Not requested]

The SES is conducted by using annual earnings data collected by Statistics Denmark. It is thus possible to conduct comparisons between the levels of earnings on an annual basis, taking into account the structural changes that may occur between the years. For example, industrial changes, but also the general change in labour force composition and classification of occupation has a great impact on comparability over time.

Specific changes in conducting the annual Danish 2010 earnings statistics:

In 2010 multimedia are included in the fringe benefits.

New classification of occupations:

Classification of employees by occupations is conducted on the basis of the nomenclature DISCO-08, which is the Danish version of the ILO's official nomenclature for occupations IS-CO-08. DISCO-08 is a revised version of the previously used nomenclature DISCO-earnings. Subsequently, DISCO-08 is used for the first time in the structural statistics on earnings.

There is no homogeneous conversion key between DISCO-earnings and DISCO-08. Consequently, earnings distributed by occupations are not comparable backwards in time.

Regions and municipalities:

New information on the payment of the 6th holiday week for 2010 showed a significantly underreporting of the utilization of the 6 holiday week. The new information forms the basis of the establishment of data on absence, which was conducted in 2010. This implies a higher level of absence, which contributes to a higher level of earnings per hour worked in 2010.

Regions and municipalities:

The special holiday rate has changed from 1.5 per cent in 2009 to 1.95 per cent in 2010, contributing to a higher level of holiday and public holiday allowances. The special holiday rate is a basic rate. For certain groups of employees, a higher rate than the basic rate is used. Unlike earlier years, this has been taken into account from 2010, which has further resulted in a higher level for holiday and public allowances in 2010.

[Not requested]

This section contains a comparison analysis between two sources of structural statistics, Structure of Earnings Survey (SES) and the National Accounts (NA) regarding earnings. The variables under consideration are:

1. Average gross annual earnings in the reference year (2010), expressed per employee.
2. Wages and salaries per employee according to the NA (2010).

A comparison must be made between the two variables broken down by NACE sections. The Average gross annual earnings (variable B41) cover remuneration in cash and in kind paid during the reference year before any tax deductions and social-security contributions payable by wage earners and retained by the employer. More specifically variable B41 contains the following:

B41= Remuneration in cash (wages)

+ Earnings related to overtime
+ Bonuses and allowances
+ Payments in kind
+ Special payments for shift work
+ Payments made by employers to their employees saving schemes

In order to derive the SES 2010 Average Gross annual earnings (AGAE), the following formula has been used:

          , where i = individual/employee

, where j = NACE Rev. 2

The number of employees in the SES is given by the grossing up variable (B52). This is a rough measurement since in some cases the grossing up variable can be very high as mentioned in section 8.1.

The NA Wages and salaries include any social contributions taxes etc. payable by the employee (National Accounts (a)).

Wages and salaries come in cash and in kind. Wages in cash consists of regular wages plus i.e. commissions, overtime payments, bonuses, payments on public holidays and payments on other holidays. Social contributions, income taxes etc. which fall on the employee are included even when they in practice are kept back for direct payment to relevant authorities by the employer.

Wages in kind – fringe benefits – consists of products which are provided freely or to reduced price by the employer to the employee as part of the conditions of employment. Fringe benefits are not necessary in the production process. If they were, they should be treated as intermediate consumption (National Accounts (b) – Sources and methods (2003, p. 169)).

The NA employment figures contain persons supplying their labour in the production of goods and services in Denmark regardless of their place of residence. The employment figures reflect the average number of employed persons in the course of the year. This number only depends on the employment rate of the persons employed and does not reflect decisions to work full time or only part time. Employment consists of persons above the age of fourteen earning a salary equivalent to at least 80 hours of work within one year as well as persons temporarily absent from their work but still have a formal attachment to the job, e.g. persons on maternity leave. Only the primary job is reflected in the statistics. (National accounts (b))

In order to derive the NA 2010 Wages and Salaries per employee, the following formula has been used:

, where j = NACE Rev. 2

The SES and NA results along with a comparison are shown in table 6.1 below. The table contains a description of the AGAE from the SES, the Wages and Salaries from the NA (both measured in Danish Kroner) and their corresponding number of employees broken down by NACE rev. 2 main sections.

Table 6.1 shows an overall negative 16.8 percentage discrepancy between the SES and NA annual earnings. The discrepancy ranges from minus 44.4 per cent (NACE section L) to 6.3 per cent (NACE section O). The smallest discrepancy (minus 0.6 per cent) occurs in NACE section J.

The overall coherence between the SES and NA thus shows a negative bias.

Annual Earnings comparison between SES and NA

(1) Number of employees

(2) Average gross annual earnings in the reference year (AGAE) (in DKR)

(3) Wages and salaries (in DKR)

(4) Percentage difference between the yearly earnings in NA and SES

The coherence illustrated above gives rise to the question of why the discrepancy is so relatively high.

The SES does not include social contributions paid by the employer. This fact however, should pull the discrepancy in a positive direction and therefor actually minimize the negative discrepancy.

The discrepancy is also affected by the large portion of data being filtered through plausibility checks. Assuming that part time employees have a lower wage than full time employees the difference will decrease if some plausibility checks are left out. This is indeed the case according to the results shown in table 6.2 where the following plausibility checks regarding part time employees has been left out

“If var.2.7. = PT, then 130*(var.2.7.1/100) < (var.3.2 – var.3.2.1) < 215*(var.2.7.1/100)”

The overall discrepancy between the NA and SES 2010 will in this case decrease from minus 16.8 per cent to minus 5.6 per cent.

Annual Earnings comparison between SES and NA (when a PT plausibility check is left out)

(1) Number of employees

(2) Average gross annual earnings in the reference year (AGAE) (in DKR)

(3) Wages and salaries (in DKR)

(4) Percentage difference between the yearly earnings in NA and SES

Bibliography

Statistics Denmark (2003), Danish National Accounts – Sources and Methods 2003:

http://www.dst.dk/pukora/epub/upload/11968/entire.pdf

National accounts;

a) http://www.dst.dk/Site/dokumentation/Declarations/compensation-of-employees--%20%20employment-and-hours-worked.aspx

b) http://www.dst.dk/pukora/epub/upload/11968/entire.pdf

[Not requested]

[Not requested]

[Not requested]

[Not requested]

-

The most detailed statistics on the structure of earnings are published in Danish in “Løn” (Statistics on earnings) appearing in the series “Statistiske Efterretninger” (Statistical News).

As the SES is close to identical to the annual national survey of structural earnings – except for a slightly different combination of variables, the SES is not published.

The main users of the Earnings statistics are the national (Employee organizations and employers´ associations) and international organisations, ministries, municipalities, counties, private business enterprises and they obtain news about upcoming statistics through “Newsletters” from Statistic Denmark which is supplied via E-mail. Alternatively information about upcoming and new statistics can be found on Statistic Denmark’s homepage.

Key figures are published once a year in “Nyt fra Danmarks Statistik” (News from Statistics Denmark) and in “Statistisk Årbog” (Statistical Yearbook), both Danish. 

The structure of earnings data is available in English via StatBank Denmark. The 2006 data was split in public and private earnings statistics while the 2010 earnings statistics combines the sectors and present them in the same tables.

The key series are: (Earnings; SLON10, SLON21, SLON30, SLON40, SLON50).

[Not requested]

Authorised research environments and analysis institutes may be given access to the Common Earnings register. Access is always given on a need-to-know-basis to no identifiable micro data in accordance with Statistics Denmark’s special external researchers’ scheme. We grant such authorisation pursuant to a concrete assessment, and the authorised researchers have the same duty of confidentiality as Statistics Denmark staff members. For educational use, we supply non-confidential, sample survey-based datasets that are constructed in such a way that it is not possible to identify persons or businesses.

If a greater level of detail or tabular cross-tabulations is required, they can be produced on request. The register is at the level of individual employees and may be used in connection with compiling more detailed statistics or in coupling data from other statistics.

The Structure of Earnings Survey 2010 is based on the Common Earnings Register. Information on formation of the Common Earnings Register, what variables it contains and on what grounds they are produced can be found in the subsection “Documentation” on Statistic Denmark’s homepage. 

Statistics Denmark (2011), Danish International Standard Classification of Occupation 2nd edition (Danish ver.):

http://www.dst.dk/da/Statistik/dokumentation/Nomenklaturer/DISCO/disco08/loenstatistikkken.aspx

Structure of Earnings statistics;

http://www.dst.dk/en/Statistik/dokumentation/Declarations/structure-of-earnings.aspx

Total Labour costs;

http://www.dst.dk/en/Statistik/dokumentation/Declarations/total-labour-costs.aspx

[Not requested]

[Not requested]

[Not requested]

[Not requested]

-

[Not requested]

[Not requested]

There are no other comments to the information on the survey.