Statistics Netherlands

Demographic and socio-economic statistics – Team labour and wages

Statistics Netherlands | Henri Faasdreef 312 | P.O. Box 24500 | 2490 HA The Hague

Statistical figures on wages, paid hours and other related indicators of employees which can be broken down to characteristics of the employees and their employers.

Statistical figures on wages, paid hours and other related indicators of employees which can be broken down to characteristics of the employees and their employers.

Nace (rev.2) sectors B to S are included in the Dutch 2018 Structure of Earnings Survey.

The Dutch Structure of Earnings Survey 2018 was based on a combination of integral register data and sample survey data. The end result of this combination yielded a sample, which needed to be weighted to obtain figures for the total population.

Two interrelated statistical units were observed.

1. Jobs of employees
2. Enterprises (instead of local units, but broken down to geographical locations)

All jobs of employees who worked in Nace (rev.2) sector B to S in the Netherlands in October 2018 and were subject to wage taxes in the Netherlands.

The Netherlands

The Dutch Structure of Earnings Survey vintages 2002, 2006, 2010, 2014 and 2018 can be found in the online Eurostat database.

The Dutch Structure of Earnings Survey vintages 2002, 2006, 2010, 2014 and 2018 can be found on the online Eurostat database.

The 2018 SES of the Netherlands is not based on a single questionnaire but on a combination of multiple sources:

• The Statistic on Employment and Wages (SEW);
  • In Dutch: Statistiek Werkgelegenheid en Lonen (SWL).
  • Key variables used: wages, paid hours, job characteristics, enterprise characteristics
  • Description: The SEW is a statistic created yearly by SN, based on integral register data, namely the ‘Register of persons insured under employee insurance schemes’ (in Dutch: Polisadministratie). In addition, this register is enriched by a number of variables (economic activity, geographical location, size of the enterprise) that were taken from SN’s General Business Register (GBR; in Dutch: Algemeen Bedrijfsregister, ABR), which in turn is based on register data received from the Dutch Chambers of Commerce.
  • Frequency/availability: reference year + 10 months
• The Labour Force Survey (LFS);
  • In Dutch: Enquête Beroepsbevolking (EBB).
  • Key variables used: occupation, job characteristics
  • Description: The LFS is a household sample survey on labour, continuously conducted by SN.
  • Frequency/availability: reference quarter + 6 weeks
• Educational Attainment File (EAF);
  • In Dutch: Bestand met hoogst behaald en hoogst gevolgd opleidingsniveau en opleidingsrichting.
  • Key variables used: educational attainment
  • Description: The EAF represents the highest attained education level of individuals at reference date October first. The educational attainment is derived by combining information on education from registers and from the Labour Force Survey conducted by Statistics Netherlands. As there is no information on educational attainment for every citizen in the Dutch population, the EAF consists of a mixture of register and sample survey (LFS) records.
  • Frequency/availability: reference quarter + 6 weeks
• The population register (PR);
  • In Dutch: De Basisregistratie Personen (BRP).
  • Key variables used: demographic characteristics
  • Description: The PR is built from the municipal population registers. SN furthermore complements this data with data from the Registration for Non-Inhabitants (RNI; in Dutch: Registratie voor Niet-Ingezetenen, RNI).
  • Frequency/availability: reference year + 11 months

The above sources contain microdata, which could be linked together (on person-level (*)). The sources were all edited and processed by SN, prior to their use for the SES 2018, specifically to make statistics.

The integral SEW 2018 served as the basis for the Dutch SES 2018. Many of the core variables for the SES, such as hourly earnings, were derived from the SEW 2018 on person-level. This data was enriched by the EAF, to derive the educational attainment on person-level. This data in turn was micro-linked to three vintages of the LFS (2018, 2017 and 2016) to derive the occupation on person-level. Because the LFS is a sample survey, the end result of this exercise also yielded a sample on person-level, with weights. (Three vintages of LFS were needed to derive a satisfactory sample size.)

*) All microdata on person-level at SN is stored pseudonymized. This makes it possible to micro-link many different social-economical datasets, while ensuring the highest standards of confidentiality.

See 3.1 Source data

See 3.1 Source data

The used sources were all edited and processed by SN, prior to their use for the SES 2018, specifically to make statistics. See also 3.1 Source data.

The weighted results regarding earnings and hours for different categories were compared to the results of the (integral) SEW and deemed of good quality. Regarding educational attainment and occupation the results were compared to other existing statistics by SN on these topics and deemed of good quality as well.

During the process of compiling the SES, several technical checks are built in to prevent technical (non-statistical) errors.

Lastly, the outcomes in the form of microdata were checked for internal consistency, using both guidelines provided by Eurostat and extra additional guidelines developed by SN.

See 3.1 Source data

Three vintages of LFS were used to derive a sample size large enough to fit the standards. This does mean however that the variable ‘occupation’, which is based on the LFS, might be outdated in regards to the corresponding earnings and hours.

The LFS is conducted only on persons in the PR. However, in our population, which is determined by the population of the SEW, there is also a group of people who are not registered in the PR (this concerns a small group of non-inhabitants who worked for a short time in the Netherlands). This was accounted for during the weighting process by incorporating an extra step: the use of ‘starting weights’. This way, this bias could be accounted for in more detail than when just the weighting scheme would have been used.

See 3.1 Data sources, 3.4 Data validation and 6.3.2. Measurement error.

The Dutch SES 2018 is deemed to be high quality. The input sources were of high quality and specifically suited for statistical processing. The outcomes have been analyzed extensively and have been compared to relevant other statistics produced by SN. Also see 3.1 Data sources, 3.4 Data validation, 6.2.1 Sampling error and 6.3.2. Measurement error.

The Dutch SES 2018 did not have significant deviations from Eurostat’s implementing arrangements for the SES.

SN are not in direct contact with national core users. The outcomes of the Structure of Earnings Survey (SES) are not published in SN’s online database ‘Statline’, because they are already published by Eurostat and there does not seem to be a distinct need by users for a separate Dutch version.

The Dutch SES 2018 did not have significant deviations from Eurostat’s implementing arrangements for the SES. Furthermore, see 5.1 User needs.

In general, the Dutch SES 2018 did not have significant deviations from Eurostat’s implementing arrangements for the SES. Furthermore, see 5.1 User needs.

However, one remark is in place here. Regarding variable 4.2.2 ‘Special payments for shift work’, SN unfortunately is not in possession of reliable and timely source data. For the SES 2014, SN conducted a research in order to make estimates for the heights of this variable, extensively assessing collective agreements between employers and employees and other related sources. The results of this research have been extrapolated to the SES 2018 with respect to developments in wages. There were no significant changes in the collective agreements between employers and employees regarding this subject between 2014 and 2018, so SN concluded respecting the developments in wages would be sufficient.

100%

Overall, the Dutch SES 2018 can be considered highly accurate. However, certain breakdowns do show a somewhat bigger covariance. See 6.2. Sampling error.

See underlying

Coefficients of variation (CV), standard deviation (s, square root of the variance) and mean are listed in tables 1 and 2. In table 1 for variable B42 ‘Gross earnings in the reference month’ and in table 2 for variable B43 ‘Average gross hourly earnings in the reference month’. The key-variables, B42 and B43, are broken down to:

- full-time (separately for men and women) and part-time employees,
- economic activity (NACE Rev.2),
- occupation (ISCO-08 at the 1-digit level),
- age band,
- geographical location (NUTS level 1),
- highest level of education (ISCED 2011),
- size of enterprise.

Note: the figures shown in tables 1 and 2 are based on all company size classes, including 1 to 9 working persons. The data shown in the Eurostat database do not account the company size class 1 to 9 working persons.

See underlying

See underlying

There is no under- or over-coverage in the Dutch SES 2018.

With regards to under-coverage, it can be noted that the LFS, which determines the final sample, is conducted only on persons in the PR. However, in our population, which is determined by the population of the SEW, there is also a group of people who are not registered in the PR (this concerns a small group of non-inhabitants who worked for a short time in the Netherlands). This bias however was accounted for during the weighting process by using extra starting weights.

The Dutch SES 2018 sample is a result of micro-linking various sources (also see 3.1 Data sources). Records/persons that did not have an entry in one or more of these sources were discarded before estimating the weights.

Three vintages of LFS (2018, 2017 and 2016) were used to derive a sample size large enough to fit the standards. This does mean however that the variable ‘occupation’, which is based on the LFS, might be outdated in regards to the corresponding earnings, hours and educational attainment, which in turn were based on register data regarding the year 2018.

Usually, official register data are of a high quality. Nevertheless, administrative errors do occur. When registers are processed at SN, for instance to create the SEW, extensive analyses and editing rules usually are applied. Many automatic rules are applied to check and correct for inconsistencies at micro-level. Additionally, the data is analyzed by researchers using a top down method. This ensures the resulting data is of very high quality on meso and macro aggregate levels and usually also on micro-level. However, some administrative errors on micro-level might not get caught in the processing and still persist in the end result.

See underlying

In the already edited sources used for the SES, unit non-response is not an issue. In the register data it is usually not a significant issue to begin with.

In the already edited sources used for the SES, item non-response was not an issue, except for one variable: 4.2.2 ‘Special payments for shift work’. SN unfortunately is not in possession of reliable and timely source data regarding this variable. For the Dutch SES 2014, SN conducted a research in order to make estimates for the heights of this variable, extensively assessing collective agreements between employers and employees and other related sources. The results of this research have been extrapolated to the SES 2018 with respect to developments in wages. There were no significant changes in the Dutch collective agreements between employers and employees regarding this subject between 2014 and 2018, so SN concluded respecting the developments in wages would be sufficient.

See underlying

Because the Dutch SES 2018 was created from sources that were already edited for statistical processing, no imputations were done during this research, except for the variable 4.2.2 ‘Special payments for shift work’, which was imputed completely based on previous research. See also ‘6.3.3.2 Item non-response – rate’.

Three vintages of LFS (2018, 2017 and 2016) were used to derive a sample size large enough to fit the standards. This does mean however that the variable ‘occupation’, which is based on the LFS, might be outdated in regards to the corresponding earnings, hours and educational attainment, which in turn were based on register data regarding the year 2018.

No seasonal adjustment was performed.

The Dutch SES 2018 data are deemed definitive, as Eurostat does not impose a fixed revision policy. Furthermore, because the research was based on pre-processed sources that were also deemed definitive, there is no need for a fixed revision policy. In case significant errors would come to light in the future, SN might consider revising in consultation with Eurostat.

N/A

N/A

See underlying

The reference period for the Dutch SES 2018 is October 2018 for the variables describing a month and the year 2018 for the variables describing a year. The Dutch SES 2018 was first submitted to Eurostat on 30 June 2020. Afterwards, a technical error was corrected and the data were resubmitted on 2 October 2020. The results were published by Eurostat on 13 October 2020. Therefore, the time-lag is roughly two years.

See 7.1.1 Time-lag first result.

See underlying

Eurostat’s deadline of 30 June 2020 was met. However, a technical error needed to be corrected which resulted in a resubmission on 13 October 2020. Also see 7.1.1 Time-lag first result.

See underlying

For most countries, the SES is a survey targeted on and describing local units. However, the information about units in the Dutch SES 2018 does not refer to local units, but to enterprises (containing all the employees) at NUTS-1 level.

For the difference or the absolute difference of inbound and outbound flows between a pair of countries divided by the average of these two values, please refer to Eurostat, as Statistics Netherlands does not have this information.

See underlying

The Dutch Structure of Earnings Survey vintages 2002, 2006, 2010, 2014 and 2018 can be found on the online Eurostat database. These vintages are largely comparable over time, however it needs to be noted that (re)sources and methods to compile the SES have changed slightly over time. Trend breaks between two following vintages have been analyzed, but trend breaks over longer periods have not.

SES vs. Labour Cost Index

SES 2018/SES 2014 growth rates of the mean hourly earnings for all (part-time and full-time) employees including apprentices, working in an enterprise of 10 employees or more and belonging to NACE Rev.2 sections B to S excluding O were compared with the corresponding 2018/2014 growth rates of the annual LCI (dataset: lc_lci_r2_a in Eurobase, variable ‘Wages and salaries (total)’, NACE Rev. 2 B to S). See table 3.

Differences in growth rates between the two domains do exist, although they do not surpass 7 percentage points.  These differences can be explained by the following factors:

-          The mean hourly earnings of the SES relate to October 2014 and 2018, while the LCI relates to the whole years 2014 and 2018.

-          The LCI accounts for structural effects (changes in the population over time), the SES does not.

-          The SES hourly earnings are based on a sample taken from integral register data (SEW). The LCI is based on the full population of the same SEW.

-          The SES does not include enterprises with size class < 10 employees, the LCI includes all size classes.

Some research has been done to explore whether one of these factors could be the main explanation for the observed differences. This seems not to be the case. The four factors seem to interact with each other which makes it difficult to give a weight to each one of these factors without doing extensive research. Such a research however is beyond the scope of this Quality Report.

SES vs. Labour Force Survey (employees)

The 2018/2014 growth rate in the number of employees taken from SES, for all (part-time and full-time) employees including apprentices, working in an enterprise of 10 employee or more and belonging to NACE rev.2 sections B to S excluding O was compared with the corresponding 2018/2014 growth rate from LFS (dataset: lfsa_eegaed in Eurobase, variable: ‘Total number of employees’, all ISCED levels, age 15 to 64). See table 4.

Furthermore, the 2018/2014 growth rate in the number of hours paid taken from SES for all (part-time and full-time) employees including apprentices, working in an enterprise of 10 employees or more and belonging to NACE rev.2 sections B to S excluding O was compared with the corresponding 2018/2014 growth rate from LFS (dataset: lfsa_ewhun2 in Eurobase, variable: ‘Average number of usual weekly hours of work in main job’, employees, total worktime, all NACE sections). See table 5.

Note: The number of hours for the SES depicted in table 5 differ from the number of hours depicted in the Eurostat database. The numbers in the Eurostat database are corrected for the share of a full-timer’s normal hours (variable 2.7.1). The numbers depicted in table 5 are not.

It is assumed that the hours paid/hours worked ratio is broadly stable over time so that changes in hours paid (taken from SES) can be compared with changes in hours worked (recorded in LFS).

Table 4 shows a difference of 4 percentage points for the growth factors of the total number of employees in the Netherlands. The highest absolute difference on sector level depicted in table 5 is 4 percentage points (sector M).

The differences in the growth factors between the SES and the LFS can be explained by the differences in the methodologies of the two domains:

-          Different sources:

• The LFS is based completely on a sample survey. The SES is based on a combination of sources (see 3.1 Source data).

-          Different populations:

• The LFS numbers represent employees aged between 15 and 64. The SES numbers contain all ages.
• The LFS describes Dutch resident citizens only. The SES describes all employees working in the Netherlands, including non-resident employees, as long as their wages are subject to Dutch wage taxes.
• The LFS does not count institutionalized employees (e.g. working prisoners). The SES does describe this group, as long as their wages are subject to Dutch wage taxes.
• The LFS literally counts the number of employees. The SES actually counts the number of ‘jobs of employees’. One employee can have multiple jobs or even have a job < 1 when the job did not exist for the entire statistical period.

-          The SES describes October 2018, the LFS describes the whole year 2018.

SES vs. Structural Business Statistics

Both SES2018 levels and SES 2018/SES 2014 growth rates of the mean annual earnings, for all (part-time and full-time) employees including apprentices, working in an enterprise of 10 employee or more and belonging to NACE Rev.2 sections B to N excluding K, were compared with the corresponding Structural Business Statistics (SBS) data (dataset: sbs_na_sca_r2 in Eurobase, variable ‘Wages and salaries – million euro’ divided by variable ‘Employees – number’, NACE sections B to N excluding K). See table 6.

Furthermore, the 2018 level and the 2018/2014 growth rate in the number of employees taken from SES, for all (part-time and full-time) employees including apprentices, working in an enterprise of 10 employee or more and belonging to NACE rev.2 sections B to N excluding K were also compared with the corresponding SBS data (dataset: sbs_na_sca_r2 in Eurobase, variable: ‘Employees-number’, NACE sections BtoN excluding K). See table 7.

As can be seen in table 6, the differences between SES and SBS in the growth rate of the mean annual earnings are quite small in certain sectors. However, in several other sectors (B, E, N), the difference seems quite substantial.

As can be seen in table 7, there are also differences between SES and SBS in the growth rates of the number of employees. For most sectors, this difference is <= 5 percentage points. However, sector I shows a bigger difference of 8 percentage points.

The differences in growth rates observed in tables 6 and 7 can be explained by the following factors:

-          In the SES, both the number of employees and the wages and salaries are derived from the same source (the SEW), which is considered to be complete (since it is based on a complete register with all enterprise and employees). The SBS also makes use of the SEW for the number of employees, but to derive the wages and salaries, the SBS makes use of a different sample survey based source (Production Statistics).

-          Even though both SBS and SES both use the SEW as a source, SES uses a sample of the SEW and is based on a combination of sources (see 3.1 Source data).

-          For section E specifically: This section contains a lot of employers that are non-market producers. The Production Statistics and therefore the SBS, exclude these non-market producers when counting the wages. The SEW and therefore the SES, do not exclude these employers when counting the wages.

-          Section N is the section that includes temporary employment agencies. In the SES all employees working for temporary employment agencies are assigned, together with their wages, to Section N. In the SBS however, due to partial usage of the Production Statistics, for a part of the working employees for temporary employment agencies, the wages are assigned to other sections.

-          The populations of the two SES vintages are based on October 2014 and 2018, while the populations of the two SBS vintages are based on the whole years 2014 and 2018. This explains the differences in number of employees, since these are based on the same source material (SEW). These differences in number of employees are also reflected in the mean annual earnings.

There are no annual nor sub annual versions of the Dutch SES 2018.

SES 2018/SES 2014 growth rates of the mean hourly earnings for all (part-time and full-time) employees including apprentices, working in an enterprise of 10 employees or more and belonging to NACE Rev.2 sections B to S excluding O were compared with the National Accounts (NA) data (dataset: nama_10_lp_ulc in Eurobase, variable ‘Compensation of employees per hours worked, in euro’). See table 8.

Although in the NA, compensation of employees include social contributions payable by the employer, it is assumed that they show the same pattern over time as the wage and salaries (i.e. gross earnings) component. Indeed, table 8 shows absolute differences no larger than 1 percentage point in the growth rates.

Likewise, SES 2018/SES 2014 growth rates of the mean annual earnings, for all (part-time and full-time) employees including apprentices, working in an enterprise of 10 employee or more and belonging to NACE Rev.2 sections B to S excluding O, were compared with the NA data (dataset: nama_10_lp_ulc in Eurobase, variable ‘Compensation per employee, in euro’). See table 8 as well.

The Dutch SES 2018 was checked extensively for internal coherence. Eurostat provides a broad set of internal consistency rules to be checked before the data is transmitted to Eurostat and which are also checked by Eurostat after transmission. Additionally, SN conducted an additional set of internal consistency checks.

After the transmission, Eurostat sent an Error Report to SN with the result of their consistency checks. These checks were either met by the Dutch SES data or explained by SN when they were not met but still plausible. Feel free to contact SN to obtain the results (and explanations) of these checks.

Publications regarding the Dutch SES 2018 are done by Eurostat. SN did/will not publish a specific news release regarding the SES.

Publications regarding the Dutch SES 2018 are done by Eurostat. SN did/will not publish a specific publication regarding the SES.

The Dutch SES 2018 is disseminated by Eurostat on their online database in aggregated form, together with the SES’s of other European countries. SN does not publish the results of the SES on its own online database.

N/A for Dutch publications regarding the SES. For consultations of Eurostat tables, please contact Eurostat.

Under strict conditions, the microdata are available for researchers through Eurostat or SN (see 1. Contact).

N/A

For additional documentation on or any questions about the methodology of the Dutch SES 2018, please feel free to contact SN (see 1. Contact).

For additional documentation on quality management regarding the Dutch SES 2018, please feel free to contact SN (see 1. Contact).

Extensive documentation regarding metadata is available through SN (see 1. Contact).

N/A

The Dutch SES 2018 did not impose an additional burden on respondents, as the research re-used data that was already available to SN (see 3. Statistical processing). The manpower needed by SN to create the SES 2018 was 2 FTE (full time full year, including a post for technical maintenance).

The Dutch SES 2018 was transmitted to Eurostat in microdata form (on person level, pseudonymized). Eurostat has applied confidentiality rules when disseminating the data in aggregated form on their online database. This might result in suppressed values, preventing the dissemination of confidential information on persons or their employers.

Microdata, especially regarding employee characteristics, are highly sensitive and confidential data and is therefore not available to the broad public. Under strict conditions, the microdata are available for researchers through Eurostat or SN (see 1. Contact).

See 11.1 Confidentiality - policy

Feel free to contact SN (see 1. Contact) when there are questions.