Data basis for register-based employment statistics
The base files for obtaining data from the register-based employment statistics are produced for use in statistics. Choices are made with a view to providing usable data at table level and for relatively aggregated levels of some variables. At the micro level, errors will occur, and for some variables the errors may be extensive.
Data is organised as year files. The unit in the files is ‘person’, and the unit contains information about employers (industry, sector, location) and employment (occupation and hours of work). Only information about an employee’s main employment in November is included in the files; there is no information about any second jobs or jobs held at other times of the year.
1. Data sources
Until the end of 2014, data for the register-based employment statistics was based on several different registers. The most important of these were the Register of Employers and Employees (Aa Register) held by the Norwegian Labour and Welfare Service (NAV), the End of the Year Certificate Register (Lønns- og trekkoppgaveregisteret) and the Register for Personal Tax Payers (Selvangivelsesregisteret) administered by the Directorate of Taxes, the Registers of Conscripts and of Conscientious Objectors (Verneplikts- og Siviltjenesteregisteret) and the Central Coordinating Register for Legal Entities (Enhetsregisteret). The Register of Employers and Employees is the main source of data on employees, but the End of the Year Certificate Register constitutes an important supplement as it captures aspects of employees that are not reported in the Register of Employers and Employees. The unit in both registers is ‘employment’ (jobs).
New data basis from 2015
From 2015, the reporting by employers to NAV’s Register of Employers and Employees and some reporting to the Norwegian Tax Administration and Statistics Norway was conflated into a new joint reporting solution known as ‘a-ordningen’. This solution is a harmonised collection of data on employment, income and tax deductions from the Norwegian Tax Administration, NAV and Statistics Norway.
The reference date is the third week in November. Up to the end of 2014, the total number of employees was determined such that the national figure for wage earners and self-employed persons respectively would be equal to that indicated in the Labour Force Survey (LFS) for these groups in the fourth quarter. The LFS was used due to the fact that we have not had good-quality registers that cover all employees. Since the introduction of a-ordningen, the LFS no longer determines the total number of wage earners in the register-based employment statistics. The change in the time series is described in more detail in the document Details of the relationship between old and new statistics (in Norwegian).
2. Who is considered to be an employee?
Employees are defined as persons who performed paid work of at least one hour in the reference week, as well as persons who had such work but were temporarily absent due to illness, holiday leave, paid leave or similar. Persons undertaking military or civilian national service are considered to be in employment. Persons on government employment initiatives who receive wages from an employer are also classified as employees. This follows the recommendations of the International Labour Organization (ILO).
The ‘kyrkstat’ (link in Norwegian) variable denotes the source of information for employment. The best quality at the micro level will be for those with code 1 or 2. Here, the source is NAV’s Register of Employers and Employees, where employment start and end dates are given. The quality will also be good for those who have code 6 (source: Register of Conscripts). For the remainder, we know that they have received income as employees or pensionable income as self-employed during the year, but not whether they have been employed in the reference week for the statistics. It is mainly the amount of the annual wages earned/income from self-employment that determines who is included as wage earners/self-employed persons for those who do not have code 1, 2 or 6 in kyrkstat. For these, analyses that follow persons over time will therefore be subject to greater uncertainty.
The reporting in a-ordningen includes some employment that Statistics Norway does not class as active in the statistics period, such as seasonal workers, on-call temporary workers etc. These should, in principle, be reported with an end date as soon as the job ends, but this is rarely done. It is mainly information about wages that determines whether the employment is to be considered as active or not during the statistics period.
Operationalisation of the population delimitation in the register-based employment statistics is as follows:
- Start date and, where applicable, end date for employment means that the subject has a job in the reference week (the week with the 16th in it)
- Information about wages and type of wage payment is important for determining whether work should be considered to be employment and whether it is active in the statistics period
- Persons who have not received wages, but who are temporarily absent due to illness, holiday leave, paid leave and such like should also be considered as employed. In order to identify this, information on payment of benefits (sick pay, maternity benefit and parental benefit), leave and lay-offs is used. Only leave/lay-offs of less than 90 days are categorised as employment.
In addition, employment without wage and absence information is controlled for various time lags, e.g. there may be a time lag in the payment of wages for new employees, or missing wage payments for hourly paid workers on holiday leave. Employment with a new start date and employment with unpaid wages are therefore also included in month t-1 and t+1 in the population.
3. The industry, occupation, place of work and hours of work variables
For employees who do not have kyrkstat (link in Norwegian) 1 or 2, the quality of these variables will often be relatively poor. More details are given about this in the information on each variable.
For the ‘occupation’ and ‘hours of work’ variables, changes have been made since the introduction of a-ordningen. The ‘place of work’ and ‘industry’ variables are retrieved from Statistics Norway’s business and enterprise register. The actual variables have not been changed from previous years as a result of a-ordningen, but as noted above, there is a break in the time series. This also applies to employees broken down into these variables.
Although variables such as occupation and industry are found at the most detailed level, Statistics Norway has not performed any controls on data at this level, nor does it publish tables on ssb.no at this level.
Industry for a person/job is determined by the industry of the business in which they work. There are three main reasons why industry may be categorised incorrectly:
- The business has the wrong industry code. Such errors are relatively rare. Time lags may be found in the controls of whether businesses should have changed their industry code. Such checks are carried out annually for all large businesses, as well as for a selection of the small ones, by Statistics Norway personnel who produce industry statistics. For industries without industry statistics, the quality will be poorer. The same goes for small businesses that are not included in the samples for the industry statistics.
- The second and more common error is that the person is not associated with the right business in enterprises with multiple businesses. For example, the employer may have reported all employees only on the main business of the enterprise. However, if all businesses have the same industry, this is no problem in terms of the industry categorisation, but could be a problem in relation to the person’s geographical place of work. In general, industry is only controlled at table level and down to a 2-digit industry level. The control also takes place mainly by looking at major changes at municipality and county level. Small errors will not therefore be detected. For municipal administration and central government, the control is more detailed. Nevertheless, Statistics Norway can only be expected to discover relatively large errors here.
- The third reason for errors in industry codes can arise where the enterprise has not distinguished all businesses with a separate organisation number. If the businesses have different industry codes, some of the employees will have the wrong industry. Such errors are controlled at the micro level for enterprises that are included in industry statistics.
A person’s occupation is determined by their work duties.
- Direct information on occupation titles is reported to the Register of Employers and Employees, but with exceptions for central government employees. Here job codes are entered for the state. The occupation code is determined by the employer from the occupation title that the employer believes is held by the employee. Some employers’ use of occupation titles differs from common practice, and employees may therefore have the wrong occupations codes.
- For central government employees, there is a state job code. Statistics Norway has recoded these to occupation codes, but this has resulted in a poorer quality partly because some job codes are too general, such as ‘consultant’.
- For employees with an employment source solely from the End of the Year Certificate Register or for those who are self-employed, Statistics Norway imputes occupation codes based on different variables. The quality of these will be acceptable at table level and for rough occupational grouping, but will have major errors at the micro level and are generally not recommended. For use at micro level, therefore, only use the variable ‘yrke_imp’ (‘yrke_publ’ for STYRK08) for those with the values A and B in the variable ‘Kyrke_imp’. The others will then get an unspecified occupation. For some of these, the combination of education and industry could give a certain indication of occupation.
- Occupation is reported for all employment in a-ordningen. It is reported at a 7-digit level, but is only controlled for the first four digits in a-ordningen. The control only ensures that a valid value is reported. Incorrectly reported occupations are not intercepted.
- The occupation of central government employees is also reported in a-ordningen. Until the end of 2014, only job codes were reported for this group, and these were recoded to occupation by Statistics Norway (see above).
A person's place of work is determined, in principle, by the location address of the business where the person works. In the same way as for industry, errors in the place of work can be due to an incorrect address for the business or an employee being registered on the wrong business. The latter is the most common error. There are some exceptions from the address of the business determining the workplace municipality. In addition, the actual place of work will not be registered for some occupations.
- For self-employed persons who are not registered with an organisation number, the municipality of residence is used as the workplace municipality. There are also some groups of employees whose municipality of residence is registered in the field for workplace municipality. This applies to those who work at sea and the Armed Forces.
- Some actual places of work do not require a separate business to be created for the location. This applies to short-term workplaces, such as construction sites.
- Another group is workplaces with only one or very few employees. In such cases, the employees will mainly be registered on the business to which they administratively belong in the enterprise.
- Those who work for employment agencies will have the workplace municipality for the administrative unit they belong to and not for the business they work at.
- Some employees are in occupations that entail them to be mobile, such as many of those working in transport activities. These will normally be linked to the administrative unit they belong to.
- With regard to controlling workplace information, this is partly done by checking that the enterprises have established a separate organisation number for each business they have and partly by ensuring that the employees are distributed in line with this. However, there are generally no controls beyond the municipal level. The quality of data at the district level and for the basic statistical unit will therefore be poorer. Statistics Norway does not generally publish employment data at the basic statistical unit level. However, supplying such data to certain research/analysis projects may be justifiable when the basic statistical units are to be used as building blocks in larger groups across municipal boundaries.
- In addition to the above points, when using the employment data for analyses of commuting and transport it must be noted that there is no information on how often employees travel from their home to their workplace. This is particularly relevant for part-time employees. We also point out the recognised problems with the residence registration of students. Many will be registered as living in their parents’ municipality and not in the municipality they actually live. Many of the questionably long commuter flows will be due to such factors.
Hours of work
This applies to contractual hours of work in the reference week. Here too, we only have direct information about the employment that uses the Register of Employers and Employees as a source. The remainder are imputed. The main model is that the LFS shows the relationship between contractual hours of work (average weekly hours of work for self-employed persons) and annual wages earned (or income from self-employment). It is assumed that there is a corresponding relationship between income and hours of work for employees in the register (for those without hours of work from the Register of Employers and Employees). This obviously gives rise to some errors at the micro level. There may particularly be errors for self-employed persons due to the fact that income from self-employment and hours of work are not as closely correlated as wages income and hours of work.
In a-ordningen, ‘expected hours of work’ is replaced by ‘percentage of position’, ‘number of hours full time per week’ and ‘paid hours’ for hourly paid workers. The contractual/expected hours of work are extrapolated from the first two variables.
There are a number of quality challenges related to the reporting of hours of work. This is partly due to missing values, partly due to default values being given (0 or 100 for the percentage of the position), and partly because extremely low or high values are given. Where possible, we correct for missing/incorrect information.
- Some payroll and personnel systems used 0 as the default value for hourly paid workers in 2015. When this was changed in 2016, many reported 100 instead. Thus, the change in full-time employment between 2015 and 2016 does not present a true picture. Changes in position percentages and full-time/part-time proportions between 2015 and 2016 must therefore be interpreted with caution.
- There is no position percentage for hourly paid workers. This variable is therefore calculated from the number of paid hours and the number of hours full time per week.
- In some cases, position percentages that are obviously incorrect are reported. The maximum percentage per employment is set at 120, and missing percentages are calculated. Negative percentages are set to zero. If an employee has more than one position in the same business, the employment is summarised. If two or more jobs are 100, the total percentage is set at 100.
- Persons with a total position percentage of more than 160 will have the percentage reduced in one or more of the jobs, such that the sum does not exceed 160 per cent.
- The ‘reported percentage position’ variable is the percentage as reported, without calculations for missing values or adjustments for unrealistic values.
- The ‘number of hours full time per week’ variable is also adjusted for obvious errors and omissions. Excessively high hours are set at 37.5. hours. If the field is empty and a shift arrangement has been entered, this value will be used.
- The ‘contractual hours of work’ variable is thus extrapolated and not directly comparable with previous years.
The variable that denotes full time/part time (arb_heldeltid) is extrapolated from the position percentage (arb_stillingspst). Position percentage >= 100 = full time and position percentage < 100 = part time.
4. Category variables in general
In a-ordningen (from 2015), missing values are replaced with ‘-2’. Invalid values according to the code list are replaced with ‘-1’. Blank fields (or ‘.’) also appear and some code lists have separate codes for unspecified values.
5. Correlation with previous years
The new data basis for register-based employment statistics means that the number of employees/employment figures from the fourth quarter of 2015 are not comparable with previous years. There are two main reasons for this:
- A-ordningen is a new data basis as from 2015.
- The LFS is no longer used to determine the total number of employees (see under ‘Data sources’).
This means that the differential between published figures for 2014 and 2015 does not correspond to the actual change in employment. Changes must therefore be interpreted with caution.
A-ordningen generally provides a better data basis because it is more accurate at the individual level and covers more employee variables than the Register of Employers and Employees. The change in the time series is described in more detail in the document Details of the relationship between old and new statistics (in Norwegian).
For more details about the employment data that forms the basis of the register-based employment statistics up to the end of 2014, see Register-based employment statistics. Documentation (Documents 2010/8).