Notes on the ILO labour market statistics from reference month September 2007
With its press release of 30 October 2007, the Federal Statistical Office has replaced the telephone survey – which was used since January 2005 as an interim solution for monthly unemployment reporting – by a new time series. The new time series is based on the labour force survey which is harmonised at EU level. Unemployment data for Germany are thus available for the first time from that statistics that is harmonised at EU level. Although the new time series from the labour force survey differs in levels, it confirms the general decline of unemployment since 2005.
The methodology for changing over to the new time series has been developed in close co-operation between the Federal Statistical Office (Destatis) and the Statistical Office of the EU ropean Communities (Eurostat). The results published by Eurostat correspond to those of the Federal Statistical Office.
This paper contains background information on the new time series, the differences in results involved in the changeover to the new data source, the method applied in estimating the seasonally adjusted results, and the method of calculating unemployment rates.
1. Goals of the monthly ILO labour market statistics
The monthly ILO labour market statistics of the Federal Statistical Office provides monthly labour market data according to the labour force concept of the International Labour Organization (ILO). As the statistics is based on that internationally accepted and widespread concept, it is ensured that the German labour market data are comparable with those of other countries. Also, since the definitions do not depend on national provisions in terms of social security law, it is largely ensured that no breaks will occur over time resulting from changes in laws and possibly making it difficult to interpret the data.
Since January 2005, the ILO labour market statistics provided not only monthly data on the employment trend but also monthly data on unemployment in Germany. Until April 2007, the absolute number of persons in employment, the absolute number of unemployed persons, as well as the employment/population ratio, the underemployment rate and the unemployment rate were regularly published. Unemployment rates were also shown in a breakdown by sex and age and for eastern and western Germany.
In a transition phase from May to August 2007, the unemployment data were published only in a restricted form. From reference month September 2007, the publication programme has been extended again. Apart from the employment data, the following information on unemployment has been re-integrated in the range of data published: Seasonally adjusted unemployment figures and rates as totals and broken down by age and sex, unadjusted figures on unemployed persons by age and sex.
2. Previous sources of the ILO labour market statistics
The most important source of the ILO labour market statistics regarding unemployment was, as an interim solution, the telephone survey “labour market in Germany” since its introduction in January 2005. The unemployment data published were taken from that survey. The telephone survey was limited in terms of time and was last conducted for reference month April 2007. The legal basis of the survey, which is the employment statistics ordinance, expired at the end of April 2007. The employment data, however, were calculated as part of national accounting on the basis of all relevant employment statistics sources, especially the statistics of employees subject to social insurance contributions and marginal part-time workers, the labour force survey / microcensus, the public service personnel statistics and the short, medium and long-term statistics for economic sectors.
The monthly ILO labour market statistics was continued after the telephone survey had been discontinued, and it was planned that the labour force survey, which has continuously been conducted since 2005, should replace the telephone survey as the source of unemployment data. However, it was not possible to change over at once to that new data source because work was still under way to assure the quality of the survey results. For a transition period from reference month May 2007 to August 2007, reporting on the ILO labour market statistics was therefore restricted. Data on the seasonally adjusted number of unemployed persons and on the unemployment rate were updated – as was the case until the end of 2004. Changing over to the new time series from the labour force survey was done when the data for reference month September were published on 30 October 2007.
3. Change-over of the unemployment time series to the labour force survey (microcensus)
After the telephone survey “labour market in Germany” had been discontinued in April 2007 and after a transition phase until August 2007, the changeover to the new time series from the labour force survey – which is harmonised at EU level and is the only data source for monthly unemployment data according to the ILO concept – was performed for reference month September 2007. In Germany, the labour force survey is part of the microcensus, which is a continuously conducted household survey. The microcensus is the largest official household survey in Europe and involves compulsory response. It is a multi-purpose survey and is used not only to provide structural data as part of the monthly ILO labour market statistics but also to obtain monthly longitudinal data.
3.1 Extent of the revision
For the period from January 2005 to April 2007, unemployment data are available both from the telephone survey and from the labour force survey. Although the lists of questions of the two surveys are quite similar, the data sources do not provide coherent results. The changeover of the data source from the telephone survey to the labour force survey has led to a marked upward revision of the unemployment figures. Chart 1 shows the seasonally adjusted unemployment series from 2000 to September 2007, first, on the basis of the telephone survey results and, second, on the basis of the labour force survey.
Chart 1 clearly shows several things: First, the series after the changeover shows an average of nearly 800,000 more unemployed persons than the series before the revision. Second, the trends of the two sources are largely consistent. Third, the results of the telephone survey show a more marked downward trend than does the current series.
3.2 Causes of the difference
Although the lists of questions of the telephone survey and the labour force survey are quite similar, there are a number of methodical differences between the two surveys, which may have contributed to the differences in results. Some of those differences are listed below. This does not allow to finally and comprehensively assess the differences (which requires further analyses and studies); it is only possible here to mention possible causes of the differences in results:
- The labour force survey (microcensus) is a one-stage area sample, and all persons living in the selected households are questioned. For the telephone survey, however, a two-stage method was applied: First, telephone numbers generated at random (fixed-line network) were used to choose a household and then a person at working age living in that household was selected for questioning. Especially persons in smaller households had a lower chance – depending on the time of the call – to get into the sample.
- What is used in the labour force survey is mainly computer-assisted face-to-face interviews (CAPI), complemented by self-administered questionnaires, and telephone interviews. The telephone survey used only computer-assisted telephone interviews (CATI), with continuous interview control in a central telephone studio.
- Due to compulsory response involved, the labour force survey shows a non-response of just 5% of the annual overall sample. The telephone survey, however, had a non-response of just under 50% in the first interview and just under 10% in the follow-up interviews, which cannot be avoided in voluntary surveys as experience shows.
- In both surveys an expansion is done, using the same demographic variables (age, sex, region, and citizenship). However, the breakdowns used are slightly different. The data on registered unemployment collected in the telephone survey for specific subpopulations were additionally adjusted to the number of registered unemployed as shown by the Federal Employment Agency.
- The list of questions on the employment status is practically identical in the two surveys, except for differences required by interview techniques. For example, longer lists of answer categories must be broken down to individual questions in the telephone studio. Altogether, the list of questions of the microcensus as a multi-purpose survey covers many more topics in other areas than does that of the telephone survey.
4. Methodical notes
4.1 Structure of the time series
Monthly reporting from the labour force survey from 2007
The monthly unadjusted figures on unemployed persons from reference month January 2007 are published directly from the labour force survey and are adjusted to the quarterly data generally four months after the end of any quarter.
Back calculation for 1992-2005
Before 2005, the labour force survey was conducted annually with a fixed reference week in spring. Therefore, no infra-annual data are available from the labour force survey for the period before 2005. In order to obtain a longer monthly series of unemployed, the annual results of the labour force survey before 2005 were calculated backwards in combination with the average monthly changes of the years 2005, 2006 and 2007. On the basis of the monthly data from that model, the seasonally adjusted time series is compiled according to the method described below. The seasonally adjusted series is published from 1992, although the results for the period from 1992 to 2006 are estimates and should be interpreted with caution. The back-calculated series not adjusted for seasonal variations is not published because of the low informative value of the data.
The unadjusted results can be published only from January 2007 because the monthly trends of the years 2005 and 2006 are influenced by unusual effects in the context of introducing the continuous labour force survey in Germany.
4.2. Seasonal adjustment
To perform seasonal adjustment, the infra-annual trend of unemployment figures must be known to a sufficient extent. For that purpose, a time series should be available that covers at least four years. As such valid monthly data are available only from January 2007, no sufficient time series is available yet for seasonal adjustment. For unemployment in Germany, it will not be available before 2011. Until that time, seasonally adjusted results must transitionally be estimated, using the information on infra-annual trends from other sources. For that purpose, the data of the Federal Employment Agency on the monthly seasonally adjusted number of registered unemployed are used. Although their definition differs from that of unemployed persons according to the ILO concept, there is a large overlap between the two groups and the seasonal trends are similar. Therefore, the seasonally adjusted monthly number of registered unemployed can be used to determine the seasonally adjusted number of ILO unemployed. A similar method is also applied in other European Union member states.
In this context, a correction factor is calculated from the moving 12-month average of unemployed from the labour force survey and of the registered unemployed. That correction factor is multiplied by the seasonally adjusted monthly figure of registered unemployed. The correction factor is updated every month when the new unemployment figure from the labour force survey becomes available, while for the six most recent months it is kept constant. This means that the seasonally adjusted monthly figures are provisional for six months. If the infra-annual trends of the figures on registered unemployed differ from those on unemployed from the labour force survey, the data will be revised.
4.3 Source of employment figures and calculation of the unemployment rate
Data on employment and unemployment in Germany are published monthly from the ILO labour market statistics. The results on employment are obtained by employment accounts on the basis of all major sources of employment statistics. It turned out in the past that employment accounts provide a reliable and stable time series of monthly employment figures. The data on employment in Germany obtained over the last few years by the labour force survey and by employment accounts differ considerably. Many improvements in the labour force survey, especially with a view to covering marginal part-time workers, have markedly reduced those differences. The exact causes of the differences that have remained despite the improvements are subject to current studies. For the calculation of the unemployment rate, the number of persons in employment from the employment accounts is used2, because this is the figure covering employment most comprehensively. For the number of unemployed, the result of the labour force survey (microcensus) is used. The unemployment rate is calculated as follows:
The efforts being made to improve the coherence between employment accounts and the labour force survey also serve the long-term purpose of putting the monthly unemployment rate on a single data source.
5. Outlook
The current changeover of the monthly unemployment statistics altogether leads to a stronger harmonisation with the methods applied in other EU countries. Nevertheless, studies are being carried out to further improve the measurement of the employment status in Germany. This refers in particular to the following items:
- The Federal Statistical Office, in co-operation with the statistical offices of the Länder, is currently performing extensive methodical studies on the survey methods of the labour force survey because so far it has been possible to explain just part of the differences observed in the results. Apart from the implementation of the ILO labour force concept in the interview process, which has been subject to repeated fundamental discussion, such activities refer to a further in-depth analysis of the field work, including especially the knowledge obtained from follow-up surveys and other studies.
- What should be taken into account for further activities is that the labour force survey has been designed as a multi-purpose survey whose main purpose is to provide annual structural results and, at the same time, information on monthly trends. It is currently being examined what measures must be taken to optimise the survey for the goal of providing monthly results. Studies here focus on the analysis of the influence of varying monthly sample sizes and on issues of adjusting the sample design to short-term presentations of data, e.g, through rotation of sample units, which is common for time series.
Although the current changeover of the monthly unemployment statistics altogether leads to a stronger harmonisation of estimation methods with those applied in other EU countries, further improvements are emerging for the future, which may have an impact on the results. For example, intensive efforts continue to be made to further improve the coherence of different labour market data in Germany, such as the labour force survey compared with the employment accounts as part of national accounts. It should also be noted that the sufficiently stable seasonal pattern – which will presumably be available in about four years – of the new monthly series of unemployed persons that has started now can be used as a basis for direct seasonal adjustment.
______________________
1 As the time series is not sufficiently long for direct seasonal adjustment, an estimation method is applied transitionally to obtain seasonally adjusted results. The seasonal adjustment method is explained under 4.
2 Resident concept, i.e. persons whose place of residence is in Germany. Also, the result is adjusted according to the harmonised definitions of the labour force survey and refers to persons living in households and aged 15 years or over (excluding persons performing compulsory military or alternative service).
For further information please contact:
Labour market statistics
Phone: +49 611 75 4868
Time of Service: Monday to Thursday 8am to 5pm, Friday 8am to 3pm
