Revision of the monthly harmonised unemployment figures for Germany

With the release of the data for March 2011, Statistical Office of the EU Communities (Eurostat) and the Federal Statistical Office (Destatis) partly modify the previous, provisional method of calculating harmonised monthly unemployment figures for Germany. The modified method continues to be based on the German Labour Force Survey and leads to marginal revisions of the unemployment series published hitherto. The changes concern some of the source data used for the denominator of the unemployment rate and the estimation method to adjust for seasonal variations.
The unemployment rate measures the share of the unemployed in the labour force (the labour force is defined as the sum of unemployed plus employed persons). The employment data used in the denominator is now taken from the same source as the unemployment data, the German Labour Force Survey (LFS). This improves international harmonisation as well as the internal consistency of the rate. Due to this change, the unemployment rate for Germany will rise by about 0.2 percentage points. The general development will however not be affected.
The use of auxiliary information from the German unemployment register for the purpose of seasonal adjustment will be discontinued. However, as a number of methodological issues remain to be addressed for seasonal adjustment directly on the LFS series, LFS based trend estimations will temporarily be published in lieu of seasonally adjusted series.
The changeover to the new method has been prepared jointly by the Federal Statistical Office of Germany and Eurostat. The figures published by Destatis are identical to those released by Eurostat.
This paper describes the magnitude and underlying causes of revisions of previously released data, and provides information on the new method. This note is divided into four main sections. The first section describes the approach to calculate internationally comparable monthly unemployment figures in the European Statistical System. The second section examines the revision resulting from the changeover to LFS employment data for calculating German unemployment figures. The third section, as a background, describes specific aspects of the new calculation method. Finally, the last section gives a short overview of the outstanding future work.

1. Harmonised unemployment rates in the European Statistical System

Eurostat produces harmonised monthly unemployment rates for the EU, the euro area and Member States with some time series dating back to 1983. The backbone of the calculations is data from the EU LFS, measuring among other things the labour status according to the recommendations of the International Labour Office – ILO labour market statistics. The LFS is regulated in the EU by Council Regulation (EC) No 577/98 of 9 March 1998, setting the common framework for a continuous survey in the Community providing quarterly results.
An important step towards improved comparability of unemployment estimates across Europe, was Commission Regulation (EC) No 1897/2000 of 7 September 2000 which created a legal basis for a common definition of unemployment and of a common approach to the formulation of the questions on the labour status in the LFS. Regulation (EC) No 1991/2002 of the European Parliament and of the Council of 8 October 2002 made the continuous, quarterly LFS mandatory for all Member States from 2003, providing, amongst other things, a transitional period for Germany until 2005.
According to the legal requirements, the LFS is targeted at the production of quarterly results. In order to provide unemployment rates with a monthly frequency, for several Member States short-term trends from unemployment registers have to be used assuming similar short-term (within-quarter) developments with the unemployment based on ILO labour market statistics. The LFS definitions. In some countries, like Germany, the sample size of the LFS allows observing directly monthly trends. The Federal Statistical Office of Germany started estimating monthly unemployment directly from the LFS in October 2007. The method applied for this estimation is considerably improved by the current revision.
In the German case, monthly values are available from the German LFS from early 2005 onwards. However, as a result of phasing-in problems, the monthly fluctuations of the data are not considered reliable enough before 2007 to allow for the publication of this data. Therefore, original values from the continuous LFS are published from reporting month January 2007 onwards. The monthly data published for the reference months January 1991 to December 2004 are based on the annual LFS and an interpolation of monthly seasonal patterns based on the available monthly data from 2005 to 2010.
The monthly values derived from this back calculation are used as intermediary figures in order to derive a trend estimation as described below. Only the trend estimation can, however, be published, and not the intermediary data. The figures from the trend estimation are published from January 1991 onwards, but the estimates based on the back calculation should be treated with caution.

2. Methodological aspects of the new approach

When monthly unemployment estimates based on the German LFS were first published in 2007, a number of methodological issues needed further investigation. Provisional solutions were applied at the time. They can now be replaced by a more solid approach, which nevertheless needs to be added by further steps in the mid-term future. This section provides background information on the changes.

Calculation of monthly unemployment rates

In Germany, the Labour Force Survey is the only source measuring unemployed persons on a monthly basis according to ILO labour market statistics definitions. For employed persons (used in the denominator of the rate), an alternative figure is available from the system of Employment Accounts that provide a reliable and stable series of employment results on a monthly basis. As a number of methodological issues regarding the monthly LFS figures still needed further investigation, in the past, only the estimate of the unemployed had been taken as measured with the LFS, while the German employment figures were taken from the Employment Accounts.
After a series of methodological improvements to the monthly LFS results since 2005, now both the unemployed as well as the employed persons as measured with the LFS are being used for the calculation of the unemployment rate. This brings the approach for Germany in line with the standard practice in the European Statistical System and improves international harmonisation. At the same time this approach also guarantees that the results for employment and unemployment are internally consistent. The following formula is used after the revision for the calculation of the unemployment rate:





As the number of employed persons according to the LFS is lower than in the Employment Accounts, the denominator of the rate becomes smaller, and, as a result, the rate increases somewhat.

Use of trend estimates as a substitute for seasonal adjustment

In order to gain a clearer picture of the general trend in unemployment for the purpose of business cycle analysis, regular variations within a year, such as an increase in unemployment due to hibernal weather conditions, must be removed from the series. This is a standard, well-know procedure for all types of infra-annual data. This process of adjustment, however, is only possible after having gained sufficient knowledge on the infra-annual variations from the data.
Actual monthly data on the number of unemployed for Germany from the LFS is only available since early 2005, which (in 2007 when the release of monthly data started)) was not a sufficiently long time series to enable reliable seasonal adjustment using standard techniques. Therefore, for a transitional period, an alternative method was applied, which used the number of registered unemployed as an auxiliary source for the estimation of seasonally adjusted results.
Now, more than three years later, a sufficiently long time series is available, so that the interim solution can be discontinued.
However, as the monthly results from the German Labour Force survey do not yet show stable seasonal patterns, it is still not possible to use standard seasonal adjustment procedures. Those would lead to a highly volatile seasonally adjusted time series, which would be very difficult to interpret. Such a procedure would also risk resulting in implausible seasonal effects. Therefore, as practiced in other Member States in a similar situation, until a more stable seasonal pattern is observed, Eurostat and the Federal Statistical Office Germany provisionally publish a trend estimate (trend-cycle component) as a substitute for seasonally adjusted unemployment figures. The trend-cycle component is estimated on the basis of the non-adjusted monthly LFS-data. The use of a trend estimation¹ removes not only seasonal effects, but also irregular effects as well as variations due to sampling errors and methodological biases from the series. As can be expected from a trend estimation, the resulting series are highly smoothed. The possible drawback of this procedure is that month-to-month movements in the unemployment data might be smoothed out and that there might be significant revisions in case of turning points.

3. The revised estimates

The revisions are mainly due to the above described changes in source data of employment figures for the calculation of the denominator of the unemployment rate and the approach to adjust for seasonal variation. In addition, figures on the absolute level of unemployment change marginally due to slight improvements of the weighting scheme of the quarterly LFS which are introduced at the same time.

Overall, the unemployment rate for Germany increases by 0.2 percentage points on average for the period since 2007. The increase ranges from 0.1 to 0.3 percentage points in most months months. At the same time, the long-term developments as well as the short-term trends are only marginally affected (see table/figure).
The trend estimation to obtain the results adjusted for regular seasonal variation as well as irregular components does not change the overall level of unemployment. Also the general development since 2005 is nearly identical to the initially published series. The month-to-month changes can however differ.

4. Future work

The new estimation method significantly contributes to further harmonising the calculation of monthly unemployment rates in the European Statistical System. Using the LFS as unique source is bringing the German unemployment series in line with international standards and guidelines. Nevertheless, further improvement efforts are needed.
The Federal Statistical Office of Germany, together with Eurostat, therefore will continue their joint work on the monthly estimates from the German Labour Force Survey. Upcoming improvement projects will examine whether changes in the sampling design, the weighting scheme and the fieldwork processes could contribute to a reduction of the volatility of the time series. At the same time, opportunities to enhance the trend estimation procedure in order to reduce the average revisions will be studied.

---

¹ Trend estimation is done on the basis of the Berlin method BV4.1. For a description of the method please refer to Methods – Approaches – Developments, 1/2011; Rengers, Martina: Monthly unemployment statistics - trend estirnation using BVA4.1. (only in German). See also Speth, Hans-Theo: The BV4.1 Procedure for Decomposing and Seasonally Adjusting Economic Time Series (only in German), Wiesbaden: Federal Statistical Office, 2004

Labour market statistics

Phone:  +49 611 75 4868

E-mail: arbeitsmarkt@destatis.de

Time of Service: Monday to Thursday 8am to 5pm, Friday 8am to 3pm