The German Employment Accounts for National Accounts and Labour Market Statistics

Transcription

1 Dipl.-Volkswirt Stephan Lüken The German Employment Accounts for National Accounts and Labour Market Statistics Experiences and findings of six decades The number of persons in employment in Germany is a key element of the country s labour market reporting and National Accounts. The German Federal Statistical Office has for many years provided appropriate information from the Employment Accounts for both of these purposes. The present paper traces the historical development of the Employment Accounts in the system of labour market statistics and in the National Accounts. Special attention should be paid to the major revisions of the National Accounts, which have frequently also formed the framework for important revisions of the Employment Accounts. This paper devotes a whole section to the latest revision, carried out in 2011, the main reason for which was the changeover to the Classification of Economic Activities, 2008 edition (WZ 2008) used in the German National Accounts. 1 The status of the Employment Accounts in the system of employment statistics and National Accounts The calculation of average employment figures for the period under review has a long tradition in German official statistics. The results obtained from the Employment Accounts are included in reports of labour market participation of the population (demographic statistics) and are used for analyses of the labour supply and short-term cyclical changes in labour markets (labour market statistics). In the National Accounts the results obtained from the Employment Accounts are used in particular as reference data for the determination of productivity and average income figures. Information on employment and labour productivity was included in the German National Accounts for the first time in 1954 and on-going annually since 1956; since at that time during the immediate post-war years it was considered more important to provide for the consumption needs of the population than to treat production activities. In the early post-war years the macroeconomic aggregates (at that time national income and national product) were therefore calculated in relation to the respective total population. In addition to this there were at that time still no reliable annual or even infra-annual employment statistics which would have made possible documenting of information on the development of employment (Fritsch/Voy 2009). However, over the subsequent years significant improvements in the supply of employment statistics data were achieved. The milestones were > the introduction of the Microcensus in 1957, a representative statistics on the population and the labour force which for the first time gave a general overview of the employment situation by way of an on-going annual reporting system, > the introduction of the European Labour Force Survey in 1960, which was integrated into the Microcensus sample in Germany, 1 and > the introduction of an integrated social insurance reporting procedure for employers in 1973, which served as basis for the creation of a comprehensive employee 1 In an interdisciplinary working group of the Federal Statistical Office in cooperation with the federal states on a new system of household statistics, ideas for a new conception are currently being developed (see Hochgürtel 2013). Statistisches Bundesamt (Federal Statistical Office) German version published in Wirtschaft und Statistik 5/2012, page 385 et seq. 385

2 and wage statistics system based on administrative data (Herberger 1975). 2 Nevertheless, the data on employment available for economic sub-sectors from the above and other sources have not yet resulted in a comprehensive reporting system capable of fully meeting the temporal and functional demands. On the one hand a reporting system on labour market was needed and on the other hand the requirements of National Accounts had to be met with respect to consistent statistical picturing of the production factor labour in combination with additional economic aggregates. Official statistics in Germany have therefore supplemented from the employment statistics available primary and secondary statistical information with calculations of average results for the period under review for numbers of persons in employment by economic industry and status in employment which were specially designed to meet the demands of labour market statistics and National Accounts. Starting point of calculation of employment figures for the purposes of both labour market monitoring and National Accounts in the former territory of the Federal Republic of Germany were the results of the population and non-agricultural local unit censuses of 1961, 1970 and 1987, which were extrapolated based on all the appropriate short-term data sources on employment. The most important function of Employment Accounts is to merge the information available from various sources of employment statistics into a complete picture as gapless as possible and without overlaps. In order to achieve this objective, conversions and supplementary estimates are necessary but subject to certain uncertainties depending on the timeliness, completeness and composition of the available original data. Nevertheless, the estimation method, using in principle all existing sources and relevant informations, makes it possible to considerably reduce the degree of uncertainty. For that purpose, similar sets of circumstances are looked at from different points of view (surveys of individuals and households, business surveys, and administrative data) accompanied by independently calculated macroeconomic aggregates of the National Accounts and other labour market statistical indicators (Wollny/Schoer 1982). 2 Concepts and definitions of employment statistics Conceptually speaking the presentations of the results obtained from the Employment Accounts for the purposes of the demographic and labour market statistics on the one hand and the National Accounts on the other differ in particular in respect of the regional classification of persons in employment (by place of residence or place of work) as well as in respect of publication of the employment figures by industry in respect of the classification of economic activity (according to the main kind of activity of the local unit or that of the enterprise). 2 See Herberger (1975) among other things on the relaunch of the Microcensus and the Labour Force Survey. Within the framework of the demographic and labour market statistics the labour market participation of the domestic population thus of the resident population is the focal point. In the labour market monitoring concept the Federal Statistical Office therefore publishes employment figures primarily delimited according to the place-of-residence concept (national concept). Where the employment figures are reported according to the national concept in a breakdown by economic activity, the main factor in the industrial classification is the respective local unit (local unit or establishment concept). 3 The National Accounts include the results obtained from the Employment Accounts as mentioned in particular as reference variables for the determination of productivity, average earnings and other income and distribution figures. Therefore, depending on the intended purpose, employment figures delimited according to place of work or to place of residence are used in the presentations of the National Accounts. With the place-of-work concept (domestic concept) all persons in employment working in domestic economic units are taken into account. The numbers of persons in employment according to the national concept and the domestic concept differ by the size of the balance of inward and outward commuters over the borders of the respective economic territory (net commuter flow). By economic activity the results obtained from the Employment Accounts are always reported in the National Accounts according to the domestic concept. At the same time classification of persons in employment by industry carried out according to the respective main kind of activity of the enterprise (enterprise concept). Differences in the industrial classification according to the main kind of activity of the enterprise or establishment occur with multi-unit enterprises where the focus of economic activity of individual local units differs from that of the enterprise as a whole. ILO definition of employment A common denominator in the calculations for the purposes of labour market monitoring and the National Accounts is the use of a single definition of employment based on the Labour Force Concept of the International Labour Organization (ILO). 4, 5 3 At macroeconomic level and broken down according to the sex of the persons in employment, for the purposes of labour market monitoring the place of residencebased employment figures were provisionally supplemented with estimates of average unemployment and labour force figures (as a total of persons in employment and the unemployed) for the period under review. The figures on registered unemployment provided by the Federal Employment Office formed the main basis for estimating the numbers of unemployed. Only very much later, it was possible to replace these unemployment figures from the Federal Employment Agency with internationally comparable unemployment figures from the European Labour Force Survey (see remarks in Section 3). 4 Resolutions of the 13th International Conference of Labour Statisticians in respect of statistics concerning the total labour force, employment, unemployment and underemployment (Geneva, October 1982), revised by means of the resolutions of the 16th International Conference of Labour Statisticians on the measurement of underemployment and inadequate employment situations (Geneva, October 1998). According to the Labour Force Concept a distinction is made between persons in employment and the unemployed on the one hand and members of the inactive population on the other. and the unemployed together constitute the population group known as the labour force; on this basis this group includes all persons who either have a (paid) job (persons in employment) or are seeking work and are available to work (unemployed). A comprehensive explanation and appraisal of the Labour Force Concept can be found in Rengers (2004). 5 A definition of persons in employment already geared to the internationally agreed Labour Force Concept was applied for the first time in the 1961 population census [see Fritsch/Voy (2009:178]. 386 Statistisches Bundesamt (Federal Statistical Office) German version published in Wirtschaft und Statistik 5/2012, page 385 et seq.

3 In principle the definition of persons in employment pursuant to the Labour Force Concept of the ILO includes all work carried out for the purpose of economic gain or for payment even if of the most limited scope. This definition, primarily based on economic categories, is in line with the rules of the National Accounts, whose function is to include any work which contributes to the macroeconomic production of a national economy. In this manner an unambiguous relationship is formed between production and labour market participation of the population such that total production is exclusively rendered by persons in employment (and not also by the unemployed and/or members of the economically inactive population) (Fritsch/Lüken 2004: 140). What work to be included in the measurement of the output of a national economy in concrete terms depends again on the binding requirements of the European System of National Accounts (ESA) in respect of the production boundary of the National Accounts. Pursuant to the prevailing ESA 1995, therefore, persons in employment in the National Accounts include all persons engaged in gainful activity within the production boundary of ESA as employees or as self-employed persons (including unpaid family workers). 6 In line with the definition of employment pursuant to the Labour Force Concept of the ILO, the duration of the activity is of no importance for classification as persons in employment. Thus persons are considered to be in employment if as little as one hour has been worked in a given period. In addition, persons in employment also include persons with a formal job attachment who are currently not working, such as persons on parental leave or in the release phase of partial retirement. By contrast, persons with several jobs are only included once, that is with their main job according to working time (personrelated concept). In principle, this very broad-based definition of persons in employment, which also includes part-time jobs and other types of (paid) occasional jobs, makes it necessary to supplement the calculated employment headcounts for example for productivity analyses with indicators of the volume of employment and in some cases also of labour intensity (Asef/Wanger 2011). In the National Accounts this takes place at national level through the publication of volume of labour data adapted to the results of the Employment Accounts of the Federal Statistical Office, which the Institute for Employment Research (IAB) of the Federal Employment Agency makes available for these purposes. Regarding demographic and labour market statistics, the Microcensus and the Labour Force Survey in particular have the task of augmenting the total employment figures from the Employment Accounts to include structural data, which provide an insight into the breakdown of the data on persons in employment by demographic characteristics (for example sex and age) as well as into changes in the world of work (for example in respect of growth of part-time work and of other forms of employment differing from the normal employment relationship). 6 European System of National Accounts (ESA) 1995, Section 11, Sub-section Review of further developments in revisions of Employment Accounts since the beginning of the 1980s Due to the close link with the National Accounts fundamental revisions and further methodological developments of the Employment Accounts are only with few exceptions integrated in terms of time and content in the general revisions of the National Accounts, and these are usually carried out at intervals of five to ten years. The major National Accounts revisions are used in particular to thoroughly check the calculations, to incorporate results from censuses only carried out at intervals of several years, to implement new methods and processes, and, where necessary to adapt the conception to over the course of time changing requirements. In this framework any necessary adjustments to account for new classifications of economic activity are carried out as well. Summary 1 gives a chronological overview of the revisions of the Employment Accounts carried out since the 1980s. Below first of all content-wise important developments from previous revisions are summarised before the latest National Accounts revision of 2011 is looked at in view of these historical developments. The quality of the Employment Accounts is entirely dependent on the original data available. At the time of the revision in 1990/1991 for the last time, data on the labour market participation of the population was presented with a detailed regional and functional breakdown being available from a population and establishment census of non-agricultural local units. According to the results of the population and establishment census of 1987 the time series from 1970 published up to that point in time by the Federal Statistical Office from the Employment Accounts for the territory of the Federal Republic of Germany, which had last been revised in the course of the 1982 revision, underestimated the prevailing level of employment by more than a million persons. However, the results of a population and establishment census available again for the first time in almost 20 years were not only the reason for a readjustment of the time series, as they also formed the basis for the future co-ordinated national, state and district calculations of employment based on a new integrated concept for the calculation of employment figures (Schoer/Hanefeld 1990; Schoer/Winter 1992). As an initial measure for this purpose co-ordinated original data were determined by economic sector and by status of the persons in employment in all required categorisations and delimitations for the national, state and district regional levels. The key data on persons in employment according to status in employment were for the most part taken from the population census. The basis for the further breakdown of the employment figures by industry was the results of the establishment census, which were supplemented to account for missing parts and adjusted for cases of multiple employments in order to ensure the Employment Accounts persons-related concept was respected. Conversion of the employment figures for the various publication concepts national/regional labour market monitoring and National Accounts was car- Statistisches Bundesamt (Federal Statistical Office) German version published in Wirtschaft und Statistik 5/2012, page 385 et seq. 387

4 Summary 1 Revisions to the Employment Accounts since the 1980s Subject of the revision / Further development of the Employment Accounts 1982 > Revision of the previous employment time series based on the 1970 population and establishment census > Incorporation of the results from complete counts for economic sub-sectors (for example census of distributive trade and the hotel and restaurant industry, crafts census) 1986 > Implementation of monthly Employment Accounts for the territory of the Federal Republic: First publication of national monthly employment figures according to the domestic and national concept 1990/1991 > Incorporation of the results of the 1987 population and establishment census into the Employment Accounts of the federal government > Transition to the integrated concept for the calculation of employment figures : Joint computational model for national, state and district determination of co-ordinated employment figures for the purposes of labour market reporting and the National Accounts > Implementation of Employment Accounts for the new German federal states comparable to those of the former territory of the Federal Republic based in particular on employment surveys 1999 > Modification of the National Accounts to ESA 1995 and conversion of the 1979 classification of economic activities to European NACE Rev. 1 > Derivation of new initial figures for pan-german monthly Employment Accounts pursuant to the Classification of Economic Activities, 1993 edition (NACE Rev. 1), and the domestic concept of the National Accounts > Separate inclusion of marginally employed persons based on the Microcensus supplemented by data from unofficial surveys > Enlargement of the Employment Accounts by economic industry with a sectoral breakdown as internal estimation variables for the National Accounts 2000 > Incorporation for the first time in the Employment Accounts (interim revision) of the data available from the reporting procedure for social insurance for marginally gainful employed persons 2005 > Integration of persons with one-euro-jobs (work opportunities for welfare recipients) in the Employment Accounts and separate calculation of the number of persons with marginal short-term employment > Publication for the first time of annual data on marginal employment from the Employment Accounts > Modification of the Employment Accounts according to the Classification of Economic Activities 2003 (NACE Rev. 1.1) > Establishment of monthly flash estimates of the Employment Accounts: Publication of the national results for Germany 30 days (previously 70 days) after the month under review together with unemployment figures/rates pursuant to the ILO concept 2011 > Revision/updating of the Employment Accounts time series > Conversion of the time series on the Classification of Economic Activities, 2008 edition (NACE Rev. 2) Maximum need for revision m; % (1976/77) + 1.2m; % (1989) + 2.0m; % (1998) + 1.8m; % (1999) + 0.4m; % (2004) + 0.3m; % (1999) 1 Data on need for revision refers to the total annual average number of persons in employment in the former federal republic respectively Germany. 2 Marginal employed persons are wage earners and salaried employees who are not engaged in a job subject to full social insurance contributions, but apply to the Labour-Force concept of the ILO as persons in employment. ried out by means of reconciliation items, which had to be determined among other things for the differences between the establishment and the enterprise concept and for the calculation of the net commuter flows, being the difference between the national and the domestic concept. For the territory of the Federal Republic the joint calculation model provided results in a breakdown according to status in employment and industry delimited pursuant to both the domestic and the national concept. 7 As no further classical population and establishment censuses in the form of complete counts have taken place since 7 For the purposes of labour market monitoring the Federal Statistical Office published in addition the average annual, semi-annual and quarterly employment figures delimited according to the national concept broken down by sex. 388 Statistisches Bundesamt (Federal Statistical Office) German version published in Wirtschaft und Statistik 5/2012, page 385 et seq.

5 then, in all subsequent revisions employment statistics sources have had to be used for the level adjustment and determination of the new initial figures for updating, which provide as complete an overall picture as possible in respect of the breakdown of the persons in employment by industry sector over all statuses in employment. These included in particular the employment statistics of the Federal Employment Agency (wage earners and salaried employees), the statistics on public service personnel (civil servants), and the Microcensus (self-employed persons and family workers). Reconciliation of the initial figures broken down by industry with the National Accounts enterprise concept was henceforth carried out by means of data from on-going enterprise surveys and from censuses for certain industries, including in particular the crafts census and the census of distributive trade and the hotel and restaurant industry. From the 1999 revision separate initial figures pursuant to the national concept in a breakdown by industry could no longer be derived with the source data available. Therefore, following integration of the new states into the calculation model, the Federal Statistical Office continued the calculations according to the national concept only at macroeconomic level for the purposes of labour market monitoring. This appeared justified, among other things, of the fact that from 1991 onwards annual data on employment were also available from the Microcensus on an on-going basis for the whole of Germany in a regional breakdown according to the national concept for a large number of structural characteristics among other things classified by industry. Implementation of pan-german Employment Accounts Following the opening of the inner German border and the subsequent German unification the labour market in the new states and in East Berlin was marked by extensive structural changes. The requirement for data empirically demonstrating this development was great. At the beginning of 1995 the first results of Employment Accounts were provided for the new states which were methodologically and conceptually comparable to the calculations made for the former territory of the Federal Republic (Statistische Ämter des Bundes und der Länder 2004: 9 seq.). These calculations were in particular based on the results of the new federal states employment survey of November 30, 1990, which after certain modifications approximately fulfilled the function which the 1987 establishment census has taken over for the Employment Accounts in the former territory of the Federal Republic. In the 1999 revision the Federal Statistical Office placed the National Accounts, including the Employment Accounts, on a pan-german basis. The key figures on employment in the former territory of the Federal Republic and in the new states published on an on-going basis by the Federal Statistical Office up to that point in time were based on separate calculations. With the transition to pan-german monthly employment accounts the separate calculations for the former territory of the Federal Republic (including West Berlin) and for the new states and East Berlin were discontinued. Regarding this territorial breakdown the Federal Statistical Office only calculated and published revised results for 1991, and these formed the key data for the subsequent back-calculation of the revised time series to 1970 for the former territory of the Federal Republic. 8 Full coverage of all persons in employment pursuant to the ILO definition As explained earlier, under the 1995 ESA a broad definition of persons in employment is to be applied in the Employment Accounts, which requires that in accordance with the recommendations of the ILO persons who are only occasionally or marginally employed must also be included in the calculation of the average employment figures for the period under review. 9 This requires that this group of persons is included in the sources being used for the Employment Accounts. This can be carried out either implicitly, by means of employment statistics sources in which marginally employed persons are included in the total number of persons in employment, or explicitly with the help of data sources from which separate information is available for the respective groups of persons in employment. Marginal employment 10 was increasingly brought to public attention in the 1990s in connection with the debate about the social insurance arrangements for this form of employment. In this context an increasing volume of data was collected on marginal jobs. 11 The available statistics indicated very dynamic growth in marginal employment. The available information also led to the conclusion that the number of marginally employed persons in the employment accounts from the data sources for wage earners and salaried employees had previously been under-represented. In view of this, in the 1999 revision the Federal Statistical Office considered it necessary to broaden its calculation model for employment to include differentiated separate calculations for marginal employment. The results of the Microcensus were initially used as a basis, its survey programme having since 1990 included a question on the performance of marginal employment. The Microcensus results for the number of exclusively marginally employed persons were supplemented in the Employment Accounts by data from the available unofficial surveys on marginal jobs and secondary jobs, which in particular more fully depicted the grey area of domestic personnel. In 2000 it nevertheless became necessary to revise the internal calculation methods for marginally employed persons. After marginal employment had been included in the 8 Data for East and West Germany respectively with and without Berlin have since then been available from the regional calculations of the working groups Erwerbstätigenrechnung des Bundes und der Länder ( and VGR der Länder (www. vgrdl.de). 9 Persons not in employment for the entire period under rewiew are to be taken into account on a pro rata basis for the purposes of averaging. 10 Marginal employment is dealt with in Section 8, Fourth Book, Social Code (SGB IV). A job is considered to be marginally remunerated where the resulting remuneration is regularly below EUR 400 per month (so-called 400 euro job). Employment is deemed to be short term where the activity is limited to two months or 50 working days within a calendar year, unless it is engaged in professionally and the monthly remuneration exceeds EUR 400 (since the year 2013: EUR 450). 11 Worthy of particular mention in this connection are the studies on marginal employment and secondary occupations carried out by the Institute for Applied Social Sciences (ISG) in Cologne in 1987, 1992 and 1997 on behalf of the Federal Ministry of Labour and Social Affairs. Statistisches Bundesamt (Federal Statistical Office) German version published in Wirtschaft und Statistik 5/2012, page 385 et seq. 389

6 employers social insurance liability in the wake of the new regulations 12 in April 1999, the first results of evaluations from the reporting procedure became available in the summer of The new figures from that register, at about four million marginally gainful employed persons, were about a third above those recorded up to this point in time in the Microcensus sample. 13 After completion of the 1999 revision the annual average employment figure for Germany for 1998 was above the previous result by about two million persons (+ 6.0 %) and for 1991 by 1.2 million persons (+ 3.4 %). Due to the 2000 interim revision on recalculating the number of marginally employed persons the 1999 annual average employment figures for Germany had to be increased again by up to 1.8 million persons (+ 5.1 %). In order to be able to adequately reflect further growth of marginal forms of employment in the Employment Accounts, the calculation model was again enhanced in the 2005 revision in two respects. Firstly, from this point in time onwards separate calculation methods were also used for persons marginally employed on a short term contract. 14 However, updating of the results was at times based on model calculations, after the Federal Employment Agency had to temporarily suspend publication of the data available from the social insurance reporting procedure due to cumulative effects. Following the 2011 revision it was possible to carry out the calculations for persons marginally employed on a short term contract on a new, improved basis. The second enhancement, which was implemented in the 2005 revision, related to the explicit inclusion of persons taking advantage of a work opportunity with compensation for additional expenses 15 (so-called one-euro-jobs ). Pursuant to the criteria of ESA and ILO this group similarly belongs to the number of persons in employment, but in the absence of suitable data sources could not previously be taken into account in the Employment Accounts. The referred to enhancements to the Employment Accounts finally resulted in publication of the number of marginally employed persons, which includes all persons who, as employees, while not participating in activity subject to full social insurance contributions, are nevertheless deemed to be persons in employment pursuant to the Labour Force Concept. As calculation of the number of marginally employed persons is in some cases still dependent on estimates, publication in the subject matter series of the Federal Statistical Office (Statistisches Bundesamt 2012: table 12 Law of 24 March 1999 on new regulations for marginal employment relationships (Federal Law Gazette I, page 388). 13 The differences in the results between the Microcensus and the employment statistics of the Federal Employment Office, in particular in respect of marginally employed persons, were recently the object of a study jointly executed by the Federal Statistical Office and the Federal Employment Agency with financial assistance from the European Commission. A whole series of possible causes of inconsistencies between the two sets of statistics was identified. Among other things it was confirmed that under-representation of marginal activities in the Microcensus (and in other population surveys) is possible because respondents tend not to mention smaller activities at all, especially if they do not mainly define themselves as persons in employment. Also see Körner et al. (2011) Körner/Puch (2012). 14 See footnote See Section 16d, Second Book, Social Code (SGB II) ) continues to be limited to reporting of the national annual figures from 2003 onwards. Periodicity and timeliness of the employment figures: the current two-stage calculation model of the Employment Accounts of the Federal Statistical Office The National Accounts traditionally provide quarterly results for the on-going short-term economic monitoring and detailed annual results for the comprehensive analysis of economic activity in Germany. For on-going labour market reporting in the framework of the business cycle analysis, monthly data which are as up-to-date as possible are required in addition. For many years only the monthly data of the Employment Agency, in particular on the number of registered unemployed, were available for these purposes. In 1986 the Federal Statistical Office for the first time issued total monthly employment figures for the former territory of the Federal Republic from Employment Accounts, which were published in parallel to the labour market reporting of the then called Bundesanstalt für Arbeit (the former name of the Federal Employment Agency) (Becker/Schoer 1986). With the 1990/1991 revision the entire Employment Accounts for the territory of the Federal Republic in all categorisations were rearranged according to a monthly method of calculation. The first results of the differentiated monthly Employment Accounts were available about 70 days after the end of the respective month under review. In the course of the work on the 2005 revision the monthly Employment Accounts were expanded into a two-stage calculation model. Since then a distinction has been made between initial provisional estimates at national level for the purposes of short-term labour market monitoring (flash estimates), whose results can be published as early as 30 days after the end of a month under review (see Summary 2), and the still differentiated monthly calculations Summary 2 The flash estimates are closely related to the introduction of current, internationally comparable monthly labour market statistics based on the concepts and definitions of the International Labour Organization. From 2005 the so-called ILO labour market statistics filled a gap in the data in respect of the provision of internationally comparable monthly unemployment figures and rates (Janke/Riede 2003; Rengers et al 2008). Until August 2007 the unemployment figures for these statistics were provided by a telephone survey of limited duration, and since then they have been based on monthly evaluations of the Labour Force Survey. By contrast, the monthly employment figures used to calculate the denominator of the unemployment rate initially came from the Employment Accounts and the flash estimates, whose results, together with the monthly unemployment figures and rates for Germany, are published in a very timely manner which means 30 days after the end of the month under review. Since March 2011 the monthly unemployment rates have exclusively been based on data from the Labour Force Survey. 390 Statistisches Bundesamt (Federal Statistical Office) German version published in Wirtschaft und Statistik 5/2012, page 385 et seq.

7 Summary 3 The components of the two-stage calculation model of the monthly Employment Accounts in the territory of the Federal Republic Method of calculation Conception Computational detail of the source or indicatorbased calculations Published results 1st stage Flash estimate Current month under review (t 1) Macroeconomic expert estimates (sourceor indicator-based) supplemented by mathematical and statistical procedures and co-ordination of the results determined by different methods Status in employment > Self-employed persons (with/without employed persons) > Family workers > Wage earners/salaried employees (without marginally employed persons) > Marginally employed persons > Marginally gainful employed persons > Marginally short-term employed persons > Persons in work opportunities > Civil servants (including military personnel) National economy > Monthly total employment figures (national/domestic concept) 2nd stage Industry sector estimate Current month under review (t 1) and previous periods Differentiated source- and indicator-based calculations with detailed breakdown of calculation ranges (bottom-up method) Status in employment > as 1st stage Industry sectors > Currently: Up to 100 two- and three-digit sectors of WZ Institutional sectors > Non-financial corporations/ households > Financial corporations > General government sector > Non-profit-institutions serving households > Monthly/quarterly total employment figures (national/ domestic concept) > Quarterly/annual employment figures (domestic concept) by occupational status (persons in employment/ employees) and by industry sector as well as for the general government sector (total employees) 1 Classification of Economic Activities, 2008 edition. in a breakdown by industry and institutional sector, whose results are included in National Accounts and published in a breakdown by industry aggregated into quarterly and annual average figures for persons in employment (industry sector estimates). 16 Summary 3 provides an overview of the most important elements of the current two-stage calculation model. The conception of the monthly flash estimates at the first calculation stage has already been reported on in detail in this journal before (Fritsch 2006; van der Wijst 2008). At this juncture the three components or pillars of the flash 16 The results obtained from calculation of the Employment Accounts by institutional sector primarily serve as internal estimation variables for processing in the National Accounts. Currently only annual data on the number of employees in the general government sector are publicly accessible [(see Statistisches Bundesamt (2012: Table )]. estimate as they are portrayed in Chart 1 should therefore just be briefly outlined: The main pillar is comprised of source-based expert calculations, carried out on a differentiated basis according to the status in employment, but exclusively at aggregated national level. As these calculations and estimates are based on a still very incomplete data basis, 17 in addition mathematical and statistical forecasting procedures are used which can be carried out in the short term and independently of the inclusion 17 First employment data available in the reference month are based on first results of the Microcensus, a first estimate of persons paying contributions according to the Social Code, Book III (employees subject to social insurance contributions), data from the statistics of the Federal Employment Agency on one-euro-jobs (work opportunities for welfare recipients), reports on the number of staff in specific business sectors (e.g. post and telecommunication, German Railways), reports of the Federal Ministry of Defence on the number of people in the Federal Armed Forces and other reports on people doing voluntary military service or engaged in Federal Voluntary Service (Bundesfreiwilligendienste). Chart 1 The three pillars of the flash estimate on employment (1) Expert estimation (3) Matching (2) Mathematicalstatistical procedures Estimate of employment figures for the current reference month by nine statuses in employment on macroeconomic level based on limited data sources Matching results of pillar (1) and (2) together with final evaluation of estimate taking into account monthly provisional data processing of the continuous Microcensus for the current reference month Estimation of employment of the current reference month by > seasonal ARIMA-(Auto- Regressive-Integrated-Moving- Average-)forecasting methods > indicator-based ADL- (Autoregressive-Distributed-Lag-) models (single equation case) Statistisches Bundesamt (Federal Statistical Office) German version published in Wirtschaft und Statistik 5/2012, page 385 et seq. 391

8 of employment statistics data sources. The results determined independently of each other according to the two methods are subsequently evaluated against the backdrop of general economic performance and amalgamated into a final result. At the same time the monthly results of the Microcensus sample distributed over the full calendar year, which has been carried out since 2005 which are currently available about 20 days after month s end and further macroeconomic labour market indicators are included. This mix of various methods ensures a high degree of flexibility and utilises potential synergy effects by means of comparisons. Thus the time series analysis and the regression approaches provide important support in estimation of the prevailing trends and seasonal movement in employment. By contrast peculiarities and irregular effects, for example the influence of weather factors or new statutory rules with employment policy-related effects, are taken into account in the expert calculations. Thanks to the enhancement of the Employment Accounts by means of flash estimates it became possible to create a very up-to-date monthly total employment indicator of labour market development in Germany and at the same time also comply with the European requirements in respect of timely provision of quarterly employment data. 18 The extent to which this considerable gain in timeliness has affected the quality of results can now be seen after many years of experience. Table 1 shows the most important error measures, relating firstly to the original values and secondly to the particularly interesting rates of change, in the current reporting on the flash estimates for employment figures published after 30 days. The error measures take into account all on-going revisions of the initial provisional estimation results in the 18 The quarterly total employment figure resulting from the National Accounts is one of the primary indicators for the monitoring of European economic and financial policies and should be available for these purposes after 45 days. For Germany this Principal European Economic Indicator (PEEI) from the employment flash estimates is similarly available after just 30 days. Table 1 Error measures for the monthly flash estimates on employment 1 Original values Relative year-on-year change % Percentage Points Average Revision (AR) Mean Absolute Revision (MAR) Standard deviation (RMSE) Referring to the period from February 2005 to June 2011 for the monthly number of persons in employment for Germany as released after 30 days. 2 Root Mean Square Error. period from 2005 to the summer of 2011 without the corrections resulting from the fundamental revisions of the calculations in the course of the major revisions of 2005 and The negative figures for the average revision (AR) show that the initial provisional results tended towards a very slight underestimation of the almost constantly positive growth in employment in the period of observation. However, the average correction requirement in the provisional estimation results measured in terms of the mean absolute revision (MAR) and the root mean square error (RMSE) with values of less than 0.3 % and percentage points is by no means worse than the employment figures previously published after 70 days. However, with respect to the evaluation of an estimation method it is particularly important that cyclical turning points are recognised at an early stage. In Chart 2 various publication statuses for the seasonally-adjusted monthly time series for employment in Germany are shown. The time series includes turning points at the beginning of 2006 and at the beginning and end of the short fall in employment in 2009 caused by the economic and financial crisis. The fraying of the time series at the lower turning point in 2006 indicates that this turning point in the results obtained from calculation of the Employment Accounts was only correctly Chart 2 Comparison of publication points in time of monthly time series on employment at different points of time million persons 41 June Dec 10 June Mar 09 Dec 09 Mar Mar 06 Sep 06 June in Germany (resident concept). Seasonally-adjusted results with census-x-12-arima Statistisches Bundesamt (Federal Statistical Office) German version published in Wirtschaft und Statistik 5/2012, page 385 et seq.

9 depicted after a lag of a few months, while the supposed turning points at the beginning and end of 2009 were very quickly recognised as only being a temporary interruption of the employment boom. Recognition of turning points in a time series is easier where the prognostic share of the estimates based exclusively on historical values is as small as possible and the (real) statistical information included in the estimate for the latest month under review is as great as possible. In the threepillar model of flash estimates of employment, the monthly pre-evaluations of the Microcensus in particular must provide labour market data for the current month under review. Nevertheless, due to the use of the random sampling method the monthly time series from the Microcensus resp. the Labour Force Survey are subject to relatively high volatility, which can impair the evaluation of the results from a cyclical point of view (Rengers 2011). All the employment statistics information still not available at the point in time of the flash estimates is gradually included in the differentiated estimates by industry at the second calculation stage. At the same time special account is also taken of cost structure statistics and other business surveys for economic sub-sectors which are particularly focused on the requirements of the National Accounts (domestic concept/enterprise concept). 19 The total employment figures derived by means of summation of the results by industry and institutional sector (bottom-up) replace and firm up the first provisional estimates. However, at national level one should note the coherence with the results of the employment register statistics on employees subject to social insurance contributions and marginally employed persons, with the results of the statistics on public service personnel on civil servants, and with the results of the Microcensus on self-employed persons and unpaid family workers. Introduction of new classifications of industries in the Employment Accounts Until the 1990s, when the Federal Statistical Office published industrially broken-down employment data as results obtained from calculation of the Employment Accounts not exclusively according to the domestic concept, but still also according to the national concept, the respective presentations were moreover based on various different classifications of economic activity. While the national figures published in the context of the demographic and labour market statistics were broken down in the respective versions for population and occupation censuses according to the classification of economic activities (WZ), 1970 and 1979 editions, in the presentations of the National Accounts (domestic concept) versions derived from the basic classifications were used which, besides activity-related functional criteria for the classification of economic units, also took into account the sponsorship characteristics ( Trägereigenschaft ) for a classification within economic sectors according to institutional criteria. On this basis it was possible to 19 The Appendix to this paper shows the current source statistics for calculation of the Employment Accounts in the territory of the Federal Republic. make the distinction, which was important for the National Accounts, as to whether given services were rendered by private enterprises, government institutions or private nonprofit institutions serving households. In the 1999 revision the Classification of Economic Activities, 1993 edition (WZ 93), replaced WZ 79. WZ 93 is the German version of the Statistical Classification of Economic Activities in the European Community, NACE Rev. 1, which in 1999 jointly with ESA 1995 became mandatory for the drawing up of National Accounts in all member states of the European Communities. In contrast to WZ 79 (version for the National Accounts), in WZ 93 institutional classification factors became less important. Thus it was no longer possible to derive the institutional sector breakdown directly from the classification of economic activities, which consequently in turn also made the institutional sector estimate introduced into the Employment Accounts of the Federal Statistical Office in the course of the 1999 revision in addition to the industry sector estimate more difficult. This from the point of view of the National Accounts detrimental aspect also applies to the two classifications of economic activity that followed WZ 93 and NACE Rev. 1. In the 2005 revision it was WZ 2003 and NACE Rev. 1.1 which differed from their respective predecessors, albeit only in respect of a few new items and designations. In contrast, the implementation of WZ 2008 and NACE Rev. 2 in the 2011 revision caused substantial changes. 4 Changeover of the Employment Accounts to WZ 2008 and revision of the results for the period 1991 to 2010 in the National Accounts revision of 2011 For several reasons changeover from WZ 2003 to the Classification of Economic Activities, 2008 edition (WZ 2008), carried out in the 2011 revision represented a special challenge for all sub-systems of the National Accounts that had their calculation being based on a detailed breakdown by industry. Most importantly one should mention the considerable structural differences already referred to between WZ 2008 and the previous WZ These are extensively described in the comprehensive article on National Accounts on the 2011 revision in this journal. 20 Therefore the more pronounced differentiation between the service industries and the creation of a new section, Information and Communication, which among other things also includes the publishing industry, that was previously included in the manufacturing industries, is referred to as an example. A further problem to be solved was that data originally collected following the new WZ 2008 from structural statistics and other important calculation bases at the point in time of the revision were for the most part only available for one 20 See Räth/Braakmann (2011). Besides much other information on the 2011 revision the article contains a detailed summary of the new subdivision of the National Accounts results to be published under WZ 2008 with all the aggregation stages together with a description of the most important innovations of WZ 2008 compared to WZ Statistisches Bundesamt (Federal Statistical Office) German version published in Wirtschaft und Statistik 5/2012, page 385 et seq. 393