The Regular Labour Force Survey as a Quality Survey of the Finnish 1985 Census by Aarno Laihonen' Central Statistical Office ofFinland The paper deals with experiences from using the sample- based regular monthly Labour Force Survey as a quality survey of census data on the economic activity of the population. The method uses record linkage at the micro level between the data of persons in the Labour Force Survey sample and the data of the same persons in the census file. An exact linkage is made possible by the uniform personal identifier used in the Finnish popula- tion registration system. The use of the Labour Force Survey as a reference quality survey of the census was possible because the survey week and the census week coincided, and because the survey and the census measured the same variables according to the same concepts and classifications. In this way, savings were achieved in the cost of the census quality survey. Fot the purpose of the quality survey, after the regular survey interview a few additional questions were asked of a sub-sample consisting of one- fifth of the Labour Fwce Survey sample (about 2,300 persons). To ensure as errorless results as possible, the data of the sub-sample were reprocessed after being entered and coded as usual. The additional information on the sub- sample was utihzed when estimating the final errors on the basis of a micro comparison between the census data and the survey data of the whole Labour Force Survey sample. Introduction In Finland, modem population and housing censuses have been carried out in 1950, 1960, 1970 and 1980, and so-called mid-decade censuses in 1975 and 1985. The latest census of 1985 was carried out within extremely tight budget constraints imposed on the Central Statisti- cal Office by the Ministry of Finance. The tight budget constraints also affected the production of census data. New cost-saving devices had to be used. A general outline of the 1985 census To give an idea of the cost frame, the direct costs of the 1980 census amounted to about 80 million marks (about 17 milhon US dollars) and those of the previous mid- decade census of 1975 to about 26 million marks, in 1986 prices. The total expenditure of the 1985 census was not to exceed 18 million marks. The central point of departure for the planning of the system solutions of the November 1985 census was to minimize the amount of manual work in census data collection and processing. First of all, data collection was minimized by an exten- sive use of registers and administrative records when gathering the basic census data. This was largely made possible by the comprehensive, high-quality population registration, taxation and social security systems charac- teristic of all the Nordic countries. The use of registers and administrative records in population and housing censuses has increased steadily since the 1970 census. This development has been aided by the widespread use of the uniform personal identifier in different registers and administrative records. A significant improvement in the register situation, which helped a great deal in bringing down the cost of the latest census, was the establishment of a building and dwelling register for Finland on the basis of the 1980 population and housing census. This register is operated in connection with the Central Population Register, and it allows the Unkage of persons and dwellings. Therefore, no questions on housing were needed on the 1985 census form. The census of November 1985 used only one question- naire, namely for gathering employment data. The Central Population Register was used as the mailing list for the population of working age. The questions and instructions on the censusform are presented in Annex 1. The questionnaires were sent out by mail from the Central Statistical Office (CSO) and were returned by mail to the CSO. No local census organization was used. Another special feature of this census was that the census form was preprinted, not only with the respondent's name and address but also, for about half the population, with the name of the respondent's workplace as it appeared in the 1980 census and with the respondent's occupational title obtained from the Central Population Register. Persons obligated to respond needed only to report any changes or errors that had occurred in this information. The census forms went directly to data entry, which was carried out as key entry. Naturally, only changes and additions on the forms had to be keyed. In this way ASSIST Quarterly complete 'pictures' of the census forms were converted to machine-readable form before any other processing operations were performed. This enabled batch mode checking and correcting of the form data, leaving only about 10 per cent of the forms to be checked and cor- rected manually, on terminals. Next, extensive automatic coding was applied to workplace and occupation data. Thus, the number of forms requiring manual processing was drastically reduced in all phases of processing. This reduced the total cost of the census and allowed prelimi- nary publication of the most essential census data as early as December 1986. In the final phase of data collection, register data were also used to obtain, by imputation, the census form data of non-respondents. In this way, a satisfactory 98.6 per cent total coverage was achieved for the central data of the census form. This also contributed to the relatively small regional variation in coverage, even though the census was carried out as a direct mail-out, mail-back system without any local census organization and with only one reminder sent to non-respondents. Register imputation of questionnaire data was tried for about 139,000 persons (3.7 per cent of the population of working age), 84,000 of whom were non-respondents and the rest persons whose responses were incomplete. The regular Labour Force Survey, the study week of which coincided with the census week, and the 1985 household survey were used as quaUty surveys of the census. According to the quality surveys, the general quality of the 1985 census data is significantly better than that of the previous mid-decade census of 1975. The quality of employment data is in part slighdy inferior to the quality of employment data in the 1980 census. The setup of the quality surveys of the 1985 census Because of the high-quality, up-to-date information obtainable from the Central Population Register (CPR) in Finland, the main purpose of the population and housing census is not to count the population, but to produce data on the economic activity and housing conditions of the whole population. The resident population of the country as registered in the CPR was the population of the census. Thus, from the point of view of the quality of the census data, there was not, by definition, any undercount or overcount of the population. The aim of the census quality surveys was to analyze the quality of the data produced on different attributes of persons, dwellings and buildings. However the problem of under- and overcount was still relevant for dwellings and buildings because of the shortcomings of these data in the CPR. The quality surveys of the 1985 census fall into two categories: those analyzing the quality of the CPR data (especially the data on household-dwelling units and dwellings) and those analyzing the census form data on the economic activity of the population. The data on household-dwellings units and dwelUngs, for instance, were analyzed by comparing the census data with corre- sponding data from the 1985 Household Budget Survey, an interview-based sample survey of 12,(XX) households. Another source of dwelling data was a sample survey of dwellings registered as unoccupied in the CPR. This survey provided information on the overcount of dwell- ings in the CPR. The quality of the data on the economic activity of the population was analyzed with the help of processing error studies (data entry errors and errors in coding and editing) and a special quahty siuA'ey in which the final census data on persons were compared with the checked and corrected data of the interview-based regular Labour Force Survey. Some experiences from this survey and the methodology of the survey will be discussed in this paper. A short general description of the regular Labour Force Survey will be presented in the next chapter. The Finnish monthly Labour Force Survey The Finnish Labour Fwce Survey (LFS) is a sample survey based on a random sample of 12,000 persons selected from among the population aged 15-74 years. Data collection takes place mainly in personal interviews carried out by the CSO's interview organization. The person interviewed is asked questions about his labour force participation (current activity), employment, unemployment, workplace, occupation, industrial status, time use, days and hours actually worked, overtime and secondary jobs, and normal hours of work. About 94 per cent of the interviews are telephone interviews and five per cent personal interviews. About one per cent of the answers are obtained using a mail questionnaire. The average non-response rate of the survey is about 4.7 per cent Structurally, the survey is a so