1990 POPULATION CENSUS Detailed data based on a 2 per cent representative sample (1992)

III. THE METHOD OF SAMPLING

were förmed in each county by the small (i.e. "not big") tonns and by the villages, repectively; the process of sampling in these strata was facilitated by a list of the localities sorted by size, which alsó contained a compound characteristic of the residential area which reflects in somé way the quality of housing. Sorting the localities by size has the effect of implicit stratification and improves the representation of data. Sampling of ED's in imer areas of localities was carried out independently of that in the outskirts; the two types of ED's were numbered in a way enabling easy distinction in this respect. Beside the above aspects the characteristic of residential areas played an important role in selecting ED's in the inner areas of localities. Owing to structural dissimilarities the residential areas defined in Budapest and in the "big" towns were different from those in other localities; e.g. traditional downtown areas and modern blocks are typical in the first group but practically non-existing in the second tfiile the opposite statement holds in the case of farms or village centres. On the other hand, industrial areas, recreational areas, mixed areas etc. may be found in small towns and villages as well as in big cities. i Residential areas among institutional ED's were as follows: - workers' hostels, - students' hostels, - nursing homes, children's homes - old people's homes. Selection of the 2 per cent sawple In the first step a 2.5 per cent sample was determined by selecting each eighth dwelling of the ED's with representative questionnaire; in the case of institutional households the procedure was slightly modified and each eighth person was selected. The selection process was centrally controlled — the starting random numbers for the county subsamples were generated at the Computer Centre of the CSO in Budapest —, but the phases of coding, data entry and preliminary checking were decentralized: they were accomplished at the régiónál offices of the CSO. The reason for selecting first a 2.5 per cent sample was the assumption that such a sample would have a subset which is a 2 per cent sample for the rtnellings and persons (including those living in institutions) of 78 area breakdowns. It was agreed that the final sample should have these properties, which imply the use of a common inflation factor of 50 — valid for each area unit considered — in the course of processing the sample survey. The breakdowns included the county seats and the counties themselves, furthermore the capital city and somé other cities of specific importance, etc. From the 2.5 per cent initial sample the final 2 per cent sample was obtained by a reduction and adjustaent process delineated below. In the course of reducing the initial 2.5 per cent sample the following aspects were considered: - data of persons who answered only the questions of the basic questionnaire and left those of the represesentative questionnaire unanswered were deleted; - if deleting the data of a person in the way just described led to confusion in the composition of the family, household or dwelling involved, all data of the units in question, were alsó deleted. After this reduction further adjustment was needed to obtain the final sample; for this purpose the preliminary data on dwellings and population were used as benchmarks. The final sanple was determined by a computer code which for somé fixed number of dwellings reduced the number of persons through a minimai number of changes of cfcellings. The fact that persons and dwellings were changed simultaneously by the computer code maintained the consistency of the sample in the sense that each person was associated with the dwelling in which he (she) actually lived. 238