Most of our surveys are being made on nation-wide representative 1000 or 1500-adult population random sample. The construction of this sample is following:
The area of the country was divided into 12 macroregions consisting of from 3 to 5 neighbouring provinces (voivodships). The starting point for establishing the macroregions was division of the country used for social and economic planning. Historical ties were also taken into consideration. Then each macroregion was divided into four "areas" - it was a result of differential analyses in sociological studies conducted by CBOS. In effect, we obtained a division of the country into 48 layers (12 macroregions x 4 "areas").
From the Central Statistical Office (GUS) we copied set of data on regions and districts created for the purposes of the National Census conducted in 1988. The data are updated annually and serve to create drawing operands of representative samples in studies made by various institutions. The set of GUS data is used to create a drawing operand of the first and second degree, that is, of statistical region and homes.
A three-degree drawing formula was accepted. The drawing unit of the first degree is the statistical region. Units of the second degree are homes, of the third - adult persons. The numerical size of the sample consisting of 250 or 375 statistical regions was divided proportionally to the total population in layer. Then using the formula of dependent drawing, the statistical regions are drawn, and from each statistical region - 6 homes. The selection of units of the third degree - persons to be interviewed - is made by a trained interviewer according to a precisely defined procedure. The interviewer draw one person from all adult persons actually residing in the home drawn whose address he/she received from CBOS. For drawing the respondent the interviewer uses a table of random numbers generated previously by a computer for the home drawn. The interviewer conducts an interview with person thus selected.
The sample realised does not encompass person drawn who were not present at the time the study was conducted, and neither does it include persons who refused to be interviewed, etc. Sometimes a reserve list is used. For these reasons the size of the realised sample is smaller than 1000 (for 250 regions) or 1500 (for 375 regions). The sample drawing method accepted is in original CBOS idea. Only certain elements of selecting the respondent in home were taken from one of the German public opinion and market study institutions. A detailed description of the algorithm of drawing the sample in the form of instructions for the programmer is available in the CBOS Samples Workshop.
We also carry out the surveys on other kinds of samples, i.e. quota samples, random-route samples, etc.