Influence of the Calibration Weights on Results Obtained from Czech SILC Data Jitka Bartoˇ sov´ a and Vladislav B´ ına University of Economics in Prague, Faculty of Management, Jaroˇ sovsk´a1117/II, 37701 Jindˇ rich˚ uv Hradec, Czech Republic, {bartosov, bina}@fm.vse.cz Abstract. The purpose of income sample survey is to obtain a representative data concerning level and structure of incomes and fundamental social–demographic characteristics of households and their members in the Czech Republic. The sur- vey results are generalized to the whole population using the calibration weights created by Czech Statistical Office. An important question in the process of gener- alization of conclusions arising from the analysis of survey data is the influence of calibration weighting on the obtained results. The paper concerns the importance of calibration process in measuring the monetary poverty of Czech households (see also BARTO ˇ SOV ´ A, J. (2009) or BARTO ˇ SOV ´ A, J. and B ´ INA, V. (2009a)). Var- ious definitions of consuming unit are used in order to compare the influence of weighting on the relative measure of poverty in dependence on the chosen scale. Keywords: Calibration weights, EU-SILC, household incomes, poverty. 1 Introduction The results of Mikrocensus and EU - SILC surveys cannot be directly gener- alized on the whole population since they does not constitute a representative sample. This conclusion follows from the comparison with results of the Czech population census ”Sˇ c´ ıt´an´ ı lidu, byt˚ u a dom˚ u (SLBD)”. Hence, the simple generalization of information obtained from sample data could lead to sig- nificant distortion of acquired information. There are few reasons leading to the corruption of sample representativeness. The sample representativeness is primarily corrupted by various rates of success of interviewers in different regions (see BARTO ˇ SOV ´ A, J. and B ´ INA, V. (2009b)). But there is also apparent non-uniformity of successfully re- turned forms in different social groups (the highest rate of success is among the households of retired and the lowest rate is among the self-employed). Also the mean size of households differs from the size ascertained in SLDB. Accordingly, it is impossible to accomplish the generalization of results using simple coefficients taking into account only the number of surveyed house- holds in region in correspondence with its total count of inhabitants. And hence, for the construction of coefficients the iterative method of weight cal- ibration was used. This method minimizes the difference of estimated and