The article refers to the shift in methods to conduct a population census: from a conventional enumeration through a sample survey and a mixed approach to administrative data, as a new standard in statistics. The paper compares two Polish censuses of 2002 and 2011. It is aimed at quality assessment in the case of both: the traditional method (2002 census) and the combined approach (2011 census). The quality of census data is discussed with essential aims and objectives to provide reliable information on the population age and sex structure in detailed territorial division. Therefore, quality assessment is provided for the whole country and at regional level. First of all, coverage errors are considered. We use multiple sources of data and non-matching methods, in particular: demographic analysis based on previous censuses, vital statistics and a comparison with other existing sources. Different cross-sections according to sex, age and place of residence are considered. In each separate domain adequacy and divergence assessments are accompanied by an attempt to provide substantive explanations.