FAQ

The households participating in the HFCS were randomly selected. Every household in Austria had a positive probability of being drawn for the sample. Sampling was carried out applying a scientific statistical method. The addresses of privately used dwellings were selected from mailing address registers maintained by Austrian Post AG (“Adress.Certified” and “data.door”). Geographical categories of households were created (stratified two-stage sample design) to ensure that households in each Austrian province took part in the survey. A total of 6,280 households were drawn for the sample (third wave). Participation in the survey was voluntary, and a great effort was made to convince households to take part. Eventually, 3,072 households participated in the Survey (third wave), providing information on a wide range of sensitive and complex issues.

Representativeness is not, strictly speaking, a statistical concept. However, the term is generally used to express that the analysis of data from a sample can be expected to yield the same results as the analysis of the target population.

The sample design (based on privately used postal addresses) used in the HFCS ensures that all households in Austria have a positive probability of participating in the survey. Stratification is a cost-efficient way to capture different households as accurately as possible. Distortions in sampling caused by complete or partial nonresponse were corrected using state-of-the-art scientific methods. Therefore, the representativeness of the HFCS can be considered to be relatively high. Full representativeness in all dimensions of the survey is an unattainable goal. However, the methods used allow correct variances of each estimator to be calculated; these variances, in turn, make it possible to assess the degree of uncertainty attached to the estimations.

Face-to-face interviewing in the HFCS allowed respondents to provide informed answers; they were able to ask interviewers for clarification and to consult records, documents and contracts if necessary. Automated consistency checks made it possible to review answers even during the interview. Thorough training and preparation of interviewers ensured that the latter interacted with respondents on a high professional level and helped errors to be detected early on. In addition, expert evaluations were carried out to identify potential errors after the interviews and to conduct follow-up investigations, including contacting households again, if necessary.

All interviewers of the HFCS in Austria were experienced and, additionally, received intense training in the run-up to the survey. Routine checks of both interviewers and households were carried out after each interview. Furthermore, random checks in which households were contacted by telephone to review their interviews were made. Each interviewer’s performance was evaluated on the basis of a set of criteria; weaknesses were pointed out, and if severe transgressions were identified, the interviewer was dismissed. If an interviewer was excluded from the survey, the households that had already been interviewed by this interviewer were contacted again. In addition, experts reviewed all household data collected. Potential problems were identified and followed up, i.e. the households concerned were contacted again and the information compiled was reviewed. In those few cases in which it was impossible to correct an error, the data of the respective households were not used. All these mechanisms ensure that data from interviews that were not conducted correctly did not enter the HFCS dataset.

Imputations fill missing data. If the right methods are applied (like in the HFCS), the associations between all variables in the dataset are preserved. Multiple imputations also capture the uncertainty attached to imputed values. If no imputations were carried out, the survey results would be biased, as the group of households refusing to answer a specific question is not purely random. Imputation is an internationally scientifically recognized approach that has been used in numerous international surveys. Imputation models similar to those used in the HFCS in Austria have been applied to comparable datasets compiled by the Banco de España (in the EFF) or the Federal Reserve System (in the SCF).

Other major national surveys include, for instance, EU-SILC (Statistics on Income and Living Conditions), Statistics Austria’s Household Budget Survey and Micro census, and SHARE (Survey of Health, Ageing and Retirement in Europe), the Austrian part of which is coordinated by Johannes Kepler University Linz in collaboration with the Austrian Academy of Sciences. All these surveys have different objectives.

The HFCS aims at capturing households’ entire balance sheets, covering both flows (such as income and consumption) and stocks (such as debt and wealth). No other survey in Austria has this specific focus. While EU-SILC captures different types of income, it does not take into account debt and wealth. The SHARE project, in turn, covers only older age groups, as opposed to the HFCS, which covers all households in Austria. The information compiled in the HFCS makes it possible for the first time to analyze households’ complete balance sheets and net wealth.