The sample of households – as established in line with the specific sampling design – allowed us to estimate parameters with a known degree of uncertainty for all households in Austria.
To make sure that this estimation – for instance an estimated mean – is unbiased and subject to as little uncertainty as possible, we used weighted samples. Each household in the sample represents a particular number of households in the frame population. However, given the sampling design, different participation rates and possible deviations from known characteristics of the frame population, not every household in the sample represents the same number of households from the target population. Here is where weighting comes in: Its purpose is to determine how many households a single household in the sample will reflect in order to ensure that the estimations based on the survey will be of the highest possible quality.
Specifically, we use three types of weights for the HFCS in Austria.
First, we constructed design weights, which correct the bias resulting from the unequal probability of a household being selected given the sample design. The design weights ensure that parts of the frame population that would otherwise be underrepresented in the sample will be weighted higher, and vice versa.
Second, nonresponse weights help overcome the bias that arises from the fact that not all households in the sample could actually be interviewed successfully. Some types of households and households in some regions will be less accessible and/or willing to participate in the survey. Hence, types of households or regions with higher nonresponse rates will be given higher weights, and vice versa.
Third, poststratification weights serve to bring the sample in line with the characteristics that the target population is known to have on the basis of other sources. The idea is to put more weight on underrepresented households, and vice versa.
Together, these weights determine the final weight of each household, which indicates how many households of the target population each household represents in the end.
For detailed information on weighting in the first wave of the HFCS in Austria, see the methodological notes.