Current quarterly indicators on private household consumption

| Last update: 18.08.2023

Bild – experimental statistics


During the COVID-19, the restrictions imposed on everyday life and the economy had a considerable impact on the household budget. In light of this situation, it was important to use all available data sources to be able to best estimate the effects of the pandemic on consumption in particular, but also on the household budget in general.

The Household Budget Survey (HBS) is particularly suited to an in-depth and temporal analysis of the effects of the pandemic, as it is continuous and covers every day of the year. This means that its latest data could be analysed and used very quickly.



These highly topical results are of obvious interest to the public and are therefore made available here. Furthermore, it is the first time for these results to be published with a time dimension of less than one year. The quarterly form was chosen because it allows the development of consumption and other components of the household budget to be studied with a sufficient number of observations. The quarterly averages for the years 2015 to 2017 combined are used as a benchmark for comparison for each quarter.

Since these are estimates of the budget of all households in Switzerland that are based on a relatively small sample (900 households per quarter), particular attention must be paid to the uncertainty associated with these estimates. Thus, confidence intervals are provided with the results as a measure of this uncertainty to allow a better assessment of their quality. They make it possible to identify significant changes in relation to the baseline indicators.

The results are available in full in EXCEL table format for both manual and visual access and in CSV format for automated analyses.


The tables (see the links below) show all the items listed in the household budget which are affected in varying ways by the pandemic. To illustrate these effects, here is the evolution of two items as an example (see below for the example of reading the confidence intervals).

Expenditure on attendance of cultural establishments has been drastically impacted, particularly as a result of their enforced closure for extended periods. We can see these effects in the graph showing expenditure on the cinema.

In other areas, the closures and bans did not have a direct effect on consumption, but the related consumption was still indirectly affected, as shown for example by expenditure on fuel.

Example of reading a graph with confidence intervals

In these graphs, the 95% confidence interval is a range that contains the true mean value of the population with a probability of 95%. The width of the confidence interval depends on the sample size and the dispersion of the values measured. The larger the sample size and the narrower the spread of values, the smaller the confidence interval.

What is the purpose of a confidence interval?

  • To measure the precision of an estimator, and therefore its quality
  • To determine whether two estimators are significantly different from each other (see below)

Lesebeispiel für eine Graphik mit Vertrauensintervallen

In the second reference quarter, the estimated average monthly amount spent by a household on meals in restaurants, cafés and bars is CHF 191 per household, with a 95% confidence interval of ± CHF 12. This means that the true average amount spent per household is between CHF 179 and CHF 203, with a probability of 95%. In the same quarter of 2020, the estimate is CHF 69, with a 95% probability that the true amount is between CHF 58 and 80, the confidence interval being ± CHF 11. As the two intervals are very far apart (more than six times their width), it can be concluded that the average monthly amount spent by a household in Q2 2020 was significantly lower than the amount spent in the same reference quarter.

Where can this information be found in the corresponding EXCEL table?

  Year 2020
2nd quarter 3rd quarter
  A Q CI95 P A Q CI95 P
5311.01: Meals in restaurants, cafés and bars  69 d ± 11 0.7% 167 d ± 21 1.7%
A: Monthly amount in CHF per household, on average
Q: Quality of the estimate
CI95: Confidence interval at 95%
P: Distribution as % of the gross income

Example for Q2 2020 values:

  • Average monthly amount per household (CHF 69): Column A
  • 95% confidence interval (± CHF11): Column CI95
  • Upper limit of the confidence interval at 95% (CHF 80): Column A + Column CI95
  • Lower limit of the confidence interval at 95% (CHF 58): Column A − Column CI95



The reference population for the HBS is the permanent resident population in Switzerland in a private household. The survey is based on monthly random samples stratified by the seven main regions of Switzerland. To obtain a sufficient number of households in all regions, a non-proportional draw is made with an over-representation of Ticino. In 2020 and 2021, approximately 3600 households per year have participated or are participating in the survey, i.e. 900 households per quarter.

Collection of data

The HBS is carried out using detailed paper questionnaires with telephone support and is supplemented by telephone interviews. During the survey month, households complete the daily booklet. Telephone support is provided by a personal adviser throughout the survey period to support households in their task. At the end of the survey month, two other booklets (a booklet for recording regular expenses as well as major expenses in the months preceding the survey month, and a booklet for recording income) are completed and returned by the household to the institute, which will then plausibilise and enter the information. In case of any doubts or questions, clarifications are made with the households. The data entered are transmitted daily to the FSO, which then verifies and plausibilises the data and if necessary requests further information from households.

Stability of results

The availability of data received by the FSO almost in real time allows for the analysis and exploitation of household data as soon as these are complete and plausible. This makes it possible to start producing results on the initial data two or three weeks following the survey month. Obviously, these quick results are based on a limited number of respondent households immediately after the survey month, but this number increases as the data are entered and verified. The results therefore stabilise and hardly change at all roughly three months after the survey month.

However, there is an exception to this rule for “major” expenses from the previous month. This concerns several topics, in particular the purchase of vehicles and bicycles recorded retrospectively over 12 months, and also expenditure exceeding CHF 300 per good such as the purchase of a plane ticket or a piece of furniture, which is recorded over 6 months. This surveying over a longer period of time makes it possible to obtain six (or even twelve) times as many events and to thus considerably improve the quality of the corresponding area estimates. For these topics, particularly transport, the results stabilise after a longer period than the 3 months mentioned above.


As was the case for the standard exploitation of the HBS results, an optimised weighting method is used to reduce the risk of any potential bias related to the different response rates of households by their characteristics. The weighting is applied separately every quarter. This approach allows estimation of the consumption and budget of private households in Switzerland to be made for each quarter, with the corresponding confidence intervals. Weighting is carried out in three stages:

  1. Initial weights
    Households in the Household Budget Survey are selected using a random sample stratified by major region. The initial weights reflect the household’s probability of being selected according to the month and stratum. The Ticino region is over-represented. All households within a stratum receive the same initial weights for each monthly sample.

  2. Correction for non-response
    As with all surveys, some of the selected households do not respond. Furthermore, non-response varies according to household’s characteristics. A model is therefore used that contains these household characteristics so that responding households reflect the whole sample as closely as possible.

    The following two characteristics have proved to be the most relevant in the model:
    •   Household composition (on its own but also combined with age characteristics)
    •   Household income

    A more detailed description of the model actually used can be found in the documentation below. Needless to say, it is more complex and contains more characteristics than the two mentioned above.

  3. Calibration
    Once the initial weights have been corrected for non-response, calibration is carried out to adjust the household sample to make it as close as possible to all private households in Switzerland. The weights of each responding household are adjusted using additional characteristics that are available for all private households.

    The most relevant characteristics for calibration are similar to those used in the correction for potential distortion due to non-response (see above).

The weighting model for the quarterly sampling is similar to that used for standard HBS data analyses. However, as the size of the quarterly sample is only a quarter of the annual sample, this model has been slightly simplified.


Other documents (tables) are available on this page in French and German. Please change the language on the top right of this page.