Statistical processing

Source data

The Household Budget Survey is calculated at household level, and is based on a combination of interviews and accounting of the participating households. All households are simply randomly selected.

In areas where data are already known through registers, data are taken from those registers. The survey used records from:

• Income Register
• CPR register
• BBR register
• Training Register and The Employment Classification Module
• Hospital Utililisation statistics

Frequency of data collection

Data are collected annually. Households participate continuously throughout the year in the survey. In this way we ensure that seasonal consumption are represented in the survey.

Data collection

An external service provider is responsible for collecting data for the Household Budget Survey. Households that are randomly selected receive an invitation letter via Digital Post (e‑Boks) and are subsequently contacted by telephone to encourage participation. When a household agrees to participate, it must keep a 14‑day consumption diary and complete a questionnaire (12‑month account) on the household’s fixed expenses and major expenditure items during the past year. The 14‑day diary is digital but can also be completed on paper. From 1994 to 2019, the 12‑month account was conducted exclusively through interviewer‑administered personal interviews (CAPI). From 2019 to 2021, data were collected through a combination of telephone interviews (CATI), web‑based interviews (CAWI), and personal interviews (CAPI). Since 2022, data have been collected solely through telephone interviews (CATI) and web‑based interviews (CAWI).

In 2024, the data collection tool was expanded so that households can scan or photograph their receipts, which are uploaded automatically. Items and amounts are categorized by purpose using AI.

Data from administrative registers are retrieved as of 31 December in the reference year, or from the most recent available year. If data are retrieved from an earlier year than the reference year, they are adjusted for price and volume to match the price level of the reference year.

Data validation

Interview data is validated both during and immediately after the visit interview. The validation during the interview consists partly of logical and partly of probable checks, while the validation after the interview is done manually. A logical check could be, for example, whether the household has a TV, but has not reported expenses for a license or antenna association, or that the household has a car, but does not report expenses for weight tax, car insurance, etc. A likely check could be, for example, that very high or low amounts are investigated directly in the program used for the interview and that the household is confronted with this and must deal with whether it is correct.

When data is received in Statistics Denmark, it goes through a validation which, for example, involves assessing the household's consumption in relation to its size. If, for example, there is only one person and a very high water consumption, or there are, for example, two adults with children, where it has not been reported how many months have been used for daycare and school, the household will be contacted to clarify the accuracy of the information. Some corrections are made without contacting the household, where the description of the purchase and the amount seem contradictory. It could be, for example, that a liter of milk is registered with an amount of DKK 1,000. This will be corrected to DKK 10.00.

The 14-day accounts are validated continuously when they are received, and collectively when the collection of accounts for a year has been completed. In the overall validation, it will be checked, for example, whether all purchases are coded correctly according to the classification.

Data compilation

When we have finished the validation of the interviews and accounts booklets the registry variable are linked in the data set. Sometimes it's difficult to find the household in the sample in the register data, this kind of difficulties can often be attributed to differences in the calculation date. When this happens we make manual imputation of for example, an individual's level of education.

After finishing the processing of Micro-data the enumeration process of making the data representative for the entire country begins. The figures in all tables are weighted this is done in order to partially resolve the gaps, as different dropout and pure random coincidences leads. Those types of Household where the risk for not participating in the survey is relatively large, which therefore results in too few households in the survey are assigned a relatively large weight, while household types, as there are too many of, is assigned a relatively small weight.

Information about both the enumerated number of households in Denmark after the weighting and on the actual number of households in the survey can be found in most tables. This last statement is relevant to assessing the sampling uncertainty, since a small number of households results in a relatively large uncertainties.

The weights are calculated using a regression estimate. The focus is on each characteristics of the relationship between sample and population. The advantage of this method is that many more features are considered than in the former method were post-stratification was used. Following characteristics are involved in the estimation:

• Household size and composition
• Income
• Main Income Recipient's socio-economic status
• The household owns or rents the dwelling
• What type of urban household lives in
• Education
• Gender
• Geography

Adjustment

We do not make other corrections of data besides those corrections described during data validation and data processing.