Skip to content

Accuracy and reliability

Contact info

Government Finances, Economic Statistics
Ulla Ryder Jørgensen

urj@dst.dk

Get as PDF

Income and consumption distribution in the household sector (experimental statistics)

The ability of the national accounts to accurately describe economic reality depends partly on the uncertainty associated with the sources and partly on the model assumptions underlying the preparation. Some parts can be calculated more precisely than others, as there is better access to source data. The first estimates of a period's national accounts will be more uncertain than the final version, which comes after three years, as they are continuously revised when new sources become available. The participation rate for the Consumption Survey in the years 2018-2022 has varied with . This creates uncertainty, not least for detailed consumption groups. For the total consumption, this means that there is an uncertainty margin of +/- 1.2 per cent. while for bread, for example, it is 2 per cent. For Food and non-alcoholic beverages, it is 1.2 per cent. , while for Alcoholic beverages and tobacco it is 4.8. There is under-reporting in a number of areas such as alcohol, tobacco, prostitution and undeclared work. The uncertainty is greater when data is based on accounting instead of interviews, and it will be greater if you look at smaller subgroups of households. In these statistics, we have chosen a more general level to ensure greater consistency between the national accounts and the consumption survey, as well as to minimize uncertainty.

Overall accuracy

Data for the national accounts comes from so many different sources that it is difficult in practice to calculate a figure for the uncertainty. The difference between the acquisition of receivables, calculated net from the financial side of the national accounts and the acquisition of receivables, calculated net from the non-financial part of the national accounts, can however be seen as an expression of the overall uncertainty in both the financial and the non-financial part of the national accounts. 2,301 households out of 17,730 surveyed households chose to participate in the Consumption Survey in 2022. For the total consumption for the entire population, there is an uncertainty margin of +/- 1.2 per cent, so the total consumption, which is estimated at DKK 359,661. with 95 percent certainty is expected to be between DKK 348,434. and DKK 370,888. For individual groupings in the population and product groups that are rarely traded, the uncertainty is greater. The consumption is thus more precisely calculated for e.g. households with higher incomes or where the main income recipient is of Danish origin, is in work, is over 60 or has education beyond primary school, as there are relatively more of these groupings who participated in the survey.

Sampling error

This is not relevant for the national accounts. The total sample for the Consumption Survey 2022 consisted of a total of 17,730 households drawn from 2021 (9551) and 2022 (8,179). In 2021, 1,123 households participated, while 1,178 households participated in 2022, a total of 2,301 for FU2022. The participation rate for FU2022 was thus 13 per cent. Sampling uncertainty is calculated for the individual consumer product groups based on the households that have chosen to participate in the survey. In FU2022, sampling uncertainty is based on the responses from the 2,301 participating households. The sampling uncertainty is quantified by the coefficient of variance (the relative standard error). The coefficient of variance for the total consumption per household in FY2022 was 1.2 per cent. This means that 95 per cent the confidence interval for the total consumption per household is DKK 359,661. +/- DKK 5728 There is a large variation in the sampling uncertainty between the total consumption and the specific consumer goods. Consumer goods that are bought frequently have a lower sampling uncertainty than consumer goods that are bought infrequently. Bread, for example, is a product group that most households often buy during the two weeks of accounting they have had in connection with participation in the survey. The coefficient of variance for bread for an average household is 2 per cent, while for, for example, preserved milk, it is 28 per cent. Correspondingly, the coefficient of variance is 3.9 per cent. in the capital region, while it is 7.4 per cent. in region Zealand for bread.

Non-sampling error

With approx. 2,200 complete annual households and more approx. 2.8 million households in total, where two thirds of the variables are imputed, a high degree of uncertainty must be expected. The result of an imputation is therefore not suitable for predicting the income and consumption level of the individual household, but can be used to provide a reasonable estimate of the total income and total consumption in large groups of households. The imputation itself is carried out using the imputation method "The Missing Forest Imputation", developed by Stekhoven and Buehlmann (2012). It is a non-parametric implementation of the more familiar Random Forest algorithm implemented in R. Since it is non-parametric, there is no need to manually train the model, unlike most other supervised machine learning methods. The MissForest algorithm is iterative and tries to optimize the precision of the predictions of the variables with missing values.

Quality management

Statistics Denmark follows the recommendations on organisation and management of quality given in the Code of Practice for European Statistics (CoP) and the implementation guidelines given in the Quality Assurance Framework of the European Statistical System (QAF). A Working Group on Quality and a central quality assurance function have been established to continuously carry through control of products and processes.

Quality assurance

Statistics Denmark follows the principles in the Code of Practice for European Statistics (CoP) and uses the Quality Assurance Framework of the European Statistical System (QAF) for the implementation of the principles. This involves continuous decentralized and central control of products and processes based on documentation following international standards. The central quality assurance function reports to the Working Group on Quality. Reports include suggestions for improvement that are assessed, decided and subsequently implemented.

Quality assessment

The ESA 2010 regulation requires Eurostat to assess the quality of data reported under the ESA transmission programme. This is done on the basis of the countries' quality reports, which are not published independently by Eurostat. The report is prepared annually. The consumption survey is a sample survey that is combined with a large number of register variables. The survey is based on a simple random sample. The survey's annual sample is not large enough to provide a safe estimate of consumption alone. Therefore, sample data from the previous year is included in the total calculation for the year. The survey thus takes on the character of a kind of "moving average".

Data revision - policy

Statistics Denmark revises published figures in accordance with the Revision Policy for Statistics Denmark. The common procedures and principles of the Revision Policy are for some statistics supplemented by a specific revision practice.

Data revision practice

Preliminary annual non-financial sector accounts are prepared, which are published twice a year: at the end of March and at the end of June. Final annual national accounts are published once a year at the end of June 3 years after the reference year. Starting with the publication of the consumption survey 2017, there is only one annual publication, and revisions will therefore not be made in the future. These statistics will follow revisions in the national accounts, but will only be published once a year after the publication of the consumption survey.