People living in households with very low work intensity by group of country of birth (population aged 18 to 64 years) ilc_lvhl16n

Expand all

The data include several units of measure, depending on the variable. For more information, see the methodological guidelines and description of EU-SILC target variables available in CIRCABC. Most indicators are reported as shares. Some are reported in other units (e.g. per cent, thousands of persons, monetary units, etc.). More information is available in Eurobase, living conditions section, to the database.

The reference period is the survey year. The nucleus or annual variables are collected yearly using the reference period, as specified in Annex II to, and Article 7(1) of, Regulation (EU)  2019/2242 as well as in the methodological guidelines.

For all countries, the reference period for income variables in EU-SILC is the previous calendar year. Until 2019, Ireland collected income information for the 12-month period immediately preceding the sample household’s interview date. For most countries, the fieldwork was carried out from January to July 2024. The lag between the income reference period and data collection period varies across countries. However, for most countries, the data collection took place at the beginning of 2024 (Figure 2).

 

Figure 2: Lag between the income reference period and data collection, 2024

Field work duration, Lag between income and other variables

Source: EU-SILC Micro database 2024 (November 2025)

Annual

EU-SILC was designed to limit the burden on respondents, to avoid a high non-response rate and to ensure the quality of the information collected. The method of interview has a significant impact on interview duration. In 2024, the average interview lasted more than 60 minutes for Bulgaria, Luxembourg, Romania, Poland and Germany, while for some countries the interview duration per household was less than 30 minutes (Netherlands, Italy, Latvia and Denmark).

Annex 2 – Mode of data collection and fieldwork provides more information about the length of interview by country.

The metadata are issued in the SIMS format for EU-SILC.

EU comparative quality reports and national quality reports can be found in CIRCABC as well as on the Eurostat website.

19.1 Specific descriptions from 2024 national quality reports

Reference population and household membership

Belgium: Tertiary students’ usual residence is their private address in their university town, although they return to their parental home during the weekend. They remain officially registered at their parents’ address. In BE-SILC, they belong to their parents’ household.

Estonia: Persons living in collective households are included in the reference population. The share of persons living in collective households and who are not members of another private household is likely to be very low. Moreover, there is no feasible way to estimate the share of these persons in the total population. Thus, excluding them is unlikely to affect the comparability and reliability of the estimates.

Methods of calculating the variance, standard error and confidence interval for the main indicators (AROPE, AROP, SMDS, LWI)

Some countries use their national method to estimate precision requirements for the main indicators.

Belgium: standard errors are estimated using the jackknife repeated replication (JRR) method. The clusters are the groups, with the strata made by two (or three) groups, using the sampling order.

The design effect for the median equivalised disposable income is 1.16.

Bulgaria: In the 2024 survey year, Bulgaria used the bootstrap variance estimation technique to obtain the standard error for the main indicator of interest, AROPE. In total, 100 replicate samples were drawn. The replicate weights were obtained using a calibration approach, which was identical to the calculation of RB050 weights. This additional procedure provides an unbiased estimator for the population totals and increases precision. The standard error was calculated using information from the 100 replicate samples, where the PSU sample size is a reflation of the sample size of the main survey for the corresponding year.

Germany: variance estimation for cross-sectional indicators is performed in Germany using Statistics Sweden’s SAS estimation software ETOS 2 - Estimation of Totals and Order Statistics.

France: for the variance at national level, estimates with a linearisation of the AROPE and persistence in poverty were tested with no significant gain. In a conservative approach, more traditional precision estimators (Horvitz-Thompson-Narain estimators or calibrated estimators) were used.

For variance at the level of NUTS 2 regions, the small area method used is equivalent to regressing interest variables on the auxiliary variables used to construct the small area weights. This equivalence between the synthetic estimator and a regression is exploited when calculating the variance by calculating the variance of the estimate using the linearisation approach on regression coefficients (this allows a variance to be calculated on the small area weights, which can be negative).

Croatia: the Sampling, Statistical Methods and Analyses Department calculates the indicators and the variance in the R programming language.

Italy: sampling errors of the main indicators are calculated using national methods in order to take into account the sampling design effects.

Latvia: the Central Statistical Bureau uses its own methodology for calculating sampling errors using R (package vardpoor).

Lithuania: variance for the main indicators (AROPE, etc.) is calculated using a different method from Eurostat (for reference see Annex A to Concept 13.2.1). Standard errors and calibrated weights are calculated using the SAS macro CLAN (CLAN97.sas) developed by Statistics Sweden. CLAN computes an estimate of a parameter q and an estimate – based on Taylor linearisation – of the standard error . The generalised regression (GREG) estimator is used with auxiliary information.

Hungary: the standard error calculation is based on the linearisation method, which takes into account the characteristics of the sampling plan and the effects of calibration.

Malta: the standard error of the main indicators is calculated in R using the vardomh() function, which estimates the variance of sample surveys within domains using the ultimate cluster method. This approach applies a jackknife-like technique where clusters (usually primary sampling units) are systematically omitted to create replications, approximating the sampling variance.

The persistent-risk-of-poverty ratio is calculated in SPSS. The dataset is filtered to include only the survey year, and weighted using the four-year longitudinal weight. The standard error is computed with the final value being adjusted by the Kish Effect. The value is then used to obtain the upper and lower confidence intervals.

Netherlands: the national method is used to calculate variance for the main indicators (AROPE, etc.). These variances have been estimated taking into account the sampling design, various stages of non-response attrition and the weighting to known population control totals. The weighting not only reduces bias due to selective non-response, but also reduces the variances of most SILC indicator estimates.

Austria: the calibrated bootstrap procedure is used with the R Packagesurveysd’ developed by the Methods Department of Statistics Austria.

Poland: sampling errors of indicators for the quality report were estimated using the ultimate cluster method and linearisation. Calibration of the weights was also taken into account. The R package vardpoor was used in the calculations.

Portugal: the standard error estimates are obtained using the jackknife resampling method, which makes it possible to calculate variances for totals (linear estimators) and for the quotient of totals and differences of quotients (non-linear estimators). This method is recommended when a complex sampling design and calibrated estimators are involved, as is the case with implementing the Portuguese EU-SILC survey, called ICOR.

Romania: at national level, variance estimation is performed using the Taylor linearisation method implemented in the Regenesses package in R, considering the characteristics of the sample design.

Finland: Sampling errors for the main indicators of the Finnish data are calculated using an estimation technique based on the with rescaled bootstrap by taking both the sampling design and weighting into account, with 1,000 replications in total by panels. The calculations are performed using SAS programs.

Sweden: note that all the standard errors published by Statistics Sweden, as well as the standard errors of the main indicators used for the quality assessment and presented in Annex A of the EU-SILC quality report, were calculated in line with the national framework, with stratification, clustering as well as household size taken into account. Calculations were made using SAS macro ETOS - Estimation of Totals and Order Statistics, which was designed to compute point and standard error estimates of totals and order statistics (parameters) from sample surveys, as well as rational functions of those parameters. Based on the ETOS 2.0 user guide (2012), the estimating equations (EE) technique was used to estimate the variance of the order statistics, and the Taylor linearisation method was used for the variance estimation of non-linear functions like ratios and products.

Serbia: Standard errors were calculated by using the bootstrap replication method, which represents a type of repeated resampling method. The general characteristic of repeated resampling method and therefor also of the bootstrap is to draw (sub-) samples from the original sample and to calculate the population parameter of interest from each sample. The variance estimation is then based on the distribution of the several estimates.

Break in time series by year

Indicators: for Portugal, the methodology for calculating household type differs from that used by INE (available on the website of the Portugal Institute National of Statistics).

Coverage: France: since 2022, the survey has also covered four overseas departments: Guadeloupe, Martinique, French Guiana and Réunion.

Methodological changes

Denmark: in 2024, the questions relating to general health variables were changed to better align with Eurostat recommendations (PH010 and PH030).

Croatia: in 2024, several methodological changes were implemented. Imputation was significantly reduced, and the questionnaire was shortened. Moreover, the collection of some income sources from administrative data affected both income variables and the length of the questionnaire.

Italy: To ensure comparability with other countries and to comply with the methodological guidance provided, the question on severe limitations was revised. As a result, a break in the series was reported for variable PH030 in 2024.

Luxembourg: design adjustments affected by the COVID-19 crisis, as well as the introduction of a mixed mode of data collection, resulted in a break in the series for 2020 and the absence of four-year longitudinal data for 2020-2023.

Portugal: administrative data relating to the personal income tax on old-age pensions in the contributory system were used for the first time in 2024, in order to improve the consistency and quality of information before deducting taxes and social contributions. In addition, the variable PH030 (Limitation in activities because of health problems) was revised according to EU-SILC guidelines.

Sweden: in 2022, the data collection method in the Swedish SILC was changed from CATI to a mixed mode combining CAWI and CATI. This may have led to breaks in the time series for certain 2024 module variables (RC380, RC390, RCH010, HD120, HD140, HD160 and HD190).

 

Weighting

Belgium: BE SILC recalculated the weights for the years 2019–2024. This revision was carried out to improve the estimation of the representative population, after an overestimation of the 0–15 age group was identified.

Hungary: the annual datasets for the 2019-2024 collections were revised in the second quarter of 2025. One of the objectives of the revision was to reweight the datasets using benchmark figures derived from the 2022 census. Methodological procedures were further developed, in particular with regard to fine-tuning the grossing, imputation and correction methods for income data.

For detailed information about significant changes and breaks in time series, please see the overview of breaks in series.