Croatian Bureau of Statistics (CBS)

Department of Eudcation, Culture and Information Society

Ilica 3, 10000 Zagreb, Croatia

The EU survey on the use of ICT in households and by individuals is an annual survey conducted since 2002. In Croatia, it has been conducted since 2007.

In 2022, the survey collects data on the access to information and communication technologies (ICT), on the use of the internet, e-government, e-commerce, internet of things, as well as green ICT.

National language: Primjena informacijskih i komunikacijskih tehnologija u kućanstvima i od pojedinaca

English: Usage of information and communication technologies in households and by individuals

Questionnaire(s) in national language(s) and the translation in English are available in the Annexes.

The following common concepts and definitions apply under the Integrated European Social Statistics (IESS):

• the International Standard Classification of Education (ISCED) 2011 published in the following breakdowns: low (ISCED levels 0-2: no formal education, primary education or lower secondary education), medium (ISCED levels 3-4: upper secondary or post-secondary non-tertiary education) and high (ISCED levels 5-6: tertiary programmes which normally need a successful completion of ISCED 3 or 4, or second-stage tertiary education leading to an advanced research qualification);
• the International Standard Classification for Occupation ISCO-08 at the 2-digit level;
• the Classification of Economic Activities (NACE Rev.2-2008), at section level;
• the Common classification of territorial units for statistics (NUTS 1) – finer granularity of NUTS 2 is provided on optional basis by some Member states;
• the SCL - Geographical code list;
• information about household income is provided at lower level of detail. 

Additional classifications used in the national questionnaire:

No additional classifications were used.

The ICT survey in households and by individuals covers those households having at least one member in the age group 16 to 74 years old. Internet access of households refers to the percentage of households that have an internet access, so that anyone in the household could use the internet.

No differences in scope.

The survey is collecting data of internet users, individuals who have used the internet in the three months prior to the survey. Regular internet users are individuals who used the internet, on average, at least once a week in the three months prior to the survey.

This annual survey is used to benchmark ICT-driven developments, both by following developments for core variables over time and by looking in greater depth at other aspects at a specific point in time. While the survey initially concentrated on access and connectivity issues, its scope has subsequently been extended to cover a variety of subjects (for example, the use of e-government and e-commerce) and socio-economic analysis (such as regional diversity, gender specificity, differences in age, education and the employment situation). The scope of the survey with respect to different technologies is also adapted so as to cover new product groups and means of delivering communication technologies to end-users.

For more details on the methodology applicable in each survey year, please consult the Methodological Manual for the respective year on CIRCABC - Methodological Manual - Information society statistics (europa.eu).

Deviations from standard ICT concepts:

No deviations from standard concepts.

Households and Individuals

In the ICT usage survey, the target population for the different statistical units is:

- individuals: all individuals aged 16 to 74;

- households: all (private) households with at least one member aged 16 to 74. 

Target population composed of households and/or individuals:

• Number of households: 1 287 641
• Number of individuals:  3 098 321

Non-compulsory age groups also included in the target population:

Whole territory of the country was covered by the survey.

Year 2022

Not applicable

Percentages of ‘Households’ and Percentages of ‘Individuals’

For questions where statement "in the last 3 months" applies, referenced period is 3 months prior to the day of interview. Accordingly, term "in the last 12 months" covers period of 12 months prior to the day of interview. Other questions are referenced to the day when interview took place.

Data collection started on 09/05/2022 and was completed on 04/06/2022. Subsequently, some non-response issues were resolved by 30/06/2022.

The legal basis for the 2022 EU survey on the use of ICT in households and by individuals is the Regulation (EU) 2019/1700 of the European Parliament and of the Council of 10 October 2019 establishing a common framework for European statistics relating to persons and households, based on data at individual level collected from samples (OJ  L 261 I, 14.10.2019, p. 1), as implemented by the Commission Implementing Regulation (EU) 2021/1223 of 27 July 2021 specifying the technical items of the data set, establishing the technical formats for transmission of information and specifying the detailed arrangements and content of the quality reports on the organisation of a sample survey in the use of information and communication technologies domain for reference year 2022 pursuant to Regulation (EU) 2019/1700 of the European Parliament and of the Council (OJ L 2269, 27.07.2021, pp. 1-45).

Complementary national legislation constituting the legal basis for the survey on the use of ICT in households and by individuals: 

No additional national legislation was required outside of regular National Law on Statistics.

ICT data sharing is handled upon request on case-by-case basis. There are no pre-defined procedures in this regard.

Formal policy mandates that all identification variables which could lead to person identification are disclosed and will not be released to anyone outside of statistical office. Territorial data is limited to NUTS3 regions.

Identification data on household is limited to NUTS3 region only, data on dwelling address, municipality and settlement are removed from microdata set and will not be disclosed. Accordingly, certain individual data is also removed from microdata set to prevent identification such as name/surname and telephone number. 

Dissemination calendar is publicly available on CBS website.

https://dzs.gov.hr/UserDocsImages/dokumenti/18-Kalendar%202021%20-%20azuriranje%2020.6.2022.pdf

Dissemination policy follows the schedule published in annual release calendar. All the planned publications are available to the public via website in the form of publications and online databases. Date and time of publishing is strictly followed and every publication is available to all users simultaneously. Users who have subscriptions to certain releases receive them as planned in release calendar. Data releases to the press also follow the schedule of release calendar. There are no deviations in dissemination policy in regard to ICT statistics.

Annual

No press/news releases were published.

Results of ICT survey is published on annual basis: https://podaci.dzs.hr/2022/hr/29624

Not available.

Not available.

Microdata is available to external entities upon request.

Not available.

Not available.

Not available.

Not available.

Quality management is currently work in progress. There are several ongoing projects through which quality management system will be developed. No documentation is currently publicly available.

Quality management is currently work in progress. There are several ongoing projects through which quality management system will be developed. Pilot programme is completed resulting in development of quality indicators database for different statistical domains within NSI. The goal is to link several different sources of data to create unified data source for quality indicators and metadata information.

In general, the quality of survey conduct is satisfactory as all the required quality criteria were satisfied. CBS will try to improve the response rate although 60% is considered quite high when compared to other social statistics surveys.

Users: Central agency for Development of Digital Society, Ministry of Regional Development and EU funds, Researchers

Data used by other government agencies is used to assist the government in creating policies in regard to development of national IT infrastructure and e-government services.

At present moment, there are no tools to measure user satisfaction.

All the required variables were included in microdata set.

All the variables had 100% completeness rate except for household income variable which had 85% completeness rate.

CBS did not find any significant systematic errors which would compromise survey results.

Sampling process ensured that sample carries a good representation of the general population and subgroups within that population. Over-coverage rate was kept to the minimum. Sampling error showed that data is of high quality as sampling error for whole population and different subgroups is meeting the quality requirements.

Response rate was quite high compared to other social surveys in our office. CBS did not find any patterns among non-respondents which would point to systematic errors.

Item non-response rate was not a problem as only problematic variable was household income with response rate of 85% due to sensitivity of the question. Respondents are sometimes reluctant to reveal household income.

Sampling error for key statistics was calculated with the usage of the SAS procedure SURVEYMEANS, which enabled that the complex sample design characteristics were taken into account. PROC statement in SURVEYMEANS procedure defined Taylor linearization series as the method for calculation of variance estimation. The following characteristics of the sample design were taken into account:

• Stratification: NUTS2 regions and Type of settlement (42 strata)
• PSU: The sampling units that were defined during the sampling stage. The sampling units were determined on the basis of enumeration area. Some of the smallest enumeration areas were merged in order to get the units with size appropriate for sampling purposes.
• Weight: Only the final weight was taken into account. By this approach the procedures like non-response adjustment and calibration were only indirectly included into the standard error estimation, but CBS considers such approach sufficient. 

The variance estimator  has to be adjusted to take unit non-response into account. The influence of unit non-response was taken into account by taking only the responding units as the sample size assuming MAR mechanism and by considering the final weights (in surveymeans procedure), where the re-weighting for non-response was also reflected.

See statistical computation section of SURVEYMEANS procedure in SAS 9.2 User's guide, available at: http://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_surveymeans_a0000000212.htm

Precision estimates for the question "Individuals having ordered goods or services for private use over the internet in the last 12 months" (individuals who ticked 'Within the last 3 months' or 'Between 3 months and a year ago' in question D1 of the 2022 model questionnaire):

Number of respondents (absolute value for ‘Yes’ answers):   1 648

Estimated proportion (in %):   56.15262758

Standard error (in percentage points):   1.2021424

Details of the breakdowns are available in the file "Standard errors and questionnaire structure" in the Annexes. 

See more details on non-sampling error below.

Data from Census 2011 was used as a sampling frame as it is mandatory source to be used in all social statistics surveys. Although the data is 10 years old the number of ineligible elements is small and is typically less than 100 households from 5 500 selected for survey.

0.6%

Not requested in the ICT survey.

1)       Measurement errors:  

None were recorded.

2)       Questionnaire design and testing:  

Both questionnaires (CAWI and CATI) are rigorously tested to ensure that both datasets are recorded in the same way with identical variable coding which insures that no data post-processing is required when both datasets are merged in a final dataset.

3)       Interviewer training:  

Interviewer training is performed as a 3-day course. First day consists of methodological training where interviewers are getting familiarized with each question, what kind of data is expected to be collected and to make sure that respondents understand what is being asked. Second day consists of practical training where interviewers interview each other in order to make sure that they fully understand each question and to clear any misunderstandings. During the first 2 days, course is conducted in questionnaire testing phase in order to discover any potential errors in questionnaire design. Third day training is performed on a day before actual interviewing starts in a practical manner just to make sure that everything is in order and interviewing can take place.

4)       Proxy interview rates:  

Proxy interviewing is allowed but the occurrence is so rare that we do not keep track, share of proxy interviews is below 0.5%.

Information about non-respondents:  

Unit non-response rate was measured at 38.2% which is considered as very good compared to other social statistics surveys. Most common type of unit non-response was refusal, 28.9% of respondents declined to participate in the survey. It was followed by non-contact type, 9.3% of respondents were not available for interview as the interviewers were unable to contact the respondents although the number of invalid telephone numbers was very low (less than 2%).

Coverage error was rather small. Share of ineligible households was low, 0.6% of households fall into this category, mostly due to household members being to old to participate in survey. Data from Census 2011 was used as a sampling frame. Although the Census data is 10 years old the number of ineligible elements is typically less than 100 households from 5500 selected for survey.

As for item non-response, the only variable with lower response rate is household income with response rate of 85%.

The unit response rate is the ratio of the number of in-scope respondents (= the number of achieved interviews or the net sample size to the number of eligible elements selected from the sampling frame).

Unit non-response rate for

• Households:  38.2% 
  
• Individuals (aged 16-74): 38.2%

Comments, if any: No comments.

1)       Methods used for dealing with unit non-response: 

No imputations were used for non-responding units.

2)       Methods used for minimizing unit non-response:  

Advance notification in the form of a letter was used as a method for minimizing unit non-response. Letter provided respondents with basic information on the survey, its goals and how the data they are providing will be used. CATI interviewers called at least 4 times in an attempt to contact respondent.

3)       Substitution permitted:  Yes.

4)       Substitution rate (in %):

Less than 0.5 %. In very rare cases where selected household is rented to another family interviewers were instructed to allow proxy interviewing and interview the family who is currently occupying the selected household. 

Items with low response rates (observed rates in %):  

Only variable with lower response rate is household income with response rate of 85%.

No significant processing errors were discovered in the survey. In very rare instances there may be some invalid variable coding errors due to software glitch. No imputations were performed.

Not requested for ICT Survey

06/12/2022

0 days

0 days

No problems were discovered.

Not relevant

Possible limitations in the use of data for comparisons over time:  Not relevant

The length of comparable time series depends on the module and variable considered within each of the modules of the survey.

Not applicable

Not applicable

Not applicable

All statistics are coherent within the dataset.

MANDATORY questions in the Eurostat model questionnaire 2022:

The table in the file "Standard errors and questionnaire structure" in the Annexes lists the questions that do not reflect the coverage of subjects and characteristics of Annex 2 of the Commission Delegated Regulation (EU) 2021/1898 of the 20 July 2021. 

Adoption of OPTIONAL questions and items in the Eurostat model questionnaire 2022:

The table in the file "Standard errors and questionnaire structure" in the Annexes lists the optional questions from the annual Eurostat model questionnaire 2022 included in the national questionnaire and their coverage for age groups beyond the standard scope. 

Additional questions introduced in the national questionnaire: 

Effects of deviations from the routing used in the Eurostat model questionnaire:  No deviations from Eurostat’s routing otherwise.

Although the CBS employees make maximum efforts to reduce errors prior to the data release, some might occur occasionally. It is important to preserve the confidence in the official statistics, as well as in the CBS as the institution whose highest goal is dissemination of official statistics. For that reason, it is crucial that the CBS acknowledges and documents the error, and that a candid and professional explanation is given.

Given that there is a high risk of users retrieving and using the data before errors have been detected and corrected, the errors must be published in the shortest period possible, and users must be enabled to see clearly what has been corrected.

Having all of this in mind, corrections of material errors in the CBS statistics shall always be indicated with a note pointing out the amendments and that the incorrect data released must not in any case be replaced with the revised version without providing prior access to both versions. This note must never be removed. If errors occur in larger publications (this includes all except First Releases), the CBS is obliged to issue a list of all required corrections (corrigenda) which clearly indicates which publication has an error and where exactly the amends have been made: on which pages of the original publication and in which tables. The corrigenda are published in an electronic format together with the publication. If they are available on the CBS website, they will be accompanied by a notice stressing it being a case of a revised version, and the release date of the corrections. Printed corrigenda are inserted into the printed publications available for sale, and to all copies of publications available for use in the CBS library.

Procedures for handling errors of the released data are adapted to the gravity of error. Excluding proofing/printing errors, the basic procedure following detection of an error is as follows:

1. a correction of a detected error, and reissuing the data/publications or corrigenda only in case of a larger publication on the CBS website in shortest period of time possible
2. a notice stating the case of an amendment next to the link with the revised data or a notice stating availability of the corrigenda in cases of larger publications together with the release data of amended data
3. an e-mail to the subscribers of the data/publications in which an error has been detected containing electronic version or the revised data or corrigenda in cases of larger publications, together with an apology letter and explanation as to why the error/errors occurred
4. sending printed version of the revised data to the subscribers, regardless of whether it was the case of data processing, purchase of publications, or just the corrigenda, if it is a larger publication together with an apology letter and explanation of the error 

In the case of a serious error that requires a period longer than one working day to correct, the incorrect data/publication shall be removed from the CBS website, and a message stating when the corrected version or corrigenda will be posted on the homepage. With the release of the amended version of the data/publication or corrigenda, the first released version is released simultaneously, in which the amendment is displayed clearly. In other words, released data containing an error shall never be replaced by the correct version, without being clearly displayed.

If an error is detected and corrected on the day of the release, the correction notice must state precise release time of the correction. Corrections of proofing/printing errors shall be made without issuing a correction notice. The errors that are discovered in released publications must be conveyed to the relevant statistics department without any delay, to correct the error in the shortest period possible. Apart from the statistics department in charge of the statistics, the Information, Services and Publications Directorate must be involved as well in all stages of preparation and issuing of the amendment or corrigenda. The employees of the specialist department in charge of the statistical area in which the error was detected, must, first of all, focus on correcting the detected error, while the ISPD employees are in charge of giving a notice to internal and external users and informing the Director General of the CBS.

Not relevant

The source of the raw data is described with more details in the paragraphs below.

Census 2011 database is used as a sampling frame. Telephone directories from telephone service providers are matched with census data to obtain telephone numbers for each sampled household.

The sample is not drawn from another survey.

Survey is a stand-alone survey.

Participation in survey is mandatory, but individuals are free to refuse participating.

There are no known shortcomings.

The sampling design is a probability design with probabilities proportional-to-size and with stratification effect. 

It is a 2-stage design. In the first stage Primary sample units are adjusted enumeration areas. The smallest enumeration areas were merged to the neighbouring areas in order to become more appropriate for sampling purposes. Primary sample unit is the same for the whole population. In the second stage, household is the sampling unit. Fixed numbers of households are selected from each of the Primary sample units.

Annual

1) Methods used to gather data: 

CATI and CAWI methods are used.

2) Short description of the survey method: 

In the first phase, web interview (CAWI) is used to collect data on households. Households which did not complete web questionnaire are transferred for CATI interviewing. 81% of collected interviews were completed via CATI while 19% were collected via CAWI method.

3) Variables completed from an external source:

No variables were collected from external source.

Validation check was completed during the data transmission.

No imputations were used in the data compilation process. 

Grossing-up procedures and adjustment for non-response for households and individuals were integrated in a single procedure. Design weights for households were calculated as the inverse value of inclusion probability. Non-response adjustment factor is calculated for each stratum (strata were defined by NUTS 3 regions and type of settlement). By calibrating the weights to the (census) population counts for each stratum, base weights for households are obtained. Base weight for individuals was calculated for each unit as the product of household base weight and number of individuals in the household  aged between 16 and 74 years. Raking procedure was used in order to calibrate the population (individuals) counts for following categories: gender, four age groups, three education classes.

Margin population counts by gender and age groups were taken from the last available demographic projections. Counts by education classes were obtained by taking the shares by classes (in percentage) from the population census and multiplying them with the estimated population counts (projections). Raking procedure was performed for each stratum. Weights coming out from the raking procedures were considered as the final individual’s weights. Final household weights were calculated by dividing individual’s weight with number of individuals, aged between 16 and 74 years, in the household.

For the target indicator "Individuals having ordered goods or services for private use over the internet in the last 12 months" (individuals who ticked 'Within the last 3 months' or 'Between 3 months and a year ago' in question D1 of the 2022 model questionnaire):

Imputation rate (% of observations): 0%

Imputation rate (share of estimate): 0%

Methods used to impute item non-response:

In rare cases deductive imputations were used to collect missing data.

Grossing up procedures have been applied to: Individuals and Households

Description of the weighting procedures:

Grossing-up procedures for households and individuals were “integrated” in one procedure; hence both procedures are described under this item:

• Design weights for households were calculated as the inverse value of inclusion probability.
• Non-response adjustment factor is calculated for each stratum (strata were defined by NUTS 3 regions and type of settlement).
• By calibrating the weights to the (census) population counts for each stratum, base weights for households are obtained.
• Base weight for individuals was for each unit calculated as the product of household base weight and number of individuals, aged between 16 and 74 years, in the household.

Raking procedure was used in order to calibrate the population (individuals) counts for following categories:

• Gender
• Four age groups
• Three education classes

Margin population counts by gender and age groups were taken from the last available demographic projections. Counts by education classes were obtained by taking the shares by classes (in percentage) from the population census and multiplying them with the estimated population counts (projections). Raking procedure was performed for each stratum.

• Weights coming out from the raking procedures were considered as the final individual’s weights.

Final household weights were calculated by dividing individual’s weight with number of individuals, aged between 16 and 74 years, in the household. 

Not relevant

Not relevant

CBS has no additional comments.