Federal Statistical Office of Germany

H22

Statistisches Bundesamt, 65180 Wiesbaden

The Continuing Vocational Training Survey (CVTS) collects information on enterprises’ investment in the continuing vocational training of their staff. Continuing vocational training (CVT) refers to education or training measures or activities which are financed in total or at least partly by the enterprise (directly or indirectly). Part financing could include the use of work-time for the training activity as well as financing of training equipment.

Information available from the CVTS is grouped around the following topics:

- Provision of CVT courses and other forms of CVT (training/non-training enterprises)

- CVT strategies

- Participants in CVT courses

- Costs of CVT courses

- Time spent in CVT courses

- Characteristics of CVT courses

- Assessment of CVT activities

The CVTS also collects some information on initial vocational training (IVT).

For further information see the CVTS 6 legislation (http://ec.europa.eu/eurostat/web/education-and-training/legislation) and the CVTS 6 implementation manual (http://ec.europa.eu/eurostat/web/education-and-training/methodology).

The main groupings for enterprises are by economic activity (NACE), size group and training/non-training enterprises.

CVTS 6 covers all economic activities defined in sections B to N and R to S of NACE Rev. 2.

Definitions as well as the list of variables covered are available in the CVTS 6 implementation manual (http://ec.europa.eu/eurostat/web/education-and-training/methodology).

Enterprise.

Enterprise definition is compliant with Council Regulation (EEC) No 696/93.

288 361 enterprises.

Variable A2tot (persons employed) refers to 31 December 2020.

Germany

2020; 2015; 2010; 2005; 1999; 1993

Not applicable.

Number, EUR.

The reference year for CVTS 6 is the calendar year 2020.

At European level

Basic legal act: Regulation (EC) No 1552/2005 of the European Parliament and the Council

Implementing act: Commission Regulation (EU) No 1153/2014, amending Commission Regulation (EC) No 198/2006

At national level:

No specific national legislation

Not applicable.

The federal statistical law (BStatG) rules how to handle confidentiality. In principle it is ruled in § 16 Abs. 1 related to § 18 Abs. 1 BStatG. Exceptional transmission of individual data, for example to universities, is ruled in "§ 16 Absatz 6 BStatG".

We do only publish data in percentages or averages, not in absolute numbers. We consider standard deviations. Any information coming from less than 5 enterprises in the sample is not published at all.

See https://www.destatis.de/DE/Presse/Pressemitteilungen/2022/08/PD22_349_215.html, where further publishing procedure is announced.

See https://www.destatis.de/DE/Presse/Pressemitteilungen/2022/08/PD22_349_215.html, where further publishing procedure is announced.

See https://www.destatis.de/DE/Presse/Pressemitteilungen/2022/08/PD22_349_215.html, where further publishing procedure is announced.

Every 5 years.

17.08.2022: press release (first results).

See "Description of methods CVTS 6" in annexes.

https://www.destatis.de/DE/Presse/Pressemitteilungen/2022/08/PD22_349_215.html

https://www.destatis.de/EN/Press/2022/08/PE22_349_215.html 

Not applicable.

No microdata disseminated.

None.

Not applicable.

See "Description of methods CVTS 6" in annexes.

Not applicable.

Not applicable.

Specification and procedures of implementing and handling plausibility at data entry and recording and the data imputation process are coded and running in a comprehensible way which requires technical understanding.

There is a good quality of qualitative information about structure and organisation of CVT. But especially the questions on total costs of CVT courses and the cost components were very problematical for some enterprises. Since the costs of measures for CVT are often not collected in separate accounts in the accounting system, many enterprises had difficulties to answer the questions under C7. A lot of enterprises could declare only roughly estimated values for total costs of CVT courses and cost components.

No specific analysis of statistical user needs available for CVTS 6 data.

The needs of the main users have been met.

No deviation.

Data are complete.

High participation increases accuracy: In Germany the survey was voluntary. An enterprise could decide if it provided answers to the questionnaire or not. Due to the many questions on different topics (structural questions, questions on CVT and IVT) most enterprises decided not to answer.

As the survey was not compulsory and unit non response occurred, the estimation was conducted in a two-step approach: After treatment of unit non-response the actual estimation was done via regression estimation with the number of employees in the business register in the sampling frame as auxiliary variable. Concerning the non-response we only know the strata and the number of employees in the business register at the date of the preparation of the sampling frame. These variables can be used to reduce the non-response bias. We suppose that unit non-response within a stratum occurs at random and with the same probability. This means that the denominator not only includes responding enterprises but also the non-eligible ones (over-coverage) as these have to be valued as valid answers (with a value of zero for the target variables) from the point of view of sampling methods. This response probability together with the inclusion probability serves as basis for the actual estimation in the second step.

By using an auxiliary variable the precision of estimates often can be raised considerably compared with a Horvitz-Thompson estimation (where the estimation is done only by the reciprocal values of the inclusion probability (if necessary, corrected for non-response)) if the target variable (for example the number of participants in courses) and the auxiliary variable are highly correlated. The number of employees in the business register at sampling time serves as an auxiliary variable. The resulting weight is added to the micro data allowing – as usual – flexible data processing for the different target variables. The SAS macro package CLAN of Statistics Sweden was used for the calculation. The SAS macro package CLAN of Statistics Sweden was used for variance estimation, too. The variance estimation was done for 8 variables and 5 ratios derived thereof, broken down by NACE categories and size classes.

These are in detail: Total number of enterprises that provided CVT, Total number of enterprises that provided CVT courses, Total numbers of persons employed, Total number of persons employed in enterprises that provided CVT, Total number of participants in CVT courses, Total costs of CVT courses, Ratio of total number of enterprises that provided CVT to the total number of enterprises, Ratio of the total number of enterprises that provided CVT courses to the total number of enterprises, Ratio of the total number of participants in CVT courses to the total number of persons employed, Ratio of the total number of participants in CVT curses to the total number of in enterprises that provided CVT, Total number of enterprises providing IVT, Total number of participants in IVT, Ratio of the total number of enterprises providing IVT to the total number of enterprises.

See table 13.2.1 "Sampling errors - indicators" in annex "DE - QR tables CVTS 2020 (excel)".

Additional information is not necessary, see annexes.

The German business register is a regularly updated database of in 2020 actually about 3 million enterprises and local units with a taxable turnover from deliveries and output and employees subject to social insurance contributions. Evaluations of business register data on the number of enterprises and local units and their employees, who are subject to social insurance contributions, and their sales (turnover) reveal economic structures in Germany.

The time lag of address data was only two months, but the time lag of data about the number of employees was 2 years.

See table 13.3.1 "Coverage error" in annex "DE - QR tables CVTS 2020 (excel)".

See table 13.3.1.1 "Over-coverage - rate" in annex "DE - QR tables CVTS 2020 (excel)". 

Not applicable.

The quality and functionality of the design of the questionnaire and web-formula is professionally proofed referring to applicable standards.

In advance of the start of the data collection the federation of enterprises received an information letter about the survey. They were asked to inform and promote for participation. The first postal reminder including a new questionnaire was sent to all enterprises which did not answer in April 2021. A second postal reminder without a new questionnaire was sent to all enterprises which did not answer in the end of May 2021. Information in advance and the two reminders increased the general response but failed the aim of 25% (only reaches about 18%). One telephone campaign followed contacting enterprises which did not answer to the first and second reminders. The telephone campaign particularly concentrated on a certain number of special units for raising the unit-response moderate successfully. As not all missing enterprises could be contacted we chose enterprises in stratums with a low response rate and huge enterprises with many persons employed, and moreover, those enterprises which had promised to send the questionnaire and had requested a later date for sending the questionnaire back.

See table 13.3.3.1 "Unit non-response - rate" in annex "DE - QR tables CVTS 2020 (excel)".

See table 13.3.3.2 "Item non-response - rate" in annex "DE - QR tables CVTS 2020 (excel)".

After receiving the completed questionnaires the Länder Statistical Offices did an intensive manual checking of their contents. Therefore we had developed a detailed handbook for the staff in the Länder Statistical Offices which described how to find out missing data and inconsistencies, how to correct them (most often in direct contact with the enterprises concerned) and how to code the answers. We developed also a national solution for data editing and controls. As for CVTS 2 to CVTS 5 we chose a BLAISE application which was based mostly on the checks included in the CVTS 6 checking rules with quite some additions which were deemed to be necessary. The BLAISE application was programmed in the Statistical Office "Nordrhein-Westfalen". We held a seminar for our colleagues from the Länder Statistical Offices with hands-on training.

In the web-based survey checks concerning the key variables were also implemented.

We also developed some small codes for data imputation and for the calculation of participants for enterprises which had only entered participant cases in the questionnaire. The codes mentioned in the last sentence only ran in our office after reception of the Länder data.

After we had received the Länder data we did some more intensive checking, among other things by doing test tables, concerning e. g. Labour costs per person employed, Labour costs per working hour, Working hours per person employed, Training course hours per person employed, Training course hours per participant, CVT costs per person employed, CVT costs per participant, CVT costs per training course hour, IVT participants compared with persons employed on 31 December 2020. As a consequence in quite a lot of cases we had to go back to the Länder which in turn had to ask some of the concerned enterprises again. This process of cleaning the material was very time-consuming.

Not applicable.

T+18 months.

17.08.2022 - no time lag.

See 8.1.

Countries should transmit data to Eurostat no later than 18 months after the end of the reference year.

No time lag.

See table 14.2 "Project phases - dates" in annex "DE - QR tables CVTS 2020 (excel)".

Not applicable.

No differences, see table 15.1 "Comparability - geographical" in annex "DE - QR tables CVTS 2020 (excel)".

No additional variables related to COVID-19 were collected.

Not applicable.

See table 15.2 "Comparability - over time" in annex "DE - QR tables CVTS 2020 (excel)".

Not applicable.

See table 15.3 "Coherence - cross-domain" in annex "DE - QR tables CVTS 2020 (excel)".

Not applicable.

Not applicable.

CVTS results for a given reference year are based on the same microdata and results are calculated using the same estimation methods, therefore the data are internally coherent.

Not available.

Not applicable.

Not applicable.

Not applicable.

See table 18.1 "Source data and data collection" in annex "DE - QR tables CVTS 2020 (excel)".

Every 5 years.

See table 18.1 "Source data and data collection" in annex "DE - QR tables CVTS 2020 (excel)" and "Description of methods CVTS 6" in annexes. 

After receiving the questionnaires, filled out by the enterprises, the Länder Statistical Offices intensively checked the contents manually. This could be done by a detailed handbook we developed for the staff in the Länder Statistical Offices. It described how to find out missing data and inconsistencies, how to correct them (most often in direct contact with the enterprises concerned) and how to code the answers. The values, for example, are proofed to be located in a stratum-specific range derived from CVTS 4 data and for example current labour cost statistics. We developed also a national solution for data editing and controls. As for CVTS 2 to CVTS 5, we chose a BLAISE application which was based mostly on the checks included in the CVTS 6 checking rules with some additions which were deemed to be necessary. The BLAISE application was programmed in the Statistical Office Nordrhein-Westfalen. We informed the Länder Statistical Offices in a workshop with hands-on-training. In the web-based survey checks concerning the key variables were also implemented. We integrated measures of corrections, calculation of participants for enterprises which had only entered participant cases in the questionnaire in SAS-Codes. After we had received the Länder data we did some more intensive checking, among other things by doing test tables, concerning e. g. Labour costs per person employed, Labour costs per working hour, Working hours per person employed, Training course hours per person employed, Training course hours per participant, CVT costs per person employed, CVT costs per participant, CVT costs per training course hour, IVT participants compared with person employed on 31 December 2020. As a consequence in quite a lot of cases we had to go back to the Länder which in turn had to ask some of the concerned enterprises again. This process of clearing the material again consumes a lot of time.

In the web-based survey checks concerning the key variables were also implemented.

See table 18.5.1 "Imputation - rate" in annex "DE - QR tables CVTS 2020 (excel)".

We integrated measures of corrections, calculation of participants for enterprises which had only entered participant cases in the questionnaire in SAS-codes.

Not applicable.

None.