The German Panel of Teacher Education Students: Surveying (Prospective) Teachers from Higher Education into Working Life

2.1 Study design

As mentioned above, the main general objective of the NEPS is to collect longitudinal data on the development of competencies, educational processes, educational decisions, and returns to education in formal, nonformal, and informal contexts throughout the life span. In order to make relevant information on educational transitions available as soon as possible, the educational biography has been divided into eight stages with a particular focus on critical transitions. Within these eight stages, the NEPS started with six different cohorts in order to observe educational careers and transitions. This multicohort-sequence design (Blossfeld et al., 2009; von Maurice et al., 2016) makes it possible to collect comparative data on competence development and educational pathways over the entire life course, while at the same providing the data in a relatively short timeframe. Theoretical dimensions, so-called ‘pillars’, provide a conceptual framework and integrate data collection in these six starting cohorts.

One of the starting cohorts mentioned refers to the stage of higher education. For this cohort, first-year students, enrolled at a German higher education institution (HEI) in the winter term of 2010/2011, were selected (Aßmann et al., 2019). HEIs include universities and equivalent institutions such as colleges of art or music, as well as universities of applied sciences. Students attending private HEIs and students enrolled in a teaching degree programme were oversampled. As mentioned above, the latter form the basis of the LAP project. For details on the sampling design see Chapter 2.4.

In Starting Cohort First-Year Students several modes of data collection were employed: self-administered paper-and-pencil questionnaires (PAPI), computer-assisted telephone and personal interviewing (CATI and CAPI), online surveys (computer-assisted web interviewing; CAWI), and competence testing.

Following the initial PAPI in the winter term of 2010/2011, annual CATIs (and one CAPI in wave 12) focus primarily on (an update of) the life course, including the educational and employment history of the participants. Central longitudinal measures of the NEPS pillars can be found in CATI/CAPI as well as in CAWI, which are conducted annually from 2011 until 2014 and biennially thereafter. In addition, questions particularly aimed at higher education students are integrated into the CAWIs. All survey instruments were basically also addressed to (prospective) teachers. They already open up a wide range of possibilities for research on teacher training and the teaching profession. When the LAP project started, an additional survey programme specifically for (prospective) teachers has been implemented in the NEPS SC5 surveys since wave 8 in autumn 2014, with the LAP study fully integrated into the survey design.

The data collected alongside the life course information within a wave can be interpreted as cross-sectional data. However, since the central constructs are usually collected in more than one wave, this data often contains longitudinal information that makes it possible to observe intra-individual developments.

A variety of modes were used for the competence test (Brachem et al., 2019). The first competence test in wave 1 was administered as a paper-based assessment (PBA). For the second test of competencies in wave 5, a mode experiment was conducted with individual web-based testing (CBWA), as well as three test modes in group settings: conventional paper-based assessment (PBA), paper-based assessment with digital pencils (E-Pen), and computer-based assessment (CBA). In wave 7, a subject-specific competence test in business administration was administered as an individual paper-and-pencil test to students or graduates of business administration and economics. The last competence test of NEPS SC5 was conducted in wave 12, again as a mode experiment in two modes: CBA during a CAPI survey and CBWA after a CATI survey (for more details, see the ‘Information on Competence Tests’ and ‘Field Reports’ at: https://www.neps-data.de/Data-Center/Data-and-Documentation/Start-Cohort-Students/Documentation).

2.2 Time of data collection

Between 2010 and 2022, 19 panel waves were conducted. The CATIs with their focus on the life course always started in spring, with the online surveys usually starting in autumn. Only the nineteenth wave, in 2022, deviates somewhat from this pattern; wave 19 is a combined CATI and CAWI survey, where all study participants who took part in the telephone interview were invited to the online survey immediately afterwards. Although unscheduled, the study participants of all NEPS starting cohorts were also asked online about the effects of the Corona pandemic in May/June 2020.² A detailed list of the studies conducted can be found in Table 1. The table also shows when the competence tests took place. The data published to date comprises data from the first 17 panel waves and the additional 2020 Corona survey.

Please note that for funding reasons the initial sample of teacher education students was randomly divided into a “LAP basic sample” and a “LAP oversample”. The relative size of the basic sample corresponds to the proportion of teacher education students in the population, and the size of the oversample is the surplus. The two groups were sometimes treated differently.

2.3 Location of data collection

All data was collected across all German federal states. In addition, study participants who belong to the initial sample were also surveyed abroad when they moved to another country.

2.4 Sampling, sample and data collection

The target population of NEPS Starting Cohort First-Year Students consists of all students who enrolled at a public or state-approved HEI in Germany for the first time in the winter term of 2010/2011 and were aiming to complete one of the following degrees: bachelor’s degree, a state examination (Staatsexamen) in medicine, law, pharmacy, or teaching, or a diploma or master’s degree in theology. However, first-year students studying at higher education institutions run by federal ministries or federal states for members of their own public services were excluded. During sampling, special emphasis was placed on students with non-traditional entrance qualifications, i.e., students without a school-leaving certificate qualifying for higher education. Furthermore, students at private HEIs and teacher education students were overrepresented in the sample, the latter building the initial sample of the LAP. More details on the population and sampling are described in Zinn et al. (2017) and Aßmann et al. (2019).

The sample for the first wave was drawn using a stratified cluster sampling approach. A particular field of study at a specific higher education institution represented one cluster (primary sampling unit), and all students within each cluster were surveyed. To oversample teacher education students (and students at private HEIs), a first stratification level distinguished between clusters in terms of type of HEI, institutional control of HEI, and teacher education programme. This first stratification level consisted of four strata (h₁ = clusters linked to teaching tracks; h₂ = all other fields of study at public universities; h₃ = all other fields of study offered at public universities of applied sciences; h₄ = all study tracks at private universities or private universities of applied sciences). To reduce sampling error, a second stratification level was introduced, which was defined using aggregated fields of studies. A detailed report of the stratified cluster sampling approach as well as a detailed description of the clusters and strata can be found in Zinn et al. (2017).

In order to achieve higher response rates, two separate approaches were used for recruiting study participants: In a first step, each student was invited by letter to participate. In a second step, field workers were sent into central first-year courses to personally recruit participants for the survey. This method had been previously tested in a pilot study and resulted in both better participation rates and higher panel attendance (Brachem et al., 2019).

In the first wave, a total of 17,909 first-year students participated in the NEPS student cohort including 5,554 teacher education students (according to variable tg02001; see Chapter 3.10) who form the initial LAP target group. The various school types of the German school system are also mirrored in different teacher education programmes. In the initial sample, 13% of the teacher education students had taken up a training course for primary education, 19.5% for lower secondary education (excluding lower secondary education at a Gymnasium), 54.5% for upper secondary education (including lower secondary education at a Gymnasium but excluding vocational education), 5% for special education, and 6.5% for vocational education. One and a half per cent of the teacher education students stated that their degree programme did not differentiate between school types. As is common in teacher education programmes, most students are female (75.5%).

However, educational and occupational trajectories are not linear, and dropping out or changing study programmes is common. Hence, the target population changes over time due to people dropping out of or moving into teacher education, or entering the teaching profession without having completed a teaching degree (e.g., lateral entrants). The LAP target population was initially limited to teacher education students, trainee teachers in preparatory service and working teachers. From wave 11 onwards, the definition of the target population was expanded to include respondents who (a) have completed preparatory service and intend to work as teachers, but are not yet employed, (b) have completed the first phase of teacher education and intend to complete preparatory service but have not yet started it, and (c) have temporarily interrupted employment as a teacher, e.g., due to parental leave.

Table 2 reports the number of participants from wave 8 to wave 17 who belong to the LAP target group. The table also shows the auxiliary variables generated from wave 8 onwards to filter the teacher-specific questions and determine the LAP sample. More information on identifying the LAP target population, also in earlier waves, and auxiliary variables provided for this purpose is given in Chapter 3.10. A full description of the student cohort’s recent panel development can be found in Ziesmer (2022).

The analysis shows that participation rates are consistently lower in CAWI compared to CATI. The proportion of students who could be reached for an interview or test declined significantly over time. The main reasons for panel attrition are missing contact information, continuous non-participation over a period of three years (so-called ‘final dropouts’), and withdrawal of panel consent (Aßmann et al., 2019). There is evidence that accessibility and participation of study participants depend on both the topic of the survey and personal availability as well as on socio-demographic factors (Liebeskind & Vietgen, 2017; Zinn et al., 2018). To account for selective study participation, weights have been estimated for every wave. More details on participation development, panel attrition, sample selectivity and weights can be found in Zinn et al. (2018) and, for the most recent Scientific Use File (SUF), Ziesmer (2022).

2.5 Materials and survey instruments

2.6 Quality Control

2.7 Data anonymisation and ethical issues

To participate in the panel study, students had to give prior informed written consent. Moreover, informed consent to take part in the study was also given by the educational institutions, so their students were allowed to be contacted. Data protection and security officers of the LIfBi and DZHW approved the consent procedure.

The anonymity of the participating students and the respective HEIs is guaranteed by following two principles of the NEPS anonymisation concept: First, disclosure of participants’ identities should be impossible and, second, a high utility of the data should be preserved (Schier et al., 2019). Therefore, information that would allow identification of individuals or HEIs (e.g., names, addresses) is not provided in the Scientific Use Files. Direct identifiers are separated from the data and are not delivered from the survey institute to the LIfBi. In addition, the anonymous identifiers used by the survey institute are replaced by new ones in the SUFs.

To provide secure and comfortable data access for the scientific community, a combination of five approaches has been implemented (see Schier et al., 2019).

Organisational data protection: The data is only available to the scientific community. Access to the data is managed by the LIfBi Research Data Center, which checks the status of the data user, their connection to a university or a research institute and the scientific interest of the data request.

Legal data protection: Data users are informed about data protection and data security. They are committed to data protection and must sign a contract with corresponding regulations.

Statistical data protection: Techniques of statistical data protection are used to ensure the respondents’ privacy by generating factually anonymous data. Various modification approaches exist and are applied in the NEPS, e.g., aggregating data or slightly changing variable values. Which approach is chosen primarily depends on the way in which data access is realised (see below).

Informational data protection: LIfBi staff offer a special training programme to data users that explains the complex data structure and gives information on data protection and data security. Furthermore, detailed documentation is provided that includes information about data protection and anonymisation measures.

Technical data protection: As the data collected in LAP/NEPS is digital, technical data protection in the form of hardware and software solutions is essential. One area of technical data protection concerns data dissemination. Depending on the sensitivity of the data, LIfBi offers three different modes of access to the Scientific Use Files (see Chapter 3.7). Regardless of the access mode, all data is anonymised using random person and institution identifiers so that neither individual study participants nor HEIs can be identified. Information that is more sensitive is only available under restricted conditions following data protection rules. For example, information on federal states and places of residence of participants is not available in the downloadable dataset but in a data-secured environment that can only be accessed with an additional data usage contract (Fuß & Wenzig, 2019). The anonymisation manual on neps-data.de provides information on which variables are available for which access mode (see https://www.neps-data.de/Data-Center/Data-and-Documentation/Start-Cohort-Students/Documentation).

2.8 Existing use of data

All publications that make use of the NEPS data can be found on the LIfBi website https://www.neps-data.de/Project-Overview/Publications. In December 2022, 140 publications could be counted for the cohort of first-year students, of which 19 relate to (prospective) teachers; they are listed in Table A12 in the Appendix. In addition, another paper has been accepted and will be published in 2023. Based on the Data Use Agreements concluded with the LIfBi for accessing the NEPS and LAP data (see https://www.neps-data.de/Data-Center/Research-Projects), the database is used by a large number of research projects. In total, there are 4,446 contracts for all NEPS Starting Cohorts combined. Consequently, more research results will be published shortly.

3.1 Repository location

All data collected by the LAP project is fully integrated in the datasets of NEPS SC5, which are published as Scientific Use Files by the LIfBi Research Data Center.

The most recently released SUF of NEPS SC5 and LAP data is version 17.0.0 (doi:10.5157/NEPS:SC5:17.0.0; NEPS Network, 2022). Data users are recommended always to use the most recently published issue. Older versions and their accompanying materials remain accessible.

After having signed the mandatory Data Use Agreement (see also Chapter 3.7), the datasets can be downloaded from the website https://www.neps-data.de/Data-Center/Data-and-Documentation/NEPS-Data-Portfolio. More information on how to access the data can be found at: https://www.neps-data.de/Data-Center/Data-Access.

3.2 Object/file name

Due to the complex structure and the rich amount of NEPS/LAP data, the data is stored in not just one but several different datasets (see Table 3). To help data users with this complexity, the LIfBi Research Data Center provides a merging matrix which gives an overview on how to link information from different datasets. In addition, users may find the semantic data-structure file useful. This file does not contain actual data but meta-data such as variable names, labels and scheme options. With this information data users can explore the data structure without signing a Data Use Agreement. For more information on the data structure and its complexity, see the data manual. All the materials mentioned are listed in Table 4.

3.3 Data types

The NEPS SC5/LAP Scientific Use File contains different types of data but mostly primary data from the target persons. Information on other (context) people such as partners and children was collected from the participants themselves. These data files include many derived variables, for example, classifications of occupations (e.g., ISCO, EGP) and education (e.g., ISCED, CASMIN), aggregated test scores and plausible values, and auxiliary variables.

In addition to primary data, methods data, context data, and process-generated employment-related social welfare data (administrative data) is also available. Context and administrative data can only be analysed on-site at the LIfBi and, in the case of the last-mentioned data, at the Institute for Employment Research (IAB) in Nürnberg or at some of the other IAB locations. In the methods data files users find paradata such as occurrence of problems during interviewing or testing, length of interview or survey, participation, and, in the case of telephone interviews and competence tests, information about the interviewer.

For NEPS SC5/LAP two types of context data are provided: Information for all 413 higher education institutions listed in the codebook of the Federal Statistical Office in 2010/2011, and information on the study programmes (aggregated to subject areas) offered by these HEIs (see Weber, 2014). Examples of variables to be found in this dataset are size and institutional control of HEI, gender composition and financial resources for each subject area in a HEI, and economic and social structure of the regional context of a HEI. The second type of context data provides information on the area of the respondents’ home such as age distribution, house type, milieus, family structure, and probability of payment default (see Schönberger & Koberg, 2017).

The last dataset to be mentioned is the NEPS-SC5-ADIAB (doi:10.5164/IAB.FDZD.2112.en.v1; Bachbauer et al., 2022). It is provided jointly by LIfBi and IAB and consists of the NEPS SC5/LAP data and the administrative data of the IAB. This administrative data includes, for example, information on employment history, benefit recipient history, and jobseeker history.

3.4 Format names and versions

The download version of the SUF is provided as Stata and SPSS files and data users need their own software licences. If users analyse the data via remote access (RemoteNEPS) or on site, they use the statistical software implemented on the LIfBi server system. In this case, Stata, SPSS, and R are available but not, for example, Mplus.

3.5 Language

Most of the documentation and supplementary materials are provided in English. The Stata files contain German and English variable and value labels. Stata users can easily choose between the languages (de, en) with the Stata command “label language languagename”.

3.6 License

All NEPS/LAP data published as SUFs is only accessible for scientific use by researchers, due to German data protection laws and corresponding participation agreements with all survey participants. As data access is regulated by Data Use Agreements, the application of (open) licenses is not required.

3.7 Limits to sharing

The LIfBi Research Data Center aims to make the NEPS/LAP data available to researchers as quickly as possible and tries to publish the Scientific Use File within 18 months of the end of fieldwork. Data access is granted to the scientific community for scientific purposes only and on condition that a Data Use Agreement is concluded (see below and Fuß & Wenzig, 2019). The amount of sensitive information that researchers can analyse depends on the access mode: The downloadable SUF version is characterised by the highest degree of data modification and anonymisation and contains little sensitive information. Datasets available via remote access provide more sensitive information than the download SUF. Finally, the on-site version, which requires a guest stay at the LIfBi, has the lowest level of anonymisation and the highest degree of sensitivity. If data users want to analyse the remote access SUF data or the on-site data version, they need to sign supplements to the NEPS Data Use Agreement. Using the data via remote access also requires a biometric authentication (keystroke biometrics). Further information on which variables may be restricted in use due to sensitivity and are therefore only accessible via remote access or on site can be found in the anonymisation report, which is released with every data issue (https://www.neps-data.de/Data-Center/Data-and-Documentation/Start-Cohort-Students/Documentation; short overview: https://www.neps-data.de/Data-Center/Data-Access/Sensitive-Information).

3.8 Publication date

The most recently published Scientific Use File has been available since 7 November 2022. With each new version the data will be extended and updated (e.g., by adding additional survey waves, data sets or corrections). Information on the publication date of earlier SUF releases can be found online (https://www.neps-data.de/Data-Center/Data-and-Documentation/Start-Cohort-Students/Data-and-Citation). By clicking on the respective link in the table, the corresponding documentation is accessible.

Since the last NEPS SC5/LAP survey was conducted in 2022, there are only a few panel waves that have not yet been published. According to the publication guideline mentioned above, publication of the 2021 CATI can be expected around spring 2023 (SUF version 18) and the release of the combined online and telephone survey conducted in 2022 is due in spring 2024 (SUF version 19). The data release schedule on the NEPS website (https://www.neps-data.de/Data-Center/Overview-and-Assistance/Zeitplan-en-US) is constantly updated.

3.9 FAIR data/Codebook

Findability: LAP data is integrated in the NEPS SC5 datasets and is therefore easy to find via the DOI (digital object identifier). Each subsequent data release is given its own and thus unique DOI. Part of this identifier code is the release number of that Scientific Use File, which is also used to link all related materials to a specific issue. In this way, data users can easily connect documentation material such as the data manual or codebook to the specific data. The LIfBi Research Data Center provides a wealth of information and documentation materials on their website (https://www.neps-data.de/Data-Center/Data-and-Documentation/Start-Cohort-Students/Documentation).

Accessibility: Access to the data is free of charge and managed by the LIfBi Research Data Center. Each researcher who wants to use the data has to sign a Data Use Agreement (see Chapter 3.7 for more information). It is recommended for new users to participate in one of the NEPS data training courses, which take place on a regular basis and are offered free of charge by the LIfBi Research Data Center. A signed Data Use Agreement is not limited to a specific SUF release; if a new SUF version is released during the contract period, data users are free to use the new issue. Information connected to outdated releases is still accessible via the LIfBi Research Data Center website.

Interoperability: Interoperability is ensured by providing SUFs in widely-used data formats (e.g., Stata or SPSS). For more information on how to access the NEPS/LAP data see Chapter 3.7.

Reuse: The LIfBi Research Data Center provides rich documentation materials (like codebooks, questionnaires and data manuals) and help for data users (like NEPSplorer, NEPSforum, data training courses). The website lists a huge variety of open access publications (e.g., NEPS Survey Papers, LIfBi Working Papers) and moreover, each data usage contract is listed with a short description on the website, which gives an opportunity to view ongoing research. The Data Use Agreement necessary to use the NEPS and therefore the LAP data clearly defines the regulations that all data users must follow.

Relevant meta-data: In addition to those materials mentioned in chapter 2.5.1, there are some further helpful sources of information: methods datasets; samples and weights; via NEPSplorer: information on what items were part of which survey(s), to what theoretical construct they belong, and which references have to be cited (BibTeX); and soon: documentation of the LAP survey instruments.

3.10 How to analyse the LAP data – technical advice, special variables and more

Due to the complexity of the study design and the longitudinal information on the teacher education students, working with the data can be a challenge. In particular, identifying the LAP target persons such as teacher education students, future teachers in preparatory service and in-service teachers can be difficult since no separate dataset for LAP exists, but rather all information is integrated in the NEPS SC5 SUF.

To facilitate data use, specific variables can be used to identify LAP target persons and the phase of teacher education they were in at the time of (selected) surveys.

Strengths and limitations of the data and data collection process

The LAP project has created an important and unique database on (prospective) teachers that has significantly expanded the existing data for empirical educational research in Germany. Combined with the NEPS study, it provides longitudinal information from over ten years of panel surveys, covering the entire teacher education phase and early years in the teaching profession with a large sample of (prospective) teachers in Germany. Unlike other large teacher surveys in Germany, the LAP data is not limited to certain German federal states, school types or subjects taught. Consequently, they allow for a more comprehensive analysis of teacher training and the teaching profession as well as comparisons between different teaching programmes. Through the publication of the Scientific Use Files (currently SUF 17-0-0), the data has the potential of being used for a variety of (secondary) analyses regarding relevant educational processes within the different phases of teacher education and aspects of professional activities of teachers.

Not only does the data provide the empirical basis for numerous questions on teacher education, the transition to the teaching profession, and the first years in the teaching profession, but furthermore the measurement instruments (further) developed in the project are also beneficial for future research on these issues. In order to cover a variety of relevant topics without overburdening respondents, many short scales were developed based on existing instruments. Sometimes, the use of short instruments is viewed critically; however, their quality was examined in detail and is therefore ensured. Documentation on the items and scales used in the LAP study is being prepared and will be published in the near future.

As is common in longitudinal studies, the NEPS/LAP study is confronted with significant panel attrition. Due to the wide range of personal information about the respondents, their psychological characteristics and life circumstances, interesting methodological analyses of panel attrition are possible.

Research potential

The data can be used to analyse the educational and occupational trajectories as well as the professional situation, professional practices, and self-assessed competencies of (prospective) teachers. A wide range of research questions about teacher education can be answered, such as the factors that influence educational decisions and trajectories, educational outcomes, and the importance of learning environments for competence development and educational decisions.

Regarding educational choices, little is known about the decision to stay in teacher education, to move to non-teaching programmes, to drop out of higher education or to move from non-teaching programmes into teacher education. Furthermore, moves within teacher education, e.g., changing the teacher training track or the teaching subject, have rarely been examined. Because the NEPS collects detailed event history data on educational biography, including moves to other higher education institutions, it provides a unique opportunity to examine the course of studies of teacher education students.

Regarding the teaching profession, for example, it is possible to examine the probation in the teaching profession or the influence of educational experiences in higher education and preparatory service on professional practices. The data also allows for analysing in-post learning, as the third phase of teacher education in Germany. In this context, it is possible, for example, to study the determinants of participation in further education or to investigate how collegial cooperation, the leadership behaviour of principals, and the participation in continuing professional education influence the professional competencies, well-being, resilience, emotional exhaustion, and career retention of teachers at the beginning of their career.

The sample also includes participants who chose teaching as a second career and did not pursue a traditional teaching degree. The data makes it possible to examine the factors that lead to the career decision of these teachers and to compare their career choice motives and self-assessed competencies with their traditionally certified colleagues.

By linking the LAP data with NEPS data, it is also possible to address a wide range of further research questions, including comparisons between different career choices and professions. For example, one can compare the professional well-being between teachers, lawyers, physicians and other occupational groups in order to check whether teachers represent a particularly stressed occupational group. One can examine how the choice for the teaching career or other career paths is influenced by interests, personality traits and school performance. It can also be analysed as to whether students from different disciplines differ in terms of study duration and dropout risk, or whether and to what extent the transition to the labour market after graduation and salaries depend on the field of study.

Implications for practice and policy

LAP data can be used to reduce research gaps in empirical teacher education research and expand knowledge in the field of teacher-related research. From the project’s research findings published and presented at conferences so far, as well as from further research with the data, recommendations can be derived for practice (e.g., the design of teacher education) as well as for education policy and monitoring. The findings may also provide the basis for reform measures in teacher education. For example, studies on predictors of dropout and career change can help to investigate some of the causes of the (internationally prevailing) shortage of teachers and to derive recommendations on how to support trainee teachers in successfully completing their teacher training and to make the teaching profession more attractive again. And by examining second career teacher more closely it will be possible to better assess their potential and risks and identify their support needs, regarding, for example, the transition into the teaching profession and their teaching quality.