+++ Registration closed +++ Late registration via Contact +++

Poster Session


  • The panel study “Health Behavior and Injuries in School Age” (GUS) - Overview and potentials for educational research
    Robert Lipp (Frankfurt University of Applied Sciences)
    ≡ Abstract

    Most studies in the social sciences are designed to answer distinct research questions. However, the range of topics covered in a survey or interview may also include areas relevant to fields of research not primarily targeted by the scientists conducting the study. The study “Health behavior and Injuries in School Age” (GUS) presents such a case. Its primary goal was to identify causes of injuries in students in secondary education schools. However, with around 80 questions per student-questionnaire including different aspects of the personality, family environment, academic performance and social situation of the respondents, it provides a valuable asset for researchers from different disciplines and with various research interests.
    The data was collected by the Research Centre of Demographic Change (Forschungszentrum Demografischer Wandel, FZDW) at the Frankfurt University of Applied Sciences as a panel survey over the course of six annual waves from 2014 to 2020. The dataset is available for re-use at the Research Data Center at the Leibniz Institute for Educational Trajectories (Forschungsdatenzentrum, RDC LIfBi) and contains information from over 40,000 questionnaires. It also includes data from the heads of schools and the interviewers present during the classroom surveys.
    The presentation gives insights into the methods and field work of the GUS study and serves as an introduction into working with the data and documentation materials. It especially highlights the research potentials with a focus on topics relevant for educational research. Important variables both surveyed and generated are presented and examples of possible analyses are given.

  • A growing data treasure for labor market and educational research: NEPS survey data linked to administrative data of the IAB (NEPS-ADIAB)
    Nadine Bachbauer(LIfBi Bamberg / IAB Nuremberg)
    Clara Wolf (LIfBi Bamberg / IAB Nuremberg)
    ≡ Abstract

    NEPS-ADIAB is a large-scale data product that provides a unique source for the empirical analysis of life course research questions. It is jointly administered by the Institute for Employment Research (IAB) and the Leibniz Institute for Educational Trajectories (LIfBi). NEPS-ADIAB includes survey data from the National Educational Panel Study (NEPS) and administrative data from the IAB, combined at the individual level using a record linkage procedure. While the NEPS data cover issues such as educational trajectories and decisions, competencies and skills, learning environments, attitudes, etc., the administrative data consist of detailed employment histories dating back to 1975 and up to 2020, plus extensive information on establishments.
    NEPS-ADIAB is particularly suitable for analyses of the nexus between educational and employment careers, short- and long-term returns to education, transitions from school to vocational training and working life, adult education and lifelong learning, as well as discontinuities in professional careers due to parenthood or unemployment. So far, four of the six NEPS Starting Cohorts are linked and released as NEPS-ADIAB data products. The four cohorts available as NEPS-ADIAB are SC1 (mothers of newborns), SC4 (students from grade 9 on), SC5 (first-year university students), and SC6 (adults). By the end of the year 2022, NEPS-SC3-ADIAB (students from grade 5 on) will be accessible for analyses, too.

  • The influence of different allday-schooling programs on the development of elementary school students
    Fabian Siegel (LMU Munich)
    ≡ Abstract

    Since there is no clearly unique definition of an all-day school (Wiere, 2011a, p. 35), this project tries to take a more differentiated look, contrary to the usual minimal definition (c.f. Strietholt et al., 2015, p. 745). Based on Fend's theory (1981), it will be emphasized that school is a diverse instance of socialization. Also, the time aspect of socialization will be addressed in order to compare the possible influences of different time-intensive schooling types. For this purpose, Bourdieu's theory of capital will be used, which emphasizes the importance of cultural capital for school success (Bourdieu, 1983; 2001). Interestingly the accumulation of capital, specifically cultural capital, primarily "takes time" (Bourdieu, 1983). Although a difference is assumed due to the time spent in the respective socialization instances, no simple linear relationship is assumed (Holtappels, Radisch, Rollett & Kowoll, 2010; Steinmann, Strietholt & Caro, 2019). The actual state of research rather suggests that a possible effect of allday-schooling is dependent much more factors, such as the individual participation intensity (Fischer, Sauerwein, Theis & Wolgast, 2016). In summary, a differentiated comparison of the types of institutionalized care in the primary school sector will be made.
    The focus on the primary level lends itself to this type of school due to the more multi-layered tasks of this schooling type. Additionally, there is an increasing proportion of all-day schooling in this schooling type (Fölling-Albers, 2019, p. 488). Likewise, theoretically this schooling type can be seen as a crucial phase in which the compensatory effect, described as catching up time (Rohlfs, 2011, p. 88), should be enabled. For the analysis the data of the NEPS (start cohort 2) is used, which allows not only the comparison of the traditional halfday-schooling versus allday-schooling. Within the first group, a distinction can be made based on participation in a "Mittagsbetreuung" or in the second group based on participation in a "Hort".
    Furthermore, the dataset allows to look at the individual development of students. Therefore causal-effect relationships can be considered in more detail, for example the remedial use of all-day schooling (Holtappels et al., 2010). In addition to the ability in reading, as an essential condition for academic success (Fischer et al., 2016; Tillmann et al., 2018), and in mathematics, as one of the "core competencies" of schooling (Steinmann et al., 2019, p. 11), other constructs will be analysed. The previous achievement focus of allday-schooling research (Wiere, 2011b) will thus be expanded to include social and motivational constructs. The primary goal of this project is not to highlight the "effectiveness" of one type over the others, but to raise awareness of the complexity of the research object.

  • Reading literacy prediction based on the bacground PIAAC data
    Katarzyna Chyl (ERI Warsaw, Poland)
    Aleksander Molak (Tel Aviv, Israel)
    ≡ Abstract

    Background: Reading literacy is an important skill, crucial in everyday modern life. Nonetheless, in many languages, including Polish, there is still no adequate tool to measure reading comprehension in adults, especially on the functional level, which could be useful in both research and practice. In the 2011 PIAAC study, Poland scored lower than the OECD’s mean in reading comprehension, emphasizing the importance of adequate diagnosis of reading problems in Polish adults. Even though PIAAC tools are confidential, extensive datasets, including reading comprehension scores and background questionnaire responses are made available on OECD’s website. We aimed to answer the question which items were the most predictive for reading comprehension scores.
    Data Source: We used publicly available Polish PIAAC questionnaire data including the responses of ~9500 participants. All Python scripts used in the analysis are shared on the project’s OSF.
    Methods: To assess which independent variables have the strongest association with the dependent variable, we used a tree-based XGBoost model (Chen, Guestrin, 2016). The strength of association has been evaluated using Shapley values – a game-theoretic approach to complex model interpretation (Lundberg, Lee, 2017). This approach is more powerful than using a classic linear model in three ways. First, it allows to model non-linear relationships. Second, it allows for automatic extraction of complex interactions between variables and – third – non-linearities and interactions are taken into account when assessing the strength of association between independent and dependent variables. To assure the best fit to the data, a hyperparameter tuning procedure was used. All relevant hyperparameters were optimized simultaneously using Tree Parzen Estimator, a multi-dimensional Bayesian optimization technique. A Python package Hyperopt was used for this purpose (Bergstra, Komer, Eliasmith, Yamins, Cox, 2015). To avoid overfitting (James, Witten, Hastie, Tibshirani, 2013) hyperparameter tuning was conducted using a 10-fold cross-validation split.
    Findings: We identified background PIAAC questions having the highest impact on the dependent variable PVLIT (reading comprehension score in PIAAC) in terms of Shapley values. In the top 5, we found questions about education (B_Q01a: Education: highest qualification level), skill use in everyday life (H_Q05f: using word processor H_Q03c: calculations, H_Q05a: using e-mails), and socioeconomic background (J_Q08: number of books at home).
    Conclusions: The results reflect both the antecedents and consequences of low reading literacy (eg. employment status or occupation may reflect low reading literacy). Finding the correlates can be useful in predicting risk groups for policy interventions or screening instruments to identify such groups for research purposes.

  • New task format in the NEPS reading competence test
    Kathrin Thums (LIfBi Bamberg)
    Karin Gehrer (LIfBi Bamberg)
    ≡ Abstract

    The National Educational Panel Study (NEPS) investigates, among various other issues, the development of different competences across the life span (Weinert et al., 2019). One of the competences that is investigated longitudinally and coherently in the NEPS across all starting cohorts is participants' reading competence. Reading competence is defined in NEPS as functional ability to comprehend written texts (Gehrer et al., 2012).
    With regard to the assessment of reading competence, the conversion from paper-based to computer-based assessments (CBA) in the NEPS allowed for implementation of new task formats. So far, three types of tasks formats were included in the reading competence test: simple multiple choice (MC) items, complex multiple choice (CMC) items, and matching (MA) items. As a new format in the CBA reading competence assessment text-enrichment-task (TET) items were added. The task requires to complement a text meaningfully with three to four additional sentences. That is, participants have to find the correct position in a text where each additional sentence fits best. To successfully answer TET items, participants have to understand the story, line of arguments, or action described in the text before they can make a decision about the correct position of the additional argument. This task might be particularly challenging for weak readers, as they first need to create a situation model of the text before they can make a correct decision.
    In order to investigate the quality of the new TET items the paper reports item fit statistics. Moreover, it addresses the following questions: How do different readers perform on the TET, and can differences between good and poor readers be identified? The NEPS reading competence test with the TET was administered for the first time in a NEPS study with 17,972 participants (53% women) from three different Starting Cohorts (SC4 N = 6,871; SC5 N = 4,816; SC 6 N = 6,441). Initial results show that particularly good and younger readers have higher reading competence scores in TET than poorer and older readers. Due to the wide age range of the sample (18 to 73 years), further analyses should be conducted separately for age groups. Because the test was computer-based, in addition to reading competence, the handling of the technique could also play a role especially for older participants. Therefore, the results will not only be discussed in terms of participants’ competence, but the measurement of reading competence will also be discussed in more detail.

  • Digital natives = digital experts? Designing a novel instrument to measure digital competence in NEPS-SC8 secondary school students
    Sümeyra Tural (LIfBi Bamberg)
    Mariann Schwaß (LIfBi Bamberg)
    ≡ Abstract

    With the digitization of modern day society, the competence of handling information and communication technologies (ICT) has become increasingly significant. Newer generations that grow up in a continuously more digitized world from a young age and almost inevitably familiarize themselves with the technologies and devices of this new environment, are labelled as ‘digital natives’ (Prensky, 2001), ‘generation internet’ or ‘whiz-kids’ (Lee & Chae, 2007). Often, these terms are used to imply a certain level of competence in operating with the internet and online media that is thought to be automatically acquired by frequent usage. On the other hand, many pieces of anecdotal and scientific evidence indicate the challenge that the manifold information of modern media and the internet can pose for growing children and adolescents. “They are not only called ‘whiz-kids’ […] but also ‘risk-kids’”, state Valcke et al. (2010) and emphasize the importance of teaching children a critical and reflected approach.
    In order to capture the aspect of critical reflection and social communication within the internet, a new test to measure ‘digital competence’ has been developed for Starting Cohort 8. The test aims to reflect the knowledge and abilities that are needed to responsibly approach the possibilities but also the risks and consequences of the ubiquitous accessibility of the internet and to reflect not only on the content of digital media, but also to identify their modes of action. In this way, the newly developed test complements the aspect of handling technology and programmes that is already established in the NEPS and that is represented by the competence domain ‘ICT’.
    This poster presentation aims to introduce the conceptual framework for this novel test domain. In a synthesis and extension of former research (e.g. Calvani, Fini & Ranieri, 2009; Carreto, Vuorikari & Punie, 2017; Digital Dannelse, 2009; Fraillon et al., 2019), four main facets with multiple subfacets were defined for the digital competence test of the NEPS: (1) Information competence, (2) Detection of intentions and strategies, (3) Communication and interaction competence, and lastly, (4) Data handling competence.
    The computer based test to measure digital competence will first be applied in a pilot study for the NEPS-SC8 in 2022. Data will be gathered in regular schools and schools for children with learning difficulties, in grade 6 and 8, respectively. In the long run, the NEPS data will allow to investigate the development, influencing factors and consequences of digital competence in secondary school children.

  • How useful are interviewer observations? Assessing data quality and future participation in the National Educational Panel Study
    André Pirralha (LIfBi Bamberg)
    ≡ Abstract

    Longitudinal surveys are gaining importance as sources of information for academics and policymakers. However, the usefulness of this kind of survey is directly linked to the quality of the data collected: poor-quality survey data can produce biased and misleading results. While assessing survey data quality is a challenge, recently a lot of research is directed to paradata and what this tells us about the survey data collection processes. While various types of paradata are available, here I focus on post-survey interviewer observations of respondents (interviewer observations).
    It is very common for the interviewer to answer some kind of close-ended questions after the interview about the respondent and the conduct shown during the interview. The idea is that respondents registered with negative observations by the interviewers will provide data of lower quality. More recently, research has also shown that interviewer observations provide valid indications regarding the respondents’ willingness to participate in the next survey waves. However, while interviewer observations are a useful and cost-efficient tool to evaluate data quality, these subjective judgments may also induce error. Some studies have questioned whether interviewers have the skills and motivations to evaluate respondent performance instead.
    The focus of this presentation is on the usefulness of interviewer observations and assessing how they are related to actual survey response and subsequent survey participation. Using data from the National Educational Survey Panel (NEPS), I assess to what extent interviewer observations predict short-term participation (subsequent wave) and long term participation patterns in Computer-Assisted Telephone Interviews (CATI).
    Interviewers record three kinds of post-survey interviewer observations in the NEPS CATI interviewing: first, one question capturing whether there were disturbances during the interview; second, whether the interviewer perceives the respondents’ answers to be reliable; and third, whether the interviewer is tired. The results of this research will be used to determine whether interviewer observations are valid indicators of future survey participation behavior in the NEPS.

  • CILS4NEPS - A harmonized dataset bases on CILS4EU and NEPS SC4
    Jörg Dollmann (MZES Mannheim)
    Lena Arnold (MZES Mannheim)
    Andreas Horr (LIfBi Bamberg)
    Victoria Kerzner (MZES Mannheim)
    Regine Schmidt (University of Bamberg)
    Hanna Soiné (MZES Mannheim)
    Markus Weißmann (MZES Mannheim)
    ≡ Abstract

    In this project, we harmonised the datasets „Children of Immigrants Longitudinal Survey in Four European Countries“ (CILS4EU) and starting cohort 4 of the National Educational Panel Study (NEPS), thus creating new potential for analyses previously not available. By combining both data sources, analyses for Germany benefit from increased case numbers for social and ethnic subgroups as well as evens such as the transition to work and further education. Since CILS4EU was conducted in four countries, the harmonised data additionally allow international comparisons of school- and early work-careers of youths from NEPS SC4 with those of England, the Netherlands, and Sweden.

back to top

Conference Host

Leibniz Institute for Educational
Trajectories (LIfBi)
Wilhelmsplatz 3
96047 Bamberg