Measuring education across different institutional settings: Lessons from longitudinal and cross-national studies |
|
| Chair | Ms Kerstin Hoenig (Leibniz Institute for Educational Trajectories (LIfBi) ) |
| Coordinator 1 | Professor Jon Miller (University of Michigan) |
| Coordinator 2 | Dr Silke Schneider (GESIS – Leibniz Institute for the Social Sciences) |
| Coordinator 3 | Mr Frank Goßmann (Leibniz Institute for Educational Trajectories (LIfBi)) |
Longitudinal studies of two or three decades that begin with students at birth or in secondary school provide a unique and important source of information about the various educational pathways that young adults follow. The traditional model of students in formal schooling, work, and retirement (in that sequence) is increasingly not applicable to large segments of the adult population of modern societies. Although this paper will utilize US longitudinal and cross-sectional studies, similar patterns and observations can be made from the British Birth Cohort Studies, the German National Education Panel Study (NEPS), and Swiss longitudinal studies, and numerous other European, Australian, and Korean longitudinal studies.
Almost all major cross-sectional national and international studies seek to measure respondent educational attainment. These measures are often simplistic and measure either highest degree earned or age at the time of the end of formal schooling. There have been some efforts to expand the educational questions in the World Values Surveys and other cross-national surveys, but few of those studies have developed or employ sophisticated measures of education attainment in a life-course context.
This paper will utilize the 30-year longitudinal record of the US Longitudinal Study of American Life to demonstrate the various educational patterns and pathways from approximately age 13 (7th grade in the US) to the mid-40's in 2015 and 2016. The data will document that various levels of education sought and the success of individual respondents in attaining the level aspired to. The data will also demonstrate that education has become a life-long endeavor and that most US workers (professional and non-professional) engage in various kinds of continuing education or technical upgrading throughout their work careers. These post-secondary educational experiences are important in understanding the economic standing of individual respondents and their ability to participate in the culture and politics of their society.
These longitudinal models will be contrasted with a 30-year time series of national adult surveys in the US. These cross-sectional studies have routinely attempted to measure educational attainment for each individual respondent and have included a variety of experimental items to assess post-secondary and educational experiences. The results will demonstrate that most cross-sectional surveys underestimate the level of adult educational attainment.
Finally, the February cycle of a national cross-sectional study conducted by the University of Michigan will be utilized to test some additional education items derived from the earlier analysis of longitudinal patterns and outcomes. The paper will make recommendations about items that might be used in cross-sectional surveys to improve the accuracy of measures of educational attainment and experiences over the life course. The recommendations should be applicable to both US and European studies.
The variable education is one of the central independent and explanatory variables in the analysis of survey data. In general, the educational attainment is coded using the International Standard Classification of Education. The most recent version is from 2011. The problem of ISCED is, first of all, the collection of the required information about the survey respondent’s qualification and educational characteristics. We suggest two separate questions. The first question registers the attainments from general education. The second question collects the qualification from the vocational and academic educational system.
The answer categories for both questions are the national degrees available in the national education system. We group the general education diploma into “no qualification“, the “first, basic general qualification“, the “second general qualification” – and if provided in the national school system the third qualification, and finally the “general entrance qualification to collage or university education”. The vocational education is grouped into “no qualification”, “qualification from dual systems”, “full-time vocational school degree”, “diploma from vocational collage”, “qualification from collages or universities of applied sciences”, and “university diploma”, ending with “doctorate”. Crossing the general with the vocational dimension, we get matrix where the survey answers are filled in and coded. These codes can be the actual ISCED 2011 categories or any other weighted scale. Now it is possible to compare and to interpret the cells from that matrix across the countries and educational systems.
We show with an example from Germany using data of the European Social Survey, how the instrument is to be applied and can be used in comparative survey research.
It is well established that survey data is afflicted by measurement errors. This is of course also true for educational measurements as for the variable 'school attended'. Unfortunately, there are hardly any possibilities to validate this information properly. Furthermore, there has been little research on determinants of false responses considering the kind of school attended. Generally speaking, recall errors and social desirability are plausible explanations for false responses. This study aims to reveal the proportion of false responses and to work out their determinants. Data of the National Educational Panel Study, Starting Cohort 3, wave 2 (sixth graders), is used. Since students were sampled in class context, the individual school-type information from the parents' CATI can be validated by the responses of the classmates' parents. As an approximation, it is assumed that the majority of parents’ responses within a class are identical to the true response. Hence, true responses can be identified by the statistical mode of the school type information, the false responses by deviations from the statistical mode. Using this dummy as dependent variable, binary logistic regressions are applied to discover determinants of false responses. Regarding recall error the following hypotheses are tested: parents have a higher probability to give false answers if: (1) they do not know the respective school type from their own experience due to migration or educational reform.and (2) their child changed schools often. Considering social desirability hypotheses tested are: parents give false answers if: (3) their child attends the lowest track especially (4) if they have high idealistic aspirations regarding their child's education.