The aim of the paper in hand is rather to focus on research methods and the problems arising with the use of particular research methods than to discuss the theoretical underpinnings of the issues discussed here.
The paper will deal with youth transitions. Under that term the transitions of young people from school to further and higher education and to the labour market is understood. Further discussions of the different concept of transitions will be omitted here in order to reserve more space and time for the discussion of the research methods applied here. However, the recent report of Furlong and colleagues (2003) in which they discuss the transitions of young people in West Scotland and which will form the base of the paper in hand gives good and detailed insights in the current discussion of that topic.
Eastern Germany and West Scotland are for both states (Germany and the U.K.) problematic areas: for example in both areas the unemployment rate is much higher than in the overall average of the state but the average income is significantly lower. Theoretically it can be argued that youth transitions into the labour market are more likely to be problematic in economically underdeveloped areas where there is already a larger competition amongst adult labour market participants and hence inexperienced youth is more likely to fail in competing with them. Hence, the comparison of the areas is a worthwhile venture. If there were structural common features in the transitions of West Scottish and Eastern German young people (of course especially in the cases of those with difficult transitions), effects of policies to address the issue of difficult transition could be compared and it would potentially be possible to learn from each other’s mistakes or successes.
In the paper in hand, first the youth transitions in Eastern Germany will be described using descriptive statistics and will be compared to the findings from West Scotland with is presented by Furlong et al. and to data from Western Germany were appropriate. A second step will, following the methodology of the Scottish researchers, aim to identify clusters of transition pattern in Eastern Germany and their description. A brief discussion of the appropriateness of a concept of linearity of transitions, which has been suggested for the Scottish case, follows at the end of the paper.
The data Furlong and his colleagues used for their analysis as well as the data, which will be used in this paper, is panel data, a special case of longitudinal data1. Whilst the definition of longitudinal data only requests that the data for at least two different points in time is representative for the same base population, panel data provides data for the same individuals for more than one point in time. The advantage of longitudinal data is that it offers the opportunity to analyse social change. However, in a longitudinal analysis that does not use panel data, the results can be misleading. For example if a researcher finds that in a population there was an unemployment rate of 15 per cent at time t1 and an unemployment rate of 15 per cent at t2 as well, it could be interpreted as there was no change in unemployment. However, in an extreme it might have happened that all people who were unemployed at t1 found a job in t2 and all the unemployed of t2 were working in t1. This kind of social change can only be examined with panel data that offers data on the individual level for each point in time. The major disadvantages of panel data are that it is extremely costly to collect and the dropouts (“panel mortality”) of individuals may bias the sample harming its representativeness for the base population. Therefore drop out analysis should be performed as a first step of an analysis with panel data to examine whether certain groups of respondents were more likely to drop out of the panel over the time. Due to time restrictions, however, this step was omitted in the analysis for the paper in hand.
The data used in this paper is the German Socio Economical Panel (GSOEP). This panel was started in 1984 and has now already gathered data of twenty years. After the reunification of Germany, a new additional sample was added to the survey. This “Sample C” was first surveyed in 1990; 2,179 households were surveyed in the first wave, sampled with a very elaborated random sampling method (described in Haisken-DeNew, Frick 2004: 140) which is regarded as having produced a highly representative sample at least for the first wave. The sampling units of the GSOEP are households, which fact results in the common problems of survey based on household sampling for example the exclusion of institutionalised people, who do not have a formal household, e.g. members of the armed forces, prisoners, asylum seekers2 or even students in University accommodation. However, since the tracking rules of the GSOEP are very extensive – basically every person which every lived in a panel household is tracked every year and when one of these persons founds a new household this household (and all of its members) becomes a panel household as well) those groups are no major problem for the GSOEP3.
However, it proved to be a certain difficulty to select the appropriate cases for the analysis. Furlong et al. did use a data set that has only been gathered for their project. The GSOEP, however, is a dataset that is designed to be used for as many as possible different uses from different subjects as sociology, economics, political sciences or geography. Therefore young people in Eastern Germany are only one of hundreds of potential base populations, which may be addressed by research with the GSOEP. To conduct comparative analysis with the West Scotland Twenty-07 Study the partial sample of the GSOEP used for this analysis shall be as similar to the Twenty-07 sample as possible. The main features of the panel group in the analysis are their age, their birth cohort and the length of their stay in the panel: Furlong et al. (2003: 2) draw a sample of 1,009 15-year-olds in 1987 (birth cohort 1971/72) and followed them for 14 or 15 years.
For the East German sample of the GSOEP data for 14 years is available to date. Hence it would be possible to use the data of individuals who where 16 years old4 when the panel started and stayed in the sample until the last wave surveyed so far. This has not been done for two reasons, one is a practical reason and the other is rather a theoretical consideration. The practical reason is that the sample size at the end of the period in question would be far too small to allow for any meaningful analysis as will be shown in some more detail later. The theoretical consideration is even more severe. Since the aim of this paper is to analyse the youth transitions in Eastern Germany, the generation which has been in age 16 in 1990 (birth cohort 1974) is not the best choice since this group (and basically everyone in Eastern Germany especially in the first few years after the reunification) is most likely strongly affected by the transitions from one political and economic system to the other as well which would interfere with the youth transition of that cohort5. Hence a compromise has to been taken between the demand that the target cohort must be young enough to be not severely affected by the post-communist effects anymore but old enough on the other hand to be not too far remote in birth year from the Twenty-07 sample and old enough to have enough waves for the analysis.
A first attempt has been made with birth cohort 1978/79; 16 years old in 1994 – 4 years after the unification but only seven years after the first wave of the Twenty-07 Study. In the Eastern German sample, there are only N=160 members of the 1978/79 birth cohort for whom data from their 16th year is available, 61 of which drop out before age 23 (which is the age Furlong et al (2003) use as the end of the early transition phase). Although the percentage of individuals remaining in the panel is with 61.9 per cent quite close to the figure, which Furlong et al quote (they receive an answer of 68 per cent of their original sample at age 23), the absolute number of respondents is too low to allow for meaningful analysis. Since there is no other one-year birth cohort that has significantly better case numbers, the cohort itself has been broadened. The aforementioned compromise between avoiding post-communist period effects and having a sufficiently large sample close to the Twenty-07 sample has been taken into account when the decision was made to include all individuals in the sample that were first surveyed (i.e. turned 16) between the years 1992 and 1997. This results in a birth cohort from 1975 to 1980. The breadth of this cohort is not an optimal outcome but has been accepted in favour for a number of 304 cases now in the analysis6.
Since the data of the GSOEP is stored for each wave separately and since the members of the different cohorts reached the different ages in different waves, the information had to be reorganised from wave specific information to life event (i.e. age) specific information. Here, it shows that not for all observations in the sample the information in the monthly diary is complete for every month. Only in 136 (44.7%) of the cases for all 96 months information is available; in 72 (23.7%) further cases information is available for 84 months – these cases are mainly from birth cohort 1979 for which the information of their 23rd year of life is not in the panel yet. However, in many cases the information gap is only one or two months long. These gaps have been filled with the information of the previous month. Whenever a gap was longer than two month no data manipulation has been carried out. In one case the information for the first month (January of the sixteenth year of life7) was missing, thus information of the previous month did not exist. Here the information from the second month has been used to fill the gap. Using that approach the number of complete datasets (i.e. information for 96 months or 84 months for 1979 born respectively) increases to 253 (83.2%).
Figures 1 and 2 below show the statuses recorded for each case in May of each year. The number of observations for each bar in the chart (and column in the table respectively) is different and varies from N=220 in May of the 23rd year to N=298 in the 16th year. In Figure 2 the main statuses of the participants of the Twenty-07 study are shown as well to allow for direct comparison.
Figure 1 shows that young people in Eastern Germany remain relatively long in Education. In the May of the year they turn 20 still nearly two third of the young people are in education. However, this figure drops very quickly in the next two years of live, leaving only about a quarter left with age 22 and 23. Employment is not a broadly taken up until age 20, when a fifth of the sample is employed. Here, it is worth stressing, that apprenticeships that include an employment8 are not counted as employment but rather as education although they were in fact working part time as well.
Fig. 1: Status in month May by average age in Eastern Germany
illustration not visible in this excerpt
Source: Own analy sis based on data from the German Socio Economic Panel 2003 (GSOEP 20)
The participation in youth training schemes9 is also counted as “education” and does therefore not appear as an individual item. As a relatively important item appears the military or community service as an alternative to military service which is obligatory for most male young people in Germany. Although the GSOEP records the voluntary social or ecological year (FSJ/FÖJ), which is open for both sexes from the 2001 wave in the same category, there are only male sample members in the category “military/ community service” – at age 21 as many as 21.4 per cent of the male sample were doing their compulsory service. Hence the male and female youth transitions in Germany must be regarded as very different; a fact that must be taken into account by a comparative analysis.
1 It is worth noting, that there is a difference between longitudinal data and longitudinal design. Longitudinal data can also be gathered retrospectively using cross-sectional survey design. In turn, it is possible to analyse only cross-sections of data gathered with a longitudinal design.
2 This issue was “solved” by defining the base population for sample C as “German Residents in the GDR”. Hence the sample covers persons in private households where the head of household was a citizen of the GDR (the sample was drawn 3 months before the reunification was completed). This means that 1.7% of the population in the GDR at this time was excluded from the sample as foreigners (who were mostly institutionalised).
3 Members of the armed forces for example or prisoners, who lived in a panel household before getting institutionalised were tracked and survey (if they agreed).
4 In the GSOEP the minimum age to be surveyed as a panel participant is sixteen; Furlong and colleagues started their project with 15 year olds. However, since Furlong et al. mainly track their respondents’ transitions from age 16 and even in their non-linear group the first year under observation is clearly dominated by school, this is not regarded as a problem.
5 Technically spoken, in this group and period, it is most likely that period effects (living in the period of transition from the socialist GDR to the capitalist FRG) strongly interfere with cohort effects (being a member of birth cohort 1973/74).
6 There were 463 Eastern German cases of the mentioned birth cohorts in the sample; 159 (34.3%), however, dropped out before turning 23.
7 In fact it is not necessarily the January in the sixteenth year of life but rather that year in which the respondent turned sixteen. That can be the January of the seventeenth year of life if a respondent turns 16 in January.
8 The so called “Dual System” of occupational training which is broadly employed in Germany consists of equal times in working in a company and in learning in a vocational school; normally changing between work and school every two weeks.
9 These are mainly Berufsvorbereitungsjahr (vocational preparatory year) and Berufsgrundbildungsjahr (primary