THE MEASUREMENT OF SCHOOL-TO-WORK TRANSITIONS AS

PROCESSES

ABSTRACT: Measuring individual transitions means capturing a process with a specific time
dimension. The established analysis of school-to-work transitions focuses on single status
changes, such as those between education and unemployment. As longitudinal datasets
became increasingly available, the periodical character of transitions has deserved attention,
mostly in terms of studies that used event history models. But even these kinds of studies
continued to focus on single status changes, which are not determined by theory but by the
respective research question, or by data availability. Hence, the analysis of micro-level
transitions remains selective, because there is no common idea how to explore them. As a
result, school-to-work transition research is in danger to overlook important aspects of this
life-course trajectory. This paper argues that the reason is the missing theoretical definition of
a transition. Given the growing complexity of school-to-work transitions, the status change
concept becomes inappropriate for their analysis. The predominance of hypothesis-testing
methods together with an underrepresentation of explorative methods lead to a disregard of
the process character of school-to-work transitions. This considerably limits the gain of new
scientific insights. Recent methodological developments regarding the explorative analysis of
longitudinal processes, namely sequence analysis, offer the possibility to cope with the
complexity of school-to-work transitions. The paper aims at comparing the advantages and
drawbacks of two methods in analysing transitions, and advocates a combined research design
between explorative and hypothesis-testing methods.

1. Introduction
The transition from school to work remains a question of enduring relevance, despite the fact
that there is no shortage of policy activities and research in this field within the past 30 years.
It remains obvious that labour market mechanisms are incapable of solving the youth
integration problem without government or corporate intervention. A smooth transition into
the labour market preserves labour market supply, prevents loss of human capital, and
provides life perspectives and independence for young people. Failure at this point may have
long-term consequences for peoples employment career and for social inequality (e.g., Mroz
and Savage 2006; Steijn et al. 2006; Bell and Blanchflower 2010; Julkunen 2010). However, the
development of youth unemployment in the OECD within the last two decades shows no
improvement. With the economic crisis since 2008, the situation for young people
deteriorated in many countries. Additionally, some researchers observed processes of
polarisation or segmentation in the youth labour market that is, increasing risks of social
exclusion for specific disadvantaged groups such as immigrants (e.g., Kogan 2004) or low-
skilled school leavers (e.g., Solga 2002; Fenton and Dermott 2006). For these reasons, both
policy makers and social scientists continue to be interested in school-to-work transition
research.
Although a huge body of existing research on school-to-work transitions suggests that there is
nothing left to be examined, there are fundamental research questions that remain to be
addressed: first, a theory that appropriately defines transitions does not exist. Second,
individual transitions are only analysed as single events, not as processes and, third, the
predominance of hypothesis-testing methods undermines the conceptualisation of school-to-
work transitions.
Shanahan (2000: 683) suggests a couple of methodological innovations, one of which is
sequence analysis, a methodological tool that uses the potential of longitudinal data and is
capable of dealing with the complexity of the transition process. Sequence analysis is an
exploratory rather than a hypothesis-testing method. Given the increasing amount of
longitudinal data available, computational tools that aim at reducing complexity in a
meaningful manner have increasingly gained importance. Sequence analysis allows for
detecting structure in a seemingly chaotic mass of information, while having a holistic view,
because sequences are treated as entities. Sequence analysis enables researchers to go
beyond simplified definitions of what a transition is, where it starts, or where it ends.
However, it has to be applied cautiously and unfolds its full power only within a meaningful
research design, at best when combined with other (causal) methods.
The next section deals with theoretical approaches that address school-to-work transitions in
order to illustrate a theoretical gap regarding the definition of micro-level trajectories between
school and work. Thefollowing section reviews school-to-work transition research with a focus
on the operationalisation of the transition process. The final section assesses the advantages
and the drawbacks of different methods of transition analysis and advocates a strategy of
analysing school-to-work transition from different angles.

2. Theorising school-to-work transitions

[...] transition-system research often appears theoretically eclectic and fragmented.
(Raffe 2008: 278)

Despite the central importance of young peoples labour market entry, an integrated theory of
the transition from school to work is not available. One can argue that this not necessarily
constitutes a problem, but borrowing theories that naturally have a different and more
general focus, most probably tend to overlook important aspects of the research object. This
section exemplifies a couple of these theories, which researchers apply to school-to-work
transitions.
Life course research aims at formulating adequate and sound theoretical rules for pathways,
and there have been some attempts, for example the work on individual transition types by
Sackmann and Wingens (2003), who examine school-to-work transitions from a life course
perspective; or human capital theory, which aims to analyse the relation between education
and labour market outcomes. But these approaches have not been able to provide satisfactory
explanations of the country differences regarding the micro-level transition from school to
work. The standard explanations provided refer to cultural, political, or institutional factors.
Related theories suffer from the fact that they are either too general or too specific in nature.
In the former case, theories are too abstract that they hardly provide valuable hypotheses for
the cases under observation, whereas in the latter case, specific theories are meaningless if
applied to all the cases of the basic population.
From the perspective of institutionalism, the Varieties of Capitalism (VoC) approach argues
that social protection is strongly related to individual skill investment and, therefore, also
related to different skill formation systems. The argument is that firms only invest in specific
skills if social protection institutions, such as employment and unemployment protection,
safeguard their returns. The absence of such institutions leads to investment in transferable
skills by employees and employers (Estevez Abe et al. 2001). The form of collective skill
formation systems can be explained well by the VoC approach, but individual transition
patterns cannot be hypothesised.
The theoretical foundations on which research in the field of micro level school-to-work
transitions is based, therefore, remain eclectic (Raffe 2008). Researchers typically borrow
theoretical pieces from different disciplines and neighbouring research fields, such as
economic labour market theories. For example, human capital theory is used for explaining the
effects of educational credentials on labour market outcomes (Becker 1962, 1975). But
although this theory is mainly focused on the individual level, it is hardly able to explain how
institutions determine the process of the school-to-work transition. Segmentation theory, in its
original version, assumes the existence of two segments within the labour market which differ
in terms of wages and employment characteristics, while there is no mobility between them
(Doeringer and Piore 1971; Edwards et al. 1975; Reich 2008). In later variants of this theory,
certain groups, among them youth and school leavers, were identified that constitute labour
market segments of their own. Although this theory includes the institutional level by
connecting labour market segmentation to the collective skill formation system (Sengenberger
1992), it does not allow for explaining country differences (Ashton 1988). Segmentation theory
is often applied to school-to-work transitions, mainly with respect to the incidence of non-
standard forms of employment among labour market entrants (e.g., de Vries and Wolbers
2005). But, again, the concrete shape of micro-level school-to-work transitions cannot be
explained.
Very close to the original segmentation theory is Marsdens dichotomy of internal vs.
occupational labour markets (Marsden 1990, 1999), which is frequently used in school-to-work
research as well. It assumes that the situation of young labour market entrants depends to a
large extent on the fact that labour markets are structured either internally or occupationally.
However, recent studies have questioned the explanatory power of this dichotomy (Gangl
2003; Brzinsky-Fay 2007).
Additionally, some researchers have employed power resources approaches such as insider-
outsider theory (Lindbeck and Snower 1989), which refers primarily to the political origins of
certain groups being disadvantaged regarding labour market access or employment conditions.
The argument here is that job holders the insiders are organised in employee organisations
and have advantages over job seekers the outsiders when it comes to negotiating their wages
and working conditions. In the long run, the situation of the outsiders and school leavers are
by definition outsiders deteriorates. But again, the insideroutsider theory has not been able to
explain country differences regarding the school-to-work transition in a satisfactory way.
Apart from human capital theory, these theories aim at explaining micro-level effects (e.g.,
labour market entry) by macro-level causes (institutions), but they are not able to explain
school-to-work transitions in their whole complexity. As a consequence of this theoretical gap,
fixed definitions of the key concepts that are usually applied in transition research, such as
transition or trajectory (cf. Brzinsky-Fay 2010: 7), do not exist in these theoretical frameworks.
The only theoretical framework that directly focuses on individual transitions is the life course
perspective.
The term transition is understood as a change between an initial and a destination status. The
duration of a transition is not determined explicitly; therefore, it is used to describe either a
very short status change or a prolonged process that involves many status changes. Regarding
school-to-work transitions, it can be assumed that they become increasingly protracted and
involve many status changes; as a result, transitions must be examined in a longitudinal way.
Because the definition as single status change is very clear and straightforward, life course
analysis to a certain extent became synonymous to transition analysis in a sense of status
change analysis (cf. Sackmann and Wingens 2001). The term trajectory represents a
structuralist view, implying that labour market destinations were largely determined by social
forces that are outside of the control of individual social actors (Evans and Furlong 1997: 18).
Other authors prefer the use of the term pathway instead (e.g., Shanahan 2000). One can
understand trajectories as normative transitions arising from certain institutional
arrangements, but finally, the concept of trajectory remains quite fuzzy, because of its
simultaneous usage to describe longer transition periods. It is not determined, whether
trajectories analytically serve as a tool for inter-individual comparison (trajectory as an
aggregation of individual processes) or as a tool for intra-individual comparison (trajectory as a
certain part within the life course) (George 2009: 164).
The life course perspective conceives life courses as endogenous causal relations (Mayer
1990), implying that one status depends on the preceding status(es), no matter what the
causal relation looks like. An alternative understanding is that of a holistic logic of the life
course, which assumes that life courses are a meaningful composition as a whole. This
understanding is seen as a more suitable concept for analysing some of the core propositions
of life course research (Aisenbrey and Fasang 2010: 422), namely the standardisation (Kohli
1985), the individualisation (e.g., Buchmann 1989), the de-standardisation (Widmer and
Ritschard 2009) and the de-institutionalisation of life courses (Held 1986). Apart from that,
researchers have attempted to postulate life-course regimes in connection with those
institutional characteristics, that are seen as fundamental within the VoC approach (Mayer
2005). In liberal market economies life courses are based on weaker societal relationships
because of the lack of coordination, whereas in coordinated market economies, life courses
are based more on long-term commitments. This explains the existence of high investments in
vocational skills as well as it allows for hypothesising the form of certain life course
trajectories. This approach remarks a promising step towards theorising individual school-to-
work transitions, but it remained on a too abstract level. Closely related to these propositions,
there was a tension between individual agency and institutional structure, which is also
reflected in the methods applied within the life course concept.
The above list of theories employed by researchers illustrates the theoretical gap regarding the
definition of transitions and trajectories that constitutes a serious problem for analysis,
because central propositions of life course research cannot be examined properly. The same is
true for the question how institutional frameworks influence school-to-work transitions.
Theoretical answers to these questions can only be obtained by including as many countries as
possible (cf. Bynner et al. 1997: 6). There is still need for more exploratory work on micro-level
school-to-work transitions, because theoretical statements about how they are influenced by
institutions can only be made if the concept of a transition and trajectory is clear, in order to
be operationalised appropriately.

3. The analysis of transition as processes

It is important to recognize that the transition is a process that occurs over time. The
initial education-occupation association is not necessarily equally meaningful in all
cases.
(Kerckhoff 2000: 463)

Measuring transitions means capturing a process with a specific time dimension, whose extent
needs to be determined by the research question, or by data availability. Analyses of labour
market transitions usually examine single status changes, for instance between employment
and unemployment, or between education and employment. As I have shown in the previous
section, this is also due to lack of theoretical clarity regarding the transition concept. As a
consequence, transition analysis is limited to status change or time point analysis, which
mostly applies event history models. These models are extremely powerful in looking at the
conditions and/or effects of status changes, but they limit the trajectory between education
system and labour market to a status change. There have only been few attempts that treat
school-to-work transitions as sequences composed of more than one or two statuses (Berger
et al. 1993). Usually, cross-sectional data are used for the calculation of aggregate measures
(e.g., OECD 2010), whereas longitudinal data are employed predominantly for individual
measures.
At the individual level, the basic indicator used in school-to-work transition research is the first
transition into employment, but it is not necessarily meaningful. Researchers have tried to
detect the crucial status change by constructing concepts such as the first significant job
(Russell and OConnell 2001; Korpi et al. 2003) that lasts at least 6 months, or the first job
after leaving school for the last time (e.g., Arum and Hout 1998), for example. This limitation
serves the purpose to exclude short, probably erratic or irrelevant employment periods.
However, the determination of the time period that has to be regarded as not significant
remains to a large extent arbitrary a problem that cannot even be eliminated by increasing the
quality of available data. Kerckhoff explains that [...] the problem of defining the first job
becomes the most troubling when studies are based on the very best possible longitudinal
data (2000: 471).
Unemployment duration of school leavers is also taken as an indicator to describe the quality
of the transition (cf. Mu l ler and Gangl 2003), but for the same reasons, its validity seems
questionable. The duration of unemployment after leaving school simultaneously measures
the duration of search processes, the general availability of jobs on the (youth) labour market
as well as mismatch of any kind. Apart from their arbitrariness, both measures have the main
disadvantage of seeking to qualify school-to-work transitions by focusing on only one single
status change, which in this case is a passage from unemployment to employment. Other
labour market statuses such as inactivity (military service, household and childcare activities)
or education beyond compulsory schooling or participation in active labour market
programmes are disregarded, and it is not clear if the school-to-work transition period ends
with the first incidence of employment. Transition periods involving more than one
employment episode are very different from those showing continuous employment regarding
both their nature and their impact on future employment prospects. Because the increase in
complexity of labour market entry (Berger et al. 1993) takes place in every industrialised
country, the difference in the effects of institutional arrangements on individual school-to-
work transitions can only be assessed by taking into account this complexity that is, by
applying longitudinal indicators to longitudinal information.
Researchers also use more elaborated individual-level measures, for example the risk of having
a fixed-term job (cf. Gebel 2009; de Vries and Wolbers 2005; Scherer 2004). Fixed-term
employment, on the one hand, is seen as a flexible form of employment that allows employers
to screen new employees while keeping the flexibility to lay them off when either the
economic situation requires retrenchment or the performance of the young employee does
not meet productivity expectations. On the other hand, fixed-term contracts imply a shift of
economic risk towards labour market entrants in terms of decreasing employment security.
The question is, whether fixed-term employment appears as a bridge or as a trap (e.g., Gash
2008). Apart from these implications, the indicator of fixed-term employment risk, like the
aforementioned indicators, only makes sense if observed and analysed in a longitudinal
perspective.
Considering school-to-work transitions as periods also helps avoiding certain problems
inherent to the comparative analysis of school-to-work transition systems: for example, the
definition of what has to be regarded as education or work, which is important for
vocational apprenticeship systems. Kerckhoff (2000: 463) states: [...] a decision has to be
made as to whether the period spent in the dual system is time in school or at work. That is,
does the transition from school to work take place before or after the period in the dual
system? Another problem for comparative analysis results from country differences regarding
the degree of coordination between education and the labour market. When comparing
Germany and the USA, Kerckhoff (2000: 465) mentions that multiple entries into the labour
force are more common in the USA. Paying attention only to the first status change between
education, unemployment, or inactivity on the one hand, and employment on the other,
means overestimating the integrative potential of the transition system in the USA, because
later exits and re-entries into employment remain disregarded.
Despite all individual-level indicators provide the possibility to analyse school-to-work
transitions comprehensively, they all suffer from the time point problem. It can be assumed
that both the frequency and the nature of the status changes within the school-to-work
transition period vary between countries. Therefore, transition analysis that refers only to one
single status change most probably leads to biased conclusions. Shanahan (2000: 683) lists a
couple of problems arising from researchers decision to limit their analysis of the pathway
from youth to adulthood to such life course markers. To address these problems, a couple of
methodological innovations are suggested, one of which is sequence analysis.
Apart from these technical questions of how to measure a time process, the question arises
whether improving data quality has effects on methodological and theoretical development.
One can observe a three dimensional increase in the size of datasets: first, the number of cases
increases because of progress made in survey methodology. Second, the number of variables
increases because of growing complexity of research questions. And third, the number of time
points for which observations and variables are available increases because of the growing
availability of longitudinal datasets. This requires the application of algorithmic explorative
methods capable of sorting information in meaningful ways in order to extract crucial
commonalities and/or differences in individual pathways.

4. Extending methodology

Since the late 1980s, event history analysis meritedly has become the method naturally
connected to (quantitative) life course analysis. However, some of its drawbacks should be
noted: first, its reliance on discrete events and its assumption that life courses are
stochastically generated lead to an emphasis of individual agency for the formation of life
courses. Second, event history models are very sensitive on how the event of interest is
operationalised. And third, the predictors of certain events are to a large extent dependent on
the occurrence of different kinds of events. These drawbacks have led to critical disputes
among scholars since the 1990s (Abbott 1990, 1995), such as the BreimanCox debate in
Statistical Science (Breiman 2001), which contributed to the spreading of sequence-analytical
methods in the field of transition and life course analysis.
Sequence analysis was originally invented by biologists in order to find out the extent to which
two DNA sequences are homologous, or, in other words, to determine the distance between
them (Kruskal 1983). The established degree of similarity then allowed for drawing conclusions
about a common ancestor of two DNA strands. The first sociologist to used sequence analysis
was Andrew Abbott, who analysed musicians careers and ritual dances (Abbott 1983; Abbott
and Forrest 1986). Sequence analysis was seen as a qualitative tool in the context of historical,
narrative sociology. Due to the limited capacity of computers in those years, analysis was
restricted to few cases with short sequences. Since the 1990s, researchers have begun to focus
on individual sequences, such as class careers (Halpin and Chan 1998), employment
biographies (Abbott and Hrycak 1990; Blair-Loy 1999; Pollock et al. 2002), family histories
(Elzinga and Liefbroer 2007), school-to-work transitions (Scherer 2001; Schoon et al. 2001;
McVicar and Anyadike-Danes 2002; Brzinsky-Fay 2007) and life-course trajectories (Billari and
Piccarreta 2005; Wiggins et al. 2007; Martin et al. 2008). As the technical situation improved
with the implementation of sequence analysis in the statistical software packages Stata1 and
R,2 researchers from different disciplines became able to compare sequences of a large
number of individuals, finding out similarities, quantifying certain characteristics or grouping
them into ideal types. The increasing number of applications also led to a discussion about the
potential and limitations of sequence analysis methods. In recent years, a number of
researchers have worked on enhancements of the method itself (Gauthier et al. 2009; Hollister
2009; Studer et al. 2011), some of which are presented in a special issue of Sociological
Methods & Research (Brzinsky-Fay and Kohler 2010), as well as an increasing number of
sequence analysis applications within the social sciences (Huang et al. 2007; Kogan 2007;
Shoval and Isaacson 2007; Quintini and Manfredi 2009; Gauthier et al. 2010; Salmela-Aro et al.
2011; Simonson et al. 2011).
Sequences are ordered listings of elements (MacIndoe and Abbott 2004; Brzinsky-Fay et al.
2006), in which an element can be a certain status (e.g., labour market status) or a physical
object (e.g., base pair of DNA) or an event (e.g., dance step). These elements are tied to either
fixed points of time (e.g., status in a certain month), or to fixed positions (e.g., protein at
position 12). Their specific order is of crucial importance and cannot be changed. Individual
school-to-work transition sequences are composed of different labour market statuses, such as
employment, unemployment, education, apprenticeship, and inactivity. The resulting
complexity of empirical transitions composed from these five statuses is provided in Figure 1,
which shows the monthly labour market entry sequences for the first 60 months after leaving
school in three countries (France, Spain and Germany). Two things appear quite clear: There
are obvious differences between countries; and the variance within countries is remarkably
high. In this situation, the application of event history analysis would require the definition of
the event of interest, which necessarily must remain arbitrary and disregards meaningful
information.
The main task of sequence analysis is to reduce complexity by comparing, sorting and grouping
these sequences. The result is a typology of transition types that apart from giving interesting
insights could be used as a well-grounded argument for choosing certain events of interest.3
The comparison of sequences includes visual inspection and calculation of simple descriptive
indicators for specific sequence characteristics such as the length of the sequence, the
number of episode changes within a sequence, or the number of different elements in the
sequence. Elzinga (2003) proposed measures for turbulence and the frequency of certain
subsequences to qualify sequence characteristics. The most frequently used technique for
comparing sequences is optimal matching (OM). OM defines the distance between two
sequences as the number of operations substitution and deletion/insertion it takes to
transform one sequence into the other. The resulting distance measure is called the
Levenshtein distance (Levenshtein 1966), and the values are computationally achieved by the
NeedlemanWunsch algorithm (Needleman and Wunsch 1970). After having established the
similarity between each pair of sequences, the resulting distance matrix can be used as input
for a cluster analysis or multidimensional scaling. The result is a classification that consists of a
manageable number of sequence types, as shown in Figure 2. These eight ideal types of
transitions appear in all countries, but their proportions vary. They can be used to create
meaningful subgroups, which can be analysed separately. This kind of longitudinal exploration
also provides event history models with good descriptive information, from which this
powerful hypothesis-testing method could start, while avoiding unfounded pre-decisions.
The application of OM requires the ex-ante definition of costs for the basic operations used to
transform one sequence into the other. The supposed subjectivity or arbitrariness of this
definition is one of the main objections against the method (Levine 2000; Wu 2000). In fact, in
many cases there is no theoretical basis for choosing the numerical values of the costs of
substitution, insertion, or deletion. However, in some cases the computation of these costs via
transition frequencies might be a solution, while heuristic testing of automated cost
calculation, as recently proposed by Gauthier et al. (2009), might be another. There is no true
cost structure, but in order to find patterns in the data, different cost structures can be tested
iteratively, while the requirements of validity and reliability are considered (Bernard 2000).
Another drawback of OM is the one-dimensionality of the elements that the sequences were
composed of. The analysis of parallel or multiple sequences (e.g., employment career and
family formation simultaneously) has been a serious handicap of the method until lately, a
couple of attempts tackled this challenge, mainly in the field of life-course research (Aassve et
al. 2007; Piccarreta and Billari 2007; Pollock 2007). The application of sequence analysis
provides important potential as Raffe (2009: 111) points out: The recent quantitative work
comparing transition sequences [...] represents an important line of development.
Sequence analysis cannot replace event history models, because it is not a causal method and
sequence analytical methods contribution to answer the question why? is limited by nature.
It has a holistic perspective, which reflects the one-sided view on the structural character of
life courses as a whole, while at the same time ignoring individual agency, which is better
reflected by event history analysis. For these reasons, sequence analysis should be seen as a
complement to event history modelling. The most promising future methodological and
theoretical development in transition analysis is the integration of both approaches. For
instance, the explorative potential of sequence analysis could serve as a basis for the definition
of meaningful events for a confirmatory event history model. Another possibility is that the
ideal types that result from the grouping process also allow for group-wise analyses with
different events of interest. This allows taking into account long transition processes and
increasing the validity of confirmatory results.

5. Conclusion

This article aims at overcoming a serious limitation in contemporary school-to-work transitions

research, which has its origins in an unclear definition of the central concepts transition and
trajectory and in the limitation of school-to-work transition research to hypothesis-testing
methods, in the first line event history analysis.
The prolongation of educational processes and the increasing flexibilisation of labour markets
in western societies have led to higher complexity of what is used to be referred to as school-
to-work transition. From a life-course research perspective, this period must be
conceptualised as a trajectory rather than a transition. The concentration on single status
changes has shown that this complexity exists, but it must fail when it comes to a qualitative
exploration of how these complexity looks like in detail.
It is shown in this article that all relevant individual-level indicators, which are used in school-
to-work transition research, suffer from this time-point problem. Even if durations are
focused on, analyses are limited to single episodes while disregarding other episodes, their
sequence, or their composition. In order to avoid arbitrariness and to detect qualitative
changes and variations in individual school-to-work transition processes, explorative methods
must find their way into the methodological toolbox of researchers. One of these methods,
which provide a good and established instrument for the exploration of longitudinal
information, is sequence analysis. Within the field of school-to-work transition, some studies
already exist, but they must be integrated with each other in a meaningful manner. A coming
new wave of school-to-work transition research can yield promising new insights into these
processes of increasing complexity, if exploration and hypothesis testing are carried out hand
in hand.