Open Access

Scarring effects of early unemployment among young workers with vocational credentials in Switzerland

Empirical Research in Vocational Education and Training20146:12

https://doi.org/10.1186/s40461-014-0012-2

Received: 10 July 2014

Accepted: 5 November 2014

Published: 16 December 2014

Abstract

Background

Using a unique, longitudinal survey that follows school-to-work transitions of pupils who participated in PISA 2000, this paper investigates adverse consequences, so-called scarring effects, of early unemployment among young adults who acquired vocational credentials in Switzerland.

Methods

As social, individual and contextual factors influence both early unemployment and later employment outcomes, taking into account endogeneity is of utmost importance when investigating scarring effects. In this regard we make use of nearest-neighbour propensity score matching and set up statistical control groups.

Results

Our results suggest that young adults who hold vocational credentials are more likely to be neither in employment nor in education, and to earn less and be more dissatisfied with their career progress later in work life than they would be, had they not experienced early unemployment.

Conclusions

We conclude that unemployment scarring also affects young adults with vocational credentials in a liberal labour market setting that otherwise allows for smooth school-to-work transitions. This finding runs counter to expectations that standardised vocational degrees, a liberal and flexible labour market structure, and predominantly short unemployment spells protect young skilled workers from scarring in case they happen to experience early career instability.

Keywords

Scarring effects Early unemployment School-to-work transition Vocational education and training (VET)

Background

The 2008–09 recession has hit youth hard in terms of their employment prospects across the OECD (OECD [2010]). With regard to its low youth unemployment rates, Switzerland is often recognized as a country that performs well in international comparison (OECD [2010], [2013b]; Hoeckel et al. [2009]). Institutional settings such as comparatively weak employment protection legislations (EPL) (see OECD [2013a]) that allow school leavers to more easily set foot within the labour market (see Breen [2005]) as well as the strong vocational orientation of the upper secondary education system both have a share in this. Vocational education and training (VET) is the dominant form of upper secondary education in Switzerland and equips the majority of youth with vocation specific certificates that foster smooth transitions to the labour market. About two thirds of youth pursue a vocational education at upper secondary level in Switzerland while only about one fourth takes up a general education (SERI [2014]; Stalder and Nägele [2011]). In international comparison, Switzerland stands out in terms of the share of school leavers who pursue a dual-vocational education (Hoeckel et al. [2009]: 10–11), which is an apprenticeship that combines school and workplace-based training. Most apprenticeships take three to four years to complete and are awarded with a federal certificate of vocational education and training that qualifies one for taking up work in the occupation trained in. About a quarter of the students who pursue vocational education at upper secondary level further obtain a federal vocational baccalaureate, which is an optional qualification that allows them to enrol in tertiary vocational education such as at the universities of applied sciences (Fuentes [2011]; Hoeckel et al. [2009]; Stalder and Nägele [2011]). Overall, the Swiss VET system can be described as highly standardised as it equips the young with vocational training and certificates that meet similar standards nationwide (see Müller and Shavit [1998]: 6–7). These vocational certificates send clear signals to employers about the occupation-specific skills and suitability of the young entrants for particular jobs (Breen [2005]), further fostering smooth transitions.

Even though skilled workers, particularly those who have undergone vocational education and training, fare comparatively well when transiting to the labour market, they are nevertheless increasingly facing problems at labour market entry in Switzerland (Salvisberg and Sacchi [2013], [2014]). The fact that early unemployment may not only be a temporary blemish but can evolve into permanent scars (Ellwood [1982]) is theoretically and empirically well established. However, while negative long-term consequences of unemployment experiences, so-called scarring effects, have been the focus of research conducted in a lot of industrialised and industrialising countries (see e.g., Arulampalam et al. [2000]; Mroz and Savage [2001]; Nordström-Skans [2004]; Gregg and Tominey [2005]; Luijkx and Wolbers [2009]; Vandenberghe [2010]; Nilsen and Holm Reiso [2011], Cruces et al. [2012]), little is known about unemployment scarring concerning those who pursued vocational education and training. To help towards closing this gap, this paper looks at scarring effects of early unemployment among young workers who have completed vocational education and training in Switzerland. As institutional settings such as weak employment protection legislations (OECD [2013a]) and high standardisation of VET presumably do not provide strong incentives for employers to base their hiring and wage-setting decisions on the young applicant’s short work histories and transition processes, we expect that in the Swiss context VET graduates will not be substantially affected by unemployment scarring. This assumption finds further support in that unemployment spells of young skilled workers are generally of short duration in Switzerland (see Sacchi and Salvisberg [2012]). Whether these potentially advantageous institutional settings indeed manage to offset unemployment scarring will be put to the test within this paper.

Besides focusing on a group of skilled young workers that has so far been assumed to be mainly sheltered from various labour market risks, we go beyond previous research in that we not only investigate unemployment scars with respect to objective dimensions of labour market outcomes but also take individual’s assessments of their career progress into account. Young adults who are to some extent worse off in the mid-term due to the exposure to early unemployment may still be satisfied with their careers, if the experience of early unemployment was a prolonged job search or welcome career-adjustment such that they are happy with the job-matches found and the way their career unfolded. For this reason we complement our analysis of unemployment scarring on objective employment outcomes with the analysis of how the young themselves perceive their career progress. This allows for a more comprehensive understanding of the impact of early unemployment in terms of both objective and subjective notions. Against this backdrop, we investigate the following research questions:

1.) Are young skilled adults who experienced unemployment in early work life disadvantaged in terms of their subsequent employment outcomes (such as labour market integration and wages) compared to a similar group of young skilled adults who did not experience early unemployment?

2.) Do young skilled workers who experienced unemployment in early work life perceive their career progress as less satisfying compared to a similar group of young skilled workers who did not experience early unemployment?

In order to investigate these research questions we first briefly review theoretical explanations and expectations as well as we report the current state of research concerning unemployment scarring. Following this, we introduce the analytical strategy employed, which is propensity score matching, and move on to describe the data, model and variables used. We then present our findings and add some concluding remarks. Finally we outline the need for future research and discuss some limitations of our analysis.

Theoretical considerations and empirical findings

What drives unemployment scarring?

Different competing explanations have been proposed in order to frame unemployment scarring, such as economic assumptions about returns on accumulation of human capital (Becker [1964]), human capital depreciation (see e.g. Pissarides [1992]; Edin and Gustavsson [2008]) or statistical discrimination (Aigner and Cain [1977]) and signalling (Spence [1973]). Becker’s ([1964]) economic approach to returns on human capital assumes that specific on-the-job training is an investment in human capital and thus goes together with higher wages in later career. Those who experience spells of early unemployment can be seen as missing out on the possibility of acquiring job specific training and thus have a lower market value when hired compared to their cohort members (see Becker [1964]). Over and beyond differing job prospects due to missed accumulation of human capital, it is further assumed that human capital depreciates during periods of economic inactivity. Loss of work specific skills due to unemployment periods is seen as leading to lower productivity and thus to worse occupational prospects when returning to the labour market (see Edin and Gustavsson [2008]; Mooi-Reci and Ganzeboom [2012]). Yet, as skilled labour market entrants in Switzerland mainly experience short spells of unemployment (see Sacchi and Salvisberg [2012]), human capital depreciation is presumably not very important as missed work experience may not be extensive. The duration of unemployment is further considered to drive unemployment scarring from an application-behavioural viewpoint. Following wage search perspectives (Mortensen [1986]), unemployed are thought to lower their reservation wages with search tenure. As time spent in unemployment passes, individuals become more prone to apply for jobs and to accept job offers that do not match their initial expectations. Thus not only the incidence of unemployment but also its duration may be seen as mechanisms driving unemployment scarring.

Statistical discrimination (Aigner and Cain [1977]) and signalling theory (Spence [1973]) draw on the assumption that hiring is an uncertain investment since the worker’s productivity is unknown to the employer. Therefore the employer has to rely on observable characteristics, so-called signals (Spence [1973]), which serve as a proxy for worker’s unknown productivity. These signals then determine to a great deal job offers and the job quality attached to these (see Spence [1973]). In line with this theoretical argument, employers may not attribute early unemployment to structural problems and shortages in the demand for newcomers, but rather take prior unemployment as a “signal” about these worker’s lower capabilities, productivity or motivation. Following Aigner and Cain ([1977]) employer’s assumptions about differential reliability of credentials to proxy true productivity across different groups of workers in combination with risk-aversion may be thought to result in statistical discrimination. Thus, irrespective of their individual ability and productivity, young adults who experience unemployment at an early stage of career formation can be assumed to have worse employment prospects since they are less attractive to hire. Particularly in the absence of reliable information on ability as well as if costs to overcome imperfect ability assessments are high, firms may be assumed to take diverse individual characteristics as proxies for ability. On the other hand, where standardised and reliable credentials of individual skills are available, incentives to use and gather additional information may be weak (see Müller and Wolter [2014]).

In addition, experiences of unemployment may also leave a psychological scar. Following Erikson ([1959]), a healthy psychological ego development during the transition from adolescence to adulthood comprises identity formation that encompasses the formation of an occupational identity (see Erikson [1959]: 94–100). In this perspective, early unemployment may be regarded as conflicting with the formation of an occupational identity, harming perceptions of self-worth (see Goldsmith et al. [1997]). Additionally, social comparison processes adversely impact self-esteem and foster depression among unemployed (Sheeran et al. [1995]). Drawing on Seligman ([1975]), the exposure to events perceived as uncontrollable, such as unemployment, bring about a feeling of helplessness, which diminishes psychological well-being and self-esteem and further undermines the motivation to control outcomes (Seligman [1975]; Goldsmith et al. [1997]). Such poor psychological adjustments are thus in turn thought to affect job search strategies and success so that a vicious circle may ensue (Furnham [1985]).

In contrast to these theoretical explanations concerning unemployment scarring, there exist assumptions about potentially positive effects of early unemployment. Following e.g. Kahn and Low ([1982]) it is argued that off-the-job searches (searching for jobs while economically inactive), may be more efficient, resulting in better job matches that go together with higher re-employment wages and good career prospects. According to this theoretical perspective, early episodes of economic inactivity and career-adjustment may be regarded as prolonged matching processes leading to better job matches in the longer run. In this sense, periods of early economic inactivity are not necessarily damaging with regard to future occupational outcomes.

Focusing on the proposed mechanisms driving unemployment scarring while taking into account the institutional setting of Switzerland it is ambiguous whether or not these mechanisms really impact the careers of VET graduates who experienced bumpy transitions. Firstly, in a country with high standardisation of VET, where high vocational specificity and transparency of skills allows employers to rely on credentials of occupation-specific skills (see Müller and Shavit [1998]: 6–7), one may argue that employers will not further base their hiring and wage-setting decisions on the young applicants’ short and uninformative work histories. Where vocational training systems provide tight linkages between education and occupational status, post-unemployment job matches should be of higher quality preventing severe scarring (see Dieckhoff [2011]: 237). Furthermore, the Swiss labour market is rather liberal, with weak employment protection of the workforce (see OECD [2013a]). Low employment protection allows labour market outsiders such as entrants and unemployed to more easily (re-)enter the labour market (see van der Velden and Wolbers [2003]; Breen [2005]; Bukodi et al. [2006]) as there are no high dismissal costs and work contracts can be terminated easily in case employers are not satisfied with their newly hired employees. In line with this, investment in screening based on prior work histories as well as stigmatisation of those formerly unemployed may be assumed to be not as pronounced in liberal and flexible labour markets. In a similar vein it has been argued that prolonged job searches may indeed be thought to increase job quality in more liberal labour markets compared to rigid labour markets with strict employment protection legislation (EPL) (see Schmelzer [2012]). Therefore prolonged job search and career-adjustment in early work life may not negatively impact subsequent employment outcomes of the young in Switzerland. The assumption that there are no distinct scarring effects further finds support in the fact that, for many skilled young entrants in Switzerland, unemployment spells are short in duration (see Sacchi and Salvisberg [2012]). Thus foregone work experience due to early unemployment among young skilled entrants is generally low, and no profound psychological scars may be expected in this regard. Against this backdrop, we expect no or only negligible scarring effects regarding vocationally skilled workers in countries with highly standardised VET and liberal labour market structures such as Switzerland.

If this hypothesis were to be rejected then this would suggest that even within an advantageous institutional context, early unemployment inflicts scars. Which would mean that either firms already consider short unemployment spells among VET graduates to be negative signals or that the experience of short unemployment may be enough to exert a psychological scar and to alter the application behaviour and job search success of the young.

Scarring effects: current state of research

Controlling for selection bias, evidence for several industrialised and industrialising countries suggests that unemployment at both labour market entry as well as in later career is in general scarring. For Germany, Britain, New Zealand, the Netherlands, Sweden, Norway, Finland and the U.S. as well as the developing countries Argentina and Brazil, past unemployment and early economic inactivity have been found to be positively related to the risk of future unemployment, suggesting significant state dependence (see Niedergesäss [2012]; Arulampalam et al. [2000]; Maloney [2004]; Luijkx and Wolbers [2009]; Nordström Skans [2004]; Nilsen and Holm Reiso [2011]; Hämäläinen [2003]; Mroz and Savage [2001]; Cruces et al. [2012]). Focusing on VET graduates in Germany, Schmillen and Umkehrer ([2013]) find that the total length of unemployment experienced in early career significantly predicts unemployment tenure later in career. Scarring further not only emerges in form of state dependence but also with regard to future occupational outcomes such as earnings. Evidence for the U.S. points towards positive effects of job stability in early career on adult wages (Neumark [1998]) and spells of unemployment in early work life have been found to go together with reduced future earnings (Mroz and Savage [2001]). Results for several western European countries, such as e.g. Britain, Sweden and Germany, as well as for Latin American countries, similarly suggest that unemployment experiences have a damaging effect on post-unemployment earnings (see Arulampalam [2000]; Gregg and Tominey [2005]; Nordström Skans [2004]; Schmelzer [2012] Gangl [2006]; Cruces et al. [2012]). Conducting a field experiment, Eriksson and Rooth ([2011]) find that Swedish employers do not take past unemployment of young applicants as a negative signal when deciding upon invitations to job interviews. They find, however, that employers sort young applicants based on their prior work experience. In addition to monetary scarring effects, there also exists evidence suggesting a psychological impact of past unemployment on psychological distress and life satisfaction in later life (see Daly and Delaney [2013]; Clark et al. [2001]; Young [2012]; Lucas et al. [2004]). On the one hand, individual resources such as self-efficacy, self-esteem and a positive attitude towards life shape transition processes (see e.g., Rüfenacht and Neuenschwander [2013]; Neuenschwander [2012]; Neuenschwander and Kracke [2011]; Gerber-Schenk et al. [2010]). On the other hand, the very same individual resources may also be adversely influenced by unemployment experiences, further manifesting themselves in worse occupational outcomes (Furnham [1985]).

With respect to Switzerland, there exist several studies pointing towards increased state dependence amongst the unemployed (see e.g., AMOSA [2010]; Sheldon [1999]). However, these studies do not fully take into account the potential confounders among the unemployed. Djurdjevic ([2005]) identified two types of unemployed for the Swiss context. The “higher risk” unemployed include less-skilled, foreign and female workers, while young, skilled, Swiss, male workers constitute the “lower risk” unemployed. The “lower risk” unemployed have better employment prospects and leave unemployment more quickly for employment. They further succeed in moving to better paid jobs after some initial earning losses (Djurdjevic [2005]: 57–58).

Even though a lot of studies conducted in different institutional settings point into the direction of unemployment scarring in early and later career, scars may be heterogeneous regarding different groups of workers (see e.g., Arulampalam [2000]; Gangl [2006]; Burgess et al. [2003]; Schmelzer [2011]; Schmillen and Umkehrer [2013]). So far, only little is known about the existence of unemployment scarring concerning young workers with vocational credentials. While Schmillen and Umkehrer ([2013]) shed some light on unemployment scarring affecting VET graduates in the rigid labour market of Germany, it still remains an open question whether a liberal market structure combined with highly standardised VET not only allows for smooth transitions but may also protect the young from scarring. This question will be investigated in the following.

Methods

Essentially, the idea of this paper is to identify the causal effects of early unemployment on subsequent employment outcomes among VET graduates. The problem we face is that both early unemployment as well as subsequent employment outcomes may depend on similar social, individual and contextual characteristics that bias the estimation of causal effects. Thus it is crucial to properly control for endogeneity when assessing causal effects of early unemployment. In this regard we make use of propensity score matching, which allows for the estimation of causal effects based on non-experimental survey data (Guo and Fraser [2010]; Rosenbaum and Rubin [1983]). The question we investigate using this technique is whether or not young people who pursued a vocational education would have had more favourable career outcomes had they not experienced early unemployment. However, we do not know what would have happened to these young entrants had they not experienced early unemployment. Addressing this problem, propensity score matching allows us to set up a statistical control group of young workers who experienced smooth transitions but who are comparable in relevant characteristics to those who experienced early unemployment. The employment outcomes of the control group then serve as a proxy for the potential career outcomes of those who experienced early unemployment, had they not experienced it.

Overall, propensity score matching may be thought of as a two stage procedure. First, propensity scores, that is conditional probabilities of experiencing early unemployment, are estimated based on a set of potentially confounding variables. Second, based on these propensity scores, a comparable control sample of individuals who have not experienced early unemployment but who are similar with regard to their propensity of experiencing early unemployment, and with this their relevant characteristics, is created (for a more thorough discussion of this equalisation see Rosenbaum and Rubin [1983]; Heckman et al. [1997]). The estimate we focus on to assess scarring is the average treatment effect on the treated (ATT). The ATT can be thought of as the estimated (hypothetical) difference between the employment outcome of respondents who experienced early unemployment (the treated) and their potential outcome had they not experienced unemployment. This unobserved and thus hypothetical outcome is also known as the counterfactual outcome and is estimated based on the matched control group. Two assumptions are highly important for the estimation of causal effects within the context of propensity score matching. First, the stable unit treatment value assumption (SUTVA) assumes that there exists no spill-over between treated and controls such that e.g. processes within the group of controls affect the outcome among treated and vice versa. If this is the case then causal inference may not be possible (see Gangl and di Prete [2004]). With regard to the conditional independence assumption (CIA) it is assumed that after controlling for confounding covariates based upon which propensity scores are estimated, treatment-assignment is independent of potential outcomes so that a counterfactual outcome can be estimated without bias based on the matched controls. In the literature on propensity score matching, this assumption is also referred to as “ignorability assumption” (see Gangl and di Prete [2004]; Rosenbaum and Rubin [1983]; Heckman et al. [1997]). The rich data at hand allows us to control for a wide variety of potentially confounding variables in order to make the CIA assumption plausible. Yet, irrespective of the fact that we control for a diverse set of potentially relevant characteristics, unobserved factors that are not related to the variables based on which we balance the sample always pose a potential risk when assessing causal effects based on non-experimental survey data.

Defining treatment: early unemployment

We define all young adults who are economically inactive as unemployed if they are either officially registered as unemployed or if they are actively in search of a job. By this definition, inactive young adults in search of an upper secondary education are not regarded as unemployed. This definition of unemployment corresponds to the ILO definition (see ILO [1982]).

Data

We make use of data provided by TREE (Transition from Education to Employment), which is the first longitudinal survey aimed at surveying patterns of post-compulsory educational pathways and transitions to the labour market of young people in Switzerland. TREE is an ongoing prospective panel survey based on a sample of approximately 6000 youths who participated in the PISA survey (Programme for International Student Assessment) in the year 2000 and who left Swiss compulsory schools in the same year. The sample was followed in annual surveys from 2001–2007, while an additional survey took place in 2010 (Stalder et al. [2011]; TREE [2013]). We restrict our analysis to a subsample of labour market entrants who have completed a first vocational education by the year of 2007, which encompasses the vast majority of those who pursued a vocational education after compulsory schooling in the cohort investigated. Concerning vocational education, both standard three to four year apprenticeshipsa and vocational baccalaureates are included. We focus on early unemployment experiences after completion of upper secondary vocational education that occur during the main transition period for the cohort investigated – which is between the years of 2003–2007 when the young are about 19–23 years old (see Bertschy et al. [2008]). We then assess whether or not unemployment within the first (one to four) years after graduation from VET between the ages of 19–23 adversely affects employment outcomes of the young in 2010 when they are 26 years oldb. Young adults who report no unemployment experiences for the time observed but for whom there exist no full records on their transition period as well as a few unclear casesc are excluded from the analysis as we do not know whether or not they really have experienced smooth transitions. In order to increase the homogeneity within treated and non-treated groups (see Imbens [2003]) we further excluded 18 cases who experienced unemployment when trying to set foot in the labour market before finishing a first vocational educationd.

The estimation of propensity scores is based on an extended sample including all youth and young adults who have completed a first upper secondary vocational education by the year 2007, regardless of whether or not they respond to the TREE survey in 2010. The final sample we derive the propensity scores from consists of 1588 young persons (777 women and 811 men) with vocational credentials, of whom 344 (152 women and 192 men) have experienced unemployment in early career. For matching and the investigation of scarring effects we restrict our sample to those young adults who respond to the TREE survey in 2010, when they are about 26 years of agee.

The main sample based upon which we investigate whether or not previously unemployed young adults are more likely to be neither in employment nor in education (NEET) consists of 1269 skilled young persons. With regard to these respondents, 244 young persons have experienced a spell of unemployment in early career at the age of 19–23 years. Overall, of those formerly unemployed 31 persons are NEET in 2010 while of those who experienced smooth transitions 35 are NEET in 2010. With regard to the analysis on wage scarring and subjective career assessment, we further restrict our analysis to skilled young adults who have succeeded in finding a job in the labour market and who respond to the complementary employment questionnaire of the TREE survey in 2010. The main sample based on which we investigate wage scarring and persisting differences in career assessment consists of 773 skilled young workers. Within this group of respondents, 119 young persons have experienced early unemployment.

Variables used

A necessity for yielding unbiased estimates when applying propensity score matching is a thoroughly considered model based on which the propensity scores are estimated (see Shadish and Steiner [2010]; Bryson et al. [2002]; Smith and Todd [2005])f. Since the TREE survey has been designed particularly with respect to the investigation of school-to-work transitions, it offers a rich source of data to estimate the propensity scores and balance samples. The following variables are included in the estimation of propensity scores based on which we then match a control group: We include social categories such as gender (coding is 1 = male and 0 = female) and migrant status modelled in terms of the language spoken at home (coding is 1 = national language vs. 0 = foreign language) as well as the country of birth (coding is 1 = Switzerland, Luxembourg vs. 0 = other country). As educational qualifications further play an important role in terms of labour market integration and future job prospects we include measures of competencies (PISA literacy scores and mark in math) and the type of track at lower secondary education (coding is 1 = basic requirements, 2 = extended requirements/pre-gym and 3 = no formal tracking). As vocational education and training should not be seen as a uniform level track in Switzerland (Stalder [2011]), the level of vocational education and training is furthermore taken into account (coding is 1 = low-skill VET, 2 = medium-skill VET, 3 = high-skill VET and 4 = vocational maturity). The coding of the level of demand of the vocational education is based on a coding scheme proposed by Stalder ([2011]). As a successful transition and career advancement are also influenced by family background (see e.g., Buchmann [2011]; Neuenschwander and Kracke [2011]) we include socio-economic status (Isei), familial possessions of poetry literature (coding is 1 = poetry at home vs. 0 = no poetry at home), a PISA index for social communication within the family (Soccom), family structure (coding is 0 = nuclear family, 1 = single parent and 2 = mixed/other) and the number of siblings (coding is 0 = no siblings, 1 = one sibling, 2 = two siblings and 3 = three or more siblings). Important individual resources that determine labour market integration of the young (see Rüfenacht and Neuenschwander [2013]; Neuenschwander [2012]; Neuenschwander and Kracke [2011]; Gerber-Schenk et al. [2010]) are also taken into account by the inclusion of measures for self-efficacy, persistency and a positive attitude towards life (measured prior to labour market entry in order to avoid bias due to reversed causality). The dummy-variables are coded 1 if the individual resource is available vs. 0 if it is not. Attendance and punctuality are important behavioural resources with regard to a successful entry into an apprenticeship (see Hupka et al. [2006]) and thus they may also be regarded as indicators for transitional outcomes and success in later work life. Therefore we include measures for behaviour at lower secondary level such as skipping classes (coding is 1 = yes vs. 0 = no) and finishing homework on time (coding is 1 = yes vs. 0 = no) in the estimation of propensity scores. As the demand for newcomers varies according to the field trained in (see Sacchi and Salvisberg [2012]), we also include measures for the field of vocational education and training. The classification of educational fields is based on an international ISCED-coding scheme proposed by Andersson and Olsson ([1999]). Overall, different traditions concerning the orientation of educational systems across linguistic regions in Switzerland exist (Seibert et al. [2009]; Meyer [2009]). Thus, we further include the region of residence (coding is 1 = German speaking part, 2 = French speaking part and 3 = Italian speaking part) as well as average cantonal youth unemployment rates between 2003–2007, which is for the main entry period we look at.

Outcome variables, based on which we assess scarring are monthly full-time equivalent gross earnings in 2010, a dichotomous variable coded 1 if a person is neither in employment nor in education or training (NEET) in 2010 and coded 0 otherwise, as well as a binary variable indicating whether or not a person is dissatisfied with the progress made in career towards meeting his/her overall career goals. The variable on career dissatisfaction is coded 1 if young adults report to be not satisfied or only partially satisfied with their career progress and it is coded 0 if young adults report to be satisfied with their career progress. A more detailed report of the variables included in the analysis can be found in the Appendix.

Analysis

In a first step we estimate propensity scores based on a logistic regression, taking into account the complex survey design of the PISA/TREE sample regarding variance estimationg. We further apply customized weightsh, which we derived based on sampling weights provided by TREE (see Sacchi [2011]). Besides the estimation of propensity scores, the logistic regression allows for insight into the determinants of early unemployment among VET graduates in Switzerland. Following the estimation of propensity scores, we move on to matching and the estimation of average treatment effects on the treated (ATT). For matching and for the estimation of average treatment effects on the treated (ATTs), nearest-neighbour matching of at most 5 nearest-neighbours (with replacement) within a caliper of 0.05 is applied. We further impose the criteria of a common support. This means we set up a statistical control group by matching up to five young adults who experienced smooth transitions, who are closest with respect to their propensity scores, to each respondent who experienced unemployment (treated). However, we do not match anyone whose propensity score falls outside a common support and we further restrict the number of potential matches for a respondent to a caliper, meaning a limited range of propensity scores close to the treated respondent, in order to avoid matches of poor quality (see Bryson et al. [2002]). The same control individual may be matched multiple times (for more details see e.g., Guo and Fraser [2010]).

In order to assess the quality of sample balance between treated and matched controls, we follow recommendations by Stuart ([2010]) and test for differences in mean and variances of propensity scores across treated and control groups. We find no significant mean-differences and the ratio of variances of propensity scores (with a value of 1.3) further indicates good sample balance.

Regarding significance testing of average treatment effects on the treated (ATT), commonly applied bootstrap procedures have proven to be not valid (see Abadie and Imbens [2008]). Therefore, we apply a bootstrap procedure proposed by de Luna et al. ([2010]). We bootstrap estimated individual causal effects (EICE) over circular block-differences (under the less rigid assumption of heterogeneous causal effects) (see de Luna et al. [2010]). In order to obtain estimates that can be generalised to the population of young skilled adults, we apply a sampling weight (wt8_kal) for the TREE sweep in 2010 when estimating ATTs and their sampling distribution (see Du Goff et al. [2014]). This sampling weight corrects for both the disproportionality due to sampling-design and panel attrition. We further report results when applying a weight (basewt), which only corrects for the disproportionality of the sample but does not adjust for panel attrition. The analysis is run in the software framework of STATA (12) using the programme package psmatch2 (see Leuven and Sianesi [2003]).

Results and discussion

In Table 1 we present findings from logistic regression analysis, modelling the risk of early unemployment among skilled young adults. We report average marginal effects (AME), which are, in contrast to odds-ratios, intuitively interpretable and robust towards unobserved heterogeneity due to scaling issues that arise in non-linear regression analysis (see Best and Wolf [2012]; Mood [2010]). Based on this model we also estimate propensity scores for the following analysis on scarring effects.
Table 1

Logistic regression

Early unemployment

AME

SE

N = 1351

  

Social categories

  

Gender (Male)

0.11**

0.04

Migration background (foreign language)

0.22**

0.08

Country of birth (abroad)

0.05

0.06

Social and familial background

  

Socio-economic background (Isei)

0.00

0.00

Poetry at home

−0.01

0.03

Social communication with parents

0.00

0.02

Family structure (single parent)a

0.18*

0.08

Family structure (mixed/other)a

0.21**

0.07

Number of siblings (1)b

−0.06

0.07

Number of siblings (2)b

−0.05

0.08

Number of siblings (3)b

−0.13(*)

0.07

Education

  

Lower secondary education (extended requirements)c

0.01

0.04

Lower secondary education (no formal tracking)c

−0.01

0.07

Reading Literacy (Wleread)

−0.00

0.00

Math mark (at average)d

0.08(*)

0.05

Math mark (below average)d

−0.06

0.04

Level of VET (medium)e

−0.09

0.05

Level of VET (high)e

−0.11*

0.05

Level of VET (occupational baccalaureate)e

−0.11(*)

0.06

Field of printing and design (2)f

0.12

0.11

Field of business and sale (3)f

0.22***

0.05

Field of applied computer science (4)f

0.19*

0.09

Field of agriculture and gardening (6)f

−0.04

0.06

Field of medical assistance (7)f

−0.02

0.06

Field of gastronomy and hairdressing (8)f

0.05

0.05

Field of other (9)f

−0.08

0.05

Individual resources

  

Self-efficacy

−0.11

0.07

Positive life attitude

−0.15*

0.06

Persistency

−0.01

0.06

Skipping classes

−0.05

0.05

Finishing homework on time

−0.08(*)

0.04

Demographics

  

Region of residence (French)g

0.13*

0.06

Region of residence (Italian)g

0.12

0.09

Cantonal unemployment rate

0.01

0.02

(*) <0.1 *< 0.05 **< 0.01.

a)Ref. category: nuclear family, b)Ref. category: no siblings, c)Ref. category: basic requirements, d)Ref. category: above average, e)Ref. category: Low skill VET, f)Ref. category: field of mechanics and electronics (5), g)Ref. category: German speaking part.

Results on the determinants of early unemployment among VET graduates reveal that young men are on average 11% more likely to experience early unemployment controlling for the other confounding covariates included in the model. This finding may reflect differing job opportunities for male and female VET graduates according to their often highly gender-specific occupational credentials (Kriesi et al. [2010]). Our findings further suggest that young persons with vocational credentials who speak a foreign language at home (migrant background) are on average 22% more likely to be exposed to unemployment in early work life. We further find that young adults who were raised in single parent or mixed/other households are about 18-21% more likely to experience early unemployment compared to young adults who grew up in nuclear families. Young adults who have three or more siblings are 13% less likely to experience early unemployment compared to young persons who have no siblings (weakly statistically significant at p <0.1). Looking at the impact of educational credentials and school competencies our results suggest that those who achieved average marks in math at lower secondary level are by 8% somewhat more likely to experience unemployment in early career compared to skilled young workers who achieved above average marks (significant at p < 0.1). With regard to differing levels of demand across vocational education and training tracks we find that young adults with high levels of upper secondary vocational education (VET) are 11% less at risk of experiencing unemployment compared to young entrants who hold a low-skill VET diploma. This finding also holds for young adults who absolved an occupational baccalaureate (significant at p < 0.1). In addition, labour demand (for newcomers) varies considerably across occupational segments in Switzerland (Sacchi and Salvisberg [2012]). This variation in demand across occupational segments is also reflected in our findings, which show that not only the level of vocational education but also the field trained in matters in terms of smooth transitions to work. Young skilled adults who pursued an apprenticeship in the field of (3) business and sale and those who pursued their training in the field of (4) [applied] computer science are 22% resp. 19% more likely to experience a bumpy labour market entry period compared to those who did their training in the field of (5) mechanics and electronics. Our results further support that individual resources play an important role in determining the risk of early unemployment. We find that young graduates who have a positive attitude towards life (measured prior to the labour market entry period) are 15% less likely to experience unemployment in early career. One may note that if we do not control for a positive attitude towards life then self-efficacy becomes significant. Furthermore young entrants who have been sincere and reliable students in terms of finishing their homework on time at lower secondary level are on average 8% less likely to experience unemployment during their career formation (weak statistical significance at p < 0.1). In addition to social and individual resources, the region of residence also matters. Young entrants from the French speaking part are on average 13% more likely to experience early unemployment compared to young labour market entrants from the German speaking part. One may note that not including the region of living in the logistic regression results in a statistically significant finding for average cantonal youth unemployment rates, indicating that these variables are somewhat confounded.

Overall, these findings suggest that social, structural as well as individual resources simultaneously shape transition processes of skilled young adults. Therefore, accounting for heterogeneity in the exposure to unemployment when investigating scarring effects seems to be a necessity in order to yield unbiased results.

Unemployment scarring

Looking at scarring effects of early unemployment based on balanced samples from propensity score matching, we find that young skilled adults are by 9% somewhat more likely to be neither in employment nor in education or training (NEET) at the age of 26 years in 2010, compared to a comparable group of young workers who experienced smooth transitions. Yet, using a weight (wt8_kal) correcting for panel attrition the result we find is not statistically significant. One may note, however, that we find a significant effect if we applied a weight (basewt) correcting only for the disproportionality of the sample. All in all, we interpret these results as evidence pointing towards a scarring effect on later labour market attachment. Looking at young persons who are engaged in the labour market in 2010 our results further suggest that young skilled adults would earn higher wages at the age of 26 years, had they not experienced unemployment at the early stage of career formation. On average young adults who experienced some unemployment in early work life earn about 360 CHF (monthly gross wage) less compared to a comparable group of young skilled workers who did not experience early unemployment. This wage scar amounts to 7% of their average monthly gross earnings of 4861 CHF. Hence, young skilled adults would earn about 7% higher wages at the age of 26 years, had they not experienced unemployment in early work life, between the ages of 19–23 years. Furthermore we find that those who experienced early unemployment are on average about 18% more dissatisfied with the progress they made toward meeting their overall career goals compared to a similar group of young skilled workers. Therefore young skilled workers do not only earn less but scarring effects are also reflected in their subjective assessment of career progress. Main results are the same if we apply a weight not correcting for panel attrition (see Table 2).
Table 2

Average treatment effects on the treated

Outcomes:

Weighta)

NT

NC

OutcomeT

OutcomeC

ATT

SE

ATT/SE

NEET-ratio

wt8_kal

205

560

0.15

0.06

0.09

0.06

1.5

basewt

205

560

0.10

0.04

0.06*

0.02

3.0

Wage level

wt8_kal

95

289

4861

5221

−360*

143

−2.5

basewt

95

289

4887

5255

−368*

139

−2.6

Career dissatisfaction

wt8_kal

103

320

0.45

0.27

0.18*

0.06

3.0

basewt

103

320

0.41

0.24

0.17*

0.06

2.8

*p <0.05, Trefers to Treated, Crefers to matched Controls.

a)wt8_kal corrects for both, panel attrition and the disproportionality of the sample, basewt corrects only for the disproportionality of the sample.

Conclusion

Dual-education systems that equip the young with standardised diplomas and liberal employment protection legislations, such as in Switzerland, are well known to allow for smooth transitions from school-to-work (see Fuentes [2011]; De Lange et al. [2014]; van der Velden and Wolbers [2003]; Gangl [2003]; Breen [2005]). However, as our results show, this does not necessarily mean that standardised vocational credentials and liberal market structures protect the young from adverse consequences in case they happen to experience unemployment in early work life.

Investigating causal effects of early unemployment, it is crucial to control for endogeneity. Based on balanced samples from nearest-neighbour propensity score matching, our results suggest that vocationally skilled young workers, who experience unemployment in early work life, are more likely to be neither in employment nor in education or training (NEET). They also earn less later in their careers compared to a similar group of skilled young workers who did not experience early unemployment. In addition, we find that also from a subjective point of view, skilled young workers who experience early unemployment are more dissatisfied with the progress they made toward meeting their overall career goals than they would be had they not experienced early unemployment. In this way, our findings run counter to the expectation that early unemployment does not have much of an impact when vocational degrees are sufficiently standardised, dismissal costs are low and unemployment is of generally short duration.

Drawing this conclusion, we must keep in mind that we only observe that young workers with vocational credentials who experienced some unemployment are worse off in the mid-term compared to a similar group of young workers who experienced smooth transitions. Mechanisms behind unemployment scarring still remain a black box and may only be theorised about. It may be that employers take transitory unemployment as a relevant sorting criterion even if degrees are sufficiently standardised and dismissal costs are low. As early unemployment experiences may also exert a psychological impact, scarring effects may also stem from reduced self-efficacy and self-esteem due to the experience of early unemployment, which in turn influence future job search effort and success (see Furnham [1985]). Over and above this, those who experience early unemployment may be more prone to apply for and accept jobs that offer worse career advancement and employment prospects (see e.g., Mortensen [1986]). In this case it may be self-selection rather than sorting by employers that is at play. Whatever the driving forces of unemployment scarring are, our results clearly contradict assumptions about improvement in job-matching quality as proposed by search and matching theories (see Kahn and Low [1982]). Therefore, we conclude that even a potentially advantageous institutional setting, which on the one hand allows for smooth transitions, does on the other hand not protect the young from adverse consequences in case they happen to experience bumpy transitions. As vocationally skilled entrants are not sheltered from unemployment scarring even within a favourable institutional setting, they should not be left unattended within research and policy considerations addressing labour market vulnerability.

In order to understand scarring effects more comprehensively, future research is needed with regard to the mechanisms driving unemployment scarring among (vocationally skilled) young workers and the development of employment scars in the longer run. It is further not exclusively clarified in which way standardised and vocation specific degrees affect unemployment scarring. Where unemployment is less prevalent, one may hypothesise unemployment to be more stigmatising (e.g., Nilsen and Holm Reiso [2011]). Since unemployment rates for young entrants holding standardised vocational degrees are comparatively low, one may for this reason also assume scarring to be more pronounced among VET graduates. In addition, scarring effects among VET graduates may be heterogeneous. Focusing on potential gender differences with respect to scarring our results (not reported in this paper) point in the direction that wage scars are more pronounced among male graduates while differences in career satisfaction are to some extent more distinct among female graduates. However, due to small group sizes, we cannot stably confirm these results and need to leave the investigation of potentially heterogeneous scarring effects among VET graduates for future research. Paying attention to differences in wage growth and unemployment experiences, comparing those who stay with the same employer where they pursued vocational education and those who need to search for a job at a new firm after graduation (see e.g., von Wachter and Bender [2006]), may also allow for further insight in future research.

Last but not least we must keep in mind that assessment of causation is a risky business. As we have a lot of information on social and individual characteristics of the young entrants, we have the advantage of removing bias based on a comparably comprehensive set of relevant variables. Nevertheless, estimates of causal effects based on non-experimental data may always be affected by unobserved confounders and therefore should be interpreted with caution.

Endnotes

a39 young adults completed a two year pre-apprenticeship program. They are not included in the analysis in order to increase sample balance and reliability of results.

bThis definition of early unemployment includes all unemployment spells observed between 2003–2007, the main transition period of the cohort investigated. This includes unemployment experienced directly after graduation as well as unemployment experienced in the first one to four years of working life, depending on the date of graduation. Scarring effects are then estimated at a cross-section in 2010, which is not the same amount of time after graduation for all respondents. A design in which we focused on unemployment and employment outcomes after the same time span following graduation was, even though preferable, not feasible because no interviews were conducted in 2008 and 2009 as well as because we face small sample sizes concerning those who experienced early unemployment and of whom we observe subsequent employment outcomes (treatment group). Due to these restrictions, the effects we look at should to some extent be seen as combined effects, assessing scarring in a universal way. They do not, however, indicate the exact size of scarring for a certain elapsed period after the experience of early unemployment.

c4 cases do not report any unemployment spells during the main transition period but retrospectively indicate that they actually have been unemployed at some stage during their transition period or report having been unemployed shortly after transition in 2008.

dThese young adults may be seen as interim drop-outs as they were trying to set foot in the labour market after lower secondary education and therefore may not be classified as either unemployed after secondary education nor can they be classified as “never unemployed during transition”.

eAs substantial results are the same whether or not we exclude those who retrospectively indicate having experienced some unemployment in 2008/2009, we leave those who experienced unemployment in 2008/2009 in both treatment and control group.

fNote: A perfect or “close to perfect” prediction of treatment assignment would cause problems in finding a control group (see Heckman et al. [1997]; Bryson et al. [2002]). The aim is not only to find good predictors to model the risk of early unemployment but to include particularly those variables in the estimation of propensity scores which are further expected to be related to future employment outcomes. However, over-parameterized models should be avoided due to arising support problems and possibly increased variation in the matching estimates (see Bryson et al. [2002]).

gThe initial sample of the TREE-Panel was a disproportionally stratified two-stage sample of ninth-graders, where schools and school classes served as primary sampling units (PSU’s). Hence, in order to obtain unbiased population inference and correct estimates of their sampling variance it is necessary to take into account survey weights, stratification, multi-stage-sampling, and finite population corrections (PSU-stage). We use Stata’s svy-commands to accomplish this.

hFor the logistic regression analysis (estimation of propensity scores), we apply a customized and truncated survey weight (for details cf. Sacchi [2011]: 44). For cases with unemployment spells, the panel weight of the first panel wave with a record of unemployment is used. For the censored cases without unemployment spells up to TREE wave 7 (end of the observation period), we employ the panel weights of this wave. Main results regarding the analysis of scarring effects are the same if no weights are included when estimating the propensity scores.

Appendix

Variables used

Gender and Migrant background

Gender, coding is 1 = male and 0 = female

Migration background modelled by language spoken at home (st17n01), coding is 0 = National languages and High German vs. 1 = other language

Migration background operationalised by country of birth (st16n01), coding is 1 = Switzerland vs. 0 = abroad

Family background

Socio-economic status (ISEI)

Familial possession of Poetry literature (st21q10), coding is 1 = yes vs. 0 = no

PISA Index of Social Communication within family (Soccom)

Family structure (Coding is 0 = nuclear family, 1 = single parent and 2 = mixed/other family structure)

Number of siblings (Coding is 0 = no siblings, 1 = one sibling, 2 = two sibling and 3 = three or more siblings)

Individual resources

Self-efficacy (measured prior to the labour market entry period)

Question: To what extent does the following statement apply to you? I can always manage to solve difficult problems if I try hard enough. Coding is 1 = moderately or exactly true vs. 0 = not at all or hardly true.

Persistency (measured prior to labour market entry period)

Question: To what extent does the following statement apply to you? Even if I encounter difficulties, I persistently continue. Coding is 1 = moderately or exactly true vs. 0 = not at all or hardly true.

Positive life attitude (measured prior to the labour market entry period)

Question: In general, what do you think about your life? My life seems to be meaningful. Coding is 1 = rather, very or totally right vs. 0 = totally, very or rather wrong.

Lower secondary educational track/School competencies

Type of school, coding is 1 = extended, 0 = basic (reference category), 2 = no formal tracking

PISA Literacy skills (wleread)

Mathematical skills measured by self-reported school mark, coding is 0 = below average (reference category), 1 = average, 2 = above average

Upper secondary Education (VET)

Classification of vocational education

The classification of upper vocational diploma is based on Stalder ([2011]) “The Intellectual Demands of Initial Vocational Education and Training in Switzerland”. Stalder ([2011]) distinguishes between 6 different levels of demand of vocational education in Switzerland. We summarise these 6 different levels of vocational education into 3 levels and add a 4th level including vocational maturity.
  1. 1

    Low VET comprises level 1 and 2 according to Stalder ([2011])

     
  2. 2

    Medium VET comprises level 3 and 4 according to Stalder ([2011])

     
  3. 3

    High VET comprises level 5 and 6 according to Stalder ([2011])

     
  4. 4

    Vocational Maturity (BMS)

     

Field of occupational training or education

The categories of occupational training are coded into 9 broad fields of education and training (one-digit) and one rest-category. This coding is based on the ISCED-97 classification scheme revised by Andersson and Olsson ([1999]). The fields 0 (general programmes) and 1 (education) are not included as the focus of this paper is on VET graduates and thus no-one is observed pursuing an educational track in these fields. One may note that the classification of educational fields as proposed by Andersson and Olsson ([1999]) differs from the ISCED-97 coding scheme insofar as the coding into educational fields is independent of educational level as well as it is more detailed.

The coding of the field of occupational training is based on the first occupational credential achieved. For those who indicated having absolved a vocational maturity, which is not a field of occupational training on it’s own, we base coding on the vocational education pursued in 2002. We renamed the ISCED blocks of educational fields so that they fit to the dominant forms of vocational education and training in our sample which make up these educational fields.

The following blocks of broad educational fields are built:
  1. 2

    Humanities and Arts (referred to as printing and design)

     
  2. 3

    Social Sciences, Business and Law (referred to as business and sale)

     
  3. 4

    Science, Mathematics and Computing (referred to as [applied] computer science)

     
  4. 5

    Engineering, Manufacturing and Construction (referred to as mechanics and electronics)

     
  5. 6

    Agriculture and Veterinary (referred to as agriculture and gardening)

     
  6. 7

    Health and Welfare (referred to as medical assistance and care)

     
  7. 8

    Services (referred to as gastronomy and hairdressing)

     
  8. 9

    Field of occupational training unknown (referred to as other)

     

Deviant behaviour

Finishing home work in time during lower secondary education, coding is 0 = never or sometimes vs. 1 = mostly or always

Skipping classes during lower secondary education, coding is 0 = never during the last two weeks, 1 = one or more times during the last two weeks

Demographics

Language region, Coding is: 1 = German(reference category), 2 = French and 3 = Italian speaking part

Average cantonal youth unemployment rates for the period of 2003–2007, calculations are based on SECO ([2013])

Outcomes

NEET in 2010

We measure labour market attachment based on a dichotomous variable coded 1 if a person is neither in employment nor in education or training (NEET) in 2010 and coded 0 otherwise. The category of NEET includes among others unemployed (whether officially registered or not), housewives/housemen, those who are in search of a job or education place as well as those engaged in voluntary work and who are neither in paid work, education nor training besides. Thus the category of NEET comprises heterogeneous situations. (Due to sample size restrictions we could not investigate separate groups but need to use this more global measure to address scarring concerning the future labour market attachment of the young).

Wage

Calculation of monthly fulltime equivalent gross earnings for the TREE data is based on the BELODIS project (see Bertschy et al. [2014]).

Career dissatisfaction

The subjective assessment of career progress is measured based on agreement vs. disagreement regarding the following statement: “I am satisfied with the progress I have made toward meeting my overall career goals”. Respondents were asked to indicate how strongly they agreed with this statement based on an ordinal scale of 5 categories. If they reported to strongly disagree, disagree or only half and half agreed with this statement we take this as an indication of dissatisfaction with the career progress and coded our binary outcome variable on career dissatisfaction as 1. If the respondents reported to agree or to strongly agree with this statement we take this as an indication of satisfaction with the career progress and correspondingly coded our binary outcome variable on career dissatisfaction as 0. (One may note that about two-thirds of the sample investigated reports to be satisfied or very satisfied with the progress made towards meeting their overall career goals).

Declarations

Authors’ Affiliations

(1)
Social Research and Methodology Group, University of Basel
(2)
TREE, University of Basel

References

  1. Abadie A, Imbens GW: Notes and comments. On the failure of the bootstrap for matching estimators. Econometrica 2008, 76(6):1537–1557. 10.3982/ECTA6474View ArticleGoogle Scholar
  2. Aigner DJ, Cain GG: Statistical theories of discrimination in labor markets. Indus Lab Rel Rev 1977, 30(2):175–187. 10.2307/2522871View ArticleGoogle Scholar
  3. Wiederholte Arbeitslosigkeit. Situation und Massnahmen. Verband Schweizerischer Arbeitsämter, Regionalkonferenz Ost, Zürich; 2010.Google Scholar
  4. Andersson R, Olsson AK: Fields of Education and Training. Manual. Statistical Office of the European Communities (EUROSTAT), Statistics Sweden; 1999.Google Scholar
  5. Arulampalam W: Is Unemployment Really Scarring? Effects of Unemployment Experiences on Wages. Institute for the Study of Labor, Bonn; 2000.Google Scholar
  6. Arulampalam W, Booth AL, Taylor MP: Unemployment persistence. Oxford Econ Pap 2000, 52(1):24–50. 10.1093/oep/52.1.24View ArticleGoogle Scholar
  7. Becker GS: Human Capital. A Theoretical and Empirical Analysis With Special Reference to Education. Columbia University Press, New York/London; 1964.Google Scholar
  8. Bertschy K, Böhni E, Meyer T (2008) Young people in transition from education to labour market. Results of the Swiss panel survey TREE, update 2007. Accessed 15 Feb 2014, [https://tree.unibas.ch/en/results/publications/] Bertschy K, Böhni E, Meyer T (2008) Young people in transition from education to labour market. Results of the Swiss panel survey TREE, update 2007. Accessed 15 Feb 2014
  9. Bertschy K, Walker P, Baeriswyl A, Marti M: Lohndiskriminierung beim Berufseinstieg. Eine quantitative Analyse für die Schweiz. SZfS 2014, 40(2):279–305.Google Scholar
  10. Best H, Wolf C: Modellvergleich und Ergebnisinterpretation in Logit- und Probit-Regressionen. KZfSS 2012, 64: 377–395. 10.1007/s11577-012-0167-4View ArticleGoogle Scholar
  11. Breen R: Explaining cross-national variation in youth unemployment. Market and Institutional Factors. Eur Sociol Rev 2005, 21(2):125–134. 10.1093/esr/jci008View ArticleGoogle Scholar
  12. Bryson A, Dorsett R, Purdon S: The use of Propensity Score Matching in the Evaluation of Active Labour Market Policies. Crown Copyright. Working Paper no. 4, Department for Work and Pensions, London; 2002.Google Scholar
  13. Buchmann MC: School-to-Work Transitions. In Encyclopedia of Adolescence, vol 2. Edited by: Brown BB, Prinstein MJ. Elsevier, London; 2011:306–313. 10.1016/B978-0-12-373951-3.00085-5View ArticleGoogle Scholar
  14. Bukodi E, Ebralidze E, Schmelzer P, Relikowski I: Increasing flexibility at labor market entry and in early career – a new conceptual framework for the flexCareer project. University of Bamberg, Bamberg; 2006.Google Scholar
  15. Burgess S, Propper C, Rees H, Shearer A: The class of 1981: the effects of early career unemployment on subsequent unemployment experiences. Labour Econ 2003, 10(3):291–309. 10.1016/S0927-5371(02)00138-0View ArticleGoogle Scholar
  16. Clark AE, Georgellis Y, Sanfey P: Scarring: The Psychological impact of past unemployment. Economica 2001, 68: 221–241. 10.1111/1468-0335.00243View ArticleGoogle Scholar
  17. Cruces G, Ham A, Viollaz M: Scarring effects of youth unemployment and informality. Evidence from Argentina and Brazil, CEDLAS, University de la Plata, Lat Plata, Argentina; 2012.Google Scholar
  18. Daly M, Delaney L: The scarring effect of unemployment throughout adulthood on psychological distress at age 50: Estimates controlling for early adulthood distress and childhood psychological factors. Soc Sci Med 2013, 80: 19–23. 10.1016/j.socscimed.2012.12.008View ArticleGoogle Scholar
  19. De Lange M, Gesthuizen M, Wolbers MHJ: Youth Labour Market Integration across Europe. Eur Soc 2014, 16(2):194–212. doi: 10.1080/14616696.2013.821621 doi: 10.1080/14616696.2013.821621 10.1080/14616696.2013.821621View ArticleGoogle Scholar
  20. De Luna X, Johansson P, Sjöstedt-de Luna S: Bootstrap Inference for K-Nearest Neighbour Matching Estimators, IFAU Working paper 13. Institute for Evaluation of Labour Market and Education Policy, Uppsala; 2010.Google Scholar
  21. Dieckhoff M: The effect of unemployment on subsequent job quality in Europe: A comparative study of four countries. Acta Sociol 2011, 54(3):233–249. 10.1177/0001699311412798View ArticleGoogle Scholar
  22. Djurdjevic D: Unemployment and Underemployment. The Case of Switzerland. SchZfVuS 2005, 141(1):1–22.Google Scholar
  23. Du Goff EH, Schuler M, Stuart EA: Generalizing observational study results: applying propensity score methods to complex surveys. Health Serv Res 2014, 49(1):284–303. 10.1111/1475-6773.12090View ArticleGoogle Scholar
  24. Edin P-A, Gustavsson M: Time out of work and skill depreciation. Ind Labor Relat Rev 2008, 61(2):163–180.View ArticleGoogle Scholar
  25. Ellwood DT: Teenage Unemployment: Permanent Scars or Temporary Blemishes? In The Youth Labor Market Problem: It’s Nature, Causes and Consequences. Edited by: Freeman RB, Wise DA. NBER, Chicago; 1982:349–390.Google Scholar
  26. Erikson EH: Identity and the Life Cycle: Selected Papers. International University Press, New York; 1959.Google Scholar
  27. Eriksson S, Rooth DO: Do Employers Use Unemployment as a Sorting Criterion When Hiring ? Evidence from a Field Experiment. Institute for the Study of Labor, Bonn; 2011.Google Scholar
  28. Fuentes A (2011) Raising Education Outcomes in Switzerland. OECD Economics Department Working Paper no. 838, OECD Publishing. doi: 10.1787/5kgj3l0wr4q6-enGoogle Scholar
  29. Furnham A: Youth unemployment: a review of literature. J Adolescence 1985, 8: 109–124. 10.1016/S0140-1971(85)80041-5View ArticleGoogle Scholar
  30. Gangl M: The Structure of Labour Market Entry in Europe: A Typological Analysis. In Transitions from Education to Work in Europe. The Integration of Youth into EU Labour Markets. Edited by: Müller W, Gangl M. Oxford University Press, Oxford; 2003:107–128. 10.1093/0199252475.003.0004View ArticleGoogle Scholar
  31. Gangl M: Scar effects of unemployment: an assessment of institutional complementaries. Am Sociol Rev 2006, 71(6):986–1013. 10.1177/000312240607100606View ArticleGoogle Scholar
  32. Gangl M, DiPrete TA: Kausalanalyse durch Matchingverfahren. In Methoden der Sozialforschung. Edited by: Diekmann A. Westdeutscher Verlag, Opladen; 2004:396–420.Google Scholar
  33. Gerber-Schenk M, Rottermann B, Neuenschwander MP: Passungswahrnehmung, Selbstkonzept und Jugendarbeitslosigkeit. In Schulübergang und Selektion - Forschungsbefunde - Praxisbeispiele - Umsetzungsperspektiven. Edited by: Neuenschwander MP, Grunder H-U. Rüegger, Chur; 2010:121–130.Google Scholar
  34. Goldsmith AH, Veum JR, Darity W: Unemployment, joblessness, psychological well-being and self-esteem: theory and evidence. J Soc Econ 1997, 26(2):133–158. 10.1016/S1053-5357(97)90030-5View ArticleGoogle Scholar
  35. Gregg P, Tominey E: The wage scar from male youth unemployment. Labour Econ 2005, 12: 487–509. 10.1016/j.labeco.2005.05.004View ArticleGoogle Scholar
  36. Guo S, Fraser M: Propensity Score Analysis. Statistical Methods and Applications, SAGE, Los Angeles; 2010.Google Scholar
  37. Hämäläinen K: Education and Unemployment: State Dependence in Unemployment among Young People in the 1990s. Government Institute for Economic Research, Helsinki; 2003.Google Scholar
  38. Heckman J, Ichimura H, Smith J, Todd P: Characterizing selection bias using experimental data. Econometrica 1997, 66(5):1017–1098. 10.2307/2999630View ArticleGoogle Scholar
  39. Hoeckel K, Field S, Grubb NW: Learning for Jobs. OECD Reviews of Vocational Education and Training. OECD Publishing, Switzerland; 2009.Google Scholar
  40. Hupka S, Sacchi S, Stalder BE: Does the Swiss VET System Encourage Inequity? European Research Network on Transition in Youth (TIY). Paper Presented at Workshop on Vocationalisation of Education: how, Where, When, why and in What Sense Does it Matter? Marseilles. TREE-Transitions from Education to Employment, Bern; 2006.Google Scholar
  41. ILO (1982) Resolution Concerning Statistics of the Economically Active Population, Employment, Unemployment and Underemployment, Adopted by the Thirteenth International Conference of Labour Statisticians (October 1982) Accessed 10 Jan 2014. ILO (International Labour Organization) , [http://ilo.org/global/statistics-and-databases/standards-and-guidelines/resolutions-adopted-by-international-conferences-of-labour-statisticians/WCMS_087481/lang--en/index.htm]
  42. Imbens G: Sensitivity to exogeneity assumptions in program evaluation. Am Econ Rev 2003, 93(2):126–132. 10.1257/000282803321946921View ArticleGoogle Scholar
  43. Kahn LM, Low SA: The relative effects of employed and unemployed job search. Rev Econ Stat 1982, 64(2):234–241. 10.2307/1924302View ArticleGoogle Scholar
  44. Kriesi I, Buchmann M, Sacchi S: Variation in Job opportunities for men and women in the Swiss Labor Market 1962–1989. RSSM 2010, 28(3):309–323.Google Scholar
  45. Leuven E, Sianesi B: PSMATCH2: Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing. 2003.Google Scholar
  46. Lucas RE, Clark AE, Georgellis Y, Diener E: Unemployment alters the set point for life satisfaction. Research article. Am Psychol Sci 2004, 15(1):8–13. 10.1111/j.0963-7214.2004.01501002.xView ArticleGoogle Scholar
  47. Luijkx R, Wolbers MHJ: The effects of non-employment in early work life on subsequent chances of individuals in the Netherlands. Eur Sociol Rev 2009, 25(6):647–660. 10.1093/esr/jcp002View ArticleGoogle Scholar
  48. Maloney T: Isolating the Scarring Effects Associated with the Economic Inactivity of Youth in New Zealand: Evidence from the Christchurch Health and Development Study. Report to the Labour Market Policy Group New Zealand Departement of Labour, Oakland; 2004.Google Scholar
  49. Meyer T: Can Vocationalisation of education go too far? The case of Switzerland. EJVT 2009, 46(1):28–40.Google Scholar
  50. Mood C: Why we cannot do what we think we can do, and what we can do about. Eur Sociol Rev 2010, 26(1):67–82. 10.1093/esr/jcp006View ArticleGoogle Scholar
  51. Mooi-Reci I, Ganzeboom HB: Unemployment Scarring by Gender: Human Capital Depreciation or Stigmatizing? Longitudinal Evidence from the Netherlands - 1980–2000. VU, Amsterdam; 2012.Google Scholar
  52. Mortensen D: Job search and Labor Market Analysis. In Handbook of Labor Economics 2. Edited by: Ashenfelter O, Layard R. Elsevier, Amsterdam; 1986:849–919. 10.1016/S1573-4463(86)02005-9View ArticleGoogle Scholar
  53. Mroz T, Savage T: The Long-term Effects of Youth Unemployment. Employment Policies Institute, Washington; 2001.Google Scholar
  54. Müller W, Shavit Y: The Institutional Embeddedness of the Stratification Process. A comparative Study of Qualifications and Occupations in Thirteen Countries. In From School to Work. Edited by: Müller W, Shavit Y. Clarendon Press, Oxford; 1998:1–48.Google Scholar
  55. Müller B, Wolter S: The role of hard-to-obtain information on ability for the school-to-work transition. Empir Econ 2014, 46: 1447–1471. 10.1007/s00181-013-0709-2View ArticleGoogle Scholar
  56. Neuenschwander M: Selektionsprozesse beim Übergang in die Sekundarstufe I, die Berufsbildung und die tertiäre Ausbildung. In Bildung-Arbeit-Erwachsenwerden. Edited by: Bergman MM, Hupka-Brunner S, Meyer T, Samuel R. Ein interdisziplinärer Blick auf die Transition im Jugend- und jungen Erwachsenenalter. Springer, Basel; 2012:180–201.Google Scholar
  57. Neuenschwander M, Kracke B: Career Development. In Encyclopedia of Adolescent. Edited by: Brown B, Prinstein M. Elsevier, Oxford; 2011:97–105. 10.1016/B978-0-12-373951-3.00007-7View ArticleGoogle Scholar
  58. Neumark D: Youth Labor Markets in the U.S.: Shopping Around vs. Staying Put. National Center for Postsecondary Improvement, Technical Report no 2–11. Stanford University, Stanford, CA; 1998.View ArticleGoogle Scholar
  59. Niedergesäss M: Duration Dependence, Lagged Duration Dependence, and Occurrence Dependence in Individual Employment Histories. University of Tübingen Working Papers in Economics and Finance, no. 26, Tübingen; 2012.Google Scholar
  60. Nilsen OA, Holm Reiso K: Scarring Effects of Unemployment. IZA Discussion Paper no. 6198. Institute for the Study of Labor, Bonn; 2011.Google Scholar
  61. Nordström Skans O: Scarring Effects of the First Labour Market Experience: A Sibling Based Analysis. Working Paper. Institute for Labour Market Policy Evaluation, Uppsala; 2004.Google Scholar
  62. Off to a Good Start? Jobs for Youth. OECD Publishing, Paris; 2010.Google Scholar
  63. OECD Employment Outlook. OECD Publishing, Paris; 2013.Google Scholar
  64. Youth unemployment rate. Employment and labour markets: Key tables from OECD. 2013b.Google Scholar
  65. Pissarides CA: Loss of skill during unemployment and the persistence of employment shocks. Q J Econ 1992, 107(4):1371–1391. 10.2307/2118392View ArticleGoogle Scholar
  66. Rosenbaum PR, Rubin DB: The central role of the propensity score in observational studies for causal effects. Biometrika 1983, 70(1):41–55. 10.1093/biomet/70.1.41View ArticleGoogle Scholar
  67. Rüfenacht M, Neuenschwander MP: Jugendarbeitslosigkeit- Risikofaktoren und erfolgreicher Einstieg in die Erwerbstätigkeit. In Selektion in Schule und Arbeitsmarkt. Forschungsbefunde und Praxisbeispiele. Edited by: Neuenschwander MP. Rüegger, Zürich/Chur; 2013:203–224.Google Scholar
  68. Sacchi S: Construction of TREE Panel Weights. Documentation for the Panel Waves from 2000–2010. TREE & cue sozialforschung, Bern/Zürich; 2011.Google Scholar
  69. Sacchi S, Salvisberg A: Berufseinsteiger-Barometer 2011 (ausführliche Version); Report im Auftrag des Bundesamtes für Berufsbildung und Technologie (BBT). Universität Zürich, Stellenmarkt- Monitor Schweiz; 2012.Google Scholar
  70. Salvisberg A, Sacchi S: Labor Market Prospects of Newly Graduated Apprentices in Switzerland. SMM Working Paper no. 2013–1. University of Zürich, Zürich; 2013.Google Scholar
  71. Salvisberg A, Sacchi S: Labour market prospects of Swiss career entrants after completion of vocational education and traning. Eur Soc 2014, 16(2):255–274. doi: 10.1080/14616696.2013.821623 doi: 10.1080/14616696.2013.821623 10.1080/14616696.2013.821623View ArticleGoogle Scholar
  72. Schmelzer P: Unemployment in early career in the UK: a trap or a stepping stone? Acta Sociol 2011, 54(3):251–265. 10.1177/0001699311412626View ArticleGoogle Scholar
  73. Schmelzer P: The consequences of job mobility for future earnings in early working life in Germany-placing indirect and direct job mobility into institutional context. Eur Sociol Rev 2012, 28(1):82–95. 10.1093/esr/jcq049View ArticleGoogle Scholar
  74. Schmillen A, Umkehrer M: The Scars of Youth. Effects of Early-Career Unemployment on Future Unemployment Experience. Institute for Employment research, Nürnberg; 2013.Google Scholar
  75. SECO (2013) Arbeitslosenquote nach Kantonen. Accessed 20 Nov 2013, [http://www.amstat.ch/v2/index.jsp] SECO (2013) Arbeitslosenquote nach Kantonen. Accessed 20 Nov 2013
  76. Seibert H, Hupka-Brunner S, Imdorf C: Wie Ausbildungssysteme Chancen verteilen. Berufsbildungschancen und ethnische Herkunft in Deutschland und der Schweiz unter Berücksichtigung des regionalen Verhältnisses von betrieblichen und schulischen Ausbildungen. KZfSS 2009, 61: 595–620. 10.1007/s11577-009-0084-3View ArticleGoogle Scholar
  77. Seligman MEP: Helplessness: On Depression, Development and Death. W.H. Freeman and Company, San Francisco; 1975.Google Scholar
  78. Facts and Figures. Vocational and Professional Education and Training in Switzerland 2014. State Secretariate for Education, Research and Innovation SERI, Biel; 2014.Google Scholar
  79. Shadish WR, Steiner PM: A primer on propensity score analysis. NAINR 2010, 10(1):19–26.Google Scholar
  80. Sheeran P, Abrams D, Orbell S: Unemployment, self-esteem, and depression: a social comparison theory approach. Basic Appl Soc Psych 1995, 17(1&2):65–82. 10.1080/01973533.1995.9646132View ArticleGoogle Scholar
  81. Sheldon G: Die Langzeitarbeitslosigkeit in der Schweiz. Diagnose und Therapie. SECO, Paul Haupt, Bern; 1999.Google Scholar
  82. Smith J, Todd P: Does matching overcome LaLonde’s critique of nonexperimental estimators? J Econometrics 2005, 125: 305–353. 10.1016/j.jeconom.2004.04.011View ArticleGoogle Scholar
  83. Spence M: Job market signaling. Q J Econ 1973, 87(3):355–374. 10.2307/1882010View ArticleGoogle Scholar
  84. Stalder BE: Das intellektuelle Anforderungsniveau beruflicher Grundbildungen in der Schweiz. Ratings der Jahre 1999–2005. Institut für Soziologie der Universität Basel/TREE, Basel; 2011.Google Scholar
  85. Stalder BE, Nägele C: Vocational Education and Training in Switzerland: Organisation, Development and Challenges for the Future. In Transitionen im Jugendalter. Edited by: Bergman MM, Hupka-Brunner S, Keller A, Meyer T, Stalder BE. Ergebnisse der Schweizer Längsschnittstudie TREE, Seismo, Zürich; 2011:18–39.Google Scholar
  86. Stalder BE, Meyer T, Hupka-Brunner S: TREE Project Documentation. In Transitionen im Jugendalter. Edited by: Bergman MM, Hupka-Brunner S, Keller A, Meyer T, Stalder BE. Ergebnisse der Schweizer Längsschnittstudie TREE, Seismo, Zürich; 2011:66–85.Google Scholar
  87. Stuart EA: Matching methods for causal inference: a review and a look forward. Stat Sci 2010, 25(1):1–21. 10.1214/09-STS313View ArticleGoogle Scholar
  88. TREE Project Documentation 2000–2012. TREE, Basel; 2013.Google Scholar
  89. Van der Velden RK, Wolbers MHJ: The Integration of Young People into the Labour Market: The Role of Training Systems and Labour Market Regulation. In Transitions from Education to Work in Europe. The Integration of Youth into EU Labour Markets. Edited by: Müller W, Gangl M. Oxford University Press, Oxford; 2003:187–211.Google Scholar
  90. Vandenberghe V: The Long-Term Effects of Recessions on Youth. Reviewing the Evidence. Université catholique de Louvain, Louvain-la-Neuve; 2010.Google Scholar
  91. Von Wachter T, Bender S: In the right place at the wrong time: the role of firms and luck in young workers’ careers. Am Econ Rev 2006, 96: 1679–1705. 10.1257/aer.96.5.1679View ArticleGoogle Scholar
  92. Young C: Losing a job: the nonpecuniary cost of unemployment in the United States. Soc Forces 2012, 91(2):609–634. 10.1093/sf/sos071View ArticleGoogle Scholar

Copyright

© Helbling and Sacchi; licensee Springer. 2014

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.