3.

Data, variables and methods

The micro-data used for our analysis of transition strategies and labour market

integration of university graduates come from a nation-wide survey carried out

by the Network of the Careers Offices of Greek Universities in 2005 on a

representative sample of 13,615 graduates belonging to the 1998-2000

cohorts.1 The individuals of the sample were questioned 5 to 7 years after

graduation about their current labour market status, job characteristics and

career aspirations as well as, retrospectively, on topics related to their studies

and the transition process from university to work.

To explore the efficacy of different transition strategies of the individuals or/ and

their families on labour market integration, we have used the micro-data of the

survey to explore the impact of a number of variables corresponding to these

strategies on the odds, 5 to 7 years after graduation, of being (a) employed vs.

unemployed if active (b) in permanent vs. temporary employment if dependent

worker (c) well-paid vs. medium or low-paid if dependent worker

(d) holding a job matching vs. not matching the content of studies if dependent

worker.

For all the above-mentioned cases we have estimated the coefficients of the

predictor variables of dichotomous logistic regression models of the form:

 Log [pi(Y=1)/ (1-pi(Y=1)] = a + b1X1i + b2X2i +…+ b kXki (1)

1 The dataset does not include graduates from Higher Technological Education Institutes (ATEI), which are
also part of the Greek higher education system.

12
The general hypothesis we have tested through statistical analysis is that, 5-7

years after graduation, the labour market integration and job characteristics of

Greek university graduates are mainly accounted for by sex, having child(ren),

family background, motivation for studies and ability, human capital

accumulation before and after graduation, job mobility, the field of study, the

private/ public sector of employment and the size of firm. “Parental income”,

the “father’s” or “parental educational attainment level” are the variables that

we have used to capture the impact of family background. “Interest for the field

of studies at the entry of university” was used as a proxy for the motivation for

studies while the “grade of degree” for ability. Ho wever, we have also

assumed that the latter does not only - or mainly - depend on innate ability, but

also - and mostly - on social origin, motivation for studies and individual

strategies regarding the transition from education to work, affecting the

decision about working while studying.

To control for human capital accumulation in addition to education we have

used a great number of variables, such as “postgrad uate studies”, “work

experience during undergraduate studies and type of work experience”,

“traineeship during undergraduate studies organised by the university”,

“participation to an ALMP scheme”. To capture varia tions in the accumulation

of work experience and job opportunities since graduation due to time spent in

the labour market, we have used as a proxy variable the “time lapse since

graduation”.
13
Table 1: Dependent and independent variables.

Dependent variables Values

Odds of being employed when active Employed=1, unemployed=0
Odds of being a permanent worker Permanent=1, temporary=0
when dependent worker
Odds of being paid more than 1,100 € per Wages >1,1 00€ =1
month when dependent worker Wages ≤1,100€=0
Odds of having full or rather good match Full or rather good job match=1
with studies in job when dependent
worker Little or no job match=0

Independent variables
Age Number of years
Sex Man=1, woman=0
Having a child or more No=1, yes=0
Sex * having a child Man without children=1,
Woman with children=0
≤10,000€=2, 10,001-30,000€=1,
Level of annual parental income >30,000€=0
Father's educational attainment level* Low=2, medium=1, high=0
Parental educational attainment
(continuous)** 2,3,4,6,7,10
Field of study Ten groups of fields***
Great scientific interest=3, small=2,
Motivation for studies at entry in
university no=1

I knew nothing about the field=0

Grade of degree (continuous) From 5 to 10 points
Grade of degree (categorical) Good=2, very good=1, excellent=0

Post-graduate studies No=1, yes=0

Participation to traineeship programme
during No=1, yes=0

undergraduate studies
Work experience during undergraduate No or occasional experience =1,
studies continuous=0
Potential work experience since
graduation Time lapse since graduation in months

Job mobility (all graduates) Number of jobs before current state

 Job mobility (dependent workers) Number of jobs before current job
Up to 1 month=4, 1-6 months=3, 6-12
Joblessness period prior to current job
spell months=2

1-2 years=1, more than 2 years=0

Part-time worker=1, full-time
Full/part-time work worker=0
Type of contract Temporary=1, permanent=0
Sector of employment and size of private
firm Public sector=2, private firm with <50
employees=1, private firm with >50
employees=0
No match=3, little match=2, rather
Degree of match between job and studies good
match=1, full match=0

*Low=primary school or below, medium=lower or upper secondary education,

high=higher education, Masters or doctoral degree. ** All combinations between
father's and mother's educational levels (low, medium, high). ***Law, humanities,
engineers, economics and business, positive sciences, social and political sciences,
life and health sciences, agricultural and environmental sciences, fine arts,
physical education and sports.

According to our approach, the human capital variables listed above, except for

the last one, correspond to different transition strategies employed by

individuals to achieve labour market integration. The scores that individuals

14
obtain in these variables operate as signals of potential productivity that

influence the hiring and job assignment decisions of employers. The indicators

of job mobility and the grade of degree perform the same function. The impact

of labour demand on the degree and quality of labour market integration is

captured by proxies such as the field of study, the public/private sector of

employment and the size of firm. All the dependent and independent variables

of all the regression models and their definition appear in Table 1.

The results of regression analyses are presented in the Appendix, which

provides the coefficients of only the statistically significant independent

variables for each regression model. A report on the variables that were found

statistically insignificant appears in the footnotes of the Table. The model-

building process was stepwise and used as a guide. The final model was

checked to exclude collinearity by comparing results from univariate and

multivariate analyses and by checking the K-agreement coefficient or the

correlation coefficient, depending on the nature of the dependent variables. For

continuous covariates we have alternatively used linear functions or categorical

transformations to check for the appropriate functional form. To compare

nested models for each regression model we have used the likelihood-ratio test.

For the overall goodness of fit of the final model we have used and provide on

the tables the Hosmer-Lemeshow test, which is considered more robust in the

case of logistic regression than the traditional chi-square test, particularly if

continuous covariates are included in the model. A finding of non-significance

is needed to conclude that the model adequately fits the data, which is the case

15
in all our regression models. In logistic regression classification tables should

not be used as goodness-of-fit measures, because they ignore actual predicted

probabilities and instead use dichotomized predictions based on a cutoff (in our

case 0.5). However, we also provide on our tables the percentage of correct

classification for each regression model along with measures of the Cox and

Snell and the Nagelkerke pseudo R-square.

Before discussing the results of statistical analysis, we will use some general

indicators to describe the degree and quality of the graduates’ integration and

draw the basic features of their transition from university to work.

APPENDIX

Table: Logistic regression results.

Unemployed=0 Temporary worker = 0

DEPENDENT VARIABLES (odds) Employed = 1 Permanent worker = 1

EXPLANATORY VARIABLES Beta S.E. Beta S.E.

Sex (woman) man 1.204*** 0.314

Child (yes) no 0.512*** 0.117

Child (yes) by sex (woman) -1.394*** 0.324

Grade of
degree 0.133*** 0.029

Work experience during undergraduate studies

(continuous) no experience/occasional -0.614*** 0.133

Traineeship during undergraduate studies (yes) -0.369*** 0.086

Post-graduate studies (yes) no 0.189** 0.092

Participation to ALMP (no) yes -0.466*** 0.084

Time lapse since graduation (months)

0.015*** 0.004 0.016*** 0.003

Number of jobs before current state 0.027*** 0.003

Number of jobs before current job -0.054** 0.023

Joblessness prior to current job spell

(more than 2 years) up to 1 month 0.436*** 0.120

1-6 months 0.138 0.132

6-12 months 0.124 0.162

1-2 years 0.250* 0.149

Sector of employment and size of firm

(private sector and firm with ≥50 employees)

public sector -0.449** 0.123

private sector and firm with <50 employees -1.293** 0.129

Field of studies (Law)

Humanities -1.055*** 0.425 -1.051*** 0.354

Engineers 0.018 0.445 -0.587* 0.345

Economics and business -0.710 0.433 0,152 0.374

Positive sciences -0.445 0.437 -1.382*** 0.346

Social and political sciences -1.336*** 0.432 -0.603* 0.362

Life and health sciences -0.630 0.453 -2.147*** 0.374

Agricultural and environmental sciences -1.355*** 0.438 -1.015*** 0.390

Fine arts 0.162 0.553 0,292 0.643

Physical education and sports -0.600 0.487 -0.979** 0.409

Constant

2.575*** 0.519

6.103 (df Sig. 8.602 (df

Hosmer and Lemeshow Test 8) 0.636 8) Sig. 0.377

Cox and Snell R square 0.044 0.287

Nagelkerke R square 0.111 0.39

Correct classifications 93.0% 74.5%

Number of observations 10,436 3,140

37
Table: Logistic regression results (cont.)

Wages ≤1,100€ = 0

DEPENDENT VARIABLES (odds) Wages > 1,100€ = 1

EXPLANATORY VARIABLES Beta S.E.

Sex (woman)

man 0.398*** 0.065

Age 0.112*** 0.010

Work experience during undergraduate studies

(continuous)

no experience/occasional -0.527*** 0.084

Post-graduate studies (yes)

no -0.363*** 0.063

Number of jobs before current

job -0.065*** 0.019

Full-time worker part-

time worker -1.371*** 0.135

Type of contract (permanent)

temporary -0.284*** 0.075

Sector of employment and size of firm

(private sector and firm with ≥50 employees)

public sector -0.712*** 0.080

private sector and firm with <50 employees -0.596*** 0.088

Job matching with studies (full match)

no match -0.727*** 0.104

little match -0.516*** 0.097

rather good match -0.290*** 0.076

Field of studies (Law)

 Humanities -3.240*** 0.313

Engineers -1.240*** 0.319

Economics and business -2.761*** 0.316

Positive sciences -2.471*** 0.324

Social and political sciences -2.747*** 0.322

Life and health sciences -1.861*** 0.342

Agricultural and environmental sciences -2.564*** 0.334

Fine arts -3.697*** 0.386

Physical education and sports -2.999*** 0.382

Hosmer and Lemeshow Test 6.074 (df 8) Sig. 0.639

Cox and Snell R square 0.298

Nagelkerke R square 0.397

Correct classification 74.6%

Number of observations 6,456

Notes: 1. S.E. = Standard Error; 2. Reference categories in parentheses; 3. Level of statistical

significance: *=0.10, **=0.05, ***=0.01. Non-significant variables: 1st model: parental income,
father’s and parental education, grade of degree, post-graduate studies; 2nd model: sex, parental
income, father’s and parental education, participation to traineeship or work experience during
undergraduate studies, participation to ALMP after graduation; 3rd model: parental income,
father’s and parental education, grade of degree.