The determinants of overeducation: Evidence from the...

26
The determinants of overeducation: Evidence from the Italian labour market Preliminary Draft Clemente Pignatti Morano * Research Department International Labour Organization Geneva – 1211 +41 22 799 6251 The author is grateful for the comments received at the Winter School “New Skills and Occupations in Europe: Challenges and possibilities”, organized by the Centre for European Policy Studies (CEPS) in Brussels, 25-27 November 2013. * [email protected]

Transcript of The determinants of overeducation: Evidence from the...

Page 1: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

The determinants of overeducation:

Evidence from the Italian labour market

Preliminary Draft

Clemente Pignatti Morano*

Research Department International Labour Organization

Geneva – 1211 +41 22 799 6251

The author is grateful for the comments received at the Winter School “New Skills and Occupations in Europe: Challenges and possibilities”, organized by the Centre for European Policy Studies (CEPS) in Brussels, 25-27 November 2013.

* [email protected]

Page 2: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

Abstract: The paper analyses the determinants of overeducation in the Italian

labour market for workers with an MSc degree, using data from the National

Labour Force Survey during the 2006-2011 period. The results confirm some of

the findings obtained by previous studies and they show that overeducation is

higher among youths, couples and foreign citizens. Interaction terms added to

the baseline regression reveal that higher overeducation among youths mostly

concern female workers and it is related to youth employment with temporary

contracts. Moreover, the study shows that overeducation is lower for workers

employed in big firms and for those with full-time contracts. The subject of study

also plays an important role, as workers who have obtained a more scientific

(e.g. engineering, scientific sciences) or a very specific (e.g. medicine,

architecture) degree are less likely to be overeducated. Previous conditions in

the labour market also affect current matching, with overeducation positively

correlated with previous unemployment status and negatively correlated with

previous student status. The paper finally examines whether labour market

conditions affect the likelihood of being overeducated. The results show mixed

evidence across age categories, but they reveal that unemployment increases

overeducation for workers below the age of 25.

Page 3: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

1. INTRODUCTION 4

2. LITERATURE REVIEW 5

2.1 The beginning of the literature on overeducation 5

2.2 The determinants of overeducation at the individual level 5

2.3 Overeducation and macroeconomic conditions 7

3. MEASURING OVEREDUCATION 7

4. DATA AND METHODOLOGY 9

5. THE RESULTS 11

5.1 Main microeconomic results 11

5.2 Interaction of microeconomic variables 15

5.3 The role of unemployment in explaining overeducation 16

6. CONCLUSIONS 19

BIBLIOGRAPHY 23

Page 4: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

4

1. Introduction

An extensive literature documents the microeconomic determinants of overeducation

in the labour market and it aims at assessing its consequences, especially in terms of

earnings’ losses (Leuven and Oosterbeek, 2011). Competing theories have been

proposed, for example considering overeducation only as a temporary status related

to occupational mobility (Johnson, 1978; Jovanovic, 1979; Sicherman, 1991; Groot,

1996); or rather as a long-term phenomenon connected to labour market rigidities or

individual unobserved heterogeneity (Clarck et al., 2012; Chevalier, 2003; Rubb,

2003). The first objective of this paper is to contribute to this literature by examining

the microeconomic determinants of overeducation in the Italian labour market, where

the phenomenon has been seldom analysed (Di Pietro and Urwin, 2003).

From a policy point of view, research on the microeconomic determinants of skills

mismatch and overeducation is motivated by the considerable increase in the

proportion of workers with a graduate degree in developed economies, which could

potentially lead to a reduction in the returns to education (Dolado et al., 2003; Leuven

and Oosterbeek, 2011).

The paper then aims at expanding the literature on overeducation by studying

whether labour market conditions affect the likelihood of being overeducated. Indeed,

while there is a growing literature providing evidence of the role played by labour

market conditions at the time of graduation in affecting future career paths,

considerably less attention has been paid to explaining the dynamics that at the

beginning of the career trigger these long term consequences (Liu et al, 2012). In

particular, the overeducation literature has not extensively targeted the question

whether negative labour market conditions increase the likelihood of being

overeducated, by for instance reducing the number of available jobs.

The answer to this question would widen the understanding of the long-term labour

market consequences of economic downturns and it would also help in the design of

active labour market policies (e.g. school to work transition) and unemployment

benefit schemes.

The rest of the paper is organized as follows. Section 2 reviews previous studies and

theories on overeducation; Section 3 examines the measurement issues related to the

concept of overeducation; Section 4 presents the data and methodology used in this

paper; Section 5 analyses the results; Section 6 concludes.

Page 5: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

5

2. Literature review

2.1 The beginning of the literature on overeducation

Skills mismatch and overeducation have started being extensively analysed by the

academic literature in the 1970s, when concerns emerged in the US over the fact that

the supply of skilled workers seemed to outpace its demand (Berg, 1970). “The

Overeducated American” by Freeman (1976) represents one of the most influential

studies of that period. The author compared the average income of college graduates

and high school students at the time of entry in the labour market and reported a

decrease in the relative wage premium by 24 per cent between 1969 and 1974.

The topic immediately triggered an extensive debate and Smith and Welch (1978)

soon challenged the results obtained by Freeman. In particular, they added four years

to the analysis and accounted for the differences in age at the time of entry into the

labour market between high school and college graduates. They reported a

considerably smaller fall in the wage premium – equal to 8 per cent – and concluded

that the evidence was more in line with an overcrowded labour market.

The attention on skills mismatch was then revitalised by Duncan and Hoffman (1981),

who presented one of the first studies based on the comparison between the amount

of education supplied by a worker and the one required for his/her job. In particular,

the wage equation they introduced allowed to distinguish the returns to education

between years of required education, overeducation and undereducation (Leuven and

Oosterbeek, 2011). Their results showed that the returns of an additional year of

required education was twice as much as the return to an additional year of

overeducation.

Since then, an important part of the literature has followed the seminal study of

Duncan and Freeman by analysing the magnitude and persistence of earnings’ losses

associated to overeducation. Numerous other studies have instead used binary

outcome models to examine the determinants of overeducation at the individual level.

The following subsection reviews this second field of research.

2.2 The determinants of overeducation at the individual level

It is difficult to conduct a comprehensive review of the studies on the determinants of

overeducation at the individual level, since the specifications of the models greatly

differ. 1 Moreover, little motivation is generally presented for justifying the variables

that are included as controls and those who are not. Similarly, some of the included

1 See Leuven and Oosterbeek (2011) for a comprehensive review of the literature on skills mismatch.

Page 6: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

6

variables such as job tenure might be endogenous, while different strategies have

been followed to measure other variables such as individual ability. However, it is

possible to draw conclusions on the relation between overeducation and some

individual variables such as age, gender, ethnicity and ability.

Age: There is a general agreement in the literature over the fact that the probability of

being overeducated decreases with age. Different labour market theories can explain

this empirical finding. First, labour market search theories predict that the quality of

job matches increases throughout the career (McGuinness and Wooden, 2009).

Moreover, young workers are more likely to be overeducated because employers may

compensate for their shorter working experience with additional education

(Sicherman and Galor, 1990). Finally, theories of career mobility predict that the skills

acquired as overeducated may increase the probability of being promoted, thus

presenting overeducation as an optimal strategy at the beginning of the career

(Leuven and Oosterbeek, 2011).

Gender: Different studies have documented that women are more likely to be

overeducated than men. The motivation generally relates to the fact that when man’s

wage represents the primary source of income, key households’ decisions – such as

house location and childcare – are taken in order to maximise his labour market

opportunities (Dolton and Silles, 2001). The choices available to women in the labour

market are therefore necessarily restricted – for instance only to part-time work –

leading to a higher probability of being overeducated (Frank, 1978; Ofek and Merrill,

1997).

Ethnicity: The analysis of the role of ethnicity in explaining overschooling has been

relatively constrained by the difficulties related to comparing educational systems

across countries. However, different studies have found a positive relation between

being a minority and employment as overeducated (Green et al., 2007; Battu at al.,

2004; Sharaf, 2013). The reasons advocated behind this phenomenon do differ. For

example, additional education might be required to compensate for other

shortcomings such as the lack of proficiency in the country’s language (Battu et al.,

2004). Alternatively, this might be simply the result of labour market discrimination,

of either taste based or statistical form (Altonji and Blank, 1999).

Ability: Being employed as overeducated can also be the effect of lower individual

ability, which is compensated by the employer with additional schooling. Even in this

case, the results of the literature have been limited by the difficulties in having access

to information on individuals’ skills. However, studies that include some measures of

individual ability generally find a negative correlation with overeducation (Allen and

Van der Velden, 2001; Chevalier and Lindley, 2009; Green and McIntosh, 2007).

Page 7: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

7

2.3 Overeducation and macroeconomic conditions

Considerably less attention has been paid in the literature to analysing whether

overall macroeconomic conditions influence the probability of being overeducated.

The hypothesis is that negative labour market conditions restrict the number of

opportunities that an individual looking for a job faces, while raising the risk of

remaining unemployed. This increases the probability of accepting a job below the

level of education he/she has acquired (Dolton and Silles, 2001).

The theoretical literature has recently started modelling this mechanism. The first

contributions that have introduced matching models with heterogeneous agents are

of Pissarides (1999) and Acemoglu (1999). One of the first studies specifically

targeting job competition by educational level has been conducted on Spain and finds

evidence of an increase in overeducation during the 1980s as a result of biased

technological changes (Collard et al., 2002). Other contributions have expanded the

model by considering on-the-job search by mismatched high skilled workers (Gautier,

2002; Dolado et al., 2003) and endogenous skills requirements (Albrecht and Vroman,

2002).

The empirical literature has mostly studied whether labour market conditions at the

time of graduating play any effects on future careers’ paths. Kahn (2010) and

Oreopoulos et al. (2012) provide evidence that in America college students graduating

during a recession are affected by persistent earnings’ losses. Only recently, there

have been some attempts to explain these negative shocks by referring to

overeducation and skills mismatch. Liu et al. (2012) use a panel database on college

graduates from Norway to show how skills mismatch at the beginning of the career

plays a significant role in explaining future career losses. In particular, they find that

overeducation has a large countercyclical trend and that an increase in

unemployment rate by 1 per cent increases the probability of being overeducated by

3.4 per cent. Similarly, Hagedorn and Manovskii (2010) show how labour market

tightness affects the quality of job matches. Dolton and Silles (2001) also look at

unemployment rate as a potential determinant of over-education among college

graduates in the UK, however without finding a statistically significant result. 2

3. Measuring overeducation

An important feature that characterizes any study on overeducation relates to the

measure of overeducation that it is used. The issue is to compare individuals’

educational attainments with the educational requirements of their job. The more

challenging task is then related to defining the required level of education for a

2 Some policy papers have also recently documented the rise in skills’ mismatch and reported the lack of research on its macroeconomic determinants (CEDEFOP, 2010).

Page 8: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

8

particular job or occupation and three methods have been mainly used in the

literature (Leuven and Oosterbeek, 2011).

Self-assessment: The first method uses answers to surveys that ask workers the

educational requirement associated to their job or occupation. This method has been

widely used and its main strength relates to the fact that it is potentially based on all

the information that is needed to define overeducation. However, the first problem is

that workers tend to overstate the educational requirements of their job (Hartog,

2000). Additionally, the questions reported in the surveys differ substantially. Some

interviews refer to recruiting standards (Duncan and Hoffman, 1981; Galasi, 2008);

while others to the skills needed to perform the job (Hartog and Oosterbeek, 1988;

Alba-Ramirez, 1993). Moreover, some studies include both formal and informal

education and training, while others refer specifically to the highest educational

degree obtained. Coherently, it has been reported that the same person responds

differently to these similar questions (Green et al, 1999).

Job analysis: A second approach is based on databases – such as the Occupational

Information Network in the US – that report information on occupational categories.

External analysts determine the required level of education related to each occupation

and this is translated into the corresponding years of education. The positive feature

of this measurement technique is that it is less subject to individual bias and that the

assessments are based on the levels of technology associated with each occupation

(Dolton and Silles; Leuven and Oosterbeek, 2011). However, updates in the

classification are infrequent and this might reduce the validity of the benchmarks

(Hartog, 2000). Moreover, these classifications are not conducted in many countries,

thus limiting comparability of studies. Additionally, the translation of educational

requirements into years of schooling is subject to debate (Halaby, 1994).

Realized matches: The last method deducts the required level of education from the

realized job matches in the labour market and it has first been proposed by Verdugo

and Verdugo (1989). In particular, the required level of education is derived as the

mean level of education of the workers in the same occupation. Most commonly, the

three-digit categorization of professions is used. Workers are then classified as over

or under qualified if their educational attainment is at least one standard deviation

above or below the mean educational level in their occupation (Verdugo and Verdugo,

1989).3 This method is criticized because it is the result of demand and supply forces,

rather than a genuine comparison of required and acquired levels of education

(Leuven and Oosterbeek, 2011). Moreover, the one standard deviation benchmark is

arbitrary and the results could be subject to outliers in small databases. Generally, it

3 Kiker et al. (1997) propose to use the mode rather than the mean for computing the required level of education.

Page 9: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

9

has been documented that the realized matches approach underestimates

overschooling (Groot and van den Brink, 2000).

4. Data and Methodology

The present study uses data from the Continuous Labour Force Survey (Rilevazione

Continua delle Forze di Lavoro) conducted by the Italian National Institute of Statistics

(ISTAT). This survey is not directly available to the public, but it can be provided upon

request. The Continuous Labour Force Survey represents the main source of

documentation of the Italian labour market and it has been conducted since 1959. In

order to comply with European standards, the methodology has been reviewed in

2004. The main innovation has been the adoption of a new frequency for the

interviews, which were previously conducted during one single week in each quarter

and are now distributed throughout all weeks of the year. The reference sample

consists of all households resident in Italy.

The ISTAT Continuous Labour Force Survey does not ask individuals about the level of

education required for their job, thus precluding the use of the self-assessment

method for measuring overeducation. Similarly, there are not reliable databases in

Italy in which external analysts convert jobs and occupations into levels of required

education, thus preventing the use of the job analysis method. However, the ISTAT

Labour Force Survey collects information on the highest degree obtained by each

worker. This paper therefore uses the realized matches approach to measure

overeducation. The categorization of workers in occupational groups follows other

studies and it is based on the 3-digit code. For each occupational group, the mean level

of education is computed and a worker is considered overeducated if his/her highest

educational attainment is one standard deviation above the average level of education

of the relevant occupational group.

Notwithstanding the limitations of the realized matches approach, some

considerations should justify its use in this specific context. First, the very large

dimension of the sample – around 1.4 million employed individuals – considerably

limits the risk that outliers will influence the mean values obtained for the required

level of education. This is one of the main concerns when the analysis is conducted

with small panel databases, as the three-digit categorization of professions may lead

some sub-groups to have a very small number of units.4 Second, the short time period

under consideration – from 2006 to 2011 – considerably reduces the risk that the

measures of required level of education are inconsistent across years due to

substantial technological changes. Finally, the measure of overeducation obtained in

this paper can be compared with the results presented by other studies. In particular,

4 In order to avoid this problem, all occupational groups with less than 100 observations have been eliminated. However, this has generated the exclusion of only 3 occupational groups.

Page 10: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

10

13.75 per cent of the working age population between 20 and 64 years old results

being overeducated. Leuven and Oosterbeek (2011) report that the studies of

overeducation that use the realized matches approach present on average a share of

overeducation equal to 13 per cent.

After having computed overeducation, we restrict the analysis to workers with an MSc

degree – around 120,000 employed individuals in the database. This is done in order

to avoid pooling together categories of workers for which the nature and

determinants of overeducation might differ. This problem has been recently raised in

the literature of overeducation by Arcidiacono et al. (2010). The MSc category is

chosen among the other graduate categories because of its significantly higher

dimension. In particular, 80.73 per cent of workers with a degree above the high

school level in the database have obtained an MSc degree. However, we confirm the

results obtained for MSc students by running a regression that also includes BA and

PhD graduates and that is presented in the Appendix (Table 1C).

We exclude workers employed in the armed forces due to the atypical nature of their

profession and we restrict the analysis to workers between 20 and 64 years old. This

represents a slight modification as compared to the traditional definition of working

age population (15-64 years old), needed to include only workers who have obtained

a postgraduate degree. Summary statistics for workers with an MSc degree are

provided in Table 1A in the Appendix.

The results should be interpreted relatively to the base worker, which is a man

between 50 and 54 years old. He is an Italian citizen, with an MSc degree in human

sciences, living in a couple – either married or not. He works in the service sector in a

firm with less than 10 employees, having the professional status of clerical worker. He

has a full-time and permanent job and he was already employed one year before the

interview. Regional dummies corresponding to the 19 regions reported in the survey

have been introduced, together with quarterly dummies for each quarter from 2006

to 2011. Although these dummies are not included in the results, it is worth noting

that the base worker lives in the region of Lombardy in the first quarter of 2008.

Employment and unemployment rates are regional quarterly data available from

ISTAT database. Finally and in order to account for the possible differences in the

effect of unemployment on overeducation across age groups, we interact

unemployment rate with the different age categories.

We use a probit model reporting marginal results with robust standard errors, as this

is the most common approach used in the literature (Leuven and Oosterbeek, 2011).

Sampling weights have been used in order to take into account of the design of the

survey. The resulting regression is described by the following equation:

Page 11: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

11

where Y is the dependent variable equal to 1 if the worker is overeducated and 0

otherwise; α is the intercept; Θ represents the time dummies; C the regional dummies;

is a set of individual characteristics; Z unemployment rate and its interaction with

the age groups; ε is the error term.

Interactions between microeconomic variables are added to the baseline regressions

at different stages to test additional hypotheses. The resulting regressions take the

following form:

where X and W are two sets of individual characteristics and their product represents

the interaction term.

5. The results The analysis of the results reveals that many individual determinants of

overeducation behave according to the predictions and the paper confirms most of the

findings obtained by previous microeconomic studies. Interaction terms are then

added at different stages and they reveal important combined effects on

overeducation of age and gender and age and the nature of the employment relation.

Finally, the results related to the effects of unemployment on overeducation differ

across age groups, but evidence suggests that unemployment increases the likelihood

of being overeducated for workers below the age of 25.

5.1 Main microeconomic results Different theories predict that the quality of job matches increases with age and the

empirical literature on skills mismatch has coherently documented how the

probability of being overeducated is higher among youths. Accordingly, the results of

this paper show how the likelihood of being overeducated decreases with age, with all

coefficients being statistically significant. As expected, the coefficients are stronger for

those age groups that are more distant from the category that has been dropped (50-

54 years old). However and quite surprisingly, the magnitude of the coefficient

remains fairly constant up to the age of 40 – suggesting that the generational penalty

is rather persistent (Table 1).

Foreign workers are also expected to be employed as overeducated more often than

national citizens. This might reflect either employers’ rational expectations about

their work quality – statistical discrimination – or rather employers’ distaste for

foreign workers – taste based discrimination. Additionally, lack of comparability of

university degrees across countries may disadvantage foreign citizens that have

Page 12: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

12

completed their studies in their home country. According to these predictions, the

results show how foreign workers – defined as workers without the Italian citizenship

– are more likely to be overeducated than Italian workers. The effect is particularly

strong, with the probability of being overeducated increasing by 13 per cent among

foreign citizens in all the different specifications.

Family composition can also play a role in explaining overeducation by modifying the

opportunity choices from which an individual looking for a job can draw. For example,

workers in a couple are likely to be constrained in their job search by the decisions of

the other member of the couple. Consistently, the results show how being single

decreases the likelihood of being overeducated with respect to workers who are in a

couple – either married or not. 5 The coefficient related with other forms of family

composition – e.g. family groups – is instead not statistically significant.

An additional factor potentially explaining overeducation is the subject of the degree

obtained. This relation has been rarely analysed in the literature, while it can be of

particular interest as it is consistent with the contemporaneous presence of a

generally overeducated workforce and the shortage of skilled workers in some

specific sectors. The hypothesis in this case is that less specific degrees could lead to

an increase in the likelihood of being overeducated, as the worker is not perceived as

having acquired a specific set of knowledge. Alternatively, we would expect to have

higher overeducation among the students of those subjects for which there is an

oversupply of workers. The results confirm these hypotheses and reveal that –

compared to the omitted group of those with a degree in human sciences – workers

who have obtained a more scientific degree like sciences and engineering are less

likely to be overeducated. Similarly, degree qualifications such as architecture and

medicine, which have a direct link to specific professions, decrease the likelihood of

being overeducated.6

Examining the role played by previous conditions inside or outside the labour market

allows us to determine whether previous career paths influence the quality of future

matches – e.g. through signalling mechanisms. The results show that having been

unemployed in the previous year increases the likelihood of being overeducated on

the current job with respect to the omitted category of those who were employed in

the previous year. This is consistent with the idea that previous unemployment

condition may reveal the low quality of the applicant and it thus increases the

likelihood that the employer requires additional schooling for the same position. On

the other hand, having been a student in the previous year decreases the likelihood of

5 The interaction between the family composition and the female dummy reveals how single women are relatively more likely to be overeducated than single men (not reported). 6 In Model 5 in the Appendix (Table 1B) we interact the subject of study with the female dummy. The results show that women are relatively less overeducated in those areas (e.g. engineering, science) where they are underrepresented in our database. The opposite applies for degrees where there is a majority of women graduates (e.g. medicine).

Page 13: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

13

being overeducated on your current job. This might be related to the fact that recent

students are more likely to be able to apply the skills that they have just acquired and

their knowledge should not be outdated. Not having been in the labour force in the

previous year has instead a generally positive but statistically not significant effect on

overeducation.

Given the relevance of micro enterprises in the Italian economy – equal to 94.5 per

cent of total enterprises – it is worth examining whether firm size plays any role in

determining the likelihood of being mismatched. The results show how indeed

overeducation decreases with firm size. This is consistent with the idea that bigger

firms have more accurate recruitment techniques that reduce the risk of hiring a

worker that does not match the educational requirements associated with the vacancy

(Dolton and Silles, 2001). Moreover, in big firms there is a wider range of positions

that enables the management to internally relocate workers in case of mismatch. The

only exception is represented by firms between 11 and 19 employees, where the

likelihood of being overeducated is slightly higher than in firms with less than 10

employees. However, the two categories of firms are likely to have similar

organizational strategies and this might explain the result.

Additionally, the results show how the probability of being overeducated is higher for

workers with part-time contracts with respect to their colleagues in full-time

employment. This is consistent with the idea that workers looking for a part-time job

are more constrained in their job search with respect to their colleagues that are

willing to work full time. An additional explanation relates to the fact that employers

may decide to prioritise – for instance in terms of training and career opportunities –

workers employed on a full time basis within the firm.7

Similarly and as expected (Dolton and Silles, 2001), the results related to the career

position reveal that workers at the top of the career ladder (directors and managers)

are less likely to be overeducated than those at the bottom (blue collar workers). The

only exception is represented by trainees – who are less likely to be overeducated

than the omitted category of clerical workers. This result can be explained by the

generally high-skilled nature of traineeships for workers with an MSc degree.

Finally, the results show that the likelihood of being overeducated significantly differs

across economic sectors. In particular, workers in the service sector are less likely to

be overeducated than those in the industry or in the agricultural sector. These results

can be interpreted by the different nature of employment in the three economic

sectors and the relatively less skilled nature of jobs in agriculture and industry.

7 In model 6 in the Appendix (Table 1B) we interact part-time employment with the female dummy. The results reveal how the positive relation between overeducation and part-time employment holds uniquely for women – which indeed represent 80 per cent of all part-time workers.

Page 14: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

14

Age categories (50-54) 20-24 0.26633 *** (0.0069267)

(0.0310587) Agriculture 0.110089 ***

25-29 0.1902011 *** (0.0342618)

(0.0094174) Firm size (1-10) Firm 11-19 0.019767 **

30-34 0.1621838 *** (0.0086609)

(0.0078102) Firm 20-49 -0.07515 ***

35-39 0.157469 *** (0.0072935)

(0.0074218) Firm 50-250 -0.15446 ***

40-44 0.1123465 *** (0.0068573)

(0.0075801) Firm >250 -0.09104 ***

45-49 0.044284 *** (0.0079431)

(0.0077861) Subject of study (human science) Social sciences 0.296735 ***

55-59 -0.0368199 *** (0.0075454)

(0.0083244) Languages 0.084947 ***

60-64 -0.1041801 *** (0.0085047)

(0.0123402) Economics-Statistics 0.277781 ***

Personal characteristics No citizen 0.1370148 *** (0.006791)

(0.014712) Law 0.195905 ***

Single -0.0291957 *** (0.0079826)

(0.006799) Sciences -0.16239 ***

Other types of family -0.0226445 (0.0080331)

(0.0207206) Engineering -0.03347 ***

Female -0.0157498 *** (0.0092743)

(0.0048278) Architecture -0.02588 *

Career position (clerical) Director -0.3324635 *** (0.0137344)

(0.0059069) Medicine -0.10219 ***

Manager -0.2972218 *** (0.0100393)

(0.0044825) Other education 0.139797 ***

Blue collar worker 0.4815742 *** (0.0072574)

(0.0066239) Previous condition (employed) Unemployed previous year 0.053039 ***

Trainee -0.1328942 *** (0.0125812)

(0.0317739) Student previous year -0.06155 ***

Employment relation Part-time 0.0395409 *** (0.0183588)

(0.007273) Not labour force previous year 0.0556

Temporary -0.1092668 *** (0.0432931)

(0.0068519) Other previous year -0.02424

Sector (service) Industry 0.0906355 *** (0.0279634)

Table 1 Dependent variable is overeducation (equal to 1 if overeducated). Categories dropped are in parenthesis. Regional and quarterly dummies not reported

Probit regression reporting marginal results. Observations=119854; Pseudo R2=0. 0.2517; Log pseudo likelihood=-62155.376. Robust standard errors in parenthesis. Marginal effects

calculated at mean values for continuous variables and using discrete differences for dummy variables. Sampling weights have been used. ***/**/* Significance at 1/5/10 per cent.

Page 15: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

15

5.2 Interaction of microeconomic variables

In order to test whether the age effects on overeducation described above are

similar for men and women, in Table 2 we present the results of the interaction

of the age categories with the female dummy – the entire regression results for

this specification are reported in Model 3 in Table 1B in the Appendix. The

results reveal how the higher probability of being overeducated among youths

mostly uniquely concerns female workers. Indeed, after having added the

interaction terms, the coefficients of the age categories up to 35-39 years old get

significantly reduced and they partially lose their statistical significance. By

contrast, the interaction terms between female and the age categories up to 35-

39 years old are all positive and statistically significant. Gender instead does not

seem to play a particular role in explaining overeducation above the age of 40,

probably reflecting how gender based discrimination decreases with age.

We then test whether the age effects on overeducation are related to the

different forms of employment that characterize youths and adults. In particular,

in Italy temporary employment is significantly concentrated among youths – 41

per cent of the workers below the age of 25 in our database holds a temporary

job. At the same time, temporary employment may affect overeducation, since

the short-term nature of the employment relation creates lower incentives for

both the employer and the employee to achieve an appropriate matching. To test

20-24 -0.044739 (0.0113552)

(0.0922506) Female 20-24 0.2847078 ***

25-29 0.0535541 ** (0.0573017)

(0.0228578) Female 25-29 0.1846796 ***

30-34 0.068661 *** (0.0167845)

(0.0191599) Female 30-34 0.1527845 ***

35-39 0.092077 *** (0.0148089)

(0.0182279) Female 35-39 0.1164717 ***

40-44 0.1048212 *** (0.0146354)

(0.0186714) Female 40-44 0.0291873 *

45-49 0.0201378 (0.0154397)

(0.0192299) Female 45-49 0.0303639 *

55-59 -0.0742792 *** (0.0155513)

(0.0204691) Female 55-59 -0.0037173

60-64 -0.1571762 *** (0.016606)

(0.027987) Female 60-64 0.0257851

Female -0.925454 *** (0.0255561)

Table 2: Age effects by gender

Note: Dependent variable is overeducation. The control variables included in the

regression are otherwise the same as in Table 4. Refer to Model 3 in Table 1B in the

Appendix for the entire regression results.

Page 16: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

16

whether the predominance of temporary employment among young workers

explains their higher overeducation, we interact temporary employment with

the age categories.

The results confirm the prediction and show how for workers below the age of

25, their employment as temporary workers is the main driver of overeducation.

By contrast, temporary employment slightly reduces the probability of being

overeducated for workers in the middle of the age distribution. This can be

related to the different nature of temporary employment in the middle of the

career for a worker with an MSc degree – i.e. temporary external consultant.

Quite surprisingly, the interaction of temporary employment and the age

category 60-64 has a positive and statistically significant coefficient. This can be

related to the very atypical nature of temporary employment for the workers of

that age category – only 8 per cent of the workers between 60 and 64 years old.

5.3 The role of unemployment in explaining overeducation The paper finally looks at whether labour market conditions affect the likelihood

of being overeducated. The underlying hypothesis is that an increase in the

unemployment rate decreases the set of opportunities available to a worker,

increasing in this way the probability of accepting a job below the acquired level

of education. In order to account for the possible differences in the effect of

20-24 0.0675125 (0.0305698)

(0.092936) Temporary 20-24 0.1402548 *

25-29 0.1246447 *** (0.079654)

(0.0199708) Temporary 25-29 0.05316

30-34 0.1531035 *** (0.0338896)

(0.0163466) Temporary 30-34 -0.0309137

35-39 0.1517392 *** (0.0332822)

(0.0158178) Temporary 35-39 -0.0798889 **

40-44 0.1207718 *** (0.0334458)

(0.0163121) Temporary 40-44 -0.0787477 **

45-49 0.0355082 ** (0.0349841)

(0.0169869) Temporary 45-49 -0.0430384

55-59 -0.0740685 *** (0.0394297)

(0.018377) Temporary 55-59 0.0032069

60-64 -0.1559045 *** (0.0537485)

(0.026763) Temporary 60-64 0.2235773 ***

Temporary -0.084912 *** (0.0582132)

Table 3: Age effects by employment contract

Note: Dependent variable is overeducation. The control variables included in the regression are

otherwise the same as in Table 4. Refer to Model 4 in Table 1B in the Appendix for the entire

regression results.

Page 17: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

17

unemployment on overeducation across age groups, unemployment rate is

interacted with the age categories. Indeed, the quality of job matches is likely to

be influenced by labour market conditions substantially more for workers that

are entering the labour market with respect to their colleagues with longer job

tenures. Employment rate is included as control variable in order to take into

account other macroeconomic fluctuations.

The results reveal how unemployment rate has a small, negative and generally

statistically not significant effect on overeducation. However, the coefficient of

the interaction terms between unemployment and the age group 20-24 is

positive and statistically significant in all the different specifications – from

Model 2 to Model 6 in Table 1B in the Appendix. Moreover, Wald tests reject the

null hypothesis that the sum of the coefficients of unemployment rate and its

interaction with the age group 20-24 is equal to zero. The coefficient of the

interaction between unemployment rate and the age group 25-29 is also positive

and statistically significant, but the coefficient is considerably smaller.

Page 18: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

18

Age categories (50-54) 20-24 0.1390118 * Subject of study (human science) Social sciences 0.2968321 ***

(0.0728929) (0.0075512)

25-29 0.1527348 *** Languages 0.0852359 ***

(0.0188564) (0.0085021)

30-34 0.1516376 *** Economics-Statistics 0.2780281 ***

(0.0162025) (0.006792)

35-39 0.1481328 *** Law 0.1963445 ***

(0.0157488) (0.0079811)

40-44 0.1182333 *** Sciences -0.1623382 ***

(0.0162447) (0.008033)

45-49 0.0343125 ** Engineering -0.0332193 ***

(0.0169377) (0.009274)

55-59 -0.0740966 *** Architecture -0.0255594 *

(0.0182885) (0.013744)

60-64 -0.1392787 *** Medicine -0.1020426 ***

(0.0268741) (0.0100512)

Personal characteristics Female -0.0153279 *** Other education 0.1399372 ***

(0.0048227) (0.007257)

No citizen 0.1364267 *** Previous condition (employed) Unemployed previous year 0.0514686 ***

(0.0146993) (0.0126076)Single -0.0243246 *** Student previous year -0.0621656 ***

(0.0054378) (0.0182742)

Other types of family -0.0229392 Not labour force previous year 0.0572343

(0.0207357) (0.0433064)

Occupational status (clerical) Director -0.3327083 *** Other previous year -0.023703

(0.005912) (0.0279473)

Manager -0.2976268 *** Macroeconomic variables Employment rate 0.0023673

(0.0044844) (0.002708)

Blue collar worker 0.4814957 *** Unemployment rate -0.0017981

(0.0066321) (0.0032794)

Trainee -0.1337331 *** Unemployment 20-24 0.024098 **

(0.0316647) (0.009458)

Employment relation Part-time 0.0392702 *** Unemployment 25-29 0.0060054 ***

(0.0072902) (0.0022888)

Temporary -0.1084648 *** Unemployment 30-34 0.0014238

(0.0068627) (0.0019549)

Sector (service) Industry 0.0903209 *** Unemployment 35-39 0.0012109

(0.0069306) (0.0018497)

Agriculture 0.1107724 *** Unemployment 40-44 -0.0010618

(0.0341727) (0.0018718)

Firm size (1-10) Firm 11-19 0.0200799 ** Unemployment 45-49 0.0013471

(0.0086571) (0.0018728)

Firm 20-49 -0.0747855 *** Unemployment 55-59 0.0049242 **

(0.0072923) (0.0019891)

Firm 50-250 -0.1542278 *** Unemloyment 60-64 0.0047017

(0.0068583) (0.0030279)

Firm >250 -0.0909885 ***

(0.0079419)

Probit regression reporting marginal results. Observations=119854; Pseudo R2=0. 0.2519; Log pseudo likelihood=-62141.666. Robust standard errors in parenthesis. Marginal effects calculated

at mean values for continuous variables and using discrete differences for dummy variables. Sampling weights have been used. ***/**/* Significance at 1/5/10 per cent.

Table 4 Dependent variable is overeducation (equal to 1 if overeducated). Categories dropped are in parenthesis. Regional and quarterly dummies not reported

Page 19: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

19

6. Conclusions

This paper has analysed the determinants of overeducation in the Italian labour

market for workers with an MSc degree using data from the National Labour

Force Survey during the 2006-2011 period.

Most results related to the microeconomic variables are in line with those

obtained by previous studies, although the conclusions are still relevant due to

the scarcity of studies on overeducation in Italy (Di Pietro and Urwin, 2003). In

particular, youths are more likely to be overeducated than adults and this effect

is particularly important for female and temporary young workers. The paper

also confirms that previous conditions in the labour market affect current

matching, with overeducation positively correlated with previous

unemployment status and negatively correlated with previous student status. An

important extension of this study is related to the analysis of the effect of the

subject of study on overeducation. The results reveal how workers who have

obtained a more scientific (e.g. engineering, scientific sciences) or a very specific

(e.g. medicine, architecture) degree are less likely to be overeducated. This result

provides for a more accurate answer to the generic question faced by individuals

on whether their investment in higher education pays off.

Finally, the paper has analysed the role of labour market conditions on

overeducation and it finds evidence of a positive effect of unemployment on

overeducation for young workers below the age of 25. This result seems to

provide evidence for an asymmetric effect of negative labour market conditions

on the quality of job matches. Possible policy considerations related to this result

may concern the design and implementation of active labour market policies

aimed at favouring the school-to-work transition of graduates entering in the

labour market (e.g. job matching). The government might be willing to focus

these policies during economic downturns or in regions characterized by high

unemployment rates. Longer unemployment benefit schemes that encourage

looking for an appropriate job match may also limit overeducation at the

beginning of the career.

Page 20: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

Number Percentage Number Percentage

Age category 20-24 597 0.3% Professional status Director 15,262 12.3%

25-29 13,950 8.0% Manager 37,122 29.9%

30-34 26,248 15.1% Clerical worker 65,861 53.1%

35-39 30,255 17.4% Blue-collar worker 5,442 4.4%

40-44 27,251 15.7% Trainee 377 0.3%

45-49 24,528 14.1% Employment relation Full-time 153,597 88.3%

50-54 23,390 13.4% Part-time 20,341 11.7%

55-59 19,629 11.3% Permanent 197,608 92.3%

60-64 8,090 4.7% Temporary 16,435 7.7%

Personal characteristics Female 88,772 51.0% Subject of study Human sciences 19,476 11.9%

Male 85,166 49.0% Social sciences 10,594 6.5%

Foreign citizen 6,970 4.0% Law 21,191 13.0%

National citizen 166,968 96.0% Economics/Statistics 25,631 15.7%

Single 32,236 19.2% Scientific sciences 19,235 11.8%

Couple 133,619 79.7% Engineering 19,937 12.2%

Other type of family 1,885 1.1% Architecture 8,875 5.4%

Economic sector of activity Agriculture 1,449 0.8% Medicine 14,980 9.2%

Industry 19,007 10.9% Other degrees 23,279 14.3%

Service 153,482 88.2% Condition previous year Unemployed 6,021 3.5%

Firm size <10 28,287 20.8% Employed 163,369 94.3%

11-19 employees 15,132 11.1% Student 2,708 1.6%

20-49 employees 28,363 20.8% Other 1,086 0.6%

50-250 employees 43,385 31.9%

More than 250 employees 20,937 15.4%

Table 1A Descriptive statistics for the population between 20-64 with an MSc degree

Page 21: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

20-24 0.2662959 *** 0.1390118 * -0.044739 0.0675125 0.1154219 0.1393574 *

25-29 0.1903321 *** 0.1527348 *** 0.0535541 ** 0.1246447 *** 0.145333 *** 0.1529568 ***

30-34 0.1619348 *** 0.1516376 *** 0.068661 *** 0.1531035 *** 0.1479009 *** 0.1518206 ***

35-39 0.1570213 *** 0.1481328 *** 0.092077 *** 0.1517392 *** 0.1476106 *** 0.1480166 ***

40-44 0.1120436 *** 0.1182333 *** 0.1048212 *** 0.1207718 *** 0.116242 *** 0.1178334 ***

45-49 0.0442103 *** 0.0343125 ** 0.0201378 0.0355082 ** 0.0304436 * 0.0342881 **

55-59 -0.0367968 *** -0.074097 *** -0.0742792 *** -0.0740685 *** -0.077175 *** -0.0735186 ***

60-64 -0.1044033 *** -0.139279 *** -0.1571762 *** -0.1559045 *** -0.1453259 *** -0.1390101 ***

20-24 female 0.2847078 ***

25-29 female 0.1846796 ***

30-34 female 0.1527845 ***

35-39 female 0.1164717 ***

40-44 female 0.0291873 *

45-49 female 0.0303639 *

55-59 female -0.0037173

60-64 female 0.0257851

Female -0.0152305 *** -0.015328 *** -0.0925454 *** -0.0151413 *** -0.0139219 -0.0182053 ***

No citizen 0.1362358 *** 0.1364267 *** 0.1302172 *** 0.1386206 *** 0.1341353 *** 0.1365823 ***

Single -0.0240502 *** -0.024325 *** -0.0182479 *** -0.0242351 *** -0.0266215 *** -0.0239429 ***

Other types of family -0.0226776 -0.022939 -0.0216014 -0.0227205 -0.0239355 -0.0232689

Director -0.332519 *** -0.332708 *** -0.3427779 *** -0.3342707 *** -0.3341278 *** -0.3330507 ***

Manager -0.2973528 *** -0.297627 *** -0.3005567 *** -0.2982952 *** -0.2987212 *** -0.2977066 ***

Blue collar worker 0.4815727 *** 0.4814957 *** 0.4840649 *** 0.4815768 *** 0.4822032 *** 0.48179 ***

Trainee -0.1333408 *** -0.133733 *** -0.1350213 *** -0.1573886 *** -0.133327 *** -0.1335614 ***

Director female

Manager female

Blue collar worker female

Trainee female

Firm 11-19 0.0198933 ** 0.0200799 ** 0.0238443 *** 0.0205507 ** 0.016252 * 0.0203505 **

Firm 20-49 -0.074936 *** -0.074786 *** -0.0693815 *** -0.074317 *** -0.0776483 *** -0.0745468 ***

Firm 50-250 -0.1543631 *** -0.154228 *** -0.1489313 *** -0.1537694 *** -0.1554977 *** -0.1539925 ***

Firm >250 -0.0910563 *** -0.090989 *** -0.0871389 *** -0.0914261 *** -0.0927908 *** -0.0908761 ***

Part-time 0.0397372 *** 0.0392702 *** 0.03694 *** 0.0409218 *** 0.0421094 *** -0.0043226

Part-time female 0.0519615 **

Temporary -0.1091496 *** -0.108465 *** -0.1149274 *** -0.084912 *** -0.1094127 *** -0.1075495 ***

Temporary 20-24 0.1402548 *

Temporary 25-29 0.05316

Temporary 30-34 -0.0309137

Temporary 35-39 -0.0798889 **

Temporary 40-44 -0.0787477 **

Temporary 45-49 -0.0430384

Temporary 55-59 0.0032069

Temporary 60-64 0.2235773 ***

Social sciences 0.2966998 *** 0.2968321 *** 0.2911389 *** 0.295454 *** 0.293918 *** 0.2965805 ***

Languges 0.0851126 *** 0.0852359 *** 0.0847683 *** 0.0856845 *** 0.0822339 *** 0.0848096 ***

Economics-Statistics 0.2778641 *** 0.2780281 *** 0.2760245 *** 0.2767232 *** 0.2912766 *** 0.2776277 ***

Law 0.1961233 *** 0.1963445 *** 0.190504 *** 0.1948342 *** 0.2151909 *** 0.19622 ***

Sciences -0.1622704 *** -0.162338 *** -0.1657393 *** -0.1640022 *** -0.0675345 *** -0.1627132 ***

Engineering -0.0331814 *** -0.033219 *** -0.0296776 *** -0.0343503 *** -0.0250111 * -0.0340419 ***

Architecture -0.0258143 * -0.025559 * -0.0327165 ** -0.0244297 * 0.0049152 -0.0254183 *

Medicine -0.1019428 *** -0.102043 *** -0.1107441 *** -0.1034551 *** -0.172262 *** -0.1022281 ***

Other education 0.1400429 *** 0.1399372 *** 0.135769 *** 0.138757 *** 0.0481838 *** 0.1397926 ***

Social science female 0.0083019

Languages female 0.0028933

Economics-Statistics female -0.0338179 **

Law female -0.0387611 **

Sciences female -0.1556326 ***

Engineering female -0.0520891 **

Architecture female -0.060506 **

Medicine female 0.1459181 ***

Other education female 0.1473159 ***

Unemployed previous year 0.0533527 *** 0.0514686 *** 0.0511004 *** 0.0495423 *** 0.0500733 *** 0.0516332 ***

Student previous year -0.0610797 *** -0.062166 *** -0.0509131 *** -0.0798217 *** -0.0618537 *** -0.0612817 ***

Not labour force previous year 0.0552413 0.0572343 0.0534599 0.0502196 0.0574696 0.0554576

Other previous year -0.0244333 -0.023703 -0.0290015 -0.0248903 -0.0238251 -0.0241656

Industry 0.0907206 *** 0.0903209 *** 0.0901992 *** 0.0896821 *** 0.088207 *** 0.0898788 ***

Agriculture 0.1102183 *** 0.1107724 *** 0.1109193 *** 0.1115381 *** 0.1307082 *** 0.1102366 ***

Employment rate 0.0023673 0.0020448 0.0023358 0.0024398 0.0024037

Unemployment rate -0.001798 -0.0017515 -0.0019701 -0.0013286 -0.001747

Unemployment 20-24 0.024098 ** 0.0238159 *** 0.0209416 ** 0.0263892 *** 0.0239634 **

Unemployment 25-29 0.0060054 *** 0.0055611 ** 0.0057815 ** 0.0067013 *** 0.0059473 ***

Unemployment 30-34 0.0014238 0.0013209 0.0015594 0.0016031 0.0013969

Unemployment 35-39 0.0012109 0.0005814 0.0018237 0.0011114 0.0012

Unemployment 40-44 -0.001062 -0.0010766 -0.0006432 -0.0010859 -0.001039

Unemployment 45-49 0.0013471 0.0012816 0.0014108 0.0016674 0.0013278

Unemployment 55-59 0.0049242 ** 0.0051177 ** 0.0049383 ** 0.0048999 ** 0.0048632 **

Unemloyment 60-64 0.0047017 0.0047606 0.0056523 * 0.0048806 0.0046759

Model 6

Table 1B Dependent variable is overeducation (equal to 1 if overeducated). Model 1 is equal to Table 1 in the paper; Model 2 to Table

4 . Other models present interaction of microeconomic variables discussed in the text. Regional and quarterly dummies not reported

Model 1 Model 2 Model 3 Model 4 Model 5

Pseudo R2: 0.2520

Observations:

119845

Observations:

119846

Observations:

119848

Observations:

119850

Observations:

119852

Observations: 119854

Pseudo R2: 0.254 Pseudo R2: 0.2545 Pseudo R2: 0.2527 Pseudo R2: 0.2561 Pseudo R2: 0.2496

Page 22: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

Age categories (50-54) 20-24 0.11977 *** Subject of study (human science) Social sciences 0.2280723 ***

(0.0326864) (0.0066149)

25-29 0.1062568 *** Languages 0.0640241 ***

(0.0148828) (0.0077361)

30-34 0.1506994 *** Economics-Statistics 0.242992 ***

(0.0129789) (0.0057183)

35-39 0.1528434 *** Law 0.1582693 ***

(0.0125742) (0.0070611)

40-44 0.1244839 *** Sciences -0.1294504 ***

(0.0130519) (0.0076826)

45-49 0.0572893 *** Engineering -0.0486209 ***

(0.0138478) (0.0083712)

55-59 -0.0659085 *** Architecture -0.0172139

(0.0159445) (0.012427)

60-64 -0.1095487 *** Medicine 0.2532391 ***

(0.0246553) (0.0054754)

Personal characteristics Female -0.0057045 Other education 0.0606685 ***

(0.0039961) (0.0064561)

No citizen 0.1643596 *** Previous condition (employed) Unemployed previous year 0.0581391 ***

(0.0108794) (0.0102474)

Single -0.008902 * Student previous year -0.0073996

(0.0045388) (0.0141673)

Other types of family 0.0206747 Not labour force previous year 0.0155393

(0.0172173) (0.0357681)

Career position (clerical) Director -0.2792865 *** Other previous year 0.0009343

(0.0058441) (0.023581)

Manager -0.265383 *** Macroeconomic variables Employment rate 0.0019909

(0.0041779) (0.0022732)

Blue collar worker 0.3943744 *** Unemployment rate -0.0021485

(0.0056793) (0.0027571)

Trainee -0.0628508 ** Unemployment 20-24 0.0130575 ***

(0.0267467) (0.004936)

Employment relation Part-time 0.0341 *** Unemployment 25-29 0.0096339 ***

(0.0059685) (0.001849)

Temporary -0.1070939 *** Unemployment 30-34 0.0038041 **

(0.0059792) (0.0016507)

Sector (service) Industry 0.0605196 *** Unemployment 35-39 0.0028166 *

(0.0059462) (0.0015617)

Agriculture 0.0886725 *** Unemployment 40-44 0.0004401

(0.0274922) (0.0015771)

Firm size (1-10) Firm 11-19 0.0146986 ** Unemployment 45-49 0.0005979

(0.007323) (0.0015728)

Firm 20-49 -0.0553496 *** Unemployment 55-59 0.003626 **

(0.006352) (0.0016935)

Firm 50-250 -0.1074838 *** Unemloyment 60-64 0.0024187

(0.006025) (0.0026024)

Firm >250 -0.0288584 ***

(0.0067396)

Table 1C Dependent variable is overeducation (equal to 1 if overeducated). Categories dropped are in parenthesis. Regional and quarterly dummies not reported.

Compared to other regressions, all workers whose educational attainment is equal or above the BA level have been included

Probit regression reporting marginal results. Observations=154166; Pseudo R2=0. 0.2005; Log pseudo likelihood=-94736.69. Robust standard errors in parenthesis. Marginal effects

calculated at mean values for continuous variables and using discrete differences for dummy variables. Sampling weights have been used. ***/**/* Significance at 1/5/10 per cent.

Page 23: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

Bibliography Acemoglu, D. 1999. “Changes in Unemployment and Wage Inequality: An Alternative

Theory and Some Evidence,” American Economic Review, 89 (5): 1259-78.

Addabbo, T. and G. Solinas. 2012. Non-Standard Employment and Quality of Work: The Case of Italy, Physica-Verlag AIEL Series in Labour Economics.

Alba-Ramirez, A. 1993. “Mismatch in the Spanish labor market: overeducation?,” Journal of Human Resources, 28 (2): 259–278.

Albrecht, J. and S. Vroman. 2002. “A Matching Model with Endogenous Skill Requirements,” International Economic Review, 43 (1): 283-305.

Allen, J. and R. Van der Velden. 2001. “Educational mismatches versus skill mismatches: effects on wages, job satisfaction, and on-the-job search,” Oxford Economic Papers, 53 (3): 434–452.

Altonji, J. and R. Blank. 1999. “Race and gender in the labor market,” in O. Ashenfelter and D. Card (eds.) Handbook of Labor Economics, Edition 1, Volume 3, Chapter 48, pages 3143-3259, Elsevier.

Arcidiacono, P.; P. Bayer and A. Hizmo. 2010. “Beyond Signaling and Human Capital: Education and the Revelation of Ability,” American Economic Journal: Applied Economics, American Economic Association, 2 (4): 76-104.

Battu, H.; C.R. Belfield and P. Sloane. 2000. “How Well Can We Measure Over- Education and Its Effects?,” National Institute Economic Review, 171 (1): 82-93.

Battu, H.; P. Sloane; E. Building; D. Street and S. Park. 2004. “Over-education and ethnic minorities in Britain,” Manchester School, 72 (4): 535–559.

Berg, I. 1970. Education and Jobs: The Great Training Robbery, Praeger

CEDEFOP, 2010. “The skill matching challenge: Analysing skill mismatch & policy implications,” European Centre for the Development of Vocational Training.

Chevalier, A. 2003. “Measuring Over-education,” Economica, 70 (279): 509-531.

Chevalier, A. and J. Lindley. 2009. “Overeducation and the skills of UK graduates,” Journal of the Royal Statistical Society: Series A, 172 (2): 307–337.

Collard, F.; R. Fonseca and R. Munoz. 2002. "Spanish Unemployment Persistence and the Ladder Effect," CEP Discussion Papers No. 538, Centre for Economic Performance, LSE.

Clarck, B.; C. Joubert and A. Maurel. 2012. “Overeducation and skill mismatch: a dynamic analysis,” Preliminary Draft.

Di Pietro, G. and P. Urwin. 2003. "Education and Skills Mismatch in the Italian Graduate Labour Market," Royal Economic Society Annual Conference 2003 59, Royal Economic Society.

Page 24: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

Dolado, J.J.; M. Jansen and J.F. Jimeno. 2003. “On-the-Job Search in a Matching Model with Heterogeneous Jobs and Workers,” IZA Discussion Paper No. 886, Institute for the Study of Labor.

Dolton, P. and M. Silles. 2001. “Over-Education in the Graduate Labour Market: Some Evidence from Alumni Data,” CEE Discussion Papers 0009, Centre for the Economics of Education, LSE.

Duncan, G. and S. Hoffman. 1981. “The incidence and wage effects of overeducation,” Economics of Education Review, 1 (1): 75–86.

Frank, R. 1978. “Why women earn less: the theory and estimation of differential overqualification,” American Economic Review, 68 (3): 360–373.

Freeman, R. 1975. “Overinvestment in college training?,” Journal of Human Resources, 10 (3): 287–311.

Freeman, R. 1976. The Overeducated American. Academic Press.

Galasi, P. 2008. “The effect of educational mismatch on wages for 25 countries,” Budapest Working Papers on the Labour Market BWP - 2008/8.

Gautier, P.A. 2002. “Unemployment and Search Externalities in a model with Heterogeneous Jobs and Heterogeneous Workers,” Economica, 69 (273): 21-40.

Green, C., P. Kler, and G. Leeves. 2007. “Immigrant overeducation: Evidence from recent arrivals to Australia,” Economics of Education Review, 26 (4): 420–432.

Green, F. and S. McIntosh. 2007. “Is there a genuine under-utilization of skills among the over-qualified?,” Applied Economics, 39 (4):427–439.

Green, F., McIntosh, S., and A. Vignoles, 1999. Overeducation’ and skills: clarifying the concepts. Technical report.

Groot, W. 1996. “The incidence of, and returns to overeducation in the UK,” Applied Economics, 28 (10): 1345–1350.

Groot, W. and H.M. van den Brink, 2000. “Overeducation in the labor market: a meta-analysis,” Economics of Education Review, 19 (2): 149-158.

Hagedorn, M. and I.Manovskii. 2010. “Job selection and wages over the business cycle,” Mimeo, University of Pennsylvania.

Halaby, C. 1994. “Overeducation and skill mismatch,” Sociology of Education, 67 (1): 47–59.

Hartog, J. 2000. “Over-education and earnings: where are we, where should we go?” Economics of Education Review, 19 (2): 131–147.

Hartog, J. and H. Oosterbeek, 1988. “Education, allocation and earnings in the Netherlands: Overschooling?” Economics of Education Review, 7 (2): 185–94.

Johnson, W. R. 1978. “A theory of job shopping,” Quarterly Journal of Economics, 92 (2): 261–277.

Page 25: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

Jovanovic, B. 1979. “Job matching and the theory of turnover,” Journal of Political Economy, 87 (3): 972–990.

Kahn, L. B. 2010. “The long-term labor market consequences of graduating from college in a bad economy,” Labour Economics, 17 (2): 303–316.

Kiker, B., M. Santos, and M. de Oliveira. 1997. “Overeducation and undereducation: evidence for Portugal,” Economics of Education Review, 16 (2): 111–125.

Leuven, E. and H. Oosterbeek, 2011. “Overeducation and Mismatch in the Labor Market,” IZA Discussion Paper No. 5523, Institute for the Study of Labor.

Lindley, J. and S. McIntosh, 2009. “A Panel Data Analysis of the Incidence and Impact of Over-education,” Sheffield Economic Research Paper Series 2008009, The University of Sheffield.

Liu, K.; K.G. Salvanes and E. Sørensen. 2012. “Good Skills in Bad Times: Cyclical Skill Mismatch and the Long-term Effects of Graduating in a Recession,” IZA Discussion Paper No. 6820, Institute for the Study of Labor.

Manfra, P. 2002. “Entrepreneurship, firm size and the structure of the Italian economy,” Journal of Entrepreneurial Finance, 7 (3): 99-111.

McGuinness, S. and M. Wooden. 2009. “Overskilling, Job Insecurity, and Career Mobility,” Industrial Relations, 48 (2): 265-286.

Ofek, H. and Y. Merrill. 1997. “Labor mobility and the formation of gender wage gaps in local markets,” Economic Inquiry, 35 (1): 28-47.

Oreopoulos, P.; T. von Wachter and A. Heisz. 2012. “The short and long-term career effects of graduating in a recession,” American Economic Journal: Applied Economics, 4 (1): 1–29.

Pellegrino, M. 2002. “Entrepreneurship, firm size and the structure of the Italian economy,” Journal of Entrepreneurial Finance, 7 (3): 99-11

Pissarides, C. 1999. “Search Unemployment with On-the-Job Search,” Review of Economic Studies, 61 (3): 457-475.

Quinn, M. and Rubb, S. 2006. “Mexico’s labor market: The importance of education-occupation matching on wages and productivity in developing countries,” Economics of Education Review, 25 (2): 147–156.

Rubb, S. 2003. “Overeducation: a short or long run phenomenon for individuals?,”

Economics of Education Review, 22 (4): 389–394.

Sharaf, M. F. 2013. “Job-Education Mismatch and Its Impact on the Earnings of Immigrants: Evidence from Recent Arrivals to Canada,” ISRN Economics, Vol 2013.

Sicherman, N. 1991. “Overeducation in the labor market,” Journal of Labor Economics, 9 (2): 101–122.

Sicherman, N. and O. Galor. 1990. "A Theory of Career Mobility," Journal of Political Economy, 98 (1): 169-92.

Page 26: The determinants of overeducation: Evidence from the ...2014.economicsofeducation.com/user/pdfsesiones/058.pdf · Labour Force Survey during the 2006-2011 period. The results confirm

Smith, J. and F. Welch. 1978. “The overeducated American? A review article,” Technical report, Rand Corporation.

Verdugo, R. and N. Verdugo. 1989. “The impact of surplus schooling on earnings: some additional findings,” Journal of Human Resources, 24 (4): 629-643