Metodološkizvezki,Vol. 14,No. 2,2017,19–48
ItalianPh.D.HoldersandMismatchinEducation
andSkills: Empiricalevidence
RosaliaCastellano GennaroPunzo
1
AntonellaRocca
Abstract
Ph.D.educationisakeyelementininnovationandthegenerationofnewknowl-
edge. Nevertheless, in Italy, the share of doctoral graduates is still lower than the
average for OECD member countries. This paper investigates the effectiveness of
doctoraleducationandtheextenttowhichtheItalianlabourmarketproperlyabsorbs
therisingflowofPh.D.holders. Theeffectivenessisassessedfromthetwofoldper-
spectiveoftheformalrelevanceofaPh.D.qualificationinthelabourmarketandthe
substantialapplicabilityofskillsacquiredtodifferentoccupationsinsideandoutside
university. Logit models enable sketches of the main determinants of overeducation
and overskilling among Italian Ph.D.’s, whereas log-earnings equations allow as-
sessmentoftheroleofeducationalandskillsmismatchesintermsofwagepenalties.
Oaxaca-Blinderdecompositionshelpanalysesomecausesofthesemismatches. The
different patterns of overeducation and overskilling among Ph.D. holders working
insideandoutsideacademeleadtodifferentdegreesofpaypenalties.
1 Introduction
In recent years, the increasing expansion of higher education in Europe (OECD, 2013)
has included doctoral education and training (Auriol, 2010). The importance of Ph.D.
education - defined as the third stage of higher education by the Bologna Process (1999)
- has become more explicit in the EU agenda thanks to the role of such education in
contributingtolong-rungrowthandinnovationinaknowledge-basedeconomy(Brinkley,
2006;Leadbeater,1999;Fumasolietal.,2015). Inaddition,becausetheyarespecifically
trainedincertainfieldsofknowledge,doctorateholdersareorientedtocarryoutscientific
researchthatmaycontributetosocialandeconomicdevelopment(Bitusikova,2010).
Although universities traditionally offer doctoral education, other public and private
research-oriented institutions and professional organisations are currently offering Ph.D.
programmes with the aim of establishing a European Higher Education Area (EHEA),
as called for by the ministers of education and university leaders of 29 countries in the
Bologna Process. In the Bologna Seminar titled “Doctoral Programmes for the Euro-
peanKnowledgeSociety”(Salzburg,3-5February2005),Europeanministersemphasised
the importance of research and interdisciplinarity in enhancing the competitiveness of
higher education across EU countries. The 2010 Vienna Declaration officially launched
1
Department of Economic and Legal Studies, University of Naples “Parthenope”, Naples, Italy; gen-
naro.punzo@uniparthenope.it

20 Castellanoetal.
the EHEA, expressing the need for a worldwide dialogue to explore the role of higher
education from a global perspective. In this context, for some years, many governments
have been supporting doctorate courses to increase the supply and enrolment of students
on the one hand and the performance of such programmes on the other hand. Their spe-
cific aim is to attract international talent and to encourage the creation of knowledge in
accordance with the principles of internationality, interdisciplinarity and intersectorality
(European Commission, 2010). Other specific reforms are intended to widen Ph.D. stu-
dents’ skills and abilities and to favour their transition into the private sector, where they
act as agents of creativity and innovation for the benefit of the global society (LERU,
2007). Many strategies have been inspired by the need for doctorate holders to increase
their so-called soft skills (e.g., problem solving, interpersonal and leadership skills, and
criticalreasoning),whichenablethemtomakeadifferenceintheworkplacebyimproving
theircapacityformanagement,teamwork,projectingandfundraising. Inthenearfuture,
in Italy, other reforms will aim to valorise the role of Ph.D. holders in the labour market
and to increase actual career opportunities at the highest levels of public administration.
In addition, in line with the objective of doctoral degree enhancement, recent school re-
formshaverecognisedtheadvantagesforPh.D.holdersintheevaluationofrequirements
forteachinginhighschools,forresearchgrantsandfornationalscientificaccreditation.
The upward trend of doctorate courses and programmes concerns the majority of
OECD(OrganisationforEconomicCo-operationandDevelopment)countries. However,
although employment opportunities and earnings usually increase with education level
(Rehme, 2007; Andersen and wan de Werfhorst, 2010; Castellano and Punzo, 2016),
Ph.D. holders are more likely to experience occupational mismatches. The insufficient
“absorptive capacity” of the national productive structure (Di Paolo and Mañé, 2013)
may mean that the increasing supply of Ph.D. holders is not completely absorbed by the
simultaneouscreationofacademicandresearch-orientedjobsorthattheyaremismatched
in the labour market (Garcia-Aracil and Van der Velden, 2008; Hakala, 2009; Auriol et
al., 2013). Ph.D. holders may be mismatched in the labour market in two respects - ed-
ucation and skills - that, although related, have different analysis and policy implications
(Desjardins and Rubenson, 2011). In fact, as argued by the OECD (2013), “more educa-
tion does not automatically translate into better skills”, which is why a joint analysis of
educationandskills attheindividual level isareliable approachtoprovidea bettercom-
prehensiveunderstandingoftherelationshipbetweeneducationandskillsmismatchesin
thelabourmarket.
This paper addresses the outcomes of the doctoral process in Italy, their impact on
graduate careers and the adequacy of the competencies developed. It assesses the extent
towhichtheItalianlabourmarketproperlyabsorbstherisingflowofPh.D.’sandhowef-
fectivelythesedegreeholdersrepresentkeyelementsforinnovationandthegenerationof
new knowledge in the economy. More precisely, our research hypothesis aims to investi-
gatetheeffectivenessofthedoctoraltrainingprocessfromthetwofoldperspectiveofthe
formalrelevanceofthePh.D.qualificationinthelabourmarketandthesubstantialappli-
cabilityofskillsacquiredfordifferentoccupationsinsideandoutsideacademe. Sincethe
doctoraldegreeisthehighestlevelofformaleducationinItaly-itisrankedastheeighth
level in the new International Standard Classification of Education (ISCED-2011) - this
paper focuses uniquely on upward mismatches in the labour market. It sketches a profile
ofPh.D.holdersatriskofovereducationandoverskillingbyexaminingtheirmaindeter-

ItalianPh.D.HoldersandMismatch... 21
minantsandtheamountbywhichrewardsarelowerforeducationallymismatchedwork-
ersthanfortheirmatchedpeers. Differencesintheprobabilityofexperiencingmismatch
ineducationandskillsaredecomposedinto(i)theendowmenteffect,whichcapturesthe
share due to differences in employee characteristics, and (ii) the return effect, which is
related to the ability of the national system (education vs. labour market) to transform
these characteristics into skills and to reward workers differently for the same individual
endowments. Finally, the decomposition of the wage gap by groups of doctorate holders
allowsustoevaluatethecontributionofeachfactortowagepenalties.
This paper is organised as follows. Section 2 describes the conceptual framework of
overeducationandoverskillinginlightofthemaintheoreticalconstructsthathelpexplain
occupational mismatches. Section 3 analyses the upward trend of Ph.D. education in the
Italian educational system. Section 4 addresses the data and some descriptive statistics
on well-matched/mismatched Ph.D. holders, whereas Section 5 shows the methodology
and the groups of covariates tested. Section 6 discusses the main results, and Section 7
concludes.
2 OccupationalMismatchesbetweenOvereducationand
Overskilling
Inrecentyears,thegrowingdifficultyinmanagingthetransitionfromuniversitytowork
has stimulated an extensive literature on occupational mismatches, especially in terms
of overeducation and overskilling. After a doctoral degree is completed, overeducation
occurs if this high level of qualification exceeds the requirements to obtain a certain job
(Sicherman,1991;Battuetal.,1999;DoltonandVignoles,2000;McGuinnesandBennet,
2007;Bárcena-Martínetal.,2011). Overskillingoccurswhenthecompetenciesacquired
duringthedoctoralprogrammeareuselessinperformingthejob(DoltonandSilles,2008;
McGuinnesandSloane,2011;GreenandZhu,2010;Mavromarasetal.,2013). Inpartic-
ular,iftheskilllevelsofovereducatedworkersarelinkedtojobsatisfaction,thegenuinely
overeducated, who are actually dissatisfied with their occupation, may be distinguished
fromthosewhoareapparentlymerelyovereducated(Chevalier,2003).
Researchershavealsotriedtoinvestigatethepotentialrelationshipsbetweenoveredu-
cationandoverskilling. McGuinnessandBennett(2007),forexample,studiedtheextent
towhichtheincidenceandimpactofovereducationarespecifictoindividualsofparticu-
larabilitylevels,andAllenandVelden(2001)examinedtherelationbetweeneducational
and skill mismatches with different effects on wages and other labour market outcomes
andconcludedthatskillmismatchesaremuchbetterpredictorsofjobsatisfactionandjob
searchesthaneducationalmismatches.
However, both dimensions of job mismatch are rarely assessed as regards in terms of
doctoral education, which remains an under-researched area compared to undergraduate
education. In other words, while increasing attention is being devoted to the matching of
education(andskills)levelwiththejobperformedforworkerswithagraduateandunder-
graduate education, such matching is only marginally assessed for postgraduate workers
(i.e.,master’sdegreeandPh.D.holders).
AsdiscussedbyCaroleoandPastore(2017),humancapitaltheory(LeuvenandOost-

22 Castellanoetal.
erbeek, 2011) and the job competition model (Thurow, 1979) are the main constructs
that help explain overeducation. The first approach holds that overeducation is a signal
of a lack of the work-related component rather than a waste of human capital. The sec-
ond model considers excess schooling a consequence of the competition for jobs in the
presence of the rigidity of the demand for highly educated labour that leads graduates to
accumulateeducation. Theassignmenttheory(Sattinger,1993),whichattemptstorecon-
cilethetwomodels,holdsthatovereducationarisesbecausewagesarenotentirelyrelated
to acquired schooling and other personal characteristics (as in the human capital model)
ortothenatureofthejob(asinthejobcompetitionmodel).
Pioneering studiesof overeducation andoverskilling primarily consideredthe United
States (Freeman, 1976). Subsequent analyses also covered some European countries
(Alba-Ramirez, 1993; Dolton and Vignoles, 2000; Büchel et al., 2003; McGuinness,
2003; 2006; McGuinness and Bennett, 2007; Quintini, 2011; Aleksynska and Tritah,
2013; Verhaest and van der Velden, 2013) and other OECD countries (Manacorda and
Petrongolo,2000;McGowanandAndrews,2015). Muchoftheresearchexploredthede-
terminants of overeducation from a cross-country perspective with many difficulties due
to the lack of comparative data (Leuven and Oosterbeek, 2011; Mavromaras et al., 2013;
Sgobbi and Suleman, 2013). Some studies also assessed the impact of a large range of
individual characteristics on the “effectiveness of the university degree” in providing a
job that matches the individual education and skill levels (Franzini and Raitano, 2012;
Cutillo and Di Pietro, 2006). In particular, Manacorda and Petrongolo (2000) showed
higherovereducationintheEUthanintheUSA,withamoredramaticscenarioinsouth-
ern Europe, especially in Italy, where there has also been extensive growth in the human
capital supply (Cainarca and Sgobbi, 2009). McGowan and Andrews (2015) found that
differences in skill mismatch across OECD countries are related to differences in pub-
lic policies. Much research has also focused on penalties in earnings and employment
prospects (Allen and Velden, 2001; Sloane, 2003; Brynin and Longhi, 2009; Leuven and
Oosterbeek, 2011; Franzini and Raitano, 2012). These studies demonstrated that a large
shareofearningdifferentialsdependsonthemismatchbetweenindividualskillleveland
jobrequirements(McGuinness,2003)andthatthewagepenaltyforoverskillingislower
than that for overeducation (McGuinness and Sloane, 2011). Neumann et al. (2009)
found that earnings are associated with the quality of an employer-employee job match
andthatbetter-matchedworkersareusuallymoreproductiveandreceivehigherearnings.
Nordin et al. (2010), who examined the income penalty for education-occupation mis-
matches for high-educated workers in Sweden, revealed that for mismatched men, the
penaltyisapproximatelytwiceaslargeasfortheirUScounterparts,whereasforwomen,
itisapproximatelythesameasfortheirUSpeers. Basedonthefindingsofotherscholars
(Dolado et al., 2002; Ortiz, 2010), overeducated and/or overskilled workers are affected
bywagepenaltiesbecausetheydonotreachthewageleveltypicallyassociatedwiththeir
qualification, with inevitable consequences related to productivity, job satisfaction and
psychologicalstrain. Moregenerally,theunderutilisationofhumancapital(Feldmanand
Turnley,1995;Feldman,1996)andinefficiencyofpublicexpenditureoneducation(Groot
and Massen van den Brink, 2000; McGuinness, 2006) may cause a waste of resources in
thesocietyasawhole.
Doctoral education is still an under-researched area compared to undergraduate ed-
ucation, and issues related to the overeducation and overskilling of Ph.D. holders have

ItalianPh.D.HoldersandMismatch... 23
rarelybeenassessed. Oneofthemainnovelelementsofthisworkisthatitlinksresearch
on the forces driving occupational mismatches of Ph.D. recipients to the decomposition
oftheprobabilityofbeingovereducatedratherthanoverskilledaswellasthecontribution
ofeachcovariatetothewagegapforspecificgroupsofdoctorateholders.
3 The Upward Trend of Ph.D.’s and their Role in the
LabourMarket
AsdocumentedbyEurostat(http://ec.europa.eu/eurostat),in2004,morethan525000stu-
dentswereattendingaPh.D.courseinEU-25(approximately1.15per1000inhabitants),
accounting for just 3.3% of tertiary students. In the same year, more than 93,000 Ph.D.
students (0.21 per 1000 inhabitants) received their doctoral degree, twice as many as in
theUnitedStatesandsixtimesmorethaninJapan. Inabroaderinternationalcontext,itis
worthstressingtheconstantdevelopmentofhighereducationandresearchsystemsacross
OECDcountries,wherethenumberofadvancedresearchqualificationshasincreasedby
56% over the period 2000–2012. However, doctorate programmes still represent a small
share of all tertiary programmes, even though on average across OECD countries, 1.6%
ofyoungpeoplein2012wereexpectedtoattainthePh.D.degreeovertheirlifetime,com-
pared to only 1% in 2000 (OECD, 2014) and 1.5% in 2009. In addition to Switzerland
(extra-EU),Sweden,Portugal,FinlandandGermanyshowedthehighestgraduationrates
at the doctoral level in 2009 with more than 2.5% of Ph.D. holders, whereas Italy is still
below the OECD average (1.6%) but in line with the EU-27 average. The expansion of
doctorates from 2000 to 2012 is due partially to the increasing presence of women, who
wereawardedonaveragealmosthalf(46%)oftheOECD’snewdoctoratedegreesin2009
(Boarini,2009;OECD,2011). However,in2009,womenrepresentedlessthan40%ofto-
tal Ph.D. recipients in most OECD countries (OECD, 2013), and in 2012, they were less
likely than men to earn an advanced research qualification (OECD, 2014). Italy shows
a tendency to buck the prevailing trend with a higher presence of women in doctorate
programmes.
In the 1980s, the high employment rates (93% on average) for doctorate holders in
most OECD countries, even greater than those for all tertiary graduates (81%), pointed
to the strong attractiveness of Ph.D. graduates as job market candidates, even in times of
economicdownturn (OECD,2013). However, inthe2000s, theoccupational situation of
doctoraterecipientshasbeenlessfavourable. Morespecifically,inthe2000s,employment
ratesforPh.D.holdersdependonthefieldofstudy(e.g.,higherforPh.D.’sinengineering
and social and medical sciences and lower for Ph.D.’s in humanities) and vary consider-
ably across career paths (e.g., the uncertainty of having indefinite contracts is higher for
Ph.D. graduates than for all employees). Particularly in Portugal but also in Germany
and the Netherlands, Ph.D. recipients, especially in humanities, are in precarious and in-
formal situations in the labour market with temporary and/or part-time contracts or even
short-term positions (e.g., postdoctoral positions), which detract from the attractiveness
ofresearchcareers. Nevertheless,over90%ofworkingPh.D.’sareeitherprofessionalsor
managers,especiallyintheeducationsector,withsignificantearningsdifferentialsacross
OECD countries. However, except for France, whose unemployment rates for Ph.D. re-

24 Castellanoetal.
cipients were higher than for graduates at a lower level of education (Harfi and Auriol,
2010), employmentrates for Ph.D.’sare higher(approximately 3 percentagepoints) than
forothertertiary-levelgraduates,confirmingthatemploymentprospectsusuallyimprove
withahigherlevelofeducation.
4 DataSource: APreliminaryAnalysis
Our analysis draws upon the most recent census data (2014) from the Istat Survey on
Doctorate Holders’ Vocational Integration; to be useful to scientists and policy makers,
it is based on information that is as up-to-date as possible. We performed the analysis
on data from the last edition of the survey, which was conducted between February and
July 2014 with doctorate recipients who earned their Ph.D. degree in Italy in 2010. The
mainobjectiveofthesurveyistodetecttheemploymentconditionsofPh.D.’sfouryears
aftergraduation;therefore,2014istheyeartowhichtheinformationrelates
2
. Thesurvey
alsocollectsalargesetofdataonthesubjectiveopinionsofPh.D.’sabouttheireducation
andlabourmarketexperiences,university-to-worktransitionprocess,familybackground
and other personal information. It gains strength by being a total survey — the target
population is composed of 11240 Ph.D. holders — in which weights allow correction
for bias due to potential nonresponses. The survey adopts a mixed approach (Groot and
Massen van den Brink, 2000; Desjardins and Rubenson, 2011; Quintini, 2011) to detect
occupational mismatches of Ph.D. holders. An objective approach allows the evaluation
of matching in education, and a more subjective approach relies on direct questions pre-
sentedtoworkersabouttheirperceptionofmismatchinskills
3
.
Among graduates who earned a Ph.D. in 2010 for whom the employment situation
was evaluated in 2014, those who are overeducated (anyone whose Ph.D. degree was
neither required by law nor useful to the current job) amount to 30.57%. Those who
are overskilled (anyone who does not consider the Ph.D. education effectively neces-
sary to perform the current job) have a more significant share that is nearly twice that of
the overeducated (60.96%). Doctorate holders who are simultaneously overeducated and
overskilled are 29.27% of the total. However, the incidence of mismatched workers who
enteracareeroutsideuniversityishigherthan thatofmismatchedworkerswhocontinue
to work within university: percentages of overeducation are 37.19% outside academe vs.
5.26%inside,whereaspercentagesofoverskillingare73.02%outsidevs. 14.79%inside.
This imbalance in mismatches inside and outside academe draws attention to the inabil-
ity of the private sector to benefit fully from the high potential of individuals who are
specifically trained for research. Only 8.46% of Ph.D.’s are still unemployed four years
after graduation, and they are prevalently women (62.81% vs. 37.19% men). More than
one-fifth of unemployed Ph.D. recipients (22.96%) are waiting to begin a job or remu-
2
This is the second edition of the census survey on doctorate holders’ vocational integration and also
includes Ph.D.’s who earned their doctoral degree in 2008. Istat conducted the first edition of the survey
between December 2009 and February 2010 with Ph.D.’s who had graduated in 2004 and 2006, with the
aimofdetectingtheiremploymentconditionsthreeandfiveyearsaftergraduation.
3
The questionnaire envisages two kinds of questions to evaluate the presence of overeducation and/or
overskilling. They are, respectively, “Was the Ph.D. degree an explicit requirement to access the current
job?” and“Inyouropinion,isthePh.D.educationeffectivelynecessarytoperformthecurrentjob?”

ItalianPh.D.HoldersandMismatch... 25
nerated training; more than one-third (34.04%) are still looking for a job and one-tenth
(10.91%) are seeking a satisfactory job, whereas 10.52% are unable to work for personal
reasons. Theproportionofindividualswhodeclaredtheyhadnotyetfoundasatisfactory
jobisratherbalancedbetweengenders,whereasofthosewhocitedpersonalreasonsthat
preventedthemfromworking,90of100arewomen.
Although overeducation and overskilling are observed exclusively for Ph.D.’s who
were employed at the time of the survey, the selection effect arising from this process
- due to unobservable individual characteristics that are potentially linked to job search
propensity - is not significant; after all, only slightly more than 10% (10.91%) of Ph.D.
graduates are unemployed after four years. The different dynamics of overeducation and
overskilling that characterise Ph.D. holders’ careers inside and outside university justify
their separate treatment. However, in Italy, for most Ph.D. graduates, the professional
prospect is the academic career for which the doctorate degree is typically recognised or
requiredbylaw. Asillustratedbefore,workingatuniversityproducesagreaterguarantee
of equality because of the more standardised job contracts; for example, there is a lower
gender-basedpaygap(7.96%)thanforPh.D.graduateswhoworkoutsideacademe(18%)
(seeSection6).
In general, doctorate holders who begin a professional career outside university earn,
onaverage,morethanothers. Overeducationisslightlymorefrequentforwomen(34.29%)
than for men (26.72%); in detail, 41.09% of women (vs. 32.96% of men) working out-
sideacademeareovereducated,and5.36%ofwomen(vs. 5.15%ofmen)workinginside
academeareovereducated.
Outsideuniversity,overeducatedmalesearn,onaverage,evenmorethanfemaleswho
are well matched in education (Table 1). However, the pay gap — computed as the dif-
ference between the mean incomes of overeducated and well-matched Ph.D.’s compared
to those who are only well matched — is more severe for men (18.99% vs. 13.85%).
Both inside and outside academe, pay differentials are in favour of well-matched Ph.D.
holders, meaning that well-matched individuals earn, on average, more than their overe-
ducated counterparts irrespective of the geographical area of the country in which they
work.
In terms of the field of study, well-matched individuals in education who work in
academe generally earn more than their peers who work outside it. The only exceptions
arethePh.D.’sinhumanitiesandlawwho,evenwhentheyarewellmatchedineducation,
earn, on average, less than their colleagues in other disciplines. In brief, overeducated
Ph.D.’s show some important differences between their mean incomes that are generally
in favour of those who undertake a career outside academe. This evidence is partly in
line with the majority of other European countries and the United States, where Ph.D.
graduates usually receive higher earnings when they do not work as researchers (Auriol,
2010). The largest wage gap (more than 40 percent) exists among doctorate holders who
workinsideacademeinthefieldsofphysical,socialandlifesciences. Incontrast,outside
academe,thewagegapsareusuallylessextensiveexceptforPh.D.’sinlaw.
It is worth noting that, except for Ph.D.’s in law, overeducation is generally more se-
vere for the remuneration of doctorate holders who started an academic profession (e.g.,
their pay gaps are consistently higher than those of their colleagues who work outside
academe). In contrast, overskilling more seriously affects the earnings of Ph.D. hold-
ers working outside university in several fields except for physical, social and life sci-

26 Castellanoetal.
Table1: MeanincomeofPh.Ds. (overeducatedvs. well-matchedineducation)bytypeofcareer(insideandoutside
academe)andbyothermaincharacteristics
Academiccareer Outsideuniversity
Maincharacteristics Overeducated
Well-matched
ineducation
Gap Overeducated
Well-matched
ineducation
Gap
Gender
Female 1102 1886 41.57 1418 1646 13.85
Male 1331 2032 34.50 1625 2006 18.99
Geographicalarea
North-west 1466 2055 28.66 1591 2007 20.73
North-east 1590 1830 13.11 1549 1861 16.77
Centre 1251 2039 38.65 1523 1844 17.41
South 788 1705 53.78 1434 1594 10.04
Islands 2183 1393 1914 27.22
Fieldofstudy
Physicalsciences 1150 2039 43.60 1401 1739 19.44
Lifesciences 1263 2141 41.01 1656 1796 7.80
Engineering 1550 2010 22.89 1583 2014 21.40
Humanities 1686 1833 2356 22.20
EconomicsandStatistics 1700 2132 20.26 1598 1964 18.64
Law 1200 1708 29.74 1177 1814 35.12
Socialsciences 1010 1740 41.95 1228 1451 15.37
Source:Authors’elaborationon2014censusdata

ItalianPh.D.HoldersandMismatch... 27
ences (Table 2). Therefore, overskilling is widespread, mainly outside academe, where
the shares of the overskilled reach 77.66% (vs. 18.06% inside) for women and 68%
(vs. 11.91% inside) for men. Particularly in academic professions, the income penalty is
higherformismatchedwomen-overskilled(34.43%vs. 9.05%)orovereducated(41.57%
vs. 34.50%). Both inside and outside university, the wage gap due to mismatches in
skilling is consistently positive along the entire Italian territory, although it is negligible
inthenortheastandevennegativeintheinsularregionsforPh.D.’swithanacademicca-
reer. An anomaly is represented by overskilled workers who achieved their Ph.D. degree
in the Isles of Italy and are currently pursuing an academic career: they earn approxi-
mately50%morethantheirwell-matchedcolleagues.
Intermsofthefieldofstudy,professionsoutsideacademerewardwell-matchedPh.D.’s
in humanities, economics and statistics, while engineering shows less severe wage gaps
between overskilled and well-matched Ph.D.’s. Inside academe, well-matched doctorate
holders in social, physical and life sciences have higher rewards than their overskilled
colleagues. In the engineering field, mismatched Ph.D.’s who pursue an academic career
showanadvantageintermsofwagescomparedwiththeirwell-matchedpeers.
The distributions of mean incomes and wage gaps of Ph.D.’s who are simultaneously
overeducated and overskilled are not very different from those of Ph.D.’s who are only
overeducated. However, in general, differentials in mean incomes between unmatched
workersappeartobelargerandtherelativewagegapsmoreseverecomparedtotheirpeers
whoareperfectlymatchedinthelabourmarket(Table3). WagedifferentialsacrossItalian
macro regions are even more pronounced. In other words, the wage gaps of mismatched
workers in both dimensions (education and skill) are consistently higher than those of
theircolleagueswhoaremismatchedonlyineducation(Table1)orinskill(Table2). The
mostsignificantexceptionsarethePh.D.’swhoholdadoctorateinlaw,whoappeartobe
severelypenalisediftheyareonlyovereducated.
5 Methodology
Inthefirststep,withtheaimofunderstandingtheleadingdeterminantsunderlyingovere-
ducation and overskilling (y
i
, manifest variables) and the probability that these events
occur among Ph.D. graduates (y
∗
i
, latent variables) who have been pursuing a career in-
side or outside academe, maximum likelihood logit models (Allen, 2000), chosen in the
sphere of binary response models, are performed. These models are tested on a set of
covariates that are grouped as follows: (i) sociodemographic characteristics (gender, co-
habiting, children, age at the date of Ph.D. attainment); (ii) family background (father’s
education level, macroarea of residence); (iii) educational path (final grade at university
degree, type of secondary school attended); (iv) doctoral characteristics (mobility from
regionwherethedegreewasattained,timetoearndoctorate,fieldofstudy);and(v)doc-
toral tutorial path (seminars, laboratory activities, schools, experience abroad, teaching).
Therefore,
y
∗
i
= γz
i
+u
i
whereγ isthevectorofcoefficientsoftherelatedcovariatesz
i
andu
i
istheerrorterms.
Inthesecondstep,totesthowovereducationandoverskillingaffectindividualwages,
log-earnings functions are tested, also controlling for other factors that are likely to ex-

28 Castellanoetal.
Table2: MeanincomeofPh.Ds. (overskilledvs. well-matched)bytypeofcareer(insideandoutsideacademe)andby
othermaincharacteristics
Academiccareer Outsideuniversity
Maincharacteristics Overskilled
Well-matched
inskills
Gap Overskilled
Well-matched
inskills
Gap
Gender
Female 1274 1943 34.43 1467 1855 20.92
Male 1839 2022 9.05 1810 2067 12.43
Geographicalarea
North-west 1415 2056 31.18 1742 2176 19.94
North-east 1788 1828 2.19 1626 2003 18.82
Centre 1351 2074 34.86 1661 1943 14.51
South 1188 1742 31.80 1498 1667 10.14
Islands 3171 2081 −52.38 1542 2156 28.48
Fieldofstudy
Physicalsciences 1487 2064 27.96 1535 1847 16.89
Lifesciences 1620 2164 25.14 1649 2046 19.40
Engineering 2121 2000 −6.05 1839 2024 9.14
Humanities 1502 1706 11.96 1966 2825 30.41
EconomicsandStatistics 2121 1688 2146 21.34
Law 1456 1709 14.80 1426 1775 19.66
Socialsciences 1069 1804 40.74 1294 1593 18.77
Source:Authors’elaborationon2014censusdata

ItalianPh.D.HoldersandMismatch... 29
Table3: MeanincomeofPh.Ds. (bothovereducatedandoverskilled)bytypeofcareer(insideandoutsideacademe)andfor
well-matchedinbotheducationandskillandbyothermaincharacteristics
Academiccareer Outsideuniversity
Maincharacteristics
Overeducated
andoverskilled
Well-matched
ineducation
andskills
Gap
Overeducated
andoverskilled
Well-matched
ineducation
andskills
Gap
Gender
Female 1102 1943 43.28 1398 1853 24.55
Male 1203 2206 45.47 1624 2095 22.48
Geographicalarea
North-west 1250 2059 39.29 1577 2215 28.80
North-east 1590 1828 13.02 1524 1987 23.30
Centre 1187 2078 42.88 1521 1953 22.12
South 788 1742 54.76 1413 1672 15.49
Islands – 2081 – 1355 2181 37.87
Fieldofstudy
Physicalsciences 1150 2064 44.28 1385 1868 25.86
Lifesciences 1186 2176 45.50 1640 2051 20.04
Engineering 1550 2000 22.50 1586 2051 22.67
Humanities – 1706 – 1833 2825 35.12
EconomicsandStatistics – 2132 – 1535 2158 28.87
Law 1200 1709 29.78 1177 1775 33.69
Socialsciences 1010 1804 44.01 1220 1593 23.41
Source:Authors’elaborationon2014censusdata

30 Castellanoetal.
plain differences in generating income. Indeed, beyond overeducation or, alternatively,
overskilling, a semi-log functional form of earnings equation is performed for other con-
trol variables, including the same personal background, family background and doctoral
characteristicsandthefather’sprofession:
lnW
i
= βX
i
+
i
whereW
i
isthepersonalwage,X
i
isthecovariatesandβ therelatedcoefficients,and
i
is
theerrorterms. Thecomparisonbetweentheunconditional(wageisexclusivelyregressed
on overeducation or overskilling) and conditional models (including all the covariates)
showstheroleplayedbythesemismatchesinrelationtowages.
Inthethirdstep,toanalyseindepththedeterminantsofdifferentoutcomesonwages
andontheprobabilityofovereducation/overskillingforspecificgroupsofdoctorates,the
threefold Oaxaca-Blinder (OB) decompositions are performed on both the logit and log-
earnings regressions. As argued by Gomulka and Stern (1990), Even and Macpherson
(1990), Yun (2004) and Fairlie (1999; 2005), the OB procedure may also be applied to
nonlinearregressionmodels. Inparticular,weperformtheextensionoftheOBprocedure
developedbyBauerandSinning(2008)tologitmodels.
Toexplorehowmuchofthemeanwagegapisaccountedforbydifferencesinthepre-
dictors,thewagegapisdecomposedintothreemaineffects(WinsboroughandDickinson,
1971;JonesandKelley,1984;DaymontandAndrisani,1984):
Δ
j
= [E(X
j
)−E(X
k
)]β
k
+E(X
k
)
0
(β
j
−β
k
)+[E(X
j
)−E(X
k
)]
0
(β
j
−β
k
) (5.1)
The first component (e.g., the difference in the average for each predictor weighted
bytheslopeofthewageearningsequationforthereferencegroupk)representstheshare
of wage gap that can be explained because of different average characteristics between
two subsets of individuals (e.g., males vs. females; northern vs. southern Italy; hard vs.
soft subjects
4
; overeducated vs. well matched in education; overskilled vs. well matched
in skills). This component is called the endowment effect and reflects the more or less
favourable endowment of observable characteristics measured by the explanatory vari-
ablesforthegroupofreferencecomparedtoeachothergroup. Therefore,theendowment
effectmeasureshowmuchdoctorateholdersfromthejthgroupwouldearndifferentlyif
theyhadexperiencedthesameeducationprocessasthedoctorateholdersofthegroupof
reference.
The second component, obtained as difference in the slopes weighted by the average
ofcharacteristicsofthegroupofreference,amountstotheproportionofwagegaprelated
to different production processes (that is, the transformation of inputs into educational
achievement) between the two groups under consideration. This component is called the
return effect and reflects more or less efficiency of the group of reference in producing
performance compared to each other group. In other words, the return effect measures
how much doctorate holders from the jth group would earn differently if they had the
sameaveragecharacteristicsasthedoctorateholdersofthereferencegroup.
4
Todistinguishthetypesoffieldofstudyinrelationtotheirquantitativecontent,wegroupedphysical,
lifeandengineeringsciencesashardsubjects,andhumanity,social,economicandstatistic,andlawsciences
assoftsubjects.

ItalianPh.D.HoldersandMismatch... 31
The third component, the interaction effect, is the residual part of decomposition and
captures the leverage produced by both effects occurring simultaneously. Standard er-
rorsofindividualcomponentsarecomputedaccordingtoJann(2008),whichextendsthe
earliermethoddevelopedinOaxacaandRansom(1998)toaddressstochasticregressors.
Regarding probability, Oaxaca and Ransom (1998) proposed a more generalised for-
mulation:
Δ
j
= [E(X
j
)−E(X
k
)]β
∗
+E(X
j
)
0
(β
j
−β
∗
)+E(X
k
)(β
∗
−β
k
)
whereβ
∗
istheweightedaverageofthecoefficientvectorsβ
j
andβ
k
:
β
∗
= Ωβ
j
+(1−Ω)β
k
Ω is the weighting matrix and I is the identity matrix. Therefore, for Ω = 1, the weights
usedaregivenbycoefficientsoftheso-calledadvantagedgroup(thegroupcorresponding
tolowerprobabilityofbeingmismatched),whereasforΩ = 0,theweightsusedaregiven
bycoefficientsofthedisadvantagedgroup(thegroupwithhigherprobability). Following
thisapproach,theOBdecompositionrepresentsaspecialcaseofthisgeneralisedequation
in which Ω is the null or the identity matrix so that it assumes the general simplified
formulationexpressedin(5.1).
6 MainResults
Beforediscussingthemainresultsofouranalysis,itisnecessarytoreflectupontheissue
ofpotentialendogeneitythatmayoccur,especiallyinthecaseoflog-earningsequations.
For this reason, we have also controlled for exogeneity that is the orthogonality of the
regressorsandthedisturbancetermsofthemodels. Theverificationofthisassumptionis
required because in the presence of endogeneity, the use of ordinary least squares (OLS)
leadstobiasedandinconsistentparameters(NakamuraandNakamura,1981).
To validate that the OLS model corresponds to our data, the Durbin-Hausman-Wu
(DHW) test, which evaluates the consistency of an estimator (OLS) when compared to
a less efficient estimator that is already known to be consistent, was conducted (Green,
2012). Inparticular,theDHWtest,whichcanbeusedtocheckfortheendogeneityofone
ormorecovariates,comparesinstrumentalvariableestimatestoOLSestimatesanddeter-
mines whether the errors are correlated (endogenous) or not correlated (not endogenous)
withtheregressors. Thenullhypothesisisthattheyarenot,andinthiscase,thepreferred
modelisOLS.Withap-valueof0.1253,theDHWtestisnotstatisticallysignificant,and
the hypothesis that regressors are uncorrelated with the disturbance terms cannot be re-
jected. Therefore, the OLS models may be considered efficient and consistent. We have
also controlled for all the correlations between each regressor with error terms, and the
correspondingassociationindexesareconsistentlybelow0.05.
Table 4 shows the results of the logit models, estimated separately for Ph.D. holders
who are pursuing a career inside and outside university. The impact of gender on the
probability of being overskilled is higher for Ph.D. holders who work outside universi-
ties and consistently favours men. Cohabiting with a partner is associated with a lower
probabilityofovereducationandoverskillingbutahigherprobabilityofbeingoverskilled

32 Castellanoetal.
outsideuniversity. Youngerdoctorateholdersarelesslikelytobeovereducatedandover-
skilled, whereas living in northern Italy reduces the probability of being overeducated
withinacademiccareersandtheprobabilityofbeingoverskilledoutsideuniversity. Most
likely, doctorate holders have the highest opportunities for adequate jobs in northern and
central Italy. Similarly, doctorate holders who have moved and are currently working
outside the region where their doctorate degree was attained are less likely to be both
overeducated and overskilled. Surprisingly, a higher final grade increases the probability
ofbeingovereducatedandoverskilled,asdoeshavingattendedahighortechnicalschool.
Regarding the field of study, overeducation is less likely for Ph.D.’s in physical and life
sciences,engineeringandlawoutsideuniversityandonlyforPh.D.’sinphysicalsciences
insideacademe.
In general, the qualitative content of Ph.D. courses with respect to schools, seminars,
teaching activities and experience abroad improves individual skills, allowing access to
job positions that are well matched to the educational background, especially outside
academe.
Toassesstheroleplayedbyeducationalandskillmismatchesinwagepenalties(Har-
tog, 2000; McGuinness, 2006), log-earnings equations are estimated in two alternative
specifications, including overeducation and overskilling (Table 5). Here, it is important
tonotethattheestimatedequationsdonotcontrolfortheclassicalhumancapitalcharac-
teristics, such as the educational level, which is the same for all the employees, and job
experience, as workers were analysed four years after they earned the doctorate. In this
light,thePh.D.holders’earningsdependnotonlyonhumancapitalendowment,personal
characteristics and parental background but also on the potential existence of overedu-
cation and overskilling. However, both overeducation and overskilling act negatively on
personal wages, especially inside university. Pay penalties increase (or at least do not
decrease) if the characteristics of doctorate holders are not controlled for (unconditional
estimates), suggesting that the covariates included in the model help identify the main
causesofmismatchedjobs.
Higher rewards are related to male doctorate holders who live in northern and cen-
tral Italy and who attained the doctorate in any field of study related to social areas
(except for humanity sciences inside and outside university and law inside). Generally,
having a higher-level family background, proxied by father’s educational level and pro-
fession, increases the expected wages for careers (e.g., a Ph.D. holder whose father has
a medium/high level of education). However, the results denote rather different earnings
dynamicsbetweenPh.D.holderswhoentercareersinsideoroutsideacademe.
Regarding the analyses of differences in the probabilities of experiencing mismatch
in education and skill (Table 6) as well as differences in wages (Table 7), the estimations
from the logit and log-earnings models have been used to decompose these differences
through extensions of the threefold OB technique in relation to the most important per-
sonalcharacteristics.
Being female and having earned the doctorate degree in southern Italy increase the
probabilityofbeingoverskilledandovereducated. However,thispenaltyisclosetozero,
especially for overeducation, within university, whereas it is more pronounced outside
university. For gender, the return effects are more than double the endowment effects,
denotingthatthemajorityofthisgapisattributabletohowpersonalanddoctoralcharac-
teristics are rewarded and therefore could include a potential discriminatory component.

ItalianPh.D.HoldersandMismatch... 33
Table4: Determinantsofovereducationandoverskillingfouryearsafterdoctorate
Academiccareer Outsideuniversity
Overeducation Overskilling Overeducation Overskilling
Socio-demographiccharacteristics
Gender(1ifmale) −0.15 −2.06 −2.79 −3.79
Cohabiting(1iflivingwithapartner) −0.69 −1.78 −0.78 1.23
Children(1ifwithout) 0.74 0.86 −0.46 −2.06
Age(1if30yearsorless) −1.07 −1.88 −2.67 −0.18
Familybackground
Father’seducationallevel(ref.: low)
Medium(secondaryschool) −1.97 −2.22 −0.42 0.57
High(universitydegree) −0.03 −0.09 −1.19 1.53
Residencearea(ref.: SouthofItalyandIslands)
North-West −0.22 −0.85 1.31 −1.85
North-East −1.03 0.56 0.95 −1.75
Centre 1.03 1.11 −0.86 −1.17
Educationalpath
Finalgrade(ref.:≤104/110)
Medium(105-107) 2.17 1.52 8.79 7.38
High(>107) 2.58 3.98 6.10 6.69
Secondaryschool(ref.: professional)
Highschool 11.11 1.11 2.22 2.13
Technicalschool 10.38 1.56 2.43 2.57
Doctoralcharacteristics
Mobility(1ifmovedtootherregion) −2.47 −3.07 −0.19 −0.31
Timetogetdoctorate(1ifnotontime) 0.21 1.02 0.56 0.58
continued...

34 Castellanoetal.
...continued
Academiccareer Outsideuniversity
Overeducation Overskilling Overeducation Overskilling
Fieldofstudy(ref.: socialarea)
Physicalsciences −0.62 0.12 −1.39 −0.17
Lifesciences 0.42 0.33 −0.62 0.04
Engineering 0.11 −0.33 −1.10 0.33
Humanities 0.66 0.22 0.41 2.11
EconomicsandStatistics 0.19 −1.11 0.16 −0.10
Law – 0.24 −0.25 1.38
Doctoraltutorialpath
Seminars −0.67 −0.83 −0.03 −1.08
Laboratoryactivities 1.70 0.07 −2.86 −3.22
Schools 0.19 0.46 −1.53 0.61
Experiencesabroad −0.13 −3.10 −2.10 −3.02
Teaching −0.11 0.23 −0.97 −0.17
N 567 610 2468 2468
Loglikelihood −140.84 −301.79 −1983 −1738
Source:Authors’elaborationon2014censusdata

ItalianPh.D.HoldersandMismatch... 35
Table5: Determinantsofwage-penalty
Logofnetmonthlyincome
Conditionalestimates Academiccareer Outsideuniversity
Overeducation −0.4697 −0.1573 –
Overskilling −0.2769 – −0.1762
Individualcharacteristics
Gender(1ifmale) 0.0294 0.0123 0.1617 0.1604
Cohabiting(1iflivingwithapartner) −0.0379 −0.0525 0.0013 0.0037
Children(1ifwithout) 0.0466 0.0531 −0.0086 −0.0154
Age(1if30yearsorless) 0.0129 0.0084 −0.0393 −0.0365
Familybackground
Father’seducationallevel(ref.: low)
Medium(secondaryschool) 0.0958 0.0897 0.0407 0.0423
High(universitydegree) 0.1060 0.1079 0.0547 0.0586
Father’sprofession(ref.: elementary)
Legislator,seniorofficialandmanager −0.0084 −0.0224 0.0826 0.0883
Experttechnician −0.0119 −0.0388 0.0196 0.0225
Technician −0.0028 0.0008 −0.0432 −0.0384
Clerkandqualifiedprofession −0.1736 −0.1547 −0.0319 −0.0166
Skilledoperator 0.1212 0.1017 −0.0122 −0.0221
Residencearea(ref.: SouthofItaly)
North-West 0.0719 0.0487 0.1091 0.1065
North-East 0.0063 0.0184 0.0774 0.0644
Centre 0.0884 0.0764 0.0724 0.0739
Doctoralcharacteristics
Mobility(1ifmovedtootherregion) 0.0897 −0.0647 0.0138 0.0530
continued...

36 Castellanoetal.
...continued
Logofnetmonthlyincome
Conditionalestimates Academiccareer Outsideuniversity
Timetogetdoctorate(1ifnotintime) −0.0591 0.1030 −0.0016 0.0029
Fieldofstudy(ref.: socialarea)
Physicalsciences 0.1582 0.1548 0.0655 0.0778
Lifesciences 0.2142 0.2877 0.1313 0.1504
Engineering 0.1365 0.1364 0.1696 0.1851
Humanities −0.0051 −0.0248 −0.0990 −0.0862
EconomicsandStatistics 0.1527 0.1229 0.01801 0.1830
Law −0.0468 −0.0375 0.3122 0.3311
Constant 7.2215 7.2639 7.1638 7.2252
N 355 355 1297 1297
AdjustedR
2
0.1403 0.1306 0.1785 0.1830
Unconditionalestimates
Overeducation −0.4636 – −0.2026 –
Overskilling – −0.3009 – −0.2153

ItalianPh.D.HoldersandMismatch... 37
The greater rewards associated with hard sciences are confirmed by the high return ef-
fects related to the decomposition of gaps in the probabilities of being overeducated and
overskilled, especially outside university. Little differences among these decompositions
arise according to the weighting scheme (Ω = 0,1) as can be seen from the residual part
ofthedecompositioncapturedbytheinteractioncomponent.
These results may be a relevant matter in the debate on policy developments to im-
prove the performance of the doctoral process and to offer the labour market workers
with the most adequate knowledge. The wage gap decompositions in every case show
a high incidence of return effects (Table 7). With the exception of the field of study,
the penalties based on the rewards are always higher for Ph.D. holders working outside
university, suggesting that the existence of different contractual forms and professional
framingsallowmajorsubjectivitythatcanalsohidediscrimination. Thegenderwagegap
for careers outside university is more than double the corresponding one for academic
careers,andoutsideuniversity,theincidenceofthereturneffectisalsodoublethatwithin
university. Conversely, overeducation determines a greater wage gap within university,
probably because in this context, it corresponds to administrative and low-paid jobs. Re-
garding overskilling, surprisingly, those who declare that they suffer from this condition
haveonaveragelowerendowmentsthantheremainingwell-matcheddoctorateholders.
When we examine the specific contribution of variables to the gap (Table 8), when
the groups are identified in relation to gender, the field of study has the most importance
in the endowment effect inside and outside university, while mismatch is important only
outside university. Demographic variables suggest higher characteristics on average for
personslivinginsouthernItalyforalltypesofcareer,whilewithreferencetogenderand
thefieldofstudy,highercharacteristicsexistonlyforcareersoutsideuniversity.
7 Conclusions
Currently,Ph.D.programmesrepresentanessentialmeansofdevelopmentofaknowledge-
based economy. Education and research form the so-called knowledge triangle and play
akeyroleinintroducinginnovation. Thecreationofnewknowledgeandadvanceofeco-
nomicactivitiesaredirectlyrelatedtothecapacitytodrawhumanresourcesintoresearch.
Hence, investing in research and innovation drives the availability of a highly qualified
workforce, which is the primary requisite for stimulating economic growth. In Italy, the
share of doctoral graduates is still lower than in most European countries; nevertheless,
Ph.D.’s obstacles in finding jobs adequate to their skills and competencies, especially
outside academe. This problem is relevant because it encourages the “brain drain” and
the consequent impoverishment of the country in an economic framework in which the
turnoverwithinacademiaisalsoobstructedbythescarcityofresources.
Inthispaper,weproposedananalysisoftheoccupationaloutcomesofItaliandoctor-
ateholdersfouryearsafterthecompletionoftheirPh.D.programmeswithaspecialfocus
on overeducation and overskilling and their potential consequences for wages. Highly
diversified scenarios arise in labour conditions and remuneration, especially when we
distinguish between doctorate holders’ careers within and outside university. These dif-
ferencessuggestedthatthetwosituationsshouldbetreatedseparately. However,someof
thesedifferencesareduetocharacteristicssuchasgender,fieldofstudyandthegeograph-

38 Castellanoetal.
Table6: Decompositionoftheprobabilityofbeingoverskilledandovereducatedforgroupsof
doctorateholdersbasedongender,geographicareaandfieldofstudy
Overeducation Overskilling
Academic
career
Outside
university
Academic
career
Outside
university
Gender(ref.: female)
Ω = 1
EndowmentEffect 0.0010 −0.0217 −0.0143 −0.0191
ReturnEffect 0.0003 −0.0536 −0.0671 −0.0662
Interaction −0.0037 −0.0059 0.0199 −0.0113
Ω = 0
EndowmentEffect −0.0026 −0.0276 0.0056 −0.0304
ReturnEffect −0.0034 −0.0595 −0.0472 −0.0775
Interaction 0.0037 0.0059 −0.0199 0.0113
Gap −0.0024 −0.0812 −0.0615 −0.0966
Area(ref.: SouthItaly)
Ω = 1
EndowmentEffect −0.0059 −0.0293 −0.0317 −0.0115
ReturnEffect −0.0029 −0.0207 0.0211 −0.0361
Interaction −0.0022 0.0151 −0.0145 −0.0012
Ω = 0
EndowmentEffect −0.0082 −0.0143 −0.0463 −0.0127
ReturnEffect −0.0051 −0.0057 0.0065 −0.0373
Interaction 0.0022 −0.0151 0.0145 0.0012
Gap −0.0110 −0.0350 −0.0252 −0.0488
Fieldofstudy(ref.: softsciences)
Ω = 1
EndowmentEffect 0.0049 −0.0464 0.0197 −0.0316
ReturnEffect 0.0101 −0.0653 −0.0157 −0.0861
Interaction −0.0146 0.0150 −0.0230 0.0081
Ω = 0
EndowmentEffect −0.0097 −0.0314 −0.0033 −0.0236
ReturnEffect −0.0045 −0.0503 −0.0387 −0.0780
Interaction 0.0146 −0.0150 0.0230 −0.0081
Gap 0.0004 −0.0967 −0.0190 −0.1097
Source:Authors’elaborationon2014censusdata

ItalianPh.D.HoldersandMismatch... 39
Table7: Decompositionofthewagegapforgroupsofdoctorateholdersbasedongender,geographicarea,fieldofstudy,overeducationand
overskilling
Groups Meanlog
income
Gap
(1)
Endowment
(2)
Return
(3)
Interaction
(4)
[(3)+(4)]/(1)
Gender
Academiccareer
Males 7.5272 0.0905 0.0509 0.0256 0.0140 0.4376
Females 7.4367
OutsideUniversity
Males 7.4672 0.2003 0.0419 0.1494 0.0089 0.7903
Females 7.2669
Area
Academiccareer
North-Centre 7.5055 0.0833 0.0317 0.0693 −0.0177 0.6194
South 7.4222
Outsideuniversity
North-Centre 7.3951 0.0951 0.0127 0.0971 −0.0148 0.8654
South 7.3000
Field
Academiccareer
Hardscience 7.5473 0.2126 0.0119 0.1698 0.0309 0.9440
Softscience 7.3347
Outsideuniversity
Hardsciences 7.4027 0.1801 0.0480 0.1018 0.0303 0.7335
Softsciences 7.2226
continued...

40 Castellanoetal.
...continued
Groups Meanlog
income
Gap
(1)
Endowment
(2)
Return
(3)
Interaction
(4)
[(3)+(4)]/(1)
Overed
Academiccareer
Wellmatched 7.5073 0.4636 0.2457 0.3613 −0.1433 0.4702
Overeducated 7.0437
Outsideuniversity
Wellmatched 7.4342 0.1962 0.0867 0.1133 −0.0038 0.5581
Overeducated 7.2381
Oversk
Academiccareer
Wellmatched 7.5189 0.3009 −0.0643 0.1887 0.1765 1.2137
Overskilled 7.2180
Outsideuniversity
Wellmatched 7.5184 0.2097 0.0773 0.1487 −0.0164 0.6309
Overskilled 7.3088
Source:Authors’elaborationon2014censusdata

ItalianPh.D.HoldersandMismatch... 41
Table8: Contributiontothewagegapofcovariatesforgroupsofdoctorateholdersbasedongender,geographicareaandfieldofstudy
Gender Area Fieldofstudy
Inside
academe
Outside
academe
Inside
academe
Outside
academe
Inside
academe
Outside
academe
Logwage
Prediction_1 Male 7.5272 7.4672 North-Centre 7.5055 7.3951 Hard 7.5473 7.4027
Prediction_2 Female 7.4367 7.2669 South 7.4222 7.3000 Soft 7.3347 7.2226
Gap 0.0905 0.2003 0.0833 0.0951 0.2126 0.1801
EndowmentEffect
Gender – – −0.0010 0.0111 0.0038 0.0240
Demographic 0.0118 −0.0027 −0.0202 −0.0011 0.0097 −0.0095
Geographic −0.0034 0.0086 – – −0.0020 0.0034
Familybackground 0.0023 −0.0008 0.0328 0.0113 0.0065 0.0001
Timetoget 0.0049 −0.0004 0.0040 0.0002 −0.0216 0.0017
Fieldofstudy 0.0264 0.0158 0.0021 −0.0134 – –
Mismatch 0.0088 0.0213 0.0141 0.0047 0.0154 0.0284
Total 0.0509 0.0419 0.0317 0.0127 0.0119 0.0480
ReturnEffect
Gender – – 0.0079 0.0148 0.0470 −0.0070
Demographic −0.0798 −0.0083 0.0133 0.0785 −0.0532 0.0110
Geographic −0.0818 0.0140 – – −0.0735 0.0239
Family-background −0.1218 −0.0634 0.0327 0.0116 0.0738 −0.0350
Timetoget 0.0194 0.0090 −0.0280 0.0104 0.0038 −0.0019
Fieldofstudy 0.3496 −0.0123 −0.3729 −0.0487 – –
Mismatch 0.0155 0.0644 0.0170 0.0363 0.0125 0.0431
Constant −0.0755 0.1461 0.3993 −0.0059 0.1594 0.0677
Total 0.0256 0.1494 0.0693 0.0971 0.1698 0.1018
continued...

42 Castellanoetal.
...continued
Gender Area Fieldofstudy
Inside
academe
Outside
academe
Inside
academe
Outside
academe
Inside
academe
Outside
academe
Interaction
Gender – – 0.0008 −0.0018 −0.0180 0.0020
Demographic −0.0007 0.0029 0.0018 −0.0097 −0.0011 0.0198
Geographic 0.0088 −0.0003 - - 0.0191 0.0024
Family-background 0.0065 0.0019 −0.0361 −0.0132 0.0145 −0.0079
Timetoget −0.0052 0.0007 0.0098 −0.0046 0.0056 −0.0037
Fieldofstudy −0.0125 −0.0005 0.0002 0.0132 – –
Mismatch 0.0170 0.0042 0.0059 0.0014 0.0107 0.0174
Total 0.0140 0.0089 −0.0177 −0.0148 0.0309 0.0303
Source:Authors’elaborationon2014censusdata(Istat)

ItalianPh.D.HoldersandMismatch... 43
ical area where the doctorate degree was earned. In brief, although education and career
pathwaysvary,asdothecareertrajectoriesfollowedafterthePh.D.holdersgraduatefrom
theirdoctoralprogrammes,somecommoncharacteristicscanbeidentified.
The persistence of significant mismatches and wage differentials in the occupations
reservedforthehighest-educatedworkerscoulddenotetheinefficiencyofthelabourmar-
ket in acquiring and valorising the available human capital. In particular, the dynamics
within the academic career tend to contain the onset of inequalities in skills and wages
when the different groups of doctorates are controlled for. Female Ph.D. holders and
those who acquired the doctorate degree in southern Italy are more likely to experience
both overeducation and overskilling even though the dynamics are different within and
outside academe, with greater disparities outside. Furthermore, in assessing the implica-
tionsforwages,wefindthatbeingfemaleincursmorepenaltiesthanbeingovereducated:
on average, well-matched women in education earn less than overeducated men. Dif-
ferent patterns are also identified for the rewards reserved for Ph.D.’s in relation to the
field of specialisation: humanities and social sciences show the lowest wages, followed
byengineeringoutsideuniversity.
As highlighted by the OECD (2014), in Italy, not only are Ph.D.’s often better paid
when they do not work as researchers, but inside university, the availability of tenured
positions is consistently reduced. This situation favours the growth of less stable types
of posts and reduces the attractiveness of research careers due, for example, to the recent
freeze of employee turnover or to wage restraints and cuts. Therefore, to stimulate eco-
nomic growth and produce knowledge and innovation, the challenges for policy makers
concern finalising actions to improve the work conditions and attractiveness of research
careers, to increase the wages paid for doctoral and postdoctoral fellowships, and to im-
prove in the access of doctorate holders to both academic and no academic employment.
In addition, fiscal incentives to attract these highly skilled workers to research could be
anefficaciousinducementforenterprisestoinnovateandtopromoteeconomicgrowth.
Ourresultssuggestextendingtheanalysistoalongerperiodandamorein-depthanal-
ysis of some of the characteristics connected to the outcomes for Ph.D. holders, specifi-
cally those who have still not found a job, controlling for specific sample selection. Fi-
nally, career opportunities within and outside academe should be further investigated to
identify the specific professional profiles needed by the labour market that could be en-
richedbyPh.D.competencies.
Acknowledgements
The authors would like to thank the Editor and the reviewers for their comments and
suggestions on general and technical issues, which have helped to improve the quality of
thepapersignificantly.
References
[1] Alba-Ramirez, A. (1993): Mismatch in the Spanish labor market: Over-education.
JournalofHumanResources,28,259–278.

44 Castellanoetal.
[2] Aleksynska,M.andTritahA.(2013): Occupation-educationmismatchofimmigrant
workersinEurope: Contextandpolicies.EconomicsofEducationReview,36,229–
244.
[3] Allen, D. (2000): Social networks and self-employment, Journal of Socio-
Economics,29,487–501.
[4] Allen, J. and Velden, R. (2001): Educational mismatches versus skill mismatches:
Effectson wages, jobsatisfaction, andon-the-job search. Oxford Economic Papers,
53(3),434-452.
[5] Andersen,R.andvandeWerfhorst,H.G.(2010): Educationandoccupationalstatus
in14countries: Theroleofeducationalinstitutionsandlabourmarketcoordination.
TheBritishJournalofSociology,61(2),336–355.
[6] Auriol,L.(2010): Careersofdoctorateholders: Employmentandmobilitypatterns.
Retrievedfromhttp://dx.doi.org/10.1787/5kmh8phxvvf5-en(13January2018).
[7] Auriol, L., Misu, M. and Freeman, R. A. (2013): Careers of doctorate
holders: Analysis of labour market and mobility indicators. Retrieved from
http://dx.doi.org/10.1787/5k43nxgs289w-en(13January2018).
[8] Bárcena-Martín,E.,Budría,S.andMoro-Egido,A.I.(2011): Skillmismatchesand
wagesamongEuropeanuniversitygraduates.AppliedEconomicsLetters,8,15.
[9] Battu, H., Belfield, C. and Sloane, P. (1999): Overeducation among graduates: A
cohortview.EducationEconomics,7,21–38.
[10] Bauer, T. K. and Sinning, M. (2008): An extension of the Blinder-Oaxaca decom-
positiontononlinearmodels.AdvancesinStatisticalAnalysis,92,197–206.
[11] Bitusikova,A.(2010): DoctoraleducationinEurope: TrendsandperspectivesCom-
municationsinAgriculturalandAppliedBiologicalSciences,75(1),23–27.
[12] Brinkley I. (2006): Defining the Knowledge Economy. Retrieved from
http://bit.ly/2D7pbgB(13January2018).
[13] Büchel, F., De Grip, A. and Mertens, A. (Eds.) (2003): Overeducation in Europe.
Cheltenham: EdwardElgar.
[14] Brynin, M. and Longhi, S. (2009): Overqualification: Major or minor mismatch?
EconomicsofEducationReview,28,114–121.
[15] Cainarca, G. C. and Sgobbi, F. (2009): Quanto paga studiare: La relazione
fraistruzioneeretribuzioneinItalia.EconomiaPolitica,26(1),53–84.
[16] Caroleo, F. E. and Pastore, F. (2017): Overeducation at a glance. Determinants and
wageeffectsoftheeducationalmismatchbasedonAlmaLaureadata.Retrievedfrom
http://bit.ly/2FyKXqV(13January2018).

ItalianPh.D.HoldersandMismatch... 45
[17] Castellano, R. and Punzo G. (2016): Patterns of earnings differentials across three
conservative European welfare regimes with alternative education systems. Journal
ofAppliedStatistics,43(1),140–168.
[18] Chevalier,A.(2003): Measuringovereducation.Economica,70(279),509–531.
[19] Cutillo,A.andDiPietro,G.(2006): TheeffectsofovereducationonwagesinItaly:
A bivariate selectivity approach. International Journal of Manpower, 27(2), 143–
168.
[20] Daymont, T.N. and Andrisani, P.J. (1984): Job preferences, collegemajor, andthe
gendergapinearnings.TheJournalofHumanResources,19,408–428.
[21] Desjardins,R.andRubenson,K.(2011): Ananalysisofskillmismatchusingdirect
measures of skill. Retrieved from http://dx.doi.org/10.1787/5kg3nh9h52g5-en (13
january2018).
[22] Di Paolo, A. and Ma´ né, F. (2013): Determinants and Consequences of Qual-
ification and Skills mismatch among recent Ph.D. Graduates. Retrieved from
http://bit.ly/2AXBXIm(13January2018)
[23] Dolado,J.J.,Garcˇ ea-SerranoC.andJimenoJ.F.(2002): Drawinglessonsfromthe
boomoftemporaryjobsinSpain.TheEconomicJournal,112(480),270–295.
[24] Dolton P. and Silles, M. A. (2008): The effects of overeducation in earnings in the
graduatelabourmarket.EconomicsofEducationReview,27(2),125–139.
[25] Dolton, P. and Vignoles, A. (2000): The incidence and effects of overeducation in
theUKgraduatelabourmarket.EconomicsofEducationReview,19,179–198.
[26] European Commission (2010): Europe 2020 Flagship Initiative Innovation Union.
Retrievedfromhttp://bit.ly/1x4k1mk(13January2018)
[27] Even,W.E.andMacpherson, D.A.(1990): Plantsizeandthedeclineofunionism.
EconomicsLetters,32,393–398.
[28] Fairlie, R. W. (1999): The absence of the African-American owned business: An
analysisofthedynamicsofself-employment.JournalofLaborEconomics,17,80–
108.
[29] Fairlie, R.W. (2005): An extension of the Blinder-Oaxaca decomposition technique
tologitandprobitmodels.JournalofEconomicandSocialMeasurement,30,305–
316.
[30] Feldman,D.C.(1996): Thenature,antecedents,andconsequencesofunderemploy-
ment.JournalofManagement,22,385–407.
[31] FeldmanD.C.andTurnleyW.H.(1995): Underemploymentamongrecentbusiness
collegegraduates.JournalofOrganizationalBehavior,16,691–706.

46 Castellanoetal.
[32] Franzini, M. and Raitano, M. (2012): Few and underutilized? Overeducation of
Italian graduates. In: E. Mandrone E (Ed): Labour economics: PLUS empirical
studies.Roma: ISFOL.
[33] Freeman, R. B. (1976): The overeducated American. New York, NY: Academic
Press.
[34] Fumasoli, T., Goastellec, G. and Kehm, B. M. (2015): Introduction: Understanding
changeintheacademicprofessionthroughtheperceptionsofacademicsandinstitu-
tionalleadership.InT.Fumasoli,G.Goastellec,B.M.Kehm(Eds): AcademicWork
andCareersinEurope: Trends,Challenges,Perspectives,1–12.Cham: Springer.
[35] Garcia-Aracil,A.andVanderVelden,R.(2008): CompetenciesforyoungEuropean
higher education graduates: Labor market mismatches and their payoffs. Higher
Education,55,219–239.
[36] Gomulka,J.andStern,N.(1990): TheemploymentofmarriedwomenintheUnited
Kingdom,1970–1983.Econometrica,57,171–199.
[37] Green,W.H.(2000): EconometricAnalysis.UpperSaddleRiver: PrenticeHall.
[38] Green, F. and Zhu Y. (2010): Overqualification, job dissatisfaction, and increasing
dispersion in the returns to graduate education. Oxford Economic Papers, 62(4),
740–763.
[39] Groot,W.andMassenvandenBrink,H.(2000): Overeducationinthelabormarket:
Ameta-analysis.EconomicsofEducationReview,19(2),149–158.
[40] Hakala, J. (2009): The future of the academic calling? Junior researchers in the
entrepreneurialuniversity.HigherEducation,57(2),173–190.
[41] Harfi,M.andAuriol,L.(2010): TheEmployabilityProblemsofdoctorateHolders:
Explaining a “French Exception”. Retrieved from http://bit.ly/2r4j4nH (13 January
2018).
[42] Hartog, J. (2000): Over-education and earnings: Where are we, where should we
go? EconomicsofEducationReview,19(2),131–147.
[43] Jann, B. (2008): Oaxaca: Stata module to compute the Blinder-Oaxaca decomposi-
tion.Retrievedfromhttp://bit.ly/2FBmAZZ(13January2018).
[44] Jones, F. L. and Kelley, J. (1984): Decomposing differences between groups. A
cautionary note on measuring discrimination. Sociological Methods and Research,
12,323–343.
[45] League of European Research Universities (2007): Doctoral studies in Europe: Ex-
cellence in researcher training. Retrieved from http://bit.ly/2mAO8qi (13 January
2018).
[46] Leadbeater,C.(1999): LivingonThinAir.London: Penguin.

ItalianPh.D.HoldersandMismatch... 47
[47] Leuven, E. and Oosterbeek, H. (2011): Over-education and mismatch in the labor
market. In E. Hanushek, S. Machin and L. Woessmann (Eds): Handbook of the
EconomicsofEducation,283–326.Amsterdam: Elsevier.
[48] Manacorda, M. and Petrongolo, B. (2000): Skill mismatch and unemployment in
OECDcountries.WirtschaftspolitischeBlatter,47(1),72–82.
[49] Mavromaras, K. G., McGuinness, S., O’Leary, N. C., Sloane, P.J. and Wei, Z.
(2013): Job mismatches and labour market outcomes: Panel evidence on univer-
sitygraduates.EconomicRecord,89(286),382–395.
[50] McGowan, M. A. and Andrews, D. (2015): Skill mismatch and public policy in
OECDcountries.Retrievedfromhttp://bit.ly/2DPqoF0(13January2018).
[51] McGuinness, S. (2006): Overeducation in the labour market. Journal of Economic
Surveys,20(3),387–418.
[52] McGuinness S. (2003): Graduate over-education as a Sheepskin effect: Evidence
fromNorthernIreland.AppliedEconomics,35,597–608.
[53] McGuinnessS.andBennettJ.(2007): Over-educationinthegraduatelabormarket:
Aquantileregressionapproach.EconomicsofEducationReview,26,521–531.
[54] McGuinness, S. and Bennett J. (2007): Overqualification and the graduate labour
market: A quantile regression approach. Economics of Education Review, 26(5),
521–531.
[55] McGuinness,S.andSloane,P.J.(2011): LabourmarketmismatchamongUKgrad-
uates: An analysis using REFLEX data. Economics of Education Review, 30, 130–
145.
[56] Nakamura, A.andNakamura, M.(1981): Ontherelationshipsamongseveralspec-
ification error test presented by Durbin, Wu, and Hausman. Econometrica, 49(6),
1583–1588.
[57] Neumann, G. R., Olitsky, N. H. and Robbins, S. B. (2009): Job congruence, aca-
demicachievement,andearnings.LabourEconomics,16(5),503–509.
[58] Nordin, M., Persson, I. and Rooth, D. O. (2010): Education-occupation mismatch:
Isthereanincomepenalty? EconomicsofEducationReview,29,1047–1059.
[59] Oaxaca, R. L. and Ransom, M. (1998): Calculation of approximate variances for
wage decomposition differentials. Journal of Economic and Social Measurement,
24,55–61.
[60] Boarini, R. (2009): Economic Survey of Italy 2009: Towards bet-
ter schools and more equal opportunities for learning. Retrieved from
http://dx.doi.org/10.1787/220864240618(13January2018).
[61] OECD (2011): Building Knowledge. Retrieved from
http://dx.doi.org/10.1787/sti_scoreboard-2011-en(13January2018).

48 Castellanoetal.
[62] OECD (2013): Education at a Glance 2013. Retrieved from http://bit.ly/2qYM8wJ
(13January2018).
[63] Ortiz, L. (2010): Not the right job, but a secure one over-education and temporary
employmentinFrance,ItalyandSpain.Work,Employment&Society,24(1),47–64.
[64] Quintini, G. (2011): Over-qualified and under-skilled: A review of existing litera-
ture.Retrievedfromhttp://dx.doi.org/10.1787/5kg58j9d7b6d-en(13January2018).
[65] Rehme, G. (2007): Education, economic growth and measured income inequality.
Economica,74,493–514.
[66] Sicherman, N. (1991): Overeducation in the labor market. Journal of Labor Eco-
nomics,9(2),101–122.
[67] Sloane,P.(2003): Muchadoaboutnothing? Whatdoestheovereducationliterature
really tells us? In Büchel, F., De Grip, A., Mertens, A. (Eds): Overeducation in
Europe,11–45.Cheltenham: EdwardElgar.
[68] Thurow,L.(1979): Ajob-competitionmodel.InM.Piore(Ed): Unemploymentand
inflation,17–32.WhitePlains,NY:M.E.Sharpe
[69] Sattinger,M.(1993): Assignmentmodelsofthedistributionofearnings.Journalof
EconomicLiterature,31(2),851–880.
[70] Sgobbi, F. and Suleman, F. (2013): A methodological contribution to measuring
skill(mis)match.TheManchesterSchool,81(3),420–437.
[71] Verhaest, D. and Van der Velden, R. (2013): Cross-country differences in graduate
overeducation.EuropeanSociologicalReview,29(3),642–653.
[72] Yun,M.S.(2004): Decomposingdifferencesinthefirstmoment.EconomicsLetters,
82,275–280.
[73] Winsborough, H. H. and Dickinson, P. (1971): Components of negro-white income
difference. In E. G. Goldfield (Ed): Proceedings of the Social Statistics Sections,
6–8.Washington,DC:AmericanStatisticalAssociation.