European Sociological Review, 2022, 38, 904–919

European Sociological Review, 2022, 1–16

https://doi.org/10.1093/esr/jcac004
Advance access publication 7 February 2022 https://doi.org/10.1093/esr/jcac004
Original article Original article

Credential Inflation and Decredentialization:

Re-examining the Mechanism of the Devaluation

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

of Degrees
1,
Satoshi Araki * and Takehiko Kariya2
1
Department of Sociology and Social Policy, Lingnan University, Tuen Mun, New Territories, Hong Kong
and 2Department of Sociology, University of Oxford, Oxford OX1 1JD, UK
*Corresponding author. Email: [email protected]
Submitted August 2020; revised September 2021; accepted January 2022

Abstract
Sociologists have long used credential inflation theory to explain the devaluation of tertiary education
degrees as the consequence of the excessive supply of educated personnel. However, the literature
has inadequately examined two fundamental conditions: the combination of degrees/skills that indi-
viduals possess and the level of degrees. In this article, cross-country multilevel regressions reveal
lower-level degrees (i.e. short-cycle tertiary) are devalued due to the larger extent of lower-level ter-
tiary expansion in a society, regardless of degree holders’ skills level. This is consistent with the con-
cept of credential inflation. In contrast, alongside the proliferation of higher-level tertiary education
(i.e. bachelor and above), individuals with such degrees are penalized only when they lack high skills.
Put differently, higher-level degree holders retain their rewards despite their diminishing scarcity as
long as they possess high skills. Meanwhile, high skills unaccompanied by tertiary degrees lose their
premium merely in connection with lower-level tertiary expansion. These results suggest credential-
ism is intensified and credential inflation operates in societies where the extent of lower-level tertiary
expansion is relatively large, whereas ‘decredentialization’ emerges along with the larger extent of
higher-level tertiary expansion in a way that devalues credentials as such whilst relatively enhancing
the role of skills in reward allocation.

Introduction the heterogeneity among the highly educated, suggesting

The past few decades have witnessed a remarkable ex- that the premium for education and its elasticity result-
pansion of education worldwide. By 2019, approximate- ing from educational expansion differs depending on
ly 38 per cent of 25–64 year-olds in OECD countries fields of study, prestige of education institutions, and
had attained tertiary education (OECD, 2020). In re- socio-economic backgrounds (Kariya, 2011; Bills, 2016;
sponse, the existing literature argues that the association Ortiz and Rodriguez-Menes, 2016; Di Stasio, 2017;
between educational credentials and economic rewards DiPrete et al., 2017; Posselt and Grodsky, 2017; Tholen,
generally weakens as a result of credential inflation (e.g. 2017). In doing so, it offers a counter-argument to func-
Collins, 1979, 2011; Brown, 2001; Bills and Brown, tionalist perspectives, which have long held that higher
2011). Recent work in this vein has further focused on levels of education are linked with higher productivity

© TheAuthor(s)
C The
V Author(s) 2022.
2022. Published
Published by Oxford
by Oxford UniversityUniversity
Press. Press.
This
This isisan
anOpen
Open Access
Access article
article distributed
distributed underunder the terms
the terms of the of the Creative
Creative CommonsCommons
AttributionAttribution License (https://creativecommons.
License (https://creativecommons.org/licenses/by/4.0/),
org/licenses/by/4.0/), which
which permits unrestricted permits
reuse, unrestricted
distribution, reuse, distribution,
and reproduction and provided
in any medium, reproduction in any
the original medium,
work provided
is properly cited. the original work is
properly cited.
European
2 Sociological Review, 2022, Vol. 38, No. 6 9050
European Sociological Review, 2022, Vol. 00, No.

that ensures economic rewards, regardless of whether and the societal level (i.e. the distinction between
education genuinely nurtures ability (i.e. human capital higher-level and lower-level tertiary expansion). Indeed,
theory) (Mincer, 1958, 1974; Schultz, 1961; Becker, as detailed in the following sections, the result of our
1964) or merely signals high potential/trainability (i.e. analysis reveals credentialism is intensified and hence
signalling/labour queue theory) (Spence, 1973; Thurow, the nominal level of degrees, rather than skills, does
1975). matter within educational expansion of lower-level ter-
Although prior research has elucidated the nuanced tiary credentials. In contrast, what we call ‘decredential-

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

structure of progressive devaluations of educational cre- ization’ operates in association with higher-level tertiary
dentials, two important dimensions have remained em- expansion in a way that devalues credentials per se and
pirically elusive. First, while credential inflation has penalizes degree holders only when they lack high skills.
been primarily conceptualized as the consequence of the Put differently, the economic return to tertiary degrees
excessive supply of educated human resources as com- does not simply decline, despite their diminishing scar-
pared to labour demand (e.g. Brown, 2003; Bills and city, so long as these degrees are accompanied by high
Brown, 2011), there has been little empirical evidence skills. These findings, in tandem with the reconceptuali-
focused on labour supply: the potential difference in eco- zation of the approach and its attendant methods we
nomic rewards between those who merely possess high- present here, are of great importance both to social sci-
level credentials unaccompanied by actual high skills ence research and social policy.
(i.e. nominal degree) and those with both high-level cre- As such, this paper revisits credential inflation with
dentials and high skills. This research gap is particularly close attention to (i) whether individuals’ degrees are
critical given that sociologists have traditionally merely nominal or accompanied by high skills and (ii)
explained variance in the devaluation of credentials how differential expansion of higher-level and lower-
among equally qualified individuals (e.g. tertiary gradu- level tertiary education relates to the devaluation of
ates) by investigating social backgrounds (i.e. the advan- individuals’ tertiary degrees. After reviewing previous
taged are more likely than their disadvantaged research and describing hypotheses in the next section,
counterparts to obtain more prestigious educational we outline our data and methods. We then report
qualifications, thus retaining preferable rewards even results of our empirical analysis, followed by further
amongst the expansion of tertiary education) (Tholen, discussion around our intended theoretical contribu-
2017). Consequently, these foci have obscured one pos- tion: ‘decredentialization’.
sibility that heterogeneous returns to degrees may be
attributed to the actual skill level of nominally equiva-
lent degree holders.
Devaluation of Educational Credentials
Second, in terms of the proliferation of tertiary edu- The positive association between educational creden-
cation as a driving force of credential inflation, the lit- tials and economic benefits has been confirmed by a
erature has paid little attention to the distinction vast literature. Regardless of whether higher levels of
between higher-level tertiary (i.e. International Standard education genuinely nurture skills (Mincer, 1958,
Classification of Education: ISCED 2011 Level 6 and 1974; Schultz, 1961; Becker, 1964) or merely signify
above) and lower-level tertiary (i.e. ISCED 2011 Level high potential (Spence, 1973; Thurow, 1975), highly
5). The long-standing approach to analyse the conse- educated people are more likely than less educated
quences of educational expansion has been to quantify counterparts to acquire preferable rewards. This rela-
the percentage of degree holders in a way that conflates tionship may be intensified, from the perspective of so-
both higher-level and lower-level or limits itself to only cial closure, once the privileged exclude others from
higher-level degree holders. As long as both levels of accessing economic assets by using educational creden-
educational expansion demonstrate the same effect on tials and other forms of capital as criteria to hold mem-
the depreciation of degrees, this conflation is not neces- bership within the select community (Murphy, 1988;
sarily cause for concern. However, it is logical to assume Weeden, 2002; Rivera, 2011; Smyth and McCoy,
that these two dimensions differently affect the scarcity 2011; Brown, 2013; Bol and Weeden, 2015; Posselt
and hence the signalling value of individuals’ education- and Grodsky, 2017; Tholen, 2017).
al credentials. However, in paying attention to the value of educa-
One may therefore detect the nuanced mechanism of tional credentials as positional goods, credential infla-
devaluation of degrees by incorporating the aforemen- tion theory has cast doubt on the idea that higher levels
tioned perspectives at both the individual level (i.e. of education automatically result in better economic sta-
whether tertiary degrees are accompanied by high skills) tus (Collins, 1979, 2011; Van de Werfhorst, 2009;
906
European Sociological Review, 2022, Vol. 00, No. 0 European Sociological Review, 2022, Vol. 38, No. 63

Horowitz, 2018). The fundamental contribution of this between overeducation and overskilling, implying skills
line of research has been its focus on the depreciation of levels should be taken into account in analysing the (de)-
educational credentials due to the larger share of edu- valuation of education. In addition, Araki (2020) reveals
cated people in a given society (Bills and Brown, 2011). labour market rewards are allocated based on the com-
This phenomenon can, in theory, be explained by both posite of credentials and skills in a nuanced way, depend-
demand side and supply side issues. ing, in part, on societal-level educational conditions.
Sociologists have historically favoured the demand While this line of work has helped advance under-

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

side story: credential inflation occurs when growth in standing of differentials in economic returns to educa-
the number of people with high credentials outpaces the tion, little attention has been paid to the distinction
increase in job opportunities that require high educa- between higher-level (ISCED 2011 Level 6 and above)
tional attainment (Brown, 2003; Bills and Brown, and lower-level tertiary degrees (ISCED 2011 Level 5).
2011). This has also been investigated as a problem of Indeed, the literature has revealed that the premium
overeducation (Verhaest and Van Der Velden, 2013; Di for individual credentials differs depending on their
Stasio, Bol and van de Werfhorst, 2016; Di Stasio, 2017; levels as well as types (Bills, 2016; Di Stasio, 2017;
Nieto and Ramos, 2017). In a similar vein, recent re- DiPrete et al., 2017). However, when it comes to the
search has further argued returns to education differ de- societal-level expansion of tertiary education, the
pending on fields of study and prestige of education long-standing approach has been to simply quantify
institutions, as well as socio-economic statuses of indi- the percentage of total tertiary degree holders (includ-
viduals (Kariya, 2011; Bills, 2016; Ortiz and Rodriguez- ing both higher-level and lower-level) or that of only
Menes, 2016; Di Stasio, 2017; DiPrete et al., 2017;
higher-level tertiary graduates. This approach may be
Posselt and Grodsky, 2017; Tholen, 2017).
questioned because the proliferation of higher-level
Although impressive findings have emerged within
tertiary education and that of lower-level tertiary edu-
these demand-side studies, the supply side story has been
cation differently affect the scarcity and hence the sig-
inadequately accounted for: higher degrees are signifi-
nalling power of specific credentials in a given society,
cantly devalued by the (lowered) quality of highly edu-
resulting in heterogeneous economic returns to educa-
cated people. That is, the existing literature has long
tion. Specifically, the devaluation of higher-level
attempted to delineate the link between various types of
degrees may happen only when higher-level tertiary
credentials and economic rewards based on the assump-
expansion progresses, whereas lower-level degrees
tion that credentials reflect actual skills (Araki, 2020).
may lose their premium in association with lower-level
Meanwhile, as reviewed, sociological research often
tertiary expansion. Furthermore, such devaluation
explains the heterogeneity in returns to education as a
may be mitigated or even offset when degree holders
product of variances in socio-economic statuses.
Consequently, one fundamental possibility has remained also possess high skills. Nevertheless, if the exclusion
empirically untested: the mechanism of devaluation of of less educated people is strengthened by educational
credentials differs depending on whether or not degree expansion as closure theory argues, higher-level ter-
holders actually possess high skills. For example, one tiary expansion may lead to devaluation of lower-level
may assume individuals with a high degree and high degrees regardless of attendant skill levels.
skills may retain their rewards, whereas those who Filling this research gap would significantly contrib-
merely possess a high degree without high skills (i.e. a ute to better understanding not only the mechanism of
nominal degree) are penalized as societal-level educa- devaluation of credentials as such but also the nuanced
tional expansion continues apace. process of stratification in various societies through the
Indeed, recent studies demonstrate the importance of lens of education and labour market outcomes.
distinguishing between educational attainment and actual However, these aspects have been obscured under the in-
skills. For example, Hanushek and colleagues have fluential notion of credential inflation. To this end, the
revealed skills, rather than years of schooling, play im- current paper re-examines credential inflation via empir-
portant roles in achieving economic success (Hanushek ically analysing the economic return to tertiary degrees
and Woessmann, 2015; Hanushek et al., 2015). with particular attention to (i) whether individuals’
Sociologists have also detected the distinct impact of skills degrees are accompanied by high skills; and (ii) the dis-
and credentials on socio-economic outcomes (Kerckhoff, tinction between higher-level and lower-level tertiary ex-
Raudenbush and Glennie, 2001; Gesthuizen, Solga and pansion. In so doing, based on the concept of credential
Künster, 2011; Doren and Grodsky, 2016). Among inflation and relevant theories reviewed above, we aim
others, Flisi et al. (2017) shed light on the discrepancy to test the following three hypotheses:
European
4 Sociological Review, 2022, Vol. 38, No. 6 9070
European Sociological Review, 2022, Vol. 00, No.

H1: Tertiary degrees are devalued in association with the As returns to degrees are significantly influenced by
expansion of tertiary education (including both lower- both individual and societal level factors, we conduct a
level and higher-level). Yet, degrees accompanied by high cross-sectional multilevel analysis using the PIAAC data
skills are not devalued or at least face relatively small de- and other country-level data. Admittedly, a country-
preciation compared to nominal degrees without high specific analysis would have been effective approach to
skills. Meanwhile, high skills unaccompanied by tertiary delineate the linkage between societal-level educational
degrees are also devalued due to intensified closure. expansion, individual-level degrees, and their economic

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

rewards. In particular, given the nature of educational
H2: In association with lower-level tertiary expansion,
expansion that incorporates changes over time within a
only lower-level degrees are devalued while higher-level
society (and indeed much of the research reviewed previ-
degrees retain their returns. Yet, lower-level degrees
ously has analysed country-specific trends), the longitu-
accompanied by high skills are not devalued or at least
dinal approach focused on particular cases is essential.
face the relatively small depreciation compared to nom-
Nevertheless, country-specific analyses do not necessar-
inal lower-level degrees. Meanwhile, high skills un-
ily provide generalizable findings beyond the boundary
accompanied by tertiary degrees are also devalued due
of states. In this regard, as demonstrated by several pre-
to intensified closure.
vious studies (e.g. Araki, 2020; Heisig, Elbers and Solga,
H3: In association with higher-level tertiary expansion, 2020), cross-country multilevel models with fundamen-
higher-level degrees are devalued, but it is not the case tal societal variables as controls have the potential to re-
(or occurs in a quantitatively reduced manner) for those veal broader trends that might not be detectable within
with high skills. Meanwhile, lower-level degrees (regard- the country-specific approach. Furthermore, given the
less of their skills level) as well as high skills unaccom- fact that PIAAC has been conducted only once in each
panied by tertiary degrees are also devalued due to country (except for the United States) and thus we can-
intensified closure. not undertake internationally comparable longitudinal
analyses incorporating skills measures, the present paper
adopts the cross-country model. Potential problems with
Data and Methods this analytic strategy, as well as possible solutions, are
further discussed below.
We use data from the Programme for the International
As regards the outcome variable, we use the
Assessment of Adult Competencies (PIAAC) conducted
International Socio-Economic Index (ISEI) of
by the OECD, which supports international compari-
Occupational Status, which permits a transparent lin-
sons of analyses of the link between educational attain-
ment, skills, and labour market outcomes. PIAAC is ear model, as distinct from dichotomized measures
primarily composed of assessments of cognitive skills such as skilled/semi-skilled occupations and earnings
and questionnaires for adults aged from 16 to 65 years. quantiles. The continuous measure of earnings would
The major components of the PIAAC assessment are lit- have been an alternative here, but a number of PIAAC
eracy and numeracy, both of which are measured on a countries merely make quantiles data publicly avail-
standardized scale from 0 to 500. The scores are also able to ensure anonymity of participants. Although
linked to the concept of proficiency levels, among which PIAAC does not directly include ISEI either, one may
Level 4 or 5 (equivalent to test score 326 points and use the International Standard Classification of
above) are interpretable as high skills (OECD, 2019). In Occupations (one digit) defined by the International
addition to cognitive skills data, PIAAC questionnaires Labour Organization for all participants to create the
provide wide-ranging variables, such as educational at- scale of ISEI (see Ganzeboom, 2010 for more details
tainment, gender, age, and social backgrounds. For our about ISEI).
analysis, participants in the prime working age (i.e. be- Per predictor variables, educational credentials are
tween 25 and 54 years) are extracted given the possibil- quantified by two dummy variables for the highest edu-
ity that many of the youngest cohort (i.e. aged 16–24 cational qualification attained, namely ISCED 2011
years) are still in education and that the oldest cohort Level 5 (i.e. lower-level degrees) and ISCED 2011
(i.e. aged 55–65 years) is more likely to be affected by Level 6 and above (i.e. higher-level degrees). Another
various work–life experiences including lifelong learn- potential operationalization would have been to divide
ing. As detailed in Table 1, we utilize data for 91,217 degrees not only by their levels but also by fields of
individuals from 26 countries participating in PIAAC study and prestige of education institutions.
between 2011 and 2015. Unfortunately, such data are not available for many
908
European Sociological Review, 2022, Vol. 00, No. 0 European Sociological Review, 2022, Vol. 38, No. 65

Table 1. Target countries (PIAAC round and the number of samples)

Country Round Respondents Country Round Respondents

Austria 1 2,998 Japan 1 2,742

Belgium 1 2,747 South Korea 1 3,777
Canada 1 14,869 Netherlands 1 2,798
Chile 2 2,780 New Zealand 2 3,287
Czech Republic 1 2,957 Norway 1 2,581

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

Denmark 1 3,593 Poland 1 3,298
Finland 1 2,972 Slovak Republic 1 2,971
France 1 3,694 Slovenia 2 2,857
Germany 1 3,118 Spain 1 3,462
Greece 2 2,580 Sweden 1 2,430
Ireland 1 3,580 Turkey 2 2,297
Israel 2 2,601 United Kingdom 1 4,856
Italy 1 2,579 United States 1 2,793
Total number of respondents 91,217

Notes: PIAAC has been conducted three times thus far: Round 1 (2011–2012), Round 2 (2014–2015), and Round 3 (2017). ‘Round’ in the Table is the round of
PIAAC in which each country participated. ‘Respondents’ indicates the number of valid cases used in this analysis.
Source: PIAAC data (https://webfs.oecd.org/piaac/puf-data/) [accessed 8 May 2021].

samples in the current analysis, thus future research also argued that general information-processing skills
must incorporate these heterogeneities. measured by PIAAC operate as the key to economic suc-
To define degrees with/without high skills, standar- cess (Hanushek et al., 2015; OECD, 2016; Araki,
dized scores of cognitive skills (literacy and numeracy) 2020). In addition, PIAAC is the only available dataset
are utilized. Among 10 plausible values (PV) in the embracing adult cognitive skills in conjunction with edu-
PIAAC public use data, the first PV, which has often cational attainment and labour market outcomes, each
been employed to investigate the economic return to of which is crucial for examining the devaluation of
skills (e.g. Hanushek et al., 2015), is extracted to calcu- degrees with/without high skills in an internationally
late the mean score of literacy and numeracy.1 Referring comparable way. We therefore use PIAAC data as the
to the definition of proficiency levels set by the OECD, a foundation for further elaboration.
respondent whose mean score is 326 points and above is To precisely examine the association between cre-
regarded as holding high skills. Combining the measures dential/skill combinations and ISEI, the following
of tertiary degrees and high skills, those who possess individual-level variables are accounted for: gender
both a degree and high skills are assigned 1 as ‘degree (men are assigned 1 with women as the reference),
holders with high skills’, whereas those with simply a cohorts (25–34 year-olds and 35–44 year-olds are
tertiary degree but lacking high skills are assigned 1 as assigned 1, respectively, with 45–54 year-olds as the ref-
‘nominal degree holders’. In addition, to better explain erence), immigration (first-generation immigrants are
the function of credentials including its role as closure, assigned 1 with others as the reference), cultural capital
individuals with high skills unaccompanied by a tertiary (sum of standardized scores of mother’s education,
degree are also assigned a dummy to distinguish them father’s education, and the number of books at home),3
from those who possess neither of them. This strategy is and years of paid work experience. The reference year of
aligned with the concept of four credential/skill combi- individual-level variables is the year when each country
nations proposed by Araki (2020).2 participated in PIAAC (i.e. between 2011 and 2015).
Here, we must bear in mind that skills assessed by In terms of societal-level variables, the extent of edu-
PIAAC represent merely one dimension of broader con- cational expansion refers to the percentage of the popu-
cepts of competences. Indeed, social and emotional skills lation (aged from 25 to 64 years) who have completed
as well as occupation specific skills, which are not tertiary education. As with individual-level degrees,
included in PIAAC, have proved to be promoters of eco- lower-level and higher-level expansions are quantified
nomic rewards (Heckman et al., 2010; Kautz et al., using the classification of ISCED 2011 (i.e. Level 5 for
2014; OECD, 2015). Nonetheless, prior studies have lower-level; Level 6 and above for higher-level). One
European
6 Sociological Review, 2022, Vol. 38, No. 6 9090
European Sociological Review, 2022, Vol. 00, No.

limitation of this approach is that these measures do not work experience), the current model adopts the assump-
directly capture any changes over time, a key dimension tion that degrees/skills affect ISEI.
within educational expansion. Nonetheless, the differ-
Yij ¼ b0j þ b1 Mij þ b2 Að25�34Þ ij þ b3 Að35�44Þ ij þ b4 Iij
ence in the said measure across countries (i.e. the per-
þ b5 Cij þ b6 Wij þ b7 SwoDij þ b8 DwoSij
centage of tertiary graduates at the lower-level and
þ b9 DwSij þ eij :::;
higher-level at one point in time as the status quo) can
(1)
be taken as a quasi-measure for the progress of educa-

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

tional expansion when employing cross-country multi- where i ¼ Level 1 (individual), j ¼ Level 2 (country), Yij
level models. In addition to this approach, one ¼ ISEI for individual i in country j, bn ¼ coefficient of
alternative operationalization is to use the cross-cohort individual-level predictors, Mij ¼ men dummy, A(25–34)ij
variance within countries (i.e. the difference in the share ¼ 25–34-year-old dummy, A(35–44)ij ¼ 35–44-year-old
of tertiary graduates among older versus younger dummy, Iij ¼ first-generation immigrant dummy, Cij ¼
cohorts), which implies the extent to which each country cultural capital, Wij ¼ years of paid work, SwoDij ¼
has enhanced access to higher levels of education, albeit high skills without a tertiary degree, DwoSij ¼ a tertiary
still measured at one point in time. Indeed, recent re- degree without high skills, DwSij ¼ a tertiary degree
search using the PIAAC data to analyse the returns to with high skills, and eij ¼ residual for individual i in
education and skills (e.g. Araki, 2020) employs the country j.
cross-cohort indicator in conjunction with the simple In Model 2, country-level variables and six cross-
cross-country measure to verify the robustness of ana- level interaction terms are added to Model 1. Interaction
lysis results and implications. The current paper there- terms are created based on three individual-level varia-
fore incorporates both the percentage of the entire bles (i.e. SwoD, DwoS, and DwS in equation 1) and two
population with tertiary degrees and its cross-cohort country-level variables (i.e. the extent of educational ex-
variation between those aged from 55 to 64 years and pansion and its cross-cohort difference in each country).
those from 25 to 34 years. In this model, educational expansion is quantified by the
Alongside educational expansion, country-level inde- percentage of the population with total tertiary degrees
pendent variables include GDP (per capita, purchasing including both lower-level and higher-level. These inter-
power parities), union density, and the strength of track- action terms in tandem with main effects show how the
ing in education system. GDP and union density are devaluation of degrees due to educational expansion, if
employed given that economic returns to education are any, differs depending on the possession of high skills.
significantly influenced by macroeconomic factors and In doing so, random slopes for credential/skill combina-
the extent to which people are collectively protected in a tions are incorporated in accordance with Heisig and
given society (Weeden, 2002; Smyth and McCoy, 2011; Schaeffer (2019). This analysis corresponds to H1.
Bol and Weeden, 2015). Furthermore, it is essential to
b0j ðin equation 1Þ ¼ c00 þ c01 Ej þ c02 DEj þ c03 Tj
adjust for tracking as it affects the value of credentials þ c04 Gj þ c05 Uj þ u0j :::; (2a)
considerably in tandem with the labour market (Bol and
b7 ðin equation 1Þ ¼ c70 þ u7j :::; (2b)
van de Werfhorst, 2013; Levels, van der Velden and Di
b8 ðin equation 1Þ ¼ c80 þ u8j :::; (2c)
Stasio, 2014; Di Stasio, Bol and van de Werfhorst, 2016;
Di Stasio, 2017; DiPrete et al., 2017; Bol et al., 2019). and
The reference years of these data range from 2010 to
b9 ðin equation 1Þ ¼ c90 þ u9j :::; (2d)
2014 (see Table 2 for descriptive statistics and details
about reference years). where c00 ¼ average intercept, c0n ¼ coefficient of
Using these individual-level and country-level varia- country-level predictors, Ej ¼ the extent of educational ex-
bles, we conduct the following four models of multilevel pansion, DEj ¼ the cross-cohort difference in the extent of
linear regressions. Model 1 employs only individual- educational expansion, Tj ¼ the index of tracking, Gj ¼
level predictor variables to investigate the overall effects GDP per capita, Uj ¼ union density, u0j ¼ country (j) de-
of tertiary degrees with/without high skills after control- pendent deviation, cn0 ¼ average coefficient of three
ling for social backgrounds. The variables concerning individual-level credential/skill variables, and unj ¼ coun-
degrees here include both lower-level and higher-level. try dependent deviation of the slopes for three credential/
Although there is the possibility of reverse causation be- skill variables. We substitute equations 2(a–d) into
tween credential/skill combinations and ISEI (i.e. people equation 1 and denote bn by cn0. Six cross-level interaction
may enhance their educational attainment and skills via terms between three credential/skill combinations and two
910
European Sociological Review, 2022, Vol. 00, No. 0 European Sociological Review, 2022, Vol. 38, No. 67

societal conditions (i.e. the extent of educational expansion expansion, DLE ¼ the cross-cohort difference in the ex-
and its cross-cohort difference) are also added as follows: tent of lower-level tertiary expansion, and DHE ¼ the
cross-cohort difference in the extent of higher-level ter-
Yij ¼ c00 þ c10 Mij þ c20 Að25�34Þ ij þ c30 Að35�44Þ ij þ c40 Iij
tiary expansion.
þ c50 Cij þ c60 Wij þ ðc70 þ u7j ÞSwoDij Finally, as a robustness check, Model 4 simply omits
þ ðc80 þ u8j ÞDwoSij þ ðc90 þ u9j ÞDwSij þ c01 Ej three country-level measures (i.e. T, G, and U) from
þ c02 DEj þ c03 Tj þ c04 Gj þ c05 Uj þ c71 SwoDij Ej Model 3 given the possible bias caused by the excessive
þ c81 DwoSij Ej þ c91 DwSij Ej þ c72 SwoDij DEj

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

number of Level 2 variables as compared to the sample
þ c82 DwoSij DEj þ c92 DwSij DEj þ u0j þ eij
size at the societal level.
¼ c00 þ c10 Mij þ c20 Að25�34Þ ij þ c30 Að35�44Þ ij þ c40 Iij
þ c50 Cij þ c60 Wij þ c70 SwoDij þ c80 DwoSij
þ c90 DwSij þ c01 Ej þ c02 DEj þ c03 Tj þ c04 Gj
Results
þ c05 Uj þ c71 SwoDij Ej þ c81 DwoSij Ej þ c91 DwSij Ej Table 3 shows the results of our multilevel linear regres-
þ c72 SwoDij DEj þ c82 DwoSij DEj þ c92 DwSij DEj sions. In Model 1 where only individual-level variables
þ u0j þ u7j SwoDij þ u8j DwoSij þ u9j DwSij þ eij :::: are included without distinguishing higher- versus
(2e) lower-level degrees, all the predictors including degrees
In Model 3, individual-level tertiary degrees are clas- with/without high skills demonstrate significant coeffi-
sified into lower-level (ISCED 2011 Level 5) and higher- cients at the 0.1 per cent level (i.e. b7¼0.173, b8¼0.341,
level (ISCED 2011 Level 6 and above). In addition, and b9¼0.430 in Model 1). This means one may expect
country-level educational expansion measures are also higher ISEI by holding a tertiary degree and/or high
distinguished between two levels for both the status quo skills. It is worthy of note that there are statistically sig-
and the cross-cohort difference, thus resulting in four nificant differences in the effect size across these three
educational variables at the societal level. Accordingly, educational variables,4 suggesting degree holders who
12 cross-level interaction terms are generated by multi- also possess high skills are more likely than others to ob-
plying four country-level variables and three credential/ tain higher ISEI, followed by nominal degree holders un-
skill combinations. As with Model 2, these interactions accompanied by high skills and then highly skilled
alongside individual-level educational variables are people without a tertiary degree.
focused on to test Hypotheses 2 and 3: The significant coefficients of individual-level predictors
are observed even after accounting for country-level varia-
Yij ¼ c00 þ c10 Mij þ c20 Að25�34Þ ij þ c30 Að35�44Þ ij þ c40 Iij bles and cross-level interactions between educational expan-
þ c50 Cij þ c60 Wij þ c70 SwoDij þ c80 LDwoSij sion and individuals’ credential/skill combinations (Model
þ c90 LDwSij þ c100 HDwoSij þ c110 HDwSij 2). However, one substantial inconsistency detected is that
þ c01 LEj þ c02 HEj þ c03 DLEj þ c04 DHEj þ c05 Tj the difference in effect size between a tertiary degree with
þ c06 Gj þ c07 Uj þ c71 SwoDij LEj and without high skills becomes insignificant in Model 2
þ c81 LDwoSij LEj þ c91 LDwSij LEj
(i.e. c70 ¼0.311, c80 ¼0.523, and c90 ¼0.553). That is, in line
þ c101 HDwoSij LEj þ c111 HDwSij LEj
þ c72 SwoDij HEj þ c82 LDwoSij HEj with prior studies (e.g. Araki, 2020), the possession of high
þ c92 LDwSij HEj þ c102 HDwoSij HEj educational credentials (i.e. tertiary degrees) plays an essen-
þ c112 HDwSij HEj þ c73 SwoDij DLEj tial role in realising preferable labour market outcomes,
þ c83 LDwoSij DLEj þ c93 LDwSij DLEj whereas the contribution of high skills is tangible only
þ c103 HDwoSij DLEj þ c113 HDwSij DLEj among those without a tertiary degree.
þ c74 SwoDij DHEj þ c84 LDwoSij DHEj Nevertheless, all of three credential/skill combina-
þ c94 LDwSij DHEj þ c104 HDwoSij DHEj tions are likely to be devalued in societies where the
þ c114 HDwSij DHEj þ u0j þ u7j SwoDij
share of tertiary graduates is relatively large. Indeed,
þ u8j LDwoSij þ u9j LDwSij þ u10j HDwoSij
þ u11j HDwSij þ eij ::: three interaction terms between the extent of education-
al expansion (i.e. the percentage of the population with
(3)
tertiary degrees) and the said three individual-level edu-
where LDwoSij ¼ a lower-level tertiary degree without cational variables show negative coefficients and they
high skills, LDwSij ¼ a lower-level tertiary degree with are statistically significant at the 0.1 per cent or 1 per
high skills, HDwoSij ¼ a higher-level tertiary degree cent level (i.e. c71 ¼ �0.004, c81 ¼ �0.005, and c91 ¼
without high skills, HDwSij ¼ a higher-level tertiary de- �0.003). Meanwhile, other interactions between the
gree with high skills, LE ¼ the extent of lower-level ter- cross-cohort difference in the proportion of tertiary
tiary expansion, HE¼ the extent of higher-level tertiary graduates in each country and credential/skill
European
8 Sociological Review, 2022, Vol. 38, No. 6 9110
European Sociological Review, 2022, Vol. 00, No.

Table 2. Descriptive statistics

Variables Mean S.D. Min. Max.

Individual level
International socio-economic index (ISEI) 44.26 15.35 18.00 65.00
Gender (male dummy) 0.49 0.50 0.00 1.00
Cohort: 25–34 years old 0.32 0.47 0.00 1.00
Cohort: 35–44 years old 0.34 0.47 0.00 1.00

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

Cohort: 45–54 years old 0.33 0.47 0.00 1.00
First-generation immigrant 0.12 0.32 0.00 1.00
Cultural capital 0.11 2.29 �3.28 17.14
Years of paid work 17.04 9.61 0.00 47.00
Higher-level degree with high skills 0.08 0.26 0.00 1.00
Higher-level degree without high skills 0.21 0.41 0.00 1.00
Lower-level degree with high skills 0.02 0.12 0.00 1.00
Lower-level degree without high skills 0.12 0.33 0.00 1.00
High skills without tertiary degree 0.03 0.16 0.00 1.00
Country level
Higher-level tertiary expansion 22.02 6.31 12.02 35.15
Lower-level tertiary expansion 9.30 6.48 0.00 24.17
Difference in higher-level tertiary expansion 12.92 6.37 1.08 27.74
Difference in lower-level tertiary expansion 2.53 7.00 �13.69 24.51
Index of tracking �0.01 0.97 �1.31 1.79
GDP per capita (USD, PPP) 35,080.98 9,065.94 20,562.26 57,998.85
Union density 28.48 17.96 7.74 68.61
Observations: individuals ¼ 91,217, countries ¼ 26

Notes: At the individual level, ‘Higher-level degree with high skills’ means the possession of both a higher-level tertiary degree (ISCED 2011 Level 6 and above)
and high skills (the mean score of literacy and numeracy in PIAAC is 326 and above), whereas ‘Higher-level degree without high skills’ is a dummy for those who
hold a higher-level tertiary degree without high skills. ‘Lower-level degree with high skills’ and ‘Lower-level degree without high skills’ are dummies for individuals
who possess a short-cycle tertiary degree (ISCED 2011 Level 5) with/without high skills, respectively. At the country level, ‘Higher-level tertiary expansion’ and
‘Lower-level tertiary expansion’ are the percentage of adults whose highest educational attainment is ‘ISCED 2011 Level 6 and above’ and ‘ISCED 2011 Level 5’, re-
spectively. ‘Difference in higher-level tertiary expansion’ and ‘Difference in lower-level tertiary expansion’ are the difference in ‘Higher-level tertiary expansion’ and
‘Lower-level tertiary expansion’ between two cohorts (aged 55–64 and aged 25–34) in each country. The reference year of individual-level data is when each country
participated in PIAAC (i.e. 2011–2012 for Round 1 countries, 2014–2015 for Round 2 countries). Educational expansion and union density refer to the data in 2010,
whereas the reference year of GDP per capita is 2010 for PIAAC Round 1 countries and 2013 or 2014 for PIAAC Round 2 countries. For the tracking index, see Bol
and van de Werfhorst (2013).
Source: PIAAC data, OECD.Stat (http://stats.oecd.org/) [accessed 8 May 2021], and Bol and van de Werfhorst (2013).

combinations at the individual level do not demonstrate ISECD 2011 Level 6 and above) at both the individual
any substantial signs. This suggests the devaluation of and country levels. In terms of main effects of
degrees primarily emerges in association with the status individual-level credential/skill combinations, clear
quo of the scarcity of such educational assets among the stratification is observed according to degree level.
entire population in each society rather than the extent While the magnitude of high skills unaccompanied by
to which it differs between older and younger age tertiary degrees is relatively small (albeit significantly
groups at one point in time. In other models that follow, positive), degree holders are likely to enjoy higher ISEI
it is also the status quo (i.e. LE and HE), not the cross- regardless of their skills level. In particular, the effect
cohort variation (i.e. DLE and DHE), that indicates the size of higher-level degrees is substantially larger than
significant link with the diminishing return to degrees that of lower-level (i.e. c70 ¼ 0.246, c80 ¼ 0.389, c90 ¼
and high skills. H1 is thus partially supported in that ter- 0.306, c100 ¼ 0.577, and c110 ¼ 0.572). This corrobo-
tiary degrees, as well as high skills unaccompanied by rates the aforementioned argument that credentials, ra-
degrees, are devalued within educational expansion. ther than cognitive skills as such, operate as the key
Moreover, the devaluation of degrees is not mitigated determinant of socio-economic success.
even when they are accompanied by high skills. Herein, cross-level interactions between lower/
In Model 3, tertiary degrees are divided into lower- higher-level tertiary expansion and lower/higher-level
level (i.e. ISECD 2011 Level 5) and higher-level (i.e. degrees with/without high skills indicate a more nuanced
912
European Sociological Review, 2022, Vol. 00, No. 0 European Sociological Review, 2022, Vol. 38, No. 69

Table 3. Multilevel regression of ISEI

Predictor variable Model 1 Model 2

B S.E. B S.E.

*** ***
Intercept 3.463 0.010 3.371 0.061
Level 1 (individual) Men (dummy) 0.013*** 0.002 0.011*** 0.002
25–34 years old 0.045*** 0.004 0.043*** 0.004

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

35–44 years old 0.032*** 0.003 0.031*** 0.003
First-generation immigrant 0.074*** 0.003 0.071*** 0.003
Cultural capital (composite of parental education 0.020*** 0.000 0.019*** 0.000
and books)
Years of paid work 0.005*** 0.000 0.005*** 0.000
High skills without tertiary degree 0.173*** 0.007 0.311*** 0.050
Tertiary degree without high skills 0.340*** 0.002 0.523*** 0.033
Tertiary degree with high skills 0.430*** 0.004 0.553*** 0.035
Cross-level Educational expansion
interactions  High skills without tertiary degree 0.004** 0.001
 Tertiary degree without high skills 0.005*** 0.001
 Tertiary degree with high skills 0.003** 0.001
Difference in educational expansion
 High skills without tertiary degree 0.000 0.001
 Tertiary degree without high skills 0.000 0.001
 Tertiary degree with high skills 0.001 0.001
Level 2 (country) Educational expansion 0.002 0.002
Difference in educational expansion 0.002 0.001
Index of tracking 0.008 0.013
GDP per capita (USD, PPP)/10,000 0.016 0.014
Union density 0.000 0.001

Variance Covariance structure (intercept) 0.002** 0.001

(random effect) High skills without tertiary degree 0.001
Tertiary degree without high skills 0.001**
Tertiary degree with high skills 0.001*
Model fit 2LL 145,621.936 144,885.348
AIC 145,625.936 144,895.348

Model 3 Model 4

Intercept 3.353*** 0.079 3.401*** 0.054

Level 1 (individual) Men (dummy) 0.014*** 0.002 0.014*** 0.002
25–34 years old 0.052*** 0.004 0.052*** 0.004
35–44 years old 0.036*** 0.003 0.036*** 0.003
First-generation immigrant 0.081*** 0.003 0.081*** 0.003
Cultural capital (composite of parental education 0.015*** 0.000 0.015*** 0.000
and books)
Years of paid pork 0.006*** 0.000 0.006*** 0.000
High skills without tertiary degree 0.246*** 0.053 0.250*** 0.052
Lower-level degree without high skills 0.389*** 0.055 0.391*** 0.055
Lower-level degree with high skills 0.306*** 0.054 0.305*** 0.054
Higher-level degree without high skills 0.577*** 0.048 0.576*** 0.048
Higher-level degree with high skills 0.572*** 0.046 0.573*** 0.046
Cross-level interactions Lower-level tertiary expansion
 High skills without tertiary degree 0.006** 0.002 0.006** 0.002
 Lower-level degree without high skills 0.005* 0.002 0.005* 0.002
 Lower-level degree with high skills 0.005** 0.002 0.005** 0.002
 Higher-level degree without high skills 0.003 0.002 0.003 0.002

(continued)
European
10 Sociological Review, 2022, Vol. 38, No. 6 9130
European Sociological Review, 2022, Vol. 00, No.

Table 3. (Continued)
Model 3 Model 4

� Higher-level degree with high skills �0.002 0.002 �0.002 0.002

Higher-level tertiary expansion
� High skills without tertiary degree �0.001 0.002 �0.001 0.002
� Lower-level degree without high skills �0.005* 0.002 �0.005* 0.002
� Lower-level degree with high skills 0.003 0.002 0.003 0.002

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

� Higher-level degree without high skills �0.005* 0.002 �0.005* 0.002
� Higher-level degree with high skills �0.004 0.002 �0.004 0.002
Difference in lower-level tertiary expansion
� High skills without tertiary degree �0.002 0.002 �0.002 0.002
� Lower-level degree without high skills 0.002 0.002 0.002 0.002
� Lower-level degree with high skills 0.002 0.002 0.002 0.002
� Higher-level degree without high skills 0.000 0.002 0.000 0.002
� Higher-level degree with high skills �0.001 0.002 �0.001 0.002
Difference in higher-level tertiary expansion
� High skills without tertiary degree 0.002 0.002 0.002 0.002
� Lower-level degree without high skills 0.000 0.002 0.000 0.002
� Lower-level degree with high skills �0.003 0.002 �0.003 0.002
� Higher-level degree without high skills �0.001 0.002 �0.001 0.002
� Higher-level degree with high skills 0.000 0.002 0.000 0.002
Level two (country) Lower-level tertiary expansion 0.002 0.003 0.002 0.002
Higher-level tertiary expansion 0.003 0.003 0.003 0.002
Difference in lower-level tertiary expansion �0.002 0.002 �0.003 0.002
Difference in higher-level tertiary expansion �0.002 0.002 �0.002 0.002
Index of tracking 0.011 0.017
GDP per capita (USD, PPP)/10,000 0.012 0.017
Union density 0.000 0.001

Variance Covariance structure (intercept) 0.000 0.000

(random effect) High skills without tertiary degree 0.001 0.001
Lower-level degree without high skills 0.002* 0.002*
Lower-level degree with high skills 0.000 0.000
Higher-level degree without high skills 0.002** 0.002**
Higher-level degree with high skills 0.001* 0.001*
Model fit �2LL 142,076.549 142,052.229
AIC 142,090.550 142,066.230

***P < 0.001,

**P < 0.01,
*P < 0.05 (two tailed).
N: individual ¼ 91,217, country ¼ 26.
Notes: Data are weighted using the full sample weight. ‘Tertiary degree’ and ‘Educational expansion’ include both higher- and lower-levels at the individual and
societal levels, respectively. The outcome variable is the logarithm of ISEI.

structure of economic returns to various credential/skill signs. That is, as stated in H2, lower-level degrees are
combinations. Firstly, interaction terms between the ex- likely to be devalued and individuals who possess merely
tent of lower-level tertiary expansion (i.e. the percentage high skills without tertiary degrees are penalized in soci-
of the population with short-cycle tertiary degrees) and eties where the share of lower-level degree holders is
individuals’ lower-level degrees (both with and without relatively large. In contrast, higher-level degrees are not
high skills) as well as high skills unaccompanied by necessarily depreciated. Nonetheless, as confirmed in
degrees show negative signs, and they are statistically Model 2 and contrary to our hypothesis, the devaluation
significant (i.e. c71 ¼ �0.006, c81 ¼ �0.005, and c91 ¼ of lower-level degrees is not cancelled out even when
�0.005). Meanwhile, interactions incorporating individ- they are accompanied by high skills. One may argue that
uals’ higher-level degrees do not indicate any substantial this result supports the soundness of conventional
914
European Sociological Review, 2022, Vol. 00, No. 0 European Sociological Review, 2022, Vol. 38, No.11
6

framework of credential inflation, which puts emphasis mechanisms behind these results and implications in the
on the excessive supply of human resources with higher next section.
levels of education as compared to the labour demand,
rather than paying attention to the variance in skills lev-
Discussion and Conclusion
els among educated personnel.
However, secondly, the interactions between the ex- This article has re-examined the devaluation of degrees,
tent of higher-level tertiary expansion (i.e. the share of a phenomenon long conceptualized as credential infla-

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

tertiary graduates with bachelor degrees and above) and tion by social scientists. Previous research has elucidated
credential/skill combinations demonstrate an interesting the variance in economic returns to different types of
contrast to the ones involving lower-level tertiary expan- credentials when considering the influence of education-
sion. In terms of individuals’ higher-level degrees, their al expansion, primarily from the demand side perspec-
depreciation is confirmed only when they are unaccom- tive: credential inflation occurs when growth in access
panied by high skills, whereas those with high skills are to higher levels of education outpaces an increase in la-
not significantly devalued despite the negative coeffi- bour market opportunities that genuinely require higher
cient of the interaction term with the extent of higher- educational attainment. However, that body of research
level tertiary expansion (i.e. c102 ¼ �0.005 and c112 ¼ has inadequately incorporated the supply side issue: the
�0.004, but only c102 is statistically significant at the depreciation of degrees may occur due to the lowered
0.5 per cent level). Meanwhile, the economic return to quality (i.e. skills levels) of highly educated people.
Consequently, while the dispersion in economic returns
lower-level degrees in the form of ISEI also becomes
to degrees has been explained with attention to the influ-
small in societies where the percentage of the population
ence of socio-economic statuses of degree holders as
with higher-level degrees is relatively high (i.e. the larger
well as types of credentials, the internal factor (i.e.
extent of higher-level tertiary expansion), but this is the
whether they are accompanied by high skills) has been
case only when lower-level degrees are unaccompanied
insufficiently investigated. Even among several pioneer-
by high skills (i.e. c82 ¼ �0.005 and c92 ¼ 0.003, and
ing studies that did incorporate the distinction between
c82 is statistically significant at the 0.5 per cent level).
educational attainment and skills, these still largely over-
Furthermore, the interaction between the extent of
look the distinctive roles of different levels of tertiary
higher-level tertiary expansion and individuals’ high
education, resulting in the inadequate explanation of the
skills without degrees does not indicate a substantial
diminishing value of degrees.
negative sign (i.e. c72 ¼ �0.001 but this is statistically
To address this gap, we analyse the structure of de-
insignificant). These results suggest that the larger share
valuation of tertiary degrees associated with educational
of higher-level tertiary graduates in a society is associ-
expansion with a focus on (i) the heterogeneity across
ated with the devaluation of individuals’ degrees un- credential/skill combinations and (ii) the distinction be-
accompanied by high skills, regardless of the nominal tween lower-level (ISCED 2011 Level 5) and higher-
level of credentials. Put differently, so long as individu- level (ISCED 2011 Level 6 and above) degrees at both
als possess high skills, they are not explicitly penalized the individual and societal levels. Multilevel linear re-
in the labour market even when lacking tertiary degrees. gression analyses, using PIAAC data for 91,217 individ-
This finding is also supported by an additional robust- uals in 26 countries, reveal that tertiary degrees and high
ness check using the continuous skills measure instead of skills generally contribute to obtaining higher ISEI even
credential/skill combinations (see Supplementary after adjusting for other individual-level and country-
Appendix). level predictors. In particular, regardless of skills level,
The same structure is confirmed in Model 4 that those with higher-level degrees are more likely than
omits several country-level variables unused for cross- others to enjoy economic returns, followed by lower-
level interactions. H3 is therefore supported as far as the level degree holders and then highly skilled people with-
depreciation of lower-/higher-level degrees without high out tertiary degrees. This result corroborates recent
skills is concerned. In addition, higher-level degrees with arguments that educational credentials rather than cog-
high skills retain their returns as stated in H3, but we nitive skills play the more significant role in the distribu-
also find that individuals who possess high skills with tion of labour market outcomes (e.g. Araki, 2020).
lower-level degrees or below are not penalized either. However, we may add nuance to the recognition that
This is a distinct social phenomenon, which is apparent- tertiary degrees are devalued in association with the
ly different from the one observed in relation to lower- larger extent of educational expansion. First, the eco-
level tertiary expansion. We discuss the potential nomic return to lower-level degrees becomes smaller in
European
12 Sociological Review, 2022, Vol. 38, No. 6 9150
European Sociological Review, 2022, Vol. 00, No.

countries where the share of tertiary graduates with Table 4. Multiple regression (OLS) of the percentage of
lower-level degrees is relatively large and hence such highly skilled people
degrees are not scarce. This devaluation is confirmed
B S.E. b
even when lower-level degrees are accompanied by high
skills. Furthermore, high skills unaccompanied by ter- Constant 0.769 3.330 –
tiary degrees are also depreciated. These results suggest Lower-level tertiary expansion 0.198 0.143 0.246
that, due to the lower-level tertiary expansion, (i) cre- Higher-level tertiary expansion 0.370* 0.146 0.448
Adjusted R2 0.250

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

dential inflation is at work in a way that devalues lower-
level degrees whose supply exceeds labour demand; and
***P < 0.001,
(ii) social closure is intensified such that individuals who **P < 0.01,
do not possess tertiary degrees are penalized (in addition *P < 0.05 [two tailed] [N ¼ 26].
to their originally smaller premium as compared to cre- Notes: Data are unweighted. This table is the result of a multiple regression
of the percentage of population with high skills among adults aged 25–65 years,
dential holders). This phenomenon does corroborate the using country-level data. Predictor variables are two measures: the percentage
conventional notion of credential inflation. of those with higher-level tertiary degrees and that with lower-level tertiary
Second, in contrast, individuals’ higher-level degrees degrees.

are devalued along with the larger extent of higher-level

tertiary expansion (i.e. the higher percentage of the To sum up, two types of educational expansion, one
population with higher-level degrees in a society) only
at lower-level and the other at higher-level, lead to two
when such degrees are unaccompanied by high skills
distinguishable social phenomena: credential inflation
(i.e. nominal degrees). That is, so long as individuals
and ‘decredentialization’. As to the mechanism behind
possess both higher-level degrees and high skills, they
these relationships, one may provisionally hypothesize
are not penalized by the diminishing scarcity of their cre-
that lower-level and higher-level tertiary expansions dis-
dentials. At the same time, the economic return to
tinctively affect the sensitivity to high skills in the labour
lower-level degrees without high skills also declines due
market, resulting in different types of devaluation of
to higher-level tertiary expansion, but this devaluation is
degrees. Specifically, as indicated in Table 4, the share
not confirmed for lower-level degrees accompanied by
of higher-level tertiary graduates in a society is signifi-
high skills. Moreover, highly skilled people without ter-
cantly correlated with that of the population with high
tiary degrees are not penalized in societies where the
skills, while the proportion of lower-level tertiary gradu-
proportion of tertiary graduates with higher-level
ates does not show a significant link with the accumula-
degrees is relatively large. Here, the results are not inter-
tion of high skills.5 Therefore, in a society where the
pretable based on the traditional concept of credential
inflation, according to which higher-level degrees should extent of higher-level tertiary expansion is large, the per-
be devalued even though they are accompanied by high centage of highly skilled people among the population is
skills. Likewise, from the perspective of closure theory, more likely to be high, and hence employers become
one would expect the depreciation of lower-level degrees more sensitive to the distinction between nominal
and non-tertiary graduates despite their high skills. Yet, degrees and degrees with high skills. As a result, ‘decre-
our analysis does not confirm this. Instead, we find nom- dentialization’ operates in a way that penalizes tertiary
inal degree holders who do not possess high skills are graduates without high skills whilst maintaining
significantly penalized by the larger extent of higher- rewards for highly skilled people regardless of educa-
level tertiary expansion. In the sense that credentials as tional credentials. This also suggests, unlike credentials
such lose their economic values only when they are un- that lose economic value due to their diminishing scar-
accompanied by actual skills and hence the role of skills city, high skills retain their rewards even though their
rather than nominal degrees is relatively enhanced in re- rarity deteriorates. This interpretation is aligned with
ward allocation, one may call this phenomenon ‘decre- the discussion by Araki (2020), which argues credentials
dentialization’. This is a distinct dimension of possess relative values, whereas high skills operate as ab-
devaluation of degrees, which cannot be explained by solute assets.6
the traditional notion of credential inflation that antici- Meanwhile, Table 4 suggests the extent of lower-
pates the declining returns to credentials regardless of level tertiary expansion is not explicitly associated with
their skills level. We therefore argue ‘decredentializa- the accumulation of highly skilled people (at least in
tion’ is detectable only by distinguishing (i) nominal terms of general information-processing skills as meas-
degrees from the ones with high skills; and (ii) lower- ured by PIAAC). In this case, sensitivity to high skills
level versus higher-level tertiary expansion. remains relatively low amongst employers, and hence
916
European Sociological Review, 2022, Vol. 00, No. 0 European Sociological Review, 2022, Vol. 38, No.13
6

the value of nominal degrees may persist. Consequently, occupation-specific skills as the key to economic out-
in a society saturated with lower-level degree holders, comes (Heckman et al., 2010; Kautz et al., 2014;
higher-level degrees are not devalued even though they OECD, 2015). Third, in scrutinizing the values of edu-
are unaccompanied by high skills. Put differently, cre- cation with/without skills, outcome measures can also
dentialism is at work in this type of society. be stretched to non-economic ones (e.g. health and sub-
The aforementioned hypothetical argument further jective well-being) and even societal-level consequences
underscores the significance of the concept of ‘decreden- such as economic growth and social cohesion (Araki,

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

tialization’. ‘Decredentialization’, which emerges in con- 2021; Rappleye and Komatsu, 2021; Schofer, Ramirez
junction with higher-level tertiary expansion, seemingly and Meyer, 2021).Finally, heterogeneity across individ-
suits a functionalist argument that highly skilled people ual social backgrounds is another important question.
make good match with preferable economic rewards. Although key individual-level measures are controlled
However, it is important to note that this matching
for in our analysis, this does not provide adequate evi-
through the devaluation of nominal degrees (i.e. decre-
dence concerning how returns to different types of
dentialization) is not a monotonous process but a theor-
degrees vary in accordance with social backgrounds.
etically rich phenomenon. This is because, while in our
This investigation would lead to further elucidation of
analysis lower- and higher-level tertiary expansions are
questions surrounding social mobility and inequality.
independently dealt with, there are—in reality—more
Despite the said room for further elaboration, the
complex combinations among them. For example,
contribution of this paper is significant as the empirical
lower-level tertiary expansion may be dominant in some
analysis focused on the devaluation of degrees with/
societies, whereas higher-level tertiary education may
without high skills and on the difference between
prevail in other societies. Depending on this composite
of lower- and higher-level tertiary expansions, either higher-level and lower-level tertiary education at both
intensified credentialism (i.e. anti-functionalist phenom- the individual and societal levels. We argue that these
ena), nullified credentialism (i.e. pro-functionalist phe- findings in conjunction with the accompanying theoret-
nomena), or their mixture may emerge. In this regard, ical discussions would pave the way for a better under-
what we call ‘decredentialization’ becomes of key theor- standing of credentialism and much broader social
etical importance as it distinguishes these phenomena mechanisms of reward allocation.
conceptually and empirically. Put differently, answering
questions about to what extent and how each society Notes
has promoted/undermined ‘decredentialization’ would 1 Literacy and numeracy are combined in accordance
promote our understanding of the mechanism of devalu- with Araki (2020). Also, it is important to note the
ation of degrees and much broader societal structures,
analysis results are robust even when 10 PVs are
which would be otherwise undetectable only through
incorporated instead of using the first PV only.
the conventional concept of credential inflation.
2 Another potential operationalization is to use the
To further clarify how and why these heterogeneous
mean score of literacy and numeracy as the continu-
phenomena operate, additional empirical analyses and
ous measure, instead of dichotomizing it to create
theoretical sophistication are required. First, while we
categorical variables (i.e. the combination of creden-
incorporate 26 countries in multilevel regressions with
tials and skills). However, this strategy does not ne-
country-level indicators as controls, one may assume
cessarily provide clear evidence to test the
there is a variation in the association between degrees
and economic rewards across societies, and importantly, aforementioned hypotheses focused on the heteroge-
such structures may change over time (Di Stasio, 2017). neous mechanisms of devaluation of degrees when
Thus, in addition to a cross-country approach demon- comparing them with and without high skills in rela-
strated here, country-specific analyses using longitudinal tion to societal-level educational conditions (i.e. the
data are essential to obtain further insights. Second, primary focus of this research is not to mathematic-
given that returns to credentials vary depending on fields ally calculate the returns to degrees/skills and their
of study and selectivity of education institutions (Bills, interactions at the individual level). Nevertheless, an
2016; Ortiz and Rodriguez-Menes, 2016; Posselt and analysis with the continuous skills measure can be
Grodsky, 2017), educational variables should be further used as a robustness check, albeit partially. We thus
detailed. Likewise, while general information-processing show the model incorporating the mean score of lit-
competences are herein used to quantify skills, their eracy and numeracy, lower-level tertiary degrees,
scope should be extended to non-cognitive and and higher-level tertiary degrees in Supplementary
European
14 Sociological Review, 2022, Vol. 38, No. 6 9170
European Sociological Review, 2022, Vol. 00, No.

Appendix for reference. Note that the results support Acknowledgements

the main arguments that follow. We would like to thank Jeremy Rappleye and Richard Breen for
3 Mother’s education and father’s education are quanti- their invaluable comments and suggestions.
fied by a tertiary education dummy (tertiary education
¼ 1), respectively, while the number of books at home
is based on six categories (10 or less ¼ 1, 11–25 ¼ 2, References
26–100 ¼ 3, 101–200 ¼ 4, 201–500 ¼ 5, more than Araki, S. (2020). Educational expansion, skills diffusion, and

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

500 ¼ 6), which are used as continuous variables. the economic value of credentials and skills. American
4 The significance of the difference in effect size is Sociological Review, 85, 128–175.
tested by replacing the reference (i.e. possessing nei- Araki, S. (2021). Does education make people happy?
ther a tertiary degree nor high skills) with other cre- Spotlighting the overlooked societal condition. Journal of
Happiness Studies (Online First). DOI: 10.1007/s10902-
dential/skill combinations, using the conventional 5
021-00416-y
per cent level. Hereafter, the difference is examined
Becker, G. S. (1964). Human Capital: A Theoretical and
via the same approach. Empirical Analysis with Special Reference to Education. New
5 In Table 4, a multiple regression analysis is con- York: National Bureau of Economic Research.
ducted, using country-level data on the percentage of Bills, D. B. (2016). Congested credentials: the material and pos-
individuals with high skills (as the outcome), that itional economies of schooling. Research in Social
with lower-level and higher-level tertiary degrees, re- Stratification and Mobility, 43, 65–70.
spectively (as predictors). The result shows the share Bills, D. B. and Brown, D. K. (2011). New directions in educa-
of higher-level tertiary graduates is significantly tional credentialism. Research in Social Stratification and
associated with the outcome. Mobility, 29, 1–4.
Bol, T. et al. (2019). School-to-work linkages, educational mis-
6 With regard to the logic behind the positive associ-
matches, and labor market outcomes. American Sociological
ation between ISEI and high skills measured by the
Review, 84, 275–307.
level of information-processing competencies,
Bol, T. and Weeden, K. A. (2015). Occupational closure and
there are two theoretical explanations. On the one wage inequality in Germany and the United Kingdom.
hand, from the perspective of human capital the- European Sociological Review, 31, 354–369.
ory, high skills per se are required for prestigious Bol, T. and van de Werfhorst, H. G. (2013). Educational systems
jobs, and hence employers favour highly skilled and the trade-off between labor market allocation and equal-
people who can work productively by directly ity of educational opportunity. Comparative Education
using such skills. On the other, based on signalling Review, 57, 285–308.
and labour queue theory, information-processing Brown, D. K. (2001). The social sources of educational credenti-
alism: status cultures, labor markets, and organizations.
skills may function as signals of high productivity/
Sociology of Education, 74, 19–34.
trainability, which is appreciated by the labour
Brown, P. (2003). The opportunity trap: education and employ-
market. Should this be the case, by incorporating
ment in a global economy. European Educational Research
the consequence of lower-level and higher-level Journal, 2, 141–179.
tertiary expansion, we may also conceptualize the Brown, P. (2013). Education, opportunity and the prospects for
multi-stage signalling. In societies with the larger social mobility. British Journal of Sociology of Education, 34,
proportion of lower-level degree holders, nominal 678–700.
levels of credentials operate as signals of high Collins, R. (1979). The Credential Society: An Historical
productivity and consequently individuals’ lower- Sociology of Education and Stratification. New York:
level degrees are devalued (first-stage signalling). Academic Press.
Collins, R. (2011). Credential inflation and the future of univer-
Meanwhile, as higher-level tertiary education
sities. Italian Journal of Sociology of Education, 3, 228–251.
expands, the signalling value of nominal degrees
Di Stasio, V. (2017). Who is ahead in the labor queue?
deteriorates as part of ‘decredentialization’ and
Institutions’ and employers’ perspective on overeducation,
subsequently information-processing skills start to undereducation, and horizontal mismatches. Sociology of
function as signals (second-stage signalling). Education, 90, 109–126.
Di Stasio, V., Bol, T. and van de Werfhorst, H. G. (2016). What
makes education positional? Institutions, overeducation and
Supplementary Data the competition for jobs. Research in Social Stratification and
Supplementary data are available at ESR online. Mobility, 43, 53–63.
918
European Sociological Review, 2022, Vol. 00, No. 0 European Sociological Review, 2022, Vol. 38, No.15
6

DiPrete, T. A. et al. (2017). School-to-work linkages in the Murphy, R. 1988. Social Closure: The Theory of
United States, Germany, and France. American Journal of Monopolization and Exclusion. Oxford: Clarendon.
Sociology, 122, 1869–1938. Nieto, S. and Ramos, R. (2017). Overeducation, skills and wage
Doren, C. and Grodsky, E. (2016). What skills can buy: trans- penalty: evidence for Spain using PIAAC data. Social
mission of advantage through cognitive and noncognitive Indicators Research, 134, 219–236.
skills. Sociology of Education, 89, 321–342. OECD. (2015). Skills for social progress: the power of social
Flisi, S. et al. (2017). Measuring occupational mismatch: over- and emotional skills. In OECD Skills Studies. Paris: OECD
education and overskill in Europe—evidence from PIAAC. Publishing. DOI: 10.1787/9789264226159-en.

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

Social Indicators Research, 131, 1211–1249. OECD. (2016). Skills Matter: Further Results from the Survey of
Ganzeboom H. B. G. 2010. International standard classification Adult Skills. Paris: OECD Publishing.
of occupation: ISCO-08 with ISEI-08 scores, available from: OECD. (2019). The Survey of Adult Skills: Reader’s
http://www.harryganzeboom.nl/isco08/isco08_with_isei.pdf, Companion. 3rd edn. Paris: OECD Publishing.
[accessed 1 December 2018]. OECD. (2020). Education at a Glance 2020: OECD Indicators.
Gesthuizen, M., Solga, H. and Kunster, R. (2011). Context mat- Paris: OECD Publishing.
ters: economic marginalization of low-educated workers in Ortiz, L. and Rodriguez-Menes, J. (2016). The positional value
cross-national perspective. European Sociological Review, 27, of education and its effect on general and technical fields of
264–280. education: educational expansion and occupational returns to
Hanushek, E. A. et al. (2015). Returns to skills around the education in Spain. European Sociological Review, 32,
world: evidence from PIAAC. European Economic Review, 216–237.
73, 103–130. Posselt, J. R. and Grodsky, E. (2017). Graduate education and
Hanushek, E. A. and Woessmann, L. 2015. The Knowledge social stratification. Annual Review of Sociology, 43,
Capital of Nations: Education and the Economics of Growth. 353–378.
Cambridge, MA: The MIT Press. Rappleye, J. and Komatsu, H. (2021). Is knowledge capital the-
Heckman, J. J. et al. (2010). The rate of return to the high/scope ory degenerate? PIAAC, PISA, and economic growth.
perry preschool program. Journal of Public Economics, 94, Compare, 51, 240–258.
114–128. Rivera, L. A. (2011). Ivies, extracurriculars, and exclusion: elite
Heisig, J. P., Elbers, B. and Solga, H. (2020). Cross-national dif- employers’ use of educational credentials. Research in Social
ferences in social background effects on educational attain- Stratification and Mobility, 29, 71–90.
ment and achievement: absolute vs. relative inequalities and Schofer, E., Ramirez, F. O. and Meyer, J. W. (2021). The soci-
the role of education systems. Compare, 50, 165–184. etal consequences of higher education. Sociology of
Heisig, J. P. and Schaeffer, M. (2019). Why you should always include Education, 94, 1–19.
a random slope for the lower-level variable involved in a cross-level Schultz, T. W. (1961). Investment in human capital. American
interaction. European Sociological Review, 35, 258–279. Economic Review, 51, 1–17.
Horowitz, J. (2018). Relative education and the advantage of a Smyth, E. and McCoy, S. (2011). The dynamics of credentialism:
college degree. American Sociological Review, 83, 771–801. Ireland from bust to boom (and back again). Research in
Kariya, T. 2011. Credential inflation and employment in Social Stratification and Mobility, 29, 91–106.
‘Universal’ higher education: enrolment, expansion and (in)- Spence, M. (1973). Job market signaling. The Quarterly Journal
equity via privatisation in Japan. Journal of Education and of Economics, 87, 355–374.
Work 24(1–2):69–94. Tholen, G. (2017). Symbolic closure: towards a renewed socio-
Kautz, T. et al. (2014). Fostering and Measuring Skills: logical perspective on the relationship between higher educa-
Improving Cognitive and Non-Cognitive Skills to Promote tion, credentials and the graduate labour market. Sociology,
Lifetime Success. OECD Education Working Papers, No. 110. 51, 1067–1083.
Paris: OECD Publishing. Thurow, L. C. 1975. Generating Inequality. New York: Basic
Kerckhoff, A. C., Raudenbush, S. W. and Glennie, E. (2001). Books.
Education, cognitive skill, and labor force outcomes. Van de Werfhorst, H. G. (2009). Credential inflation and educa-
Sociology of Education, 74, 1–24. tional strategies: a comparison of the United States and the
Levels, M., van der Velden, R. and Di Stasio, V. (2014). From Netherlands. Research in Social Stratification and Mobility,
school to fitting work: how education-to-job matching of 27, 269–284.
European school leavers is related to educational system char- Verhaest, D. and Van Der Velden, R. (2013). Cross-country dif-
acteristics. Acta Sociologica, 57, 341–361. ferences in graduate overeducation. European Sociological
Mincer, J. (1958). Investment in human capital and personal income Review, 29, 642–653.
distribution. Journal of Political Economy, 66, 281–302. Weeden, K. A. (2002). Why do some occupations pay more than
Mincer, J. 1974. Schooling, Experience, and Earnings. New others? Social closure and earnings inequality in the United
York: National Bureau of Economic Research. States. American Journal of Sociology, 108, 55–101.ss
European
16 Sociological Review, 2022, Vol. 38, No. 6 9190
European Sociological Review, 2022, Vol. 00, No.

Satoshi Araki is an Assistant Professor at the Takehiko Kariya is a Professor in the Sociology of
Department of Sociology and Social Policy, Lingnan Japanese Society, Nissan Institute of Japanese Studies
University, Hong Kong. His research focuses on social and Department of Sociology, and a Fellow of St
stratification and mobility, inequality, the sociology of Antony’s College, University of Oxford. His research
education and work, multi-dimensional well-being, and focuses on the sociology of education, the sociology of
research methods. In addition to his academic work Japan, social stratification and mobility, inequality, and
including articles published in the American Sociological sociological theory. His articles appear in the American

Downloaded from https://academic.oup.com/esr/article/38/6/904/6523878 by University of Hong Kong user on 05 January 2023

Review and the Journal of Happiness Studies, he has Journal of Sociology and Sociology of Education,
published more than 50 policy-oriented research papers. among others.