Journal of Economics and Economic Education Research (Print ISSN: 1533-3590; Online ISSN: 1533-3604)

Research Article: 2021 Vol: 22 Issue: 5

Determinants of Education-Job Vertical Mismatch in Urban Ghana

Prince Adjei, Koforidua Technical University

William Baah-Boateng, University of Ghana


This paper explores the determinants of education-job vertical mismatch in urban Ghana. It uses cross section data from the World Bank Skills toward Employment and Productivity (STEP) surveys of working age urban population and applies the self-Assessment method of measuring the incidence of education-job mismatch. The study employs the method of multinomial logit to ascertain whether over and undereducated individuals possess a relatively worse bundle of skills than workers who are adequately matched to their jobs in terms of formal education. It identifies gender, marital status, education, skills, occupation and time to proficiency as significant determinants of mismatch. The study adduce evidence to the transient nature of mismatch as reflected in time to proficiency’s negative relationship with overeducation and positive link with undereducation. Given the competitiveness of the national and global economic environment, the study provides some policy thoughts towards addressing challenges of skill mismatch.


Education-Job Mismatch, Overeducation, Undereducation, Multinomial Logit, STEP


Labour market analysis predicts a perfect alignment of workers acquired skills, education, and those demanded by firms as the outcome of education and training. This notion is however far from reality as empirical studies and evidences point to a divergence in the skill set of work-ers and the skill needs of employers. The challenges of ensuring the convergence of skills and education outcomes in the labour market has thus triggered renewed interest in across the world to ascertain the various forms of mismatches that are always present in labour markets, underly-ing determinants and effects on labour market outcomes.

OECD (2012) describes skills as “the new global currency” of the 21st -century economies, stressing the fact that no serious economy will survive with substandard workforce and skill-deficient human capital. The critical role that education and skills play in development has been identified as a priority area in internationally agreed development goals. For instance, the Sus-tainable Development Goals (SDG) 4 projects that by the year 2030, all countries must increase the number of working youth and adults to access relevant skills, including technical and voca-tional skills for employment, decent jobs and entrepreneurship.

The issue of skills is also worth considering given the macroeconomic and microeconomic dividends that it inures. Perry et al. (2014) posit the view that better and developed skills enable individuals to perform better and improve economic processes within a defined set-up or eco-nomic structure. Beyond the productivity enhancing trait of individuals, better skills affect post-hire outcomes in the form of higher wage premium; improved skill-set serves as an escape tool from unemployment (see Hanushek et al., 2014). At the aggregate level, better skills facilitate technological adoption and innovation and enhance potential for faster economic growth (Cic-cone & Papaioannou, 2009)

Given the fact that education is essential to improvements in poverty and welfare (Nsowah-Nuamah, 2010; Rollestone, 2011), facilitates entry into more lucrative occupations and improve earnings and earnings inequality (e.g. Kingdon & Söderbom, 2008), it can be assumed that an individual chooses a particular level and type of education that will manifest the expected out-come. This expectation is hardly met for the ordinary Ghanaian youth and urban job seeker who instead of picking up decent and career-oriented jobs, roam on the streets in search of their dream but unavailable jobs. Confounded with this situation, the observed worker optimises his or her choice variables either by succumbing to a low-profile indecent job that is unrelated to his level and type of education or upgrades him or herself through additional educational investment as a hedge against unemployment (Szirmai et al., 2013; Hyéfouais, 2016). This presupposes that the employment problem in Ghana does not manifest itself as open unemployment but skill un-deremployment, job mismatch and vulnerable employment and thus raises questions about the market relevance of education in harmonising labour market transitions.

The Government of Ghana through the help of donor partners have initiated several poli-cies and programmes such as Free Compulsory Universal Basic Education (FCUBE) and Free Senior High School (Free SHS) aimed at expanding schooling, at all levels. This move has been grounded on the perceived high private returns to education in general, and at higher levels of education, and the importance of education as an equity enabler (Ackah et al., 2014; Nsowah-Nuamah, 2010; Sackey, 2008). The ability of the Ghanian economy to absorb the increased sup-ply of skilled labour remains a concern to policy makers and researchers. However, studies on the incidence and determinants of education-job mismatch in Ghana rarely exist despite the fact that a large body of literature on over / undereducation is available (e.g. Hartog, 2000; McGuin-ness, 2006; Leuven & Oosterbeek, 2011).

Although the debate on education- job mismatch predates time, the empirical evidence is usually grounded within a developed economy context, with relatively sparse evidence for de-veloping countries, and in particular for Sub-Saharan Africa (SSA). This study attempts to fill this empirical gap and to limited literature of the subject on Ghana which mirrors many SSA countries. Similar work by Darko and Abrokwah (2020) examined the effects of educational mismatch on earnings in Ghana using cross sectional data to show increased incidence of under-education over the period 1998-1999 to 2012-2013.

From SSA perspective, the work of Herrera and Merceron (2013) on ten African found that 14.8-25.0% of employed workers aged 15 years and older are undereducated while 20.7-21.3% is overeducated between 2001-2005. This studies seeks find answers to the following questions: what factors predispose individual workers towards over / under education? To what extent are the mismatches transitory or permanent? What is the role of skills (cognitive and tech-nical skills) in determining mismatch?

This study aims to build on previous research to investigate the determinants of education-job vertical mismatch in Ghana using cross-sectional data from the World Bank STEP survey collected in 2013. The worker-self assessment method is used to derive the measures of overedu-cation and undereducation.

Walking through the Literature

Theoretical Perspectives on Skills Mismatch

A plethora of theories abound for explaining mismatch but in this study, the argument is limited to four main theories namely human capital theory, assignment theory, heterogeneous skill theory and institutional theory. Human capital theorist suggests that individuals’ productivi-ty determines his or her earnings rather than the essential attributes of the job they occupy. The worker productivity - the value of their marginal product is a function of their human capital composition that is made up of elements such as education, experience, innate ability among oth-ers (Becker, 1975). By virtue of this, investment in human capital is considered optimal if the net present value of future earnings is positive. Differences in labour market outcomes and success on the job from a human capital perspective is thus seen as a function of differences in education, work experience, skills, level of training among others. Overeducation is therefore seen as a short run phenomenon for an observed worker given the fact that further investments can change one’s position on the mismatch ladder.

In the assignment theory, educational mismatches imply skill mismatches (Allen Van Der Velden, 2001) Thus, workers underreport their required education due to poor matching triggered as a result of disconnect between possessed knowledge or skills and required ones for the job. Consequently, workers whose level of education exceeds the job requirement threshold are unable to fully utilise their skills given the fact that they are prone to be less productive than their peers whose educational level commensurate the requirements of the job. The heterogen e- ous skill theory on its part suggests a weak link between education and skills mismatch. Pre m- ised on the assumption that skill endowment and ability vary among individuals even in the face of equal level of educational attainment, the theory postulates that it is possible to have overed u- cated workers being under skilled and vice versa.

The institutional theory projects the view that job characteristics are the primer for earnings and labour market success. The foundational underpinning of the theory’s argument is on the fact that due to the encumbrances employers face in assessing indivi dual productivity at the point of hiring; they resort to the use of job specific characteristics to make their decision. For this reason, successful job match is an interplay of employer employee actions. Whereas an individual’s employability is contingen t on the amount of training ex post or ex ante his or her recruitment that of the employer is dictated by the training cost they will have to bear after initial recrui t- ment. Consequently, the more educated the prospective worker is at the point of initial hiring, the lower the associated training cost, and the greater the chances of being employed. The spill over effect of such interplay is that prospective workers may intentionally choose to be overeducated as a defence against unemployment and thus mainta ins one’s place in the queue for the desired job.

Empirical Literature

Empirically, whilst some perceive mismatch as transitory on the grounds of substitutability between experience and education, there is also the argument of “no escape trap” for overeducat-ed workers especially and thus mismatch seen as a long-term phenomenon. Key proponents of the transient argument (Sicherman (1991); Alba-Ramirez (1993); Buchel (2002) and their argu-ment gained ground within the career mobility theory. Its predictions are that you do not stay at the same position forever and that being overeducated does not mean it is so indefinitely. There- fore, overeducated workers have the opportunity to progress in their career and move to jobs that matches their skill and qualification set.

The advocates of permanency of the mismatch phenomenon posit the evidence that over-education correlates with lower job satisfaction and negatively affects worker productivity and welfare in general. Consequently, overeducated workers may be prone to high labour turnovers and lower earnings, a situation that is demotivating and career retrogressing (Rubb, 2003; Dolton and Vignoles, 2000). In terms of demographics, no clear pattern emerges for males and females relative to marriage. In Groot and Maassen van den Brink’s (2000) meta-analysis, they find overeducation to be more common among females than males. Sloane et al. (1996); Dolton and Vignoles (2000) observe similar patterns. Groot (1996), on the other hand, finds that married males are more likely to be over-educated. Frank (1978), finds higher incidence of overeducation among women and posits that women often do not have much of a choice as male husbands de-termine their location based on job offers available to them. Hung (2008) finds that marriage re-duces the probability of being overeducated for women while it has no effect on men. Sloane et al. (2000) find that the presence of children in the household present differential effects across gender in that younger children reduce overeducation for males and raise overeducation for fe-males, as they make more compromises in the labour market.

Overeducation is often seen as a short-term problem resulting from a lack of co-ordination in the adjustment of schooling requirements and schooling investments between firms and indi-viduals (Duncan & Hoffman, 1981). However, some studies have found that for a large group of workers overeducation is a long-term phenomenon (e.g. Sloane et al., 1999; Dolton & Vignoles, 2000). Overeducated workers earn less than equally educated workers who are employed in a job that matches their education, whereas undereducated workers who are employed at a job level that is higher than their level of education, earn more (Hartog, 2000). In the literature on overed-ucation, it is often argued that, apart from the attained level of education, job characteristics also determine a worker’s productivity (e.g. Sicherman, 1991).

Skill mismatch can arise from structural changes in the economy via innovation and tech-nological change (Perry et al., 2014). This suggest that individuals lacking those skills become unemployed or at worst accept jobs that do not match their skill portfolio. Socio-demographic factors also affect the level and incidence of mismatch. Desjardins and Rubenson (2011) posit the view that the more experience a worker is, the greater the likelihood of avoiding entry-level jobs and the better able they are to signal their skills for high paying and highly skilled jobs. They argue that women have a greater propensity of being under-skilled relative to men in the presence of labour market discrimination, or where childbearing and home-keeping activities are rife. When higher-skilled workers are employed in a lower-level job, their productivity will be restricted, whereas being employed in a higher-level job contributes to a worker’s productivity. However, others state that the lower productivity of the overeducated workers may indicate the relatively lower ability of these workers compared to the higher-skilled workers who found a job at a proper level (Sloane, 2003).

Econometrics Analysis

Conceptual Issues

Essentially, different methods and techniques have been employed to conceptualise the measurement of mismatch. Whilst no unanimity rule exists for a particular choice, the three key methods of worker-self assessment, realised matches and job assessment approach have had wide empirical appeal. Though fraught with its peculiar shortcomings, the worker-self assess-ment has been used extensively in studies that measure educational mismatch as proxy for skill and occupational mismatch (e.g. Korpi and Tahlin (2009); Buchel (2002); Oosterbeek (1991).

The statistical or realised match approach has also had strong appeal from scholars such as Verdugo and Verdugo (1989); Bauer (2002); Croce and Ghignoni (2012). The key issue in the usage of this approach has been whether to use the mean or mode in estimating the required years of education. This has however been less contentious as both approaches have been em-ployed in some cases for robustness checks.

Interestingly, the methods that seem to be less subjective in nature has not been vigorously pursued and utilised by researchers. Specific mention can be made of Rumberger (1987); McGoldrick and Robst (1996) as key actors having employed the approach. That notwithstand-ing, some studies (e.g Chevalier & Lindley, 2009); Groeneveld and Hartog (2004) have em-ployed a mixed approach-composition of job assessment, realised matches and worker-self as-sessment in their estimations.

Data Source

The paper uses dataset drawn from the World Bank STEP Survey collected in 2013 to pr o- vide quantitative estimates of the determinants of education job mismatch. The Ghana Skills Toward Employment and Productivity ( Measurement Program was collected by the World Bank in 2013 and consists of two survey instruments that collect information on the su p- ply and demand for skills. The STEP household survey is thus a collection of background info r- mation on randomly selected individuals within the household aged bet ween 15 to 64 detailing their history of acquired skills, educational attainment, work status and history, family bac k- ground, and health.(World Bank, 2015).

An important aspect of the STEP surveys which was administered in 12 other countries i n- cluding Kenya, Colombia and China is the use of different skill concepts that extends beyond educational attainment to include human capital indices more comprehensively. The sampling frame for the Ghana STEP is made up of urban households and individuals aged bet ween 15 and 64 years who are either working or otherwise. It is an exclusive supply side data as it is limited to only the employee. Three broad types of skills are measured namely, cognitive, non cognitive and job relevant skills. Cognitive skills are def ined as the “ability to understand complex ideas, to adapt effectively to the environment, to learn from experience, to engage in various forms of reasoning, to overcome obstacles by taking thought.” Whilst cognitive skills involve numeracy, literacy and t he ability to solve abstract problems, non cognitive skills relate to multiple traits spanning social, emotional, personality, behavioural and attitudinal factors. The third strand of skills are task related (such as computer use, driving skills use) which is a mixture of cognitive and soft skills.

Model Specification and Estimation

The study examines the determinants of mismatch within a multivariate framework where each regressor is evaluated in terms of its causal effect, holding all other factors constant. Fo l- lowing Leuven and Oosterbeek (2011), all the responses which reported required education to be higher than the actual attained education are coded as under- education, all those which say that the required education are less than the worker’s attained education are coded as over-education, and the remaining responses are interpreted as correct matches. The option of a simultaneous es- timation of the probability of overeducation and undereducation makes the specification of the multinomial logit model a preferred choice.

Given that there is no explicit ordering of the dependent variable’s categories, the study adopts the unordered model. The set of regressors ranges from personal biological traits, human capital characteristics, skill endowments, household background as well as traits of occupation and sector of economic activity. A multinomial model is thus estimated with the matched catego-ry as the base outcome and the results are interpreted as the likelihood of being in either of the remaining mismatch categories relative to being well matched.

Following Aleksynska and Tritah (2013); Herrera and Merceron (2014), the conceptual model for the determinants of labour skills mismatch(herein referred to as education-job mis-match) is given below with specific modifications:

image (1)

Where; Yij is the probability that a worker i is in one of the three jth education-job match categories: undereducated, correctly matched, or overeducated. The vector Xi includes individual specific characteristics. The general expression is given as

image (2)

image (3)

Where; pr (i) is the probability of the decision maker choosing alternative “i” and Vj is the systematic component of the utility of alternative “j”

Assume E0, Eu, Er where E0 is overeducation, Eu is undereducation and is Er required edu-cation;

The probability of each alternative is given as:

image (4)

image (5)

image (6)

Where pr (EO), pr(ER), pr (EU) are the probabilities of the decision maker being over, ade-quately or undereducated respectively.

image are the systematic components of the utility for overeducation alone, un-dereducation and adequately educated respectively.

In a more explicit form, the model is specified as

image (7)

Two different empirical models are estimated: a baseline model without controls and a skill-augmented model. They are presented in equations 8 and 9.

Model 1: Baseline model (without controls)

image (8)

As already indicated, the dependent variable is polytomous in nature as it assumes three (3) possible states - well matched (reference category), overeducated and undereducated. Equa-tion 8 is the baseline model where there are no controls. image is a vector of explanatory variables such as gender, time to proficiency, occupational groups, education, marital status, socioeconom-ic status at age 15, age and parental involvement in education. imagerepresents a vector of coeffi-cients of explanatory variables.

Model 2: Model with control for Skills

image (9)

In equation 9, the baseline model is augmented with skills to examine how an observed worker’s skill endowment predisposes him or her to being mismatched. image represents two set of skills: cognitive skill, extrapolated from a factor analysis of numeracy, literacy and writing skills and computer skill intensity (decomposed into low, medium and high usage, with no usage as reference category). image represents the coefficients of the skills variable and thus examines the par-tial impact of the skill variables on the predisposition of an observed worker being mismatched, ceteris paribus.

Measurement of Variables for Estimation

Dependent Variable

The study adopts the self-assessment approach for measuring mismatch, which has a strong appeal (Van der Velden and Van Smoorenburg, 1997) as it focuses on how an individual per-ceives the jobs in the labour market (Duncan & Hoffman, 1991; Sloane et al., 1999; Linsley, 2005). There are however some limitations associated with this measure including workers giv-ing wrong information about their education levels as observed by Hartog (2000) and the failure to effectively analyse the context of education level that a worker has with respect to the job that he or she performs (Battu et al., 2000). Under the self-assessment approach, if a worker’s highest educational attainment is the same as that defined by him/her as required by a given job, he/she is classified as well matched (see Farooq, 2011; Verhaest et al., 2015). If an educational attain-ment is higher (lower) than that required by a job, he/she is classified as overeducated (undered-ucated).

Independent Variables

The model controls for demographic, productive and socioeconomic characteristics of the individuals such as age, gender, marital status, education, occupation, time to proficiency, cogni-tive skills and computer literacy (Table 1).

Table 1: Definition and Measurement of Variables
Variable Operational Definition and Measurement
Dependent Variable  
EDUMATCH1 Educational mismatch 0-Perfectly matched; 1-Undereducation; 2-Overeducation
Explanatory Variables  
Age Age of the individual measured in years
Agesq Age squared measured in years
Female Gender variables as dummy:0=Male; and 1=Female
Married Marital dummy:0=Unmarried and 1=Married
Education Categorical dummies: No formal education, Primary, Lower secondary, Upper Secondary, and Tertiary
Time proficiency How long it takes for an observed worker to catch up on the job measured in months
Socioeconomic status The variable is captured in three categories: “low ses”, “middle ses”, and “high ses” and is captured in the form of three categorical dummies with low ses as the reference dummy
Occupational type Based on the International Standard Classification of Occupation (ISCO, 2008) classification and STEP aggregation, five main occupational groups are captured; Highly Skilled White Collar, Low Skilled White Collar, Craft and Related Trade Workers, Elementary Occupations and Skilled Agriculture.  
Parental involvement It is captured as a dummy with zero (0) being "No Active Involvement" and one (1) “Active Involvement
Cognitive skills Measured as composite index constructed through a factor analysis of three skill measures-reading, writing and numeracy skills.
Computer use It is the intensity score of computer usage. It is captured in three categorical dummies namely low usage, medium usage and high usage, and is a composite and aggregated score of the computer usage was constructed. The reference category is no usage (0).

Discussion of Empirical Results

Prior to the multinomial estimation, we carry out Independence of Irrelevance Alternative (IIA) test to validate the choice of reference category of the dependence variable in the estima-tion process (Hausman & McFadden, 1984). The results of the IIA test reported in the appendix suggest that IIA has not been violated, meaning the multinomial logit is appropriate. The study reports the marginal effects and outcome of variables that are statistically significant only. Mod-el (1) is the baseline model without controls and Model (2) shows the controls for skills in de-termining mismatch.

Baseline without controlling for skills

Table 2 presents multinomial logistic regression results for the baseline model in a form of marginal effect to suggest that age and parental involvement have no significant effect on educa-tion mismatch in urban Ghana. In contrast, gender, marital status, level of education, time to pro-ficiency and occupation has significant effect on education mismatch in urban Ghana. From gen-der perspective, females are less likely to be undereducated but more likely to be overeducated relative to males with stronger marginal effect. This suggests that in urban Ghana, female work-ers who are well matched in terms of formal education are more likely to be overeducated rela-tive to men but are less predisposed to be undereducated and thus corroborating similar study on Macedonia by Kupets (2015). It however contrasts findings in Georgia that records a lower like-lihood of overeducation among women and another by Herrera and Merceron (2014) who con-cluded from a study among urban workers in selected African countries that men have a higher probability than women of being mismatched.

Table 2: Marginal Effects Estimate of the Determinants of Education –Job Mismatch in Ghana
    Model 1: Baseline Model
VARIABLES Required education Undereducation Overeducation
Female -0.023 -0.087*** 0.111***
Time to Proficiency 0.023** 0.021*** -0.044***
Age -0.002 0.006 -0.004
Agesq -0.000 -0.000 0.000
Occupation(High skilled white collar as reference dummy)  
Low Skilled White Collar -0.186*** -0.076* 0.261***
Crafts &Related Trades -0.169*** -0.124** 0.294***
Elementary Occupation -0.303*** -0.162*** 0.465***
Skilled Agriculture -0.304*** -0.184*** 0.488***
Education: No formal education as reference dummy    
Primary Education -0.624*** 0.105** 0.518***
Lower secondary -0.382*** -0.118** 0.500***
Upper secondary -0.343*** -0.227*** 0.570***
Higher(Tertiary) -0.391*** -0.272*** 0.662***
Married 0.079* 0.114 -0.091*
Parental involvement 0.053 0.032 -0.061
Number of Observations is 2,987. *** p<0.01,     ** p<0.05,       * p<0.1;n=2063

Difference in gender probabilities might stem from duplicity of factors ranging from work-family balance, gender discrimination in the labour market, occupational choices and field of study. Traditionally, the multiple roles of home care/management and work-family life reconcili-ation by Ghanaian women tends to push them into a trap of statistical discrimination as lauded to by hiring agents who may be profit optimising. Consequently, most women find themselves rep-resented in non-standard employment and might switch from full-time to part-time employment (Spareboom, 2014; Connolly and Gregory (2008) which in most cases involves occupation downgrading making them highly prone to the risk of overeducation relative to their male coun-terparts. Again, there is the possibility that some fields of study (such as Humanities, Arts and Law), which are more prone to be exposed to overeducation tendencies in the labour market may have greater women representation (Barone & Ortiz, 2010; Betti et al., 2011; Jauhiainen, 2011; Wirz & Atukeren, 2005). These evidences potentially explain the higher overeducation but lower undereducation tendencies for females relative to males in urban Ghana.

In urban Ghana, married individuals are more likely to be undereducated than single and never married (Table 2). Being married do not significantly affect the chances of being overedu-cated; the story assumes a variant posture in the case of undereducation. In all cases of mismatch (over and undereducated), being divorced/separated relative to those who are single and never married tends to increase the chances of being mismatched either in the form of undereducation or overeducation. Within the category of divorced/widowed workers, the tendencies for overedu-cation are more pronounced than undereducation (9.3% as against 3.8% as evident in Table 2).

This finding contrasts the study by Kupets (2015) who found in Armenia that married indi-viduals are less likely to be overeducated than both single and divorced workers but gains sup-port with empirical findings from Georgia where marital status is statistically significant for un-dereducation. The implication of our findings is that married people may have the burden of economic responsibility shared among the partners and therefore an observed married worker may not necessarily overeducate herself since the income earned from the required education could be supplemented by additional family income. In such a case, they could leverage other human capital endowments like experience rather than additional educational investment. Re-sultantly, their tendencies to be overeducated are less and undereducation is high. The reverse might hold for single and divorced workers who may require additional educational investment beyond the minimum required level in order to boost earnings and maintain job/income security.

Irrespective of how long or short it takes urban workers to learn on the job, the results af-firm the statistical robustness of time to proficiency in explaining mismatch. For both over and undereducated, time to proficiency explains mismatch but in opposite directions. Whereas it in-creases the incidence of undereducation, it minimises the extent of overeducation. From Table 2, relative to those with less than one-month proficiency in their field of work, workers who spent over six months to catch up have 13.7% likelihood of being undereducated and 25.1% less chance to be overeducated. This means that the longer it takes urban workers to upgrade their skills, the lower the degree of being mismatched. The observed positive relationship with under-education and negative with overeducation suggests that workers that are more educated may require less on-the-job training and therefore have shorter time to proficiency while less educated workers may supplement their formal education with job experience often acquired through on-the -job training and thus requiring longer time to proficiency. This could imply that overeduca-tion might be transitory since workers with different years of education could be made to per-form the same or similar task if their human capital can be augmented to perform the same job. This finding corroborates the argument of Sloane et al. (1996) with the view that workers that require a long time to proficiency in view of potential losses from a bad match are more placed to be overeducated.

There appears to be mixed evidence on the role of education in determining mismatch. Whilst at all levels of education, there is a significant probability of being overeducated; educa-tion seems to reduce the likelihood of being undereducated as indicated in the baseline model (Table 2). In the absence of controls, observed urban workers with primary education relative to those with no formal education are 48.9% more likely to be overeducated compared to 11.1% chances of being undereducated. A key observation is that the higher the level of education, the greater the likelihood of being overeducated (48.9% for primary, 49.5% for lower secondary, 57.4% for upper secondary and 68.9% for tertiary level-as evident in Table 2). Intuitively, work-ers respond to the signalling prospects and benefits of higher education-higher incomes, job se-curity, secured tenure etc- and therefore might undertake more educational investment to main-tain their position in the job queue and potential for career progression. This finding is line with Herrera and Merceron (2014) work on ten African countries confirm that having a high level of education correlates with overeducation whereas urban workers with low level of education cor-relates with undereducation. The evidence also confirms Flisi et al. (2014) whose studies on se-lected European countries that having a higher level of education implies higher possibilities of overeducation rather than being matched.

The type of one’s occupation greatly affects the mismatch probabilities in urban Ghana. The directional impact of this variable seems to be at opposites with respect to undereducation and overeducation relative to being well matched (Table 2). All occupational groups significant-ly have lower mismatch tendencies with regard to undereducation but in the case of overeduca-tion, the tendencies are high in determining mismatch relative to highly skilled white-collar jobs. Whereas low skilled white-collar workers have the least incidences of over and undereducation (8.4% and 24.5% in Table 2), the skilled agriculture group have the highest predisposition of mismatch-both in the case of over and undereducation (19.7% and 48.1% respectively in Table 2).

Further evidence from the baseline results shows that urban workers in low-skilled white-collar occupations have higher predisposition of being overeducated relative to their counterparts in the high skilled group. This evidence gains support from Flisi et al. (2014) in a study on se-lected European countries that the higher the skill of the occupation, the lower the likelihood of being overeducated rather than matched. This suggests that those in the lower-skilled group may need additional educational investment to enhance their signalling potentials and use overeduca-tion as a hedge for job security as postulated by the job competition theory. Consequently, rather than suffering low job satisfaction/remuneration and potential of being displaced, overeducation becomes a defensive necessity for the low-skilled worker. Conversely, the high-skilled category workers might as a choice leverage their high human capital signalling trait and not necessarily pursue higher education beyond what is required for the job since the higher echelons of job ben-efits may accrue to them.

The Effect of Controlling for Skills

Table 3 presents the results of the multinomial logit with controls for two types of skills-cognitive skills and job-specific skills. The cognitive skill variable is a composite index con-structed through a factor analysis of three skill measures-reading, writing and numeracy skills. The job-specific skill is indexed by the intensity of computer skill usage on a scale of zero (0) to four (4) indicating no skill usage to high skill usage.

Table 3: Marginal Effect Estimates of the Determinants of Education-Job Mismatch in Ghana
    Model 2: Control for Skills
VARIABLES Required education Undereducation Overeducation
Female 0.001 -0.066*** 0.065***
Time to Proficiency 0.021** 0.019*** -0.041***
Age 0.001 0.010** -0.011*
Agesq -0.000 -0.000* 0.000*
Occupational Groups(High skilled white collar as  reference dummy
Low Skilled White Collar -0.172*** -0.047 0.220***
Crafts &Related Trades -0.145** -0.095* 0.239***
Elementary Occupation -0.277*** -0.134** 0.412***
Skilled Agriculture -0.279*** -0.149** 0.429***
Education(No formal education as reference dummy)  
Primary Education -0.564*** 0.098** 0.465***
Lower secondary -0.325*** -0.156*** 0.482***
Upper secondary -0.309*** -0.297*** 0.607***
Higher(Tertiary) -0.410*** -0.338*** 0.748***
Married 0.013 0.003 -0.016
Cognitive Skills 0.033* 0.043** -0.076***
Computer Skills Intensity(No usage as reference dummy)  
Low Usage 0.028 0.048 -0.077*
Medium Usage 0.066 0.067 -0.133*
High Usage 0.093* 0.117* -0.211***
Number of Observations is 2,987. *** p<0.01,       ** p<0.05,       * p<0.1;n=2058

Cognitive skills significantly explain overeducation but not so for undereducation. The re-sults show that the more adept urban workers are in their cognitive ability (reading, writing and numeracy), the less likely they are to be overeducated and more prone to being well matched. Indeed, with no statistically significant effect of cognitive skills on undereducation urban Ghana-ian workers are responsive to their skill endowments in terms of their status of being mismatched or otherwise.

Whilst computer skill use does not significantly explain undereducation tendencies, it showed statistically significant effect on overeducation in an inverse direction. Relative to urban workers without computer usage, those who use computer in medium and high intensities have significantly have lower chances of being mismatched specifically being overeducated (Table 3). In terms of the impact on overeducation, the likelihood increases with the intensity of the skill use. Of all the computer skill use groups, it is only the high-skilled intensity category that shows positive and significant likelihood of being well matched. This supports the view that workers with better skills tend to obtain matched jobs, while less talented counterparts are more likely to get jobs for which they are overqualified in terms of formal education. This evidence confirms Autor et al. (2003) findings that the adoption of computer-based technologies alters job skill de-mands to favour workers who hold a comparative advantage in computer skills.

Another compelling evidence in the face of controls for both cognitive skills and computer skills generates mixed results for the explanatory power of time to proficiency in explaining mismatch possibilities. Whereas the impact diminishes in the face of cognitive skill control, it magnifies in the presence of computer skills control. This variable captures the temporariness or otherwise of mismatch. If it is positively related to undereducation and negatively related to overeducation, then it implies that overeducation is transitory and the reverse is valid. As we control for skills in the model, females tend to exhibit positive overeducation traits relative to males, with shrinking magnitude and no statistically significant effect. The impact on underedu-cation tendencies however remains statistically significant and maintains the same direction, al-beit with lower magnitude effect compared to the baseline model.

Similar to the findings in the baseline model, married workers relative to unmarried demonstrate positive and significant undereducation traits with an even higher magnitude effect of 0.041 in the skill-controlled model against 0.038 in the baseline model. In terms of overeduca-tion however, the signs differ with lower likelihood for married workers relative to the reference category, although statistically not significant. On the contrary, those who are widowed relative to the single or never married appears to have greater mismatch possibilities both for over and undereducation. Intuitively, it could be inferred that possession of the right cognitive ability and computer skills reduces the job-matching friction and therefore dissipates the potential for educa-tion over and above the required level.

Even in the face of controls, occupational groups still matter in determining mismatch. For all occupational groups except the low skilled white collar, the size of the marginal effect on overeducation diminishes compared to the baseline model (although the sign and statistical sig-nificance remains consistent). Being in a low skill white-collar job relative to high skilled white collar consistently increases the chances of being overeducated. In a similar fashion, being in crafts and related trades significantly and consistently decreases the chances of urban workers being mismatched. This effect however increases in magnitude with further controls for cogni-tive and computer skills usage in the case of undereducation. For overeducation, the effect seems to be in congruence with the findings in the baseline model.

Education continues to be a decisive element in influencing mismatch possibilities. In the face of skills control, workers with primary education and lower secondary relative to those with no education demonstrate positive and significant overeducation traits but the magnitude of this effect is lower compared to the baseline model. This means the addition of skills compensates for education at that level and hence reduces the likelihood of being excessively mismatched (see Table 3). For upper secondary and tertiary level workers, the introduction of skills exacerbates their mismatch tendencies given their higher overeducation marginal effects compared to the baseline model (76.4% as against 68.9%). The intuitive induction for this observation is in tan-dem with the arguments proffered in the baseline model.

Conclusions and Policy Thoughts

We have examined the determinants of educational mismatch for employed workers in the urban Ghanaian labour market. The empirical analysis was based on cross-sectional data from the World Bank STEP survey conducted in 2013. The worker-self assessment approach to meas-uring educational mismatch was used.

The present analysis has revealed that job mismatch is a multidimensional phenomenon in-fluenced by a set of personal, employment and job specific characteristics. The recorded inci-dence of overeducation in most cases does not lend credence to the fact that the educational ex-penses be lowered. The key implications from this study are that the issue of mismatch present itself more of overeducation within the Ghanaian context. This finding suggests that there are more people within the Ghanaian labour market space whose level of education is higher than what is required on the job and consequently cannot find the appropriate job placement. As a policy measure in this regard, it is proposed that demand for higher skill jobs be stimulated so that talents and excess education do not go waste. This could be done through investments in new technologies and products activated through financial incentives. The evidence also shows the significance of time to proficiency in determining mismatch, implying that no matter the lev-el of education attained, there is still some level of on-the-job training and competency-based upgrading that must be done to fit the worker to the job. On hindsight of overeducation coupled with the above evidence, it can be implied that the focus of training agents (educational and vo-cational bodies) should be more on the quality of attained skills and not necessarily about large numbers that are turned out.

One potential limitation and grey area for future research is to examine the mobility di-mension of mismatch. Given the possibility that some available jobs may require certain skills which may be available but cannot be filled due to labour immobility, policymakers at both the national and the local level must, of course, factor mobility in when assessing mismatches and pondering an appropriate policy response and employers also have a role to play in overcoming mismatch caused by lack of mobility.

Another area of interest in the future will be to examine the relationship between education mismatch among graduates and different field of studies over time. This is however contingent on the availability of a panel data which requires more than one wave of data collection. This will provide grounds for comparative evidence on how mismatch differs along the course of study and degree of welfare implications that may be associated.


  1. Ackah, C., Adjasi, C., Turkson, F., & Acquah, A. (2014). Education, skill, and earnings: Further evidence from Ghana. Working Paper (No. 2014/073), UNU-WIDER.
  2. Adeyemi, M. (1997). An Investigation into Graduate Unemployment in Botswana. University of Botswana: National Institute of Development Research and Documentation.
  3. Alba-Ramirez, A. (1993). Mismatch in the Spanish Labour Market: Overeducation? Journal of Human Resources, 27(2), 259-278.
  4. Alba-Ramírez, A., &Blázquez, M. (2003). Types of job match, overeducation and labour mobility in Spain. Overeducation in Europe, 65-92.
  5. Aleksynska, M., &Tritah, A. (2013). Occupation–education mismatch of immigrant workers in Europe: Context and policies. Economics of Education Review36, 229-244.
  6. Allen, J., & Van der Velden, R. (2001). Educational mismatches versus skill mimsatches: effects on wages, job satisfaction, and on-the-job-search. Oxford Economic Papers, 53(3), 434-452.
  7. Ama, N. (2008). Transition from higher education to employment: A case study of graduates of faculty of social sciences. Educational Research and Review, 3(8), 262-274
  8. Autor, D., Levy, F., & Murnane, R. (2003). The skill content of recent technological change:An empirical exploration. The Quarterly Journal of Economics, 118(4), 1279-1333.
  9. Baah-Boateng, W. (2012). Gender Dimension of Labour Market Discrimination in Ghana. Lambert Publication.
  10. Baert, S., Cockx, B., &Verhaest, D. (2013). Overeducation at the start of the career: stepping stone or trap?. Labour Economics25, 123-140.
  11. Barone, C., & Ortiz, L. (2011). Overeducation among European University Graduates: a comparative analysis of its incidence and the importance of higher education differentiation. Higher Education61(3), 325-337.
  12. Battu, H., Belfield, C., & Sloane, P. (2000). How well can we measure graduate over-education and its effects? National Institute Economic Review, 171(1), 82-93.
  13. Bauer, T. (2002). Educational mismatch and wages: a panel analysis. Economics of Education Review, 21(3), 221-229.
  14. Becker, G. (1975). Human Capital. Chicago: University of Chicago Press.
  15. Becker, G. (1964). Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education. New York: Columbia University Press.
  16. Betti, G. (2007). Education Mismatch of Graduates: A Multidimensional and Fuzzy Approach. Working Paper.
  17. Boateng, K., & Ofori-Sarpong, E. (2002). An analytical study of the labour market for tertiary graduates in Ghana. World Bank(NCTE and NAB Project Report).
  18. Büchel, F. (2002). The effects of overeducation on productivity in Germany: The firms’ viewpoint. Economics of Education Review, 21, 263–275.
  19. Chevalier, A. (2003). Measuring over-education. Economica, 70(279), 509–531.
  20. Chevalier, A., & Lindley, J. (2009). Overeducation and the skills of UK graduates. Journal of the Royal Statistical Society: Series A (Statistics in Society)172(2), 307-337.
  21. Ciccone, A., & Papaioannou, E. (2009). Human capital, the structure of production, and growth. The Review of Economics and Statistics, 91(1), 66-82.
  22. Croce, G., & Ghignoni, E. (2012). Demand and supply of skilled labour and overeducation in Europe: a country-level analysis. Comparative Economic Studies54(2), 413-439.
  23. Cutillo, A., & Di Pietro, G. (2006). The effects of overeducation on wages in Italy: a bivariate selectivity approach. International Journal of Manpower27(2), 143-168.
  24. Darko C.K, and Abrokwa K.K.(2020) “Do you really need it? Educational mismatch and earnings in Ghana”. Rev Dev Econ.;00:1–28. rode.12681
  25. Desjardins, R., & Rubenson, K. (2011). An Analysis of Skill Mismatch using Direct Measures of Skills. OECD Working Papers, 63.
  26. Dolton, P., & Silles, M. (2001). Overeducation in the Graduate Labour Market:Some Evidence from Alumni Data. Centre for the Economics of Education.
  27. Dolton, P. J., &Silles, M. A. (2008). The effects of over-education on earnings in the graduate labour market. Economics of Education Review27(2), 125-139.
  28. Duncan, G., & Hoffman, S. (1981). The incidence and wage effects of overeducation. Economics of Education Review, 1(1), 75-86.
  29. Farooq, S. (2011). Mismatch between education and occupation: A case study of Pakistani graduates. The Pakistani Development Review, 531-552.
  30. Fenesi, B., & Sana, F. (2015). What is your degree worth?The relationship between post-secondary programs and employment outcomes. The Canadian Journal of Higher Education, 45(4), 383.
  31. Flisi, S., Goglio, V., Meroni, E., Rodrigues, M., & Vera-Toscano, E. (2014). Occupational mismatch in Europe: Understanding overeducation and overskilling for policy making. JRC Science and Policy Report, Luxembourg: Publication Office of the European Union.
  32. Frank, R. (1978). Why women earn less: the theory and estimation of differential overqualification. The American Economic Review, 68(3), 360-373.
  33. Geo-Jaja, M. A., & Mangum, G. (2001). Structural adjustment as an inadvertent enemy of human development in Africa. Journal of Black Studies32(1), 30-49.
  34. Green, F., McIntosh, S., & Vignoles, A. (2002). The utilisation of education and skills:evidence from Britain. The Manchester School, 70(6), 792-811.
  35. Groeneveld, S., & Hartog, J. (2004). Overeducation, wages and promotions within the firm. Labour Economics11(6), 701-714.
  36. Groot, W., &Maasen Van Den Brink, H. (1997). Allocation and the returns to over-education in the UK. Education Economics5(2), 169-183.
  37. Groot, W. (1996). The Incidence of, and Returns to Overeducation in the United Kingdom. Applied Economics, 28, 1345-1350.
  38. Handel, M.J., Valerio, A., & Puerta, M.L.S. (2016), Accounting for Mismatch in Low- and Middle-Income Countries: Measurement, Magnitudes, and Explanations. World Bank
  39. Hanushek, E. A., & Woessmann, L. (2011). The economics of international differences in educational achievement. In Handbook of the Economics of Education (Vol. 3, pp. 89-200). Elsevier.
  40. Hanushek, E. A., Schwerdt, G., Wiederhold, S., & Woessmann, L. (2013). Returns to Skills around the World: Evidence from PIAAC (No. w19762). National Bureau of Economic Research
  41. Hartog, J. (2000). Overeducation and earnings:where we are, where should we go? Economics of Education Review, 19(2), 131-147.
  42. Herrera, J., & Merceron, S. (2013). Underemployment and Job Mismatch in Sub-Saharan Africa. In P. D. Vreyer, & F. Roubaud (Eds.), Urban Labour Markets in Sub-Saharan Africa (pp. 83-107). Washington DC: World Bank.
  43. Hersch, J. (1991). Education match and job match. The Review of Economics and Statistics, 140-144.
  44. Hung, C. Y. (2008). Overeducation and undereducation in Taiwan. Journal of Asian Economics, 19, 125-137.
  45. Hyefouais, N. (2016). Characteristics and Determinants of Underemployment in Cameroon. African Economic Research Consortium.
  46. Kingdon, G., & Söderbom, M. (2008). Education, Skills, and Labour Market Outcomes: Evidence from Ghana. Working Paper Series 12, World Bank, Washington DC.
  47. Korpi, T., & Tahlin, M. (2009). Educational mismatch, wages, and wage growth: Overeducation in Sweden. Labour Economics, 16, 183-193.
  48. Kupets, O. (2015). Skill mismatch and overeducation in transition economies. IZA World of Labour, 224.
  49. Leuven, E., & Oosterbeek, H. (2011). Overeducation and mismatch in the labour market. Handbook of Economics of Economics of Education, 4, 283-326.
  50. Levels, M., van der Velden, R., & Allen, J. (2014). Educational mismatches and skills:new empirical tests of old hypotheses. Oxford Economic Papers, 66(4), 959-982.
  51. Linsley, I. (2005). Causes of overeducation in the Australian labour market. Australian Journal of Labour Economics8(2), 121.
  52. McGoldrick, K., & Robst, J. (1996). Gender differences in overeducation: A test of the theory of differential overqualification. The American Economic Review, 86(2), 280-284.
  53. McGuinness, S. (2006). Overeducation in the labour market. Journal of Economic Surveys, 20(3), 387-418.
  54. Morsy, H.  & Mukasa, A.N. (2019), Youth Jobs, Skill and Educational Mismatches in Africa, Working Paper Series N° 326, African Development Bank, Abidjan, Côte d’Ivoire.
  55. Nsowah-Nuamah, N., Teal, F., & Awoonor-Williams, M. (2010). Jobs, Skills and Incomes in Ghana: How was poverty halved? Working Paper Series (No. 2010-01), University of Oxford., Centre for the Study of African Economies.
  56. Oppong, S., & Sachs, P. R. (2015). Managing Graduate Unemployment in Emerging Economies: Critical Analysis of the Skills Mismatch and Oversupply Theses. Polovna Brief1, 1-14.
  57. Oosterbeek, H. (1991). Essays of Human Capital Theory, PhD. PhD Dissertation, University of Amsterdam.
  58. Perry, A., Wiederhold, S., & Ackermann-Piek, D. (2016). How can skill mismatch be measured? New approaches with PIAAC. methods, data, analyses8(2), 38.
  59. Robst, J. (2007). Education and Job Match: The relatedness of college major and work. Economics of Education Review, 26(4), 397-407.
  60. Rolleston, C. (2011). Educational access and poverty reduction: The case of Ghana 1991-2006. International Journal of Educational Development, 31(4), 338-349.
  61. Rubb, S. (2003a). Overeducation: A short or long run phenomenon for individuals. Economics of Education Review, 22(4), 389–394.
  62. Rumberger, R. W. (1987). The impact of surplus schooling on productivity and earnings. Journal of Human Resources, 22(1), 24-50.
  63. Sackey, H. A. (2008). Private returns to education in Ghana: implications for investments in schooling and migration. Research Paper (No. 174)., African Economic Research Consortium.
  64. Sattinger, M. (1993). Assignment Models of the Distribution of Earnings. Journal of Economic Literature, 31(2), 831-880.
  65. Senarath, S., & Patabendige, S. (2014). Job-Education Mismatch among the Graduates: A Sri Lankan Perspective. Ruhuna Journal of Management and Finance, 1(2), 1-16.
  66. Sicherman, N. (1991). Overeducation in the Labor Market. Journal of Labour Economics, 9, 101-122.
  67. Sloane, P., Battu, H., & Seaman, P. (1996). Overeducation and the formal education/experience and training trade-off. Applied Economics Letters, 3(8), 511-515.
  68. Sloane, P., Battu, H., & Seaman, P. (1999). Overeducation, Undereducation and the British Labour Market. Applied Economics, 31(11), 1437-1453.
  69. Sloane, P. (2003). Much Ado About Nothing? What does the Overeducation Literature Really Tell Us. In F. Büchel, A. De Grip, & A. Mertens (Eds.), Overeducation in Europe: Current issues in theory and policy (pp. 11-45). Edward Elgar.
  70. Sloane, P., Battu, H., & Seaman, P. (1999). Overeducation, Undereducation and the British Labour Market. Applied Economics, 31, 1437–1453.
  71. Sparreboom, T., &Nübler, J. (2013). Productive transformation, employment and education in Tanzania.
  72. Szirmai, E., Gebreeyesus, M., Guadagno, F., & Verspagen, B. (2013). Promoting Productive Employment in Sub-Saharan Africa: A Review of the Literature.
  73. Teal, F. (2001). Education, incomes, poverty and inequality in Ghana in the 1990s. Working Paper Series (No. 2001-21), University of Oxford, Centre for the Study of African Economies.
  74. Thurow, L. (1975). Generating Inequality. Basic Books.
  75. Twumasi Baffour, P. (2015). Determinants of Urban Worker Earnings in Ghana: The Role of Education. Modern Economy, 6, 1240-1252.
  76. van Smoorenburg, M., & van der Velden, R. (2000). The Training of School-leavers. Economics of Education Review, 19(2), 207-217.
  77. Van der Velden, R., & Van Smoorenburg, M. (1997). The Measurement of Overeducation and Undereducation:Self Report vs. Job-Analyst Method. Maastricht: Research Centre for Education and the Labour Market.
  78. Verdugo, R., & Verdugo, N. (1989). The impact of surplus schooling on earnings: some additional findings. Journal of Human Resources, 629-643.
  79. Verhaest, D., Schatteman, T., & Van Trier, W. (2015). Overeducation in the early career of secondary education graduates: An analysis using sequence techniques. Young23(4), 336-356.
  80. Verhaest, D., & Omey, E. (2006). The Impact of Overeducation and its measurement. Social Indicators Research, 77(3), 419-448.
  81. Verhaest, D., Schatteman, T., & Van Trier, W. (2015). Overeducation in the early career of secondary education graduates: An analysis using sequence techniques. Young23(4), 336-356.
  82. Verhaest, D., & Omey, E. (2006). The Impact of Overeducation and its measurement. Social Indicators Research, 77(3), 419-448.
  83. World Bank. (2015). World Development Indicators. World Bank online data


HAUSMAN IIA TEST (On the basis of the Reported Results)

Test: Ho: difference in coefficients not systematic

chi2(36) = (b-B)'[(V_b-V_B)^(-1)](b-B)

= 4.16

Prob>chi2 = 1.0000

(V_b-V_B is not positive definite)


A significant test is evidence against Ho.

If chi2<0, the estimated model does not meet asymptotic assumptions.

From the test, since the chi2>0, the estimated model meets asymptotic assumptions and thus IIA has not been violated, meaning the multinomial logit is appropriate.

Get the App