Education occupation mismatch in vietnam determinants and effects on earnings

Analyzing survey data from 267 workers, we find that the mismatch between schooling major and working field which is caused by unavailability of job in the schooling field demand-related

UNIVERSITY OF ECONOMICS ERASMUS UNVERSITY ROTTERDAM

HO CHI MINH CITY INSTITUTE OF SOCIAL STUDIES

VIETNAM – THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

EDUCATION-OCCUPATION MISMATCH IN VIETNAM:

DETERMINANTS AND EFFECTS ON EARNINGS

BY

PHAN THI THANH THAO

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY, August 2016

UNIVERSITY OF ECONOMICS ERASMUS UNVERSITY ROTTERDAM

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

EDUCATION-OCCUPATION MISMATCH IN VIETNAM: DETERMINANTS AND EFFECTS ON EARNINGS

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

By

PHAN THI THANH THAO

Academic Supervisor:

Dr TRUONG DANG THUY

HO CHI MINH CITY, August 2016

ACKNOWLEDGEMENTS

The process of writing a thesis is a collaborative experience involving the support and helps from many people I want to express my gratitude to those who give me the tremendous support to complete this thesis

I am deeply indebted to my parents for their invaluable supports and constant reminders The sentence I hear every day is “lose weight and finish your thesis, daughter” I really appreciate for their efforts in reminding a very lazy girl like me And their boundless love are motivation for my endeavor in building up my life more interesting and valuable

I wish to express my heartfelt gratitude to my supervisor Dr Truong Dang Thuy for his valuable suggestions during the time I write this thesis He has also encouraged and reminded me to pursue this topic from the initial ideas to the final completion I am really thankful him for his guidance and patience

Finally, after finishing this thesis, I realize that each success is a process of continuous effort And more difficulties you overcome, more values you get for your life

Phan Thi Thanh Thao

August, 2016

ABSTRACT

We examine the education-occupation mismatch in horizontal and vertical respects; and their impacts on earnings of Vietnamese workers We start by clarifying definitions and causal reasons of mismatch between education and occupation: in major and level Analyzing survey data from 267 workers, we find that the mismatch between schooling major and working field which is caused by unavailability of job in the schooling field (demand-related horizontal mismatch) has a negative effect on earnings And the mismatch between schooling major and working field caused by remaining reasons (supply-related horizontal mismatch mismatch) has no statistically significant impact on earnings Interestingly, a horizontal mismatch because of supply-related reasons for workers who learned science major has a positive effect on earnings Furthermore, when examining the effect of vertical mismatch, a negative effect of under-education on wage

is found whereas over-educated years have no significant effect on wage

From policy perspective, we recommend that people should avoid major mismatch for best earnings However, in case individuals learn science and work in mismatched career voluntarily, their earnings will be better than ones in adequate career Moreover, students should avoid over-education to reduce the waste of resources unless they want to study

more for their own preferences

Contents

INTRODUCTION 1

1.1 Problem statement 1

1.2 Research objectives 3

1.3 Main research questions 4

1.4 Organization of the study 4

LITERATURE REVIEW 5

2.1 Mismatch in major between career and schooling (horizontal education-occupation mismatch) 5

2.1.1 Definition 5

2.1.2 Determinants of horizontal education-occupation mismatch 6

2.2 Over-education and under-education (Vertical education-occupation mismatch) 8

2.2.1 Definition 8

2.2.2 Determinants of vertical education-occupation mismatch 11

2.3 Effect of education-occupation mismatch on earnings 14

2.3.1 Mincer’s earnings model 14

2.3.2 Wage effect of horizontal education-occupation mismatch 17

2.3.3 Wage effect of vertical education-occupation mismatch 18

METHODOLOGY AND DATA 20

3.1 Empirical models 20

3.1.1 Horizontal mismatch and earnings 20

3.1.2 Vertical mismatched and earnings 25

3.2 Data source 28

RESULTS 29

4.1 Descriptive statistics 29

4.1.1 Horizontal education-occupation mismatch 29

4.1.2 Vertical education-occupation mismatch 40

4.2 Regression results 48

4.2.1 Horizontal education-occupation mismatch 48

4.2.1.1 Determinants of horizontal education-occupation mismatch 48

4.2.1.2 Effect of horizontal education-occupation mismatch on earnings 53

4.2.2 Vertical education-occupation mismatch 59

4.2.2.1 Determinants of vertical education-occupation mismatch 59

4.2.2.2 Effects of vertical education-occupation mismatch on earnings 63

CONCLUSION AND POLICY IMPLICATION 71

5.1 Conclusions 71

5.2 Policy implications 73

5.3 Limitations 74

REFERENCES 76

APPENDIX 81

LIST OF TABLES

Table 4 1: Descriptive statistics of continuous variables 29

Table 4 2: Age among horizontal mismatched groups 30

Table 4 3: Schooling years among horizontal mismatched groups 30

Table 4 4: Experience in current firm among horizontal mismatched groups 31

Table 4 5: Experience in current working field among horizontal mismatched groups 31

Table 4 6: Reasons for mismatch among horizontal mismatched groups 32

Table 4 7: Education level among horizontal mismatched groups 33

Table 4 8: Schooling major group and horizontal mismatched groups 33

Table 4 9: Gender and horizontal mismatched groups 34

Table 4 10: Marital status among horizontal mismatched groups 34

Table 4 11: Number of children and horizontal mismatched groups 35

Table 4 12: Mobility status among horizontal mismatched groups 35

Table 4 13: Long-term health status among horizontal mismatched groups 36

Table 4 14: Firm type and horizontal mismatched groups 37

Table 4 15: Working place among horizontal mismatched groups 37

Table 4 16: Immigration status among horizontal mismatched groups 38

Table 4 17: Earnings level among horizontal mismatched groups 39

Table 4 18: Fulltime/part-time job and horizontal mismatched groups 39

Table 4 19: Age of vertical mismatched groups 40

Table 4 20: Schooling years among vertical mismatched groups 41

Table 4 21: Experience in current firm among vertical mismatched groups 41

Table 4 22: Experience in current field among vertical mismatched groups 42

Table 4 23: Gender and vertical mismatched groups 42

Table 4 24: Education level of vertical mismatched groups 43

Table 4 25: Schooling major group among vertical mismatched groups 44

Table 4 26: Marital status among vertical mismatched groups 44

Table 4 27: Number of children among vertical mismatched groups 45

Table 4 28: Firm type and vertical mismatched groups 46

Table 4 29: Fulltime/part-time job among vertical mismatched groups 46

Table 4 30: Mobility status among vertical mismatched groups 47

Table 4 31: Long-term health status among vertical mismatched groups 47

Table 4 32: Working place among vertical mismatched groups 48

Table 4 33: Immigration status among vertical mismatched groups 48

Table 4 34: Determinants of horizontal mismatched education: Ordinal logistic

regression 50

Table 4 35: Marginal effect of determinants of horizontal mismatched education 51

Table 4 36: Effects of horizontal mismatched education on earnings 55

Table 4 37: The earnings effects of mismatch by schooling majors 58

Table 4 38: Determinants of over-educated years 61

Table 4 39: Effects of vertical mismatched education on earnings (Duncan and Hoffman model) 64

Table 4 40: Effects of vertical mismatched education on earnings (Verdugo and Verdugo model) 69

LIST OF GRAPHS

Graph 4 1: Distribution of over-education (Duncan and Hoffman model) 60Graph 4 2: Distribution of under-education (Duncan and Hoffman model) 60Graph 4 3: Effects of vertical mismatched education on earnings 67Graph 4 4: Effects of vertical mismatched education on earnings for male and female 68

LIST OF APPENDICES

APPENDIX 1: t-test for determinants of horizontal mismatched education 81APPENDIX 2: Chi-squared test for determinants of horizontal mismatched education 82APPENDIX 3: t-test for determinants of vertical mismatched education 83APPENDIX 4: Chi-squared test for determinants of vertical mismatched education 84APPENDIX 5: Questionaire 85

This large increase in highly educated labor force causes an unbalance in labor market where supply excess demand This disequilibrium in labor market makes high educated workers accept unskilled job or a mismatched job to avoid to be unemployed In an interview set up by Hiep Pham (2013), Le Duy Luong – the human resource director of

a Japanese electronics company in Hoa Cam Industrial Zone – said that hundreds of collar worker in his company had university degrees Furthermore, Hiep Pham (2013) also noticed that over-education is rising in Vietnam The job employees are working does not require as much knowledge as they learned in school and it seems a waste when they are over-educated (a vertical education-occupation mismatch)

blue-Another result of supply excess in labor market is that employees have to work in an unrelated job to his schooling major (horizontal education-occupation mismatch) At the

time when individual chose university major, he expected that he could work in the field

of that schooling major in the future However, there are many indicators affecting his decision in choosing the studied major: expected wage, changes in labor market equilibrium, non-price orientations, and the probability of graduation of that major And

it seems many young people do not know clearly what they want, what they can and what they should So that this is also a reason for the mismatch between career they are working and the university major they learned

These education-occupation mismatches are not only a waste in money and human capital but also a reflection of labor market failure There are many studies about mismatch in education grade and in schooling major including Tsang and Levin (1985), Sicherman (1990), Bauer (2002), Björklund and Kjellström (2002), Büchel and Mertens (2004), Robst (2007), Dolton and Silles (2008), Nordin, Persson and Rooth (2010) Bender and Heywood (2011) However this research issue is quite new in Vietnam

As mentioned above, individuals have tendency to learn more because they believe in a higher future earnings But is it true that wage will change with the change of education level? When considering the effect of over-education and under-education on earnings,

it is found that employees with over-education earn less than ones with adequate education level (Kiker et al., 1997; Dolton, 2008) However, how much over-education

or under-education affect earnings? It can be 35-40 percent declining in earnings for over-educated person as Dolton (2008) found from data of one large civil university in the UK Furthermore, Kiker et al (1997) examined a sample of 50,000 Portuguese individuals and found that over-educated workers earn approximately 8 percent less than similar workers with the same education level who are working in an adequate job

On the other hand, some studies found that the effect of an additional year of schooling is positive Duncan and Hoffman (1981) also revealed that return to an additional year of over education can be positive for US workforce Nevertheless, they

over-also found that this estimated return to an additional year of over-education is only a half

of return to an additional year of required education

According to Kiker et al (1997), workers with less education than requirement for the job earn 16.3% more than those with the same education level who are working in an adequate job Bauer (2002) used a large panel data set of Germany in period 1984-1998

to examine workers with similar job but different education levels, and he concluded that under-educated employees bear a penalty for an additional year of deficit education is 6-

11 percent The same result is also found by Duncan and Hoffman (1981) with 4.2% decrease in earnings for an additional year of deficit education

Furthermore, using data from National Survey of American College Graduates in 1993,

he found that workers with mismatched between schooling major and career bear a decrease in earnings 10-12% depending on mismatched type (Robst, 2007) Another evidence comes from study of Nordin et al (2010) for Swedish people from 28-36 years old which indicates an earnings penalty of 12-20% for mismatched workers

Although the field is widely investigated, there are very few studies about this issue in Vietnam This study will give a basic overview about horizontal and vertical mismatched education and their impacts on earnings Particularly, this study examines the determinants of education mismatch and its impacts on wage, using data from a survey

of 267 respondents

1.2 Research objectives

My main research objectives are two-fold Firstly, I examines determinants which affect probability of education-occupation mismatch both in two respects: horizontal and vertical These determinants include demographic factors, job characteristics and schooling majors Secondly, the effects of over-education/under-education (vertical mismatch) and mismatched in major (horizontal mismatch) on earnings are examined in more details

Based on the research results, some policies are suggested to solve existing over/under education and mismatched between schooling major and career

1.3 Main research questions

There are three main questions which need to be clarified in this research:

 Firstly, which determinants have statistically significant impact on probability of education-occupation mismatch?

 Secondly, how do over-education and under-education affect earnings?

 Thirdly, how does mismatched between schooling major and career affect earnings?

1.4 Organization of the study

This thesis consists of five chapters After this introduction chapter, the remaining of this thesis is arranged as follow Chapter 2 is the theoretical framework to have a basic knowledge in this topic It discusses about over/under education and mismatched major

in details: definition, determinants, impact on earnings and several model which were applied by previous researches Chapter 3 is research methodology This chapter presents methods I apply to estimate the determinants of vertical and horizontal mismatch, and the effect of mismatched education on earnings Chapter 4 is the result which indicates descriptive statistics and regression results Chapter 5 concludes the study with policy implications and limitations

CHAPTER 2

LITERATURE REVIEW 2.1 Mismatch in major between career and schooling (horizontal education-

occupation mismatch)

2.1.1 Definition

When choosing a career, individual tries to maximize his satisfaction based on the compromise between his opportunities and limitations of environment The trade-off in this case causes a great number of people work in unrelated job with their schooling major

According to Robst (2007), there are two main reason categories which explain why an individual choose an unrelated career: supply-related reasons and demand-related reasons Supply-related reasons which are defined as voluntary by workers, including pay-promotion opportunities, change in job interests, working environments and conditions, working location and family and social related reasons Demand-related reason, which can be regarded involuntary, is the unavailability of job in the schooling field While supply-related reasons do not suggest labor market failure, demand-related reasons indicate the inefficiency in market

Supply related factors are suggested base on the change in individual’s preferences, constrains or changes in information about career characteristics Two subcategories of supply related indicators are classified: career oriented and amenity Pay and promotion opportunities and change in career interest are considered as job oriented factors A job with high salary and chances to higher position in career is attractive and hard to be refused although it is unrelated to their learned knowledge and skills from school For example, a person with a bachelor in sales works as a marketing manager because this job gives him higher wage and better position According to Nordin et al (2010), primary investment decision is taken based on expectation about future earnings and occupational

characteristics which can be changed after working in matched jobs Besides, changes in other occupation’s information can attract workers to these jobs These reasons indicate that workers are interested in their existing job characteristics and their mismatch is completely active

The other subcategory is amenity which includes working condition, job location and family-related reasons As indicated by Sicherman (1990), women’s utility is more influenced by non-market conditions than men’s The non-market conditions include working hours, household duties, illness in the family For example, a married woman

in a small labor market has less chance to find a related job because of disadvantages in geographic and time These reasons constrain workers from finding an interesting job Sicherman (1990) indicated that men change job more often because of career oriented reasons and better opportunities while women change job because of amenity (Robst, 2008)

In demand factors, the inability to find a related job is considered as an excess supply problem in labor market It means that there are more graduates than jobs in field they want to work in Another problem in demand side reasons is that there are jobs in market but labors cannot get such position caused by incomplete information in job search Furthermore, low ability and other non-educational individual characteristics are also reasons for demand-related horizontal mismatch In a research of Robst (2007), unlike his expectation, the result indicates that there are more men who accept to work in unrelated job because of these reasons than women These reasons indicate an inefficiency of labor market

2.1.2 Determinants of horizontal education-occupation mismatch

There are not much literatures discussing about the mismatch based on major of schooling Following Robst (2007), Nordin et al (2010), Bender and Heywood (2011) career mismatch can be classified into three categories: related career, partly related career and completely unrelated career Zhu (2014) suggested three main determinants

which affect mismatch probability: demographic variables, job characteristics and major they learned

First of all, with demographic variables, Robst (2007) used age, married status, the highest degree, race, disabled His result indicated that the likelihood of mismatch increases with age, disability and probability to be horizontal mismatched is higher for single individuals than married ones Opposite expectation comes from Madamba and

De Jong (1997), job mismatch is expected to be more common among younger than older workers The explanation is that when a worker older, he has more time to find the suitable job to his major Furthermore, following Robst (2007), workers with high-level degree such as Master, Professional, and Doctor have less likelihood of being mismatched than workers with Bachelor degree only

Secondly, there are three main subcategories are considered in job characteristics, including working sector, career stage and working city These determinants is clarified

receive PhD degree in science or engineering The first indicator which is mentioned is working sector The employees who work in government or business sector face more tendency to have mismatched career compared to academic sector The career stage is also considered with three stages: early stage (less or equal 10 years since degree), middle stage (11-24 years since degree) and late stage (25 or more years since degree) The results indicated that early in their career is more likely to be matched between education and occupation compared with later stage Furthermore, these mismatch may

be not the result of inefficiency in labor market

Another determinant is working city presented by Zhu (2014) including: type of province where the employees are working, the same between work and home province, the same between work and college-located place The explanation is suggested by Abel (2012) that there is a causal relationship between job matching and agglomeration which uses population size or employment density as its proxies Abel and Deitz (2015) clarified

that in a big city with more concentrated labor market, the cost of searching job is lower and career opportunity range is wider, workers have more chances to have a related job with their schooling major, so they are more likely to match their human capital to job The research from Abel and Deitz (2015) indicated that the college-educated individuals work in more agglomerated metropolitan areas have higher proportion of working in career related to their schooling major More precisely, career matching increase by about 0.15 percentage point as metropolitan area population increase by one million people The employment density is also researched with conclusion of an increase by

100 workers per square mile causes by about 0.25 percentage point in probability of working in a related job

Thirdly, the learned major in school is expected to have a significant impact on the probability of having mismatched job The probability of mismatch job of the major with general knowledge and skills is expected lower than the major with specific knowledge and skills Workers who learned in general majors have more choices for their career because there are many jobs can be suitable with their major Following Robst (2008), there are some majors which focus on occupation requiring specific skills (such as architecture, doctor) so that there is less chance for them to have a related job and less probability for completely match career in these majors On the other hand, the occupation general skills and knowledges which focus on general human capital increase the transferable in career and workers can work in many related field with their schooling major

2.2 Over-education and under-education (Vertical education-occupation

requirement More academic definitions are noticed by Rumberger (1981) in two ways Firstly, over-education is considered as a decrease in the economic position of white-collar workers relative to historically higher levels, particularly in monetary aspect Secondly, over-education can be defined as unrealized expectations related to benefit of education This definition is considered with a conception that every students have their own expectation about future job but this expectation may not be come true after graduation However, both two definitions from Rumberger (1981) are quite weak because they ignored some important components such as: non-monetary aspect of schooling, change over time of expectation and difficulties in measuring individual’s expectation Thusly, the first definition from Dolton and Silles (2008) seems the best one which notices that productivity and earnings associated with job characteristics, not individual performance

Above definition of over/under-education is identified through a comparison between attained education and required education Thus the question need to be clarified is what required education is and how to measure it Hartog (2000), Dolton and Silles (2008) summarized three ways to measure required education in their own researches The first

is an analyst of skills or knowledge requirements for each occupation The second is worker’s self-assessment in survey which reflect own thinking about his educational requirement The third using mean or mode of education level across a range of occupations as the basic to classify over-education In this method, average or modal value of education for the occupational group are considered as the bases in there a worker is considered over-educated if he has education level varying by one or two standard deviations from the bases

The existence of over-education can be explained by neoclassical economic theory where enterprises make input decisions and production with given technology and relative prices With the assumption of zero-information cost, the labor market will respond promptly to a change in relative labor supply and reach new equilibrium price

of labor In other words, an increase in graduate supply will reduce relative wage and enterprises can adjust production structure to take advantage of cheaper and more abundant skilled labor force Nevertheless, in worker’s aspect, they will redesign their investment plan and expectation if they recognize that additional investment in education has a smaller rate of return than their belief or alternative investment After adjustments

of enterprises and labor, the skills of employees will be fully utilized in the long run In other words, over/under education only exist in short term when there is a temporary disequilibrium between supply and demand in labor market

However, how long is the long run for this adjustment of labor market? According to Tsang and Levin (1985), in spite of lower wage, individuals continue to invest in education if they think their private rate of return of this investment stay high enough This can happen if wages of lower education levels are turning down and increasing speed of unemployment of these levels are equal or higher than that of the higher level Another potential is that if workers have higher education, they have higher probability

to be in the upper tail of earnings distribution The last potential is expressed by worker’s expectation that they think falling in rate of return to higher-education as a temporary trend and it will be better in long-run In conclusion, the long-run equilibrium of labor market only can be reached in a distant future

Furthermore, the existence of over-education can be also explained as a symptom of human capital deficits A worker can use his over-schooling to substitute or compensate for deficiencies in other fields of human capital which are not only simply knowledge in school but also work experience and on-the-job training (Sloane et al., 1996) As mentioned in occupational mobility theory, if these deficiencies can be corrected by experience and on-the-job training, over-education will be eliminated day by day Nevertheless, over-education is a long-term problem which is correlated to differences

in permanent ability across graduates

According to job-competition model of Thurow (1975), imaging there are two queues: one for jobs and one for candidates Each job in the job queue has specific requirements, productivity characteristics and payment range In candidate queue, each person has his own position which depends on his education and experience The higher position in queue, the higher probability to have expected job in job queue So that, individual tries

to increase his education level to have a better position in candidate queue with desire to have a job which even be underemployed his knowledge and skills

There is another explanation for the existence of over-education in labor market such as the study of Jovanovic (1979) This study indicated that the imperfect in information of labor market which causes mismatching between employer and employee about worker’s productivity is the root of this problem A worker can temporarily accept a job which requires less knowledge and skills in order to express his truly productivity Another one came from Frank (1978) and McGoldrick and Robst (1996) that geographical differences cause over-education Employees who are restricted in a specific labor market have higher risk to work in an over-educated job than workers in a large labor market The last one mentioned here came from Robst (1995) which mentioned quality of studying He argued that if a person attends a low-quality university get less knowledge and skills than person in a high-quality university So that, over-education does not mean over-qualified for the job, an over-education can be necessary

to meet the requirement of job

2.2.2 Determinants of vertical education-occupation mismatch

Dolton and Silles (2008) examined the determinants of probability of education, including: studying major, class of degree (denotes how well a student passes the final exam: at the first, second or third time), studying qualifications (denotes what education level a person are being: post-graduate, degree, sub-degree, no qualifications) and professional qualifications (including academic, professional or vocational) Furthermore, he noticed some determinants related to the existing job, including:

over/under-working sector, occupation (denotes manager, professional, associate professional), firm size, employment characteristics (self-employed, part-time and full-time), training, experience for current job and squared experience Moreover, macroeconomic variables are found to have significant impact: unemployment, national statistics of graduates (include university participation rate and graduate unemployment rate), labor market mobility (job-motivated change) And the last group is personal characteristics (includes gender, age, partner and child)

Dolton and Silles (2001) indicated that studying major have an important impact on the likelihood of mismatched education There is a theory that graduates in less vocationally-oriented qualification faculties have more tendency to work in an over-educated job For instance, Dolton and Silles (2001) found that graduates in faculty of education have more probability of working in adequate job with their educated level than people in other faculties He also pointed out a higher likelihood of being over-educated of graduates in arts, humanities and languages Furthermore, class of degree reflects many unmeasured characteristics of workers, especially ability Most employers have the tendency to hire workers with high ability which is considered as higher class of degree to reduce on-the-job training cost

Following Dolton and Silles (2008), employees who are working in a small enterprise with less than 25 workers have more probability in over-educated job In his own research in 2001, he argued that there is a more professional recruitment system and more jobs in large firms than in small firms Therefore, probability of mismatched between education level and job’s requirements is reduced Furthermore, working sector

is also thought as a significant determinant which affects mismatched probability through differences in nature of competition For instance, employees in public sectors may have more risk to be over-educated because of low competitive working environment

Beside job related indicators, human related characteristics also have a remarkable part

in job-match probability Firstly, different types of family engagement have different effect on over/under education or in choosing a related job For instance, men may have more responsibility for financial support in family so they can choose a high-earnings job which does not make full use of their skills and abilities (underemployed job) On the other hand, women seem to have more responsibility for taking care of their family’s members so they have tendency to choose an underemployed job with flexible working time Moreover, Dolton and Silles (2001) indicated that graduates who face family engagement at an early age may have larger influence on mismatched likelihood than individual with this engagement at older age The reason can come from the trade-off between high search cost and current consumption demand so fresh graduates have to find any job despite of mismatched education level Secondly, impact of marital status

on the probability of over-education can be explained through the effect of his partner’s job For example, a worker can be limited in job-searching range because of his partner’s work place Thirdly, health status has a mixed impact on mismatched career On the one hand, disable graduates can face many difficulties in searching an adequate job because

of limitation in moving or passing working stress On the other hand, they can find a related job easier than a graduates with good health because of subsidize policies of government For instance, article no.35 (Vietnamese disability law, 2010) encourages enterprise hiring disable workers whereas these firms can have priority such as exempting enterprise income tax, borrowing capital with low interest rate, lessening land rental cost, and many other policies following the proportion of disable workers in firm Finally, debt problem in studying process may make graduates find an immediate career

to finance these loans although this career is not fit to his education level

Another determinant is the imperfect information whereas candidates cannot know exactly the actual required education level of job they are applying for The only thing they know is required level to entry job which is estimated by company This difference

in requirement of job causes many workers realize that they are over/under educated in

existing job

2.3 Effect of education-occupation mismatch on earnings

2.3.1 Mincer’s earnings model

Mincer (1958) applied the concept of compensating differences as the main idea to clarify why individuals with distinct schooling levels received distinct earnings In this research, he assumed there are the identical abilities and opportunities among individuals, so that difference in required occupation’s level hence of in their training has different compensation And size of this compensation differential is calculated by subtracting cost of different human capital investment from present value of earnings flow Mincer (1958) argued that human capital investment expenses depend on the length of schooling years in two aspects: the deferral of earnings in training period and educational services/equipment cost (include tuition fee and books cost) In this research, educational services/equipment cost is assumed to be zero to simplify the calculation Furthermore, individual’s wage is assumed to be unchanged during working life With above assumptions, the present value of life-earnings is equated:

V n = a n∫ (𝑒𝑛𝑙 −𝑟𝑡)𝑑𝑡 = 𝑎𝑛

𝑟 (𝑒−𝑟𝑛− 𝑒−𝑟𝑙)

Where:

Vn is present value of earnings flow with s years of schooling

an is annual earnings of individual with s years of schooling

r is externally interest rate at which future earnings are discounted

l is length of working life plus length of schooling for all persons, it means length of working life of persons without education

By comparing present value of life-time earnings, Mincer (1958) found that differences

in annual earnings is the result of differences in the length of training (schooling years) Equating the present values of two individuals with different schooling years Vn = Vn-d,

the ratio of annual earnings kn,n-d indicated that individuals with more training require higher earnings

Up to Mincer (1974), this relationship is equated by setting the logarithm of wage as a function of education years and potential working experience years(age minus schooling years minus six) and quadratic experience According to the human capital theory, earnings are not affected by requirements of a particular job, but affected by characteristics of workers And Mincer model directly reflects this theory through its independent variables which include worker’s characteristics (education, experience) Moreover, schooling coefficient in this model is closely related to marginal internal rate

of return to education – an important measure when comparing to return of other investments

Ln w(s,x) = α0 +ρss + β0x + β1x2Where: x is working experience years

s is schooling years

To have above equation, Mincer (1974) focus on life-cycle dynamics of earnings and the relationship among observed earnings, potential earnings and human capital investments (include schooling training and on-the-job training) More precisely, observed earnings are collected by subtracting human capital investment expenses from potential earnings where potential earnings in time t+1 depend on investments in time t

As the summary from Lemieux (2003), simple Mincer model remains an effective and suitable standard in a stable environment where educational achievement develops smoothly over consecutive labor force However, in current period, Mincer model has some troubles in fitting data because of unstable environment More precisely, there are some main problems in basic Mincer model, including (1) endogeneity due to an ability bias; (2) non-linear relationship between earnings and education; and (3) distinct experience premiums for individuals with distinct education level

Firstly, Harmon et al (2003) argued that students with higher abilities have tendency to receive more schooling and more earnings It causes a correlation between schooling and wage that is not explained in causal link of Mincer model Many solutions are considered and the most popular one is to use instrument variable (IV)

Secondly, Mincer (1997) argued that relationship between education premiums and earnings is a convex function This non-linear effect can be explained by two potentials: the logical reaction to a relative increase in human capital demand and “sheepskin” effects “Sheepskin” effects indicate that getting a certificate is more valuable than non-certificated education For instant, an individual completes a four-years course with a certificate can get very larger earnings premiums than one with only three out of four years of the same course

Thirdly, experience premium is assumed to constant for everybody regardless of their education level in basic Mincer model However, Heckman et al (2003), Belzil (2008) found evidences to reject the parallel relationship in the earnings-experience profile for all education levels They said that there is the divergent wage growth paths but they cannot give the solution The reasons for this non-parallel growth paths can come from the relatively high compensation for highly educated workers in type of high experience premium Another possible reason is the higher productivity growth of high-educated workers which is reflected in experience premium

Although this model remains relatively accurate in relationship among earnings, schooling and experience, it needs some extensions These changes are not for only robustness of model in new environment, but also for new using purposes To solve endogeneity due to an ability bias, some common IVs are applied, including: distance to schooling and spouse’s education which correlate closely with schooling but are not correlated with ability or wages And to solve non-linear relationship between earnings and education, a set of dummy variables which reflects different completed education levels is added to model

Furthermore, an adjustment is mentioned in almost researches which is about effect of over-education/under-education on earnings In this case, Mincer’s model is used with a small change in variable structure More precisely, return to required years of education and returns to years of educational mismatch are used instead of return to attained years

of education Moreover, some additional control variables are added to model such as: demographic indictors (gender, age, marital status, ect), schooling major And a special independent variable in this research is mismatch career which indicate the matching status between schooling major and current job

2.3.2 Wage effect of horizontal education-occupation mismatch

This section investigates the economic literature on the impacts of mismatch between occupation and schooling major on wage Following Shaw (1984), working in unrelated job with the university field, the less learned skills are applied in career, the lower wage for mismatch workers

More precisely, the level of wage decrease depends on the studied majors In university, some majors teach specific skills in one field which are not transferable to other fields

It causes the negative effect on wage level when working in unrelated field with schooling major Besides, there are some majors with general skills which are transferable such as foreign languages, social sciences, liberal arts In these majors, the wage effects of mismatched career are smaller than in majors with specific skills

Furthermore, in majors with lower mismatch proportion, the cost of mismatch in form

of lower wage is greater

Furthermore, there are many kinds of mismatch career based on the reasons for accepting the mismatched position, which have different effects on earnings Following Robst (2007), among supply-related reasons, wage effects from career oriented reasons are expected to be smaller than amenity reasons Moreover, Robst (2007) indicates that workers who work outside of schooling majors because of pay and promotion opportunities have higher wage compared to workers who works in well-matched jobs This finding give no surprise when the reason they accept these jobs is higher earnings The career mismatch because of demand-related reasons which means workers cannot find a related job is expected to be negative in his research

About the impacts of demographic indicators on wage effects of mismatched career, the difference in gender have a significant difference wage effects Robst (2007) did confirm this difference at the level of 10% using the data of USA graduates

2.3.3 Wage effect of vertical education-occupation mismatch

Effect of mismatched education on earnings level can be clarified based on the reasons

of mismatched in different theories In human capital theory, earnings of individual are paid based on workers’ productivity which is determined by their human capital instead

of job characteristics (Dolton and Silles, 2001) In the above section, one of reasons for over-educated workers is excess of skilled-labor force supply which causes relative decrease in their wage According to job competition theory, wage is determined by job characteristics where individuals’ position in person queue is determined by his education and experience level In this model, employers have a tendency to hire over-educated worker to save training cost

By using over/under educated years, Duncan and Hoffman (1981) found that returns to over-education are significantly positive but it has half size compared to returns to required education In this research, Duncan and Hoffman (1981) applied self-

assessment where individuals reported about required education level of their job This lower returns to over-education than returns to required education is also found by Rumberger (1987) when he use U.S data set in 1969, 1973 and 1977

Another model is used by Verdugo and Verdugo (1989) for U.S sample in 1980 which use dummy variables for over/under education They found that overeducated employees have lower wage than those who have adequate education Furthermore, undereducated employees are found to have higher earnings than overeducated ones It can be explained

by more excellent performance on the jobs of under-educated workers while educated workers do not mean higher productivity

over-There are not much of the literature investigating the effects of over/under education on earnings in developing countries Quinn and Rubb (2006) made research in Mexico labor market from 1987 to 1997 and indicated consistent results with researches in developed countries They designated a lower returns to over-education than returns to required education Another research is applied on Pakistan for period from 1998 to 2004 by Abbas (as cited by Bedir, 2014) He also found the positive returns to over-education whereas negative returns to under-education Moreover, both values of returns to over- and under-education is lower than returns to required education

However, there is a paper of El-Hamidi (2010) that examined data from Egypt in the period of 1998 and 2006 for the private sector had some different results He represented

a higher earnings of over-educated workers compared to adequate educated ones This result is explained by an argument that employers consider more education years as an indicator of less on-the-job training cost in the future So that employers tend to hire over-educated workers to save future training cost

Based on previous researches, there are some expectations about effect of over/under education on earnings Firstly, the return to over-education is positive but lower than return to required education Secondly, the return to under-education is negative and its absolute value is lower than return to required education

on earnings

3.1.1 Horizontal mismatch and earnings

In this thesis, a process with two steps is applied: identify factors which affect probability

of mismatch career and its impact on earnings

Step 1: To have a clearly view about indicators which affect mismatch job’s probability, the ordinal logistic regression is estimated:

Pr (Mismatch)i = Xi β + Z α + εi (1) where the dependent variable is ordinal: 1 = related job, 2 = partial mismatch job, 3 = completely mismatch job

X includes demographic variables: age, male, married status, child quantity, no activity limitation, health status, highest degree level (include vocational certificate, college, bachelor, master, doctor), working firm type (include state/government institution, state owned enterprise, private enterprise, FDI enterprise, self-employed business), immigration (equal 1 if working place is born place and 0 otherwise), fulltime job (equal

1 if full-time job and 0 if part-time job), working place (equal 1 if work in Ho Chi Minh city and 0 otherwise) for individual i

Z is the schooling field Details on variables and their expected sign on probability to be

in completely mismatched career are presented in Table 3.1

Robst (2006) argued that the schooling fields which provide general skills have less mismatch chance than fields which provide occupation specific skills Also based on this argument reality in Vietnam, four main groups with details major are classified in this thesis as below

 Business: Economics, Admin - Human resources, Travel-Hotel and Catering

 Liberal Arts: Communication, Architect – Art, Linguistic, Legal profession, Environment, Cuisine, Sport

 Science: Information Technology, Agriculture science, Biological – Chemical science, Natural science research (Mathematic, Physic), Medical – pharmaceutical industry, Technical (Electric, mechanic), Construction

 Education

Furthermore, Canes and Rosen (1995) indicated that there is a significant difference in choice of schooling majors and career based on gender Based on this, the regressions in this thesis are also run separately for men and women

Table 3.1: Determinants of horizontal mismatch probability

sign

male dummy take value 1 if respondent is male and 0 otherwise +/- married dummy take value 1 if respondent married and 0 for single one - child children Number of children of respondent +/-

active dummy take value 1 if respondent has no activity limitation and 0 otherwise +/- health dummy take value 1 if respondent has good health and 0 otherwise +/- college dummy take value 1 if highest degree of respondent is college certificate and 0

liberal art dummy

take value 1 if respondent has schooling major is liberal art (include Communication, Architect – Art, Linguistic, Legal profession, Environment, Cuisine, Sport) and 0 otherwise

-

take value 1 if respondent has schooling major is science (include Information Technology, Agriculture science, Biological – Chemical science, Natural science research (Mathematic, Physic), Medical – pharmaceutical industry, Technical (Electric, mechanic), Construction) and 0 otherwise

Step 2: To estimate the wage effect of mismatch, a vector of demographic and kinds of mismatch are considered as explanation variables A standard wage regression is applied:

Where X includes demographic variables: age, male, married status, child quantity, no activity limitation, health status, highest degree level (include vocational certificate, college, bachelor, master, doctor), working firm type (include state/government institution, state owned enterprise, private enterprise, FDI enterprise, self-employed business), working place (equal 1 if work in Ho Chi Minh city and 0 otherwise), fulltime job (equal 1 if full-time job and 0 if part-time job) for individual i

Z is the schooling field

Y denotes for a dummy category of mismatch kinds, including:

 Completely related job

 Partial mismatch job because of demand related reasons

 Partial mismatch job because of supply related reasons

 Completely mismatch job because of demand related reasons

 Completely mismatch job because of supply related reasons

Details on variables and their expected signs on earnings are presented in Table 3.2

Table 3.2: Variable list – effect of horizontal education mismatch on earnings

sign

male dummy take value 1 if respondent is male and 0 otherwise + married dummy take value 1 if respondent married and 0 for single one + child children Number of children of respondent + active dummy take value 1 if respondent has no activity limitation and 0 otherwise + health dummy take value 1 if respondent has good health and 0 otherwise + college dummy =1 if highest degree of respondent is college certificate and 0 otherwise + bachelor dummy =1 if highest degree of respondent is bachelor certificate and 0 otherwise + master dummy take value 1 if highest degree of respondent is master certificate and 0

Human resources, Travel-Hotel and Catering) and 0 otherwise +/-

liberal art dummy

take value 1 if respondent has schooling major is liberal art (include Communication, Architect – Art, Linguistic, Legal profession, Environment, Cuisine, Sport) and 0 otherwise

+/-

science dummy

take value 1 if respondent has schooling major is science (include Information Technology, Agriculture science, Biological – Chemical science, Natural science research (Mathematic, Physic), Medical – pharmaceutical industry, Technical (Electric, mechanic), Construction) and 0 otherwise

+/-

dempartmis dummy = 1 if worker is partly mismatched because of demand side reasons, 0

suppartmis dummy =1 if worker is partly mismatched because of supply side reasons, 0 otherwise +

demmis dummy Take value 1 if worker is fully mismatched because of demand side reasons and

supmis dummy Take value 1 if worker is fully mismatched because of supply side reasons and

govins dummy =1 if respondent is working in a state/government institution and 0 otherwise +/- stateent dummy =1 if respondent is working in a state owned enterprise and 0 otherwise +/- indent dummy =1 if respondent is working in an private enterprise, 0 otherwise +/- fdi dummy take value 1 if respondent is working in a FDI enterprise and 0 otherwise +/- fulltime dummy take value 1 if respondent is working a full-time job and 0 otherwise + workhcm dummy take value 1 if respondent is working in Hochiminh city and 0 otherwise +

3.1.2 Vertical mismatched and earnings

Firstly, the determinants of over/under education are estimated based on definition of dependent variables as below:

𝐸𝑖𝑜is years of over-schooling 𝐸𝑖𝑜 = {𝐸𝑖 − 𝐸𝑖𝑟 , 𝑖𝑓 𝐸𝑖 > 𝐸𝑖𝑟

0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

𝐸𝑖𝑢 is years of under-schooling 𝐸𝑖𝑢 = {𝐸𝑖𝑟 − 𝐸𝑖, 𝑖𝑓 𝐸𝑖 < 𝐸𝑖𝑟

0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒Where :

 𝐸𝑖𝑟 is required years of schooling

 E i is actual years of schooling

There is left-censored distribution of over-education and under-education because of this definition of vertical mismatched So that, Tobit model is applied to examine determinants of over/under education

Z is the schooling field

There are two basic methods which are applied to determine effects of over-education and under-education on earnings The first model is developed by Duncan and Hoffman (1981), comparing mismatched worker with worker in the same occupation

ln W i = β0 + β1𝐸𝑖𝑟+ β2𝐸𝑖𝑜 + β3𝐸𝑖𝑢+ X i γ + εi (4)

Where W i denotes the log of wages of individual i, 𝐸𝑖𝑟 is required years of schooling; and

E i is actual years of schooling

In this model, β1 is return to years of required education, and β2, β3 are return to an additional year schooling over requirement and to a year of schooling under requirement, respectively Applying this model, there are two main results that returns to surplus schooling is positive (β2 >0) but smaller than returns to required schooling (β1 > β2) and returns to deficit schooling is negative but its absolute value is smaller than returns to required schooling (β1 > - β3) These expected signs are presented in table 3.3

The second model is developed by Verdugo and Verdugo (1989) which use dummy variables to denote mismatched in schooling level

ln W i = α0 + α1E i + α2OV i + α3UN i + X i γ + εi (5)

where W i denotes the log of wage of individual i E i is years of education actually

attained OV i is dummy variable which takes the value 1 if an individual is over-educated

and UN i are dummy variable which takes the value 1 if an individual is under-educated

X i is a vector of other explanation indicators

In this model, mismatched workers are compared to similar workers who have the same schooling level and work in a job that requires exactly his attained education level In this case, if productivity and wages are determined by only required education level, α2

and α3 are expected to be zero Besides, if wages are determined by required education level of a job, over-educated schooling years are unproductive and there is no reward for these exceeding years It means over-educated workers are expected to earn less than similar workers who have same schooling level but are working in a suitable job and thus α2 is expected to be negative In the same way, an under-educated worker is expected to have higher earnings than similar worker who has same education level but

is working in his fixed-schooling level job and α3 is expected to be positive (Table 3.3)

Table 3.3: Variable list – effect of vertical mismatched education on earnings

sign

male dummy take value 1 if respondent is male and 0 otherwise + married dummy take value 1 if respondent married and 0 for single one + child children Number of children of respondent + active dummy take value 1 if respondent has no activity limitation and 0 otherwise + health dummy take value 1 if respondent has good health and 0 otherwise + college dummy =1 if highest degree of respondent is college certificate and 0 otherwise + bachelor dummy =1 if highest degree of respondent is bachelor certificate and 0 otherwise + master dummy =1 if highest degree of respondent is master certificate and 0 otherwise + doctor dummy =1 if highest degree of respondent is doctor certificate and 0 otherwise + majorexp year experience years in current working field + majorexp2 squared experience years in current working field - business dummy =1 if respondent has schooling major is business (include Economics, Admin

- Human resources, Travel-Hotel and Catering) and 0 otherwise +/-

liberal art dummy

take value 1 if respondent has schooling major is liberal art (include Communication, Architect – Art, Linguistic, Legal profession, Environment, Cuisine, Sport) and 0 otherwise

+/-

science dummy

take value 1 if respondent has schooling major is science (include Information Technology, Agriculture science, Biological – Chemical science, Natural science research (Mathematic, Physic), Medical – pharmaceutical industry, Technical (Electric, mechanic), Construction) and 0 otherwise

3.2 Data source

The survey is collected from 267 respondents via two ways: face to face interview and interview through internet Respondents is collected randomly though survey on internet; and face to face interview is set up for workers who are working in Ho Chi Minh city Respondents are employees who have at least vocational degree in a specific major In case of a person has more than one schooling major, the latest major is collected because

it is most close to his career The collected wage is net wage respondents earn per month Earnings in survey is divided into nine intervals; then average value for each interval is used as wage level of respondent This method of questionnaire give more exactly result and make respondent more comfortable when responding their wage level The survey was implemented from September to December of 2015 The questionnaire is presented

in Appendix 5

CHAPTER 4

RESULTS 4.1 Descriptive statistics

4.1.1 Horizontal education-occupation mismatch

The data is collected from 267 respondents with averaged 28 years old and 16.73

schooling years in average as shown in Table 4.1

Table 4 1: Descriptive statistics of continuous variables

With the classification of mismatched job in three level: completely matched, partly mismatched and completely mismatched, the percentage of people who are working in related job to their schooling major is only 48% It means there is over 50% worker in this survey are working in unrelated field or partly related field to their schooling majors Applying t-test for variable “age”, there is a significant difference in age level between completely matched group and partly matched group with mean of 29.09 years old and 27.31 years old, respectively (table 4.2) Nonetheless, these difference is not considerable between completely matched and mismatched group