過度教育與學用不符對台灣高等教育畢業生的薪資與工作滿意度影響 - 政大學術集成

全文

(1)國立政治大學社會社學學院應用經濟與社會發展英語碩士學位學程 International Master’s Program of Applied Economics and Social Development, College of Social Sciences National Chengchi University. 碩士論文政治. 立 Master’s Thesis. 大. ‧. ‧ 國. 學 er. io. sit. y. Nat. 過度教育與學用不符對台灣高等教育畢業生的薪資與工作滿意度影響. n. al The Effect of Overeducation and Skill v Mismatch on i n Ch Wage and Job Satisfaction Graduates in i U e n g cofh College Taiwan. Student：梁家瑜 Chia-Yu Liang Advisor：莊奕琦 Yih-Chyi Chuang 中華民國 108 年 1 月 January 2019. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(2) Acknowledgments In this work, I would like to grant the greatest gratitude to my advisor, Professor Yih-Chyi Chuang. Though he was mostly tied up, he has supported me for all academic consultancy and was willing to discuss my thesis as frequently as I wanted. I have learned considerably from his advice and realized his demand on writing a dissertation facilitates the organization and the precision of my research. To alleviate my econometric and STATA technical proficiency, Professor Po-Chun Huang has also helped a lot. On the other hand, I appreciate Chen-Kai Liu has assisted me the adjustment of layout to make my thesis readable. Plus, the psychological supports from my family and friends, I-Ning Chen, Che-Hsu Chang, Kai-Chieh Huang and so on. Without any of these people around me, I could barely succeed to finish my study. Master degree is a totally different journey in my life. I left my comfort zone and. 政治大. pushed myself to limitation. It nurtures me to be a more perseverant and open-minded person.. 立. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(3) Abstract Overeducation and skill mismatch are two incidents commonly occurring when job-seeking. Imperfect information and the failure of market mechanisms can result in such mismatch. Job mobility and job competition theories state that overeducation can be voluntary, as jobseekers want to be more competitive than other candidates or expect a more promising career. Human capital theory points out that it is likely that graduates choose to be overeducated because they have too little experience and feel unprepared to do an adequate job. Overeducation, in particular, has been an issue since the expansion of higher education. Assignment theory explains that an increase in skilled labor brings about a lower wage, on average. These issues are strongly associated with the education and labor markets. This has definitely been a concern in Taiwan since the 1980s. These two items have the potential to. 政治大 Heckman two-step, and inverse probability-weighting (IPW) estimates will be used to 立 analyze wage effect. Probit models will be used to study job satisfaction. The first section of influence both wages and job satisfaction. In this research, ordinary least squares (OLS),. ‧ 國. 學. the research will use three methods and examine the effects on the entire highly-educated labor of the Taiwan Education Panel Survey-Beyond (TEPS-B) from the 2014 interview.. ‧. After comparing three estimates, one will be adopted in further research of wage effect, and a. y. Nat. probit model will be used for job satisfaction. This further research will stratify highly-. sit. educated graduates into six groups: gender, ranking, profession, institution, ownership, and. al. er. io. industries. In the specification of gender, women claim less overeducation but have a higher wage penalty than men. Top 10, medical, and public graduation have stronger effects on. n. v i n C that wages and satisfaction. This indicates elite universities have an impairment h eTaiwanese ngchi U regarding educational distribution and therefore not every graduate takes a job as he or she. expects. Third industry workers have a lower effect on wage penalty and are more willing to be overeducated or skill mismatched. Keywords: Taiwan, TEPS-B, wages, job satisfaction. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(4) Table of Contents. 1. Introduction. 1. 2. Literature review. 2. 3. Data. 4. 4. Methodology. 13. 5. Result. 16. 6. Discussion. 29. 7. Conclusion. 33. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(5) List of Tables Table 1: Covariates. 5. Table 2: OLS wage equations. 16. Table 3: Wage equations of Heckman two-step method. 19. Table 4: IPW model of wage. 22. Table 5: Probit model of satisfaction. 23. Table 6: IPW model of satisfaction. 25. Table 7: Genders. 26. Table 8: Rankings. 27. Table 9: Professions Table 10: Institutions. 立. 27 28. ‧ 國. 學. Table 11: Ownership. 政治大. Table 12: Industries. 28 29. ‧. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(6) List of Figures Figure 1: Ratio of overeducation and skill mismatch of genders. 9. Figure 2: Ratio of overeducation and skill mismatch of rankings. 10. Figure 3: Ratio of overeducation and skill mismatch of three professions. 11. Figure 4: Ratio of overeducation and skill mismatch of two institutions. 11. Figure 5: Ratio of overeducation and skill mismatch of different ownership. 12. Figure 6: Ratio of overeducation and skill mismatch of two industries. 12. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(7) 1. Introduction Educational mismatch is the disparity between actual education level and job-required education. Undereducation and overeducation both belong to this criterion. Between the two, overeducation indicates a potential issue in the labor market. It is likely to imply a problem of signaling between job-seekers and employers. Alternatively, it possibly results from a lack of experience on the part of new entrants into the labor market. Job-seekers may want to impress potential employers with more robust qualifications than other candidates. It may even be a serious incidence of labor misallocation related to the expansion of higher education. The expansion of higher education in most developed and developing countries has given rise to the problem of overeducation. The rocketing supply of qualified workers graduated from universities does not match the level of demand in the labor market. Therefore, the excess of. 政治大. highly-educated laborers either emigrate or are underemployed. Such workers are considered. 立. overeducated or overqualified (Lin and Wang, 2015; Hartog, 2000; Croce and Ghignoni,. ‧ 國. 學. 2012).. Skill mismatch is called “actual mismatch” by Allen and Van der Velden (2001). It is a. ‧. mismatch of skills learned and skills utilized in one’s career. It may reflect a considerable. sit. Nat. and labor allocation can be problems leading to skill mismatch.. y. issue of the labor market. Excepting the preference of job-seekers, information asymmetry. al. er. io. In a U.K. study, 36 percent of highly-educated graduates claimed overeducation in their initial jobs compared to 14 percent in other European countries. In the same study, 33 percent. n. v i n Cpercent felt skill mismatched compared to 17 European counterparts (McGuinness h e nofgtheir chi U and Sloane, 2011).. In Asia, Taiwan is famous for low tuition fees for tertiary education. Since the expansion of higher education in the 1980s, most Taiwanese are able to attend a university. Compared to western European countries, the Taiwanese economy is not as strong. The more prosperous an economy is, the higher the demand for labor is, and highly-skilled laborers are well-aware of this. Enrolment in tertiary education is therefore incredibly high. This implies that overeducation may be a serious issue in Taiwan. According to a study by Lin and Wang (2015), overeducation rose from 26.6 percent in 1993 to 35.0 percent in 1999. Overeducation and skill mismatch not only influence wages but also job satisfaction, which can affect the mental health of laborers. In terms of both economic conditions and labor, psychological health is always of importance in society. 1. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(8) University graduates possess diverse potentialities having received various sorts of education from their universities. The resulting differences affect employment opportunities and wages. Graduates of top university receive a higher quality education and may be more competitive than those who did not graduate from a top university. However, the phenomenon of overeducation continues to worsen. Will this more seriously affect graduates of top universities? Also, are graduates from different systems affected differently by wages? In this research, the influence of overeducation and skill mismatch on wages and job satisfaction among highly-educated graduates in Taiwan is the focus of study. The introduction of the Taiwan Education Panel Survey and Beyond (TEPS-B) database and the ratio of overeducation and skill mismatch generalizes the fundamental condition of Taiwan in 2014, after a literature review. Later, the first part of the research will examine original. 政治大 conduct further study in different categories. At the end of this thesis, a conclusion and policy 立 implications will be addressed. models by estimating three methods. One of the three will then be selected in order to. ‧ 國. 學. 2. Literature review. ‧. In related research, determining whether an individual is overeducated can be either objective. y. Nat. or subjective. The U.S. example of an objective measure is supported by professional analysts. sit. to determine the required educational levels for jobs in the Department of Labor’s Dictionary. er. io. of Occupational Titles (DOT). Thus, those not having the same level of education recorded in. al. the DOT are regarded as having an educational mismatch. The subjective measure evaluates. n. v i n C questions educational mismatch through survey as “what level of education does your h e n gsuch chi U job require in your opinion?” A different measure will be employed in this paper as no. professional database or direct questionnaires are available. This measure emphasizes the education level of each subject, and actual years of schooling above the distribution mean are deemed to be overeducated. However, this way of measuring will lead to a different interpretation of results since it is based on actual labor market data, and therefore endogeneity plays a role (Hartog, 2000; Lin and Wang, 2005). There are three theories explaining why new labor market entrants tend to be overeducated. First, under the competition model, new entrants are willing to be overeducated in order to compete for a better job (Thurow, 1975; Hung, Yin, and Tao, 2015; Caroleo and Pastore, 2014). The second is signaling theory (also called search and match theory). On this view, new entrants are inclined to have information asymmetry in job seeking (Hartog, 2000; Allen 2. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(9) and Van der Velden, 2001). Employers judge candidates by observed characteristics, usually the universities from which new entrants have graduated (Thurow, 1975; Allen and Van der Velden, 2001). Human capital theory offers another explanation: recent graduates have too little job experience, and therefore they tend to be overeducated and earn lower wages (Groot and Massen van der Brink, 1996; Allen and Van der Velden, 2001; Caroleo and Pastore, 2014). All three theories conclude that overeducation is temporary and will alleviate with time spent in the labor market. Based on the study by Hung, Yin, and Tao (2015), using panel data that included master’s degree graduates, highly-educated Taiwanese usually choose to be overeducated to substitute for experience. They become more adequately educated as they stay in the workplace longer. Job mobility theory is taken into consideration as well when an individual opts for. 政治大 easier time earning promotions or increasing their wages in the future (Caroleo and Pastore, 立 2014; Leuven and Oosterbeek, 2011; Sicherman and Galor, 1990).. overeducation. Those choosing to be overeducated or skill mismatched may expect to have an. ‧ 國. 學. According to Duncan and Hoffman (1981), education level reflects productivity. An increase in the supply of highly-skilled labor will definitely lead to lower wage equilibrium. Taking. ‧. the same job requires the same level of productivity. Though those who are overeducated are. y. Nat. more productive, they should be paid at the same level as those who are adequately educated.. sit. The surplus education that overqualified workers possess has a positive effect on individual. al. er. io. wages. However, in the same study, adequately-educated workers have wages twice as high. v i n C hand Van der VeldenU(2001) posit that skill mismatch is resources. Based on this theory, Allen engchi n. as those who are overeducated. This indicates a possible misallocation of education. also an aspect of this since skill mismatch is related to productivity and has an effect on wages. As a result, educational mismatch has a stronger influence on wages than skill mismatch. Human capital theory, however, casts doubt on this. From the premise of the. theory, people with the same level of education will be screened out in order to obtain a more fitting job. Hence, unobserved characteristics may be important factors for determining wages. Skill mismatch, on the other hand, is a stronger indicator of negative job satisfaction, which provides an incentive to switch jobs. On the other hand, it is likely that as progress occurs in a certain type of job, the required educational level moves upward but is still recognized in the old paradigm. Otherwise, lower educational qualifications can be blamed for overeducation and attributed to the inefficiency of market mechanisms (McGuinness and Sloane, 2011; Verdugo and Verdugo, 1989; Caroleo 3. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(10) and Pastore, 2014). When estimated separately, overeducation accounts for an almost 40 percent and overskill has a 22 percent wage penalty; while the effects are lower and become 34 percent and 10 percent when estimated in the same regression. Female U.K. graduates experience a 37 percent wage penalty, meaning that their overeducation has more impact than for males (31 percent). In contrast, the wage penalty of skill mismatch for males is worse, at 10 percent, than for women, at only 1 percent. Though most of the previous research has found overeducation to be a wage disadvantage, job satisfaction has not necessarily been found to align similarly with the effects of overeducation. However, skill mismatch has been found to cause more discontent. One likely interpretation is that overeducation is voluntary, but skill mismatch is not (McGuinness and Sloane, 2011). The outcomes of the research in this paper are similar to the above studies, but it does not offer a clear argument for the assumption. It only implies that assignment theory surely plays a role.. 3. Data. 立. 政治大. ‧ 國. 學. Database and main variables. This research employs the database of Taiwan Education Panel Survey and Beyond (TEPS-. ‧. B). TEPS-B is supervised by Academia Sinica of Taiwan. It has collected four-wave panel. y. Nat. surveys of two Taiwanese cohorts born from 1984 to 1985 and from 1987 to 1988 so far.. sit. Continuing the panel survey from Taiwan Education Panel Survey (TEPS), working on the. al. er. io. same cohorts and investigating the observations, their parents and teachers in junior or/and senior high school, TEPS-B aims at the follow-up academic enrollment and connects to the. n. v i n C hborn from 1987 toU1988 will be the research objects labor market. However, only the cohort engchi in the study because there are no associated questions regarded as the indicators of overeducation or skill mismatch. The cohort born from 1987 to 1988 has received investigation from junior high school and been divided into two categories, Panel-CP1 and Panel-NCP1, in senior high school. Only Panel-CP1 has offered the survey of TEPS in senior high schools. In the following panel studies, phone interviews have implemented on both Panel-CP1 and Panel-NCP1 in different times without providing the response of overeducation or skill mismatch. Until the recent studies in 2014, two panels are under the same interviews at the same age and offer valid information for the research. Only the employed observations with tertiary educational diplomas and without a master or Ph.D. degree will be the target. The on-study students will be excluded from this research. The number of valid observations is 2,580 in total.. 4. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(11) In the panel survey of TEPS-B, the measure of overeducation, the adequate education and undereducation of jobs will all depend on the subjective self-assessment of workers themselves (by the question of “which educational level do you consider is required for the job you’re taking”). The answer will be junior high school (9 schooling years), senior high school (12 schooling years), junior college (14 schooling years), bachelor degree (16 schooling years), master degree (18 schooling years) or doctoral degree (22 schooling years). Then compared to the educational level of the bachelor degree the observations, if the schooling year of responding educational level is less than 16 years, the participants will be classified as overeducated, vice versa as undereducated, and adequately-educated when it is equal to the university level. And skill mismatch will adopt from the valuation of “you can utilize what you have learned on the current/last job.” The scale from 1 to 5 represents the level of match from the “fairly strong align (scale 1),” “align (scale 2),” “indifferent (scale. 政治大. 3),” “not align (scale 4)” and “completely not align (scale 5)”. The variable of mismatch. 立. adopts the scale 4 and 5, which represents the required job skills are associated and strongly. ‧ 國. 學. associated with what the individuals have learned on campus respectively. The two dependent variables are wage and satisfaction. Wage is based on the monthly wage. ‧. of current jobs and will be transformed into the logarithm. Satisfaction toward the current job is sourced from the survey of “are you satisfied with the job,” which response could be. Nat. sit. y. “fairly satisfied,” “satisfied,” “indifferent,” “not satisfied” and “totally not satisfied.” Only. er. io. the answers of “fairly satisfied” and “satisfied” are considered as the criteria of satisfaction (equal to 1 of the binary variable); otherwise, the rest of them is identified as 0 of the. n. al. Ch. i n U. v. satisfaction. The following list demonstrates all the covariates utilized in this study. Table 1: Covariates Variable Wage Job Satisfaction Skill Mismatch. engchi. Description monthly salary (NTD) of individuals, converted to logarithm job satisfaction of individuals. 1 as satisfied, 0 as not. skill mismatch of individuals. 1 as skill matched, 0 as not.. Mean. Standard Deviation. 10.3072. 0.3675. 7.6009. 12.6115. 0.5574. 0.4968. 0. 1. 0.3220. 0.4673. 0. 1. 0.5113. 0.5000. 0. 1. Min. Max. Overeducation (required education Overeducation. level of a job is less than the completed education level). 5. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(12) Gender. Description Gender of individuals. 1 as male, 0 as female.. Experience. Total working years as experience. Square of. the square of working experience is. experience. for outweighing. Mean. Standard Deviation. 0.4792. 0.4997. 0. 1. 1.9957. 1.1439. 0. 8.25. 0.2487. 0.6932. -0.4167. 6.75. 0.8758. 0.3299. 0. 1. 0.9309. 0.2536. 0. 1. 0.6466. 0.4781. 0. 1. -0.0282. 1.0136. -3.6081. 3.2695. 0.1479. 0. 1. ‧. Variable. 0. 1. 0. 1. 0.4499. 0. 1. 0.4283. 0. 1. 0.1660. 0.3722. 0. 1. 0.0058. 0.0762. 0. 1. 0.0739. 0.2617. 0. 1. 0.0730. 0.2601. 0. 1. 0.3654. 0.4816. 0. 1. Min. Max. the individual has critical thinking Critical Thinking. ability. 1 as yes, 0 as no the individual has problem fixing ability. 1 as yes, 0 as no. Supervisor. Coordinator Sale Personnel Primary Producers Specially Skilled Manual Personnel and Unskilled. Job type dummy. Supervisors and managers. Job type dummy. Professionals. Job type dummy.. al. Professional assistants. Job type dummy. Coordinators.. Ch. 0.0039. 0.0623. 0.1284. 0.3346. 0.2818. n. Assistant. 0.0224. Military-related jobs.. io. Professional. Job type dummy.. Nat. Professional. in junior high school in 3p score. U e n g c h i0.2419. Job type dummy. Service and sale personnel. Job type dummy. Producers of primary industries. Job type dummy. Specially skilled personnel.. y. Military. 立. the score of innate ability test taken. 學. ability. 1 as yes, 0 as no. sit. Junior high innate. 政治大. the individual has leadership.. ‧ 國. Leadership. er. Fix Problem. v ni. Job type dummy. Manual personnel and unskilled labor (reference group) Working location dummy.. Northern. Northern Taiwan: Taipei, New Taipei, Keelung. 6. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(13) Variable. Description. Mean. Standard Deviation. 0.1226. 0.3280. 0. 1. 0.2529. 0.4348. 0. 1. 0.0937. 0.2915. 0. 1. 0.1184. 0.3231. 0. 1. 0.0292. 0.1683. 0. 1. 0.0023. 0.0476. 0. 1. 0.0227. 0.0476. 0. 1. 0.1524. 0.3595. 0. 1. 0.4556. 0. 1. 0.2964. 0. 1. 0.2020. 0.4016. 0. 1. 0.0233. 0.1510. 0. 1. 0.0895. 0.2855. 0. 1. 0.8366. 0.3698. 0. 1. 0.0366. 0.1879. 0. 1. Min. Max. Working location dummy. Northern Central. Northern Central Taiwan: Taoyuan, Hsinchu, Miaoli Working location dummy.. Central. Central Taiwan: Taichung, Changhua, Nantou Working location dummy.. Southern Central. Southern central Taiwan: Yunlin, Chiayi, Tainan. Southern. Working location dummy. Southern Taiwan: Kaosiung, Pintong working location dummy.. Eastern. 政治大. Eastern Taiwan: Ilan, Hualien, Taitung (reference group). 立. working location dummy.. ‧ 國. Island area: Penghu, Kinmen,. 學. Island Area. Lianjiang. Discipline of studies dummy. Major in Social Science.. n. al. 0.2938. Science Engineering. Discipline of studies dummy.. Ch. Major in Science-related.. Major in Engineering-related.. y. i n U. 0.0973. engchi. Discipline of studies dummy.. sit. Major in Humanity and Art.. er. Discipline of studies dummy.. io. Social Science. Major in Education.. Nat. Humanity and Art. Discipline of studies dummy.. ‧. Education. v. Discipline of studies dummy. Farming. Major in Farming-related. (reference group). Medication. Discipline of studies dummy. Major in Medical studies. Status of employment dummy.. Status of employment. 1 as being full-time employed workers with monthly wage greater than NTD$ 20,000, 0 otherwise.. Marital status. Marital status of individuals. 1 as married, 0 as single.. 7. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(14) Variable Live with parent(s). Description. Mean. Standard Deviation. 2.3557. 1.7835. 0. 1. 0.0600. 0.2375. 0. 1. 0.5415. 0.4984. 0. 1. 0.3051. 0.4605. 0. 1. 0.3079. 0. 1. 0. 1. 0. 1. 0.0180. 0. 1. 0.0596. 0. 1. Min. Max. The individual lives with parents or not. 1 as live, 0 as do not The individual lives with mates or. Live with spouse. not. 1 as live, 0 as do not The individual is locals of the job. Local. location. 1 as locals, 0 as not. Job-search by informal channel Seek over 3 months. The job opportunity is via family, relatives or teacher introduction. 1 as introduction, 0 as not. 政治大. whether the individual sought The current/last job for over 3 months.. 立. 0.1060. 1 as over 3 months, 0 as not. ‧ 國. Rely on self. income or savings.. 0.9604. 學. The economic source is from own. 0.1950. 1 as own, 0 as not. ‧. The economic source is from family Rely on family. or relative supports.. 0.2280. 0.4196. Nat. benefit. Financial Market. pay.. al. sit 0.0003. Ch. U i e h n c g The economic source is via financial 1 as unemployed payment, 0 as not market.. 0.0036. er. unemployed pension or severance. n. unemployment. The economic source is from. io. Living by. y. 1 as supported, 0 as not. v ni. 1 as financial market, 0 as not. In the beginning, regressions run for the overall highly-educated observations and compare three methods. Then the most proper method will be used for the second part of the research. The later part decomposes the database into six categories based on the genders, the ranking, the three main professions, the two different institutions, the two different ownership and the two industries. The specification facilitates to figure out the different influence of overeducation and skill mismatch among groups.. 8. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(15) Among the six groups, the genders are straightforwardly defined as male and female of the entire highly-educated graduates. The ranking takes premise on the research of Shen and Lin (2018) and the common stratification in Taiwan 1. The three main professions are humanityrelated, general types and medically-related schools based on diverse professions of universities (Global Views Research, 2018; Ministry of Education, 2018) 2. The two different institutions are typical universities and science and technology universities. Typical universities are academics-oriented and mostly recruits senior high school graduates. And science and technology universities are vocation-oriented and usually recruits vocational high school graduates. The specification of different ownership divides universities into either public or private ones. And the last category set highly-educated graduates apart by differentiating their employed industries, the secondary and third industries.. 政治大 In this study, there are 2,580 samples 立 of highly-educated graduates. 49.3 percent of them Descriptive statistics. ‧ 國. 學. claim overeducation and 30.4 percent are skill-mismatched. The average wage of the entire employed graduates is NTD 33,119.86 per month. 66.63 percent of them satisfy with their current jobs.. y. 30 20. sit. er. al. 35.5. 42.41. n. 40. 57.03. io. 50. Nat. 60. ‧. Figure 1: Ratio of overeducation and skill mismatch of genders. Ch. engchi. i n U. v. 25.75. 10 0. HE male Overeducation. HE female Skill Mismatch. Top 10 universities are NTU(台大), NCCU(政大), NTHU(清大), NCTU(交大), NCKU(成大), NSYSU(中山), NCU(中央), NTNU(台師大), NTUST(台科大), NYMU(陽明), and the remaining is in the categories of non-top universities. 2 The primary professional realm of each university is determined by the related departments taking over 60% of the university. The three main categories are humanity-related, generality and medically-related in Taiwan. Humanity-related: Humanity, language, social science, commerce, law, education, art and so on. Generality: the universities can’t be classified into humanity- and medically- related are generality. 1. 9. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(16) There are 1,217 male samples. Almost 60 percent of them feel overeducated and 35.5 percent feel skill mismatched. In comparison, among 1,363 employed females, 42.4 percent of them claim overeducation and 25.8 percent of them feel skill mismatched in their jobs. The two rates are relatively lower among women than men. The average monthly wage of males is NTD 34,597.5 and 64.5 percent claim satisfaction of their jobs. The female labors have a monthly wage of NTD 31,800.5 on average and 68.5 percent of them feel satisfied with the jobs. The average earning of males is higher but the job satisfaction is less than the female counterparts.. Figure 2: Ratio of overeducation and skill mismatch of rankings 60 50. 立. 40. y. Nat. Non top. io. Overeducation. al. Skill Mismatch. sit. Top 10. er. 0. 18.55. ‧. 10. ‧ 國. 20. 30.94. 27.42. 學. 30. 政治 50.41 大. n. v i n The top-10 graduates have less thanC 5 percent in the tertiary graduates (the h e n population i U h c g number of observation is 124). There is 27.4 percent of them feel overeducated and 18.6 percent state their skills mismatch in the current jobs. And half of the non-top university graduates have higher rates of overeducation and skill mismatch. Top-10 labors have NTD 36,449.6 average wage and 67.7 percent feel satisfied with their jobs. Non-top graduates earn NTD 32,951.8 monthly wage on average and have 66.6 percent of satisfaction. The average wage of the top-10 graduates is higher, while the job satisfaction is quite closed.. 10. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(17) Figure 3: Ratio of overeducation and skill mismatch of three professions 60 50 40 30. 50.85. 50.11. 39.63 31.04. 30.97. 21.95. 20 10 0. Humanity. Generality Overeducation Skill Mismatch. Medical. Among the three main professions, half of humanity-related and general-type graduates argue. 政治大 Medically-related graduates have less overeducation (39.6 percent) and skill mismatch (22.0 立 percent). The average monthly wage is NTD 32,883.8 (humanity-related), NTD 32,774.1. they are overeducated. One in three of them also claims skill mismatch in their jobs.. ‧ 國. 學. (general-type), and NTD 33,597.8 (medically-related). Wages, then, are not significantly different. Satisfaction toward current jobs is almost equal (66.8, 66.6, and 67.7 percent).. ‧. 30. 39.98. al. n. 40. 30.17. sit. io. 50. 57.7. er. 60. Nat. 70. y. Figure 4: Ratio of overeducation and skill mismatch of two institutions. Ch. engchi. i n U. v. 30.52. 20 10 0. University Overeducation. Science & Technology Skill Mismatch. The two institutions have a disparity of the percentage of overeducation. Typical universities have approximately 40 percent and science and technology schools have a higher amount, almost 60 percent. One in three of the graduates from typical universities claim they are skill mismatched, so is the group of science and technology graduates. The graduates of typical universities have a higher wage of NTD 35,136.7 than of science and technology. 11. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(18) counterparts, which has NTD 31,333.2 on average. 65.2 percent of typical university graduates and 67.9 percent of science and technology ones feel satisfied with the jobs. Figure 5: Ratio of overeducation and skill mismatch of different ownership 60 50 40. 52.85 39.88 31.63. 27.08. 30 20 10. 政治大 Private Overeducation Skill Mismatch 立. Public. 學. ‧ 國. 0. The public support university graduates have a 40 percent claim of overeducation and 27 percent of them feel skill mismatched. For private school graduates, both indicators have. ‧. higher ratios than public graduates. Over half of them feel overeducated and a slightly larger. y. Nat. amount of skill mismatch (31.6 percent). The average of public school graduates has NTD. n. al. er. io. evenly in both categories, 67.3 and 66.4 percent individually.. sit. 34,682.7 and of private ones has NTD 32,555.2 per month. The job satisfaction is distributed. i n U. v. Figure 6: Ratio of overeducation and skill mismatch of two industries 60. 53.97. Ch. engchi. 47.25. 50 40. 36.44 26.8. 30 20 10 0. Secondary Industry Overeducation. Third Industry Skill Mismatch. 12. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(19) Secondary industry has NTD 31,359.04 on average and 61.78% of satisfaction. Over half of secondary industry workers have claimed overeducation and more than one third feel skillmismatched. Third industry has a monthly salary, NTD 33,043.25 on average and higher job satisfaction of 68.97%. Both overeducation and skill mismatch of third industry workers are lower than their secondary counterparts.. 4. Methodology Wage effect There are three methods – ordinary least square (OLS), Heckman two-step method, and inverse probability weighting (IPW) for estimating wage effect in the study. Each of them regresses on the two issues individually then together. Ordinary least square. 立. 政治大. First, the ordinary least square (OLS) model is implemented to do the simple and. ‧. ‧ 國. follows,. 學. straightforward regression for the entire highly-educated observations. The OLS model is as. (1) ln 𝑊𝑊𝑖𝑖 = 𝛽𝛽0 + 𝛽𝛽1 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑖𝑖 + 𝛽𝛽2 𝑆𝑆𝑆𝑆_𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖 + 𝛽𝛽3 𝑍𝑍𝑖𝑖 + 𝛽𝛽4 𝐸𝐸𝑖𝑖 + 𝛽𝛽5 𝑃𝑃𝑖𝑖 + 𝑒𝑒𝑖𝑖. sit. y. Nat. Where 𝑊𝑊𝑖𝑖 is the individual i’s monthly wage. 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑖𝑖 is the phenomenon of. er. io. overeducation. 𝛽𝛽1 is the target coefficient and the return of overeducation on wage.. 𝑆𝑆𝑆𝑆_𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖 is the skill mismatch binary variable. 𝛽𝛽2 is the return of skill mismatch. 𝑍𝑍𝑖𝑖. n. al. i n U. v. represents the primary control variables (gender, marital status, working experience, and. Ch. engchi. square of working experience), which generally conforms to Mincer earnings function. 𝐸𝐸𝑖𝑖 includes personal characteristics (𝐸𝐸1𝑖𝑖 : critical thinking, problem-solving ability and. leadership; 𝐸𝐸2𝑖𝑖 : innate ability in junior high). 𝑃𝑃𝑖𝑖 is the vector of binary variables (types of. job, locations of organization, and graduated professions). Heckman two-step method. Second, to avoid the possible selection bias of individuals’ decision of employment and take the consideration of those unemployed, two-step Heckman estimates are adopted. In the first step of Heckman, the binary dependent variable is the current status of employment with the additional covariates in the vector 𝐶𝐶𝑖𝑖 (marriage, live with parent, live with mate, local to. workplace, job introduced by relatives or family, length of job-seeking over 3 months, self-. 13. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(20) financial support, financial support from relatives or family, financial market support and unemployment benefit) and regresses by probit model as below, (2) 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖 = 𝛾𝛾0 + 𝛾𝛾1 𝑍𝑍𝑖𝑖 + 𝛾𝛾2 𝐶𝐶𝑖𝑖 + 𝜀𝜀𝑖𝑖. (3) ln 𝑊𝑊𝑖𝑖 = 𝛿𝛿0 + 𝛿𝛿1 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑖𝑖 + 𝛿𝛿2 𝑆𝑆𝑆𝑆_𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖 + 𝛿𝛿3 𝑍𝑍𝑖𝑖 + 𝛿𝛿4 𝐸𝐸𝑖𝑖 + 𝛿𝛿5 𝑃𝑃𝑖𝑖 + 𝛿𝛿6 𝜆𝜆 + 𝜇𝜇𝑖𝑖. And the second step of Heckman (3) adopts the same covariates of OLS (1) above. 𝜆𝜆 is the. inverse Mill ratio and computed by the mean of the entire first set. This method examines the potential selection bias of the original database and controls to achieve a more precise. statistical analysis. If the selection bias significantly alters the consequence of OLS estimates, the inverse Mill ratio 𝜆𝜆 will be critical and have to be included in the models to adjust the coefficient. To differentiate the effect, the observations are also the all highly-educated graduates in the Heckman estimate. Inverse probability-weighted. 立. 政治大. The last method of analyzing wage effect is inverse probability-weighted (IPW). IPW works. ‧ 國. 學. for controlling the self-selection bias and the heterogeneity of observations. The effect of the average treatment effect on the treated (ATT) are implemented in this study. Attached to the. ‧. topic, overeducation and skill mismatch are two treatments. Two will be included at the same. y. Nat. equation but only one effect can be estimated from IPW. Therefore, for each of the case,. sit. there are two individual consequence for overeducation and skill mismatch. The outcome. er. io. variable is the logarithm of wage. The first part of this method is allocating a score, so-called. al. propensity score (PS), as presented in equation (4), which represents the probability of. n. v i n C h based on theUcontrol variables. The weights are receiving the treatment, for each observation e n g1c h i 1 different between the treated and the untreated. for the treated and for the untreated. 𝑃𝑃𝑃𝑃. 1−𝑃𝑃𝑃𝑃. The control vectors are the 𝑍𝑍𝑖𝑖 and 𝐸𝐸𝑖𝑖 from OLS and Heckman models and should be. independent on outcome and treatment covariates. The observed control variables are used for controlling heterogeneity in a non-parametric way. (4) 𝑃𝑃(𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 ) = 𝑃𝑃𝑃𝑃(𝐷𝐷𝑖𝑖 = 1| 𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 ) = 𝐸𝐸(𝐷𝐷𝑖𝑖 |𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 ). After allocating a score for each observation from probit estimates, the weights makes the result more precise, and the participation of treatments won’t only fall on either 0 or 1 but in a more compromise way of targeting the effect of treatments. 𝑃𝑃𝑃𝑃𝑃𝑃 (5) τ𝐴𝐴𝐴𝐴𝐴𝐴 = 𝐸𝐸𝑃𝑃(𝑍𝑍𝑖𝑖 ,𝐸𝐸𝑖𝑖 )|𝐷𝐷𝑖𝑖 =1 (𝜏𝜏|𝐷𝐷𝑖𝑖 = 1) = 𝐸𝐸[ln 𝑤𝑤(1)|𝐷𝐷𝑖𝑖 = 1, 𝑃𝑃(𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 )] − 𝐸𝐸[ln 𝑤𝑤(0)|𝐷𝐷𝑖𝑖 =. 1, 𝑃𝑃(𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 )]. 14. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(21) 𝑃𝑃𝑃𝑃𝑃𝑃 In the equation (5), τ𝐴𝐴𝐴𝐴𝐴𝐴 is the average treatment of treated with propensity score matching.. It takes the counterfactual assumption 𝐸𝐸[ln 𝑤𝑤(0)|𝐷𝐷𝑖𝑖 = 1, 𝑃𝑃(𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 )] into account to see if there is self-selection bias from choosing treatment and control groups. If: (6) 𝐸𝐸[ln 𝑤𝑤(0)|𝐷𝐷𝑖𝑖 = 1, 𝑃𝑃(𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 )] − 𝐸𝐸[ln 𝑤𝑤(0)|𝐷𝐷𝑖𝑖 = 0, 𝑃𝑃(𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 )] = 0. It proves the sample selection is random and means there is no treatment group participator 𝑃𝑃𝑃𝑃𝑃𝑃 having the wage of control group. And the average treatment of the treated τ𝐴𝐴𝐴𝐴𝐴𝐴 will be no. difference from the original OLS model with heterogeneity control. If not, the consequence will be adjusted.. Addressing unobserved characteristics is of importance since distinguishing personal features from educational institutions can influence the research. This will make it more likely that we. 治政 can be positive that the unobserved personal characteristics大 are the underlying issue. Therefore, IPW is necessary to立 the study as well.. will know that the effects on an individual’s wages are from the school per se. Otherwise, we. ‧ 國. 學. Satisfaction. ‧. Two approaches are about to implement for job satisfaction. Probit model and inverse probability weighting (IPW) will estimate for the effect of overeducation and skill mismatch. y. sit. n. al. er. io. Satisfaction model. Nat. on job satisfaction.. v. At the section of analyzing the effect of overeducation and skill mismatch on satisfaction,. i n C probit estimate will be adopted and runhthe i U efollowing n g c hmodel:. (7) Satisfaction = 𝛾𝛾0 + 𝛾𝛾1 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑖𝑖 + 𝛾𝛾2 𝑆𝑆𝑆𝑆_𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖 + 𝛾𝛾3 ln 𝑊𝑊𝑖𝑖 + 𝛾𝛾4 𝑍𝑍𝑖𝑖 + 𝛾𝛾5 𝐸𝐸𝑖𝑖 + 𝛾𝛾6 𝑃𝑃𝑖𝑖 + 𝜀𝜀𝑖𝑖. The most different part of the model from the wage effect equations is adding the logarithm wage into it and the dependent variable alters to the binary covariate of job satisfaction, which is considered to have influence on the job satisfaction. The rest of equation hold constant as the wage equations above. Inverse probability-weighted (8) 𝑃𝑃(𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 ) = 𝑃𝑃𝑃𝑃(𝐷𝐷𝑖𝑖 = 1| 𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 ) = 𝐸𝐸(𝐷𝐷𝑖𝑖 |𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 ). 𝑃𝑃𝑃𝑃𝑃𝑃 (9) ω𝐴𝐴𝐴𝐴𝐴𝐴 = 𝐸𝐸𝑃𝑃(𝑍𝑍𝑖𝑖 ,𝐸𝐸𝑖𝑖 )|𝐷𝐷𝑖𝑖 =1 (𝜔𝜔|𝐷𝐷𝑖𝑖 = 1) = 𝐸𝐸[Satisfaction(1)|𝐷𝐷𝑖𝑖 = 1, 𝑃𝑃(𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 )] −. 𝐸𝐸[Satisfaction(0)|𝐷𝐷𝑖𝑖 = 1, 𝑃𝑃(𝑍𝑍𝑖𝑖 , 𝐸𝐸𝑖𝑖 )]. 15. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(22) The inverse probability-weighted estimate works similarly as the previous estimate to do the control of unobserved characteristics based on the observed controls (𝑍𝑍𝑖𝑖 and 𝐸𝐸𝑖𝑖 ). The same. score allocating then continues to compute the average treatment of the treated on job satisfaction in equation (9).. 5. Result All highly-educated graduates: Wage effect Table 2: OLS wage equations Column 1-2. -0.1021*** (0.01). Column 3-1. Column 3-2. Column 4-1. Column 4-2. -0.1093***. -0.0974***. -0.1028***. -0.0615***. -0.0649***. (0.01). (0.01). (0.01). (0.01). (0.01). Skill Mismatch. -0.0204. -0.0275*. -0.0086. -0.0149. (0.01). (0.01). (0.01). (0.01). (0.01). (0.01). 0.0961***. 0.0823***. 0.0852***. 0.0945***. 0.0981***. 0.0792***. 0.0822***. (0.01). (0.01). (0.01). (0.01). (0.01). (0.01). (0.01). (0.01). 0.0423. 0.0492. 0.0336. 0.0406. 0.0439. 0.0511. 0.0328. 0.0384. (0.03). (0.03). (0.03). (0.03). (0.03). (0.03). (0.03). (0.03). 0.0238***. 0.0257***. 0.0222***. 0.0244***. 0.0242***. 0.0262***. 0.0210***. 0.02236***. (0.01). (0.01). (0.01). (0.01). (0.01). (0.01). (0.01). (0.01). -0.006. -0.0064. -0.0040. -0.0042. -0.0060. -0.0064. -0.0057. -0.0060. (0.01). (0.01). (0.01). (0.00). (0.01). (0.01). (0.01). io. (0.01). al. n. Critical Thinking. 0.0321** (0.01). Fix Problem. Leadership. Junior high innate ability. Ch. y. sit. Nat. Square of Experience. -0.0561***. ‧. Experience. 政治大. -0.0465***. 學. Marital Status. 立. Column 2-2. 0.0930***. ‧ 國. Gender. Column 2-1. er. Overeducation. Column 1-1. i n U. v. 0.0287. 0.0312*. (0.02). (0.01). (0.01). engchi. 0.0333**. 0.0519**. 0.0608***. 0.0507**. 0.0328. (0.02). (0.02). (0.02). (0.02). 0.0511***. 0.0572***. 0.0498***. 0.0437***. (0.01). (0.01). (0.01). (0.01). -0.0037. -0.0031. -0.0036. -0.0043. (0.01). (0.01). (0.01). (0.00). Job search by. 0.0006. informal channel. (0.01) Military. 0.2524***. 0.2666***. (0.03). (0.03). 16. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(23) Column 1-1. Column 1-2. Column 2-1. Column 2-2. Column 3-1. Column 3-2. Supervisor. Professional. Professional Assistant. Coordinator. Sale Personnel. Primary Producers. 立. ‧ y. sit. n. al. er. io. Central. Nat. Northern Central. ‧ 國. Northern. 學. Specially Skilled. 政治大. Southern Central. Southern. Island Area. Education. Humanity and Art. Social Science. Ch. engchi U. v ni. Column 4-1. Column 4-2. 0.5850**. 0.6008*. (0.24). (0.25). 0.1369***. 0.1435***. (0.03). (0.03). 0.0753***. 0.0820**. (0.03). (0.03). -0.0565*. -0.0559*. (0.03). (0.03). 0.0183. 0.0228. (0.03). (0.03). 0.0288. 0.0374. (0.09). (0.08). 0.0175. 0.0166. (0.03). (0.03). -0.0349. -0.0353. (0.04). (0.04). 0.0247. 0.0258. (0.04). (0.04). -0.1175***. -0.1139***. (0.04). (0.04). -0.1289***. -0.1318***. (0.04). (0.04). -0.0422. -0.1390***. (0.02). (0.04). 0.1056*. 0.1140. (0.19). (0.18). 0.0689. 0.0665. (0.04). (0.04). -0.0072. -0.0073. (0.02). (0.02). 0.0091. 0.0110. (0.02). (0.02). 17. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(24) Column 1-1. Column 1-2. Column 2-1. Column 2-2. Column 3-1. Column 3-2. Science. Engineering. Medicine. constant. observation. Column 4-1. Column 4-2. -0.0316. -0.0310. (0.02). (0.02). -0.0211. -0.0228. (0.02). (0.02). 0.0459*. 0.0426. (0.05). (0.02). 10.20***. 10.31***. 10.17***. 10.28***. 10.21***. 10.32***. 10.25***. 10.34***. (0.02). (0.02). (0.02). (0.02). (0.02). (0.02). (0.05). (0.05). 2580. 2563. 2580. 2563. 2580. 2563. 2580. 2563. Table 2 reports the OLS estimates. Column 1-1 is the specification of including only. 政治大 column 1-2 has a similar explanatory 立 variables but takes 𝐸𝐸 as the fixed effect instead.. overeducation, the primary control covariates and the variables of 𝐸𝐸𝑖𝑖1 personal characteristics; 𝑖𝑖2. ‧ 國. 學. Column 2-1 is the specification of including only skill mismatch and the rest of observatory variables are the same as column 1-1. Column 2-2 exchanges the personal characteristics of. ‧. column 2-1 from 𝐸𝐸𝑖𝑖1 to 𝐸𝐸𝑖𝑖2 . Column 3-1 and column 3-2 remain the same set of main control variables and contains both overeducation and skill mismatch with 𝐸𝐸𝑖𝑖1 and 𝐸𝐸𝑖𝑖2 fixed effect. y. Nat. control respectively. The last one, column 4-1 and column 4-2 include all covariates in. sit. column 3-1 and 3-2 as well as extra 20 binary variables (occupational types, locations and. al. er. io. graduated majors). The control of personal characteristics in column 4-2 is added the extra. n. v i n C h Skill mismatch percent wage penalty in the separate estimate. e n g c h i U has a negative 4.7 percent on. variable of job search by informal channel. According to column 1-1, overeducation has 10 wage in column 2-1. While including both together, overeducation has decreased its effect. insignificantly about 0.3 percent, and skill mismatch declines almost 3 percent to 2 percent and becomes not significantly associated with wage. As we take more dummies into control, both overeducation and skill mismatch even weakens to 6 percent and 1 percent respectively. The first section of table 2 (column 1-1, 2-1, 3-1 and 4-1) takes 𝐸𝐸𝑖𝑖1 , the vector of critical. thinking ability, problem solving and leadership, as the personal characteristics control, and the second part takes 𝐸𝐸𝑖𝑖2 , the innate ability score in junior high school. Though there is no. significantly different from two sections, the regressions inclusive of 𝐸𝐸𝑖𝑖2 have a more. prominent influence than those including 𝐸𝐸𝑖𝑖1 . This explains the innate ability may also. contain critical thinking ability, problem solving, leadership and even other unobserved. 18. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(25) characteristics we can’t think of. And those we can’t figure out effect the overeducation and skill mismatch. Therefore, the rest of research will take 𝐸𝐸𝑖𝑖2 as the control of fixed effect. Table 3: Wage equations of Heckman two-step method Column 1. Column 3. Column 4. -0.1029***. -0.0965***. -0.0608***. (0.02). (0.02). (0.02). -0.0537***. -0.0271. -0.0147. (0.03). (0.03). (0.02). 0.0890***. 0.0784***. 0.0912***. 0.0800***. (0.03). (0.03). (0.03). (0.02). 0.0079. 治政 -0.0030 0.0101 大. Skill Mismatch. (0.07). 0.0008. -0.0003. 0.0023. (0.04). (0.04). (0.03). (0.03). -0.0000. -0.0001. -0.0001. -0.0001. (0.01). (0.01). (0.01). (0.00). -0.0035. -0.0028. -0.0034. (0.01). (0.01). Nat. io. Junior high innate ability. al. n. (0.01) Job search by informal channel. Ch. engchi U. (0.05) 0.0113. ‧. Square of Experience. (0.08). 學. Experience. 立. ‧ 國. (0.07). 0.0104. y. Marital status. sit. Gender. -0.0042. er. Overeducation. Column 2. v ni. (0.01). 0.0026 (0.02). Military. 0.255*** (0.07). Supervisor. 0.606*** (0.14). Profession. 0.144*** (0.05). Professional Assistant. 0.081** 19. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(26) Column 1. Column 2. Column 3. Column 4 (0.38). Coordinator. -0.0542 (0.04). Sale Personnel. 0.0208 (0.41). Primary Producers. 0.0357 (0.11). Specially Skilled. 0.0157. Northern. ‧ 國. ‧. (0.05). -0.0970. al. er. sit. y. (0.05). n. Southern. 0.0277. io. Southern Central. (0.05). Nat. Central. 0.0254. 學. Northern Central. 立. (0.04). 政治大. Ch. n U engchi. iv. -0.1201** (0.06) -0.1303** (0.05). Island Area. 0.117 (0.17). Education. 0.0652 (0.06). Humanity and Art. -0.0079 (0.03). Social Science. 0.0122. 20. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(27) Column 1. Column 2. Column 3. Column 4 (0.03). Science. -0.0316 (0.04). Engineering. -0.0247 (0.03). Medicine. 0.0356 (0.04) 10.45***. 10.43***. (0.08). (0.09). (0.09). 0.1052 (0.08). (0.08). (0.08). (0.08). 0.4238***. 0.4238***. 0.4238***. (0.09). (0.09). (0.09). (0.09). -0.0508***. -0.0508***. -0.0508***. -0.0508***. (0.02). (0.02). (0.02). (0.02). 0.3774. 0.3774. n. (0.29) Live with parent(s). Live with spouse. Local. al. -0.0654***. (0.29) C (0.29) hengchi U. 0.4238***. y. sit. ‧ 國. 0.3774. 0.1052. ‧. io. Marital status. Nat. Square of Experience. 立. 學. Experience. 10.41***. (0.08) 治政大 0.1052 0.1052. Status of employment Gender. 10.45***. er. constant. v ni. 0.3774 (0.29). -0.0654***. -0.0654***. -0.0654***. (0.03). (0.03). (0.03). (0.03). -0.0100. -0.0100. -0.0100. -0.0100. (0.18). (0.18). (0.18). (0.18). 0.2346***. 0.2346***. 0.2346***. 0.2346***. (0.08). (0.08). (0.08). (0.08). -0.0170. -0.0170. -0.0170. -0.0170. (0.09). (0.09). (0.09). (0.09). Seek search by informal channel. 21. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(28) Seek over 3 months. Rely on self. Rely on family. Column 1. Column 2. Column 3. Column 4. -13.09. -13.09. -13.09. -13.09. (.). (.). (.). (.). 1.0721***. 1.0721***. 1.0721***. 1.0721***. (0.15). (0.15). (0.15). (0.15). -0.4840***. -0.4840***. -0.4840***. -0.4840***. (0.09). (0.09). (0.09). (0.09). 4.108. 4.108. 4.108. 4.108. (.). (.). (.). (.). 5.278. 5.278. Living by unemployment benefit. Financial Market. 立. (.). -0.713***. -0.663***. (0.22). (0.23). (0.21). 3067. 3067. 3067. (.) indicates the standard deviation is considerable.. (.) -0.478*** (0.16). 3067. Nat. y. ‧. observation. -0.666***. 5.278. 學. lambda. ‧ 國. mills. 政治5.278大 (.) (.). sit. Table 3 describes the regression of Heckman two-step. The observations are used more than. er. io. OLS because this model is required to include both the employed and the unemployed. al. v i n C penalty overeducation has 10 percent of wage segregated equation. Skill mismatch has h e n ingthe chi U n. samples so as to analyze whether sample selection bias affects the database. In column 1,. almost a negative 5 percent effect on wage. And when combining two into one equation (i.e. column 3), both weaken slightly and skill mismatch becomes insignificant to wage. After adding up dummies, overeducation drops approximately 3 percent to 6.1 percent and skill. mismatch has less than 1 percent influence on wage. The volatility of the Heckman method is consistent and similar to the OLS model. Compared the column 4 of table 1 and table 2, the adjustment of the wage penalty of two predictors is not quite significant. It implies the influence of sample selection bias is slight among the employed and unemployed observations. Table 4: IPW model of wage wage. Overeducation. Skill Mismatch. -0.0998***. -0.0257** 22. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(29) observation. (0.01). (0.01). 2563. 2563. Table 4 indicates the consequence from IPW. Because of the limitation of designing IPW models, only primary control covariates and personal characteristics (i.e. 𝐸𝐸𝑖𝑖2 ) can be. included. Therefore, we do the comparison with the OLS estimates from column 3-2 of table 2 and the Heckman methods from column 3 of table 3. IPW has similar results with the two other models, but skill mismatch here is more significant than the others. All highly-educated graduates: Job Satisfaction Table 5: Probit model of job satisfaction Column 1 -0.1930**. 立. 0.3280***. 0.2962***. -0.3272***. (0.06). (0.06). (0.06). 0.4838***. 0.5037***. 0.465***. 0.576***. (0.10). (0.10). (0.10). (0.11). -0.1104*. -0.0924. -0.1177**. n. (0.06) Marital Status. Experience. Square of Experience. Junior high innate ability. al. (0.06) C(0.06) hengchi U. sit. y. ‧. -. io. Gender. (0.06). -. Nat. Wage. (0.06). 學. Skill Mismatch. ‧ 國. (0.05). er. Overeducation. Column 3 Column 4 政治 -0.1252** 大 -0.1956***. Column 2. -0.0407. v n i(0.07). 0.3735**. 0.3887**. 0.4033**. 0.4363***. (0.16). (0.16). (0.16). (0.16). -0.0784. -0.0616. -0.0643. 0.0140. (0.07). (0.07). (0.07). (0.03). 0.0178. 0.0155. 0.0165. 0.0233. (0.01). (0.01). (0.01). (0.04). 0.046. 0.048. 0.047. 0.050. (0.03). (0.03). (0.03). (0.03). 23. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(30) Column 1. Column 2. Column 3. Column 4. Job search by informal channel. 0.080 (0.06). Military. -0.5136** (0.20). Supervisor. 0.2226 (0.46). Professional. -0.2094 (0.14). Professional Assistant. ‧ 國. (0.12). 0.1647 (0.13). -0.2567 (0.33). n. al. er. io. sit. y. Nat. Primary Producers. -0.0229. ‧. Sales Personnel. (0.12). 學. Coordinator. 立. 政治大 -0.1667. Specially Skilled. Northern. Ch. engchi U. v ni. -0.1717 (0.14) -0.0536 (0.16). Northern Central. -0.0252 (0.17). Central. 0.0465 (0.16). Southern. 0.0202 (0.17). 24. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(31) Column 1. Column 2. Column 3. Southern Central. Column 4 -0.0589 (0.17). Island Area. 0 (.). Education. 0.2390 (0.21). Humanity and Art. -0.0832 (0.10). Social Science. 立. ‧ 國. -0.1244 (0.10). -0.2444** (0.12). er. io. sit. y. Nat. Medicine. (0.11). ‧. Engineering. -0.0773. 學. Science. 政治大 -0.0834 (0.09). al. observation. 2563. n. constant. v i n -4.367*** C-4.592*** h e n g c-4.145*** i U -5.197*** h (1.01) (1.00) (1.01) (1.12) 2563. 2563. 2563. Based on table 5, overeducation has a negative 19 percent on job satisfaction when computing separately. In column 2, skill mismatch has an almost 33 percent influence. While putting two predictors together, the effect of overeducation has decreased much more than skill mismatch. They are 13 percent and 30 percent individually in column 3. Then, both become worse on job satisfaction after including dummies. Table 6: IPW model of job satisfaction Overeducation. Skill Mismatch. 25. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(32) satisfaction observation. -0.0505. -0.1111***. (0.02). (0.02). 2563. 2563. In the estimate of IPW, overeducation has a negative 5.1 percent reduction on job satisfaction, while skill mismatch is more influential and have a percentage of 11.1 than overeducation. Compared to table 5, the probit model, two coefficients adjusted differently. Overeducation is more serious in the IPW than the probit, skill mismatch is weaker in table 6. Therefore, after controlling unobserved characteristics, overeducation has a stronger impact on job satisfaction, while skill mismatch changes in an opposite way. Six categories: Wage effect and job satisfaction. 政治大. From the above estimates, we are unable to determine the best choice. OLS and Heckman are. 立. likely to have the problem of overfitting when analyzing the small samples of the five. ‧ 國. 學. categories—gender, ranking, three professions, two institutions, two ownership and two industries—according to the one-to-ten ratio of control variable selection to sample size, the. ‧. so-called rule of thumb in statistics. In addition, the effect on satisfaction is significant but the effect on wages is less so. It is necessary to choose a similar control in order to maintain. y. Nat. consistency and arrive at a convincing interpretation. In addition, because taking unobserved. n. al. Male. Wage. Satisfaction. observation. er. io. Table 7: Genders. sit. characteristics into consideration is significant, IPW is chosen for use in further research.. Ch. v i n U Female. e n gOvereducation chi. Overeducation. Skill Mismatch. Skill Mismatch. -0.1002***. -0.0081. -0.1048***. -0.0493***. (0.02). (0.02). (0.01). (0.02). -0.0369. -0.1182***. -0.0680**. -0.1038***. (0.03). (0.03). (0.03). (0.03). 1208. 1208. 1355. 1355. For the male graduates, overeducation has a similar wage penalty with women; while the skill mismatch of men has a much lower effect than of women. Overeducation has a more. 26. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(33) considerable influence on women’s job satisfaction than men by 3 percent. Skill mismatch of both reduces job satisfaction in a closed level. Table 8: Rankings Top 10. Wage. Satisfaction observation. Non-top. Overeducation. Skill Mismatch. Overeducation. Skill Mismatch. -0.1795***. 0.0485. -0.0939***. -0.0264**. (0.05). (0.05). (0.01). (0.01). -0.0249. -0.1365. -0.0512**. -0.1120***. (0.11). (0.13). (0.02). (0.02). 122. 122. 2441. 2441. 政治大 In table 8, the wage of top-10 university graduates has been negatively influenced much more 立 from overeducation than skill mismatch. It reaches 18 percent wage penalty but a positive 5. ‧ 國. 學. percent from skill mismatch. Non-top graduates have less stronger wage penalty effects from both overeducation and skill mismatch. The satisfaction of top-10 graduates is affected. io. Overeducation. Satisfaction. observation. -0.0983*** (0.02). Skill Mismatch. a-0.0288 l C (0.02) h. n. Wage. Generality Overeducation -0.0785*** (0.02) eng chi. Skill Mismatch. er. Humanity. sit. y. Nat. Table 9: Professions. ‧. negatively by overeducation and skill mismatch.. -0.0110 iv n U (0.02). Medical Overeducation. Skill Mismatch. -0.1674***. -0.1208***. (0.04). (0.04). -0.0610**. -0.1138***. -0.0388. -0.1144***. -0.0873. -0.1337. (0.03). (0.03). (0.04). (0.03). (0.10). (0.10). 1300. 1300. 1054. 1054. 164. 164. From table 9, we see that the wage penalty from overeducation on general type university graduates is the weakest while it is the strongest on medically-related college graduates at 16.7 percent. Humanity-related graduates experience the weakest wage effect from skill mismatch, and medical school graduates bear the strongest effect. The effect of overeducation on general-type university graduates is the weakest influence on satisfaction. Both experience insignificantly different levels of effect than humanity-related school students. For medically-. 27. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(34) related graduates, overeducation has considerably more effect on satisfaction than does skill mismatch. Table 10: Institutions Typical University. Wage. Satisfaction observation. Science & Technology University. Overeducation. Skill Mismatch. Overeducation. Skill Mismatch. -0.1126***. -0.0561***. -0.0617***. -0.0052. (0.02). (0.02). (0.02). (0.02). -0.0217. -0.1013***. -0.0805***. -0.1121***. (0.03). (0.03). (0.03). (0.03). 1192. 1192. 1344. 1344. 政治大 The wage effect of typical universities has been influenced by overeducation and skill 立. mismatch more potently than of science and technology universities by the difference of 5. ‧ 國. 學. percent on each. Overeducation has caused a weaker job discontent of typical university graduates. Skill mismatch have alike effects on the job satisfaction of typical university. io. Wage. Satisfaction observation. -0.1283***. Skill Mismatch. (0.03). Overeducation. Skill Mismatch. a l -0.0445 -0.0802*** i v n Ch U (0.03) (0.02) engchi. n. Overeducation. Private University. er. Public University. sit. y. Nat. Table 11: Ownership. ‧. graduates.. -0.0168 (0.01). -0.0489. -0.1234***. -0.0638**. -0.1091***. (0.06). (0.05). (0.02). (0.03). 667. 667. 1882. 1882. According to table 11, overeducation and skill mismatch of the public schools have a larger influence on the wage penalty than of the private ones and are above the average. When it comes to satisfaction, the effect from skill mismatch on private universities is more prominent than public ones. The effect from overeducation toward job satisfaction is quite alike in both types of schools.. 28. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(35) Table 12: Industries Secondary Industry. Wage. Satisfaction. observation. Third Industry. Overeducation. Skill Mismatch. Overeducation. Skill Mismatch. -0.0815***. 0.0156. -0.0932***. -0.0457***. (0.02). (0.02). (0.02). (0.02). -0.0435. -0.1486***. -0.0494. -0.0802***. (0.04). (0.04). (0.03). (0.03). 746. 746. 1643. 1643. Overeducation has a similar effect on the wage of both secondary and third industries, while. 治政大have a similar influence from effect but a wage penalty on third industries. Third industries 立 overeducation on the job satisfaction but less from skill mismatch compared to secondary skill mismatch has a different story. Skill mismatch causes secondary workers a positive. ‧ 國. 學. industries.. ‧. 6. Discussion. y. Nat. All highly-educated graduates. sit. Overall, highly-educated Taiwanese graduates experience a 6 to 10 percent wage penalty. n. al. er. io. from overeducation and a 1 to 4 percent penalty from skill mismatch. U.K. graduates have a. i n U. v. 34 to 39 percent penalty from overeducation and 10 to 22 percent penalty from skill. Ch. engchi. mismatch (McGuinness and Sloane, 2011). In Italy, there is a 12 to 20 percent penalty from overeducation and a 7 to 16 percent penalty from skill mismatch (Caroleo and Pastore, 2014). The two issues for the whole highly-educated graduates are still mild in Taiwan. For the same group, Taiwanese graduates tend to have a 5 to 17 percent impact on job satisfaction from overeducation and an 11 to 30 percent effect from skill mismatch. Skill mismatch usually has a more potent effect than overeducation on job satisfaction. This overall condition is consistent with previous studies (McGuinness and Sloane, 2011; Caroleo and Pastore, 2014; Allen and Van der Velden, 2001). Nevertheless, the impact of both are still weaker in Taiwan than in other countries, though there are more claims of overeducation in Taiwan. Skill mismatch in Taiwan is similar or slightly above the level of other countries. This study also found that the wage gap between overeducated and non-overeducated graduates is not significantly different. The average wage of overeducated graduates is NTD 31,516.01 and of 29. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(36) non-overeducated graduates is NTD 34,679.57. Therefore, it is likely that there are other causes for wage differences. The effect of overeducation on wages is consistent with previous studies, which implies that overeducation has a larger effect on wages than skill mismatch. Assignment theory also supports this result. An increase in productivity leads to lower wages at equilibrium. The effects on job satisfaction are not as prominent as expected. In addition to considering other aspects affecting job satisfaction, job competition theory and job mobility theory may provide explanations for this. Therefore, it is possible that most highly-educated graduates are anxious to be competitive and look forward to promising careers. Accordingly, overeducation will only be temporary. Genders. 政治大. In this study, the females claim overeducation for 42 percent and skill mismatch for 27. 立. percent of them, while the males have higher proportions of each. Based on the research by. ‧ 國. 學. Robst (2007), it is possible that men are more willing to admit overeducation or skill mismatch than women. In my opinion, perhaps males may be inclined to overestimate their. ‧. own ability and thereby claim overeducation or skill mismatch easily. Either one argument can support the idea that, if a woman feels overeducated or skill mismatched, she must be. sit. y. Nat. very unfulfilled. On the other hand, the result of female wage effect is more considerable than of male. It may result from the higher average wage among men no matter whether there is. io. n. al. er. overeducation or skill mismatch. Plus, according to Robst (2007) and Martin and Shehan. i n U. v. (1989), men and women have different values when seeking a job and accepting. Ch. engchi. overeducation or skill mismatch. Men tend to look into the extrinsic part, such as promotion and career development; while women care more intrinsic aspects, such as company welfare and flexibility on work. The diverse values differ the average wage and the penalty effect from overeducation and skill mismatch. Though there are less female claiming overeducation, the wage penalty is similar to men. It is distinguishable if overeducation or skilled mismatch is claimed by women. Also, job mobility theory more fits men than women in this case. Hence, it is more likely that the male overeducation will alleviate, but not certainly for women. The effect of job satisfaction from skill mismatch is similar with both groups and not in line with the previous studies (Clark, 1997; McGuinness and Sloane, 2011), which point out overeducation and skill mismatch of females tend to have lower influence on job satisfaction. 30. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(37) because they cling to have a lower job expectation. The consequence explains Taiwanese female graduates have a similar expectation to males. Rankings Top schools usually offer more high-quality resources, environment, and networking than non-top ones. Therefore, school branding can give their graduates signal to employers as new labor market entrants. However, there are still 27 percent of overeducation and almost 20 percent skill mismatch. If the signaling theory doesn’t work for them, the education quality on the top school graduates can be doubtful. And the large disparity between those with signaling working well and those not reflect on the wage penalty. On the other hand, it is possible that there is imperfect information among those overeducated graduates from top schools, and thereby arise higher discontent toward the job because they consider to be deceived.. 立. Professions. 政治大. ‧ 國. 學. Among the division, medically-related graduates have less proportion of overeducation and skill mismatch but possess a highest negative wage effect and job dissatisfaction from. ‧. overeducation. This can infer to the knowledge-intensive medical schools may not cultivate. y. Nat. entire graduates well. There are part of them claiming overeducation and showing discontent. sit. toward the job because they can’t be treated as their other elite counterparts. They have high. al. er. io. job expectation. Also, they may still have high expectation to have a more decent job but. n. imperfect information drives them choose the non-ideal ones under the circumstance of the inefficient market mechanism.. Ch. engchi. i n U. v. In comparison, general universities cultivate a variety of backgrounds in different industries. Particularly, technology, computer science, engineering and so on professions are part of them. The graduates are easy to seek a job since the demand is still high in the market. However, it is possible there are graduates not from the profitable backgrounds. Therefore, general university students are inclined to have less wage effect from overeducation and skill mismatch. According to the same table, humanity-related school graduates have a stronger wage effect from overeducation because of assignment theory. An alike level of discontent from both predictors toward jobs indicates imperfect information and high expectation. Otherwise, the possible low education quality cultivates too much unworthy human capital.. 31. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(38) Overeducation has caused much less unpleasure than skill mismatch for general university graduates. It implies job mobility or job competition theory applicable for them. On the other hand, it is possible for them to consider other aspects of work. In an opposite way, medical graduates have a more considerable dissatisfaction from overeducation than skill mismatch. Their discontent causes from the uneven treatment as their elite counterparts or information asymmetry. Institutions Compared to typical universities, the graduates from science and technology universities have a smaller gap between education and the labor market. Skill mismatch is expected to be smoother to them, but not in the research. However, half of them feel overeducated, which is a larger amount than typical university graduates. Though reflecting on the results on wage. 政治大 technology university graduates. 立What they have learned is more practical and easy to link effect, both overeducation and skill mismatch have a smaller impact on science and. ‧ 國. 學. with work, and they have had an internship in the time of university, which applies to human capital theory. They need less training than typical university graduates. Plus, signaling theory works positively, and their graduate institutions cause them less wage penalty. And if. ‧. we look at both wage effect and satisfaction, it is likely science and technology graduates are. y. Nat. looking forward to being competitive or getting promotion in the future because they are. n. al. er. io. satisfaction.. sit. more willing to be overeducate according to a lower influence from overeducation on job. i n U. v. For the typical university graduates, learning academic-oriented knowledge may not be handy. Ch. engchi. for joining a labor market. Therefore, overeducation and skill mismatch have a more considerable effect on them. This reflects to human capital and signaling theory, they need more on-job training because of lack of practical experience. Two predictors have similar effects on job satisfaction of university graduates. Though overeducation has caused a higher wage penalty than skill mismatch, university graduates do not have significantly more unpleasure toward it. Both predictors arise the discontent of them. They possibly quit a job for either incidence. On the other hand, typical university graduates have a higher discontent from overeducation than science and technology university ones. This points out they have higher expectation to get a better position. Ownership. 32. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(39) Public schools in Taiwan tend to have more resources for students in comparison with private ones. This explanation is similar to the category of rankings above. Public university graduates are more able to find a job paired their education as signaling works well; while private university graduate has a much higher rate of overeducation. This incidence follows the signaling theory as well but in a negative way. For public schools accordingly, they have no much significant unpleasure from either overeducation or skill mismatch, which indicating the applicability of job mobility theory or job competition theory, or there are other aspects they care more than the previous two. However, private school graduates gain more discontent from skill mismatch, which aligns with most of the previous studies. It implies the potential imperfect information during job seeking and probably drives them to quit.. 政治大. Industries. 立. Bothe secondary and third industries have a more considerable effect from overeducation on. ‧ 國. 學. their wage. Assignment theory describes the wage penalty from overeducation. Secondary industries have more prominent effects from overeducation on wage than third industries.. ‧. Skill mismatch has a positive effect on secondary industry workers. It describes a better pay for skill mismatch and the skill-mismatched workers may have higher productivity on the. sit. y. Nat. jobs. Also they have higher proportions of claiming overeducation and skill mismatch. This indicates college graduates in secondary industries are inclined to be overeducated and skill. io. n. al. er. mismatched. The skill and techniques are less likely to require their knowledge learned from. i n U. v. tertiary education. The lower overeducation and skill mismatch of third industries implies the similar incident.. Ch. engchi. The higher discontent from skill mismatch of secondary industries can be reluctancy of choosing a job, though they have a positive wage effect from it. In comparison, third industry workers have less negative influence on their satisfaction. This explains they are more voluntary to be overeducated or skill mismatched. Job mobility theory or job competition theory can reflect the results as well.. 7. Conclusion Generally, most of the results are in line with the previous studies ‒ overeducation is a better predictor for wage effect and skill mismatch is better for predicting satisfaction ‒ but the effects are overall weaker in the case of Taiwan. When it comes to the specifications, there is. 33. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(40) more information from them. Men tend to feel more overeducated and skill mismatched to their jobs than women. And women have more serious job penalty because their wage is lower than men on average. Based on the research of Robst (2007) and Martin and Shehan (1989), though women are less sensitive to overeducation than males, the overeducation they have claimed may influence them more seriously. With the same level of schooling, their education brings not the same return as their male counterparts. However, the job satisfaction level is similar between two genders. This is different from most of the previous studies, which usually indicate women are happier than men. Taiwanese women have the similar expectation as males. Their discontent can be explained by job mobility theory or job competition theory. Under the assumption of signaling theory and well-functioning market mechanism, top 10,. 政治大 inclined to have lower overeducation and skill mismatch but higher negative return from the 立 two indicators. This implies elite schools may also produce less qualified graduates and medical and public schools are famous for cultivating elite graduates. The graduates are. ‧ 國. 學. thereby they have higher wage and job satisfaction effect from overeducation and skill mismatch. Job competition and job mobility theory can describe the less discontent from. ‧. overeducation. This also points out they are more willing to be overeducated. Human capital theory only applies to science and technology university graduates in the case because they. Nat. sit. y. usually have internship before entering labor market. They have more practical experience. er. io. than typical university graduates. Taking premise on my findings, strengthening the connection between higher education and labor market may improve the wage effect and the. al. n. job satisfaction.. Ch. engchi. i n U. v. Rather than the typical way of panel survey to control the heterogeneity by adopting the time series data, this study applies inverse probability weighting (IPW) method to take place of it. Because of the inconsistency of TEPS-B in the different waves of investigation, it is unlikely to take advantage of every previous survey to do panel study. Another limitation of the research results from lack of further survey after 2014 on the same cohort. Taking the premise from prior studies, the overeducation will reduce with the length of an individual in labor market extended because the accumulation of working experience enhances the professions and the skills of a job, and thereby the individual can be in a relatively suitable position (Caroleo and Pastore, 2013). Due to the incapability of the database, this cannot be convinced in the study. However, it is likely to count on the release of a future survey from TEPS-B and enables to facilitate further research.. 34. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.

(41) Also, as IPW models adopted in the research for controlling unobserved characteristics, there are some limitations encountered. It’s possible to lose other valid control from the predictable covariates. The limitation of this research lacks the chance to see the same issue in other aspects. And the confined studies about varied professions, institutions and ownership result in the insufficiency of a proper interpretation. It is expected to have more delicate work in the related topics in the future.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 35. DOI:10.6814/THE.NCCU.IMES.002.2019.F06.