Размещено на http: //www. allbest. ru/

FEDERAL STATE EDUCATIONAL INSTITUTION

OF HIGHER EDUCATION

NATIONAL RESEARCH UNIVERSITY

HIGHER SCHOOL OF ECONOMICS

SAINT PETERSBURG SCHOOL OF ECONOMICS AND MANAGEMENT DEPARTMENT OF MANAGEMENT

Bachelor's thesis

In the field 38.03.02 `Management'

Educational programme `Management'

Overqualification and youth self-employment in Russia

Reviewer Position, degree

Kurbonov Orifjon Olimjon ugli

Academic supervisor Senior Lecturer

E. A. Zazdravnykh

Saint-Petersburg 2020

Table of contents

• Abstract
• Introduction
• 1. Theoretical foundation
• 1.1. Theory
• 1.2. Background
• 1.3. Measuring mismatch
• 2. The statement of the research question
• 3. Methodology
• 3.1. Data and sample description
• 3.2. Definition and measurement of variables
• 3.3. Analytic strategy
• 4. Description of the results
• Conclusion
• References
• Appendix

Abstract

The increase of demand for education in recent times caused much instability in labour markets. The number of young people working in jobs which are inconsistent with their education, or, more seriously, number of unemployed youths is rising rapidly across all regions of Russian Federation. The previous studies have merely focused on effects of overqualification as wage penalties, job dissatisfaction etc. This paper is aimed to expand the understanding of overqualification problem in Russia and explore whether Russian self-employed tend to experience education-job mismatch more than wage employees. Research is based on data from Comprehensive monitoring of living conditions surveillance data conducted by The Federal State Statistics Service in 2018. Following the availability of data in the dataset, this paper applies objective method of measuring overqualification. Using sample data for Russian youth, the probit regression model was chosen to explore the main goal of the study. The results of the work observe that self-employed Russian youth are more likely to experience under- or overqualification compared to their colleagues working in paid jobs. Moreover, results suggest stronger indication of underqualification among Russian self-employed youth.

Keywords: education-job mismatch, overqualification, underqualification, self-employment, wage and salary employees.

overqualification employee education

Introduction

With the growth of young people and development of economy in the world, education has been developing as a major factor of one's socio-economic well-being in society. Policy makers and academics have put forward education as a key instrument in gaining main socio-economic benefits, such as labor market competitiveness, strong economic growth, low rates of unemployment and other indicators of economic welfare of countries. From the perspective of history, there have been many scholars arguing that education is a major driver of income equality and economic growth (for instance Chiswick & Mincer, 1974; Kuznets, 1955; Tinbergen, 1977).

According to the Report of OECD, (2019) “Education at a glance 2019: OECD indicators”, the number of young people attaining education has been accelerating nonstop. More importantly, the report highlights Russian Federation as a country with one of the highest rates of educational attainment. Hereby, indicators of tertiary education attainment in Russia show that 63% of young people aged 25-34 had attained tertiary education, the second largest proportion after South Korea and way higher than the average indicator among other members of OECD which is roughly 40%; and countries of G20, where current indicator is 38%. Large share of entrants in bachelor's degree was found in the mathematics, science, engineering and technology fields. If to compare attainment of master's or equivalent degree, rates are also higher for Russia (34%) compared to average OECD indicators (14%). Aside from higher education, upper secondary and vocational programs participation in Russia accounted 46% of young people in 2017 compared to 40% of average in OECD, who are found mostly in broad fields such as manufacturing, construction and engineering (38%) followed by services, business, administration and law (18%). Post-secondary non-tertiary levels of education are also indicated to be relatively popular in Russian Federation with 20% of such educational attainment among 25-64 years-old, compared to 19% in OECD average. Despite the high level of educational attainment, International Labour Organization data shows that unemployment rate among Russian youth was at the level of 14,79%.

Continuous expansion of education among young people may assume that there is an unmet demand for supply of educated youth, basically because of the technological change, meaning that demand for young people with education and specialization can be unable to handle the relative expansion of supply. Consequently, they might be working in jobs where workers cannot fully utilize their qualifications, skills and abilities and end up being overqualified (Sйamus McGuinness, 2006). For many years, academics have been arguing on the negative outcomes of situation of being mismatched. Relative Deprivation Theory (Stouffer, 1949) suggests that overqualified workers can be more dissatisfied with their jobs, which can, in turn, lead to lateness, absenteeism, on the job search, and turnover (Allen & van der Velden, 2001; Belfield & Harris, 2002; Hersch, 1991; Sicherman, 1991). In a country level, the expansion of overqualified workers means that the full potential of economic development and productivity of countries will not be met (Bьchel 2002; Sйamus McGuinness, 2006) and it could increase the probability of equilibrium unemployment and intervene the governments' long-term development plans and strategies (Sianesi & Reenen, 2003). Despite the theories and predictions of scholars, overqualification has been frequent and long-term existing across countries (Chevalier, 2003; Sйamus McGuinness, 2006; Verhaest & Omey, 2010; Wen & Maani, 2019).

Despite the long history and popularity of the subject, most empirical and theoretical evidence on overqualification phenomenon concentrates on wage and salary workers and effects of overqualification on work-related aspects. (Bender & Roche, 2013; Congregado et al., 2016; Shevchuk et al., 2015) argue that there has been no research on the relationship between mismatch and self-employment, however the study of overqualification and self-employment would help to enlarge the understanding of mismatch and labour economics. Hereby, there is no enough evidence whether self-employed can also be overqualified and if the answer is yes, to what extend the incidence of overqualification is higher/lower in comparison to employees with salaries. Russian Federation encompasses many self-employed people by entrepreneurship and, mostly, not because of self-interest in field but the economic condition in the country. As Lazear, (2005) argues, self-employed are not coming from the top of the ability distribution, they are rather individuals with no alternatives and with deep necessity, most self-employed are what is left-over. However, this argument may contradict the entrepreneurs with billion-worth companies, but most of the self-employed are distinct from those successful entrepreneurs.

Following the previous gap in the literature, current study aims to observe the relationship between self-employed young people and overqualification in Russian Federation. The research results and strategy can be somewhat different from previous evidence, because this study adopts overqualification and self-employment terms for Russian socio-economic situation. As previously mentioned, the share of young people participating in education in Russia is one of the highest in terms of all post-school education types. The large number of young people with education is spread across all regions of Russian Federation, however, the development rate of each region is diverse. Much of the technological development and economic opportunities are centralized in large regions and cities of Russia, such as Moscow, Moscow region (Oblast) and Saint-Petersburg. This study assumes that educated young people, living in less developed regions face serious difficulties with employment in labour market. Difficulties may be related to employment procedures, the length of job search or, most obviously - with low level of salaries and financial freedom. Having experienced unsuccessful tries in the labour market, those young people decide to enter self-employment, normally, in different sectors of economy then their education, meaning that they are more prone to experience mismatch being self-employed. This may be a person holding a degree in science, but being self-employed in sector of wholesale or, vice versa, the person having an education in services, but being self-employed in construction sector (being underqualified).

Hereby, to understand the incidence of overqualification among Russian self-employed youth, this study uses data from `Federal Statistical Observations on Socio-Demographic problems' which is conducted by Rosstat. This section of observations include variety of datasets on different issues of which `Comprehensive Monitoring of The Population's Living Conditions' data conducted in 2018 was used as the main source of information. Initial number of observations of the dataset includes 60 000 households and roughly 130 000 individuals answering on different life-related questions. For this study purposes it was decided to use observations only for individuals following the logic of work aim. In order to conduct the analysis, firstly, dataset is cleaned for the observations necessary for the paper, such as excluding economically inactive population etc. Secondly, in consistency with the literature (Green & Zhu, 2010; McGuinness, Pouliakas, et al., 2018), mismatch is measured applying Empirical Method, which is believed to be more accurate and less subjective. Finally, hypothesis is tested using the main model. Also, for the purpose of confidence and robustness check, some other models are presented.

The major contribution of the current work is believed to be scientific rather than practical. First of all, current study is argued to be unique in Russian Federation, as there have been found very few numbers of papers studying overqualification in Russia. Most of the works were somewhat theoretical and applied no empirical evidence, while the others were not related to the field of overqualification and self-employment. Hereby, by answering the questions put forward in this paper, the results can be useful for future research, in terms of studying in a narrower pace and by applying different strategies. Undoubtedly, overall strategy and results of the work have some limitations, which can be easily explained by the novel introduction of empirical analysis of self-employed youth and mismatch.

The layout of the paper is as follows. Explanations, reasons and results of various previous studies for education-job mismatch and entrepreneurship are given in the Theoretical Foundation. Methodology part presents the data and explains the methodology of current study by stating the research question and hypothesis. Results part include results drawn from analysis. Finally, Conclusion part summarizes all the results and discusses further research directions and recommendations within stating main limitations faced in this study.

1. Theoretical foundation

1.1 Theory

Looking at the history of the topic being discussed, there are several various theories on the occurrence of overeducation which may contradict to each other and that cover only employees and employers in terms of supply and demand (Congregado et al., 2016). Talking about supply side of the problem, Human Capital Theory (Becker, 1962) for instance, suggests that overqualified workers tend to obtain skills in jobs for which they are mismatched to shift to higher level positions in the future; Theory of Career Mobility (Sicherman & Galor, 1990), as the replication of previous theory, argues that people intentionally choose jobs and positions where they do not fully utilize their qualifications in order to gain experience, training in order to have a stable career development further; Compensation Theory (Sicherman, 1991) assumes that people accept to be overqualified to compensate shortcomings in other types of human capital such as quality of education and work experience; Spatial mobility theory (Bьchel & Van Ham, 2003), argues that occurrence of overqualification can be explained by limitations in peoples' abilities or family conditions, meaning that workers lack sufficient opportunities in the labor market due to paucity of financial opportunities

On the contrary, demand side of the problem suggests some other theories, for instance Screening (or Signalling or Filtering) Theory (Spence, 1973; Stiglitz, 1975; Wolpin, 1977) or Educational Credential Hypothesis (van der Meer & Wielers, 1996) put forward the problem when employers typically require to `present' skills and abilities in order to signal the level of their productivity within showing educational credentials which are quite hard in measuring real productivity of the worker; Job Competition Theory (Thurow, 1975) is similar to former theories of demand-side of the problem when workers are ranked by their educational level and credentials as an indicator of good job performance; Job Search Model (McCall, 1970; Mortensen, 1970) or Matching Model (Jovanovic, 1979; Viscusi, 1980) theories discuss both supply and demand side of overqualification. The main idea is that employers and employees search for each other in a definitely uncertain labor market conditions and if the cost of information is high, both of them might limit the area of search and as a result worker end up being overqualified; Assignment Theory (Sattinger, 1993) states that normally heterogenous employees search for heterogenous jobs and, consequently, the probability of perfect match is less likely to happen.

Having analyzed these theories from the past, the following subsection discusses the fundamental studies conducted on overqualification and overall mismatch for the last several decades.

1.2 Background

European scholars were one of the first to show interest in education-job mismatch topic, where the concentration was mainly on European workers (Nordin et al., 2010; Witte & Kalleberg, 1995; Wolbers, 2003). Gradually, interest in overqualification has started to spread in the United States (Bender & Heywood, 2011; Robst, 2007). And just a few years ago, the research of mismatch expanded to post-transition countries, such as Slovenia (Domadenik et al., 2013), China (Zhu, 2014), Ukraine (Olga Kupets, 2016) and Russia (Shevchuk et al., 2015).

Despite the versatility of education-job mismatch topic, vast majority of previous research has been concentrating on the outcomes of being engaged in jobs and positions that do not fully match workers' education and qualification (Bender & Roche, 2013). Outcomes include wage penalties, job dissatisfaction, career mobility, high turnover rates and other job-related factors. Hereby, most of the scholars argued that mismatch is strongly correlated with wage penalties (see, for example, Borghans & de Grip, 2000; Chevalier, 2003; Groot & Maassen van den Brink, 2000). On the other hand, Allen & van der Velden, (2001); Wolbers, (2003) found that workers with education-job mismatch are more likely to have higher rates of job quits and turnover. Other wider group of scholars have discovered that overqualification can potentially lead to situations when worker is not satisfied with their jobs (Belfield & Harris, 2002; Bender & Heywood, 2006). Moreover, results for job dissatisfaction were identical even in case of using panel data (KeithA. Bender & Heywood, 2011; Mavromaras et al., 2010; Mavromaras & McGuinness, 2012; McGuinness & Wooden, 2009; Moshavi & Terborg, 2002; Verhaest & Omey, 2010). Allen & van der Velden, (2001); Di Pietro & Urwin, (2006); Groot & Van Den Brink, (2003); Quintano et al., (2008); Wolbers, (2003) connected overqualification with higher incidence of on the job search - the term used for situations, when employee is always trying to find a new jobs, that is consistent with their qualifications.

Interestingly, most of the literature analyzed for this study purpose considered mismatch only from perspective of wage and salary employees. There is still a gap in literature on whether overqualification differs between employment groups - namely whether there is a difference across self-employment and salaried jobs (Bender & Roche, 2013). As the first researches that shed a light on self-employment and overqualification, (Bender & Roche, 2013; Congregado et al., 2016; Shevchuk et al., 2015) argue that almost all research on overqualification has stressed exclusively on the wage and salary employees and self-employment has been truly ignored all these years. There are only a few studies, which included self-employment as a control variable in a regression models, however, without putting any more focus on it (Allen & van der Velden, 2001; Badillo-Amador & Vila, 2013).

One may argue whether self-employed can suffer from overqualification, because in fact, self-employed are those who contribute to their own work and there is no obligation to show any skills, knowledge and credentials to the employer. However, according to Bender & Roche, (2013) there might be various reasons to consider self-employed as one of the groups prone to experience overqualification. First of all, if decision of becoming self-employed is voluntary, it might be a way of finding a place or job where to utilize education more than in paid jobs. But still the self-employed is considered mismatched. Overqualification might be also higher in situations when transition to self-employment is caused due to difficulties with finding a job in a labour market. Also possible reason might be low salaries and wages, which make people to self-employ. In his work, Lazear, (2005) defines entrepreneurs as `jack of all trades' meaning that entrepreneurs possess wide range of various skills, which can potentially be an answer on why they tend to work in fields not related to their educations. Also, Lazear, (1977) had a theory that most of the self-employed are professionals such as dentists, scientists, physicians, lawyers and they tend to invest much in professional education because credentials of those self-employed could show quality of their products. Scholars also suggest that self-employed in the United States are prone to have higher levels of education compared to wage and salary employees. Also, Djankov et al., (2005) argue that most of the Russian entrepreneurs are people with higher education and, in general, those entrepreneurs can have much higher levels of investment in education if compared to paid employees.

First opener in this area, Bender & Roche, (2013), observe that self-employed are more likely to be mismatched, particularly if they are women. Authors also find that mismatch among self-employed leads them to a greater earnings penalty compared to wage and salary workers, however, earning declines are not associated with further job dissatisfaction. Perhaps, this may be partially explained by Hamilton, (2000), that self-employed have lower earnings, but they tend to continue working on their businesses without being stressed. Moreover, Bender & Roche, (2013) derives that self-employed men tend to be mismatched more often due to working conditions, while self-employed women are reported to be mismatched due to family reasons. Previously, some authors (see, for instance, Boden Jr., 1996; Connelly & Connelly, 1992; Hundley, 2000) also proved that there are different reasons behind being mismatched by gender. Shevchuk et al., (2015) in their study of education-job mismatch among self-employed Russian-language internet freelancers applying horizontal mismatch measurement, also observe consistent results with Bender & Roche, (2013). Specifically, they confirm that 40% of Russian internet freelancers work in fields that are not related to the area of their study. This figure is twice as higher than reported by Bender & Roche, (2013) for US self-employed in spheres as science or engineering.

1.3 Measuring mismatch

Based on the analyzed wide range of previous literature, there is a strong understanding that authors have been measuring mismatch in various ways. According to Capsada-Munsech, (2019), the methodological debate on how to measure the overeducation phenomenon has been present since 1980s and still nowadays scholars has not come to one consensus (Battu et al. 2000; Chevalier, 2003; Groot & Maassen van den Brink, 2000; Kucel et al., 2016; Hartog 2000; Sйamus McGuinness, 2006; Verhaest & Omey, 2006). Often the type of measurement is derived from data availability and in best it is recommended to implement several indicators to overcome existing limitations of each measurement type.

The recent analysis conducted by McGuinness et al., (2018) has summarized available mismatch measurement approaches in papers written between 2006-2017. In his paper, McGuinness et al., (2018) concludes that there are several ways of measuring mismatch such as: `Overeducation and Undereducation', `Overskilling and Underskilling', `Horizontal mismatch', `Skill Obsolescence', `Skill Gaps', `Skill Shortages', `Macroeconomic Indicator of Skill Mismatch'. Similarly but in more narrow aspects, Capsada-Munsech, (2019) made a summary of different methods for measuring `Overeducation' and `Undereducation'. However, the latter study considers only `Overeducation' and `Undereducation', it gives detailed description for each method and draws vital conclusions on the advantages and shortcomings of the methods by supporting them with empirical analysis. Current paper, precisely, relies upon first method of measuring mismatch, namely - `Overeducation' and `Undereducation'. The following subpart contains brief information on `Overeducation' and “Undereducation, `Overskilling' and `Underskilling' types of mismatch following the popularity of those two broad methods. Also some advantages and drawbacks of each methods will be discusses according to McGuinness et al., (2018).

Overeducation and Undereducation. There are 3 methods of measuring under- and overeducation: subjective method, the empirical method and the job evaluation method (Leuven & Oosterbeek, 2011). Even if each three methods are designed to measure the same mismatch, the magnitude of estimates can vary accordingly (Barone & Ortiz, 2011). The most frequently used methods are Subjective and Empirical method. Subjective method implies using workers' opinion on whether the skills or educational level matches the requirements of the job. Duncan and Hoffman (1981) were the first to propose this method and followed by (Sicherman, 1991; Sloane et al., 1999; Verhaest & Omey, 2010). The question may be direct when individual is asked whether he/she feels overeducated in the job or it might be indirect when the worker is asked about the required education `to do' or `to get' the job in comparison to his/her actual education level (Verhaest & Omey, 2006). Further, results from self-assessment are put together with the highest level of education obtained by the employee. It shads light on whether the worker is matched (has a level of education required by an employer), overeducated (with education level exceeding the required by employer), undereducated (does not possess required credentials or education required by employer). Typically, subjective method implies using binary variables in cases when there is an information on years of attained education. As one of the main advantages of subjective method. scholars conclude that it allows to apply survey data relatively easier than other methods. Subjectivity bias, however, is the principal downside of current method. Because respondents (workers) can feel over emotional when answering the questions of the survey, the survey result may be biased (Sйamus McGuinness, 2006). The empirical method is one of the objective methods of measuring the overqualification. Current method was first introduced by Clogg et al., (1984), developed by Verdugo & Verdugo, (1989). It measures the educational requirements of the job by calculating the modal or mean level of education within a job, dividing workers with required education above (or below) the average as being overeducated (or undereducated). The main advantage of current method is that it can be easily applied for any existing micro-data containing questions and answers on both educational attainment and occupational level such as national labor force surveys (McGuinness et al., 2018). As for drawbacks of empirical method, it does not include any information on skill requirement of the occupation, it, mainly, reflects average credentials or level of required qualification and helps to understand how `to win' and get the job, not `to do' the job (McGuinness et al., 2018; Capsada-Munsech, 2019). Verhaest & Omey, (2010) suggested that differences in required qualifications across different occupations can be classified within broad groups. In details, when empirical method is applied, the mode `educational' level is typically obtained from broad occupational groups such as `professionals in health', however it does not classify the job by more individual occupations such as `nurse' or `doctor' (McGuinness et al., 2018). Job evaluation method is another type of objective mismatch assessment when state of mismatch is defined by professional job analysts or experts. It is a type of approach based on correspondence of education and occupation and particularly made for professional dictionaries with occupational classifications. There are both advantages and disadvantages of this method. Results obtained by this method are judged to be the most pure and vivid because they can represent the real situation in the labor market and, additionally, it is based on expertise testing. And the high costs of carrying out those expertise assessments are known to be the only disadvantage of job evaluation method (McGuinness et al., 2018).

Overskilling and Underskilling. Overskilling is the term defining the situation when the worker possesses the skills above the level of required for their current jobs, while underskilling describes the opposite situation, when worker's set of skills do not meet the requirement of their current occupations. According to McGuinness et al., (2018), both concepts are measured subjectively by asking separate questions. As an example, Reflex Project data is popular among researchers, it includes various sets of questions such as “to what extent your skills are utilized in this job”, with a scale of response from 1 up to 5. Typically, the highest (5) and the lowest (1) responses are used in studies. Since it is impossible to measure underskilling through overskilling question, there is a separate question on underskilling in Reflex Project data, which is “to what extent does this work require more knowledge and skills that you can actually offer”. Measurement of overskilling has been argued to be more accurate than overeducation on the grounds that (a) job skill content is accurately reflected by job entry requirements, (b) worker qualifications truly reflect their total work related human capita (McGuinness & Wooden, 2009), whilst overeducation measurement does consider requiring more credentials and it misses the fact that individual human capital will also consist of skills and abilities acquired in labor market (Mavromaras et al., 2009). Besides, the main disadvantage of overskilling questions (subjectivity bias), occur when it is not clear that respondents are thinking about job-related skills answering the question, it also has some arguments against on the grounds that how the question is formulated.

Observing the statistics from McGuinness et al., (2018), of 94 overeducation research papers 21,5% uses subjective approach, 25,9% uses realized matches (empirical) approach and 25,5% uses job evaluation approach. Accordingly, undereducation and underskilling are studied within overeducation and overskilling, respectively. Hereby, of 30 papers of undereducation 10,7% implies subjective method, 26,2% uses empirical method and 15,8% uses job evaluation method. The average incidence of measuring undereducation using empirical method is statistically significantly greater than other two methods of measurement. Of 22 papers, 14 papers examine overskilling and overeducation together, and 8 study exclusively - overskilling. In these papers (McGuinness et al., 2018; Capsada-Munsech, 2019). Hard not to mention that of 39 countries, present in the literature, there is no incidence of measuring overqualification or overskilling in Russia.

As a brief conclusion of theoretical foundation and literature review, it is vital to mention that overqualification topic has been developing for several decades and it has been showing novelty findings which can be potentially useful in the future. However, there is a very shortage of studies concentrating on self-employment and overqualification, whereas the topic is argued to be important. Of all papers analyzed for this work, only a few (3) have been somehow related to self-employment. And, since there is a shortage of studies on self-employed authors, Bender & Roche, (2013); Congregado et al., (2016); Shevchuk et al., (2015) suggest to look deeper in the topic of overqualification within different employment types. Also, they assume that methods of measurement are not still definitely clear because there is no certain consensus on what is right and what is wrong, hence further research should be carried out by calculating various proxies for mismatch in order to check the robustness of the models.

2. The statement of the research question

Based on the analysis of the literature it was found that practically all researchers' attention was concentrated on overqualification and its effects on economic and social aspects of life such as job satisfaction, wage penalties, career mobility, job turnover etc. It can be easily noticed that interest in overqualification-entrepreneurship issues has been developing only these days. In general, of all analyzed literature, there have been very few researches conducted on how overqualification is connected with self-employment. What is more important, according to McGuinness et al., (2018), up to 2017 there have been almost no overqualification papers indexed in Russian Federation. However, during the literature review for current work, it was possible to find some papers on overqualification in Russia, but still 1-2 papers are not sufficient to draw solid conclusions. Based on the results of literature review, following research question has been proposed:

Research question:

`Is it true that most young people with education-job mismatch are self-employed?'

It may be assumed that research question is double interesting when applied in socio-economic conditions of Russian, because Russian self-employment is occurred not due to self-interest in working as an entrepreneur, but more with intentions of `surviving' and enhancing the financial conditions, or, more probably, to earn enough money to feed families. Preexisting scholars argue that incidence of overqualification has been increasing mostly among young people for last several decades and causing various hardships in the labor market (Barone & Ortiz, 2011; Chevalier, 2003; Kucel & Vilalta-Bufн, 2019; Quintano et al., 2008). The incidence of overqualification and other types of mismatch have been present in Russian labor market for long periods in a veiled occurrence and only minor theoretical studies tried to shed a light to this topic. Hereby, excluding large central cities of Russia such as Moscow, Moscow region (Oblast) and Saint-Petersburg, where labor economics and overall economy operates more or less like in a developed country and the share of wage and salary employees is higher in comparison to self-employed, current research focuses on other less developed Russian regions, where the income distribution in uneven (Shevchuk et al., 2015) to check if the results are robust across models. Consequently, one part of young people, after graduating from schools, colleges, universities are known to move to more developed cities as Moscow and Saint-Petersburg and the second part - to stay and continue living in other regions of Russia. Those young people, living in regions other than central, tend to be employed as a salary worker in the beginning, but unfortunately, the low level of wages make most of them to transfer to be self-employed. Indeed, current trend may occur mostly among graduates of universities. For example, the graduate of physics faculty in some region of Russia, probably, will not work in jobs related to physics and make any research, because there will be very few opportunities for that. Opportunity, to a great extent, is weighted in financial outcomes. Having found no financial outcome in the labour market, the physics graduate will decide to be self-employed and open an enterprise, where he will have more financial stability. However, from the second point of view, the graduate will not be self-employed in physics, but in a field, which does not require the knowledge of being physician. Hence, in this situation the mismatch issue occurs. In view of the above and in order to answer the research question, following hypothesis was put forward:

Hypothesis 1.

Self-employed youth are more likely to experience mismatch (under-, overqualification) than paid employees.

In order to test the hypothesis, this study stresses on the Job evaluation (Empirical) method of measuring mismatch (McGuinness, Pouliakas, et al., 2018). Using this method, education-job mismatch can be measured by cross tabulating the worker's education level with the level of education required for the job. Further, mismatch variables (variable of interest) and the state of being self-employed (response variable) are put into probit regression model in order to test the relationship between them. Also, as a robustness check, some other models are provided within showing marginal effects, which could help to discuss the concrete magnitude of change. In general, methods are consistent with the literature, however, there are some minor corrections made into models in order to be relevant with the market conditions of Russia. Next parts of the paper will discuss the dataset and give the detailed information on the methodology used in the study.

3. Methodology

3.1 Data and sample description

Current study uses data from the Comprehensive Monitoring of Living Conditions (CMLC) or in Russian - `Kompleksnoe Nablyudenie Uslovij Zhizni Naseleniya' for 2018. The CMLC is a standardized and annual longitudinal surveillance conducted at the level of regions (Oblast) of Russian Federation. The surveillance has been coordinated by Federal State Statistics Service of Russian Federation (Rosstat) in accordance with the Resolution of the Russian Federation Government № 946 from 27 November 2010 “On organization of series federal statistical observations on socio-demographic problems and monitoring of mortality loses, morbidity and disability of the population”. Rosstat, in turn, is a governmental organization which manages governmental statistical observations, and is guided by the Ministry of Economic Development of Russian Federation. According to the official document, the surveillance has been carried out from 2012 annually until now. The target population of the CMLC includes 130 610 individuals living in 60 000 households. The group of respondents is divided into group of children under the age of 15 and adults aged from 15. The foremost reason of the monitoring is to obtain annual statistical information on factual living conditions of Russian households and individuals which include questions on favorable living environment, healthy lifestyle and problems with health system, situation in the labor market, the upbringing and development of children, housing conditions, professional and educational situation in the living area, social mobility and many other socio-demographic indicators. Besides, surveillance includes various questions related to personal attributes of respondents.

As previously mentioned, the surveillance has data for both households and individuals separately. For this study purposes, it was decided to use data for individuals with 130 610 observations because in consistency with the research question, the main target group is young people, who are included in a group of individuals. Initially, the dataset included respondents who do not participate in labor market, who were then excluded from the final dataset. These are, mainly, pensioners, children and people with disabilities. Moreover, minority of people self-employed in agriculture and farming are excluded from the dataset.

3.2 Definition and measurement of variables

The main goal of this study is to identify the relationship between overqualification and self-employed youth. Undoubtedly, the clear definition of youth can be the vital aspect of the research. Having analyzed broad scope of literature, it was unclear what is the most appropriate age group for youth. Some scholars (Crecente-Romero et al., 2018; Williams, 2004) restrict to 20-24 years old, others (Verhaest & Van der Velden, 2013) argue that qualified and skilled workers are those aged 25 and older, Kucel & Vilalta-Bufн, (2019) suggest that on average by the age of 25, people finish their studies at the university. On the other hand, different reports of OECD define youth as 20-34 years old. Therefore, accepting the little uncertainty in the literature, this study defines youth as individuals aged 20-29, however, the sample is restricted for 20-34 years old in order to make further robustness check and analyze whether different age groups can influence the results. The logic behind starting the interval of youth from 20 is based on a simple fact that most young people by this age are supposed to finish upper secondary and postsecondary education and have a stable job; young people who participate in tertiary education, which lasts normally 4 years in Russia (bachelor's), are supposed to finish their education by the age of 22-23. 29 years old was chosen as the threshold for youth interval, because some of the youth tend to participate in postgraduate studies. Besides, there are many youths, studying in Medicine, where education lasts up to 11 years cumulatively.

The definition of self-employed is also one of the most important parts of this work because the individual's type of employment is considered to be the response variable. The dataset derived from CMLC has already included readily calculated variable indicating the employment status of an individual. This way, the response variable is a dummy, which identifies the current working status of an individual - 1 if the respondent is self-employed and 0 if respondent is not self-employed (wage or salary employee). In order to make it clear who is considered self-employed and who is not, it is most appropriate to cite Rosstat. It defines self-employed as “Individuals who perform work defined as “work in own business”. This is a job in which the renumeration directly depends on the income received from the production of goods and services. A person makes production decisions related to the activities of the enterprise or delegates these powers, leaving responsibility for the well-being of the business”. Individuals who are not self-employed are supposed to work in jobs with salaries, hence without feeling any strong responsibility or having a linear relationship between their contribution to the enterprise and the level of salary. The amount of money they generate is fixed in a paper agreement with an employer.

With respect to prior literature, the measurement of mismatch is undertaken within several more complex procedures (Green & Zhu, 2010). As previously mentioned, overqualification is defined by means of both subjective and objective calculations in literature (Congregado et al., 2016). It was decided to use the empirical method (objective) of mismatch measurement, because it is considered less subjective than self-evaluation method presented in the literature review. Besides, this method best fits the CMLC data available for the study. According to the empirical method, an individual is known to be overqualified when his level of attained education is higher than the level of education required by the occupation. This way, education-job mismatch was estimated by comparing educational level and the level of occupation of an individual. Variable indicating the level of education initially consisted of 9 different categories of education present in Russia (see Table 1), which were then combined into 2 wide categories to differentiate between individuals with a higher (specialist) and other (post-secondary, secondary, no qualification) degrees of education. The level of occupation also consisted of 8 categories, which reflect the stepped levels of qualification and wide job positions. Similar to the previous grouping method, this variable was also divided into 2 groups (see Table 1). The first indicating the positions with a requirement of high qualifications and the second group with medium or low levels of qualification. Latter 2 categories further will be referred as graduate jobs and non-graduate jobs.

As presented in Table 1, Rosstat has divided respondents into detailed categories with levels of education and the levels of occupations. Categories begin with the top levels (highly qualified workers and jobs requiring high qualifications) and gradually end with low level of education and jobs. The lowest level of jobs, evidently, does not require any qualification. Current approach of evaluation is adapted due to the absence of alternative measures which could be made within Rosstat dataset. However, the approach used in this study is nearly similar to that adopted in previous studies of scholars (Croce & Ghignoni, 2012; Ghignoni & Verashchagina, 2014; McGuinness, Bergin, et al., 2018). Besides, there is a confirmation from previous studies stating that the choice of overqualification measure tend to have a little consequence in terms of the estimated impacts (McGuinness, 2006).

In this stage of the analysis, it is vital to stress on the division of occupational levels used in the survey because it may face some critique on being subjective. Hereby, those classifications are derived from the official governmental document “Russian Classification of Occupations OK 010-2014 (MSK3-08) from 01 July 2015” developed by the Ministry of Labor and Social Protection of the Russian Federation. The Russian Classification is developed based on the International Standard Classification of Occupations 2008 (ISCO-08). Table 1 with various occupational levels is fully consistent with the official classification, however the original document includes some more detailed divisions at specific levels of occupations. The general division is based on worker's level of education, their qualifications and work experience.

Table 1 Ungrouped categories of variables

Overqualification occurs when an individual with higher educational degree and most probably, is working in non-graduate jobs. The term underqualification, in turn, occurs in a very opposite situation, when a person with middle or low education is working in occupations which require high levels of qualification or education. Perfect match is observed only when individuals with a higher education are working in graduate jobs and individuals with lower education are working in non-graduate jobs. Table 2 shows the general logic of qualification comparison in a way of cross tabulation, which is derived from the work of Green & Zhu, (2010).

Table 2 Typologies of mismatch

As it was mentioned earlier, variables for Educational level (EL) and Occupational level (OL) consist of 2 categories, say 1/0, where 1 - higher education/graduate jobs, 0 - other education/non-graduate jobs. One is supposed to be matched in their positions when EL=1, OL=1 and EL=0, OL=0. However, overqualification occurs when EL=1 and OL=0, and underqualification occurs when EL=0 and OL=1, right opposite to overqualification.

Because previous studies divided control variables into broad categories (Shevchuk et al., 2015), this study also uses two categories of variables as a control for regression analysis: demographic (gender, age, marital status, parenthood, region of living, living in city or rural places); education (years of schooling). Age was measured as several years from the birth date in a form of continuous variable. Additionally, to avoid non-linear effects, age is used as a squared term. Marital status consisted of different groups, which were then combined into two wide groups indicating the person to be married (official marriage and civil marriage) and to be single (divorced, widow, has never been married). Parenthood was derived from a question “How many children aged under 9 do you have?”, the answers including and above 1 (having 1 child or more) are coded into dummy category “Have children” and 0 are coded into another category “Do not have children”. Years of schooling is presented as a continuous variable. Dummies for regions are also used motivated by the fact that some regions appear to be more advantageous than others in terms of self-employment.

3.3 Analytic strategy

The foremost objective of this paper is to observe the relationship between self-employed youth and education-job mismatch. As mentioned in the previous parts, the initial steps fall in with analyzing the papers of previous scholars and understand the best way to measure the mismatch. This was the most important step of the work as mismatch is known to be a subjective measure, which has many drawbacks and critiques from experts. As the main tool for making manipulations with data and building the regression model to test the hypothesis it was decided to use the STATA statistical package.

In order to test the hypothesis it was essential to choose the best method of regression analysis. Overall, the study has nominal response variable which identifies the status of being self-employed or wage employee (1-self-employed, 0-wage employee). There are also several covariates (variables of interest), which include categorical indicators of experiencing education-job mismatch. This overqualification variable includes three categories: if a person is matched, overqualified or underqualified. Also, there are other control variables on social, demographic, education indicators. Among control variables, Age and Years of schooling are continuous variables, while others are categorical variables.

To observe the linkage between self-employment and the fact of being overqualified, there has been made deep research on the substantive (econometric) aspect of the topic. The choice of the appropriate regression model was quite easy to make, however, there have been some challenging situations. At a glance, one can argue that this hypothesis can be tested by using simple LMP (Linear probability model) of OLS regression to forecast the probability of positive outcome in the response variable. However, despite its simplicity, many experts assume that OLS is a not a very useful tool for predicting the outcomes of the nominal response variable (Gujarati, 2011, p. 248) . Nominal response variable, unlike ratio variables, is known to have two categories: positive and negative. Models with nominal response variables are also known as binary or dichotomous or dummy dependent models (Gujarati, 2011, p. 248). There follows a detailed description of why the OLS regression is not the best fit model for analyzing model with nominal response variables (Baum 2006). Firstly, LMP assumes that the probability of positive outcome of response variable (the probability of being self-employed in terms of this paper) transposes linearly with the value of explanatory variable, however it is not true, because the connection between those variables are not always linear. Secondly, LMP omits the constraints of the probability which should be between 0 and 1. By default it can predict the probabilities beyond the restriction level. Consequently, it makes the model hard to interpret. Thirdly, the common practice, when the error term has a normal distribution, cannot work for cases when response variable takes only 0 and 1.