Do Students Expect Any Returns on Effort Applied to Studying? An Econometric Analysis of Determinants of the Amount of Expected Wage after Graduation Marina Telezhkina1[0000−0003−2724−8973] , Andrey Maksimov[0000−0001−6307−1656] , and Nataliya Maksimova[0000−0001−5383−6088] National Research University Higher School of Economics, Nizhny Novgorod, Russia, 1 m.s.telezhkina@mail.ru Abstract. Students’ perception of the labor market makes a great deal in students’ decisions concerning effort to study, work during university studies, etc. The aim of the research is to define whether students iden- tify significant returns on effort with respect to wage after graduation. Moreover, it seems reasonable to single out other factors that students expect to influence their wage significantly. With the use of the data of Russian students’ questionnaire conducted in 2012 within the frame- work of the Monitor of Economics of Education project the regressions with the use of instrumental variables and stochastic frontier approach are estimated. The results suggest effort is considered as an influential factor in determining wage by Russian students if students’ incomplete awareness about labor market is taken into account. Besides, university quality, abilities, wage received by working students, region, specialty, family’s income and gender make the difference in the amount of wage expected by students. For additional analysis the 20% and 80% quantile regressions are built. According to the results, persons having the high- est wage forecasts base them on the amount of wage offered to working students on the labor market and do not correct them subject to their effort, university quality and abilities. At the same time another group of students, keeping similar basis for expectation formation as a previously analyzed group, expect significant contribution of effort and abilities. Keywords: expected wage, wage after graduation, returns on effort, quantile regression, stochastic frontier approach 1 Introduction Expectations are the base of decisions and actions of persons. One of the key decisions during university studies is time allocation between study, work and leisure. One can assume that among the determinants of students’ time alloca- tion are their wage expectations, in particular, expectations about factors which contribute to wage after graduation. For instance, effort dedicated to study, uni- versity quality, working experience can be among the factors that are expected to influence wage after graduation. If a person supposes that his/her effort to study influences in a great extent the wage after graduation then he/she will decide to spend more time studying. If a student expects that university quality makes a great deal in his/her positions on the labor market after graduation and determines his/her wage then he/she probably will dedicate less effort to study- ing relying on university reputation on the labor market. In the case a student thinks working experience is the most valuable asset to an employer and study- ing in a university is just a social norm or parents’ wish then he/she will also spend as less time studying as possible. Thus, having known the relevance of ef- fort to study in determining student’s positions on the labor market assumed by students, undertaking appropriate measures, for instance inviting employers to make a talk about vitality of knowledge got in university, can stimulate students to dedicate more time to study. This can be treated as an argument support- ing the relevance of the research. Consequently, the aim of the research is to determine the influence of their effort on wage after graduation as expected by students. Moreover, it makes sense to reveal other factors, which students sup- pose to influence their wage after graduation, and measure the degree of their influence. The research is based on the data of Russian university students’ question- naire in 2012 conducted within the framework of project Monitor of Economics of Education organized by National Research University Higher School of Eco- nomics, the analytical center of Yurij Levada under control of the Ministry of Education of Russia. The dependent variable is the amount of expected wage of Russian university students. The authors analyze the factors that are tradition- ally considered in this kind of research among which are gender, year of study, family’s disposal income, specialty, region. The list of regressors is supplemented by the proxies for effort, university quality, abilities and the amount of wage received by working students. Regression approach is applied to the conditional expectation of the dependent variable and 20% and 80% quantiles of the dis- tribution. The said approach includes estimation with the use of instrumental variables and stochastic frontier method which is used to control for students’ inefficiency in estimates and limited knowledge about labor market. The results suggest the existence of the difference of expectation formation between two groups of students with the lowest and the highest wage expecta- tions. The latter build their expectations on the amount of wage they are offered working during studies. They do not associate the variability of the sum of wage after graduation with the amount of effort applied to studying and university quality. Abilities are not a significant factor as well. They are the students al- ready having a well-paid job to the time they enter university. They usually do not apply much effort. Persons with the lowest wage expectations also take into account the wage offered to working students but still they assume effort makes the difference in formation of wage after graduation. So do abilities. University quality is also not admitted as significant in explanation of variation in values of expected wage of this group of students. The results of the regression for all students of the sample propose effort, university quality, abilities and wage of working students are significant if we take into account “individual inefficiency”. In terms of the above stated aim of the research, it means that students realize vitality of their effort for their positions on the labor market. Thus, it is worth to coordinate university policy so that it drives students’ decisions to work harder which, in turn, will result in benefits for both students and university. Having revised the factors influencing students’ wage forecasts we have real- ized that the parameter is prone to labor market shocks connected with changes in an average level of wage and that offered by employers to students; fiscal shocks expressed in variations in transfers and tax rates; personal shocks that cover all possible changes in life of a student, his/her preferences, perception of labor market, etc. Section 2 gets through the literature on the research question. Section 3 reveals methodological and data aspects. In section 4 the results are presented and discussed. Section 5 discusses the influence of different kinds of shocks on students’ wage expectations. Section 6 provides a conclusion. 2 Literature Review The pool of research conducted on the data of different countries (USA, Den- mark, Czech, Britain, Cyprus, Switzerland, etc.) found the similar set of variables determining the expected wage after graduation and revealed similar patterns of influence of the factors on the dependent variable. The findings can be sum- marized as follows. Firstly, scholars point out that wage expectations play a great role in decision- making concerning specialization of studying and choosing the amount of effort to study [6, 8, 9, 18, 19]. Secondly, younger students have higher expectations about wage after gradu- ation than elder ones. This is explained either by the rate of inflation taking into consideration by younger students or by the fact that they do not have enough information about labor market [5–7]. Thirdly, female students have lower expectations about wage after graduation in comparison with male students [4, 6, 7, 10]. Fourthly, high family income increases students’ wage expectations [1, 6, 14] relatively to persons from low-income families who tend to underinvest in their education because of significant financial resources deficiency etc. [11]. Fifthly, specialty, level of parent’s education, abilities and working experience are also among factors influencing students’ expectations about wage [5,7,11,14]. The researchers provide controversial evidences concerning accuracy of stu- dents’ wage estimates. [6, 9, 10, 15–17, 19]. It is necessary to mention that despite the fact that the influence of wage expectations on students’ decisions about the amount of effort allocated to study is accounted by scholars, the contribution of effort to wage assumed by persons is not mentioned or verified. Furthermore, the influence of university quality on students’ wage expectations has not received enough attention. Our research is aimed at filling the gaps. 3 Methodology and Data We are going to consider the following determinants of the amount of expected wage in the research1 (Table 1): Table 1. Independent variables under consideration Factor Variable (name) Variable (description) effort dedicated to t_self the amount of hours dedicated to study self-education (hours per a week) university quality share_oregion share of students from other regions studying in the university gender female binary variable, = 1 for female students (control variable) year of study year_of_study dummy variable, takes values from 1 to 4 (control variable) family income yd dummy variable for family’s monthly disposal (control variable) income (in rubles), yd = 1, if income ∈ [0, 2000], yd = 2, if income ∈ (2000, 4000), yd = 3, if income ∈ (4000, 7000], yd = 4, if income ∈ (7000, 10000], yd = 5, if income ∈ (10000, 15000], yd = 6, if income ∈ (15000, 20000], yd = 7, if income ∈ (20000, 30000], yd = 8, if income more than 30000 rubles specialty techn binary variables, (control variable) ped techn = 1 for students of technical specialties med ped = 1 for students of pedagogical specialties med = 1 for students of medical specialties region Moscow binary variable, = 1 for students of Moscow (control variable) universities working experience w1 wage a person gets working during university studies (in rubles) abilities ege_av the average score of all state exams a student took graduating from school 1 The descriptive statistics of variables under consideration is presented in Table 3 in Appendix (except dummies) It is necessary to mention that because of the fact that effort, university quality, working experience and abilities are latent or vague to interpret we use proxy variables mentioned in the table instead of them. As for effort, it might be argued that it is necessary to use the overall studying time which includes time spent on lectures and seminars and time to self-education and preparation of tasks. However, the former depends not only on student’s willingness to study but also on time of commuting to studies and other factors. At the same time, one might suppose that the latter clearly demonstrates student’s effort to understand a discipline. Evidently, we need to control for abilities here. Certainly, exam results depend not only on person’s abilities but also on his/her characteristics, for instance concentration, stress resistance, self-confidence, etc., and teacher’s effort. Nevertheless, the existence of positive correlation between exam results and abilities justifies using the chosen proxy. As for university quality, we suppose that demand on university educational service clearly reflects the quality of that service. We consider share of students from other regions arrived to study in a particular university as a parameter revealing demand on university educational services. However, this parameter demonstrates not only quality of university but also quality of region in comparison with native one for an individual. However, positive correlation between university quality and share of students from other regions is an argument in favor of using the aforementioned proxy for university quality. Working experience can be treated in terms of its presence or not, years of working, years of working on specialty, whether work on the last workplace is connected with specialty, etc. We take into consideration the first meaning. In addition to this, we take into account amount of wage a student receives working during studying because probably this value will be, firstly, the base for formation of wage expectations, secondly, it probably will be his/her reserve wage entering labor market after graduation. For the purpose of our research we use the data of students’ questionnaire conducted in the framework of the Monitor of Economics of Education by NRU Higher School of Economics together with the analytical center of Yurij Levada under control of the Ministry of Education of Russia in 2012. It is necessary to point out that the sample includes only bachelor students whose study is financed by the government. This is the limitation of the research. The final sample includes 553 observations. The dependent variable is logarithm of expected wage after graduation (ln_w2). 4 Results Having conducted the regression analysis with the use of two-step least squares method, we chose the following regression specification (1) basing on economet- ric criteria and theoretical assumptions. The results are presented in Table 2 (Column “IV”). ln_w2i = β0 + β1 t_self i + β2 t_self 2i + β3 share_oregioni + + β4 femalei + β5 year_of_studyi + β6 ydi + β7 techni + + β8 pedi + β9 medi + β10 Moscowi + β11 w1i + β12 ege_avi + εi (1) Table 2. The results of regression analysis of logarithm of expected wage with the use of different methods Quantile Quantile Stochastic ln_w2 IV regression, regression, frontier 20% 80% approach t_self 0.00004 -0.029∗∗∗ 0.003 -0.017∗∗∗ (t_self)2 0.0001 0.0007∗∗∗ -0.00006 0.0004∗∗∗ share_oregion 0.003∗ 0.002 0.002 0.02∗∗ female -0.118∗∗ -0.159∗∗∗ -0.168∗∗∗ -0.147∗∗∗ year_of_study -0.068∗∗∗ -0.109∗∗∗ -0.118∗∗∗ -0.094∗∗∗ yd 0.056∗∗∗ 0.055∗∗∗ 0.022∗∗ 0.047∗∗∗ techn 0.084 0.128∗ 0.133∗∗ 0.109∗∗ ped -0.163 -0.212 -0.304∗∗∗ -0.305∗∗∗ med -0.112 -0.136 0.005 -0.048 Moscow 0.432∗∗∗ 0.474∗∗∗ 0.457∗∗∗ 0.479∗∗∗ w1 · 10−5 0.826∗∗∗ 0.717∗∗ 1.400∗∗∗ 0.637∗∗∗ ege_av 0.008∗∗∗ 0.008∗∗∗ 0.002 0.005∗∗∗ Const 8.961∗∗∗ 8.976∗∗∗ 10.111∗∗∗ 9.279∗∗∗ R-squared 0.380 0.220 0.215 - sigma_v2 - - - 0.426 sigma_u34 - - - 0.228 Firstly, it is necessary to point out the presence of endogeneity in the model with endogenous variables of time dedicated to self-education and the square of that variable. The reason is that not only effort expressed by the endoge- nous variables has a presupposed influence on wage but wage expectations do affect effort stimulating or de-stimulating a student. An instrumental variable is amount of time allocated to writing course papers, a bachelor project, etc. The instrument is correlated with amount of time for self-education as far as they characterize a student as more or less diligent. However, the amount of time for writing course papers, etc. is hardly connected with expected wage. The results correctly reflect labor market trends in Russia. In particular, according to the estimates, wage expectations of Moscow universities students 2 sigma_v denotes the estimate of standard deviation of the stochastic component νi 3 sigma_u denotes the estimate of standard deviation of the stochastic component ui 4 LR test of sigma_u = 0: Prob ≥ chibar2 = 0.008 are approximately 40% higher than that of students of other universities. In addition to this, despite the influence of specialty on value of expected wage is insignificant, the signs of coefficients are correct. In particular, technical specialty increases while pedagogical and medical specialties decrease the level of expected wage in average in comparison with graduates of other specialties. Furthermore, tendencies mentioned in other research works are supported by our results. The coefficient reflecting the influence of years of study on wage has a negative sign. The fact means that students of earlier years of studies expect to receive higher wage after graduation than elder students. Gender has negative wage contribution for female students as have been already noticed by researchers. The influence of family’s disposal income on the amount of expected wage is also in correspondence with findings of other researchers. The influence of share of students from other regions and average exam score on the value of expected wage match with the theoretical rationale. In particular, according to the estimates, higher university quality and higher abilities result in increase in the amount of expected wage. Although the coefficient of the variable of time dedicated to self-education turned to be insignificant, we leave it in the regression as far as it is in the focus of study. Moreover, we assume the effect of the proxy variable for effort on wage to be nonlinear. For the purpose of the research the regressions for 20% and 80% quantiles of the dependent variable are considered (Fig. 1). The aim is to check whether the influence of effort on expected wage differs for the abovementioned groups of individuals with high and low values of expected wage. The results are in Table 2 (Columns “Quantile regression, 20%”, “Quantile regression, 80%”). Fig. 1. Histogram for values of wage expected after graduation by university students (with the use of kernel of standard normal distribution) In general, the estimates of the coefficients are rather similar in the re- gressions. However, there are some differences in the influence of time to self- education, average exam score, wage for working students and medical specialty on the dependent variables in regressions for 20% and 80% quantiles. The coef- ficient of the variable w1 in the regression for the group of students expecting the highest wages after graduation is greater than in the regression for students from another group. In addition to this, students constituting 80% quantile of the dependent variable do not assume any significant influence of effort to study, university quality and their abilities on their wage. If it turns to individuals with the lowest wage expectations, this group of students is characterized by significant influence of effort and innate abilities on the amount of wage. To sum up, students do differ in the way they form their expectations and the weights they associate with factors of the amount of expected wage. Speaking about students expecting high wages after graduation, they evidently base their forecasts on the wage they get working in university. Students from the cohort with the lowest expectations also pay attention to the amount of wage they re- ceive working during university studies in formation of their wage predictions. Nevertheless, they expect positive returns on effort and abilities with respect to wage after graduation. In order to find explanations for the existing difference in formation of wage forecasts we will analyze the two groups of students in more details. It is necessary to pay attention to the fact that the majority of individuals belonging to the group with the least wage predictions are not stu- dents of Moscow universities. In addition to this, there are much more students of technical specialties (18%) among the students with the highest 20% of wages than in the group of students with the lowest 20% of wages (10% of students are getting technical specialty). At the same time, the former group includes significantly less amount of students of pedagogical specialty (0.84%) than the latter group (4.4%). Moreover, it is interesting to notice that the average amount of time dedicated to self-education by this group of students (11.19 hours in a week) is higher than the average amount of time to self-education of students with the highest wage expectations who in fact attribute no significance to the influence of effort on wage (10.12 hours a week). Finally, the average wage of working students in the group of individuals with the highest 20% of expected wages is approximately equal to 21,000 rubles that is almost two times greater than the average wage of the last 80% of students which is around 12,000 rubles. Thus, we can conclude that students forming the group with the lowest wage expectations are persons who are studying not in Moscow, who are getting peda- gogical specialties; they mainly do not receive high wage working during studies and pay quite a lot of attention to study. The group of students expecting to get the highest wages after graduation includes students of technical specialties and students who already have a well-paid job and do not apply too much effort to study. The effect mentioned in the last sentence has been already noticed by the statisticians of the Ministry of Education of Russia and is also described in the publication of the media source “Komsomolskaya Pravda” [13]. They call the reader’s attention to the fact that students of universities demonstrating low quality of education and having low ranks according to the official university rankings get the highest wages after graduation. The author of the publication comments that this kind of universities are highly attractive to individuals who already have a substantial working experience but they need a university diploma for further career promotion. They usually work full-time and do not have an opportunity and willingness to study hard. One can think of fallacies of particular individuals which can be expressed in incorrect knowledge about “law” of wage formation on labor market, wrong assumptions about value of particular skills and knowledge of an individual to an employer, errors in estimates of time allocated to self-education etc. These are named as “individual inefficiency” [2, 3, 12] and can be taken into account with the use of stochastic frontier regression. The idea of the approach is in considering noise as consisting of two components: one is a traditional stochas- tic component that is assumed to be normally distributed νi , another reflects individual inefficiency ui and usually is assumed to be distributed either ac- cording to truncated normal or exponential distribution functions. The classical formulation of regression specification is expressed by equation yi = α + Xi0 β + νi − ui , (2)  2 6 where i = 1, . . . , n, νi ∼ iid N (0, σν2 )5 , ui ∼ iid N+ 0, σu or ui ∼ iid expo- nential. The fact is that the idea of stochastic frontier was firstly presented for application to estimation of a production function. It was assumed that firms demonsrate inefficiency for some reasons and are not able to reach the maxi- mum of their production opportunities. Thus, an inefficiency component has a negative sign because of the assumption that it always results in decrease of firm’s production. However, things differ in our case. Student’s inefficiency can lead either to increase of the value of expected wage or decrease in comparison with the level of “justified expected wage”. Consequently, we need to introduce changes in the regression specification according to equation yi = α + Xi0 β + νi + ui (3)  2 where i = 1, . . . , n, νi ∼ iid N (0, σν2 ), ui ∼ iid N+ 0, σu or ui ∼ iid exponential. The estimation results are presented in Table 2 (Column “Stochastic frontier approach”). We considered an inefficiency term to be distributed exponentially. According to the results, all variables do significantly influence the depen- dent variable except the dummy variable for medical specialty. However, taking into consideration specific features of the Russian labor market we will leave this variable in the regression. It is easy to notice that an average student pays attention to his/her current wage, corrects his/her expectations subject to time dedicated to self-education and university quality. Abilities also matter in for- mation of wage predictions. Influence of other variables on value of expected 5 N (0, σν 2 ) denotes normal distribution with mean value equal to 0 and dispersion equal to σν2 6 N+ 0, σu2 stays for zero-truncated normal distribution with mean value equal to 0 and dispersion equal to σu2 wage after graduation is in agreement with the stylized facts, perception of their interrelation and findings of other researchers. It is necessary to mention that the stochastic frontier approach has been developed for production function modelling based on the idea of uniformity of technological process. Using this approach in the research we propose that students form their expectation in the same manner, or in other words, having the same “rule” in their head. This is evidently the limitation of our research. 5 Discussion It seems interesting to think of factors that may change student’s expectations, or in other words, kinds of shocks that can affect value of expected wage after graduation. Turning to the list of independent variables that are considered in the regression analysis we can divide them into several groups. Firstly, it is possible to single out fixed factors, which do not change, and con- ditionally fixed factors, which value can be assumed permanent under somewhat conditions. The former group includes gender and abilities. Specialty, region and university quality can be attributed to the latter group. The structure of data and the scope of the research allow to think that students do not change their specialty or university if they have chosen it once. We also do not consider the case of students’ expulsion. Moreover, university quality is suggested to be con- stant in short run and change in long run gradually as a result of substantial changes in university policy if they take place. Secondly, it is possible to account for variable factors of expected wage for- mation. It goes without saying that along with growth of a person and increase in the number of studying years, insights about labor market advance and become more realistic. Wage offered to working students on labor market is often regarded as a benchmark for formation of wage forecasts. That is why revision of value of students’ labor implemented by employers will lead to reassessment of wage after graduation. Family’s disposal income amounts to parent’s wages considering transfers and taxes. Consequently, changes in this parameter may be attributed either to labor market or fiscal shocks that can be the result of changes in the economy in a whole or in the state of a particular region. These alterations will supply variations in expectations concerning wage after graduation. An individual can apply more or less effort to study that is measured by the amount of time dedicated to self-education in our research. This parameter can undergo substantial and sudden shifts during the period of study driven by rearrangement of person’s preferences. Keeping assumed contribution of effort to wage constant, changes in amount of time devoted to self-education can initiate shifts in anticipated amount of wage that will be offered to a student after graduation. At the same time it is necessary to keep in mind that wage beliefs themselves play a significant role in decision about amount of effort. This means that if wage expectations are revised by a student due to changes on labor market or in economy in general then a person is to correct the expected contribution of effort and knowledge to wage after graduation and finally to redefine the time allocation between self-education, lectures and seminars attendance, work and leisure. All in all, the amount of expected wage after graduation is susceptible to the influence of several kinds of shocks. They are the following: – labor market shocks expressed in changes of wage offered to students and overall level of wage as well as changes in value of graduates’ knowledge to employers foster adjustments in the level of expected wage after graduation; – fiscal shocks, in particular, changes in transfers and tax rates may also lead to revision of expectations of all economic agents in general and reevaluation of expected wage by students too; – individual shocks associated with alterations in personal life of an individual, in his/her point of view, preferences may result in changes in the amount of effort to study or redefining of the role of factors in the law of the expected wage formation and reassessment of level of expected wage after graduation as a consequence. Additionally, it seems reasonable to point out that any changes in the princi- ples of building wage forecasts, in perception of labor market, will cause not only changes in expected wage itself but also in the amount of effort taken because of their interdependence. 6 Conclusion Formation of wage expectations of Russian university students is scrutinized in the paper. Apart from the traditional for this kind of research factors of expected wage, the authors analyze the influence of effort, university quality, abilities and working experience on value of the dependent variable. Two different ways of expectation formation can be detected for 20% of individuals with the highest and 20% with the lowest wage expectations. The latter are not studying in Moscow in majority. Effort is recognized as decisive by individuals from the second group. They also take into consideration the wage they get working during university studies; abilities do also influence the value of expected wage of this group of students. The primary role in formation of the highest wage expectations is played by the wage offered to working students. Effort and abilities do not influence the expected wage of this group of students. Thus, the evidence is provided for the fact that the highest wage expectations in Russia have students who are having a well-paid job, apply mostly to low quality universities and devote as less effort to study as possible to get a diploma. If we consider all the students together estimating the regression for the conditional expectation of the dependent variable then it appears that time allocated to self-education characterizing student’s effort has a distinct influence on the amount of expected wage in the case we account for inefficiency of students connected with their limited awareness about labor market. Returning to the research question, we may conclude that an average student expects his/her effort positively affects the amount of wage he will be offered after graduation. So do the students outside the Moscow region having the lowest wage expectations. Thus, it makes sense for university to take actions towards rising students’ effort and stimulate them to allocate more time to study. In general, it will result in improvement of quality of university graduates and will have positive impact on university reputation. Additional analysis suggests the existence of several kinds of shocks that may affect the amount of expected wage after graduation, in particular, labor market, fiscal and individual shocks. References 1. Abbiati, G., Barone, C.: Is university education worth the investment? The ex- pectations of upper secondary school seniors and the role of family background. FBK–IRVAPP Working Paper 2015–13 (2015) 2. Aigner, D., Lovell, C., Schmidt, P.: Formulation and estimation of stochastic fron- tier production function models. Journal of econometrics 6, 21–37 (1977) 3. Aigner, D., Chu, S.: On estimating the industry production function. The American economic review 58(4), 826–839 (1968) 4. Anchor, J.R., Fišerová, J., Maršı́ková, K., Urbánek, V.: Student expectations of the financial returns to higher education in the Czech Republic and England: Evidence from business schools. Economics of Education Review 30(4), 673–681 (2011) 5. Andonova, M., Mojsoska-Blazevski, N.: Factors influencing the earnings expecta- tions among macedonian students: A comparative perspective with the EU stu- dents. Croatian Economic Survey 17(1), 71–110 (2015) 6. Betts, J.R.: What do students know about wages? Evidence from a survey of undergraduates. Journal of Human Resources 31(1), 27–56 (1996) 7. Brunello, G., Lucifora, C., Winter-Ebmer, R.: The wage expectations of euro- pean business and economics students. Journal of Human Resources 39, 1116–1142 (2004) 8. Delavande, A., Zafar, B.: University choice: the role of expected earn- ings, non-pecuniary outcomes and financial constraints. ISER Working Pa- per Series 2014-38, Institute for Social and Economic Research (Oct 2014), https://ideas.repec.org/p/ese/iserwp/2014-38.html 9. Diaz-Serrano, L., Nilsson, W.: The reliability of students’ earnings expectations. IZA Discussion Papers Series 10700, IZA – Institute of Labor Economics (Apr 2017), http://ftp.iza.org/dp10700.pdf 10. Frick, B., Maihaus, M.: The structure and determinants of expected and actual starting salaries of higher education students in Germany: identical or different? Education Economics 24(4), 374–392 (2016) 11. Gamboa, L.F., Rodrı́guez, P.A.: Do Colombian students underestimate higher ed- ucation returns? Documentos de trabajo 012050, Universidad del Rosario (Aug 2014), https://ideas.repec.org/p/col/000092/012050.html 12. Greene, W.H.: Econometric Analysis. Pearson Education International, New York (2012) 13. Konukhova, K.: The paradox of domestic education: the highest wages have graduates of the weakest universities. Komsomolskaya Pravda, https://www.nnov.kp.ru/daily/26396.5/3273116/, accessed: 30.07.2017 14. Menon, M., Pashourtidou, N., Polycarpou, A., Pashardes, P.: Students’ expecta- tions about earnings and employment and the experience of recent university grad- uates: Evidence from Cyprus. International Journal of Educational Development 32(6), 805–813 (2012) 15. Pfeifer, G., Witte, S.: Students’ wage expectations in Germany – new evidence considering tax adjusted estimates. Conference Paper D13–V2, ZBW – Deutsche Zentralbibliothek für Wirtschaftswissenschaften, Leibniz–Informationszentrum Wirtschaft, Kiel und Hamburg (2014), http://hdl.handle.net/10419/100354 16. Simsova, J., Reissova, A.: How much will i earn? Expectations versus reality. Eco- nomics and Management XIX(2) (2016) 17. Webbink, D., Hartog, J.: Can students predict starting salaries? Yes! Economics of Education Review 23(2), 103–113 (2004) 18. Wiswall, M., Zafar, B.: How do college students respond to public information about earnings? Staff reports 516, Federal Reserve Bank of New York (Sep 2011), https://www.newyorkfed.org/research/staff_reports/sr516.html 19. Wolter, S., Zbinden, A.: Rates of return to education: The view of students in Switzerland. IZA Discussion Papers Series 371, IZA – Institute of Labor Economics (Sep 2001), http://ftp.iza.org/dp371.pdf Appendix Table 3. Descriptive statistics of variables used in the regression analysis Variable w2 ln_w2 t_self share_oregion yd w1 ege_av Mean value 24857.01 9.882 10.06 19.25 5.557 12715.3 68.00 Minimum 2500 8.294 1 0 2 200 32 Maximum 200000 11.918 56 68 10 200000 100