LEXONOMICA Vol. 14, No. 1, pp. 29–48, June 2022 https://doi.org/10.18690/lexonomica.14.1.29-48.2022 © 2022 University of Maribor, University Press PRIVATE RETURNS TO EDUCATION IN DEVELOPING COUNTRIES: EVIDENCE FROM KOSOVO Accepted 25. 8. 2021 Revised 5. 1. 2022 Published 21. 6. 2022 MIC UKAJ, AVDULLAH HOTI, RAHMIJE TOPXHIU University of Prishtina "Hasan Prishtina", Prishtina, Kosovo micukaj@hotmail.com, avdullah.hoti@gmail.com, rahmije.topxhiu@uni-pr.edu CORRESPONDING AUTHOR micukaj@hotmail.com Keywords human capital, rate of return, education, wages, labour market Abstract This paper aims to assess the reward that employees receive from education in the Kosovo labour market in terms of a wage premium. The incentive to address this issue comes as a result of the high increase in participation in formal education in recent decades, especially in higher education. The research is mainly focused on private benefits from education, in terms of wage increases, assessed through a private return to investment in education. For this research, microdata from the Labour Force Survey has been used, conducted by the Kosovo Agency of Statistics. We used only the sample of wage employees, which includes 9,300 individuals. The econometric model used in processing the empirical results is based on Mincer's wage equation. The results presented in section 5 reveal that the rate of return on education investment has been sufficient to keep the demand for education investment alive. In comparison to the other two levels of education (primary and secondary), the results indicate a very high rate of return to tertiary education, justifying the high enrollment in this level of education. Females receive a higher rate of return than males. 30 LEXONOMICA. 1 Introduction Education is one of the main forms of investment in human capital; hence the interest among researchers in assessing its impact on the economy, particularly in the labour market, has been quite high. The effects it produces for the individuals or society are mainly evaluated through the calculation of the rate of private or social return from education (Borjas, 2013: 267). Kosovo is ranked among countries where research is lacking in many fields. Investing in human capital, as well as the effects that it produces for the economy, is one of the least researched areas. On the other hand, the importance of research in this area is quite high, given the fact that the country has the youngest population in Europe; therefore, investment in human capital remains one of the main resources for sustainable long-term development. In recent decades, a significant increase in higher education participation has been evident in Kosovo. According to the data from the Kosovo Agency of Statistics, in the period between 2007 and 2019, the number of participants in higher education has increased from 25,840 to 104,579 people (KAS, 2020: 71-75). The increase in higher education participation has been significantly affected by the increase in opportunities because of the establishment of several new public and private universities and colleges. Therefore, given this high interest in education investment (especially in higher education), it has been considered with interest to test the economic rationality of such an investment. In this research, we test only for the rationality of investing in education, evaluated in terms of its impact on wages. Using the Labour Force Survey data from the Kosovo Agency of Statistics, this research estimates the private rate of return to investment in education based on Mincer's wage equation. The empirical analysis includes wage-earning individuals only. The results indicate that the rate of return to education was sufficient to justify the large influx of Kosovars to invest in education, especially in higher education. However, the low opportunity cost of attending higher education may affect the high demand for attending higher education. The remainder of the paper is organized as follows. In Section 2, we discuss the theory of investing in human capital, the main methods used to estimate the rate of return from education and some of the empirical results around the world. Section 3 presents a description of the data used in this research. Section 4 discusses the estimation methodology utilized in this research, while the findings are presented in M. Ukaj, A. Hoti, R. Topxhiu: Private Returns to Education in Developing Countries: Evidence from Kosovo 31. Section 5, where a few comparisons are made with other developing countries. Our conclusions are drawn in the final section. 2 Theoretical background According to Schultz (1961:11) and Becker (1993: 17), investing in human capital refers to the investment of current resources in exchange for higher future benefits. These investments can be in the form of monetary expenditures and time dedicated to education, training or other forms of human capital investment, such as health and migration (Schultz, 1961:1). According to Becker (1993:17), education is one of the most important forms of investment in human capital, which will be of primary interest in this research. The benefits of investing in education can be monetary and non-monetary, which can be of benefit to the individuals who make the investment as well as to society as a whole. Benefits for individuals can be in the form of direct benefits such as increased wages and employment probabilities, better job positions, etc., as well as in the form of indirect benefits, which can be manifested with better health care, active participation in social life, elimination of bad habits such as smoking tobacco, alcohol, etc. On the other hand, the education social benefits can also be numerous, ranging from the effect on economic development and increased welfare (thus resulting from the higher productivity of the more educated workforce), continuing with a range of other indirect benefits, which may be as following: crime reduction, environmental protection awareness, active participation in social life, etc. Furthermore, for benefits and costs of education, see Ukaj and Mustafa-Topxhiu (2020:194,195). In the empirical evidence, the complete discounting method and Mincer's wage equation are used to estimate the rate of return on schooling (Psacharopoulos and Patrinos, 2018: 3). The first method can be used to estimate the private and social rate of return; however, in this research, our focus will be only on private return to education. This method requires rich data on education, wages, years of experience and so on. Given that few countries possess such data, researchers have preferred the second method, which calculates the rate of return based on cross-section data to estimate the average living income profile for people with different levels of education (Woodhall, Hernes and Beeby, 2004: 21). As a result, in the empirical literature, the majority of studies that look at wage disparities between workers with 32 LEXONOMICA. different levels of education have used this method to calculate the rate of return (Psacharopoulos, 2018: 6; Hampf, Wiederhold and Woessmann, 2017: 5) Mincer (1974) used this method to estimate the rate of return on education using data from the 1960 US census. He discovers that the rate of return on education is about 10%, while the rate of return on experience is around 8% (Harmon, Oosterbeek and Walker, 2003: 117). Since then, Mincer's wage equation has been one of the most widely used models in empirical economics. Psacharopoulos and Layard (1979: 485) conducted a similar study for Britain using a random sample of 7,000 employed males and found a similar rate of return (10 per cent). Despite the fact that various studies across the world produce different results in terms of the rate of return, the global average does not vary significantly from Mincer's (1974) study for the United States and Psacharopoulos and Layard's (1979: 485) study for the United Kingdom. There are several summaries of studies in the literature that include data on the global rate of return, with a wide range of surveys and a number of countries (with different levels of economic development). According to the findings of these summary studies, the average rate of return ranges between 8-10% (Patrinos and Psacharopoulos, 2020: 56; Psacharopoulos and Patrinos, 2018: 7; Montenegro and Patrinos, 2013: 7). The rate of return on education varies depending on gender, the degree of economic growth of the countries, the level of education, and a variety of other social or even political factors, according to the empirical literature. The majority of studies that have looked at gender differences in the rate of return have shown that females have a higher rate of return (Psacharopoulos and Patrinos, 2004: 346, 2018: 10; Montenegro and Patrinos, 2013: 17; Peet, Fink and Fawzi, 2015: 70, Patrinos, 2016: 4; Patrinos and Psacharopoulos 2020: 57). If we focus on the rate of return according to the economic development level of the countries, the ones leading are the developing countries, especially those in Sub-Saharan Africa and Latin America (Psacharopoulos and Patrinos, 2018: 452). Whereas, in terms of education levels, primary and tertiary education offer higher rates of return (Psacharopoulos and Patrinos, 2018: 11). "Latterly, we have an increase in the rate of return to tertiary education, as a result of rising demand for this level of education in the labour market" (Montenegro and Patrinos, 2013: 10). In some countries, the economic and political background can also have a substantial impact on the rate of return. For more about the theory of human capital, the costs and benefits of investing in M. Ukaj, A. Hoti, R. Topxhiu: Private Returns to Education in Developing Countries: Evidence from Kosovo 33. education, and some of the empirical results in countries with different levels of economic development, see Ukaj and Mustafa-Topxhiu (2020). 3 Data and descriptive statistics This research uses data from the Labour Force Survey (LFS) implemented by the Kosovo Agency of Statistics (KAS). LFS has been implemented since 2001, providing data on labour market categories in Kosovo. Since 2012, this survey has been conducted according to the Eurostat methodology, where each selected household is surveyed four times within 12 months (one survey and three re-surveys) (KAS, 2017: 2). LFS provides data on employees' wages as well as the highest amount of education achieved, acting as a valuable resource for determining the private return on investment in education. In this research, we use data for the period 2015-2017. During this period, 75,248 people were surveyed through LFS; 49,949 of them were of working age (15-64), of which 13,454 were employed, 6,542 unemployed, and the others were inactive. Out of the 13,454 employed, 9,300 are wage employees, and 9,100 of them (6,962 males and 2,138 females) have declared information on their wages, level of education and the number of working hours, the information necessary for this research. We use only the sample of wage employees (9,100 individuals), mainly for two reasons. First, the LFS results lack data on the wage variable for almost all non-wage employees (self-employed, family business employees, etc.), and second, as Montenegro and Patrinos (2014: 6) point out, "it is difficult for the self-employed to distinguish between the return to investment in education and the return to capital", therefore, as suggested, we excluded them from the sample. The assessment of the rate of return is done by using the hourly wage instead of the monthly or annual wage since the number of working hours differ significantly among persons with different levels of education. The hourly wage is found by comparing the monthly wage with the number of working hours declared by the employees. In the table of descriptive statistics (Table 1), we present some of the main characteristics of the sample which is used in this research. In the first part (I) of Table 1, we have presented the characteristics of age, gender, marital status and levels of education for all individuals involved in the LFS research, whereas in the second part (II), only the characteristics of individuals belonging to working age (16-64) are presented. Following the table of descriptive statistics (Table 1), the sample will be 34 LEXONOMICA. presented more extensively on issues like wages, working hours, thus relating education with some of the benefits in the labour market such as increasing participation in the labour force, increasing employment, reduction of overtime working hours, finding stable jobs, reduction of gender differences (in terms of labour force participation and employment), etc. About 94% of employees work full time, 70% work for pay (the rest are self- employed or work for their families). Nearly half of employees (42%) work more than 40 hours a week, while overtime compensation is almost non-existent. About 99% of overtime hours are not paid (result not reported). Table 1: Some descriptive statistics of the sample Variables Number of Observations Mean Standard Deviation Minimum Maximum Total I. Age 0-15 75248 0.243 0.428 0 1 18290 15-24 75248 0.194 0.396 0 1 14616 25-54 75248 0.386 0.487 0 1 29026 55-64 75248 0.084 0.277 0 1 6307 65+ 75248 0.093 0.290 0 1 7009 Male 75248 0.505 0.500 0 1 37993 Level of education No formal education 56886 0.053 0.224 0 1 3001 Primary 56886 0.415 0.493 0 1 23622 Secondary 56886 0.414 0.493 0 1 23543 Tertiary 56886 0.118 0.323 0 1 6720 II. Labour market indicators (15-64) Labour force participation 49949 0.400 0.490 0 1 19915 Employed 49949 0.268 0.443 0 1 13453 Full-time 13453 0.938 0.241 0 1 12623 Working hours (paid workers) <40 9300 0.072 0.259 0 1 670 <=40 9300 0.583 0.493 0 1 5422 >40 9300 0.417 0.493 0 1 3878 Monthly wage 9275 367.086 155.609 50 3000 3404725 Hourly wage 9100 2.231 1.210 0.260 26.041 20304.76 Wage between 175-1000 9275 0.983 0.131 0 1 9114 M. Ukaj, A. Hoti, R. Topxhiu: Private Returns to Education in Developing Countries: Evidence from Kosovo 35. The average net wage of wage employees is around 367.08 € per month. Males earn a higher average wage per month, 367.95 €, compared to females, 364.21 €, with a slight difference; however, when the wage is converted to working hours, females earn 10.6% more than males (2.40 / 2.17), as they, on average, work fewer working hours (see Table 2). Table 2: Working hours and average wages by sector Sector Average working hours Average monthly wage (€) Average hourly wage (€) Males Females Total Males Females Total Males Females Total (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Private 48.34 43.95 47.52 333.55 310.55 330.2 1.76 1.83 1.77 Public 41.84 36.14 40.56 421.58 410.26 418.24 2.80 2.88 2.83 Total 44.48 38.89 43.30 367.95 364.21 367.08 2.17 2.40 2.23 If we look at the relationship between education and labour market indicators, Table 3 shows a close relationship between them. As can be seen from column 4, labour force participation for individuals with no formal education is only 5.4%, while this percentage increases progressively as high as the level of education, reaching 70.3% for those with tertiary education. The same holds for employment; while the employment rate for employees with no formal education is only 2.5%, for those with higher education, it reaches 54.4%. Meanwhile, as expected, unemployment appears to be inversely linked to educational attainment (see Table 3, column 8). Table 3: The link between education and labour market indicators Level of education Working- age (15-64) Active population % Employed % Unemployed % Not active % (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) No formal education 1,192 65 5.4 30 2.5 35 53.8 1,127 94.5 Primary 19,533 4,023 20.5 2,380 12.1 1,643 40.8 15,510 79.4 Secondary 22,864 11,398 49.8 7,547 33.0 3,851 33.7 11,466 50.1 Tertiary 6,297 4,429 70.3 3,425 54.4 1,004 22.6 1865 29.6 Total 4,9883 19,915 40 13,382 26.8 6,533 32.8 29,968 60 The level of education (especially higher education) has a greater effect on females, both in terms of participation in the labour force and in terms of employment. Figure 1 shows that labour force participation, as well as female employment, increases by a higher percentage compared to males when moving from one level of education to another, especially with a transition from primary to secondary and tertiary 36 LEXONOMICA. education. In tertiary education, the difference with males is significantly reduced, especially in labour force participation. Figure 1: Gender differences in labour force participation and employment by level of education Besides closing the gender gap in terms of labour force participation and employment opportunities, as mentioned above, education also plays a significant role in reducing gender differences in labour market compensation. While males with primary education benefit about 40.2% more than females with the same level of education, in secondary and tertiary education, the gap is reduced to only 11.1 per cent and 6.3 per cent, respectively. Table 4: Average wage by the level of education and gender Levels of education Total Males Females Emplo yed Average wage Emp loyed Average wage Wage per hour Empl oyed Average wage Wage per hour (1) (2) (3) (4) (5) (6) (7) (8) (9) No formal education 8 234.37 7 228 1.17 1 275 1.43 Primary 1148 309.73 983 323.04 1.58 165 230.45 1.35 Secondary 5089 327.22 4182 333.154 1.84 907 299.86 1.78 Tertiary 3027 456 1942 466.28 3.17 1085 438.57 3.07 The findings also indicate that education plays an important role in reducing overtime hours. According to the findings of this research, employees with higher education in Kosovo work on average about 39 hours per week, significantly less than the average of employees with lower levels of education, where the average M. Ukaj, A. Hoti, R. Topxhiu: Private Returns to Education in Developing Countries: Evidence from Kosovo 37. working hours reach up to 50 hours per week, especially for those with no formal education. 4 Research method In this research, we apply Mincer's wage equation to assess private returns to investment in education. The model relates to the logarithm of wages with years of education, experience and experience squared: Ln (wi)= β0 +β1Si +β2Xi + β3Xi2+ µi When Ln (wi) is the natural wage logarithm, S is the number of years of schooling, Xi is the number of years of labour market experience, and μi is the error term that captures other unobserved variables that influence income and may also have an impact on the education decision (Montenegro and Patrinos, 2014: 4, 18). In this equation, the coefficient β1 estimates the average rate of return from one additional year of education. According to Montenegro and Patrions (2014: 4), in countries where the education variable is expressed in the form of the highest level of education completed, such as the case of Kosovo and many other developing countries, the model can be adapted by using dummy variables for education levels. The model takes the following form: Ln(wi) = β0βpDpi + βsDsi+ βtDti +β1Xi + β2Xi2+µi where D, presents dummy variables for education levels. The Ordinary Least Squares (OLS) approach is used to calculate the rate of return on investment in education. As noted by several researchers, the use of OLS may overestimate the rate of return, mainly because of innate ability impact (Blundell, Dearden and Sianesi, 2001: 6; Gunderson and Oreopoulos, 2010: 39). However, most studies that have looked into this problem have found that treating endogeneity with different natural instruments increases the rate of return as compared to the OLS, implying that the OLS does not overestimate but rather underestimates the rate of return (Walker and Zhu, 2001: 16-17; Psacharopoulos and 38 LEXONOMICA. Patrinos, 2004: 1; Montenegro and Patrinos, 2014: 18; Bingley, Christensen and Walker, 2005: 10; Ashenfelter and Kruger, 1994: 1164; Card, 2001: 36). According to Peet et al. (2015: 72), finding methods to eliminate the bias resulting from using OLS is not easy; therefore, estimating the rate of return using this method (OLS) remains a useful alternative. In addition, Harmon (2011: 1) points out that the use of different strategies for treating endogeneity in education has not had any significant success empirically. Ukaj and Mustafa-Topxhiu (2019: 197) accomplished a summary of studies that have dealt with the treatment of endogeneity using different strategies. This study finds a number of drawbacks that result from the use of these strategies. 5 Empirical findings The first part of the findings will focus on the private rate of return from education, reflected in one additional year of education, as well as gender gaps, job tenure, and rate of return by age group. The following are the rate of return evaluation results according to the levels of education, classifying them into three main categories (primary, secondary and tertiary). 5.1 Rate of return from one additional year of education Since the data for the education variable in LFS is expressed in the form of education levels, for calculating the rate of return on a year of additional schooling, the variable of years of education is adjusted by converting education levels into several years possessed by each level. For a similar approach to defining the variable of years of education, in situations where education is declared by levels, see, for example (Pischke Wachter, 2008: 10; Peet et al. 2015: 71; and Montenegro and Patrinos, 2014: 5). In addition, the adjustment of the variable of years of education has been made by taking into account the implications of the reforms in education that have occurred at all levels after 2000. Thus, for example, if the respondents answered that they had completed primary education, they have been converted into 8 or 9 years of completed education, depending on the period in which they had completed a certain level of education (i.e., depending on whether they were affected by the reform when the implementation of the ninth grade had begun or not). M. Ukaj, A. Hoti, R. Topxhiu: Private Returns to Education in Developing Countries: Evidence from Kosovo 39. Table 5 shows the results of three models with different specifications for calculating the rate of return from an additional year of schooling. In Model 1 (column 2), where the rate of return is calculated by keeping only job tenure constant, the results show that an additional year of education in Kosovo, on average, brings around 8.6% of additional income. If we control for gender and sector (as in Model 2, column 3), it can be noticed that the rate of return drops to 7.3%. Table 5: Rate of return from one additional year of education Explanatory variables Model 1 Model 2 Model 3 (coefficient) (coefficient) (coefficient) Years of education 0.086*** (0.002) 0.073*** (0.002) 0.082*** (0.002) Job tenure 0.033*** (0.001) 0.020*** (0.001) 0.033*** (0.001) Tenure2 -0.001*** (0.000) -0.000*** (0.000) -0.001*** (0.000) Male – 0.026*** (0.009) – Public sector – 0.260*** (0.009) – Female – – -0.261*** (0.048) Female * years of education[1] – – 0.019*** (0.004) Constant -0.628*** (0.020) -0.500*** (0.022) -0.570*** (0.022) Number of observations 9100 9095 9100 R2 0.371 0.420 0.373 Standard errors in parentheses “* p<0.05, ** p<0.01, *** p<0.001” To find the differences in the rate of return by gender, as Wooldridge (2016: 275) points out, it is necessary to include the interaction term between gender (female) and education variables in the model. In this case, the model takes the following form: Ln (wi)= α +β1Si +β2Xi + β3Xi2 + β4femra + β5 female*education + µi 40 LEXONOMICA. After including the interaction term (Model 3, column 4), it can be shown that males have an 8.2 per cent rate of return, while females have a rate of return that is 1.9 percentage points higher than males; in other words, females get a return rate of 10.1% (8.2 +1.9). Approximately the same results are achieved if the assessment is done with separate samples for males and females (results are not reported). After including the interaction term (Model 3, column 4), it can be shown that males have an 8.2 per cent rate of return, while females have a rate of return that is 1.9 percentage points higher than males; in other words, females get a return rate of 10.1% (8.2 +1.9). The higher rate of return for females is in line with the findings of a similar study conducted by Hoti (2011: 77) for Kosovo, but also with most studies in both developed and developing countries. Moreover, the difference in the rate of return between males and females is almost identical to the global gender difference presented in the studies of Montenegro and Patrinos (2013: 17) and Psacharopoulos and Patrinos (2018: 10), which provide data for more than 90% of the world countries. It should be noted that due to the nature of the data (in the absence of data, specifically, for the last year of completion of a certain level of education), it has been impossible to find the rate of return for the final year of a certain level of education, the year in which the rate of return is expected to be higher than previous years, due to the degree effect, a phenomenon known in the literature as the `sheepskin effect` (Harmon et al. 2003: 127). Consequently, in this study, all years of education are assumed to bring the same rate of return. Due to some difficulties in precisely identifying the education variable, reflected in the number of years completed, the possible effect of extremely high or low wages on the rate of return, and some other details, it was deemed appropriate to conduct some additional tests to ensure that the findings presented in table 5 were consistent. Tests were done by removing extreme wages (under 100 € and over 1500 €), professions in which determining the exact number of years was difficult, as well as estimating the rate of return expressed in euros (not through the log wage). It should be noted that even after these tests, the results do not differ almost at all from those presented in Table 5. M. Ukaj, A. Hoti, R. Topxhiu: Private Returns to Education in Developing Countries: Evidence from Kosovo 41. 5.1.1 Rate of return from job tenure Mincer (1974), in his model, in addition to education, also evaluates the role of experience on wages. Unfortunately, the Kosovo Labour Force Survey does not have data on employee experience. In such circumstances, in the absence of experience data, Mincer (1974) proposed potential experience, calculated as current age minus years of education minus age of starting schooling. On this occasion, Mincer assumed that the transition from school to work is immediate. However, in the case of Kosovo, because the transition from school to work can take years because of high unemployment, it is considered more appropriate to use the job tenure as a proxy for potential experience. According to the results presented in Table 5 (always using tenure as a proxy for experience), one year of additional work experience in Kosovo brought about 3.3% additional return, expressed in income per working hour, keeping fixed the years of education. If we use age instead of job tenure as a proxy for work experience, as some studies do (Hoti, 2011; Gjipali and Kristo, 2011: 34; Peet et al., 2015: 82), it turns out that one year of experience brings about 4% of additional income (result not reported). 5.1.2 Return to education by age group Findings in this research show that employees of the age group 15-24 years get a lower rate of return (7.1%) compared to the other two age groups, 25-54 and 55-64, in which the rate of return is around 8.7 % and 8.6%, respectively. The lower rate of return of the age group 15-24 is quite comprehensible, as, during this age, the first attempts to enter the labour market commence, where education is not considered in its full role. Furthermore, there may be a gap in the entry period into the labour market for people with different levels of education in this age group; as a result, the wage may be influenced more by work experience, i.e., passing the probationary process, than by education level. For example, employees with secondary education (who have not proceeded to higher education) potentially enter the labour market earlier; consequently, the wage difference between these employees and employees with higher education may be greatly influenced by the experience that the former already gained in the workplace (reference). The higher rate of return in the age groups 25-54 and 55-64, compared to the reference age group (15-24), seems to be 42 LEXONOMICA. influenced by the accumulation of experience. This can be confirmed when we check for the influence of age on wage, keeping the years of education fixed. To do this, a dummy variable is used for the three age groups mentioned above. The results show that, for the same level of education, employees in the age groups 25-54 and 55-64 earn, on average, about 20% more compared to the reference group (15-24). While the rate of return for males is positively correlated with age, surprisingly, quite the opposite happens for females. In all cases, the coefficients are statistically significant (see Figure 2). Figure 2: Rate of return by age group and gender The high rate of return for females in the 15-24 age group may be because females with higher skills are more likely to be working in high-paying employment in this age group; consequently, when these are compared to employed females who have less education, this creates a large difference in income per working hour. However, it is possible that the smaller number of observations in the 15-24 and 55-64 age groups will influence the rate of return of females. 5.2 Rate of return according to levels of education The results presented in Table 7 (column 2) show that employees with primary education get a rate of return that is 27.4% higher compared to employees with no formal education, while the rate increases to 41.1% and 88.6%, respectively, for employees with secondary or tertiary education. These results are achieved by keeping only job tenure fixed. Progressive increases in the rate of return as well as in the reliability of results (expressed in levels of significance) while transitioning M. Ukaj, A. Hoti, R. Topxhiu: Private Returns to Education in Developing Countries: Evidence from Kosovo 43. from one level of education to another suggest that continuing to invest in education constitutes good decision making. From column 3 of Table 7, it can be seen that the rate of return by the level of education does not differ much even after controlling for gender and sector. Whereas, in order to find the difference in the rate of return by gender, we use the interaction term between the variable of education and that of gender (female) just as we did when calculating the rate of return by one additional year of education (Table 5). The results presented in column 4 of Table 7, using the interaction term, suggest that females with secondary education get a rate of return which is 7.9 percentage points higher compared to males with the same level of education. The gender difference increases even more at the tertiary level, where females with this level of education get a rate of return that is 12.1 percentage points higher compared to males. Almost identical results are achieved even if the evaluation is done with separate samples for males and females (results are not reported). Table 7: Rate of return according to levels of education Explanatory variables Model 1 Model 2 Model 3 (coefficient) (coefficient) (coefficient) Primary education 0.274* (0.129) 0.255* (0.124) 0.257* (0.124) Secondary education 0.411** (0.128) 0.365** (0.124) 0.356** (0.124) Tertiary education 0.886*** (0.129) 0.777*** (0.124) 0.754*** (0.124) Job tenure 0.031*** (0.001) 0.019*** (0.001) 0.019*** (0.001) Tenure2 -0.001 (0.000) -0.000*** (0.000) -0.000*** (0.000) Public sector – 0.241*** (0.009) 0.242*** (0.009) Male – 0.041*** (0.009) – Female – – -0.132*** (0.03) Female * secondary education – – 0.079*** (0.033) Female* tertiary education – – 0.121*** (0.003) Constant -0.065 (0.128) -0.057 (0.124) 0.005*** (0.124) Number of observations 9100 9095 9095 R2 0.398 0.439 0.441 Standard errors in parentheses “* p<0.05, ** p<0.01, *** p<0.001” 44 LEXONOMICA. If we compare each level of education with the previous level (by setting the previous level in the constant), we can see a U shape, where primary and tertiary education bring a higher rate of return compared to the second one (see Figure 3). The difference is significantly more pronounced when comparing tertiary education with secondary education (47.3%). The U-shaped form, where secondary education brings a lower rate of return compared to the other two levels, is well known in the empirical literature (Heckman Lochner and Todd, 2003: 10; Psacharopoulos and Patrinos, 2018: 11; Peet et al., 2015:74). Figure 3: Rate of return from one level of education to another 6 Comparison of results with other developing countries In comparison to other countries in the region, Kosovo's rate of return from one year of additional education is not substantially different (see Table 8). The difference in the rate of return between these countries can be affected to some degree by the period (year) of the research. M. Ukaj, A. Hoti, R. Topxhiu: Private Returns to Education in Developing Countries: Evidence from Kosovo 45. Table 8: Rate of return compared to countries of the region Country Year of study Rate Bosnia and Herzegovina 2002 8.1 Bulgaria 2012 7.8 Romania 2012 10.3 Kosovo 2020 8.6 Montenegro – – Macedonia 2005 5.7 Serbia 2014 10 Albania 2012 8 Note: Data for countries of the region (excluding Kosovo) are taken from a review of studies by Psacharopoulos and Patrinos (2018) and Montenegro and Patrinos (2014). In countries where there is more than one study, the results of the last study are taken into account. If we compare the results with other developing countries, the results show significant differences depending on the country or region. Psacharopoulos and Patrinos (2018: 20-23) and Montenegro and Patrinos (2014: 20-36) summarize the most serious studies by different authors for the majority of developed and developing countries worldwide. According to the findings of these authors' studies, the rate of return is not always inversely interrelated to the degree of economic development of the countries, as noted by Perkins et al. (2013: 274) and Card (2001). In some of the developed countries like Germany, Singapore, South Korea, the USA, etc., the rate of return is, on average, over 12%, thus being higher than the world's average of about 10%. On the other hand, there are quite a few developing countries (even with low per capita income) with a rate of return that is slightly lower than the global average. Some of the developing countries where the return rate is between 2.2% and 7% are Armenia, Yemen, Cambodia, East Timor, Syria, Congo, Bangladesh, etc. However, it should be noted that the countries leading in terms of the highest rate of return globally are low-income countries, such as Ethiopia (18.5%), Uganda (16.4%), Burundi (17.3%), Rwanda (18.2%), Tanzania (16.4%), Zambia (14.9%) etc. Peet et al. (2015: 81) estimate the rate of return for 25 developing countries, using 61 household surveys for the period 1985-2012. In most of the countries included in the study, primary and tertiary education result in higher rates of return compared to secondary education. For the countries of the region, the results of this research show a higher rate of return to primary and tertiary education in countries such as Serbia and Bosnia, while in Albania, the rate of return is the lowest in primary 46 LEXONOMICA. education while the highest is in the tertiary one. The same pattern (where primary and tertiary education bring a higher rate of return) is also found in a review of studies conducted by Montenegro and Patrinos (2014: 12). According to the findings of this study, the highest rate of return at these two levels of education is observed in all regions of the world and both genders. This study also confirms that tertiary education brings the highest rate of return to education. Also, Psacharopoulos and Patrino (2018: 11) found that primary and tertiary education result in a higher rate of return. 7 Conclusion The research aimed to assess the reward that employees receive from education in the Kosovo labour market in terms of a wage premium. The findings presented in this research (Section 5) indicate that investing in education in Kosovo is a reasonable investment in terms of its impact on employee income. According to the results of this research, every additional year of education in Kosovo brings an average of 8.6% additional earnings from the labour market. When we look at the rate of return by the level of education, we can see that workers with a primary education earn 27.4 per cent more than those with no formal education, while the rate rises to 41.1 per cent and 88.6 per cent, respectively, for those with secondary or tertiary education. Tertiary education brings the highest rate of return compared to the other two levels (primary and secondary), even more than both levels combined. The average rate of return from one additional year of education is not far from the world's average, as well as the countries of the region. References Ashenfelter, O., Kruger, A. (1994) Estimates of Return to Schooling from a New Sample of Twins, The American Economic Review, 84(5), p. 1157-1173. Becker, Gary S. (1993) Human Capital: A Theoretical and Empirical Analysis with Special Reference to Education, Third Edition (Chicago: The University of Chicago Press). Bingley, P., Christensen, K., and Walker, I. (2005) Twin-Based Estimates of the Returns to Education: Evidence from the Population of Danish Twins. Unpublished paper. Blundell, R., Dearden, L., Sianesi, B. (2001) Estimating the Returns to Education: Models, Methods and Results (Centre for the Economics of Education London School of Economics). Borjas, G. (2013) Labor Economics, Sixth Edition (McGraw-Hill). Card, D. (2001) Estimating the Return to Schooling: Progress on Some Persistent Econometric Problems, Econometrica, 69(5), 1127-1160. M. Ukaj, A. Hoti, R. Topxhiu: Private Returns to Education in Developing Countries: Evidence from Kosovo 47. Gjipali A., Kristo S. (2011) Mincer Rates of Return to Education in the Emerging Economy of Albania, Studies in Business & Economics, 6(2). Gunderson, M., Oreopoulos, P. (2010) Returns to Education in Developed Countries: University of Toronto, ON, Canada. In Brewer, D. J. and McEwan P. J. (ed), Economics of Education, p. 37. Hampf, F., Wiederhold, S., and Woessmann, L. (2017) Skills, Earnings, and Employment: Exploring Causality in the Estimation of Returns to Skills, Large-scale Assessments in Education, 5(1), p. 1-30. Harmon, C. (2011) Economic Returns to Education: What We Know, What We Don't Know, and Where We Are Going - Some Brief Pointers, Economics of Education Review, 30(6). Harmon, C., Oosterbeek, H., and Walker, I. (2003) The Returns to Education: Microeconomics, Journal of Economic Surveys, 17(2), p. 115-156. Heckman, J., Lochner, L. and Todd, P. (2003) 'Fifty Years of Mincer Earnings Regressions.' NBER Working Paper No. 9732. Hoti, A. (2011) Returns to Education in Kosovo: Estimates of Wage and Employment Premia, South East European Journal of Economics and Business, 6(1), p. 71-84. Kosovo Agency of Statistics (KAS) (2017) Labor Force Survey. Kosovo Agency of Statistics (KAS) (2020): Statistical Yearbook of Kosovo. Mincer, J. (1974) Schooling, Experience and Earnings (New York: Columbia University Press for National Bureau of Economic Research). Montenegro, C. E., & Patrinos, H. A. (2013). Returns to Schooling Around the World, Background Paper for the World Development Report, 8258024-132095074719. Montenegro, C., Claudio and Patrinos, H. (2014) "Comparable Estimates of Returns to Schooling Around the World", World Bank Policy Research Working paper no. 7020, World Bank Group, Washington, DC. Patrinos, H. (2016) Estimating the Return to Schooling Using the Mincer Equation, IZA, World of Labor, Vol. 278. Patrinos, H. A., and Psacharopoulos, G. (2020) Returns to Education in Developing Countries, The Economics of Education (Academic Press). Peet, E. D., Fink, G., and Fawzi, W. (2015) Returns to Education in Developing Countries: Evidence from the Living Standards and Measurement Study Surveys, Economics of Education Review, 49, p. 69-90. Perkins, D., Radelet, S., Lindauer, D., Block, S. (2013) Economics of Development, Seventh Edition (New York: W. W. Norton & Company). Pischke, J. S., and Von Wachter, T. (2008) Zero Returns to Compulsory Schooling in Germany: Evidence and Interpretation, The Review of Economics and Statistics, 90(3), p. 592-598. Psacharopoulos, G. (2018) Benefits and Costs of the Education Targets for the Post-2015 Development Agenda. In B. Lomborg (ed.), Prioritizing Development: A Cost-Benefit Analysis of the United Nations' Sustainable Development Goals (Cambridge: Cambridge University Press), pp. 119-142. Psacharopoulos, G., and Layard, R. (1979) Human Capital and Earnings: British Evidence and a Critique, The Review of Economic Studies, 46(3), p. 485-503. Psacharopoulos, G., and Patrinos, H. A. (2018) Returns to Investment in Education: A Decennial Review of the Global Literature, Education Economics, 26, p.1-14. Psacharopoulos, G., Patrinos, H. (2004): Returns to Investment in Education: A Further Update, Education Economics, 12(2), p. 111-134. Schultz, T. W. (1961) Investment in Human Capital, The American Economic Review, 51(1), p. 1-17. Ukaj, M., and Mustafa Topxhiu R. (2019) The Returns to Investment in Education: Some Theoretical and Empirical Insights. Annals of the University Dunarea de Jos of Galati: Fascicle: I, Economics and Applied Informatics, 25(1), p. 193-203. Walker, I., and Zhu, Y. (2001) The Returns to Education: Evidence from the Labor Force Surveys, Research Report RR 313 (Department of Economics: University of Warwick). 48 LEXONOMICA. Woodhall, M., Hernes, G., and Beeby, C. E. (2004) Cost-Benefit Analysis in Educational Planning, Fundamentals of Educational Planning, 80. Wooldridge, J. M. (2016) Introductory Econometrics: A Modern Approach (South-Western, Cengage Learning).