The positive relationship between education of a worker and his/her earnings and productivity has been witnessed in many countries (Schultz, 1988). Microeconomic empirical studies show that if education enhances the productivity and earnings of labor, government and families will increase the amount of money to invest in children’s education.

Vietnam’s achievements in education are far in advance of most countries at a similar development level (ADB, 1998). Economic reforms initiated in 1986 which translated into high economic growth, stabilized rate of inflation, dramatic decline of rate of poverty and the like, have also directly affected the education sector in terms of state expenditure on education, structure of financing and administration and ultimately changes in enrollment rate. Employment and the labor market of Vietnam in this period have also undergone significant changes owing to the Doi moi policy such as the retrenchment in state sector, the development of private sector, changes in salary and wage system, etc.

The education attainment of worker, their participation in the labor market and their earnings are basically well related. As the economy is moving toward a more market orientation, the impact of education on earnings is expected to be more significant. In addition, educational attainment and earnings inequality still prevails between men and women, rural and urban areas and also between the public and private sector. These issues require further understanding and analysis of education - earnings relationship.

1.1 Research objective

This study aims to examine the main features of reforms in education and employment sectors, from that to analyze changes in educational impacts on earnings during the transition period, to identify rates. It also aims to identify differences of this educational impact between different geographical regions, economic sectors and gender. To fulfill these objectives, rates of return to education are used as an instrument to measure educational impacts on earnings.

1.2 Conceptual limits and scope of the research

The research is confronted with certain conceptual limitations. First, wage earners aged over 14 years are the only targeted group. Second, only the private rate of return to education will be investigated. Finally, we will focus on the impact of education on earnings or monetary benefits of education.

1.3 Significance of the research

Experiences from other developing countries and countries in transition raise the need to evaluate education-earnings relation in Vietnam where little research has been undertaken, except the World Bank’s study in 1998. There is also an urgent need for a comprehensive and comparative study on education’s impacts on earnings in terms of policy implications. In addition, given the availability of the two national household surveys it is possible for a comparison in the rates of return to education between two periods of time (1993-1998).

1.4 Research questions and hypothesis

The following sub-questions will be examined in this research:

The objective of this chapter is to provide theoretical arguments and framework to examine the effects of education/training on earnings. The theory presentation is to work out the most proper methods to assess impact of education on earnings the in case of Vietnam. The impacts of education will be examined in some developing or in transition countries and also Vietnam.

2.1 Key definitions

Human Capital  refers to a different kind of capital such as schooling, a training course etc in the sense that they improve health, raise earnings or add to a person’s appreciation of literature over much of his or her lifetime (Becker, 1993).
Rate of Return refers to the compensatory nature of earnings on investment in human capital (Eatwell et al, 1991). Income must compensate for the cost and effort required to practice this occupation or profession.

2.2 The theory of human capital

The general analysis of investment in human capital was written by Becker from his first edition of “Human capital” in 1964. On-the-job training is thoroughly presented because it illustrates the effect of human capital on earnings. Suppose that a firm hires employees for a specified time period. The firm with profit-maximizing target would be in equilibrium when marginal products of worker equals wages. However, over n periods, expenditures and receipts during all periods would be interrelated and the present value of expenditures and receipts would be equalized as bellow equation where E_t and R_t represent expenditures and receipts during period t, i presents the market discount rate.

The equilibrium means that the present value of marginal products stream would have to equal the present value of the wage stream. Another implication from this theory is that marginal productivity would greater of less than wage is the cost of training is greater of less than the returns from training.

A school can be defined as an institution specializing in the production of training. A student may work while he is in school. The difference between earnings that could be got and the amount that he actually earns represents an important indirect cost of schooling. Meanwhile, direct cost of schooling clearly includes tuition, fees, books, and the alike. Net earning can be defined as the difference between actual earnings and direct school costs. Net earnings can also be presented as the difference between potential earnings and total costs. Schooling, therefore, would have implications as general on-the-job training. Schooling would steepen the age-earning profile, introduce a negative relation between the permanent and current earnings of young persons, and implicitly provide for depreciation on its capital.

Basing on the human capital theory, he next sections will illustrate methods to identify the relationship between earnings and education.

2.3 Cost and Benefit Analysis of education investment (Net Present Value method)

Cost-benefit analysis (CBA) is a technique by which the factors of costs, benefits, time scale and uncertainty or risks can be compared systematically for the purpose of evaluating the profitability of any proposed investment. In most cases of applying CBA in education, calculations have been based on the internal rate of return. If the costs are C a year and the investment is expected to yield benefits of B a year over n years then the rate of return (r) is the rate of interest at which the sum of present value of costs C_t / (1+r)^t  from year 0 to year n is exactly equal in value to the discounted sum of benefits B_t / (1+r)^t from year 0 to year n.
 
The crucial task in any CBA is to identify and measure the costs and benefits. Economic analysis of investment in education tries to measure the total cost including opportunity cost. The wages and salaries that students must forgo in order to enroll in education often represent a large part of private cost to education. Direct cost includes tuition fees and expenditure on book in general. In measuring the benefits of education, education yields direct and indirect benefits to both individual and society. The direct benefit of education for individual is higher lifetime earnings and for society it is the higher productivity of educated workers and the additional contribution to national income over their entire working lives. Education also brings indirect benefits which are often called externalities which will not be focused on in this study.

2.4 Age-earnings profile method

Age-earnings profile method is regarded as “elaborate method” as it works with detailed age-earnings profile by level of education (Psacharopouslos, 1993) to find out the discount rate. The annual stream of benefits is typically measured by the earnings advantage of a graduate of the educational level in comparison with the earnings of control group of lower level graduates. The stream of costs consists of the forgone earnings of the individual while in school measured by the mean earnings of the educational level that serves as control group.  
 
Where (W_u - W_s) is the earnings differential between a university graduate (subscript u) and a secondary school graduate (subscript s, the control group). C_u represents the direct cost of university education (tuition and fees, books, etc…) and W_s denotes the student's forgone earnings or indirect costs. A similar calculation can be made for other levels of education. However, primary school children, mostly aged 6 to 12 years, do not forgo earnings during the entire length schooling, but only for one or two years.

2.5 Human Capital Earning Function method

The “human capital earnings function” (HCEF) with schooling version was first developed in Chiswick and Becker (1966) and extended to include on-the-job training in Mincer (1974). 
2.5.1 Derivation of and interpreting the coefficient

Chiswick (1966) made an interpretation of coefficient of schooling in the HCEF by the deriving the equation: LnE_s  = LnE₀  + (r₀K₀) S where E₀ = Earnings if there is no schooling, E_s  = Earnings received each year after obtaining s years of schooling, r_t = Rate of return on investments in year t of schooling, K_t = C_t/E_t-1 = Investment in year of schooling t, relative to a full-year’s potential earnings if investment was not made in this year of schooling
Estimating the regression equation, the regression coefficient of schooling is estimated as b for example and K is known prior, the rate of return to schooling is: r = b/K. If we make an assumption that investment in one year of schooling equals to full-year potential earnings or K=1, the coefficient b is the rate of return to one year of schooling. The value of K need not be constant across schooling levels. Let denote the number of schooling years at each corresponding level as P for primary. M for middle or secondary and H for tertiary or higher education, then:  LnE_s  = LnE₀  + (r_pK_p) P + (r_mK_m) M + (r_hK_h) H            
2.5.2      Mincerian earnings functions

The simple version of human capital earning function was further extended by Mincer (see Mincer, 1974) to include on-the-job training. A great part of the term “on-the-job training” is associated with learning from experience. Mincer (1974) has shown that in the United States earnings are more closely correlated with years of working experience. Moreover, Mincer (1974) developed the rationale for the standard quadratic form for the experience variable EXP². Measures of employment or labor supply (number of working hours per week or number of working weeks per year) can also be incorporated into the analysis in logarithm form.
Basic earnings function
lnY_i = a + bS_i + g₁ EXP_i + g₂ EXP_i² + l lnH_i  + ε_i  as where i refers tp individual i.

Mincerian earnings functions involve in fitting the log-wages (lnY) using years of schooling (S), adding years of labor market experience (EXP) and its square (EXP²) and log-time worked (lnH_i)) as independent variables. The coefficient on years of schooling can be interpreted as the average private rate of return to one additional year of education in general.
Extended earnings functions
lnYi = a + b₁PRIM_i + b₂SEC_i  + b₃VOC_i  + b₄UNIV_i + g₁ EXP_i + g₂ EXP_i² + l lnH_i + ε_i
PRIM_i , SEC_i  , VOC_i  , UNIV_i are dummy variables for each level of schooling completed by individual i.: primary, secondary academic, secondary vocational and university education After fitting this extended earnings function, the private rates of return to different levels of schooling can be derived by comparing adjacent dummy variable coefficients:
r_(PRIM)  =  b₁ / S_PRIM                       r_(VOC)  =  (b₃ - b₁) / (S_VOC  – S_PRIM)
r_(SEC)  =  (b₂ - b₁) / (S_SEC – S_PRIM)       r_(UNIV)  =  (b₄ - b₂) / (S_UNIV - S_SEC)

where S_PRIM , S_SEC, S_VOC and S_UNIV represent for the total numbers of schooling years for corresponding four levels of education. Again, care should be taken regarding the forgone earnings of primary school-aged children, therefore we divide b₁ by only 1 or 2.

2.6 Other methods and method selection

There are another methods to estimate returns to education such as: Short-cut method and Reverse cost-benefit method.  However, HCEF method is preferred in this study because this method can provide several desirable features such as requiring less data, flexibility to include compensatory variables, facilitating comparison across space from countries to countries etc.

2.7 Empirical analysis

This section is to review empirical studies on the impact of education on earnings in some developing, transitional economies and also in Vietnam. Special attention is paid to those studies, which have estimated the rate of return based on Mincerian earnings function method.

2.7.1 Empirical studies on impact of schooling and education levels to earnings

Schooling is an important determinant of wage and earnings. Mincerian rate of return to year of schooling is estimated at 11.2 percent for low-income countries. Van der Gaag and Vijverberg (1989) showed returns to an additional year of schooling was very high in Cote d’Ivoire, 20 percent. The impact of education on wage may differ by schooling level. Among the main three levels of education, primary education exhibits higher rate of return to secondary or higher education (Psacharopouslos, 1991). There are several studies revealing findings supporting the above pattern: King (1990) in Peru, (Stelcner et al, 1987) in Latin America. However, the world pattern is not always the case in some countries. During 1970-80 period in Brazil, the lowest private rates emerged for primary education, and the highest for higher education (Tannen and Michael, 1991).

Doubts have been repeatedly raised, regarding the economic profitability of vocational education. Examining the world pattern, Psacharopouslos (1993) confirmed a quite counter-intuitive finding that returns to the academic/general school track are higher than the vocational track: 11.7 percent versus 10.5 percent. This pattern is supported by study of King (1990) on Peruvian women’s earnings. 8 percent as compared with 5 percent.

2.7.2 Empirical studies on rates of return to education over time

Overtime evidence of short-term changes in returns to education associated with the process of economic transition seems to be showing that successful reform will eventually lead to higher returns. Orazem and Vodovipec (1995) examined returns to education from 1987 to 1991 in Slovenia showing that average returns to years of education have risen dramatically following transition. A similar result was found by Flanagan (1993) in Czech Republic.

There are also evidences of long-term changes in rate of return to education. Evidence in Hungary indicates that the private rate of return almost doubled in secondary education between 1971 and 1993 (Varga, 1995). Lachler (1996) in his study on education and earnings inequality in Mexico shows that the overall rate of return to education had increased as a direct consequence of the increased dispersion in wages that took over 1984-94 period.

In contrast, there are several studies that reveal findings of long-term changes in rate of return to education which are opposite to the above increasing pattern. Over 12 year period where private returns to years of schooling have declined 2 percentage points on average (estimation by Mincerian method) (Psacharopouslos, 1993). Examining changes in educational rates of return in South Korea during the 1970s and 1980s, it is suggested that the payoff to lower schooling levels declined substantially in absolute terms (Ryoo and Jai-Kyung, 1993).

2.7.3      Empirical studies on gender disparity in rate of return to education

Overall, most of empirical studies show that the returns to female education are higher than those for males (Psacharopouslos, 1990 and 1993), about 12.4 and 11.1 percent. This is similar to the pattern found in China in the mid 1980s (Jaminson and Van der Gaag, 1987), in Latin American countries (Psacharopouslos and Tzannatos, 1992) and in Peru where return for men are lower about 4 to 10 percent than returns for women (King, 1990).

However, Van der Gaag and Vijverberg (1989) found that there was no discrepancy in return to year of schooling or training between gender groups in Cote d’Ivoire. The evidence on returns to education difference between men or women at individual levels of education is more mixed. Psacharopouslos applied Mincerian method in 1993 study found that at primary level and higher education, returns for women are lower than for men: 13 percent versus 20 percent at primary and 12.7 percent versus 13.4 percent at higher education. The opposition happens at secondary level where return for women is greater than for men.

2.7.4      Empirical studies on regional disparity on rates of return to education

Holding other factors constant, place of residence and working determines significantly the earnings. Estimating earnings functions for women in Nicaragua, Behrman et al (1985) found relatively high rates of return to schooling 13 and 12 percent for women in the central metropolis of Nicaragua and other urban areas, in contrast with 5 percent for the rural samples. Knight and Lina (1993) find greater returns to education in the less institutionalized, rural labor market than in the urban labor market in China.

2.7.5      Empirical studies on economic sector of employment disparity

Mincerian approach is based in the assumption that wage are set equal to the marginal productivity of the wage earners. However, non-competitive market forces may influence the wage structure in many ways such as minimum wage legislation, government’s distributional and other political policies, etc, especially in developing countries.

Most of the existing findings show mixed results of wage differentials between the public and the private sector. Psacharopouslos (1985) reports world pattern of difference in returns to schooling between the “competitive private” sector and the “non-competitive public”: 13 and 10 percent respectively. Brown (1992) finds similar results in Estonia. These studies estimated wage differentials by using one or more dummy variables to indicate the sector. However, in Stelcner’s study, using switching regression model, the entire model is estimated by full information maximum likelihood (FIML). The results showed that secondary education yields higher return in the public sector: 9.4 percent per year as compared to 3.8 percent in private sector. Post-secondary education yields 2 percent more per year in the private sector. Applying a similar approach, Van der Gaag and Vijverberg (1989) found that years of elementary schooling show no effect in both sectors.

2.7.6      Empirical studies on impact of education on earnings in Vietnam

Since the initiation of renovation policy known as “Doi moi” in 1986, little specific work has been done on the impact of education on earnings. The only two studies, which examined the relation between education and earnings in Vietnam, were those of the World Bank (1996) and Moock et al (1998). The later was based and developed from the former while estimation results and implications remain the same. The results estimated by Mincerian method show that on average, the estimates rates of returns for Vietnam are still low compared with the returns to education estimated for other developing countries. Return to one additional year of schooling is 5 percent in average, 13 percent per year to primary education, 11 percent per year to university education. Secondary and vocational education earns lower returns of 4 to 5 percent per year. By fitting earnings function separately to different sub-groups, these studies found disparities in rate of return to education between male and female, between workers in public and private sector, between regions and between young and experienced wage earners. However, in these studies, it is not clear whether the above disparities were tested for their statistic significance. There are few points to be noted from these studies. First, education level was not strictly defined as level completed with diploma. Second, only post secondary vocational training was considered in these studies. There are several other studies, which also mentioned the return to education. However, rates of return to education are not specifically calculated.

3.1       Economic reform in Vietnam from early 1990s

In 1986, Vietnam launched an economic reform process aimed at renovation and open door policy. The initial success was very impressive (ADB, 1998) as some major indicators in table 3-1. By 1998, Vietnam’s GDP per capita had increased to $350 (1998 US dollars), (World Bank, 2000). Vietnam has been one of the fastest growing economies in the world with an impressive real GDP growth of 8.4 percent per annum on average. Inflation rate was strictly under control at about 6 percent in 1996 and 8 percent in 1998 (IMF, 1999). The foreign investment inflow rose dramatically in 1990s at $2 billion per year during 1995-97 period. This rapid growth also led to a dramatic decline in poverty, from 58 percent in 1992-93 to 37 percent in 1997-98 (World Bank, 1999).

Table 3-1: Major economic indicators of Vietnam in 1993-98



Sectoral structure of the economy also changed in 1990s. Vietnam remained primarily an agricultural economy although the share of agriculture was reducing from 40 percent in 1991 to 24 percent in 2000 and the share of service and industrial sectors rose. During the first decade of the reform process, Vietnam reduced the number of state-owned enterprises (SOEs) by a half from 12,000 to 6,000 enterprises (ADB, 1998; Koch, 2001). Nevertheless, non-state sector have begun to show their positive effect with share of employment increasing from 85 percent in 1986 to 91.3 percent in 1999 (IMF, 1999) with annual growth at over 3 percent.
3.2      Education sector
3.2.1.   Structure of education system

By 1989, the twelve-year system of general education from primary to upper secondary was fully implemented countrywide. Crèche (nursery) and kindergarten for children from 3 years old to the age of primary school are not compulsory. Primary education which includes grade 1 to grade 5 was made compulsory and remained fully subsidized by the government. Primary education has been universalized all over the country except some remote areas. The diversification in curriculum starts after lower secondary level where students can select four types of training or education. Tertiary education in Vietnam includes college and university. Most of the tertiary students enroll in 65 public universities and some other private or semi public universities (World Bank, 1996).

In regard to the administration of education, the education administrative authority operates at three levels: central, provincial and district levels. Despite this division of responsibilities, the education system in Vietnam has generally remained centralized in many aspects.

3.2.2.   Enrollment achievement

 The overall picture of education sector in Vietnam is very bright over the past years with a lot of impressive progress in expanding access to basic education (Nga, 2000). In early 1990s, adults literacy rate of Vietnam was already high 92 percent for men and 84 percent for women (ADB, 1998). The highest literacy rate for adult from 15 years old was recorded highest at 95 percent in 1998. These numbers are higher than those in regional countries.

Over five-year period 1993-98, net enrolment rate significantly increased at all levels, especially at higher education levels: 4 times at upper secondary (from 7 percent to 28 percent) and 3 times at post-secondary level (from 3 percent to over 9 percent) after a fall in 1989-92.

Figure 3‑1: School enrollment trend in Vietnam, 1986-98
 

Sources: Ministry of Education and Training
Table 3‑2: Disparities in school enrollment in Vietnam, 1993-98
    Considering gender difference in net enrollment, it is found that the disparities in access to education between males and females was not very significant at primary and secondary levels (see table 3-2). In general, access to education at all levels was lower in rural than in urban areas and the enrollment gaps in 1998 were smaller than that in 1993.

3.2.3.   Education provision and the rise of private sector

One of the critical reforms designed in education sector to promote government education targets was the development of non-public school. Three main types of non-public school are classified as follows according to World Bank report (1996): Semi-public schools, People/community founded school and Private school. In principle, non-public schools are encouraged to operate at all levels, but only legalized at pre-primary and vocational levels. Private university was not allowed until 1993. However, a limited number of students have enrolled in these non-public schools, only 5 percent in total. Data from two surveys VLSS 1992-93 and 1997-98 revealed that share of private sector in providing school service increased at upper secondary education, from 2 percent in 1993 to 5 percent in 1998 (Nga, 2000).

Evidently, we can find that the education system in Vietnam has been more deeply decentralized by the participation of private sector. Privatization can be implemented in two aspects: Private provision and private financing. In provision aspect, policy instrument of privatization has encouraged non-public providers to play a larger role in education/ training.

3.2.4.   Education financing and cost recovery policy

There is diversity in resource mobilization either from Government subsidies, from cost recovery in public education or from private sector. User fees have been applied in education service which is provided by Government, private sector was also liberalized in 1989 (World Bank, 1996).

Public sector spending
The role of public financing as a whole has decreased from 52 percent in 1993 to 50 percent in 1998. However, the proportion education expenditure in GDP and in total public expenditure has been substantially increased of 6 percent over the period 1992-98. Comparing the proportion of public expenditure on education of GDP in Vietnam with other countries, Vietnam’s proportion is a little lower. The allocation of public spending for education is also different between levels. Public spending on primary education had dramatically increased relatively to other education levels, increasing four folds from 1.14 trillion dong in 1993 to 4.6 trillion dong in 1998. Changes in policy of budget allocation strongly affect the cost of education that students or their families must incur.

Private education spending
Estimation from VLSS 1992-93 and 1997-98 shows that private sector financed almost 50 percent of overall education expenses in both time periods. However, private spending was larger than public subsidies at all levels except for primary education. On one hand, it means that education sector has been successful in mobilizing a considerable volume of private source to finance education. On the other hand, it implies that private costs already play an important role in rationing access to school.

Cost recovery policies
In transition to market–oriented economy, cost-recovery policy is allowed in public schools. Basing on estimation by Nga (2000) from VLSS 1993 and VLSS 1998, it is suggested that private expenditure on education has increased in triple in real term (all converted to January 1998 price) over 5 years. Schooling fee, a main part of cost recovery, it is different among education levels and expenditure groups. In overall, private expenditure paid by households accounts for about 43 percent of total education expenditure, varying at different levels, 12 and 19 percent for vocational and university education, 48 percent, 59 percent and 62 percent for primary, lower and upper secondary levels respectively. It reflects the financing education policy in Vietnam as similar as those in other developing countries favoring the rich.

3.3       Employment and labor market
3.3.1. Employment situation and structural change in labor market

Total employment of Vietnam increased rapidly from 32.7 million in 1993 to 38.2 million persons in 1998 (IMF, 1999; ADB 1999) in which about 5 millions employment increase was in this non-state sector. Among wage earners, almost 3 million persons are working at the state sector. Throughout the period of economic reforms, economic forces led to structural change in Vietnam’s dual labor markets in terms of ownership. There was shrinkage of state sector in its employment share partly due to SOE reform with a large retrenchment of up to 1.5 million workers. However, there was no huge rise in unemployment as more than 5 million jobs from 1993 to 1998 was created in non-state sector. Structure changes also happened in rural-urban labor market. It is estimated that unemployment rate in rural is about under 3 percent. High urban growth rates are attracting rural workers to the cities.

3.3.2.   Major reforms in employment and labor market

In 1990, the government of Vietnam began the process of dismantling the old public wage structure in which salaries and wages of state employees would no longer base on seniority, length of services and lifetime employment would no longer be guaranteed. In 1993, the government passed two regulations 25/CP and 26/CP dated 23 May 1993 specifying the exact “basic wage” to be paid to all state sector employees calculated on a multiple of the minimum wage rate (McCarty, 1999). At the same time, the government also announced a plan of monetizing wage, fringe benefits of all employees. Real income and employment in Vietnam, as a result, was negatively affected in several ways by country economic reforms. However, thanks to dramatic economic growth in later years of reforms, real wage was gradually improved.

Government also took initiatives to develop non-state sector including cooperative and private sector. This sector has shown positive effects with increasing share of employment. Law promulgated and regulations implemented like Law on Enterprise, privatization policies provided impetus for private sector to emerge. With the emergence of private sector, wages are more freely set without interference from government, especially in private enterprises.

In private sector, although government also takes an administration management by setting minimum wage, wages are more freely negotiated. Minimum wage in state-owned enterprises was 210,000 dong per month from early 2001.

Labor mobility is also an issue raised by economic reform process. In reform period, labor mobility from rural to urban areas is not officially prohibited but working permit is still required in some cases (McCarty, 1999). The social security system used to be a barrier to labor mobility, as there was no nation-wide coverage. Since 1995, as the social insurance system was established nation wide allowing employees to maintain their social insurance even when their job is shifted from public to private sector and thus further increased labor mobility among sectors and regions.

In this chapter, we analyze the impact of education on earnings using data from the VLSS 1992-93 and VLSS 1997-98. It addresses several important questions. First, do years of schooling and schooling levels significantly determine earnings? Second, is the impact of education on earnings changed as the economy becomes more market-oriented? Third, is there any disparity in earnings and in the impacts of education on earnings between regions, sectors of employment and gender? To analyze these issues, first, descriptive analysis is used and then two regression models are estimated.

4.1      Data set

The data are drawn from the two Vietnam Living Standard Measurement Surveys (VLSS) in 1992-93 and 1997-98. Our sample is limited only to wage earners aged from 14 years having job as their primary and main activity during the past 7 days. As a result, the size of the sample from VLSS 1992-93 is 2,245 and from VLSS 1997-98 is 3,179. Concerning education, the highest education level completed by an individual is determined if he/she completed the last grade and obtained a diploma of that level.

4.2      Variables definition and measurement

Dependent variable – Logarithm of monthly earnings. Monthly earnings are the total of wage or salaries including value in cash or in kind payment, plus other income, excluding contributions to pension, health insurance. Monthly earnings is adjusted all to January 1998 price and for regional price deflators.

Independent variables
Years of schooling (Yearsch). It includes years of general/academic schooling completed plus years of vocational training if any.

Education level completed (Level). One is classified as completing an education level only if he/she has finished the last grade at that level and received a level diploma. Each education level will then be separated as dummy variables. However, certificate and diploma of these education levels are cumulative.

Vocational training (Voc). This is also a dummy variable. Vocational training is then divided into 4 sub-variables as in table 4-1.

Years of experience (Exper). In the absence of direct information on work histories, year of experience is defined as exper = age – yearsch – 6. If exper is negative, the value is 0.

Other variables are defined in table 4-1.

Table 4‑2: Variable definition and measurement
Variables                                            Definition and measurement

Dependent variable

Logarithm of monthly earnings   Logarithm of monthly earnings in VND of wage earners

Independent variables

Years of schooling                                Years of academic and vocational schooling

Education level                                   Education level completed with diploma

Level 2                                     Equal 1 if individual graduated from primary school, 0 otherwise

Level 3                                     Equal 1 if individual graduated from L-secondary school, 0 otherwise

Level 4                                     Equal 1 if individual graduated from U-secondary school, 0 otherwise

Level 5                                     Equal 1 if individual graduated from university, 0 otherwise

Vocational training                             Equal 1 if individual took vocational training, 0 otherwise

Voc 2                                                   Equal 1 if individual took vocational training after primary education,

0 otherwise

Voc 3                                                   Equal 1 if individual took vocational training after lower secondary education, 0 otherwise

Voc 4                                                   Equal 1 if individual took vocational training after upper secondary education, 0 otherwise

Years of experience                              Years of working experience (= age – years of schooling – 6)

Experience square                                Quadratic form of years of experience

Logarithm of working hours                  Logarithm of number of working hours per week

Gender                                                 Equal 1 if male, 0 if female

Sector                                                  Equal 1 if public sector, 0 if private sector

Urban                                                   Equal 1 if in urban areas, 0 if in rural areas

North                                                   Equal 1 if in the north, 0 if in the south

4.6               
4.3      Descriptive analysis and hypothesis
Descriptive statistics for major characteristics of wage earners indicate some changes between the two surveys. First, education status is generally improved, mean years of schooling increased by almost one year. The proportion of low education level graduates reduced while proportion of workers with higher education levels increased: by 2 percent for upper secondary and by 3 percents for university and college level. Second, mean of monthly earnings has risen by VND 221,000 or 64 percent in real terms (January 1998 price).

Earnings difference by gender
There is a typical feature that the mean earnings of males is often higher than that of females in both surveys, 20-30 percent higher in 1992-93 and 50 percent at university level in 1997-98 survey. Moreover, there was little disparity in mean earnings between different education levels in 1992-93 survey. Workers, either men or women without education or with primary, secondary education receive quite similar earnings. This phenomenon in 1992-93 survey may support our previous suggestion that education had a greater impact on earnings in 1997-98 rather than before.

Earnings difference by economic sectors
Disparity in mean earnings between workers of different economic sectors is not consistent in two surveys. In the first survey 1992-93, the average monthly earnings in public sector seem to be higher than in private sector, though not much. However, in the second survey 1997-98, private sector workers usually earn more than their public counterparts on average, especially at university and college level. There is also a similar pattern in earnings difference where the mean of monthly earnings in 1992-93 survey is quite similar to workers with different levels of education. In contrast, there is a significant change in survey 1997-98, workers with higher education levels earn much larger amount of monthly earnings on average than less educated workers. Between upper secondary and lower secondary, the difference is 25 percent higher in private sector and 18 percent higher in public sector. University graduates earn a striking 150 percent higher than upper secondary graduates in private sector, and corresponding 50 percent in public sector. This may imply a changing impact of education on earnings in two surveys.

Earnings difference by region
Earnings are apparently higher in urban areas than rural area. The gap is more considerable in 1997-98 survey, especially at university level with 26 percent. Data also reveals a similar comment that education seems to strongly impact earnings in later survey than the former.

4.4      Econometric specification

Selecting good econometric regression models involves in the extending the basis Mincerian model to include other relevant variables such as gender, economic sector, urban/rural and north/south (for the particular condition of Vietnam) location into the models.  In addition, we also test to include one more dummy variable “young” for the argument that studying two different groups aims may find unequal effects of education to earnings of young persons who entered directly into reformed economy and more free labor market and to the ones who came into labor force when the economy was still more centrally planned.

However, each variable added to the model needs careful examination on its significance. Likelihood ratio (LR) test will be used here to evaluate the significance of adding variables. We start with the full model of all potential variables (lnwkh, gender, sector, urban and north) because the literature has indicated such a difference in earnings and rates of return between men and women, rural and urban workers, public and private economic sector (see chapter 2). Descriptive analysis in section 4.3 also reaffirms the possible influence of these above variables on earnings and raises some hypotheses of disparity which need to be tested.

From full or unrestricted model above, one by one variable is then dropped to test its significance. After these tests, we run two regressions originating from Mincerian earnings functions separately for each survey 1992-93 and 1997-98.

The first is basic model: Lnearn_i = a + b yearsch_i + g₁exper_i + g₂expersq_i + g₃lnwkhour_i + l₁gender_i + l₂sector_i + l₃urban_i + l₄north_i + u_i
The second is extended model: Lnearn_i = a + b₁exper_i + b₂expersq_i + b₃lnwkhour_I + g₁Level2_i + g₂Level3_i + g₃ Level4_i + g₄Level5_i + g₅Voca_i + l₁gender_i + l₂sector_i + l₃urban_i + l₄north_i + u_I
In addition, we need test for change in rate of return to education between 1992-93 and 1997-98, other tests have also been done to find out whether there is a significant disparity in returns to education for different groups of gender, economic sectors and regions.

4.5      Discussion of results
4.5.1      Rate of return estimation

investment in schooling equals to full year potential earnings, the coefficient of variable (yearsch) can be interpreted as rate of return to one year of education in general which are 2.7 percent in 1992-93 survey and nearly 4 percent in 1997-98 survey.

Lnearn_i = 3.25 + 0.027yearsch_i + 0.03exper_i - 0.001expersq_i + 0.44lnwkhour_i + 0.26gender_i - 0.3sector_i + 0.14urban_i –0.33north_i            (1992-93 survey)
Lnearn_i = 3 + 0.039yearsch_i + 0.02exper_i - 0.001expersq_i + 0.66lnwkhour_i + 0.19gender_i - 0.65sector_i + 0.22urban_i –0.32north_i               (1997-98 survey)
Coefficients of yearsch are positive and statically significant at 1 percent level implying a strong positive impact of schooling on monthly earnings and thus evidently prove the first hypothesis of this study to be true. However, this rate is very low when comparing with average rate of return to additional year of schooling over the world which is about 10 percent in general, but it is similar to that in other transitional economies: 5 percent in China in 1985 and 2.9 percent in Poland in 1986 (Psacharopouslos, 1993). 

We can obtain the rate of return to one year of schooling at each education level simply by dividing the estimated coefficients of education level variables the official number of years of schooling at corresponding levels: lower secondary (4), upper secondary (3), university/college (4), by the following formulas.

r _primary = g₁ /1                      r _{lower secondary} = g₂ / 4
r _{upper secondary} = g₃ /3             r _university = g₄ /4               
Estimation results indicate that most of coefficients of education level variables are positive and statistically significant at 1 or 5 percent. Primary education shows a strong positive impact on monthly earnings. Upper secondary education also significantly determines monthly earnings of wage earners. The relationship between university/college education and earnings is also very statistically significant. A worker with university/college diploma would earn monthly earnings of 19 percent higher in 1992-93 and 47 percent higher in 1997-98 in comparison with upper-secondary graduated worker with similar characteristics. The only exception in this regression is the coefficient of lower secondary level which is not statistically significant. This implies that there is a little difference in mean earnings between primary and lower-secondary graduated wage earners. It is consistent with the fact that workers with primary and lower secondary diploma are treated similarly as low-skill work force and the private cost for lower secondary education is much higher than that of primary level.

Table 4‑5: Rates of return to one year of schooling by level of education


This table depicts the consistencies in rates of return to education in Vietnam to some world patterns. Among basic levels of education, primary education exhibits the highest private profitability over secondary levels for both periods. Lower secondary education with insignificant impact on earnings results in 1 percent rate of return per year. University and college education shows a very strong impact on earnings in the second period with the highest rate of 9.6 percent per year, reflecting an increasing demand for well-educated and high skills laborers as the economy develops.

The coefficient of vocational training variable is positive but not statistically significant in 1992-93 survey. In 1997-98 survey, its coefficient is positive and statistically significant at 1 percent level. When vocational training is split into different levels, coefficient of post upper secondary vocational training is statistically significant at 5% level. Other thing being equals, an upper secondary graduates with vocational training can earn the rate of return of about 6.4 percent per year for his trainings in 1997-98 survey. The interpretation is that in short term, there is demand for low educated but skilled labor who finished training. However, there’s also a trend of recruiting high educated workers to give them on-the-job training. Upper secondary graduates can assimilate quickly in on-the-job training to increase productivity and earnings.

4.5.2      Structure change test

In this section, we need to test whether increases in rates of return in above section are statistically significant and whether there exists a structure change in the relationship between education and earnings. There are several methods to test for structure stability between two periods, that is the Chow test (Gujarati, 1995) and dummy variable test. Dummy variable technique is used in this study due to its advantages such as simplicity, testing variety of hypothesis and increases the relative precision of the estimated parameters. We start the test by pooling all observations in two data sets 1992-93 and 1997-98 together into one of 5,424 observations. To do so, a new dummy variable is generated, namely “survey”.
Estimating basic earnings function with “survey” and interaction term “yearsur” (yearsch x survey) variables added, both coefficients are statistically significant at 1% level. Other things being equal, survey period partly explains for the increment of monthly earnings. Moreover, the positive coefficient of interact term “yearsur” implies a higher coefficient of years of schooling in 1997-98 survey compared to that of 1992-93 survey. We, therefore, can reject the null hypothesis that the two regressions in two periods have the same coefficients for years of schooling, meaning that there exists a significant increase in impact of education on earnings.

 Regression of structure change test for Mincerian extended earnings model reveals that interaction effects between survey period and education levels are statistically significant in general. These positive coefficients reflect an increase in return to some education levels from 1992-93 survey to 1997-98 survey: 10.4 percent to upper secondary, 23.7 percent to university/college and 9.8 percent to vocational training. However, changes in coefficients of primary and lower secondary levels are not statistically significant, meaning that although there is a difference in absolute rates of return, these changes are not statistically meaningful.

Findings in this study are quite similar with increasing patterns found in other transitional economies: Slovenia (Orazem and Vodopivec, 1995), Hungary (Vargan, 1995) where from 1971 to 1993 there was a 3.5 times increase in return to higher education The results in the line with evidence of short-term changes associated with the process of economic transition. There may be three arguments for increasing rates of return. First, economic reforms in general and labor market reform in particular have generated a large amount of new employment and much of that are high-skills or managerial jobs. In addition, the development of private sector has created over 5 million employment in this period 1993-98 (IMF, 1999). Second, along with the development of the country is the expansion of scientific and technical knowledge that raises productivity of labor in production. Education and training is very helpful in these periods in coping with changing technologies and advancing productivity in both manufacturing and service sectors. Therefore, it results in an increasing demand for highly educated workers. Third, labor mobility is encouraged in this period since several major barriers to labor mobility have been removed such as social security system.

4.5.3      Gender disparity in returns to education

Estimation results show that the coefficient of gender interaction effect with years of schooling is not statistically significant in 1992-93 survey, meaning that the return to one year of schooling is almost indifferent between men and women. However, return to an additional year of schooling for a woman is 1.2 percent higher than return to an additional year of schooling for an identical man in 1997-98 survey. This conclusion is consistent with the world pattern and results found in many countries (Psacharopouslos, 1993). Nevertheless, disparity in return to education at different education levels for male and female are not always consistent. Our estimation implies similarity. The only exception is in 1992-93 survey where gender interacts strongly with upper secondary education to increase earnings for female graduates over male graduate, resulting in a higher return of 5 percent to an additional year of schooling at this level. This result may be due to the fact that it is easier for upper secondary female graduates to find a job with better earnings than their male graduates, for example jobs in some labor intensive industries like garment, textile and footwear.

4.5.4      Economic sector disparity in returns to education

Estimating basic model allowing for the interaction between “sector” and years of schooling, it is showed in 1992-93 survey returns to additional year of schooling in general are similar for wage earners in public and private sector. However, the situation had changed in 1997-98. Economic sector strongly interacts with year of schooling, making a significant distinction in return to schooling for public wage earners over their private counterparts of 1.4 percent. Our finding is in contrast to world pattern given by Psacharopouslos (1993) that private or competitive sector receives about 3 percent higher in return than pubic/non-competitive sector. This result may be due to the fact that private sector in Vietnam is somewhat different from that in other developed countries. It consists of mainly small and medium enterprises with simple production tools and technology, using skilled labor but not new advantaged ones as compare to the public sector.

Negative coefficients of interaction effects between economic sector and primary, lower secondary education imply that returns to primary and lower secondary graduates in private sector are higher than to identical worker in public sector in 1992-93 survey. In contrast, we found that upper secondary graduates in public sector earn higher return of 7 percent to an additional year of schooling at this level than identical graduates working in private sector. Returns to higher education (university/college) are insignificantly different in public and private sectors. Surprisingly, the later survey 1997-98, the pattern has dramatically changed. Economic sector has no statistically significant interaction effect with lower education levels while it strongly interacts at 1 percent significant level with university/college education to have impact on earnings. A striking negative coefficient of –0.49 shows that returns to university graduates working in private sector (mostly managerial and technical job) are much higher than return to their counterpart in public sector, 12 percent per year of university schooling.

Although the results are mixed, there are several similarities found in other developing countries. Stelcner’s study (1987) on public-private wage differential for males in Peru showed that post-secondary education yields 2 percent more per year in private sector, whereas secondary education yields higher return in public sector, 9.4 percent per year compared to 3.8 percent in private sector. Brown (1992) found similar results in Estonia where returns to education are larger in emerging private sector than in the state sector.

The results for the case of Vietnam can be apparently explained by the boom of private sector over 5 year period, labor market also turns to be more competitive where wages and salaries are more freely determined to attract university graduates. Good university graduates now prefer to work in private sector because the offer of managerial and highly technical work is higher. Evidence from the recently completed Higher Education Graduates Tracer Study (World Bank, 1997) suggests that change in trend is occurring rapidly, more and more university graduates are taking position with private sector. In addition, high demand for better-educated workers to suit new production technologies pushed the return to university and college education to high level.

4.5.5      Regional disparity in returns to education

Regional differences are explored in two perspectives: urban versus rural and the north versus the south. Generally, it is found that the coefficients of interaction term of “north” variable with years of schooling are insignificant in 1992-93 survey. It implies that there is no significant disparity between the north and the south in returns to additional year of schooling in this period. In contrast, in 1997-98 survey, positive coefficients of interaction results in 1 percent higher rate of return to an additional year of schooling to worker in the north than to identical worker in the south.

Regarding urban-rural comparison, statistically significant positive coefficient of interact term between primary level and “urban” shows that return to an additional year of primary education for an urban worker is much higher (19 percent) than for identical worker in rural areas. A reasonable explanation for this difference is the better quality of primary education in most of urban areas than in rural areas. In 1997-98 survey, the rate of return to one additional year of lower secondary education for a wage earner in urban areas is 5 percent higher than for an identical wage earner in rural areas. Similarly, the rate of return to an additional year of schooling for university graduates in urban areas is 4.5 percent higher than that for identical rural graduates.

Comparing returns to education between the north and the south, the most surprising result is that rate return to an additional year of primary education is 24 percent higher for worker in the south than for an identical one in the north. Statistically significant coefficient of university education’s interact term in 1997-98 survey at 1 percent level also reveals the disparity in returns to university education between the north over the south which is about 7 percent higher per year.

4.6 Full model estimation

Estimation results of full model are almost similar to the partial effects estimation. Particularly, it is found that the most of the coefficients which are statistically significant in partial regression are also statistically significant in full model, the sign and the magnitude of interact terms’ coefficients are also maintained, showing unchanged disparity in rates of return to education between groups.

There are only few exceptions. First, partial effect examination found that the rate of return to one additional year of upper secondary school is about 5 percent higher for women than for identical men in 1992-93 survey. However, the full model expressed no gender disparity in return to education at upper secondary level in this period. The discrepancy suggests that when all interact terms are taken into account or when we control for all characteristics as well as their interaction effects with education variables, gender does not significantly interact with upper secondary education to determine earnings. Second, interaction effects between years of schooling with “sector” and with “north” variables also turn to be insignificant in full model in the second survey 1997-98. It means there is no disparity in return to one year of general schooling for wage earners between the north and south as well as between public and private sector in 1997-98 survey.     
 

This chapter concludes some main findings and suggests some policy changes which are necessary for further reforms in education sector and labor market to improve the efficiency of education as an investment and may ensure better equalities of education achievement and income between different groups in the society.
5.1 Main findings and policy implications

1. In general, education strongly determines earnings in both survey periods 1992-93 and 1997-98. Returns to additional year of schooling are 2.8 percent in the first survey and 4 percent in the second survey. Empirical results indicate that primary education exhibits very high rate of return to one year of primary schooling which is 8.5 percent in 1992-93 and 9 percent in 1997-98 whereas secondary education has very minor impact on earnings. Therefore, the expansion of primary education will offer the highest rate of return than any other investment in education. The fact in Vietnam shows that primary school is the level that most of the poor can attain. From preliminary analysis of education financing in chapter 3, it is shown that the allocation of public spending on education in Vietnam is favor higher education although this bias has been lessen over 5 years. It means that public spending has benefited the rich rather than the poor. For the equality consideration, one should maintain and gradually increase the current levels of public expenditure on primary education.

2. Surprisingly, lower secondary education does not show significant and strong impact on earnings capacity over primary education in both periods, only 1 percent rate o return per year. It is likely that lower secondary diploma does not value much to employers. The other possibility is that private cost for lower secondary education is much higher than that of primary level. 1 percent rate of return to one additional year of lower secondary may also indicate the low quality of this level of education in Vietnam. As Vietnam moves towards the achievement of universal lower secondary education and to improve its investment return, more attention can be paid to the issue of education quality such as: increased teacher subject knowledge, increased instructional time, textbook availability, instructional material availability and some other factors such as professionalism among teachers, school and class facilities improvement.

3.Vocational training after upper secondary education has considerable impact on monthly earnings. The rate of return to one year of vocational training after upper secondary education has reached at the level of 6 percent which is much higher than the rate for lower secondary and upper secondary education. Policy concerns should pay attention to this good type of investment. Vocation training schools and centers need be upgraded or encouraged to open more in order to provide this crucial human resources. High return to vocational training after upper secondary education also implies that university and college is not the only-one choice for upper secondary graduate to secure a good-earnings job.

4. The rate of return to year of schooling in general and to year of schooling at different education levels are still low in comparison with the rates in other developing countries and transitional economies. Estimation of rates of return to education in two survey VLSS 1992-93 and 1997-98 shows that there is an increase in rates of return to an additional general year of schooling as well as at some levels of education. This reflects the effects of economic reforms over the five-year period in both the education sector and the labor market.

These conclusions have some important policy implications. Education can have little impact on earnings unless people can use education in competitive and open markets. With the target to a more developed labor market, some factors can be considered as: the openness of the market, the removal of labor mobility and restructuring the distorted wages and price.

5. Generally, gender has a positive effect on earnings which means that men have higher earnings and women. Gender disparity in rates of return to education was found in 1997-98 survey where female wage earners received 1.2 percent higher return to one additional year of schooling than identical male wage earner, other things being equal. The implication from this result has two folds. One, education investment in women brings higher return than investment in men’s education thus access to education for girls should be promoted. Two, as our result is interpreted upon condition of having a wage-earnings job, women should be encouraged to participate in labor force and at wage job in particular.

6. Difference in rates of return to education between workers in the public and private sector is found prominently at university level in 1997-98 survey. The rate of return to one year of university schooling for private wage earner is higher than the rate for identical public counterpart 12 percent. The disparity in return to education between economic sectors suggests that labor market in Vietnam is not perfectly competitive but rather segmented. Wages of workers with similar skills employed in different sector are not equal conclude that there exist structural differences across sectors.

7.Regional disparity in return to education is most prominent between rural and urban area. In general, return to one additional year of schooling for urban worker is 3 percent higher than the rate for an identical worker in rural areas. The evident implication is that education quality, especially lower education levels is very poor in rural area. Further measures similar to ones in implication 2 should be implemented to improve education quality. 

5.2 Limitations and areas for further research

There are also some limitations at this study. As mentioned in chapter 1 on the scope of the thesis, our sample is confined to only the wage earners so that all results are conditional upon being a wage earner. An analysis of the other sectors of the labor market will give more complete picture of the impact of education but is beyond the scope of this thesis. The implication on returns to vocational training is restricted to only general assessment. As the theory and international research suggests that it is possible to examine the effect of vocational training before taking the job and on-the-job training. This can be done later when this type of information will be available in later survey.

Due to the limited size of the extracted data set, the research has not analyzed though taking into consideration, the differences between seven geographical regions and between industries. Sector impact on worker’s earnings may be caused by different work intensity of the industry. Industry impact also might come from price distortions due to various policies, most notably trade protection policies.

Another limitation is in the analysis of economic sector disparities. The rate of return to education to the public sector worker and private sector worker should take into account the selection bias of sector choice. Probability of working in an economic sector should be considered as endogenous factor and full information maximum likelihood should be estimated. Disparities in the rate of return to education between gender, sector or region can also be further analyzed to determine the possible and significant factors that cause such differences. The results of these studies will clarify recommendations and measures for the government to narrow the income and education gap between different groups.

The availability of data in future surveys also creates an opportunity for research to be carried out on examining the structural change of education’s impact on earnings. The five-year period in this study is only able to access structural change in the short-term. Long-term assessment could be of grater value.