Returns to education in vietnam a clustered data approach

2003 did an attempt to analyze the returns to education inVietnam by using Mincer earnings function based on the 1992–93 Vietnam LivingStandards Survey VLSS data.. In this paper, I repli

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT

ECONOMICS

RETURNS TO EDUCATION IN VIETNAM: A CLUSTERED DATA

APPROACH

BY:

NGUYEN THI NGOC THANH

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HOCHIMINH CITY, DECEMBER 2012

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT

ECONOMICS

RETURNS TO EDUCATION IN VIETNAM: A CLUSTERED DATA

Assoc Prof Dr NGUYEN TRONG HOAI

Dr PHAM KHANH NAM

HOCHIMINH CITY, DECEMBER 2012

ACKNOWLEDGEMENTFirst of all, I would like to express my sincere thank to the Vietnam –Netherlands Programme (VNP) for such a challenging but interesting programme,whereby I enjoyed unforgettable time beside my classmates and broadened

my networking via class

I am much grateful to famous whole-hearted professors at home and abroadfor advanced knowledge and updated information they gave us in class andbeyond the class-time Specially, I would like to deeply thank two supervisors:Assoc Prof Dr Nguyen Trong Hoai and Dr Pham Khanh Nam for their helpfuland valuable advices on the last but utmost duty, this thesis, that helps me fulfill

my study career

From the bottom of my heart, I always feel thankful to my Family for theirdaily care, daily worries, daily happiness with every failure or achievement Iget in life I keep looking for chances to bring them happiness

To my C16 Classmates, I can say that two-year was a great memory when I

am with you all Thank you for your kindness, sharing and support Especially,

I cannot forget the enthusiastic disinterested help from Mr Le Anh Khang –our class “Hero” before every final exam He has inspired and motivated me alot I would like to take this opportunity to say thanks to him formally

Life is still ahead of us, let’s just stop a moment to celebrate ourachievement today and keep going forward afterward I wish you all goodhealth, happiness and success for the coming New Year 2013 Cheers !

ABSTRACTMoock et al (2003) did an attempt to analyze the returns to education inVietnam by using Mincer earnings function based on the 1992–93 Vietnam LivingStandards Survey (VLSS) data In this paper, I replicate the job of Moock et al.(2003) to re- estimate the returns to education by using the 2008 VietnamHousehold Living Standard Survey (VHLSS) and Mincerian earnings functions,but with a different regression method, called clustered data at household level usingpanel commands.

The study reveals that (1) an additional year of schooling associates with8.95% increasing in the average rate of return to education, comparing withonly 5% in 1992/1993 In terms of gender gap, females experience higher returns

to school than males (11.47% vs 8.33%) This pattern is unchanged whenreferring to result in 1992/1993 (6.8% vs 3.4%); (2) workers in public sectorget higher rates of return to education than those in private sector (9.95% vs.5.59%) However, foreign sector is the one has the highest rates of return amongthe three, 11.9%; (3) university is the best option for schooling investment withthe rate of return of 19% higher than upper secondary level while this numberwas 11% in 1992/1993 Primary level brings back 16% rate of return vs no level(13% in 1992) The rates are 10% for vocational vs primary (4% in 1992); 8%for upper secondary vs lower secondary; while only 2% for lower secondary vs.primary

Key Words: return to schooling, education, Vietnam, Human Capital,Mincer earnings function, clustered data, random effect model

4

TABLE OF CONTENTS

CHAPTER 1 INTRODUCTION 6

1.1 Problem Statement 6

1.2 Research Objectives 7

1.3 Research Questions 8

1.4 Research Methodology 8

1.5 Structure of the Thesis 8

CHAPTER 2 LITERATURE REVIEW 9

2.1 Definition 9

2.2 A Standard Model of Human-Capital Investment 10

2.3 Empirical Studies on Estimating Returns to Education 12

2.3.1 Selective Empirical Studies in the World 12

2.3.2 Empirical Studies in Vietnam 15

2.4 Analytical Framework 19

2.5 Chapter Remarks 19

CHAPTER 3 RESEARCH METHODOLOGY 21

3.1 Data 21

3.2 Research Methodology 23

3.3 New Approach - CLUSTERED DATA APPROACH in Estimating the Returns to Education 24

3.4 Empirical Models of the Returns to Education 27

3.5 Variable Coding 29

CHAPTER 4 RESEARCH FINDINGS AND DISCUSSION 32

4.1 Descriptive Statistics 32

4.1.1 Distribution of the Dependent and Explanatory Variables 32

4.1.2 Descriptive Statistics of the Dataset 37

4.2 Regression Results 38

4.3 Chapter Remarks 43

CHAPTER 5 CONCLUSION AND POLICY RECOMMENDATION 45

5.1 Conclusion of the Study 45

5.2 Policy Recommendation 46

5.3 Limitations of the Study 47

5.4 Suggestion for further Studies 48 REFERENCE

LIST OF TABLES

Table 2.1: Empirical studies in Vietnam utilizing Mincer earnings function over the

period 1992-2008 17

Table 3.1: Sample of cross-sectional data 26

Table 3.2: Sample of clustered data 27

Table 3.3: Description of the Variables and Variable Coding 30

Table 4.1: Descriptive statistics 38

Table 4.2: Earnings function by years of schooling 39

Table 4.3: Earnings function by sector of employment 40

Table 4.4: Earnings function with schooling levels (for all, males, and females) 41

Table 4.5: Private rates of return to schooling by level of education (%) 42

LIST OF FIGURES Figure 4.1: Histograms of log of earnings (by gender) 32

Figure 4.2: Histograms of log of earnings (by sector) 33

Figure 4.3: Histograms of years of schooling and log of hours worked/week 34

Figure 4.4: Scatterplots of monthly earnings and years of schooling 35

Figure 4.5: Scatterplots of monthly earnings and education levels 36

Figure 4.6: Scatterplots of monthly earnings and years of experience 36

LIST OF ABBREVIATIONS

ADB : Asian Development

Bank GSO: General Statistics

Office IV : Instrument Variable

OLS : Ordinary Least Squares

VHLSS : Vietnam Household Living Standard

Survey VLSS : Vietnam Living Standard Survey

CHAPTER 1 INTRODUCTIONThis chapter explains the context of the thesis, its objectives and researchquestions In addition, a brief of methodology is also mentioned in this part.Finally, the structure of the thesis is presented.

1.1 PROBLEM STATEMENT

Education plays an important role in modern labor markets Hundreds ofstudies in many different countries and time periods have confirmed that bettereducated individuals earn higher wage than the less-educated ones1 A variety ofstudies have been started with the seminal work by Mincer (1974) who was thefirst to derive an empirical formulation of earning over the lifecycle In hisbasic formulation, the logarithm of earnings can be interpreted as years ofschooling, years of experience and squared years of experience

In Vietnam, since the Vietnam Living Standards Survey (VLSS) firstlyconducted in 1992–93 till present, many studies have employed the VLSS dataand the Mincerian earnings function to examine rates of return to education inVietnam, such as: Glewwe

& Patrinos, 1998; Gallup, 2002; Moock et al., 2003; Liu, 2006; Nguyen XuanThanh, 2006; Vu Trong Anh, 2008; Vu Thanh Liem, 2009; Doan & Gibson,2010; etc The results are also diverse

The most cite study is from Moock et al (2003), in which the authorsattempt to analyze the returns to education in Vietnam by using Mincerianearnings function based on the data of VLSS 1992–93 The authors find that theestimated rates of return are quite low (4.8%) In particular, on average, therates of return to primary and

1 Psacharopoulos and Patrinos (2002) contains rate of return estimates for 98 countries spanning more than

30 years; Trostel, Walker and Woodley (2002) contains estimates for 28 countries; Polachek (2007) contains estimates for 42 countries; etc.

university education are 13% and 11% But these rates are just 4% and 5% atsecondary and vocational levels For higher education (colleges, universities orabove), the returns are higher for females (12%) than for males (10%).

Now, 20 years have passed, I return to the issue and question for now, whatare the returns to education in Vietnam? How have the returns changed?Especially, in term of gender gap, between males and females who receivehigher returns to education? In term of sectoral gap, among public, private, andforeign sectors, any discrepancies among these three? The findings are importantimplications for policy makers in directing the wage and educational policies

I would like to replicate the job of Moock et al (2003) to answer thesequestions by using the Vietnam Household Living Standard Survey (VHLSS),conducted by General Statistic Office (GSO), in 2008 and Mincerian earningsfunction, but with a different regression method which is first time applied in thiskind of estimation, called Clustered data at household level using panelcommands2, instead of using a simple standard cross-sectional OLS estimator.From the results, I would like to suggest some policy implications

1.2 RESEARCH OBJECTIVES

There are 03 main objectives in this study:

(1) To estimate private returns to education by years of schooling and by levels ofschooling for both sexes, for males and females; and in private, public andforeign sectors recently;

(2) To assess the variation in returns to education by comparing with the findings fromMoock et al (2003);

2 Please refer to the chapter on methodology (Chapter 3) for more details.

(3) To propose some policy options.

1.3 RESEARCH QUESTIONS

The research questions are proposed:

(1) What are the rates of return to education by years of schooling, by levels ofeducation for both sexes, for males and females; and in private, public, andforeign sectors recently?

(2) How are the rates of return to schooling different comparing with 15 years ago?Should the rates increase or decrease?

(3) What are policy recommendations?

1.4 RESEARCH METHODOLOGY

In the study, I use the VHLSS conducted by GSO in 2008 and the HumanCapital Model developed by Mincer (1974), with the regression method so-calledClustered data at household level using panel commands, instead of using asimple standard cross-sectional OLS estimator

1.5 STRUCTURE OF THE THESIS

The paper is structured as follows: Chapter 2 provides the literature reviewand empirical studies over the world and in Vietnam Chapter 3 describes the datasamples and specifies the research methodology The results based on descriptivestatistics and econometric models are presented in Chapter 4 The last chaptercomes up with conclusion, policy recommendation, limitations of the study, andsuggestion for further studies

CHAPTER 2 LITERATURE REVIEW

The first part of this chapter gives preliminary definition of main terms used inthe context The next part comes to provide theoretical foundation for empiricalresearch A standard model of human-capital investment by Mincer (1974) isintroduced to briefly explain how to form up the standard Mincerian earningsfunction Some of selective empirical studies on returns to education in theworld and in Vietnam are then recalled to summarize the empirical results found

by different researchers This chapter also is going to build up analytical frameworkfor the study

2.1 DEFINITION

Human Capital

Human capital is " the skills, knowledge, and experience possessed by anindividual or population, viewed in terms of their value or cost to an organization orcountry" (Oxford Dictionaries April 2010 Oxford University Press)

Rate of Return

Rate of return is " the gain or loss on an investment over a specifiedperiod, expressed as a percentage increase over the initial investment cost.Gains on investments are considered to be any income received from the security

Return to Education (Schooling)

The return to education is captured only indirectly by different methodsdepending on which level the study is examined at Specifically, at societylevel, the return to education is presented as the investment in educationrelative to national wealth; At enterprise level, it is the investment in training ineffect with enterprise

performance; In term of individual, the return to education is described as years

of schooling relative to life income

At individual level, the "individual return to education" is also termed as

"private return to education" to distinguish with "social return to education" atsociety level This study covers at individual level

2.2 A STANDARD MODEL OF HUMAN-CAPITAL INVESTMENT

Mincer (1974) proposed the standard Human Capital model in which thelog of observed earnings of an individual is interpreted by years of schooling,experience in labor market and squared of experience The theoreticalfoundations behind this standard model are briefly presented as follows:

Mincer contends that potential earning at time t depend on investment inhuman capital made at time t-1 Let Et be potential earnings at time t.Assuming that an individual uses kt as a share of his/her potential earningswith rt as a return in each period t to invest in human capital Then the potentialearning at time t+1 is as follows:

(2.2)

By assuming that schooling is the number of years, s, spent in full-timeinvestment (k0=…=ks-1=1), which is assumed to arise at the beginning of lifeand to produce a rate of return rs which is constant over all years of schooling

14

(r0=…=rs-1=β) and the return to post-schooling investment is constant over time(rs=…=rt-1=λ), we can rewrite equation (2.2) as follows:

15

t1 ln Et  ln E0  s ln(1   )  t1 ln(1  k j )  ln E0  s 

k j

2.3)

Where, the last approach is for small value

of β, λ, and k

To link between potential earnings and

experience z from labor market, the post-

schooling investment is assumed to be linearly

decreased over time

ln Y t

.8)

Replacing (2.8) into (2.7), we got the standard Mincerian earnings equation:

2.3 EMPIRICAL STUDIES ON ESTIMATING RETURNS TO EDUCATION2.3.1 Selective Empirical Studies in the World

There are a huge number of studies in the world relying on the Mincerianearnings function in estimating returns to education In spite of sample selectionbias as serious limitation, OLS regression are worldwide applied Numeroussupplementary variables are often fitted in the estimation, such as: gender,regional dummy variables, ethnicity, race, marital status, union membership, etc.These variables serve as exogenous “control variables” which may shift theearnings function upward or downward depending on their signs

Johnson and Chow (1997) estimate rates of return to schooling in China

by using OLS regression and data from a survey of Chinese individuals in 1988.The study also includes gender, race and Communist Party affiliation as controlvariables The authors find that the rates of return to education in China is4.02% in the rural and that 3.29% in the urban In the urban areas, females’rate of return is significantly higher than that for males (4.46% vs 2.78%).Additionally, members of Communist Party in urban areas have significantly lowerreturns to schooling than those of non-members (2.42% vs 3.68%)

Onphanhdala and Suruga (2007) assess the returns to education in Lao byusing Lao Expenditure and Consumption Survey in 2002/2003 (LECS 3) Dummyvariables for gender, area, ethnicity, type of business and region are included in theregression Interpreting the OLS estimator, the authors present that the returns toschooling in Lao are still very low, but have increased significantly with theeconomic transition (from 3.2% in 1997-1998 to 5.2% in 2002-2003) Specially,young workers obtain higher returns (7.0%) than older workers (3.9%),indicating that returns to education will continue to rise when the marketreforms have full effect Although workers with high levels of education are paidlarge earnings premiums, but primary level still indicate it

as the most profitable investment in education Furthermore, wage differentials are found significant between public sector (2.2%) and private one (5.2%).

To correct the sample selection bias caused by nonrandom data, Heckman(1979) introduces a two-step simultaneous model which has become a populartechnique in many fields of study Siphambe (2008) applies this model in his studywhen estimating the educational returns in Botswana in 2002-2003

Siphambe (2008) uses the Household Income and Expenditure Surveydata (HIES) in 2002-2003 to examine the returns to education in Botswana.The author includes such variables as age, education, and marital status in probitequation to create the selection variable, the Inverse Mill Ratio, which is theninserted into the earnings function The author then re-estimates the equation.The results show that the average rates of return to education in the 2002-2003period is 15%, representing 1% decline compared with the 1993-1994 period(16%) In term of schooling levels, details are reported that the biggest fall isfor upper secondary at 28% points (8% in the later period vs 36% previously).The university education, however, has the rates of return rise at more than 50%(24% vs 11%) Except the upper secondary, the pattern of rates of return toeducation keeps similar to the findings in Siphambe (2000) In term of wagedifferentials, the results show that the females and males enjoy the same rates ofreturns on education (around 15%) in 2002-2003, which is much different fromSiphambe (2000), where the average rates were higher for females than for males.Another critical problem when studying educational returns is the endogeneity

To deal with unobserved heterogeneity, in his review works, Card (1999)summarizes three broad approaches: (i) using instrumental variables based oninstitutional features of the education system (typically, Angrist and Krueger,1991); (ii) using family background as instrument for schooling (Ashenfelter andRouse, 1998; Nakamuro and

Inui, 20123); (iii) estimating based on the schooling and earnings of twins(Ashenfelter and Krueger, 1994) These works generally focus on theestimation of the average impact of education on earnings, by means of bothOLS and IV techniques.

Angrist and Krueger (1991) reason that because of school start age policyand compulsory school attendance laws, individuals born in the beginning ofthe year usually start school at an older age, and can therefore drop out aftercompleting less schooling than individuals born near the end of the year

The estimation draws on a variety of data sets constructed from the PublicUse Census Data in 1970 and 1980 The samples focus on males of 16 years oldborn in the US to specify the 1920-1929 corhort (in 1970 Census); and 1930-

1939 corhort and 1940-1949 corhort (in 1980 Census)

Using the interaction between quarter-of-birth and year-of-birth as instrumentfor education, the athors evaluate the effect of compulsory schooling laws oneducation across cohorts After controlling for age in quadratic, race, maritalstatus and urban residence, the difference-in-difference approach suggests thatthe returns to an additional year of schooling is 10% for men born in 1920-1929,6% for 1930-1939 men, 7.8% for 1940-1949 men

Ashenfelter and Krueger (1994) use primary data collected at the AnnualTwins Days Festival in Twinsburg (Ohio) in 1991 to state that the workers’ability (or other characteristics) and schooling are uncorrelated, hence cause

no direct effect on earnings The final sample contains 298 pairs of identical twins4

who are assumed to have the same ability but for some random reason vary inthe amount of school they

3 This empirical study is not included in the review of Card (1999) but in line with the work of Ashenfelter and Rouse (1998), so I add in.

4 Identical twins (or Monozygotic twins) who come from the same egg and sperm and are genetically identical, hence are hypothesized to share the same innate ability; vs Dizygotic twins (or fraternal twins)

who come from two eggs and two sperm and are not genetically identical, hence are more likely to be affected by omitted ability bias.

obtain By using each sibling’s report on his/her sibling’s education level as aninstrumental variable for his/her sibling’s education level, the authors find out that

an additional year of schooling raises wages by 12-16%

Ashenfelter and Rouse (1998) utilize the data conducted at the AnnualTwins Days Festival in Twinsburg (Ohio) (the so-called Princeton Twins Survey) for

3 years 1991-1993, including 340 twin pairs (680 twins) of identical twins Theauthors control for age (rather than experience as in traditional Mincerianequation) and use the difference between twin 2's report of twin 1's educationand twin 2's report of his/her own education as instrumental variable The resultsare fitted by fixed-effect estimator estimating that the annual returns to schoolingattained for identical twins is about 9% on average

In the very recent study, Nakamuro and Inui (2012), following Ashenfelterand Rouse (1998), measure the causal effect of education on earnings by usingsample of twins in Japan The final results regressed on the data of 2,257identical twin pairs collected through a web-based survey After correcting themeasurement errors by the IV method, the authors obtain 9.3% as the averagereturns to education in Japan

2.3.2 Empirical Studies in Vietnam

In Vietnam, there are a few articles written on the Mincerian function Most ofthe studies use OLS regression (one round or two rounds) (Glewwe & Patrinos,1998; Gallup, 2002; Moock et al., 2003; Vu Trong Anh, 2008; Vu Thanh Liem,2009; some use Heckman two-stage approach to correct the sample selectionbias (Liu, 2006); some use Heckman one single step model (Doan & Gibson,2010); some use the difference-in-difference approach (Nguyen Xuan Thanh, 2006).Glewwe & Patrinos (1998) use the VLSS 1992–93 to examine the nature

of attending private schools in Vietnam As a result, some importantconclusions are

made: (1) among public, private and semi-public schools, better-off householdstend to send their children to private schools rather than to semi-public ones; (2) ofthe same school attainment, individuals attending private schools get higher wagesthan ones from public schools; (3) the return to schooling in Vietnam is 1.6% in1992–93.

Gallup (2002), while calculating wage inequality among such controlledvariables as different sectors, regions, or genders in Vietnam in 1993 and 1998,finds that although the rate of return to schooling in Vietnam increases from 2.9%

in 1993 to 5.0% in 1998, it is still very low coefficient The results are retrievedfrom the VLSS 1992–93 and 1998 data, and two-round OLS regression

Moock et al (2003), in their attempt to analyze the returns to education inVietnam by Mincerian function based on the VLSS 1992–93, find that theestimated rates of return are quite low (4.8%) Specifically, on average, therates of return to primary and university education are 13% and 11%respectively However, these rates are just 4% and 5% at secondary andvocational levels For higher education (colleges, universities or above), thereturns are higher for females (12%) than for males (10%)

Nguyen Xuan Thanh (2006) is the pioneer in applying the difference approach to investigate the rate of return to schooling in Vietnam Theresult is derived from the VLSS 2002 He documents that an additional year ofschooling is associated with 7.32% increase in wage in 2002

difference-in-Liu (2006) exploits the data of VLSS 1992–93 with Heckman two-stageapproach and the VLSS 1998 with OLS regression He reports a higher coefficient

on schooling for males (5.9%) than females (4.2%) for year 1992–93, but acontrast results are seen for year 1998 when males are rewarded with 3.5%for each additional year of education, while females are rewarded with 4.8%

Vu Trong Anh (2008) uses the data of VHLSS 2004 to point out that therate of return to schooling in Vietnam is 7.4% in 2004 For the same objectivebut with a different data set - VHLSS 2006, Vu Thanh Liem (2009) shows thereturn to be 7.63% in 2006.

Different from other authors who just suggest the returns for a specificyear, Doan & Gibson (2010) utilizing VLSS 1998, 2002, 2004, 2006 and 2008examine the trend in the rate of return to schooling in Vietnam over 10 years1998-2008, by using OLS and Heckman selection estimator (Maximum Likelihoodapproach) The returns are found 2.9% for year 1998, 7.6% for 2002, 8.6% for

2004, 8.8% for 2006, and 9.5% for 2008, showing a clear rising trend over thementioned period and reach their peak around 2004-2008

The Table 2.1 below summarize the above-mentioned empirical studies in amore visual way (The research for 2009-2010 has not been found out till this study

is done, therefore, not included in the study)

Table 2.1: Empirical studies in Vietnam utilizing Mincerian earnings function over the period 1992-2008 5

Gallup (2002)

1992-931998

2.9%

Experience, experience squared

1992-931998

1.9%

Experience, experience squared, gender, minority, Chinese, non-agricultural

5 The comparision may be inappropriate due to different methodology apllied and diverse control variables incorporated.

employment, private, employer, HCMC, Hanoi

Moock et al

Experience, experience squared, log weekly hours worked

Nguyen Xuan

Thanh (2006) 2002

7.3% OLS Experience, experience

squared11.4%

in-differenceapproach

Difference-Experience, experience squared, gender, geography, agricultural/non-agricultural job, sectoral ownership

Experience, experience squared, married, migrant, urban, regions, majority, state employees, SOEs employees, industries1998

2004 9.6% one single squared, gender, household

2006 9.5% step model size, non-wage income

Clustered data: VHLSS 2008

The logarithm of monthly earnings

(1)

Schooling: divided by years of schooling and levels of schooling including primary, secondary, vocational education, bachelor and above.Years of experience

Squared years of experience

The logarithm of hours work per week

(2)(3)(4)

Models are fitted for all, male, and female; private, public, and foreign sectors

Random Effects model, clustered on household

Private returns to education for all, male, female; private, public, foreign sectors

2.4 ANALYTICAL FRAMEWORK

Figure 2.1 Analytical Framework2.5 CHAPTER REMARKS

The study employed the standard Human Capital model developed by Mincer

(1974) to build up its conceptual framework Under the framework, the

logarithm of observed monthly earnings of an individual is explained by years of

schooling, years of

experience in labor market, squared years of experience, and the logarithm of hours work per week.

In order to examine the returns to education in Vietnam, most of theexisting studies use OLS regression as their first or final modeling However, thismethod does face problems of underestimating standard errors within the samehousehold and ignoring the mean variation between different households

Specifically, individuals/employees in the same household are likely to sharethe same unobservable household characteristics such as culture, specific geneticsthat may affect their earnings ability Therefore, the error terms for individualsfrom the same household will be correlated through a common household-levelcomponent, and if ignored this may lead to substantially underestimated standarderrors (Glick and Sahn, 2000, p.69-70)

Moreover, OLS estimator ignores the mean variation between households Forinstance, OLS results a common intercept at state level, say, all individualshave a common intercept, regardless of households It is unlikely to be true inreality as individuals from different households may hold different intercepts

By addressing the above issue, instead of using a simple standard sectional OLS estimator along with cluster-robust standard errors, I transfercross-sectional data to clustered data at household level, and then fitted

cross-by random-effect estimator By doing so, I allow for such correlation, asmentioned above, in the model through a random effect for the residuals

27

CHAPTER 3 RESEARCH METHODOLOGYThis chapter describes the source of data used, the way to collect and extract

to final results The methodology applied to analyze the data sample is presented

in the next part, followed by empirical models The introduction of newapproach - CLUSTERED DATA APPROACH is the highlight of this chapter.Variable coding is the final part to show how I code the dependent andindependent variables from the data set

3.1 DATA

The data for this study is the Vietnam Household Living Standard Survey(VHLSS) conducted in 2008 by the General Statistical Office (GSO) of Vietnam.The surveys contain detailed information of 9,189 households from 3,063communes Samples were weighted basing on the statistics of Vietnam PopulationCensus in 1999 with approximately 70% of Vietnamese households lived in ruralareas The communes were randomly selected from a total of proximately 10,000communes in 646 districts, and 64 provinces and cities in Vietnam, and then anaverage of 3 households were randomly selected for interview in each commune

In this research, I am going to estimate returns to education for onlyindividuals who are employed for salary6 Only individuals in ages from 15 to

60 for male and 15 to 55 for female are considered Earnings are calculated bymonthly earnings in labor market Earnings/month (1,000 VND) Individuals whowork for their household are dropped out of the sample

6 VHLSS separates employment into wage employment, farm employment, and non-farm employment In this study, I consider only wage earners Earnings are proxied only by salary/wages received, including payment in kind, from the work being done (refer to Table 3.3 “Description of the Variables and Variable Coding” for more details).

self-Years of schooling are collected from general education system7 This isthe highest class that he/she has been completed For example, a person who is ingrade 10, only recorded grade 9 is the highest grade completed Another man was

in grade 9 and dropped out of school, write the grade 8 is the highest classfinished For individuals who are at College level, years of schooling equal 15years; 17 years for Bachelor; 19 years for Master; and 22 years for PhD level (LeThi Nhat Phuong, 2008; Le Anh Khang, 2012)

While data on schooling attainments for each individual is obtainable,information on post-school investment is not available in VHLSS Therefore,following Mincerian earning function, difference in quantities of post-schoolinvestment among employees are measured by differences in years ofexperience which is proxied by age of employee (in years) minus years ofschooling

Hours of work per week are affixed as a compensatory instrument (Moock etal., 2003, p.504) Mincer (1974, part 1, p.22) states that the annual earningsprofile is affected when hours of work vary over the life cycle For instance, in thecircumstance of certainty where individual wealth is considered as fixed thecost of time increases with experience until reaching the peak of earningcapacity If so, the ascent and descent of earning capacity is likely to trigger acorresponding pattern of working hours provided for market Thus, it seems to beoverestimated of investment in human capital or the rates of return if we use theobserved annual earnings as dependent variable Hours worked per week arethen added as a compensatory factor for the above overestimated matter

After consolidated to remove errors and inconsistencies, the sample dataremains 6,956 individuals/employees, living in 4,335 households

7 VHLSS divides education into general education and vocational education.

3.2 RESEARCH METHODOLOGY

Returns to education are estimated based on the Human Capital Model developed

by Mincer (1974) which is formulated by an earning function as follows:

ln Yi    Si   1 EXPi   2

Where,

lnYi is the logarithm of the monthly earnings for

individual i Si is the number of schooling years of

individual i

EXPi is the number of years of working experience of individual i

EXPi2 is the squared of experience of individual i

ui is the error term

The squared of experience (EXP2) in equation (3.1) implies that earningsshould increase along with the years of experience but at a diminishing rate and itscoefficient (γ2) is expected to have a negative sign

To measure the average returns to education at different levels ofschooling, various dummy variables are created by converting the continuousvariable years of schooling The extended earning function is as follows:

ln Yi    1 PRIM i   2 SECi  3VOCi   4UNIVi   1 EXPi 

then divided by the number of years of schooling at that level For example, therate of return to the kth level (rk) is calculated as follows:

rk   k   k1

nk

(3.3)Where

nk is years ofschooling at

Specifically,

at primarylevel,

individual

complete,hence nPRIM =

parameter is

secondary(nSEC = 7), 6for

vocationaleducation(nVOC = 6),and 4 foruniversity

Moock et al.(2003,

I separatesecondaryeducationinto lowersecondary(LOWSEC)and uppersecondary(UPPSEC)education

By doing so,the years ofschooling atlower

secondarylevel will be 4years

and uppersecondarywill be 3years

3.3 New Approa

However, thismethod doesface

problems ofunderestimating standarderrors within

household andignoring themean variationbetweendifferenthouseholds.Specifically, in terms

e householdcharacteristi

cs that mayaffect theirearningsability Forinstance,the culture,genetics of

a specifichouseholdmakeindividuals

householdperformwell, getmore

knowledge

in school,more

earnings inthe labor