education income distribution and growth

Trang 1

INFORMATION TO USERS

This manuscript has been reproduced from the microfilm master UMI films the text directly from the original or copy submitted Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer

The quality of this reproduction is dependent upon the quality of the copy submitted Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted Also, if unauthorized copyright material had to be removed, a note will indicate the deletion

Oversize materials (ẹg., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand corner and continuing from left to right in equal sections with small overlaps

Photographs included in the original manuscript have been reproduced xerographically in this copỵ Higher quality 6” x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an ađitional chargẹ Contact UMI directly to order

Beil & Howell Information and Leaming

300 North Zeeb Road, Ann Arbor, MI 48106-1346 USA 800-521-0600

Trang 3

Education, Income Distribution and Growth

Ana Corbacho

Submitted in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

in the Graduate School of Arts and Sciences

Trang 4

UMI Number: 3005706 ® UMI UMI Microform 3005706

Copyright 2001 by Bell & Howell Information and Learning Companỵ All rights reserved This microform edition is protected against

unauthorized copying under Title 17, United States Codẹ

Bell & Howell Information and Learning Company 300 North Zeeb Road

Trang 6

Abstract

Education, Income Distribution and Growth Ana Corbacho

The dissertation studies the role of education in economic development for the case of Argentinạ

The first chapter reports returns to education in the Metropolitan Region of Buenos Aires during 1985-1997 In 1991, several important macroeconomic re- forms were enacted Previous to these reforms, returns at all levels of schooling followed similar trends Since 1992, however, returns have risen particularly at the higher education level

Between 1974 and 1997 education and income inequality have increased in Argentinạ While children from a low-income family stop their schooling at the el- ementary level, wealthier students make it all the way to the university, enjoying the highest returns The second chapter quantifies the impact of parents’ education, family income, gender and other factors on educational achievement Income trans- fers would be effective in promoting attainment However, these transfers would have a different impact depending on family background While the distribution of education of disadvantaged children would shift towards the high school level, the distribution of education of wealthier students would shift towards the university

Trang 9

Confents

Ïntroduction QC ch nh nh nạ 1

1 Gender Profiles in Returns to Education 2

GẠ na nh đạ ee epee pectetenyebeeeueeyencẹ 2

1.2 Econometric Model ch nh hs 5

1.3 Data Description .0.0 2 ốnn6ẲẮ<eádđ Ẽ_ 1

1.4 Average Retums 2 6ẹ I

1.5 Returns by Level 00 cece cece cae cree ees eenetnetncunrenees 13 1.6 Final Comments 0.0.00000 00 cc ccc ccc cece ccc e ee uc ene cn eee eteebebecbeeeccẹ 15 1.7 Tables and Figures 00 0c cece cece cece cee ev cece veceaenttaureenens 17 2 Family Background and Educational Attainment 32 2.1 Introduction 0 0c e cece cence eevee eens ra eeuttetvertbrnberncẹ 32 2.2 Theoretical Framework 0.00.00 00 000 ccc cece c cee ce cece ceeteeteeeceeeceeeeẹ 35

NG nh gạ vnnenerneetereanaes 39

2.4 Data Description nến ene s eee eeeeeaetnnaus 42

5 “e6 ẽ nsH 44

“ 5 n6 nạ 46

Trang 11

3 The Allocation of the Education Budget 59 3.1 IntroducHon ¬ eee ene e ence eee tees ee eeeavunes 39 3.2 Basic Framework ¬ HH Hy HN Nà ki ki KH kg vn ky kà 66 3.3 Government Spending «ĐH HA Hy Ko ĐK nà Ki renee eee te veneer ee eeeaences 73 3.4 Re-allocation Policy ¬_ 79 3.5 SkillPremium ¬ 87 3.6 Application to Latin America ¬ reeset ee een tenes 89 3.7 Final Comments k k t Ho Hàn tk n kh Vy cv xế 94

SA an eee cence cence ees ¬ cece eee eee eenee 97

3.9 Tables and Figures ¬ eeeeets ¬ 99

References Q.22 chọ 111

Trang 13

Acknowledgments

I am grateful to my advisors Alessandra Casella, Kenneth L Leonard and Aavier Sala-i-Martin for valuable comments and suggestions I am also grateful to Francisco Rivera-Batiz and Stephen P Zeldes for participating in the Dissertation Defensẹ Mitali Das and Stephen Cameron provided useful guidancẹ

I thank my friends Adriana, Bikas, Francisco, Manuel, Alex, Alberto, Nuno and Anne for their companionship and thoughtful advice throughout the Ph.D pro- gram

My family deserves special recognition for always being there for me, even if thousands of miles awaỵ Ma, Pa, Matu, Alvaro, Sebas, Vally: Gracias por todo! Los quiero muchọ

I am certain that without the love and patience of my husband, Andrés, I would not have been able to finish this dissertation He has been my strongest supporter I will always be indebted to him

Trang 17

Introduction

Human capital, as measured by the educational attainment of the population, has consistently emerged as an essential factor for economic growth and development Edu- cation contributes to economic performance through several channels At the micro level, more educated individuals are more productive and earn higher wages At the macro level, endogenous growth theories predict that investments in schooling will promote growth Ed- ucation is also seen as a great equalizer: a better distribution of education contributes to a more equal distribution of income and enhances intergenerational mobilitỵ

Trang 19

Chapter 1

Gender Profiles in Returns to Education in the Argentine Metropolitan Region

1.1 Introduction

Education is a key factor in economic development It can help to reduce inequality, im- prove efficiency and promote social and political advancement A critical issue when de- signing sound economic policies is the estimation of the economic value of education.! Rates of return to schooling can be useful to determine optimal levels of educational eX- penditures, investments in research and development and appropriate financing schemes for public education

The estimation of rates of return falls within the human capital approach to school- ing.” The key assumption in human capital theory is that individuals choose the level of education that maximizes the present value of lifetime earnings Ađitional schooling im- plies opportunity costs such us foregone eamings and direct expenses (tuition, books, etc.) An individual will undertake extra years of education if the expected lifetime earnings com- pensate for these costs He will choose the level of schooling that equates its return to the retum from other forms of investment The standard human capital earnings function is:

L See for instance Glewwe (1996)

See Becker (1993) and Mincer (1974) Other models of schooling demand include signaling and con- sumption See Stiglitz (1975) for an exposition of the signaling model and Lazear (1977) for an analysis on education as a consumption versus production good See Willis (1986) for a general discussion

2

Trang 21

L.1 Introduction 3

J

InYi = Ø, + đái + Bgzi + Byx7 +S Bykyi + v; (1.1)

j=4

where Y is a measure of income, earnings or wage rates, s is a measure of educational attainment, like number of years of education, z is years of experience in the labor market, k; represents other observed factors that affect wages (gender, tenure, industry or type of occupation), and v is a disturbance, representing other unobserved forces.°

The concavity of observed earnings profiles is captured by the quadratic experience terms ({3, is expected to be positive and (34 to be negative) The schooling coefficient, By, provides an estimate of the rate of return to education It measures the percentage extra eamings a worker would have for an extra year of education

Returns to education have been estimated for a number of countries." The evidence for the case of Argentina, however, is quite scarcẹ One exception is Pessino (1996), who estimates standard rates of return to schooling in the Metropolitan Region of Buenos Aires between 1986 until 1993 Major emphasis is given to the effects of hyperinflation and price stabilization on wage profiles This study does not include a number of controls that have been shown to be important in wage regressions like industry, occupation and size of the firm effects This results in an overestimate of the return to education Ađitionally it does not control for potential self-selection bias, arising from the use of a non-random sample in

3 Mincer's original earnings function (Mincer, 1974) included only education, experience and experience squared as independent regressors Subsequent research in human capital promoted the inclusion of other factors, such as tenure, industry, occupation and size of the firm

Trang 23

1.1 Introduction 4

wage regressions (workers who choose to participate in the labor market) Finally, it only concentrates on returns for men

This paper makes several contributions It estimates rates of return to education for

the Metropolitan Region for a longer period of time, from 1985 until 1997, and for both female and male workers The econometric model follows Heckman (1979) and controls for self-selectivity into the labor market using an efficient Maximum Likelihood procedurẹ Wage equations include industry, occupation and size of the firm effects The inclusion of these controls decreases the return to education by almost three percentage points compared to Pessino’s estimates

The return to average years of education increased from 1985 until 1989, reaching 7.6% It plummeted in 1990 and started to recover considerably since 1991, particularly at the higher education level Females have made great improvements in their labor market position Between 1991 and 1997 female labor force participation increased by nearly 20 percentage points This was accompanied by a narrowing gender gap both in wage levels and returns to education The return to education for female workers even surpassed that of male workers in 1995 and 1996

Trang 25

1.2 Econometric Model 5

respectively, the return at the college level has risen considerably, reaching 10% in 1997 This rise in the return to university education was experienced mostly by female workers

The rest of the paper is organized as follows Section 1.2 discusses the econometric model Section 1.3 describes the data and summary statistics for female and male workers Section 1.4 presents rates of return to years of education for all workers and by gender, while section 1.5 estimates returns by level of education Section 1.6 concludes

12 Econometric Model

Returns to education will be estimated via a Heckman Maximum Likelihood® procedure that controls for self-selection bias It is a generalization of the model presented in equa- tion 1.1 that takes into consideration the non-random sample of workers used in any wage regression OLS regressions that do not take self-selection into consideration yield biased estimates

In the standard labor supply model, an individual will choose to work (labor sup- ply decision at the extensive margin) if the market wage exceeds her reservation wagẹ Therefore we can observe a positive wage for such individual For those individuals whose reservation wage is lower than the market wage, we observe either a zero wage or nothing at all This implies that individuals select themselves into the labor market

Consider the model:

Yi; = In ¥, —In(A,) = X} eb 4+ dị “participation equation”

Trang 27

1.2 Econometric Model 6

YP =lnM, = MPR 40? “wage equation”

In Y; is the logarithm of real wages and R; is the reservation wage corresponding to person 2,

If Y;*! > 0 then person 7 is active in the labor market and E; = 1, where E stands for

“employed” For this person Y*? = In Y; is observed

If Y,*' < 0 then E; = 0 Wages for this person are not observed, but the exogenous variables X? and X? arẹ This is an example of censored regression models

A simple least squares regression of In Y on X° yields biased estimates:

E(iny, / Ey =1) = X73? + E(ue jal > —xX}34) (1.2)

The last term is usually non-zerọ Under the assumption of normality of the errors,

with (tỷ tỷ) distributed as bivariate normal [0,0,1,< p] the expected value of observed

€arnings equals:

E(ny, / Ej = 1) = X73? + E(ủ Ju} > —X}8") = X78? + pod(X} 8") (13)

where Ặ) are the “inverse Mills’ Ratios.”© The parameter “p” can result either positive or negative according to the type of selection that occurs

There are two ways of correcting this selection-bias: the standard Tobit model or Heckman’s selection model, also known as the “Generalized Tobit”.? The Generalized Tobit estimates equation 1.3, either through Maximum Likelihood or two-step consistent

Trang 29

1.3 Data Description 7

procedures The advantage of this model is that the coefficients and the variables in the participation equation are not restricted to be equal to the ones in the wage equation, as is the case in the censored Tobit model.® Strictly speaking, there is no need for X? to be dif- ferent from X', since the inverse Mills’ ratio is a non-linear function of the explanatory variables However, in practice, it is preferable to have some identifying restrictions to re- duce the potential collinearity between the Mills’ ratio and the rest of the variables included in the wage equation

1.3 Data Description

Micro data sets for Argentina are scarcẹ In fact, the only source of public systematized data is the Permanent Household Survey, collected by INDEC (National Institute of Statis- tics and Census), for the Buenos Aires Metropolitan Region The Metropolitan Region includes the Capital District and 19 counties of the Greater Buenos Aires, concentrating approximately one third of the total population of the countrỵ For other areas data exists in a much less regular basis The Survey is conducted in May and October of each year This study uses the October data onlỵ

The sample includes men and women, aged 18 to 65 The dependent variable is the logarithm of real hourly wages from the main occupation The independent variables to be

included in the wage equation (X*) are:

Trang 31

làn

to

1.3 Data Description 8

Schooling dummies, corresponding to the highest degree attained by the individual (primary complete or incomplete; secondary complete or incomplete; college complete or incomplete), or alternatively: Years of education (yed), constructed using the average number of years spent at each stagẹ Years of labor market experience, constructed as experience = age — yed — 6, and experience squared

Dummies for tenure in the job (for less than a year, between I and 5 years, between 5 and 10 years, and over 10 years)

Occupation and industry dummies Size of the firm dummies

Gender

The independent variables to be included in the participation equation are (X°): Age and age squared

Trang 33

1.43 Data Description 9

Years of education squared, for the wage equations that use schooling dummies as the measure of educational attainment.®

Schooling dummies, for the wage equations that use years of education as the measure of educational attainment.!©

Figure 1.3.1 displays the behavior of real wages There is a clear turning point in

1990 From 1985 till 1989, real wages followed a downward trend, decreasing by 38%

When disaggregating by gender, males suffered a loss of 40% and women of 30% From 1991 on, real wages started nsing, reaching in 1997 a higher level for females and on average than in 1985 On the other hand, male wages in 1997 remained at lower levels than at the beginning of the period The average male-female wage differential was almost 70 cents between 1985 and 1989 From 1990 until 1997, this differential dropped to only 14 cents,

Labor force participation rates also display different trends before and after 1990, par- ticularly for females Between 1985 and 1990, female labor force participation was barely above 35%, one of the lowest of Latin Americạ Since 1991, however, female labor force participation increased by almost 20 percentage points, reaching 52% in 1997 The growth in male labor force participation was more moderate throughout the whole period These

3 Years of education (not squared) turned out to be insignificant in the participation equations and was therefore excluded to decrease the potential collinearity between the Mills’ Ratios and the other explanatory variables in the wage equation

Trang 35

1.3 Data Descnption 10

considerable changes in female labor force participation make selection controls in wage equations more important Figure 1.3.2 displays the trends in labor force participation

Between 1985 and 1989, inflation was very high in Argentina (188% average) and there was little to no GDP growth In the period 1989-1990 there were two hyperinflations, with an average inflation of 2700% A currency board was adopted in 1991, pegging the local currency to the US dollar on a one-to-one basis (the so-called Convertibility Plan) This plan was very successful in stopping inflation Between 1991 and 1994, GDP growth was of 32% and pnices stabilized, reaching a one digit inflation rate in 1994 1995 was an exception in this spectacular performancẹ The effects of the Mexican devaluation in December of 1994 were severe, with a fall in GDP of 5% The Argentine economy expe- rienced positive growth again in 1996 and 1997 Figures 1.3.3 and 1.3.4 present the trends in inflation and GDP growth

There were major structural changes in the economy during the 90’s, including pri- vatization of public enterprises, reorganization of the revenue system, rationalization of public expenditures and liberalization of tradẹ It is fair to say that, from 1991 onwards, Argentina moved towards a more market-oriented economy,

Trang 37

1.4 Average Retums i]

educated than working men, their salaries have been lower, especially during the first five

years under analysis.!!

The following sections present the returns to schooling for males and females to see how they have changed over time and whether macroeconomic shocks have had a different impact on them

1.4 Returns to Average Years of Education

This section reports the results from the wage equations that include average years of edu- cation as the measure of educational attainment.!*

Table 1.4.1 corresponds to the sample of all workers, while Tables 1.4.2 and 1.4.3 correspond to male and female workers respectivelỵ The Wald Test rejects the null hy- pothesis of all coefficients being jointly equal to zero at the 1% level of significance for all years The test for self-selection bias corresponds to the significance value of the corre- lation coefficient between errors in the participation and wage equations (Rho) Selection bias turned out significant for several years.!3 The human capital variables are significant at the 1% level in most regressions Experience and experience squared have their expected Signs, positive and negative respectivelỵ The female dummy is negative and significant ex- cept for some years in the 90’s, showing that women earn on average 10% to 40% less than men

"The reader should note that these averages in years of education are for the sample of workers When looking at total populations, men have on average two more years of education than women

12 Results for the participation equations are available upon request

Trang 39

1.4 Average Retums 12

Figure 1.4.1 plots the coefficient on years of education for all workers The return to education increased during the inflation period between 1985 and 1989, when it reached a maximum of 7.5% Then it declined sharply in 1990 and started to recover since 1992 By 1997, it was roughly at the 1985 level, around 6.5% One explanation for this trend is that in periods of high inflation education is useful to deal with real wage indexation mechanisms, and this is reflected in an increase in the return to education

In order to establish the effect of inflation on the level and slope of wage profiles, Table 1.4.4 presents a model that estimates the same equations as before on a pooled sample for all years The first column controls for the level of inflation while the second column ađs an interaction term between inflation and education to capture the effect of inflation on the return The results suggest that inflation has a negative effect on the level of wages but a positive effect on the return to education An increase of 100 percentage points in the annual inflation rate reduces wages by about 1.5% and increases the return to education by 0.2%