Gender wage gap and returns to schooling in Argentina

(1)

Gender wage gap and

returns to schooling in

Argentina

Virginia Giordano

Student Number: 10824839

Supervisor: Dr. Carmine Guerriero

Date: June 30

th

_{, 2015}

Master Thesis

Faculty of Economic and Business

Universiteit van Amsterdam

(2)

2

A

BSTRACT

Gender wage gap and returns to schooling

in Argentina

In Argentina women’s salaries are lower than men’s. Different rates of returns to schooling are a possible explanation for the existence of the gender wage gap. In this paper the returns to schooling for male and female workers are estimated using various methodologies (Ordinary Least Squares, Heckman Two Step Procedure, Panel Data Fixed-Effect Approach and Quantile Regression Analysis). Data form the Argentine Permanent Household Survey conducted by the

Instituto Nacional de Estadisticas y Censos (INDEC) is used. The Minceran Earning

Functions reveal that women used to have lower returns to education, but the trend has reversed. Now women experience higher returns at every level, except primary school. The results show that women do not earn less than men because the labor market rewards them less, however, the Oaxaca-Blinder decomposition of the gender wage gap reveals that if women had men’s education and work experience, their wages should decrease, meaning possible differentiated treatment by employers.

(3)

List of Tables

Table 1: Educational Attainment of the Argentine Workforce by Gender .. 9

Box 1: The World Economic Forum Global Gender Index ...10

Table 2: Labor Market Indicators by Gender ...11

Table 3: The Gender Wage Gap ...14

Table 4: Evolution of the returns to schooling by gender ...22

Table 5: The Retruns to Schooling: Panel Fixed Effect Approach ...25

Table 6: Quantile Regressions, returns to education by gender ...26

Table 7: Oaxaca Decomposition ...29

Table 1a: Description of Explanatory Variables ...34

Table 2a: Descriptive Statistics ...35

Table a4: Determinants of Earnings OLS, 2004 ...41

Table a14: Determinants of Earnings Heckman, 2004 ...51

(5)

I. Introduction

In Argentina, as in most parts of the world, women have made major progress towards achieving social and economic equality. Female labor market participation has increased substantially since the 70s, a process that has been linked to decline in fertility rates and the expansion of the schooling system (Duryea et al, 2003). Women have also made considerable gains in the area of education, and have even surpassed men in some fields; they outperform men at every level and are more likely to enroll in university (DiPrete et al., 2013). Even though there are more and better-educated women in the labor market today than before, it’s a well-documented fact that women are still in a disadvantaged position relative to men: they participate less, they are subject to a higher degree of unemployment and labor informality and still earn less than men in many segments of the labor market. Part of the wage gap may be explained by differences in human capital between men and women, but part of it responds to unobserved factors or “wage discrimination”- women earn less than men simply because they are women (International Labor Organization, 2015). The aim of this paper is to test one possible explanation for the existence of the gender wage gap: there are differential returns to schooling for women than for men in the Argentine labor market.

There are many previous estimates that have measured the returns to education in Argentina, but most of them have reported results for male wage-employees in the Greater Buenos Aires. For instance, Pessino (1995) estimates the returns to education for the 1986-1993 period for males in Great Buenos Aires. She finds that after an inflationary shock, the returns to schooling increased from 10

(6)

6 percent to 12.5 percent in 1989, but dropped once again to 10 percent once inflation dropped as well. Also estimating results for male workers in Great Buenos Aires, Margot (2001) calculates the returns to schooling using a dynamic cohort analysis for the period 1980-1999. She shows that workers who have finished secondary school experience returns of around 13 percent on average for the whole period. Workers with complete higher education experience much higher retruns, 19.1 percent on average. Gasparini and Acosta (2007) also find increasing returns for workers with higher education during the 90’s.

Some authors have reported results by gender. Psacharopoulos (1994) estimates that in 1989 the urban argentine the labor force had 9.1 years of schooling on average and the private rate of return to another year of schooling was 10.3 percent. Returns to schooling for women were slightly higher than for men: in 1985, 9.1 for men and 10.3 percent for women; and in 1989, 10.7 for men and 11.2 percent for women.

The two most recent and consistent estimates are by López Boo (2010) and Giovagnoli, Fiszbein and Patrinos (2005). López Boo (2010) employs a variety of methodologies to estimate the returns to schooling for 1992-2003 and uses macroeconomic time series to explore shifts in earnings after the 2001 argentine crisis. She find that the returns to education for men are higher than for women at every level of education and confirms the reversal for the trend found in the 1980’s and 1990’s by Psacharopoulos (1994). The author also finds evidence of a positive selection bias, meaning that the wage distribution observed for paid women is higher than would be found for comparable women workers who chose not to participate.

(7)

Giovagnoli et al. (2005) estimate the returns to schooling in urban Argentina for 1992-2002, using OLS and a Quantile Regression Analysis approach. They find results that are in line with López Boo (2010). The rates of return for men and women were 9.1 percent and 8.1 percent in 1992, and increased to 12 percent for men and 10.8 percent for women in 2002. The returns to all level of education are much higher for men than for women at every level of education. For the Quantile Regression Analysis they find that men in higher quantiles of the distribution have higher returns to schooling compared to those in the lower quantiles. For women it’s the exact opposite, returns are highest at the lowest quantile than in the highest quantile.

Overall, the previous literature shows two main results: (i) returns to schooling increase with the level of education (Fiszbein, Patrinos, and Giovagnoli 2005; Gasparini and Acosta 2007), which goes against the traditional view that earnings function is concave (Fiszbein, Patrinos, and Giovagnoli 2005), and (ii) women used to have higher returns to schooling in Argentina (Psacharopoulos 1994), but the trend reversed in the 90s. During this decade, men had higher returns to education compared to females for every level of education (Fiszbein, Patrinos, and Giovagnoli 2005; López Boo 2010).

The objective of this paper is to estimate the returns to schooling by gender in a consistent manner for the 2004-2013 period, a decade not covered by any of the previous estimates. Mincer’s (1974) semi-logarithmic approach, already broadly used in the literature (Psacharopoulos, 1994, Psacharopoulos and Patrinos, 2004) that links individual earnings to an additional year of schooling will be employed. Two issues need to be takeninto account to consistently estimate the relationship between earnings and schooling are sample selection bias and

(8)

8 omitted variable bias. In order to take into account these issues and consistently estimate the returns to schooling, four methods of estimation are employed: i) Standard Ordinary Least Squares, ii) Heckman two-step procedure to deal with sample selectivity issues; (iii) a Quantile Regression Analysis and (iv) a Panel Data Fixed Effect Approach. In all methods parameters are allowed to differ between men. Data form the Argentine Permanent Household Survey (EPH) conducted by the Instituto Nacional de Estadisticas y Censos (INDEC) will be used. The rest of the paper is structured as follows. Section II makes a brief analysis of the current labor market making gender comparisons; Section III discusses the data, Section IV examines the empirical strategy, Section V shows the results and finally section VI concludes.

II. Education and Wages by Gender: 2004-2013

The Argentinean educational system is one of the most advanced of Latin America: data from the 2010 census shows that Argentina’s illiteracy rate of 1.9 percent is the second lowest in Latin America. In 2013, average years of schooling of the whole population were 9.8, higher than the regional average of 8,2 (Barro and Lee, 2013). The proportion of workers with complete university raised from 18 percent in 2004 to 22 percent in 2013. These figures show an improvement in the human capital level of the argentine workforce. Also, the gap between female and male university graduates has increased 4 points: 14 percent of males had completed university in 2004, while 24 percent of females did. The figures for 2013 are 16 and 30 percent respectively. This could be signaling some selectivity issues in the labor market: more educated women self-select into the labor force (Table 1).

(9)

Table 1: Educational Attainment of the Argentine Workforce by Gender

2004 2013

Men Women Men Women

Incomplete University 6% 6% 4% 3% Complete Primary 25% 19% 19% 13% Incomplete Secondary 20% 15% 20% 14% Complete Secondary 20% 21% 28% 26% Incomplete University 13% 15% 13% 15% Complete University 14% 23% 16% 30%

Source: EPH, INDEC.

Labor market conditions evolved favorably during the period analyzed (Table 2): unemployment rate1_{decreased by 8 percent and informality}2_{by 15 percent. On} the other hand, the activity rate3_{decreased by 2 percent, possibly indicating} some weaknesses in the labor market. The decline in the unemployment rate has been attributed not to a massive creation of jobs, but to a “discouragement effect”, where many of unemployed tired of looking for a work decide to quit (a decrease in the activity rate) coupled with the creation of, often catalogued as unproductive, public sector jobs (IDESA, 2014).

Male economic activity rate decreased 1 point. On the other hand, female economic activity rate decreased its double. Lower activity rates in young are driving the decrease in the activity rate, therefore young females are deciding to stay longer in the educational system, or it is the outcome of one of the biggest welfare plans implemented in the country’s history: the Universal Child Allowance (AUH, for its acronym in Spanish). From the year 2008 onwards the government implemented the AUH, a welfare plan that requires no labor

1_{The unemployment rate is percentage of the total labor force that is unemployed but actively}

seeking employment, it is calculated as the ratio between the unemployed and the active population.

2_{The informality rate is the percentage of the workforce that is not registered in a formal job.}

With the EPH, it is defined as the workers who do not contribute to the formal pension system.

3_{The activity rate is the number of active persons (employed or unemployed) and the total}

(10)

10 counterpart and that the person receiving it does not obatain the formal minimum wage. Parents with children fewer than 18 years old and up to 5 children per household can apply, the only prerequisite is that the child has to be registered in school and vaccinated according to government requisites. It was a plan supported by most part of the society because it covers the basic needs of a huge part of the population and raises the enrolment rates and improves health coverage. But warning sings should rise if the gap between women and men joining the labor markets widens, especially in the segment of young females.

Box 1: The World Economic Forum Global Gender Index

The Global Gender Gap Index ranks countries according to their gender gap and looks to find how well they are dividing their resources and opportunities among men and women. It takes into account the following areas:

 The Economic Participation and Opportunity Index looks at women and men

participation rates in the labor market, the gender wage gap, wage equality for similar work and the ratio of women among legislators, senior officials and mangers an the ratio of women among technical and professional workers.  The Educational Attainment Index measures women and men’s access to

education by looking at ratios in school and literacy rates.

 The Health and Survival Index takes into account sex ratio at birth and the

gap between women and men’s health life expectancy.

 The Political Empowerment: takes into account the ratio of women and men

in minister-level and parliamentary positions.

Argentina is ranked 31st_{in the world with a score of 0.73 out of 1 and among the}

upper middle-income countries is one of the leaders. In the Economic Participation and Opportunity Index it receives a score of 0.631 and it’s ranked 98 in the world. The country’s weakest areas seem to be the wage equality for similar work and the amount of women among legislators, senior officials and managers, but it is doing fairly well in the area of technical and professional workers. In the Educational Attainment Index and Health and Survival Index, the country receives a score of 1, meaning almost full equality between genders in this area. Concerning the Political Empowerment the country gets a score of 0.320 having very few women in ministerial positions or working in the parliament.

(11)

Table 2 shows that the labor market trend is oriented towards higher degrees of formality but still, labor market conditions and performance between men and women are very heterogeneous. Overall, men and women perform quite differently: women participate less in the labor market, have higher unemployment rates, lower employment rates and are subject to higher degrees of informality. Women are also less likely to be business owners and to be formal salaried workers than men. Demand factors and unequal occupational profiles offered by households are probably driving this outcome. The unfavorable insertion conditions for women seem to be partly compensated with a higher participation in the public sector, which is usually associated with higher salaries relative to the private sector.

Table 2: Labor Market Indicators by Gender

2004 2008 2013

Men Women Men Women Men Women

Activity Rate 80% 57% 79% 54% 79% 55%

Unemployment Rate 13% 17% 7% 10% 6% 9%

Employment Rate 69% 47% 73% 49% 74% 50%

Public Sector Worker 12% 20% 12% 19% 13% 21%

Formal Salaried Work 32% 24% 39% 32% 39% 33%

Informal Salaried Work 28% 35% 23% 30% 23% 27%

Self Employed 23% 18% 20% 15% 21% 17%

Boss/Owner 5% 3% 6% 3% 5% 3%

100% 100% 100% 100% 100% 100%

Source: EPH, INDEC.

Changes experienced during the decade in real wages (2013 prices) and hours worked by gender and type of employment are shown on Table 3. On an aggregate level, real wages increased sharply between 2004 and 2008, and continued to increase but this time at a much slower pace until 2013. This is consistent with the fact that after the 2002 economic crisis and devaluation of its currency, the country experienced high economic growth rates due to its quick

(12)

12 recovery, but was later affected by the global slowdown caused by the Great Financial Crisis and since then the economy keeps growing below average. Apart from the external factors, Argentina has also one big internal problem: inflation. The inflation rate for 2013 was 30 percent. The high increase the aggregate level of prices could be also slowing down the real wage growth trend experienced between 2003 and 2008.

The monthly gender wage gap is always positive (Table 3). Varying through out the years and depending on the type of occupation women earn between 15 and 42 percent less than men every month. The highest monthly gender wage gap is observed for informal salaried workers, who earn between 34 and 42 percent less than men. This is a highly unfavorable situation for women, because around one third of the female work force are informal workers. They do not only earn less while working, but they are also in the risk of poverty and vulnerability once they reach the age to retire because of lack of contributions to the pensions system.

Most of the difference in monthly wages can be explained by the fact that women work fewer hours than men in every occupation. Men work on average between 10 and 13 hours more than women every week. On the other hand, in 2013 women spent on average 3 hours more per day than men on unpaid domestic activities. This unequal gender distribution shows the persistence of cultural patterns and gender stereotypes, where the reproductive work was assigned to women and the productive work to men (INDEC, 2014).

(13)

Box 1: Unpaid Domestic Work

 According to INDEC in 2014, 75 percent of the population invested time in

unpaid domestic work.

 Of the total time spent on unpaid domestic work, 76 percent corresponds to

women and 24 percent to men.

 The highest participation and time dedication rate is for women between 30

and 59 years old.

 The highest participation gap between men and women is observed for

individuals between 18 and 29 years old, the lowest for individuals who are 60 or older.

 Women who are either married or coupled spend the most time doing unpaid

domestic work. When couples divorce, men’s participation increases, while women’s decreases.

 The highest participation rate is observed for women who have children that

are 6 years old or younger. Having kids increases 1.6 percent the amount of time spent by men and 4.4 percent the time spent by women per day.

Correcting the monthly wage by the amount of hours worked yields the hourly wage gap (Table 3). Women who are public sector workers or formal salaried employees are in a more or less equal economic condition as their male peers, and in some periods even better, and this has remained constant through out the years. On the other hand, women who are subject to labor informality, self employed or business owners are always in a more unfavorable condition than their male counterparts, and their relative position fluctuates a lot more.

A striking feature is that not only women who are inserted into the informal labor segment (and one would expect less educated and coming for poorer households) are exposed to gender inequality, but also women who are business owners or bosses at a company are in somewhat similar situation. This means that the type and quality (size, sector, degree of informality) of company that men and women work in varies significantly.

(14)

Table 3: The Gender Wage Gap

2004 2008 2013

Monthly Wage Men Women Difference Men Women Difference Men Women Difference

Public Sector Worker 3695 2591 30% 9303 7867 15% 8794 7216 18%

Formal Salaried Work 4343 3434 21% 9246 7801 16% 8479 6926 18%

Informal Salaried Work 2032 1345 34% 4718 2724 42% 4724 2815 40%

Self Employed 3003 1904 37% 6626 4467 33% 5465 3904 29%

Boss/Owner 6347 4647 27% 12561 10430 17% 8870 7541 15%

Hours worked Gap

Public Sector Worker 41 28 13 43 29 14 40 34 14

Formal Salaried Work 52 38 14 48 38 10 46 38 8

Informal Salaried Work 45 33 12 44 28 16 44 33 16

Self Employed 51 45 6 47 44 3 47 34 13

Boss/Owner 58 47 11 52 47 5 49 58 5

Hourly Wage

Public Sector Worker 23 23 -3% 54 68 -25% 55 53 3%

Formal Salaried Work 21 23 -8% 48 51 -7% 46 46 1%

Informal Salaried Work 11 10 10% 27 24 9% 27 21 21%

Self Employed 15 11 28% 35 25 28% 29 29 1%

Boss/Owner 27 25 10% 60 55 8% 45 33 28%

(15)

III. Methodology

:

The Minceran Wage Earning Model

The earning functions are calculated by estimating the Minceran rates of return (1974). The relationship between earnings and schooling can be estimated in the following way:

ln 𝑊_𝑖 = 𝛼 + 𝛽₁𝑆_𝑖+ 𝛽₂𝑋_𝑖+ 𝛽₃𝑋2 𝑖 + 𝜇𝑖

Where LnW is the natural logarithm of the earnings for the ith individual; S are the years of schooling, X is the labor market experience; 𝑋2_{is experience} squared, and 𝜇 is a random disturbance that reflects unobserved abilities. The 𝛽₁ coefficient measures the rate of return of each additional year of schooling. The continuous years of schooling variable can be converted into a series of dummy variables that represent each levels of schooling:

𝐿𝑛𝑊_𝑖 = 𝛼 + 𝛽₁𝑃𝑟𝑖𝑐_𝑖+ 𝛽₂𝑆𝑒𝑐𝑖_𝑖+ 𝛽₃𝑆𝑒𝑐𝑐_𝑖+ 𝛽₄𝑈𝑛𝑖𝑖_𝑖+ 𝛽₅𝑈𝑛𝑖𝑐_𝑖+ 𝛽₆𝑋_𝑖+ 𝛽₇𝑋_𝑖2+ 𝜇_𝑖

Where Pric, Seci, Unii and Unic denote dummy variables for complete primary school, incomplete secondary school, complete secondary school, incomplete university and complete university. The variable Pric is equal to 1 if a person has achieved 7 years of schooling, Seci is equal to 1 if an individual has achieved between 8-11 years of schooling, Secc equals 1 if 12 years of schooling have been achieved, Unii is 1 if 13-16 years have been completed of schooling and Unic is 1 if the individual has achieved 17 years or more. The returns to schooling by level are:

𝑟(𝑃𝑟𝑖𝑐) = 𝛽1 𝑆_{𝑃𝑟𝑖𝑐} ⁄

(16)

16 𝑟(𝑆𝑒𝑐𝑖) = (𝛽2− 𝛽1) (𝑆𝑆𝑒𝑐𝑖− 𝑆𝑃𝑟𝑖𝑐) ⁄ 𝑟(𝑆𝑒𝑐𝑐) = (𝛽3 − 𝛽1) (𝑆_{𝑠𝑒𝑐𝑐𝑐}− 𝑆_{𝑃𝑟𝑖𝑐}) ⁄ 𝑟(𝑈𝑛𝑖𝑖) = (𝛽4− 𝛽3) (𝑆_{𝑈𝑛𝑖𝑖}− 𝑆_{𝑆𝑒𝑐𝑐}) ⁄ 𝑟(𝑈𝑛𝑖𝑐) = (𝛽5− 𝛽2) (𝑆𝑈𝑛𝑖𝑐 − 𝑆𝑆𝑒𝑐𝑐) ⁄

Where S represents the number of years of schooling for each level of education (7, 10, 12, 15 and 17 years of education respectively). This model assumes that the difference in wages among workers with different levels of education is constant through the period and that the only cost of keeping studying is the sacrificed earnings during that period (Giovagnoli et al., 2005).

The earning functions for both men and women are firstly estimated using OLS. However, three issues usually arise when trying to establish the relationship between education and earnings, these are: sample selection, omitted variables bias and measurement error bias. Sample selection bias occurs because only earnings for those employed are observed, but not for those who are not participating the labor market. Therefore, workers receiving a wage are a non-random draw from the population. In this context, it could possibly imply that more educated female workers may be self-selecting into the labor market. The result is that the conditional mean for the subsample exceed the mean of the whole distribution, making OLS estimates inconsistent (Mondino and Montoya, 2000). The usual way the literature has dealt with this issue is by applying Heckman’s Two Step Procedure. Therefore, the returns to schooling will also be estimated using this procedure.

(17)

The second problem to be dealt with is omitted variables. The disturbance term captures unobservable individual effects, the classic example is unobserved ability, and these may influence the schooling decision and create a correlation between education and the error term in the Mincerean earning function (Harmon et al., 2000). Unobservable ability determines both schooling and earnings, causing endogeneity in the earnings function and thus yielding biased estimates with OLS. To remove the pernicious effect of omitted variable bias, a panel data fixed effect approach to estimate the returns to schooling is used as well. Finally, in order to estimate the identify differences in the returns across the whole distribution the retruns will be calculated using Quantile Regression Analysis following Giovagnoli et al., 2005. This methods allows to test weather education is independent or not of the unobservable variables, and if it is not, if it complements or compensated them.

IV. Data and Descriptive Statistics

Data from the Argentine Continuous Permanent Household Survey or Encuesta

Permanente de Hogares carried out by the National Institute of Statistics and

Census (INDEC) will be used. The EPH is conducted quarterly in major urban centers in the country every three months. The design of the EPH is a stratified sample with rotating panel, where 25 percent of the sample is replaced each round. The survey includes 18,000 households, representing 31 urban centers and approximately 70 percent of the urban population of the country.

The EPH gathers information on the individual situation and personal characteristics as well as information on employment status taking as reference the week before the month preceding the survey. In terms of personal,

(18)

18 demographic and economic information on each household member: employment status (employed, unemployed or inactive), relationship to head of household, age, gender, marital status, hours worked in the past week, task type, company size and sector of activity, type and amount of income, education level, number of children, hours worked, etc. People who work as employees respond to the type of benefits they receive making it possible to establish the formality of the labor relationship (Montoya and Giordano, 2012).

To calculate the Mincerean Earning Functions, data from all second trimesters between 2004 and 2013 is used. From the third trimester of 2006 onwards, 3 new conglomerates were introduced, therefore for comparability reasons, only conglomerates available for all years are considered4_{. Earning functions are} calculated for men and women first jointly and then separately. The sample includes all workers between 14 and 65 years. The definition of the variables used are presented in Appendix 1a and the means and standard deviations in Appendix 2a.

V. Econometric Results

A. Ordinary Least Square Results

The OLS Mincerean earning functions for 2004-2013 are presented in the Appendix. The earning functions is a by levels function, including one dummy variable for each educational level. Other variables are included as controls, these are: public employment, self employment and formal employment

4_{San Nicolás-Villa Constitución, Viedma-Carmen de Patagones and Rawson-Trelew}

(19)

dummies and 5 dummies that represent 5 out of 6 of the Argentine Regions (Gran Buenos Aires, North West, North East, Cuyo, Pampa and Patagonia). The first column shows the estimates for the whole sample (includes a binary variable indicating sex) and the next two columns show the estimates for men and females respectively.

The coefficients on educational levels are always positive and increase with each level of education for both genders, indicating a convex relationship between education and earnings. This finding is in line with Galiani and Sanguinetti (2003), Fiszbein, Patrinos and Giovagnoli (2005) and Gasparini and Acosta (2007). Table 3 shows the evolution of the return of an additional year of schooling over time. The returns to schooling for primary school level are on average 1.5 for the whole decade, this is not surprising considering the universality of primary education and the presence of a highly educated workforce. Workers got on average 5.1 higher earnings percent for each additional year of high school, and the returns to education to university are the highest: 10.3 percent for the whole period.

Overall, the returns to schooling have decreased for the whole period analyzed, for all levels and all genders: the overall returns to schooling were 8.4 in 2004 and this number has dropped to 6.1 in 2013. However the decline experienced by male workers with university level has been much sharper than for the rest: its returns decreased by 4.9 percent during the decade. On the other hand, women with university level, experienced a decrease of 1 percent. Neumeyer (2014) and Levy Yeyati (2013) have already reported evidence on the decline of the returns to education during the 2000s. They attribute this behavior to the decline of wage inequality experienced by the country during the decade: the

(20)

20 difference between the wages of the most educated and the least educated fell. The decline is explained by a decrease in the demand for educated workers, rather than by an increase in supply, which is the opposite of what has been happening in the developed world where the returns to education are increasing. In an analysis by gender, it can be seen that in 2004 men experienced higher returns for all levels of education, this is in line with the findings of Fiszbein, Patrinos, and Giovagnoli et al. (2005) and López Boo (2010) for the early 2000s. Overall returns to schooling were almost 1 percent higher for men than for women. The trend appears to have reversed after 2006 and by 2013 the returns for male workers were 2 percent lower. Women now perceive higher returns to education for middle (5 percent) and high education (9 percent), but still observe lower returns for primary school (0.1 percent).

In conclusion, I find that: (i) overall returns to schooling, and for each level of education, have decreased in the period analyze as reported by Numeyer (2014) and Levy Yeyati (2013); and (ii) women used to have lower returns to schooling at every level as stated by López Boo (2010) and Fizbein, Patrinos and Giovagnoli (2005) during the early 2000s, but the trend has reversed for middle and higher education.

B. Heckman Two-Step Procedure: Controlling for Sample Selection

In order to control for sample selection bias, the earnings function is also estimated using Heckman’s (1979) two-step procedure. Sample selection bias arises when working with individuals participating in the labor market and this could be a non-random draw form the entire population. To correct for selection bias, the Heckman method specifies a preliminary equation with a binary

(21)

dependent variable indicating whether the worker is in or out of the labor force, and then treats this equation and the earning’s function simultaneously.

The model assumes that there are is an underlying regression relationship: ln 𝑊𝑖 = 𝛽𝑋𝑖+ 𝜇1𝑖

This is the regression equation. However, the dependent variable, is not always observed. It depends on the selection equation:

𝛾_𝑖𝑍 + 𝜇_2𝑖 > 0

The selectivity corrected earnings functions include the standard variables, these are: education, age, age squared, public employment, self-employment and formal employment dummies and the 5 regional dummies. Two household demographic variables are included as restrictions: Single equal to 1 if the household head is not married and Qkids equal to the number of children in the household. These variables are believed to determine the decision to participate in the labor force, but not to have direct influence on wages. All are individually statistically significant. The earnings functions are presented in the appendix, while the results for the returns to schooling are given in Table 3.

Correcting for selection produces higher returns for females and lower returns for male workers and the returns to schooling also decrease during the whole period analyzed. Women still have higher returns to education for secondary and university level and lower returns for primary school.

(22)

Table 4: Evolution of the returns to schooling by gender Whole Sample 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Overall 8,39% 8,39% 8,78% 8,04% 7,80% 7,74% 8,02% 7,35% 6,38% 6,09% Primary 1,16% 1,17% 1,94% 1,59% 1,99% 1,41% 1,34% 1,64% 1,71% 0,69% Secondary vs. Primary 5,80% 5,36% 5,80% 5,62% 4,60% 4,32% 5,74% 4,94% 4,76% 4,00% University vs Secondary 10,98% 11,42% 11,76% 10,46% 11,00% 11,16% 10,30% 9,76% 8,00% 8,18% Men Overall 8,83% 9,03% 8,65% 7,58% 7,18% 7,21% 7,55% 6,61% 5,74% 5,31% Primary 1,39% 1,67% 2,04% 2,34% 2,60% 1,59% 2,01% 1,66% 2,11% 0,94% Secondary vs. Primary 5,84% 4,96% 6,26% 4,82% 3,66% 4,34% 5,46% 4,34% 4,40% 3,70% University vs. Secondary 11,82% 13,10% 11,04% 10,34% 10,70% 10,08% 9,64% 8,88% 7,08% 6,92% Women Overall 8,06% 7,94% 8,85% 8,77% 8,55% 8,34% 8,46% 8,27% 7,24% 7,07% Primary 0,94% 0,50% 1,86% 0,34% 0,87% 1,10% 0,17% 1,59% 0,99% 0,14% Secondary vs. Primary 5,78% 5,98% 5,20% 7,32% 6,36% 4,60% 6,14% 6,24% 5,60% 4,78% University vs. Secondary 10,34% 9,90% 12,50% 10,22% 10,74% 12,08% 10,78% 10,30% 8,88% 9,36%

Men corrected for selectivity 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

Overall 7,75% 7,91% 7,75% 7,37% 6,56% 6,35% 6,55% 5,87% 5,26% 4,88%

Primary 1,61% 1,90% 1,70% 2,31% 1,36% 1,60% 1,89% 1,74% 1,70% 1,40%

Secondary vs. Primary 5,04% 4,48% 5,80% 5,36% 4,10% 4,30% 4,56% 3,88% 3,94% 3,30%

University vs. Secondary 10,46% 11,34% 9,70% 9,38% 9,02% 8,40% 8,54% 7,86% 6,58% 6,46%

Women corrected for selectivity 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

Overall 8,3% 9,7% 10,3% 9,7% 9,9% 10,2% 7,5% 9,6% 9,5% 8,8%

Primary 1,5% 1,7% 2,4% 1,7% 2,7% 2,4% 0,8% 2,4% 1,5% 2,1%

Secondary vs. Primary 5,8% 7,0% 7,1% 7,3% 7,6% 7,3% 5,5% 7,4% 7,6% 6,7%

University vs. Secondary 10,9% 12,5% 13,6% 12,1% 12,1% 13,0% 9,4% 11,9% 11,5% 10,8%

(23)

The Mills ratio or the selection term is positive and highly significant in the women’s equation, indicating a positive selection bias in women’s employment. This suggest that those women who decide to participate in the labor market get higher wages than a random drawing from the population of women with a comparable set of characteristics would get. This is in line with the increasing levels of women’s education evidenced in the descriptive statistics.

C. Panel Data: Fixed Effects

As previously mentioned, one of the key problems when working with Mincerean retruns to schooling is that it does not account for endogeneity in the schooling variable. This is, unobservables such as ability, skills and motivation could be correlated both to schooling and wages, therefore biasing the coefficients used to calculate the returns to schooling.

Following Harmon et al. (2000), the literature has dealt with this issue in three ways. The first way is by incorporating a proxy that measures ability (for example by including IQ scores) in order to account for the unobserved effects. By including a measure of ability the coefficient of education should be smaller, since ability is already controlled for. Some studies, like Lackburn and Neumark (1993) have confirmed that OLS estimates are indeed upwards biased. A second way to deal with omitted variables is by using instrumental variables that affects schooling but not wages, like the quarter of birth of birth of the individual or the distance to school. However, this method has been argued to produce poor estimators due to the use of weak instruments. The third way is the use of panel data fixed effect approach where the unobserved individual effect is considered time-invariant. Some of the downsides of using panel data is the panel data often suffers from issues like attrition and that representing a fraction of the labor force because data is costly to collect (Himaz and Aturupane, 2012). When lacking individual panel data the literature has resorted to the

(24)

24 use of within-twins or within-siblings differences, under the assumption that ability, skills and motivation are common between them.

In order to have an alternative to the Mincerean approach using cross-sectional data, in this section the results to schooling are calculated using a panel data fixed-effect approach. The standard Mincerean earning function can be specified for periods of time t and individual i as follows:

ln 𝑊_𝑖𝑡 = 𝛼 + 𝛽₁𝑆_𝑖𝑡+ 𝛽₂𝑋_𝑖𝑡+ 𝛽₃𝑋2

𝑖𝑡+ 𝛼𝑖𝑡+ 𝜇𝑖𝑡

Where once again, LnW is the natural logarithm of the earnings for the ith individual for period t; S are the years of schooling for individual i at time t, X is the labor market experience for individual i at time t; 𝑋2_{is experience squared for individual i at time t,} and 𝜇 is a random disturbance. The term 𝛼_𝑖𝑡 reflects unobserved individual variables such ability and motivation. If 𝛼_𝑖𝑡 is not correlated with 𝑆_𝑖𝑡 and 𝑋_𝑖𝑡, then OLS estimators are consistent. However, it is quite likely that unobservables are correlated to both schooling and experience causing OLS estimators to be upward biased. The 𝛼𝑖𝑡 intercepts can be treated as an individual fixed effect.

The fixed effects methods solves the omitted variable issue that result from unobserved heterogeneity. The panel is built by incorporating all trimesters between 2004 and 2013 for the EPH5_{. The results of the earnings function for the whole sample, men and women} are presented in the appendix. Table 5 shows the returns to schooling.

(25)

Table 5: The Retruns to Schooling: Panel Fixed Effect Approach

Whole Sample Men Women

Overall 0,59% 0,57% 0,61%

Primary vs none 0,40% 0,39% 0,41%

Secondary incomplete vs Primary 0,57% 0,53% 0,63%

Secondary vs Primary 0,90% 0,90% 0,90%

University incomplete vs Secondary 0,73% 0,73% 0,73%

University vs Secondary 1,12% 1,04% 1,18%

Results calculated by OLS indicated that the returns to schooling for the period were 7.7 percent for the whole sample, while the fixed effects approach yields results that are much smaller, around 0,6 percent. This suggests that unobervables such as ability, skills and motivation are upward biasing the returns to education in cross section analysis. The results confirm the trend shown by cross sectional estimates, where women have now higher returns to schooling than men.

D. Quantile Regression Analysis

The returns to schooling using OLS estimate the effect of education on wages, assuming that the return to education is common among different individuals. To estimate the effect of schooling on income at different places of the wage distribution, the Quantile Regression method is used. The distribution of wages reflects not only education, but could also be reflecting some unobservable factors like ability and skills. The individuals at the end of the distribution are probably less educated but also probably have less endowment of unobservable skills. An interesting point is if education is independent of unobservable skills or if complements or compensates the. If education is independent of the unobserved ability and skills, then the returns to education should be the same through out the distribution. If education complements ability then those at the top of the distribution should have higher returns, on the contrary if it compensates the opposite should be true.

(26)

26 Following Giovagnoli et al. (2005) the Quantile Regression model is:

𝑙𝑛𝑊_𝑖 = 𝑋_𝑖𝛽_𝜃+ 𝜇_𝜃𝑖

𝑋_𝑖𝛽_𝜃 = (𝑄𝑢𝑎𝑛𝑡𝑖𝑙𝑒 𝜃)(𝑙𝑛𝑊_𝑖⁄ ) 𝑋_𝑖

Where 𝑋𝑖 is a vector of a vector of exogenous variables, 𝛽0 is the vector of parameters and (𝑄𝑢𝑎𝑛𝑡𝑖𝑙𝑒 𝜃)(𝑙𝑛𝑊𝑖/𝑋𝑖) is the θth (0< θ<1) conditional quantile of lnw given X. The θth quantile is derived as follows:

𝑀𝑖𝑛 ∑ 𝜌_𝜃(𝑙𝑛𝑤_𝑖 − 𝑋_𝑖𝛽_𝜃)

Where 𝜌_𝜃 is 0ε if ε≥0 and 𝜌_𝜃= (θ-1) ε is ε ≤0. The estimation is done simultaneously by bootstrapping, where the data is resampled and the coefficients remain the same, the results are presented in table 6.

Table 6: Quantile Regressions, returns to education by gender

MALES FEMALES Year q20 q40 q60 q80 q20 q40 q60 q80 2004 5,9% 7,4% 8,3% 9,1% 5,8% 6,8% 6,3% 8,5% 2005 6,2% 7,3% 8,1% 9,2% 5,6% 6,8% 7,6% 7,8% 2006 6,4% 7,2% 7,9% 8,8% 7,4% 8,4% 8,2% 8,6% 2007 6,0% 6,8% 7,8% 8,4% 6,8% 7,6% 8,2% 8,7% 2008 5,3% 6,3% 7,0% 7,9% 7,2% 7,3% 7,6% 8,2% 2009 5,4% 6,2% 6,6% 7,2% 7,0% 7,5% 7,7% 8,0% 2010 5,9% 6,0% 6,7% 7,3% 7,0% 7,6% 7,8% 7,8% 2011 4,9% 5,2% 6,0% 6,7% 6,6% 7,3% 8,7% 7,8% 2012 4,7% 5,0% 5,8% 6,5% 6,5% 6,7% 7,2% 7,4% 2013 4,1% 4,7% 5,1% 5,7% 6,1% 6,1% 6,6% 7,0% Source: EPH.

There are significant differences between those at the bottom of the income distribution and those at the top. Women now have higher overall returns to schooling than men and the returns of schooling have been decreasing over time, as the OLS and Selectivity models indicated. It can be observed that the decline is steeper for men in the highest quantiles (3 points) than for men in the lowest quantiles (1.9 points). The OLS results

(27)

showed the biggest decline for men with university level education, so there is preliminary evidence for complementarity between education and unobserved variables. Women’s returns on the on the other hand, have fluctuated less and the biggest decline is observed for the highest quantiles (1.6 points). The big declines in the returns to schooling for the highest quintiles could be a plausible explanation for the decline in income inequality experienced in the country in the last 15 years.

For both men and women returns are higher towards the end of the distribution indicating complementarity between education and unobserved ability and skills. This means that the expansion of the schooling system could raise inequality, however through the years the difference between returns of those at the top and at the bottom has decreased, which goes in line with the observed reduction in inequality. Higher returns towards the end of the income distribution are also found in many other countries like the US and UK (Martins and Pereira 2004).

E. Further Assessment: Does The labor market discriminate against women?

In this section the factors that explain the differences in male and female labor outcomes are explored. The most widely used methodology to assess the determinants of mean wage differences between to groups of workers (sex, race, ethnicity, etc.) is the Blinder-Oaxaca decomposition.

Through the Blinder-Oaxaca method the wage gap can be decomposed as follows: 𝑌_𝑖 = 𝛽_𝑖𝑋_𝑖

Where i=male, female. X is the vector of average characteristics of i and 𝛽 is the vector of estimates parameters of i.

(28)

28 𝑀𝑒𝑎𝑛 𝑊𝑎𝑔𝑒 = 𝛽𝑀𝐸(𝑌𝑀) − 𝛽𝐹𝐸(𝑌𝐹)

𝑀𝑒𝑎𝑛 𝑊𝑎𝑔𝑒 = {𝐸(𝑋_𝑀) − 𝐸(𝑋_𝐹)}′_𝛽

𝐹+ 𝐸(𝑋𝐹)´ (𝛽𝑀− 𝛽𝐹) + {𝐸(𝑋𝑀) − 𝐸(𝑋𝐵𝐹)}´(𝛽𝑀− 𝛽𝐹) 𝑀𝑒𝑎𝑛 𝑊𝑎𝑔𝑒 = 𝐸 + 𝐶 + 𝐼

The first term, denoted by E is the “endowments effects”, and accounts for the difference explained by the predictors. The endowments effect measures the expected change in female’s mean outcomes if they had male predictor levels. The second term, denoted by C is the “coefficients effect” and measures the contribution of differences in the coefficients. This is, it measures the expected change in female outcomes if females had male coefficients. And I stands for the “interaction effect”. The results of the decomposition are reported in Table 7.

The wage gap is always positive but not statistically significant every year; the endowment effect always negative and the coefficients always positive and both statistically significant. This is, for example, for 2006 the mean log of wages is 2.87 for men and 2.84 for women, yielding a wage gap of 0.04 in favor of men. The decomposition shows that if women had the same characteristics as men in terms of education, work experience, etc. they should see their wages decrease by 0.098 points. And if women had their characteristics and men’s coefficients, then their wages should increase by 0.15. This results show that even though now women have higher returns to schooling than men there is some evidence that there is labor market discrimination, however due to the possible but in the presence of unobservable variables, ability and skills among them, it cannot be affirmed. This analysis could be interesting for future lines of research to check if there is indeed labor market discrimination.

(29)

Table 7: Oaxaca Decomposition 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Male 2.488*** 2.616*** 2.874*** 3.090*** 3.359*** 3.546*** 3.752*** 4.007*** 4.272*** 4.507*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Females 2.448*** 2.602*** 2.825*** 3.049*** 3.317*** 3.538*** 3.742*** 3.993*** 4.262*** 4.496*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) difference 0.040** 0.014 0.049*** 0.042*** 0.042*** 0.009 0.010 0.014 0.010 0.011 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) endowments -0.098*** -0.092*** -0.122*** -0.138*** -0.150*** -0.142*** -0.154*** -0.155*** -0.141*** -0.141*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) coefficients 0.119*** 0.099*** 0.150*** 0.156*** 0.160*** 0.125*** 0.145*** 0.137*** 0.117*** 0.124*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) interaction 0.019** 0.007 0.021*** 0.023*** 0.031*** 0.025*** 0.020*** 0.031*** 0.034*** 0.028*** (0.01) (0.01) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)

(30)

VI. Conclusion

Women in the Argentine labor market are at disadvantage in comparison to men: they participate less, have higher unemployment rates, are subject to higher degrees of informality and earn less in many sectors of the labor market. This study tries to find out if differential returns to schooling for women than for men in the Argentine labor market explain the gender wage gap.

Three methods are used to consistently estimate the returns to schooling: i) Ordinary Least Squares; ii) Heckman Two-Sept Procedure and iii) Quantile Regression Analysis. OLS and selectivity corrected results point that the returns to education for an additional year of schooling are between 4.8 and 5.3 percent for men and between 7 and 8.8 percent for women in 2013. This signals a reverse of the trend of lower returns to schooling perceived by women in the 90s shown by Giovagnoli et al. (2005) and López Boo (2010), but again in line with Psacharopoulos (1994) that reported higher returns to education for women in the 80s. In addition to the gender differences in returns to schooling, it is found the overall returns to schooling have been falling in time, consistent with the fact that income inequality has declined as well. To account for omitted variables returns to schooling were calculated using a Panel Fixed Effect approach. This methodology yields results that are much smaller, around 0,6 percent. This suggests that unobervables such as ability, skills and motivation may be upward biasing the returns to education in cross section analysis. The results confirm the women have higher returns to schooling.

(31)

In addition to Mincerean Earning Functions, the Quantile Regression estimates for the same period confirm the previous results: returns to education have been decreasing throughout the period and women have now higher returns to schooling than men. Also, both men and women in the higher quantiles have higher returns compared to those in the lowest quantiles, showing complementarity between unobserved ability and education as reported in many other countries.

Overall, the results suggest that women do not earn less than men because the labor market rewards them less, but because there is some other form of labor market discrimination, or simply because they choose they work less and in more precarious and lower paid jobs in order to take care of domestic activities. . To further assess on this issue, the Oaxaca-Blinder decomposition method was used. The results suggest some elements of potential gender discrimination, in the form of pregnancy and maternity discrimination for example. Still, caution is needed when interpreting these results, since unobserved characteristics of both males and females may not be controlled for. The existence of gender discrimination is an interesting topic for future lines of research.

(32)

References:

Andres, L. (2003): “Private Education versus Public Education: The Effects of the Quality of the Alternative Schooling in Argentina,” Discussion paper, University of Chicago, unpublished manuscript.

Aslam, Monazza (2005): “Rates of Return to Education by Gender in Pakistan”, GPRG-WPS-064, Global Poverty Research Group.

Barro R.J. and J.-W. Lee. (2000): “International Data on Educational Attainment: Updates and Implications.” Harvard University, Center for International Development Working Paper No. 42.

Barro R.J. and J.-W. Lee. (2013): “A new data set of educational attainment in the world, 1950–2010”. Journal of Development Economics. Volume 104, September 2013, Pages 184–198.

Rojo Brizuela S. and Tumini L. (2008): “Inequidades de género en el mercado de trabajo de la Argentina: las brechas salariales”. Observatorio de Empleo y Dinámica Empresarial (OEDE), Revista de Trabajo.

DiPrete T. and Buchmann C. (2013): “The Rise of Women: The Growing Gender Gap in Education and What It Means for American School”. Russell Sage Foundation.

Duryea, S., A. Cox-Edwards, y M. Ureta. 2001. Women in the LAC Labor Market: The Remarkable 1990s. Document presented at “Women at Work: A Challenge for Development”, Interamiercan Developement Bank.

Fiszbein, A., H. A. Patrinos, and P. I. Giovagnoli (2005): “Estimating the returns to education in Argentina: 1992-2002,” Discussion paper, The World Bank, Policy Research Working Paper Series: 3715 2005.

Galiani, S., and P. Sanguinetti (2003): “The impact of trade liberalization on wage inequality: evidence from Argentina,” Journal of Development Economics, 72, 497–513.

Gasparini, L., and P. Acosta (2007): “Capital Accumulation, Trade Liberalization and Rising Wage Inequality: The Case of Argentina,” Economic Development and Cultural Change, 55(4), 793–812.

Gasparini, L., M. Marchionni, and W. Sosa Escudero (2001): “Characteriza- tion of inequality changes through microeconometric decompositions: The case of Greater Buenos Aires,” in the microeconomics of income distribution dynamics: A comparative analysis of selected developing countries. IADB. Washington DC.

(33)

Harmon, C., Oosterbeek, H. and Walker, I. (2000): ·The Returns to Education: A Review of Evidence, Issues and Deficiencies in the Literature”, London School of Economics.

Heckman, J. J. (1979): “Sample Selection Bias as a Specification Error,” Econometrica, 47(1), 153–61.

International Labour Organization (2015): “Global Wage Report 2014/15, Wages and income inequality”. International Labour Office, CH-1211 Geneva 22, Switzerland.

López Boo, F. (2010): “Returns to Education and Macroeconomic shocks: Evidence from Argentina”, Discussion paper series, IZA DP No. 4752.

Margot, D. (2001): “Rendimiento a la educacion en Argentina: un analisis de cohortes,” Discussion paper, Documento de trabajo numero 33, Universidad Nacional de la Plata.

Martins, P.S. and P.T. Pereira. 2004. “Does Education Reduce Wage Inequality? Quantile Regressions Evidence from 16 Countries.” Labour

Economics 11(3): 355-371.

Mincer, J. (1974): “Schooling, Experience and Earnings,” in Human Behavior and Social Institutions. Columbia University Press, New York.

Mondino G. and Montoya S. (2000): “The effects of labor market regu-lations on employment decisions by firms: Empirical evidence for Argentina”. IADB Research Network Working Paper no. R-391. Washington, D.C.: Inter- American Development Bank.

Montoya S. and Giordano V. (2012): “Immigrants Wage Gap in the Great Buenos Aires Labor Market: How Important Are Differences in Human Capital?”. Facultad de Ciencias Económicam s  Escuela de Economía “Francisco Valsecchi”, Universidad Católica Argentina. Documento de Trabajo N 36.

Pessino C. (1996): “Returns to Education in Greater Buenos Aires 1986-1993 From Hyper- inflation to Stabilization and Beyond,” Cuadernos de Economia (Pontifical CatholicUniversity of Chile), 33(99), 205–26, CEMA, Instituto Universitario, Buenos Aires.

Psacharopoulos, G. (1994): “Returns to Investment in Education: A Global Up- date,” World Development, 22(9), 1325–43.

Psacharopoulos, G., and H. A. Patrinos (2004): “Human Capital and Rates of Return,” in International Handbook on the economics of education. U.K. and Northampton, Mass.

(34)

VII. Statistical Appendix

Table 1a: Description of Explanatory Variables

Variable Definition

Education Maximum level attained

Pric 1 if individual finished primary school, 0 otherwise

Seci 1 if individual did not finish secondary school, 0 otherwise

Secc 1 if individual finished secondary school, 0 otherwise

Unii 1 if individual did not finish university, 0 otherwise

Unic 1 if individual finished university, 0 otherwise

Age Age of the person, continuous form 14 to 65

AgeSquared Age squared

Male 1 if individual is a male, 0 if female

Labor Market Insertion

Public 1 if individual works in the public sector, 0 otherwise

SelfEmp 1 if individual is self-employed, 0 otherwise

FormalEmp 1 if individual contributed to the pension system, 0 otherwise

Regional

GranBA 1 if individual works in Gran Buenos Aires, 0 otherwise

NW 1 if individual works in the North West of Argentina, 0 otherwise

NE 1 if individual works in the North East of Argentina, 0 otherwise

Pampa 1 if individual works in the Pampa Region, 0 otherwise

Personal Characteristics

Single 1 if individual is not married, 0 otherwise

(35)

Table 2a: Descriptive Statistics

_Variables Whole Sample Men Women

Mean SD Mean SD Mean SD

2004 Hourly Wage 15,69 18,25 16,37 20,27 14,79 15,10

Log Hourly Wage 2,47 0,83 2,49 0,85 2,45 0,82

Primary Incomplete 0,06 0,25 0,07 0,25 0,06 0,24 Primary Complete 0,22 0,41 0,24 0,43 0,19 0,39 Secondary Incompelte 0,19 0,39 0,21 0,41 0,16 0,36 Secondary Complete 0,21 0,41 0,21 0,41 0,21 0,41 University Incomplete 0,14 0,34 0,13 0,33 0,15 0,36 University Complete 0,17 0,38 0,13 0,34 0,23 0,42 Age 38,20 12,37 38,20 12,62 38,20 12,04 Age Squared 1612,31 991,41 1.618,27 1.011,91 1.604,45 963,72 Public Employment 0,23 0,42 0,18 0,38 0,30 0,46 Self Employment 0,20 0,40 0,23 0,42 0,16 0,36 Formality 0,40 0,49 0,41 0,49 0,38 0,49 Single 0,46 0,50 0,46 0,50 0,47 0,50 Amount of Children 0,34 0,68 0,37 0,72 0,30 0,64 N 17724 10079 10079 2005 Hourly Wage 17,87 22,16 18,20 22,71 17,42 21,38

Log Hourly Wage 2,61 0,81 2,62 0,81 2,60 0,80

Primary Incomplete 0,07 0,25 0,07 0,26 0,06 0,23 Primary Complete 0,22 0,41 0,24 0,42 0,19 0,39 Secondary Incompelte 0,18 0,38 0,20 0,40 0,14 0,35 Secondary Complete 0,22 0,41 0,22 0,41 0,21 0,41 University Incomplete 0,14 0,35 0,14 0,34 0,15 0,36 University Complete 0,18 0,38 0,13 0,34 0,24 0,43 Age 38,43 12,44 38,29 12,67 38,62 12,13

(36)

36 Age Squared 1.631,19 1.000,96 1.626,79 1.019,22 1.638,73 978,08 Public Employment 0,22 0,42 0,18 0,38 0,29 0,45 Self Employment 0,20 0,40 0,22 0,42 0,16 0,36 Formality 0,41 0,49 0,42 0,49 0,40 0,49 Single 0,48 0,50 0,49 0,50 0,47 0,50 Amount of Children 0,34 0,68 0,38 0,72 0,29 0,63 N 17745 10215 7530 2006 Hourly Wage 22,78 24,44 23,15 24,83 22,29 23,90

Log Hourly Wage 2,85 0,81 2,87 0,79 2,82 0,84

Primary Incomplete 0,06 0,24 0,07 0,25 0,05 0,23 Primary Complete 0,21 0,40 0,23 0,42 0,17 0,38 Secondary Incompelte 0,18 0,38 0,20 0,40 0,15 0,36 Secondary Complete 0,22 0,42 0,23 0,42 0,22 0,41 University Incomplete 0,14 0,35 0,14 0,34 0,15 0,36 University Complete 0,18 0,38 0,13 0,34 0,25 0,43 Age 38,29 12,52 38,18 12,76 3.844 1.217,81 Age Squared 1.622,96 1.004,80 1.620,83 1.024,03 1.626 978,36 Public Employment 0,22 0,41 0,17 0,38 0,28 0,45 Self Employment 0,19 0,39 0,21 0,41 0,15 0,36 Formality 0,43 0,50 0,44 0,50 0,42 0,49 Single 0,49 0,50 0,50 0,50 0,49 0,50 Amount of Children 0,33 0,67 0,37 0,72 0,27 0,60 N 18628 10688 7940 2007 Hourly Wage 28,03 28,70 28,23 28,21 27,74 2,94

Log Hourly Wage 3,06 0,81 3,07 0,79 3,04 0,85

Primary Incomplete 0,06 0,23 0,07 0,25 0,05 0,21

Primary Complete 0,20 0,40 0,22 0,42 0,16 0,36

Secondary Incompelte 0,18 0,39 0,21 0,41 0,15 0,35

(37)

University Incomplete 0,14 0,35 0,125 0,330 0,16 0,37 University Complete 0,18 0,39 0,126 0,332 0,26 0,44 Age 38,24 12,44 38,15 12,72 38,36 12 Age Squared 1.616,65 997,11 1.617 1.020 1.616,38 965 Public Employment 0,22 0,41 0,17 0,38 0,277 0,448 Self Employment 0,17 0,38 0,20 0,40 0,139 0,346 Formality 0,46 0,50 0,47 0,50 0,460 0,498 Single 0,51 0,50 0,52 0,50 0,493 0,500 Amount of Children 0,31 0,64 0,35 0,68 0,254 0,568 N 22867 13284 9583 2008 Hourly Wage 36,39 35,18 36,52 34,73 36,21 35,79

Log Hourly Wage 3,33 0,80 3,35 0,76 3,31 0,85

Primary Incomplete 0,054 0,226 0,06 0,24 0,04 0,20 Primary Complete 0,191 0,393 0,22 0,41 0,15 0,36 Secondary Incompelte 0,180 0,384 0,20 0,40 0,15 0,35 Secondary Complete 0,251 0,433 0,26 0,44 0,24 0,43 University Incomplete 0,135 0,342 0,12 0,33 0,15 0,36 University Complete 0,186 0,389 0,13 0,33 0,26 0,44 Age 38,47 12,44 38,33 12,72 38,67 12,04 Age Squared 1.634,78 1.002,33 1.630,53 1.024,88 1.640,59 970,69 Public Employment 0,21 0,41 0,17 0,37 0,27 0,45 Self Employment 0,17 0,38 0,20 0,40 0,14 0,35 Formality 0,48 0,50 0,48 0,50 0,48 0,50 Single 0,51 0,50 0,52 0,50 0,50 0,50 Amount of Children 0,29 0,62 0,33 0,66 0,24 0,55 N 22419 12946 9473 2009 Hourly Wage 45,29 66,22 45,33 77,49 45,23 46,34

Log Hourly Wage 3,53 0,80 3,53 0,78 3,52 0,84

(38)

38 Primary Complete 0,18 0,39 0,21 0,40 0,15 0,36 Secondary Incompelte 0,18 0,38 0,21 0,41 0,14 0,35 Secondary Complete 0,25 0,43 0,26 0,44 0,24 0,43 University Incomplete 0,14 0,34 0,13 0,33 0,15 0,36 University Complete 0,19 0,39 0,14 0,34 0,27 0,44 Age 38,57 12,31 38,30 12,54 38,94 11,99 Age Squared 1.639,26 997,02 1.624,20 1.015,31 1.659,98 970,97 Public Employment 0,22 0,42 0,18 0,39 0,28 0,45 Self Employment 0,18 0,38 0,20 0,40 0,15 0,35 Formality 0,49 0,50 0,50 0,50 0,49 0,50 Single 0,51 0,50 0,53 0,50 0,49 0,50 Amount of Children 0,29 0,61 0,33 0,65 0,25 0,55 N 21563 12488 9075 2010 Hourly Wage 54,96 56,14 54,82 54,06 55,16 58,96

Log Hourly Wage 3,74 0,80 3,74 0,78 3,73 0,82

Primary Incomplete 0,05 0,21 0,06 0,23 0,04 0,19 Primary Complete 0,18 0,38 0,20 0,40 0,15 0,36 Secondary Incompelte 0,18 0,38 0,21 0,41 0,14 0,34 Secondary Complete 0,25 0,43 0,27 0,44 0,24 0,42 University Incomplete 0,14 0,34 0,13 0,33 0,15 0,36 University Complete 0,20 0,40 0,14 0,35 0,29 0,45 Age 38,49 12,25 38,27 12,48 38,81 11,93 Age Squared 1.631,91 994,59 1.620,29 1.011,80 1.648,27 969,67 Public Employment 0,22 0,41 0,18 0,38 0,27 0,44 Self Employment 0,18 0,38 0,20 0,40 0,14 0,35 Formality 0,49 0,50 0,49 0,50 0,49 0,50 Single 0,53 0,50 0,54 0,50 0,51 0,50 Amount of Children 0,30 0,61 0,33 0,64 0,26 0,55 N 21310 12458 8852

(39)

2011 Hourly Wage 70,80 87,75 70,40 89,25 71,35 85,62

Log Hourly Wage 3,99 0,79 4,00 0,76 3,99 0,83

Primary Incomplete 0,04 0,21 0,05 0,22 0,03 0,18 Primary Complete 0,17 0,38 0,19 0,39 0,14 0,35 Secondary Incompelte 0,18 0,38 0,21 0,41 0,14 0,34 Secondary Complete 0,26 0,44 0,27 0,44 0,24 0,43 University Incomplete 0,14 0,34 0,13 0,33 0,15 0,36 University Complete 0,21 0,41 0,15 0,35 0,29 0,45 Age 38,54 12,31 38,26 12,50 38,92 12,01 Age Squared 1.636,38 999,91 1.620,19 1.015,82 1.658,86 977,00 Public Employment 0,22 0,41 0,18 0,38 0,28 0,45 Self Employment 0,17 0,38 0,19 0,39 0,14 0,35 Formality 0,51 0,50 0,52 0,50 0,51 0,50 Single 0,54 0,50 0,55 0,50 0,52 0,50 Amount of Children 0,30 0,62 0,34 0,66 0,26 0,56 N 21629 12572 9057 2012 Hourly Wage 90,08 81.886,00 89,16 81,32 91,33 82,63

Log Hourly Wage 4,26 0,77 4,26 0,72 42,51 0,82

Primary Incomplete 0,04 0,20 0,05 0,22 0,03 0,17 Primary Complete 0,17 0,38 0,20 0,40 0,14 0,34 Secondary Incompelte 0,17 0,38 0,20 0,40 0,13 0,34 Secondary Complete 0,27 0,44 0,28 0,45 0,25 0,43 University Incomplete 0,14 0,34 0,12 0,33 0,15 0,36 University Complete 0,21 0,41 0,14 0,35 0,29 0,46 Age 38,77 12,35 38,48 12,60 39,16 11,99 Age Squared 1.655,42 1.006,42 1.639,44 1.026,46 1.676,95 978,44 Public Employment 0,22 0,42 0,18 0,39 0,28 0,45 Self Employment 0,18 0,38 0,20 0,40 0,15 0,36 Formality 0,51 0,50 0,52 0,50 0,51 0,50

(40)

40

Single 0,55 0,50 0,56 0,50 0,53 0,50

Amount of Children 0,30 0,61 0,33 0,64 0,25 0,56

N 20976 12038 8938

2013 Hourly Wage 112,21 116,39 112,53 97,90 112,53 97,90

Log Hourly Wage 4,49 0,74 4,48 0,78 4,48 0,78

Primary Incomplete 0,04 0,20 0,03 0,17 0,03 0,17 Primary Complete 0,16 0,37 0,12 0,33 0,12 0,33 Secondary Incompelte 0,17 0,38 0,14 0,34 0,14 0,34 Secondary Complete 0,27 0,45 0,26 0,44 0,26 0,44 University Incomplete 0,14 0,35 0,16 0,36 0,16 0,36 University Complete 0,21 0,41 0,30 0,46 0,30 0,46 Age 39,00 12,34 39,43 11,91 39,43 11,91 Age Squared 1.673,55 1.008,32 1.696,81 973,87 1.696,81 97,39 Public Employment 0,22 0,41 0,275 0,447 0,275 0,447 Self Employment 0,18 0,39 0,157 0,364 0,157 0,364 Formality 0,51 0,50 0,518 0,500 0,518 0,500 Single 0,55 0,50 0,530 0,499 0,530 0,499 Amount of Children 0,27 0,58 0,239 0,529 0,239 0,529 N 20329 8488 8488 Source: EPH.

(41)

Table a4: Determinants of Earnings OLS, 2004

Pric 0.079** 0.095** 0.062 (0.02) (0.03) (0.04) Seci 0.248*** 0.288*** 0.192*** (0.03) (0.03) (0.04) Secc 0.371*** 0.390*** 0.354*** (0.03) (0.03) (0.04) Unii 0.609*** 0.624*** 0.595*** (0.03) (0.04) (0.04) Unic 0.916*** 0.978*** 0.867*** (0.03) (0.04) (0.04) Age 0.045*** 0.045*** 0.044*** (0.00) (0.00) (0.00) AgeSquared -0.000*** -0.000*** -0.000*** (0.00) (0.00) (0.00) Male 0.128*** (0.01) Public 0.080*** 0.049* 0.091*** (0.02) (0.02) (0.02) SelfEmp -0.081*** 0.033 -0.263*** (0.02) (0.02) (0.03) FormalEmp 0.294*** 0.343*** 0.250*** (0.01) (0.02) (0.02) GranBA 0.093*** 0.076*** 0.121*** (0.01) (0.02) (0.02) NW -0.239*** -0.239*** -0.237*** (0.02) (0.03) (0.03) NE -0.359*** -0.393*** -0.307*** (0.03) (0.04) (0.04) Pampa -0.178*** -0.199*** -0.141*** (0.03) (0.03) (0.04) _cons 0.906*** 0.962*** 0.996*** (0.06) (0.08) (0.10)

(42)

42

Pric 0.079*** 0.115*** 0.030 (0.02) (0.03) (0.04) Seci 0.210*** 0.263*** 0.133*** (0.02) (0.03) (0.04) Secc 0.347*** 0.363*** 0.330*** (0.02) (0.03) (0.04) Unii 0.612*** 0.615*** 0.615*** (0.03) (0.03) (0.04) Unic 0.917*** 1.016*** 0.824*** (0.02) (0.03) (0.04) Age 0.045*** 0.041*** 0.051*** (0.00) (0.00) (0.00) AgeSquared -0.000*** -0.000*** -0.000*** (0.00) (0.00) (0.00) Male 0.104*** (0.01) Public 0.017 0.008 0.017 (0.01) (0.02) (0.02) SelfEmp -0.121*** -0.042* -0.245*** (0.02) (0.02) (0.02) FormalEmp 0.335*** 0.366*** 0.312*** (0.01) (0.02) (0.02) GranBA 0.097*** 0.092*** 0.105*** (0.01) (0.02) (0.02) NW -0.292*** -0.280*** -0.304*** (0.02) (0.03) (0.03) NE -0.325*** -0.366*** -0.267*** (0.03) (0.04) (0.04) Pampa -0.132*** -0.131*** -0.132*** (0.02) (0.03) (0.04) _cons 1.033*** 1.149*** 0.990*** (0.06) (0.08) (0.09)

(43)

Pric 0.134*** 0.140*** 0.130*** (0.02) (0.03) (0.04) Seci 0.285*** 0.307*** 0.254*** (0.02) (0.03) (0.04) Secc 0.425*** 0.454*** 0.390*** (0.02) (0.03) (0.04) Unii 0.706*** 0.689*** 0.731*** (0.02) (0.03) (0.04) Unic 1.009*** 1.003*** 1.010*** (0.02) (0.03) (0.04) Age 0.038*** 0.040*** 0.034*** (0.00) (0.00) (0.00) AgeSquared -0.000*** -0.000*** -0.000*** (0.00) (0.00) (0.00) Male 0.165*** (0.01) Public 0.033* 0.008 0.046* (0.01) (0.02) (0.02) SelfEmp -0.099*** -0.052** -0.194*** (0.01) (0.02) (0.02) FormalEmp 0.341*** 0.331*** 0.354*** (0.01) (0.02) (0.02) GranBA 0.036** 0.026 0.051** (0.01) (0.02) (0.02) NW -0.347*** -0.368*** -0.316*** (0.02) (0.03) (0.03) NE -0.388*** -0.405*** -0.362*** (0.03) (0.03) (0.04) Pampa -0.179*** -0.183*** -0.176*** (0.02) (0.03) (0.03) _cons 1.337*** 1.444*** 1.412*** (0.05) (0.07) (0.09)

Gender wage gap and returns to schooling in Argentina

Gender wage gap and

returns to schooling in

Argentina

Virginia Giordano

Student Number: 10824839

Supervisor: Dr. Carmine Guerriero

Date: June 30

, 2015

Master Thesis

Faculty of Economic and Business

Universiteit van Amsterdam

A

BSTRACT

Gender wage gap and returns to schooling

in Argentina

Contents

List of Tables

I.

Introduction

II.

Education and Wages by Gender: 2004-2013

III.

Methodology

:

The Minceran Wage Earning Model

IV.

Data and Descriptive Statistics

V.

Econometric Results

VI.

Conclusion

VII. Statistical Appendix

_{, 2015}